HLA class I and II diversity contributes to the etiologic heterogeneity of non-Hodgkin lymphoma subtypes

Abstract

A growing number of loci within the human leukocyte antigen (HLA) region have been implicated in non-Hodgkin lymphoma (NHL) etiology. Here, we test a complementary hypothesis of “heterozygote advantage” regarding the role of HLA and NHL, whereby HLA diversity is beneficial and homozygous HLA loci are associated with increased disease risk. HLA alleles at class I and II loci were imputed from genome-wide association studies (GWAS) using SNP2HLA for: 3,617 diffuse large B-cell lymphomas (DLBCL), 2,686 follicular lymphomas (FL), 2,878 chronic lymphocytic leukemia/small lymphocytic lymphomas (CLL/SLL), 741 marginal zone lymphomas (MZL), and 8,753 controls of European descent. Both DLBCL and MZL risk were elevated with homozygosity at class I HLA-B and -C loci (OR DLBCL=1.31, 95% CI=1.06–1.60; OR MZL=1.45, 95% CI=1.12–1.89) and class II HLA-DRB1 locus (OR DLBCL=2.10, 95% CI=1.24–3.55; OR MZL= 2.10, 95% CI=0.99–4.45). Increased FL risk was observed with the overall increase in number of homozygous HLA class II loci (p-trend<0.0001, FDR=0.0005). These results support a role for HLA zygosity in NHL etiology and suggests that distinct immune pathways may underly the etiology of the different NHL subtypes.

INTRODUCTION

Genome-wide association studies (GWAS) have identified a growing list of common susceptibility loci modestly associated with risk of non-Hodgkin lymphomas (NHLs) including several HLA (human leukocyte antigen) genetic variants on chromosome 6p21, a region that is critical for innate and adaptive immune responses. Putative NHL susceptibility loci either directly implicate genes within the Major Histocompatibility Complex (MHC) or appear in strong linkage disequilibrium (LD) with extended HLA haplotypes (1–5). Interestingly, there is little convincing overlap of the identified HLA susceptibility loci among the NHL subtypes, diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), marginal zone lymphoma (MZL) and chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), suggesting that disparate aspects of the MHC and resulting immune responses are involved in the etiology of each NHL subtype.

The HLA genes are the most polymorphic in the human genome and specific HLA loci determine the antigens that are bound by antigen presenting cells (e.g., B cells and dendritic cells) and presented to T cells to elicit immune responses. Functionally, HLA molecules are critical for the host immune response. HLA class I molecules present foreign antigens primarily to cytotoxic T-cells that in response kill these target cells, while HLA class II molecules stimulate antibody production in response to specific antigens.

Reduced diversity, as defined by homozygosity at each co-dominant HLA loci, might adversely affect the host’s ability to recognize a more diverse array of foreign antigens and thereby increase subsequent disease burden. This concept is supported by a priori research that has examined effects of HLA zygosity on infectious disease, whereby a lack of HLA class I and II diversity has been associated with increased risk HIV and hepatitis B virus infection (6–8).

Given the growing evidence that genetic variation within HLA genes play in the etiology of NHL subtypes (1–4, 9), we specifically aimed to test whether lack of HLA diversity - as measured by HLA homozygosity – was associated with increased NHL risk. Specifically, we posit that associations with HLA Class II, which primarily presents peptides derived from extracellular sources, would implicate a role in infectious disease etiology. On the other hand, associations with HLA Class I, which primarily presents peptides derived from intracellular sources, would suggest a role in related conditions, such as autoimmune or atopic conditions. We present here results from a pooled analysis of 25 studies from North America, Europe, and Australia where we measured the associations between HLA class I and/or class II zygosity and four main NHL subtypes.

MATERIALS AND METHODS

Study sample

Our study sample comprises the same study participants of European descent that were included in the original GWAS efforts from which 25 studies participated. Specifically, adults diagnosed with incident, non-HIV-related B-cell NHL of mostly European descent, ascertained from cancer registries, clinics, or hospitals or through self-report were included and where diagnoses were verified by medical and pathology reports (1–4). Study designs included prospective cohort studies, population- and hospital-based case-control studies, and clinic-based studies. Original details of design methods for each study and of each GWAS have been described previously (1–4).

This study was approved by the City of Hope Institutional Review Board. Each participating study obtained approval from human subjects review committees and written informed consent from all participants. A de-identified pooled dataset with individual-level data on genotypes, demographic characteristics, and NHL subtypes of cases was provided by the InterLymph Data Coordinating Center (Mayo Clinic, Rochester, MN).

Genotyping

GWAS platforms used include the Illumina 317K, Illumina HumanHap 610K, Illumina HumanHap 660W, Illumina Human CNV370-Duo BeadChip, Affymetrix SNP 6.0, and the Illumina OmniExpress (Table 1). Quality control metrics employed (e.g., QQ plots and Eigenstrat results) and main results of each GWAS have been previously described in-depth (1–4).

Table 1.

Genome wide association studies (GWAS) included in the evaluation of human leukocyte antigen (HLA) homozygosity and risk of four non-Hodgkinlymphoma (NHL) subtypes: diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), chronic lymphocytic leukemia/small lymphocytic lymphoma(CLL/SLL), and marginal zone lymphoma (MZL).

Study Name	Study Abbreviation	GWAS Platform	NHL Cases				Controls (n=8753)
			DLBCL (n=3617)	FL (n=2686)	CLL/SLL (n=2878)	MZL (n=741)
Alpha-Tocopherol, Beta-Carotene Lung Cancer Prevention Study	ATBC	Illumina OmniExpress	43	17	50	1	238
British Columbia Non-Hodgkin Lymphoma Study	BCCA	Illumina OmniExpress	92	98	26	40	109
American Cancer Society Cancer Prevention Study-II Nutrition Cohort	CPS-II	Illumina OmniExpress	188	141	251	52	220
Treatment program of DLBCL patients from the Groupe d’Etude des Lymphomes de l’Adulte (GELA) consisting in LNH03-1B, 2B, 3B, 39B, 6B and 7B.	GELA	Illumina HumanHap 610K	549	0	0	0	0
Epidemiology & Genetics Unit Lymphoma Case-Control study	ELCCS	Illumina OmniExpress	229	182	0	0	245
Environmental and genetic risks factors study in adult lymphoma	ENGELA	Illumina OmniExpress	56	30	44	5	63
European Prospective Investigation into Cancer, Chronic Diseases, Nutrition and Lifestyles	EPIC	Illumina OmniExpress	46	46	72	8	265
Epilymph case-control study in six European countries	EpiLymph	Illumina OmniExpress	198	123	158	59	211
Genetic Epidemiology of CLL (GEC) Consortium	GEC	Affymetrix 6.0	0	0	391	0	296
Health Professionals Follow-up Study	HPFS	Illumina OmniExpress	12	5	19	5	85
Iowa-Mayo SPORE Molecular Epidemiology Resource	IOWA-MAYO SPORE	Illumina OmniExpress	146	228	242	112	0
Multicenter Italian study on gene-environment interactions in lymphoma etiology: translational aspects	Italian GxE	Illumina OmniExpress	16	16	5	6	45
Mayo Clinic Case-Control Study of NHL and CLL	MAYO-Case-Control	Illumina OmniExpress	25	245	132	75	343
Mayo Clinic Case-Control Study of NHL and CLL	MAYO-Case-Control	Illumina HumanHap 660W	393	0	0	0	172
The Melbourne Collaborative Cohort Study	MCCS	Illumina OmniExpress	71	58	57	8	75
Memorial-Sloan Kettering Lymphoproliferative Disorders Study	MSKCC	Illumina OmniExpress	175	174	36	47	4
National Cancer Institute-Surveillance, Epidemiology, and End Results Interdisciplinary Case-Control Study of Non-Hodgkin’s Lymphoma	NCI-SEER	Illumina OmniExpress	251	217	86	62	270
Nurses’ Health Study	NHS	Illumina OmniExpress	28	24	18	12	88
New South Wales non-Hodgkin lymphoma study	NSW	Illumina OmniExpress	115	146	13	34	154
New York University Women’s Health Study	NYU-WHS	Illumina OmniExpress	8	11	10	6	53
Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial	PLCO	Illumina OmniExpress	153	115	278	26	3076
Scandinavian Lymphoma Epidemiology Study	SCALE	Illumina OmniExpress	405	0	395	64	291
Scandinavian Lymphoma Epidemiology Study	SCALE	Illumina HumanHap 317K	0	376	0	0	791
Molecular Epidemiology of non-Hodgkin lymphoma 1	UCSF1	Illumina OmniExpress	38	7	22	91	10
Molecular Epidemiology of non-Hodgkin lymphoma 1	UCSF1	Illumina HumanCNV370-Duo	254	210	213	0	749
Molecular Epidemiology of non-Hodgkin lymphoma 2	UCSF2	Illumina OmniExpress	0	119	0	0	349
Utah Chronic Lymphocytic Leukemia Study	UTAH	Illumina HumanHap 610K	0	0	321	0	405
Population-based NHL case-control study in Connecticut women	YALE	Illumina OmniExpress	126	98	39	28	146

HLA imputation

As reported by Skibola et al (2), classical HLA alleles were imputed at HLA class I (HLA-A, HLA-B, HLA-C) and class II loci (HLA-DQA1, HLA-DQB1, HLA-DRB1, HLA-DPA1, HLA-DPB1 ) using SNP2HLA and a reference panel from the Type 1 Diabetes Genetics Consortium that comprised 5,225 individuals of European descent who were typed for HLA-A, B, C, DQA1, DQB1, DRB1, DPA1, DPB1 4 digit alleles. We note that the SNP2HLA reference panel is typed both for a panel of MHC SNPs and using classical HLA typing; the imputation algorithms used thus rely on both methodologies particularly when only SNPs are available. A comparison of imputed HLA alleles to 4-digit HLA sequencing data available for a subset of samples showed high concordance: HLA-A (97.3%), B (98.5%), C (98.1%) and DRB1 (97.5%). In all, 201 classical HLA alleles (two- and four-digit resolution) were successfully imputed (info score r²>0.3 for alleles) and available for analysis. Because of the strong LD between the HLA class II A1 and B1 loci (e.g., HLA-DQA1 and DQB1), we present results for each of the B1 loci (HLA-DQB1, HLA-DRB1, HLA-DPB1) since there were fewer homozygous B1 loci than A1 loci. For each HLA locus, individuals were coded as homozygote (for any allele) or heterozygote, as determined from the imputed alleles. All results presented are based on four-digit resolution.

NHL Classification

NHL subtypes were harmonized at the InterLymph Data Coordinating Center using the InterLymph Pathology Working Group guidelines (10,11), which are based on the World Health Organization classification (12).

Final analytic sample

Data for HLA loci were directly imputed from the original GWAS SNP panels and evaluated for the 3,617 DLBCL, 2,686 FL, 2,878 CLL/SLL, 741 MZL, and 8,753 controls. We note that, as with the original GWAS manuscripts, the specific numbers of controls differed by NHL subtype, due to different study inclusion and control selection criteria for each NHL subtype analyses, as described by the original GWAS publications (enumerated in Table 2).

Table 2.

Effect of homozygosity at the three HLA class I loci -A, -B and -C and three HLA class I loci -DRB1, DQB1, and DPB1 on susceptibility to four NHL subtypes (DLBCL, FL, CLL/SLL, and MZL) in Caucasian participants within participating lymphoma genome-wide association studies (analyses adjusted for sex, study or region, age, and ancestry/principal components).

		Controls (n=6912)		DLBCL (n=3617)			Controls (n=7880)		FL (n=2686)			Controls (n=7441)		CLL/SLL (n=2878)			Controls (n=5991)		MZL* (n=741)
		n	%	n	%	OR (95% CI)	n	%	n	%	OR (95% CI)	n	%	n	%	OR (95% CI)	n	%	n	%	OR (95% CI)
Class I locus
HLA-A	Heterozygote	6039	89	3096	86	1.00 (ref)	6843	88	2330	88	1.00 (ref)	6504	89	2460	87	1.00 (ref)	5244	89	649	89	1.00 (ref)
	Homozygote	756	11	484	14	1.14 (0.98–1.34)	923	12	313	12	1.03 (0.88–1.21)	821	11	378	13	1.19 (1.02–1.38)	646	11	78	11	1.06 (0.82–1.38)
HLA-B	Heterozygote	6430	93	3297	91	1.00	7330	93	2469	92	1.00 (ref)	6916	93	2656	92	1.00 (ref)	5576	93	675	91	1.00 (ref)
	Homozygote	476	7	318	9	1.22 (1.01–1.47)	544	7	216	8	1.14 (0.94–1.38)	519	7	221	8	1.04 (0.87–1.26)	411	7	66	9	1.34 (1.01–1.78)
HLA-C	Heterozygote	6238	90	3182	88	1.00	7112	90	2383	89	1.00 (ref)	6719	90	2576	90	1.00 (ref)	5414	90	651	88	1.00 (ref)
	Homozygote	674	10	435	12	1.20 (1.02–1.41)	768	10	302	11	1.13 (0.96–1.34)	722	10	301	10	1.10 (0.94–1.29)	577	10	90	12	1.33 (1.04–1.70)
Total # of homozygous Class I loci	0	5535	80	2792	77	1.00	6266	80	2121	79	1.00 (ref)	5965	80	2225	77	1.00 (ref)	4805	80	586	79	1.00 (ref)
	1	950	14	524	14	1.05 (0.90–1.21)	1120	14	361	64	0.98 (0.84–1.13)	1009	14	457	70	1.19 (1.03–1.36)	822	14	94	13	0.97 (0.76–1.23)
	2	297	4	187	5	1.33 (1.05–1.69)	342	4	132	5	1.18 (0.93–1.51)	323	4	130	5	1.08 (0.85–1.37)	256	4	37	5	1.16 (0.80–1.68)
	3	130	2	114	3	1.31 (0.95–1.81)	152	2	72	3	1.29 (0.93–1.79)	144	2	66	2	1.16 (0.83–1.62)	108	2	24	3	2.13 (1.33–3.42)
	p-trend					0.0008					0.12					0.0518					0.026
	OR per locus					1.11 (1.03–1.19)					1.06 (0.98–1.15)					1.08 (1.00–1.16)					1.08 (1.00–1.16)
Class II locus
HLA-DRB1	Heterozygote	6331	92	3173	88	1.00 (ref)	7212	92	2339	87	1.00 (ref)	6810	92	2583	90	1.00 (ref)	5500	92	663	89	1.00 (ref)
	Homozygote	561	8	435	12	1.51 (1.27–1.78)	648	8	338	13	1.54 (1.31–1.82)	608	8	286	10	1.19 (1.00–1.42)	480	8	78	11	1.45 (1.12–1.89)
HLA-DQB1	Heterozygote	6137	89	3055	84	1.00 (ref)	6999	89	2255	84	1.00 (ref)	6603	89	2494	87	1.00 (ref)	5310	89	638	98	1.00 (ref)
	Homozygote	773	11	561	16	1.30 (1.12–1.51)	879	11	431	16	1.42 (1.23–1.65)	836	11	384	13	1.20 (1.03–1.39)	681	11	10	2	1.20 (0.95–1.52)
HLA-DPB1	Heterozygote	5544	80	2817	78	1.00 (ref)	6292	80	2064	77	1.00 (ref)	5972	80	2320	81	1.00 (ref)	4809	80	582	79	1.00 (ref)
	Homozygote	1356	20	798	22	1.05 (0.93–1.19)	1576	20	620	23	1.24 (1.10–1.40)	1455	20	554	19	0.92 (0.81–1.04)	1176	20	158	21	1.13 (0.93–1.38)
Total # of homozygous Class II loci	0	4889	71	2341	65	1.00 (ref)	5545	71	1694	63	1.00 (ref)	5255	71	1994	70	1.00 (ref)	4247	71	501	68	1.00 (ref)
	1	1428	21	830	23	1.08 (0.95–1.22)	1656	21	660	25	1.28 (1.14–1.45)	1543	21	586	20	0.95 (0.84–1.08)	1241	21	159	21	1.08 (0.89–1.31)
	2	426	6	344	10	1.51 (1.25–1.83)	493	6	239	9	1.47 (1.21–1.78)	462	6	224	8	1.20 (0.99–1.46)	363	6	60	8	1.42 (1.05–1.91)
	3	136	2	91	3	1.30 (0.92–1.82)	153	2	82	3	1.89 (1.37–2.61)	143	2	62	2	1.10 (0.77–1.57)	123	2	20	3	1.48 (0.89–2.43)
	p-trend					<0.0001					<0.0001					0.1924					0.0124
	OR per locus					1.15 (1.07–1.23)					1.24 (1.15–1.32)					1.04 (0.97–1.12)					1.15 (1.03–1.28)
Total # of homozygous Class I or Class II loci	0	3972	59	1866	52	1.00 (ref)	4486	58	1390	53	1.00 (ref)	4275	59	1569	56	1.00 (ref)	3446	59	407	56	1.00 (ref)
	1	1750	26	992	28	1.11 (0.98–1.25)	2029	26	710	27	1.13 (1.00–1.28)	1880	26	767	27	1.07 (0.95–1.20)	1528	26	181	25	1.03 (0.85–1.25)
	2	625	9	407	11	1.32 (1.11–1.58)	741	10	293	11	1.22 (1.03–1.45)	683	9	284	10	1.11 (0.94–1.32)	538	9	71	10	1.16 (0.87–1.53)
	3	232	3	128	4	0.96 (0.73–1.27)	259	3	130	5	1.55 (1.19–2.00)	247	3	106	4	1.10 (0.84–1.44)	204	3	39	5	1.52 (1.04–2.21)
	4	98	1	82	2	1.92 (1.30–2.81)	114	1	58	2	1.94 (1.32–2.85)	111	2	49	2	1.06 (0.71–1.57)	83	1	16	2	1.84 (1.04–3.27)
	5+	84	1	92	3	1.72 (1.19–2.49)	100	1	50	2	1.50 (1.02–2.22)	87	1	49	2	1.57 (1.04–2.38)	70	1	12	2	1.64 (0.86–3.13)
	p-trend					<0.0001					<0.0001					0.029					0.0024
	OR per locus					1.10 (1.06–1.16)					1.13 (1.08–1.18)					1.05 (1.01–1.10)					1.12 (1.04–1.20)

^*adjusted by geographic region (continent) rather than study

Statistical analysis

Heterozygosity and homozygosity at each individual HLA locus and the number of homozygous loci for class I loci (A, B, C) and class II loci (DQB1, DRB1, DPB1) were determined; odds ratios (ORs) and 95% confidence intervals (CIs) were calculated as estimates of NHL risk with heterozygotes as the referent category, adjusted for sex, age, study, GWAS platform, and ancestry (with principal components as conducted for each subtype-specific GWAS and previously published (1–4). For analyses of MZL, adjustment by geographic region was conducted due to sample size restrictions (instead of by individual study). In addition to calculating the risk estimates for each additional number of homozygous loci, we further calculated the p-trend.

To further describe associations of zygosity by loci, we conducted joint effects analyses for HLA class I loci and class II loci. Each HLA loci (class I or II) was conducted in a stratified manner whereby heterozygotes for all loci were the referent groups and all combinations of homozygosity among the loci were evaluated. For example, to pinpoint whether HLA class I associations were attributable to HLA Class I B or C loci, we modeled as one covariate, 4 levels/combinations for HLA-B and -C (e.g., homozygous for both HLA-B and –C, homozygous only for HLA-B, homozygous for only HLA-C, and heterozygous for both), with heterozygote for both HLA-B and -C as reference (Table 3). For the associated p-trends reported in Table 3, each category is modeled based on ordinal variable in the order listed in the table, with heterozygosity at all loci as the referent group in a logistic regression model. For each p-trend, we also present the linearized additive relative-risk-per-locus, reflecting the slope of the trend-line.

Table 3.

Effects of zygosity by individual HLA Class I and Class II loci, for DLBCL, MZL, FL, and CLL/SLL (analyses adjusted for sex,age, study/region, and ancestry/principal components).

			Controls (n=6912)		DLBCL (n=3617)			Controls (n=7880)		FL (n=2686)			Controls (n=5991)		MZL* (n=741)			Controls (n=7441)		CLL/SLL (n=2878)
			n	%	n	%	OR (95% CI)	n	%	n	%	OR (95% CI)	n	%	n	%	OR (95% CI)	n	%	n	%	OR (95% CI)
Class I locus
HLA-B	HLA-C
Heterogyzote	Heterozygote		6133	89	3127	87	1.00 (ref)	6992	89	2350	88	1.00 (ref)	5321	89	637	86	1.00 (ref)	6608	89	2520	88	1.00 (ref)
Heterogyzote	Homozygote		297	4	170	5	1.07 (0.83–1.36)	338	4	119	4	1.01 (0.79–1.29)	255	4	38	5	1.28 (0.89–1.85)	308	4	135	5	1.19 (0.94–1.50)
Homozygote	Heterozygote		100	1	53	1	0.89 (0.58–1.38)	115	1	33	1	0.81 (0.51–1.28)	89	1	14	2	1.27 (0.70–2.30)	106	1	55	2	1.10 (0.75–1.62)
Homozygote	Homozygote		376	5	265	7	1.31 (1.06–1.60)	429	5	183	7	1.23 (1.00–1.52)	322	5	52	7	1.38 (1.01–1.90)	413	6	166	6	1.04 (0.84–1.29)
		p-trend					0.02					0.1258					0.02					0.43
		p-trend OR					1.08 (1.01–1.15)					1.05 (0.99–1.13)					1.12 (1.02–1.24)					1.03 (0.96–1.10)
Class II locus
HLA-DPB1	HLA-DQB1	HLA-DRB1
Heterozygote	Heterozygote	Heterozygote	4889	72	2341	65	1.00 (ref)	5545	71	1694	63	1.00 (ref)	4247	71	501	68	1.00 (ref)	5255	71	1994	70	1.00 (ref)
Heterozygote	Heterozygote	Homozygote	52	1	42	1	2.10 (1.24–3.55)	62	1	48	2	2.60 (1.66–4.06)	42	1	9	1	2.10 (0.99–4.45)	61	1	39	1	1.76 (1.09–2.86)
Heterozygote	Homozygote	Heterozygote	239	3	149	4	1.01 (0.77–1.33)	263	3	122	5	1.33 (1.03–1.73)	213	4	24	3	0.80 (0.51–1.26)	257	3	110	4	1.15 (0.88–1.49)
Heterozygote	Homozygote	Homozygote	345	5	277	8	1.54 (1.25–1.91)	403	5	195	7	1.42 (1.14–1.76)	297	5	48	6	1.43 (1.03–2.00)	378	5	172	6	1.12 (0.90–1.39)
Homozygote	Heterozygote	Heterozygote	1137	17	639	18	1.05 (0.92–1.21)	1331	17	490	18	1.21 (1.06–1.39)	986	17	126	17	1.11 (0.89–1.37)	1225	17	437	15	0.88 (0.76–1.01)
Homozygote	Heterozygote	Homozygote	28	0	24	1	1.44 (0.77–2.71)	30	0	13	0	1.33 (0.66–2.68)	18	0	1	0	0.62 (0.08–4.82)	26	0	13	0	1.15 (0.54–2.48)
Homozygote	Homozygote	Heterozygote	53	1	43	1	1.38 (0.82–2.33)	60	1	31	1	1.88 (1.14–3.10)	48	1	11	1	1.54 (0.75–3.16)	58	1	39	1	1.78 (1.11–2.85)
Homozygote	Homozygote	Homozygote	136	2	91	3	1.30 (0.92–1.82)	153	2	82	3	1.89 (1.37–2.61)	123	2	20	3	1.48 (0.90–2.44)	143	2	62	2	1.10 (0.77–1.57)
		p-trend					0.0091					<0.0001					0.04					0.95
		p-trend OR					1.04 (1.01–1.06)					1.07 (1.05–1.10)					1.04 (1.00–1.09)					1.00 (0.97–1.03)

Platform-specific results are shown in a Supplemental Table 1. Additional sensitivity analysis included evaluation of potential confounders, including evaluation of associations by previously implicated autoimmune conditions and HLA loci associated with specific NHL subtypes. We conducted stratified analysis to evaluate whether HLA zygosity associations were present among participants with and without autoimmune conditions (generally, and by specific conditions); similarly stratified analyses were conducted among participants with and without previously identified SNPs associated with NHL subtypes. We further calculated the risks, adjusting for autoimmune conditions and for all reported genetic susceptibility loci (for each NHL subtype). As neither variable altered the odds ratio >10%, those data are not presented. Analyses that restricted studies to population-based controls only also did not have measurable effect on the results. Finally, to evaluate the probability that some of our results could be due to chance, we used the Benjamini-Hochberg method to calculate the false discovery rate (FDR) and applied it to the p-trends as this allows for the fewest number of comparisons and thus degrees of freedom to assess the additive model.

Unconditional logistic regression models were applied using SAS 9.4 (SAS Institute). All tests of statistical significance were 2-sided.

RESULTS

The numbers of European cases and controls from each of the 25 studies in North America, Europe, and Australia for which HLA class I and II loci were evaluated are detailed in Table 1.

DLBCL

Elevated DLBCL risks of 20–50% were observed for homozygosity for individual HLA class I (B and C) and/or class II loci (DRB1 and DQB1) (Table 2). DLBCL risk also increased with increasing number of homozygous class I loci (p-trend=0.0008; FDR p=0.003) and class II loci (p-trend<0.0001; FDR p=0.0005) (Table 2). Although homozygosity for HLA-A had a borderline non-significant effect for increasing DLBCL risk, joint analyses suggested that the 30% risk increase observed with two or more homozygote loci (Table 2) was attributable to homozygosity at the HLA-B and -C locus (OR=1.31, 95% CI=1.06–1.60, Table 3). Similarly, for class II loci, joint analysis showed statistically significant associations for homozygosity specifically at the HLA-DRB1 locus (OR=2.10, 95% CI=1.24–3.55) as significantly increased risk was observed only in combination with homozygous HLA-DRB1 locus (Table 3).

FL

There were no significant associations between zygosity at HLA class I loci and FL risk (Table 2). Statistically significant 24–54% increases, however, were observed for FL risk for each of the three HLA class II loci. Further, FL risk increased with the total number of homozygous HLA class II loci (p-trend<0.0001; FDR p=0.0005), with an odds ratio of 1.89 (95% CI=1.37–2.61) for those fully homozygous compared with those fully heterozygous at all three HLA class II loci. Joint analyses additionally supported a statistically significant increased risk for FL with overall homozygosity at the HLA class II loci (p-trend<0.0001; FDR p=0.0005, Table 3).

MZL

Homozygosity at HLA class I loci HLA-B (OR=1.34, 95% CI=1.01–1.78) and –C (OR=1.33, 95% CI=1.04–1.70) but not -A (OR=1.06, 95% CI=0.82–1.38) increased MZL risk (Table 2). Stratified analysis supported independent associations for both HLA-B and –C and MZL (Table 3). Homozygosity at HLA class II loci increased MZL risk (Table 2), but only the association with HLA-DRB1 reached statistical significance (OR=1.45, 95% CI-1.12–1.89, Table 2). Analyses considering single locus homozygosity provided evidence of a role for HLA-DRB1 in increasing MZL risk (Table 3).

CLL/SLL

Modest CLL/SLL risk increases were observed for HLA-A (OR=1.19, 95% CI=1.02–1.38), HLA-DRB1 (OR=1.19, 95% CI=1.00–1.42) and HLA-DQB1 (OR=1.20, 95% CI=1.03–1.39) (Table 2). Increasing CLL/SLL risk was not observed with increasing number of homozygote class I or class II loci, though when evaluating total numbers of class I and II loci altogether, a borderline significant increased risk was observed for those with all five homozygote class I and II loci (OR=1.57, 95% CI=1.04–2.38, p-trend = 0.029; FDR=0.055) (Table 2). We were unable to isolate CLL/SLL associations with HLA zygosity to any singular locus (Table 3).

DISCUSSION

Based on the largest number of NHL subtypes to date for whom imputed HLA data is available, we demonstrate that HLA homozygosity plays a role in four B-cell NHL subtypes, and that the associations between homozygosity at HLA Class I and/or Class II loci are distinct by these subtypes. Specifically, FL risk was associated with homozygosity at HLA class II loci, but not Class I loci. CLL/SLL risk appeared to be associated (borderline) with homozygosity at either HLA Class I or Class II loci. In contrast, while both DLBCL and MZL were associated with zygosity at HLA Class I and Class II loci, the associations appeared specific to Class I HLA-B and –C loci and to the Class II HLA-DRB1 locus. We note that the p-trends evaluated for each additional homozygous loci remained statistically significant after adjust for multiple comparisons, with exception of that for CLL/SLL. Our results add to the growing body of literature implicating different roles for HLA class I and II loci, key modulators of human immune response, in the heterogeneous etiologies of B-NHL subtypes (1–4). Our results also add to the current literature which points to similarities in the etiologic profiles of DLBCL and MZL (13). Overall, these data support the importance of HLA diversity in NHL etiology, with the type of HLA diversity potentially varying by NHL subtype.

The underlying hypothesis regarding the role of HLA zygosity and disease is that homozygosity at HLA loci reduces the diversity of peptides that can be presented, with the hypothesis that these peptides can reflect etiologic agents such as infectious diseases, self-antigens for atopic or autoimmune conditions, and even cancerous cells. At present, there is a growing body of literature supporting that HLA heterozygotes are more resistant to infectious diseases, and the corollary, that HLA homozygotes are more susceptible to infectious diseases. Specifically, HLA class I heterozygote advantage (e.g., presenting greater diversity of antigenic peptides to CD8+ cytotoxic T lymphocytes) has been demonstrated for slowing progression to AIDS (6), whereas heterozygotes at HLA class II loci appear to have greater ability in clearing HBV infection (8) and HCV infection (14) than homozygotes. HLA-DRB1 heterozygosity has also been reported to confer favorable outcome (e.g., against end-stage liver disease) among HCV-infected liver transplant recipients (15). There are also reports evaluating HLA zygosity as a key contributor in autoimmune conditions. For example, reports of heterozygote advantage for class II loci and inflammatory bowel disease (16) and for class I loci and psoriatic arthritis (17) have both been published. Specific associations between HLA zygosity and NHL have been limited to reports of CLL. Evidence of the importance of HLA zygosity include reports that homozygosity at HLA-A, -B, and -DRB1 are associated with CLL (18) and with CLL disease progression (19–20), with the hypothesis that limited HLA diversity provided an advantage of the tumor to escape the immune response.

HLA heterozygote advantage is posited to work in concert with specific allele associations (as opposed to exclusively) (21); our results thus complement ongoing efforts that have identified the most role that specific HLA alleles have on NHL subtype risk. In sensitivity analysis, we evaluated the effect of known HLA associations and, in stratified and adjusted analysis, did not find that these associations diminish the reported association between HLA zygosity and NHL subtypes. Further evaluation into how these complementary associations act in concert are thus warranted and inclusion of HLA zygosity in the construct of genetic risk scores for each NHL subtype should be considered.

Further research to understand the association – or independence - between HLA zygosity with infections and autoimmune conditions and NHL risk are also needed (21–24). For example, efforts to evaluate autoimmune conditions linked to class II alleles (e.g., Sjögren syndrome, systemic lupus erythromatosus, and rheumatoid arthritis) (23) with class II zygosity in relation to FL risk could provide potential insight regarding immune mechanisms modulating FL risk. A particularly pressing research question is understanding what are the underlying mechanisms of individual allele-associations and how are they distinct from HLA zygosity associations. Similar efforts to identify commonalities between autoimmune and infectious disease associations with HLA loci and zygosity among other NHL subtypes are also warranted. Finally, extension of these efforts towards understanding the genetic and structural variants and HLA expression are also required to fully understand the implication of HLA-allelic associations in the context of overall class I or II zygosity.

Study strengths include the large sample size available to evaluate individual NHL subtypes which no studies have been able to do adequately to date (25). Potential study limitations include possible misclassification of HLA alleles due to imputation, although direct comparison of a subset with genotyped HLA alleles showed >97% concordance (2). While the present analysis leverages the available GWAS data through imputation of HLA alleles, we recognize that confirmation with direct HLA allelotyping may provide additional levels of information not ascertained in imputed data.

Our study’s restriction to individuals of European ancestry requires our results to be replicated for other racial or ethnic groups, as the associations may not apply universally to all ethnic groups. However, as demonstrated for HLA associations in autoimmune conditions, fine-mapping studies show that the same amino acid changes contribute to disease in both European and Asian populations (26), implicating similar underlying biologic mechanisms for disease etiology. Studies limitations also include our inability to evaluate heterogeneity within NHL subtypes, either defined molecularly, by infectious etiology, or by organ site.

In summary, our results add to the growing evidence of HLA alleles as susceptibility loci in the etiology of B-cell NHL subtypes. In addition to ongoing fine-mapping studies being conducted as follow-up to GWAS, our results here suggest that functional studies aiming to understand the underlying biology of zygosity and NHL subtype risk will also be important. Additional efforts to evaluate larger-scale zygosity, such as of immune genes and perhaps the entire genome may prove important in understanding the full extent of the role diversity of the immune response plays in lymphoma etiology.