Germline variation in complement genes and event-free survival in follicular and diffuse large B-cell lymphoma

Authors


  • Conflict of interest: The authors have no known conflicts of interest.

Abstract

The complement pathway plays a central role in innate immunity, and also functions as a regulator of the overall immune response. We evaluated whether polymorphisms in complement genes are associated with event-free survival (EFS) in follicular lymphoma (FL) and diffuse large B-cell (DLBCL) lymphoma. We genotyped 167 single nucleotide polymorphisms (SNPs) from 30 complement pathway genes in a prospective cohort study of newly diagnosed FL (N = 107) and DLBCL (N = 82) patients enrolled at the Mayo Clinic from 2002 to 2005. Cox regression was used to estimate hazard ratios (HRs) for individual SNPs with EFS, adjusting for FLIPI or IPI and treatment. For gene-level analyses, we used a principal components based gene-level test. In gene-level analyses for FL EFS, CFH (P = 0.009), CD55 (P = 0.006), CFHR5 (P = 0.01), C9 (P = 0.02), CFHR1 (P = 0.03), and CD46 (P = 0.03) were significant at P < 0.05, and these genes remained noteworthy after accounting for multiple testing (q < 0.15). SNPs in CFH, CFHR1, and CFHR5 showed stronger associations among patients receiving any rituximab, while SNPs from CD55 and CD46 showed stronger associations among patients who were observed. For DLBCL, only CLU (P = 0.001) and C7 (P = 0.03) were associated with EFS, but did not remain noteworthy after accounting for multiple testing (q>0.15). Genes from the regulators of complement activation (CFH, CD55, CFHR1, CFHR5, CD46) at 1q32–q32.1, along with C9, were associated with FL EFS after adjusting for clinical variables, and if replicated, these findings add further support for the role of host innate immunity in FL prognosis. Am. J. Hematol. 2012. © 2012 Wiley Periodicals, Inc.

Introduction

There were an estimated 65,540 new cases of NHL and 20,210 deaths due to this disease in the United States in 2010 [1]. The most common NHL subtypes include diffuse large B cell lymphoma (DLBCL) and follicular lymphoma (FL). The more aggressive DLBCL is currently treated with the standard rituximab, cyclophosphamide, hydroxydaunorubicin, vincristine, and prednisone (R-CHOP). While this treatment has improved progression-free and overall survival, some patients are refractory to this therapy and relapse or develop resistance to anti-CD20 [2]. FL on the other hand is an indolent B-cell lymphoma with diverse initial management approaches including observation, chemotherapy (single agent or combination), immunochemotherapy, immunotherapy (e.g., rituximab), radioimmunotherapy, and radiation [3–5]. Regardless of treatment type, patients will often experience repeated relapses or transformation to aggressive lymphoma [6]. Mechanisms involved in treatment failure are not clearly understood, so further investigation of factors that potentially influence response to treatment is necessary.

Unlike traditional chemotherapeutic agents, rituximab is not generally thought to be toxic to tumor cells per se, but rather relies on host immune actions for its cytotoxic effects [7, 8]. In addition to mediating tumor clearance in the absence of rituximab, constituents of the complement pathway are potentially key factors in the immune response related to removal of CD20 positive cells via this therapy, either directly through complement-dependent cytotoxicity (CDC) [8–11] or, to a lesser extent, indirectly through complement-dependent cellular cytotoxicity (CDCC) [12, 13]. However, several studies have found a contradictory role for complement pathway components, which are also capable of protecting tumor cells through inhibition of NK-mediated cell death [14, 15] and enhancing myeloid-derived suppressor cell migration and activity [16].

The complement cascade is kept in check by production of natural inhibitors of the complement system, known as complement regulatory proteins (CRPs). These include membrane bound CD46, CD55, and CD59, as well as soluble CFH, CFHR1, and CFHR5, shown to be expressed in several different malignancies [17–19]. Functional importance of these CRPs has been reported by at least two groups, who found enhanced CDC in prostate cancer, breast cancer, and erythroleukemia cell lines with siRNA knock-down of CD46, CD55, and CD59 [20] and increased susceptibility to rituximab induced CDC with CD55 knock-down in rituximab resistant primary lymphoma cells [21].

To date, aside from one study of a single C1qA polymorphism [22], no one has extensively examined the relationship of polymorphisms in the complement pathway with NHL prognosis. Prior studies by our group found a significant association with SNPs in components of the complement pathway and risk of developing NHL [23]. Given the role of the complement pathway in immune mediated destruction of lymphoma cells, we evaluated whether genetic variation in this pathway might be associated with event-free survival (EFS) in FL and DLBCL patients.

Methods

Study population

This study was reviewed and approved by the Human Subjects Institutional Review Board at the Mayo Clinic, and written informed consent was obtained from all participants. Full details of this prospective cohort study of lymphoma outcomes have been previously published [24]. Briefly, since 2002 we offered enrollment to consecutive, newly diagnosed (within 9 months of first diagnosis) patients age 18 and older into the Molecular Epidemiology Resource, which is part of the University of Iowa/Mayo Clinic Lymphoma Specialized Program of Research Excellence (SPORE). All pathology was centrally reviewed; baseline clinical, laboratory, and treatment data were abstracted using a standard protocol. Participants provided a peripheral blood sample, and DNA was extracted using a standard procedure (Gentra, Minneapolis, MN). All patients were systematically followed every 6 months for the first 3 years, and then annually thereafter. Disease progression, retreatment, and deaths were verified through medical record review.

This analysis included FL (N = 107) and DLBCL (N = 82) patients enrolled onto the Mayo component of the Molecular Epidemiology Resource through 2005 and who were genotyped as part of another project to assess the role of immune and inflammatory genes in the etiology of NHL [25].

Genotyping

The SNPs from the complement genes reported here were from the ParAllele (now Affymetrix) Immune and Inflammation SNP panel that included 1,253 genes and 9,412 SNPs; full details on methods and quality control have been published earlier for the entire genotyping project [25] as well as the 30 complement genes [23]. Briefly, tagging SNPs were selected using release 16 of HapMap data. Tagging SNPs covered 5 kb up and downstream of each gene with minor allele frequency of (MAF) of ≥0.05 and a pairwise r2 threshold of 0.8. Genotyping was conducted at the Affymetrix facility using molecular inversion probe technology [26]. Overall sample success rate was 98.75%, the assay call rate was 99.13%, and the concordance rate (based on 48 blind duplicates) was 98.95%.

Statistical analysis

EFS was defined as the time from diagnosis to disease progression, re-treatment, or death due to any cause. Patients without an event were censored at time of last known follow-up. A subtype-specific clinical risk score was used to adjust for a number of prognostic factors in the analyses. For FL, the risk score combines the effects of FLIPI and treatment category. For DLBCL, the risk score combines the effects of IPI and treatment category. The treatment categories are provided in Supporting Information Table S1.

We first conducted a gene-level test of association with EFS for each subtype. Briefly, the SNPs from each gene were examined in a principal components analysis, and then the top principal components were incorporated into a Cox proportional hazards regression model. A multiple degree of freedom likelihood ratio test was used to assess significance of the principal components with survival, after adjustment for the clinical risk score.

To further examine significant gene results, hazard ratios (HR) and 95% confidence intervals (CI) for each SNP were estimated using Cox proportional hazards regression, adjusted for the subtype-specific clinical risk score. HRs and CIs were estimated separately for those with one copy and two copies of the minor allele versus those with two copies of the major allele. The test for trend was based upon the Wald chi-square test, in which the ordinal SNP variable was examined in a model. These analyses were conducted with SAS version 9.2 (SAS Institute). Haploview version 4.2 was used to visually examine LD patterns within a gene. Genes with P-value less than 0.05 were considered significant. Q-values were also calculated using R version 2.12 to assess the effects of multiple testing separately by subtype.

Results

Patient characteristics

There were 107 newly diagnosed FL patients and 82 DLBCL patients available for analysis. The median age at diagnosis for FL was 61 (range 25–85) years and 67 (range 25–83) years for DLBCL. There were slightly more males in both subtypes (52% in FL; 57% in DLBCL). Forty-two percent of FL patients had a FLIPI score of 0 or 1, 36% a score of 2, and 22% a score of 3–4, while 38% of DLBCL patients had an IPI of 0–1, 26% a score of 2, 28% a score of 3, and 9% a score of 4–5 (Supporting Information Table S1). The most common initial therapeutic approaches in FL were observation (41%), CVP based chemotherapy (22%), anthracycline-based therapy (14%), and rituximab monotherapy (8%). Thirty-three percent (N = 35) of FL patients received rituximab, either alone or in addition to other therapies. For DLBCL, nearly 80% of patients were initially treated with immunochemotherapy (R-CHOP). The median follow-up of patients still alive was 63 months (range 2–99), and 56% of FL (N = 60) and 48% (N = 39) of DLBCL patients had an event.

Complement gene-level associations with FL and DLBCL EFS

In gene-level analysis (Supporting Information Table S2), CFH, CD55, CFHR1, CFHR5, CD46, and C9 were all significantly (P < 0.05) associated with FL EFS after adjustment for the clinical-risk score, which includes FLIPI and treatment. These six genes all had a q < 0.15, suggesting these associations have a low likelihood of being false positives. For DLBCL EFS, only CLU and C7 had P-values <0.05 at the gene level and both had q > 0.15.

Based on gene level results, we next focused on SNP-level associations for FL (Table I) and DLBCL (Supporting Information Table S3) for genes with P < 0.05 in the gene level tests; all SNP level associations for FL and DLBCL are reported in Supporting Information Tables S4 and S5, respectively.

Table I. SNP-Level Associations with FL EFS (P-Trend <0.05) from Significant Genes (P < 0.05 from Gene-Level Test)
GeneSNPaMAFGenotypeAll follicular patientsFollicular patients treated with any rituximabbFollicular patients observed
N patients% EventsHR (95% CI)cPdN patients% EventsHR (95% CI)cPdN patients% EventsHR (95% CI)cPd
  • a

    Reference sequence (rs) ID from dbSNP.

  • b

    Patients received rituximab, either alone or in combination with other therapies.

  • c

    Adjusted for clinical risk score, which accounts for the effects of treatment type and FLIPI (FL) or IPI (DLBCL).

  • d

    P-trend.

  • e

    CFHR5 SNP rs12092294 in LD with rs6694672 (HR and P value are the same for both).

  • f

    CD55 SNP rs4844591 in LD with rs2564978 (HR and P value are the same for both).

  • g

    CD9 SNP rs700228 was associated with EFS (P = 0.006) but only one patient was a minor allele carrier (heterozygote).

CFHrs37664040.14TT79481.00 (reference) 26191.00 (reference) 30631.00 (reference) 
TC26772.13 (1.22, 3.71) 8758.69 (2.30,32.82) 14711.30 (0.60, 2.82) 
CC21005.54 (1.29,23.79) 110025.07 (1.24,505.6) 0NANA 
Total10756 0.0013534 7E-044466 0.51
TC + CC28792.25 (1.31, 3.87) 9789.49 (2.59,34.82) 14711.30 (0.60, 2.82) 
CFHrs13294230.29AA53641.00 (reference) 18501.00 (reference) 19681.00 (reference) 
AG44520.55 (0.32, 0.94) 13230.35 (0.09, 1.31) 22640.70 (0.32, 1.52) 
GG9330.26 (0.08, 0.86) 30NA 3670.58 (0.13, 2.61) 
Total10657 0.0043435 0.0604466 0.32
AG + GG53490.49 (0.29, 0.82) 16190.29 (0.08, 1.10) 25640.68 (0.32, 1.44) 
CFHrs10654890.18GG73621.00 (reference) 24461.00 (reference) 29691.00 (reference) 
GT30470.47 (0.25, 0.87) 9110.19 (0.02, 1.55) 13620.65 (0.26, 1.57) 
TT4250.25 (0.03, 1.86) 20NA 2500.60 (0.08, 4.58) 
Total10756 0.0083534 0.0914466 0.33
GT + TT34440.44 (0.24, 0.81) 1190.15 (0.02, 1.26) 15600.64 (0.27, 1.50) 
CFHR1rs4367190.28AA48671.00 (reference) 15531.00 (reference) 19631.00 (reference) 
AC59470.57 (0.34, 0.96) 20200.35 (0.10, 1.18) 25681.09 (0.50, 2.34) 
CC0NANA 0NANA 0NANA 
Total10756 0.0343534 0.0904466 0.83
AC + CC59470.57 (0.34, 0.96) 20200.35 (0.10, 1.18) 25681.09 (0.50, 2.34) 
CFHR5ers66946720.07TT91521.00 (reference) 29241.00 (reference) 38631.00 (reference) 
TG16812.63 (1.41, 4.92) 6836.00 (1.59,22.67) 6832.15 (0.81, 5.71) 
GG0NANA 0NANA 0NANA 
Total10756 0.0033534 0.00834466 0.13
TG + GG16812.63 (1.41, 4.92) 6836.00 (1.59,22.67) 6832.15 (0.81, 5.71) 
CD55frs25649780.27CC55641.00 (reference) 18281.00 (reference) 25801.00 (reference) 
CT46500.57 (0.33, 0.97) 16381.15 (0.34, 3.85) 15530.28 (0.11, 0.69) 
TT6330.25 (0.06, 1.07) 11003.81 (0.31,46.31) 4250.13 (0.02, 1.02) 
Total10756 0.0093534 0.484466 0.0024
CT + TT52480.52 (0.30, 0.88) 17411.29 (0.40, 4.13) 19470.25 (0.10, 0.59) 
CD46rs24665710.43AA40451.00 (reference) 15331.00 (reference) 14431.00 (reference) 
AC42571.26 (0.68, 2.34) 13230.54 (0.13, 2.30) 19682.43 (0.88, 6.71) 
CC25721.96 (1.02, 3.77) 7572.17 (0.57, 8.24) 11914.78 (1.60,14.33) 
Total10756 0.0493534 0.404466 0.0044
AC + CC67631.49 (0.86, 2.61) 20350.96 (0.30, 3.05) 30773.01 (1.16, 7.81) 
C9grs14210940.36GG43671.00 (reference) 13461.00 (reference) 16811.00 (reference) 
GA50500.56 (0.33, 0.97) 17240.41 (0.11, 1.48) 21620.38 (0.16, 0.89) 
AA13460.44 (0.18, 1.08) 5400.67 (0.13, 3.41) 6500.34 (0.10, 1.21) 
Total10657 0.0233534 0.384367 0.032
GA + AA63490.54 (0.32, 0.90) 22270.47 (0.15, 1.49) 27590.37 (0.17, 0.82) 

Complement SNPs associated with FL EFS

For FL, the two most significant SNPs from the CFH gene (rs3766404 and rs1329423, r2 = 0.061) were intronic SNPs; rs3766404 was associated with inferior EFS (dominant model HR, HRDom = 2.25; 95% CI 1.31–3.87) and rs1329423 was associated with superior EFS (HRDom = 0.49; 95% CI 0.29–0.82). A coding non-synonymous CFH SNP (rs1065489) in exon 19, which causes a missense change of glutamic acid to aspartic acid at position 936, was also significantly associated with superior EFS in FL (HRDom = 0.44; 95% CI 0.24–0.81). However, this substitution was predicted to be benign using PolyPhen and non-damaging using SIFT analysis. When restricting to patients who had received any rituximab, the HRs for rs3766404 (HRDom = 9.49; 95% CI 2.59–34.8), rs1329423 (HRDom = 0.29; 95% CI 0.08–1.10), and rs1065489 (HRDom = 0.15; 95% CI 0.02–1.26) all strengthened, keeping in mind that these estimates were based on small numbers (N = 35). The two other coding SNPs and six other intronic CFH SNPs assessed in this study were not significantly associated with FL EFS (Supporting Information Table S4).

The CFHR1 SNP rs436719 was associated with superior FL EFS (HRDom = 0.57; 95% CI 0.34–0.96). However, we noted that this SNP was significantly out of HWE (P = 0.00008), although inspection of the clustering plots showed no obvious problems with the genotyping calls. For CFHR5, there were two statistically significant SNPs associated with inferior FL EFS, rs6694672 in the 5′ region of the gene (HRDom = 2.63; 95% CI 1.41–4.92) and rs12092294 in an intron (HRDom = 2.63; 95% CI 1.41–4.92); these two SNPs were in complete LD (Fig. 1). The associations in the CFHR1 and CFHR5 regions were stronger among the FL patients treated with any rituximab for both rs436719 (HRDom = 0.35; 95% CI 0.10-1.18) and rs6694672 (HRDom = 6.00; 95% CI 1.59-22.7).

Figure 1.

Gene structure and tagSNP mapping for SNPs in genes associated with EFS. (A) Chromosome 1q31.3–1q32 genes CFH, CFHR1, and CFHR5 and (B) chromosome 1q32 genes CD55, CR2, CR1, and CD46. Top: −log10 P-value for trend across tagSNPs for FL (dark squares) and DLBCL (open circles) EFS. Bottom: linkage disequilibrium plot of SNPs genotyped in this analysis (darker shading indicates higher r2 correlation values between SNPs; numbers are |D′| values).

In CD55, rs2564978 in the 5′ region of the gene (HRDom = 0.52; 95% CI 0.30–0.88) and rs4844591 in an intron (HRDom = 0.52; 95% CI 0.30–0.88) were associated with FL EFS, and these two SNPs were in strong LD (r2 = 0.995). The SNP rs2466571 in CD46 was associated with inferior FL EFS, although the HR estimate from the dominant model was not statistically significant (HRDom = 1.49; 95% CI 0.86–2.61). The FL patients who were observed appear to be driving the association for the CD55 and CD46 SNPs, as restriction of the analysis to this subset strengthened the HRs for rs2564978 and rs4844591 (HRDom = 0.25; 95% CI 0.10–0.59) and rs2466571 (HRDom = 3.01; 95% CI 1.16–7.81).

Finally, the intronic SNP rs1421094 in C9 (HRDom = 0.54; 95% CI 0.32–0.90) was significantly associated with FL EFS, and the associations were similar for patients receiving any rituximab or observation. A second intronic SNP in C9, rs700228, was also associated with FL EFS (P = 0.006), although only one patient had a minor allele (as a heterozygote), limiting the interpretation of this finding.

There were nine additional SNPs that had associations with FL EFS at a P-trend of <0.05, but from genes with a gene level P-value of ≥0.05. These SNPs were located in genes including CR1, CR2, MBL2, C4BPA, and C7 (Supporting Information Table S4). Of note, three of nine of the SNPs were from genes in the RCA: two SNPs from C4BPA (rs1126618, 9943268) and one SNP from CR1 (rs1408077).

Complement SNPs associated with DLBCL EFS

For DLBCL, SNP-level results for the genes significant at P < 0.05 in the gene level tests are reported in Supporting Information Table S3, noting that q-values for these genes were all >0.15, suggesting they are likely to be false positive results from multiple testing. A SNP in an untranslated region of the C7 (rs324058) was associated with DLBCL EFS (HRDom = 1.66; 95% CI 0.87–3.17), as was a SNP in an untranslated region of CLU, rs3087554 (HRDom = 0.46; 95% CI 0.21–1.00). Individual SNPs that were associated with DLBCL EFS at P-trend <0.05, but in genes with non-significant gene level results, included C5, MBL2, CD46, MASP2, C8B, and CFB (Supporting Information Table S5).

Figure 1 shows an LD plot for the region of 1q31–32 where CFH and CFHR1-CFHR5 genes are located, and shows that the SNP associations were spread across this region and included several LD blocks. CD55 is farther downstream, but is also part of the 1q31–32.1 region termed the RCA region, along with C4BPA (rs1126618, 9943268) and CR1 (rs1408077; Fig. 1).

Discussion

In this study of FL patients diagnosed from 2002 to 2005 and followed through 2010, we identified six noteworthy genes – CFH, CFHR1, CFHR5, CD55, CD46, and C9 – that predicted EFS after adjustment for clinical factors and accounting for multiple testing. Interestingly, CFH, CFHR5, CFHR1, CD55, and CD46 all belong to the “Regulators of Complement Activation” (RCA) located in the 1q32 region, with the primary function of keeping the complement cascade in check. We also identified two genes – C7 and CLU – that predicted DLBCL EFS after adjustment for clinical factors, although not after adjustment for multiple testing, suggesting a high probability of false positive findings. Complement pathway polymorphisms appeared to play a larger role in FL than DLBCL EFS. This could be due to the fact that the study had 20% more FLs and therefore somewhat greater power to detect associations. Important biological differences between the two subtypes are another likely explanation. For example, FL seems to rely heavily on factors in the microenvironment to thrive [27], whereas DLBCL may be less dependent on the environment for survival. In addition, differing treatment regimens might explain some of the variation, in context of DLBCL being more uniformly treated.

Because FL treatments were heterogeneous, we carried out a sensitivity analysis in subgroups of patients treated with any rituximab and those who received no initial treatment (observation). Although this analysis had limited power, SNPs in CFH, CFHR1, and CFHR5 showed stronger associations for patients receiving any rituximab; SNPs in CD55 and CD46 showed stronger associations for patients in the observation group; and C9 SNP association was similar in both groups. It is important to emphasize the limited numbers of rituximab treated (N = 35) and observed (N = 44) FL patients in this study. This also limited our ability to address differences in specific chemotherapies, or combination versus monotherapy, making these results preliminary. Nonetheless, these findings do raise the hypothesis that the CFH family of proteins modifies the response to rituximab mediated killing, while CD55 and CD46 may only impact progression in patients that do not receive treatment.

CFH is the central negative regulator of the alternative pathway, and limits the common complement pathway as well by hindering amplification by C3b [28]. The importance of this protein in containing complement mediated inflammation has been demonstrated in pigs deficient in CFH that develop uncontrolled plasma C3 activation [29]. One of the CFH SNPs related to superior FL EFS in our study, rs1065489, has also been shown in a recent GWAS study to reduce risk of meningococcal disease [30]. This polymorphism results in an amino acid change from glutamic acid to aspartic acid. Although our PolyPhen and SIFT alignment analysis did not predict this CFH variant to be damaging, the disease associations suggest it could have reduced function and therefore allow increased complement activity, which would explain both improved complement mediated bacterial clearance and tumor killing. CFHR1, CFHR4, and CFHR5 are located near CFH (Fig. 1) and encode proteins closely related to CFH. CFHR5 is most structurally similar to CFH and was shown to bind C3b in vitro [31]. In our study, we observed statistically significant associations with CFHR1 and CFHR5 variants and EFS in FL, although little is known about any putative functions for these SNPs.

CD46, also known as membrane cofactor protein, regulates complement by binding C3b and C4b, targeting them for degradation by Factor I [32]. We observed a modest association with CD46 intronic SNP rs2466571 and inferior FL EFS. This association was stronger when we restricted the study to patients who received no initial therapy. In addition to its role in complement regulation, CD46 also functions in T regulatory 1 (Tr1) cell differentiation [33]. Regulatory T cells have been associated with suppression of anti-tumor immune responses and poor prognosis in other malignancies [34], although the association of these cells with FL survival has been variable [35–37]. Little is known about the function of rs2466571, but we speculate that if it affects CD46 expression or function, it may influence EFS in untreated patients through either traditional complement regulatory activities or by altering Tr1 numbers or function. This SNP is also in fairly high LD with rs2488255, found in one previous study to increase risk of invasive pneumococcal disease [38]. In this study, rs2488255 was related to inferior FL EFS with borderline significance, although restricting to untreated patients did strengthen the association (data not shown).

CD55, or decay accelerating factor, is a membrane protein that regulates the complement system by increasing the rate of decay of C3/C5-convertase and thus provides a mechanism for tumor cells to evade complement attack. Indeed, CD55 expression on colorectal and breast cancer cells is associated with poor prognosis [39, 40]. In addition, this factor has been shown to abrogate NK cell-mediated cytotoxicity [41] and increase resistance to CDC in rituximab-treated cells [21]. The CD55 rs2564978 T/rs3841376 insertion haplotype has been reported to have higher transcriptional activity in an adenocarcinoma cell line than the rs2564978 C/rs3841376 deletion haplotype [42]. In our study, we did not assess rs3841376, but did find that the T variant of rs2564978 was associated with better prognosis, an association that was especially notable in untreated patients. If this variant is associated with elevated CD55, we speculate based on findings by Markiewski et al. [16], that it might improve EFS by reducing levels of C3/C5-convertase and one of the resultant byproducts of its activity, C5a, which has the ability to recruit and promote suppressive actions of MDSCs.

C9 is the final component of the membrane attack complex (C5b-9) which forms a pore in the cell membrane that can lead to cell lysis. Individuals with C9 deficiencies largely due to nonsense mutations have increased susceptibility to meningococcal infections [43, 44], although among infected individuals, mortality is lower in those with the C9 deficiency when compared with those with normal C9 [45]. We found an association with EFS and C9 at the gene level, and observed superior FL EFS with the minor allele of rs1421094 C9 SNP, which did not appear to vary between those treated with rituximab versus observation. Little has been reported about the potential function of this intronic variant.

In our prior case–control analysis C5 and C9 complement pathway genes were most strongly associated with risk of developing NHL overall, and risk associations were similar for FL and DLBCL [23]. C5 is near TRAF1, and both of these are in a region also associated with rheumatoid arthritis risk [46]. Most of the SNPs from C5 associated with risk were not associated with EFS for either FL or DLBCL. One C5 SNP, rs1017119, which was not associated with FL risk, was associated with inferior FL EFS, although only three patients carried a copy of the minor allele. Two SNPs in C9 (rs261752 and rs3776526) that were related to reduced risk of NHL [23], were associated with inferior EFS in DLBCL (Supporting Information Table S5). This raises the working hypothesis that genetic variation in the complement pathway may differentially impact lymphoma risk versus prognosis.

Strengths of this study included the prospective enrollment of consecutive, newly diagnosed FL and DLBCL patients and central pathology review of all cases. We also had detailed collection of key clinical prognostic and treatment data, extensive quality control of the genotyping, and prospective, systematic follow-up of all patients for EFS, and multivariate adjustment for FLIPI (FL) and IPI (DLBCL). We were able to evaluate the major genes in the complement pathway, although SNP coverage varied and was low for some genes. For low coverage genes in particular, we cannot rule out gene-level associations with EFS. There may also be other genetic mechanisms (e.g., copy number variation) that could predict EFS that we did not assess. The major limitation was the lack of replication in an external dataset. FL treatments were heterogeneous, and power was low for treatment-specific analyses. We did not assess overall survival due a limited number of events and relatively short follow-up time. The study was observational and unmeasured confounding could impact the results, although we were able to adjust for the standard prognostic factors. Unlike the more ideal standardized evaluation performed in clinical trials, treatment, and follow-up were based on routine clinical practice. The study population also lacked racial and ethnic diversity, which increases internal validity (limits population stratification), but decreases generalizability beyond white patients of European descent.

In conclusion, the results of this study highlight the importance of the complement pathway in FL and perhaps DLBCL EFS. Variants in C9 and complement regulatory genes CFH, CFHR5, CFHR1, CD46, and CD55 were associated with FL EFS, while C7 and CLU were associated with DLBCL EFS. Preliminary results further suggest that CFH, CFHR5, CFHR1 polymorphisms had a stronger influence on EFS in rituximab treated FL patients, while C9, CD46, and CD55 polymorphisms had a stronger influence on EFS in FL patients that received no initial treatment. If replicated, these findings may be important from a clinical perspective, as they suggest that patients with different complement pathway genotypes may respond differently to rituximab therapy or may have shorter EFS when no treatment is administered. These initial results also support prior laboratory findings and raise questions about interactions between these genotypes and treatment. We are currently conducting follow-up studies in a larger patient population, where we will be able to address some of these questions.

Acknowledgements

The authors thank Biospecimens Accessioning Processing (BAP) Laboratory.

Ancillary