R E V I E W S
`
`Glycosylation as a strategy to improve
`antibody-based therapeutics
`
`Roy Jefferis
`
`Abstract | To date, more than 20 recombinant immunoglobulin G (IgG) antibody therapeutics
`are licensed for the treatment of various diseases. The mechanism of action of recombinant
`monoclonal antibodies (rMAbs) has been extensively investigated and several distinct
`pathways have been defined; selective activation of specific pathways may optimize clinical
`outcomes for different diseases, such as cancer and chronic inflammation. Human IgG is
`a glycoprotein with oligosaccharides attached at a single site. These are essential to the
`mode of action of rMAbs, and the antibody efficacy can vary depending on the particular
`oligosaccharide that is attached. Methods are now becoming available that allow the
`production of rMAbs bearing pre-selected oligosaccharides — glycoforms — to provide
`maximum efficacy for a given disease indication. This Review summarizes current knowledge
`of these methods and avenues for their exploitation in the clinic.
`
`Recombinant monoclonal antibody (rMAb) therapeutics
`are exemplars of translational medicine. The rMAbs cur-
`rently licensed represent a significant success in terms
`of the clinical benefit delivered and the revenue gener-
`ated in the biopharmaceutical industry. Additionally, it
`is estimated that ~30% of new drugs that are likely to be
`licensed during the next decade will be based on anti-
`body products1–3. High-volume production of these large
`biological molecules, combined with the maintenance
`of their structural and functional fidelity, results in high
`costs, which can limit their availability to patients owing
`to the strain that it puts on national and private health
`budgets. Although the hallmark of an antibody is its
`specificity for the target antigen, the antigen–antibody
`complex formed must also be able to be removed and
`destroyed. The humoral antibody response can be
`mediated by antibodies from nine antibody classes
`and subclasses, each of which has unique mechanisms
`of action (known as effector functions). To date, all
`licensed rMAbs have been of the immunoglobulin G
`(IgG) class; however, each of the four human IgG sub-
`classes also exhibits unique effector functions. Therefore,
`when developing an rMAb therapeutic, it is important
`to select the IgG subclass that is anticipated to have the
`most potent activity for a given disease indication. The
`presence of oligosaccharides attached at a single site on
`each of the IgG heavy chains is essential for the antibody’s
`effector functions, and efficacy can vary depending on the
`particular oligosaccharide that is attached. Methods that
`allow the production of rMAbs bearing homogeneous
`
`oligosaccharides (glycoforms) are now becoming available.
`It is hoped that delivery of rMAbs that are optimized for
`specificity and effector functions, together with patient
`profiling, will have a significant impact on the cost of
`treatment.
`Human IgG isolated from normal serum is com-
`prised of multiple glycoforms owing to the addition of
`diverse complex diantennary oligosaccharides in the IgG Fc
`(crystallizable fragment) region. The presence or absence
`of IgG Fc oligosaccharides does not affect antigen binding
`but has a profound effect on the biological mechanisms
`that are activated by the immune complexes formed. Basic
`research has allowed the generation of a series of homo-
`geneous antibody glycoforms and demonstrated signifi-
`cant differences in their effector-function profiles and/or
`efficacy. Production vehicles, including mammalian cells,
`yeasts and plants, have been engineered to maximize the
`production of selected IgG glycoforms. The biopharma-
`ceutical industry now faces the challenge of translating
`these laboratory protocols into the manufacture of selected
`homogeneous antibody glycoforms. In contrast to modu-
`lation of effector functions by protein engineering, a major
`benefit of this optimization approach is that production of
`an rMAb of a single naturally occurring glycoform does
`not contribute to potential immunogenicity.
`
`The development of antibody therapeutics
`The original technology for generating MAbs of defined
`specificity was developed in the mouse and immedi-
`ately applied to the generation of MAbs with specificity
`
`Diantennary
`oligosaccharide
`An oligosaccharide that
`has two mannose ‘arms’;
`as opposed to triantennary,
`tetra-antennary and so on.
`
`Division of Immunity
`and Infection, University
`of Birmingham, School of
`Medicine, Edgbaston,
`Birmingham, B15 2TT, UK.
`e-mail: r.jefferis@bham.ac.uk
`doi:10.1038/nrd2804
`
`226 | MARcH 2009 | VoluMe 8
`
` www.nature.com/reviews/drugdisc
`
`© 2009 Macmillan Publishers Limited. All rights reserved
`
`1 of 9
`
`Fresenius Kabi
`Exhibit 1022
`
`

`

`R E V I E W S
`
`for human proteins4. Although early clinical success
`was reported for the treatment of an acute episode of
`kidney graft rejection with mouse IgG2a antibodies
`reactive against the cell-surface protein cD3, exposing
`humans to mouse antibodies provokes a human anti-
`mouse antibody (HAMA) response. The application
`of genetic engineering to the development of chimeric
`mouse–human rMAbs and, later, ‘humanized’ and/or
`‘fully human’ rMAbs promised reduced immunogenicity;
`however, a proportion of patients develop human anti-
`chimeric antibodies or human anti-human antibodies,
`respectively, to these antibodies5. Such antibody
`responses can prejudice treatment if they are neutral-
`izing, if they result in clearance of the therapeutic as
`immune complexes or if they sensitize the patient for
`severe reactions on re-exposure5–8. Importantly, these
`technologies allow the human antibody constant region
`isotype — which determines the effector functions acti-
`vated by immune complexes — to be selected based on
`its perceived suitability for a given disease indication
`in vivo9–11. To date, all approved rMAbs have been of the
`human IgG isotype, and they have predominantly been
`of the IgG1 subclass.
`The potential for immunogenicity imposes a demand
`for vehicles that can deliver products with absolute struc-
`tural fidelity, including appropriate post-translational
`modifications. However, inherent to the uniqueness of
`an individual antibody product is an inevitable potential
`for immunogenicity. The required fidelity cannot be
`realized using current production vehicles — that is,
`chinese hamster ovary (cHo) and mouse (NS0 and
`Sp2/0) cell lines. consequently, these production cell
`lines are being engineered for both product quantity and
`quality. A particular focus is glycosylation, as it has been
`shown that biological activities vary between different
`natural glycoforms and that non-natural glycoforms can
`be immunogenic.
`
`The basic structure of human IgG antibodies
`In humans the IgG antibody class predominates in the
`blood and equilibrates with the extravascular space;
`these characteristics are key to the design of antibody
`therapeutics with appropriate pharmacodynamics. In a
`natural immune response the pathogen–IgG immune
`complexes that are formed activate a wide range of
`effector functions, resulting in the killing, removal and
`destruction of the pathogen. Four subclasses of human
`IgG are defined and enumerated according to their rela-
`tive concentrations in normal serum: IgG1, IgG2, IgG3
`and IgG4, which respectively account for approximately
`60%, 25%, 10% and 5% of serum IgG. each IgG sub-
`class exhibits a unique profile of effector functions when
`evaluated by in vitro assays9–11.
`In its simplest form an individual IgG molecule is
`composed of two identical light chains and two identical
`heavy chains, which are in turn comprised of repeating
`structural motifs (homology regions) of ~110 amino-
`acid residues. The tertiary structure of each homology
`region defines the immunoglobulin fold or domain9–11.
`Domains of the light and heavy chains pair in covalent
`and non-covalent association to form three independent
`
`protein moieties connected through a flexible linker (the
`hinge region) (FIG. 1). Two of these moieties, referred to
`as Fab (antigen-binding fragment) regions, are of identi-
`cal structure and each form the same specific antigen-
`binding site; the third, the Fc, forms interaction sites for
`ligands that activate clearance and transport mecha-
`nisms. These effector ligands include three structurally
`homologous cellular Fc receptor types (FcγRI, FcγRII
`and FcγRIII), the c1q component of complement and
`the neonatal Fc receptor (FcRn)9–14; the latter influences
`the catabolic half-life of the antibody and its transport
`into the extravascular space and across the placenta.
`Activation of Fc receptors and of the c1q component
`of complement initiates inflammatory cascades that
`combat and resolve episodes of infection. These activi-
`ties are crucially dependent on IgG Fc N-linked glyco-
`sylation and vary between antibody glycoforms9–20. By
`contrast, binding to FcRn, and hence the catabolic half-
`life, is not dependent on antibody glycoform10; data are
`not currently available for transport across the human
`placenta.
`The IgG Fc region is a homodimer comprised of
`covalent inter-heavy chain disulphide-bonded hinge
`regions and non-covalently paired cH3 domains;
`the cH2 domains are glycosylated through covalent
`attachment of oligosaccharide at Asn297. X-ray crys-
`tallographic analysis reveals that the structure of the
`oligosaccharide is defined (as opposed to undefined
`owing to continuous mobility), is integral to the IgG Fc
`structure and forms multiple non-covalent interactions
`with the protein surface of the cH2 domain; thus, the
`conformation of the IgG Fc glycoprotein moiety results
`from reciprocal interactions between the protein and the
`oligosaccharides10,20,21. There is evidence that interaction
`sites on IgG Fc for FcγRI, FcγRII, FcγRIII and c1q
`effector ligands are comprised principally of only the
`protein moiety; however, generation of the interaction
`sites for these ligands is dependent on IgG Fc protein–
`oligosaccharide interactions. Thus, effector mechanisms
`mediated through FcγRI, FcγRII, FcγRIII and c1q are
`severely compromised or ablated for aglycosylated or
`deglycosylated forms of IgG9–20.
`It is generally reported that IgG Fc oligosaccharides
`of normal human IgG are predominantly devoid of ter-
`minal sialic acid residues21–24; however, recent studies
`have suggested that IgG Fc sialylated molecules (<10%
`of total IgG Fc molecules) may have beneficial prop-
`erties. Small but possibly significant differences in the
`affinities of sialylated and non-sialylated IgG Fc for
`FcγR have been reported25,26; for example, in a com-
`plex mouse model of inflammation, polyclonal human
`IgGs bearing Fc sialyl sugar residues were shown to be
`anti-inflammatory27,28. This anti-inflammatory activity
`was proposed to be mediated by an FcγR-independent
`mechanism. It is thought that this model might reflect
`the situation in human patients in which high doses of
`polyclonal IgG alleviate the symptoms of inflammatory
`autoimmune conditions — that is, they exhibit anti-
`inflammatory activity. We await the results of further
`studies of this phenomenon in experimental models that
`are more relevant to humans.
`
`NATuRe ReVIeWS | Drug Discovery
`
` VoluMe 8 | MARcH 2009 | 227
`
`© 2009 Macmillan Publishers Limited. All rights reserved
`
`2 of 9
`
`Fresenius Kabi
`Exhibit 1022
`
`

`

`R E V I E W S
`
`Fab
`
`Fab antigen-
`binding site
`
`Hinge
`
`Asn297
`
`Oligosaccharide
`
`CH2 domain
`
`CH3 domain
`
`Fc effector functions
`
`Figure 1 | The α-carbon backbone structure of the immunoglobulin g (igg) molecule.
`This structural representation illustrates the sequence of antiparallel β-pleated sheet
`Nature Reviews | Drug Discovery
`domains that constitute the immunoglobulin fold. The oligosaccharide is integral to the
`protein structure and has a defined conformation. CH, constant heavy; Fab, antigen-
`binding fragment; Fc, crystallizable fragment.
`
`Glycoform profiles of serum-derived IgG
`The oligosaccharides present in the IgG Fc of normal
`poly clonal IgG, attached at Asn297, are of the complex
`diantennary type and are comprised of a core heptasaccha-
`ride with variable addition of fucose, galactose, bisecting
`N-acetylglucosamine and sialic acid (FIG. 2A); sialylation is
`modest, with <10% of structures being monosialylated or
`disialylated11,21–24. The relative yields and structures of the
`neutral oligosaccharides that may be released from normal
`IgG Fc by incubation with the glycosidase PNGase F are
`respectively shown in FIG. 2b,c. each heavy chain may
`bear one of a total of 32 unique oligosaccharides, and
`random pairing of heavy-chain glycoforms could gen-
`erate ~500 glycoforms. Given the paucity of sialylation,
`neutral oligosaccharides predominate, but they alone
`allow for the generation of 128 unique glycoforms; the
`possible number of combinations (16 × 16) is divided by
`2 because of the symmetry of the molecule.
`A shorthand system of nomenclature for the oligo-
`saccharides uses G0, G1 and G2 for oligosaccharides
`bearing zero, one or two galactose residues, respectively.
`When fucose is present G0F, G1F or G2F is used; when
`bisecting N-acetylglucosamine is present a B is added —
`for example, G0B or G0BF. The glycoform of the whole
`molecule is the oligosaccharides present on each heavy
`chain and may be represented as (G0–G0), (G0–G0F),
`(G0F–G0F), (G0–G1), (G0F–G1) and so on.
`
`Differential scanning
`micro-calorimetry
`A thermoanalytic technique
`in which the difference in the
`amount of heat required to
`increase the temperature of
`a sample and a reference is
`measured as a function of
`temperature.
`
`The glycoform profile of normal polyclonal IgG is
`subject to variation with age, over the term of pregnancy
`and in a number of inflammatory conditions11,22–24. The
`major change reported is the extent of galactosylation,
`usually reported as the percentage of [G0 + G0F] oligo-
`saccharides released. Thus, the levels of [G0 + G0F]
`reported for young children and elderly adults exceed
`those reported for adults between 20 and 50 years of age.
`Increased levels of [G0 + G0F] are reported in a number
`of inflammatory diseases that have an autoimmune com-
`ponent — for example, rheumatoid arthritis, crohn’s
`disease and vasculitis. Although they do not equate to
`the aetiology of the disease, these changes may repre-
`sent ‘acute-phase reactants’ that reflect inflammation
`and may contribute to an exacerbation of inflamma-
`tion (see below). Analysis of monoclonal human IgG
`proteins isolated from the sera of patients with multiple
`myeloma revealed individual (and unique) glycoform
`profiles, including major differences in the levels of
`galactosylation, fucosylation and the addition of bisect-
`ing N-acetylglucosamine residues29,30. It may be antici-
`pated, therefore, that the IgG response to an individual
`antigen (pathogen) may be comprised of predominant
`glycoforms.
`Approximately 30% of polyclonal human IgG molecules
`bear N-linked oligosaccharides in the IgG Fab region, in
`addition to those attached at the conserved glycosylation
`site at Asn297 in the IgG Fc11,31–33. When present they
`are attached to the variable regions of the kappa (Vκ) or
`lambda (Vλ) light chains, to the heavy (VH) chains or to
`both. In the immunoglobulin sequence database ~20% of
`the IgG variable regions have N-linked glycosylation con-
`sensus sequences (Asn-X-Thr/Ser, where X can be any
`amino acid except proline). Interestingly, these consen-
`sus sequences are mostly not encoded in the germ line;
`rather, they mostly result from somatic mutation — sug-
`gesting positive selection for improved antigen binding.
`Analysis of polyclonal human IgG Fab reveals the pres-
`ence of diantennary oligosaccharides that are extensively
`galactosylated and substantially sialylated, in contrast
`to the oligosaccharides released from IgG Fc11,31–33.
`The functional significance of IgG Fab glycosylation of
`polyclonal IgG has not been fully evaluated, but data
`emerging for monoclonal antibodies suggest that Vκ,
`Vλ or VH glycosylation can have a neutral, positive or
`negative influence on antigen binding11.
`It is generally observed that the oligosaccharide
`present in glycoproteins contributes to their solubility
`and stability. The influence of glycosylation on the ther-
`mal stability of human IgG1 Fc was demonstrated for a
`series of truncated glycoforms using differential scanning
`micro-calorimetry20,34. It is possible, therefore, that IgG
`Fab glycosylation may be beneficial when form ulating
`IgG therapeutics at concentrations of 100–150 mg
`per ml. Such high-concentration formulations allow the
`development of self-administration protocols, and the
`high serum concentrations that are achieved can reduce
`dosing intervals, resulting in reduced cost of treatment.
`controlling glycoform fidelity at two sites (Fc and Fab)
`offers a further challenge to the biopharmaceutical
`industry.
`
`228 | MARcH 2009 | VoluMe 8
`
` www.nature.com/reviews/drugdisc
`
`© 2009 Macmillan Publishers Limited. All rights reserved
`
`3 of 9
`
`Fresenius Kabi
`Exhibit 1022
`
`

`

`R E V I E W S
`
`Glycoforms of recombinant IgG antibodies
`As Fc receptor and complement binding, and activa-
`tion, are essentially ablated for aglycosylated IgG Fc,
`recombinant antibody therapeutics should have fully
`occupied glycosylation sites. The most widely used pro-
`duction cell lines are cHo, NS0 and Sp2/0, in various
`engineered forms35,36. These cell lines produce IgG Fc
`glycoforms bearing the top eight oligosaccharides shown
`in FIG. 2c; however, the proportion of galactosylated and
`non-fucosylated IgG Fc is low relative to normal IgG Fc
`(FIG. 2b). Although these cell lines do not add oligosac-
`charides containing bisecting N-acetylglucosamine
`residues, they do add sugars that are not present in
`normal serum-derived IgG, and these can be immuno-
`genic. A particular concern is the addition, by NS0 and
`Sp2/0 cells, of galactose in α(1–3) linkage to galactose
`linked β(1–4) to the N-acetylglucosamine residues (gal
`α(1–3) gal). Humans and higher primates do not have
`a functional gene encoding the transferase that adds
`galactose in α(1–3) linkage; however, owing to continual
`environmental exposure, humans have an IgG antibody
`that is specific for this antigen37. Similarly, cHo, NS0 and
`Sp2/0 cells can add an α(2–3)-linked N-glycylneuraminic
`acid that is not present in humans and that might also be
`immunogenic38, and cHo cells can add N-acetyl neuraminic
`acid in α(2–3) linkage rather than the human α(2–6)
`linkage.
`The licensed antibody therapeutic cetuximab (erbitux)
`bears an N-linked oligosaccharide at Asn88 of the VH
`region; interestingly there is also a glycosylation motif at
`Asn41 of the Vl region, but it is not occupied39. Analysis
`of the IgG Fc and IgG Fab oligosaccharides of cetuximab,
`which is produced by Sp2/0 cells, revealed highly signifi-
`cant differences in composition. Whereas the IgG Fc oligo-
`saccharides are typical — that is, they are predominantly
`fucosylated non-galactosylated diantennary oligosaccha-
`rides — the IgG Fab oligosaccharides are extremely hetero-
`geneous and include complex diantennary and hybrid
`oligosaccharides as well as galactose in α(1–3) linkage to
`galactose and N-glycylneuraminic acid residues.
`A recent study reported that 25 of 76 patients
`treated with cetuximab had hypersensitivity reactions
`to the drug, and Ige antibodies against gal α(1–3) gal
`were detected in 17 pre-treatment samples from these
`patients40. Interestingly, environmental factors seem
`to influence the development of Ige gal α(1–3) gal
`responses, as the incidence varied significantly between
`treatment centres and predominant infectious agents
`present in local environments.
`A detailed analysis of the glycoforms of a humanized
`IgG rMAb, which is also expressed in Sp2/0 cells, bearing
`oligosaccharides at Asn56 of the VH and Asn297
`revealed the expected IgG Fc profile of predominantly
`G0F oligosaccharides. However, 11 oligosaccharides
`were released from the IgG Fab, including diantennary
`and triantennary oligosaccharides bearing gal α(1–3)
`gal, N-glycylneuraminic acid and N-acetylgalactosamine
`residues41. The consistent observation of higher levels of
`galactosylation and sialylation for IgG Fab N-linked oligo-
`saccharides than for IgG Fc N-linked oligosaccharides is
`thought to reflect increased exposure and/or accessibility
`
`of the glycosylation sites on Fab fragments. Given the
`findings of these studies and the potential for NS0
`and Sp2/0 cells to add gal α(1–3) gal residues, it seems
`likely that these cells will cease to be developed for
`therapeutic antibody/glycoprotein production, unless
`they are engineered to inactivate the gal α(1–3) and
`N-glycylneuraminic acid transferases.
`The double challenge of producing rMAbs with appro-
`priately glycosylated IgG Fc and IgG Fab sites has led to
`some companies engineering out VH or Vl glycosylation
`motifs when present in candidate rMAbs42; however,
`present reports suggest that cHo cells can glycosylate
`VH and/or Vl motifs in a manner similar to that
`observed for normal polyclonal IgG43.
`
`IgG Fc glycoforms: structure and function
`Structure. Protein engineering, using alanine scanning,
`has been used to map amino-acid residues deemed
`to be crucial for FcγR and c1q binding. These studies
`map the binding site for all four of these ligands to the
`hinge-proximal and lower hinge region of the cH2
`domain14,16,44–48. X-ray crystallographic analysis suggests
`that the lower hinge region of IgG Fc is mobile and with-
`out defined structure21, which might seem incompatible
`with the suggestion that this region is directly involved
`in generating structurally distinct interaction sites for
`the FcγR and c1q ligands. However, the suggestion
`may be rationalized by the proposal that this region is
`comprised of multiple conformers in equilibrium, with
`individual conformers being compatible with specific
`ligand recognition11,14. X-ray crystallographic analysis
`of IgG Fc in complex with a soluble recombinant form
`of the FcγRIIIb receptor provides proof of the direct
`involvement of the cH2 lower hinge regions and hinge-
`proximal regions in ligand binding45,46. Residues in the
`hinge region that do not form part of a discrete struc-
`ture in the ‘free’ IgG Fc are shown to interface with the
`receptor and assume discrete conformations with ligand
`bound. Both heavy chains are involved in forming an
`asymmetric binding site. consequently, monomeric IgG
`is univalent for the FcγR; if monomeric IgG were divalent
`for FcγR it could cross-link cellular receptors and hence
`constantly activate inflammatory reactions.
`Study of a series of truncated IgG Fc glycoforms revealed
`that non-covalent interactions between sugar residues
`of the α(1–6) arm of the IgG Fc oligosaccharide and the
`inner protein surface of the cH2 domain determine the
`overall conformation of the protein–oligosaccharide
`complex. If the oligosaccharide moiety is progressively
`truncated a closing of the ‘horseshoe-like’ structure is
`observed20,34,48. Reduced structural integrity and function-
`ality was demonstrated when the oligosaccharide was
`truncated to only the initial trisaccharide20,34, and func-
`tionality was completely ablated when the oligosaccharide
`was further truncated to only the initial covalently bound
`N-acetylglucosamine residue48.
`
`Function: ADCC. The effectiveness of rMAbs in oncol-
`ogy depends on sensitizing target cells for subsequent
`killing by the mechanisms of antibody-dependent cell ular
`cytotoxicity (ADcc) or complement-dependent cytotoxicity
`
`Alanine scanning
`A genetic manipulation that
`sequentially replaces wild-type
`amino acids with alanine to
`determine the impact on
`the protein’s structure and
`function.
`
`Conformer
`A discrete, defined
`conformation in three-
`dimensional space.
`
`Antibody-dependent
`cellular cytotoxicity
`(ADcc). cell death that results
`when the Fc fragment of an
`antibody bound to a target cell
`interacts with Fc receptors
`on monocytes, macrophages
`or natural killer cells that are
`consequently activated to kill
`the target.
`
`NATuRe ReVIeWS | Drug Discovery
`
` VoluMe 8 | MARcH 2009 | 229
`
`© 2009 Macmillan Publishers Limited. All rights reserved
`
`4 of 9
`
`Fresenius Kabi
`Exhibit 1022
`
`

`

`R E V I E W S
`
`(cDc) and/or the induction of apoptosis. It is unequivo-
`cally established that ADcc and cDc are dependent
`on appropriate glycosylation of the rMAb11,16–20, where as
`induction of apoptosis may require only cross-linking
`of cell-surface antigens. Glycosylation has been a focus
`
`of interest for the biopharmaceutical industry for the
`past several years, and cell lines have been engineered
`in efforts to optimize antibody products for ADcc and
`cDc by the differential addition of fucose, galactose,
`bisecting N-acetylglucosamine and sialic acid.
`
`A
`
`–Gln–Tyr–Asn297–Ser–Thr–Tyr–Arg–
`
`B
`
`GlcNAc
`
`Fuc
`
`GlcNAc
`
`α(1–6)
`
`Man
`
`α(1–3)
`
`Man
`
`GlcNAc
`
`Gal
`
`Neu5Ac
`
`GlcNAc
`
`Man
`
`GlcNAc
`
`Gal
`
`Neu5Ac
`
`h
`
`F
`
`M–GN–GN–pr
`
`GN–M
`
`GN–M
`
`' /
`
`e
`
`f
`
`g
`
`GN–M
`
`GN–M
`
`M–GN–GN–pr
`
`' /
`
`d
`
`c
`
`a
`
`b
`
`10
`
`20
`
`30
`Elution time (minutes)
`
`M–GN–GN
`
`Ca
`
`GN–M
`
`/
`
`GN–M
`
`c
`
`GN–M
`
`M–GN–GN
`
`b
`
`G–GN–M
`
`/
`
`GN–M
`
`d
`
`G–GN–M
`
`e
`
`F
`
`M–GN–GN–pr
`
`GN–M
`GN
`GN–M
`
`' 7
`
`p
`
`n
`
`o
`
`50
`
`m
`
`40
`
`M–GN–GN
`
`M–GN–GN
`
`F
`
`F
`
`M–GN–GN
`
`G–GN–M
`
`e
`
`GN–M
`
`GN–M
`
`g
`
`GN–M
`
`G–GN–M
`
`/
`
`/
`
`/
`
`i
`
`GN–M
`M–GN–GN
`GN
`7
`GN–M
`
`k
`
`GN–M
`M–GN–GN
`GN
`7
`G–GN–M
`
`F
`GN–M
`M–GN–GN
`GN
`7
`GN–M
`
`'
`'
`'
`'
`'
`'
`'
`'
`
`m
`
`o
`
`F
`
`GN–M
`M–GN–GN
`GN
`7
`G–GN–M
`
`'
`'
`'
`'
`'
`'
`'
`'
`
`M–GN–GN
`
`M–GN–GN
`
`F
`
`F
`
`M–GN–GN
`
`M–GN–GN
`
`M–GN–GN
`
`F
`M–GN–GN
`
`F
`
`M–GN–GN
`
`G–GN–M
`
`f
`
`G–GN–M
`
`GN–M
`
`h
`
`G–GN–M
`
`G–GN–M
`
`/
`
`/
`
`/
`
`j
`
`l
`
`n
`
`p
`
`G–GN–M
`GN
`7
`GN–M
`
`G–GN–M
`GN
`7
`G–GN–M
`
`G–GN–M
`GN
`7
`GN–M
`
`G–GN–M
`GN
`7
`G–GN–M
`
`a
`
`10
`
`20
`
`30
`Elution time (minutes)
`
`40
`
`50
`
`Complement-dependent
`cytotoxicity
`(cDc). cell death that results
`when the IgG Fc regions of an
`antibody bound to a target cell
`activate the c1 component
`of complement, initiating a
`cascade of reactions that lead
`to the formation of a complex
`that disrupts the cell membrane.
`
`Figure 2 | igg Fc diantennary-complex oligosaccharide composition. a | The oligosaccharides present in the
`Nature Reviews | Drug Discovery
`immunoglobulin G (IgG) Fc (crystallizable fragment) region of normal polyclonal IgG attach at Asn297. They form a
`diantennary complex comprised of a core heptasaccharide (sugars in blue boxes) and outer arms (sugars in red boxes)
`constructed by variable addition of fucose (Fuc), galactose (Gal), bisecting N-acetylglucosamine (GlcNAc), sialic acid and
`N-acetylneuraminic acid (Neu5Ac). b | A high-performance liquid chromatography profile of the complex diantennary
`oligosaccharide structures released from normal human IgG Fc24 (top) and from recombinant IgG produced in Chinese
`hamster ovary cells (bottom). c | The potential library of all neutral complex diantennary oligosaccharides that can be
`released from normal human IgG Fc. The oligosaccharides are labelled a–p, corresponding with the peaks in part b.
`Oligosaccharides i–l are collectively <3% of the total amount and therefore cannot be identified in the profile for human
`serum IgG. F, fucose; G, galactose; GN, N-acetylglucosamine; M, mannose; Man, mannose; Pr, protein.
`
`230 | MARcH 2009 | VoluMe 8
`
` www.nature.com/reviews/drugdisc
`
`© 2009 Macmillan Publishers Limited. All rights reserved
`
`5 of 9
`
`Fresenius Kabi
`Exhibit 1022
`
`

`

`R E V I E W S
`
`The first rMAb to be licensed for the treatment of can-
`cer was rituximab (Rituxan), a chimeric mouse–human
`IgG1 antibody against cD20 produced in cHo cells. It is
`approved for the treatment of non-Hodgkin’s lymphoma,
`in which the target is neoplastic cD20-expressing B lym-
`phocytes49,50. Rituximab was shown to sensitize B cells for
`killing by ADcc, cDc and the induction of apoptosis
`in vitro. Parameters that influence killing by ADcc and
`cDc include epitope specificity and the level of expres-
`sion of the cD20 antigen and the decay-acceleration
`factors cD55 and cD59. Rituximab proved inefficacious
`in the treatment of other B-cell malignancies in which
`the level of expression of the cD20 antigen is low — for
`example, chronic lymphocytic leukaemia. This rMAb has
`been subject to studies designed to develop glycoforms
`with enhanced effector functions.
`extensive investigation of the mechanism(s) by
`which rituximab can kill cD20-positive lymphocytes
`in vitro and in vivo has established that recruitment and
`activation of FcγRIIIa-expressing cells — for example,
`natural killer (NK) cells — is a dominant pathway49,50.
`Thus, target cD20-positive tumour cells that are sen-
`sitized by bound rituximab engage FcγRIIIa receptors
`expressed on NK cells, with subsequent activation and
`release of granzymes. The efficacy of NK cell recruitment
`and activation seems to be directly determined by the
`affinity of the IgG Fc for FcγRIIIa which, in turn, varies
`with the presence or absence of fucose on the oligosac-
`charide of the IgG Fc. Present rMAb production vehicles
`(cHo, NS0 and Sp2/0 cells) generate >90% fucosylated
`forms of IgG Fc. The cHo cell line has been engineered
`to produce non-fucosylated IgG Fc, and new glycoforms
`of several licensed rMAbs are nearing the clinic.
`An early report compared the ability of alemtuzumab
`(campath-1H), a chimeric IgG1 rMAb against cD52,
`produced from cHo, NS0 and YB2/0 cells to mediate
`ADcc, and concluded that alemtuzumab produced
`in the rat YB2/0 cell line had the highest efficacy. The
`only structural difference between the products was the
`presence of complex oligosaccharides bearing bisecting
`N-acetylglucosamine residues51 in the YB2/0 antibody.
`Subsequently, two groups engineered the cHo cell
`line to express the GnTIII enzyme that adds bisecting
`N-acetylglucosamine residues, and demonstrated that
`ADcc improved by two or more orders of magnitude
`for a rituximab-like product52,53. enhanced ADcc was
`also reported for rituximab produced from cHo cells in
`which the α(1–6) fucosyl transferase enzyme had been
`knocked out18,36,54. It is now appreciated that the addition
`of the N-acetylglucosamine sugar residue is a relatively
`early event in glycoprotein processing in the Golgi appa-
`ratus, and the presence of this residue inhibits the later
`addition of fucose to the primary N-acetylglucosamine
`residue19. The absence of fucose from the primary
`N-acetylglucosamine results in the IgG1 antibody having
`increased binding affinity for the FcγRIIIa receptor,
`with consequent increased efficacy of NK cell-mediated
`ADcc55. Thus, it seems that it is the absence of fucose,
`not the presence of bisecting N-acetylglucosamine
`per se, that results in enhanced ADcc17. The enhanced
`ADcc observed for afucosylated IgG Fc results, in part,
`
`from the increased affinity for FcγRIIIa — this allows
`the afucosylated IgG Fc to overcome the competition
`from high concentrations of fucosylated IgG in nor-
`mal serum17,54–57. The findings reported for rituximab
`have recently been extended with the demonstration
`of improved ADcc for afucosylated trastuzumab
`(Herceptin) in a whole-blood assay 58. The presence
`or absence of fucose has been shown to have minimal
`impact on effector functions mediated through other
`FcγRs or complement.
`Non-fucosylated forms of therapeutic antibodies such
`as rituximab and trastuzumab may be anticipated to have
`clinical advantage over the fucosylated forms owing to the
`increased sensitization of target cells to NK cell-mediated
`ADcc; however, it is important to approach this goal
`with caution. These reagents should be treated as new
`therapeutics, as their increased potency could result
`in enhanced cell killing at first dose with consequent
`increased release of cytokines and possible adverse reac-
`tions. In addition, the presumed lesser need for target
`sensitization could result in increased killing of normal
`cells that express the target antigen.
`The industry is beginning to respond to the perceived
`advantages of non-fucosylated antibodies for cancer
`therapies, but a recent report suggests that there are
`contrary indications59. Two preparations of an antibody
`specific for epidermal growth factor receptor (eGFR)
`that differed in their level of fucosylation were shown
`to be equivalent in a whole-blood assay. However, when
`ADcc mediated by isolated peripheral blood mono-
`nuclear cells (PBMcs) and polymorphonuclear neutro-
`phils (PMNs) were compared, it was shown that low
`levels of fucosylation favoured PBMc-mediated ADcc
`whereas high levels of fucosylation favoured PMN-
`mediated ADcc59. It was further