A calibrated functional patch-clamp assay to enhance clinical variant interpretation in KCNH2-related long QT syndrome

Summary

Modern sequencing technologies have revolutionized our detection of gene variants. However, in most genes, including KCNH2, the majority of missense variants are currently classified as variants of uncertain significance (VUSs). The aim of this study was to investigate the utility of an automated patch-clamp assay for aiding clinical variant classification in KCNH2. The assay was designed according to recommendations proposed by the Clinical Genome Sequence Variant Interpretation Working Group. Thirty-one variants (17 pathogenic/likely pathogenic, 14 benign/likely benign) were classified internally as variant controls. They were heterozygously expressed in Flp-In HEK293 cells for assessing the effects of variants on current density and channel gating in order to determine the sensitivity and specificity of the assay. All 17 pathogenic variant controls had reduced current density, and 13 of 14 benign variant controls had normal current density, which enabled determination of normal and abnormal ranges for applying evidence of moderate or supporting strength for VUS reclassification. Inclusion of functional assay evidence enabled us to reclassify 6 out of 44 KCNH2 VUSs as likely pathogenic. The high-throughput patch-clamp assay can provide moderate-strength evidence for clinical interpretation of clinical KCNH2 variants and demonstrates the value of developing automated patch-clamp assays for functional characterization of ion channel gene variants.

Graphical abstract

Introduction

The revolution in genome sequencing has long promised to usher in an era of precision medicine. Though it is now easy to sequence and discover variants, it has proved more difficult to determine their impact on clinical phenotypes. There are now more than 1 billion genetic variants that have been discovered in the human genome and are available in the build 155 version of dbSNP.¹ Of the 1.15 million variants classified in ClinVar, 41% are listed as “variants of uncertain significance” (VUSs)² and therefore cannot be used to aid clinical decision-making.³ Thus, there is an urgent need to develop accurate and high-throughput assays to assess the functional consequences of gene variants.⁴

Under the current American College of Medical Genetics and the Association for Molecular Pathology (ACMG/AMP) variant classification framework,⁵ variants are classified into one of five categories: pathogenic, likely pathogenic, VUSs, likely benign, or benign. All variants start as VUSs and can be actively moved toward a pathogenic or benign classification only if there is sufficient evidence, based on several lines of evidence including frequency in affected individuals, segregation data, population allele frequency, in silico predictions, and functional data from in vivo or in vitro assays. The issue of VUSs is particularly acute for rare diseases, which are defined by a prevalence of less than five in 10,000.⁶ By their very nature, it is rare to have clinical phenotyping data from a sufficiently large cohort of individuals with the same variant to establish strong evidence for pathogenicity. Yet it is in rare diseases that a genetic diagnosis can be especially valuable for detecting asymptomatic cases during family screening and, in some cases, initiating preventative treatment.

Inherited long QT syndrome (LQTS [MIM: 192500]) is an autosomal dominant cardiac disorder with an estimated incidence of 1 in 2,000.⁷ It predisposes otherwise healthy individuals to arrhythmias and sudden cardiac death.⁸ A clinical diagnosis of LQTS is based on electrocardiogram (ECG) abnormalities (most notably a heartrate-corrected QT interval of over 470 ms in women and 450 ms in men) and suspicious personal or family history.⁹ Once diagnosed, LQTS can be effectively treated with lifestyle modifications, beta blockade, or cardiac denervation and, in more severe cases, with an implantable cardioverter-defibrillator (ICD).⁹ The major difficulty with managing individuals and their families lies with diagnosing the disease prior to the onset of a potentially lethal arrhythmic event. This is complicated by the highly variable presentation of LQTS. For example, the hallmark QT prolongation is absent in ∼30% of genotype-positive siblings of probands with the disease.¹⁰

The two major subtypes of LQTS are caused by variants in KCNQ1 (MIM: 607542) and KCNH2 (also called human ether a-go-go-related gene, hERG; MIM: 152427), which encode for cardiac potassium channels. We now have a detailed understanding of how defects in these ion channels cause LQTS.^11,12 Furthermore, ion channels are amenable to high-throughput functional analysis using automated patch-clamp (APC) platforms.13, 14, 15 Given that (1) LQTS is an actionable disease where improved diagnostic yields from genetic testing will bring tangible benefits to individuals and their families, (2) the mechanisms linking gene-level defects to clinical phenotypes are well understood, and (3) there are high-throughput assays available for characterizing variants in ion channels, LQTS is an excellent candidate for the development of clinically actionable functional genomics assays.

To help standardize the evaluation and application of functional evidence, the Clinical Genome (ClinGen) Sequence Variant Interpretation (SVI) Working Group recently published recommendations on validating functional assays.¹⁶ Accordingly, the aims of this study were to validate a patch-clamp assay for KCNH2 missense variants and to determine the strength of evidence that it can provide for clinical variant classification.

Material and methods

KCNH2 patch-clamp assay

Selection of KCNH2 variants

A total of 44 KCNH2 (GenBank: NM_000238.4) missense variants listed in the ClinVar database as either benign/likely benign or pathogenic/likely pathogenic underwent internal variant curation, by experienced cardiac genetic counselors, using the ACMG/AMP criteria⁵ and variant interpretation guidelines.¹⁷ This included the following levels of evidence for pathogenicity: (1) Prevalence of the variant in affected individuals (PS4) as supporting (for 2 probands), moderate (4 probands), strong (8 probands), very strong (>16 probands).¹⁸ Proband counts are from ClinVar, published literature, and our internal database. Care has been taken not to double count any probands across these sources. (2) Absence/rarity in the population (PM2) as supporting evidence.¹⁹ (3) Co-segregation with disease in three or more affected family members (PP1). (4) Deleterious in silico predictions (PP3) were derived from Meta scores including MetaLR,²⁰ MEtaSVM,²⁰ MetaRNN,²¹ and REVEL.²² Individual scores were also used, all of which were included in one or more of the Meta scores and were cross-checked for consistency. (5) Mutational hotspot evidence (PM1) for transmembrane/linker/pore specific for KCNH2.²³ Evidence for benign variants included abundance of population allele frequencies (BA1 or BS1), multiple lines of in silico evidence suggesting a benign effect (BP4), and multiple reputable labs reporting the variant as benign/likely benign in ClinVar (BP6). Out of the 44, 14 variants met benign/likely benign criteria, and 17 met pathogenic/likely pathogenic classification without applying functional evidence (PS3).¹⁶ These 31 variants were used as variant controls for KCNH2 assay validation (see Table S1).

Generation of KCNH2 variant cell lines

Methods for the assay pipeline have been described in detail previously.²⁴ In brief, DNA plasmids containing each variant, confirmed by Sanger sequencing, were ordered from GenScript Inc. (Pistcataway, NJ, USA) and subcloned into a bicistronic expression plasmid. This expression plasmid enabled co-expression of variant and wild-type (WT) KCNH2 alleles to recapitulate the heterozygous expression of variants in individuals with LQTS2. The plasmids were used to generate Flp-In HEK293 (Thermo Fisher, cat. #R78007) stable cell lines, which were assayed using an APC electrophysiology platform (SyncroPatch 384PE, Nanion Technologies, Munich, Germany).

Sample-size determination for reliable current-density estimation

To establish sample-size requirements and the dynamic range of the assay (i.e., normal function versus complete loss of function), positive (WT Flp-In HEK293 stable cell line) and negative (Flp-In HEK293) controls were assayed across six full and two half-full 384-well plates, using three separate batches of cells.

Controls and replications

Following recommendations from the ClinGen SVI Working Group, all assay plates included WT cells as positive controls and blank cells as negative controls, which left sufficient wells to assay ten heterozygous variant cell lines per plate. Two technical replicates for each batch of cells were performed, and this was repeated for a fresh batch of cells grown from a new frozen stock for the same construct. Thus, each variant was assayed on four separate plates.

Data analysis

Quality control (QC) measures were applied so that only high-quality patch-clamp recordings were included: seal resistance >300 MΩ, capacitance of 5–50 pF, and leak-corrected current at −120 mV < 40 pA of baseline.²⁴ A suite of voltage protocols used in patch-clamp electrophysiology experiments was used to measure activation, deactivation, and inactivation gating, as described in detail previously.²⁴ Current density was also measured by first stepping the voltage to +40 mV for 1 s and then measuring peak tail current at −50 mV, the voltage at which the rapid delayed-rectifier current, I_Kr (which is encoded by KCNH2), peaks during repolarization of the cardiac action potential.¹¹

The cell-cell variation in channel expression resulted in peak current densities with a non-Gaussian distribution. The peak tail current densities from each plate were transformed to a normal distribution by using a square-root (sqrt) function, thereby allowing the use of mean and standard deviations (SDs). We then normalized data to the mean value of WT from the respective plate (normalized peak tail current densities_sqrt) to control for any plate effects. To determine the sensitivity and specificity of the peak tail current density measurements for differentiating between benign and pathogenic variants, we used 2 SD from the mean of all benign/likely benign variants to establish the threshold for functionally normal versus abnormal variants. The odds of pathogenicity (OddsPath) was then calculated based on the performance of the assay using the formulas derived by Brnich et al.¹⁶ assuming a perfect binary readout.

$$OddsPath=\frac{P2\times \left(1-P1\right)}{\left(1-P2\right)\times P1},$$ (Equation 1)

$$\text{where}P1=\frac{numberofpathogenicorlikelypathogeniccontrols}{totalnumberofcontrolvariants}$$

$$\text{and}P2=\{\begin{array}{l}\frac{numberoffunctionallyabnormalvariants}{numberoffunctionallyabnormalvariants+1},\text{for}\text{pathogenic}\text{variants}\hfill \\ \frac{1}{numberoffunctionallynormalvariants+1},\text{for}\text{benign}\text{variants}\hfill \end{array}$$

VUS reclassification

Selection of variants for reclassification

A total of 45 KCNH2 variants were selected for reclassification based on the following criteria: (1) clinical significance listed in the ClinVar database as either a VUS or not provided and (2) published functional data from a previous study²⁵ indicating loss-of-function phenotype according to the criteria established in this study. As BS3_supporting would not be sufficient to reclassify these VUSs as likely benign, we did not investigate these VUSs further, but rather focus only on those VUSs that have loss of function to investigate whether there was sufficient evidence strength from the functional assay to support reclassification as likely pathogenic.

Following baseline assessment of the 45 variants, one was classified as likely pathogenic and 44 were confirmed as VUSs using the following criteria: absence/rarity in the population (PM2_supporting); prevalence of the variant in affected individuals (PS4) as supporting (for 2 probands), moderate (4 probands), strong (8 probands), or very strong (>16 probands); deleterious in silico predictions (PP3); and co-segregation with disease in three or more affected family members (PP1). Functional evidence was applied at moderate level for severe loss of function (> 4 SD of mean) (PS3_moderate) or supporting level for partial loss of function (between 2 and 4 SD of mean) (PS3_supporting), and we assessed the resulting impact on the baseline classification.

Results

Sample-size determination for reliable current-density estimation

The raw current densities measured at −50 mV from 1,198 recordings (from a possible 2,496 wells) that passed all quality control measures (see material and methods) shows the distribution of data, which reflects the cell-to-cell variability in protein expression levels and is positively skewed (Figures S1Ai and S1Aii). The data can be converted to a normal distribution using a square-root transformation. The SD for this transformed distribution was 33.4% (Figure S1Bii). This translates to requiring a minimum of 14 cells per variant to detect a 50% difference or 54 cells per variant to detect a 25% difference at 90% power with 95% confidence interval (CI) (Figure S2A).

Based on these results, we designed the assay layout to contain 32 wells for each variant, as well as 32 wells each for the WT and negative controls. Each variant was assayed on at least four different assay plates to ensure adequate sample size and reproducibility (Figure 1A). On average, 67 ± 16 (SD) recordings per variant passed all QC criteria (see material and methods). Example peak tail currents for WT control wells are shown in Figure 1Bi, and the peak tail current density for all WT and negative control recordings from the 24 plates are summarized as violin plots in Figures 1Bii and 1Biii. The WT and negative controls showed a good dynamic range despite some variation between plates.

Figure 1.

Implementation of best-practice recommendations for KCNH2 patch-clamp assay

(A) A total of 12 flasks were cultured in parallel, which includes WT, 10 variant lines, and a negative control (i). The function of different KCNH2 variants expressed in HEK293 cells was assessed using an automated patch-clamp platform with two technical replicates for each run (ii). This process was repeated once with a different batch of cells cultured independently, hence each variant was assessed four times (see material and methods) with WT and negative controls present on each plate (iii). Overall, 49% of cells passed all quality control measures, resulting in 67 ± 16 (SD) successful recordings for each variant across the four plates (iv).

(B) Violin plot of current density at −50 mV for WT (i) and negative control (ii). Insets show the peak tail current for WT and negative controls. Shown in (iii) is the dynamic range of the assay, showing a clear separation between the positive (WT, filled) and negative control (clear). Data are shown as mean ± 95% CI.

Reproducibility of the KCNH2 patch-clamp assay

An example family of current traces for WT was shown in Figure 2A, and example current traces and violin plot summaries for the current density measurements at −50 mV for a benign variant (p.Lys897Thr [c.2690A>C]), a likely pathogenic variant (p.Gly584Ser [c.1750G>A]), and a pathogenic variant (p.Ala561Val [c.1682C>T]) are shown in Figure 2B. The current density at −50 mV reflects the number of channels reaching the plasma membrane as well as impacts of variants on the activation and inactivation kinetics and any ion-selectivity changes.¹³ The normalized peak tail current densities_sqrt for these variants, presented as violin plots with each replicate plate highlighted by different colors in Figure 2B, show that the assay can detect loss of current density caused by these variants.

Figure 2.

Reproducibility and replication of the KCNH2 patch-clamp assay

v(A) Steady-state deactivation voltage protocol and an example family of current traces for WT channels, which were depolarized to +40 mV for 1 s before repolarizing to voltages between +20 mV and −150 mV for 3 s to deactivate the channel. The tail current traces recorded at −50 mV were used to derive the current density and are highlighted in black in the current trace. The red box on the entire trace illustrates the section shown for each variant in (B).

(B) Example peak tail current of WT:WT, p.Lys897Thr:WT (benign), p.Gly584Ser:WT (likely pathogenic), p.Ala561Val:WT (pathogenic), and negative control corresponding to the highlighted region (red box) within the voltage protocol. The quantified peak tail current densities were transformed using square-root function and normalized to the respective WT from the same plate. Four replicates were acquired for each variant, and the corresponding WT and negative controls for those plates were also shown. The circles are mean of each replicate, and error bar is the standard error of mean. The total numbers of patch-clamp experiments from four replicates are indicated within the bracket.

Determination of the functional evidence strength for KCNH2 patch-clamp assay

A set of 31 variants, 14 benign/likely benign and 17 pathogenic/likely pathogenic, that had been classified in the absence of functional data (see Table S1) was used to establish a threshold for separating functionally normal and abnormal variants. The normalized peak tail current densities_sqrt for the benign/likely benign and pathogenic/likely pathogenic variants were clearly segregated (Figure 3A). The representative current traces used to measure current density for benign/likely benign and pathogenic/likely pathogenic variant controls are shown in Figures S3 and S4, respectively. The functionally normal region was defined as the mean ± 2 SD for the normalized peak tail current densities_sqrt for the WT and 14 benign/likely benign variant controls (blue region, Figure 3A). The pathogenic/likely pathogenic variant controls all reside outside the functionally normal region. Based on the threshold for normal and abnormal current density established using this set of variants, the KCNH2 patch-clamp assay achieved 100% sensitivity and 93% specificity (Figure 3B). The CI for the mean value for p.Arg148Trp (c.442C>T) classified as likely benign in this study using ACMG/AMP criteria crossed over into the abnormal range. Using the formula for calculating the OddsPath provided by ClinGen SVI Working Group¹⁶ (see Equation 1, material and methods), the assay achieved 0.063 for benign (corresponding to BS3_moderate) and 14.8 for pathogenic (corresponding to PS3_moderate; Table S2). For those pathogenic variants that have sufficient current density, we also analyzed gating phenotypes. p.Leu552Ser (c.1655T>C) showed faster deactivation and a hyperpolarized shift in the V_0.5 of steady-state activation (Figure S5), whereas p.Gly584Ser and p.Asp774Tyr (c.2320G>T) both showed enhanced inactivation, which would contribute to the reduced current density measured at −50 mV for these two variants (Figure 3). The gating parameters for these three variants, as well as for all benign/likely benign variant controls, are summarized in Table S4.

Figure 3.

Establishment of the ACMG BS3/PS3 strength using validation variants

(A) Summary plot of the peak tail current density measurements at −50 mV with square-root transformation applied and normalized to WT current density on the respective plates. Data are shown as mean ±95% CI. The mean and SD of the mean for WT and the 14 benign controls (blue circles) are 1.00 ± 0.11. A variant is defined as having normal function if its mean value lies within 2 SD of the mean of all the benign/likely benign variants (highlighted by the blue region). Functionally abnormal variant for loss-of-function (brown circles) is defined as having mean value more than 2 SD below the mean of the benign variants (PS3_moderate). An additional axis on the right was added to show the % of WT for the respective current density without the square-root transformation.

(B) Confusion matrix showing high specificity and high sensitivity for the classification of these variant controls based on the normal and abnormal threshold in (A).

Application of functional data for variant reclassification

A total of 45 variants reported in the ClinVar database were found to have reduced current density.¹⁹ These VUSs can be assigned with PS3_supporting for partial reduction or PS3_moderate for severe reduction in current density based on the calibrated KCNH2 assay (Figure 4A). Of the 45 variants chosen for the test cohort, one was reclassified as likely pathogenic based on the current ACMG/AMP criteria, and 44 remained as VUSs before applying functional data (Figure 4B; Table S3, column P). Analysis of current-density data for these 44 variants supported application of PS3_moderate for 31 of 44 VUSs and PS3_supporting for 13 of 44. Integrating these functional data with existing evidence allowed 6 of 44 VUSs to be reclassified as likely pathogenic (Figure 4C; Table S3, columns U/V). If the point-based system for variant classification proposed by Tavtigian and colleagues²⁶ were to be adopted, then an additional 9 VUSs could be reclassified as likely pathogenic (Figure 4D).

Figure 4.

Reclassification of KCNH2 VUSs that have abnormal function using a revised evidence strength

(A) The assignment of evidence strength for 45 VUSs in KCNH2 by incorporating the revised evidence levels based on the SD from the mean of benign controls: (1) PS3_supporting (2–4 SD smaller than the mean) for partial loss of function, (2) PS3_moderate (>4 SD smaller than the mean) for severe loss of function, and (3) BS3_supporting (within 2 SD from the mean). The respective % of WT current density was shown as right y axis. Data are shown as mean ± 95% CI.

(B) Categorial classification without functional data.

(C) Categorial classification with functional data applied as moderate or supporting (Table S3).

(D) Point-based classification²⁶ with PS3_moderate being applied as 3 points. Red is likely pathogenic and gray is VUSs.

Discussion

Patch-clamp assays have been a mainstay for assessing the functional impact of genetic variants in ion channel genes, including KCNH2 (reviewed in Delisle et al.²⁷). Traditionally, these assays have been very labor intensive, with typical throughputs of 10–20 cells per day. However, with the development of APC platforms, this throughput has increased up to thousands of cells per day,²⁸ thus making it feasible to start addressing the deluge of VUSs in ion channel genes. Here, we have formally validated a KCNH2 APC assay based on the recommendations proposed by the ClinGen SVI Working Group.¹⁶ Following recommendations and results of our preliminary study, we designed the assay to incorporate the following: (1) inclusion of positive and negative controls on each 384-well plate, (2) allocation of 32 wells per variant on at least four different assay plates to ensure adequate sample size and reproducibility, (3) normalization to WT (positive control) on each plate, and (4) transformation of data to ensure a normal distribution of the data. Using a cut-off between functionally normal and functionally abnormal of 2 SD below the mean value of the 14 benign/likely benign variant controls, the assay produced concordant results for 13 of 14 benign/likely benign variants and 17 of 17 pathogenic/likely pathogenic variant controls. Based on the OddsPath statistical analysis developed by Brnich et al., our assay is currently capable of providing functional evidence at a moderate level. However, for those variants that were only 2–4 SD lower than the mean of the benign variants, we have adopted a more conservative classification of supporting-level evidence as recommended by Brnich et al. for variants with partial loss of function.¹⁶ With this evidence, we could reclassify 6 of 44 VUSs (13.6%) as likely pathogenic.

Current density at −50 mV captures the phenotype of KCNH2 variants

Variant function was determined by measuring peak tail current density at −50 mV, a measure that reflects (1) the number of channels that have trafficked to the plasma membrane, (2) the proportion of channels that are activated during the depolarized voltage at +40 mV, (3) the proportion of channels that have recovered from inactivation at −50 mV,²⁴ and (4) the ionic selectivity and hence reversal potential for current flow. The vast majority (>80%) of loss-of-function missense variants in KCNH2 that have been characterized to date are known to reduce current density via trafficking defects.²⁹ There are, however, a small number of variants that reduce activation,³⁰ enhance inactivation,³¹ or alter ion selectivity,¹³ which will also be detected as reduced current density in our assay. Thus, this single assay interrogates the vast majority of mechanisms by which missense KCNH2 variants can affect current density produced by K⁺ channels containing KCNH2 subunits. ²⁷ and was able to detect known gating defects in likely pathogenic variant controls, i.e., the inactivation gating defect in p.Gly584Ser³¹ and deactivation defect in p.Leu552Ser³² (Figure S5). As abnormal inactivation and V_0.5 of activation would reduce current density, this should not be considered as further evidence beyond that assigned for reduced current density. However, accelerated channel deactivation could occur in the absence of reduced current density, in which case it could be assigned evidence strength at PS3_supporting.

It is also notable that the current density values across the cells were not normally distributed, which is consistent with what has been seen for cell-cell variation in single-cell gene-expression studies.³³ The positively skewed raw data could be converted to a normal distribution by applying a square-root transformation, and so this transformation was applied to all current density measurements.

Based on our analysis of the 14 benign/likely benign variant controls, the lowest threshold to be considered as functionally normal was 0.78 (for square root of current density), which is equivalent to 61% of WT current density (Figure 3A). This is slightly higher than the predicted 50% loss of function caused by heterozygous nonsense variants that are well established for causing LQTS2, albeit with incomplete penetrance.¹⁰

Functional patch-clamp assay has high sensitivity and specificity

All pathogenic/likely pathogenic variant controls fell in the functionally abnormal range. This indicates that our KCNH2 patch-clamp assay is highly sensitive for identifying functionally abnormal variants. It is also highly specific, which is useful for discerning the status of variants that are identified as incidental or secondary genetic findings. However, there is a caveat that assays based on transfected cDNAs in heterologous expression systems may miss variants that affect splicing or interactions with other proteins. The one likely benign variant that had its CI fall outside the normal range was p.Arg148Trp (c.442C>T). This variant was classified internally as likely benign mainly due to its high minor allele frequency (gnomAD v2.1.1 AF: 0.001133; gnomAD v3 AF: 0.0006441). However, there are conflicting classifications in ClinVar (five benign, three likely benign, two uncertain significance), and there are numerous reports linking p.Arg148Trp to LQTS2.34, 35, 36 Thus, it is possible that p.Arg148Trp, though insufficient to cause LQTS2 alone, does have a deleterious effect on function and may be a risk allele.³⁷

Incorporation of functional data for variant classification in LQTS genes

APC data can provide functional evidence for KCNQ1, KCNH2, and SCN5A (MIM: 600163) variant classification.13, 14, 15^,38 However, factors such as a lack of detailed guidance on how functional evidence should be evaluated and differences in the application of the functional evidence strength (PS3/BS3) criterion has contributed to discordances in variant interpretation between diagnostic laboratories.¹⁶ The KCNH2 patch-clamp assay reported in this study was stringently assessed to provide evidence for normal and abnormal function at moderate levels (BS3_moderate or PS3_moderate; see Table S2). However, lack of an in vitro functional defect does not exclude the possibility that a variant might be abnormal in vivo. For example, the HEK cell assay described here may not fully recapitulate all of the protein interactions that occur in native cardiomyocytes, such as the interaction between hERG1a and hERG1b isoforms.³⁹ Given the potential lethality of LQTS2, applying BS3_moderate to these variants could risk causing harm to individuals. Accordingly, we suggest that the evidence provided for functionally normal variants should be applied at supporting level (BS3_supporting) when classifying clinical KCNH2 variants. For functionally abnormal variants, we suggest applying evidence at moderate level (PS3_moderate) for KCNH2 variants with severe loss of function (>4 SD from the mean of benign/likely benign variants) or supporting level for KCNH2 variants with partial loss of function (2–4 SD from the mean of benign/likely benign variant controls) (Figure 4A; Table S3, column S).

Another limitation/barrier to developing functional assays for assessing LQTS variants is the lack of definitively benign variants in LQTS genes (KCNH2, as well as SCN5A or KCNQ1), which restricts the strength of evidence that can be provided by a functional assay to a moderate level (PS3_moderate).⁴⁰ Previous studies have used patch-clamp data at strong-level evidence for variant classification,^15,38,41 although this contradicts the advice in the current ACMG/AMP categorial classification system. The OddsPath ratio for our KCNH2 assay for PS3 is 14.3:1, which does not quite reach the level required for a strong level of evidence (minimum OddsPath of 18.7:1 for strong and 4.3:1 for moderate). If a point-based classification system²⁶ was used for PS3, our assay would be given 3 points rather than 2 (moderate) or 4 (strong), which would result in an additional nine VUSs being reclassified as likely pathogenic.

Significance

Deciphering the clinical impact of VUSs is a key step to unlocking the full potential of genomic medicine. High-throughput functional assays, such as the KCNH2 patch-clamp assay described here, can improve the diagnostic yield of genetic testing. However, only ∼16% of the VUSs were able to be reclassified when functional evidence was incorporated using the current categorical variant classification system. This relatively low rate of reclassification is due to their being only a limited amount of other evidence (e.g., small number of cases, lack of segregation data) for each variant. The rate of reclassification would have been increased to 36% if a recently proposed point-based system²⁶ were adopted (see Figure 4B). A major strength of our assay is that it has a very high specificity for detecting variants that cause long QT syndrome type 2 (LQTS2). Thus, there is the potential to improve the evidence strength of our assay to strong, which would considerably increase the reclassification yield. To achieve this will require identification of more benign variants so that we can increase the number of benign controls in our assay. Results from this assay will be an important resource for the ClinGen Variant Curation Expert Panel (VCEP) when establishing gene-specific criteria for KCNH2. Our result has important implications for the management of patients with LQTS2 in terms of diagnosis and the provision of gene-specific advice on lifestyle modifications and family screening.^9,42 Our study also provides a blueprint for the validation of future APC assays in other ion channel genes.

Published: June 9, 2022

Data and code availability

The datasets supporting the current study have been deposited in Dryad (https://doi.org/10.5061/dryad.m905qfv38). The Python codes that were used to perform the QC assessment and data analysis are available at https://git.victorchang.edu.au/projects/SADA/repos/syncropatch_automated_analysis/browse.