- Email: [email protected]
HLA DR and DQ alleles and haplotypes associated with clinical response to glatiramer acetate in multiple sclerosis
Multiple Sclerosis and Related Disorders (2013) 2, 340–348
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/msard
HLA DR and DQ alleles and haplotypes associated with clinical response to glatiramer acetate in multiple sclerosis Suhayl Dhib-Jalbuta,n, Reuben M Valenzuelaa, Kouichi Itoa, Michael Kaufmanb, Mary Ann Piconec, Steve Buysked a
Department of Neurology, UMDNJ-Robert Wood Johnson Medical School, 683 Hoes lane, Piscataway, NJ 08554, USA b Carolina Medical Center, Multiple Sclerosis Center, Charlotte, NC, USA c Multiple Sclerosis Center at Holy Name Hospital, Teaneck, NJ, USA d Rutgers University, Department of Statistics and Biostatistics, Piscataway, USA Received 17 September 2012; received in revised form 23 January 2013; accepted 14 February 2013
KEYWORDS Multiple sclerosis; Glatiramer acetate; Glatiramer Acetate Biomarkers; HLA-DR; HLA-DQ; MS Biomarkers
Abstract Objective: Clinical response to immunomodulatory therapies in multiple sclerosis (MS) is variable among patients. Currently, there are no validated biomarkers of clinical response to any of the approved treatments for MS. The objective of this study was to determine if HLA-class II alleles predict the clinical response to glatiramer acetate (GA). Methods: This was a prospective study of 64 MS patients with relapsing-remitting disease. Patients were HLA-typed and classified as GA-responders or non-responders after 2 years of treatment based on a clinical criterion. Statistical models were used to determine whether HLA-DR and DQ alleles and haplotypes predict the clinical response to GA. Results: Tests of association of response singled out four alleles and two haplotypes with nominal po0.01. The presence of alleles DR15 or DQ6 or the absence of DR17 and DQ2 alleles was associated with favorable clinical response. The presence of the DR15 DQ6 haplotype and the absence of the DR17–DQ2 haplotype were also associated with favorable treatment response. A best fitting two-haplotype model resulted in the identification of three prognostic categories (good, neutral, and poor). A DR15-DQ6 positive but DR17-DQ2 negative combination was strongly predictive of a favorable clinical response (71%). Conversely, a DR15-DQ6 negative but DR17-DQ2 positive combination was strongly predictive of poor clinical response to GA (17%).
Abbreviations: MS, Multiple sclerosis; GA, Glatiramer Acetate; EDSS, Expanded Disability Status Scale; ARR, Annualized Relapse Rate; ROC, Receiver Operating Characteristics n Correspondence to: Department of Neurology, 125 Paterson Street, Suite 6200, New Brunswick, NJ 08901, USA. Tel.: +1 732 235 7335; fax: +1 732 235 8115. E-mail addresses: [email protected] (S. Dhib-Jalbut), [email protected] (R. Valenzuela), [email protected] (K. Ito), [email protected] (M. Kaufman), [email protected] (M. Ann Picone), [email protected] (S. Buyske). 2211-0348/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msard.2013.02.005
Clinical response to glatiramer acetate in multiple sclerosis
341
Conclusion: HLA-DR and DQ typing may prove to be useful biomarkers of predicting response to GA in MS and may help select patients appropriate for this treatment. & 2013 Elsevier B.V. All rights reserved.
1.
Introduction
Genetic susceptibility in multiple sclerosis (MS) has been inferred from its association with HLA class-II major histocompatibility antigens which control cell mediated immunity (Schmidt et al., 2007). DR15, DQ1 (molecularly defined as DRB1n1501, DQB1n0602) is the most frequent haplotype associated with MS (Sawcer et al., 2011). Interestingly, MS patients with the HLA-DR15 allele have heightened T-cell responses to the immunodominant myelin basic protein (MBP) peptide (84–102) (Martin et al., 1990; Pette et al., 1990; Smith et al., 1998), which underscores the interplay between HLA-related susceptibility and myelin antigens in MS. Glatiramer acetate (GA), an approved drug for the treatment of MS (Comi et al., 2001; Martinelli Boneschi et al., 2003), is a random copolymer of four amino acids (Glutamic acid, Lysine, Alanine, and Tyrosine), which are highly represented in MBP. GA competes with myelin peptides for binding to HLA-DR molecules (Fridkis-Hareli et al., 1999) as well as to the T-cell receptor (Kim et al., 2004) which results in reduction of myelin-reactive T-cell proliferation. GA also modulates antigen presenting cells (APC) (Vieira et al., 2003; Weber et al., 2007) which underlies the induction of T helper 2 (Th2) type cells and regulatory Tcells that presumably migrate to the brain and lead to in situ bystander suppression (Aharoni et al., 2003; DhibJalbut et al., 2003; Duda et al., 2000; Hussien et al., 2001; Vieira et al., 2003). Since neither the clinical response nor the induction of GA-specific Th2 cells is universal among GA treated MS patients (Farina et al., 2002), it would be important to identify predictors of clinical response to the drug. Studies by Fusco et al. (2001) and Gross et al. (Gross et al., 2011) suggested an association of HLA DRB1n1501 with a favorable clinical response to GA. We therefore conducted a prospective study to determine if HLA-DR and DQ alleles and haplotypes predict the clinical response to GA after at least 2 years of therapy.
2.
receive GA therapy were given the opportunity to participate in this study. Enrollment criteria included RRMS with at least one clinical relapse in the year preceding enrollment. Patients were examined and an Expanded Disability Status Scale (EDSS) was determined at baseline. Annualized relapse rate (ARR) at enrollment was determined historically based on chart reviews for at least the preceding 2 years. ARR and EDSS were determined again at the end of the 2 year treatment period. Sixty-four patients completed treatment with GA for at least 2 years (Table 1). Seven patients signed informed consent but did not return to participate in the study.
2.2. Classification of clinical responders and non-responders After 2 years on GA therapy, patients were classified as GAresponders (GA-R) (n= 30) or GA-non-responders (GA-NR) (n =34) based on clinical criteria that were more stringent than those reported in the literature (Cohen et al., 2004). A responder is a patient with no relapses and no evidence of disease progression as measured by EDSS at the end of two years of treatment with GA. A non-responder is a patient with one or more relapses or with progression in the EDSS of at least 1 point sustained for 6 months. The demographics of the responders and non-responders are shown in Table 2. Mean disease duration was 10.476.1 years for the GA-R group and 12.278.2 years for GA-NR group. Mean treatment duration was 26.476.6 months for the GA-R group and 26.175.9 months for the GA-NR group. Baseline ARR and EDSS were not significantly different between the two groups. However, the ARR and EDSS were significantly lower at the end of the treatment period in the GA-R group compared to the GA-NR group consistent with the classification criteria used. Overall, a 100% reduction in the ARR and a 0.45 point decrease in the EDSS occurred in the GA-R group, whereas a 10.3% reduction in ARR and a 0.35 point increase in EDSS occurred in the GA-NR group during treatment (Table 2).
Methodology 2.3.
2.1.
Cells and HLA typing
Subjects
Seventy-one patients diagnosed with definite relapsingremitting MS (RRMS) according to the McDonald criteria (McDonald et al., 2001) were enrolled and followed up at four MS centers: 15 from the University of Medicine and Dentistry of New Jersey (UMDNJ)-Robert Wood Johnson Medical School MS Center, 36 from the University of Maryland Center for MS in Baltimore Maryland, 8 from the Gimble MS Center in Teaneck, New Jersey and 12 from the Carolinas Medical Center-MS Center, Charlotte, North Carolina. The study was approved by the Institutional Review Board for Human Subjects at each Center. Patients who elected to
Approximately 60 cc of heparinized blood was obtained by venipuncture from each MS patient at enrollment. Peripheral blood mononuclear cells (PBMC) were purified using Ficoll-Hypaque gradient as described in the supplier’s protocol (ICN Biomedicals Inc. Ohio, USA). Samples from collaborating centers were sent via overnight delivery at room temperature and processed immediately upon arrival. Intermediate HLA typing for DR and DQ alleles was performed at the Tissue Transplantation Laboratory at the University of Pennsylvania using DNA-based technology as described (Gourley et al., 2002). Briefly, genomic DNA was purified from PBMC using the QIAamp blood kit (QIAGEN Inc.
342
Table 1 ID no
Age/Sex
35/M 47/F 53/F 29/F 59/F 53/F 52/F 34/F 31/F 38/F 28/F 49/F 40/F 42/F 27/M 37/M 40/M 42/F 33/M 49/F 53/F 33/M 24/M 17/F 37/F 46/F 31/F 32/F 45/M 45/F 32/F 29/F 60/F 53/F 53/F 43/F 52/F
HLA DR
8,15 7,17 15 4,13 4,7 4,15 15,17 7,17 15,18 1,15 4,13 15 11,13 1,15 13,17 11,15 1,11 7,15 4,15 7,17 7,17 13,15 11,14 11,17 7,15 13,17 13,17 4,18 11,15 1,17 11,17 11,13 1,4 13,17 4,13 11,13 9,11
HLA DQ
4,6 2 6 2,8 8,9 6,8 2,6 2 4,6 5,6 7,8 6 6,7 5,6 2,6 6,7 5,7 2,6 6,8 2 2 6,7 5,7 2,7 6,9 2,7 2,7 4,8 6,7 2,5 2,7 6,7 5,8 2,6 6,8 6,7 2,6
EDSS
ARR
Clinical response
Pre-Rx
On-Rx
Pre-Rx
On-Rx
2.5 3.0 3.0 2.0 2.5 2.0 1.0 2.5 3.5 2.5 2.5 1.0 4.0 1.0 3.5 2.0 2 2.5 1.0 4.0 1.5 1.0 1.0 1.0 2.5 2.0 2.0 1.0 2.0 1.0 2.0 2.0 1.5 2.0 2.0 1.0 1.5
0 2 1.0 3.0 2.5 2.0 2.0 2.5 2.5 3.5 2.0 1.0 4.0 1.0 2.0 2.0 3.0 2.0 1.0 6.0 2.5 1.0 2.0 1.0 2.0 2.5 2.0 2.0 1.0 1.0 2.0 2.0 2.0 2.5 2.0 1.0 1.5
0.5 0.5 0.5 2.0 0.5 0.5 0.5 1.0 0.5 1.5 0.5 0.5 1.0 0.5 1.0 1.0 1.0 1.0 0.5 1.0 1.0 1.0 0.5 0.5 1.0 1.0 0.5 1.0 1.0 0.5 1.0 0.5 1.0 0.5 0.5 0.5 1.0
0 0 0 2.0 1.0 0 0.5 0.5 0 0.5 0 0 0 1.0 0 0.5 0.5 0.5 0 0 0 0 0.0 0.5 0 1.5 0.5 1.5 1.5 0.5 1.5 0.5 0.5 0.5 0.5 0.5 1.0
R R R NR NR R NR NR R NR R R R NR R NR NR NR R NR NR R NR NR R NR NR NR NR NR NR NR NR NR NR NR NR
S. Dhib-Jalbut et al.
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37.
MS patients HLA alleles and their clinical response to Glatiramer acetate.
40/F 44/F 51/F 51/F 40/M 60/M 29/F 22/F 53/F 20/F 63/M 39/F 36/F 28/F 36/F 32/F 53/F 50/F 38/F 32/M 27/F 31/F 42/M 40/F 45/F 42/M 40/F
8,15 1,11 8,9 7 7,15 8,15 7,13 4,11 11,13 1,15 15 1,3 3,13 1,13 17 8,11 4,7 11,15 13,15 15 4,15 1,7 4,7 1,14 8,15 7,15 1,12
4,6 5,7 2,7 2,9 2,6 4,6 2,6 3,6 3,6 5,6 6 2,5 2,3 3,5 2 3,4 2,8 6,7 6 6 6,8 5,9 2,8 5 4,6 2,6 3,5
2.0 1.5 2.0 2.0 2.0 1.5 2.0 1.0 2.0 2.0 2.0 1.0 2.5 2.0 1.5 1.0 1.0 2.0 1.0 1.0 1.0 2.0 1.0 1.0 2.0 2.0 3.0
2.0 1.5 2.5 2.0 2.5 1.0 2.0 1.0 1.0 2.0 1.0 1.0 2.0 1.0 2.0 1.0 2.0 1.0 1.0 1.0 1.0 1.0 1.0 2.0 2.5 2.0 2.5
0.5 0.5 1.0 1.0 1.0 1.0 1.0 1.0 0.5 1.5 1.5 0.5 1.0 1.0 0.5 0.5 1.0 1.0 1.0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 2.0
0 0 1.5 0 1.0 0 0 0 0 0 0 0.0 1.5 0.0 0.5 0 1.5 0 0 0 0 0 0.5 0.0 0.5 0.0 1.0
R R NR R NR R R R R R R R NR R NR R NR R R R R R NR NR NR R NR
Clinical response to glatiramer acetate in multiple sclerosis
38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64.
ARR: Annualized Relapse Rate; EDSS: Expanded Disability Status Scale; R: clinical responder; NR: clinical non-responder, Rx: Treatment.
343
344
S. Dhib-Jalbut et al.
Table 2 Demographics, mean ARR and mean EDSS in the GA-responders and non-responder groups. Demographics
GA-R
GA-NR
p-value
Number of patients Mean age Sex M F Mean duration of illness (years) Mean ARR Pre-Rx 24 months Rx Mean EDSS Pre-Rx 24 months Rx
30 34 38.9711.2 41.8710.0 NS NS 8 6 22 28 10.476.1 12.278.2 NS
0.7570.31 0.8570.40 NS 0.0070.00 0.7670.53 o0.0001 1.9370.86 1.8170.70 NS 1.4870.75 2.1670.92 0.0019
Valencia, CA). HLA–DR and DQ alleles were determined using the One Lambda LABType SSO kit. The laboratory personnel were blinded as to whether the patients were clinical responders or non-responders. Haplotypes were inferred based on published HLA DR–DQ haplotype frequencies (Klitz et al., 2003).
2.4.
Statistical analysis
The R statistical environment was used for statistical analysis (R, 2011). Summary performance statistics were computed for each allele or haplotype individually. If the presence of the allele or haplotype predicted non-response, rather than response, the predictor was reversed for analysis so that all summaries relate to the prediction of response. Logistic regression with the likelihood ratio test was used to test allelic and haplotypic associations. Logistic regression with lasso regularization, a coefficient shrinkage and variable selection method, was used to analyze models with multiple predictors (Friedman et al., 2010).
3.
Results
Of the 74 enrolled subjects, 64 completed the treatment period and both their HLA alleles and response/nonresponse status were determined. The remaining 7 did not start the study and therefore were dropped from further consideration. Of those who completed the study, 30 (46.9%) were classified as responders.
3.1. GA
Figure 1 Title: Relative frequency of HLA-DR and DQ alleles by responder status. Legend: Non-responders (N =34) are in lightgray on the left in each pair; Responders (N = 30) are dark-gray on the right in each pair. Panel A shows DR alleles, while Panel B shows DQ alleles. Vertical axes indicate the proportion of nonresponders or responders with the allele present.
individual HLA alleles. As predictors of clinical response status, only DR15, DQ2, and DQ6 had both sensitivity and specificity above 0.5. Nominal p-values for association with clinical response, uncorrected for multiple testing, are also shown in Table 3. DR15, DR17, DQ2, and DQ6 were nominally significant. DR15 and DQ6 were associated with responder status (p=0.020 and 0.014, respectively) whereas DR17 and DQ2 were associated with non-responder status (p=0.012 and 0.002, respectively). No allele was significant following a Holm correction for multiple testing (Wright) except for DQ2, which approached statistical significance with an adjusted p-value of 0.0583. Other than the pair DR15 and DQ6, none of these alleles were in high linkage disequilibrium (r2o0.4), suggesting that, except for pair DR15 and DQ6, the four tests of association of these alleles with responder status have low correlation.
Individual alleles as predictors of response to
The relative frequencies of DR and DQ alleles by responder status are shown in Figure 1. As expected, DR15 and DQ6 were more common relative to other alleles, which is consistent with what has been reported in the literature (Schmidt et al., 2007). In addition these two alleles are known to be in strong linkage disequilibrium, which was also the case in our sample (r2 =0.53). Table 3 summarizes the response status by
3.2. Individual haplotypes as predictors of response to GA The relative frequencies of the most common DR–DQ haplotypes by responder status are shown in Figure 2. The DR15–DQ6 haplotype was the most common (present in 39% of all individuals) as well as the best individual haplotype predictor of treatment response. Table 4 summarizes the
Clinical response to glatiramer acetate in multiple sclerosis
Table 3
Summary of treatment response by HLA-specificities.
HLA-specificities DR1 DR3 DR4 DR7 DR8 DR9 DR11 DR12 DR13 DR14 DR15 DR17 DR18 DQ2 DQ3 DQ4 DQ5 DQ6 DQ7 DQ8 DQ9
345
NR 27 33 27 26 32 32 25 33 26 32 26 23 33 15 32 32 26 21 23 27 33
R 25 29 25 24 26 30 24 30 22 30 13 28 29 24 26 25 25 8 25 26 27
+NR
+R
Predictor
Sensitivity
Specificity
PPV
NPV
AUC
Nominal p
7 1 7 8 2 2 9 1 8 2 8 11 1 19 2 2 8 13 11 7 1
5 1 5 6 4 0 6 0 8 0 17 2 1 6 4 5 5 22 5 4 3
– + – – + – – – + – + – + – + + – + – – +
0.83 0.03 0.83 0.80 0.13 1.00 0.80 1.00 0.27 1.00 0.57 0.93 0.03 0.80 0.13 0.17 0.83 0.73 0.83 0.87 0.10
0.21 0.97 0.21 0.24 0.94 0.06 0.26 0.03 0.76 0.06 0.76 0.32 0.97 0.56 0.94 0.94 0.24 0.62 0.32 0.21 0.97
0.48 0.50 0.48 0.48 0.67 0.48 0.49 0.48 0.50 0.48 0.68 0.55 0.50 0.62 0.67 0.71 0.49 0.63 0.52 0.49 0.75
0.58 0.53 0.58 0.57 0.55 1.00 0.60 1.00 0.54 1.00 0.67 0.85 0.53 0.76 0.55 0.56 0.62 0.72 0.69 0.64 0.55
0.52 0.50 0.52 0.52 0.54 0.53 0.53 0.51 0.52 0.53 0.67 0.63 0.50 0.68 0.54 0.55 0.53 0.68 0.58 0.54 0.54
0.6876 0.9284 0.6876 0.7328 0.3052 0.1077 0.5407 0.2581 0.7725 0.1077 0.0062 0.0077 0.9284 0.0028 0.3052 0.1637 0.4939 0.0044 0.1436 0.4396 0.2372
NR =Non-responders negative for allele, R=Responders negative for allele, +NR=Non-responders positive for allele, +R=Responders positive for allele, Predictor indicates whether presence (+) or absence ( ) of allele is associated with positive response, PPV= positive predictive value, NPV= negative predictive value, AUC=area under the Receiver Operating Characteristics (ROC) curve. Nominal p indicates the p-value from logistic regression for association of the allele with response uncorrected for multiple testing. In the case of alleles with ’Predictor=-’, all statistics are shown for the reversed formulation (i.e., absences of the allele predicts response). The bold numbers are those with significant p-value.
Figure 2 Title: Relative frequency of HLA DR–DQ haplotypes by responder status. Legend: Non-responders (N=34) are in light-gray on the left in each pair; Responders (N=30) are dark-gray on the right in each pair. Vertical axes indicate the proportion of non-responders or responders with the allele present. Text below the haplotype names indicates the total number of patients with haplotype.
response status by individual HLA DR–DQ haplotypes. As a predictor of clinical response status, only DR15–DQ6 had both sensitivity and specificity above 0.5. Haplotypes DR15– DQ6 and DR17–DQ2 were both significant even after a Holm’s
correction for multiple testing; their adjusted p-values were 0.0437 and.0462, respectively. The presence of the DR15– DQ6 haplotype and the absence of the DR17–DQ2 haplotype were associated with treatment response.
346
S. Dhib-Jalbut et al.
Table 4 Haplotype DR1–DQ5 DR4–DQ8 DR7–DQ2 DR11–DQ7 DR13–DQ6 DR15–DQ6 DR17–DQ2
Summary of treatment response by haplotype. Only haplotypes with observed frequency above 0.05 are shown. NR 27 27 27 26 30 26 23
R 25 26 26 27 25 13 28
+NR
+R
Predictor
Sensitivity
Specificity
PPV
NPV
AUC
Nominal p
7 7 7 8 4 8 11
5 4 4 3 5 17 2
– – – – + + –
0.83 0.87 0.87 0.90 0.17 0.57 0.93
0.21 0.21 0.21 0.24 0.88 0.76 0.32
0.48 0.49 0.49 0.51 0.56 0.68 0.55
0.58 0.64 0.64 0.73 0.55 0.67 0.85
0.52 0.54 0.54 0.57 0.52 0.67 0.63
0.6876 0.4396 0.4396 0.1447 0.5738 0.0062 0.0077
NR= Non-responders negative for haplotype, R= Responders negative for haplotype, +NR=Non-responders positive for haplotype, +R= Responders positive for haplotype, Predictor indicates whether presence (+) or absence ( ) of haplotype is associated with positive response, PPV=positive predictive value, NPV=negative predictive value, AUC=area under the Receiver Operating Characteristics (ROC) curve. Nominal p indicates the p-value from logistic regression for association of the haplotype with response uncorrected for multiple testing. In the case of haplotypes with ’Predictor=-’, all statistics are shown for the reversed formulation (i.e., absences of the haplotype predicts response). The bold rows are those with significant p-values.
3.3.
More complex models
Age, sex, and initial disease severity were not significant as covariates and were not used in any of the analyses in this report. Neither ARR nor EDSS at baseline differed significantly (p40.05) by allele haplotype status (Supplemental Table 1), suggesting that the observed effects are not due to regression to the mean. Cross-validation of lasso-regularized logistic regression indicated that the best fit involved equal magnitude coefficients (but opposite signs) for the DR15–DQ6 and DR17– DQ2 haplotypes. This is equivalent to a score with 3 levels: 0 (DR15–DQ6 absent but DR17–DQ2 present), 1 (DR15–DQ6 and DR17–DQ2 either both present or both absent), and 2 (DR15– DQ6 present but DR17–DQ2 absent). This model has an AUC of 0.72 and is summarized in Table 5.
4.
Table 5
Simplified haplotype model.
Prognostic profile
Haplotypes
NR/R (% R)
‘‘poor predicted response’’ ‘‘neutral predicted response’’ ‘‘good predicted response’’
(DR15–DQ6) absent & (DR17– DQ2) present (DR15–DQ6) present & (DR17– DQ2) present or (DR15–DQ6) absent & (DR17–DQ2) absent (DR15–DQ6) present & (DR17– DQ2) absent
10/2 (16.7%) 17/11 (39.3%) 7/17 (70.8%)
Response rates can be compared to the overall rate of 46.9%. The AUC of the model is=0.72.
Discussion
This prospective study identified four HLA class-II alleles that can potentially predict the clinical response to GA. Tests of association of response with one allele at a time singled out four alleles with nominal po0.01: DR15, DR17, DQ2, and DQ6. The presence of the allele is associated with positive response for DR15 and DQ6, while the absence of the allele is associated with positive response for DR17 and DQ2. These associations are nominally significant in the sense that, with a Bonferroni correction for testing 21 alleles, the required critical value for p is a =0.05/21=0.0024 and none of the pvalues achieve that; DQ2 is close at p=0.0028. However, DR15 is significant if regarded as a test of replication of the results reported by Fusco et al. (2001) and Gross et al. (2011). Gross et al. suggested that the DRB1n1501 effect followed a recessive pattern since only patients homozygous for this allele had a longer event free time on GA. While our sample size is insufficient to confirm this observation, it is worth pointing out that all four patients homozygous for DR15, and all five patients homozygous for DQ6 were GAresponders. Similarly, four of the five patients homozygous for DQ2 were non-responders. Our findings differ from those of Fusco et al. and Gross et al. First, we show a complete analysis for the major HLA class-II alleles and haplotypes as shown in Tables 3 and 4,
while Fusco et al. presented results for DR15 only (when considering treatment success) and Gross et al. tested only DR15 for GA response. Second, We detected deleterious effects for DR17 and DQ2 beyond the beneficial effect for DR15 noted by Gross et al. and Fusco et al. Third, we included an analysis of the DR–DQ haplotypes. The four HLA alleles perform similarly (0.63–0.68) for the AUC, or area under the ROC curve, which can be considered a measure of the discriminating ability of the model. The AUC runs from 0.5, equivalent to a coin toss, to 1.0, corresponding to perfect prediction. Larger differences are found in the sensitivity, specificity, positive predictive value, and negative predictive value. Although it might seem that using all of the allelic information would result in better prediction, the AUC for a model including all alleles is only slightly better than that for the best two allele model, and worse if one corrects for overfitting (data not shown). The two haplotypes best at predicting treatment response, DR15–DQ6 and DR17–DQ2, give equivalent predictions to the alleles DR15 and DR17, respectively, since those alleles are found only in the two haplotypes. The other haplotypes containing DQ2 and DQ6 are not good predictors, suggesting that the performance of DQ2 and DQ6 as allelic predictors is due to their association with the DR15–DQ6 and DR17–DQ2 haplotypes.
Clinical response to glatiramer acetate in multiple sclerosis
347
Analysis of prognostic models using multiple haplotypes suggested a two-haplotype prognostic model with a simple interpretation (Table 5). Patients with two predictors of successful response, namely the presence of DR15–DQ6 and the absence of DR17–DQ2, can be considered to have a good predicted response (71% are responders). Patients with neither (that is, they are DR15–DQ6 negative but DR17-DQ2 positive) have a poor predicted response (17% are responders); patients with DR15–DQ6 and DR17–DQ2 either both present or both absent have a neutral prediction of response (39% are responders). Our findings are consistent with the mechanism of action of GA. This copolymer binds HLA class-II molecules, and therefore, influences signaling through the T-cell receptor and antigen presenting cells (Gran et al., 2000; Lalive et al., 2011) leading to the generation of Th2 cells and regulatory T-cells. Allelic differences in HLA class II molecules could result in conformational differences in the GA/class-II/TcR complex that influence immune responsiveness to GA and subsequently its clinical efficacy. Our findings need to be replicated in other data sets before clinical recommendations can be made. If validated, it is possible based on the model presented in Table 5 that a recommendation can be made to patients considered candidates for treatment with GA to be HLA-typed for DR and DQ alleles before initiation of therapy. A predicted response to GA for a particular patient can then be classified as good, neutral or poor. In a disease with variable responses to immunomodulatory agents, our findings represent a step forward towards personalized therapy in MS
Cohen BA, Khan O, Jeffery DR, Bashir K, Rizvi SA, Fox EJ, et al. Identifying and treating patients with suboptimal responses. Neurology 2004;63:S33–40. Comi G, Filippi M, Wolinsky JS. European/Canadian multicenter, double-blind, randomized, placebo-controlled study of the effects of glatiramer acetate on magnetic resonance imagingmeasured disease activity and burden in patients with relapsing multiple sclerosis. European/Canadian glatiramer acetate study group. Annals of Neurology 2001;49:290–7. Dhib-Jalbut S, Chen M, Said A, Zhan M, Johnson KP, Martin R. Glatiramer acetate-reactive peripheral blood mononuclear cells respond to multiple myelin antigens with a Th2-biased phenotype. Journal of Neuroimmunology 2003;140:163–71. Duda PW, Krieger JI, Schmied MC, Balentine C, Hafler DA. Human and murine CD4 T cell reactivity to a complex antigen: recognition of the synthetic random polypeptide glatiramer acetate. Journal of Immunology 2000;165:7300–7. Farina C, Wagenpfeil S, Hohlfeld R. Immunological assay for assessing the efficacy of glatiramer acetate (Copaxone) in multiple sclerosis. A pilot study. Journal of Neurology 2002;249:1587–92. Fridkis-Hareli M, Aharoni R, Teitelbaum D, Arnon R, Sela M, Strominger JL. Binding of random copolymers of three amino acids to class II MHC molecules. International Immunology 1999;11:635–41. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. Journal of Statistical Software 2010;33:1–22. Fusco C, Andreone V, Coppola G, Luongo V, Guerini F, Pace E, et al. HLA-DRB1n1501 and response to copolymer-1 therapy in relapsingremitting multiple sclerosis. Neurology 2001;57:1976–9. Gourley IS, Kearns J, McKeen M, Birkos S, Brown M, Kamoun M. HLA Class I typing of volunteers for a bone marrow registry: QC analysis by DNA-based methodology identifies serological typing discrepancies in the assignment of HLA-A and B antigens. Tissue Antigens 2002;59:211–5. Gran B, Tranquill LR, Chen M, Bielekova B, Zhou W, Dhib-Jalbut S, et al. Mechanisms of immunomodulation by glatiramer acetate. Neurology 2000;55:1704–14. Gross R, Healy BC, Cepok S, Chitnis T, Khoury SJ, Hemmer B, et al. Population structure and HLA DRB1 1501 in the response of subjects with multiple sclerosis to first-line treatments. Journal of Neuroimmunology 2011;233:168–74. Hussien Y, Sanna A, Soderstrom M, Link H, Huang YM. Glatiramer acetate and IFN-beta act on dendritic cells in multiple sclerosis. Journal of Neuroimmunology 2001;121:102–10. Kim HJ, Ifergan I, Antel JP, Seguin R, Duddy M, Lapierre Y, et al. Type 2 monocyte and microglia differentiation mediated by glatiramer acetate therapy in patients with multiple sclerosis. Journal of Immunology 2004;172:7144–53. Klitz W, Maiers M, Spellman S, Baxter-Lowe LA, Schmeckpeper B, Williams TM, et al. New HLA haplotype frequency reference standards: high-resolution and large sample typing of HLA DR-DQ haplotypes in a sample of European Americans. Tissue Antigens 2003;62:296–307. Lalive PH, Neuhaus O, Benkhoucha M, Burger D, Hohlfeld R, Zamvil SS, et al. Glatiramer acetate in the treatment of multiple sclerosis: emerging concepts regarding its mechanism of action. CNS Drugs 2011;25:401–14. Martin R, Jaraquemada D, Flerlage M, Richert J, Whitaker J, Long EO, et al. Fine specificity and HLA restriction of myelin basic proteinspecific cytotoxic T cell lines from multiple sclerosis patients and healthy individuals. Journal of Immunology 1990;145:540–8. Martinelli Boneschi F, Rovaris M, Johnson KP, Miller A, Wolinsky JS, Ladkani D, et al. Effects of glatiramer acetate on relapse rate and accumulated disability in multiple sclerosis: meta-analysis of three double-blind, randomized, placebo-controlled clinical trials. Multiple Sclerosis 2003;9:349–55. McDonald WI, Compston A, Edan G, Goodkin D, Hartung HP, Lublin FD, et al. Recommended diagnostic criteria for multiple
Conflict of interest Dr. S. Dhib-Jalbut received grant support and consultant fees from TEVA Pharmaceuticals, Bayer Health, Biogen-Idec and EMD-Serono. Dr. R.M. Valenzuela has nothing to disclose. Dr. K. Ito has nothing to disclose. Dr. M. Kaufman received grant support and served on the speakers’ bureau for TEVA Neuroscience. Dr. M. Picone has nothing to disclose. Dr. S. Buyske has nothing to disclose.
Acknowledgment Supported by Grants from the Wadsworth Foundation and TEVA Pharmaceuticals Inc.
Appendix A.
Supporting information
Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.msard. 2013.02.005.
References Aharoni R, Kayhan B, Eilam R, Sela M, Arnon R. Glatiramer acetatespecific T cells in the brain express T helper 2/3 cytokines and brain-derived neurotrophic factor in situ. Proceedings of the National Academy of Sciences USA 2003;100:14157–62.
348 sclerosis: guidelines from the International Panel on the diagnosis of multiple sclerosis. Annals of Neurology 2001;50:121–7. Pette M, Fujita K, Wilkinson D, Altmann DM, Trowsdale J, Giegerich G, et al. Myelin autoreactivity in multiple sclerosis: recognition of myelin basic protein in the context of HLA-DR2 products by T lymphocytes of multiple-sclerosis patients and healthy donors. Proceedings of the National Academy of Sciences USA 1990;87:7968–72. R. R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; 2011. Sawcer S, Hellenthal G, Pirinen M, Spencer CC, Patsopoulos NA, Moutsianas L, et al. Genetic risk and a primary role for cellmediated immune mechanisms in multiple sclerosis. Nature 2011;476:214–9.
S. Dhib-Jalbut et al. Schmidt H, Williamson D, Ashley-Koch A. HLA-DR15 haplotype and multiple sclerosis: a HuGE review. American Journal of Epidemiology 2007;165:1097–109. Smith KJ, Pyrdol J, Gauthier L, Wiley DC, Wucherpfennig KW. Crystal structure of HLA-DR2 (DRAn0101, DRB1n1501) complexed with a peptide from human myelin basic protein. Journal of Experimental Medicine 1998;188:1511–20. Vieira PL, Heystek HC, Wormmeester J, Wierenga EA, Kapsenberg ML. Glatiramer acetate (copolymer-1, copaxone) promotes Th2 cell development and increased IL-10 production through modulation of dendritic cells. Journal of Immunology 2003;170:4483–8. Weber MS, Prod’homme T, Youssef S, Dunn SE, Rundle CD, Lee L, et al. Type II monocytes modulate T cell-mediated central nervous system autoimmune disease. Nature Medicine 2007;13:935–43.
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/msard
HLA DR and DQ alleles and haplotypes associated with clinical response to glatiramer acetate in multiple sclerosis Suhayl Dhib-Jalbuta,n, Reuben M Valenzuelaa, Kouichi Itoa, Michael Kaufmanb, Mary Ann Piconec, Steve Buysked a
Department of Neurology, UMDNJ-Robert Wood Johnson Medical School, 683 Hoes lane, Piscataway, NJ 08554, USA b Carolina Medical Center, Multiple Sclerosis Center, Charlotte, NC, USA c Multiple Sclerosis Center at Holy Name Hospital, Teaneck, NJ, USA d Rutgers University, Department of Statistics and Biostatistics, Piscataway, USA Received 17 September 2012; received in revised form 23 January 2013; accepted 14 February 2013
KEYWORDS Multiple sclerosis; Glatiramer acetate; Glatiramer Acetate Biomarkers; HLA-DR; HLA-DQ; MS Biomarkers
Abstract Objective: Clinical response to immunomodulatory therapies in multiple sclerosis (MS) is variable among patients. Currently, there are no validated biomarkers of clinical response to any of the approved treatments for MS. The objective of this study was to determine if HLA-class II alleles predict the clinical response to glatiramer acetate (GA). Methods: This was a prospective study of 64 MS patients with relapsing-remitting disease. Patients were HLA-typed and classified as GA-responders or non-responders after 2 years of treatment based on a clinical criterion. Statistical models were used to determine whether HLA-DR and DQ alleles and haplotypes predict the clinical response to GA. Results: Tests of association of response singled out four alleles and two haplotypes with nominal po0.01. The presence of alleles DR15 or DQ6 or the absence of DR17 and DQ2 alleles was associated with favorable clinical response. The presence of the DR15 DQ6 haplotype and the absence of the DR17–DQ2 haplotype were also associated with favorable treatment response. A best fitting two-haplotype model resulted in the identification of three prognostic categories (good, neutral, and poor). A DR15-DQ6 positive but DR17-DQ2 negative combination was strongly predictive of a favorable clinical response (71%). Conversely, a DR15-DQ6 negative but DR17-DQ2 positive combination was strongly predictive of poor clinical response to GA (17%).
Abbreviations: MS, Multiple sclerosis; GA, Glatiramer Acetate; EDSS, Expanded Disability Status Scale; ARR, Annualized Relapse Rate; ROC, Receiver Operating Characteristics n Correspondence to: Department of Neurology, 125 Paterson Street, Suite 6200, New Brunswick, NJ 08901, USA. Tel.: +1 732 235 7335; fax: +1 732 235 8115. E-mail addresses: [email protected] (S. Dhib-Jalbut), [email protected] (R. Valenzuela), [email protected] (K. Ito), [email protected] (M. Kaufman), [email protected] (M. Ann Picone), [email protected] (S. Buyske). 2211-0348/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msard.2013.02.005
Clinical response to glatiramer acetate in multiple sclerosis
341
Conclusion: HLA-DR and DQ typing may prove to be useful biomarkers of predicting response to GA in MS and may help select patients appropriate for this treatment. & 2013 Elsevier B.V. All rights reserved.
1.
Introduction
Genetic susceptibility in multiple sclerosis (MS) has been inferred from its association with HLA class-II major histocompatibility antigens which control cell mediated immunity (Schmidt et al., 2007). DR15, DQ1 (molecularly defined as DRB1n1501, DQB1n0602) is the most frequent haplotype associated with MS (Sawcer et al., 2011). Interestingly, MS patients with the HLA-DR15 allele have heightened T-cell responses to the immunodominant myelin basic protein (MBP) peptide (84–102) (Martin et al., 1990; Pette et al., 1990; Smith et al., 1998), which underscores the interplay between HLA-related susceptibility and myelin antigens in MS. Glatiramer acetate (GA), an approved drug for the treatment of MS (Comi et al., 2001; Martinelli Boneschi et al., 2003), is a random copolymer of four amino acids (Glutamic acid, Lysine, Alanine, and Tyrosine), which are highly represented in MBP. GA competes with myelin peptides for binding to HLA-DR molecules (Fridkis-Hareli et al., 1999) as well as to the T-cell receptor (Kim et al., 2004) which results in reduction of myelin-reactive T-cell proliferation. GA also modulates antigen presenting cells (APC) (Vieira et al., 2003; Weber et al., 2007) which underlies the induction of T helper 2 (Th2) type cells and regulatory Tcells that presumably migrate to the brain and lead to in situ bystander suppression (Aharoni et al., 2003; DhibJalbut et al., 2003; Duda et al., 2000; Hussien et al., 2001; Vieira et al., 2003). Since neither the clinical response nor the induction of GA-specific Th2 cells is universal among GA treated MS patients (Farina et al., 2002), it would be important to identify predictors of clinical response to the drug. Studies by Fusco et al. (2001) and Gross et al. (Gross et al., 2011) suggested an association of HLA DRB1n1501 with a favorable clinical response to GA. We therefore conducted a prospective study to determine if HLA-DR and DQ alleles and haplotypes predict the clinical response to GA after at least 2 years of therapy.
2.
receive GA therapy were given the opportunity to participate in this study. Enrollment criteria included RRMS with at least one clinical relapse in the year preceding enrollment. Patients were examined and an Expanded Disability Status Scale (EDSS) was determined at baseline. Annualized relapse rate (ARR) at enrollment was determined historically based on chart reviews for at least the preceding 2 years. ARR and EDSS were determined again at the end of the 2 year treatment period. Sixty-four patients completed treatment with GA for at least 2 years (Table 1). Seven patients signed informed consent but did not return to participate in the study.
2.2. Classification of clinical responders and non-responders After 2 years on GA therapy, patients were classified as GAresponders (GA-R) (n= 30) or GA-non-responders (GA-NR) (n =34) based on clinical criteria that were more stringent than those reported in the literature (Cohen et al., 2004). A responder is a patient with no relapses and no evidence of disease progression as measured by EDSS at the end of two years of treatment with GA. A non-responder is a patient with one or more relapses or with progression in the EDSS of at least 1 point sustained for 6 months. The demographics of the responders and non-responders are shown in Table 2. Mean disease duration was 10.476.1 years for the GA-R group and 12.278.2 years for GA-NR group. Mean treatment duration was 26.476.6 months for the GA-R group and 26.175.9 months for the GA-NR group. Baseline ARR and EDSS were not significantly different between the two groups. However, the ARR and EDSS were significantly lower at the end of the treatment period in the GA-R group compared to the GA-NR group consistent with the classification criteria used. Overall, a 100% reduction in the ARR and a 0.45 point decrease in the EDSS occurred in the GA-R group, whereas a 10.3% reduction in ARR and a 0.35 point increase in EDSS occurred in the GA-NR group during treatment (Table 2).
Methodology 2.3.
2.1.
Cells and HLA typing
Subjects
Seventy-one patients diagnosed with definite relapsingremitting MS (RRMS) according to the McDonald criteria (McDonald et al., 2001) were enrolled and followed up at four MS centers: 15 from the University of Medicine and Dentistry of New Jersey (UMDNJ)-Robert Wood Johnson Medical School MS Center, 36 from the University of Maryland Center for MS in Baltimore Maryland, 8 from the Gimble MS Center in Teaneck, New Jersey and 12 from the Carolinas Medical Center-MS Center, Charlotte, North Carolina. The study was approved by the Institutional Review Board for Human Subjects at each Center. Patients who elected to
Approximately 60 cc of heparinized blood was obtained by venipuncture from each MS patient at enrollment. Peripheral blood mononuclear cells (PBMC) were purified using Ficoll-Hypaque gradient as described in the supplier’s protocol (ICN Biomedicals Inc. Ohio, USA). Samples from collaborating centers were sent via overnight delivery at room temperature and processed immediately upon arrival. Intermediate HLA typing for DR and DQ alleles was performed at the Tissue Transplantation Laboratory at the University of Pennsylvania using DNA-based technology as described (Gourley et al., 2002). Briefly, genomic DNA was purified from PBMC using the QIAamp blood kit (QIAGEN Inc.
342
Table 1 ID no
Age/Sex
35/M 47/F 53/F 29/F 59/F 53/F 52/F 34/F 31/F 38/F 28/F 49/F 40/F 42/F 27/M 37/M 40/M 42/F 33/M 49/F 53/F 33/M 24/M 17/F 37/F 46/F 31/F 32/F 45/M 45/F 32/F 29/F 60/F 53/F 53/F 43/F 52/F
HLA DR
8,15 7,17 15 4,13 4,7 4,15 15,17 7,17 15,18 1,15 4,13 15 11,13 1,15 13,17 11,15 1,11 7,15 4,15 7,17 7,17 13,15 11,14 11,17 7,15 13,17 13,17 4,18 11,15 1,17 11,17 11,13 1,4 13,17 4,13 11,13 9,11
HLA DQ
4,6 2 6 2,8 8,9 6,8 2,6 2 4,6 5,6 7,8 6 6,7 5,6 2,6 6,7 5,7 2,6 6,8 2 2 6,7 5,7 2,7 6,9 2,7 2,7 4,8 6,7 2,5 2,7 6,7 5,8 2,6 6,8 6,7 2,6
EDSS
ARR
Clinical response
Pre-Rx
On-Rx
Pre-Rx
On-Rx
2.5 3.0 3.0 2.0 2.5 2.0 1.0 2.5 3.5 2.5 2.5 1.0 4.0 1.0 3.5 2.0 2 2.5 1.0 4.0 1.5 1.0 1.0 1.0 2.5 2.0 2.0 1.0 2.0 1.0 2.0 2.0 1.5 2.0 2.0 1.0 1.5
0 2 1.0 3.0 2.5 2.0 2.0 2.5 2.5 3.5 2.0 1.0 4.0 1.0 2.0 2.0 3.0 2.0 1.0 6.0 2.5 1.0 2.0 1.0 2.0 2.5 2.0 2.0 1.0 1.0 2.0 2.0 2.0 2.5 2.0 1.0 1.5
0.5 0.5 0.5 2.0 0.5 0.5 0.5 1.0 0.5 1.5 0.5 0.5 1.0 0.5 1.0 1.0 1.0 1.0 0.5 1.0 1.0 1.0 0.5 0.5 1.0 1.0 0.5 1.0 1.0 0.5 1.0 0.5 1.0 0.5 0.5 0.5 1.0
0 0 0 2.0 1.0 0 0.5 0.5 0 0.5 0 0 0 1.0 0 0.5 0.5 0.5 0 0 0 0 0.0 0.5 0 1.5 0.5 1.5 1.5 0.5 1.5 0.5 0.5 0.5 0.5 0.5 1.0
R R R NR NR R NR NR R NR R R R NR R NR NR NR R NR NR R NR NR R NR NR NR NR NR NR NR NR NR NR NR NR
S. Dhib-Jalbut et al.
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37.
MS patients HLA alleles and their clinical response to Glatiramer acetate.
40/F 44/F 51/F 51/F 40/M 60/M 29/F 22/F 53/F 20/F 63/M 39/F 36/F 28/F 36/F 32/F 53/F 50/F 38/F 32/M 27/F 31/F 42/M 40/F 45/F 42/M 40/F
8,15 1,11 8,9 7 7,15 8,15 7,13 4,11 11,13 1,15 15 1,3 3,13 1,13 17 8,11 4,7 11,15 13,15 15 4,15 1,7 4,7 1,14 8,15 7,15 1,12
4,6 5,7 2,7 2,9 2,6 4,6 2,6 3,6 3,6 5,6 6 2,5 2,3 3,5 2 3,4 2,8 6,7 6 6 6,8 5,9 2,8 5 4,6 2,6 3,5
2.0 1.5 2.0 2.0 2.0 1.5 2.0 1.0 2.0 2.0 2.0 1.0 2.5 2.0 1.5 1.0 1.0 2.0 1.0 1.0 1.0 2.0 1.0 1.0 2.0 2.0 3.0
2.0 1.5 2.5 2.0 2.5 1.0 2.0 1.0 1.0 2.0 1.0 1.0 2.0 1.0 2.0 1.0 2.0 1.0 1.0 1.0 1.0 1.0 1.0 2.0 2.5 2.0 2.5
0.5 0.5 1.0 1.0 1.0 1.0 1.0 1.0 0.5 1.5 1.5 0.5 1.0 1.0 0.5 0.5 1.0 1.0 1.0 0.5 0.5 0.5 0.5 0.5 0.5 0.5 2.0
0 0 1.5 0 1.0 0 0 0 0 0 0 0.0 1.5 0.0 0.5 0 1.5 0 0 0 0 0 0.5 0.0 0.5 0.0 1.0
R R NR R NR R R R R R R R NR R NR R NR R R R R R NR NR NR R NR
Clinical response to glatiramer acetate in multiple sclerosis
38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64.
ARR: Annualized Relapse Rate; EDSS: Expanded Disability Status Scale; R: clinical responder; NR: clinical non-responder, Rx: Treatment.
343
344
S. Dhib-Jalbut et al.
Table 2 Demographics, mean ARR and mean EDSS in the GA-responders and non-responder groups. Demographics
GA-R
GA-NR
p-value
Number of patients Mean age Sex M F Mean duration of illness (years) Mean ARR Pre-Rx 24 months Rx Mean EDSS Pre-Rx 24 months Rx
30 34 38.9711.2 41.8710.0 NS NS 8 6 22 28 10.476.1 12.278.2 NS
0.7570.31 0.8570.40 NS 0.0070.00 0.7670.53 o0.0001 1.9370.86 1.8170.70 NS 1.4870.75 2.1670.92 0.0019
Valencia, CA). HLA–DR and DQ alleles were determined using the One Lambda LABType SSO kit. The laboratory personnel were blinded as to whether the patients were clinical responders or non-responders. Haplotypes were inferred based on published HLA DR–DQ haplotype frequencies (Klitz et al., 2003).
2.4.
Statistical analysis
The R statistical environment was used for statistical analysis (R, 2011). Summary performance statistics were computed for each allele or haplotype individually. If the presence of the allele or haplotype predicted non-response, rather than response, the predictor was reversed for analysis so that all summaries relate to the prediction of response. Logistic regression with the likelihood ratio test was used to test allelic and haplotypic associations. Logistic regression with lasso regularization, a coefficient shrinkage and variable selection method, was used to analyze models with multiple predictors (Friedman et al., 2010).
3.
Results
Of the 74 enrolled subjects, 64 completed the treatment period and both their HLA alleles and response/nonresponse status were determined. The remaining 7 did not start the study and therefore were dropped from further consideration. Of those who completed the study, 30 (46.9%) were classified as responders.
3.1. GA
Figure 1 Title: Relative frequency of HLA-DR and DQ alleles by responder status. Legend: Non-responders (N =34) are in lightgray on the left in each pair; Responders (N = 30) are dark-gray on the right in each pair. Panel A shows DR alleles, while Panel B shows DQ alleles. Vertical axes indicate the proportion of nonresponders or responders with the allele present.
individual HLA alleles. As predictors of clinical response status, only DR15, DQ2, and DQ6 had both sensitivity and specificity above 0.5. Nominal p-values for association with clinical response, uncorrected for multiple testing, are also shown in Table 3. DR15, DR17, DQ2, and DQ6 were nominally significant. DR15 and DQ6 were associated with responder status (p=0.020 and 0.014, respectively) whereas DR17 and DQ2 were associated with non-responder status (p=0.012 and 0.002, respectively). No allele was significant following a Holm correction for multiple testing (Wright) except for DQ2, which approached statistical significance with an adjusted p-value of 0.0583. Other than the pair DR15 and DQ6, none of these alleles were in high linkage disequilibrium (r2o0.4), suggesting that, except for pair DR15 and DQ6, the four tests of association of these alleles with responder status have low correlation.
Individual alleles as predictors of response to
The relative frequencies of DR and DQ alleles by responder status are shown in Figure 1. As expected, DR15 and DQ6 were more common relative to other alleles, which is consistent with what has been reported in the literature (Schmidt et al., 2007). In addition these two alleles are known to be in strong linkage disequilibrium, which was also the case in our sample (r2 =0.53). Table 3 summarizes the response status by
3.2. Individual haplotypes as predictors of response to GA The relative frequencies of the most common DR–DQ haplotypes by responder status are shown in Figure 2. The DR15–DQ6 haplotype was the most common (present in 39% of all individuals) as well as the best individual haplotype predictor of treatment response. Table 4 summarizes the
Clinical response to glatiramer acetate in multiple sclerosis
Table 3
Summary of treatment response by HLA-specificities.
HLA-specificities DR1 DR3 DR4 DR7 DR8 DR9 DR11 DR12 DR13 DR14 DR15 DR17 DR18 DQ2 DQ3 DQ4 DQ5 DQ6 DQ7 DQ8 DQ9
345
NR 27 33 27 26 32 32 25 33 26 32 26 23 33 15 32 32 26 21 23 27 33
R 25 29 25 24 26 30 24 30 22 30 13 28 29 24 26 25 25 8 25 26 27
+NR
+R
Predictor
Sensitivity
Specificity
PPV
NPV
AUC
Nominal p
7 1 7 8 2 2 9 1 8 2 8 11 1 19 2 2 8 13 11 7 1
5 1 5 6 4 0 6 0 8 0 17 2 1 6 4 5 5 22 5 4 3
– + – – + – – – + – + – + – + + – + – – +
0.83 0.03 0.83 0.80 0.13 1.00 0.80 1.00 0.27 1.00 0.57 0.93 0.03 0.80 0.13 0.17 0.83 0.73 0.83 0.87 0.10
0.21 0.97 0.21 0.24 0.94 0.06 0.26 0.03 0.76 0.06 0.76 0.32 0.97 0.56 0.94 0.94 0.24 0.62 0.32 0.21 0.97
0.48 0.50 0.48 0.48 0.67 0.48 0.49 0.48 0.50 0.48 0.68 0.55 0.50 0.62 0.67 0.71 0.49 0.63 0.52 0.49 0.75
0.58 0.53 0.58 0.57 0.55 1.00 0.60 1.00 0.54 1.00 0.67 0.85 0.53 0.76 0.55 0.56 0.62 0.72 0.69 0.64 0.55
0.52 0.50 0.52 0.52 0.54 0.53 0.53 0.51 0.52 0.53 0.67 0.63 0.50 0.68 0.54 0.55 0.53 0.68 0.58 0.54 0.54
0.6876 0.9284 0.6876 0.7328 0.3052 0.1077 0.5407 0.2581 0.7725 0.1077 0.0062 0.0077 0.9284 0.0028 0.3052 0.1637 0.4939 0.0044 0.1436 0.4396 0.2372
NR =Non-responders negative for allele, R=Responders negative for allele, +NR=Non-responders positive for allele, +R=Responders positive for allele, Predictor indicates whether presence (+) or absence ( ) of allele is associated with positive response, PPV= positive predictive value, NPV= negative predictive value, AUC=area under the Receiver Operating Characteristics (ROC) curve. Nominal p indicates the p-value from logistic regression for association of the allele with response uncorrected for multiple testing. In the case of alleles with ’Predictor=-’, all statistics are shown for the reversed formulation (i.e., absences of the allele predicts response). The bold numbers are those with significant p-value.
Figure 2 Title: Relative frequency of HLA DR–DQ haplotypes by responder status. Legend: Non-responders (N=34) are in light-gray on the left in each pair; Responders (N=30) are dark-gray on the right in each pair. Vertical axes indicate the proportion of non-responders or responders with the allele present. Text below the haplotype names indicates the total number of patients with haplotype.
response status by individual HLA DR–DQ haplotypes. As a predictor of clinical response status, only DR15–DQ6 had both sensitivity and specificity above 0.5. Haplotypes DR15– DQ6 and DR17–DQ2 were both significant even after a Holm’s
correction for multiple testing; their adjusted p-values were 0.0437 and.0462, respectively. The presence of the DR15– DQ6 haplotype and the absence of the DR17–DQ2 haplotype were associated with treatment response.
346
S. Dhib-Jalbut et al.
Table 4 Haplotype DR1–DQ5 DR4–DQ8 DR7–DQ2 DR11–DQ7 DR13–DQ6 DR15–DQ6 DR17–DQ2
Summary of treatment response by haplotype. Only haplotypes with observed frequency above 0.05 are shown. NR 27 27 27 26 30 26 23
R 25 26 26 27 25 13 28
+NR
+R
Predictor
Sensitivity
Specificity
PPV
NPV
AUC
Nominal p
7 7 7 8 4 8 11
5 4 4 3 5 17 2
– – – – + + –
0.83 0.87 0.87 0.90 0.17 0.57 0.93
0.21 0.21 0.21 0.24 0.88 0.76 0.32
0.48 0.49 0.49 0.51 0.56 0.68 0.55
0.58 0.64 0.64 0.73 0.55 0.67 0.85
0.52 0.54 0.54 0.57 0.52 0.67 0.63
0.6876 0.4396 0.4396 0.1447 0.5738 0.0062 0.0077
NR= Non-responders negative for haplotype, R= Responders negative for haplotype, +NR=Non-responders positive for haplotype, +R= Responders positive for haplotype, Predictor indicates whether presence (+) or absence ( ) of haplotype is associated with positive response, PPV=positive predictive value, NPV=negative predictive value, AUC=area under the Receiver Operating Characteristics (ROC) curve. Nominal p indicates the p-value from logistic regression for association of the haplotype with response uncorrected for multiple testing. In the case of haplotypes with ’Predictor=-’, all statistics are shown for the reversed formulation (i.e., absences of the haplotype predicts response). The bold rows are those with significant p-values.
3.3.
More complex models
Age, sex, and initial disease severity were not significant as covariates and were not used in any of the analyses in this report. Neither ARR nor EDSS at baseline differed significantly (p40.05) by allele haplotype status (Supplemental Table 1), suggesting that the observed effects are not due to regression to the mean. Cross-validation of lasso-regularized logistic regression indicated that the best fit involved equal magnitude coefficients (but opposite signs) for the DR15–DQ6 and DR17– DQ2 haplotypes. This is equivalent to a score with 3 levels: 0 (DR15–DQ6 absent but DR17–DQ2 present), 1 (DR15–DQ6 and DR17–DQ2 either both present or both absent), and 2 (DR15– DQ6 present but DR17–DQ2 absent). This model has an AUC of 0.72 and is summarized in Table 5.
4.
Table 5
Simplified haplotype model.
Prognostic profile
Haplotypes
NR/R (% R)
‘‘poor predicted response’’ ‘‘neutral predicted response’’ ‘‘good predicted response’’
(DR15–DQ6) absent & (DR17– DQ2) present (DR15–DQ6) present & (DR17– DQ2) present or (DR15–DQ6) absent & (DR17–DQ2) absent (DR15–DQ6) present & (DR17– DQ2) absent
10/2 (16.7%) 17/11 (39.3%) 7/17 (70.8%)
Response rates can be compared to the overall rate of 46.9%. The AUC of the model is=0.72.
Discussion
This prospective study identified four HLA class-II alleles that can potentially predict the clinical response to GA. Tests of association of response with one allele at a time singled out four alleles with nominal po0.01: DR15, DR17, DQ2, and DQ6. The presence of the allele is associated with positive response for DR15 and DQ6, while the absence of the allele is associated with positive response for DR17 and DQ2. These associations are nominally significant in the sense that, with a Bonferroni correction for testing 21 alleles, the required critical value for p is a =0.05/21=0.0024 and none of the pvalues achieve that; DQ2 is close at p=0.0028. However, DR15 is significant if regarded as a test of replication of the results reported by Fusco et al. (2001) and Gross et al. (2011). Gross et al. suggested that the DRB1n1501 effect followed a recessive pattern since only patients homozygous for this allele had a longer event free time on GA. While our sample size is insufficient to confirm this observation, it is worth pointing out that all four patients homozygous for DR15, and all five patients homozygous for DQ6 were GAresponders. Similarly, four of the five patients homozygous for DQ2 were non-responders. Our findings differ from those of Fusco et al. and Gross et al. First, we show a complete analysis for the major HLA class-II alleles and haplotypes as shown in Tables 3 and 4,
while Fusco et al. presented results for DR15 only (when considering treatment success) and Gross et al. tested only DR15 for GA response. Second, We detected deleterious effects for DR17 and DQ2 beyond the beneficial effect for DR15 noted by Gross et al. and Fusco et al. Third, we included an analysis of the DR–DQ haplotypes. The four HLA alleles perform similarly (0.63–0.68) for the AUC, or area under the ROC curve, which can be considered a measure of the discriminating ability of the model. The AUC runs from 0.5, equivalent to a coin toss, to 1.0, corresponding to perfect prediction. Larger differences are found in the sensitivity, specificity, positive predictive value, and negative predictive value. Although it might seem that using all of the allelic information would result in better prediction, the AUC for a model including all alleles is only slightly better than that for the best two allele model, and worse if one corrects for overfitting (data not shown). The two haplotypes best at predicting treatment response, DR15–DQ6 and DR17–DQ2, give equivalent predictions to the alleles DR15 and DR17, respectively, since those alleles are found only in the two haplotypes. The other haplotypes containing DQ2 and DQ6 are not good predictors, suggesting that the performance of DQ2 and DQ6 as allelic predictors is due to their association with the DR15–DQ6 and DR17–DQ2 haplotypes.
Clinical response to glatiramer acetate in multiple sclerosis
347
Analysis of prognostic models using multiple haplotypes suggested a two-haplotype prognostic model with a simple interpretation (Table 5). Patients with two predictors of successful response, namely the presence of DR15–DQ6 and the absence of DR17–DQ2, can be considered to have a good predicted response (71% are responders). Patients with neither (that is, they are DR15–DQ6 negative but DR17-DQ2 positive) have a poor predicted response (17% are responders); patients with DR15–DQ6 and DR17–DQ2 either both present or both absent have a neutral prediction of response (39% are responders). Our findings are consistent with the mechanism of action of GA. This copolymer binds HLA class-II molecules, and therefore, influences signaling through the T-cell receptor and antigen presenting cells (Gran et al., 2000; Lalive et al., 2011) leading to the generation of Th2 cells and regulatory T-cells. Allelic differences in HLA class II molecules could result in conformational differences in the GA/class-II/TcR complex that influence immune responsiveness to GA and subsequently its clinical efficacy. Our findings need to be replicated in other data sets before clinical recommendations can be made. If validated, it is possible based on the model presented in Table 5 that a recommendation can be made to patients considered candidates for treatment with GA to be HLA-typed for DR and DQ alleles before initiation of therapy. A predicted response to GA for a particular patient can then be classified as good, neutral or poor. In a disease with variable responses to immunomodulatory agents, our findings represent a step forward towards personalized therapy in MS
Cohen BA, Khan O, Jeffery DR, Bashir K, Rizvi SA, Fox EJ, et al. Identifying and treating patients with suboptimal responses. Neurology 2004;63:S33–40. Comi G, Filippi M, Wolinsky JS. European/Canadian multicenter, double-blind, randomized, placebo-controlled study of the effects of glatiramer acetate on magnetic resonance imagingmeasured disease activity and burden in patients with relapsing multiple sclerosis. European/Canadian glatiramer acetate study group. Annals of Neurology 2001;49:290–7. Dhib-Jalbut S, Chen M, Said A, Zhan M, Johnson KP, Martin R. Glatiramer acetate-reactive peripheral blood mononuclear cells respond to multiple myelin antigens with a Th2-biased phenotype. Journal of Neuroimmunology 2003;140:163–71. Duda PW, Krieger JI, Schmied MC, Balentine C, Hafler DA. Human and murine CD4 T cell reactivity to a complex antigen: recognition of the synthetic random polypeptide glatiramer acetate. Journal of Immunology 2000;165:7300–7. Farina C, Wagenpfeil S, Hohlfeld R. Immunological assay for assessing the efficacy of glatiramer acetate (Copaxone) in multiple sclerosis. A pilot study. Journal of Neurology 2002;249:1587–92. Fridkis-Hareli M, Aharoni R, Teitelbaum D, Arnon R, Sela M, Strominger JL. Binding of random copolymers of three amino acids to class II MHC molecules. International Immunology 1999;11:635–41. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. Journal of Statistical Software 2010;33:1–22. Fusco C, Andreone V, Coppola G, Luongo V, Guerini F, Pace E, et al. HLA-DRB1n1501 and response to copolymer-1 therapy in relapsingremitting multiple sclerosis. Neurology 2001;57:1976–9. Gourley IS, Kearns J, McKeen M, Birkos S, Brown M, Kamoun M. HLA Class I typing of volunteers for a bone marrow registry: QC analysis by DNA-based methodology identifies serological typing discrepancies in the assignment of HLA-A and B antigens. Tissue Antigens 2002;59:211–5. Gran B, Tranquill LR, Chen M, Bielekova B, Zhou W, Dhib-Jalbut S, et al. Mechanisms of immunomodulation by glatiramer acetate. Neurology 2000;55:1704–14. Gross R, Healy BC, Cepok S, Chitnis T, Khoury SJ, Hemmer B, et al. Population structure and HLA DRB1 1501 in the response of subjects with multiple sclerosis to first-line treatments. Journal of Neuroimmunology 2011;233:168–74. Hussien Y, Sanna A, Soderstrom M, Link H, Huang YM. Glatiramer acetate and IFN-beta act on dendritic cells in multiple sclerosis. Journal of Neuroimmunology 2001;121:102–10. Kim HJ, Ifergan I, Antel JP, Seguin R, Duddy M, Lapierre Y, et al. Type 2 monocyte and microglia differentiation mediated by glatiramer acetate therapy in patients with multiple sclerosis. Journal of Immunology 2004;172:7144–53. Klitz W, Maiers M, Spellman S, Baxter-Lowe LA, Schmeckpeper B, Williams TM, et al. New HLA haplotype frequency reference standards: high-resolution and large sample typing of HLA DR-DQ haplotypes in a sample of European Americans. Tissue Antigens 2003;62:296–307. Lalive PH, Neuhaus O, Benkhoucha M, Burger D, Hohlfeld R, Zamvil SS, et al. Glatiramer acetate in the treatment of multiple sclerosis: emerging concepts regarding its mechanism of action. CNS Drugs 2011;25:401–14. Martin R, Jaraquemada D, Flerlage M, Richert J, Whitaker J, Long EO, et al. Fine specificity and HLA restriction of myelin basic proteinspecific cytotoxic T cell lines from multiple sclerosis patients and healthy individuals. Journal of Immunology 1990;145:540–8. Martinelli Boneschi F, Rovaris M, Johnson KP, Miller A, Wolinsky JS, Ladkani D, et al. Effects of glatiramer acetate on relapse rate and accumulated disability in multiple sclerosis: meta-analysis of three double-blind, randomized, placebo-controlled clinical trials. Multiple Sclerosis 2003;9:349–55. McDonald WI, Compston A, Edan G, Goodkin D, Hartung HP, Lublin FD, et al. Recommended diagnostic criteria for multiple
Conflict of interest Dr. S. Dhib-Jalbut received grant support and consultant fees from TEVA Pharmaceuticals, Bayer Health, Biogen-Idec and EMD-Serono. Dr. R.M. Valenzuela has nothing to disclose. Dr. K. Ito has nothing to disclose. Dr. M. Kaufman received grant support and served on the speakers’ bureau for TEVA Neuroscience. Dr. M. Picone has nothing to disclose. Dr. S. Buyske has nothing to disclose.
Acknowledgment Supported by Grants from the Wadsworth Foundation and TEVA Pharmaceuticals Inc.
Appendix A.
Supporting information
Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.msard. 2013.02.005.
References Aharoni R, Kayhan B, Eilam R, Sela M, Arnon R. Glatiramer acetatespecific T cells in the brain express T helper 2/3 cytokines and brain-derived neurotrophic factor in situ. Proceedings of the National Academy of Sciences USA 2003;100:14157–62.
348 sclerosis: guidelines from the International Panel on the diagnosis of multiple sclerosis. Annals of Neurology 2001;50:121–7. Pette M, Fujita K, Wilkinson D, Altmann DM, Trowsdale J, Giegerich G, et al. Myelin autoreactivity in multiple sclerosis: recognition of myelin basic protein in the context of HLA-DR2 products by T lymphocytes of multiple-sclerosis patients and healthy donors. Proceedings of the National Academy of Sciences USA 1990;87:7968–72. R. R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; 2011. Sawcer S, Hellenthal G, Pirinen M, Spencer CC, Patsopoulos NA, Moutsianas L, et al. Genetic risk and a primary role for cellmediated immune mechanisms in multiple sclerosis. Nature 2011;476:214–9.
S. Dhib-Jalbut et al. Schmidt H, Williamson D, Ashley-Koch A. HLA-DR15 haplotype and multiple sclerosis: a HuGE review. American Journal of Epidemiology 2007;165:1097–109. Smith KJ, Pyrdol J, Gauthier L, Wiley DC, Wucherpfennig KW. Crystal structure of HLA-DR2 (DRAn0101, DRB1n1501) complexed with a peptide from human myelin basic protein. Journal of Experimental Medicine 1998;188:1511–20. Vieira PL, Heystek HC, Wormmeester J, Wierenga EA, Kapsenberg ML. Glatiramer acetate (copolymer-1, copaxone) promotes Th2 cell development and increased IL-10 production through modulation of dendritic cells. Journal of Immunology 2003;170:4483–8. Weber MS, Prod’homme T, Youssef S, Dunn SE, Rundle CD, Lee L, et al. Type II monocytes modulate T cell-mediated central nervous system autoimmune disease. Nature Medicine 2007;13:935–43.