Progressive multifocal leukoencephalopathy genetic risk variants for pharmacovigilance of immunosuppressant therapies

Abstract

Background: Progressive multifocal leukoencephalopathy (PML) is a rare and often lethal brain disorder caused by the common, typically benign polyomavirus 2, also known as JC virus (JCV). In a small percentage of immunosuppressed individuals, JCV is reactivated and infects the brain, causing devastating neurological defects. A wide range of immunosuppressed groups can develop PML, such as patients with: HIV/AIDS, hematological malignancies (e.g., leukemias, lymphomas, and multiple myeloma), autoimmune disorders (e.g., psoriasis, rheumatoid arthritis, and systemic lupus erythematosus), and organ transplants. In some patients, iatrogenic (i.e., drug-induced) PML occurs as a serious adverse event from exposure to immunosuppressant therapies used to treat their disease (e.g., hematological malignancies and multiple sclerosis). While JCV infection and immunosuppression are necessary, they are not sufficient to cause PML.

Methods: We hypothesized that patients may also have a genetic susceptibility from the presence of rare deleterious genetic variants in immune-relevant genes (e.g., those that cause inborn errors of immunity). In our prior genetic study of 184 PML cases, we discovered 19 candidate PML risk variants. In the current study of another 152 cases, we validated 4 of 19 variants in both population controls (gnomAD 3.1) and matched controls (JCV+ multiple sclerosis patients on a PML-linked drug ≥ 2 years).

Results: The four variants, found in immune system genes with strong biological links, are: C8B, 1-57409459-C-A, rs139498867; LY9 (alias SLAMF3), 1-160769595-AG-A, rs763811636; FCN2, 9-137779251-G-A, rs76267164; STXBP2, 19-7712287-G-C, rs35490401. Carriers of any one of these variants are shown to be at high risk of PML when drug-exposed PML cases are compared to drug-exposed matched controls: P value = 3.50E-06, OR = 8.7 [3.7-20.6]. Measures of clinical validity and utility compare favorably to other genetic risk tests, such as BRCA1 and BRCA2 screening for breast cancer risk and HLA-B^*15:02 pharmacogenetic screening for pharmacovigilance of carbamazepine to prevent Stevens-Johnson Syndrome and Toxic Epidermal Necrolysis.

Conclusion: For the first time, a PML genetic risk test can be implemented for screening patients taking or considering treatment with a PML-linked drug in order to decrease the incidence of PML and enable safer use of highly effective therapies used to treat their underlying disease.

Keywords: JC virus; PML; immunodeficiency; multiple sclerosis; natalizumab; pharmacovigilance; progressive multifocal leukoencephalopathy; serious adverse event.

Figure 1

Case and control recruitment and study design. (A) Prior study for genetic discovery and validation using Whole Exome Sequencing (WES), 669 candidate immune response genes, and gnomAD 2.1 (WES + WGS) population controls (45). (B) New recruitment of PML cases and matched controls (JCV+ MS patients exposed to a PML-linked drug ≥2 years). All PML cases and matched controls were genotyped for the prior study's 19 candidate PML risk variants. Matched controls without JCV serostatus were assayed (see Methods). Excluded cases: four were found to be duplicates of the prior study (see Methods). Excluded controls: 152 JCV seronegative (JCV-) patients; one QC failure for genotyping assays due to low quality DNA. (C) Drug-exposed analysis is the pooled subgroup (n = 110) of total PML cases (n = 336) compared to matched controls. Drug-exposed study results are reported in Tables 2, 5 and genes for the top 4 variants are listed.

Figure 2

Predictive risk comparison to BRCA screening tests, odds ratio (OR) vs. diagnostic (Dx) yield. Results for a 4-variant PML risk test are shown in comparison to the BRCA1/BRCA2 breast cancer risk prediction test. The proposed 4-variant PML risk test data point (●) is based on the total drug-exposed PML cases (Table 5). The BRCA1 (Δ) and BRCA2 (□) risk test data points are based on results for over 95,000 women reported in Kurian et al. (53). More Utility is defined as higher OR and higher Dx yield and Less Utility is defined as lower OR and lower Dx yield.

Figure 3

Predictive risk comparison to HLA-B*1502, positive predictive value (PPV) vs. number needed to treat (NNT). Results for a 4-variant PML risk test are shown in comparison to the HLA-B*15:02 test that is required in Asian populations before administering carbamazepine (CBZ). CBZ is a cause of the serious adverse event Stevens-Johnson Syndrome (SJS) and Toxic Epidermal Necrolysis (TEN). The proposed 4-variant PML risk test data point (●) is based on the total drug-exposed PML cases (Table 5) and a PML incidence rate of 3% (17). The HLA-B*15:02 SJS/TEN risk test data point (Δ) is based on results reported in Shi et al. (54). More Utility is defined as higher PPV and lower NNT and Less Utility is defined as lower PPV and higher NNT.