Validation of Large-Scale
Chromatographic Processes, Part 1
FDA requires
chromatography
procedures to be
validated for reuse,
cleanability, sanitization,
robustness, and
(potentially) for virus
removal. This case study
follows the development
of a validation process
that combines reuse and
characterization studies,
defines acceptable
running parameters,
reduces costs, and meets
GMP requirements.
MILLIPORE CORPORATION (WWW.MILLIPORE.COM)
16 BioPharm MAY 2002
Case Study of Neuleze Capture on Macroprep High-S
Timothy N. Breece, Ellen Gilkerson, and Charles Schmelzer
fter a chromatographic process is
A
developed for purifying a new drug,
FDA requires that the procedure’s
performance be validated (1). Validat-
ing that performance involves demon-
strating that the chromatographic media
can be reused, cleaned, and sanitized
and that the process, as defined, is robust. FDA
prefers prospective process validation when
appropriate.
VALIDATION TESTS REQUIRED
Although cleaning of full-scale equipment must
be documented, performing all studies at process
scale is impractical. Production chromatography
columns are therefore validated at a minimum of
two different scales. Scale equivalence is
achieved by maintaining constant column bed
height and linear flow rate while varying column
diameter. Buffer and load amounts are scaled to
the column volume.
At large-scale, two validation studies are
required. Pools from several runs are tested for
impurity clearance and yield. Then the cleanabil-
ity of the resin and that of associated large-scale
equipment is verified by analyzing the pool from
a mock run for the presence of both product and
selected impurities.
At small-scale, three studies are required. A
reuse study is needed to verify continued resin
performance after multiple use, followed by a
small-scale carry-over run to confirm cleanability
with multiple use. A characterization study
proves the process is robust and defines the
acceptable limits of operation. Another study
must demonstrate that regeneration is sufficiently
sanitizing and that the storage solution is bacte-
riostatic and fungistatic (1).
A process used to purify products derived from
mammalian cell cultures must show clearance of
both potential adventitious and endogenous viruses
(2). Because viral testing is expensive, once a pu-
rification step is validated for viral clearance, no
changes to the process should be made that would
require new testing. Because FDA does not con-
sider it necessary to evaluate every step of a
process for viral clearance (2), maximal process
flexibility is maintained by minimizing the number
of steps required to adequately validate viral clear-
ance. The choice of small-scale column size de-
pends on knowing whether virus removal will be
validated. If so, then enough end-of-use resin
should be generated to pack several smaller-scale
columns for viral validation studies.
REDUCING IMPURITIES
This case study details the validation of a cation-
exchange column, Macroprep High-S, used as the
capture step in the purification of Neuleze, a nerve
growth factor (NGF). The product is produced in
mammalian (Chinese hamster ovary, CHO) cell
cultures, where it is secreted by the cells and har-
vested by tangential-flow microfiltration before
loading onto the column. The column run protocol
comprises six steps: equilibration, load, wash with
equilibration, elution, regeneration, and storage.
The resin binds the product while most of the
CHO protein (CHOP), host-cell DNA, and other
cell culture components flow through.
The capture step is followed by four more
chromatography steps and formulation by ultra-
filtration. Column performance is evaluated by
percent yield (% yield) and by reductions of
DNA, CHOP, and a proprietary media compo-
nent (PMC) used as a small molecule representa-
tive. Because this is the first of several purifica-
tion steps, we used impurity reduction as a metric Table 1. Factors in the experimental design
rather than specific impurity levels.
Reuse study. Several considerations must be Center Factorial Points Axial Points
made before embarking on the small-scale reuse Factor Point Low High Low High
study. Although it is unnecessary to determine Flow rate (cm/h)a 375 150 600 100 750
the actual number of reuses for a resin, a suffi- Load volume L (HCCF/L) of resin 62 24.6 98.5 16.3 124
cient resin lifetime should be determined. An Equilibration pHa 7.0 6.5 7.5 6.0 8.0
Equilibration HEPES (mM) 50 42 58 33 100
appropriate number of reuses can be chosen by Equilibration sodium acetate (M) 1.5 1.25 1.75 1.0 3.0
balancing the resin cost with the time required to Elution pHa 7.0 6.5 7.5 6.0 8.0
do the reuse study and with the anticipated annual Elution HEPES (mM) 50 40 60 34 67
number of production runs. We expected a maxi- Elution sodium chloride (M) 1.5 1.2 1.8 1.0 2.0
Elution TMAC (M) 0.25 0.20 0.30 0.17 0.33
mum of 12 runs per year. Because we hold in Elution thiodiglycol percent 2 1.6 2.4 1.3 2.67
inventory enough resin to replace the production- Pooling OD Starta 0.4 0.2 0.6 0.1 0.8
scale column and because material control con- Pooling OD End 0.2 0.1 0.3 0.05 0.4
cerns led to a three-year replacement schedule, a For convenience the combined factor is referred to by this name
we chose a reuse rate of approximately 36 runs.
Given the cost of the resin, reuse levels were jus-
A major savings of cost and time can be gained
tified economically as well.
by combining the reuse and characterization studies
Characterization study. To decide what factors to
into a single protocol. Maximum information is ob-
test in the characterization study, we made a list of
tained by combining the studies and varying the
possibilities. That list consisted of equilibration
factors using a fractional, factorial design. By inter-
and elution salt/buffer molarities; equilibration and
spersing control runs amid characterization runs, it
elution pH; flow rate; load volume, product titer,
is possible to verify that column performance re-
variation (in a fermentation lot), pH, and conduc-
mains undegraded by reuse and characterization
tivity; optical density of pool collections; wash
conditions. To guarantee the validity of the reuse
times; and operation temperature. The total num-
results, it is important to balance sources of poten-
ber of potential factors was too great to be tested
tial column degradation when designing the com-
adequately, so we reduced the number studied
bined study so that total exposures are the same as
based on experience gained during process devel-
would have been seen in a separate reuse study.
opment. In addition, we reduced the number of
The two main sources of potential column
factors to a reasonable level by combining factors
degradation in this study are the load or the
and treating them as single variables (3).
regeneration. Load exposure is balanced by
The variables were linked so as to ensure that
ensuring that the total load for the combined
their effects would be additive. Hence, molarity
study runs (though not the same every cycle) is
and pH of equilibration and elution were combined
greater than the total load that would have been
because, at higher levels, both factors inhibit bind-
used in a reuse study made up entirely of control
ing to cation-exchange resins. Load volume and
runs. Regeneration buffer exposure was con-
flow rates were combined because at higher levels
trolled by keeping every run the same and by val-
both would tend to reduce yield. And the increased
idating sanitization separately.
residence time and absence of displacement effects
observed at higher loading would suggest that at METHODS AND MATERIALS
lower levels both load volume and flow rate would Instruments and hardware for production and validation.
increase the relative amount of impurities found. We used a production-scale, 120 cm  20 cm
Combining load volume and flow rate effectively (depth  height), stainless steel column from East-
keeps loading times constant. ern Rivers, Inc. (Chattanooga, TN). It was run on a
When combining variables, it is important to proprietary, custom-designed, automated produc-
be aware of what you have confounded in doing tion skid consisting of a controller, pump, auto-
so. We treated the rest of the load variables as a mated valves, and inline ultraviolet (UV), pH, flow,
single uncontrolled variable of the fermentation and conductivity monitors. The reuse–characteriza-
lot. Wash times were considered to be optimized
tion study used a 1.6 cm  20 cm column packed
and controllable, so we omitted them. Likewise,
in an Amicon Vantage glass column (Millipore,
temperature was omitted because during develop-
www.millipore.com). The study was performed on
ment no temperature effects were observed and
a Pharmacia (www.pharmacia.com) BioPilot liquid
because production operations are conducted in a
chromatography system. The viral clearance study
controlled environment.
columns were 0.5 cm  20 cm packed in Pharma-
BioPharm MAY 2002 17

Log removal virus = Log 10 Virus titer in load × load volume
Equation 1. Calculation for log
removal of virus
Resin Test Organisms
To test resin sanitization and
storage, resin samples were
independently challenged
with five test organisms listed
by USP for preservative
effectiveness and with an
environmental isolate:
Aspergillus niger
(ATCC 16404)
Burkholderia picketti
(Genentech isolate)
Candida albicans
(ATCC 10231)
Escherichia coli
(ATCC 8739)
Pseudomonas aeruginosa
(ATCC 9027)
Staphylococcus aureus
(ATCC 6538).
18 BioPharm
of the spiked load and resulting pools were sent to
BioReliance (www.bioreliance.com) to be tested
by a viral infectivity assay. The “Equation 1” box
Virus titer in pool × pool volume
shows the calculation used for the log removal of
the viruses.
cia HR glass columns. The study was performed on
Column cleaning. Production-scale carry-over
a Pharmacia ÄKTA 100 chromatography system.
lots were run on the columns currently in use at
Reagents and resins. The Macroprep High-S
the manufacturing plant. The carry-over runs
cation-exchange resin came from Bio-Rad
were performed according to appropriate manu-
Laboratories (www.bio-rad.com). The buffer
facturing tickets (with the exception of the load).
salts and reagents were from the Genentech
Mock loads of phosphate-buffered saline (PBS)
(www.gene.com) GMP released material facility
were loaded on each column in the Neuleze-
in South San Francisco, CA.
recovery process. Mock pools were collected at
Column running protocols. The flow rate was
the point in which Neuleze pools would normally
375 cm/h (70.6 L/min) for all steps except the
have been collected based on historical elution
regeneration step, which ran at 120 cm/h. The
times and volumes.
column was equilibrated with three CVs of
At the end of the small-scale reuse–characteri-
50-mM HEPES, 1.5-M sodium acetate, pH 7,
zation study, a mock pool was collected from a
then loaded with 12,500–14,000 L of harvested
carry-over run. Impurity levels were evaluated as
cell culture fluid (HCCF), or 55–62 L HCCF/L of
nanograms of impurity per milligram of product
resin. The column was washed with five CVs of
using the lowest concentration of Neuleze in a
equilibration buffer and eluted with five CVs of
Macroprep High-S production run pool.
50-mM HEPES, 1.5 M sodium chloride, 0.25 M
Resin sanitization and storage. Production run
tetramethylammonium chloride (TMAC), and
resin was repacked in a 1.1 cm  20 cm Amicon
2% thiodiglycol, pH 7. The optical density (OD)
Vantage column. The column was cycled as
of the element was monitored, and the pool was
described and scaled to the column. After elution,
collected from 0.4–0.2 outer diameter (o.d.). The
the column was unpacked into storage solution
column was regenerated with a minimum of three
without being regenerated or washed. In prepara-
CVs of 0.5 normal sodium hydroxide (N NaOH)
tion for challenging the sanitizing and storage
for 30 minutes and washed with three CVs of
solutions, inoculum suspensions were prepared at
0.1 N NaOH storage solution.
about 108–109 colony forming units (CFU)/mL
Reuse–characterization. For the reuse–charac-
using the turbidimetric method for titer estima-
terization study, a response surface design was set
tion of bacteria and yeast and the plate-count
up using JMP version 3.0 statistical software
method for titer determination of mold. Resin
(SAS Institute Inc., www.sas.com). The number
samples were mixed with sanitizing and storage
of factors to vary was reduced to four by combin-
solutions to achieve final concentrations of 0.5 N
ing (confounding) and varying them as shown in
NaOH and 0.1 N NaOH. Time-point samples
Table 1. A central composite orthogonal design
were diluted to neutralize the sanitizing agent,
was chosen using the four factors from Table 1
and recovery of challenge organisms was
and the fermentation lot as a blocking factor. The
determined.
high and low load levels were balanced so that the
Controls were prepared the same way as the test
total HCCF loaded was greater than if all the runs
samples (substituting PBS) and used to establish
had been performed under normal conditions. A
the baseline against which the effect of the sanitiz-
total of 42 reuse runs were performed under the
ing and storage agents were determined. Resin
conditions shown in the “Reuse Runs” sidebar.
samples were independently challenged with five
Virus removal validation. After the 42 reuse
organisms listed by USP for testing preservative
cycles were completed, the Macroprep High-S
effectiveness and with an environmental isolate
(end-of-use) resin was removed from the small-
(see the “Resin Test Organisms” box).
scale (0.5  20 cm) column. In addition, new
Assays. Pool samples from both the manufac-
resin was packed into an equal number of
turing-scale and small-scale studies were stored
columns to serve as benchmarks. A known quan-
at lower than 20 °C until analysis. The samples
tity of three viruses — simian virus 40 (SV40),
were thawed, and sodium azide was added as a
murine leukemia virus (MuLV), and minute
preservative to a final concentration of 0.05% to
murine virus (MVM) — was added to the HCCF.
all samples except those intended for DNA analy-
The spiked HCCF was loaded into columns and
sis. All analyses were performed by the Process
run according to the protocol described. Samples
MAY 2002

ng containment in load
Log removal = Log 10
mg product in load
Equation 2. Calculation of log
removal for assay analysis
Corresponding author Timothy N.
Breece is a senior research associate
in recovery sciences, Ellen
Gilkerson, is a senior biostatistician
in the biostatistics department, and
Charles Schmelzer is a senior
scientist in recovery sciences at
Genentech, Inc., One DNA Way, South
San Francisco, CA 94080,
650.225.3292, fax 650.225.5275,
tnb@gene.com, www.gene.com
20 BioPharm MAY 2002
ng containment in pool
– Log 10
mg product in pool
Sciences Analytical Operations group at Genen-
tech. The samples were analyzed for rhuNGF
(concentration was determined using affinity
chromatography HPLC), CHOP (using a sand-
wich ELISA), DNA (using a DNA threshold
assay), and PMC (using a competition ELISA).
All log removal results were calculated using the
formula in the “Equation 2” box
Statistical methods. JMP statistical software was
used to analyze the data. Analyses were
performed separately for the response variables
of interest: % yield and CHOP, DNA, and PMC
log clearances. To investigate possible changes
over time in % yield and in CHOP log clearance,
a straight line was fit by least squares regression
for the response at the center points versus the
cycle number. To provide 95% confidence, three
standard deviations were chosen as the criterion
for significance.
To investigate the effect of the factors on the
response, a second-degree model with the main
effects (factors), two-way interactions, and the
quadratic terms for the main effects was fit to the
Info #11
data for CHOP and PMC log clearance and for
% yield. The model removed unimportant factors
by first performing stepwise regression with
factors with P0.05, followed by a screening fit
of the data. The important individual parameter
estimates were displayed in Pareto plots to ascer-
tain significant variables. Because assays were
performed on only 11 runs for DNA, the model
for this impurity included main effects only.
RESULTS TO COME . . .
As is often the case with validation studies and
other bioanalytical testing, we must wait for the
results and their interpretation. Part Two of this
article will conclude in our July issue by present-
ing them along with some discussion of GMP re-
quirements and the approach to compliance taken
in this case study.
REFERENCES
(1) Center for Biologics Evaluation and Research,
Guideline on General Principles of Process
Validation (FDA, Rockville, MD, May 1987).
(2) International Conference on Harmonisation, Viral
Safety of Biotechnology Products Derived from Cell
Lines of Human or Animal Origin (FDA, Rockville,
MD, 24 September 1998).
(3) L. Shi et al., “Experimental Design for Validation of
Chromatographic Process Limits,” presented at the
BioPharm Conference, May 1997. BP