Risk of human T-lymphotropic virus type I-associated diseases in Jamaica with common HLA types

Authors


  • The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the U.S. Government.

  • The authors have no commercial or other associations that might pose a conflict of interest.

Abstract

Human T-lymphotropic virus-I (HTLV-I) causes adult T-cell leukemia/lymphoma (ATL) and HTLV-associated myelopathy/tropical spastic paraparesis (HAM/TSP). We postulated a higher disease risk for people with common human leukocyte antigen (HLA) types, due to a narrower immune response against viral or neoplastic antigens, compared to people with uncommon types. HLA class-I (A,B) and class-II (DRB1, DQB1) allele and haplotype frequencies in 56 ATL patients, 59 HAM/TSP patients and 190 population-based, asymptomatic HTLV-I-infected carriers were compared by logistic regression overall (score test) and with odds ratios (ORs) for common types (prevalence >50% of asymptomatic carriers) and by prevalence quartile. HTLV-I proviral load between asymptomatic carriers with common versus uncommon types was compared by t-test. ATL differed from asymptomatic carriers in overall DQB1 allele and class-I haplotype frequencies (p ≤≤ 0.04). ATL risk was increased significantly with common HLA-B (OR 2.25, 95% CI 1.19–4.25) and DRB1 (OR 2.11, 95% CI 1.13–3.40) alleles. Higher prevalence HLA-B alleles were associated with higher ATL risk (OR 1.14 per quartile, ptrend = 0.02). Asymptomatic carriers with common HLA-B alleles had marginally higher HTLV-I proviral load (p = 0.057). HAM/TSP risk did not differ consistently with common HLA types. Thus, ATL risk, but not HAM/TSP risk, was increased with higher prevalence HLA-B alleles. Perhaps breadth of cellular immunity affects risk of this viral leukemia/lymphoma. © 2007 Wiley-Liss, Inc.

Human T-lymphotropic virus type I (HTLV-I) causes adult T-cell leukemia/lymphoma (ATL) and HTLV-I-associated myelopathy, also known as tropical spastic paraparesis (HAM/TSP), but over 95% of HTLV-I-infected individuals remain asymptomatic lifelong.1 Differences in human leukocyte antigen (HLA) genes are believed to affect the risk of disease,2, 3, 4, 5, 6 but reported associations have been heterogeneous. For example, HLA A26 and A24 were associated with ATL and HAM/TSP, respectively, in Japan,2 whereas A36 was associated with ATL in Jamaica.4 In Japan, HAM/TSP risk was increased with DRB1*0101 and B*5401, and it was decreased with A02 and Cw08 (reviewed in Ref.5). The exception to this heterogeneity may be the class II haplotypes, DRB1*1101-DQB1*0301 and DRB1*1501-DQB1*0602, which were associated with ATL in both Jamaica and Japan.2, 3

A unifying theory is that HLA alleles associated with HAM/TSP, in contrast to alleles associated with ATL, elicit strong cytotoxic T-lymphocyte (CTL) responses against the Tax viral oncoprotein.2, 6, 7 A complementary perspective is that greater HLA diversity conveys selective advantage against disease because the immune response is elicited by a greater variety of antigens, as described for human immunodeficiency virus (HIV) and AIDS.8, 9 With HIV, HLA class I heterozygotes progress more slowly to AIDS than do homozygotes.8 Moreover, HIV viral load is significantly lower with rare HLA class I supertypes and perhaps alleles.9, 10 HTLV-I is transmitted primarily by cell-to-cell contact,11 and transmission from mother to infant is directly related to burden of infection in the mother (proviral load)12 and is significantly increased for mothers and infants who are highly HLA concordant.13

In the present study, we examined HLA gene polymorphisms of 305 Jamaican subjects with HTLV-I infection, postulating that people with higher prevalence HLA types may develop an immune response against a narrower range of cell-associated viral or neoplastic antigens and consequently may be at higher risk of ATL or HAM/TSP.

Methods

Study subjects

Of 105 ATL and 90 HAM/TSP patients, identified through nation-wide disease registries and referrals to the university clinic in Jamaica,3 HLA typing was performed on the 52 ATL and 55 HAM/TSP patients with sufficient DNA. Four additional ATL cases and 4 additional HAM/TSP cases from Trinidad were included for a total of 56 ATL and 59 HAM/TSP cases. There were 24 with acute, 17 with lymphoma, 8 with chronic, 2 with smoldering and 5 with unclassified ATL subtypes. For comparison, from participants in a nationwide serosurvey who resided in Kingston and Clarendon parishes, HLA was performed on 199 of 201 asymptomatic carriers who had no evidence of ATL or HAM/TSP as determined by complete blood count and differential, questionnaire and physical examination.14 The cases and asymptomatic carriers were unrelated and had similar age and geographic distributions.

Informed consent was obtained from all participants. Study protocols followed the human experimentation guidelines of the US Department of Health and Human Services and Institutional Review Board approvals at the National Cancer Institute and University of the West Indies.

Laboratory assays

HTLV-I seropositivity was determined by whole virus enzyme-linked immunoassay (EIA) (Organon Teknika, Durham, NC) and confirmed by a Western blot (Cambridge Biotech, Rockville, MD, or Genelabs Diagnostics, Singapore). Antibody titers were determined by EIA with the 5-fold end-point dilution method (Genetic Systems, Seattle WA, or Cambridge-Biotech, Rockville MD).

DNA was extracted from cryopreserved lymphocytes using the PureGene DNA Isolation Kit (Gentra Systems, Minneapolis MN) according to manufacturer's instructions. Provirus load was measured by real-time PCR, using a 7700 ABI Prism Sequence Detection System, as described.15 DNA typing of HLA class I (A and B loci) alleles was performed by a modified PCR-SSP (sequence-specific primers) method. The specificity of HLA alleles was confirmed by the PCR sequence-specific oligonucleotide (SSO) probe method. High-resolution (allele level) HLA class II genotyping was performed using the SSO probe typing protocols developed by the 13th International Histocompatibility Workshop (http://www.ihwg.org/protocols/protocol.htm). HLA-DRB1 and DQB1 genes were amplified using locus-specific PCR primers flanking exon 2, the polymorphic segments of the class II genes. The 300-bp PCR products were blotted onto nylon membranes and hybridized with a panel of SSO probes. HLA alleles were assigned by the reaction patterns of the SSO probes based on known HLA sequences. For samples with ambiguous SSO results, exon 2 was sequenced.

Statistical analysis

Nine asymptomatic carriers were excluded from analysis because 1 or more of their HLA alleles could not be unambiguously assigned. The frequencies of HLA alleles and haplotypes (henceforth referred to as types) of the remaining asymptomatic carriers were compared separately to the ATL and HAM/TSP patients. Logistic regression models that assume an additive effect of the alleles and haplotypes on homologous chromosomes were the primary analysis method.16 Separate models were constructed for each case group as the dependent variable and with each HLA locus or haplotype class as the exposure. With the referent group in each model being the aggregate of types defined as rare (less than 5% for alleles; less than 4 and 2% for class II and I haplotypes, respectively, among asymptomatic carriers), a global score test was used to detect overall differences in allele and haplotype frequencies between cases and asymptomatic carriers. Each model also provided an odds ratio (OR), 95% confidence interval (CI) and two-sided p-value for each individual allele or haplotype. p < 0.05 were considered significant, and p < 0.01 were considered highly significant. Other than adjustment for all variables (the types at each locus or haplotype) in each model, no additional adjustment for multiple comparisons was performed.

A second set of logistic regression models was developed to assess whether risk of ATL or HAM/TSP was associated with more or less common HLA types. For this, each type was assigned a value equal to its prevalence in the control population, and each individual's level of exposure was the sum of the 2 prevalence values on his/her homologous chromosomes. For example, among asymptomatic carriers, 64 had HLA-A*02, 24 had A*33 and only 6 had A*29. The prevalence score for all participants with A*02, A*33 was 64 + 24 = 88; the prevalence score for those with A*33, A*29 was 24 + 6 = 30. OR and 95% CI were calculated for these prevalence scores in quartiles and dichotomized as “common” (above median) versus uncommon; trend in OR across quartiles also was tested. The dichotomized analysis was repeated for the 2 ATL subtypes (acute and lymphoma) with a sufficient number of cases.

t-tests were used to compare differences in HTLV-I provirus load (log10 transformed) between asymptomatic carriers with common (above median prevalence score) versus uncommon HLA types. Untransformed provirus load values are presented. Differences in age and sex between asymptomatic carriers with and without provirus load measurements were compared by t-test and Fisher's exact test, respectively. All p-values were two-sided. Analyses were performed with SAS version 9.1 (SAS Institute, Cary, NC).

Results

HLA class I (A and B loci) and class II (DRB1 and DQB1 loci) alleles were typed in 305 HTLV-I infected participants. Table I presents the frequencies of individual alleles and haplotypes, ordered by their prevalence in 190 asymptomatic carriers and compared to 56 ATL and to 59 HAM/TSP patients. In multivariate logistic regression models, ATL patients differed from asymptomatic carriers in the overall distributions of HLA-DQB1 alleles and class I (A-B) haplotypes (omnibus p = 0.04 and 0.03, respectively). In the same models, ATL was associated with 7 individual alleles and 1 haplotype. Two of these associations were highly significant (p < 0.01): A*03 (OR 0.24, CI 0.08–0.73) and DQB1*0501 (OR 0.24, CI 0.09–0.59).

Table I. Frequencies of HLA Alleles and Haplotypes in 56 ATL and 59 HAM/TSP Patients and 190 HTLV-I-Infected Asymptomatic Carriers in Jamaica
HLA allele or haplotypeAsymptomatic carriers (N = 380)ATL patients (N = 112)HAM/TSP patients (N = 118)
  • HLA, human leukocyte antigen; ATL, adult T-cell leukemia/lymphoma cases; HAM/TSP, HTLV-associated myelopathy/tropical spastic paraparesis cases.

  • 1

    Values within parentheses indicate percentages.

  • 2

    p = 0.012 to 0.05 by multivariate logistic regression.

  • 3

    p ≤ 0.01 by multivariate logistic regression.

  • 4

    Referent group for multivariate logistic regression analysis. Alleles with frequency <5%, class I (A-B) haplotypes with frequency <2%, and class II (DRB1-DQB1) haplotypes with frequency <4% in the 190 asymptomatic carriers. These rare alleles were: A*01, 11, 24, 25, 26, 29, 31, 32, 35, 36, 53, 66 and 80; B*08, 13, 14, 18, 27, 37, 39, 40, 41, 45, 47, 49, 50, 51, 52, 78, 81 and 82; DRB1*0101, 0103, 0304, 0305, 0401, 0403, 0404, 0405, 0410, 0415, 0802, 0806, 0901, 1001, 1002, 1102, 1104, 1118, 1126, 1134, 1201, 1202, 1301, 1303, 1304, 1305, 1307, 1312, 1313, 1314, 1315, 1318, 1401, 1402, 1404, 1409, 1410, 1414, 1415, 1502, 1504, 1601 and 1602; and DQB1*0302, 0303, 0401, 0502, 0503, 0603, 0604, 0605, 0606, 0607, 0608 and 0609.

  • 5

    Comparison of overall frequency distributions among ATL versus asymptomatic carriers and among HAM/TSP versus asymptomatic carriers by score test from multivariate logistic regression.

A*0264 (16.8)125 (22.3)28 (23.7)
A*3051 (13.4)14 (12.5)12 (10.2)
A*6846 (12.1)10 (8.9)17 (14.4)
A*2341 (10.8)6 (5.4)27 (5.9)
A*0337 (9.7)4 (3.6)316 (13.6)
A*7429 (7.6)9 (8.0)5 (4.2)
A*3324 (6.3)9 (8.0)11 (9.3)
A*3421 (5.5)6 (5.4)3 (2.5)
Rare467 (17.6)29 (25.9)19 (16.1)
Omnibus p-values5 0.140.22
B*5353 (14.0)22 (19.6)214 (11.9)
B*1547 (12.4)19 (17.0)218 (15.3)
B*3534 (9.0)8 (7.1)17 (14.4)
B*0732 (8.4)10 (8.9)6 (5.1)
B*4432 (8.4)13 (11.6)8 (6.8)
B*5722 (5.8)9 (8.0)8 (6.8)
B*5822 (5.8)4 (3.6)9 (7.6)
B*4221 (5.5)7 (6.3)6 (5.1)
Rare4117 (30.8)20 (17.9)32 (27.1)
Omnibus p-values5 0.220.54
DRB1*150137 (9.7)20 (17.9)29 (7.6)
DRB1*030133 (8.7)8 (7.1)9 (7.6)
DRB1*070131 (8.2)8 (7.1)17 (14.4)
DRB1*030228 (7.4)7 (6.3)7 (5.9)
DRB1*150326 (6.8)9 (8.0)2 (1.7)2
DRB1*110125 (6.6)10 (8.9)6 (5.1)
DRB1*080424 (6.3)7 (6.3)6 (5.1)
DRB1*010219 (5.0)5 (4.5)11 (9.3)
DRB1*130219 (5.0)4 (3.6)6 (5.1)
Rare4138 (36.3)34 (30.4)45 (38.1)
Omnibus p-values5 0.630.25
DQB1*060296 (25.3)35 (31.1)17 (14.4)2
DQB1*020184 (22.1)25 (22.3)28 (23.7)
DQB1*050168 (17.9)9 (8.0)322 (18.6)
DQB1*030164 (16.8)16 (14.3)227 (22.9)
DQB1*040229 (7.6)7 (6.3)7 (5.9)
Rare439 (10.3)20 (17.9)17 (14.4)
Omnibus p-values5 0.040.13
A*02-B*3511 (2.9)3 (2.7)9 (7.6)2
A*68-B*5313 (3.4)0 (0)4 (3.4)
A*03-B*159 (2.4)0 (0)4 (3.4)
A*36-B*539 (2.4)9 (8.0)21 (0.9)
A*02-B*078 (2.1)4 (3.6)2 (1.7)
A*30-B*428 (2.1)5 (4.5)0 (0)
A*33-B*538 (2.1)4 (3.6)3 (2.5)
Rare4314 (82.6)87 (77.7)95 (80.5)
Omnibus p-values5 0.030.29
DRB1*1501-DQB1*060237 (9.7)19 (17.0)9 (7.6)
DRB1*0301-DQB1*020131 (8.2)7 (6.3)8 (6.8)
DRB1*0701-DQB1*020130 (7.9)8 (7.1)17 (14.4)
DRB1*0302-DQB1*040225 (6.6)6 (5.4)5 (4.2)
DRB1*1503-DQB1*060225 (6.6)6 (5.4)2 (1.7)2
DRB1*0804-DQB1*030121 (5.5)5 (4.5)6 (5.1)
DRB1*0102-DQB1*050118 (4.7)5 (4.5)11 (9.3)
DRB1*1101-DQB1*060216 (4.2)8 (7.1)3 (2.5)
Rare4177 (46.6)48 (42.9)57 (48.3)
Omnibus p-values5 0.620.13

HAM/TSP did not differ from asymptomatic carriers in overall type distributions; it was associated with 2 alleles and 2 haplotypes, none of which had p < 0.01 (Table I). Two of these alleles were in the same class II haplotype (DRB1*1503-DQB1*0602) associated with a substantial but barely statistically significantly reduced risk of HAM/TSP (OR 0.22, CI 0.05–0.99).

Table II presents ATL and HAM/TSP risk by quartile and median prevalence scores of the alleles and haplotypes among the asymptomatic carriers. The common (above median) HLA-B alleles were associated with a significantly increased risk of ATL (OR 2.25, CI 1.19–4.25), including both the acute ATL subtype (OR 3.26, CI 1.24–8.58) and the lymphoma ATL subtype (OR 3.54, CI 1.11–11.23). The risk of ATL, including all subtypes, increased significantly with each quartile of B-allele prevalence score (OR 1.41, CI 1.06–1.86 per quartile, ptrend = 0.02). The common HLA-DRB1 alleles also were associated with a significantly increased risk of ATL (OR 2.11, CI 1.13–3.40), including the acute subtype (OR 3.00, CI 1.14–7.89) but not the lymphoma subtype (OR 1.83, CI 0.65–5.16). There was no trend in overall ATL risk across the quartiles of DRB1 prevalence score (ptrend = 0.15). In contrast to ATL, HAM/TSP risk differed only with the most common quartile of DQB1 alleles (OR 0.35, CI 0.14–0.88); this quartile included the DQB1*0602 allele mentioned earlier. Neither ATL nor HAM/TSP risk differed with common HLA-A alleles, A-B haplotypes or DRB1-DQB1 haplotypes (Table II).

Table II. Odds Ratio (OR) and 95% Confidence Interval (CI) for ATL or HAM/TSP by Frequency of HLA Alleles and Haplotypes among HLTV-I-Infected Asymptomatic Carriers
Allele or haplotype prevalence in asymptomatic carriers1ATLHAM/TSP
OR95% CIOR95% CI
  • 1

    Values in parentheses are the prevalence scores in the control population; quartiles, from least to most frequent alleles (A, B, DRB1, DQB1) or haplotypes (A-B, DRB1-DQB1), are presented.

  • 2

    p = 0.05.

  • 3

    p = 0.01.

  • 4

    p = 0.02.

  • 5

    p = 0.03.

A quartile 1 (2–56)1.00Referent1.00Referent
A quartile 2 (57–78)0.810.36–1.821.140.48–2.73
A quartile 3 (79–96)0.660.28–1.561.190.50–2.86
A quartile 4 (97–128)0.790.35–1.781.520.67–3.47
A, uncommon (≤ median)1.00Referent1.00Referent
A, common (> median)0.810.44–1.471.270.71–2.28
B quartile 1 (10–38)1.00Referent1.00Referent
B quartile 2 (39–54)1.080.38–3.030.550.21–1.43
B quartile 3 (55–73)2.200.87–5.561.540.70–3.37
B quartile 4 (74–106)2.4620.99–6.111.010.44–2.31
B, uncommon (≤ median)1.00Referent1.00Referent
B, common (> median)2.2531.19–4.251.650.90–3.00
DRB1 quartile 1 (6–31)1.00Referent1.00Referent
DRB1 quartile 2 (32–41)0.350.11–1.050.560.23–1.37
DRB1 quartile 3 (42–53)1.510.67–3.381.470.69–3.12
DRB1 quartile 4 (54–74)1.270.56–2.890.520.21–1.29
DRB1, uncommon (≤ median)1.00Referent1.00Referent
DRB1, common (> median)2.1141.13–3.401.270.71–2.28
DQB1 quartile 1 (17–104)1.00Referent1.00Referent
DQB1 quartile 2 (105–147)0.580.22–1.541.570.70–3.53
DQB1 quartile 3 (148–163)0.540.23–1.260.940.43–2.05
DQB1 quartile 4 (164–192)0.750.36–1.560.3550.14–0.88
DQB1, uncommon (≤ median)1.00Referent1.00Referent
DQB1, common (> median)0.790.43–1.430.5040.28–0.91
A-B quartile 1 (2–4)1.00Referent1.00Referent
A-B quartile 2 (5–7)0.520.22–1.250.220.09–0.54
A-B quartile 3 (8–11)0.720.30–1.720.410.18–0.94
A-B quartile 4 (12–24)0.650.27–1.560.510.23–1.12
A-B, uncommon (≤ median)1.00Referent1.00Referent
A-B, common (> median)1.010.56–1.830.970.54–1.74
DRB1-DQB1 quartile 1 (2–19)1.00Referent1.00Referent
DRB1-DQB1 quartile 2 (20–31)1.830.77–4.381.030.44–2.42
DRB1-DQB1 quartile 3 (32–45)0.970.40–2.371.040.47–2.27
DRB1-DQB1 quartile 4 (46–74)1.180.49–2.830.690.29–1.63
DRB1-DQB1, uncommon (≤ median)1.00Referent1.00Referent
DRB1-DQB1, common (> median)0.780.43–1.410.860.48–1.54

HTLV-I proviral load was measured in 35 (63%) ATL patients, 38 (64%) HAM/TSP patients and 133 (70%) asymptomatic carriers. Asymptomatic carriers with and without a proviral load measurement did not differ significantly with respect to age and sex. Proviral load was higher in ATL and HAM/TSP (27,835 and 11,834 copies/106 cells, respectively) than in asymptomatic carriers (2,300 copies/106 cells).1 As shown in Table III, common HLA-B alleles among asymptomatic carriers were associated with a marginally higher proviral load (3,569 versus 1,432 copies/106 cells, p = 0.057). Otherwise, mean load among asymptomatic carriers did not differ between common and uncommon types (p = 0.19 to 0.56). Data were too sparse to evaluate proviral load with individual types.

Table III. Association of Common Versus Uncommon HLA Alleles and Haplotypes with HTLV-I Proviral Load among HTLV-I-Infected Asymptomatic Carriers
HLA type, prevalenceNo. testedHTLV-I proviral load (copies/106 cells)
  • 1

    t-tests on log10 transformed values.

A, uncommon611,955
A, common722,639
 p-value1 0.54
B, uncommon641,432
B, common693,569
 p-value 0.057
DRB1, uncommon721,723
DRB1, common613,233
 p-value 0.19
DQB1, uncommon561,631
DQB1, common772,952
 p-value 0.22
A-B, uncommon751,833
A-B, common583,084
 p-value 0.27
DRB1-DQB1, uncommon722,018
DRB1-DQB1, common612,682
 p-value 0.56

Discussion

Antigen recognition by polymorphic HLA class I and class II molecules is likely to be fundamental in the immune response to and control of HTLV-I infection, thus affecting the likelihood of progression to ATL or HAM/TSP. Our results support this hypothesis. Our major finding was that higher prevalence (common) B alleles were significantly associated with a higher risk of ATL and with higher HTLV-I proviral load in the asymptomatic carriers, although the latter did not meet nominal statistical significance. Both acute and lymphoma subtypes of ATL were increased with common B alleles.

The association of ATL risk with common B alleles does not negate the possible effect of individual alleles or haplotypes. In our multivariate logistic regression analysis, we compared the overall distribution of the individual types, finding that ATL differed from the asymptomatic carriers in DQB1 alleles and A-B haplotypes. Two alleles (A*03 and DQB1*0501) were associated with a highly significant, 4-fold lower risk of ATL. These associations have not been noted previously, but differences in populations and in statistical analysis methods make comparisons with the published literature difficult. As 1 example, with our model that adjusted for the frequencies of the other class II haplotypes, DRB1*1501-DQB1*0602 was not significantly more frequent in ATL than in asymptomatic carriers, in contrast to previous unadjusted analyses.2, 3 Some other alleles were simply too rare in the Jamaican population to evaluate. A meta-analysis or large consortium study with standardized laboratory and statistical methods would be needed for better insight on the role of individual alleles and haplotypes in the pathogenesis of ATL and HAM/TSP.

In contrast to individual types, we did have good statistical power to assess whether more common types were associated with disease risk. Given that HTLV-I is tightly cell associated and viral protein is expressed on the cell surface,11 HLA can be postulated to play an important role in recognizing such cells as foreign and mounting an effective immune response against them. Indeed, Biggar et al. noted that concordance in HLA class I alleles between a mother and her infant was a strong risk factor for transmission by breast feeding.13 Specifically, among infants who were breast fed for at least 12 months, HTLV-I infection occurred in 14% of those who differed from their mother at all 3 alleles (HLA-A, B and C), 16% of those who differed at 2 alleles and 38% of those who differed at 0 or 1 allele.13

All participants in the current analysis were infected with HTLV-I, primarily a consequence of breast feeding during infancy or sexual activity as an adult.1 Although HLA may play a weaker role in controlling established infection than it does in preventing infection, we observed that HTLV-I proviral load, which is a major determinant of ATL and HAM/TSP risk, tended to be higher among our asymptomatic carriers who had common HLA B alleles. This is in the postulated direction, because individuals with common types would recognize fewer antigens as foreign.

Our findings follow those noted with HIV, for which progression to AIDS was faster among HLA class I homozygotes, probably because the immune response is narrower and elicited by a smaller variety of antigens compared to the responses of heterozygotes.8 It seems likely progression to AIDS also would be faster with common alleles, as simple probabilities imply that common alleles must be overrepresented among homozygotes. In support of this, but from the other end of the spectrum, 3 rare HLA supertypes (B7, B27 and B58) were reported by Trachtenberg et al. to be associated with a lower HIV load and consequently a reduced risk of AIDS.9 Similarly, a study in Thailand noted that HIV load was lower with uncommon class I alleles.10 The authors hypothesized that people with common HLA alleles are less able to control infection because fewer antigens in the viral envelope would be recognized as foreign and targeted for CTL attack. If our current findings are independently confirmed, it would suggest that HLA class I diversity affects the risk of ATL, perhaps by limiting the proliferation of HTLV-I-infected cells in vivo.

In contrast to ATL, HAM/TSP risk was reduced with a class II haplotype, DRB1*1503-DQB1*0602, an association that barely reached nominal statistical significance. The lack of a consistent association with more common HLA types suggests a fundamentally different pathogenesis for HAM/TSP, perhaps mediated in part by individual alleles that elicit strong CTL responses that cross-react with epitopes in the central nervous system.2, 6, 7

In addition to our study's limited ability to investigate the role of individual alleles and haplotypes, our findings may not generalize to all of Jamaica because we could not include cases with rapidly lethal disease, particularly ATL. Nonetheless, the ATL cases were ascertained and recruited from the nation-wide lymphoma registry in Jamaica and therefore should be relatively representative of the population. Survival bias would have little effect on our HAM/TSP cases, because this is a chronic disease with low short-term mortality. Population-based sampling for HTLV-I infected asymptomatic carriers is a strength for our study. A large fraction of the asymptomatic carriers lacked an HTLV-I proviral load measurement, but this did not appear to bias our results.

Differences between populations in the natural history of HTLV-I infection and in associations with HLA are well described (reviewed in Refs.1,2, and5). Across populations, elevated levels of HTLV-I provirus and serum antibodies are associated consistently with ATL and HAM/TSP.1 Our finding that common B alleles were associated with a 2-fold higher risk of ATL and a 0.4 log10 higher HTLV-I proviral load suggests that the presence of immunocytes capable of recognizing diverse epitopes may reduce the risk of evolution to frank leukemia/lymphoma, perhaps through control of HTLV-I-infected cells.17 Ultimately, a large study that takes other genetic and environmental influences into account will be needed to disentangle the roles of HLA and other host factors in controlling HTLV-I infection and in determining the risk of ATL and HAM/TSP.

Acknowledgements

The authors are grateful to Dr. Noreen Jack, Dr. Courtney Bartholomew and Dr. Farley Cleghorn for contributing the Trinidad cases; Dr. Eric Engels for advice on the analysis; and Dr. Hui-lee Wong for review of the article.

Ancillary