Different Adaptations of IgG Effector Function in Human and Nonhuman Primates and Implications for Therapeutic Antibody Treatment

The complex interplay of Fc receptors and Abs has been an area of scientific interest for the last decades. More than 30 years ago, we and others could show that different IgG Fc receptors (FcγR) exist that are highly selective in terms of subclass specificity and affinity (1, 2). The high complexity of this system is increased by the existence of activating and inhibitory FcγRs (3) and the fact that the different effector cell populations are decorated with different types and different densities of Fc receptors (4). Interaction of monomeric Abs or Ab immune complexes with Fc receptors triggers Ab-dependent cellular cytotoxicity (ADCC) and Ab-dependent cellular phagocytosis, whereas interaction with complement results in complement-dependent cytotoxicity (CDC) (5). From a signaling standpoint, activation of the cellular effector functions is mediated by ITAM-containing FcγRs interaction with IgG immune complexes (e.g., opsonized bacteria or parasites). These activities are counterregulated by the ITIM-containing FcγRIIb (CD32b), which plays a critical role in modulating IgG-mediated effector responses, such as in humoral immunity (6, 7). The relative densities of inhibitory (i.e., FcγRIIb) versus activating FcγRs, and in particular their corresponding selectivity and affinity toward different IgG isotypes, determine whether an Ab will generate an activating proinflammatory or an inhibitory anti-inflammatory response. For a given IgG isotype, Ravetch and Nimmerjahn (8) have found that the ratio of affinities of the activating ITAM-bearing FcγRs to the inhibitory FcγRIIb is a key determining factor. In the case of therapeutic Abs, IgG/FcγR interactions contribute significantly to both therapeutic effect and toxicity of these drugs (5). Due to their proximal evolutionary relationship to humans, the most appropriate preclinical model to analyze the therapeutic effect and potential toxicity of human Abs are nonhuman primates (NHP), in particular the cynomolgus monkey (9). However, recent clinical trial results have questioned whether NHP are an appropriate model. A striking example is the outcome of the TGN1412 trial in 2006, in which an anti-CD28 Ab (TGN1412) did not show severe adverse side effects in NHP, but generated a dramatic cytokine storm in all six treated human subjects (10). Possible contributions to this differential response in NHP and humans in this trial include differences in the signaling domain of CD28, the loss of Siglec-5 expression during human evolution (11, 12), or a different expression pattern of CD28 on a specific T cell subset (13). However, because triggering of a cytokine storm by TGN1412 is dependent on FcγR interaction (14), it has also been speculated that the human IgG4 Fc portion of the TGN1412 Ab might bind less efficiently to monkey than to human FcγRs (15).

In light of the growing use of IgG mAb as biotherapeutic compounds, we further investigated the interaction of human and cynomolgus IgG subtypes and FcγRs to assess the appropriateness of cynomolgus monkeys in evaluating Fc-dependent efficacy and toxicity profiles of therapeutic IgGs. Our study revealed that the effector function of humans and NHP is an adaptable system, which is shaped by species-specific constraints. We have found clear differences in FcγR expression on granulocytes, allotypic variation of IgG4 Fab arm exchange, and enhanced activity of IgG2 and IgG4 subclasses in cynomolgus monkeys compared with the human system, which may have been evolved by different needs in phagocytosis and neutralization of bacteria and parasites. Interestingly, this overall enhanced activation function in cynomolgus monkeys is counterbalanced by a stronger inhibitory capacity of cynomolgus FcγRIIb toward IgG2, an isotype with preference for repetitive carbohydrate Ags on bacteria (16, 17), and toward IgG4, which (together with IgE) is generated in antiparasite immunity and responses to allergens. In contrast, IgG1 and IgG3 subclasses, which are mainly involved in other types of the humoral immune response, show a very similar pattern of FcγR binding and effector function in the two species. To our knowledge, this study describes for the first time the complete set of FcγR/IgG interactions and functional consequences in the cynomolgus monkey in comparison with the human system, and will therefore be very helpful in interpreting the data generated with therapeutic Abs in NHP and their translation to humans.

Extracellular domains of human and cynomolgus FcγR were cloned into expression vectors harboring a 4-aa purification tag (4APP; Novartis) and an Avi biotinylation tag (GLNDIFEAQKIEWHE; Avidity), expressed in 293 cells and purified with anti-4APP affinity chromatography. The human receptors were expressed with a CD33 signal leader (P20138, 1-16) instead of the natural leader sequence to enhance protein expression. Receptor proteins expressed with the natural leader and the CD33 leader showed identical amino acid sequences as analyzed by mass spectrometry. The following receptor sequences were synthesized by Geneart: human FcγRIIIA (CD16a)158V (Uniprot: P08637, 17-199, aa 158 = V), human FcγRIIIA 158F (Uniprot: P08637, 17-199), cynomolgus FcγRIIIA, human FcγRIIA (CD32a) 131H (Uniprot: NP_067674, 34-217), human FcγRIIA 131R (Uniprot: NP_067674, 34-217, aa 131 = R), cynomolgus FcγRIIA (Uniprot: AAL92096, 1-210), human FcγRIIB (CD32b) (Uniprot: P31994, 42-223), and cynomolgus FcγRIIB (Uniprot: AAL92097, 1-226). Receptors were site directed biotinylated with BirA (Avidity). Extracellular domain of human FcγRI (CD64) protein was purchased from R&D Systems (Minneapolis, MN). Cynomolgus κ L chain (GenBank: JN984930) and γ H chain C regions (IgG1, GenBank: JN984927; IgG2, GenBank: JN984928; IgG4, GenBank: JN984929) were cloned from cynomolgus blood mRNA (primer κ L chain, 5′-ATCAAACGAGCTGTGGCTGCACCA-3′, 5′-ATTAACACTCTCCCCTGTTGA-3′; γ H chain, 5′-GCCTCCACCAAGGGCCCATCG-3′, 5′-TTTACCCGGAGACACGGAGAG-3′). Genomic DNA from the blood of 10 cynomolgus animals was sequenced to validate the γ H chain C regions (primer: 5′-GCCCAGCTCTGTCCCACACCG-3′, 5′-CCTGGGAAGTATGTACACGG-3′). Human H and L chain C regions and cynomolgus IgG3 were synthesized by Geneart, Uniprot accession numbers of human IgG1-4: P01857, P01859, P01860, and P01861; human κ L chain: P01834; cynomolgus IgG3, GenBank: DJ444798 (18). Uniprot database: http://www.uniprot.org/uniprot/; GenBank database: http://www.ncbi.nlm.nih.gov/nuccore. Human and cynomolgus H and L chain C regions were cloned directly after the anti-CD20 mAb Rituximab (19) or the anti-CD28 mAb TGN1412 (14) V region, expressed in 293 cells, and purified by protein G and size exclusion chromatography to a purity of >99% monomeric IgG. There was no sign of reaggregation during the course of the study, as assessed by analytical size exclusion chromatography. Analysis of the protein size under reducing and nonreducing conditions revealed human IgG4 half-molecule formation and cynomolgus IgG4 H-L chain dimer formation similar to a recent report (20), and all Abs showed comparable binding to CD20-expressing cells (Supplemental Fig. 1).

The 100-μl samples of whole blood were incubated for 30 min at 4°C in the dark with each of the FITC-conjugated Abs (dansyl-specific isotype control Ab clone 27–35 [mouse IgG2b, κ], 3D3-FITC, and FLI8.26-FITC [BD Biosciences]). After RBC lysis, cells were washed and analyzed by flow cytometry. Analysis was restricted to granulocytes gated by their forward light scatter versus side light scatter. The ratio of 3D3 minus isotype control geometric mean fluorescence intensity (ΔGMFI) versus FLI.8.26 ΔGMFI was used to determine the genotype of the different donors. Ratios are as follows: heterozygous donors, ∼0.5; CD32a^R/R131 homozygous donors, ∼1; CD32a^H/H131, ∼0.

For FcγRIII and FcγRII, site-directed biotinylated human and cynomolgus FcγRIII and FcγRII were bound to streptavidin sensor chips (General Electric), and the equilibrium-binding levels of the different Abs were analyzed. Equilibrium dissociation constants (K_D) were calculated by a 1:1 model. For FcγRI, the Rituximab-derived Abs were captured by an anti-rituximab idiotypic Ab (2842; AbD Serotec), and the binding to human and cynomolgus FcγRI was analyzed. Association (k_on) and dissociation (k_off) rate constants were calculated by a 1:1 model; K_D was calculated as k_off/k_on. K_A, the reciprocal value for the equilibrium dissociation constant, was calculated as 1/K_D.

For in vivo Fab arm exchange, 100 μg human rituximab-IgG4 and human TGN1412, or 100 μg cynomolgus rituximab-IgG4 and human TGN1412, or 100 μg human rituximab-IgG4 alone was injected into SCID mice, and Fab arm exchange was analyzed 48 h later. For in vitro Fab arm exchange, 1 μg human rituximab IgG4 was incubated for 24 h at 37°C with 1 μg of the different IgG4 variants of TGN1412 in the presence of 0.5 mM reduced glutathione (GSH). To analyze the generation of bispecific Abs, CD28-expressing Jurkat cells or CD20-expressing Raji cells were incubated with the serum of the injected animals or the Ab mixtures and stained with fluorescent anti-rituximab idiotype-specific Ab (MB2 A4; AbD Serotec) or CD28-Fc. To analyze the generation of a third Ab species with mass corresponding to a H-L chain dimer of each parental Ab, liquid chromatography-mass spectrometry (LC-MS) (Waters Acquity UPLC system coupled directly to a Waters Q_TOF Premier Mass Spectrometer) analysis of PNGase F (New England Biolabs)-treated Ab mixture was performed. The resulting N-deglycosylated sample was injected onto a MassPrep Micro desalting column (2.1 × 5 mm; Waters) heated at 80°C at a flow rate of 0.4 ml/min with a gradient of 2–90% solvent B. Mobile phase A: 2% acetonitrile, 0.1% formic acid; mobile phase B: acetonitrile with 0.1% formic acid. The electrospray ionization time-of-flight mass spectrometer was operated in the positive V-mode, under a source temperature of 120°C, a desolvation temperature of 250°C, a sample cone set at 40 V, and a capillary voltage of 1.5 kV. Mass spectra were acquired in the m/z range of 1000–6000. A 0.95-Hz scan rate with 0.05-s interscan delay was used to acquire data over the entire analysis. The deconvoluted electrospray ionization time-of-flight spectra using MaxEnt1 (Waters) are shown.

ADCC activity was assessed by the depletion of B cells. PBMCs were purified by Ficoll gradient, 100% Ficoll was used for the human blood separation, and 90% Ficoll plus 10% PBS for the cynomolgus blood separation. Human or cynomolgus PBMC were incubated for 1 h with the anti-CD20 Abs, stained for B cells with anti-CD40 (5C3; BioLegend) and anti-CD21 (B-ly4, BD Biosciences), and analyzed by flow cytometry. Killing was calculated as percentage of B cells without Ab to percentage of B cells with Ab.

FcγR fulfill diverse functions in the different immune cells from neutrophil Ab-mediated phagocytosis of bacteria to regulation of B cell activation.

Human granulocytes express only FcγRIIIb and FcγRIIa, and therefore become highly activated by IgG-opsonized bacteria or parasites. In contrast to humans, NHP do not express any form of FcγRIII on granulocytes (21). Surprisingly, we found instead a 3-fold higher expression of FcγRII on cynomolgus granulocytes in comparison with humans (Fig. 1). This is a specific difference in the FcγR expression pattern of neutrophils as other populations such as monocytes (Fig. 1), B cells, or NK cells (data not shown) show no differences in the expression pattern and levels between human and cynomolgus monkey. As FcγRIII has a different binding profile in comparison with FcγRII, especially for the IgG2 subclass (22), we decided to perform a complementary analysis of the whole IgG/FcγR interaction set in humans and the cynomolgus monkey.

Highly purified monomeric Abs with V_H and V_L derived from the CD20-binding Ab rituximab (19) recombinantly expressed with the four human and cynomolgus IgG subclasses were used to measure their binding against the different human and cynomolgus FcγR by surface plasmon resonance. In contrast to human IgG2 and IgG4 isotypes, which show weak interaction with the human and cynomolgus activating receptors FcγRI and FcγRIII, cynomolgus IgG2 and IgG4 showed substantial binding to corresponding FcγR of both species (Fig. 2). Human and cynomolgus FcγRI do not bind to human IgG2, but show nanomolar affinity to cynomolgus IgG2 (Fig. 2A, 2B, Supplemental Figs. 2, 3, Table I). The binding of FcγRI to cynomolgus IgG4 has a 10× higher affinity due to a slower off-rate in comparison with human IgG4 (Fig. 2B). Human and cynomolgus FcγRIII bind cynomolgus IgG2 and IgG4 with a 5–10× higher affinity than human IgG2 and IgG4 (Fig. 2D–G). Overall, human and cynomolgus FcγRI and FcγRIII show a qualitatively similar binding profile toward the different human and cynomolgus IgG isotypes (IgG1–IgG4). In humans, FcγRIII has an allotypic variation and FcγRIII-158V has a higher affinity than FcγRIII-158F (23). Interestingly, cynomolgus FcγRIII contains the more valine-like amino acid isoleucine at position 158 and accordingly shows a high-affinity phenotype (Table I). Distinct amino acid substitutions in the cynomolgus IgG sequences at the FcγR binding domains (24, 25) account for the different affinity of cynomolgus IgG2 and IgG4 (Fig. 3). The “silent” effector function of human IgG2 and IgG4 can be explained by differences to IgG1 in residues of the hinge and C_H2 domain that dramatically decrease their affinity for FcγRs and complement, therefore rendering them inactive for ADCC or CDC (26). This is not the case for cynomolgus IgG2 and IgG4, where these residues are identical to the full effector function bearing IgG1 and IgG3 isotypes (Fig. 3A). In summary, cynomolgus IgG2 and IgG4 show increased binding to human and cynomolgus FcγRIII and FcγRI in comparison with their human counterparts, whereas human and cynomolgus IgG1 and IgG3 show an almost identical binding to human and cynomolgus FcγRIII and FcγRI.

Interestingly, all the activating ITAM-bearing FcγRs of human and cynomolgus origin show almost identical binding profiles. One striking difference is that the cynomolgus inhibitory FcγRIIb binds IgG2 with substantially higher affinity than human FcγRIIb (Fig. 2C). This difference in binding is directly correlated with a different amino acid pattern in the Fc binding region of cynomolgus FcγRIIb (Fig. 3C). There is a polymorphism in the Fc binding region of the activating human FcγRIIa, which has a higher affinity to IgG2 if the amino acid at position 131 is a histidine. Accordingly, human FcγRIIa-131H and cynomolgus FcγRIIa (131H) show a higher affinity to IgG2 than human FcγRIIa 131R (Fig. 2C). In contrast to human FcγRIIb, in which the amino acid at position 131 is an arginine leading to low-affinity binding of IgG2, cynomolgus FcγRIIb has a high-affinity conferring histidine at this position (Fig. 3C). Accordingly, cynomolgus FcγRIIb shows a 20-fold higher affinity to human IgG2 and a 10-fold higher affinity to cynomolgus IgG2 compared with human FcγRIIb (Fig. 2C).

Another important feature that distinguishes cynomolgus IgG4 from human IgG4 is its inability to perform Fab arm exchange. It is now well established that human IgG4 can efficiently perform half-molecule Fab arm exchange in vivo, which is thought to further decrease the effector function of this isotype by preventing target cross-linking (27, 28). IgG4 Fab arm exchange is dependent on the presence of a serine at position 228 in the hinge of the IgG4 molecule. In contrast to human IgG4, however, we identified a cynomolgus IgG4 genotype with a proline-containing IgG1-like hinge sequence (Fig. 3A) in all 12 individual cynomolgus monkeys analyzed in this study that was resistant to Fab arm exchange (Fig. 4). We could not detect the generation of anti-CD20/anti-CD28–bispecific Abs if cynomolgus IgG4 subclass Abs were combined with either human, cynomolgus, or rhesus IgG4 Abs. In contrast, human and rhesus IgG4 anti-CD20/anti-CD28 Ab combinations generated high levels of bispecific Abs in vitro as well as in a mouse in vivo system (Fig. 4A, 4B). The cynomolgus IgG4 H chain is missing a cysteine at position 131 in comparison with human IgG4, and thus, the H and L chains are only held together via noncovalent interactions (20, 28). However, introduction of the cysteine (S131C) did not result in enhancement of Fab arm exchange (Fig. 4B). Notably, cynomolgus IgG4 with a human IgG4-like hinge region (P238S) did not increase Fab arm exchange. In contrast, the C_H3 domain of cynomolgus IgG4 seems to be important for Fab arm exchange, because the exchange of this domain by the human or rhesus C_H3 completely restored the Fab arm exchange with human IgG4 (Fig. 4B). To demonstrate the generation of bispecific Abs by Fab arm exchange by an alternative method, we used LC-MS under denaturing conditions and the cysteine-containing version of cynomolgus IgG4 (S131C). Combinations of human/human IgG4 Abs generated a third mass exactly between the masses of the two human parental Abs. In contrast, combination of human and cynomolgus IgG4 Abs did not reveal an additional mass species, further providing evidence that cynomolgus IgG4 does not perform half-molecule exchange (Fig. 4C–F).

To analyze whether the observed differences of human and cynomolgus IgG2 and IgG4 at the molecular and biochemical level are reflected in different function of these subclasses, we used human and cynomolgus PBMCs and serum to assess the ADCC and CDC potency of anti-CD20 Abs grafted on human and cynomolgus C region IgG subclasses. Cynomolgus B cells include two different populations characterized by different relative expression of CD20, CD40, and CD21 (i.e., CD20^lowCD40^highCD21⁺ and CD20^highCD40^lowCD21⁻) (29). Because the CD20^lowCD40^highCD21⁺ population in cynomolgus monkey shows similar properties as human B cells with respect to in vivo and in vitro sensitivity to rituximab treatment, we analyzed the in vitro depletion of CD40^highCD21⁺ B cells in both species to obtain comparable data. Human IgG2 and IgG4 are functionally silent not only toward human, but also toward cynomolgus PBMC, because they show similar low binding to human and cynomolgus FcγRIII (Fig. 2D, 2F), resulting in a corresponding similar low ADCC activity with human and cynomolgus NK cells (Fig. 5A, 5C). In contrast to human IgG2 and IgG4, cynomolgus IgG2 and IgG4 show potent ADCC with both human and cynomolgus PBMC as well as potent CDC with human and cynomolgus serum (Fig. 5B, 5D, Supplemental Fig. 4). In contrast, IgG1 and IgG3, the subclasses with equivalent FcγR binding, show a similar effector function profile on both human and cynomolgus PBMCs (Fig. 5). In conclusion, the observed ADCC and CDC functional activities appear to directly correlate with critical amino acids and binding data differences of the human and cynomolgus IgG subclasses. In contrast to the human IgG2 and IgG4 subclasses, cynomolgus IgG2 and IgG4 are not silent, whereas IgG1 and IgG3 behave very similar in the two species.

Despite the reported sequence homology of human and cynomolgus IgG subclasses and FcγRs (18), we could identify key differences between these two species on the level of molecular sequences, FcγR binding, receptor expression, distribution, and function. The identified differences induce changes, which can be explained by a single divergent evolutionary constraint, namely a different strategy to fight against bacteria and parasites. Cynomolgus granulocytes, consisting of neutrophils and eosinophils, do not express FcγRIIIb, but instead display FcγRII at very high levels. Although FcγRIIIb is a GPI-linked receptor, it still supports a variety of signaling events and is thought to enhance the overall avidity of IgG binding, thereby contributing to the IgG subclass specificity and FcγR-mediated function of human granulocytes (30, 31). Bacteria are mainly opsonized by carbohydrate-specific IgG2 subclass Abs (16, 17), whereas helminth parasites and allergens preferentially induce the generation of IgG4 (and IgE) Abs (32, 33). Strikingly, only IgG2 and IgG4 properties are different in cynomolgus monkey, demonstrating potent FcγR binding and effector function, whereas IgG1 and IgG3 subclasses show an almost identical profile in both cynomolgus monkeys and humans. The potential effector function of cynomolgus IgG4 is further increased by mutations in the hinge and C_H3 abrogating Fab arm exchange. Whether this is a general feature of all cynomolgus monkey subspecies and/or from different origins will require further investigation. This is in particular relevant, because, in the rhesus monkey, the critical amino acids involved in Fab arm exchange display allotypic variations (18, 34), and we could show that recombinant Abs with rhesus IgG4 C regions display similar or even increased IgG4 Fab arm exchange than human IgG4. It has also been described that human IgG4 therapeutic Abs can perform Fab arm exchange with cynomolgus serum Abs, but the IgG4 molecular sequence or allotype was not analyzed in this study (35).

Of note, we found that the increase of activating function of cynomolgus IgG2 (i.e., higher binding to FcγRI and FcγRIII and higher effector function) is counterbalanced by a strongly increased affinity toward cynomolgus FcγRIIb. Interestingly, in contrast to FcγRIIb in the NHP macaque family members cynomolgus and rhesus monkey, which contain a histidine at the position 131, all human primates analyzed (gorilla, orangutan, and chimpanzee) are characterized by an arginine at position 131. This change toward the low-affinity arginine variant of FcγRIIb is always accompanied by a corresponding low-affinity mutation in the Fc binding region of IgG2 similar to the mutation found in human IgG2 (Fig. 3B, 3C). We therefore hypothesize that the inhibitory receptor FcγRIIb and the silent isotypes coevolved and that macaque species such as the cynomolgus and rhesus monkey have developed a different strategy to balance IgG/FcγR interactions in contrast to humans and human primates. This coevolution of activating IgG isotypes and the inhibitory FcγR, FcγRIIb, supports the model of activating-to-inhibitory ratio as proposed by Nimmerjahn and Ravetch (8) based on studies in the rodent system, as it seems to play a fundamental role in the evolution and function of human effector function.

Some of our findings are in contradiction with an earlier study from Jacobsen et al. (20) showing that cynomolgus IgG1 displayed the highest binding on cynomolgus monocytes compared with other human and cynomolgus IgG isotypes. In addition, cynomolgus IgG1 was far more potent in mediating ADCC in cynomolgus whole blood. The differences to the data presented in this report can be explained by the following improved technologies applied in this study. First, the size exclusion chromatography-purified material used in our study guarantees a monomeric composition, whereas only protein A-purified material used by Jacobsen et al. (20) can result in a high percentage of aggregated, nonmonomeric Abs. For studying FcγR binding and effector function, the use of purely monomeric IgG is absolutely necessary, as FcγR shows 10- to 100-fold higher binding to aggregated IgG, and therefore, data generated with aggregated IgG would reflect the difference in the proportion of aggregated IgG rather than the mere difference in isotype selectivity. Furthermore, the affinity of IgG to the low-affinity FcγR (i.e., FcγRIIa, FcγRIIb, and FcγRIII) is micromolar, which obscures the detection of monomeric IgG in the FACS-based assay used by Jacobsen et al. (20) due to the fast off-rate displacing the IgG from the Fc receptor during the washing steps. In contrast, the surface plasmon resonance analysis of this interaction as used in our study is able to detect the low-affinity binding and is therefore the preferable method. In addition, the effector cells (i.e., monocytes) used in the Jacobson study (20) express a variety of different FcγR, and it is therefore not possible to clearly assign to which Fc receptors the different IgG subclasses bind. The functional activity of the different IgG isotypes was analyzed by Jacobsen et al. (20) in a whole blood B cell depletion assay using anti-CD20 Ab for depletion and CD19 as the marker for cynomolgus B cells. In contrast to humans, cynomolgus monkeys show two different populations of CD20-positive B cells. Only the CD40^highCD21⁺ cynomolgus monkey B cells are depleted by rituximab similar to human B cells and in a Fc-dependent manner (29, 36) (M. Warncke, unpublished observations), and therefore we used this population for our analysis. We think that the high quality of the material used in our study together with the precise methodology allowed us to identify differences between the human and cynomolgus IgG isotypes. The fact that the molecular sequence corresponds to the binding data and effector function further validates our approach.

Interestingly, position 131 of human FcγRIIa is also determining the high- and low-affinity binding of FcγRIIa to C-reactive protein (CRP). In contrast to IgG, CRP has a higher affinity to human FcγRIIa 131R than to human FcγRIIa 131H (37). As CRP is highly conserved in evolution and is mimicking important aspects of IgG, such as binding to FcγRs and complement (38), the discussed difference in NHP FcγRIIb may also affect the immune system beyond FcγR effector function.

Given the importance of FcγRIIb for the balance of the immune response, this remarkable shift in function of cynomolgus versus human FcγRIIb raises questions regarding the prediction of NHP effector function for human therapeutic Abs, as it has a direct influence on the toxicity assessment in cynomolgus or rhesus models of FcγRIIb-engaging agents (39, 40). For example, mutations in the Fc part of human IgG1 (S267E and L328F) led to >200-fold increased affinity to human FcγRIIb and FcγRIIa 131R, but not to FcγRIIa 131H. Thus, the key amino acid at position 131 is responsible for the increased affinity of this Fc variant (S267E and L328F) to FcγRIIb. However, this amino acid in cynomolgus FcγRIIb is different (Fig. 3), and therefore this mutant Fc variant (S267E and L328F) shows no enhancement in affinity to cynomolgus FcγRIIb (41).

In contrast to the differences identified in the silent IgG subclasses and FcγRIIb, the activating FcγR (FcγRI, FcγRIIa, FcγRIIIa) and the active IgG1 and IgG3 subclasses have comparable effector function in the human and the cynomolgus system: the binding characteristics of human and cynomolgus IgG1 and IgG3 toward human and cynomolgus FcγRs are comparable. IgG1 and IgG3 ADCC activity, which is mainly triggered in vitro by the ITAM-bearing receptor FcγRIIIa on NK cells (and to a lower extent FcγRIIa and FcγRI), and is largely independent on FcγRIIb, is indistinguishable whether human or cynomolgus effector cells are used. We therefore conclude that the observed species differences can be explained by key differences in the amino acid sequence composition at critical residues of the IgG2 and IgG4 Ab subclasses and the inhibitory receptor FcγRIIb in human and cynomolgus monkey. On the contrary, the activity profiles of the various activating FcγRs are by and large functionally conserved between the two species. Thus, human therapeutic Abs with a wild-type IgG1 C region are expected to have a very similar effector profile in the cynomolgus monkey as compared with humans, which qualifies the cynomolgus monkey as a relevant model for this isotype. However, modifications to the IgG1 wild type, resulting in decreased or enhanced binding (24, 42–44) to FcγRIII or FcγRIIa, which are increasingly being evaluated for therapeutic use, have to be evaluated very carefully, as the difference in FcγRIII and FcγRII expression on human and cynomolgus granulocytes could influence their overall functional behavior (45). In particular, when evaluating Fc modifications altering the binding to FcγRIIb (39, 40), equivalent FcγRIIb cross-reactivity for the selected animal model species is required to allow prediction to humans.

In summary, our findings shed new light on the evolution and plasticity of IgG effector function and allow a better extrapolation of human therapeutic Ab Fc-mediated functions from the NHP to humans.

We thank Holger Heine, Patrick Schindler, Stefan Ewert, Steffen Hartmann, Cornelius Fritsch, and Thomas Pietzonka for helpful discussion and David Lee for critical comments on the manuscript.

All authors are employees and/or shareholders of Novartis.