Genetic variation of human neutrophil Fcγ receptors and SIRPα in antibody‐dependent cellular cytotoxicity towards cancer cells

The efficacy of cancer therapeutic antibodies varies considerably among patients. Anti‐cancer antibodies act through different mechanisms, including antibody‐dependent cellular cytotoxicity (ADCC) triggered via Fcγ receptors (FcγR). This phagocyte ADCC can be promoted by interference with CD47‐SIRPα interactions, but the magnitude of this enhancement also varies among individuals. Both FcγR and SIRPα display considerable genetic variation, and we investigated whether this explains some of the variability in ADCC. Because of linkage disequilibrium between FcγR variants the interpretation of previous reports suggesting a potential link between FcγR polymorphisms and ADCC has been troublesome. We performed an integrated genetic analysis that enables stratification. ADCC by activated human neutrophils towards Trastuzumab‐coated breast cancer cells was predominantly dependent on FcγRIIa. Neutrophils from individuals with the FcγRIIa‐131H polymorphic variant displayed significantly higher killing capacity relative to those with FcγRIIa‐131R. Furthermore, ADCC was consistently enhanced by targeting CD47‐SIRPα interactions, and there were no significant functional differences between the two most prevalent SIRPα polymorphic variants. Thus, neutrophil ADCC capacity is directly related to the FcγRIIa polymorphism, and targeting CD47‐SIRPα interactions enhances ADCC independently of FcγR and SIRPα genotype, thereby further suggesting that CD47‐SIRPα interference might be a generic strategy for potentiating the efficacy of antibody therapy in cancer.


Introduction
Therapeutic antibodies are widely used for the treatment of certain forms of cancer. In addition to direct growth effects on the cancer cells, monoclonal antibodies can coat the tumor cells, thus turning them into targets for immune-mediated destruction by antibodydependent cellular cytotoxicity (ADCC), antibody-dependent cellular phagocytosis and/or complement-dependent cytotoxicity. Although the exact contribution of ADCC to antibody therapy in cancer patients is not known, the reported associations between the clinical efficacy of different cancer therapeutic antibodies, including Trastuzumab, and polymorphisms in FcγRIIa expressed on myeloid cells and FcγRIIIa expressed on NK cells, suggest a role for ADCC mediated by both cell types in patients [1]. A considerable number of genetic Fcγ Receptor (FcγR) variants have been described within the FCGR2/3 locus [2][3][4]. Preliminary studies reported that some of these, such as e.g. the FcγRIIa H/R131 single nucleotide polymorphism (SNP), are linked to responsiveness to antibody therapy in cancer patients [5,6] and also to the ADCC capacity of neutrophils [7]. However, the interpretation of such findings is not straightforward at all, mainly because many of the genetic variants are in linkage disequilibrium with others, which makes a direct comparison based on the analysis of individual variants difficult if not impossible. Thus, to obtain insight into the contribution of FcγR variation to antibody mediated cancer cell destruction, while avoiding such bias, an integrated analysis of FcγR genotype needs to be performed along with proper stratification. We have previously developed a multiplex ligation-dependent probe amplification (MLPA) assay to determine the relevant polymorphic and gene copy number variations (CNV) within the FCGR2/3 locus [2,[8][9][10]. In the present study we have applied this method to investigate direct associations between the relevant genetic FcγR variants and neutrophil ADCC capacity.
Another important aspect of cancer therapeutic antibodies is that their clinical efficacy is rather limited. In fact, despite their high degree of specificity, the potency of cancer therapeutic antibodies is generally too low to justify their use as monotherapeutics in the absence of additional non-specific treatment regimens such as chemotherapy. Chemotherapeutics themselves are carcinogenic and cause many other side effects, such as leukopenia, which would be anticipated to compromise ADCC rather than to promote it unfortunately. Therefore, there is a pertinent need to improve the efficacy of cancer therapeutic antibodies. We and others have previously demonstrated that targeting the interaction between CD47 expressed on cancer cells and the inhibitory immunoreceptor SIRPα expressed on myeloid cells substantially potentiates the capacity of anti-cancer antibodies, including Trastuzumab and Rituximab [11][12][13], for recent reviews see [14][15][16]. Consistent with this notion the clinical response of either breast cancer patients treated with Trastuzumab or Non-Hodgkin lymphoma patients treated with Rituximab was better when CD47 expression levels in the tumor cells were lower [11,13]. However, it is not exactly known how SIRPα signaling inhibits ADCC. Does it for instance do so by inhibiting all different FcγRs and their variants expressed on phagocytes, or by only affecting some of them? Furthermore, there are different polymorphic variants of SIRPα within the population [17,18] and although the two variants most commonly found in Caucasians do not differ with respect to their CD47 binding capacity [19], it is not known whether they differ in other aspects of their functioning, including their capacity to signal and to modulate ADCC.
In the present study we have evaluated 101 healthy individuals to investigate a possible association between FcγR genetics and function in neutrophil ADCC. In addition, we have explored whether the capacity to potentiate ADCC through the manipulation of CD47-SIRPα interactions is affected by SIRPα polymorphisms. While activated neutrophils express a range of FcγR, including FcγRI (CD64), FcγRIIa (CD32), FcγRIIIb (CD16) and in some individuals also FcγRIIc we have observed that the principle FcγR on neutrophils participating in ADCC towards Trastuzumabcoated breast cancer cells is FcγRIIa. Furthermore, we have identified an independent association between the FcγRIIa-H/R131 polymorphism and neutrophil ADCC, with no significant associations for the other variants, i.e. the FcγRIIIb-NA1/NA2 polymorphism and FcγRIIc-Stop/ORF.
Finally, we show that the capacity to induce ADCC through the different FcγRs and their genetic variants can be potentiated to a similar extent (ß1.5 to 2-fold) by the manipulation of CD47-SIRPα interactions, and this is not affected either by the major SIRPα polymorphisms present in the Caucasian population. These findings demonstrate for the first time unequivocally a direct link between neutrophil FcγR genetics and function in ADCC. In addition, they support the concept that interference with CD47-SIRPα interactions constitute a generic method to enhance the efficacy of antibody therapy in cancer patients.

The principal FcγR mediating neutrophil ADCC is FcγRIIa
The clinical efficacy of antibody therapy in cancer, as well effector cell ADCC capacity, varies considerably among individuals [13]. We aimed to explore the mechanistic basis underlying this variability in individual ADCC potential, and in particular the role of genetic and functional variation in FcγR expression. We did so by evaluating neutrophil ADCC capacity towards Trastuzumabcoated Her2/Neu-positive SKBR-3 breast cancer cells of 101 healthy adult Caucasian individuals. Evaluation of ADCC, performed in the presence or absence of CD47 interference, by either blocking of CD47-SIRPα interactions with antagonistic anti-CD47 F(ab ) 2 antibody fragments or by the knock-down of CD47 in tumor cells, showed a substantial inter-individual variation, ranging from 20-82% in the absence of CD47 interference ( Fig. 1A and B). Killing in the absence of Trastuzumab was always below 3%. As reported previously [13], CD47 interference gave a highly significant and also consistent enhancement of neutrophil ADCC, supporting the idea that CD47-SIRPα interactions and SIRPα inhibitory signaling restrict ADCC performed by phagocytes [13,20,21] (Fig. 1A and B). The interference with CD47-SIRPα interactions resulted in an average ß1.5 to 2-fold increase in cytotoxicity independent of the method of interference used (Fig. 1C). Neutrophils express different FcγRs, which could all contribute to neutrophil ADCC. To investigate the involvement of the various FcγRs in our neutrophil ADCC model, we used blocking antibodies and other antagonists against the different FcγRs. Neutrophils cultured in the presence of G-CSF and IFNγ, as used in this study and also in a previous study [13], express FcγRI, FcγRIIa, FcγRIIIb, and in a minority of about 15-20% of Caucasian individuals also FcγRIIc [2]. As shown in Fig. 2, blocking experiments with antagonists, i.e. monovalent Fc-fragments for blocking FcγRI (Supporting Information Fig. 1) or F(ab ) 2 -fragments of blocking antibodies against FcγRII or FcγRIII, revealed that Trastuzumab-mediated killing of SKBR-3 cells by activated neutrophils predominantly involves FcγRIIa. Of interest, there was no effect of the blocking of FcγRI, which represents the high affinity FcγR, but is expressed at relatively low levels and may potentially already be saturated with monovalent IgG from human plasma and/or the culture [22]. The blocking of FcγRIIIb (CD16b), which lacks intrinsic signaling capacity but is highly expressed on neutrophils and can potentially act as a tethering receptor [23][24][25] did not result in inhibition. Instead, if anything there might be a neglible enhancement.

The FcγRIIa polymorphism is associated with neutrophil ADCC capacity
The FcγRs are encoded on chromosome 1, of which some, in particular those of the FCGR2/3 cluster, are subject to considerable variation in terms of polymorphisms and/or surface expression variation due to either gene CNV or promotor polymorphisms [2,3]. We wanted to know whether the variation in neutrophil ADCC between neutrophil donors was related to the known genetic variations in neutrophil FcγR expression and function. Previous studies have already demonstrated that some of the relevant polymorphisms of FcγRs on neutrophils are associated with the clinical efficacy of therapeutic antibodies. In particular, the 'higher affinity' allele of FcγRIIa, FcγRII-H131, is associated with a better clinical responsiveness as compared to the alternative FcγRII-R131 allele, as has been observed for Rituximab treatment in Non-Hodgkin lymphoma [26], Trastuzumab treatment in Her2/Neupostive metastatic breast cancer [5], and Cetuximab treatment in colorectal cancer [27]. However, as indicated above, additional FcγR polymorphisms and gene CNV, which are actually known to be in linkage disequilibrium with the FcγRIIa-H/R131 polymorphism and with each other, also exist at the FCGR2/3 locus [28]. Thus, to make the most appropriate comparisons, these variations need to be evaluated in an integrated fashion. Therefore, all our neutrophil donors were genotyped by MLPA for all FcγR variants, and the relationship between FcγR genotype and target cell killing in ADCC was explored after appropriate stratification. An overview of the FCGRs SNP and CNV in our cohort can be found in Table 1.
When considering the FcγRIIa H/R131 polymorphism, the results demonstrated that there is a trend that homozygous H131 donors induce higher cytotoxicity toward target cells than homozygous R131 donors when considering the entire study population (Fig. 3A). After stratifying for the three most common Figure 2. Involvement of FcγRs in neutrophilmediated ADCC toward Trastuzumab-coated SKBR3 cells. ADCC was performed as described in detail in the materials and methods. FcγRs on neutrophils were blocked during ADCC by antagonistic antibodies against FcγRI (monovalent Fc Fragments), FcγRII (mAb 7.3 F(ab ) 2 ) and/or FcγRIII (mAb 3G8 F(ab ) 2 ). Experiments were performed with either SKBR3 cells ('control') or with SKBR3-CD47KD cells ('CD47 interference'). Data shown are means + SEM pooled from 5 independent experiments with neutrophils from n = 10 healthy individual donors with 2 donor samples per experiment. Statistics were performed by paired one way ANOVA with Sidak post-test. Significance was calculated in relation to the two control groups (Trastuzumab only). ns = non-significant, *p < 0.05, **p < 0.01 and ***p < 0.001. genotypes in the population having: (i) the FCGR2C-Stop haplotype of the p.Q57X SNP in exon3 of the FCGR2C gene, with the corresponding individuals lacking FcγRIIc protein expression, designated FcγRIIc-STOP, (ii) having 2 gene copies of FcγRIIIb termed FcγRIIIb-2x, and (iii) having no SH variant in FcγRIIIb [29], this difference was significant (Fig. 3B). To determine the cause of this difference we looked into the FcγR expression levels. We  29 (35,4%) observed no relationship between the FcγR expression level and ADCC capacity (data not shown), suggesting at least that FcγR expression levels per se are not an important determinant in this context. Secondly, as can be seen in Supporting Information Fig. 2, there were at least no differences in FcγRIIa/c expression between H131 and R131 individuals, essentially excluding this possibility too. We also determined that there is no significant difference in our ADCC assays between individuals who have FcγRIIc-ORF or FcγRIIc-STOP (Supporting Information Fig. 3).
There are various levels of potentially relevant genetic variation in FcγRIIIb, which is the only FcγRIII isoform expressed by neutrophils [23]. First, there is FcγRIIIb gene CNV, with individuals expressing 0-5 copies of the gene. The number of encoded FcγRIIIb copies is strongly associated with surface FcγRIIIb expression and function [30,31]. Because in the group that we tested the vast majority (n = 86) had 2 gene copies of FcγRIIIb gene copies (FcγIIIb-2x) we did not have the power to test for a possible relationship between FcγRIIIb CNV and killing capacity. The most common polymorphic variants of FcγRIIIb are designated FcγRIIIb-NA1 and -NA2. The FcγRIIIb -NA1 and -NA2 nucleotide sequences differ at five positions, with four amino acid differences. As a consequence, the NA2 variant has two additional N-linked glycosylation sites as compared to NA1, and this might have functional consequences. For instance, neutrophils from FcγRIIIb-NA1NA1 individuals are known to bind and phagocytize IgG-opsonized bacteria and red blood cells more efficiently than those from -NA1NA2 and -NA2NA2 individuals [32,33]. We therefore hypothesized that the FcγRIIIb NA1/NA2 polymorphism may also play a role in ADCC. We stratified our analysis to the neutrophil donors with 2 copies of FcγRIIIb (FcγRIIIb-2x). As shown in Supporting Information Fig. 4A, the various genotypes induced no significantly higher cytotoxicity to tumor cells. Further stratification for the most common FcγRIIc-Stop allele and having no SH variant gave essentially the same results (Supporting Information Fig. 4B). Again, there were no significant differences in the surface expression of FcγRI, FcγRII or FcγRIIIb among the NA1/NA2 variants (Supporting Information Fig. 5). Finally, we investigated potential functional interactions between the relevant genotypes. Due to the limited group size we were forced to restrict our analysis to the most common FcγRIIa-H/R131 and the FcγRIIIb-NA1/NA2 variations. As can be seen in Supporting Information Figs. 6 and 7 the contributions of these two variants described above were independent and were not linked to each other.

Potentiation of neutrophil ADCC by interference with CD47-SIRPα interactions
We have previously demonstrated that CD47-SIRPα interactions restrict neutrophil-mediated ADCC, suggesting that interference with such interactions could be a promising strategy for enhancing therapeutic antibody-dependent tumor cell destruction [13]. However, it is not known whether CD47-SIRPα interactions have a generalized effect on FcγR signaling, or whether the effects are restricted to one or more FcγR or their variants. To test the contribution of CD47-SIRPα interactions and inhibitory signaling to killing through the various FcγRs and their genetic variants, parallel testing in ADCC was performed with SKBR3 target cells in which either CD47 knock-down (CD47KD) was performed or when an inhibiting F(ab ) 2 antibody for CD47 was used. As can be seen in Fig. 2, CD47 interference consistently enhanced ADCC towards the tumor cell targets irrespective of the FcγRs that were functional. The same was found when the effect of CD47 interference in combination with the different FcγRIIa-H/R131 and FcγRIIIb-NA1/NA2 polymorphisms were evaluated ( Fig. 3 and Supporting Information Fig. 4). These data are consistent with a common pathway of regulation in ADCC by CD47-SIRPα that is independent of FcγR genotype.

SIRPa polymorphisms do not affect neutrophil ADCC capacity
Finally, we investigated the potential contribution of SIRPα polymorphisms. It has been reported that a considerable number of SIRPα polymorphisms exist in different ethnic groups [17], Figure 4. Polymorphisms of SIRPα do not affect neutrophil ADCC toward Trastuzumab-coated SKBR3 cells. ADCC assay was performed as described in the materials and methods. ADCC capacity for donors genotyped for the SIRPα polymorphisms were compared. (A) Results for all, including subjects with the indicated SIRPα genotypes, including homozygous α 1 /α 1 (n = 13), homozygous α BIT /α BIT (n = 40) and heterozygous α 1 /α BIT (n = 29), or (B) stratified for subjects carrying the FcγRIIc-Stop allele, 2 copies of FcγRIIIb and no SH variant: homozygous α 1 /α 1 (n = 9, homozygous α BIT /α BIT donors (n = 26) and heterozygous α 1 /α BIT (n = 18). Experiments were performed with either SKBR3 cells ('control') or with SKBR3 cells in which CD47 was manipulated 'CD47 interference' either by 'anti-CD47 F(ab ) 2 '-blocking or 'CD47KD'. The individual ratios of cytotoxicity of the indicated CD47 interference condition and control are shown in the panels on the right. Statistical significance between control and CD47 interference was performed by paired Student's t-test resulting in **** p < 0.0001 for α 1 /α 1 , α 1 /α BIT and α BIT /α BIT (not shown in graph). Data shown are means + SEM and are pooled from over 50 experiments with 2-3 donor samples per experiment. Data shown are all individual donors and represented by a single dot in the graph. Statistics were performed by one way ANOVA with Tukey's test for multiple comparisons. ns = non-significant.
but DNA sequencing and MLPA analysis of our healthy Caucasian donors (n = 82) identified only two polymorphic variants, SIRPα 1 and SIRPα BIT (Supporting Information Fig. 8), within this population with frequencies of 15.9 and 48.7% of SIRPα 1 and SIRPα BIT homozygotes, respectively, and 35.4% heterozygotes (allele frequencies: 34% (SIRPα 1 ) and 64% (SIRPα BIT )) as can be seen in Table 1. SIRPα gene CNV variation was not observed in our MLPA analysis (not shown). Although these two variants differ in 13 out of 118 amino acid residues in the N-terminal immunoglobulin-like domain responsible for CD47 binding, there appear to be no detectable differences in affinity for CD47 [19], which is perhaps not surprising as most of these polymorphisms occur in areas outside the CD47 binding site (Supporting Information Fig. 8). Nevertheless, it seemed possible that there are other differences between these variants, such as differences in expression levels or signaling capacity that could have an impact on downstream signaling and neutrophil ADCC capacity. To our knowledge functional differences in terms of downstream signaling capacity have not been investigated among SIRPα polymorphic variants before in (primary) cells. However, as can be seen in Fig.  4 there were no measurable differences in the ADCC capacity between the different SIRPα genotypes. Furthermore, the relative modulatory effect of interference with CD47-SIRPα interactions, i.e. the ß1.5 to 2-fold increase mentioned above, was consistent and did not differ among the SIRPα genotypes tested. Taken together, it seems therefore that the enhancement of ADCC by manipulation of CD47-SIRPα interactions is independent of FcγR genotype, and SIRPα genotype, which further fuels the idea that this may provide a generic method for potentiating the efficacy of cancer therapeutic antibodies.

Discussion
In the present study we have investigated whether differences in FcγR and SIRPα genetics may provide an explanation for the inter-individual variation in neutrophil ADCC capacity, and the potential of the latter to be enhanced by manipulation of CD47-SIRPα interactions, respectively. This represents the largest study thus far conducted with respect to either of these issues. Furthermore, this is the first study that investigates FcγR genetic variation in antibody-mediated destruction of cancer cells in an integrated fashion, which is important given the known linkage disequilibrium within the FcγR locus. This analysis allowed us, upon stratification for the two most frequent variations in the other neutrophil FcγR, namely FcγRIIc-Stop and FcγRIIIb-2x, to define a significant association between the FcγRIIa H/R131 polymorphism and neutrophil ADCC towards Trastuzumab-coated breast cancer cells. It should be mentioned that there is as yet no definitive mechanistic explanation available for the observed association of FcγRIIa, because the affinity of FcγRIIa-H131 (K A ß5,2 × 10 −6 ) for human IgG1 appears only slightly higher than that for FcγRIIa-R131 (K A ß3,5 × 10 −6 ), as measured by surface plasmon resonance [34]. It appears unlikely that this slight difference is the sole cause of the difference in ADCC. By comparison, the difference between the two FcγRIIa alleles for e.g. IgG2 binding is much more prominent and this also translates into differences in neutrophil ADCC capacity as observed for the IgG2 anti-EGFR antibody Panitumumab [35]. There was also no difference in expression levels between the different alleles that could explain the results, thus, receptor properties other than IgG1 binding capacity, such as membrane mobility or interactions/cooperation with other molecules that could potentially affect intracellular signaling and consequently ADCC could perhaps contribute for the differences observed. Whatever the mechanism, there is good reason to believe that the observed association is relevant in vivo as it is also linked to the clinical response to Trastuzumab [5] and other cancer therapeutic antibodies of the IgG1 subclass [6,36]. The fact that such an association between the FcγRIIa-H/R131 and clinical responsiveness to cancer therapeutic antibodies is not found in all studies [37][38][39] can, apart from an overall lack in statistical power and appropriate stratification as discussed, also be explained by the varying levels of immunosuppression achieved by concurrent chemotherapeutic regiments applied in these different studies. Finally, our findings appear to be in contrast with earlier observations obtained on the link between FcγRIIa 131H/R polymorphism in neutrophil ADCC towards Her2-positive breast cancer cells, where better neutrophil killing with the FcγRIIa-R131 was seen, but this can be explained by the use of the mouse IgG1 antibody 520C9 against Her2 [7,40]. What is also apparent from our findings is that there is at least no obvious associations between neutrophil ADCC capacity and some of the other FcγR variations that we evaluated in the current study, including the FcγRIIc-Stop/ORF and the FcγRIIIb-NA1/NA2. Whereas for some of these our study may have been somewhat underpowered, at least in some cases such as in the case of the FcγRIIIb-NA1/NA2 variant our data corroborates earlier in vivo data where no effect was found [41]. Nevertheless, these data do not necessarily exclude a role for FcγRIIIb in neutrophil ADCC and antibody therapy. In fact, we found a small but significant increase in neutrophil ADCC in our study, which may be indicative for an 'inhibitory' role of FcγRIIIb in certain cases. We believe that the role of FcγRIIIb warrants further evaluation, especially because we have used in vitro activated neutrophils in our study in which FcγRIIIb expression is relatively low.
It should be noted that when considering human neutrophils as effector cells in ADCC towards tumor cells some studies have also indicated a requirement for both FcγRII and FcγRIIIb [42,43], while others show a more exclusive involvement of FcγRII [44]. These differences are likely to be related, at least in part, to differences in target cells and opsonizing antibodies employed, and in the activation state of the neutrophils. One potential pitfall that could also have affected the results, at least in some of the reported studies, is the use of intact antibodies for FcγR blocking, which may interfere with the activity of other FcγR as well by a phenomenon that has become known as the 'Kurlander-effect' [45].
Although resting neutrophils do not express significant levels of FcγRI, it represents the only high affinity receptor for IgG and may therefore be important for successful therapy with cancer therapeutic antibodies especially when combined with SIRPα-CD47 interference therapies [46]. FcγRI is constitutively expressed on monocytes and macrophages [47], can be induced on neutrophils in patients by treatment with cytokines, such as IFNγ and/or G-CSF [26,48,49], and may even be upregulated in cancer patients during chemotherapy-induced neutropenia [50]. Since the Kurlander-effect could not be excluded when using intact monoclonal antibody 10.1, a known blocker of FcγRI, and the F(ab ) 2 fragments of this antibody are known not to be able to fully block the receptor [22] we used monovalent Fc-fragments to inhibit FcγRI. Due to the receptors high affinity to binding IgG-Fc fragments they are able to inhibit FcγRI (Supporting Information Fig. 1), and potentially also partially other receptors. However, as shown in Fig. 2, monovalent Fc fragments are unable to block FcγRIIa/c and FcγIIIb in our system. Our findings demonstrate ADCC with IFNγ and G-CSF treated neutrophils involves mainly FcγRIIa/c, also when this is further enhanced by CD47-SIRPα checkpoint blockade, with no significant role found for FcγRI. It should be noted that when using a different effector cell, like macrophages or CD16+ monocytes, that the FcγRs required for ADCC and phagocytosis could be different then when looking at neutrophils, since they express a different combination of FcγRs.
We have also studied in detail whether the beneficial effect of targeting CD47-SIRPα interactions, with either antagonists or CD47 knock-down in the tumor cells, is associated with the available FcγR (geno)type. Our findings essentially show that the potentiating effect of CD47-SIRPα targeting occurs independent of the FcγR type available (Fig. 2) and FcγRIIa or FcγRIIIb polymorphic variant(s) encoded ( Fig. 3 and Supporting Information Fig. 4), and that also the magnitude of the enhancing effect is very similar on average. We have also explored a possible role for SIRPα genetics in regulating neutrophil ADCC. Although it has been reported that there are at least 10 SIRPα polymorphic variants among different ethnic groups, including African and Asian individuals [17], the actual diversity within the Caucasian population has not been determined. We show that within our 82 healthy Caucasian neutrophil donors evaluated there are only two variants present, i.e. SIRPα 1 (also known as variant 2) and SIRPα BIT (also known as variant 1), with allele frequencies of 34 and 66% respectively. Whereas it was known that amino acid variation primarily occurs in the regions flanking the CD47 binding site within the N-terminal Ig-like domain of SIRPα and also that these particular Caucasian SIRPα variants are similar with respect to their affinity for CD47 [19], there could still be functional differences in the responses downstream. However, our current results show that the both genetic SIRPα variants have very similar ADCC suppressing capacity, which is represented by the typical ß1.5to 2-fold potentiation observed upon interference (Fig. 4). This demonstrates, for the first time, that the two SIRPα polymorphisms within the Caucasian population do not show differences in their overall function as determined here by ADCC, at least when using neutrophils.
Collectively, our findings show that FcγRIIa expressed by neutrophils can effectively trigger ADCC against antibody-coated cancer cells. They also demonstrate that the magnitude of the response is affected by the FcγRIIa-H/R131 polymorphism. Finally, we demonstrate that CD47-SIRPα interactions regulate ADCC triggered via FcγRIIa and its genetic variants to a similar extent, and independent of the SIRPα polymorphisms that are present. The latter clearly supports the idea that interference with CD47-SIRPα interactions will be a broadly applicable therapeutic strategy to potentiate antibody therapy in cancer, independently of FcγR and SIRPα genetics.

Isolation of human neutrophils from healthy donors
Neutrophils were isolated from n = 101 healthy Caucasian volunteers by density centrifugation of heparinized blood over isotonic Percoll (Pharmacia Uppsala, Sweden) followed by red cell lysis with hypotonic ammonium chloride solution at 4°C [51]. Cells were cultured in RPMI (Gibco) medium, supplemented with 10% (v/v) fetal bovine serum, 2 mM L-glutamine, 100 U/mL penicillin and 100 μg/mL streptomycin (i.e. complete RPMI) or HEPES + medium (containing 132 mM NaCl, 6.0 mM KCl, 1.0 mM CaCl 2 , 1.0 mM MgSO 4 , 1.2 mM K 2 HPO 4 , 20 mM Hepes, 5.5 mM glucose and 0.5% HSA), in the presence of 10 ng/mL clinical grade G-CSF (Neupogen; Amgen, Breda, The Netherlands) and 50 ng/mL recombinant human Interferon-γ (Pepro Tech Inc, USA) at a concentration of 5 × 10 6 cells/mL for 16 h. Afterward cell viability was determined by the amount of FITC-Annexin V (BD Pharmingen, San Diego, CA) positive cells on FACS, after which the cell concentration was corrected to have 5 × 10 6 viable cells/mL. Cells were consequently washed and prepared for analysis by ADCC assay.

Analysis of FcγR and SIRPα polymorphisms and CNV
Genotyping of 101 individuals for FcγRIIa, FcγRIIIb, and FcγRIIc polymorphisms was performed using the FCGR-specific Multiplex Ligation-dependent Probe Amplification (MLPA) assay (MRC Holland), using genomic DNA isolated from whole blood with the QIAamp R kit (Qiagen, Hilden, Germany). The MLPA assay was performed essentially as described previously [2,8,52].
In short, CNV of FCGR2C and FCGR3B was detected by genespecific MLPA probes designed on multiple sites in these genes. Probes were also included to detect the FCGR2A [p.131H/R] (rs1801274) SNP and haplotypes in FCGR2C and FCGR3B. Specific probes were included for the FCGR3B haplotypes NA1/NA2/SH, which respectively encode for the HNA1a/HNA1b/HNA1c alloantigens of human neutrophil antigen 1 [53]. To construct FCGR2C haplotypes, the assay contained a probe specific for the stop codon in exon3 of the FCGR2C gene [p. 57X] and a nonspecific FCGR2B/C probe to detect the ORF in exon3 [p. 57Q], which together can determine the rs759550223 SNP. Probes were also included for the splice site mutation at the border of exon7-intron7 in FCGR2C (rs76277413 c.798 +1 A>G), to distinguish the non-expressed 'nonclassical' FCGR2C-ORF variant from the 'classical' FCGR2C-ORF that is typically expressed on natural killer (NK) cells, monocytes and neutrophils [2]. Because the nonclassical FCGR2C-ORF variant is not expressed [2], it was grouped with FCGR2C-Stop in all analyses. An overview of the probes used is shown in Supporting Information Table 1. The frequency distribution of the different variants within our study population is provided in Table 1.
With genomic DNA from PBMC of healthy Caucasian donors (n = 82), SIRPα CNV and SNPs were determined with a SIRPα specific MLPA assay and confirmed with regular sequencing. In our donor set, SIRPα haplotype was identified by sequencing the V-Ig domain encoded by the third exon (data not shown), namely SIRPα 1 and SIRPα BIT . MLPA probes, binding to SIRPα variants, are synthetic oligonucleotides made by Invitrogen (Carsblad, CA) and were designed according to the sequencing results and the available data in http://www.ensembl.org/index.html. The regular sequencing primers were located in the introns surrounding exon 3, thus sequencing the whole exon. For an overview of the specific target sequences of the probes and the forward and reverse primer used for the regular sequencing, see Supporting Information Table 1. The MLPA assay was performed essentially as described previously [2,8,52].

ADCC assay
ADCC was measured in a 4-hour 51 Cr release assay with SKBR3 and SKBR3-CD47KD as target cells and G-CSF/IFNγ-primed human granulocyte as effector cells, as described previously [13]. In brief, human breast carcinoma cell lines were harvested by mild trypsin treatment, and washed tumor cells (1 × 10 6 cells) were collected and labeled with 100 μCі 51 Cr (Perkin-Elmer, USA) in 500 μL for 90 min at 37°C. The target cells (5 × 10 3 /well) and effector cells were co-cultured in 96-well U-bottom tissue culture plates in a ratio of E:T = 50:1, in the presence or absence of 5 μg/mL trastuzumab in RPMI or IMDM supplemented with 10% (v/v) FCS medium. Aliquots of supernatant were harvested and analyzed for radioactivity in a gamma counter. The percentage relative cytotoxicity was determined as [(experimental cpmspontaneous cpm)/ (total cpm-spontaneous cpm)] × 100%. All conditions were tested in triplicate. In case of neutrophils of a single individual tested on multiple occasions, the average measurements were used to avoid disturbing the population balance.

Rosetting assay with 293T cells and RBCs
As described earlier [22] 293T cells (ATCC) (which have no endogenous FcγR expression) were transfected with a FcGr1A1-GFP fusion construct, or GFP alone as a control. RhD positive erythrocytes, uncoated or coated with human anti-RhD IgG (RheDQuin, Sanquin, The Netherlands), stained with DiD (5 μM, Invitrogen), were combined in a ratio of 1:10 (293T:RBC) in combination with known FcγRI blocking antibody 10.1 and monovalent Fc-fragments for blocking FcγRI with various concentrations for 15 minutes at 37°C after which samples were fixed with PFA. Samples were run in an ImageStreamX flow cytometer (Amnis Corporation, Seattle, WA) to determine their blocking capacity.

Statistical analyses
Statistical differences were determined by either paired or ordinary one way ANOVA, with Sidak or Tukey's post-test, or by paired Student's t-test, as indicated in the figure legend.