Bioseparations
E. N. Lightfoot, J. S. Moscariello
Department of Chemical and Biological Engineering, University of Wisconsin,
Madison, Wisconsin; telephone: 608-262-6934; fax: 608-262-5434;
e-mail: lightfoot@ engr.wisc.edu
Received 2 May 2003; accepted 16 March 2004
Published online 7 July 2004 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/bit.20111
Abstract: Here we review key applications of separation
technology in applied biology. We first sketch out the field
as a whole, but then narrow our scope to the processing
of fermentation products, particularly to high-value biolog-
icals such as proteins and nucleotides. We go on to provide
a qualitative overview describing the importance and gen-
eral nature of this large field, major trends, and the strat-
egies that have proven most fruitful in evolving effective
separation and purification processes. We then give a de-
tailed description of individual separations equipment and
the principles governing their operation. We concentrate
throughout on making the available literature accessible to
the reader; we provide what is hoped to be a representative
set of basic references. However, these references, in turn,
include some that suggest promising new developments
as well as a number of more specialized reviews. We hope
that our overall result provides the reader with access to
the most relevant literature. B 2004 Wiley Periodicals, Inc.
Keywords: bioseparations; separations; chromatogra-
phy; membrane filtration; microfiltration; ultrafiltration;
crystallization
INTRODUCTION
Orientation
There are those of us who have seen chemical engineering
change from a rather compact field that could be defined
in terms of a few core disciplines built primarily upon
petroleum processing and petrochemistry to a large and
currently unwieldy conglomeration of which biotechnology
seems an ever-increasing part. However, there is still some
utility in considering that much of what we do is built upon
a combination of chemical transformations and separations.
The discussion below is built upon this simplified picture,
and our focus will be on separations processes operating
either on the output of bioreactors or a mixture derived
from biological sources. Such bioseparations can be
divided into the extraction or purification of natural prod-
ucts and what may loosely be called fermentation tech-
nology. This discussion will be limited for the most part to
the second to produce some coherence into what must be a
relatively short discussion.
Correspondence to: E. N. Lightfoot
B 2004 Wiley Periodicals, Inc.
Fermentation technology (Ahuja, 2000; Atkinson and
Mavituna, 1991; Flickinger and Drew, 1999; Shuler, 2001)
is commonly divided into ‘‘upstream’’ and ‘‘downstream’’
processing, which is primarily a way of saying bioreac-
tion and bioseparations. However, these two interact very
strongly: the choice of fermentation substrates has a major
effect upon separations strategy and vice versa.
One important example, the separation of starch from
corn by wet milling, has resulted in the industrial scale
availability of inexpensive and very pure glucose. This, in
turn, has had an enormous impact on a major aspect of
bioprocessing—production of commodity chemicals from
biomass. It may even have begun to reverse the steady
encroachment of petrochemical production on biochemical
production and made the availability of inexpensive corn a
major factor in plant siting. Fermentation-based ethanol and
lactic acid from glucose are now economically interesting
products, and other major sources of pure sugars may have
potential as well. Here the very large-scale processing of
milk in New Zealand leading to the availability of plentiful
supplies of crystalline lactose comes to mind.
However, wet milling and recovery of simple mono-
mers from fermentation broth do not have very significant
biological overtones, and they will not be discussed here.
More subtle changes in fermentation substrates do have
great impact on the downstream processing of the primary
subjects of this discussion—high-priced biologicals—but
they will also have to be largely ignored for lack of space
and author energy. Vitamins and antibiotics are important
biologicals often produced by fermentation, but for the most
part the processes used are mature and already well covered
in encyclopedias of industrial chemistry (e.g., Kroschwitz,
1991). We shall also have to neglect the very large area of
food processing, but we note in passing that our discussion
of membrane filtrations finds many parallels there (Ho and
Sirkar, 1992; Noble and Stern, 1995).
Bioseparations are critical to the success of modern bio-
technology and represent a major manufacturing cost for a
wide variety of products. In fact, the variety is so wide that
we must first select a relatively compact subset of products.
Next, we briefly review process strategy to reduce the
combinatorial problem faced by individual engineers to a
manageable level: the potential parameter space typically
encountered in biotechnology dwarfs that found in most et al., 2003). Computerized design techniques (Jungbauer
nonbiological applications. Moreover, the pressures for and Kaltenbrunner, 1999) continue to be developed by a
rapid process development are often extremely severe. We number of research groups (Harrison et al., 2003, Ch. 11;
then proceed to an overview of equipment design and Mustafa et al., 2003; Nagrath et al., 2003; Pieracci et al.,
the underlying thermodynamic and transport principles gov- 2003). Several are certainly effective for stoichiometric
erning equipment performance. purposes and hence for making direct comparisons of
Our primary concern will be biologically active proteins, alternative processes. However, we are doubtful that any
polysaccharides, nucleotides, viruses and vaccines, and vi- contain enough information to elaborate a complete sep-
able whole cells. However, we shall also briefly discuss the aration scheme: all have yet to agree on optimizing scale-up
separations problems encountered in processing body of a simple chromatographic step. For the most part, these
fluids, principally blood purification and recovery of val- techniques have yet to be tested realistically.
uable products from blood. Many of these operations are
performed using techniques familiar from older literature,
and we will therefore refer to reviews of this readily avail- OVERVIEW
able material where possible. Moreover, it must be rec-
Bioseparations are typically difficult. All products of interest
ognized that the wide range of topics covered limits the
are labile and thus require mild processing conditions. They
bibliography to a representative as opposed to an exhaus-
typically arrive at downstream processing contaminated
tive set of references.
with closely related species, and the required product purity
The whole field of bioseparations is presently changing
is usually very high. Safety is a major consideration. Perhaps
very rapidly, and we will emphasize these changes. Es-
the most severe problem is finding separating agents with
tablished techniques are becoming much more effective
both high capacity and selectivity for the product of in-
through advances in technology (Abraham et al., 2003;
terest; almost always, effective separations depend upon
Caren, 2003; Chae et al., 2002; Christy et al., 2002; Curling
anomalies of the system being studied.
and Smiley, 2003; Heng, 2003a, 2003b; Hjorth, 2003;
These demands typically dictate a large number of
Khosla et al., 1996; McCue et al., 2003; Palmgren et al.,
processing steps in series. Moreover, in most situations
2003; Phillips, 2003; Shpritzer, 2003; Tebbe et al., 1996; van
the feed to the separation train is either dilute or the process
Reis et al., 1997c), and new techniques are increasingly
stream becomes dilute somewhere in the process. Thus,
available (Chae et al., 2002; Heng, 2003b; McPherson,
even though the mass of product being produced may be
1999). There is increasing interest in new products such as
measured in kilograms, or even less, process volumes can
plasmids (Curling and Smiley, 2003; Deshmukh et al., 2000;
be quite large. Problems of contamination and mutation
Hahn et al., 2002; Jungbauer et al., 2003; Teeters et al.,
usually demand batch processing, and each product is
2003), vaccines (Aunins et al., 2000), and viruses. New
unique in important respects. Regulatory considerations
adsorbents are being developed (Hjorth, 2003; Palmgren
usually require that the basic separation scheme must be
et al., 2003; Sarkar et al., 2003) and used (Hahn et al., 2002;
fixed very early in the overall development process. In
Jungbauer et al., 2003; Phillips, 2003; Pieracci et al., 2003;
addition to strict standards on the purity of a product, the
Teeters et al., 2002). Even such widely accepted transport
Food and Drug Adminstration (FDA) requires validating the
mechanisms as intraparticle diffusion have been found to be
removal of various contaminants: host cell-related (DNA,
more complex than previously believed (Dziennik et al.,
endotoxin, protein, virus), product-related (oxidation, dea-
2003). Computational chemistry (DePhillips and Lenhoff,
midation, acetylation, dimerization), and process-related
2001; Koehler et al., 2000) and atomic probes (Tibbs
(antibiotic, antiform, inducing agent). Finally, the rising
and Fernandez, 2003) are increasingly helpful in choosing
cost of health care is beginning to put severe economic
extractants, and protein databases (e.g., http://www.rcsb.org/
constraints on processing cost.
pdb/index.html) are simplifying the hunt for effective
In summary, a safe and economic process must be found
separating agents.
quickly somewhere in an extremely large parameter space
Equipment performance, even for such old standbys
which simply cannot be explored in depth. Downstream
as stacked-disk centrifuges, is being examined critically
processors are forced to depend heavily upon heuristics.
(Abraham et al., 2003; Kaltenbrunner and Stokelman, 2003;
Separations design is a continuing battle to eliminate un-
Shpritzer, 2003; Staby et al., 2003; Taggart et al., 2003;
promising regions of parameter space until one finally ar-
Tebbe et al., 1996). Economics are being taken more
rives at an acceptable process. Our first step is to see what
seriously, and process details such as column packing
bioseparations have in common, and we begin that process
(Lightfoot and Cockrem, 1987; Sonnenfeld et al., 2003;
right now.
Williams et al., 2002) and the choice between relatively
minor process variations (Caren, 2003; Khosla et al., 1996;
Mustafa et al., 2003) are getting serious attention. Statistical
Primary Recovery
techniques for evaluating process performance are being
refined (Blackmore and Ryland, 2003; Larson et al., 2002, All processes considered here start with suspensions of
2003; Prior et al., 2003; Sonnenfeld et al., 2003; Taggart cells, and these are of three types: prokaryotes, i.e., bacteria

260 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004

(usually including E. coli), simple eukaryotes such as
baker’s yeast, S. cerevisiae, or live mammalian cells, for
example, those of human blood. One may also encounter
specialized mammalian cells such as immortalized Chinese
hamster ovary (CHO) cells.
The simplest of bioseparations conceptually is hemo-
dialysis (Galletti et al., 2000; Lysaght and Morgan, 2000),
used in the treatment of end-stage kidney disease. Here
the cell suspension, blood, is dialyzed against a carefully
controlled salt solution and ultrafiltered to remove water.
The membrane is a key element to success, and the cap-
ital cost of membrane production has been a major fac-
tor. Commercial cellophanes were used almost exclusively
until recently, and there is a strong tendency now to use
the body’s own membranes, i.e., peritoneal dialysis
(Lysaght and Morgan, 2000). This is a simple process from
the separations standpoint, but it is important economi-
cally. The major process engineering problems are the
costs of high-purity water and other expendables, reliability,
and control.
In almost all other situations cells must be removed from
the suspending liquid. Bacterial and yeast cells are quite
hardy, and they can be separated rather simply from the
suspending fluid by flocculation (Boonaert et al., 1999;
Sharma, 1999), mechanical filtration and centrifugation
(Axelsson, 1999; Hanley, 1999). Introductory discussions
are readily available (Harrison et al., 2003; Lydersen et al.,
1994; Wheelwright, 1991), and more extended treatments
are available in Flickinger and Drew (Flickinger and Drew,
1999). However, as indicated above these processes are
active areas of development. Mammalian cells are quite
fragile, and must often be conserved. They may, in fact, be
the desired product, as in aphoresis (Zydney, 2000), and we
shall have much more to say about them in connection with
microfiltration in the section on equipment performance
below. Increasingly however, they are being removed by
disk-stack centrifugation (Abraham et al., 2003; Axelsson,
1999; Hanley, 1999; Shpritzer, 2003; Tebbe et al., 1996).
Where the cells are the desired product, the next and final
stage is fractionation of cell types (Palsson and Bhatia,
2003), for example by flow cytometry (Shapiro, 2003).
Table I. Characteristic magnitudes.
Entity Mol. Weight Char. Length
Mammalian cells – 10 microns
E. colia – 1
1
3 micron
Inclusion bodies – 1 micron
Viruses 50 nm
DNAb 109
RNAb 106
Plasmids 3.3 – 13.2106 150 – 250 nm
Proteinsc 40 kDa 5 nm
a
Characteristic of bacterial cells and mammalian cell organelles such as mitochondria. These are corrections of data in Lehninger, Table 7 – 1 p. 123
(Lehninger, 1971). This table contains numerous errors that have been taken over uncorrected in Bailey and Ollis (1986), Table 5.6, p. 269.
b
From Lehninger (1971).
c
Approximate for albumins. The range encountered is from 10,000 to over one million.
LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS
This can be an important operation clinically, but it is be-
yond our scope here.
More commonly, however, we are interested in smaller
entities as indicated in Table I, and these may be found
either in the suspending fluid or inside the cells.
For proteins, until recently, one had only the choice
of secreting them to the suspending medium in their na-
tive state or forming them inside the host cell. Internal
protein can be in soluble form, but normally much more
can be accommodated as inclusion bodies, large granules
of nearly pure but aggregated and insoluble protein. It now
appears however, that a third choice is becoming realistic:
forming the protein of interest in the bacterial periplasm
(Chae et al., 2002) and releasing it by osmotic shock.
The resulting protein is biologically active and very nearly
pure. This is the first major decision point in the design
process and an example of interaction between upstream
and downstream processing.
Solutions of secreted proteins are normally dilute and
heavily contaminated by other proteins (O’Keefe, 2000) as
well as unrelated species, but the proteins are typically
biologically active. Inclusion bodies are concentrated and
nearly pure, but they must first be solubilized and then
renatured (De-Bernardez Clark, 2001; Middleberg, 2002).
This normally requires dilution and both breaking and
reforming disulfide bonds. Moreover, the cell wall must be
lysed, and the inclusion bodies must then be separated
from cell fragments, usually by differential centrifugation
(Harrison et al., 2003; Wheelwright, 1991). A somewhat
analogous but more complex situation exists for vaccines
(Aunins et al., 2000; Wen et al., 2003).
One also has the choice of producing proteins in E. coli,
which is normally relatively inexpensive and well under-
stood, or in mammalian cells. The latter may be necessary
if proper glycosylation or other post-translational pro-
cessing is needed.
Plasmid DNA (Curling and Smiley, 2003; Deshmukh
et al., 2000; Harve and Bajpai, 2000; Teeters et al., 2003)
exists only within cells, and here lysis is complicated by
shear degradation of genomic DNA and the high effective
viscosity of cell lysates.
Density g/cm3 Mass Other
1.07 2.41012 g
f21013 g
3,000 – 20,000 Base pairs
1.34 5.641020 g
261
Final Purification
This is the most complex aspect of bioseparations, and it
tends to take the form of a long series of unit operations
using a variety of separations mechanisms and equipment.
Elaboration of a separations scheme is still highly dependent
upon the personality and experience of the designer. We
will have much to say about this in our discussion of de-
sign strategy.
As this review is being written, fixed-bed sorption
operations, in particular various forms of chromatography
(Adams et al., 1999; Cramer and Natarajan, 1999; Guiochon
et al., 1994; Kaarsnaes, 1999; Lightfoot, 1999; McGuire and
Bothwell, 1999; Rathore and Velayudhan, 2003; Roth and
Yarmush, 1999; Thoemmes, 1999) are dominant for recov-
ery of valuable products from impure dilute solutions and
for separations of closely related materials. They will be
discussed in some detail below.
Membrane filtrations (Ho and Zydney, 2002; Kalyanpur,
1999; Noble and Stern, 1995; van Reis and Zydney, 1999;
Zeman and Zydney, 1996) also discussed further below, are
widely used means for collecting cells and other partic-
ulates, changing solvents, and performing less difficult
separations. They are also the method of choice for con-
centrating either cells or true solutions when only the de-
sired species is retained by the membrane. Moreover,
membrane filtrations are much cheaper than chromatog-
raphy, and available selectivity is rapidly increasing
(Christy et al., 2002; van Reis and Zydney, 1999; van Reis
et al., 1991, 1997a, 1997b; van Reis and Saksena, 1997;
van Reis et al., 1999). Membranes have already largely
replaced gel permeation chromatography (Kurnik et al.,
1995), and further encroachments can be expected.
For example, precipitation induced by temperature, pH,
salts, or alcohols has been an important commercial means
of capturing and purifying proteins since development of
the Cohn process for fractionation of human blood (Chang,
1988; McIntosh et al., 1994; Stryker et al., 1985). Var-
iations of this venerable process (Chang, 1988; McIntosh
et al., 1994; Stryker et al., 1985) are still being used, in
particular for human serum albumin, which can be heat
sterilized. Applications of crystallization in modern bio-
technology (Becker and Lawlis, 1991; Becker et al., 1997;
de la Rosa and McPherson, 1999; Harrison et al., 2003,
Ch. 8; Heng, 2003a, 2003b; Johns, 1999; McPherson, 1999)
are increasingly important. In addition to its uses in insulin
production (Johns, 1999) it has proven effective for the
capture and purification of industrial enzymes (Johns,
1999). It appears destined for wider use.
Liquid extraction is used where it offers special advan-
tages, for example, in penicillin production and aqueous-
aqueous systems for proteins (Harrison et al., 2003; Ch. 6.2,
p. 171). Liquid membrane systems, long considered prom-
ising, do not seem to be playing an important role at present.
Other processes find use in special circumstances. For
example, density gradient centrifugation is used in vaccine
production (Aunins et al., 2000), dielectrophoresis (see Bird
262 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004
et al., 2002; pp. 784– 785) is finding a great many ap-
plications for very small-scale separations, and magnetic-
fields appear promising (Cramer, 2002; Zborowski
et al., 2002). They are already widely used to collect solute
loaded magnetic sorbent particles. Flow cytometry is used
to fractionate cell mixtures, again on a very small scale
(Shapiro, 2003).
In the next section we provide a rationale for assembling
these individual operations into a coherent process. Some-
times the results are simple, but more often they are quite
complex. The production of biosynthetic human insulin is
an excellent example discussed in detail by Harrison et al.
(2003; p. 349). In the two-chain method developed by
Genentech (San Francisco, CA) the two chains of the in-
sulin molecule are produced separately as inclusion bodies.
The inclusion bodies are solubilized and renatured, purified
and combined to produce the desired highly purified pro-
duct. The overall downstream process contains nearly 30
major process steps.
Our subsequent discussion will emphasize primary
recovery and ‘‘final’’ purification, but packaging and related
separations are also important. Operations in this last
category include crystallization and drying, and they are
discussed at some length in Harrison et al. (2003). Auxiliary
nonbiological separations such as water purification and
solvent recovery also represent significant costs. Water
processing and other supporting operations are discussed in
Lydersen et al. (1994). Solvent-related costs often dominate
the economics of chromatographic operations, but discus-
sion of them appears to be fragmentary and scattered.
PROCESS STRATEGY
All process design is heuristic, and success depends heavily
upon adopting an effective structure for the problem at
hand. Each problem is unique in this respect, and dis-
cussion must be largely by way of example. We provide an
overview of processing dilute feeds, with emphasis on
therapeutic protein production, which is among the most
complex of bioseparations. The production of vaccines will
be left to Aunins et al. (2000) and processing of inclusion
bodies will be left to available references (De-Bernardez-
Clark, 2001; Harrison et al., 2003; Middleberg, 2002). An
excellent discussion of antibody production is provided by
Boschetti and Jungbauer (2002), and oligonucleotides are
discussed by Deshmukh and Warner (2000).
A sense of process design strategy is based on ob-
servations (Lightfoot and Cockrem, 1987) that materials
handling dominates the cost of recovery from dilute
solution, the usual situation for proteins. Typical data for
recovery of biologicals shows that overall cost of recovery
is inversely proportional to their concentration in the feed
to the separation train needed to produce them from their
original source—and independent of final purity. This
situation already holds for commodity chemicals such as
citric acid, at a few grams per liter, and is found for all
more dilute feeds.
 This simple observation has had great impact on process
design and is responsible for much of the differences
between laboratory- and commercial-scale operations.
Solvent consumption, the cost of pumps, lines and tanks,
and even waste disposal, become major considerations in
the plant while in the laboratory the main cost is labor.
Plotting separation trajectories on a plot of solute
thermodynamic activity vs. concentration as in Figure 1
provides a good illustration of this. In general, the best tra-
jectories are the most concave. We begin by looking spe-
cifically at two relatively well-known protein concentra-
tions by way of example: comparing the classic laboratory
procedure of ‘‘salting out’’ proteins, at upper and left, with a
commercial ion exchange process at lower and right. In
the first process addition of ammonium sulfate to a high salt
concentration increases the activity of a typical globular
protein at the expense of modest dilution, and it sponta-
neously concentrates as indicated. This spontaneous con-
centration will frequently form as a relatively impure
precipitate rather than a pure product as in crystallization.
Also, one now has the expense of the ammonium sulfate
plus a rather large waste disposal problem. Alterna-
tively, one can spontaneously adsorb the protein on an
ion-exchange column and then desorb it with a very small
volume of salt solution, thereby greatly reducing the volume
of the process stream.
At first sight the second is the more attractive, but
deciding between these two alternatives is a matter of
judgment. Crystallization generally produces a high-purity
product directly, but if the mother liquor is too complex
there may be a major coprecipitation of very troublesome
impurities. Crystallization, as opposed to relatively non-
selective precipitation, can be a very powerful purification
mechanism. True crystallization is difficult to achieve, but
it has proven effective in a number of commercial pro-
cesses (Heng, 2003a). Here phase separation is normally
induced by a carefully controlled decrease in temperature
rather than by adding a precipitant.
Figure 1. Separation paths.
LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS
Adsorption is still the more popular method of concen-
tration, but this is not the end of the story. As just one
example, protein produced in bacterial periplasm should be
a good candidate for crystallization because of the high
purity possible. Here again we find a strong interaction
between upstream and downstream processing.
In any event, it generally remains to separate the product
in this concentrate from closely related substances and to
remove traces of potentially harmful impurities. It is now
generally accepted that final purification involves three
conceptually different operations:
 Capture (or concentration)
 Separation (or fractionation)
 Polishing
Capture
The capture step is normally the most expensive, and it is
here that close attention should be paid to both equipment
and process design. Batch adsorption (see below) in fixed
beds is by far the most common means of capture as these
beds are simple to operate and reliable. They normally can
process many bed volumes of feed per batch and are easily
reversible. Simulated moving beds (Nicoud, 2000) are
promising because they require many fewer stages (Roper
and Lightfoot, 1993; 1994). However they suffer from two
major problems. They require a heavy capital investment,
and they are so complex that on-line control seems are yet
to be achieved. They are seldom used during capture, but
see ‘‘Separation’’ immediately below.
Liquid extraction can suffer from the formation of stable
emulsions, and it is usually difficult to reduce process
volumes by more than a factor of 5 to 10. Solvents are
generally less selective than adsorbents, and it is often
difficult to find effective reversible systems. Selective
membrane filtrations are usually not well suited to large
fractional volume reductions.
Separation
Separation or fractionation generally requires the most
chemical knowledge and familiarity with available sepa-
rating agents. It is also usually performed by adsorption in
fixed beds, often via gradient elution (see below). This is,
in part, because adsorptions can be highly selective and a
great variety of adsorbents are available commercially.
Adsorbent selection is a major problem, and much attention
is being given to it (DePhillips and Lenhoff, 2001; Hearn,
2000; Mazza et al., 2001; Staby et al., 2003; Tibbs and
Fernandez, 2003). Moreover, detailed structures are now
available for a great many proteins (e.g., http://www.rcsb.
org/pdb/ index.html).
Simulated moving beds are proving highly successful for
the resolution of optical isomers and other situations where
continuous processing is justified, and the products tend to
be pure and stable. Here they are claimed (Nicoud, 2000) to
263
reduce process development time as well as to be effective
and economical. They have not yet proven successful
for proteins.
Crystallization is a very powerful technique at this stage,
and it will probably become increasingly important.
Polishing
This should be the least expensive, but it may require the
most care. It is the last opportunity to avoid damaging
impurities. In the specific case of viruses, two or more inde-
pendent means for elimination are typically required (ICH,
1999). Membrane filtrations play a major role here, and
disposable adsorption cartridges are widely used for viruses,
DNA, RNA, and other minor but serious toxic materials such
as pyrogens. Adsorptive membranes are proving very at-
tractive here (Deshmukh and Warner, 2000), and their use
seems to be increasing rapidly. Specifications often require
reductions of many powers of 10 (‘‘logs’’ in current usage).
Changes of solvent are also common here, and diafiltration
is rapidly replacing size-exclusion chromatography.
Practical Considerations
These functions may however be combined, and it is
particularly important to carry out as much fractionation as
possible in the first step. In any serial process overall re-
covery is the product of step recoveries so minimizing the
total number of steps is important. We shall consider here
adsorption-dominated processes as a commonly encoun-
tered example, and we begin by discussing adsorbent prop-
erties. However, the choice of equipment type is important
at every stage.
Moreover, one can never know enough about the phys-
icochemical characteristics of one’s product and its chief
impurities: essentially all really good separations processes
depend upon anomalies. Detailed structures are available
for a great many proteins (e.g., http://www.rcsb.org/pdb/
index.html), and, as indicated above, much attention is
being directed to making use of this information.
A great variety of sorbents is now available (Adams et al.,
1999), and chief among them for preparative purposes are
size exclusion (SEC; Roth and Yarmush, 1999), ion-
exchange (IX; Cramer and Natarajan, 1999), hydrophobic
interaction (HIC; Kaarsnaes, 1999), and ‘‘affinity’’ (Stahl
et al., 1999) adsorbents. There are a great many subtypes
within each group, including for example, both weak and
strong acid and base exchangers —and many gradations of
affinity systems. These latter include immobilized metal
ion affinity (IMAC), dye affinity, and specialized adsorb-
ents such as protein A in addition to true immunologically
specific systems. They come in many sizes and forms, from
the common spheroids to adsorptive membranes, monoliths,
and coated tubes—and they exhibit a wide range of internal
structures (Lightfoot, 1999).
As one extreme, a highly selective sorbent can be used
on the feed stream itself. Examples include use of Protein
264 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004
A for antibody isolation and putting ‘‘tags’’ on recombi-
nant proteins. Thus, histidine tags that may be removed
later, can make the use of nickel affinity columns attractive.
However, this approach can prove to be an inefficient use
of an expensive material. Sorbent life can be reduced by
direct exposure to a complex mixture, and both selectivity
and capacity reduced by nonspecific binding of impurities.
One more commonly performs the first two functions
using complementary adsorbents and begins with one well
suited to the feed stream. A popular example is to begin with
an ion-exchange (IEX) column and to follow it with one
depending upon hydrophobic interaction (HIC). The ion
exchanger can probably adsorb proteins directly from the
feed stream with only pH adjustment, as salt concentration is
initially low. The salt needed for elution is also helpful for
adsorption on the hydrophobic interaction column. One
can change solvents between purification steps, as by gel-
permeation chromatography, or more probably diafiltration,
but each step increases cost and reduces yield. Finding
selective adsorbents is still very much an art, but knowledge
of protein composition and structure is rapidly improving
as already indicated.
Steps for the removal of contaminants, particularly DNA,
endotoxins, and virus, must also be incorporated into the
purification process. A large portion of these clearance steps
utilize negative chromatography, a mode of operation where
the product of interest is unretained while the impurities are
absorbed. Frequently, the final product stream is passed
through an anion-exchange resin for the removal of DNA
and negatively charged viruses.
In addition to the removal, the production, and purifica-
tion of vaccines, plasmid DNA and viruses have become
areas of increasing interest. These macromolecules behave
quite differently than proteins, although the same basic
heuristics appear to apply. Vaccine production is not new,
but manufacturing techniques seem less well developed than
for proteins (Aunins et al., 2000). Interest in plasmids is
rather recent, but the general outlines of a separation strat-
egy are beginning to take form (Deshmukh and Warner,
2000). A major problem is the similarity between the double-
stranded DNA plasmids, genetic DNA, and RNA, but the
large size of plasmids, with typical sizes of 3,000 to 20,000
base pairs, presents major difficulties as well. Capacity of
conventional adsorbents designed for protein separations is
typically very small, and only a very few of available ad-
sorbents appears suitable (Deshmukh and Warner, 2000;
Jungbauer and Kaltenbrunner, 1999; Teeters et al., 2002).
EQUIPMENT SELECTION, DESIGN,
AND OPERATION
Selection and Design
We now turn our attention to the design and operation
of the key process equipment: adsorption columns and
membrane filters. Though different from each other both
are typical of separations devices in their sensitivity to
process conditions as listed in Table II. The first require-
ment, high solute capacity, helps to determine the physical
and chemical nature of the process streams while the next
four tend to fix the geometry. The last is geometry sensitive
as well, but is also highly dependent on operating modes.
As a practical matter the interaction of mass and
momentum transfer tends to determine the general nature
of the design, and there are only a few ways to obtain a
satisfactory balance. Each choice leads to a ‘‘family’’ of
devices, each of the latter distinguished from the others of
its family by the way the problems of dispersion and flow
control are dealt with. The ways in which mass transport
and flow interact are quite different for chromatographic
and membrane systems, however, so we shall discuss the
two separately.
Chromatographic Apparatus
It is instructive here to look briefly at mass vs. momentum
transfer in a column packed with spherical adsorbents:
Mass transfer (Bird et al., 2002; Sect. 19.5):
@ci =@t ¼ f ðt ; r  ; IC; BCÞ ð1Þ
where
t ¼ tD im =R2 ; r  ¼ r=R ð2; 3Þ
Equation (1) is just a general expression for solute
transport showing the time dependence of solute concen-
tration ci , or mass of i per unit volume, on particle radius
R and effective diffusivity Dim within the particle. This
simple dependence is found for all initial (IC) and
boundary (BC) conditions as well as for the boundary
layer surrounding the sphere, except that in this latter case
the diffusivity is for the external solution. For all of these
situations time always appears in the dimensionless ratio
(tD im =R2 ).
Momentum transport in a packed bed of spheres (Bird
et al., 2002; Sect. 6.4):
F ¼ dðmvÞ=dt; dðmv=VÞ=dt ¼ dp=dz ð4; 5Þ
2
: 150 ð1  qÞ
dp=dz ¼ v ð6Þ
4 q3
This second set of equations shows the dependence of
the rate of momentum (mv) transfer per unit volume V on
Table II. Guidelines for separator design.
Maximize solute capacity.
Maximize lateral mass transport: small diffusion distances and large mass
transfer surface.
Minimize momentum transport (pressure drop): large diffusion distances
and small surface.
Minimize axial dispersion: small diameters, and especially so for tubes.
Maintain flow uniformity.
Minimize fouling of the transfer surfaces.
LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS
particle radius, solution viscosity A and void fraction q
for a packed bed of spheres. (Note that rates of momentum
transfer are finite even at steady state). Here v is the
superficial fluid velocity. Note that momentum transport
per unit volume is the axial pressure gradient dp/dz in
the column.
It may be seen immediately that changing particle size has
no effect on the ratio of mass to momentum transport per unit
volume, but that momentum transport can be changed
independently of mass transport by changing either the
velocity or void fraction. Moreover, momentum transport is
independent of the internal structure of the packing. It is
suggested (Table III) that these simple criteria are sufficient
to classify existing chromatographic devices into ‘‘fami-
lies’’ based on the way they deal with this ratio. We now turn
to this task. The examples listed are taken from the authors’
personal experience only and are not endorsements.
Each of these ‘‘families’’ has found commercial appli-
cations and has both generic strengths and weaknesses. The
high-pressure columns can exhibit tens of thousands of
plates, but they must be carefully maintained to do so. The
small interstices are sensitive to fouling, and cracking of the
packing particles can partially occlude the downstream
header screens and cause flow maldistribution. Adsorptive
membranes (Deshmuk and Warner, 2000; Gebauer et al.,
1997) can accommodate very large solutes and high
percolation rates. However, they tend to show flow mal-
distribution because of the wide aspect ratios, and high
pressure drops limit packed depth. Presssure limitations, in
turn, increase costs and complicate scale-up. These devices
are most popular during polishing. The ‘‘square’’ columns
(length of the order of diameter) are very widely used, and
both headers and packing systems are highly effective
(Moscariello et al., 2001). Batch adsorption and gradient
elution operations often do not require extremely high plate
Table III. Chromatographic families.
High-pressure packing of small particles: Massive columns, strong
packings, and high pressure pumps; very high resolution.
Prochrom Dynamic Axial compression
Short and wide columns: low percolation velocity and reduced pressure
gradients; increased distribution problems. Efficiency insensitive to
percoation velocity.
Adsorptive membranes (Deshmukh and Warner, 2000): Pall Mustang
Disk; Pharmacia Radial Flow
‘‘Square’’ columns: Often a good compromise with reasonably low
pressure drop and moderate header problems.
.Millipore Isopak
Upflow through expanded beds: Very low pressure drops, ability to deal
with particulate suspensions, need for dense packings and careful
alignment.
Pharmacia Streamline
Parallel bundles of internally coated tubes: Low pressure drops and high
efficiency but need for small pore diameters to minimize axial
dispersion; not yet adapted for bioseparations. Alltech Multicapillary
Modified packings: A great variety and probably still increasing. See
discussion in text.
265
counts, and these columns are probably the workhorses of
adsorptive bioseparations. The expanded bed systems are
now being substantially upgraded, (Frej, 1999; Hjorth,
2003), and they offer the major advantage of being able to
handle large particulate loads. This gives them a major
advantage at the capture stage and can eliminate one whole
operation in the separation sequence: removal of cells or
other particulates. The parallel, coated tubes represent a
major technological achievement and could be considered
highly ordered monoliths. They give very favorable ratios
of mass to momentum transfer and excellent plate counts.
They should be resistant to fouling, but they have so far, to
our knowledge, not been applied to liquid systems as yet.
There seems to be no serious obstacle to doing so.
A very wide range of packings is also available, but a
large fraction of them are spherical aggregates of very small
impermeable spheres with derivatized surfaces. These
packings are strong and stable, but they have only a modest
volumetric capacity. Moreover, the bulk of adsorbed solutes
are on these derivatized surfaces where their effective dif-
fusivity is rather small. Hydrogels have both a higher ca-
pacity and more rapid diffusion of internal solute, and they
are therefore more effective for dilute feeds. However hy-
drogels are too soft for most applications of interest here.
Their strength can be much increased by providing a rigid
skeleton, and both alumina and silica have been used to date.
Zirconia skeletons are being actively investigated to elim-
inate the sensitivity to cleaning procedure that is one of
the few weaknesses of these adsorbent systems. Adsorptive
membranes, monoliths, and perfusive adsorbents are among
the few that appear promising for plasmids, viruses, and
other very large solutes of rapidly increasing current interest.
Membrane Separations
Our primary concerns here are with micro- and ultra-
filtration, and both equipment design and operation are
covered in readily available references (Belfort et al., 1994;
van Reis et al., 1997a; Zeman and Zydney, 1996). These
are designed to provide high ratios of filtration rates (mass
transfer) to pressure drop, or drag, in the fluid flowing past
the membrane. Most depend upon laminar rectilinear, or
near rectilinear, flow of the process stream to minimize or
even eliminate form drag. Basic geometries of this type
include bundles of hollow fibers, flat plates, and spiral
wound modules. However, it is also possible to improve the
mass transfer to drag ratio by use of secondary flows, and
this can be accomplished either with vortex flows or, more
recently by inducing Dean vortices (see, for example,
Schutyser and Belfort, 2002; Schutyser et al., 2002). Taylor
vortices (Belfort et al., 1994) are also effective, but they are
not widely used at present. Flow control is also important in
the case of ultrafiltration to limit concentration polariza-
tion, and for both processes, a complex series of events known
collectively as fouling (Zeman and Zydney, 1996; Ch. 9).
Basic transport problems include minimizing flow re-
sistance across the membranes and providing mechanical
266 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004
support without occluding membrane surface unduly. These
problems are less severe for filtration processes than with
dialysis as here there is no diffusional resistance on the
downstream side of the selective membrane. This situation
permits what is the almost universal use of asymmetric bi-
layer membranes, which, in turn, go a long way to resolving
both of these difficulties. These membranes consist of a
very thin selective layer tightly bonded to a much thicker
and more porous nonselective backing. The production
of such asymmetric membranes is a major technological
achievement (Zeman and Zydney, 1996), but we will not
address it here in what is basically a user-oriented dis-
cussion. This membrane can then be supported by a variety
of supports such as screens.
Operation
In this final section we discuss the operating character-
istics of adsorption columns and membrane filters, two
major classes of commercial-scale bioseparation equip-
ment. We shall leave pumps, dead-end filters, and a number
of operations such as dialysis to the references already
provided. Adsorptive membranes (Deshmukh and Warner,
2000) are in a sense a hybrid of chromatography and fil-
tration as already indicated above. They are widely used to
remove trace impurities and are promising as chroma-
tographic media.
Chromatography
Historically the term chromatography has been used to
group three distinct but related processes: differential
chromatography, displacement chromatography, and fron-
tal analysis. To these, we will now add a fourth: gradient
elution. Differential chromatography is the basis for a
variety of analytical applications (Snyder et al., 1997) and
it is widely used to determine column efficiency. Moreover,
it is useful here to introduce the differential equations
describing all operating modes and much of the nomen-
clature used for them. Differential chromatography is also
useful for defining differential migration as the first of two
chromatographic separation mechanisms as well as the
important concept of time constants. Displacement chro-
matography is a powerful and basically efficient separation
technique, but it has not really caught on very well in
commercial bioseparations. It will be left to the literature
(Shukla and Cramer, 2000). Frontal analysis is very widely
used in the rudimentary form of batch adsorption for the
capture stage of many bioseparations, and we will provide
an introduction to its dynamics here. Gradient elution is
very effective for the fractionation stage of the overall
recovery process, and it depends heavily upon the second
separation mechanism: differential elution. Under favorable
circumstances it permits excellent separation even in very
short columns. In fact, such separations known as ‘‘on –
off’’ chromatography are more damaged by using too long
a column. We now proceed to examine each of our three approximation. We find then that for a Gaussian effluent
selected processes in turn. distribution the number of plates is just the reciprocal of
the variance of the effluent curve when normalized by the
mean residence time. This is the key result of our devel-
Differential Chromatography and General Concepts
opment, and it can be written in terms of temporal moments
This is a limiting process in which a short solute pulse is of the effluent curve:
introduced to a previously solute-free column and migrates
2 1 32
down the column as a result of a steady solvent percolation. Z1 Z1 Z
Solute concentrations are considered sufficiently low that 1=N ¼ cðt  tÞ2 dt  cdt =4 ct dt 5 dt
the entire set of equations are linear in solute concentration,
0 0 0
c. There are many ways to write the defining equations, but
the more reliable ones (e.g., Reis et al., 1979) are written ¼ m2  M0 =m21 ð14Þ
for a bed of uniform spheres and begin with a differential
mass balance for a section dz of a column with uniform Here mn is the nth temporal moment of concentration
percolation velocity: relative to the mean residence time, and Mn is the nth

 absolute moment as defined by this equation.
@cf @cf @ 2 cf @csm These definitions of N and H are consistent with that
q þv  E 2 ¼ ð1  qÞ ð7Þ
@t @t @z @t of Martin and Synge when the effluent curve is Gaussian,
but it is useful even when it is skewed (see below) (Martin
Here q is bed void fraction, z is distance of any sphere center
and Synge, 1941). For a Gaussian distribution N is just the
from the column entrance, t is time, v is now mean intersti-
square of the ratio of solute mean residence time to
tial velocity, E is a convective dispersion coefficient and c is
standard deviation.
solute concentration. The subscripts f, s, and m refer to fluid
All of the terms in the above equations have simple
phase, solid phase, and to volume mean values inside the
physical significance. Thus, u is just the equilibrium
adsorbent, respectively. The problem is how to deal with
fraction of solute in the mobile phase, and a is a simple
mass transport within the adsorbent spheres, and the sim-
distribution coefficient. The overall mass transfer coef-
plest approach is to use a lumped-parameter approximation.
ficient is for most modern column operations a slowly


4 3 @csm varying function of velocity (Athalye et al., 1992), and at
kR ¼ ð4kR2 ÞKc;tot ðacsm  cf Þ ð8Þ high velocities it and the dispersion coefficient approach.
3 @t
Here a is an interphase distribution coefficient and Kc,tot 5D is
is an overall mass transfer coefficient to describe the com- LimðKc;tot Þjv!"1" ¼ ; LimðEÞjReSc !"0"  2:5Rv
R
bined effects of boundary-layer and intra-bead diffusional
ð15; 16Þ
resistance (Bird et al., 2002; Sect. 22.4).
Equations (7) and (8) can be integrated for introduction
of a small solute pulse containing a solute mass, mo, to the Since Schmidt numbers, SC = A / U D im, are quite high
column at zero time. For all but exceptionally short columns for all separations being considered here, both limiting
the result is conditions in these two equations are frequently met and
one can approximate the prediction of plate height as
2   2 3

 ðt 1Þ  
m0 4 exp  2j2 5 R2 uvð1  uÞ
cf ðL; tÞ ¼ pffiffiffiffiffiffiffiffiffiffi ð9Þ H  2 2:5R þ ð17Þ
qA0 vt 2kj2 15aD is

Here Ao is the column cross-sectional area and Equation (12), with somewhat more accurate expressions
 for E and Kc,tot, has been found to agree within experimental
t ¼ t=t; t ¼ L=uv
error for factory-packed size-exclusion columns (Athalye
et al., 1992). However, the dispersion limit depends strongly
2t Ttot 1 H
j2 ¼ ¼ ¼ upon void fraction, and the mass transport contribution
t N L
depends upon both the reliability of Eq. (8), and the
  assumption of Fickian transport within the particles.
E RC
Ttot ¼ þ Both of these assumptions need further examination.
uv2 Kc; tot
(10, 11, 12, 13) First, one must note that the effluent curve is never truly
ð1  uÞ Gaussian, and therefore that plate height varies with time
C¼ for any given situation: the definition in Eq. (10) contains a
3a
variable, the scaled time t*. Normally, this problem is
Equations (9) through (13) provide a basic definition of resolved by assuming t* to be unity, but this is only valid
the plate height, H, valid wherever Eq. (8) is a reasonable for very sharp peaks. Frequently more serious, Eq. (8) is

LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS 267

only valid if the diffusional response time of the adsorbent
beads, usually taken to be
tdif  R2 =6D is ð18Þ
is small compared to the time scale over which the local
fluid concentration may be considered unchanging. This is
usually taken to be the (unscaled) standard deviation of the
effluent curve jt . In modern columns this requirement is
frequently not met. Fortunately, a more complete solution
of the governing equations in which Eq. (8) is replaced by
Fick’s second law has been provided in terms of temporal
moments (Schneider and Smith, 1968). It has therefore
been found, as shown in Figure 2, that Eq. (14) does yield
a linear relation between plate height and percolation
velocity in the high velocity limit as suggested by Eq. (15).
Here the reduced plate height is H/2R, and the reduced
velocity is 2Rr=D if. The upper curve was obtained from
data as shown and the moment ratio of Eq. (14). The lower
curve was obtained from a ratio of peak height to width
as indicated. The apparently low plate heights obtained
by this latter method are very misleading because of
the serious tailing found when values of (R2 =6jt
D is) be-
come appreciable.
The method of moments is however, subject to error
because of the heavy weighting given to large time data. It
is therefore often convenient to fit the data with an asym-
metric distribution, such as an exponentially modified Gaus-
sian (Jeansonne and Foley, 1992). This task is simplified by
recognizing the three moments needed, the first absolute and
second relative to mean residence time t already introduced
as well as the third relative to t are additive. Thus, one
can easily eliminate the effects of pumps and lines, and
even the headers, by simply subtracting the moments for
these ancillary structures from the observed data.
Finally it has been shown in a rather elegant way
(Klinkenberg and Sjenitzer, 1956) that contributions to the
plate height are additive as predicted by Eq. (12). Note that
the additivity of boundary layer and internal diffusional
Figure 2. Plate height estimation.
268 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004
resistance implied by using the overall mass transfer
coefficient (Bird et al., 2002; Sect. 22.4). This implies a
large degree of model insensitivity so that Eq. (12) can be
written as
E eff RC
Ttot ¼ ¼ ð19Þ
uv2 Kc; eff
for correlating data at any given percolation velocity and
particle size. There is no real advantage to doing this here,
but it can be very convenient for obtaining a fit to data in
the batch adsorption mode being considered next.
We have now concluded our basic discussion of differ-
ential chromatography, but we shall find that much of it has
application below. We note here that an important applica-
tion is the testing of columns for packing non-uniformity
(Lightfoot et al., 2003b), which can be a major practical
problem (Farkas et al., 1997; Guiochon et al., 1997; Shalliker
et al., 2002, Williams et al., 2002; Yew et al., 2003).
Another application is to provide insight as to scale-up of
laboratory data, a subject that has generated much more
heat than light. There are many proponents of maintaining
column height and percolation velocity in scale-up, but this
can lead to very wide short columns and corresponding
problems of flow distribution. We are part of a large group
advocating scaling up at constant mean solute residence
time t if the packing is to be kept identical, and using at
least the percolation velocity of the laboratory data. It can
be seen from the shapes of all van Deemter plots that it is
a safe procedure. If larger packing diameters are to be used
mean residence time should be increased to maintain the
same ratio of to the square of particle diameter. This will
be a safe procedure even if, as in some forms of affinity
chromatography, adsorption kinetics are slow.
Batch Adsorption
A typical capture step consists simply of feeding a treated
broth to an adsorption column until an appreciable amount
of product appears in the effluent, a condition known as
breakthrough. Almost always the equilibrium isotherm for
this situation is ‘‘convex,’’ which is to say the equilibrium
ratio of solute concentration in solution to that adsorbed
decreases with increasing concentration. For such satu-
rating isotherms solutes migrate faster as their concen-
tration increases, and the concentration front approaches an
asymptotic shape known as a constant form front.
There is normally little concern for the mathematics of
this situation, and the shape of the front depends only upon
the migration velocity and particle size for any given
adsorbent system. As a result scale-up of laboratory data is
relatively simple, and the rules just suggested above suffice,
except for the problems of packing nonuniformity and the
flow distribution characteristics of the headers. However,
it does appear that a bit more analysis is in order, if only to
see if the headers are working properly and if the column is
well packed.
 The shapes of nonlinear fronts, and the rate at which the
asymptotic shape is approached can in fact, be predicted in
terms of the isotherm shape and internal diffusivity. Perhaps
the most powerful description of nonlinear column processes
is that of Rhee and Amundson (1982) that can deal with
interfering solutes and such complex processes as displace-
ment chromatography. However, the simpler analysis of
Baddour et al. (1954) is sufficient for batch adsorption and
can also describe the behavior of finite bands migrating
down a column. The functions used in this development can
be handled simply by MathematicaB and other available
programs. The specific rate model used is not important
because of the model insensitivity already noted.
Many modern adsorbents saturate at solute concentra-
tions much below those in the feed (‘‘irreversible’’ in the
current jargon), and for these the concentrations take the
very simple form
1 ultrafiltration (Zeman and Zydney, 1996) in tangential flow
cf =cfeed  ð20Þ
1 þ expðL t Þ
for L* > c 5. Here
L ¼ L  ½ðki =qÞð1 þ f Þ=r ð21Þ
t ¼ tf ki =q ð22Þ
q cfeed
f ¼  ð23Þ
ð1  qÞ cs
and c*s is the solid-phase solute concentration in
equilibrium with the feed. The rate constant ki = (aKc)
and is perhaps best considered empirical.
Internal diffusion coefficients can be handled by a
number of means (see, for example, Lewus et al., 1998).
Use of batch adsorption data to evaluate column efficiency
has been discussed recently (Lightfoot et al., 2003a), and
also the rational design of headers (Yuan et al., 1999).
Gradient Elution
Gradient elution is the selective mobilization of solutes from
a preformed adsorption band by controlled increase of an
elutant in the feed stream to a column. It is perhaps the
most widely used operating mode for protein fractionation.
A common example would be the addition of increasing
amounts of salt in the feed to an ion-exchange column, but
this general process can be used for any of the adsorbent
types previously described (Cramer and Natarajan, 1999;
Kaarsnaeas, 1999; Stahl et al., 1999). This process depends
on a species-dependent response of the adsorption equilib-
rium to elutant concentration, and it thus adds a new
separation mechanism to the selective migration discussed
previously: selective elution (Coffman et al., 1994; Yama-
moto et al., 1993; Yamamoto, 1995). In favorable cases,
known as ‘‘on –off’’ separation, the desired product can be
released without any appreciable mobilization of others. In
this limiting case only a single ‘‘plate’’ is needed, and ad-
ditional plates are harmful: once all solutes are fully eluted
LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS
the corresponding bands move at the same velocity, but
they continue to spread.
Normally some combination of selective elution and mi-
gration is needed, and in principle the optimum operation
procedure is a series of step changes in elutant concen-
tration. However, uncertainty in both the adsorption equi-
libria and the control of solvent proportioning makes
continuous changes the preferred mode (Kaltenbrunner and
Stokelman, 2003).
Gradient elution found its first applications in analytical
separations, and much of the basic theory was developed
for that purpose. This work has been extensively reviewed
in Snyder et al. (1997) and in Guiochon et al. (1994).
Membrane Filtration
Here we shall concentrate our attention on micro- and
although reverse osmosis is important for ancillary pur-
poses such as water purification (Ho and Sirkar, 1992;
Noble and Stern, 1995). These processes are deceptively
simple at first sight as they consist of partially removing
solvent in flow across a membrane in modules of quite
simple geometry. Reality is much more complex partic-
ularly because of fouling and the formation of dynamic
boundary layers.
Microfiltration
Microfiltration is used to remove particulates from true
solutions, and microfiltration membranes typically exhibit
effective pore sizes between about 0.03 to 1.0 micron in
diameter. It is used primarily to remove cells and cell debris,
and these entities exhibit very low effective diffusivities.
As a result, the retained entities tend to adhere to the mem-
brane surface and thus reduce water fluxes. A variety of
techniques are used to resolve this problem, including both
introducing secondary flows (see, for example, Schutyser
et al., 2002) and time-dependent transmembrane pressures
(Belfort et al., 1994). Operation is also complicated by low
intrinsic transmembrane pressures that can even result in
flow reversal. Simplified mathematical models are of re-
latively little use, and it is perhaps best to review case his-
tories (Ho and Zydney, 2002; van Reis et al., 1991).
Ultrafiltration
Ultrafiltration has proved to be a remarkably powerful and
versatile process that is continually finding new applica-
tions. First used for concentrating macromolecules such as
proteins it became prominent if not dominant for buffer
exchange, and it is now being used for protein fractiona-
tion. In the case of buffer exchange it is common to use the
diafiltration mode where product concentration remains
nearly constant. Protein fractionation is rapidly becoming
more selective through improvements in membrane design.
Repulsive electrostatic interactions between the charged
269
proteins and the charged membrane is particularly
effective (van Reis et al., 1999). Here we can only deal
with some of the simpler aspects of water and solute
transport and leave many important aspects to the literature
references already provided.
We begin here by reviewing the basic problem of
predicting water transport in ideally selective membranes,
which is basically governed by boundary layer mass trans-
port (Bird et al., 2002; p. 713). This was demonstrated
(Kozinski and Lightfoot, 1972) using carefully purified
native bovine serum albumin, a protein impermeable mem-
brane, and a spinning disk system that produced a uni-
form boundary layer of known properties: protein rejection
during the filtration process produces a typical concen-
tration boundary layer in which back diffusion balances
convective protein transport toward the membrane. Numer-
ical integration of the boundary-layer equations taking into
account known concentration dependence of viscosity and
diffusivity enables prediction of protein concentration at
the upstream membrane surface and hence water activity
there. Water activity is normally expressed in terms of os-
motic pressure k. Water velocity out from the downstream
membrane surface is then given by
v ¼ KH ðDp  kÞ ð24Þ
where p is the drop in pressure across the membrane,
and KH is the membrane hydraulic permeability. Solution
of the boundary-layer equations gives an expression for this
water flux in terms of protein concentrations, the mass
transfer coefficient in the absence of appreciable protein,
and the appropriate equations of state. This expression can
be written for Kozinski’s system as (Lightfoot, 2003,
unpublished calculations):

o 


Uw Up0
v ¼ 1:39Q kp ln ð25Þ
Uw0 Up1
Here U is mass concentration, kU is the mass transfer
coefficient, and  is a correction for physical property
variation. The subscripts p, w, l, and 0 refer to protein,
water, surface conditions and bulk solution, respectively.
The superscript o refers to pure water.
The term in square brackets is a correction factor, and
that in rounded brackets is the classic film theory ex-
pression used in most ultrafiltration models. The coef-
ficient 1.39 is an empirical correction to the film theory
for flow-induced distortion of the protein concentration
profile. It is crude but surprisingly accurate: an asymptotic
correction of this magnitude is reached by the time polar-
ization has an appreciable effect on water flux. The prop-
erty correction
Q ¼ ðD 2rel =vrel Þ1=3 ð26Þ
may be species sensitive. Here is v kinematic viscosity,
and the subscript ‘‘rel’’ refers to the arithmetic mean of
interfacial and bulk values divided by the bulk value.
270 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004
It remains only to specify KH, and the concentration
dependence of osmotic pressure, diffusivity, and kinematic
viscosity (see Bird et al., 2002; Prob. 22C.1, p. 724
for examples).
Comparison of completely a priori predictions obtained
from this model with data in Figure 3, shows excellent
agreement. However, the generality of this result is still
unknown. Perkins et al. (Perkins, 1999) found similar good
agreement between data and a boundary-layer model and
poor predictions from the stagnant film model. However,
no extended study is available.
Most investigators use the stagnant film model because of
its simplicity, and the results of Kozinski and Lightfoot
(1972) suggest that this may be permissible for data cor-
relation, as opposed to a priori prediction. On balance, the
chief advantage of careful modeling may be as an adjunct to
early process development, to set the starting point for an
experimental program: the experimental data speak for
themselves. In any event there is increasing interest in
membranes that can select between proteins, and we return
to this point shortly.
First, however, we note that the next important ap-
plication of ultrafiltration was in solvent exchange where it
has been replacing size-exclusion chromatography (Kurnik
et al., 1995).
The most recent applications, which seem to be expanding
rapidly are in the discrimination between proteins of similar
size, a family of processes known as high performance
tangential flow filtration (Christy et al., 2002; van Reis and
Saksena, 1997; van Reis et al., 1999, van Reis et al., 1997a,
1997b, 1997c). As already indicated above this process
seems poised to compete with chromatography.
WHAT COMES NEXT?
We have done our best to summarize the present state of
bioseparations and to indicate current trends. It seems clear
from this discussion that what one might call the heroic age
of bioseparations is now over, and it has left us with a good
sense of strategy plus a very large amount of accumulated
information. We have probably identified most of the basic
Figure 3. Predicting ultrafiltration rates.
separations mechanisms and their salient characteristics.
However, there is still a large degree of empiricism in our
designs and few, if any, have been refined or optimized to
the degree familiar in more established industries. Some
obvious combinations such as countercurrent membrane
cascades have yet even to be recognized, and many other
combinations of existing mechanisms can also be expected.
Equipment and process refinement are already well
underway. They will include both more quantitative un-
derstanding of equipment performance and more effective
equipment and process designs. They will also include
more effective monitoring and control of processes. Under-
standing of basic equilibria and transport behavior at the
molecular level has already proven useful, and rapid ad-
vances can be expected. Economic analyses are becoming
more sophisticated, and they can be expected to reach
levels now standard in more mature industries. Computa-
tional programs will become more powerful (and realistic).
Competition does seem effective here.
Most important of all however, are the increased knowl-
edge and skills of those working in this area. The chemical
engineers involved know a lot more biology than their
mentors and are beginning to accept the complexity of
biological systems. They are less likely to be surprised by
lack of precision in biological measurements or the tricks
that Mother Nature always seems to have up her sleeves. The
authors are not really qualified to assess the skills of chemists
and biologists, but one of us (ENL) cannot resist a personal
comment: when I took a course in proteins and enzymes
from J.B. Sumner in the late 1940s these molecules were
characterized as either a cigar, a discus, or a balloon.
However, the 1950s were very rich in the development of
both fundamental understanding and techniques. Things
have improved! And they will continue to do so.
References
Abraham S, Bingham N, Green K, Kenworthy J. 2003. Strategies for
improving mammalian cell clarification. Abstr Pap Am Chem Soc
225:119 – BIOT.
Adams WR, Goetz V, Lee AL. 1999. Adsorption, protein, batch. In:
Flickinger MC, Drew SW, editors. Encyclopedia of bioprocess
technology. New York: Wiley.
Ahuja S. 2000. Handbook of bioseparations. San Diego, CA: Academic
Press.
Athalye AM, Gibbs SJ, Lightfoot EN. 1992. Predictability of chromato-
graphic protein separations—Study of size-exclusion media with
narrow particle-size distributions. J Chromatogr A 589:71 – 85.
Atkinson B, Mavituna F. 1991. Biochemical engineering and biotechnol-
ogy handbook. Basingstoke, UK: Stockton Press. p 873 – 904.
Aunins JG, Lee AL, Volkin DB. 2000. Vaccine production, section, XI,
part 105. In: Bronzino JD, editor. The biomedical engineering hand-
book, vol. II. Boca Raton, FL: CRC Press. p 1 – 16.
Axelsson H. 1999. Cell separation, centrifugation. In: Flickinger MC,
Drew SW, editors. Encyclopedia of bioprocess technology. New
York: Wiley.
Baddour RF, Goldstein DJ, Epstein P. 1954. Elution from partially
saturated columns. Ind Eng Chem 46:2192 – 2195.
Bailey JD, Ollis DF. 1986. Biochemical engineering fundamentals. New
York: McGraw-Hill.
LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS
Becker T, Lawlis VB. 1991. Subtilisin crystallization process. U.S. Patent
5,041,377.
Becker NT, Braunstein EL, Gros EH, Fewkes R, Heng MH. 1997.
Crystalline cellulase and method for producing same. World Patent
97/15660.
Belfort G, Davis RH, Zydney AL. 1994. The behavior of suspensions and
macromolecular solutions in cross-flow microfiltration. J Mem Sci
96:1 – 58.
Bird RB, Stewart WE, Lightfoot EN. 2002. Transport phenomena. New
York: Wiley.
Blackmore RS, Ryland JR. 2003. Tracking resin dynamic capacity in
preparative chromatography. Abstr Pap Am Chem Soc 225:71 – BIOT.
Boonaert CJP, Dupont-Gillain CC, Dengis PB, Dufrêne YF, Rouxhet PG.
1999. Cell separation, flocculation. In: Flickinger MC, Drew SW,
editors. Encyclopedia of bioprocess technology. New York: Wiley.
Boschetti E, Jungbauer A. 2002. Separation of antibodies by liquid
chromatography. In: Ahuja S, editor. Handbook of bioseparations. San
Diego, CA: Academic Press.
Caren R. 2003. Strategies for optimizing the robustness of viral filtration
steps in large-scale biopharmaceutical manufacturing processes. Abstr
Pap Am Chem Soc 225:118 – BIOT.
Chae YK, Jeon W, Cho KS. 2002. Rapid and simple method to prepare
functional pfu DNA polymerase expressed in Escherichia coli pe-
riplasm. J Microbiol Biotech 12:841 – 843.
Chang CE. 1988. Continuous fractionation of human plasma proteins by
precipitation from the suspension of the recycling stream. Biotechnol
Bioeng 31:841 – 846.
Christy C, Adams G, Kuriyel R, Bolton G, Seilly A. 2002. High-
performance tangential flow filtration: A highly selective membrane
separation process. Desalination 144:133 – 136.
Coffman JL, Roper DK, Lightfoot EN. 1994. High-resolution chromatog-
raphy of proteins in short columns and adsorptive membranes. Bio-
separation 4:183 – 200.
Cramer S. 2002. Special edition on magnetic field induced separations.
Sep Sci Technol 37:3611 – 3726.
Cramer S, Natarajan V. 1999. Chromatography, ion exchange. In:
Flickinger MC, Drew SW, editors. Encyclopedia of bioprocess
technology. New York: Wiley.
Curling J, Smiley L. 2003. Chromatography for plasmid DNA manufac-
ture—Tutorial: Novel adsorbent technology for cGMP production of
plasmid, DNA. Gen Eng News 23:42.
De-Bernardez-Clark E. 2001. Protein refolding for industrial processes.
Curr Opin Biotechnol 12:202 – 207.
De la Rosa APB, McPherson A. 1999. Crystallization, bulk, macro-
molecules. In: Flickinger MC, Drew SW, editors. Encyclopedia of
bioprocess technology. New York: Wiley.
DePhillips P, Lenhoff A. 2001. Determinants of protein retention charac-
teristics on cation-exchange adsorbents. J Chromatogr A 933:52 – 72.
Deshmukh RR, Leitch WE II, Sanghvi YS, Cole DL. 2000. Large-scale
chromatographic purification of oligonucleotides. In Ahuja S, editor.
Handbook of bioseparations. San Diego, CA: Academic Press.
Deshmukh RR, Warner TN. 2000. Adsorptive membranes for biosepara-
tions. In Ahuja S, editor. Handbook of bioseparation. San Diego, CA:
Academic Press.
Dziennik SR, Belcher EB, Barker GA, DeBergalis MJ, Fernandez SE,
Lenhoff AM. 2003. Nondiffusive mechanisms enhance protein up-
take rates in ion exchange particles. Proc Nat Acad Sci USA 100:
420 – 425.
Farkas T, Sepaniak MJ, Guiochon G. 1997. Radial distribution of the flow
velocity, the column efficiency, and the concentration profile at the
outlet of a preparative scale HPLC column. AIChE J 43:1964 – 1974.
Flickinger MC, Drew SW. 1999. Encyclopedia of bioprocess technology.
New York: Wiley.
Frej AKB. 1999. Adsorption, expanded bed. In: Flickinger MC, Drew SW,
editors. Encyclopedia of bioprocess technology. New York: Wiley.
Galletti PM, Colton CK, Lysaght ML. 2000. Artificial kidney, section,
XIII, part 130. In: Bronzino JD, editor. The biomedical engineering
handbook, vol. II. Boca Raton, FL: CRC Press. p 1 – 25.
271
Gebauer KH, Thoemmes J, Kula MR. 1997. Plasma fractionation with
advanced membrane adsorbents. Biotechnol Bioeng 54:181 – 189.
Guiochon G, Shirazi SG, Katti AM. 1994. Fundamentals of preparative
and non-linear chromatography. San Diego, CA: Academic Press.
Guiochon G, Farkas T, GuanSajonz H, Koh JH, Sarker M, Stanley BJ, Yun
T. 1997. Consolidation of particles and beds, and packing of chro-
matographic columns. J Chromatogr A 762:83 – 88.
Hahn R, Panzer M, Hansen E, Mollerup J, Jungbauer A. 2002. Mass
transfer properties of monoliths. Sep Sci Technol 37:1545 – 1565.
Hanley DD. 1999. Centrifuges, animal cells. In: Flickinger MC, Drew SW,
editors. Encyclopedia of bioprocess technology. New York: Wiley.
Harrison RG, Todd P, Rudge SR, Petrides DP. 2003. Bioseparations
science. Oxford, UK: Oxford University Press.
Harve R, Bajpai R. 2000. Separation of nucleic acids and proteins. In:
Ahuja S, editor. Handbook of bioseparations. San Diego, CA:
Academic Press.
Hearn MTW. 2000. Physicochemical factors in polypeptide and protein
purification, etc. In: Ahuja S, editor. Handbook of bioseparations. San
Diego, CA: Academic Press.
Heng M. 2003a. Method for rapidly obtaining crystals with desirable
morphologies. U.S. Patent 6,593,118.
Heng M. 2003b. Practical aspects of large-scale protein crystallization.
Abstr Pap Am Chem Soc 225:138 – BIOT.
Hjorth RA. 2003. Efficiency of expanded-bed adsorption processes. Abstr
Pap Am Chem Soc 225:140 – BIOT.
Ho WSW, Sirkar KK. 1992. Membrane handbook. New York: van
Nostrand Reinhold.
Ho CC, Zydney AL. 2002. Transmembrane pressure profiles during
constant flux microfiltration of bovine serum albumin. J Mem Sci
209:363 – 377.
ICH. 1999. Viral safety evaluation of biotechnology products derived from
cell lines of human or animal origins. Proceedings of the International
conference on harmonization of technical requirements for registration
of pharmaceuticals for human use.
Jeanesonne MS, Foley JP. 1992. Improved equations for the calculation of
chromatographic figures of merit for ideal and skewed chromato-
graphic peaks. J Chromatogr A 594:1 – 8.
Johns MR. 1999. Crystallization, proteins, kinetics. In: Flickinger MC,
Drew SW, editors. Encyclopedia of bioprocess technology. New
York: Wiley.
Jungbauer A, Kaltenbrunner O. 1999. Chromatography, computer aided
design. In: Flickinger MC, Drew SW, editors. Encyclopedia of bio-
process technology. New York: Wiley.
Jungbauer A, Zoechling A, Hahn R. 2003. Chromatography of ultra large
proteins and nanoparticles using monoliths. Abstr Pap Am Chem Soc
225:89 – BIOT.
Kaarsnaes P. 1999. Chromatography, hydrophobic interaction. In: Flick-
inger MC, Drew SW, editors. Encyclopedia of bioprocess technology.
New York: Wiley.
Kaltenbrunner O, Stokelman A. 2003. Qualification of large scale
chromatographic systems. Abstr Pap Am Chem Soc 225:95 – BIOT.
Kalyanpur M. 1999. Membrane separations. In: Flickinger MC, Drew SW,
editors. Encyclopedia of bioprocess technology. New York: Wiley.
Klinkenberg A, Sjenitzer F. 1956. Holding-time distributions of the
Gaussian type. Chem Eng Sci 5:258.
Khosla C, Caren R, Kao CM, McDaniel R, Wang SW. 1996. Evolutionary
guided enzyme design. Biotechnol Bioeng 52:122 – 128.
Koehler JA, Ulbricht M, Belfort G. 2000. Intermolecular forces between a
protein and a hydrophilic modified polysulfone film with relevance to
filtration. Langmuir 16:10419 – 10427.
Kozinski AA, Lightfoot EN. 1972. Protein ultra-filtration. AIChE J 18:
1030 – 1040.
Kroschwitz JI. 1991. Encyclopedia of chemical technology. New
York: Wiley.
Kurnik RT, Yu AW, Blank GS, Burton AR, Smith D, Athalye AM, Van Reis
R. 1995. Buffer exchange using size exclusion chromatography, coun-
tercurrent dialysis and tangential flow filtration: models, development
and industrial application. Biotechnol Bioeng 45:149 – 157.
272 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 87, NO. 3, AUGUST 5, 2004
Larson TM, Gawlizek M, Evans H, Albers U, Cacia J. 2002. Chemometric
evaluation of on-line high-pressure liquid chromatography in mam-
malian cell cultures: Analysis of amino acids and glucose. Biotechnol
Bioeng 77:553 – 563.
Larson TM, Davis J, Lam H. Carcia J. 2003. Use of process data to assess
chromatographic performance in production-scale protein purification
columns. Biotech Progr 19:485 – 492.
Lehninger AL. 1971. Bioenergetics. San Francisco: Benjamin/Cummings.
Lewus RK, Hakan AF, Carta G. 1998. Protein adsorption and desorption
on gel-filled rigid particles for ion exchange. Ind Eng Chem 37:
1079 – 1087.
Lightfoot EN. 1999. The invention and development of process chro-
matography: interaction of mass transfer and fluid mechanics. Amer
Lab 32:13 – 23.
Lightfoot EN, Cockrem MCM. 1987. What are dilute solutions? Sep Sci
Technol 22:165 – 189.
Lightfoot EN, Roper KD, Moscariello JS. 2003a. How can you tell if a
column is doing its job? Abstr Pap Am Chem Soc 225:64 – BIOT.
Lightfoot EN, Moscariello JS, Teeters MA, Root TW. 2003b. Interaction
of mass transfer and fluid mechanics. In Rathore AS, Velayudhan A,
editors. Scale-up and optimization in preparative chromatography.
New York: Marcel Dekker. p 33 – 75.
Lydersen BK, D’elia NA, Nelson KL. 1994. Bioprocess engineering. New
York: Wiley.
Lysaght MJ, Morgan J. 2000. Peritoneal dialysis equipment, section, XIII,
part 131. In: Bronzino JD, editor. The biomedical engineering hand-
book. Boca Raton, FL: CRC Press.
Martin AJP, Synge RLM. 1941. A new form of chromatogram employing
two liquid phases. 1. A theory of chromatography. 2. Application to
the micro-determination of the higher monoamino acids in proteins.
J Biochem 35:1358 – 1368.
Mazza CB, Sukumar N, Breneman CM, Cramer SM. 2001. Prediction of
protein retention in ion-exchange systems using molecular descriptors
obtained from crystal structure. Anal Chem 73:5457 – 5461.
McCue JT, Kemp G, Low D, Quinones-Garcia I. 2003. Evaluation of
protein-A chromatography media. J Chromatogr A 989:139 – 153.
McGuire J, Bothwell MK. 1999. Adsorption, proteins with synthetic
materials. In: Flickinger MC, Drew SW, editors. Encyclopedia of
bioprocess technology. New York: Wiley.
McIntosh RV, Griffin B, Leslie F, Foster PR. 1994. Bio-process scale-up
in the manufacture of factor VIII concentration from human plasma.
Second Conference on Advances in Biochemical Engineering.
Brighton, UK: Institute of Chemical Engineers.
McPherson A. 1999. Crystallization of biological macromolecules. Cold
Spring Harbor, NY: Cold Spring Harbor Laboratory Press.
Middleberg AOJ. 2002. Preparative protein refolding. Trends in Bio-
technology 20:437 – 443.
Moscariello JS, Purdom G, Coffman JL, Root TW, Lightfoot EN. 2001.
Characterizing the performance of industrial-scale columns. J Chro-
matogr A 908:131 – 141.
Mustafa MA, Washbrook J, Farid S, Lim AC, Titchner-Hooker N. 2003.
Novel software tools for evaluating integration within a downstream
process. Abstr Pap Am Chem Soc 225:96 – BIOT.
Nagrath D, Bequette BW, Cramer SM. 2003. Evolutionary operation and
control of chromatographic processes. AICHE J. 49:82 – 95.
Nicoud RM. 2000. Simulated moving bed chromatography for biomole-
cules. In Ahuja S, editor. Handbook of bioseparations. San Diego, CA:
Academic Press.
Noble RD, Stern SA. 1995. Membrane science and technology. Am-
sterdam, The Netherlands: Elsevier.
O’Keefe DO. 2000. Analysis of protein impurities in pharmaceuticals
derived from recombinant DNA. In Ahuja S, editor. Handbook of
bioseparations. San Diego, CA: Academic Press.
Palmgren RJ, van Alstine J, Kjellgren A, Wulff M. 2003. Stimulus
response polymers used in chromatographic separation. Abstr Pap Am
Chem Soc 225:90 – BIOT.
Palsson BO, Bhatia SN. 2003. Tissue engineering. Englewood Cliffs, NJ:
Prentice Hall.
Perkins TW, Saksena S, van Reis R. 1999. A dynamic film model for
ultrafiltration. J Mem Sci 158:243 – 256.
Phillips MW. 2003. Membrane adsorber technology for trace impurity
removal. Abstr Pap Am Chem Soc 225:116 – BIOT.
Pieracci J, Schiffner C, Thoemmes J, Conley LE. 2003. Membrane
adsorber process optimization through design of experiments. Abstr
Pap Am Chem Soc 225:117 – BIOT.
Prior J, Menon M, Dugopolski C, Legg K. 2003. Monitoring, diagnosis and
trouble-shooting of large scale chromatography. Abstr Pap Am Chem
Soc 225:65 – BIOT.
Rathore AS, Velayudhan A. 2003. Scale-up and optimization in
preparative Chromatography. New York: Marcel Dekker.
Reis JFG, Lightfoot EN, Noble PT, Chiang AS. 1979. Chromatography in
a bed of spheres. Sep Sci Technol 14:367 – 394.
Rhee HK, Amundson NR. 1982. Analysis of multicomponent separation
by displacement development. AIChE J 28:423 – 433.
Roper DK, Lightfoot EN. 1994. Mean residence time and variance in
steady counterflow separations. Chem Eng Sci 49:1621 – 1630.
Roper DK, Lightfoot EN. 1993. Comparing steady counterflow with dif-
ferential chromatography. J Chromatogr A 654:1 – 16.
Roth CM, Yarmush MY. 1999. Chromatography, size exclusion. In:
Flickinger MC, Drew SW, editors. Encyclopedia of bioprocess
technology. New York: Wiley.
Sarkar S, Carr PW, McNeff CV, Subramanian A. 2003. Characterization
and optimization of a chromatographic process based on ethylenedi-
amine-N,N,NV,NV-tetra(methylphosphonic) acid-modified zirconia
particles. J Chromatogr B-Anal Tech Biomed Life Sci 790:143 – 152.
Schneider P, Smith JM. 1968. Adsorptive rate constants from chromatog-
raphy. AIChE J 14:762 – 771.
Schutyser M, Rupp R, Wideman J, Belfort G. 2002. Dean vortex mem-
brane microfiltration and diafiltration of rBDNF, E. coli inclusion
bodies. Biotech Prog 18:322 – 329.
Schutyser M, Belfort G. 2002. Dean vortex membrane filtration non-
Newtonian viscosity effects. IEC Res 41:494 – 504.
Shalliker RA, Wong V, Broyles BS, Guiochon G. 2002. Visualization of
bed compression in an axial compression liquid chromatography
column. J Chromatogr A 977:213 – 223.
Shapiro HM. 2003. Practical flow cytometry. New York: Wiley.
Sharma BP. 1999. Cell separation, sedimentation. In: Flickinger MC,
Drew SW, editors. Encyclopedia of bioprocess technology. New
York: Wiley.
Shpritzer R. 2003. Evaluation of a continuous disc stack centrifuge for the
clarification of mammalian cell cultures. Abstr Pap Am Chem Soc
225:87 – BIOT.
Shukla A, Cramer SM. 2000. Bioseparations by displacement chromatog-
raphy. In: Ahuja S, editor. Handbook of bioseparations. San Diego,
CA: Academic Press.
Shuler ML. 2001. Bioprocess engineering: Basic concepts. Englewood
Cliffs, NJ: Prentice-Hall.
Snyder LR, Kirkland JJ, Glajch JL. 1997. Practical HPLC method de-
velopment. New York: Wiley-Interscience.
Sonnenfeld A, Forrester H, Thoemmes J. 2003. Packing quality and in-
tegrity of industrial chromatography columns. Abstr Pap Am Chem
Soc 225:66 – BIOT.
Staby A, Jensen IH, Bruus U, Sand MB, Hansen RG, Jacobsen JH. 2003.
Comparison of strong cation-exchange resins for [protein purification.
Abstr Pap Am Chem Soc 225:136 – BIOT.
Stahl S, Nilsson J, Hober S, Uhlen M, Nygren PA. 1999. Affinity fusions,
gene expression. In: Flickinger MC, Drew SW, editors. Encyclopedia
of bioprocess technology. New York: Wiley.
Stryker MH, Bertolini MJ, Lee HY. 1985. Blood fractionation: Proteins.
Adv Biotech Proc 4:275 – 336.
Taggart TS, Dripps DJ, Kessler T, Cameron ME, Seely JE, Todd RJ. 2003.
LIGHTFOOT AND MOSCARIELLO: BIOSEPARATIONS
Resin reuse validation and evaluation of performance over resin
lifetime. Abstr Pap Am Chem Soc 225:69 – BIOT.
Tebbe H, Lutkemeyer D, Gudermann F, Heidemann R, Lehmann J. 1996.
Lysis-free separation of hybridoma cells by continuous disc stack
centrifugation. Cytotechnology 22:119 – 127.
Teeters MA, Root TW, Lightfoot EN. 2002. Performance and scale-
up of adsorptive membrane chromatography. J Chromatogr A 944:
129 – 139.
Teeters MA, Conrardy SE, Thomas BL, Root TW, Lightfoot EN. 2003.
Adsorptive membrane chromatography for purification of plasmid,
DNA. J Chromatogr A 989:165 – 173.
Thoemmes J. 1999. Membrane chromatography. In: Flickinger MC,
Drew SW, editors. Encyclopedia of bioprocess technology. New
York: Wiley.
Tibbs TL, Fernandez EJ. 2003. Alpha-lactalbumin tertiary structure
changes on hydrophilic interaction chromatography surfaces. J Coll
Interf Sci 259:27 – 35.
van Reis R, Saksena S. 1997. Optimization diagram for membrane
separations. J Mem Sci 129:19 – 29.
van Reis R, Zydney AL. 1999. Protein ultrafiltration. In: Flickinger MC,
Drew SW, editors. Encyclopedia of Bioprocess Technology. New
York: Wiley.
van Reis R, Leonard LC, Hsu CC, Builder SE. 1991. Industrial scale
harvest of proteins from mammalian cell culture by tangential flow
filtration. Biotechnol Bioeng 38:413 – 422.
van Reis R, Brake JM, Charkoudian J, Burns DB, Zydney AL. 1999. High-
performance tangential flow filtration using charged membranes. J
Mem Sci 159:133 – 142.
van Reis R, Goodrich EM, Yson CL, Frautschy LN, Dzengeleski S,
Lutz H. 1997a. Linear scale ultra-filtration. Biotechnol Bioeng 55:
737 – 746.
van Reis R, Goodrich EM, Yson CL, Frautschy LN, Whiteley R, Zydney
AL. 1997b. Constant wall concentration ultrafiltration process control.
J Mem Sci 130:123 – 140.
van Reis R, Gadam S, Frautschy LN, Orlando S, Goodrich EM, Saksena S,
Kuriyel R, Simpson CM et al. 1997c. High performance tangential
flow filtration. Biotechnol Bioeng 56:71 – 82.
Wen EP, D’Silva LD, Murray D, Sagar SL, Lee A. 2003. Purification
of a polysaccharide conjugate vaccine using (a) microfiltration
membrane in backpulsing mode. Abstr Pap Am Chem Soc 225:
115 – BIOT.
Wheelwright SM. 1991. Protein purification. Oxford, UK: Hanser.
Williams A, Taylor K, Dambuleff K, Persson O, Kennedy RM. 2002.
Maintenance of column performance at scale. J Chromatogr A 944:
69 – 75.
Yamamoto S, Suehisa T, Sano Y. 1993. Preparative separation of proteins
by gradient- and stepwise-elution chromatography: Zone sharpening
effect. Chem Eng Comm 119:221 – 230.
Yamamoto S. 1995. Plate height determination for gradient elution
chromatography of proteins. Biotechnol Bioeng 48:444 – 451.
Yew BG, Ureta J, Shalliker RA, Drumm EC, Guiochon G. 2003.
Mechanics of column beds: II. Modeling of coupled stress-strain-flow
behavior. AIChE J 49:642 – 664.
Yuan QS, Rosenfeld A, Root TW, Klingenberg DJ, Lightfoot EN. 1999.
Flow distribution in chromatographic columns. J Chromatogr A.
831:149 – 165.
Zborowski M, More LR, Williams PS, Chalmers JJ. 2002. Separations
based on magnetophoretic mobility. Sep Sci Technol 37:3611 – 3633.
Zeman LJ, Zydney AL. 1996. Microfiltration and ultrafiltration. New
York: Marcel Dekker.
Zydney AL. 2000. Therapeutic apheresis and blood fractionation, section
XIII, part 132. In: Bronzino JD, editor. The biomedical engineering
handbook, vol. II. Boca Raton, FL: CRC Press. p 1 – 15.
273