Mapping and rewiring the MYBPC3 promoter for rescue of haploinsufficiency driven hypertrophic cardiomyopathy

Figures

• 

  Figure 1:. Saturation mutagenesis of MYBPC3’s promoter in an MPRA. MPRA was performed using a variant library containing each possible SNV in the MYBPC3 456bp core promoter sequence. (A) MPRA construct design consisted of pools of SNV, triple transversions, or wild type (WT) sequence oligos placed next to remaining MYBPC3 promoter sequence as a constant sequence. A 20bp constant spacer sequence followed, then a random 16bp barcode in the 5’ UTR of GFP. Our lentiMPRA workflow was adapted from Gordon et al.14 Each variant oligo was randomly paired with a unique 16bp barcode in a modified pLS-SceI backbone. High throughput sequencing of the plasmid pool was used to build a dictionary of variant-barcode pairs. Induced pluripotent stem cell-derived cardiomyocytes were transduced with lentivirus packaged from this plasmid pool. DNA and RNA were extracted and then high throughput sequencing targeting the barcode was used to obtain barcode counts which were interpreted using the variant-barcode dictionary. Normalizing the barcode counts in the RNA (expression) to the counts in the DNA (viral insertion events) provides a measure of variant effect on transcription. See Methods for further details.…

• 

  Figure 2:. Transcription factor binding site motif analysis demonstrates that key cardiac transcription factors are vulnerable to SNVs. (A) MPRA data for SNVs within each annotated TFBS. Variants within GATA4, and MEF2D decrease expression compared to non-TFBS SNVs, consistent with their roles as cardiac-active transcription factors. Variants within the TATA box, and initiator sequence (INR) similarly reduce expression compared to the non-TFBS SNVs. Variants within RXR increase expression compared to non-TFBS SNVs, suggesting it acts as a repressor in this context. (Dunnett test, *p<0.05, **p<0.01, ***p<0.001) (B) Vulnerability analysis for each annotated TFBS. Variants within GATA4, MEF2D, and the TATA box are enriched for significance (Fisher’s exact test, p<0.001). (C) Predicted SNV impact on individual TFBS binding affinity (Delta Score) was calculated using motifmatchr (see Methods) and compared to impact on transcription in the MPRA (Pearson correlation). Positive correlation indicates that TF is acting as an activator while negative correlation suggests it is acting as a repressor. See Supplemental Figure S3B for additional TFBSs. (D) Logo-style motifs for TFBSs were generated…

• 

  Figure 3:. Linear modeling analysis of paired SNVs to determine additive or epistatic Effects. (A) 25 ChromBPNet models were trained on different DNase and ATAC-seq datasets from human left ventricle samples. These models were used to perform in silico mutagenesis on the MYBPC3 promoter. (LFCWT for simulated ATAC-seq read counts) (B) These SNV effect predictions were correlated to the LFCWT observed in the MPRA experiment. (Pearson correlation, p<0.0001) (C) The SNVs with the 20 largest positive and negative MPRA effects were selected for analysis of the effects of pairs of SNVs in an MPRA. A weighted linear model was created for each possible pair among these 40 variants using the single SNV effects to predict the paired effect assuming additive effects. These predictions were highly correlative with actual MPRA activity (R=0.79, p<0.0001, Pearson correlation). (D) The interaction term of the model represents any epistatic effect. All three pair classes display a sub-additive behavior. (E) The ChromBPNet models predicted LFCWT for paired SNVs were compared to the summed predicted effect of the two individual SNVs. (F) The interaction term for the ChromBPNet predictions for paired SNVs were…

• 

  Figure 4:. MPRA with synthetic CREs investigates regulatory grammar. (A) Synthetic CREs consisting of 1-3 TFBS motifs were inserted within a null background sequence upstream of a “minimal” MYBPC3 promoter and measured using MPRA. Log2 fold-change compared to the null background sequence was calculated for each synthetic CRE (LFCBackground). (B) Clusters of 1-3 homotypic TFBSs spaced 5bp apart were tested to determine the effect of copy number on expression for GATA4, MEF2D, NKX2-5, and SRF. (C) These homotypic TFBS clusters as well as multiple heterotypic TFBS combinations were tested for level of transcriptional activation in the minimal MYBPC3 promoter context. Combinations with significant transcriptional upregulation are shown. See Supplemental Figure S6C for all clusters tested. (D) ChromBPNet models predicted the LFCBackground for the same clusters of 1-3 homotypic TFBSs spaced 5bp apart. (E) ChromBPNet models predicted the LFCBackground for the same homo- and heterotypic TFBS clusters. The 10 most positive are shown here. See Supplemental Figure S6D for all clusters tested. (F) Pearson correlation of MPRA data and ChromBPNet predictions for the TFBS clusters. (R=0.48, p<0.0001) (G)…