Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 1 of 28 PageID #:
`9480
`Biotechnology and Genetic Engineering Reviews
`
`ISSN: 0264-8725 (Print) 2046-5556 (Online) Journal homepage: http://www.tandfonline.com/loi/tbgr20
`
`Industrial Purification of Pharmaceutical
`Antibodies: Development, Operation, and
`Validation of Chromatography Processes
`
`Robert L. Fahrner , Heather L. Knudsen , Carol D. Basey , Walter Galan , Dian
`Feuerhelm , Martin Vanderlaan & Gregory S. Blank
`
`To cite this article: Robert L. Fahrner , Heather L. Knudsen , Carol D. Basey , Walter
`Galan , Dian Feuerhelm , Martin Vanderlaan & Gregory S. Blank (2001) Industrial
`Purification of Pharmaceutical Antibodies: Development, Operation, and Validation of
`Chromatography Processes, Biotechnology and Genetic Engineering Reviews, 18:1, 301-327,
`DOI: 10.1080/02648725.2001.10648017
`
`To link to this article: http://dx.doi.org/10.1080/02648725.2001.10648017
`
`Published online: 15 Apr 2013.
`
`Submit your article to this journal
`
`Article views: 1774
`
`View related articles
`
`Citing articles: 2 View citing articles
`
`Full Terms & Conditions of access and use can be found at
`http://www.tandfonline.com/action/journalInformation?journalCode=tbgr20
`
`Download by: [64.124.209.76]
`
`Date: 04 November 2015, At: 12:21
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 2 of 28 PageID #:
`9481
`
`12
`Industrial Purification of Pharmaceutical
`Antibodies: Development, Operation, and
`Validation of Chromatography Processes
`
`ROBERT L. FAHRNER1·, HEATHER L. KNUDSENl, CAROL D. BASEyl,
`WALTER GALAN I, DIAN FEUERHELMI, MARTIN VANDERLAAN2 AND
`GREGORY S. BLANK]
`
`IDepartlnent ofRecovery Sciences and 2Department ofAnalytical Chemistry,
`Gellelltech, 1Ilc~, 1 DNA WaYJ South San Francisco, CA 94080, U.S.A.
`
`Introduction
`
`Recombinant monoclonal antibodies are becoming a great success for the biotech(cid:173)
`nology industry. They are currently being studied in many clinical trials for treating
`a variety of diseases, and recently several have been approved for treating cancer
`(Carter et al., 1992; Anderson et al., 1996; Baselga et al~, 1996; Bodey et al., 1996;
`Longo~ 1996). Although there are several types of antibodies produced in different
`types ofcel1lines, the most clinically significant antibodies are full-length humanized
`IgG. produced in CHO cells. This review describes the methods used to purify these
`antibodies at industrial scale, focusing on chromatography processes~ and with
`particular reference to recent work at Genentech.
`Routine laboratory purification ofantibodies has been well described (for example
`see Scott et aL, 1987), but the considerations for large-scale production of pharma(cid:173)
`ceutical-grade antibodies are much different than those for laboratory scale. There are
`extreme purity requirements for pharmaceutical antibodies~ and routine large-scale
`production requires high yield and process reliability. To gain regulatory approval,
`the process must be completely validated to run consistently within specified limits,
`so the process should be designed to facilitate validation,
`Large-scale production of antibodies as pharmaceutical products is a complex
`
`*To whom correspondence nlay be addressed (fahmer.roberl@gene.com)
`
`Abbreviations: CV, column volume; HCCF,. harvested cell culture fluid; CHOP.. Chinese hamster ovary
`proteins; CHO, Chinese hamsterovary; ELISA. enzyme-linked immunosorbent assay; 8DS-PAGE, sodiunl
`dodecyl sulphate polyacrylamide gel electrophoresis; BSA, bovine serum albumin; eEt capillary electro(cid:173)
`phoresis; HPLC. high-perfonnance liquid chromatography; ppm, parts per million (nglmg); LOQ, limit of
`quantitation; SEC, size exclusion chromatography; pI. isoelectric point; GMP,. good manufacluring
`practice; gIl, when describing column loads this is grams of antibody per litre of column volume.
`
`BiotecJuzO[(lgj' (Inti Gfmetic E"gineering RC1';eW$ - Vol. 18. July 2001
`0264-8725101118/301-327 $20.00 + $0.00 © Intercept Ltd~ P.O. Box 716. Andover, Hampshire SPIO lYG. U.K.
`301
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 3 of 28 PageID #:
`302 R.L. FAHRNER et al.
`9482
`endeavour, including a manufacturing process with multiple steps and significant
`analytical SUppOIt. Antibody Inanufacturing includes cell banking and cell culture,
`recovery, filling (possibly including lyophilization)t finishing t and packaging. Product
`recovery includes harvest~ which is removal ofcells and cell debris by tangential flow
`filtration or centrifugation (van Reis et al., 1991), chromatography for antibody
`purification~ and formulation by tangential flow filtration. Here we focus on process
`chromatography, which must reliably produce highly purified antibody.
`To satisfy the stringent purity requirements for phal1~naceutical antibodies t an
`extensive analytical control system is integrated with the nlanufacturing process at all
`stepSt particularly on release of the final product. The analytical control system
`includes assays for product-related variants (including charge and glycosylation
`variants), often using ion exchange HPLC or CE (Hunt et al., 1996; Hunt and
`Nashabeh, 1999), but these variants are typically controlled during cell culture and are
`not removed during chromatography. To ensure that no variants are formed during
`purification, antibody stability is controlled during chromatography by limiting
`extrenles of pH, temperature, and other process variables to reduce the amount of
`oxidation, deamidation, aggregation, and other variant-formation routes.
`Many phannaceutical proteins require a significant clearance of product-related
`variants. An example of this is insulin-like growth factor, where several product(cid:173)
`related variants (such as a single amino acid oxidation and clipped forms) are
`removed to <1 % during purification (Fahrner et a/. t 1998, 1999b). The acceptable
`level of product..related variants is an issue which dates to the first proteins produced
`by recombinant DNA technology. The resolution and sensitivity of current analytical
`technology permits the definition of very minor differences among the product
`protein population. The fact that variants can be discovered does not autoJnatically
`indicate that they need to be removed or even controlled. For example, the DNA
`sequence for IgG J antibodies codes for a lysine at the C-terminus of each heavy chain.
`DUling cell culture, one or both of these lysines are usually removed, leading to three
`charged populations (zero, one~ or two lysines). This variability has no impact on the
`ability of the antibody to bind its target antigen or effect any biological activity.
`Therefore,
`the product definition would allow for all
`three species. The same
`approach can be extended to other product variants. It is necessary to characterize the
`Inolecular source of the variation and demonstrate that the variation has no effect on
`potency or safety.
`Froln a recovery standpoint, one of the most significant advantages to using
`antibodies produced in CHO cells is that the level of product-related variants can be
`effectively controlled during cell culture so that little or no variants must be removed
`during recovery. This level of control during cell culture allows the use of a
`streamlined, three-step recovery process. Instead of focusing on the removal of
`product-related variants~ the process is concerned with the clearance of pharma~
`ceutical impurities such as virus, DNA, host cell proteins~ endotoxin, and small
`Inolecules. This recovery process consists of protein A affinity chromatography,
`cation exchange chromatography, and anion exchange chroJnatography.
`
`Antibody recovery
`
`Since no single chromatography step can achieve the necessary antibody purity, the
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 4 of 28 PageID #:
`9483
`Industrial purification ofpharlnaceutical antibodies
`303
`
`three process steps DIUSt be integrated to meet the requirements of puritYt yield, and
`throughput. In addition, the process must be robust, reliable, and amenable to
`validation.
`The primary consideration is purity. While yield and throughput may be necessary
`for an economically viable product, without meeting the purity requiretnents for
`biological pharmaceuticals there will be no product at all. Throughput and yield are
`becoming more important as many clinical indications for antibodies require very
`high doses. At our manufacturing plant, we typically use processes that purify a 5-10
`kg antibody batch in less than three days with greater than 65% overall process yield.
`
`PURITY CONSIDERATIONS
`
`Although pharmaceutical antibodies do not require the removal of product-related
`variants that complicate the purification of some proteins, other purity requirements
`are extreme. There are six main purity considerations for the recovery of pharma(cid:173)
`ceutical antibodies.
`
`1. Host cell proteins
`
`Host cell proteins are present in high amounts (sometimes >1,000;000 ng/mg) in the
`harvested cell culture fluid. They are typically removed during purification <5 ppm,
`a total reduction of at least lOs, In our studies, the level of host cell proteins was
`lneasured quantitatively by ELISA (Chen, 1996) and qualitatively by SDS-PAGE.
`For the ELISA, affinity purified goat anti-CHOP antibodies were immobilized on
`microtitre plate wells. Dilutions of the pool samples were incubated in the wens,
`followed by an incubation with peroxidase-conjugated goat anti-CHOP. The horse(cid:173)
`radish peroxidase enzymatic activity was quantified with o-phenylenediamine.
`Samples were serially diluted 2...fold in assay diluent so that the absorbance reading
`fen within the range of the standard curve (1.5 ng/ml to 400 ng/nl1).
`To analyse the antibody by SDS-PAGE, the pool samples were run under reducing
`and non-reducing conditions on one-dimensional Novex 8-16% Tris-glycine gels.
`Samples were loaded at 2~5 J.1gllane for non..reducing conditions and 5.0 J,lg/lane for
`reducing conditions. The gels were silver..stained using the Novex silver express kit.
`The samples were compared to a reference standard for identification of product
`related bands.
`
`2. DNA
`
`The World Health Organization set a requirement for DNA in biopharmaceutical
`formulations of <10 ng/dose. DNA is present at high levels in the harvested cell
`culture fluid (> I ,000,000 pg/mg) and must be removed to <10 ng/dose levels. During
`validation studies DNA may be spiked into the load to demonstrate clearance.
`In our studies, the level of DNA was measured using the Molecular Devices
`Threshold DNA assay kit. The typical range of detection of the Threshold Total DNA
`assay was between 6.3 and 400 pg/ml. Samples were assayed at a mininlum of 3
`dilutions with and without a 100 pg spike of DNA~ This procedure was used to
`evaluate DNA recovery because sonle buffers, impurities and proteins are known to
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 5 of 28 PageID #:
`9484
`304 R.L. FAHRNER etal.
`
`inhibit the detection of DNA and inhibit spike recovery. The mean value for all
`sample dilutions falling within the range of the standard curve and meeting spike
`recovery acceptance criteria was used.
`
`3. Aggregate
`
`The main product-related variant that must be reduced is aggregated fonns of the
`antibody (mostly dimer) because of the possible immunogenicity of the aggregate.
`The aggregate content in the HCCF is about 5-15% for many antibodies, and it is
`typically reduced to below 0.5% in the final bulk. The primary step used to remove
`aggregate is cation exchange chromatography.
`In ourstudies, aggregate was measuredby size-exclusion chromatography. A BioSil
`SEC-250 7.5 x 300 mm column from BioRad was run at 1mllmin using a mobile phase
`containing 50 rnM NaH2P04/50 mMNa2HP04/0.15 MNaCI~pH 6.8. ThecoJumn was
`equilibrated with the mobile phase buffer and 20 f.lI volumes ofblank, standard, control
`and study samples were sequentially injected and run on the SEC for analysis~
`
`4. Small molecules
`
`The harvested cell culture fluid contains many small molecules, originating from the
`media components and created during cell culture by the CHO cells. Rather than
`determining the level of all small molecules t a few representative marker molecules
`are measured. Here we present the results from measurements ofinsulin and Pluronic
`F-68.
`The level of Pluronic F-68 was measured using a 500 MHz NMR. NMR detects
`hydrogen...containing molecules based on magnetic moments. Pluronic has a charac(cid:173)
`teristic peak in the spectrum with a chemical shift of 1.1 ppm, which was used for
`quantification. Peak areas in samples were compared with the standards. The Pluronic
`F-68 standard curve was run in process buffers, and covered the range of 25 Jlg/ml to
`l024lJglrnl. As controls, the conditioned protein A pool was analysed unspiked and
`spiked with 25 Jlg/ml Pluronic F-68.
`The level of insulin in the pool samples was determined by a competition ELISA~
`The monoclonal antibody to insulin was immobilized on microtitre plate wells.
`Diluted samples and biotinylated insulin were placed in the antibody immobilized
`wells. The insulin and biotinylated-insulin compete for binding to the antibody. The
`amount of bound biotinylated-insulin was detected with alkaline phosphatase(cid:173)
`streptavidin and p-nitrophenyl phosphate substrate. All samples were assayed in
`wells coated with non-immune mouse antibody in place of the specific monoclonal
`antibody. This control showed that binding to the plate is mediated by the specific
`monoclonal antibody and not by a non-specific interaction.
`
`5. Leached protein A
`
`During protein A affinity chromatography, some protein A leaches fronl the column
`and ends up in the antibody pool. Because protein A can be immunogenic and cause
`other physiological reactions (Gagnon~ 1996), leached protein A must be cleared
`during downstream chromatography.
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 6 of 28 PageID #:
`9485
`Industrial purification ofpharnlaceutical antibodie..Jt
`305
`
`The level ofprotein A in our samples was detennined by a sandwich ELISA (Lucas
`et al., 1988). Chicken anti-protein A antibody was immobilized on microtitre wells;
`Protein A binds to the coat antibody. The amount of bound protein A was detected
`with chicken anti-protein A labelled with biotin, followed by streptavidin-HRP and
`then the substrate o-phenylenediamine dihydrochloride and hydrogen peroxide. The
`reaction was stopped by adding sulphuric acid. The product was quantified by reading
`an absorbance at 490 nm. All samples were initially diluted to 0.2 mg/ml antibody in
`assay diluent. Samples were then serially diluted 2-fold with sample/standard diluent
`which contained 0..2 J.lg/ml antibody. Samples were assayed as a dilution series to
`ensure that antibody excess was reached. Values were calculated as the average of all
`results within the reporting range (O.78~25 ng/ml).
`
`6. Virus
`
`Harvested cell culture fluid may have 104 or more retrovirus-like particles per ml, and
`biological pharmaceuticals are allowed to have 1 theoretical virus particle per 106
`doses, so the recovery process must provide significant virus clearance. The valida(cid:173)
`tion and test procedures for viral clearance are complicated and are beyond the scope
`ofthis chapter. However, the process is capable of clearing virus to acceptable levels.
`In general, the protein A affinity chromatography step provides 107 ('7 logs') of virus
`clearance (104 by removal and 103by low-pH inactivation in the elution pool), and the
`anion exchange chromatography step provides 104 (4 logs) of viral clearance by
`removal. If this level of viral clearance is not sufficient, additional process steps such
`as viral filtration may be required.
`
`Purity calculations
`
`For all quantitative assays, the level of impurity in the sample is calculated by
`multiplying the measured value by the sample dilution. Since samples may be diluted
`to differing extents to avoid matrix interference, the absolute sensitivity (LOQ) of the
`assay will be influenced by the required sample dilution. Because values are often
`reported in ppm or ng of impurity per mg of product (not ng/ml)~ the reported
`sensitivity will also depend on the product concentration in the sample.
`
`THREE-STEP RECOVERY PROCESS
`
`The purity., yield, and throughput requirements can be achieved using three chromato(cid:173)
`graphy steps: protein A affinity chromatography, followed by cation exchange
`chromatography, followed by anion exchange chromatography. Protein A and cation
`exchange chromatography are run in bind-and-elute modes, while the anion exchange
`chromatography is run in flow-through mode (for antibodies with pI greater than
`about 8). Running in these modes in this order produces a high-yield process capable
`of meeting the purity requirements (Table 12.1).
`We present methods that may be applied to many antibodies, but it is important to
`note that some antibodies may have specific considerations, such as susceptibility to
`aggregation, oxidation, deamidation, or other stability problems. In these cases,
`adjustments to the process may have to be made 4 For example, an antibody that is
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 7 of 28 PageID #:
`9486
`306
`R.L. FAHRNER et al.
`Table 12.1. Typical yield and purity values for the three-step antibody recovery process
`
`Yield
`(%)
`
`Host cell proteins
`(ngllng)
`
`DNA
`(pg/ntg)
`
`Endotoxin
`(EU/mg)
`
`Protein A
`(oglIng)
`
`Aggregate
`(%)
`
`HCCF
`Protein A
`Cation
`Anion
`
`>95
`75-90
`>95
`
`250,000-1,000,000
`200-3000
`25-150
`<5
`
`100,000-1,500,000 5-100
`10D-1000
`<0.005
`<0.005
`<10
`<0.005
`<10
`
`3-35
`<2
`<2
`
`5-15
`5-15
`<0.5
`<0.5
`
`highly prone to deamidation nlay require a limit on its exposure to high pH (>8) during
`recovery. An important part ofprocess development is determining the stability of the
`antibody, since product stability will strongly influence the specific parameters used
`during recovery.
`The tirst step in the process is protein A affinity chromatography (Figure 12.1). The
`majority of the purification occurs during protein A affinity chromatography (Table
`12.1), which clears host cell proteins, DNA, and endotoxin. In addition, it removes
`insulin and Pluronic F w 68 to less than detectable levels. However, it does not clear
`aggregate, and it adds protein A into the pool..
`Protein A is a bacterial cell wall protein that binds specifically to antibodies t and it
`binds particularly well to human IgO l
`. When immobilized onto chromatography
`
`3.5
`
`10 200
`
`.. r -
`t i ,
`iii
`~ b
`~
`~I
`L-l ~I
`
`I
`
`1 1. .:
`:
`1 L...
`
`........ :
`
`.
`.
`.--.__..'..
`
`•
`I
`~
`t
`
`I
`
`I
`I
`~
`•
`
`\ . . . . .
`
`o
`
`o
`
`0
`
`..-_-
`l
`LL-._.
`o
`
`.
`
`_--
`
`------=---
`
`400
`
`._----1----_
`
`800
`
`1200
`
`Volume (ml)
`
`Figure 12.1. Chromatogram fronl a typical protein A affinity chroDlatography run. A 1.0 em inner
`diameter x 20 Clll length column was packed with Prosep Achromatography media. Four buffers were used.
`Buffer A was 25 n1M Tris. 25 tuM Nae), 5 mM EDTA, pH 7.1; buffer B was 25 mM Tris, 25 DIM Nnel,
`5 nlM EDTA. 0.5 M tetranlethyJamnlonium chloride pH 7.0; buffer C was 0.1 M acetic acid, pH 3.5; and
`buffer D was 2 Mguanidine HCI. 10 "1M Tris, pH 7.5. The colunln was equilibrated with Scolumn volumes
`of buffer A, loaded to 20 gil. washed with 3 colunln volunles of buffer A, \vashed with 3 column voluJues
`buffer B, washed with 3 colUJnn volumes of buffer A, eluted with 5 column volumes of buffer C, and
`regenerated \vith 3 column volumes of buffer D. The column was run at 550 em/h.
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 8 of 28 PageID #:
`9487
`Industrial purification ofpharmaceutical antibodies
`307
`
`3
`
`14
`
`110
`
`-
`
`I
`
`;
`
`•
`
`1-...-::...'
`.
`:
`I···
`,
`.
`I:
`t
`·
`~
`I!
`:
`·
`.
`a:
`:
`f
`:
`!
`i
`i
`~
`i
`.. ~j
`t ,
`•
`
`J
`
`\
`
`.....
`
`-_ - - , ...,...
`
`_ -
`- -
`
`t
`
`•
`
`1I\ I,
`
`I
`
`rs
`
`, t"
`
`'..... ~
`.,.J - - -\
`: :
`I
`i. .
`I
`:
`l.
`I••.j
`
`;::>
`$.
`
`e=0
`-~
`
`co
`N
`
`Q)
`u
`c::
`l'U
`..c
`loot
`0
`.!:J
`-<
`
`~
`
`m
`!.
`.€
`~
`u
`:=
`c:
`0
`U
`
`~
`
`:r:
`
`0.
`
`o
`
`0
`
`0
`
`o
`
`.... - • • - • • • • • ""
`
`o.
`
`-
`
`- . - _ • • _ .
`
`100
`
`200
`
`300
`
`400
`
`Time (min)
`
`Figure 12.2. Chronlatogram from a typical cation exchange chromatography run. The column was 0.66
`em inner dianleter x 20 em length t packed with Poros 50 HS. The column was washed with 2 CV of 0.016
`M MES/O.OO4 M NaMES/O.SOO M NaCI. pH 5.5, then equilibrated with S CV of 0.01 6 M MES/O.004 M
`NaMES/O.060 M NaCI, pH 5.5~ loaded to 40 gil. washed with 5 CV of 0.016 M MES/O.004 M NaMESI
`0.060 M NaCl, pH 5.5, eluted with 5 CV of 0,016 M MES/O.004 M NaMES/O.160 M NaCI, pH 5.5,
`regenerated with 2 CV of 0.016 M MES/O,004 M NaMES/O.SOD M NaCl, pH 5.5, sanitized with 2 CV of
`0.5 N NaOH. and stored in 3 CV ofO.t N NaOH.
`
`media1 protein A provides a technique for purifying recombinant antibodies because
`it can selectively bind antibodies in complex solutions1 allowing impurities to flow
`through (By et al., 1978; Surolia etal. 1 1982; Lindmark et a/., 1983; Reis et al., 1984).
`Protein A affinity chromatography is by far the most effective type of chroma(cid:173)
`tography for removal of host cell proteins and small molecules, and this is the nlain
`reason that it is used for antibody purification.
`In the past, the harvested cell culture fluid was often concentrated before the first
`chromatography step to decrease the loading time. With the development ofhigh-titre
`cell culture (typically >0.5 gil) and protein A affinity chromatography media capable
`ofhigh capacity at high flow rate (typically 20 gn at 40 CVIh)t the need to concentrate
`the harvested cell culture fluid has been eliminated. In our three-step process, the
`harvested cell culture fluid is loaded directly onto the protein A column. Because
`protein A affinity chromatography media is expensive, a smaller column is cycled
`several times to purify a single batch. This is possible because of the high flow rates
`that can be achieved for protein A columns.
`Cation exchange chromatography (Figure 12.2) is the second step. It uses a
`negatively charged group (typically sulphopropyl) immobilized to the chrolna(cid:173)
`tography media. Cation exchange chromatography clears host cell proteins, aggregate,
`and leached protein A (Table 12.1). The antibody binds to the negatively charged sites
`on the column, and it is eluted with a step gradient to high salt. Host cell proteins,
`aggregate, and leached protein A elute in the regeneration phase, after the antibody
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 9 of 28 PageID #:
`9488
`308 R.L. FAHRNER et OJ.
`
`200 r
`
`2
`
`14
`
`- - - -- -- - - - - - - - - -
`
`l~
`
`I I ~
`
`_....
`
`rr 1
`!
`~
`~
`.," i
`:
`:
`·
`.
`............... ~==-... j_ •••••--•••• j.__ .•- ......-r·"--····......-r.......-_.... ~-----
`·
`.
`\ -.
`
`I
`
`•
`I
`
`J
`
`I
`
`I
`•
`
`:
`
`I
`
`200
`
`400
`
`600
`
`Time (min)
`
`I'
`
`o
`
`o
`
`o
`
`o
`
`Figure 12.3. Chromatogranl from a typical anion exchange chromatography ron. The column was 0.66
`em inner diameter x 20 em length, packed with QSepharose Fast Flow. The column was washed with 3 CV
`of 0.180 M Tris HellO.07 M Tris BaseJ2.0 M NaCl, pH 8.0. equilibrated with 4 CV of 0.018 M Tris HeJI
`0.007 M Tris Base/G.OS M NaCl t pH 8.0, loaded to 100 gilt washed with 7 CV of 0.180 M Tris HeIlO.07
`MTris Basel2.0 M NaCl, pH 8.0, regenerated with 3 CV of0,25 MTris/2.0 M NaCl, pH 8.0, sanitized with
`2 CV of 0.5 NaOH, and stored in 3 eVofO.1 NaOH, 3 CV.
`
`has eluted. Cation exchange columns can be loaded to >40 gIl, which allows the batch
`of antibody to be purified in a single cycle on a reasonably sized column.
`Anion exchange chromatography (Figure 12.3) is the last chromatography step. It
`uses a positively charged group (typically quaternary amine) immobilized on the
`chromatography media. Anion exchange chromatography can be run in flow-through
`mode, which means that the antibody product flows through the column while the
`impurities bind. It removes DNA and residual host cell proteins. These impurities are
`removed from the column with a regeneration step, typically 0.5-1 M NaOH.
`These three steps together comprise a process that, while meeting stringent purity
`and throughput restrictions~ still produces a high yield of antibody (Table 12.1). The
`protein A affinity step has >95% yield, the cation exchange step has>75% yield t and
`the anion step has >95% yield t for an overall >65% process yield, which is excep(cid:173)
`tional for an industrial process with these extreme purity requirements. By choosing
`and sizing columns correctly and running them under conditions for high capacity, the
`throughput requirements can be met.
`
`PROCESS VALIDATION
`Validation is a regulatory requirement to demonstrate that a process, when operated
`within set parameters, can consistently produce a specified product. The complete
`validation plan is extensive and includes validation of process equipment, software,
`utilities, equipment cleaning~ and analytical methods.
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 10 of 28 PageID #:
`9489
`Industrial purification ofpharn1aceutical antibodies
`309
`
`Accordingto theU.5. FDA, ~Process validation is establishing documented evidence
`which provides a high degree of assurance that a specific process will consistently
`produce a product tneeting its pre-determined specifications and quality attributes'
`(FDA's Guidelines on General Principles of Process Validation, May It 1987). For
`chromatography processes, this means in part that processes must be validated at
`extremes ofoperating parameters such as load, conductivity, pH, and column lifetime.
`An iOlportant part of the validation effort is developing and writing validation
`protocols. A validation protocol is 'a written plan stating how validation will be
`conducted, including test parameters, product characteristics, production equipment,
`and decision points on what constitutes acceptable test results t (ibid.). The chroma(cid:173)
`tography validation studies should be carefully designed in advance, and data
`generated during process development is often used to determine validation ranges
`and critical process variables.
`SOlne validation studies must be performed at lnanufacturing scale. This includes
`the validation of process purity, where the levels of impurities are nleasured at each
`process step over several (usually three) runs. Validating the relnoval of impurities
`can elitninate the necessity to measure these impurities in each batch prior to release.
`The specific iInpllrities to be measured are determined in advance, and a table sitnilar
`to Table 12.1 is constructed showing the measured levels across the process. Con(cid:173)
`sistent results can then be demonstrated for consecutive runs.
`Several studies that are not practical to do at manufacturing scale may be performed
`at laboratory scale (Sofer, 1996). These include viral clearance, hold times for product
`pools and buffers used in production, and colutnn parameter and re--use. In the
`following sections, we present data from studies that validated the column operating
`ranges (parameter validation) and the column lifetime (re...use validation). This data
`also serves to illustrate the constraints under which the processes must operate, which
`may in tum affect the developlnent effort.
`Paralneter validation determines the effects of the variation of process conditions
`on the product and the process, because processes must be robust within the licensed
`operating parameters (Kelley et al., 1997). Typical variables that are studied during
`characterization are load, buffer conductivity, and buffer pH. The effect on the
`product and process is measured by yield and purity. Column lifetime should be
`prospectively detennined, and re-use validation determines a limit on the number of
`times a chromatography column may be re-used or cycled (Seely et aI., 1994).
`Both parameter and re-use validation were performed at laboratory scale. When
`llsing laboratory scale studies as part of the overall chromatography validation plan~
`every parameter except column diameter must be the same as manufacturing scale. To
`ensure comparability to the manufacturing process, all process parameters~ including
`buffers, volumes (measured in CV), and column heights were the same as the
`manufacturing process. Only the column diameter was changed. The buffers were
`prepared according to the manufacturing batch records using GMP raw materials.
`
`Protein A affinity chromatography
`
`DEVELOPMENT AND OPERATION
`The basic protocol of a protein A affinity column is straightforward: bind at neutral
`pH and elute at acid pH. This simple bind/elute chemistry does not leave much room
`
`Downloaded by [64.124.209.76] at 12:21 04 November 2015
`
`

`

`Case 1:18-cv-01363-CFC Document 79-6 Filed 03/22/19 Page 11 of 28 PageID #:
`9490
`310 R.L. FAHRNER et at.
`
`for purification optimization, but since protein A affinity chron18tography provides
`extreme purification in a single step, even an unoptimized process can produce a
`highly purified antibody. The optimization effort typically focuses not on purity but
`on throughput.
`Protein A affinity media is expensive compared to ion exchange media - more than
`30 times the cost. While the ion exchange process columns are sized so that a batch of
`antibody can be purified on a single cycle on the column, protein A affinity columns
`are sized to run several cycles to purify a single batch in order to minimize the cost of
`the colunm (as well as minimizing the cost of replacing the column if it is damaged).
`This cycling requires throughput optimization in order to purify the antibody in a
`reasonable amount of time. One important factor in optimizing throughput is the
`column capacity.
`Capacity is affected by many variables, including the type of protein A affinity
`chromatography media, ligand density. the antibody concentration in the load, the
`column temperature and column length, the buffer, conductivity) and pH of the load,
`and the flow rate (Katoh et aI., 1978; Tu et al., 1988; Fuglistaller, 1989; Kamiya et al.,
`1990; Kang and Ryu, 1991; Schuler and Reillacher, 1991; Van Sommeren et al.,
`1992). Of these variables~the simplest to control for production and the ones that will
`have the most significant impact on capacity are the column length, the flow rate. and
`the chromatography media. Bed height and flow rate both affect the breakthrough
`capacity; together bed height and flow rate detel1nine the residence time (Fahrner et
`al., 1999a).
`Several types of chromatography media are available for process applications.
`They include Sepharose Fast Flow (crosslinked agarose), Poros 50 (polystyrene(cid:173)
`divinylbenzene), and Prosep (controlled...pore glass). In a study comparing these
`sorbents (Fahrner et af., 1999c), we found that the sorbent type and flow rate had a
`strong effect on breakthrough capacity (Figure 12.4). Flow rate had the strongest
`effect on Sepharose; while both Poras and Prosep were less strongly affected by flow
`rate, Poras had a higher capacity at all flow rates. The type of media had a strong effect
`on breakthrough capacity, but it did not strongly affect the purity of the antibody
`(Table 12.2). For example, the amount of host cell proteins in the purified antibody
`pools ranged from 2.5 mg/g to 4.9 mg/g. The amount of host cell proteins in the load
`was approximately 950 mg/g (950,000 ppm), so these numbers represent a range from
`380-fold clearance to 190-fold clearance. The Paros sorbent may have the least non-
`
`Table 12.2. Comparison of protein A affinity chromatography sorbents
`
`Pressure drop (psi h em-·! x 10-3)
`Purified antibody
`Yield (0/0)
`DNA (nglnlg)
`Host cell proteins (mg/g)
`Protein A (ngllng)
`
`Poros 50
`
`3.2
`
`I04± 1
`41:t 3
`2.5 ± 0.2
`4.6 ± 0.5
`
`Prosep
`
`0.3
`
`103 ±2
`40:t4
`3.7 ± 0.2
`3.1 ± 0.5
`
`Sepharose
`
`1.1
`
`IOO± 2
`29±2
`4.9:f: 1.2
`5.7 ± 1.7
`
`(Data from Fahrner el al, 1999c.) Values for yield (percent loaded amibod)' in Ihe purified pool), host cell proleins (mg
`host cell proteins per g mllibody)~ DNA {ng DNA per mg antibody). and protein A (ng protein A per mg antibody) were
`for runs using a 10 em column lenglh and 500 cmlh flow rate (50 CVIb). loaded 10 their capacity determined at 1%
`breakthrough. Values are the aver~ge of three runs, plus or minus one stundntd deviation. Load material was clarified
`Chinese hamster ovary cell culture fluid.
`
`Downloaded by [64.124.209.76] at 12:21