Massively parallel variant characterization identifiesNUDT15alleles associated with thiopurine toxicity

Abstract

As a prototype of genomics-guided precision medicine, individualized thiopurine dosing based on pharmacogenetics is a highly effective way to mitigate hematopoietic toxicity of this class of drugs. Recently, NUDT15 deficiency was identified as a genetic cause of thiopurine toxicity, and NUDT15-informed preemptive dose reduction was quickly adopted in clinical settings. To exhaustively identify pharmacogenetic variants in this gene, we developed massively parallel NUDT15 function assays to determine the variants’ effect on protein abundance and thiopurine cytotoxicity. Of the 3,097 possible missense variants, we characterized the abundance of 2,922 variants and found 54 hotspot residues at which variants resulted in complete loss of protein stability. Analyzing 2,935 variants in the thiopurine cytotoxicity-based assay, we identified 17 additional residues where variants altered NUDT15 activity without affecting protein stability. We identified structural elements key to NUDT15 stability and/or catalytical activity with single amino acid resolution. Functional effects for NUDT15 variants accurately predicted toxicity risk alleles in patients treated with thiopurines with far superior sensitivity and specificity compared to bioinformatic prediction algorithms. In conclusion, our massively parallel variant function assays identified 1,152 deleterious NUDT15 variants, providing a comprehensive reference of variant function and vastly improving the ability to implement pharmacogenetics-guided thiopurine treatment individualization.

Massively parallel functional testing ofMSH2missense variants conferring Lynch Syndrome risk

“…Although the bioinformatic predictions were modestly correlated with our experimental measurements for variants in the functional validation set, they were markedly less concordant for other variants at those same residues, or throughout MSH2 at large ( Figure 4B). Similarly weak overall agreement has also been observed when benchmarking bioinformatic classifiers with deep mutational scans of other genes [61][62][63] . As variant effect predictors are often trained on the limited number of known variants with available classifications, their divergence with our experimental measurements may reflect overfitting, further suggesting that the comparison to a small set of functionally characterized alleles overestimates bioinformatic predictors' performance.…”

“…In terms of overall constraint, the dataset closest to MSH2 is from the transcription factor PPARG, in which 21.5% of missense variants scored as having substantial probability of being casual for lipodystrophy 76 . Most other human genes subjected to full-length mutational scans thus far have shown substantially higher overall constraint: for instance, 40.5% of missense variants in the pharmacogene NUDT15 scored as damaging 62 , as were 39% of variants in the homocysteine metabolic factor CBS 77 . A recent activity-agnostic protein stability screen showed the fraction of missense variants with substantially destabilizing effects to range from 25.1-43.1% across three different genes 62,78 ; these could be taken as a lower bound given that an unknown fraction of true LOF variants may remain stable.…”

“…Most other human genes subjected to full-length mutational scans thus far have shown substantially higher overall constraint: for instance, 40.5% of missense variants in the pharmacogene NUDT15 scored as damaging 62 , as were 39% of variants in the homocysteine metabolic factor CBS 77 . A recent activity-agnostic protein stability screen showed the fraction of missense variants with substantially destabilizing effects to range from 25.1-43.1% across three different genes 62,78 ; these could be taken as a lower bound given that an unknown fraction of true LOF variants may remain stable. For any such alleles that do behave as hypomorphs in functional assays, it remains unclear whether these meaningfully contribute to Lynch Syndrome risk 51,80 .…”

“…We instead used a close proxy for MMR dysfunction, treatment with the purine analog 6-TG, which selectively enrichies for MMR-defective cells which lack the ability to remove the resulting lesions. As with other deep mutational scans, these results may be further refined with orthogonal functional selections 62,89 or complemented by specialized assays for protein stability 78 , localization , or partner (MSH6) binding.…”

Editorial: is pharmacogenetic testing for adverse effects to IBD treatments ready for roll‐out?

“…Pharmacogenetics is the often‐hyped prototype of precision medicine. Advances in genetic sequencing platforms, bioinformatics and the collection of well‐phenotyped and adequately powered cohorts have led to a rapid expansion of novel pharmacogenetic associations 1 . Outside the field of oncology, however, few genetic biomarkers have translated into clinical care.…”