You are on page 1of 8

Peer-Review Research

Purification of Protein by
HIC: Mechanistic Modeling
for Improved Understanding
and Process Optimization
LUCRÈCE NICOUD, YOHANN LE GUENNEC, EDOUARD NICOUD,
ANTONIO CARDILLO, ELENA LIETTA, ELISA INNOCENTI, AND ALESSANDRO PIERI

ABSTRACT
Hydrophobic interaction chromatography (HIC) is a separation technique widely
used for the purification of therapeutic proteins. The main leverages identified
to operate HIC processes are solution pH and ionic strength. In this article, the
ArtemisDiana/Stock.Adobe.com

authors show that simplistic mathematical models derived from statistical analysis
cannot describe the impact of pH and ionic strength observed in HIC processes.
A new mechanistic model accounting for the protein charge in solution (varying
with pH) and the solution non-ideality (activity coefficient varying with the ionic
strength) is proposed here. This simple model is shown to accurately describe
experimental data and paves the way for numerical optimization of HIC processes.

H
ydrophobic interaction chro- be applied in a straightforward manner
matography (HIC) is widely to HIC. In another study, a HIC model
used for the purification of inspired by a complexation mechanism with
therapeutic proteins (1,2), ligands has been proposed (5). This model
Lucrèce Nicoud*, lucrece. in particular for aggregate removal (3). has the advantage of allowing the predic-
nicoud@ypso-facto.com, is Designing a HIC process requires deter- tion of the mean retention coefficient with a
head of product; Yohann Le mining numerous operating parameters rather limited number of model parameters,
Guennec is project manager; such as buffer pH, buffer ionic strength, but it does not provide a way to explic-
and Edouard Nicoud is loading volume, f low rate, and others. itly describe the impact of pH and ionic
head of customer success; Simulation can be used as a tool to investi- strength on chromatograms.
all at Ypso-Facto, France.
gate different operating conditions digitally In this article, the authors propose a sim-
Antonio Cardillo, PhD, is
and thus reduce the experimental burden. It ple chromatographic model capable of pre-
expert scientist & GSK fellow;
Elisa Innocenti is scientist;
can also be used to perform process optimi- dicting the impact of pH and ionic strength
and Alessandro Pieri, PhD, zation and reduce the environmental impact, on HIC chromatograms. This mechanistic
is scientific leader & GSK in particular by decreasing raw materials model is based on the dependence of the
associate fellow; all at GSK, consumption and waste generation. protein charge with pH and on an explicit
Italy. Elena Lietta, PhD, is Existing HIC models are often inspired expression of the protein solution activity
from Politecnico di Torino, by the protein solubility behavior observed coefficient with the ionic strength. The
Duca degli Abruzzi, Italy. in solution (1). An interesting study treats model is first tested on literature data. Then
salt-induced protein precipitation as a liq- it is applied to a set of experimental data
uid–liquid phase separation with a pro- generated by GSK. Finally, it is used to per-
To whom all correspondence
*

should be addressed.
tein-poor supernatant and a protein-rich form process optimization.
precipitate (4). A theoretical model capa-
ble of accounting for both pH and ionic MATERIALS AND METHODS
PEER-REVIEWED strength effects was derived. However, Experimental part
Submitted Sept. 15, 2022 this model counts many model parameters Materials. A mmon iu m su l fate w a s
Accepted: Dec. 8, 2022. that are difficult to estimate and cannot purchased from Carlo Erba Reagents

22 BioPharm International  April 2023  www.biopharminternational.com


2.2 Ammonium A and Buffersulfate B was set was7.0.purchased The
1−ε e flowrate was 1.16 mL/min, which corresponds to

2.2.1 There Statistical


2.1.1 isModeling
currently model
Materials a high part interest in using statistics from to Carlo describe Erba theReagents impact of key (Milan, operati I
cm/h.
phosphate The from impact Sigma of solution
Aldrich pH and Louis,
(St. ionic strength MO, (modulated
USA). A by the %
solution of B)
6 wa
mo
on 2.2
2.2.1some
2.2 Modeling
relevant
Statistical Modeling quantities part
model part (e.g., purity, yield). Typically, a Design of Experiment (“DoE”
There 2.2 is Modeling
currently this reference
part
a are high experiment.
interest in purchased
using Overall, fivetoexperimental runs were performed:
Purification ofRun pro
andAmmonium purchased
the results from
regressed
sulfate Carlo
was based Erba on statistics
Reagents
linear or quadratic
from describe
(Milan,
Carlo the
Italy).
expressions.
Erba impactReagents Inofthis key(Milan, operating
study, quadra
Italy pa
on some 2.2.1
Thererelevant at
is pH
Statistical
currently 7.0,
quantities Run a high 2
model
(e.g., (low purity,
interest ionic strength):
yield).
in using Typically, 18
statistics % B a to at
Design pH
describe 7.0, of Run
Experiment
the 3
impact (high ionic
(“DoE”)
of key strengt
opera is p
were used
2.2.1
phosphate in
We4Statistical an
define attempt
frompH): “Buffer
Sigma to
model describeA”
Aldrich the
as(St. an variations
aqueous
Louis, MO,of
Peer-Review the
solution USA). lumped of% Athis Henry
Research
0.2
solution mol/L coefficient
of potassiu
6 mol/L with
and the onresults
some Run are
relevant (low
regressed quantities based 23 %(e.g.,
onBlinear atpurity,
pH or 6.0, Run 5Typically,
quadratic
yield). (high
expressions. pH): a Design 23In B ofat study, pH 8.0.
Experiment quadratic (“DoE ex
strength
There mol/L Iis observed
currently experimentally
a high interest in
in the
using literature:
statistics to describe the impact of key oper
were in used
HIC
and purchased
There
processes
in
the an results The isammonium
attempt from
is process
currently
are extremely toCarlo
regressed was
a high
describe sulfate
Erba
monitored
high interest
based
and
Reagents
(several
the in“Buffer
by
variations
on linear online
using
mol/L) or(Milan,
of andB” itonline
pH,
statistics
the
quadratic
asis
lumped aexpressions.
Italy).tosolution
known conductivity
describe
Henry thatof the
the 0.05
coefficient and
impact
presence
In Experiment
this
mol/L
online
study, ofofkey
with UVof op
salt
pH
quadr
paa
on The some
We The pH relevant of
following
define the quantities
“Buffer Buffers
calibration A” (e.g.,
A was
as andpurity,
an B
established yield).
were
aqueous in Typically,
adjusted terms
solution of to aof
totalDesign
reach
0.2 protein mol/Lof
a targetconcentration:value
potassium (“Do (6
impacts
strength were I onthe
observed some values relevant of
experimentally pK quantities
[8, 9, 10].
in (e.g.,
theBesides, purity,
literature: it isyield).
clear Typically,
that the a
spatial Design
2 configurationof Experiment of the (“
and used the results in an are attempt a
regressed to describebased on K(I)the linear variations
= aor 0+ a1 ×
quadratic ofIthe alumped
+expressions.2×I Henry In this coefficient
study, quad wit
impact mol/L
strength of
and6apparent
its theImol/L
ammonium results
observed ofare
net potassium
sulfate
regressed
charge:
experimentally it and hydroxide
based
has “Buffer
been
in on reported
the linear or
B”or1asquadratic
literature: mol/L
that aofsolution theofpK phosphoric
a of
expressions. of 0.05 amino mol/L acid.
Inacids of
this study, pota
depend qu
0.3
( M i l a n , I t a l y) a n d p o t a s s i u m 2.2 their mL/min.
were
were
used
used The in an
in injection
12]an
attempt
attempt volume
to describe
to describe c the(g/L)
the
variations= U V (mAU the lumped
)/2450
2the lumped Henry coefficient Henry
(mAU.L/g) coefficient w
location
The
Modeling pHThe
part [11,
of crude
the and to
Buffers be
some purified
A amino
K(I)
and B= was
acid a 0 residues
were + 1 variations
obtained
aadjusted× I may + aby 2 be×of
to cell
Ipoorly
reach culture aaccessible
target
Purification and pre-purified
because
value of (6.0,
protein“bu
was strength I observed experimentally in the bliterature:
monobasic phosphate from Sigma core10 µm with athus monomer con- 2
of strength Iruns observed notexperimentally K(pH) in=the the + b1 a× ×
literature: pHInet + bacharge. 2× pH 2 To reduce the numb
of theAll
6 UV protein,
mol/Lthe saturation of reported
potassium was contributing
observed in this
hydroxide atK(I) to
study
around or= 0aapparent
were
12500 +
0 mol/L 1performed
mAU. of+phosphoric 2 × Iwith the
acid. same start
Aldrich (St. Louis, Mo.). A solution centration
parameters, where in only
KThe the is the range
the pKlumped 50–500
a values Henry and µg/ number
coefficient of and accessible a0by ,+ aspartic/glutamic [Eq. acids 1] fractions
mL and histid
concentration
The crude material to in the
beelutingpurified crude
K(pH) from was=wasthe 0.64
0obtained
bK(I) +HIC b =×column
apHg/L.
0+ aa11cell
b ,was
Size
× a×I2 ,culture
+
pH b0a,22b×
Exclusion
collected 1 , Ib2 and
2 in are2 4.7
fitting
Chromatograp
pre-purified model pa
by
of 6 mol / L of potassium hydrox- inmL. Dilution of the sample was per-atmol/L)
1 2
were
HIC pulse considered
processes experiments, is as
extremelyadjustable
the lumped
high parameters.Henry
(several coefficient K(I) and =can a0isbe
it +known 1 × I that
acalculated + a2the × 2Ipresence
from the measuredof salt ret
sign
All column
the
presence runs volume)
ofreported
three and in storedthis study-20°C =werebefore
0 + performed
SEC analysis with to
× pHthe same determine the conten
starting
ofgroups ofK(pH) species bdenoted
b0Low
b1 × pH + bMolecular Weight (LMW
ide was purchased from Carlo Erba impacts formed
whereasThe with
K activity
follows:
the is
values
and 10HMW.
the mM
lumped
of pKpotassium
coefficient a [8, 9, 10]. phos-
Henry the protein
Besides,in
coefficient and UaV
itsolution
is 0 , saturation
clear a1 ,thatwasa2 , computed
the 1 ,was
, bspatial b22  are obser
using fitting
configuration the ved at of parame
Truesdell-Jone
model the pro
pulse as 2.2 concentration
follows: product)
experiments, Modeling the and
part in
lumped the
High crude
Henry Molecular was K(pH)
coefficient 0.64Weight
ε
can= g/L.
V bbe + Size
(HMW)
b
calculated × 1 Exclusion
pH− ε + accounting
from b × thepH Chromatography
2
measured for
Purification8%, 74%
retention of
Reagents (Milan, Italy). phate at
impact itswhere
pHapparent 7.2. K isnet thecharge: lumpeditHenry has been coefficient
tRaround
reported
K(pH)
= denoted
e and
= 2500
0
colthat
ba010+ , the
1 ab1 , ×apK
+mAU. 1
2pH,eab0+ , b ,×
2
Kof b1amino 2
b2pH are 2
acidsfitting depend model strp
as follows: presence
pulse content,
experiments, of three
respectively. the groups
lumped of
The species
Henry conductivity
coefficient  Q can √ Low
of be the Molecular
calculated
ε  crude was
from Weight adjusted
the (LMW),
measured with re
“Buffer A” is defined in this article theirHIC location experiments.
product)
where [11, K
whereand
12]is and the some
K monobasic
isHigh the HIC
lumped
lumped
amino
Molecularexper-
Henry acid
Henry
residues
coefficient
εWeight
Vcol TheAand
ecoefficient
may
materia
(HMW) 2−
z1and
a0be
Inet
poorly a2 , baccessible
ε,aea01,l,accounting
aeluting
e
, a20, 
, b1 ,from
bTo 0 , breduce
b2 are
1 ,for
because
b2sulfate
are
fitting “buried
the
8%, fitting
74%
model
mo an
core of asthe
wherepulse potassium
follows:
protein,
Vcol experiments,
is the columnthus not thewithcontributing
lumped
volume, phosphate
t =
log(γ
RεHenry to the and
apparent
= −εcolumn
)external
coefficient 1 + k2 k mol/L

canwas beK
+ charge.
C 1ammonium
k ×QI the in
calculated from the the to
number
measured r rea
as an aqueous solution of 0.2 mol/L iments were performed ahigh 7.7 e the kHIC porosity and
collected flowrate.
parameters, incontent,
HIC
asThe
pulse
Buffer
only
follows:
processes
ionicthe
experiments,
respectively.
A. pK
strength The is
a values
extremely
pHI isof
theand lumped
The
the
defined number Henry
(several
conductivity
crude
as: tR
Qwas
of =
e1V+
coefficient
mol/L)
then
accessible colof
kand
Badjusted

ecan
1the I+ it −
1crude
εaspartic/glutamic is εknown
be calculated
e was
to Kreach that
 adjusted a4.7
from
acids the
targetmL
andthe with
presence measure
value
histidine ofa(6 s
potassium monobasic phosphate and where x
were 100 V mm
impacts as
potassium
considered
wherecolAof is column
follows:
the the as columnvalues (HiScreen
monobasic
adjustable of
volume, pK [8,
ε Butyl
9,
phosphate
parameters. the 10].external fractions
Besides, and ε2
porosity Q
it V is
mol/L
1 (corresponding
clearand  that
Q 1 ε
5
ammonium e the

the ε spatial
flowrate. to  the
sulfate col-
configuration
defined in Equattto reach of
k , 6Bk mol/L, Ck arepotassium a
3 fitting eparameters hydroxide and or =
R =I volume)
tbeen I1 is
e col
εe V the col1 +Cphosphoric
ionic
zk2and
strength
1 −K e
εe of asacid.
1 mol/L ammonium sulfate and “Buffer HP, Thecolumn
impact
ionic
activity
where
parameters Buffer Avolume
its
Vstrength
col A.k is
apparent
coefficient
,B k , ICcolumn
the
The k
4.654
ispH netthe
defined
of
were mL,
of charge:
volume,
the
estimated as:
protein GE
crude itεin has
eby the
was umn
solution tR =
external
then
comparing reported
wasQ computed
2 porosity
adjustedmodel that k+ the
to andεstored
1simulations using
ereach QpK the K the
a
aat -20amino °C
Truesdell-Jones
flowrate.
target
with acids depe
value
experimental (6.0, md
 Qk may be εpoorly
B” as a solution of 0.05 mol/L of potas- asHealthcare, follows:their
of
where The6
location Uppsala,
ionicisstrength
mol/L
V
[11,
potassium
the Sweden)
12]
column
and
I is volume,some
definedusing
hydroxide
aminoas:
ε I before
the
acid
or
=
1
1
external SEC
residues
mol/L z 2
C analysis
phosphoric
porosity andto e
determine
Q acid.
the
accessible
flowrate. the because
core 2.1.2
where of the
where C col protein,
kVcol isSize the
is system
the Exclusion
thus not volume,
concentration ofChromatography
contributing e
species 2to the√
k external
in solution
1LMW, k
kapparent
2and
analysisnet zkand charge.
itsand valence. Toflowrate.
reduce the nu
sium monobasic phosphate. The pH of a Interactions
chromatography with the column solid (ÄKTA phase as:The εcontent
e the
Ik =of
interactions porosity
zwith CMono, the solid Q the HMW.
phase were descr
In The
parameters, theThe ionic
literature, only
ionic
strength
the
strength pKIa is
results values
Iare defined
is oftenandpresented
defined number
as: Ak zofk2inaccessible2I1  kof kaspartic/glutamic
terms k  , which is relatedacids to Kand as fo hi
Buffers A and B were adjusted to reach a Avfollowing a nwere
where t C2The 5adsorption
, the
is CSEC y concentration
2.1.2 Size Exclusion Chromatography t i vanalysis
aisotherm:
, adjustable
U S Awas )of log(γ
sspecies
uperformed
p -k ) = k −Column
in solutionusing √ and characterization.
k+1 C
an z k 2×
ACQUITYits I valence. UPLC Blue Protein
kconsidered as parameters. 1 + Bk I I2= analysis
I = zk C
kεzk 2
t k Ckis
target value (6.0, 7.0, or 8.0) with the ported
In the by µm,
The a activity
literature,
where software
Ck is
4.6 results
mmx300 the
coefficienttrending
aremm)
concentration oftenof the tool
presented
on of
anspecies
protein dextran
ACQUITY
K =in εterms
k +in(1solution
solution and of
−UPLC k2ki )acetone
εwas , which and
computed system, kzk pulsesrelated
its valence.
usingboth tohave
the K Truesdell-J
were as follows:
purcha
AModeling B C CI kiis=the λink γtermsk Ck 1kstrength − ε
addition of 6 mol/L of potassium where (Unicorn,
2.2 as The
k ,In
follows:
USA). k
wherethe ,
Cytiva,
SECCThe are
part
literature,
k USA).
analysis
is
3 fitting
eluent
the results was parameters
was
concentration performed
are 10 often
mM of
andbeen
presented
sodium
species using per
k anfor
phosphate
in
ionic
solution
ε med
ACQUITY of to
and k t
with, est
which
z
as imate
UPLC
400
its
defined
is mM related
valence. t he in to
Protein
ammonium
Equation
Purific
K BE as f
parameters Aand , BkC, kCCkare
kwhere kwereisthe the estimated
concentration by comparing
of species − modelk inktsolution
simulations  and kwith experimental data
z its valence.
hydroxide or 1 mol/L of phosphoric acid. The µm,
where following
Cwas 4.6 mmx300 protocol mm) was on
concentrations usedan K =
ACQUITYε
of external
i +the (1 species ε and
)
UPLC
i total
in
ε√
k
system,
the porosities,
ε µm liquid
 both and
k respec-
were
solid purchased
phases,to Kresp
In0.3 the mL/min. Theareinjection volumein1 was t 10 with is a related
monomer to con
In the
k literature,
k results often presented −terms of k , which as
2.2 2.2 Modeling 2.2.2 approach
Modeling Mechanistic
part part used,model
literature, results are often presented
K contributions
= εphosphate + (1 Ak− in
zk2εterms I t of
i )from
, which
kk species is related
Purification
Purificatio K
The cr ude to be purif ied was forlumpedthe USA).reference
µg/mL. The experiment:
is
eluent
Dilution meaning
was of 10
the First,
mM thatsodium
sample the
log(γ tivelywas
k ) = i(6,7).

performed The √ 1with − following
+
with ε ε Cthe
t k 400 ×10 I mM mM values
adsorbed
ammonium
potassium on th
su
ph
Interactions with theof solid phase are The interactions withinIthe solid  phase were describe
obtained by cell culture and pre-pu- 2.2.2 located
equilibration
2.2
2.1.4
following
Solution
was in 0.3
Modeling
Mechanistic
Column
adsorption
the
with
chemistrypores
mL/min. part Buffer
model
theThe
characterization
isotherm: Theparticles
A during
following
injection averaged
were
reactions K
volume =
K
ε
=
together
obtained:
in
i + 1
was
ε
+
solution
(1
+
B−
(110
k εε− i
e
)µmε
C kt.with
=1i )−
were 0.4 ε
ε t andk εt = 0.93.
The
taken
k a
activity
into
monomer account:coefficient Purifi
conce γ
2.2 Modeling the where part
non-ideality
Equilibrium
µg/mL.
of the protein solution, pK while the coefficient
i λ quantifies
110 t
−strength
εt thus Purification
the adsorption of st
rified by two chromatographic steps. Solution 4.5 bed volumes Bk(BV); second, load- The internal
and I is porosity was found
AMechanistic , Ck are k
2.2.2 k , Dilution of 3 the
model fitting sample parameters
a was performed the with ionic mM as defined phosp
potassium in Eq
2.2 Modeling
2.1.3
chemistry  part + part
HHydrophobic +The Interaction Chromatography experiments Pur
the chromatographic medium. Ckfollowing The reactions
parameter byincomparingsolution
λk was were
estimated taken byinto comparing account: model simu

2.2 ModelingH 2O OH 14.00 Purificat
All the runs reported in this study 2.1.4 Blue
ing
2.1.4 with
Column 2.2
parameters
dextran
Column
2.2.2
Equilibrium
experimental
Solution N crude
H
Modeling
+ Mechanistic

data.
A
and
during
chemistry
N
k , Bacetone k ,part
characterization
characterization H + approximately
H
werepulses
model
+The pK
estimated
following
a 9.23
have equal
C
reactions
k
been
= λto γ
k in ε C k (ε t − ε e)/(1
performed
ki = solution
model
were
simulations
to -taken ε e) = 0.89.
estimate
into
with the experiment
account:
extern P
2.2.2 Mechanistic 3 model
5respectively +[6, 7]. −characterization
The − following values were obtained: e = experiments
εButyl 0.4 HPand εtatPurifica
=low 0.93
4
were performed with the same starting where 2.2BV,
2.1.4
HThe
CO
2 which
2.1.3
k+and

HHIC
Modeling Column
HO
Equilibrium
lumped
3 PC
+roughly
experiments
 part
OH
Hydrophobic
Henry
are H 2P O
the corresponds
coefficient +were
concentrations H 14.00 isperformed
+Interaction pK
defined
2.15 asspecieswith
the Chromatography
slope a HiScreen
ofthe theliquid adsorption isotherm column (7.7 co
eof
2.2 Modeling part athe k
in in and solid phases, respec
Puri
BlueBlue
2.1.4 dextran
Column
was dextran
Solution
Interactions
thus  and and
characterization
found 
chemistry
k4acetone
equalacetone
with
+ + to pulses
4the
ε pulses
The
solid
= ε t −ε have
following
have phase been
= been
reactions
0.89.Theperformed performed
interactions solution toanto with were
estimate estimate the taken the
solid the
into external
phase external
account: were ande
material. The total protein concentra- to
lumped 2
N mg
H
Considering
2.1.4 4 H mL,
Solution
approach
following
H
mono P2 NO/mL
O
Equilibrium
Column
GE
H
Equation
− 3is
adsorption +H
bed
H ;
chemistry
HP
used, third,
Healthcare,
+ 8,
OOH we
2−
meaning

+
characterization
isotherm: isocratic
9.23
obtain:
i H +Uppsala,
The 1−ε
that following
14.00
6.82ethe
pK Sweden)
Modeling reactions
contributions using
partin from solution the ÄKTA were
species avant
takenadsorbed into 25 (Cytiva
account:
on the sT
respectively
respectively Blue HIC [6, 2
dextran [6,
7].
experiments4 7].
The and The
− followingacetonefollowing
4
were
+ + + performed values
pulses values werea
have were
with obtained:
been obtained:
a performed
HiScreen ε ε= Butyl =
0.4 to 0.4 and estimate
HP and ε
column ε= =0.93.
the 0.93.exter
(7.7x10 The
4N 2
Equilibrium pK
+ e t
tion in the crude was 0.64 g/L. Size 2.1.4
elution H3 P inO Column
with 4H P O characterization
+ Hparticles − 2.15 →
− e t
477% 4Buffer A + 23%
located 2.1.4 HP
the H H OThe 2−
pores
O Column
 N
following
of
H P H+ O
the 3++3− H
+characterization
Hprotocol are 9.23
12.38 was
averaged K ka =together
used
Statistical λkfor γk (C the model.
k in C 0)k . There
reference The activity is
experiment: currently coefficient (1) γEqu
k tq
Blue dextran and acetone 4+ OH
pulses ε −ε have
ε −ε been 14.00 performed to estimate the external and
waswas thus thus
Hfound
respectively
2 PmL, Ofound

H2N GE2
equal
PBed H HP
O 2equal
O
[6, 
to
Healthcare,
4 O 2−
7]. ε2to
H
Hprotein+iPThe
− =H
Oε+i+OH
− =t
+
−e
following
Uppsala,
+6.82
H
t e
+= 0.89. =
2.15 0.89.
14.00 were
values
Sweden) using obtained: ank ÄKTA εetoavant = 0.4adsorption
25and εthe
(Cytiva)t =stren 0.9 su
exclusion chromatography (SEC) the
respectively Bu
2.2
2.1.4fBlue
fer
non-ideality
The
[6, B
H4.5
dextran
Modeling
4
ColumnSO
approach
7].
3
du HThe r4+and
4of ing Volumes
the and
HSO
used4N 2H
following 0and part BV;
4acetone
++
characterization
3in 4H
this(BV),
H +
fou
1−ε
solution,
1−ε article
e
values r(2)
++pulsesεtth,
e 1.92 Loading
while
can9.23
were be ahave high
the
extended
obtained: with
been interest
coefficient
C = to Crude
λ performed
εkeanyγλ = in using
during
quantifies
Ckother 0.4 andkind statistics
approximately
εtoestimate
the
of adsorption to external 5exte
isothe B
Blue was dextran
2.1.4
HP Blue
thus
O 2−
H The  dextran
P N
Column
found P
O H−+

3−
following
O  acetone
equal
+HP N H H
O+ 2−
characterization
protocol
to ε
+ pulses
acetone
+ H 12.38
H = −ε
was have
pulses
6.82 e 9.23
used
= been
have
0.89. for performed
k
the been reference kperformed toexperiment:
estimate t = 0.93.
the
estimate (1) Thethe
Equilib ine
ana lysis revea led the presencewas of thusthe chromatographic
regeneration HSO −
with4[6,
4

4SO medium. 2−
100% P+ 3
H − +The
,4Buffer parameter
+ 8.06
Bvalues λk was
describe estimated
the impact by of comparing
key model simulati
εoperating
respectively
Langmuir
found
respectively
42equalmg
H 2 Pmono
adsorption
34 O 4
to ε/mL 
H
7].4isotherm.
=O 2bed ε tO
The4−ε (3)
e+
i−
= Isocratic
HThe
following 1−ε
0.89. + 2.15
e values
concentration Elution were inwith the 77%
obtained:liquid Buffer
phase εe0.4 A
=kand
C + is 23%
0.4 and
indeed Buffer t =B
=ε5externa
simply 0
 k[6, P7]. kThe 2following were obtained: k inεthe eto= εthe 0.93.
H O − H P O ++ HLoading 2.15
−
2− 3−
+
Blue
experimental respectively
H2where
SO4.5 4HP
dextran
The Bed C net
data.O Pand
HSO 4Volumes
3charge
and 4C 
4i[6,
+P are
H
acetone
of 7]. Othe
(BV),
the
4 1−ε +2−The
eH 41.92
protein (2)
pulses following
concentrations
+ 12.38
in have
solution values
ofwith
been thewas were
species
Crude
performed
calculated obtained:
during as aliquid
approximately
estimate
function e and = 0.4 solid
of t pHand BV,
fromεt =
phases, tw
t h ree g roups of species denoted 2.22.2 was
γModeling
2.2.1
during C
kwas
Blue
HSO
kModeling
thus
.
lumped
was 3
−with
thusH 
H
Statistical
BV;
dextran
foundthus
2found
SO H
approach
SO
O
100% and
242−
2−part
4
Pfound
 O+
equal HSO
part
f
−equal
and H
HP model
ifth,
Buffer
 isto
+ acetone
HP

used,
equal
to
4 O= B +
stripping
ε
2−
ε+ifollowing+
H
during
ε
i+
meaning
8.06
H4Isocratic
to
t=−ε
+εpulses
ε+
Hofe
=
t −ε
6.82
=
1.92
1−ε
3 parameters
e BV,
6.82
−ε
t0.89.
εthat =
have 0.89.
ethe been
=
(5)
contributions
0.89. on
Stripping some
performed relevant
from A with to
the quantities
Water
estimate
species during theatconce
adsorbed 3
exte Bon
The sequence.
respectivelylumped
2 4mg HP 2mono To
Henry
O[6, 4/mLdo7].  4 so,
coefficient
P
bed The Othe43−
, (3) number
+ is
H defined
1−ε i positive
as
Elution
values
12.38e the and
slope
were with negative
of the
77%
obtained: charges
adsorption
Buffer ε = were isotherm
+0.4 23% computed
and at
Bufferε low= B each
0.93.du
as low molecular weight (LM W ), Considering with2.2 water AHSO
isModeling
and during 4Buffer
3 BV.
Bpart was4 The
set pH 7.0. The
1−ε
(e.g., purity,
e flowrate was 1.16emL/min, which
yield). aTypically, athe design t corre
2− 4O 3− + e
The
There respectively
located
the net number charge
currentlyin HP − the
4and O [6,
of
4pores
SOthe
pK a7].
2−
4high
PofThe
protein
of + the
H
each +following
+
interest in+H
particles
aminoεtsolution
−ε8.06eacid,
values
12.38
are averaged
was
and were
calculated
the two obtained:
together as
contributions C kε.ewere
infunction =of
The 0.4activityand
pHimpact
then from
summed εtthe = to
coefficie 0.
am
+ εt3in using statistics towith describe of inket
2.2 Modeling
2.2.1 Equation
Statistical
with 100% part 8,
model
Buffer we obtain:

2.2.1 was thus
Statistical H found2 SO 4 equal
model HSO to 4Bε +−iduring
H
= 1.92BV,
=e1.92 0.89. (5)the Stripping Water during 3byBV.
a
Hydrodynamics and  kinetics To simulate experimental data acquired by adsorptio
GSK
monomer (Mono, the target prod- sequence. of
on the
2.2the
was
some crude,
cm/h.
non-ideality
To
thus
The
charge.relevant do
Modeling
HSO netH foundBuffer
so,
2 SO
− The
charge

the
4 of
equal
quantities
SO A,
impact HSO
the
number
of
2− and
protein
the
part
+ to4 +i
H
of+
εBuffer
protein
(e.g.,
Hsolution
= solution,
positive
1−ε e
−ε
in
K
purity,
8.06 k of
and= pH
=
solution experiment
while
λ 0.89.
γ and
negative
yield).
k k
the
was
(C k
ionic
coefficient

− (DoE)
charges
calculated
Typically,0) strength is
were λ ask
aperformed
a
Design
(modulated
quantifies
computedfunction
ofand
theat
of each
pH
Experimen
pH
from the
2.2 the
2.2.1 A
present and
Modeling Buffer
study,
Statistical 4 − a Bpart
mixing 4was set
cell 7.0.
model The
1−ε was e used flowrate to contributions
describewas 1.16 mL/min,
the hydrodynamics which in thecorrespocolum
fmodel
2−
is2.2 Modeling part +
uct), and high molecu la r weight There
2.2.1
the
2.2 Band
There iswas
number
Linear
the
Modeling
the set
sequence. to HSO
chromatographic
thisand
currently
Calculations
currently
Statistical The
results
Driving 7.0.
reference
pK
part
aTo
model
net a The
high
are
Force
do
4aof
charge
each
so,
high
were SO low
interest
regressed
medium.
experiment.
the
of
amino
approximation
+performed
rate
number
interest
first
4 the H
in was
acid,
protein
based
The
inof
using and
was
using
onin
8.06
the
parameter
Overall, the
positive
assuming results
two and
statistics
statistics
solution
linear
used to or
fiveλthat:
was are
to was
knegative
quadratic
describe regressed
experimental
to estimated
(i)
describe
calculated describecharges
the
the kinetics
were pK based
the
asaof
expressions.
bythen
atheruns
were
of
impacton
comparing
function
of linear
were
summed
computed
impact
the
internal
amino
In ofof
model
perform
to
Purificati
of
key
thispH
mass
obtai
at
key
acid eac
ope
from
stud tra
sso
charge.The cm/h.
approach
experimental
the number The The used
net
and data. impact
in
charge
pK this
a of
of
article
of
each solution
the can
aminoprotein be pH
acid, extended
in and
solution
and theionic
to twoany
was strength
other
calculated
contributions kind (modulated as wereadsorption
a function
then by of
summed the
isotherm
pH % fr
to
(HM W ) accounting for 8%, 74%, on on 2.2
1.16some mL/min,
independent
2.2.1 Modeling
at
relevant
sequence. pH
Statistical which
7.0,
of
To the
quantities
do Run part
corresponds
ionic
so, 2
model
the (low
strength,
(e.g.,
number ionic to
purity, (ii)
of or
strength):
the quadratic
yield).
positive spatial and 18
Typically, expressions.
%
configuration
negative B at a pH
chargesDesign In 7.0,
of this
the
were Run
of study,
protein 3
Experiment
computed (highdoes at ion
noe(
There2.2some2.2.1
is
There
The
were
Langmuir relevant
2.2
Modeling key
used
this
2.2.1
Calculations
charge.
currently
This
ismodel
Statistical
adsorption
sequence.
The
provided part quantities
currently
Modeling
in
reference
lumped
a anparameters
Statistical
were
high attempt
isotherm.
To Henry
first adohigh
model
interest
aregressed
first
(e.g.,
experiment. so,
curve
are
part toThe
model
the
coefficient
performed in
purity,
interest
the
of%
number
describe
number
using
the
yield).
concentration
Overall,
is
assuming inofthe
defined
statistics
charge
using
ofpositive
vs
mixing
five
as
that:
Typically,
statistics
variations
inthe
to
pHquadratic
the cells
experimental
to and
(i)slope liquid
describe the
perform
Jof
negative apK
of to
and Design
the the
phase
the
describe
the
of
simulations. runs
charges
adsorptionthe
impact kof
characteristic
lumped
C Experiment
isamino
were theisotherm
indeed
wereTo Henry
of
impact time
simply
performed:
computed
Purification
acid
key
improve atof
side
operat
(“D
coefficoflow
rep
the ak
in
ch
and 14% of the protein mass content, a f luid velocity
Run 4 (lowof 150 pH): cm/h.23 The
B at pH quadratic
6.0, Run expressions
5 (high pH): were 23 used % B (see at pH 8.0.
a
and the
Ctransfer
and independent
γthe kresults
results
the number
tIiare
,relevant are
respectively,
regressed and pK whichof
based based
each were
on aminoon linear
determinedacid, or
and from the two
protein expressions.
contributions pulse aexperiments were In this
then study,
summed
performed q
on some
. Considering quantities (ii)linear
(e.g., purity, or quadratic
yield). expressions.
Typically, Design In were this
3 of study,
Experime qua
a
kstrength the observed
number and experimentally inin the literature:
2.2.1 There at pH of
Calculations
experimental is
Statistical 7.0,
the
currently Equation
Run
ionic
data, were apK
2strength,
model
the 8,
(low a we
first
high ofionic
assumptions each
obtain:
performed
interest amino
strength):
the were acid,
spatial
assuming
using and
progressively 18statistics
%
configurationthe
that: Batwo at contributions
pH
(i)describe
relaxed. the
to 7.0,
of describethe
pK
Indeed, Run
protein
of the
the (high
doesthen
amino
impact
ionic summ
ionic
not acid
strengt pla
of
respectively. The conductivity ofon some relevant quantities (e.g., purity, yield). Typically, Design of Experiment (“DoE
There is
charge. currently a high interest in using statistics to a the impact of key
the
were wereimpact
inThisused
and used
flowrates
HIC 2.2.1There
in
the
provided of
processes
independent
Run aninunder
results
charge. solution
The
anis(low
aStatistical
attempt
4regressed first nonprocess
attempt
currently
isofare curve
the
pH): pH
binding
extremely toionic
regressed of
23
toand
was
athe
model
describe
%
describe
conditions
high high ionic
monitored
based
charge
strength, interest
Bperformed
at (several
the pH
the
(van
variations
on
vs(ii) pH
6.0,
by
Equations
variations
Deemter
in
linear
theKto Run
online
using
mol/L) =or
perform
kspatial λ5kof 2
plot).pH,
(high
of
γstatistics
kandthe
quadratic
(C and the
k it →
−online
simulations.
configuration 3)
lumped
pH): is in
lumped
0) expressions.an
to23describe
known of attempt
conductivity
Henry
To
% the
Henry
that
improve
Bofthe protein
at to
the
the
coefficient
In
pH the
and
coefficien
impact
presen
this descr
does
8.0.
on
stunw
and the on on
results
some some Calculations
are
relevant relevant quantities quantities
were
based firston
(e.g., (e.g.,
linear purity, purity,
or assuming
quadratic
yield). yield). that: Typically,
expressions.
Typically, (i) the
a 15] DesignpK ato
In Design
this of study, of
Experiment Experim
amino quadra acid
cr ude was adjusted with a buffer strength
strength IRegarding The
Iis (modulated
observed the
following data that
experimentallyby were
calibration
the % extracted
B) was
in the from
established
describe the
literature: literature
the in
variations terms [14, of
of total
the perform 2 protein
lumped preliminary
conce
a
Hydrodynamics
impacts
strength There
experimental
were on
This the
observed
used some
independentThe
Calculations
currently
values
data,
provided in relevant
and
the
an
process of
experimentally
a
ofare akinetics
pK high
assumptions
attempt
first
the was
aquantities
were [8,
curve
ionic
first
interest
9,
to
monitored
To
of 10].
in
were
describe
strength,the (e.g.,
simulate
performed
the in using
Besides,
progressively
charge by
(ii)K(I)purity,
the
literature:
the vs
online
the
experimental
assuming
statistics
it is
variations
pH
= a
spatial
yield).
clear
relaxed.
topH, + that:
performa to
online Typically,
that
×of
configuration
data
Indeed,
I(i)
describethe thethe
simulations.
+ acquired
a
conductivity the
lumped pK
spatial
× athe
of the I aDesign
ionic by
ofimpact
To GSK
the
strength
Henry of
configuratio
improve
and
protein
in
aminoExpe
of
online
does
the
ke
enc
coeffi th a
were in used
HICand and
model,There
processes
in
the an the
The
only isresults
attempt
results approach
currently
theisresults extremely
arepeak to used
regressed regressed
acell
describe
positions inhigh
high this
basedwere based
article
interest
(several
the can
variations
reported
on onmol/L)
in
linear linear
be
using
inor extended
the or
statistics
0
and
ofquadratic
the
originalquadratic
to
it 1 is
lumped any toother
known
articles
expressions. expressions.
describe
Henry 2
(and kind
that of
notthe
the
coefficient
In the In
adsorption
impact
presence
this full this with
chrom
study, sto
iso
of
conta ining 0.4 mol / L potassium thewas
impact on then
present and
experimental
some
strength its
This studied
study,
independent
the
relevant
apparent
Ifollowing
provided observed a mixing
data, around
of
quantities
anet are
first thethe
charge: this
model
ionic
regressed
assumptions
curve(e.g.,
experimentally ref-
it was
strength, basedused
purity,
has were Henry
inbeen to
(ii)
on
the the coefficient
describe
linear
progressively
yield).reported spatial
literature: or the
Typically, relaxed.
that with
hydrodynamics
configuration
quadratic atotal
the pHIndeed,
Design pK and
expressions.of
a Cof ionic
inthe
the
ofis the ioniccolumn,
protein
Experimen
amino In thi
streng
acid do
impacts I were
Therefore,
onthe The
Langmuir
some used
values only adsorption
inof
relevant the an pK calibration
attempt
thermodynamic
aquantities
a [8,
isotherm.
9,curve toofthe the
wasThe
describe charge
established
parameters
(e.g., concentration
purity, cthe vsdescribed
pH
variations
yield). 6to
in perform
interms the
in
Typically, the liquid
of simulations.
ofparagraphs
the phase
Ilumped
a)/2450 protein
Design To
kabove improve
concentra
indeed
Henry
of were
Experim estto
sim
coe
strength observed experimentally l10].
l,in Besides,
literature: it(g/L)is clear = that
U Vperform the
(mAU spatial configuration
(mAU.L/g)
were used in an attempt to describe the variations of the 2internal
monobasic phosphate and 2 mol/L Linear
2.2 their erence
and
Modeling γ
Driving
were
the
location
This
experiment.
results
experimental
C . partused Force
provided
[11, in
are12]
approximation
data, an O
regressed
and
first
vera
attempt
the some based
assumptions
wasof
famino
ive
toK(I)
used
the
K(I)
describe
on to
charge
strength
linear
acid
=
were a= describe
+athe vs
0a+
or
residues
progressively
pH
that a1the
variations
quadratic
×
to
I ×
are may
+ Ikinetics
observed
a +
relaxed. alumped
be 2 I×
of
expressions.
× 2of
simulations.
the
poorly experimen-
Indeed,
Henry
lumped Inmass
accessible
To
this
Purification
the
coefficie
Henry
ionic
transf
improv
study bec
stren
not the
k
strength khydrodynamic Iparameters
observed / kinetic
experimentally parameters. inwere 0
the 1 2
a mmon iu m su l fate to reac h t he core
The
impact strength andmodel
key
its
experimental
were ofwere
the
apparent
experimental
strength
theused I observed
protein,
results
inruns I net
an were
are
observed
attempt
thus
charge:
data, regressed
are
experimentally the assumptions
the
performed:
notwere
number
it
experimentally
tocontributing
based
has
describe in ofthe
been
K(pH) on
mixing
ctally
the
linear
reported
toliterature:
in
in =
6the the
the
variationsbliterature:
cells or
progressively
+ Jquadratic
0aliterature: band
that
literature:
apparent 1a ×of
the
the
pHthe
relaxed.
Inet +pKexpressions.
characteristic
+lumped 2a ×
bacharge. of
Indeed,
pH amino
I 2Henry
2 time
To
Inacids
the this
of
ionicinter
coeffic
reduce
st
d
st
transfer ti ,part
UV respectively,
used saturationin which was observed determined at K(I)
(g/L)from
around = = protein U0V
2500 + (mAU1×
pulse
mAU. )/2450
2experiments 2 × (mAU.L/g) performed at
2.2 their
Modelinglocation [11, 12]anand attempt some to
amino
K(I) describe acid
= a the
residues
+ a variations
× I may
+ a
6experimental ×
be of I the
poorly lumped accessible
Purification Henry becau coe
of
same conductivity as Buffer A. The parameters, • where
strength Run
Hydrodynamics
Simulations I1only (reference):
observed the
All pK and
simulations 23%
experimentally
values BHenry
kinetics at and
presented pH To inin
number 0
the
simulate
this of1
bliterature:
article accessible
the were
2
aspartic/glutamic data acquired
2 the acids
byfitting
GSK
coreflowrates
of the strength under
protein, KThe Inonisobserved
thebinding
material
thus lumpednot a conditions
eluting
experimentally
contributing K(pH) K(pH)
(van
from coefficient
=Deemter
K(I) to b=
the
in K(I)
0thethe
+
= HIC
0b + and
plot). bcolumn
1literature:
0= ×
apparent
a2500 + a0×a+
adescribe
1 pH 0pH
1a×
,+ aaperformed
Inet
11b+
+,was
× a× 2,pH +×
baI2charge. b0apH
collected
× ,Iwith
2 2b×
2 1 To
, Ib22 in software
are
2 4.7
reduce in thethe
mL Yp
pH of the crude was then adjusted were pulse
inparameters,
HIC processes 7.0
the UV
considered
Regarding present
experiments, saturation
the isthe
study,
as
data adjustable
extremely that
the
awas mixing
were observed
lumped
high
cell
parameters.
extracted model
Henry
(several at from was
around
mol/L)the used
coefficient K(I) to
literatureand = can
6mAU. [14,
0
itb SEC
6 +
be
2
the
a
15] ×
to
calculated
isaspartic/glutamic
known 1
2
hydrodynamics
I
perform +
that a 2 × I
preliminary
from
the the
presence mea
co
teo
column
only volume)
pK a values and and stored number at -20°Cof before
accessible ×a× analysis ,× to determine
2 acids and
where Linear KThe is(low
Driving the lumped
Force Henry
approximation =K(pH)
coefficient was used and = abpH
to 0+
adescribe,+ aa 1 ,× pH
I2the,pH b(anda+ ,22, bb×
0kinetics Ibare pH are fitting mo
22of internal mass
to reach a target value (6.0, 7.0, or model,
where
impacts asThe• follows:
theKRun
only
activity
is
values 2the
the peak ionic
material
coefficient
lumped
of pK positions strength):
[8, eluting
Henry of
9, were
K(pH)
the
10]. 18%
from
reported
protein
coefficient
Besides, the
bK(I)
0in +
in HIC
and
it the
b =
solution
is
1 ×a column
original
clear , 0
0a + ,a
was
that 1b2was
articles
2 , computed
b
the +
, collected
b spatial
12not
b [Eq. in
the
using 2]
4.7
full
2fitting
configuration the mL
chromat
Trues
mode fra
were considered and asHMW. adjustable K(I) 0 =can 1
a0is  + a1Jparagraphs×0 Ithat 1+ 2
thea2characteristic 
× I the
andparameters.
The key model a lumped
parameters are theat number of and mixing cells and time
8.0) with the addition of 6 mol/L inpulse
HIC pulse
as 2.2experiments,
processes
Therefore,
follows: B
experiments,
where atonly
Modeling
column pHisK extremely
the7.0 volume)
the
is thethe
part
thermodynamic
lumped lumped high stored
Henry Henry
(several parameters
Henry coefficient
mol/L)
coefficient
K(pH) -20°C
K(pH)
coefficient described
= before ε0e=V
bcan + and bit
b0in
be
col SEC
+× bethe
0b ,1calculated
known
calculated
a pH a×analysis
1+,pH − b22pH
apK ε,+ e× from
bare to
pH from
,theabove
2 b×
0bexperiments
2 presence
b2were
determine
,pH
the
2 measur
2 Purifi
measured
are theso
of
estima
fittin
impact
whereThe
as follows:
not its
K
the apparent
transfer
activity
is the
hydrodynamic t
lumped ,
coefficient net
respectively, / charge:
Henry
kinetic of which
the it
coefficient
parameters. has
protein were beendetermined
in
and t reported
K(pH)
solution
=a , a from
, =
was a1that b , +
protein
1computed
b + ,the
b 1
b ×
, pulse
b a using
K of
+ b amino
1
fitting ×
the pH acids
perform
Truesdel
model d
pt
potassium hydrox ide or 1 molimpacts the values of pKaionic [8,the 9, 10]. Besides, itRcoefficient
is clearQthat thebespatial  configuration ofme
i 0 1 2
/as
L follows:
their 2.2 • flowrates
pulse
location Run
and
Modeling [11,3HMW.
experiments, (high
under 12]part and non strength):
some
binding lumped amino 28
conditionsHenry acid residues
(van
0

Deemter
1  can
may
2

plot).
0
be
1
calculated
ε
poorly 
2
accessible 2from the
Purificat
becau
pulseas
impact itsfollows:
experiments,
where where
apparent K the
is K the
net7.0 is
lumped the
lumped
charge: lumpedHenry Henry
it has Henry
coefficient K(pH)
coefficient
been coefficient
reported ε
can =
e V andbbe
col 0 +
thatand b
calculated
a ,
1 a ×
thea 1 , pH
, − a a ε ,+,e e a
from
b b ,2 , ×
b b the
,pH
, b b , b
measured
are 2 are
fitting fitti
retm
phosphoric acid. asthe %protein,
follows: Bwhereat pH
Regarding Kthe isdata the that lumped Henry
were textracted tRto ε=ecoefficient
K(pH) Vfrom
col = the A 0z 1and
b0kliterature
+1√
+ + 2 −
b11Iε×
0
aepK 1
2pH
,[14, a
aK 0of
1+
2
,15] ab1amino0
22, ×
2pH1
bperform
0 , breduce
acids
2
1 , bprelimin
2 are
depefi
k
QbIto 
core of thus not contributing the apparent net 0K charge. To the
as follows:
their where
pulse
Simulations
location pulse V[11,experiments,
experiments, is
12]All the column
simulations
and the some the
lumped lumped
volume,
presented
amino Henry Rε
acid ine=this
log(γ
Henry the k)
coefficient
residues =
coefficient
external
Q
article − 1were
may can can
porosity
be √poorly
performed
be ε,be e+
calculated
 Ccalculated
and kwith ×
accessible the
from
the from
flowrate.
softwarethe
because the
measu
Ypso m
SEC analysis. The SEC analysis parameters, •
in HIC Run
where
pulse
model, col
only
where 4 K
processes (low
the K
is
experiments,
only thepH):
the
pK
is is
peak
the
lumped 23
extremely
values % the
positions
lumped B and at
Henry
lumped pH
high
were
Henrynumber
coefficient
ε Henry
(several
reported
V Q
of
coefficient in
A
εand
coefficient
mol/L)
accessible e1 V
the
z 1+col
2−
and
a ,
original
B 0k
Iε a
ε a
and can a
e1I it2 is 0known
aspartic/glutamic
, a
1 −
,
be
articles
, a
b ε
,
,
calculated
e
b (and
1
,
,
b
b
,not are
2that the pre
b the
acids
are
fitting
from full thch“
and
fitti
core of asthe asThe
follows:follows:
protein, ionic thus strength nota I is
contributing defined as:
to e
the t
col =
apparent k  net e
10 + charge. 1 2  K 0
To[Eq.  1
reduce 2
was performed using an ultra-perfor- where pulse
impacts V6.0 experiments,
Therefore,
as isthe
follows: the onlycolumn the the lumped
volume,
thermodynamic tR e=
εlog(γ
Henry the k) =−R
coefficient
external
parameters 1 +described k
porosity
Q can √ be K
+ C
calculated
and
1inionic
 the × Q Ithe the
eparagraphs
ε5ethe from
flowrate.  thethe
3]above mea
were nu
Cvalues of pK [8, 9, 10]. Besides, εIit isis clear isthat spatial theconfigu
k ε1 −
eV εeflowrate.
where V iscol the εe1porosity
Vcol 1col −
were where
inconsidered
HIC pulse
A ,the B
kprocesses kas ,column
experiments, adjustable
k is are volume,
3 the
extremely fittingparameters.
alumped εhigh the
eparameters external
Henry
(several Q coefficient
tRand
mol/L) εand can
the and it be Q calculated
known strength εthat from
as the m
defined
presen
=I+ =Bkε I+
col
mance liquid chromatography (UPLC) parameters,
impact only
not
• follows:
asThe
its Runionic
apparent 5the hydrodynamic
(high
strength
netthe pK
charge: pH):
a values I 23
itεBesides,
has is /%
been and B atnumber
kinetic
defined reported pH parameters.
as: tR of = accessible
kthat the 1eV
easpartic/glutamic
col1 z 2+Ck K1 −K e acids and hi
The ionic
impactsactivity
as the
where Vstrength
follows: A coefficient
is
, Bthe k , Ivolume,
is pK
column
Cof defined
of as:10].
protein
volume, in solution
the tεit
external QV= wasQ
 2and computed
porosity
that 1−
1ε5the +and eusing
εεeflowrate. QpK the
 Kthe
a of amino
Truesdel
flowrate. a
parameters Bthe Ckvalues were a [8, 9,
εestimated eby comparing is clear model the
esimulations spatial configurati
as with exper
R 1K
system (ACQUITY UPLC Protein where
werewhere V col Aisk ,8.0.
considered
their kas,column
col
location adjustable
k are [11, 3k 12] parameters.
fittingand the
eparameters
some external
amino and porosity
where
acid Ie is
ε col
residues the
V is Q 2+Q
ionic
kthe may strength
lumped 1 −
be ε εpoorly
e Henr defined
y
accessible in
asThe follows: The ionic t = I 1= 7 e col 1z C K e
impact its apparent is,strength net Ivolume,
is
charge: defined itεinhasas: been reported kthat theand pK of amino acid
R 2 k
ionic
activitySimulations
Vstrength
coefficient isAll
ICthe simulations
defined
of the as:
protein presented e Ithe
solution int=Rthis = was article computed k, 1asimulations
were
Cporosity
zkapparent +performed K Qwith the softwar
where where V is Bthe column column volume, oftheεspecies
external externalQ porosity ,and aε2e,usingbQ the ,the theTruesdell-J
flowrate. flowrate.
Athe 2the a
BEH SEC Column [200 ˚A, 1.7 µm,parameterstheirThe
corewhere process
ofcol
where
location
k ,col
C kVwas
protein,
kis
[11,
col
k
the
is
were
monitored
12] thus estimated
concentration
the column
and bynot
someonline eby
contributing
volume,
amino
comparing
coefficient
εe1the
acid  tok external
2residuesin andmodel
k1 a
solution
Q√ 0 porosity
may 1 2be and nete0z, kand
εpoorly its bwith
b1charge. Q are experime
valence.
the
2accessibleToflowra
redu
be
as Interactions
follows: The ionic with the solid phase The interactions with the solid phase w
4.6 mm x 300 mm] on an ACQUITY core pH,
where The
online
parameters,
In
of V
the
the The
col isstrength
ionic thestrength
conductivity,
only
literature,
ionic column
the IpK is
and
results
strength
defined
Ivolume,
isonline
a values
defined
Iare
as:
often
is of e Ithe
εand
defined
as:
fitting
= external
number
presented
as: A
Izof=
model k 2
Iparameters.
Cporosity
zk1k2in accessible
 k2terms 
zkofC and k  Q For
aspartic/glutamic
kand ,charge.
whichthe pulseflowrate.
isToflowrate.
related act
where
following where C kVprotein,
adsorption is the
is the thus
concentration
column
isotherm: not contributing
volume, species
ε the 2
−to
k inthe
externalksolution
√I =√ apparent 1 porosity 2 Cand net z its Q valence.
the reduce
UPLC system, Waters, Milford, Mass.).Interactions
ultraviolet
where is(UV)
Ckconsidered thecol
with absorbance
concentration
the Iatis
solid 215 nm.
of
phase log(γ
species )experiments,
kThee =
k in I1solution
interactions
k= theIand + 1C
zklumped
with zkzkk2× itskI2Henry
the k
solidcoef-
valence. phase were
In The
were
parameters,
the The
ionic
literature,only
ionic
strength
theasresults
strength
adjustable
pK a values
defined
Iareis often andas:
parameters.
defined number
presented
as: Ain k
2
zof
+ inaccessible
B2terms
Ik1I 2 k ,εCztwhich
= kkof , kaspartic/glutamic Czkis is relatedacids to K
The eluent was 10 mM sodium phos- In
following
where
were T
the h e
where
The
Ckconsidered f
literature,
adsorption
is the o l Cl
activity o w i
is n
results
theg
isotherm:
k as coefficient
concentration c a
arel i b
concentration
adjustable r a
often t i
log(γ
of of o
species n
presented
parameters. of
thek )protein
= f
k in icient
species
−Ksolution can
kk in
terms
= εIsolution i= √
+ (1 be of
solution
andk +− calculated
1 C 2kki2)×
zεwas
k
z Citswhichand
Icomputedk
valence. from
kk  its the
related valence.
using totheK Ta
phate with 400 mM ammonium sul- A
where(Equation , B , C are was 3 fitting
established parametersin and 1 + C I B
kis = I λ
the
I
= γ C
ionic k
kεz 1 k k2−
strength
C ε  as defined in
In theasThe k In
follows:
where the
k
literature,
where Ck is
activity
k 1)
literature,
Cresults
the
is were results
areconcentration
often
concentration
the
coefficient of the are often
presented
of species
protein measured
of presented
in εterms
K species ki +in(1
k retention
2in
of
ksolution
in
k terms k
2ki ),εtsolution
ksolution whichtime
and t of
k and kkR as
ztis
t , zkshown
which
related
its valence. is related
tothe
its valence.K True
as fo
parameters A ,
where B , kC
C is the estimated
concentration by =
comparing
of 7solution
species − εwas
model k in
computed
simulations  kk and
using
with
z itsexperime
valence
fate at pH 7.5. The f low rate where
was terms
A
where , B of
Ckthe
Inwhere ,
Inandtotal
C
k k C are
k
literature,protein
3k
k are results
k concentration:
fitting parameters
the results
concentrations
are areoften K = in
and ε C +
Equation
I
ofpresented
i
presented is(1=
the species
k −
λ
the k γ
ε i
in solution )
4:
k
termsC
ionic k 1
k εofk −
in
√ k
strength ε
kthe
,εtwhich liquid
 as defined
kand in
issolid Eq
toph
 is related
askfollows: the literature, often tof
2.2.2
k
Aand InkC,the
, Mechanistic
B kC is literature,
the concentration
model results of
are species
often kmodel
presentedεεin Aεin1t − terms
in
zkk2εtermst and
I and zkof ,itswhich
kzkkand
kwith valence.
,itswhich
related
is rela
parameters
lumped kwhere
approach kwere
Ckare isisthethe
used,estimated
concentration
meaning by
K comparing
that
= εof +theK
species
(1 =
contributions
− iik)+in(1 k−
simulations
solution i )from the experimenta
species valence.
adsorb
where CkIn thewith Cliterature,
k concentrations
resultsphase are oftenThe of
i the
log(γ
presented species
k) = − in 1− k
terms in
√ the√ ε1t the
of liquid
k− 
+, εCwhichtk × I is
solid
related phasto
Interactions
2.2.2
located inMechanistic
Inthe the pores theofmodel
literature, solid
the results
particles are often interactions
presented in 2εterms
twith ofCkεkkt.solid
,εtaken  phase
which were
is related
lumped
2.2.2 Solution
approach
Mechanistic chemistryis used,model The
meaning thatare
following theK averaged
=
reactionsK εi=+ε(1
contributions A together
in
i
− =coefficient +k1
− z+ kεiB
solution
(1 √
)− Ik εin
from i )Iwere
the The
tk
species activity
 into co
accou
adsorbed
following
www.biopharminternational.com Interactions adsorption
with of,isotherm: log(γ k) =
April K
2023 BioPharm εiwith + (1 1 the− +ελ ε1tCtisolid
k) ×εtI phase
−quantifies
International 23 k the
2.2.2
the
located non-ideality
inEquilibrium
where
Solution
2.2.2
Mechanistic
the A , Bthe
kpores
chemistry
Mechanistic kmodelCksolid
the
of the protein
areThe 3phase
particles
model
solution,
fitting
following areThe
pKreactions
parameters interactions
a while
averaged
K = εin
thetogether
iand + (1IB−is
1solution
7+ k εin )Iwere
ithe
C k .kεtaken
ionic The
1tk − εactivity
strength were de
ads
ascoeffi
tinto account: defi
following
Solution
the adsorption
chemistry
thenon-ideality
chromatographic
H O  of isotherm:
H +
the + The
medium.
OH
protein following
− The
solution, reactions
parameter
14.00while Kin
the =solution
λ ε + (1
iwas estimated
coefficient
k −
were λε1 i− ) takenε
quantifies
t by k
into account:
comparing
the adsorp m
parameters
where Equilibrium
2
AMechanistic
k ,+B
Akk, Bare
k, C k, C were estimated
3k fitting pKa by comparing model k 1− simulations
εt as with ex
2.2.2 2.2.2
Equilibrium Mechanistic model+model pKparameters C k and = λkIγkisCthe k
ionic strength defined
eserved on ingmodel part at around µm, 2500 4.6 mAU. mmx300 mm)inon anwas ACQUITY thus found UPLC and system,
the results both netare were regressed purchased based from
on reported Waters
linear orduring (Milford,
quadratic MA,
expressions
owing re
experiment.
nterest
and
gressed
nry number
valuesin 2.2
based
coefficient
performed using wereon
with
Overall, statistics
ofModeling aaccessible
linear
obtained:
and HiScreen a
five
of
or , part atoεexperimental
concentration
6quadratic mol/L describe
e,aspartic/glutamic
Butyl =a 0.4
, b
HP ofexpressions.
and , theεruns
2potassiumb
column ,
the
impact
tb2 = are0.93.
were
crude
(7.7x100 of
hydroxide
acids
Infitting
performed:
2.2
this
The
respectively
was
key impact
Purification Blue
mm, and
study,
internalstrength
0.64
operating or
were
model 4.5
dextran
Modeling
column 1equal
its
histidineg/L. of
Bed
quadratic
Run
mol/L
used
apparent
porosity
[6,
protein
Iparameters
parameters. volume
Size to 17].
observed
Volumes and inof400
ε(reference):
residues
an
The
iExclusion
part
=
expressions
by
phosphoric
acetone
4.654 attempt 1−εHIC
(BV), following
charge:
experimentally e
For
=(2)
pulses23 0.89.
Chromatography
acid.
to
εtPurification
%
it values
Loading B
has
have
describe in were
been the
with
been
ctheof(g/L)
obtained:
literature:
(SEC)
Crude
performed
variations
protein =7.5. Uby V
that
analysis εeto=
HICof the
(mAU
the 0.4
approximately
estimate pK
revealed
lumped
)/2450
anda ofε tami
the
(mA the
He
=
erization
un ng
K(pH) 2 from(low = theb
ionic K(I) + HIC b
strength): = USA).
× column
a pH +
0 The+ a
18 1 b was
×
% eluent × BI2 +
pH atcollected0 awaspH × 1 10
7.0, I mM
2 in
Run sodium
4.7 3 2.2 mL
(high was
their phosphate
fractions
Blue
ionic
Modeling thus location were dextran
strength):
found part with
used (corresponding
[11, equal and in
12] 28 (3) an
to mM %acetone
and ε attempt
B ammonium
= at
some (2) pH −εto
pulses eto the
7.0,
amino =
K(I) sulfate
have
describe
0.89. =acid a beenat
the
residues
+ a pH performed
variations
× I may+ The
a ×
be of flowrate
toI 2the
poorly estimate lumped access
Puri th
aal
= smpt ε model
(e.g.,
have
hight
rameters.
Uppsala,
−ε e been
to =purity,
interest 0.89.
describe
Sweden) yield).
0performedin 1the
K(I) using
using = 0Typically,
toa
variations statistics
an
presence
estimate +The 1
ÄKTA
2
1crude
acalculated ×
ofato of
Design
the
avant Ithe +three
describe 2to external
a25
lumped be of× groups Experiment
purified
(Cytiva) Ithe 2 and
Henry ofsupported
impact wasspecies
total (“DoE”)
porosities,
of
obtained
respectively
coefficient 2denoted
key
strength by mgoperating Unicorn by
with is[6, Low
Iperformed/mL
cell observed7].
pH culture
VVV.
Molecular
parameters
The
and , (3) ifollowing and
experimentally
ionic Isocratic1−ε Weight
pre-purified
e values
(2) (LMW),
Elution inwere 0 with
by
the two Monomer
1 77%
obtained:
bliterature: chromatographic Buffer (Mono,
2 εe + A
=50-500 + 0.4 the
23% andsteps.target
Buffe
2 εt =
d(several ical
tored
H):
terization 1−εemodel
Henry 23 at% coefficient
mol/L)
-20°C
B at pH was
and
before 6.0, can 0.3
it Run 0isbe
SEC

mL/min. known
5 (highanalysis that
pH): The 2the
to from
injection
23 determine
2% presence B the at measured
volume
pH core ofthe
8.0. of was
salt
respectively strength
content
the 10 mono
retention
significantlyprotein, µm [6,
of I with observed
bed
7].
LMW, thus time a The monomer not t Rfollowing
experimentally
Monomer contributing K(pH)
concentration
values K(I) to in =
were the
the= 0ina + b1 a×range
the
obtained:
literature:
apparent + pH
× I net ε
+ 2×
bacharge.
= × 0.4 pHI 2and
massε3
To
dvalues
based
pulses
antities
experimentally
acterization 2.2 K(pH)
were
onused
have
(e.g.,
Modeling obtained:
linear
been =
purity,
in bthe
orperformed
0part +
quadratic
yield).εeAll b1statistics
literature: product) 0.4
the pH
to
Typically, and
expressions.
runsestimate +to and εbtreported
2describe= × High
aare pH
0.93.
the
Design Inexternal Molecular
in The
this2.2 of
this internal
study, 2.2.1
ExperimentModeling
and
study was Weight
quadratic porosity
totalthus Statistical
with
were (“DoE”) (HMW)
porosities,
found
UV 100%
performed part
expressions saturation ABuffer
equal model
is accounting performed withto BPurification
εiduring
was (3)
the=observed ε for −ε
same
t 38%,
e BV,
=estarting
of0.89.at74% (5)
protein
around and
Stripping
material.
by2500 14%
0HICmAU. of
with
1 The theWater totale protein 2 protein during
eotocol .
ess oefficient
ea protein
high
Besides,
was
was interestin
monitored and
it is
for
solution ainclear 0 µg/mL.
, using
by a 1
reference
, was
that
online a 2 ,
Dilution computed
b
the pH,0 , experiment:
b spatial
online
1 , of b 2 the  using (1)
sample
configuration
conductivity fitting Equilibration
the
the Truesdell-Jones
model
impact
was and performed
parameters, of was
online
with
ofparameters.
the key thus Buffer
protein
UV operating
with only found model
absorbance 10 the may
during
For
mM
equal parameters
pK [13] potassium to at
values ε 215 =1−εεet −ε
and nm. phosphate
number = 0.89. ofat pH
accessible 7.2. aspartic/glutami
V), remely
to ee
tly
lowing =
egressed (2)0.89.
a high
describe
pulses high
LoadingPeer-Review
valuesbased
have interest
2.2 were
(several
the with
been onvariations in
obtained:
linear
Modeling Crude mol/L)
performed using Research or
concentration content,
during of statistics
quadratic
part ×
andεthe eto=

it
approximately respectively.
0.4
lumped
estimate is expressions.to
and
known in × describe
athe
εthe
Henry t = that
crude 20.93.
5external The
BV, the
the
coefficient
In was
whichimpact
conductivity
Thepresence
2.2
this
Purification and 0.64 internal
study,
roughly where
with
A
total ofand
Modeling
g/L. ofkey quadratic
of porosity
salt
pH of
K
Size Buffer
corresponds
porosities, The operating
protein the
and is
significantly the
Exclusion materialcrude
ionic
B part
expressions lumped
wasby toparameters
a was set
HIC
i adjusted
Henry
7.0.
Chromatography
eluting The
1−ε
Purification
K(pH) from coefficient
e flowrate
= with
the
bK(I)
0+
a(SEC)
HIC
of buffer
and
was
bprotein
1= ×columnatopH aanalysis
01.16 0containing
+ ,by+ aa11mL/min, ,was
bHIC
× a×I2revealed,pH + b0a,20.4
collected ×1 ,mol/L
bwhich Ibthe
22 are
inco
quantities e K(pH) (e.g., purity,
ε = V b + yield). b 1 pH
Typically, ε + b pH
Design of Experiment (“DoE”) is performed (3)
ry εcoefficient 2.1.4 can Column be calculated characterization from the measured There retention
is currently time a high t interest in using statistics describe the 2impact 2o
Henry
ntg tone
tempt K has
=
mentally t −ε
calibration
[8,been
aquantities
coefficient
pulses
9,
to = 10].
in have
K(pH) was
reported
0.89. (e.g.,
the Besides, e
literature: and
established
been col
purity, = 0 performed
that b
it a 0is ,1
yield).
potassium
+ a
the
clear b 1in , × aterms
pK to
Typically,
that2 pH , e b
estimate
monobasic
0 +
the of,
of
2 b b total
1 ,
amino
spatial ×a b the
2 Design
2pH are
protein
2.2.1 external
2
acids
phosphate fitting
configuration
in of HIC concentration:
were depend
Statistical
Experiment
pulse and model
considered
and
processes
cm/h. total of 2
experiments, strongly
mol/L
the parameters.
porosities,
model
The (“DoE”)
is protein asextremely impact adjustable
ammonium on R
the is may performed
lumped
of For
high solutionparameters.
sulfate
(3) Henry
(several pH to reach
coefficient
mol/L)
and K(I)
ionic theand = same
can
strength it a 0 is +
be
2
a
conductivity
known calculated
1 ×
(modulated I +
that a 2 ×
from
the Ipr
than
by
)ollowing Isocratic
1−ε
regressed e tdescribe
Elution
values
R based =2.2 were on with
K(I) the
Modeling obtained:
linear 77% variations
= aor presence
0 1 +
Buffer
0+ 1
quadratic
part ε of
ae1 × I +expressions.A
= of + three
a
0.4 theK 23% and lumped
a2 × tI groups
Buffer
2 ε = B Henry
of
0.93.duringspecies
Inon The2.2
this coefficient
some 20 denoted
BV,
internal
study, Modeling
relevant (4) column quadratic with Low
Regeneration
porosity quantities
pH Molecular
volume) partand
expressions ionic
(4)
(e.g., and
(2) Weight stored (LMW),
Purification
purity, at
K(pH)
yield).-20°C Monomer
=
of before
b
protein
Typically, 0 + b 1
SEC × (Mono,
by pH
acomputed HIC
Design analysis + the
b 2 ×
ofare
target
to
pH
Experi dete
2
ed et following Henry values
coefficient were
2.1.3 obtained:
can Hydrophobic εee The = 0.4 and
Interaction εof t = 0.93. The
Chromatography internal porosity experiments
mino areHenry
ttempt
regressed
charge:εtacid
experimentally
iduring
= −ε3
2.1.4εetto−ε =it
ε
residues
Blue
coefficient
e BV,
V
based
has
0.89.
edescribe
(5)2.2
Column
 been
dextran
in Q on
the
Stripping and
the
may
Modeling linear
reported
literature:
product)
characterization
1 2variations
√ a0be
Buffer
and
− ,be
with
ε
or aU poorly
acetone calculated
1part
ε,A.
quadratic
that 
Water and
a(mAU 2 , the
of bHigh
the
accessible
pulsespH
0during, )/2450 pK of from
bexpressions.
1 ,Molecular
lumped ab the
have
32There BV. arethe
aminobecause
crude
2.2 impacts
been
Henry and
measured
where
2.2.1In
isWeight
fitting
The was as
acids
Modeling The
2.2
this
performed
pH
currently
thecoefficient
2.2
the K
“buried”
then
follows:
thismodel
of
results
activity
Statistical
study,
depend (HMW)
the is
values
Modeling and retention
Modeling
adjusted
reference the
a crude,
part high
are
quadratic
to
parameters.
with HMW. coefficient
lumped in
strongly
of estimate pK
accounting
regressed interest
part pH
model
the
to
Buffer experiment. time
areach and [8, Henry
part
expressions
on9, ofta10].
the
inionic
based ForRthetarget
for
using external protein
coefficient
Overall,
on
8%, value
Purification
Besides, 74%
statistics
linear
andinand(6.0,
it
five
or
solution
andisto
a 7.0
clear , aporosities,
of
experimental
total
quadratic
0 14%
describe
1 ,that
or
protein was a8.0)
of 2 ,the
expressions.
bwith
the
the
0by , runs bspatial
protein1HIC
impact
,the b2were  using
addition
In
configu
mass
of
fittin
perfo
Purifikey
this
th
2s
nand ed o ε
attempt
i =1−ε
Henry e to =
coefficient
col 0.89.
describe A
c (g/L) zthe and of I =
variations a
6 mL/min, e
Imol/L , V a , a ofwhich
potassium , 2 the b , b
lumped , (mAU.L/g)
b
hydroxide are
pulse as
Henry fitting
follows:
experiments,
or where
coefficient
1apparent mol/Lmodel K the iswith parameters. the lumped pH lumped and Henry For
Henry
ionic (1)
coefficient K(pH)
coefficient ε
can = V andb be + calculated
a b , a× 1 ,pH − a ε, + from
b b , b× pH
the
, b m a
ephosphoric acid.
some amino acid k residues k × 0 may 1 ×
be 2  0 1 2
IIpoorly accessible tobecause “buried” in the
K(I) = 0 1 2
ped
me, buting
terest
set t7.0.εlog(γ
Henry
d experimentally
R ein
1−ε
=the to
The
using
e
k) the
flowrate
respectively
2.1.4=
coefficient
external apparent

statistics 1aColumn
HIC
in + was
0 porosity
the
+ [6,
content, a1.16
experiments
to can √
1
literature: 7].
describe net be +KThe + charge.
characterization and C a2following
respectively.
calculated
kthe ×were Q impact the To
performed reduce
corresponds
values
from
flowrate.
of The
on
Table 2 some
key
impact
Purification thewere
with the
conductivity
There
I.were 2.2.1
measured
relevant
Reactions
operating ausedits
number
obtained: isof
aHiScreen
at fluid
in
pH protein
Statistical
currently
parameters in
ofvelocity
quantities
solution. an
7.0, of
the
retention εButyl
attempt
fitting
Run
=by net
crude a of 0.4 (4)HIC
(e.g.,
high (2)
2charge:
model
HP 150 (low
and
(7) was
time
to column purity,
interest describe t tit
εionic
adjusted = has
R (7.7x100yield).
in been
strength):
0.93. using
the
with The reported
RK(pH)
mm,
tTypically,= 18
ainternal
statistics
variations
e
buffer %=
column col Bathat
of to bat
porosity
containing
Design
0the + pH
10describe+the
volume 1 7.0,

blumped of pK 2pH e
Run
0.4
4.654
aK
Experiment
the
0+ofmol/L
Henry b31amino
impact 2 (high
×2pH co
umped
y f high Blue Henry
(several Q
dextran K(pH) ε
coefficient
mol/L) Vand = b
acetone
=kand ε + can b 1 × be− pulses pH ε calculated + b
have × pH been from as the
follows:
performed 2.2.1
pulsemeasured to experiments, Statistical
estimate retention the the model external (3)
time
lumped t Henry
and total coefficient porosities,
 Q can be calculated
ε  from
+a+it is× εknown that the presence of salt significantly √
rved notexperimentally inB the literature: ε −ε 2
us
tart d solution
n number
pH
contributingtwas of and
= thus atK(I)
accessible e1Typically,
ionic
mL, to+ col
found thestrength
GEpotassium 0
1aequal 0I
eapparent
aspartic/glutamic
Healthcare, 1 a(modulated
to monobasic e
Iinet =
K +Uppsala, acharge.
2t
2× by e
e and
the
=
Iphosphateacids To
0.89. %
Sweden)their
the B)
reduce
and
onresults
was
location
and then
the
Run
histidine
using 2arewhere number
mol/L
an 4studied
[11,
regressed (low
ÄKTA residues
K 12] ammonium ispH): around
of and the avant fitting
based 23 some
lumped % 25 on(4)(2) B amino
sulfate atHenry
(Cytiva)
linear RpHor toacid
6.0, reach residues
Runthe
coefficient
supported
quadratic 5Typically,
(high
by may
same
and
expressions. Unicorn apH): be
conductivity poorly
ε,aeaa01,Design ,23 a2%
eVVV. ,this bB accessibl
, at 1 ,pH
bthan b2
aIn study,
fined was as: observed around 2500 1
mAU. some relevant quantities (e.g., purity, yield). 0of Exper
l(e.g., purity, yield).  Design of Experiment (“DoE”) isIis performed εliterature:
Vstatistics 2−
ment. respectively
Overall,
R
Blue2.1.4 five [6,
dextran Q
experimental Column
7]. The and following characterization
acetone runs ε were values
pulses
1−ε
performed:
Purification were
have strength
obtained:
been
ofRun There
protein as 1 follows:
Equilibrium
performed(reference): ε by
observed
currently where
= HIC 0.4 to K
and23 experimentally
estimate ais % high ε the B = lumped
interest
0.93.
the inin
The
external Henrytheusing
internal ecoefficient
and col pK porosity
total A z1and
kporosities, to
0
Inet
describe 1 , athe 2,  bTo 0, b
impac 1,
K
umped
meters. in9, values
10].
using
theeluting
a
based Henry
external
on and
Besides,
statistics
linear number
coefficient
or 2.1.4
porosity to it
εIbK(I)
quadratic e0 V is
describe
1The of  clear
Buffer
Column and accessible
2.1.2
and following
expressions. the
a0Q that
A. ,+ 1aatheSize
5
The
impact −,was aspartic/glutamic
the
characterization
e protocol pH
aε×I2eflowrate.Exclusion
,In spatial
of
b0this of key
a,2b×
2 the 
1was study, operating
, were configuration
crude
b2defined
2 used
in core
Chromatography
are HIC
used acids
was
fitting
for of
parameters
where the
then
the
There pulse
processes
in and
an model protein,
reference The
Vattempt histidine
of
adjusted
e is experiments,
colcurrently the
is is process
parameters. the
extremely protein
thus
analysis to
experiment: to residues
column reach not
was
a high
describe t the may
high contributing
monitored
a
For target
lumped
volume, (1)
interest
(several
the tvalue
log(γ
Henryby
=
Equilibration
εlinear
einthe
variations
R to online
(6.0, the
)
k external
using
mol/L) = apparent
7.0

coefficient pH,
a 1or
with
statistics
and
orofquadratic the + itonline
8.0)
lumped
k
Buffer
porosity can √
isε expressions. with
to describe
known be K
+ charge.
conductivity
AHenry C
and the
calculated
during
kthat × addition
I
Q coefficie
the
the red
and
from
flow
imp
pre
parametersrial K(pH) from
and = the is+ε HIC
col b V the = × column
apH ionic
02 + 11bεstrength×
1t −ε − epH + collected
ε as Iionic in 2 quadratic
4.7 and on mL the
in in
some HIC expressions
fractions
results
Equation
pulse relevant processes are
experiments, (corresponding regressed
5.
quantities is (3) The extremely the based (2)
(e.g., tohigh
lumped the
on
purity, Henry
(several Q
yield). ε
coefficient
mol/L) V
Typically, and can a
it1 −
be
Design
is ε calculated
known In that
of this
Exp
ustable wel ge: ionic was thus
strength):
parameters.
ityield).
has t
respectively
been =
found 18
I coefficient
= reported% equal K(I) B
eof [6,1 at col 6Volumesto 1
pH zmol/L
7].=can + ε
itC
7.0,
ai The = +known Run2
acalculated
potassiumfollowing × K 3 = (high
Ipulses
e(2) 0.89.
+ coefficient
aLoading values
2hydroxide ×have Ipresence strength):
were
parameters, or obtained:
H The O
1performedmol/L 28 ⇄ only
+
H %
following +ionic
− B the
OH at
εstrength
phosphoric =pH
pK 0.4
calibration 7.0, values and (5)
acid. (4)
εdefined
and = (2)
was 0.93.
numberestablished The of internal
accessible
14.00 1
e
in terms +col B porosity e I
aspartic/glutamic
of total e
2 protein  ofco aE
+kthat the pK of amino acids depend strongly on
he purity,
nterest lumped in using Henry R Typically,
statistics 4.5 Bed ato Design describe 0is of be εExperiment
the
(BV), 1Henry eimpact (“DoE”)
of from key operating
the withisImeasured
performedas
Crude follows:
parameters during retention approximately a time ist10]. 5external
BV, whichtR = roughly k 1the +lumped Kto
ionic Ito as: Icorresponds
on some relevant eto quantities t (e.g., purity, ayield). Typically, aIDesign
mely high (several k
describe the Blue
variations Rmol/L) dextran of and
bthe and lumped acetone that the been of salt significantly estimate the and total porosities, 
tRun = 1−ε 2impacts with the pH values and of pK [8, 9, Besides, it is clear that spatial configur
me) s:
otein and in stored 2.2
solution at Modeling
-20°C was Q before
computed b1simulations
part SEC + eusing [Eq.
analysis ε2K
4] the to strength determine
Truesdell-Jones were observed
used the content
in model experimentally
an attempt of [13] LMW, in
Monomer
describe the (4) literature:
the variations of the Henry thc
2
ume,
mated odel
% Blinear atεpurity,
eby
pH the 6.0,
was external
comparing K(pH)
Blue
thus 5Typically, 2expressions.
dextran
found(high =porosity
model The 0pH): equal SEC
and 1 23 × and
to pH
% analysis
acetone ε B Q += at bthe tpH
a −ε× was
pulses with flowrate.
pH8.0. = experimental
performed
have
0.89. and been impacts the using as performed
results follows: data.
thean ACQUITY
are
values to
regressed a
estimate
of pK
R
(3)
UPLC
abased [8, the9, K(I)
Protein
on external
10]. linear =
Besides, BEH 0and
or +quadratic
εait
QSEC1total
V ×
is clear +
Column a
porosities,
expressions.
2that × 1 ε5the −e (200theε Å,
spatial In 1.7
d ent Henry
on
(e.g., of the or protein
coefficient quadratic
yield).  in and
2residues solution
mgis2mono a Q ,
[6,/mL a , was
a a Design , In computed
b , (3) ,
this ε b of ,
study, b
Experiment
Isocratic are e
using quadraticfitting
Elution the
where
were (“DoE”)
whereTruesdell-Jones
model V
considered
expressions
with NH A is
+ parameters.
is
77% ⇄ the
performed
, NH B as
+
Buffer , column
+ C
H adjustablemodel
ke A are For + volume,
3 [13]
23% fitting parameters.
Buffer ε the
parametersB external
during and porosity
9.23
20=internal I is 1 the and ionic Q strength flowrate. as
IBV, ε(4) Regeneration
k i and the results regressed e col e
ome [8, 9,amino
entally 10].ininthe acid
Besides, literature:
1respectively
respectively it clear 0 may
1 7]. that
conductivity 2 The be the 0 poorly spatial
followinge 1 2
1−ε accessible
econfiguration values impact
2anstrength
were because
its col of
obtained:
apparent the
Ikobtained: observed k“buried”
protein εnet =may charge: 0.4 in and the itεhas tbased
e = in 0.93.
been on reported
linear
The or=quadratic that VWaters porosity
the pK expressions. a1of −K amino
εHenry Ina
interest using statistics to describe the impact of key operating parameters εIexperimentally the literature:
tbeen 1zk2porosity
+C
bed 4 3
efined gh monitored interest as: by
Ithe = online
incoefficient
using K(pH) pH,
statistics zfound C online
= µm, bbe to [6, + 4.6 7].
describe mmx300a× The − the followingmm) and
impact b× on
online values
of ofkey UVACQUITY
were
Purification absorbance
were
impact
operating used itsinUPLC of
parameters an at
apparent protein 215
attempt system, (5) nm.t= by net 0.4 to both
HIC and describe
charge: were
εt =itεpurchasedcin0.93.
the
has
(g/L) RThe
variations = from
U internal
reported
VQ2the eof
(mAU col the that lumped
)/2450 k+ (Milford, the pK
(mAU.L e MA, of
mpedhigh lume,
ibe (several
the Henry εon
variations mol/L) external of and the kcanε2.1.2
it lumped Vpart
porosity iscol known 0b+
calculated Henry and11,pH
that εQ
the +
coefficient efrom 0b the
presence the pH
,eflowrate.
with measured salt
pH
The significantly
and POretention
ionic
activity ionic strength coefficient time is defined
of the (3) as:
protein solution t2.15 = was computed 1simulations eusing
εlumped K the floT
d lumped based Henry linear coefficient
or quadratic and 0expressions. a ,1Design aεpK ,In bExperiment
this
ε ,2t −ε study, bBV, are quadraticfitting where model
expressions V A isparameters. the ecolumn For volume, the external porosity and Q the a
harge: contributing 2.2 it e has 2.2.1 Modeling
been
was to thus Statistical
with reported
the 100% e
apparent  k that
equal Buffer model
Size the to
net BExclusion
× during
=
2charge. of amino
1 3
2.2 × =Chromatography
To 2their 0.89. 2parameters
(5)
acids
reduce
Modeling location Stripping
were
H depend the part ⇄used Hperformed
[11, −
PO
number stronglywith
+
in ,
12] +B
analysis H an ,
and Water C attempt
of on R were
some fitting during estimated
to
amino
K(I) 3
describe BV.
=acid a eby Thethecomparing
residues
+ a pH
variations
× RI of may
+ a model ×
be crude,
of I 2the
poorly Buffer accessible
Purificat wit
Hen
es
ation ion (e.g.,
volume, ofwas purity,
species where
εe the
established yield).
it2was
k t
2.2 in
VexternalK(pH)inis
= Typically,
solution 1
Modelingthe
terms = column
porosity b
of2lumped 0 aand
1
total part+ b 12volume,
z protein and i pH of
its a K Q+valence. b εthe
concentration: 2 −ε pH
flowrate. (“DoE”) 3 4 is 2 col k k k (4) (3) Purification
0 1
1 of protein
Q
2 by HIC ε
ofUSA). The eluent was e10 mM sodium phosphate with 400 ismM ammonium sulfate at
colthus found equal kto εsolid = 1−ε = 0.89.
strength their Iand
location observed [11, experimentally
12] inas: the literature: kpHmay 7.5. The 2flowrate
a√ Iand some amino acid bresidues be poorly ac
R t e e
9, ities10].
yphase inlumped (e.g.,
the Besides, purity,
literature:
The ayield).
is
interactions clear Typically,
that the with spatial abe Design the configuration i of Experiment phase ofthe thewere (“DoE”) protein described is may performed
ethe
defined oed
ues
e
describe
some
lumpedbased
often and as:
aminoK(I)
ε
on the
Henry
number
presented
Henry
2.2
linear
the
variations
=
acidexternal 0IA
or
coefficient k coefficient
of =
+quadratic
Modeling
residues
2
inand 1×
accessible 
porosity,
terms
Q
BufferIthe and + may can
zexpressions.
a k2of2√ C B ×part
0beand
aexpressions.k ,was Ia
aspartic/glutamic
k , poorlycalculated
1which
ε, 
Q
Henry
esetathe 7.0.
2 ,Inbaccessible 0this is
coefficient
,in1−ε bthisThe from ecore
study, as
,injection
related bstudy,
2 are
follows:
flowrate with
because
of
quadratic
acids to
measured
the
H
fitting POwas
strength
where
K
pH
The

and protein,
as “buried”
⇄model 1.16
HPO
expressions V ionic
2− Icol
histidine
follows:
+ionic
retention (2)
observed mL/min,
+thus
is strength
parameters. Hthe inbythe not column
residues
thetime which
experimentally
contributing (5) tFor defined
Rvolume, K(pH)
corresponds K(I) to in =Ithe
εspecies the
= b= 0to + a+
literature:
apparent
6.82
a2500
external 1fluida× 1z
2pH
k× Cvelocity Iknet
porosity ++ the 2×
bacharge.

of pH
andIz2150 QTo theredu fl
e: is
essed
the
mentally defined
it based
has
lumped
2.2.1 been
in the
e as:
There
on Henrylinear A
reported
literature:
Statistical isε
k e z currently
V or
coefficient
k Thecol 2 I
quadratic
that 1and was
model A the SEC z a0.31
and high

pK I ε
mL/min. aI0K
analysis e interest
of , charge. amino
a1Henry ,solution
awas 2, 
The
In acids
bTo 1using
, bpH
0performed depend
1 , band
statistics
are
2
volume
core
quadratic UV
strongly
4
where
fitting
using whereof to saturation
an
4
the describe
wasexpressions
C
model on k Vwith
ACQUITY protein,
is 10 the is µm
parameters. the
was
the concentration with
thus impact
observed
column UPLC a
not Forof
monomer volume,key
at
contributing
Protein of around operating
e concentration
εBEH the 0to
k SEC in parameters
the
external
2interactions mAU.
solution1 apparent
Column  in porosity and range
(200 net kand charge.
its 50-500
Å,solid Qvalen
1.7thp
not
umn f , tolog(γ
the
speciescontributing
describe
lumped
volume, )f =
low
k c the
t
2.2−
in ε =
rate.
(g/L)
log(γ
Henry solution
the to
variations cm/h.
Modeling= the
) U
external=
coefficient √ V −apparent The
of
(mAU1 k +part+ k
the C  z
porosity k
impact )/2450
can 2 √×
lumped net
its
a
be + of
valence.
and C
(mAU.L/g)
calculated × Q I the coefficient reduce
parameters,
from
flowrate.
2Interactions
the the
with ionic
HPO measured number
2− pHonly
The strength
⇄ PO and
3− the
ionic + col
ofionic
retention
+ H pK (modulated
fitting
strength (7) (4)
the values
(1) (7)time solid I and ist by phase
defined the
Purification
number % B)
as: The
of e wasof protein
accessible
12.38 Ithen = √ studied by
aspartic/glutamic HIC
z 2 around
with C the aci
me ept parameters.
to
amino
K(I) k
describe acid
= on 02.2
R
a2.2.1 some
the
e a variations
residues
+ 1×
k IQ Modeling
relevantI= may µm,
+coefficient εIaµg/mL. e21V ×
be 1col
quantities
ofIpoorly 2 kzthe
kmmx300
εmodel Cpart
eDilution
lumped 1accessible(e.g.,
− εknownk
Purification eofHenry purity,
the because ofyield).
sample
coefficient protein where
“buried”was Typically,
Inperformed
with
parameters,by 4 K the
HIC The in
(2) pH 4 is the
literature,
the material
and aioniconly lumped
Design withionic a
the eluting
10
results of pK Henry
mM a (5)
Experiment
K(pH) from
R
values
Iare coefficient
potassium
tRoften the
and 0(“DoE”)
bK(I)
=purchased + HIC
number
presented
bas:and
phosphate
1= ×column
afrom
A pH a
0kis
k
+ ,
of2in
0 + a
performed 1at
aaccessible Iwas,
bterms ×× a pH I2 pH , + b
collected
k7.2.
of ,
2 kk× b
0a2aspartic/glut 
1, I, b 2
which 2 inMA,are
2 4.7
fitt
is rm
nts,
h
extremely
nimentally
values the
Ipresented
isThere andlumped
2.2
defined
K(pH) inhigh The
isthe
number Modeling
= as:
in
Henry
1(several
2.2
ionic
currently
+
bliterature:
0terms
this
+ B Statistical
of
tliterature:
2.2 k=
strength
1 a×
bModeling Modeling
Isolution
reference
accessiblepart mol/L)
Modeling
a of 2= pH high +
k
4.6
+ B
, I ba kand
is
k
interest
which  ε× ztcan
experiment.
part I+
aspartic/glutamic
1and defined
k2 it
kpH C part
2zk
be
2is mm)
is in
based
calculated
as
using
related Kvalence.
on
Overall,

in
an
that were
statistics
to
HIC
from
acidsACQUITY
the
pulse
K five
followingas
processes
where and
considered
the
presence
to experimental
Hexperiments,
SO
follows:
measured
histidine
describe
C ⇄
UPLC
adsorption
isHSO
The −
of
as
extremely
the +
salt
adjustable
(3) +the
system,
residues
Hthe
retention
concentration runssignificantly
impact
strength
isotherm:
lumped high were both
parameters.
of
time
(severalkey
were
performed:
Purification
(4) Henryof
is
(5)log(γ
defined
operating
species ofRun
K(I)
Purification
coefficient
mol/L) )quadratic
=
k protein

in 1=
parameters
and 1.92 can
solution itazk0is
(reference):
by
of be
Waters
+ 12
√known 12aHIC
protein 1k+

calculated
and ×C Ik(Milford,
z 2+
that 23
× byits aI% 2HIC ×B2Ipresen
from
the
valence. the
ation perimentally
the
ontributing of
protein species
K(I)and
K to in = the
the= 2.2
ε k
solution
2.1.4
a R
results
apparent
+ in+ (1 − × are
Column εwas
I net2 + ) regressed part2computed
charge. × characterization
I k To its reduce using on linear
the the number or quadratic
Truesdell-Jones
were column of fitting
considered expressions.
volume) (2) model as and
adjustable In
(6) [13]storedthis study,
at Purification
parameters.-20°C
k before of I protein
expressions
SEC=×pH 1
analysis by z HIC
C to determi
zt×
2 4 4
of pK a [8, 9, 10]. Besides, USA).
1 εit Q k iis The clear 2 eluent (low that ε5the
k was
the 10
spatial mM  sodium
configuration phosphate k ofpH the7.0, with protein 400 3may mM ammonium K(pH) =sulfate b01 + bB1at pH 7.5. + b The
k k2 pH flowrate
a+ I2=
bserved at around i2500 0 mAU. −
at pH 7.0, eV 1purity, εeflowrate. , bk2and
ε1Run 2ε
able olume, parameters.εsome the external porosity and Qinterest e ionic strength): The 18 activity % B at coefficient Run of the (high
protein ionic in strength):
solution was a28 k,Icomputed b% B 1at εpH C 7.0,
3ameters fitting on e parameters
shown
and 2.2.1
There CIin
relevant is =2.2
Equation
is tathe λ
Statistical
and
currently quantities γIan ionic C
Iaspartic/glutamicis 5: col
e1
the
aVkstrength − model
(e.g.,
high ionic strength
1as εdefined
yield).
in using where
as asdefinedin
Typically, K
follows:
statistics Equationis inthe aEquation
to lumped
Design describe 5. ofThe 5. Henry Experiment
the (8)
The coefficient
impact (“DoE”)
of key and awith
is0=
operating ,in
performed ki,that , ionic kb2the kusing
are fitting thevale T m
+in2parameters
impacts the values of CpK [8, 9, 10]. Besides, itconcentration
is clear the kis
tmodel −fitting In the HSO literature, ⇄ SO + His results are often presented ε1terms
8.06  of kspatial
), which  configurati related
2+ −
where 2− +
the concentration of species solution 0protein 2
tes
are
Henry ting
ration netand often number
coefficient
charge:
from of the were
presented
K(I)
species
of
it
HIC
k
used
2.2.1
and
has
=
accessible
column
ak been
R in
+
in in , k=
was
a terms
Statistical
a solution
k=
Modeling
2.1.3 ,reported
was
× a attempt
0.3I
k , + b of
mL/min.
collected col
a , 1bz×
2Hydrophobic
and k part ,
that Cto ,
I b 2
k2+which
z
describe
in are the The
4.7
its
acids
K pK mL is
e injection
valence.
and
the
related
Interaction
model
of
fractions histidine
variations
amino
2.2 to
parameters.
volume and
K
acids4The
(corresponding
Modeling
residues of
as HMW.
Chromatography
was4 the
activity
depend follows: For
10
part klumped a
(2) µm coefficient
strongly
to with
the Henry (5)a on (4) ofcoefficient
experiments
monomer the protein Purification
K pH
solution (1and of −in εwas iprotein computed
range by kzby
HIC
k50-500its2 usin Pu
ted lumn
were K(pH)
defined
of the
by volume,
estimated
and =as:
protein
2.2
the
comparing
b +
02.1.4
on ε
results
K(I) b in1 Blue
by
Modeling
some
×
the Run
solution pH
comparing
= 0
are
Column
relevant
model a
0
external 2.2t +
4
dextran
regressed
R + 1 1 b =
(low
part ε
2
a was t ×
Q 2
Modeling
2× pHpH):
porosity
model 0
computed
and
characterization
quantities
simulations I + 2
based a
1 k 23 1
acetone part
simulations
× % and
Ion ε
k
(e.g., 2 Busing Qat
linear pulses
purity, pH
the
pulseK the
a
or as
with6.0,
where
quadratic
yield). Run
Truesdell-Jones
flowrate.have
follows:
experiments,
experimental
whereA been 5 ,
Typically, (high
B
expressions. K (3)
performed , the C
is pH): model
data.
the lumped
are a 23
lumped
Design 3 (2)In [13]
to %
fitting this B
estimate
Henry of at
Henry
study, pH
parameters (4)
Purification
Experiment 8.0.
the
coefficient K(pH)
coefficient
quadratic external ε
can
and
(“DoE”)Purification
of = V C
andb
Ibe and
protein
expressions +
is =a is b λ
calculated
the, totala ×
by
performed γ 1 ,of
pH
ionic C−
HICa ε,+
porosities,
1 k from
b −b
strength , ε b×  the
, pH b HIC
measu
are
as fi
defi
ofwith experimental data.
impact its apparent e col 0 1 e 2 2
tonet charge: it impact has been reported =that the pK ofkbamino acid
e  e k k k k
12]hparameters.
entration ed Kis =
Henry
(several εof of
strength
i+
coefficient
mol/L) species
There (1 and −+ Iis
can εacid k0)SEC
itbobserved µg/mL.
1is
icurrently be in  known solution
kcalculated
1 Q kexperimentally
athat
Dilution
k high 2the and
from presence
interest zεpoorly
the the e its in
measured sample valence.
of
insolid theusingsalt literature:
retention as
significantly
was statistics k Inwhere
follows:
performed the
k time literature,
k
describe t(3) with (6) 10the results
mM potassium are often
of tRkey K(pH) presented
=phosphate
operating bk01+ 0
atin+ parameters1
bsolution
pH terms ε×
2pH
7.2. aK of+
0 t1
2, which
× 2
zkpH is
Iand some amino residues may
stored at -20°C before analysis to 2be accessible because C“buried” in the
ntrations the species in bthe liquid ,determine kand the content phases, of LMW, respectively. Monomer is the A
concentration of species in and its vae
 monitored 1
he th are protein
coefficient
olumn often
were K(pH)
defined
in
volume, HIC
used
in andand
=
presented
solution
aas:
There
in
εthe
brespectively
processes
2.2
, 0 an
a
I
2.1.4
the results
= , was The 1is
attempt
Modeling
a in
external
× 1 , currently
HIC
is −pH
computed
b process
termsare
Column, ε
extremely
z b√ +2[6,
experiments
,to
regressed
C part
b
porosity 27]. of ε× was
describeatusing
are
pH
k The
high high
fitting
characterization which
based
andthe following
 interest
were
the(several as
on
Truesdell-Jones
model
Q is
significantly
variations
their the
by
related
follows:
in
performed
linear online
values
using
parameters
mol/L)pulse
location
parameters.
flowrate. or of to
were pH,
statistics
experiments,
impacts
with
and
quadratic the
model [11, K A For itaonline
k as
obtained:
lumped
where
12] isBisas
the
,HiScreen
[13] follows:
to
known
expressions.
k ,values
and
R kC conductivity
describe
C k the
Henry were
some εis (5) elumped
of
Butyl
that =
the pK 0.4
aminothe
estimatedthe
coefficient
In HP
concentrationand
and
impact
Henry
uid column
presence
this acid εonline
and tby
study,K =
withof solid
of
Purification
residues
= εUV
0.93.
key
coefficient
comparing
(7.7x100 +
salt
pH
of
quadratic absorbance
The
operating
phases,
Qspecies
(1 canmm,
significantly
and
may − of √ internal
ionic
ε k be
respectively.
model protein
expressions
be ) in parameters
column
poorlycalculated
ea
atporosity
ε,tsolution
simulations
 kby 215
 volume
HICA
accessible nm.
and from with
z 4.654the
its
be
ults 0]. thus
oncentration are
Besides, often
notK(pH) it K on
contributing
Blue is presented
= clear
0
2.1.4
some
ofcontributionse
species ε
dextran 1
+ that
 relevant (12
A Column
to k the

in in
and0
z
the 2
ε k
solutiont
terms
spatial
1
1 ) I 
quantities
apparent
acetone k 2 
characterization of
configuration
and k k net
pulses ,
(e.g.,
z whichcharge.
its purity,
have εQ of
valence.
−ε is the
e been related
To where protein
yield). reduce where
performed C to
Typically,
kIn may the and K Kthe number C the
literature,
to ka follows: are k
estimate Design lumped the
of fitting a
results of (6) Henry
concentrations
the Experiment are
external coefficient
ε often V of
(“DoE”)
and i the
presented and
total species
1 2 −ais 0i ,
ε in a
performed
porosities, 1 k
terms in
√ 2 , the b of 0
ε , kb
liquid

1 , , b
which 2 are
and
k
i
ing e column that the
volume, ε)eon was Theεi
some the
thus 2
following
,1relevant
k external
1found a√ i from calibration
b0abequal
porosity
quantities the to 2 species
εIi2the
k and was
=
Purification
(e.g., adsorbed
the
testablished
purity, = flowrate.
0.89.
of yield). protein in on
where terms the
Typically, by K HIC ofsurfaceis totalthe athe lumpedprotein or concentration:
Henry e coefficient
col A z and I a 1
e −, a ε , a , b , b  , b 2 a
as follows: IDesign oflumped Experiment 1(“DoE”) is zperformed
k− (8–10). Besides, aitK 1is clear that 2 spa- approach k is used, meaning t
he enry
ngth Henry
tresults
has Istrength
coefficient
solid impacts
phase
log(γ
coefficient
is
been defined were
reported Ikcan = theV
observed
as:=
used
and bThe
colbe 0e− +a
values
that 0bModeling
calculated
in interactions amL,
the × experimentally
Ian
bk1z1of ,=pH
of pK 2√
attempt GE
pKε,+ e1 from −,2Healthcare,
of + bz×
[8, b1ε
2the
aminoC,tC
with pH
to 9,
kbk 2× 10].in
measured
arethe
describe acids Besides,
fitting Uppsala,
solid
core literature:
K(I)
1−ε retention
the
depend phase
of
emodel
where itthe Sweden)
variations
= pulse
lumped a2.2.2
is
strongly clear
were time
protein,
0parameters.
V+ experiments,
approach is that
× tof
on using
Mechanistic
In
described
R the Ithus the +the
the For
columnaan (3)
lumped not
is spatial
literature,×the ÄKTA
the by
used, contributing the
lumped
volume, model Henry avant
configuration
(7)
meaning results
(5) tthis

25the
coefficient
e=
log(γ
Henry to (Cytiva)
are
that )the of
often
=
coefficient
external
the the
−K with
apparent supported
= protein
presented
+ εpH i
k+
can
porosity
contributions √net and
(1A bemay
0K
+ − in by
charge.
and ε(2)
C
t Unicorn
1terms
2ionic
calculated I×Q
i )from
2 To
I the of0 k
the VVV.
reduce
kflowrate1, whic
fromspecie th
ase
ecles
rength pK aare The
values
are
I is RK(pH)
taveraged =
oftenand
interactions
and
respectively
defined the =
2.2
number
presented
Blue 2.1.4
as: bresults
together 01+
12.1.4 2.1.3
A + [6,
dextran
+ with
B
2 ×
in
7]. are
Column pH
accessible
2 terms
I in
aregressed
The K
Hydrophobicand

Ikcharge:
+
part
the C ε k2tof
following ×
acetone
. from The
pH
characterization
solid based
aspartic/glutamic
k ,  which phase
activity values
pulses on is
Interaction linear
were
related
were
have
in coefficient
HIC
or acids
pulse toquadratic
described
obtained:
been col Chromatography
processes and as
performed
experiments, γkpK histidine
follows: (4)
ε expressions.
byquantifies
e is
2=
extremely
(3) 0.4 to residues
the and estimate Inε
experiments
lumped high
t = 0.93.
the
Henry
(several
kPurification
study, Q The
external
log(γ quadratic
ε
coefficient
mol/L) Vinternal
) = and − of
ε
and
protein
expressions
total
can
k
porosityit 1 k−
be k
isspeciesε
porosities,
√ by
calculated
1
known − + HIC
ε C ×
that I from
the
ollowing
otherm:
amino dped Henry acid reactions
impact KkBlue
coefficient
log(γ
2.1.4 strength
residues and
its
 = )QColumn = the
in
apparent
εmay i− I2.2
candextran
+ observed
√ results
solution
be (1 beModeling
k
− ε,The
kcalculated
poorly knet
√ εand
characterization are
Column
1and following
were
,)experimentally
+ acetone
,C
1accessible
kpart
regressed
zεb[Eq. ×,εC btkkI225]
taken characterization
it protocol
thehasbased
pulses intotial been
measured
in
parameters, on
have
the was
account:
Interactionslinear
configuration
reported used
been
retention
literature: or
only for quadratic
performed
that ofthe the time
withthe thereference proteintFor
aRthe pK expressions.
toof
values Ia(7) estimate of
solidmay experiment: andamino (6)
phase
numberIn
thatthe this
acidsthe (1)
externalstudy,
The
of Purification
tcontributions
depend
=
e1quadratic
Equilibration +
kand col Bktotal
strongly
interactions
accessible from e Iofwith
aspartic/glutamic
1+ + protein
expressions
the porosities,
on
with Buffer
e ε
the
K bytsolid kA HIC during
acids pha
k2because “buried” as
located The follows: in
ionic in Vthe the strength pores ε=is the defined particles as: are averaged together εin C kt .(1) The act
tration of ea
adjustable
olution, ped
Henry
Henry εespecies
coefficient
while
were
parameters.
Vwas
coefficient K
thus
the
were A krespectively
and
used = in z found
1and2ε
coefficient
used 1
asolution
− i0+
in I ε,+ a
aIan
B
in
1(1
,equal = a an − 2i[6,
attempt,I a λ
bto
εVolumes
1
attempt
07].
,
iand )− b
kz0k1iε,k2The
quantifies =its
to
b ,
valence.
to
are
t −ε
εdescribe
b following  efitting
are
describe =the fitting the
c2.2.2
0.89. model
(g/L)
values
impacts
adsorption
the
variations
modelSolution
as = parameters.
were
variations U
follows:
Mechanistic parameters.
the (mAU of
obtained:
values
strength chemistry the
of 2the )/2450 lumped
of model
For pK with
lumped =(mAU.L/g) The
[8,
Henry
0.4 (5)
9,Henryfollowing
and (6)
10].
coefficient
εthe R
Besides, reactions
=protein
coefficient 0.93. K with
Q
it = The
is with  i
pH
in
εclear 1internal
pH 
+
solution
(1 and
that
B − ε
andk ionic ithe)Iporosity
were
ionic spatial εtaken
tk  in
confi
mped ng ributing
ume,
s are pK
t 2.2
parameters
concentration =
log(γ
Henry
εoften their
a to
the Modeling
)thecol
= locationrespectively
apparent

coefficient
external
presented of
strength 12.2
and
2.1.4
species + k Blue
part
porosity k 
can
in
I
[11,
I HIC
√ Modeling
net
terms
observed k
4.5
is
Column
e0
dextran
be 12]
+K
in
k[6,Bed
charge.
the C
and I
experiments 1
calculated
solution
and
2
of=7].× ionic
Q
2
k I  The
some

the
,
experimentally To
1
kcharacterization which −ε
part
and
t z 1following
acetone
treduce
strength
from
flowrate. k ε C
were(BV),
amino
K(I)
1−ε
z kis
 the e impact
where
were
performed
the
related
its
(2)
pulses
acid
= values
measured a
number
where
as
valence.
in
Loading
following
the V its
defined
the
to have
residues
+
0consideredcol a Kapparent
A were
1with
literature:is ×
of ,
retentionthe
non-ideality
as with
been
I
adsorption
fitting
B may
in + obtained:
a
follows:
Equilibrium as (4)
, a
column net
HiScreen
C Crude
Equation
2 performed
(7)be
adjustable ×
time charge;
are I poorly
of t during
isotherm:
volume,
3 εthe Butyl
e fitting it 5.
eto
protein has
a 0.4
accessible
parameters. approximately
Purification
HPestimate
ε
The e and the
parameters adsorbed
column (5)ε
solution, t =
because
external tof
pKtthe
0.93.on5
external
(7.7x100
and
while the
BV, The
ε
“buried”
porosity I surface
which
VK
is
e the 1mm,by internal
coltheand
= HIC
and ε in
(2)
coefficient or
roughly
column
iionic total
+ the
Q located
(11 5porosity

e
the −
strength
1−
εcorresponds
porosities,
ε
volume
λ in
)
ei −quantifies
1 flowrate. ε t 4.654
as k defi to
ely lfficient
lumped high
R e of the
k
Blue
Henry
(several Q protein
εe1Vwas
dextran
coefficient
mol/L)
strength K B =Blue thus
εand εeand solution
can
I i2 + it 1found
observed
dextran
(1−
beacetone
is
k −εequal
calculated
known εwas2k/mL and
i ) experimentally
equal computed
pulses
that to,acetone
t k=
from εkifollowing
the ε=
K(pH)
have the
−ε
tpresence
ε
using
t −ε
pulses
e been in
e
=
measured =the
impact b
of
the 0.89. have
where
performed +
2.2.2
saltTruesdell-Jones
k
b
literature:retention
its
k
× been
significantly A pH
apparent
k
Mechanistic ,Ito B
+performed
time b
kestimate ,charge. R CBuffer
× kmodel
net t pH are 2charge:
2 to3(6)
the
model [13] estimate
fitting external
it parameters
has a and
been
R = Iexternal
=
total
reportedεand e Vporosities,
coland I1zinternal2 + isC
that total
the (3) the ionic k K ε1e −strength
porosities,
pK εt amin
of
ng
The
and tting
,efined reactions
parameter
number
14.00 core
parameters
as:are tR where
of of in was
the
accessible
= comparing CUV
solution
λ kis the
and+ col thus
respectively
protein,
was k C I 1kconcentration
aspartic/glutamic
mL, found
is
+ the mg
were
estimated
= λ
GE thus mono γ e[6,
ionic taken
C not
kHealthcare,
Kk[6, 7]. to
1strength
k of
contributing
by − bed Thespe-
ε
into acids
εkt ,The
i (3)
comparing account:
Uppsala, andbeen
Isocratic
as K(I)
1−ε The =
to reported
histidine
defined
eparameters
Solution 0.89.
wherevalues
the
model
=
ionicElution
0
activity
Sweden) a apparent
in0V + thatwere
1
strength
residuesEquation simulations
chemistrya
A2using with
coefficient
is
,B the
× kobtained:
 the net
k ,follows: pK
77%+
anI C(4) 5.is
columna+ÄKTA
2 of weredefined
of × The amino
R withI
the ε (8)
evolume,
estimated A=
To
avantproteinas: + 0.4 the
23%
reduce and
25εeby pores
in Bufferε
the
solution
(Cytiva)
the of
=
t external
comparing the
B 0.93.
number
t = particles
during
was
supported
QsolutionTheof
2λkporosity 20
computed
model fitting are
k BV, by 1 k+
averaged
and
simulations (4)
(2)
Unicorn
εe takenusing Regeneration
porosity
Q the K the
aVVV. True
flowrat
with ex
the chromatographic medium. Ckfollowing eεt==The
reactions
parameter in Qkwas were
estimated by into compa acc
e −
estimated 8,eresults by often kpresented saturation
respectively model in k terms was simulations observed
of 7]. which 1−ε atwith
followingεaround
is related experimental 2500
values to Hcol K k1O
mAU. were as H data.
obtained:
=kkatheir
2 + OH εsome 0.4 and14.00 εt1R = 0.93. The internal εpoorly porosity
K re eameters.
gfollowing εel concentration
9,
the
estimated
parameters
the
10]. respectively
Besides,
cies
parameters,
external
concentration C 2.2kby
and
reactions in
kwhere== of
porosity
comparing λ2.1.4
Q
Iit
Blue species
eModeling
ε2.2.1
solution, Vkis
of
is
was γ
only
1colK 2.1.4the
ink
species
[6,
clear
dextran
Cand
is and
thussolution

the Column
7].
ionic
the
kThe
kStatistical
with that
Q
model zin
k
ε5The
1Column
found
pK is
the esolution

100%
lumped andthe
following
strength
in εεeflowrate.
part
its following
valence.
values
tsimulations
were
solution
characterization
spatial
acetone
equal Buffer model  and
 characterization
Henry astaken andto
protocol Bzvalues
configuration
defined
and εpulsesiduring
knumber as
with
coefficient z
its
acids
=into follows:
was
in
its
were
valence.
t −εhave their
depend
3eofused
experimental
εEquation
account:
valence.
of
BV,
= parameters
obtained:
been
accessible
andeThe
Equilibrium
the location
0.89. strongly
for
2.2.2 (5)
a
protein
ionic performed
the
5.
, a Stripping
data.The,
ε[11, on
aspartic/glutamic
reference
strength
Mechanistic
a e mayA
, b
, 0.4
12] ,
Bto
2(8) with b
kand
I
,and
experiment:
, estimate
loca- isb 22Water
were
defined
aremodel pKacidsestimated
0.93.
together
fitting the
amino
during
as:(1) and The
external
model acid
in C
by
3histidine
BV.
Equilibration .
comparing
internal
= and
residues
The
The
parameters. λ total
2 k residues
γ porosity
activity
pH C may
with ofmodel porosities,
be
thecoeffi-
Buffer
For
e simulations
crude, A during access
Buffer
odel
he a 9.23
concentrations t K =
tRIexperiments,
= εof ε + ethe V
The was (1
colz 1 − 2+
species material
thus
C ε k ) 1 a
found −
k εεpK K ε
in
√ eluting
e the k equal εK(pH)
ε t liquid
−ε from
 eto ε= and
k the K(I)
1−ε
b=0values
is+ HIC
solid
−ε
texperimental
where =
1e ×were
bwhere Solution
column
a
=pH
phases, 00.89. NV+ 0 H
+ a +
14bis
1 was
× 
data.
respectively. chemistry
×
the I2(5) N+
pH collected
(4) a
Hthus
column
0 (6)
× +=H
1 I A + in
volume, The 4.7 following
mL
a e9.23
εspeciesIfractions
the = reactions
external k (corresponding
z C in k
k (3)
porosity solution
k
and (2) were
to Q the the taken flowrir
a2.2.2 Mechanistic model
ifound ki+ e

re,
4.00 olume,c
arge:
protein
timated
ture, as: results
model
itεin has
ewere
resultsby was
are In
solution
the pulse
been thus
R often
the
considered
external
comparing
are
reported
respectively
=
oftenliterature,
2.2
was
Blue Q presented
2K porosity
model computed
4.5
dextran
presented as A kthat
equal
Modeling Bed 1results
and
adjustable [6,
simulations andin
the ε1the to
Volumes −
eBuffer
terms
7].
using
and
in Q 2are
lumped itthe
terms
The =K the
a
acetone of
often
B of
part
parameters.
with 1−ε twaskamino
following , which
Truesdell-Jones
flowrate.
(BV), Henry
of
=
set
e experimental
k tion
pulses
(2)
,
iacids
0.89.
which7.0. (11,12),
Loading
coefficient core
K(I)
1−εThe
have
depend
related
is H model of
where
data.
related and
O =can
flowrate
with
been the to strongly
C col
some
0is [13]
HkV+ K
obtained:
Crude
be protein,
performed
to is 2as
+acalculated
was
K 1+ the
amino is× OH
as
on
follows:the
during Iconcentration
1.16 +2acid
follows:
ε3column
(4)
− ea to mL/min,
2 × res-not
approximately
from 0.4
estimate Ipresence2 contributing
and
volume,
the cient
14.00
εthetγ=
ofmeasured
which 5 0.93.
εquantifies
the
BV,
external 2to
corresponds The
the
ksignificantly
in
external
which
retention √ and
kapparent
solution
the
1 slope internal
tonon-ideality
roughly a 2fluid
porosity
totaltime and tporosity
net zvelocity
corresponds
porosities, kand charge.
its
(2)
of valence.
Q of To
the 150
to flo
t
ations
elid ed, is
some in
2.15 pK
defined
meaning
following of
aminoHIC athe as
processes
that
acid
2.2.1 the
2.2
species
1Kactivity
reactions column
There
the
residues slopeBlue
Statistical Q isModeling
=kk in is
contributions
in
extremely
ε may of
volume) +
currently
dextran
solution
ifound the the
(1A bekmodel − z ε
liquidadsorption
ε
ke and
poorly high
and a
iwerepart
)from I high (several
stored
and
acetone
accessiblethe
taken k
interest solid isotherm
at
species
K(pH)
εt −ε mol/L)
Interactions
-20°C
pulses
into in
phases,
where
because using
adsorbed
= and
before
have
account: at
The
2
0The
bInteractions
C +
“buried” low
Solution
it
statistics
H Equilibrium
lumped
respectively.
andbbeen SEC
1ionic
3× PC
on concentration.
with known
col
O pH the analysis
4strength
performed
in 
are Henry
to
+the
the
chemistry the surface H that
describe
bwith the
22pK × PAO solid to
pH the
coefficient
I to
concentrations

is
or determine
2 +defined phase
(6)
the H
estimate
The
+ impactis
following
ofpK
defined
The
the
2.15
as: k
the
e salt
of
of the
a the
external as
interactions
content
key IThe
k =the
operating
species
reactions of and LMW, k
in zin with of
Cthe
total
solution k
the
parameters Monomer
R
(3) the
 porosities,liquidadsorptionsolid
were and phas
taken solis
defined phase as:presented
where The The was
log(γis in the 2thus
interactions terms ) = cm/h.
coefficient
lumped − of k′ The equal
Henry , with
which
of √ impact
1 the to − + the
coefficient ε
is
ε C
protein = of× solid I
idues
solution in
e
and =
phase
may
solution a0.89.
parameters,
Inbe
pH , the
a kwere
poorly
and , The
was a literature, ,ionic described
only
accessible
ionic
computed
b k
, b strength strength
, b results because are using by 4the values
(modulated
fittingIare the (7)
is
the solid
often
the
defined and phase
protein
Truesdell-Jones
model number
presented
by as: A
solution,
parameters. %of
z 2
in accessible
B) terms
I
model interactions
while
was For k of
then
[13] theaspartic/glutamic
k , which
coef-
studied with isthe
around relate sol
The
ain:
+ volume,
23solidthe following
impacts ε
phase
14.00
particles Ias
the = follows:
the
are reactions
external
The on respectively
values
and
averaged z
2.2 Ck
interactions
some 2
porosity
HMW.
 k of mg in pK
Modeling
relevant mono
together solution
a [6,
and [8, /mL with 7]. 9,Q
quantities 10].
bed
in theThe were
the Cthe
, ε(3) i
tfollowing
flowrate.
part t
k .To
k
Besides,
solid The
Isocratic
taken
(e.g., 1−ε phase eitinto
following values
purity,
activity
lumped is Elution 0 account:
clear
Nwere
Considering H yield).
+
were
1 that
4isH
coefficient
approach
adsorption 
with
Solution
H
described 2(5)
Pof× N obtained:
O the H77%
Equation
Typically,
O 0 

γiskas
3 spatial
+ +
1 Hisotherm: Buffer
by H+
chemistry
HP
quantifies
used,
2+ +

theε a8, e OH
Omeaning
a
configurationA
Design=
2− we + −0.4
+ 23%
obtain:
2 log(γ H and
9.23
of The
+ Bufferε of
Experiment
thatk )6.82
t= =
following
14.00
the − B
the 0.93. during
protein
k k The
reactions
(“DoE”)
contributions
2
√The 201 may
k+
BV,
internal C isin 2×
(4) solution
performed
I (3) Regeneration
porosity were tak
zkB zkfrom the species
23 % aB
6.82 e k 1reference + IQ 
mn yot ation
that contributing
The
volume, ofthe
pK species
following contributions
εaexperiments,
2.2 the to
k Modeling
22.2 inthe
reactions
external2.2.1 solution1K apparent
respectively2.2 = porosity from
2in
εStatistical
part and + net
solution
Modeling (1 [6,
the

and charge.
kits εmodel 7].
i) valence.
species
were
model The εflowrate.
parttaken reduce
tkfollowing adsorbed
K(pH) where
into the
were values
Cnumber
account:
= on
following
considered
bkbe +five the were
the 2
2Equilibrium surfacefitting
concentration obtained:
adsorption − badjustable ,2or × ε4isotherm:
e+of = 0.4
species
(6) and
Purification
parameters. εpK
kretentiont =
in
λ3kmodel
solution0.93.
ofThe protein
Iparameters.
= and 1t1R internal
z(reference):
by k its
CHIC valence. porosity
m: del pulse asrelated follows: to kK √ as
Modeling follows the this lumped εin part Equation Henry experiment. 6εεcoefficient they t−ε
kεwere areOverall,
can “buried”
εstatistics
V 0bby
calculated in the
12−
aexperimental ,4pH core
ε,O efrom bof of the
the pH runs measured ficient
were performed:
Purification quantifies
1aK+kstrongly of Run
time γthe
I2protein adsorption by CHIC
e
d
values s15
erm: I protein
are is
phase
defined impact
often 9.23
and
defined
as:2.2The
pK
solution,
numberK
presented There
as:
itsa
pK =
A
2.2.1
where IModeling
interactions
apparentand
zλ was
is
while
of = 2
in γ
accessible
was
K
currently
the
terms
I
thus (C
Statistical
with
the K is netz=
results
thus
the
kof
ifound
with

−C 100%
coefficient k
k apart
charge:lumped
+
ifound0)
aspartic/glutamic
,
the
high
are which(1equalBuffer solid

regressed
equal
interest
it
λε1isHenry
ito )−
has B
phase
quantifies
relatedto itduring =
beenε
incoefficient
based acids
=
using
t
to
1−ε In εK 3e−ε
reported
K(pH)
ton
=
locatedthe
and
the
BV,
described
=
as
e
H 0.89.
elinear
and3=
col
adsorption
histidine
follows:
=
(5)
Pin
literature,
0 that
0.89.O b or
HP
a 4N
the
0 1 +
H
,1quadratic
Stripping +Hto Othe b
+
1
residues
pores4
the
Equilibrium
HO
1
describe × 2
strength
results
apK
 P 2(5)pH N of H awith
P 4

K H
0 + + +
expressions.
O
theare 3+
bthe
3−
b amino
+
1
with H
(9)
2
,Water
OH × H
+
bimpact
particles
often
2pH +
H −are + 2
acids
2.15 fitting
during
(6)
presentedIn of
are key
depend
this 9.23
12.38
BV.pK
operating
study,
averaged
14.00 in
aB
terms =quadraticλ
ktogether kof
pH
Isolution
= parameters
kk(C on of
, kwhich
the
k ε−
→ ztFor
expressions
in (4)
k
k
2 crude,
C 0) .
kis (3) The
related
Buffer
activit toth
are sic are 312.38 averaged
14.00
fitting k
parameters atogether k 2 k  and k in Iits Cthe issome .the The ionic k 2activity
1(low − strength εprotein, R coefficient
e
asethus where
The
defined activity C
γcontributing
kin is
2 quantifies Equation the coefficient concentration +5. of The the −+proteinof species
K with = εkeyk in
solution
+ strength):
(1 − performed was
εathe
2ki )fluid and
computed k z
k kB atusing its valence.
model
log(γ
of following
speciesas
their follows:
)= k14.00 pulse
on−
location
in some
solution
k
2.2 k 2.2.1
experiments,
There[11, √1and
relevant A Modeling
k +
12] atC
1isand
pHStatistical
and
zkzcurrently 2× Buffer
quantities I7.0, k valence. Run
lumped Bapart amino was model
(e.g.,
high Henry
set purity,
acid
interest
i ionic
t7.0. coefficient
The
1−ε
residues strength):
yield).
in  Q
using
H flowrate P not
may can
Typically,
O −√
statistics
H 18

(7)
be k be
Pwas H 4 %
HP
O poorly calculated
ε,O B
 1.16 a Oto  at 2−pH
Design H mL/min,
H to
accessible
describe 2+
4
Pwith H
O7.0,
the + from −
of + OH + Run
Experiment
the H strength
the
which
because
6.82 3measured
impact (highcorresponds
2.15 ionic
14.00
(“DoE”)
“buried”
of the
i retention chromatographic
to
is
in
operating time 28 tFor
parameters %
velocity ofpH 150 7.0,
he
medium.
OH ble +
82 −parameters.
2.15 k
The reactions
parameter
14.00 where
There were
I
B = inis 2.2.1K
used solution
λcurrently is kinthe
was C an
Statistical
k2 lumpedwere
estimated attempt
a high taken Henry byto
model
interest into describe
where coefficient
comparing ε account:
inthe
V A
using the ,
non-ideality
k In The Band
2model variations
statistics
the
k , H
1 24 C a

approach N
k SO
3
0
literature, ,
ε are
simulations
H a of 1+ 24
to  e
3ofa
the (5)
describe
2
,usedHSO ,
fitting the
N
4 b protein
resultsH
0 ,
b3in
−b
lumped 1 + ,
parameters
+ 4 the
this b
H H2
are + are
+
solution, Henry
impactoften
article fitting and
1.92while
of 9.23
presented
can
C
model
coefficient
keyI
be% k is
the= λ
the
operating
extended
γ
parameters.
with
coefficient
in k ionic
k
terms C k pH
to 1 kstrength
parameters of
kany
− and
λ ε
k tRquantifies
, other ionic
which as kind defined
is the
rela
of
2.2.1 1follows:
+coefficientStatistical I2= ελ4ztmodel as follows: A zand Inet inC kε=then λλ(7) γand Ckesti-
Ck were estimated kby comparing k k 2 4 k
tion, 1.92 while
9.23 the
asthe where Isome cm/h. Run K is kkthe
The (low
Cother kmodel
quantifies lumped
impact pH): simulations
23 of Henry the %solutionapparentB ecoefficient
adsorption
at colwith
pH pH net 16.0, experimental
where
and kcharge. Run strength
ionic C ae5− 4,strength
To is(high a1the +
reduce data.
with
 a(5) pH): ,(8)
concentration 40+
the
(modulated , 2− 23 breduce ,+% bthis2B +are
medium. +at ofbyfitting
pH species
the 8.0.
The model
parameter
kB) parameters.
solution
was iswas kstudied zFor itsaround valenc
2 ,3kN
ticle can be and extendedthe C results = kto are any
γobserved C
regressed kind
based of on adsorption
linear or quadratic 2−
,k
isotherm, 0K H −
expressions. 3−
 like 2 N  athe H +
1
In (8) H study, 9.23
quadratic
core of kiprotein, thus not inexpressions
 contributing to
+ tration
stic
ry ten
of + the+presented
model
The of
protein
pK species
a pulse
where
K
9.23
following =in terms
εon V + on
in
strength
solution 2.2 C (1
reactions
some Thereksolution
experiments, kis of
− the εwas λi )I
2kModeling
relevant ,iskk relevant
which
γcolumn inand
k kC
computed
currently kthe
solution kzis
kquantities k quantities
related
its
lumped
volume, valence.
experimentally
using
apart were high tto εlog(γ
the
(e.g.,
R Henry
takenK
= (e.g.,
einterest the
parametersas kthe
purity,)the
Truesdell-Jones purity,
follows:
=in
coefficient
external
into in −apparent
HP the
chromatographic
account:
yield).
using +yield).
A O literature:
porosity
kwhere
HSO H
can B √
model
Typically,
statistics P Typically,
be COP+ C charge.
O
and
(6) Ccalculated
[13]
were 4k is 
SO ×
medium.
to Q
the+ HP aH I42− H
the
estimated To O +Design
2concentration
Design
describe 3from
4flowrate.
H4− ++
The of 12.38
theH by
K of
the +theExperiment
6.82
measured
comparing
parameter
Experiment
impact
= 8.06 ofnumber species λkey of
ε(“DoE”)
kretention
model
(“DoE”)
ofofspecies
− iwas
k)UV fitting
operating (4)
tsimulations
estimated
solution
is time
kkperformed performed
parameters tandRby
kwith
zcomparin itsexper vale
+ H 9.23 H P O P O H +ε i2.15+ (1
2.38 col ε V 4 k 1k 4 − 44 ε
H nistic
parameter
arameters
+ 8.06 6.82 model and in λCHIC
I pulse
is 2.2.1
=
was the λ
processes 2.2 experiments,
γ estimated
ionic
There this C Statistical 1 The
Modeling
−is
strength
kreference ε extremely
currently process by the model
as lumped
comparing
experiment. defined was
part
ahigh high monitored Henry
number
(several
in interestQ
where model
Equation
Overall, Langmuir
of C
coefficient
mol/L)e by
fitting
1in +
In and
online
simulations
col 5.
using
five Bthe Theε
parameters,
and adsorption
C can
I
literature,
experimental 4 pH,
(8)
statistics it are k be is online thecalculated
known
with e only  to isotherm.
concentrations
results conductivity
the
describe
runs that from
are
were−
mated
the The the
presence
often
the concentration
of
and
by
impact
performed: the
measured
presentedonline
comparing of salt
Run key retention
in in
k
model
significantly
1terms
operating in
absorbance
√ the
(reference): the of liquid
time
simula- liquid
k ,
parameterst at
which phase
23kand
215 % is C
solid
nm.
Brel
k
H sntrationThe are
− +
coefficient 2.15
parameters,
often
concentration
+ There
14.00
of iswere
presented
species
2.2
as The
pKin
defined
k
isused
kfollows:
and HIC
onlyModeling
ktheand in
currently
ionic the
in
on
k
asin processes
in
thethe
terms
solution
k
some εstrength
the
results an
the k pKresults
slope part
attempt
ofaaliquid
relevant kvalues
high
are
and
t
is,of Iare extremely
which is
regressed
zthe interest
the to
phase and
regressed
defined
its
quantities is describe
adsorption
valence. number
related Chigh
in
based as: based
kusing isthe
(e.g., t(several
to of K
indeed= variations
accessible
on
statistics
2.2.2
isotherm
experimental
on
R linear
purity, Hlinear
as 2
k mol/L)
follows:
whereSO simply
Mechanistic
The
or at In
yield). 4HPk or
to 

low
quadratic
e
1the of
aspartic/glutamic
C net
data.quadratic
+ kO
kH
and
HSO
describe the 2−
replaced
and 3charge
concentration.
literature,
Typically,
P−it O
lumped−
4C 
4 is+
expressions. PkK the known
expressions.
model Oare
H
of H3−
4
+
by impact
the
results
atotal2Henry
the P+ 2−
Design
Othat
acids
H + + Purification
concentrations
41.92
proteinIn of
are theH
coefficient
and
In key
this
of+ oftenpresence
this histidine
12.38
instudy,
Experiment
2.15
study,
operating
solution
presentedwithofofofresidues
quadratic
protein
the salt
pH 1−
quadratic
parameters
was
(“DoE”)
A zspecies
in 2significantly
and εcalculated byionic
I expressions
texpressions
terms is
HIC
ε(4)tk of inkthe
performed
R
as aliquid
, which functi isa
h +
ncentration O
ncentrations
92
try ated
tein H
the
4 +by
concentrations H
The
12.38
solid
in6.82 2.15
following
comparing
solution of 2.15
impacts of
phase
species of
as a
the follows:
model
was reactions
the
k species
species
The
inon at values tThesimulations
solution
calculated some
pH interactions k
7.0, k
following
in of in
relevant in
solution
pK
and Run [Eq.
k the with z [8,liquid
ε6]
2 liquid
calibration
with
its were
experimental
9,
quantities
(low 10].
valence. and
theionictaken
pK and solid
lumped
Besides,
solidwas
values
(e.g., γ
solid
into
strength): phases,
data.
k C
established
phase
purity, Q
itk . phases,
approach
account:
εIand is number
 clear
were
18 H
respectively.
yield). % 2  P in that 4 isO
respectively.
described
B ofε
terms
54at used,
Typically,
− accessible
the −
pH 
HP of A
meaning
spatial
7.0, by O  a
Run
4 the A +protein
tions
H
that
configuration
2−
Design 3 (high with
ofthe
+ 6.82
K =
Experiment
ionic ε
contributions
concentration:
experimentalof i +
the
strength): (1 k−
protein
(“DoE”)data.
k ε i )from28 may % istheB kperformed
at species
pH 7.0, ad
+ K = εconsidered
+ (1 −,2.2.1 εBiI)kas kthe aas aeparametersfunction ofthe pH εfrom
Vis1statistics  (6) the 12− amino εso, acid
impacts the values  of pK [8, 9,external 10]. Besides, it− is clear H P that O the HP spatial O + Hlog(γ
configuration 6.82 −of the protein Cmay
mistry
8, HOH
ts entrations
2−weare −where
were
obtain:
+often where
+on
The
9.23 ofiV strength
the
some Awere
following
col 14.00
presented iskThere
species the
were
relevant
and in
used observed
column
1,apparent C k adjustable
used
reactions
isterms
the inStatistical

currently
in are
quantities
t results
in
an ofεvolume,
3t an experimentally
liquid
k in fitting
attempt , are parameters.
attempt
solution
which
a
high model
kand
(e.g., a
regressed the
is
to solid to
purity,
were
interest
related
describe in
describe
phases, the
ataken
based in
to and
yield). K literature:
porosity
TheK(I)
using
on the
into
as lumped
respectively.
IHSO sequence. eTypically,
follows:
lumped
variations
linear variations
=
account:
col
Vthe
H and
acol or 1SO +approach
ionic
SO To
Henry
quadratic Q
Aa2of4a1
2−
4to  the
Design
do× e of
the strength
+ IHSO
describe
12− 4 − the eflowrate.
H+
coefficient lumpedis
εor
+
aeO theof
expressions.

lumped
used,
43− × + as
Experiment
the IH
number
2 +
meaning
4defined
8.06
isimpact
Henry Henry
defined In of 1.92in ofthat
positive
coefficient
this k
as )study,
coefficient
Equation
(“DoE”) key =the
the operating
and
slope
with with
iscontributions
quadratic5.with
performed
negative
of
pH √the
The pH 1and
parameters− +adsorption
and
εε(4) k from
tcharges
ionic
tsolid ×ionic
(2) I the wereisothspwc
Cexpressions
Interactions with 2+ the solid phase + The interactions the
O +meaning Hof +are H that
6.82 6.82 the
pK 2.2.1
acontributions
− kε Statistical from the model species tpH
adsorbed
located = = in e4on
the HP 2
0the
poreszon O
C surface of the P K 2
particles + H are 12.38
averaged together1solution
+ B in Iwere (5) k .expressions The  phase activity
n
m
el
ber eaning
06
+sults isotherm:
4
defined
ning + + 1.92
that that
often
The
positive
as the
parameters
and impact
Kthe
where the
presented
=
ionic impact
activity ε
contributions
the and
slope
on i V contributions
+ its
strength
pK
results A
some (1 A and
its
coefficient
negative
is
, Run

kof in
zBk2the areε terms
apparent
the
relevanti the
, ) I
from I4C is
observed
column
regressed results
(low
adsorption
were of net
of
defined
charges
the k from
kpH):
the
quantitiesnet charge:
,volume,
species
estimated which
are proteincharge:
as: the
23
experimentally
based regressed
were % is
isotherm
adsorbed
onitspecies
(e.g.,ε B aspartic/glutamic
related
in
2.2.2
byhasitat
the
linear solution
computed been
R
based
has
comparing
purity, on to
Solution
atadsorbed
t
externalin
or
K
6.0,
been the =
low
Mechanistic
The reported
on
the
quadraticwas
locatedas
Q
yield).net Run
at
2 linear
reportedfollows:
model acids
chemistry
computed
concentration.
literature:
surfaceporosity
HSO each
charge 5 in
Typically,
k
HP that
or
(high or− 1
expressions. the and
the
kquadratic
4
pH +
model
simulations  (6)
that O
and of ε the pH):
pores
SO histidine
using surface
based
the Qthe a
pKThe
2− the 23
P
protein
Design K of4pK
the
a
+ expressions.
InO of
on
with% 3−
following or
flowrate.
the
H this amino
Bof + The
at
Truesdell-Jones
+
of amino
H
particles
in study, pH
solution
experimental
Experiment lumped
acids
+reactions
8.06In 8.0.K this
acids
12.38
are
quadratic depend
= was Henry
ε in
study,
depend
model
averaged
data.
i
(“DoE”)+ (1
calculated strongly
expressions [13]coefficient

quadratic kstrongly ε
together
is i ) as
performed
on a εon is
taken
tk
(4)
in functionC  into
. The ac
of
+
3−
erticles+ HOH H+ +H
particles −+ 12.38
2.15log(γ are k9.23
12.38 strength
)averaged
=location
14.00 − Kwere
There col ka k = Iλ used
together
is kk observed
√ γ1kk−
currently (C +inεεtkCtan in
k →
− C a0)
×experimentally
attempt
Ikphase .
high The interest to e
following
activity the Considering
describe in in
R
the the
coefficient
using
non-ideality cthe
adsorption (g/L) number
literature: Q
statistics kEquation
variations H = γ ofSO U
4 and
isotherm: V
quantifies
the 4(7) to  4 eof
(mAU pK 8, εpoorly
describe
protein HSO ethe
we
a4 2of (9) obtain:
)/2450 −
lumped each 4a
+
solution,−H
the amino
+ Henry
(mAU.L/g)
impact while acid,
1.92 of K
key
the and
coefficient εthe
=coefficient
operatingi +with two (1 −contributions ε1i )−quantifies
pH
parameters
λ andεt data k were
ionic k
the (1)
phase + n H
articlessolutionas are
8.06 The
− are 12.38
averaged
follows: their
Mechanistic
interactions
was
2.2 averaged
their calculated
Modeling togetherlocation
1processes
There + were with
B together [11,
model: The
part in
is usedII[11,the as
C 12]
currently process .a12]solid
in and
solution
in
The an
function
isand
 some
Cdefined attempt
was
activity
a
.high
some The amino
were
monitored of residues to
activity
aminopH
K(I)
coefficient
interest acid
described
describe 1 were
Hydrodynamics
from
where =by
acid
incharge.residues
Equilibrium
athe online
γusing considered
by
coefficient
the
A
k residues
+ the ka
quantifies the
,quadratic pH,
amino
B
1statistics × 2
variationsmay k
H,Iobtain 2may C+ SO as
onlinekγis
be a adjustable
and
acid
are to be ×the
quantifies of conductivity
poorly
3Idescribe
HSO kinetics
the fitting 4 accessible
lumped defined
pK
accessible
+ parameters To
+
Purification
the H andHenry
aimpact as
because the
simulate
online
because
1.92 of slope
coefficient
and
of UV the
“buried”
protein of
I “buried”
isand the
experimental
absorbance the with
by adsorption
in ionic
HICkthe pH 1 the
inparameters
(2) at and
−strength ελtas 215 ionic nm.acquas
+amino
P + − + −
following
H
O +OH
+ +acid,
H6.82
1.92 +were
reactions and
in
K The
2.15 HIC
used
= and εthe14.00
in ionic
+ in
strength two
solution
the (1 an − strength
results kε contributions
attempt i) were isobservedare
k extremely εItaken
tk to
regressed describe into
kexperimentallywere
high account:
based as: thethen
(several
Ithesequence.
on
= 2.2.2
variationssummed
linear mol/L) 0 z= The
2
or non-ideality
To
Mechanistic
C of to
anddo thenetk so, it
− lumped
charge  (6) the
k 2
4 known of
expressions. number
model the of Henry net
the protein
that of
4 protein the
coefficient
In solution,
positivepresence
this inK and
with
ksolution
study, = λkey
while
of negative
salt
pH
quadratic
γto k (C
operating
the
was significantly
(5) →
−coefficient
charges
ionic
calculated
expressions
0) were k quantifi
acompu func
kin the literature:
where iV is the column 2−
otein
sin 3in ++ 4To
+ HH
this
+
solution,
simulate
chemistry.
article core 1.92while
can
where K ofthe
9.23 on
= be the the
εC
The some
col experimental
iextended +Theprotein,
is coefficient
(1 following
C the relevant
− ε= i)
1relevant −to thus
concentration
λ εobserved
any
tγ λ
quantities
reactions
kεC data
knot kkvolume,
quantifies
other acquired
contributing kind
of10]. (e.g., εspecies
Solution e the
parameters. the
of K(pH) external
adsorption
purity, by
chromatographic
the to
adsorption HinGSK
chemistry yield).
present
the Othekapparent
solution b porosity
 0HSO
kin strength
+
isotherm, H Typically,
study, the
b+ and +×The and
frame
medium.
net OHpH azCthe withSO
(6)
likemixing + Q
following

charge.
its atotal 4ba the the
Design
of
valence. +the
× The flowrate.
cell
(8)HTo
pH
+ 2
of
reactions
2isotherm
model +Experiment
parameter
reduce
14.00 8.06
at
was inlow
the solution
λthe
used (“DoE”)
kwas
concentration.
number were
estimated
describek
ofwas is
fitting performed
taken the byinto comparing
(3)
hydrodynam account
N H
4solution,
otein
H+
ting solution,
positive
2− + pK Hin
parameters
K
+ + 1.92
while and
solution
strength
+=8.06
while
impacts
λ where
the
corenegative
and were
γto coefficient
9.23
I ofthe
was on the
observed
I kthe
usedstrength
V is some
col

− coefficient
calculated
values
the is
protein,
charges inkfollowing the
λ
ionic I of
experimentally
an
k 1 column
quantifies
k −
attempt pKthus
were
strength λ calibration
as quantities[8,not avolume,
the
quantifies experimentally
computed
9,
to function
as adsorption
contributing
in
describe
defined was
the ε
(e.g., the
Besides,
e the
the established
2
parameters
at
ofpurity,
number
literature:
in the external
K(I)strength
pH to
adsorption
each
the in
2√
it
sequence.
Equation variations 1 the
is =
yield). pH
chromatographic
from and cleara literature:
inporosity
with
A
apparent 5. HSO + based
pK 2the terms
,
1
To 4 a
that
The Bstrength
Typically,
of

do× ,
ofamino
the k I of
and
net
oneach
so, + were
SO
lumped spatial Q
charge.
2 awith
medium.
the amino
2−
acid× 2 protein
estimated
Design I
+
number flowrate.
H
Henry To
acid,
configurationofThe concentration:
reduce
of by
Experiment
and comparing
8.06
parameter
positive
coefficient the k
of two number
the and
with(“DoE”)λ model
contributions
protein negative
pH of and fitting
is simulations
estimated
may performed
charges
ionic were (2) by
then with
were cot
k (C 0)in (9)
Interactions k k k
H 12.38 kkwith the solid k phase The interactions =Calculations with × the
were solid × first phase
performed were described
assuming by the
that: (i) the pK offun
a UV saturation was observed K(I) at 0 1 2
P

del
2um.
O 3 −+ ++8.06
The H kparameter 6.82 The and ionic
λthe was strength
results estimated are Iliquid t is a defined
regressed by comparing based as: Equilibrium
on 2.2.2
model Iaround
lineark= asimulations
The +2500
0Mechanistic aznet C mAU.aI4 +
charge with aapproximation model of 4I the
pK a protein CK in = λfollows:
solution k γquadratic
k Csolution was k
calculated (5) (2) asaof ainto
+ 4H I+or  quadratic expressions. In this study, expressions
P k experimental + data. 1 + 2The
model
wing +O
otherm.
The 4reactionswas
parameter
14.00− HThe
(Table used
parameters,
in
+In 2.15
concentrationI)
the
solution
λstrength
parameters, was
Thein describe
literature,
and solution only
were
estimated
ionic the only the
taken
strength
results the thewere
results
the pK by hydrodynamics
into pK
are
a taken
comparing values
I areaccount:is phase
values
regressed often
defined and model
and The
Cnumber
presented
basedas: is Solution
activity
in
Linear
number The
A indeedthe
simulations
on N zofto
2.2.2 2
inHaccessible
approach
linear of2coefficient
column,
Driving
terms simply chemistry

accessible with
The
or N
Mechanistic kof H
quadratic used Force
net
kkwhile of
aspartic/glutamic
replaced + ,,the whichHthe
in
aspartic/glutamic
charge thisthe pro-
expressions.by is
model following
article
of relatedthe Considering
9.23 acids
can to
proteinIn reactions
was be
acids
kand
this used
as
extended
in andhistidine
study, in
solutionto
Equation
histidine
k
describe
to
quadratic residues
any
was were
8,
other
residues the
calculated
expressions we taken
kinetics
kind as ofaa
ads
OH
m.
mber
H
of e re acid,
H
3−
+
rformed the The
P
estimated
2the
+2− Oof
H and +
41.92
protein +
where
8.06
parameter
positive
following
concentrations Hthe
by
assuming
C
impact
12.38
in
= C two
comparing
λ
k 2.15
and
adsorption
solution
were
is γ theλ
its
contributions
where
C of
that: knegative
used
apparentwas
model
the I was
concentration K observed
The (i)
in estimated
isotherm:
species is simulations
calculated
an the net
charges
material
the attempt were
lumped
k pK charge:
experimentally
in a
of by
the
asthen
with
were
eluting
log(γ
species of
to Henry
a comparing
liquid it
K(pH) the
kfunction
describe
has
summed
=
k
k
experimental
) computed charge.
fromH−
inand been
coefficient
in
amino =O the k
the
solutionofsolid
theb experimental
the
independent
K(I) model
kreported
+ HIC
sequence.
pH H √acidobtain
number
data.
4
literature:
atbO
variations 1 and
phases, =
from
and +×
(8)+ eachcolumn
aU simulations
side
pH
OH C0az
that
+0the the

Toand
−+
3
itspH a data.
respectively.
of aofI1chains 1bthe was
×
do
amino net
the
valence.
pK a based
×I22× so, , pK
ionic
+
pH
a b
collected
lumped of
0a with,
the
acid a2+on
2
are
14.00 each
b of
× A,
strength,
1 amino
b
Ithe
number
2 2aminoare
in2number
Henry 4.7 acidsacid,
fitting
(ii)
of mL the
positive
coefficient
depend
and
model
fractions spatial the and
with
strongly
parameters.
(7) two configuration
(corresponding
negative
pH contributions
(3)
and
on For
charges
ionic (2) of
to were
the
wer
protein
odel
sO epK P O
4 following
eetion cana was
9.23 +
be
2− in H
into
+solution
were
used
extended kaccount.
reactions were6.82
k considered
to k in
k describe
wereto
considered was solution
k
any usedcalculated as other adjustable
the
were as in kinetics
taken
adjustable
an kind attempt as a
into
of a
parameters. function
of tein
account:
adsorption where
parameters. to internal in
Langmuir
describe The
2
C The of
c
solution
Solution H (g/L)
and I pH
Equilibrium
0lumped
key
mass
isotherm,
the =PK(I) C was
adsorption
sequence.from
1model= k
4 transfer
variations 
1
are =Henry
chemistry computed V k
z Ha
k
2the
the
like (mAU
parameters
C P+ Tocoefficient
O 2
isotherm.
a of amino
the [6,
concentrations

dousing
the+)/2450 7].
I H
The so, + are
lumped acid
a is
the (mAU.L/g)
The obtain
×
following pK
defined
2.15I of
number a
Henry Equation
concentration
the as the
species
of
reactions of slope
mixing
positive
coefficient 9: in
k
inthe in of
the
solutionand
with cells
the the liquid adsorption
J
liquid
negative
pH and
(5)
were phase
and and the charges
takenionic C isothe
charac
solid(2)(1) is
k into ppiw
k IThe lumped Henry coefficient is defined as slope of the adsorptio
+ B k
+ model HP

used,
amino
ngth,the
fficient
+43−
4O
protein
+number H ++
meaning
acid,
8.06
4(ii)
isIn 2.2
definedthe Htheir
in
theof solution
HIC
and
1.92Modeling pulse
that
spatial
positive
literature, as
location
6.82processes
the
strength the the experiments,
two was
and
slope part
column
configuration
results I
[11,
contributions is calculated
contributions
negative
observedof
12]
extremely the volume)
are
and the
adsorption
often charges some
from
experimentally lumped
of as
high
and were
presented the atheamino
stored
were K(I)
function
(severalHenry
isotherm species
then
protein
N
= 1Calculations
acid
at
K(pH)
computed
inH in
+
Interactions
+asummed
coefficient
kmol/L)
This
-20°C
terms0at
the 
+
of
charge.3 a
residues
adsorbed
Solution
H does
the low N = pH1I
literature:
O 2t=
provided
of H

and
before
number
at b not Ieach
concentration.
k from
to
+ can
,
may
+ were
it
on with
b +
k
chemistry
H
which obtain0 2be
εaSEC
is playzt×
+ 2 ak28,
the
and
pH be×
the
known
C pH
first
first kthe
4Isurface
calculated
1
is
poorly
analysis
pK
based a − amino
the
+related
performed
curve role.
b solid
that of
9.23 net
×
The
on+each
accessible
or to
pH acid
from
to
phase
2Purification
the
of determine
following
K amino
assuming
2 presence
the the
as charge because
The
measured
follows: ofof vs
the
reactions
acid,
that:
protein
salt
and pH“buried”
interactions
content retention
thein (i)
significantly
to by perform
solution
two
the
ofHIC in
(2) with
LMW, pK
the
time
contributions wereathe
(5) tof
simulations.Monomer
R
(3)
the
solid
taken am
we i
H
 he 14.00
O
e9.23
he
− mber
nt
rations
P
olid
is
number+
3−
4concentration
following
O
pK
2.15 of
definedH
+of
phase +
+
a the
H The
positive as
of
+
reactions
core
impacts
The
where
the
Thenet
12.38
where
mixing
species as The
of inslope
12.38 charge
and
activity
the
interactions k C
Kactivity
strength
the
follows:in the k
in of
cells
solution is
values
is
negative
protein,the of
the
liquidthe the coefficient
J the
liquid
Ifitting
of
with
concentration
adsorption
coefficient
lumpedand
observed
were protein
phase
thus
pK charges
and
the the
taken of
not
[8, Henry
solid
solid C in
isotherm
the
ofkinto
characteristic
experimentally
9, were
the
contributing
10].phases,
is
phase
of
protein the
lumped
coefficient
indeed
account:
Considering
species
atK
protein γ
Besides,Truesdell-Jones
computed
K(pH) low
=C transfer
ε
krespectively.
independent
were following
K(I) 4 .k
approach
iin
k simply
to H
and+
time
in it
in
concentration.
solution =
described (1
P
2 b
solutionsolution
Equilibrium
Considering
2thethe
=
is O
a0the Equation

Calculationsa
clearat
0of+i
− ,
,of 0ε
3was
4literature:
apparent
adsorption A2
is
k
replaced
+b 
anumber H
respectively,
each
)
iinternal
1by model
the ×
used,
,athatwas HP and
computed
a− 1 pH +
Equation
the× ,ionic pH O OH
computed
bnet
I
the
k
0and
z
+
meaning +
4
2−
(13)we
,by k
isotherm:
were mass bbased
bthe
a
its +
2,pK
strength,
charge.
spatial
obtain:
a × as
which
× H valence.
using
2
bbased
8,
first
2aIpH are
2
we ofon
that
using
2 were
Tothe 14.00
6.82 the
fitting
each
obtain:
(ii)
performed
configuration the
pK
reduce determined
Truesdell-Jones
amino
thecontributions
Truesdell-Jones
model spatial
the acid,
assuming
of parameters.
number
the
from
and model
configuration
protein from the
that:
protein
model
of the
two [13]
(6)
For
fitting
may (i)
pulse
species
[13] (3)
contributions
theof
(2) the
pK
experim
adsorprotof
cient
we sed number
of
protein
of
The
the + − the
each
obtain: H4is
assuming
where defined
following
charge amino of
particles
in A positive
solution
vs ,
8.06 B
that:
acid,
reactions aspHare , thewas
Cand averaged
to and
(i)
UVin slope
are calculated
and
the
perform the
solution negative
3
saturation HMW.
two of together
pK the
contributions
were as
simulations. a of charges
parameters
was a
adsorption
taken in function
the C
observed k into .
amino were
The
were and of
isotherm
account:
To H at C
pH
then computed
activity
acid
IP k
improve is
around
OHP =
experimental
from Nthe λ side
charge.
summed H k at γ
coefficient
H the
ionic
Equilibrium
+ 2500
k 0
the C
low
2 P at
chains
k amino
to
O N 
1 kmAU. each

strength
1
description H
2
data,
concentration.
obtain + ε γ  + H k are acid pH
Hquantifies
+ 1
the + 2 as assumptions
+2.15 defined
net ofconditions on in pKaEquation
were
Kfitting a =as progressively
λretention
γk (Cof5. The →
− (8)0)k . tThe relaxed. Indeed, a
b9.23
+ In the literature, results are often presented in terms of 3k ,,OH which ,2is ×related to kK follows:
−14.00 a0
H SO hthe tain:
+
4 +
amino 6.82 H solution
acid, as k follows:
1.92
and k
was2.2the k
calculated
Modeling two contributions as
part a a func-
were shown then in flowrates
Equation
K(pH) summed O = 2− under b 7: + to P non
b+ obtain
,O a×
3− t binding
pH −+ εH −
+ the bnet 12.38 2 (van
2 Purification
Deemter plot).
protein by0)HIC
located in the pores of the particles 0b are ,pH 2averaged together (3) of
ε2accessible k in C activity co
4Vand 11
ng that the 3the k4−
acontributions from the species adsorbed on surface or
kinetics To as simulate follows: the experimental kmodel
pulse experiments, the lumped λdata Henry acquiredcoefficient by H GSK
can O
charge. tbe in H the
calculated + aframe bof
from the measured
14.00 time
pK where C iswas the concentration of species in solution 4 and
where K iscalculated the alumped aHenry coefficient εfrom a ,z(9)a− its
,ionic valence. are λparameters. →
− For
4 K =
were rm: HSO determined
+ +parameters,
impact from
its apparent protein
Conly the net pulse
pK charge: values experiments and
itquantifies has This
number
been performed
provided of
independent
reported e col that bfirstat 1of different
the curve
aspartic/glutamic the pK eof of the charge
strength,
b1amino acids
acids vs (ii)pH
and the
depend tohistidine perform
kspatial strongly kγ (C
kresidues simulations.
configuration kon Toth i
+ H 1.92 ,
umber protein of solution,
positive and while negative k the coefficientcharges were computed the at adsorption
each pH based ,strength 4b + on k with
− R
each
obtain: of(ii)
erformed
2.15 2− the parameters
amino
the
4 9.23protein pKspatial
K acid,
assuming
a= in
K Akand
λ where
kand
γ ,k (C
solution
where
configuration
= B λ ,the
that: γ →
− KCkare The
(C k0) is
were
two is (i)the

− the
material estimated
contributions
0) lumped
the ofconcentration the pK eluting asHenry
kby
protein
K of
K(pH) =comparing
were
from
theε coefficient
function
of H+ =species
does then
the
amino
(1
P b tK(I)
O − K(pH)
of
+ model
=
HIC
H not
ε bpH and
ksummed
) P= Hin
Calculations
acid
× OHP play
column
a 2pH =
Oa
simulations
solution  +0(9)
O side k + 02−
a aH 10the
a + 1toH
b + was
× + P role. 1obtain
chains
× amino
and
I2+
H
O × +
pH +
wereb2that
OH
collected pH with
0+ a z,2+ ak+ KbH ×
6.82+
acid
1its ,+experimental
the
first
are I 22 ×
valence.
inare
net pH
performed
2.154.7 fitting
14.00
mL model
data.
assuming
fractions parameters.(corresponding
(6) that: (3) (i) For (4)
the (2) to pK (3)the o
O +
ptions
les
xing 4 +
are
2− H
12.38
cell tion
were
where
averaged
14.00
model +as k of
progressively
were C 8.06
was
follows:
In
k pH
together the
pulse
consideredused from k
C k
literature, in
experiments, k
kto describe the
C relaxed.
kask . the amino
The adjustable concentrations
results activity
the the acid
Indeed, are a
hydrodynamics
lumped coefficient
often
parameters. the of the
presented
Henry ionic
non-ideality
i 2
Hydrodynamics
experimental
The γ 0R
H
inRegarding
speciesquantifies
k coefficient
4 K(I)
approach
This 2 N strength
SO
i
the
3 1 in H 
1 Qof

+
provided
4 k
column,
terms
= 4 0
data,  in
the√
Calculations
ε a can
usedthe
HSO 4Nthe + √
1encountered
protein
and of
the H
ε 2 2
a data
bewhile
a k
in

liquid +
× , +
kinetics
calculated
ε
assumptions
first 4
this H
which
I H solution,
the
+ 2
were and
curve a
article were is
× solid
first
related
I
offromextracted
1.92
To2 while phases,
9.23
were
can the simulate
performed
the
be tothe K from
measured
progressively
charge extended respectively.
coefficient
the
as
C vs assumingthe
=experimental
follows:
pH toλ literature
retention
relaxed.
toany
γ λ Cperform
k A that:
other[Eq.
quantifies time [14, 9]
data
Indeed, (i)
kind 15]
t
simulations the
acquire
the (2)
of to
the
a
pK
adsoper
ads io
H+ditions SO
ch −
c the medium.
amino + (van
Hacid, their
The Deemter
and
pulse location
8.06
parameter the experiments, two
where plot). [11, λ
iscontributions K 12] was is and
the estimated some
lumped
lumped were amino by then
Henry acid
comparingsummed coefficient residues model
to obtaincan may+ 0

simulations
be be
the 3 4
calculated
1 poorly net e  with accessible
+ 2 from the because
measured “buried” retention in the
time t(3) R
ofmat
aHenry coefficient and a ε,H aanalysis ,3kpositions
a ,H bin ,+ b1adsorbed
,presence
barticle
2 are fitting inmodel
in HIC processes extremely high (several mol/L) and k parameters. For(and
,harge number 14.00 of positive and negative charges were computed at εit each is known pH based a,that the
on of salt kassignificantly
2.2.2 Mechanistic model 4 t
2
+ N H t SEC N H + 9.23 k LMW, kconfiguration
4i
eelution,
H 6.82ength,first4 the performed
2.15vs (ii) pH the K to kCspatial =In
assuming
perform λthe column
kkγ literature,
(C
configuration that:k
simulations. volume)

− (i)
0) results the and of pK
To are stored the often
ofthe
improve KHPat
the
protein presented
= -20°C
Oεamino 2−independent
εe4HSOthe
iVH + does before
The
A
(1
P bdescription
0O P acid
− in
zO −
O 2approach terms
3−
bnot ) side
Imedium.
1and 2 × − +
HP 0 play of of chains
O + 1k the
of2−
used 2 + which
ionic
× are
12.38
role. H (9)to
0pH 2is
strength,
+0determine
this related
6.82 to
(ii)
canthe K bethe
content extendedfollows:
spatial of to (6) any Monomer
Rother kind
arge this of
pproximation
article ++ article
performed of
1.92
can thesequence.
protein
lumped
while
9.23 be can the
core was
extended
assuming
where
protein be The in
approach
coefficient
of asTo solution
used
extended
the
inK = follows:
activitydo
to λ
isprotein,
solution to
that: any
γ so,
the k is
λ
where C kdescribe
towasthe
used,
quantifies
other k any
coefficient
lumped was (i)
thus K number
calculated meaning
kindisthe
other the Henry
not
calculated the theofof pKof
kinetics
adsorption
kind the
contributing
as
that
adsorption
lumped coefficient
athe
of
of
protein
as
function
of HenryK(pH)
adsorption
log(γ athe
Langmuir
tRεlog(γ
model,
contributions
strength
chromatographic
the internal present
isotherm,
function)inamino
and
to = of
solution
the−
=
coefficient
colHa
experimental
only
pH
mass
isotherm,
adsorption
of ,

Pwithk
study,
independent
like
apparent acid
a
+A
pH
4 kfrom
,
4the
k the
was √z1SO
transfer
a from
from
k 1 a
side
, (8)peak
− H
2−
blike
computed
+
pH
theInet
mixing
data, a
isotherm.ε, ethe
C P
0 +the b , of+amino
O4[6,
the
chains a
H
,
×
charge.
− bspecies
the
1 b
+
the ,The
2
I
amino 7].
cell
+ a are H
using
2 , acid
The b
parameter
model
ionic
are

assumptions To ,were
fitting
acid b
8.06 ,
2.15wasb
strength,
concentration
the1
reduce
reportedareλ on
used
Truesdell-Jones
model
2 were
k
the
the
fitting
was
(ii)to
parameters. in
progressively
number
the
surface
describe
the model
estimated the original
spatial
model
of liquid or parameters.
the
fittingFor by
(7) [13]
relaxed.
articles
phase comparing
hydrodynamic
configuration C For
Indeed is ino
e= 12− +  K (4)
H extracted 9.23 from as the k
follows:pulse literature k
experiments, k
[14, 15] the to lumpedaperform Henry preliminary )external
=
coefficient 0−3 4 4 2tests
can √ 4 2 + of C the 2× − +Inet + (7)
ofparameter
each impacts
amino acid, where the and values Vspatial the isof oftwo the pK contributions
column [8, 9,charges 10].
volume, Besides,
were then
the itkLangmuir is summed clear that porosity to the
obtain tbe kspatial andcalculated the Q configuration
the from
flowrate. the ofmeasured
the protein retention may time tR
1P  εside
0play 1first 4  k
of
+ s−To
herm. he 12.38
ionic
st
ters
.
a H
o,Hwere
the
The
+simulate
the are
strength,
Interactions
performed6.82
8.06The positive
progressively
number
located
charge
2.15
the
concentration λthe
number
Solution kvs
concentration
(ii)
assuming
ofwas
in and
2.2 positive
pH
in the the
with
experimental
of
in HIC negative
estimated
chemistry to
pulse
Modeling and
relaxed.
pores
mixing
the col the
perform
processes
in and
liquid
HMW.
that: experiments,
the cellssolid
charges
by configuration
negative
the
part phase
liquid data
comparing
Indeed, The
a(i)
J particles
simulations.
is phase
and the
were
extremely
C acquired
following
phase the
isthe pK
model are
indeed C
of
The
lumped ionic
experimental
characteristic
the
were
high by
of
averaged
reactions
acoefficient
is H
Tosimply
Linear K
kprotein
interactions
2
indeedthe
simulations
K(pH) GSK
whereSO
Therefore,
strength
computed
improve
Henry
(several
= The Q
4
ε HP 
This
amino C
together
replaced
in
Driving
H +
simply
ε in
1
net
data.
time
= kO +H
does
HSO
εP
coefficientwith
mol/L)
solution
VThis(1 b ethe
and
1
provided
O
with
B −only
encountered
3V at
the
charge
+ of
+col− acid O

εby
Force

not
provided
B
replaced
4 b
adsorption
frame
C
ineach

4I
internal
) +
Pthe
the
description
εand
kwere × C HP Oare
1 H
canof
I kpH
H .aathermodynamic
3−
pH t+
approximation

solid
it 1εthe
O of
2 The
takenby
ε
P
the−
be
k +
+a 2−
mass
is
abased O
kchains εH role.
4phase
calculated
protein
known
first
b +
e
curve
isotherm.
concentrations
1.92
activity
×
+
into
H of Hare
on
+
curve
pH
of
were
212.38
The
in
that
account:
was
6.82
the
of
The2.15
parameters
fromapproach
coefficient
solution
Purification
the the
charge
described
used
concentration
of
the
presence the
chargeto
vs
γused
measured
was by
species
kdescribe
pH
described
quantifies
calculated
of
vs of in the
protein
salt
pH
6to thiskin perform
retention
the
to in the
in by
significantly the
article as
perform the
kinetics
(6)
liquid
HICa liquid simulation
paragraph
time
function
(3)
phase
of
simulat and
t inte o
nry trength,
o
srotein + so,
article +the
2.2.2
coefficient
Hfollowing pulse
(ii)
number
can2.15 as
parameters,
the follows:
experiments,
Mechanistic
be
is defined of
extended where
spatial positive only
as K kconfiguration
the model the
to is
and slope the
any pKlumped negative lumped
aof other values the k Henry Henry
ofof
charges
kind and
adsorption the γ
number
of C
coefficient
protein
were .
adsorption
isotherm of
computed can accessible and
does be 4
at k
isotherm, calculated
alow not
at , a eeach
aspartic/glutamic
concentration. , 4
play t,pHlike from
b a+ , basedb role.
the the
, b measured
on are acids fitting and retention
model
histidine parameters.time residues t For R
i
performed impact assuming its as apparent that:
follows: (i) net the charge: pK isof it the
has k amino
been t=
reported acid i = e 2 side
col 0 that i
chains −
1k 1H + are e 2−
2  K 2 (4)(3)
concentrations
were4 1.92 +
curve + reported
12.38
in of the
solution of
charge the
in
adsorption was the The
species vs
calculated original pH ionic
isotherm: to in strength
the
performas articles a liquid function Isimulations.
aand(and defined solid not
pH HSOas:
kphases,
the
from k
To
lumped
t
not K K(pH)− full=
improve
γ
R
the
the respectively.
 C εsummed
experimental aminochromatograms).
SOH .
iapproach +
hydrodynamic= (1
P
2− 4the

b O −
1 0 +
acid + ε1the
 b
descriptionA 1) −
is data,
+ × HPεpK−2
used, pH 4aO
/ K k
the
0 of
kinetic +
meaning
8.06 b 1amino
of + 
assumptions
× H 2pHacids
+
parameters.that 6.82 depend
thewere contributions strongly
progressively on R relaxed.
from (6)
(4) the Indee
species
c+K del
mptions
y, − aH
which was
of
strength, each
+ were computed
used
the
6.82were
as amino
where(ii)to
non-ideality
follows: Equilibrium
progressively
determined acid,
the at
describe
impacts
A each
,A as Band
spatial of follows:
,in from pH
the the the
Cperform based
relaxed.
values
are twohydrodynamics
configuration
protein
protein 3 the on
contributions
of
fitting the
solution,
pulse
pK Indeed, pK [8,
parameters of
experiments while
9, in
were
The
thethe
10].
The the
sequence. R then
the protein
lumped
ionic
keyand H
column,
k
4 performed
Besides, coefficient
model
HP2 SO
k
Istrength To
Henry
experimental
is OεImay
2Q does
ite44the
4 2−
Vcando
is√ 0while
parameters  4 HSO
clear toλ
so,
coefficient
at
ionic knot
P 1 obtain
encountered 1
O the
quantifies
the
different 4[Eq.
that− 3− t data, +
playε
15the
strength number
−+tare
H 7]
the
the 4 is the
εHeflowrate. a
+
the2 net
the
defined
spatialcan
role.
as of 1.92

number
assumptionsbe
adsorption
positive
12.38
defined as extended
the
configurationofinmixing and
slopestrength
were
Equation tonegative
of any
of the
cells
progressively the with other
5. adsorption
J protein charges
and
The kind the were
may
relaxed. isotherm
character comp In
ε12col
pulse experiments, the lumped Henry coefficient e
dmber pK of each were amino V
considered kacid, is ,kthe and as column Ctheadjustable two volume, contributions parameters. εHenry the external
were then porosity
Q summed ,be
and to calculated
ε,role. Q4eobtain the from bnet the measured 6 retention time tFor
on strength, +8, H we 6.82
obtain:
(ii) 2.2.2 the Mechanistic
spatial B configuration model of athe aprotein does  not play 
m.
mic
Hn
re
8.06
he
is of
+
meaning
ve
+ the
first
defined
was of
The 1.92
assumptions
aof
parameters
Solution
the
charge
that
the
12.38
used
their
performed
concentration
positive
as
number the
charge
the
to
where vs
location
were
and
slope and
chemistry
chromatographic
contributions
described
describe vs
pH
assuming
colnegative
 pKof progressively
where kthe
pH
6to [11,
of + kK
adsorption
each
the
to
k,12]
the in
charges
The that:
medium.
from
is
perform
kthe and
amino
kinetics
are
liquid
the − relaxed.
following (i) some
were 3isotherm
simulations.
lumped
paragraphs acid,
fitting
species
the
The phase
of
amino
computed
simulations.
pK Indeed,
reactions
parameter
internal
eatparameters
adsorbed
aClow ofacid
above
The
k atTo
the
is
the located
in ε indeed
coefficient
each
concentration.
net
mass
improve
on
residues
solution
λ
Vnumber
To
amino
ionictwere
HSO
and
the
waspH
charge
improve= in IHP
transfer
surface
simply
strength
and

based
=theacid
estimated,
were
estimated
and 
is
O
1
the
of e− a2−
Vthe
4poresSO
pK0 be
theonside
taken
the ε [6,
or aawe
description
 replacedionic
zpoorly
1encountered
2− 1k 2 P
protein
+ of C
byof chains
7]. +
descriptiona− O
into
keach2and the
H
3− strength
,comparing 1b+accessible
0− + in
account: are
by
,particles
aminoK εH
bpK 1eof
+
,of
solution as are
adsorption
28.06
of model
acid,
defined
because
12.38
are fitting
wasaveraged
and
in
“buried”
model
calculated
isotherm,
simulations
the
Equation
two together parameters. such
with
contributions
in
as 5.
the
a in as
The
function C
the .R
k (5) The
were(4) of thenacti
pH
nding + H + conditions 12.38 parameters H
The (van
as 2 Oimpact
where follows:
ionic
activity Deemter A H wherekV
its
strength , + B coefficient
kis
OH
apparent
, Kplot).C
the is I were the
is
column net
defined
of lumped
estimated
the 14.00
charge:
volume, protein Considering
as: Henry by it
transfer
εto in has coefficient
comparing
ethe kcol
solution been
R Equation
t
H t, SOA 4 =reported
respectively, was z
model
Q and  8, col
I
computed HSO a e a4 ,
simulations that a 1 obtain:
which
+ , 4athe
+ ε H
using, b+were ,withb
K the , 1.92 amino
determined
b are
experimental
Truesdell-Jones fitting acids from depend
model
data. protein
model parameters. strongly
[13]pulse on
experimen
For (4)
γexternal porosity and Q the flowrate.

− a
+
umptions en:particles
inof
+ solution
ionic
amino the are
charge
strength,
8.06acid, core were wasaveraged
and vs
parameters
of calculated
where the
(ii) pH
progressively
the pulse +two the V K
together
protein, toa
experiments, =
perform
kspatial
contributions colA askλ
is the ,akthus B
inγfunction
relaxed. (C
C
,+ Cknot
kconfiguration
kcolumn k. the
simulations. were The
were of 0)
contributing activity
pH estimated
Indeed,
volume,then
lumped fromof tRεHydrodynamics
To
summed Ccoefficient
the
log(γ
Henry = improve
the
thethe eby
the = amino
the
protein
ionic
)toλ comparing
=
external obtain
coefficient
non-ideality −
i γC
the
apparent2 R 1kH acid
kdoes+
strength k
quantifies
description
the
SO 4
porosity 2
k
can √ and
model
not
4net
net
kof  be 0 K
+the play
encountered
HSO kinetics
charge.
and Ccalculated4of
simulations
1
protein× −
a4Q 2
eI+ 
role.
the (9)
0
To H To 1
+reduce
from with
flowrate.
solution, simulate
2
the
1.92 experimental
the measured
while thethe
number experimental of data.
coefficient (8)
retention 6experimental
fitting (4)were (7) data
time
λ tacquired
quantifies
were
snumber
parameters.
H
tein
sumptionsofcs
that ssumptions
+
netics
+were
protein H +first
Towere
solution,
positive and
1.92
first
in of
To
Equilibrium
performed
experimental
simulateextracted
1.92
and
were
2.2.2
the
Solution
mixing
as
solution were
performed
while N two
the
2.2.2
follows:
simulatenegative in H
progressively
6
theirHIC
the
was cellsMechanistic
4progressively
pulse 
contributions
assuming
data.
experimental
from chemistry
the
The
col Nthe
location
Mechanistic
processes
assuming
coefficient
charges
calculated Jionic H
experiments,
experimental and +that:
3literature
relaxed. were H λ
strengthdata
[11,the
as
were
is
that:
model
The
relaxed. 12]
extremely
quantifies
a computed
(i)pK then
acquired[14,
characteristic
model the
data
Indeed,
function (i)
I
the
followinga 15]
and is 9.23
lumped
the
pK
some
Indeed,
acquired
the
defined
sequence.
by
atto
high
of the where
pK
e
GSK
aperform of
reactions
adsorption
the
each
pH
k
charge.
amino
flowrates
Henry
ionic
the
time
(several
as:
from
k
by
of AQ
Simulations
in
the
present
pH
TheTo
,Hydrodynamics
kthe
k
GSK
amino
theB
based
strength
the ofin

preliminary
acidionic
HSO ,do
kunder
net
1amino
coefficient
mol/L)
strength e1 and
frame
study,

solution
V
internal
aminoin+
so,
2−
residues
colon
charge
acid Cnon
strength
Bthe 
encountered
the
of

Q
and
a acid
with eare
acid
All can
SO
frametests
I
mixing
side
were
bindingnumber
massmay
it of
1three
simulations
2−
side be − isof
kthe and
chains
taken
εknown
encountered
+ of be ethe
calculated
cell fit-
H
chains
of εpoorly
protein
conditions
+ einto
model
positive
kinetics
are Langmuir
+that are
in
account:
presented K
accessible
from
8.06
the
was (vanand
ksolution
=To theλ in
presence
used
negative
simulate
adsorption

Deemter this was
(C
because
measured
tok kof
describe →

article charges
calculated
the
isotherm.
plot).
salt 0)“buried”
retention were
significantly
the The as
performed
hydrodynamics
kin timeacomputed
the function
R data
tR witat
st sed l e curve
ionic
model
ositions in this K of
8.06 Hparameters,
summed
strength,
was k
were
the
=
The O
article
used λ 
k charge
γ
reported
H toto H
Equilibrium
lumped k (C
as
can The
(ii) +
where
obtain
describekfollows:+
be →
−only
vs
the 
inionic OH
Henry 0) pH V
extended the
the
spatial the
col
− to
strength
net is pK
coefficient perform

hydrodynamics
original the k
charge.
a values
configuration
to column I
any +isis
14.00
articles
and
simulations.
defined
other defined volume,
pK the
ting
in
(and
number
of
kind
the as: a
number theε
parameters,
the
as of
column,
not
To
t
sequence.
the
e
of
R I the
protein
adsorption
Calculations
the
==accessible
improve
chromatographic
the and
slope external
present
while
full HSO pK (9)
To
and of
4
does
chromatograms).zthe
the
the
k
k
2−
do of 1
C
Iapparent
isotherm, were  
aspartic/glutamic
porosity
study,ison
not
k + each
so, description
adsorption SO
the
4
the
play medium.
first
2−
amino
a ionic like and
mixing +K
number a
performed H Q
role.of
acid,
isotherm
the the
The
cell acids
of and
concentration 8.06
flowrate.
parameter
positive
model at and
the
assuming low two
washistidine
in and λ
contributions
used
the
concentration.
that:
k was
negative
liquid to residues
(i) estimated
describe
phase 6 the (5)
charges were
ofCkfittingpK (4)
the bythen
were
is aatofunctio compa
hydrody
of sum
the com a
3P core Owhere of H
the Pspatial O
protein, +of H thus 2.15
not contributing to Vthe net charge. inTo reduce the number
Regarding the a data that
C is the concentration of species kεchains in solution ε5and εwere zksolid a extracted
its valence. from the literature [14, 15] perfor
where 2
Cwas and as follows: 4 4
+
no eperformed
um.
umber gthe H
thecell+
determined
acid, The model
ionic
ofand
assumptions 8.06
parameter
positive
where
were Solution
strength,
the
Interactions
assuming k+
where from
two V impacts
considered
Interactionsused
and
were λC
contributions
Athat:
k is 4to
(ii)
negativewhere
kprotein was are
chemistry
the
,progressivelyB the
describe
the (i)with
as the
estimated
column
,pH
2k
C+V
values charges
adjustable
the
with col concentrations
pulse were
are 4the is
pK the the
volume,
the The
by experiments
then
relaxed.
3obtain: werepK
hydrodynamics
configuration
solid column
offunction
fitting following
comparing summed
parameters.
the
a
solid [8,
computed
εisphase of
9,
the
Indeed,
amino
eparameters
a
phase
the
volume,
10]. to
Linear
external
The species
performed
reactions
model
of Besides,
in
obtain
at
The
acid thethe the
each
approach
The eThe
εDriving
simulations
side
andionic the
protein
the
interactions 2k
pH
porosityin
column, Q
it
interactions
I
in
net
external
eInet at
is
solution
based
strength
is 1=
used the
Force √  charge
clear
thedoes 1different
with
and while
are liquid
in with
ionic porosity
approximation
that
were
not
Q
this
encountered 12of
with thethe eand
− the the
play
article
strength taken eflowrate.
the andprotein
spatial
solid into
can
solid Q

role. phases,
as the
was
phase configuration
account:
be phase
defined used
extended respectively.
solution
flowrate.
were todescribed
were
in
was
describe
to
Equation of any
described the calculated Aother protein
the
by 5. 6The
the
by kinetics
kind the
as
may of of
adsorpt inter
mation rst n protein
cle
tics curve
isotherm.
rmodynamic can was inN
of
Considering
be
To solution
H
used
extendedthe
Calculations
The
simulate 
Solution
parametersH
to chargecol N O was
concentration
describe
to
parameters, H
Equationk
the
− any Thewere
 calculated
vs + H 6 +
kchemistry
H
other
experimental
described the ionic first
+ k8, 6to
OH
onlykinetics in
kind 2−westrength as
perform
performed

thein
the
a
of a 9.23
the The
liquid
ofpK+
adsorption I
internal
data paragraphsfollowing
simulations. 14.00
values phase defined
charge.
strengthof
acquired pH
mass
isotherm,
and
model, C the as:
from
experimental
reactions
independent as
kabove To
is
transfer
t
number number
Linear
defined
byindeed
only = the
improve
like The
Iwere GSK = amino
thein
k[6,
zof
col
Driving
ε 2 ofnet
and in
simply Vdata.
solution
7].
estimated,
accessiblein theEquation charge
pK
1 acid
z the 2+
C ionicz
Force
description
replaced of
offrame C
were k 1 of
each
kand  −
aspartic/glutamic 5.
strength, the
taken
approximation
K ε amino
by
of protein
of
indeed into acid,
(ii) simplyin
account:
the solution
and
was
acids the
spatialused
replaced and two wasto
by
configuration
histidine γJkC calculated
contributions
describe . articles residues the
(5) as
(5)
(4)
of were
thea
kinetics functhe
pro
H P OIn the HP literature, O is+ Hresults 6.82 Iare often presented kthe in peak terms 1positions ,The which were is reported
related ininsurface
to Kthe as original
follows: (and not
the lumped
protein in approach
solution 3 is used, meaning athat the contributions A from e Iisotherm.
col the a k
kspecies adsorbed
e on the or
4was calculated as function
εof pH sequence. from the To amino do so, acid the number of positive and negative charges were co
2 The ionic strength is defined as:
aroximation resented
ons
h
rength,first
amino (van curve acid,
(ii) in was
Deemter
the this
of4
and
following
The used
the
spatial the article
2Equilibrium
impact ionic
activity charge to
plot).
two
adsorption its
describe
configuration
strength werevs
contributions apparent pH ,performed the
to
isotherm:
Ithe of net
perform
definedkinetics
were the charge: then
aprotein with
as: of pK
simulations.
Langmuir summeditthe
internal
The doeshas software
key been
not R
to mass To
tobtain
model
adsorption Rplay =reported
improve transfer Ypso-Ionic.
the
parameters
abased 2role. net kthat
the [6, + 7].
description the
are pK
the K of
number of
concentration amino acids
ofpositive
mixing depend
the cells liquid strongly
andk phase the C on
characteris
k is (4) indee
Jcoefficient of the protein in solution was computed eusing the Truesdell-Jones model [13]
where V is the 4 column volume, a the external Q k porosity 1first and εcoefficient Q the a flowrate.
number,ell the
the of
assumptions
number H positive
assuming 3of
following
parameters
Ppositive
O
mixing
whereN  and
H were
that: Equilibrium
+
H cells
C negative
2 P
the
A
adsorption
progressively
col isON kpK−
the ,+
H and B+ kof+charges
H
concentration CH
+
the
+ isotherm:
were
−amino
were
relaxed.
characteristic 2.15 estimated computed9.23
of The log(γ
Indeed, pK
time
K
species eby
Calculations
parameters
k= )comparing
at
charge.
of the=
kλThe sequence.
each
The
internal
kin− ionic lumped
γ1concentration.
(C
solution pH
key A were ,
2mass
strength

− B
model
model √ , To
0) Henry
and
and k+ do on Csimulations

parameters
C performed so,
encountered
zcurve 2×
were
its Ithe valence. number arewith is
assuming the
Hydrodynamics
experimental
ofnumber
defined that:as theof and
(i) slopedata.
mixingand the negative of
(9) pK the
cells
kinetics. (7)
a of charges
adsorption
Jparagraphs
andthe amino
the were iso
cha
The
he
fficient
sach /of
ata, med
tracted are the
concentration
kinetic
number
modellocated
the
the is
assuming
charge parameters.
definedof
number was
assumptions
from in
vsthe HPthe
4
used
as
that:pH
in
their
ofH were
the O to
the
pores
4 O
mixing
literature
2−
and
(i) to
slope
were 
locationk
perform 
liquid
considered
the negative
of
describe
H P6 of
4cells pK
progressively
2−H[14,
O
the 3phase
the
+
3−
[11,
simulations.JOH +
particles k
adsorption
of as
charges
and
15]the H
12]Cthe +
kadjustable
to
is
and
the amino indeed
were
hydrodynamicsare
relaxed. 12.38
isotherm
some
characteristic
perform To acid
14.00
γimprove
simply
computed
averaged parameters.
amino
Indeed,
transfer
This
kaat
Therefore,
side the replaced
low
thethe at
together provided
in
chains
acid
time number
the
tat each
kthe 
,1obtain
description only
residues
+ of
ionic
respectively, kkby
are IpHin
column,
internal
Btests = Q a
the
and C k based
strength first
I2= ofof . may pK 1
thermodynamic The mass
k
z
on
while
kkkbe
which of C activity
encountered k 2 ε
each
poorly of
e the
were
the amino charge
parameters
coefficient
accessible
determined acid, vs γ
and pH
becausefrom theto
described
quantifies perform
two “buried”
protein in
contributions simulations.
the
pulse in the (5)
experiment were To a
th
kpreliminary ,/the
as P follows: a Cpresented .ε Considering k k a k

2 The 4 ionic 4 a εand
eh, fmptionswere
each
obtain:
amino
(ii)
performed the determined
amino
the
acid,
acid
H 2acid,
non-ideality
spatial side
assuming O and
In
H
−from
H coreand
N SO the
chains
the
where PH
configuration
3
H HP
O where
the
The
that:
of+ 2two
literature,

O
protein
4from the V 6two
are
O HSO
col
contributions
Hprotein
activity
N (i) C H
6strength
independent
is + P
+ −
pulse
contributions
2
protein, k the the

of isH
O +
+the
results
+
+
thepK
H
OH
H experiments
column
+
coefficient +
solution,
thus + H6.82Iare
− were
concentration
protein of
+is
of were
the
not
defined
often
volume,
1.92
then2.15
ofwhile
doesindependent
estimated
aminothen
the
contributing
9.23
summed
14.00
kperformed
not
as:
the
not of Ithe
experimental
summed
protein
eacid the =
by
species K
coefficient
play
charge.
to
Calculations
transfer i of
comparing
in
external
hydrodynamic
side =to a tonumber
ε the
different
terms
k
role.
the obtain
solution
chains + zink2Equation
k ionic
t(1
data, λiC
Ithe
apparent ,porosity
solution
 kof−
and were
respectively,
model
the
quantifies
are
net
εk2the
was
strength,
ki )and net pK which 8,
first
andsimu- az
assumptions
computed
kinetic
net
t we kof Czperformed
k
the
k
Q
charge.
each
isobtain:
k
(ii)
which itsTo
related
the
parameters.
amino
the
valence.
using
adsorption were
simulate
flowrate.
To
spatial
to the acid,
assuming
reduceKdeterminedthe
as
progressively and
configuration
follows:
Truesdell-Jones
strength the
the
that:
experimental with
number fromtwo (i)(5)
relaxed. contributions
theofdata
protein
model
of fitting
thepKIndeed,[13]
(6) pulse
a of(5)
protein wer
the
exp
which eimation were werewas determined 4
used
progressively where to
where describe C 4
relaxed. Vwith is
protein the is 4 thethe concentration
Indeed, pulse column kinetics experiments ionic
volume, of of species
internal
strength εeandperformed the k in
mass
encountered solution 1 at transfer different z [6, its
of7]. valence.
The approach ,used 3in this article can beprovided
flowrates extended C under
k2 = aIexternalto λfirst any
γnon kkC other porosity
binding kind kand
conditions Qasthe
adsorption flowrate.
(van isotherm,
Deemter like
plot). the (8)
2 4 a i k
onditions reported (van
the in
whereDeemter
2−the
Interactions
ionic  Poriginal
Asimulate
strength;λkN ,4In
The H
plot). −B +
4
3−
ionic kC
the
col articles
N are 4
H6spatial
+2−
strength 3the 3+ fitting
H (and ++ parameters
solid
con- Iimprove
+is
notdefined This
phase
lations the
9.23 withfull
The
independent chromatograms).
charge.
flowrates C Iby
interactions
experimental kis = = k√ the λof under γterms
curve k kC
ionic
the
data. zwere 2non
with C ionic1of kstrength − the εk
the
binding , charge
tstrength, solid defined
vs
phase
conditions
acquired (ii) pH by K in
to
were
the GSK (vanEquation
=described
perform
spatial inλDeemter
kthe γwith (C
kconfiguration 5.
k −
simulations.
frame → by The
plot). 0)
of the (8)(5)of
the Tothe impr pr
HP O P O k+ H 12.38 λas:
k
kH PO  the HP literature, O + Hresults 6.82 are often presented in kof which is related
− k k2−

− kto K as follows:
and
ng ulations
re charge
ttrength, kinetics
thenumber
conditions
the
performed
vs
presented
Langmuir pH
chromatographic
(ii) the Kto
(van
assuming 4 To
HSO of = as
perform
spatial in
parameters,
H 2 In
Deemter
adsorption
mixing follows:
this
k the γO 4
that:where
k4 The (C
4

simulations.
configuration SO
article medium.
literature, the
plot).
cells Honly
ionic
(i) Cisotherm. 0) P +
theexperimental
were O
is
J 4the H −+
the
strengthand
pK The
results+ of To
performed
pK H theaparameter
concentration
The
the
of values
Iare
the protein
8.06 is data2.15
concentration
characteristic often
defined
amino
the
with
and acquired
does description
the
number
presented
of
Regarding
k acid was
as: not
species Calculations
software
Ain
sideestimated
time play of
zthe 2in GSK
the ofaccessible
kchains 2 aliquid
terms
in I ofYpso-Ionic.
data role.
solution
 in (9)
internal ε by
are kthe
of
that phase comparing
k frame
first
aspartic/glutamic,and
were which
mass C of
zperformed
kis
extractedis
itsmodel related
indeed valence. simulations
assuming
acids
to
simply
from K
the and as
replaced that:
histidine
follows:
literature by (i) [14, the
residues 15] pK to aperform of th
re
firstns To extracted
parameters
mixing were simulate
performed f iguration
H
progressively from
where
SOthe described
parameters
following
assuming
HP the C
experimental
3literature
of O
4
H
HSO 2−
and
relaxed. 3the A
P
adsorption O
that:
k − in
where
C ,protein
4 B
+P  4[14,
the
k Oare
data
H
Indeed,,2k3−
(i)H C + C k 215]
paragraphs
the P
does
the
+
4
were
isotherm: is
acquiredO to
H the
−a+
the
1.92
pK perform
+ estimated
not
concentrations ionic H
concentration
ofby
+
12.38 Hydrodynamics
log(γ
the GSKpreliminary
above
experimental
strength 2.15
Interac by
K
amino )
in This
of= =were
comparing
the the
ε of

encounteredtions
acid + tests
frame data,
species
provided
species
(1
k
Calculations
estimated,
Regarding side −k ofwith of √ and
model
ε the k 1 )
chains k a in + in kinetics
assumptions
first
the C
t
 the and
simulations
solution
the are were
× k curve
data 
solid liquid
I first
and
that k
of To and with
were
the z simulate
performed
were
present its
solidexperimental
progressively
charge valence.
extracted
study, the
phases,vs a experimental
assuming
pH from
mixing todata.
relaxed.
respectively. perform
the
cell that: literature
model (7)data
Indeed,
(6)(i)simulations.
A was acquired
the the
[14, pK ionic
15]
a by
ofT
to
γcell 2 model
where was Cto is used the to ofk4describe
concentration 6 adjustable the of hydrodynamics species kindependent
in inapproach
solution the column, and zof its while valence. ,the
k
sch his ofwere
tstrength,the
article
were
charge
experimental
extracted
canC
(ii)
kdetermined k . be
vs
Thewhere
the 4were pH
fromH
extended net
data.
spatial k
2 Poriginal
C kconsidered
4theO In
k
from
perform
charge − and the
4to 
literature
configuration
− any
k C
HP literature,
protein
simulations.
are
the
as
O
other
k 2−
4 [14, 6(and
the
protein4
+kind
pulseof H
15] concentrations
a
results
the
+
of toTo in protein6.82
perform
adsorption
experiments
improve
solution
parameters.
are model, k
Simulations
often does the
ofThe
preliminarywas
i the
presented
only
isotherm, 1not
description
performed
calculated
the species
Iplay = All
peak
like
i
I2tests of
ain 1the 1−
simulations
role. of
kthe
termsused k
positions
2at
in
zas √
εrole. kk C
2 ionic
the the
ain
different of εC function thisliquid
k strength, article
presented
were which and
k of ispH
reported solid
(ii)
can from
in(ii)
related the
be phases,
thisindescribe the
spatial
extended
article
to
the K respectively.
amino
original aswere configuration
tofollows: acid
any performed
articles other A (and (5) ofwith
kind the
notof ta
nicce umptions
were
ameters.
odel was
strength,reported
approximation play
used
HSO
were a
to
(ii)
lumped
In
in
− describe
role.
H 
the
progressively
the
lumped
the
2.2.2
was SO SOH This
spatial
approach
literature, 24used P
2−
approach
4 Mechanistic
 theO In + provided
HSO 
articles
the
hydrodynamics
configuration
relaxed.
to H is +HP
describe
results literature,

used,is + O a
Indeed,
used, H
2−
f+model
meaning
4are8.06 irst
the +of not
often
meaning H results
in the
the
+
1.92
kinetics
the the
thephase.
full
6.82
protein
that
ionic
presented
present
are
column,
that thechromatograms).
often
experimental
The
strength
of does
internal
the
study,
independent
model, +
presented
interactions
while
contributions
Kin =termsnot B
encountered
contributions ε the konly
massa+ Iplay mixing
of =
data, (1 Awith k the
− aof
in
zfrom
transfer , ε the tthe
terms
the
which peak
) εcell
from zt kk2the
Iisotherm. solid ionic
assumptions tmodel
[6,k of
positions
isthe species
k k7].
related ,usedwas
 strength,
which
species
used
towere
were
adsorbed
to is
describe
K
adsorbed
to
related
as the
progressively
reported on
follows: the spatial
to
the
on in K the
the
hydrodynamics
the surface assurface relaxed.hydrodynamics
configuration
follows:
original orsimulations. or Indeed,
articles
(6) (5)of in
thth
(a
ve
namic tions
erm. of thewere
The The
charge
parameters sequence.lumped
reported
concentration 4 vs HP 2pH
described where
The
in To
Henry
O to
6 2−
the
4in do
activity
perform the C
in
4
originalso,
coefficient
k theP is
liquid O the the
4 3−
coefficient
simulations.
paragraphs number
concentration
articles
+
phase His + defined
C ofof
(and
above To
is the positive
12.38 as
not
improve
indeed ofthe
protein
Therefore,
were species
the This
LangmuirK and
slope
simply
log(γ
estimated, =
full
the εnegative
k
of
provided
solution
onlydescription
i replaced
) +
chromatograms).
= in the
adsorption
i (1 − solution
the
and − adsorption charges
a
k
was
ε 2
thermodynamic
k byi a√ ) first
k of i and
computed 1 curve − were
+ kz
isotherm
ε Ck its of
×computed
valence.
I the
using The at
parameters low
charge at
concentration
the each
concentration.
vs
Truesdell-Jones pH
described pH to
6 in basedperform the in model
the on
liquid paragraphs [13]
(6) phase
(7) C kab iT
ameters conditions
ion was
urve of the used
curve
The
are where
the (van
chargetonumber
Interactions
located
the net of describe
HSO
number
located A Deemter
vs
the
charge ,In
in HP pH −
B4chargethein
and
4
,O

of the
to
of the
2−
C4pores with
SO
mixing
pK perform
the plot).
are
kinetics
vs.pores
2− Pof
protein
of
4
the
pH3+ Ocells
3−
4of
simulations.
fitting
eachthe
H of to + solid
the+Jinternal
per-
particles
in
amino Hand
+ k
parameters
solution
particles phase8.06
the Linear
phase
acid, massTo
are12.38 improve
waswere
characteristic
are
and TheDriving
transfer
and
averaged This CIthe
Therefore,
described
calculated
averaged
the
k
interactions
ktwo is=
[6, Force
provided
together the
time λ
7].
description kby
together
contributions γ1as ionic
kdata,approximation
only of CakB
the within
internal kthe
1following
function −
strength
Cfirst
in of kε the .C εThe
thermodynamic curve
t ,assumptions
were
t
mass
twhich
.
k
solidof Theas was
in of
activity
pH
then phase
the
defined used
the
from
activity
summed
charge
column, to
parameters
were
in
coefficient
the describe
Equation
coefficient
to
vs
described
while
amino obtain
pH the
γkfollows:
acid the
γ
to
described
5.
quantifies
the by
linear kinetics
perform
The
quantifies net the (8)driv- in the of
simulations interna
paragr
ecvely,
sumptions
odynamic parameters.Considering were
parameters
Hydrodynamics 2.2.2 as Equation
H kfollows:
Solution
progressively SO
2 where
Mechanistic the
k
described  literature,
k
and8, chemistry
C relaxed.
HSO we
a 4 in
iskinetics obtain:

the
4model the + results
Indeed, H
−concentration
+The
paragraphs To are
the following
1.92
simulate often
ionic
not above
of γ thekpresented
experimental
Creactions
strength
species
the k .
were K
hydrodynamic
experimental = kencountered ε
estimated, in in
inhydrodynamic + terms
solution +
solution
(1
εtsimulations − k/data ε of theand
kinetic
ki and
I
) kwere k
acquired z/εthe εtaken
t k
parameters.
itskvalence. is related
into
 byofGSK were account: to K
progressively
in progressively
the as frame k relaxed.
ofwith Indeed,
(6) t
were number
assumptions extracted ofparameters
form mixing
were
following from
simulations. cells
progressively Athe
Hadsorption JSO 4 and
Bliterature
4To  , kthe CC relaxed.
HSO
improve characteristic
kisotherm: 4[14, + Indeed,
the H 15] + The
to time the
adsorption perform
1.92 key
of
ionic model
internal experimental
strength
not
isotherm preliminary
(1kthe Kparameters
mass ε=
iencountered
(Equation ) εin i+ data, (1 tests are− 1athe
εdifferentithe of
)− knumber
assumptions
tkinetic ing forceparameters.mixing
were
approximation 6 cells J on andwas (6) the
used relaxed. characteristic Indeed
(6)
ented
inetic sequence.
whichinthe
parameters.this the
charge.were
where
present
The
the non-ideality
To
article 6determined
HSO
do
where
2.2.2 C net
non-ideality
study, In
k
so,
2−
Equilibrium
were
and where
kcharge
the
, the
A  a
k
Mechanistic C ,
of
SO
number
from
performed
kBkare
literature,
mixing
the
of
k 2−
,
were
C
is
the protein
+kprotein
the cell
the
H are
estimated
of
protein
protein
+
results
model with
concentrations
3
solution,
positive
concentration
+ fitting
model
pulse in
solution,
are Kthe
8.06
was
by
kK
and
solution
experiments
=
pK =comparing
while
λaεknegative
software
parameters
often
used
of
of iwhile
to
+
γpresented
kthe
the
species
was
(C describe

the →

Ypso-Ionic.
species and
model
coefficient
charges
calculated
performed ik0) coefficient
in I the − kis
terms
solution in√
were
the kλat
hydrodynamics
as
the 8):
of ε
λquantifies
ionic
computed
liquid
kk andfunction
1 quantifies
, −
which
withzεkand
strength
its experimental
the at
of
is
in solid adsorption
each
valence.
thepH
as
related
the
from
phases,
defined
pH
adsorption
column, to
the
K
data.
based strength amino
respectively.
in as
while
strength
(9)
Equation
follows: the
acid with
5.AtoThe
ere ns determined
were the
bindingToconditions description
reported
number the
sequence.
Solution
from
the chromatographic
2.2.2
Calculations and protein
in HSO
(van
chromatographicof
pK To
chemistry
Hused
4
the
6 experimenta
Mechanistic
In

do
Deemterof
4 were
k
the

pulse
original each
so, SO 4
literature,
+ the experiments
medium.
first
2−
amino
plot).
The l +
articles
number
medium. data,
model
performed H acid,
following
− results The of (andtransfer
performed
and
The 8.06
parameter
are
positive the not
reactions often
parameter
assuming
t
two at
i
K
,
the
and respectively,
different
λ
presented
=in that:
k full
contributions negative
εisolution was
λ+ A
was
(1
1 estimated
chromatograms).
(i)k−zthe in 2 ε
charges
εwere
estimated
whicht
terms were
)from I pK takent byofwere
then were of k  
comparing
byspecies
kfunction ,
into
the
determined
t
describe
summed
which
computed
comparing account:
amino is the
model
to related
at
acid
from
kinetics
obtain
model each to
side
protein
simulationsthe K
pH of
simulations
chains net
asinternal
6experimental based pulse
follows: with
are on mass
with
experiments
(6)
p
2O
lumped approach a H iskof used,+
4 OH meaning 14.00
netics itionspresented (van Linear
The in simulate
Deemterthis
approach The
Driving the
parameters
article plot). net experimental
were
Force charge in performed A
approximation
this , B the
article
k data , C with protein
k acquired
were
can the was be
flowratesinthat
estimated
software by solution
used
Simulations
extended GSK
log(γthe
Hydrodynamics
tounder by
Ypso-Ionic. contributions
in was
describe
)to =the
comparing any
non −
kcalculated
frame
All other
binding the simulationsk ofimodel and
kinetics
√ kind 1 as
−conditions
+ athe
kinetics
ε ofC asimulations of presented
×
adsorption internal
I (van To ofadsorbed
with
in pH
simulate
mass this
isotherm,
Deemter from
experimental on
transfer
article the
plot). the
like the 7were amino
[6,
the surface data.
7]. performed acid or (7) data with acqu th
charge. the the assumptions
experimental number The2.2.2 andnet were
data. charge
Mechanistic
pK progressively of of
each theamino protein
model acid, in andsolution C the kk = two was λ kγ kcalculatedC εof tas kwere athe function
transfer of
(6,7). pHdescribed
The from key the model amino (8)
param- acid
ynamic parameters described in the paragraphs above Simulations were 1contributions estimated, All Csimulations and then summed
presented to
in obtain
this γk6article the net
were performed
Interactions Equilibrium pK tof
ata cellthat were extracted from 6with  the the literature solid phase
[14, 15] to The perform interactions preliminary with tests the .solid phase were by the
independent experimental Solution of + the data.
chemistry ionic strength, + The (ii)
following the spatial
reactions configuration in solution the protein does not play abased role.
kwere  taken into account:
located inN theH pores N
a H of +
the H particles are9.23 averaged together k in The activity coefficient quantifies
tions
extracted model
presented
Langmuir Thefrom was key the used
sequence. Solution
in model
adsorption this
literature to describe
2.2.2
articleTo
parameters chemistry do
[14,
isotherm. were so, the
Mechanistic
15] the
to hydrodynamics
performed
are
3 perform
The number The
the number
concentration model
following with
preliminary of in
apositive
the thethe
ofpositive
Regarding column,
reactions
software
mixing testsin K and
present =the
the of cells εnegative
in
the while study,
Ypso-Ionic.
iliquid +
data J+
solution
(1 and B −the
that phase ka charges
εithe )Iadsorption
mixing were
were C characteristic εwere
taken
is
extractedtkcell indeed computed
model
into was
account:
time
simply
from the at
used
of each
internal
replaced
literatureto pH
describe by mass [14, 15]the onto hydrodynam
(6)
perform (6) p
2.2.2 Mechanistic 4+ 6 model
ak ticpositions
oximation parameters. relaxed.
Calculations
This charge.
following were Thesequence.
H
provided Indeed,
reported
The O lumped
adsorption 
were Equilibrium
lumped H the
aC To infirst
firstHenry + do
ionic

the OH
isotherm:
Henry curve so,
performed originalstrength
coefficient
− the coefficient
of − number
the articles assuming is
14.00
charge+ defined of
isacid,(and pK
definedvs as
that:
anot
pH the
as K
the
to and
(i)
the slope =
perform full the ε negative
slope +of pK
chromatograms). (1 the of −
simulations.
a the of charges
[Eq.
ε 1approximation) the
−adsorption k
ε8]t were amino werek eters
isotherm
Tothen computed
are
acid
isotherm
improve thesideat number
low at
chains
at to
the low each
description of
concentration. mixing
are
concentration. pH 6 based of cells on
J
.was used to describe ,the kinetics of internal mass transfer
H i [6, 7].
non-ideality 3P O of the H Pprotein O 3+ Hsolution, 2.15 while the coefficient λ 1quantifies the adsorption strength thewith
2 iliquid
ere reported
2.2.2
the γ Ctransfer
Mechanistic inthe
where the number
twhere Coriginal
i ,Interactions
model kEquilibrium
Solution
respectively, and A and ,articles Bk4 pK are
chemistry Cawhichwith the
(andof 2are each concentrations
not4were
the amino
the
fitting The determined
solid full
model, pK
following of
aLinear
chromatograms).
phase
parameters and
only the from the
reactions
the species
TheDriving
and two
protein
peak I contributions
interactions k
in Force
ispositions in solution
the pulse the ionic kexperiments with werewere
−strength εtand the taken
reported solid
solid as summed
was
into
performed phases,
phase
defined
in used
account:
the original were
inatobtain
respectively.
to describe
different
Equation described articles the 5.net
Aby kinetics
The
(and thenot of thei
encountered
thermodynamic
are the Solution
the Hnumber
Considering
independent of the
Calculations
Nparameters
Considering
experimental
number of
charge.
lumped mixing 
inH k
2HIC
Equation
ionic
Solution
N
data,
chemistry
chromatographic
4 H
approach P O
cells
O
and
H
were
− k +
described
Equation
3 
the
Jis+ H pK
processes
and + 8,
strength,
chemistry
H first
The
HP
used, + of
we
medium.OH
8,
assumptions
a the O −ineach
we
2− is
obtain:
(ii)

performed
the amino
the
The
9.23
paragraphs
+obtain:
characteristic
following
meaning HThe +were
that
acid,
spatial
following
assuming
14.00
time
reactions
parameter
6.82
the
and
above
progressively
of
in
+thewere
configuration
reactions
that:
internal two(i)
wasrelaxed.
solution
λ
contributions
in ofsolution
the k
estimated,
mass
were
estimated the
pK aprotein
Indeed,
taken of
and
into
byare
were
wereand
the k then
the
does
taken
amino
ionic
account:
comparing notsummed
characteristic
into play
acid
strength
model
k +
account:
side totime
simulations
obtain
aencountered
role.
chainsof J the net
internal
are
with
mic parameters
flowratesdescribed
underH 2O
following
parameters
charge. Equilibrium
in
2non the H
binding
A
adsorption
4 +
paragraphs
B
k , ionic
OH
k ,theconditions
C3−k4
above
isotherm:
were 14.00
were
(van
estimated
Therefore, The
pKby acontributions
estimated,
Deemter key model
and
plot).
k
comparing
only the from
parameters
model
thermodynamic the species
simulations the
parametersadsorbed
number
with of on mixing
experimental
described the insurface
cells
data.
the or the charact
and
paragraphs above
mic
hich / This
arameters. extremely
kineticprovided
independent
were determined H high
P
parameters.a
2.2.2
Equilibrium
experimental
locatedSolution
3 inNOfirst
7
from
HP
Calculations
H N
the
4 
(several
+H
H
ofO curve
 H2−
data.
pores
O
4 2
Equilibrium
+the
Mechanistic
protein N
chemistryP
were H O
Nof
of
H

mol/L),
P
4 H
pulse
+
+O +
first
the3+H +H and
charge
H++ +
strength,
model
experiments
+ + The H
particles
OH −pK
performed2.15vs where
following
a
pH
(ii)
are
9.23
K
the
9.23to C
performed
12.38 k
assuming
averagedk
transfer
14.00
= and
spatial
K =
ka
reactions
λ
perform
pK =λ
k γ
at
t λC
kγ(C
that:
k, k are →

different
in
k
together
k kk the
simulations.
configuration
γ C (C
solution
k (i)
respectively,
k
0)


intheconcen-
C 0)were
pK
.
To
of
The
which the
taken
of mass
improve
protein
the into
activity
were transfer
the
amino does
account: acid
coefficient
determined i
,
description
t respectively,
not side
γ play
from
of a
chains
quantifies
protein which
(9)
role.
(8) are(9) experim
pulse
ns presented Regarding
Hydrodynamics in this the
article
2.2.2
−and
Calculations data
2 4 that
were
Mechanistic
kinetics  3 were
performed
4+ To extracted
model
−not
simulate withfrom
the thethethe literature
software
hydrodynamic i
experimental / [14, 15]
Ypso-Ionic.
kinetic
data k to perform
parameters.
a
acquired by the preliminary
GSK in the tests
frame k of the
of chains
Owere first performed vsassuming that:kIndeed, (i) the pKionic of amino acid side of are
4
g conditionsit isThis
experimentalknown
(van
H H that
data,
provided
2 P
Deemter
O
The
independent O 4H 
the
the
+ a+H
plot).
P
Equilibrium
lumped of Opresence
first
HP
+2 4
OH O
assumptions
Henry
the 
−2−H −
curve
H +of
P− H
O +salt
of
coefficient
−+were
OH
the
+++H
+ trations
isprogressively
charge
6.82
14.00 +
pK2.15
defined of14.00
the
pH
as the species
relaxed.
to perform
slope of in
the the liq-
simulations.the
adsorption a were
ofTo determined
strength
improve
isotherm at low from
encountered
the protein
description
concentration. pulse
the The H
non-ideality
k+approach
SO 4 4of used the 2ionic
HSO protein
Pin thisstrength,
+ Hsolution,
article (ii)
can1.92the
while
be spatial
the
extended
flowrates under configuration
coefficient
to=inany
non γλ
other ofand
quantifies
binding kind the protein
the does
adsorption
adsorption
conditions (van not
isotherm,
Deemter play
strength awith
like
plot). role.
the
where 2
C H
and P2NC3
O H are H the4N 2concentrations
O
H + + 4H H 2.15of 9.23
the species
model, only the peak positions were reported indescribe
the original kk the
articles kliquid
(and not solid
the phases,
full respectively.
chromatograms). A
The 3approach used in this article cana be extended to
werethe present Solution
Interactions
study,
independent a mixing
k chemistry
4of
7 +
with 
cell
the 34+
4the
model
ionic Thesolid
was following
strength,
+ phase
used to
(ii) reactions
the The Cthe
spatial in
interactions λ kany
solution kC
hydrodynamics
configuration kother
were
with kind
taken
the
of intheofinto
solid
the adsorption
account:
phase
column,
protein were
does isotherm,
while described
not theplay likea the
byrole.
the(8)
2
tsimulations
extractedexperimental
from
HP O
the2−
Solution
H data,
P N
literature3P
O H−
O 
3−
the
+ [14,
+
++
chemistryHP N
H15] H
assumptions
O −2−
to + +H
perform
12.38
The
H +were 9.23
progressively
preliminary
following
6.82 tests
reactions relaxed.
of inthesolution Indeed,
were the
taken ionicinto strength
account: encountered
esented in the
this
lumped
Therefore,N Hprovided
presented
Considering
This
article H Hin
were
4only
Langmuir 4
approach 2N
chromatographic
Langmuir HSO
P O
H O
the H
this
adsorption −
Equation
−a

performed
P isO
adsorption
Equilibrium +
H
4 44H
article
first HP
used,
thermodynamic 8,
medium.
SO H
2−
were
OH
curve
isotherm.
with
O we
2− +
P
meaning O
isotherm.
3
4+H −+
the H9.23
performed
obtain:
of the
The
++
The
software
H
that
parameters
The6.82 with
14.00
charge
parameter
+ 8.06
the vs
concentration the
pH λ
Regarding
Ypso-Ionic.
2.15 software
to
6
contributions
described
concentration
pK k was
in perform
the
in Ypso-Ionic.
estimated
the liquid
data
from
the simulations.
by
phase
that
the
paragraphs
liquid comparing
were
species
phase C k To
is improve
extracted
Cadsorbed
above is model
indeed
were
indeed simply
from
on the the
the
estimated,
simply description
simulations
replaced with
literature
surface and
replacedorby of
[14,
by 15] to per
Linear
ions were reported
H
following
Driving
This
H PinSO
where
O N
2
Force
provided
theHP
H adsorption
34 4
H
original
C +
H O 
HSO2−
Equilibrium
and
P O
4
approximation
a−
P
3charge
N O
−first
− 
articles
HC 4++
P
4
+
2isotherm:
curve
4
are
H
O
H H+(and
3−
+
H 2+ P
the
+
4was
OH

not
1.92
2.15
used
+Simulations
ofconcentrations
the
+ H
the to
charge
+ full
12.38
9.23
adescribe
pK
K
vsofpH
2.15 All the
chromatograms).
= λ the
γ
kinetics
tosimulations
perform
species
(C →
− 0)k inof internal
presented
simulations.
the liquid in
k mass
To
and transfer
thisimprove
article
solid phases, [6, description
were
7the 7].
performed of
respectively.
(9)
with
A the so
Thenot
experimental
experimental
γ C
located
the . 2
Cin The
k .HP
the
hydrodynamic
k γkk3 4 4
HO H 2−
net
4O data,
data.
k
2pores P
2 4
O /4of
H2−P 
the
+ O k
the
4kinetic
4 −3+
3−
HPassumptions
of
OH4+O the 2−
− +
H4 −2−
particles protein
4
++are
parameters. H +were
in
12.38 6.82
averaged
14.00 progressively
solution
model, ka only was
together
k k the krelaxed.
calculated
peak
in Indeed,
as
positions
k . TheIndeed,
Ccharacteristic a
activitythe
function
were ionic of
reported strength
coefficientpH from
in the
γk massencountered
the
quantifiesamino
original acid
articles (and n
dynamic key model
parameters
24 BioPharm
H HSO
P O
parameters
experimental
described

lumped−
H 
H 
P
42H
International 
HSO
HP
O SO O data,
24
approach in
OP
2− are
Othe
H+ H
April
HSO
+H
P 
the
the
4+
is
H
O +
2023
−+ number
assumptions
paragraphs
+ HP
− OH
used,++ O
+number
HH meaning
8.06
+
6.82
of
above
+ H mixing
+were
were
1.92
2.15 6.82
14.00
that
6 cells
the J
progressively
estimated, and
contributionsthe
and relaxed. from the time
the
species of
ionicinternal
adsorbedstrength
www.biopharminternational.com
on encountered
the surface or
Thesequence.
lumped
the tnon-idealityHapproach
SO 2
H4of +24
To
Henry

2−
the do
HSO  4 so,
protein − the
coefficient
+ 3−
H + 4 is +
H 1.92
solution, of
defined positive
as
Therefore,
while thethe and
slopeonly
coefficient negative
of the
the charges
adsorption
thermodynamic were computed
isotherm at
parameters low at each pH
concentration.
described based
in the on
paragraph
netic transfer
parameters.
2
, The 4
4 2N3 HP
respectively, 4HP
O 4
4used
which
+ O
4N 2−
pores
2 P+
H
N3 in O
4were 444H
this
Pof+
+
article
Odetermined
3− ++ +
inHsolution
12.38
9.23
can be
from extended
12.38 k = to
Ccalculated
protein γλ
kany
λtogether
pulse kC
quantifies
kother
kexperiments kind the adsorption
ofperformed
adsorption strength
isotherm,
atcoefficient
different with
like the (8)
iThe located
2− charge
net Nin H

HPthe

O4HSO
HH2
the3−
HSO
P P of
O 4 +the
SOH H2−
+protein
3+ 2−
H
4 12.38the
H +
8.06particles
+ are
9.23 averaged
was
6experimental as C . The
a infunction of activity
pH from the aminoγ quantifies
acid
PO
number 4and HP
pK O of + H
each + 6.82
amino acid,λand the two contributionsk were then summed to obtain
k the net
Considering
the
Hydrodynamics
Simulations
Langmuir
flowrates under non H−
chromatographic
All
HydrodynamicsO
Equation
SO
4 
44SO
adsorption
3binding 
and
4SO
4 8,
2−
medium.
H
2simulations
and we
P
a 4 obtain:
−−
kinetics
HSO
4+OH +
The
+
44 +
kinetics
4isotherm.
− conditions
H
H
presented+
To
The
−(van in
To not
parameter
simulate
8.06
2.15
this the
simulate
concentration
Deemter hydrodynamic
was
the
1.92article estimated
were
the
plot). in performed
experimental
the /data
by
liquid kinetic
with
data
phase parameters.
comparing
acquired
the
C model
by by
GSK
k issoftware
acquired
indeed simulations
in the
Ypso-Ionic.
GSK
simply with
frame
inreplaced
the ofbyof
frame
4 H
2− 2 +HSO
3− −++ H++ k
1.92
Peer-Review Research

experiments performed at different Figure 1. Impact of pH and ionic strength on the protein retention quantified
f low rates under non-binding condi- in terms of the lumped Henry coefficient. Experimental results are represented
with symbols and calculations with lines. The experimental data were extracted
tions (van Deemter plot).
from Heinitz et al. (14) for the ionic strength impact (left) and from Xia et al. (15)
Rega rding the data that were for the pH impact (right). Results were obtained with two different proteins in
extracted from the literature (14,15) isocratic conditions. For the ionic strength impact, experiments were performed
to perform preliminary tests of the with chymotrypsinogen. A 20 mM phosphate buffer was used with varying
model, only the peak positions were concentrations of sodium chloride (NaCl) and sodium sulfate (Na2SO4). For the
reported in the origina l articles pH impact, experiments were performed with a high isoelectric point (pI) α–
amyloglucosidase using a 1 M Na2SO4 solution. Two calculation methods were
(and not the full chromatograms). used: statistical (top) and mechanistic (bottom).
Therefore, only the thermodynamic
parameters described in the earlier
paragraphs were estimated, and not
the hydrodynamic/kinetic parameters.
Simulations. All simulations pre-
sented in this article were performed
with the specific software (Ypso-Ionic).

RESULTS AND DISCUSSION


Preliminary results
based on literature data
Prel imina r y invest igat ions were
performed based on experimental
literature data obtained with chymo-
trypsinogen (14) and α–amyloglucosi-
dase (15). To do so, the protein Henry
coefficient measured at different val-
ues of the ionic strength (14) and solu-
tion pH (15) were considered. Two
approaches were used: a statistical found to increase with the salt concen- chromatograms (instead of retention
approach based on a quadratic expres- tration. This phenomenon is referred times only as in this section).
sion and a mechanistic approach to as a “salting-out” effect. In addition,
based on physical phenomena. The a maximum of retention is observed as Reference conditions
expressions used for the statistical a function of pH, the retention being The UV and conductivity prof iles
approach are given in Equations 2 maximal around the isoelectric point of obtained for the reference HIC exper-
and 3. The expressions used for the the protein (around eight for the con- iment are shown in Figure 2a. The
mechanistic approach are given in sidered protein). Considering panels (a) steps are also indicated: first, crude
Equations 7 and 9. and (b) of Figure 1, it is observed that loading at a high ionic strength; sec-
T he resu lts a re presented in the statistical model based on a quad- ond, wash at the same ionic strength;
Figure 1, where the lumped Henry ratic expression is not able to represent third elution at an intermediate ionic
coefficient is plotted as a function of the experimental trends. Indeed, some strength; fourth, regeneration at a
the ionic strength (left) and solution negative values of the lumped Henry low ionic strength; and fifth, strip-
ALL FIGURES ARE COURTESY OF THE AUTHORS.

pH (right). The dots represent the coefficient are calculated, which does ping with water. Three peaks are
experimental data and the lines the not make sense physically. observed in the UV profile, the first
calculations. In examining the exper- Considering panels (c) and (d) of t wo during the elution phase and
imental results, it is observed that the Figure 1, it is observed that the sim- the last one during the regeneration
lumped Henry coefficient goes through ple mechanistic model proposed in this phase. The perturbations of the UV
a minimum as a function of the ionic article represents well the experimental and conductivity signals at the begin-
strength. At low ionic strength, the trends of the lumped Henry coefficient ning of the experiment were caused
retention is found to decrease with the with the ionic strength and solution pH. by a temporary malfunction of the
salt concentration. This phenomenon In the rest of the article, the mechanis- system (probably due to the pres-
is referred to as a “salting-in” effect. At tic model is applied to novel experimen- ence of an air bubble) and were not
high ionic strength, the retention is tal data. It is challenged against the full observed in subsequent runs.

www.biopharminternational.com April 2023 BioPharm International 25


Peer-Review Research

Figure 2. Results of the reference experiment (a) experimental ultraviolet (UV) charge variants was guessed based
and conductivity profiles together with the percentages of low molecular on the SEC analysis of the chro-
weight (LMW), monomer (Mono), high molecular weight (HMW) species
matogram of the reference exper-
determined by size exclusion chromatography (b) simulation results showing
the species taken into account. iment. T he fol low ing resu lts
were obta ined: L M W1 (0.66%),
LM W2 (2.56%), LM W3 (4.88%),
Mono1 (10.61%), Mono2 (43.90%),
Mono2bis (14.63%), Mono3 (8.46%),
HMW1 (0.26%), HMW2 (3.11%),
and HMW3 (10.93%).
The simulation of the reference
experiment (Run 1) with all the
considered spec ies is show n in
Figure 2b. The procedure used to
deter m ine model pa ra meters is
described in the materials and meth-
ods section. Briefly, the main param-
eters to be determined for each species
k are the Truesdell-Jones coefficients
Ak, Bk, Ck and the coefficient λk of the
adsorption isotherm. To reduce the
number of fitting parameters, it was
assumed that all species eluting in a
given peak have the same Truesdell-
Jones co-eff icients (e.g., LM W1,
Mono1, and HMW1 have the same
coefficients). In addition, it was con-
sidered that λ k = 1 for all species,
except Mono2bis, for which the λk was
estimated to be 1.2. The results of the
fitting for the five runs are presented
in terms of lumped Henry coefficient
For each of the three peaks, frac- species based on different physical as a function of the ionic strength in
tions were collected and analyzed properties, namely charge/hydropho- Figure 3a. The results of the charge
by SEC to qu a nt i f y t he Mono, bicity and size. As a consequence, as a function of pH are presented in
L M W, a nd H M W species. T he representing the crude as a mixture Figure 3b. As described in the mate-
analyzed results are presented in of Mono, LMW, and HMW is not rials and methods section, the pKa val-
Figure 2a. Interestingly, Peak 1 and suff icient. It is necessary to intro- ues and number of accessible aspartic/
Peak 2 have similar compositions, duce additional species, which are glutamic acids and histidine residues
with around 90% of Mono. Peak 3 likely to be protein charge variants. were adjusted compared to their theo-
contains around 65% of impurities Accordingly, the authors considered retical values to allow a better descrip-
(LMW+HMW) and 35% of Mono. LM W1, LM W2, and LM W3 for tion of experimental results.
These results show that the three the LMW species eluting in Peak The total protein concentration
species identif ied by SEC (LMW, 1, Peak 2, and Peak 3, respectively. (experimental and simulated) for
Mono, HMW) co-elute in the three Similarly, they considered HMW1, the reference experiment is shown
peaks observed in HIC. It is thus not HMW2, and HMW3 for the HMW in Figure 4 (Run 1). It is seen that
possible to match the peaks observed species eluting in Peak 1, Peak 2, the model allows correctly predict-
in HIC with those observed in SEC and Peak 3, respectively. For the ing the overall shape of the chroma-
(e.g., associate Peak 1 with LMW, Mono, two charge variants in Peak togram, with three peaks, the last
Peak 2 with Mono, and Peak 3 with 2 were considered in an attempt to one being much narrower than the
HMW). This can be explained con- describe peak broadness. The com- first two. The retention times of the
sidering that HIC and SEC separate position of the crude in terms of peaks are accurately described, while

26 BioPharm International  April 2023  www.biopharminternational.com


Peer-Review Research

Figure 3. (a) Experimental and simulated values of the lumped Henry coefficient as a function of the ionic strength at
pH 6.0, 7.0, and 8.0 for the species eluting in peak 2. The lumped Henry coefficient is equal to the activity coefficient
(Truesdell-Jones equation) because λ is equal to 1.0. (b) Calculated protein charge as a function of pH. The dots correspond
to the protein charge fitted to describe the experimental values of the lumped Henry coefficients.

Figure 4. Impact of the ionic strength (modulated by the % B) and pH: (a) experimental chromatograms at different ionic
strength values; (b) simulated chromatograms at different ionic strength values; (c) experimental chromatograms at different
pH values; (d) simulated chromatograms at different pH values.

their broadness could be improved. erature (16); it is regarded as out of dicted thanks to the Truesdell-Jones
The separation between Peak 1 and the scope of the present study. equation (Figure 3a). This behavior
Peak 2 is indeed underestimated. A is referred to as a “salting out” effect
reduction in the characteristic time Impact of key (in contrast to the “salting in” effect
for mass transfer of Mono2 was con- operating parameters observed at low ionic strength).
sidered to improve peak resolution. Ionic s trength. The left-hand Solution pH. The right-hand side
However, the broadness of Peak 2 side of Figure 4 shows the impact of Figure 4 shows the impact of pH
was then signif icantly underesti- of the ionic strength on the chro- on the chromatograms: experimental
mated. A plausible explanation is matograms: experimental (top) and (top) and simulated (bottom). In the
that more than two charge variants simulated (bottom). It is observed investigated pH range, the higher the
need to be considered to describe the that the higher the %B (and thus the pH, the more retained are the species.
observed peak broadness. The impact lower the ionic strength), the more This is related to the variation of the
of multicomponent mixtures on peak retained are the species. This is due to charge of the protein with solution
broadness has been studied experi- an increase in the activity coefficient pH (Figure 3b). Although the curve
mentally and theoretically in the lit- at high ionic strength, which is pre- seems flat in the pH range of interest,

www.biopharminternational.com April 2023 BioPharm International 27


Peer-Review Research

Figure 5. Prediction of the impact of the ionic strength (%B) and solution pH. Conditions used to fit the model are
represented in black, while conditions fully predicted by the model are in red.

a slight decrease of the protein charge fit the model are represented in black, mediate situation obtained with pH
with pH is actually obtained: z = 5.11 while conditions fully predicted by the 6.0 and 28 %B without reducing the
at pH 6.0, z = 4.94 at pH 7.0, and model are in red. elution time. This is considered for
z = 4.32 at pH 8.0. This is sufficient The choice of “optimal” process comparison purposes but has no real
to describe the increase in retention conditions must be done consider- practical interest.
observed in HIC experiments. ing given targets and constraints. Process performances (recover y,
Overall, the peak positions and I n t he c a s e of u nder i nve st ig a- p r o du c t i v it y, s ol v e nt c on s u mp -
shapes of the five performed experi- tion, one can consider two options: tion) are quantif ied in Figure 5
ments are well represented by model the collection of Peak 2 only, lead- for the three options. When mov-
simulations. The proposed model is ing to a product of Grade 1 purity, ing from option (a) to option (c),
thus capable of describing the impact or the collection of Peak 1 and 2 t he product iv it y is inc reased
of t wo key operating parameters, simultaneously, leading to a product by a fac tor of t wo a nd t he sol-
namely ionic strength and solution of Grade 2 purity. vent consu mpt ion reduced by a
pH. The model was also shown to be To obtain Grade 1 purity, pH 7.0 factor of two.
capable of describing the impact of and 18 %B seem the most appro- This study illustrates how simu-
loading volume and f low rate (data priate conditions to ensure a good lation can help to perform process
not shown). A more ref ined ver- separation bet ween the f irst t wo optimization, increasing productivity
sion of the model would consist in peaks. This is highlighted as option and reducing solvent consumption.
improving the description of peak (a) in Figure 5. To obtain Grade 2
broadness, potentially by including purity, pH 6.0 and 28 %B seem the CONCLUSION
more charge variants. most appropriate conditions to allow A simple mechanistic model of HIC
a gain in productivity and decrease has been developed based on the
Illustration of in solvent consumption. Indeed, curve of the protein apparent charge
process optimization the elution time can be sig nif i- as a function of pH and the use of an
Once model parameters are deter- cantly reduced while still obtaining a activity coefficient in solution.
mined, the model can be used to pre- good separation between the f irst It was show n that the model
dict unexplored operating conditions. two peaks and the third peak. This is allows predicting the impact of ionic
This predictive capability is illustrated highlighted as option (c) in Figure 5. strength (%B) and solution pH on the
in Figure 5, where conditions used to Option (b) corresponds to the inter- position and shape of chromatograms.

28 BioPharm International  April 2023  www.biopharminternational.com


Peer-Review Research

The model was then used to illus- 86 (12), 1277–1283. DOI: 10.1002/ Selectivity. J. of Chromatogr. A 1988,
trate how numerical process optimi- prot.25608 443, 173–182.
13. Truesdell, A. H.; Jones, B. F. Wateq, a 15. Xia, F.; Nagrath, D.; Cramer, S. M.
zation can be performed with a view Computer Program for Calculating Effect of pH Changes on Water
to improving process performances Chemical Equilibria of Natural Waters. Release Values in Hydrophobic
such as recovery, productivity, and J. Res. US Geol. Surv 1974, 2 (2), Interaction Chromatographic Systems.
233–248. J. of Chromatogr. A 2005, 1079 (1–2),
eluent consumption. 14. Heinitz, M. L.; Kennedy, L.; 229–235.
Kopaciewicz, W.; Regnier, F. E. 16. Pfister, D.; Morbidelli, M.; Nicoud,
REFERENCES Chromatography of Proteins on R. M. A Continuum Theory for
1. Carta, G.; Jungbauer, A. Hydrophobic Interaction and Ion- Multicomponent Chromatography
Protein Chromatography: Process Exchange Chromatographic Matrices: Modeling. J. of Chromatogr. A 2016,
Development and Scale-Up, Mobile Phase Contributions to 1446, 50–58. ◆
2nd ed.; John Wiley & Sons,
Hoboken , NJ, 2020.
2. Curling, J. The Development of
Antibody Purification Technologies.
In Process Scale Purification of
Antibodies, Gottschalk, U., Ed.; John
Wiley & Sons, Hoboken , NJ, 2009; pp.
23–54. DOI:10.1002/9780470444894
3. Shekhawat, L. K. ; Rathore, A. S.
Mechanistic Modeling-Based Process
Analytical Technology Implementation
for Pooling in Hydrophobic Interaction
Chromatography. Biotechnol. Prog.
2019, 35 (2), e2758.
4. Kuehner, D. E.; Blanch, H. W.;
Prausnitz, J. M. Salt-Induced Protein
Precipitation: Phase Equilibria from an
Equation of State. Fluid Phase Equilib.
1996, 116 (1–2), 140–147.
5. Mollerup, J. M. Applied
Thermodynamics: A New Frontier for
Biotechnology. Fluid Phase Equilib.
2006, 241 (1–2), 205–215.
6. Nicoud, R.M. Chromatographic
Processes: Modeling, Simulation and
Design; Cambridge University Press,
Cambridge, UK, 2015.
7. Pfister D.; Nicoud, L.; Morbidelli,
M. Continuous Biopharmaceutical
Processes: Chromatography,
Bioconjugation, and Protein Stability;
Cambridge University Press,
Cambridge, UK, 2018.
8. Abe, Y.; Ueda, T.; Iwashita, H.; et al.
Effect of Salt Concentration on the
pKa of Acidic Residues in Lysozyme. J
Biochem. 1995, 118 (5) 946–952. DOI:
10.1093/jb/118.5.946
9. Lee, K. K.; Fitch, C. A.; Lecomte, J. T.;
García-Moreno E, B. Electrostatic
Effects in Highly Charged Proteins:
Salt Sensitivity of pKa Values of
Histidines in Staphylococcal Nuclease.
Biochemistry 2002, 41 (17), 5656–5667.
10. Kao, Y. H.; Fitch, C. A.; Bhattacharya,
S.; et al., Salt Effects on Ionization
Equilibria of Histidines in
Myoglobin. Biophys.J.
2000, 79 (3), 1637–1654.
11. Li, A. H.; Robertson, D.; Jensen, J.H.
Very Fast Empirical Prediction and
Rationalization of Protein pKa Values.
Proteins. 2005, 61 (4), 704–721. DOI:
10.1002/prot.20660
12. Pahari, S.; Sun, L.; Basu, S.; Alexov,
E. Delphipka: Including Salt in the
Calculations and Enabling Polar
Residues to Titrate. Proteins. 2018,

www.biopharminternational.com April 2023 BioPharm International 29

You might also like