Fc Gamma Receptor 3A and 2A Polymorphisms Do Not Predict Response to Rituximab in Follicular Lymphoma

Follicular lymphoma (FL) patients who are asymptomatic and with low tumor burden are candidates for a watch and wait strategy, as early treatment has not been shown to improve survival (1, 2). However, single agent rituximab is often administered to these patients, with a goal of delaying the need for chemotherapy (3–7). Rituximab, an IgG1 subclass monoclonal antibody to CD20, has revolutionized therapy of FL by improving response rates, duration of responses (DORs), and overall survival (5, 8–13). One postulated mechanism of action is antibody-dependent cell-mediated cytotoxicity (ADCC). In this process, binding of the Fc gamma receptor (FCGR) on macrophages and NK cells to the Fc portion of the rituximab antibody induces phagocytosis of the target cell, to which the antibody is bound (14).

In vitro and animal studies have suggested that variation in specific SNPs in the FCGR sequence might confer variable responses to rituximab, due to the efficacy of Fc binding and triggering of ADCC (15, 16). The FCγRIIIA receptor (CD16a) is present on NK cells, monocytes, and macrophages (17). A valine/phenylalanine (V/F) polymorphism at amino acid position 158 of FCGR3A (rs396991) has been identified in humans, with the valine allele demonstrating higher affinity to human IgG1 than phenylalanine, resulting in enhanced ADCC (18). The FCγRIIA receptor is present on monocytes and macrophages, but not NK cells (17). A histidine/arginine (H/R) polymorphism at position 131 of FCGR2A (rs1801274) affects binding affinity of IgG2, with the histidine allele binding more strongly.

The first clinical study in FL examining the influence of these SNPs in FL suggested the FCGR3A valine/valine (VV) genotype, (but not the FCGR2A HH genotype) was associated with improved response rates to single agent rituximab (3). A retrospective analysis from Stanford University then reported improved response rates and time to progression in patients with a FCGR3A VV genotype, and in patients with a FCGR2A histidine/histidine (HH) genotype (19). A prospective multicenter trial of single agent rituximab for FL found the FCGR3A VV genotype was associated with event-free survival but not response rate (4). These three studies support the hypothesis that ADCC plays an important role in cell killing and that certain polymorphisms in FCGR3A and FCGR2A may influence ADCC. However, each study had different findings on the relative importance of FCGR3A versus FCGR2A and different findings about the influence on response rate and/or response duration. Using patient derived samples from E4402, the rituximab extended schedule or re-treatment trial (RESORT), we conducted a correlative analysis, evaluating the impact of these candidate SNPs in the FCGR3A and FCGR2A genes. The goal was to definitively determine the clinical significance of these two polymorphisms.

RESORT was a multicenter randomized trial that enrolled 408 FL patients between November 2003 and September 2008 (20). Patients received single agent rituximab (375 mg/m2) in 4 weekly doses, followed by randomization for responders to re-treatment with rituximab upon progression (375 mg/m² × 4 weekly doses) versus maintenance rituximab (MR) (375 mg/m² once every 12 weeks) (Fig. 1).

SNP genotyping was performed to assess the FCGR3A SNP genotype for rs396991 [valine (V) or phenylalanine (F)] and the FCGR2A SNP genotype at rs1801274 [histidine (H) or arginine (R)]. Using banked PBMCs (N = 212) or formalin-fixed paraffin-embedded tumor tissue (FFPE) (N = 109), DNA was extracted using an automated platform (AutoGen FlexStar Qiagen chemistries), followed by quantification by UV absorbance and quality control by 260/280 OD ratio and PicoGreen, and then storage in TE buffer. For SNP genotyping, samples were plated into 96 well plates and genotyped on the Taqman (TM) platform (Applied Biosystems 7900HT Fast RealTime PCR System). Sequence data and assay conditions for rs396991 (FCGR3A) and rs1801274 (FCGR2A) are provided at SNP500 (http://snp500cancer.nci.nih.gov). Genotype data were analyzed using Applied Biosystems SDS 2.3 analysis software. Quality control samples included study replicates (5%); FCGR2A and FCGR3A known wild-type homozygotes, heterozygotes, homozygote variants; and a DNA negative control. We summarized the call rates per SNP as well as per sample, examining samples that failed in 5% or more of SNPs. SNPs that failed in 5% or more of individuals were evaluated by Sanger sequencing. Likewise, if the error rate in the subjects for whom repeated genotypes was judged to be unacceptably high (>2%), genotyping was repeated until consistent results were achieved.

The FCGR3A rs396991 SNP has been reported to be difficult to accurately genotype, due to high copy number variation and high homology (97%) between FCGR3A and FCGR3B genes. FCGR3B carries an invariant G at the position that corresponds to FCGR3A 4985T > G and co-amplification of FCGR3A and FCGR3B can lead to SNP miscalls. Transgenomics, Inc., Omaha, NE developed primers and probes specific for genotyping the FCGR3A that does not co-amplify the FCGR3B. To test the transgenomics (TG) proprietary technology, cases with available PBMC DNA (N = 212) were run in duplicate using both the commercial TM platform and the TG platform. All PBMC cases were also subjected to Sanger sequencing, with the sequencing result used as the reference gold standard to determine assay accuracy.

Fisher's exact test was used to compare response rates among the different genotypes. A logistic regression model was employed to evaluate the polymorphism effect on response rate accounting for other patient characteristics. DOR was defined as the time from documented response to documented progression and estimated using the Kaplan and Meier method. Logrank test (one-sided significance level of 0.05) was used to compare DOR. A Cox proportional hazards regression model was used to evaluate the significance of polymorphism effect on DOR after adjusting for other patient characteristics.

A total of 408 FL patients were enrolled in RESORT. Of these, 321 underwent SNP genotyping, whereas unavailable clinical material precluded genotyping in 87 patients. There were no major differences in the baseline characteristics when comparing the genotyped population to the entire population (Table 1). PBMCs were the DNA source in 212 cases and FFPE tissue was the DNA source in 109 cases. PBMC DNA was subjected to genotyping by both commercially available TM technology and TG proprietary pyrosequencing technology. Using TM for genotyping the FCGR3A SNP, three cases could not be genotyped and 18 cases were miscalled (using sequencing as the gold standard) for an accuracy rate of 90% (191/212). Using TG genotyping for the FCGR3A SNP, there were nine cases that could not be genotyped and one case miscalled for an accuracy rate of 95% (202/212). Using TM for genotyping the FCGR2A SNP, five cases could not be genotyped and three cases were miscalled for an accuracy rate of 96% (204/212). Using TG for genotyping the FCGR2A SNP, there were no failures to genotype and there was one miscall for an accuracy rate of 99% (211/212). Including the 109 FFPE cases and the 212 PBMC cases, and after adjudicating discrepant PBMC cases by sequencing, the final FCGR3A and FCGR2A genotype frequencies were VV 14%, VF 45%, FF 40% and HH 28%, HR 47%, rituximab re-treatment (RR) 22%, respectively.

The overall response rate (ORR) to initial rituximab was 71%. The likelihood of obtaining a complete response or any response was not correlated with FCGR3A genotype (VV vs. VF vs. FF, or VV vs. F carrier; Table 2). Similarly, the likelihood of obtaining a complete response or any response was not correlated with FCGR2A genotype (HH vs. HR vs. RR, or HH vs. R carrier; Table 3). Additionally, no combination of genotypes (e.g., VV/HH vs. FF/RR) was associated with complete response or ORRs (data not shown).

Genotyping was performed in 235 of 289 randomized patients, RR (n = 115) or to MR (N = 120). With a median follow-up of 5.5 years, the 3-year response duration in the RR arm and the MR arm was 50% and 78%, respectively. The FCGR3A genotype was not associated with response duration in the RR (P = 0.92) or the MR (P = 0.58) treatment arms (Fig. 2A). Similarly, the FCGR2A genotype was not associated with response duration in the RR (P = 0.19) or the MR (P = 0.61) treatment arms (Fig. 2B). Additionally, no combination of genotypes (e.g., VV/HH vs. FF/RR) influenced DOR or survival (data not shown).

Based on this analysis of a treatment-naïve population of low tumor burden FL, we conclude that the two candidate missense SNPs in FCGR3A (rs396991) and FCGR2A (rs1801274), alone or in combination, does not predict the likelihood or the durability of response to single agent rituximab. These data are in contrast to three early reports examining this question and suggesting these polymorphisms impact the efficacy of rituximab.

In a study with a patient population similar to ours, Cartron and colleagues, determined the FCGR3A and FCGR2A genotypes in 49 patients receiving rituximab (3). The objective response rates at 12 months were 90% for VV patients and 51% for F carriers (P = 0.03). Progression-free survival (PFS) at 3 years was 56% for VV patients and 35% for F carriers (P = NS). There was no impact of the FCGR2A polymorphism on outcome. A retrospective analysis of 87 FL patients receiving single agent rituximab at Stanford University demonstrated that FCGR3A V/V patients (n = 13) and FCGR2A H/H patients (n = 20) experienced higher response rates and more durable remissions compared with FCGR3A F carriers and FCGR2A H carriers, respectively (19). Finally, a retrospective analysis from the SAKK (Swiss group for Clinical Cancer Research), in which patients with follicular and mantle cell lymphoma were treated with single agent rituximab, followed by no further treatment or MR, the FCGR3A V/V genotype was associated with superior event-free survival (4). Because these studies were retrospective, relatively small with a heterogeneous FL population, the true impact of these polymorphisms remained unclear. Analysis of the RESORT patients was an ideal opportunity to evaluate these findings in a large prospective study of homogeneous FL patients. The findings from the present work should be considered definitive and highlight the importance of prospective studies in homogenous populations and with adequate power to identify predictive biomarkers.

Consistent with our findings, another large prospective study evaluating these polymorphisms in FL patients receiving single agent rituximab found no association with response rate or response duration (21). This United Kingdom sponsored intergroup trial randomized low tumor burden and asymptomatic FL to “watch and wait” versus rituximab induction alone versus rituximab induction followed by rituximab maintenance. No difference was seen in the 257 patients for whom FCGR genotyping was available with respect to CR rate, time to next treatment, or PFS.

Analyses of previously untreated high tumor burden FL patients treated with combination rituximab-chemotherapy, similarly found no impact of these SNPs on response rate, progression, or overall survival (22, 23). The PRIMA study (24), which showed an improvement in PFS after MR in patients who were treated with immunochemotherapy induction, was designed with prespecified objectives to clarify the role of FCGR polymorphisms in this context. The PRIMA investigators also found no impact of different FCGR genotypes on response rates to initial immunochemotherapy or MR, nor any difference in PFS at any timepoint (22).

Our results were somewhat unexpected given the strong preclinical data suggesting the FCGR3A 158V/F polymorphism positively influences both the binding of IgG as well as the expression of CD16 by NK cells (18, 19). It is possible that the models do not accurately recapitulate the human tumor environment or it is possible that ADCC is not the major mechanism of cell killing after rituximab therapy. Rituximab's precise mechanism of action remains ill-defined and could include components of ADCC, complement-mediated cytotoxicity, and direct killing (13, 15, 25, 26). Direct effects attributed to rituximab include inhibition of cell proliferation, induction of phosphatidylserine translocation, cell signaling via increased phosphorylation, and induction of apoptosis (27).

Many factors can influence FL response to rituximab, including biologic heterogeneity, clonal evolution, tumor bulk, prior treatments, and host factors. This analysis attempted to minimize many of these factors, by evaluating a relatively homogeneous FL population with low tumor burden, no prior therapy, and protocolized treatment and follow-up of patients. Despite this, we were unable to find a differential response based on FCGR genotype at the two candidate loci, as hypothesized by preclinical and small clinical studies. Although this does not preclude a role of other genetic variation in these genes in treatment response and outcomes, none has been identified to date. Given these definitive results, we conclude that the selection of patients for single agent rituximab therapy should not be based upon these FCGR3A or FCGR2A genotype status and these data raise additional questions regarding the mechanism of rituximab cytotoxicity.

J.R. Cerhan is a consultant/advisory board member for Genentech. R.D. Gascoyne reports receiving speakers bureau honoraria from Seattle Genetics, and is a consultant/advisory board member for Celgene and Janssen. S.J. Horning has ownership interest (including patents) in Roche. B. Kahl is a consultant/advisory board member for Roche.

This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute.

Conception and design: M. Lewis, M.E. Williams, S.J. Horning, B.S. Kahl

Development of methodology: L. Sullivan, M.E. Williams, B.S. Kahl

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.R. Cerhan, L. Sullivan, M.E. Williams, R.D. Gascoyne, S.J. Horning, B.S. Kahl

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V.P. Kenkre, F. Hong, J.R. Cerhan, M. Lewis, L. Sullivan, M.E. Williams, R.D. Gascoyne, B.S. Kahl

Writing, review, and/or revision of the manuscript: V.P. Kenkre, F. Hong, J.R. Cerhan, M.E. Williams, R.D. Gascoyne, S.J. Horning, B.S. Kahl

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.R. Cerhan, M. Lewis, M.E. Williams

Study supervision: M.E. Williams, S.J. Horning, B.S. Kahl

This study was coordinated by the ECOG-ACRIN Cancer Research Group (Robert L. Comis, MD, and Mitchell D. Schnall, MD, PhD, Group Co-Chairs) and supported in part by Public Health Service Grants CA180820, CA180794, CA180799, CA180790, CA180847, CA180816 and from the NCI, NIH, and the Department of Health and Human Services. This work was supported by a grant from the Lymphoma Research Foundation and the University of Wisconsin Forward Lymphoma Fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.