Genetic Variants Associated With Intraparenchymal Hemorrhage Progression After Traumatic Brain Injury

Key Points Question Is genetic variation in a key channel of secondary injury after traumatic brain injury (TBI) associated with patients’ risk of intraparenchymal hemorrhage (IPH) progression, a major contributor to unfavorable outcome? Findings In this genetic association study of a prospective cohort of 321 patients with severe TBI, 8 spatially clustered ABCC8 (sulfonylurea receptor 1) and TRPM4 sequence variants, all brain-specific expression quantitative trait loci, were associated with IPH progression and improved clinical models; regulatory annotations further suggest biological plausibility. ABCC8 variants that increase brain tissue ABCC8 messenger RNA expression were associated with increased IPH progression risk, whereas variant TRPM4 was protective. Meaning The findings suggest that identifying patients with risk-altering genotypes of a pivotal channel in IPH progression may guide risk stratification, prognostication, patient selection, and upcoming trial design for targeted sulfonylurea receptor 1 inhibition.

IPH specifically. Reports were searched for the terms including "expanded", "blossomed", "increased" or "new" in reference to the hemorrhagic component of the IPH only while terms such as "evolved" or "matured" were felt to be more representative of the natural course of disease rather than progression. To minimize likelihood of false positive results, reports where expansion was not mentioned were interpreted as stability.
• Comparison of the two independent metrics of hemorrhage progression yielded a percent agreement of 88.9% and a Cohen's kappa of 0.78, suggesting substantial agreement between the two metrics.

SNP Functional Potential Determination
SNPs were evaluated for impact on gene expression using the Genotype Tissue Expression (GTEx) Project data portal (www.gtexportal.org, 06/23/2020) [7]. They were interrogated for brain-specific gene expression quantitative trait loci (eQTL) in the hippocampus, non-specified cortex, frontal cortex, putamen, and cerebellum to assess breadth and consistency of impact on gene expression across cortical vs. deep brain structures. Regulatory potential was evaluated using RegulomeDB v2.0 and HaploReg V4.1 [8][9][10]. SNPs loci were explored for the ~200 bp regional chromatin state, transcription start sites, promoter histone marks, enhancer histone marks, DNAse, and protein binding (via Chromatin-ImmunoPrecipitation, ChIP, reports) in brain vs. all reported tissues. Individual SNPs were interrogated for impact on altering regulatory motifs for transcription factors using sequence logos (RegulomeDB) as well as position-weighted matrix (PWM) scores (HaploReg). PWM scores on Haploreg (determined using experimental data on JASAPR, TRANSFAC, and protein binding microarray experiments) are available as log-odds, and account for motif lengths and base-pair compositions; they reflect transcription factor binding affinity [9]. Log-odds score differences between variant allele vs reference alleles evaluate change in binding affinity [9] (positive values reflect an increase in log-odds score for variant alleles, and suggests increased transcription factor binding strength). SNPs were evaluated for reported clinical significance via systematic PubMed, Embase, and ClinVar searches.

Spatial relationship modeling between ABCC8 and TRPM4 loci and channel structure
Chromosomal locations were identified using the University of California, Santa Cruz genome browser, humangenome assembly(hg-38). Linkage disequilibrium (LD), distance from the proximal exon, peptide sequences encoded by specific exons and residue overlap splice sites were identified via Ensembl-100 [11]. Established SUR1 (5WUA) and TRPM4 (6BQV) 3-dimensional electron microscopy structures were obtained from the Research Collaboratory for Structural Bioinformatics Protein Data Bank [12][13][14][15]. University of California, San Francisco Chimera was used to generate the octameric SUR1-TRPM4 channel [16].

Statistical Analysis: Sample Size Impact Simulation
We also simulated an example to demonstrate the potential impact of a priori knowledge of patients' ABCC8 and/or TRPM4 genotypes on patient selection for clinical-trial design. We stratified our sample by genotypes, and, based on the genotype-subgroup, looked at both the sample size required and the number of patients who would need to be genotyped in order to have a 90% power to find a 30% relative risk reduction (RRR) in the risk of intra-axial hemorrhage progression within 120 hours. For this, we first calculated the probability of belonging to a specific genotype subgroup and the probability of hemorrhage progression within that subgroup for each of the following categories: 1) all comers (current trial designs), 2) at least one ABCC8 SNP with homozygous variant (risk) genotype, 3) at least one TRPM4 SNP with homozygous wildtype (risk) genotype, 4) at least one ABCC8 SNP with homozygous variant (risk) phenotype AND at least one TRPM4 SNP with homozygous wildtype (risk) genotype, 5) and at least one ABCC8 SNP with homozygous variant (risk) phenotype OR at least one TRPM4 SNP with homozygous wildtype (risk) genotype, 6) patients predicted to have a >50% probability of hemorrhage progression by the full clinical model without genotypes used for creating the ROC curves, 7) and patients predicted to have a >50% probability of hemorrhage progression by the full clinical plus genotypes model used for creating the ROC curves. The probability of hemorrhage progression for each subgroup was then used in sample size calculations to design different iterations of the hypothetical study where a treatment is expected to result in a RRR of hemorrhage progression by 30% with a power of 0.9. The obtained sample size was divided by the proportion of patients within that subgroup to determine how many patients would need to be genotyped to reach the necessary enrollment sample size. This procedure was followed for all 5 subgroups to identify the degree to which genotype-based patient selection may impact sample size based on varying risk of hemorrhage

5-day pvalue ABCC8
Additive (Reference GG) Table 1). c SNPs are significant (p<0.05) in all-comers regardless of craniectomy status (Supplemental Table 9) d SNPs previously reported to be predictive of intracranial pressure and/or acute CT edema after TBI.
Panel demonstrating violin plots of normalized mRNA expression levels (y-axis) associated with the genotypes (x-axis) of four TRPM4 SNPs significantly associated with hemorrhage progression after severe traumatic brain injury (TBI). Each subpanel shows the normalized TRPM4 mRNA expression level in the cerebellum, cortex, frontal cortex, hippocampus and putamen by SNP genotype: rs3760666 (A), rs1477363 (B), rs10410857 (C) and rs909010 (D). Shaded regions in teal or blue (alternating between brain region) of the individual violin plots indicate the density distribution of mRNA expression in the samples in each respective genotype, with the white line showing the median value. The P value provided for each SNP at each location indicates the P value for different expression levels across genotypes for that SNP in the respective tissue location. Unlike ABCC8, all four TRPM4 SNPs are brain-specific eQTLs only in the cerebellum, with rs3760666 also having significantly different TRPM4 mRNA expression levels in the brain cortex and putamen. In all cases, mRNA expression is lower with variant TRPM4 SNPs, with a dose-dependent effect noted between homozygous wild-type, heterozygotes, and homozygous variants. Schematics of sequence logos of different transcription factor binding site motifs obtained from RegulomeDB annotations for ABCC8 SNP rs2237982 (A) and TRPM4 SNP rs10410857. The y-axis in each subgraph represents a binary information tool (bits) containing two pieces of information: the height of the base pair alphabet is non linearly proportional to the frequency with which it is found at that position, and the total height of each column denotes the importance/strength of that location for transcription factor recognition and binding. Each base pair is shown in a specific color. The location altered by the SNP is highlighted in a red box. For transcription factors MEF2d, MYEF2, and MYEF2-b, ABCC8 SNP rs2237982 changes a highly conserved base-pair, and this site is highly important for determining strength of transcription factor recognition and binding. TRPM4 SNP rs10410857, only moderately affects transcription factor binding for ARHGEF12 but strongly affects BCL6. enhancers (yellow), heterochromatin (lavendar), repressed polycomb (dark gray) and quiescent/low transcription (light gray) is shown for each SNP in brain vs all-tissue. For each SNP, the ratio (R) of strong transcription and enhancers to repressed polycomb, quiescent states, and heterochromatin is shown. For example, for all ABCC8 SNPs, R is markedly higher in brain tissue vs other tissue, with most regional samples demonstrating strong transcriptional activity. This difference was not as pronounced with TRPM4 which was annotated to have strong transcription (green) in all tissues.