`WIT. Ktbnoti
`DATE --4 -
`KRAMM COURTREPOR rING
`
`UC San Diego
`UC San Diego Previously Published Works
`
`Title
`Comparison of binding characteristics and in vitro activities of three inhibitors of vascular
`endothelial growth factor A.
`
`Permalink
`https://escholarship.orci/uc/item/3408m1cv
`
`Journal
`Molecular pharmaceutics, 11(10)
`
`ISSN
`1543-8384
`
`Authors
`Yang, Jihong
`Wang, Xiangdan
`Fuh, Germaine
`et al.
`
`Publication Date
`2014-10-01
`
`DOI
`10.1021/mpsool6ov
`
`Peer reviewed
`
`eScholarshio.orq
`
`Powered by the California Digital Library
`University of California
`
`Mylan Exhibit 1106
`Mylan v. Regeneron, IPR2021-00881
`Page 1
`
`
`
`oarmaceutics
`
`pubs.acs.org/molecu la rpha rmaceutics
`
`Article
`
`Comparison of Binding Characteristics and In Vitro Activities of
`Three Inhibitors of Vascular Endothelial Growth Factor A
`Jihong Yang Xiangdan Wang Germaine Fuh, Lanlan Yu, Eric Wakshull, Mehraban Khosraviani,
`Eric S. Day, Barthélemy Demeule, Jun Liu, Steven J. Shire) Napoleone Ferrara,t and Sandeep Yadav*
`
`Genentech, Inc., 1 DNA Way, MS 56-2A, South San Francisco, California 94080, United States
`
`Ran,b,ziirnab
`
`Atibecctat
`
`U"w (UtUe
`
`Bev,ci,u mat
`
`A
`
`B
`
`C
`
`t
`
`ABSTRACT: The objectives of this study were to evaluate the relative binding and potencies of three inhibitors of vascular
`endothelial growth factor A (VEGF), used to treat neovascular age-related macular degeneration, and assess their relevance in the
`context of clinical outcome. Ranibizumab is a 48 kDa antigen binding fragment, which lacks a fragment crystallizable (Fc) region
`and is rapidly cleared from systemic circulation. Aflibercept, a 110 kDa fusion protein, and bevacizumab, a 150 kDa monoclonal
`antibody, each contain an Fc region. Binding affinities were determined using Biacore analysis. Competitive binding by
`sedimentation velocity analytical ultracentrifugation (SV-AUC) was used to support the binding affinities determined by Biacore
`of ranibizumab and aflibercept to VEGF. A bovine retinal microvascular endothelial cell (BREC) proliferation assay was used to
`measure potency. Biacore measurements were format dependent, especially for aflibercept, suggesting that biologically relevant,
`true affinities of recombinant VEGF (rhVEGF) and its inhibitors are yet to be determined. Despite this assay format dependency,
`ranibizumab appeared to be a very tight VEGF binder in all three formats. The results are also very comparable to those reported
`previously.` At equivalent molar ratios, ranibizumab was able to displace aflibercept from preformed aflibercept/VEGF
`complexes in solution as assessed by SV-AUC, whereas aflibercept was not able to significantly displace ranibizumab from
`preformed ranibizumab/VEGF complexes. Ranibizumab, aflibercept, and bevacizumab showed dose-dependent inhibition of
`BREC proliferation induced by 6 ng/mL VEGF, with average IC,,, values of 0.088 ± 0.032, 0.090 ± 0.009, and 0.500 ± 0.091
`nM, respectively. Similar results were obtained with 3 nglmL VEGF. In summary Biacore studies and SV-AUC solution studies
`show that aflibercept does not bind with higher affinity than ranibizumab to VEGF as recently reported,' and both inhibitors
`appeared to be equipotent with respect to their ability to inhibit VEGF function.
`KEYWORDS: rauibizuniab, aflibercept, bevacizu,nab, VEGF, affinity, analytical ultracentnfugation
`
`• INTRODUCTION
`
`important variables that govern the determination of affinity
`
`The determination of binding affinity of a therapeutic protein
`to a target is an integral part of pharmaceutical development. A
`widely used methodology for assessing tight interactions is
`based on surface plasmon resonance (SPR) such as Biacore.5
`One caveat in this technology is that it requires ligand
`immobilization to a surface and it has been shown that
`orientation, method of binding, and which ligand is bound are
`
`constants. 6,7
`
`Received: February 25, 2014
`Revised:
`August 22, 2014
`Accepted: August 27, 2014
`Published: August 27, 2014
`
`'rACS Publications
`
`02014 American Chemical Society
`
`3421
`
`dadolorg/10.1021/nipsooltoy I Moi. plwrmoceatic, 2014. 11.3421-3430
`
`Mylan Exhibit 1106
`Mylan v. Regeneron, IPR2021-00881
`Page 2
`
`
`
`Molecular Pharmaceutics
`
`Alternatively, analytical ultracentrifugation (AUC) can be
`used as an orthogonal free-solution technique that circumvents
`the potential artifacts of matrix/stationary phase or chemical
`modifications associated with SPR. AUC has been widely used
`in the biophysical characterization of proteins to determine
`weight-average molecular mass, sedimentation coefficient,
`frictional coefficient associated with molecular shape, bimo-
`lecular interactions involving reversible associations or complex
`formation, self-association of glycosylated and nonglycosylated
`proteins, and competitive binding of anti-IgE antibodies to IgE
`and also as an orthogonal technique to size exclusion
`chromatography to determine the presence of aggregates.' In
`addition to being able to detect the presence of protein
`aggregates, AUG analysis allows measurements directly in the
`formulation buffer or condition of interest, thereby avoiding
`common size exclusion HPLC limitations of protein-resin
`interactions and significant dilution in the elution buffer that
`can potentially alter the size distribution of the self-associates
`and aggregates, as highlighted in the above studies. One aspect
`of this work is to assess AUC as an orthogonal technique to
`SPR in evaluating the binding of therapeutic proteins,
`highlighting that caution must be exercised while relying on
`SPR results. In addition, the recent development of
`fluorescence optics in the analytical ultracentrifuge9'° com-
`bined with the use of fluorescently labeled material can provide
`definitive information about the type of complex formed.
`There have been several SPR studies and potency assess-
`ments of inhibitors of vascular endothelial growth factor A
`(VEGF),2'-'4"' 2 a key driver of the vascular leakage and
`neovascularization seen in intraocular vascular diseases
`including age-related macular degeneration (wet AMD), retinal
`vein occlusion (RVO), and diabetic macular edema (DME).'
`VEGF inhibitors such as ranibizumab, aflibercept, and
`bevacizumab are used intravitreally in patients with wet
`AMD, RVO, and DME. The US Food and Drug Admin-
`istration has approved ranibizumab for the treatment of wet
`AMD, RVO, and DME 13 and allibercept for the treatment of
`wet AMD and central RVO. 14 lievacizumab is used off-label. A
`recent SPR study concludes that aflibercept binds to VEGF
`with much higher affinity than ranibizumab.' Herein we report
`the results of the determination of affinity constants for binding
`of VEGF to ranibizumab, affibercept, and bevacizumab by SPR
`using different assay formats as well as the potency of inhibition
`of VEGF. In addition, a novel solution based competitive
`analytical ultracentrifuge method is used to support our
`conclusions.
`
`• MATERIALS AND METHODS
`
`Lucentis (ranibizumab) and Avastin (bevacizumab) were
`obtained from Genentech, Inc., South San Francisco, CA.
`Eylea (aflibercept; Regeneron, Inc., Tarrytown, NY) Was
`obtained commercially. Recombinant human VEGF 165
`(rhVEGF) was expressed and purified at Genentech. Anti-Fab
`antibody was obtained from GE Healthcare (Pittsburgh, PA),
`and protein A was obtained from Thermo Fisher Scientific,
`Pierce Protein Biology Products (Rockford, IL).
`SPR Binding Assays. Binding kinetics and affinities of
`inhibitors of rhVEGF were assessed using surface plasmon
`resonance technology on a Biacore T200 instrument (GE
`Healthcare, Pittsburgh, PA). A series of analyte concentrations
`were prepared in HES-EP running buffer (0.01 M HEPES, 0.15
`lvi NaCl, 3 mM EDTA, and 0.05% surfactant Polysorbate 20)
`and injected at a flow rate of 50 pL/min for 3 min over flow
`
`Article
`
`cells (FCs) of Series S CMS sensor chips immobilized with
`ligand molecules at various densities depending on assay
`formats.
`In format 1, rhVEGF was the ligand immobilized directly
`onto FCs at --'20 resonance units (RU) density, while the
`inhibitors were the analytes. The dissociation of inhibitors from
`the immobilized rhVEGF was allowed to proceed for S min for
`all samples except for the ranibizumab and bevacizumab
`samples with the highest concentration (200 nM), in which
`dissociation proceeded for 3 h or is mm, respectively. All
`experiments were carried out at 37 °C.
`In format 2, the inhibitors were the immobilized ligands and
`rhVEGF was the analyte in the mobile phase. The final ligand
`density was 22-45 RU. The dissociation of the analytes from
`the immobilized ligand was allowed to proceed for S min for all
`samples except for the ranibizumab sample with the highest
`concentration (200 nM), in which dissociation proceeded for 3
`h. The dissociation time for all bevacizumab samples was 20
`mm. All experiments were carried out at 37 °C.
`In format 3, the inhibitors were immobilized indirectly to the
`sensor chip using anti-human lgG Fab antibody or protein A as
`capturing molecules as previously reported.' Rigorous surface
`testing was conducted in the current study to evaluate the
`validity of the method for all inhibitors. The densities of the
`capture molecules were -11000 or 1000 RU for anti-human
`1gG Fab antibody and -S500 RU for protein A. The final
`ligand density used was -'28 RU for the indirectly captured
`ranibizumab, -'40 RU for bevacizumab, and -'50 RU for
`aflibercept. The dissociation of the analytes from the
`immobilized ligand was allowed to proceed for S min for
`ranibizumab samples except for the sample with the highest
`concentration (200 nM), in which dissociation proceeded for 3
`h. The dissociation time for all aflibercept and bevacizumab
`samples was 30 mm. Experiments were carried out at 37 °C or
`25 °C for this format.
`Sensorgrams of ranibizumab, allibercept, and bevacizunsab
`binding to rhVEGF using all three formats were analyzed to
`obtain kinetic data and affinities using Biacore T200 Evaluation
`Software (version 2.0.1; GE Healthcare). Because of the
`dimeric nature of rhVEGF and the presence of two potential
`binding sites in all inhibitors except ranibizumab, definitive
`monovalent binding affinities for rhVEGF and its inhibitors can
`be challenging to obtain. Very low immobilization densities
`were used to encourage monovalent binding, and the presence
`of such interactions were evaluated using a 1:1 Langmuir
`binding model. In all but two conditions tested, the 1:1 binding
`model was sufficient to describe interactions between rhVEGF
`and its inhibitors. The dissociation rate constant (kd) and
`association rate constant (Ic,) were obtained via kinetic fitting,
`and the equilibrium dissociation constant (K,,) was derived by
`taking the ratio of kd over k. calculated using the simplest 1:1
`binding model. Only in cases where allibercept was evaluated in
`formats I and 2 was the 1:1 binding model insufficient to
`describe interactions between the inhibitor and rhVEGF. In
`those cases a bivalent analyte binding model was used and the
`first equilibrium dissociation constant (K,,,), first dissociation
`rate constant (kdj, and first association rate constant (Ic,,) were
`reported.
`Competitive Binding Assessed by Sedimentation
`Velocity Analytical Ultracentrifugation (SV-AUC). Each
`molecule individually and preformed complexes between
`ranibizumab and VEGF and allibercept and VEGF were first
`evaluated to obtain their sedimentation coefficients. After this,
`
`3422
`
`d,.doI.o,g10102i/n,p500160v I Mel. Pharmoc.utirs 2014, 1,3421-3430
`
`Mylan Exhibit 1106
`Mylan v. Regeneron, IPR2021-00881
`Page 3
`
`
`
`Molecular Pharmaceutics
`
`Ranibizumab
`La
`La
`ts
`pM)
`67
`
`0.19
`
`0.67
`
`\\.
`
`tI
`
`
`
`04
`
`Aflibercept
`kal
`Ml
`(IC' 1/MO) )I1 I!')
`3.30
`1)0.1
`
`6,6
`)pM)
`0163
`
`Article
`
`Bevacizumab
`Si
`64
`60
`30' lAos)
`lu'
`two
`0,14
`219
`4410
`'.3
`
`A
`
`B
`
`C
`
`La
`110' I/Mt)
`
`64
`510' 1/')
`
`Op
`two
`
`64'
`La!
`210' I/Mi) 510' IfS)
`U4
`6906
`
`6<1
`(pM)
`6744
`
`La
`fiG' 1.44,)
`31.1
`
`Olt
`
`LaO
`
`rn' V.)
`
`32.9
`
`K,
`5pM)
`159
`
`La
`
`60
`
`La
`
`7.7
`
`La
`La
`510' l/MOS 110 - 1A)
`213
`203
`
`An114, ii 111 & L) Ab
`
`'
`
`rro, on A
`
`Figure 1. Binding of rhVEGF—anti-VEGF inhibitor molecules in Biacore assays. (A) rhVEGF is the ligand immobilized directly onto the flow cell
`(PC) while the inhibitors were the analytes injected over the PC at varying concentrations (format I). (B) VEGF inhibitors immobilized as ligand
`with VEGF in the mobile phase as analyse (format 2). (C) Inhibitors immobilized indirectly to the sensor chip with VEGF in the mobile phase as
`analyse (format 3). Note: limit of !Cj that can be accurately measured by the instrument is —S X io s'. To be conservative, a k4 < io s— was
`chosen.
`
`competition experiments were conducted using a preformed
`inhibitor/VEGF complex challenged with a different VEGF
`inhibitor to assess whether the previously reported —100-fold
`higher affinity of aflibercept to rhVEGF compared with
`ranibizumab4 is valid in free solution, i.e., no binding to a
`surface as in SPR measurements.
`Experiments were performed at room temperature in PBS,
`PH 7.2 (137 mM NaCl, 27 mM KCl, 8 mM Na1HPO 4, and 1.5
`mM KH 2PO4). Alexa Fluor 488 protein labeling kits were
`purchased from Molecular Probes (Eugene, OR). All chemicals
`used were reagent grade or higher. Alexa Fluor 488 labeled
`ranibizumab (denoted as ranibizumab*) was produced as
`recommended by the manufacturer.
`Sedimentation velocity experiments were performed in an
`Optima XL-I analytical ultracentrifuge equipped with absorb-
`ance optics, interference optics (Beckman Coulter, Fullerton,
`CA), and fluorescence optics (Aviv Biomedical) in centrifuge
`cells with 12 mm graphite-filled Epon centerpieces (Spin
`Analytical, Durham, NH) at 20 °C and rotor speed of 40000
`rpm. Quartz windows were used when using the absorbance
`optics, and the scans were acquired at a wavelength of 230 nm
`at 30 pm radial increments. When using the fluorescence optics,
`sapphire windows were used, and the data were acquired at 20
`pm radial increments averaging five revolutions per scan. The
`sedimentation boundaries were analyzed with SEDFIT, version
`11.3 and 11.72c.' 5 The resulting continuous, c(s), distribution
`with 70% confidence level was calculated after optimizing
`baseline, meniscus, and cell bottom positions by nonlinear
`regression. All s values obtained with the c(s) distribution in
`
`PBS were converted to s2p. with SEDNTERP (version 1.09)
`using the measured density and viscosity of PBS.
`Bovine Retinal Microvascular Endothelial Cell (BREC)
`Proliferation Assay. A BREC assay was used because bovine
`microvascular endothelial cells are well established as a cell type
`that is highly responsive to growth factors such as VEGF and
`bFGF.' 6 Unlike another cell line that has been used to assess
`potency, HUVEC, which is derived from a large vessel, BREC is
`a more physiologically relevant cell type to investigate
`angiogenesis.
`BREC proliferation assays were performed as previously
`described. 2 Cells were seeded in 96-well plates in low glucose
`DMEM (supplemented with 10% heat-inactivated calf serum, 2
`mM glutamine, and antibiotics) at a density of 500 cells/well.
`Ranibizumab and allibercept were tested from 0.004 to 10 nM,
`while bevacizumab was tested from 0.04 to 90 nM. Twenty
`minutes after addition of inhibitors, VEGF was added to a final
`concentration of 6 ng/mL (o.is nM) or 3 nglmL (0.075 nM).
`After 6 days, cell growth was assayed with the use of alamarBlue
`(BioSource). Fluorescence was measured at 530 nm excitation
`wavelength and 590 nm emission wavelength. 1C values were
`calculated using KaleidaGraph. For statistical analysis, one-way
`ANOVA was used, followed by the Tukey—Kramer HSD test
`comparing all pairs.
`
`• RESULTS
`Binding Affinities and Kinetics of VEGF Inhibitors
`Using SPR Technology. Format 1. rhVEGF, an antiparallel
`homodimer, was the immobilized ligand with inhibitors in the
`mobile phase as analyses (Figure IA). Binding of ranibizumab
`
`3423
`
`ddrA.orp/l0.10ll/rrp50Ol60, I Mat Pho,lnoreel)cs 2014,11,3421-3430
`
`Mylan Exhibit 1106
`Mylan v. Regeneron, IPR2021-00881
`Page 4
`
`
`
`Molecular Pharmaceutics
`
`Article
`
`followed a simple monovalent (1:1) analyte binding model as
`expected because the Fab molecule has only one VEGF binding
`site. This was clearly shown by the closeness of the fit to the
`experimental data (Figure IA). Allibercept and bevacizumab
`each contained two VEGF-binding sites. Using very low
`rhVEGF immobilization levels, binding of these two inhibitors
`to rhVEGF was encouraged to favor monovalent binding. 17,18
`This approach worked for bevacizumab since the 1:1 binding
`model sufficiently described the interactions between the
`inhibitor and rhVEGF (Figure 1A). However, in the case of
`aflibercept, even the lowest immobilization level of rhVEGF
`was insufficient to completely shift the interaction to
`monovalent binding, and attempts to fit the data with a 1:1
`Langmuir binding model failed (data not shown). Therefore, a
`bivalent analyte binding model was used considering each
`aflibercept molecule has two potential VEGF binding sites.
`Although the curve fits still deviated from the experimentally
`obtained results (Figure IA), the overall quality of the fit was
`much improved over that obtained from using a 1:1 binding
`model (data not shown). The challenge in fitting the aflibercept
`binding curves may be due to the global fit bivalent model that
`allows individual bulk effect correction to accommodate
`baseline drift,' 8 although other factors such as binding induced
`conformational change cannot be ruled out. Because the second
`step in the bivalent analyte binding model involves intra-
`molecular binding on a sensor chip without an increase in mass,
`only two-dimensional kinetics for the second step are obtained.
`The first step kinetics from a bivalent analyte binding model are
`most relevant in understanding the binding kinetics and
`- strength between an analyte and a ligand. Therefore, only
`first kinetic parameters (k46, k3,) and first KD, were shown for
`aflibercept binding to rhVEGF. Although the (first) association
`rate constants for all three inhibitors were similar, ranibizumab
`had a much slower dissociation rate constant (039 X la_s j'1)
`than aflibercept and bevacizumab (zso.z x io- and 21.9 )<
`lO' s', respectively). As a result, ranibizumab showed a lower
`value (67 pM) than aflibercept (9263 pM) or bevacizumab
`(4456 pM) (Figure IA, insets).
`Format 2. The inhibitors were the immobilized ligands with
`rhVEGF in mobile phase as the analyte (Figure IB). Since each
`rhVEGF has two potential binding sites for the immobilized
`inhibitor molecules, experimental conditions were again
`optimized to encourage monovalent interactions by using low
`ligand immobilization levels. Similar to format 1, the fits using
`the 1:1 binding model for both ranibizumab and bevacizumab
`showed reasonable agreement to the experimentally obtained
`results (Figure IB); for aflibercept, much discrepancy was once
`again observed between the experimental data and the fitted
`curves using the 1:1 binding model (data not shown).
`Therefore, the bivalent analyte binding model was used, and
`first kinetic parameters and dissociation equilibrium constant
`were summarized in Figure 1B, inset. Ranibizumab again
`dissociated much more slowly than the other two inhibitors,
`and a very conservative limit (i x i0 5 s') of it3 was used in
`order to confidently assess the upper limit of K0 value. Even
`with this conservative approach, ranibizumab showed a higher
`binding affinity than aflibercept and bevacizumab, with K0
`(K0,) values of <9.2, 4744, and 159 pM, respectively (Figure
`IB, insets). Comparing results obtained from format 2 and
`format I revealed some interesting observations: the (first)
`association rate constants it, for all three inhibitors were higher
`using format 2 than format I. While it is challenging to know
`exactly how different the it3 values were for ranibizumab, both
`
`formats showed very slow dissociation. The it3 (it31 ) values
`obtained from both formats were very similar for bevacizumab
`but not for aflibercept (Figure IA,B, insets).
`Format 3. The inhibitors were immobilized indirectly to the
`sensor chip using anti-human 1gG Fab antibody or protein A as
`capturing molecules following surface testing to evaluate the
`validity of the method for all VEGF inhibitors (Figure IC).
`While the levels of captured aflibercept and bevacizumab (not
`shown) stayed almost the same during the time needed for
`kinetic analysis (Figure 2A), a significant amount of
`
`520
`.2
`
`B
`
`rime, seconds
`
`1
`
`2000
`
`14.
`
`Time, seconds
`
`2
`
`4305
`
`5300
`
`0000
`
`I 2008
`
`12000
`
`14M
`
`Time, seconds
`
`Figure 2. Levels of aflibercept and ranibizumab captured by an anti-
`Fab antibody or protein A. (A) Level of aflibercept captured by protein
`A over 3 h in an indirect capturing format at 37°C. Level of protein A
`immobilized was approximately 5500 RU. Alhibercept was indirectly
`captured at approximately 50 RU. (B) Level of ranibizumab captured
`by anti-Fab antibody over 3 Ii in an indirect capturing format at 37 °C.
`Level of anti-Fab antibody immobilized on the sensor chip was
`approximately 11000 RU. Ranibizumab was indirectly captured at
`approximately 28 RU. (C) Level of ranibizumab captured by the anti-
`Fab antibody over 3 h in an indirect capturing format at 25 °C. Level
`of anti-Fab antibody immobilized on the sensor chip was
`approximately 1000 RU. Ranibizumab was indirectly captured at
`approximately 27 RU.
`
`ranibizumab dissociated from the anti-Fab antibody capture
`molecule: signal decreased nearly 100% when anti-Fab antibody
`was immobilized at 11000 RU at 37 °C (Figure 2B) and more
`than 5096 when anti-Fab antibody was immobilized at 1000 RU
`at 25 °C (Figure 2C), a condition reported in the literature.'
`These results indicated that this anti-Fab antibody is not
`suitable to capture ranibizumab for affinity measurements. We
`
`3424
`
`do.do&ocf30A011/mp500l60vI Md. Phom,oceol,c, 2014, Ii, 0411-3430
`
`Mylan Exhibit 1106
`Mylan v. Regeneron, IPR2021-00881
`Page 5
`
`
`
`Molecular Pharmaceutics
`
`Article
`
`3.7
`
`5.2
`
`(I
`Ii
`II
`II
`II
`II
`
`3.0,
`
`6.5
`
`A
`
`3
`
`2
`
`1
`
`0
`00 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0
`Sedimentation coefficient (5,,j
`
`40 0
`
`4008
`
`6.1
`
`C
`
`3500
`
`3000
`
`2500
`
`Zy 2000
`
`1500
`
`1000
`
`500
`
`0.0
`
`2.0
`
`4.0
`
`6.0
`
`3.7
`
`0 0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 ILO 12.0
`Sedimentation Coefficient (S,.)
`
`1
`
`5.7
`
`55 N
`
`j
`
`5.8
`
`D
`
`3.4
`4
`
`.
`
`.
`
`• 0
`0
`001.02.0304.05.06.0708.09,010.011.012.0 00 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0
`Sedimentation Coefficient (S)
`Sedimentation Coefficient (S,)
`
`.4.
`
`Figure 3. Analytical ultracentrifugation analysis. (A) Continuous sedimentation coefficient distribution of VEGF (solid black trace), ranibizumab
`(dashed black trace), aflibercept (green trace), 2:1 ranibizumablVtcF complex (red trace), and 1:1 aflibercept/VEGF complex (blue trace) in PBS
`at a total protein concentration of 0.1 mg/mL. The 2:1 ranibizumab/VEGF complex and 1:1 aflibercept/VEGF complex were prepared on a molar
`basis. Ranibizumab, VEGF, and aflibercept were 99% monomeric and did not reveal the presence of aggregate missed by SE-HPLC. Aflibercept
`showed a small amount of higher order species around 7.1 5, which appears to be present in 1:1 aflibercept/VEGF complex, as well at 8.7 S. The
`higher order species were present at a low enough concentration to be considered insignificant. (B) Competition experiment: the sedimentation
`profile of a solution containing 1 mol equiv of ranibizumab (dotted line) or 2 mol equiv of ranibizumab (solid line) added to preformed 1:1
`aflibercept/VEGF complex. The addition of ranibizumab to the aflibercept/VEGF complex displaced aflibercept, which appears as a monomer at 5.3
`S. (C) The continuous sedimentation coefficient distribution for labeled ranibizumab* (inset) and 2 mol equiv of ranibizumab* (labeled) added to
`preformed 1:1 aflibercept/VEGF complex in PBS measured using the fluorescence detection system (FDS) optics. The total concentration of labeled
`ranibi zumab* was maintained at 200 nM. The preformed l:t aflibercept/VEGF complex in PBS was challenged with 2 mol equiv of labeled
`ranibizumab*. The sedimentation coefficient distribution clearly shows the formation of a ranibizumab/VEGF complex. (D) The sedimentation
`profile of various complexes where a preformed 2:1 ranibizumab/VEGF complex was titrated with different amounts of aflibercept in PBS. The total
`protein concentration was maintained at 0.1 mg/mt. A preformed 2:1 ranibizumab/VEGF complex was challenged with 0.5 mol equiv (red trace), 1
`mol equiv (blue trace), and 4 mol equiv (black trace) of aflibercept. Aflibercept was not able to materially displace ranibizumab from the
`ranibizuanab/VEGF complex.
`
`were able to obtain a stable level of captured ranibizumab by
`using an antibody specific to hulgG heavy and light chain
`(Jackson ImmunoResearch Laboratories, Inc., West Grove,
`PA). The results for ranibizumab were similar to those obtained
`in format 2, with an equilibrium dissociation constant of less
`than 13.1 pM using a conservative estimation of the k value
`(Figure IC, inset). In contrast, the results for aflibercept in this
`assay format were drastically different from the ones obtained
`in the first two assay formats: first of all, a simple 1:1 binding
`model was sufficient to describe the interactions between
`aflibercept and rhVEGF using format 3; in addition, the Ic, value
`was over 500- and 10-fold higher than k, 1s obtained from
`formats I and 2, respectively; conversely, the kd value was about
`10- and 200-fold smaller than the kd s's derived from formats I
`and 2, respectively. These kinetic differences resulted in K0
`
`(K01 ) values for aflibercept ranging from 1.83 to 9263 pM
`(Figure I, insets). The K0 of 1.83 pM for aflibercept is similar
`to what was reported by Papadopoulos et al .'4 although the
`experiment reported here was conducted at 37 °C rather than
`25 °C. The association rate constant for aflibercept as
`determined in this format, 1.63 X 108 M' s, is extremely
`fast and may have exceeded the instrument's limit for reliable
`measurement. The results for bevacizumab were similar to the
`ones in format 2 with similar Ic,, Ic4, and IC0 values, obtained
`using a 1:1 binding model (Figures IB,C, insets). The K0 for
`bevacizumab was 7&4 pM as determined using format 3.
`Comparing kinetics obtained from format 3, it is interesting to
`note that both ranibizumab and aflibercept have very high
`affinities in their binding to VEGF, although the main drivers
`
`3425
`
`dx dea.orgllo.loal/rnpstoleov I Moe Phe,mcceurica 2014, 11. a421-a4a0
`
`Mylan Exhibit 1106
`Mylan v. Regeneron, IPR2021-00881
`Page 6
`
`
`
`Molecular Pharmaceutics
`
`Article
`
`for those tight interactions are quite different since ranibizumab
`dissociated very slowly, while aflibercept associated very rapidly.
`Evaluation of Ranibizumab and Aflibercept Binding
`to VEGF Using Sedimentation Velocity Analytical Ultra-
`centrifugation (SV-AUC). Each molecule and preformed
`ranibizumab/VEGF and aflibercept/VEGF complex was first
`evaluated individually to obtain its sedimentation coefficient.
`The analytical ultracentrifugation data was fitted by the direct
`boundary model, c(s), maximizing the resolution and
`sensitivity. 1"9'2° The sedimentation coefficients of VEGF,
`ranibizumab, and aflibercept in PBS, corrected for the standard
`conditions of water at 20 °C
`were 3.0, 3.7, and 5.2 5,
`respectively (Figure 3A). The sedimentation coefficients of the
`2:1 ranibizumab/VEGF complex and 1:1 aflibercept/VEGF
`complex were 6.0 and 6.5 5, respectively (Figure 3A).
`A competition assay in which a preformed 1:1 complex of
`aflibercept/VEGF was challenged with 1 mot equiv of free
`ranibizumab (Figure 3B, dotted line) showed a major peak at
`6.2 5, a peak at 3.6 5, and a raised baseline between 4 and 6 S.
`The peak at 3.6 S is consistent with free ranibizumab (Figure
`3A). The main peak at 6.25 likely represents a mixture of 2:1
`complex of ranibizumabiVhGF (6.0 5) and 1:1 complex of
`aflibercept/VEGF (6.5 5). The raised baseline indicates that the
`mixture is complex and may also contain free aflibercept.
`Interestingly, the addition of 2 mot equiv of free ranibizumab to
`a 1:1 aflibercept/VEGF complex showed a more symmetrical
`peak at 6.2 S with a distinct shoulder at 5.3 S and a peak at 3.6 S
`(Figure 3B, solid line). The shoulder at 5.3 S is consistent with
`free aflibercept and that at 3.6 S with free ranibizumab (Figure
`3A), indicating that ranibizumab was able to displace aflibercept
`from the preformed aflibercept/VEGF complex.
`To verify that the peak at 6.2 5 contains significant levels of
`ranibizumab/VEGF complex, ranibizumab was labeled with the
`Alexa Fluor 488 dye (ranibizumab*) and the competition
`experiment was performed with detection using fluorescence
`optics (Figure 3C). The sedimentation coefficient
`= 3.7
`5) for ranibizumab* alone (inset, Figure 3C) confirmed that
`the labeling procedure did not change the sedimentation
`characteristics of ranibizumab. Additionally, the potency was
`also unaffected (data not shown). The addition of 2 mot equiv
`of ranibizumab to the preformed 1:1 aflibercept/VEGF
`complex clearly showed a ranibizumab*/VEGF complex peak
`at 6.1 S in addition to the free ranibizumab* peak at 3.75. Note
`that with fluorescence optics, only the fluorescently labeled
`species (i.e., ranibizumab* and ranibizumab* complex) are
`visible, and any free aflibercept or aflibercept associated
`complex will not be detected.
`The reverse competition experiment was performed where a
`preformed 2:1 ranibizumab/VEGF complex was challenged
`with 0.5, 1, and 4 mot equiv of free aflibercept (Figure 3D). In
`this case, no labeling was used and the detection was achieved
`with UV optics. The minor peak at 3.4 S is likely a small
`amount of free ranibizumab. The free ranibizuniab peak at 3.4 S
`remained constant with increasing amount of free aflibercept
`added to the preformed complex, indicating no large
`displacement of ranibizumab by aflibercept, even at a 4-fold
`molar excess of aflibercept. The slight difference observed in
`the sedimentation coefficient of free ranibizumab (3.4 vs 3.7 5
`from Figure 3A) is likely due to the known limits of c(s)
`analysis to accurately resolve minor peaks.2' With increasing
`molar concentrations of free aflibercept, the major peak shifted
`to a smaller sedimentation coefficient, that is, from 5.8 to 5.7 to
`5.6 S at 0.5, I, and 4 mot equiv of free aflibercept, respectively,
`
`indicating that the peak may consist of a mixture of 2:1
`ranibizumab/VEGF complex and increasing levels of free
`aflibercept that are not well resolved. The following mass
`balance equations predict this behavior:
`
`s20,(obs) = [A1o,w(A)] + 1p(RV)s20,11(RV)1
`
`A) + (av) = 1
`
`(A) = C(A)/ C, 0031
`
`(I)
`
`(2)
`
`(3)
`
`where F is the weight fraction, C is the concentration in mg/
`mL, and the subscripts (A) and (RV) indicate free aflibercept
`and ranibizumab/VEGF complex) respectively.
`Equation 1 assumes that the observed sedimentation
`coefficient is composed of the weight fraction of free aflibercept
`and ranibizumab/VEGF complex cosedimenting. Equation 2
`represents the assumption that free aflibercept and ranibizu-
`mab/VEGF complex account for all sedimenting species.
`Equation 3 defines the weight fraction of free aflibercept as
`the concentration of aflibercept added to the total protein
`concentration present.
`With C,001 = 0.1 mglmL and the added concentration of
`aflibercept, (A) and F( Rv) can be calculated from eqs 2 and 3.
`With the measured sedimentation coefficients from Figure 3k
`520,w(A) = 5.2 and s2, (,) = 6.0, th e obse rved sedimentation
`coefficient [s2 (obs)] can be calculated from eq I and
`compared with the experimentally determined sedimentation
`coefficients (Table i).
`
`Table 1
`
`co)
`0.030
`O.O4
`0.770
`
`' (caic)
`
`5.76
`5.63
`5.38
`
`s,,.. (obs)
`5.8
`5.7
`5.5
`
`The calculated sw,, values are in remarkably good agreement
`with the experimentally observed 520,w' corroborating the notion
`that the observed shift in the s-value of the major peak in Figure
`3D is likely a consequence of the main peak consisting of
`unresolved 2:1 ranibizumab/VEGF complex and free afliber-
`cept and that increasing amounts of free aflibercept added to
`the preformed ranibizumab/VEGF complex does not signifi-
`cantly disrupt this comp