Integrative analysis of KCNQ1 variants reveals molecular mechanisms of type 1 long QT syndrome pathogenesis

Figures

• 

  Fig. 1.. Structural mapping of the analyzed KCNQ1 variants. The sites of the 61 variants analyzed in this work are mapped onto the structure of KCNQ1. The structure of the main channel was determined by cryo-EM [Protein Databank (PDB) ID: 6uzz, (9)], while the tetrameric helical bundle HD domain structure was determined by X-ray crystallography [PDB ID: 3bj4, (10)]. The unstructured N and C termini, as well as the loops between the HA and HB helices and between the HC and HD helices, are manually drawn. The backbone Cα atoms of analyzed variant sites are denoted as spheres. The C-alpha positions of analyzed variants are colored according to our experimental classifications based on the results of this paper: Class I (red), Class II (orange), Class III (yellow), Class IV (blue), Class V (magenta), Class VI (gray). Class definitions are discussed in the results and discussion sections. Right: Close-up views of individual KCNQ1 domain structures. The sidechains of analyzed residue positions are colored according to classification as stated above.

• 

  Fig. 2.. CETSA reports on thermal stability of KCNQ1 variants. (A) Representative Western blots of CETSA experiments for KCNQ1 WT and three VSD variants. HEK293 cells were heat shocked for 3 min in 4 °C increments, from 37 °C to 76 °C followed by cell lysis and centrifugation. The soluble fraction of the lysate was then used for Western blot analysis. Western blots were probed using an anti-KCNQ1 antibody (Alomone Labs, catalog #APC-022). The black box indicates the quantified band corresponding to KCNQ1 WT at 37 °C. The incubation temperature of the samples increases from Left to Right. For full-size western blot images, SI Appendix, Fig. S2. (B) Representative thermal unfolding curves (quantification of panel 1A). The signal intensity of each KCNQ1 band was normalized to the signal of the respective sample at 37 °C. Curves were fit to a four-parameter symmetric sigmoidal curve to determine Tagg using GraphPad Prism v10.2.2. Variant V110I is colored pink, H126L is colored purple, and E115G is colored teal. WT KCNQ1 is colored black. (C) Average Tagg values ± SD. P-values are denoted as bars with the P-value written above. N = 41 for KCNQ1 WT and N = 3 for each variant.

• 

  Fig. 3.. Thermal destabilization is a common effect of KCNQ1 pathogenic variants. (A) Tagg values for the 61 variants studied in this work. Mean Tagg values from 3 independent (4 replicates for L266R, 41 for KCNQ1 WT) are ordered from highest to lowest. Data are color-coded: Known pathogenic variant (green), VUS (dark purple), conflicting interpretations (cyan), likely pathogenic (light purple), and engineered variant (light gray). Statistically significant Tagg values relative to KCNQ1 WT are denoted with asterisks. *P-value < 0.05, **P-value < 0.01, ***P-value < 0.001, ****P-value < 0.0001. Statistical significance was determined using an ordinary one-way ANOVA followed by Dunnett’s Multiple Comparisons Test. Z-score values, which demonstrate the effect sizes achievable using this method, are indicated with dashed horizontal lines and noted on the right Y-axis. Z-scores from −2 to 2 are shaded light gray. Error bars denote mean and SD. (B) Tagg values of destabilized variants from panel A (statistically significant Tagg difference). Variants are colored according to location: VSD (red), PD (orange), HA/HB (light blue), HD (purple). Error bars denote mean and SD.

• 

  Fig. 4.. Thermal destabilization results in reduced cell surface expression for KCNQ1 variants. (A) Surface expression values for the 61 KCNQ1 variants. Data are color-coded based on initial variant classification: known pathogenic mutation (green), VUS (dark purple), conflicting interpretations (cyan), likely pathogenic (light purple), and engineered mutations (light gray). HEK293 cells were transfected with WT or variant KCNQ1 for 48 h followed by flow cytometry analysis according to (20). Surface expression, determined by the protein signal at the plasma membrane, is given as % of WT. Data are ordered from highest to lowest, Left to Right. Dashed lines are drawn at 100% WT and 65% WT [our cutoff for pathological LOF (20)]. All variants <65% WT are also statistically significantly different from WT (for P-values, SI Appendix, Table S1) via an ordinary one-way ANOVA followed by Dunnett’s Multiple Comparisons Test. Error bars denote mean and SD. (B) Comparison of Tagg and surface expression values. Variants are colored as in A. Data are ordered from highest to lowest Tagg. For Tagg (Right), vertical dashed lines are drawn at the mean WT Tagg as well as at the Tagg below which a variant is…

• 

  Fig. 5.. Pathogenic HD variants exhibit proteasome-independent impaired trafficking. (A) Cell surface expression vs. trafficking efficiency plotted for the 61 KCNQ1 variants. Surface expression was calculated as in Fig. 4A. Trafficking efficiency was calculated as the fraction of the total protein signal that accounts for surface protein: signalsurface/(signalsurface + signalinternal). Data are colored according to domain location: VSD (red), PD (orange), HA/HB (light blue), HC (dark blue), HD (purple). Values are given as %WT. Mean and SD are plotted. A linear regression (calculated in GraphPad) is designated with a solid line, and the 95% CI of the fit is given as dashed lines. The coefficient of determination for the regression is written as R2. (B) Surface expression vs. total expression. Total expression was calculated as the sum of surface plus internal signal. Values are given as % WT. Data are colored as in panel A. Error bars denote mean and SD. The linear regression, 95% CI, and coefficient of determination are also given as in panel A. (C) The effect of bortezomib treatment on KCNQ1 total expression. HEK293 cells were treated with 1 µM bortezomib or 0.1% DMSO for 20 h followed…

• 

  Fig. 6.. Loss of KCNQ1 channel function is often independent of mistrafficking. Comparison of mean current density and cell surface expression. Data are color-coded based on initial variant classification: known pathogenic mutation (green), VUS (dark purple), conflicting interpretations (cyan), likely pathogenic (light purple), and engineered mutations (light gray). For mean current density (Right panel), dashed vertical lines indicate mean current density values of 100%, 65% (LOF), and 25% (severe LOF) WT. Error bars denote mean and 95% CI. All variants with current density values below the 65% cutoff value are significantly different from WT (via ordinary one-way ANOVA and Dunnett’s multiple comparison’s test). For P-values, SI Appendix, Table S1. For surface expression (Left panel), dashed vertical lines indicate 100% and 65% WT. Error bars denote mean and SD. For P-values (calculated from ordinary one-way ANOVA with Dunnett’s multiple comparison’s test), SI Appendix, Table S1.

• 

  Fig. 7.. Performance of pathogenicity prediction tools varies with KCNQ1 sequence conservation. (A) Overall agreement of the eight tested models. Models were scored against the experimental data for 143 KCNQ1 variants: 61 variants in this work, 50 variants from our previous work (20), and 32 known benign variants (34). A model was considered in agreement if it predicted an experimentally dysfunctional variant as pathogenic and a benign variant as benign. Models are ordered from highest to lowest agreement. (B) Model agreement for individual classification types. The average agreement of the seven tools selected for further analysis (MetaLR was excluded) is plotted along with the SD. (C) Mean agreement vs chemical similarity for the dysfunctional experimental classes. Chemical similarity was determined by aligning the KCNQ1 sequence to the rest of the KCNQ family (KCNQ2-5) using UniPRot Align, with subsequent manual inspection. Mean % agreement is plotted as the average with the SD. The linear regression is shown with a solid line. The coefficient of determination is written as R2.