Human KCNQ5 de novo mutations underlie epilepsy and intellectual disability

Abstract

We identified six novel de novo human KCNQ5 variants in children with motor/language delay, intellectual disability (ID), and/or epilepsy by whole exome sequencing. These variants, comprising two nonsense and four missense alterations, were functionally characterized by electrophysiology in HEK293/CHO cells, together with four previously reported KCNQ5 missense variants (Lehman A, Thouta S, Mancini GM, Naidu S, van Slegtenhorst M, McWalter K, Person R, Mwenifumbo J, Salvarinova R; CAUSES Study; EPGEN Study; Guella I, McKenzie MB, Datta A, Connolly MB, Kalkhoran SM, Poburko D, Friedman JM, Farrer MJ, Demos M, Desai S, Claydon T. Am J Hum Genet 101: 65-74, 2017). Surprisingly, all eight missense variants resulted in gain of function (GOF) due to hyperpolarized voltage dependence of activation or slowed deactivation kinetics, whereas the two nonsense variants were confirmed to be loss of function (LOF). One severe GOF allele (P369T) was tested and found to extend a dominant GOF effect to heteromeric KCNQ5/3 channels. Clinical presentations were associated with altered KCNQ5 channel gating: milder presentations with LOF or smaller GOF shifts in voltage dependence [change in voltage at half-maximal conduction (ΔV₅₀) = ∼-15 mV] and severe presentations with larger GOF shifts in voltage dependence (ΔV₅₀ = ∼-30 mV). To examine LOF pathogenicity, two Kcnq5 LOF mouse lines were created with CRISPR/Cas9. Both lines exhibited handling- and thermal-induced seizures and abnormal cortical EEGs consistent with epileptiform activity. Our study thus provides evidence for in vivo KCNQ5 LOF pathogenicity and strengthens the contribution of both LOF and GOF mutations to global pediatric neurological impairment, including ID/epilepsy.NEW & NOTEWORTHY Six novel de novo human KCNQ5 variants were identified from children with neurodevelopmental delay, intellectual disability, and/or epilepsy. Expression of these variants along with four previously reported KCNQ5 variants from a similar cohort revealed GOF potassium channels, negatively shifted in V₅₀ of activation and/or delayed deactivation kinetics. GOF is extended to KCNQ5/3 heteromeric channels, making these the predominant channels affected in heterozygous de novo patients. Kcnq5 LOF mice exhibited seizures, consistent with in vivo pathogenicity.

Keywords: KCNQ5; M current; channelopathy; epilepsy; intellectual disability.

Graphical abstract

Figure 1.

Human KCNQ5 de novo mutations identified from patients with intellectual disability (ID)/epilepsy alter highly conserved residues distributed throughout the channel subunit. A: location of de novo KCNQ5 mutations mapped to the structural domains of the KCNQ5 subunit defined by cryo-electron microscopy (cryo-EM) of the homologous Xenopus KCNQ1 subunit (57). R178X and R443X (blue) encode predicted truncated proteins, prematurely terminated at the base of S2 (R178X) and between Helix-A and Helix-B (R443X) in the cytosolic COOH-terminal tail. K289N, P369T, R443Q, and H576R (red) encode missense mutations distributed at the extracellular edge of S5 (K289N) and throughout the COOH-terminal tail (P369T, R443Q, H576R). Previously reported de novo KCNQ5 missense mutations (47) are depicted in orange: V145G in S1, L341I in S6, P369R in Helix-A, and S448I in the loop between Helix-A and Helix-B. B: alignment of amino acid residues for each KCNQ5 residue affected by de novo mutations (highlighted in yellow), with the sequences of all 5 human KCNQ subunits. KCNQ5 mutations alter residues highly conserved in all 5 KCNQ subunits. K289N neutralizes a positive charged residue located between 2 flanking negatively charged residues, E288 and D290 (highlighted in blue). C: structural homology model of a tetrameric KCNQ5 channel based on the structure of KCNQ1 (57), viewed obliquely from the extracellular side. The ribbon structure of a single subunit is shown (blue) with the surface volume representations of the other 3 subunits and a ring of 4 calmodulin accessory subunits (yellow). Locations of residues affected by de novo mutations are shown in the ribbon structure, using the same color scheme as in A.

Figure 2.

Human KCNQ5 channels carrying de novo missense mutations from patients with intellectual disability (ID)/epilepsy result in gain of function (GOF) expressed in HEK293 cells. Left: current traces evoked from HEK293 cells transfected with wild-type (WT) KCNQ5 (black) and KCNQ5 missense mutants (K289N, P369T, R443Q, H576R) (red), under voltage clamp from a holding potential of −80 mV, stepped through a family of 3-s voltage steps from −100 mV to 60 mV and returned to a poststep potential of −60 mV. Right: conductance/voltage (G/V) plots are shown for each construct, calculated from isochronic current measurements at the end of 3-s voltage steps and the Nernst potassium reversal potential based on the recording solutions. All KCNQ5 missense mutations (K289N, P369T, R443Q, H576R) resulted in channels negatively shifted in voltage dependence of activation toward hyperpolarizing potentials relative to WT (+). All transfections were performed in a background of cotransfected catalytically inactive BACE1(D289N) to assist channel expression levels (49) (but see text). Plots displayed as means ± SE, with fits to single Boltzmann functions and values for fitted voltage at half-maximal conduction (V₅₀) and N (in brackets). G/G_max, conductance normalized to maximal conductance; V_m, membrane potential.

Figure 3.

Previously reported de novo KCNQ5 mutations associated with intellectual disability (ID)/epilepsy [Lehman et al. (47)] result in channels with gain of function (GOF) expressed in HEK293 cells. Left: current traces evoked from HEK293 cells transfected with wild-type (WT) KCNQ5 (black) and KCNQ5 missense mutations (V145G, L341I, P369R, S448I) previously reported by Lehman et al. (47). Currents evoked under voltage clamp from a holding potential of −80 mV stepped through a family of 3-s voltage steps from −100 mV to 60 mV and returned to a poststep potential of −100 mV. Right: conductance/voltage (G/V) plots are shown for each construct, measured from tail currents at −100 mV. All mutations (V145G, P369R, S448I) resulted in channels shifted in voltage dependence of activation toward hyperpolarized potentials relative to WT (+), except for L341I, which exhibited a voltage at half-maximal conduction (V₅₀) (−14.1 mV) slightly depolarized to WT (−20.3 mV). However, L341I, along with V145G and P369R, all exhibited significant constitutive conductance at the holding potential of −80 mV, evident from their G/V plots and instantaneous currents at the beginning of voltage steps. Both properties of hyperpolarized voltage dependence of activation (V145G, P369R, S448I) and constitutive conductance at the holding potential (V145G, L341I, P369R) contributed to GOF for all these mutations. Plots displayed as means ± SE, with fits to single Boltzmann functions and values for fitted V₅₀ and N (in brackets). G/G_max, conductance normalized to maximal conductance; V_m, membrane potential.

Figure 4.

Delayed deactivation kinetics contribute to gain of function (GOF) for de novo KCNQ5 mutations V145G, L341I, and P369R. A: examples of tail current traces recorded at −100 mV, following a 3-s activation voltage step to 0 mV for wild-type (WT), V145G, L341I, P369R, and S448I. Delayed deactivation results in GOF due to accumulation of open channels with repetitive depolarizations. Tail currents fitted with either single- or double-exponential functions (blue): I_tail(t) = A₁exp(−t/τ₁) + A₂exp(−t/τ₂) + C (baseline offset). B: examples of exponential fits of tail currents (top), superimposed upon recorded tail current traces (gray) normalized to maximal current and baseline current at −100 mV (bottom). C: fitted taus of deactivation (τ₁, τ₂). V145G, L341I, and P369R fitted by the sum of 2 exponentials; WT and S448I fitted by single exponentials. The primary components of deactivation (τ₁) for V145G, L341I, and P369R were 2- to 4-fold slower compared to WT, with no significant difference between WT and S448I. Plots displayed as medians and interquartile ranges (IQRs), with values for the median and N (in brackets) for each sample set. Pairwise Mann–Whitney tests for significance: ***P < 0.0001; ns, nonsignificant. D: the secondary fast component of deactivation (Α_2, τ₂) contributes to 10–20% of total tail currents for V145G, L341I, and P369R.

Figure 5.

Coexpression of wild-type (WT) KCNQ5 with BACE1(D289N) and in combinations of KCNQ5(P369T) and WT KCNQ3: dominant gain-of-function (GOF) effect of P369T on heteromeric KCNQ3/KCNQ5(P369T) channels and lack of effect with BACE1(D289N). A, left: current traces from Chinese hamster ovary (CHO) cells transfected with WT KCNQ5 and WT KCNQ3 [denoted KCNQ5(+) and KCNQ3(+)] controls (black) compared with KCNQ5(+)+BACE1(D289N) (red) and KCNQ5(P369T) (blue). Right: similar current traces from KCNQ5(+)+KCNQ3(+) control (black) compared with KCNQ5(P369T)+KCNQ3(+) (red) and KCNQ5(P369T)+KCNQ5(+) (blue). For all recordings, currents evoked under voltage clamp from a holding potential of −80 mV stepped through a family of 3-s voltage steps from −110 mV to 60 mV and returned to a poststep potential of −100 mV. B: conductance-voltage (G/V) plots derived from tail currents for KCNQ5(+), KCNQ5(+) + BACE1(D289N), and KCNQ5(P369T). G/V plots for KCNQ5(+) and KCNQ5(+) + BACE1(D289N) are statistically indistinguishable. KCNQ5(P369T) is left-shifted relative to both KCNQ5(+) and KCNQ5(+) + BACE1(D289N). Single Boltzmann fitted values: KCNQ5(+) [voltage at half-maximal conductance (V₅₀) = −10.1 mV, N = 10], KCNQ5(+) + BACE1(D289N) (V₅₀= −18.4 mV, N = 5), KCNQ5(P369T) (V₅₀ = −52.0 mV, N = 7). C: G/V plots for KCNQ5(+) + KCNQ3(+), KCNQ5(P369T) + KCNQ3(+), and KCNQ5(P369T) + KCNQ5(+). G/V plots for KCNQ5(P369T) + KCNQ3(+) and KCNQ5(P369T) + KCNQ5(+) are both significantly left-shifted from KCNQ5(+) + KCNQ3(+), demonstrating a dominant GOF effect by the mutant KCNQ5(P369T) subunit. Single Boltzmann fitted values: KCNQ5(+) + KCNQ3(+) (V₅₀ = −8.9 mV, N = 8), KCNQ5(P369T) + KCNQ3(+) (V₅₀ = −21.0 mV, N = 7), KCNQ5(P369T) + KCNQ5(+) (V₅₀ = −20.8 mV, N = 12). G/G_max, conductance normalized to maximal conductance; V_m, membrane potential.

Figure 6.

KCNQ5(P369T) subunits slow deactivation when coexpressed in homomeric KCNQ5 or heteromeric KCNQ5/3 channels. A: tail currents recorded from Chinese hamster ovary (CHO) cells transfected with KCNQ3(+) + KCNQ5(+) and KCNQ5(+) controls (black, left) compared with KCNQ3(+) + KCNQ5(P369T) and KCNQ5(+) + KCNQ(P369T) (red, right). Coexpression of KCNQ5(P369T) significantly slowed deactivation kinetics only when coexpressed with KCNQ3(+), compared with controls. Coexpression of KCNQ5(P369T) + KCNQ5(+) exhibited a trend toward slower deactivation relative to KCNQ5(+), but this trend did not reach statistical significance. Tail currents recorded at −100 mV under voltage clamp, following a family of 3-s voltage steps from −110 mV to 60 mV, from a holding potential of −80 mV. B: taus of deactivation (τ₁, τ₂) for tail currents from KCNQ5(+) (Q5), KCNQ5(+) + BACE1(D289N) (Q5+BACE1), KCNQ5(P369T) (P369T), KCNQ5(P369T) + KCNQ5(+) (P369T+Q5), KCNQ5(+) + KCNQ3(+) (Q5+Q3), and KCNQ5(P369T) + KCNQ3(+) (P369T+Q3) fitted to either single- or double-exponential functions: I_tail(t)= A₁exp(−t/τ₁) + A₂exp(−t/τ₂) + C (baseline offset). Mean τ₁ and τ₂ values with SE shown for each condition, with N in brackets. Deactivation rates (τ₁) between KCNQ5(+) and KCNQ5(+) + BACE1(D287N) were statistically indistinguishable. Wild-type (WT) KCNQ5/3 heteromeric channels [Q5(+) + Q3(+)] exhibited 3-fold accelerated deactivation (τ₁ = 58 ms) relative to KCNQ5(+) homomeric channels (τ₁ = 178 ms). Coexpression of KCNQ5(P369T) with KCNQ3(+) [P369T + Q3(+); τ₁ = 336 ms] significantly slowed deactivation compared with WT KCNQ5/3 currents (τ₁ = 58 ms). Coexpression of KCNQ5(P369T) with KCNQ5 [P369T + Q5(+); τ₁ = 216 ms] produced tail currents with deactivation kinetics that were insignificantly different from KCNQ5(+) (τ₁ = 178 ms). C: the secondary fast component of deactivation (A₂, τ₂) accounted for ∼3–20% of total tail currents, although most tail currents were well fitted by single exponentials. Pairwise Mann–Whitney test for significance: ***P < 0.001, **P < 0.002, *P < 0.05; ns, nonsignificant. WT KCNQ5 and WT KCNQ3 denoted by KCNQ5(+) and KCNQ3(+), respectively.

Figure 7.

De novo KCNQ5 nonsense mutations R178X and R443X are loss-of-function (LOF) alleles. No functional currents were recorded from HEK293 cells transfected with R178X or R443X. A: examples of current traces from untransfected (HEK293) and GFP-, R178X-, and R443X-transfected cells. R178X- and R443X-transfected cells produced currents indistinguishable from untransfected and GFP-transfected controls. Currents evoked from a holding potential of −80 mV stepped through a family of 3-s voltage steps from −100 mV to 60 mV and a poststep potential of −100 mV. B: R178X and R443X current densities (blue) are indistinguishable from GFP-transfected controls (green). By contrast, wild-type (WT, gray) and all de novo KCNQ5 mutations (red and orange) expressed 10- to 20-fold higher current densities (I/C) compared with either GFP controls or these missense mutations. C: R178X and R443X exhibited membrane potentials (V_m) indistinguishable from untransfected and BACE1(D289N)- or GFP-transfected controls (∼−20 mV). By contrast, WT and all de novo KCNQ5 mutations produced cells with hyperpolarized V_m (−50 mV to −70 mV). Plots displayed as median and interquartile range (IQR). Pairwise Mann–Whitney tests for statistical significance: ***P < 0.0001, **P < 0.008, *P < 0.020; ns, nonsignificant.

Figure 8.

Murine Kcnq5 loss-of-function (LOF) deletion alleles created by targeted Cas9/single guide RNA (sgRNA). A: site of Cas9/sgRNA targeted deletions in a diagram of the Kcnq5 subunit. A sgRNA was selected to target Cas9 to exon 5 encoding the NH₂-terminal edge of the pore α-helix (P). Two targeted deletion mouse lines were recovered (Q5d1, Q5d2). B: genomic DNA chromatograms of wild-type (WT) and homozygous Kcnq5 deletion alleles (Q5d1, Q5d2) sequenced through the Cas9-targeted site. Targeted deletion of 37 base pairs (Q5d1) and 33 base pairs (Q5d2) observed, deleting the native 3′ splice site of exon 5 and the 5′ end of the following intron 5. C: DNA sequences and predicted open reading frames (ORFs) of Kcnq5 exon 5 for WT (Q5 WT, green ORF) and deletion alleles (Q5d1, blue ORF; Q5d2, yellow ORF), with the sgRNA shown in red. Both deletion alleles ablate the native donor splice site of exon 5 and generate predicted ORFs, which truncate the channel protein prematurely. These deletion alleles likely result in LOF from in vivo nonsense-mediated decay (see Fig. 9) (61).

Figure 9.

Kcnq5 Cas9-generated alleles are loss of function (LOF) by selective loss of mRNA and protein. Kcnq5 LOF mice exhibit late-onset seizures. A: quantitative RT-PCR (qRT-PCR) assessment of transcript abundance for Kcnq1–5 relative to Gapdh, with whole brain RNA from wild-type (WT) (BL6) and homozygous Kcnq5 LOF alleles (Q5d1, Q5d2). Both Kcnq5 LOF alleles selectively reduce Kcnq5 transcript abundance by ∼10,000-fold relative to WT, without appreciably altering the expression levels of Kcnq1-4. Each measurement plotted as mean ± SE, for N = 3 or 4. dCt = (sample Ct − GAPDH Ct). B: Western blot of membrane lysates from cerebral cortex from WT (BL6) and homozygous Kcnq5 LOF alleles (Q5d1, Q5d2) probed with a commercial anti-KCNQ5 antibody raised against full-length KCNQ5 peptide. KCNQ5 protein detected as an ∼103-kDa band in WT lysate, which is absent in lysates from both Kcnq5 LOF alleles. N = 3, with similar results. (See Supplemental Fig. S4 for full molecular mass range.) C: example of handling-induced tonic-clonic seizure observed in a Kcnq5 LOF adult. Neck and tail flexion observed at 2.20 (s), loss of standing posture at 3.73 (s). (See Supplemental Fig. S3, Supplemental Video.) D: summary of age of onset of handling-induced seizures observed for homozygous Kcnq5 LOF alleles outcrossed to C57BL/6J for 1–2 (N1/2) and 5 generations (N5). Onset of seizures occurred in mature adults of both LOF alleles at ∼25 postnatal wk with a penetrance of ∼20–24%, in N1/2 animals. Onset of seizures was significantly increased to ∼40 postnatal wk in N5 animals, without altering penetrance (23–37%). Plots displayed as median and interquartile range (IQR). Statistical significance determined by pairwise Mann–Whitney tests: *P = 0.0135, ***P < 0.0001.

Figure 10.

Kcnq5 loss of function (LOF) mice exhibit abnormal interictal electroencephalographs (EEGs) and thermal-induced seizures. A: power spectral density plots of cortical EEGs from C57BL/6J [wild type (WT)] and Kcnq5 LOF adult mice, recorded during the resting awake state for 2 h. Kcnq5 LOF mice exhibited persistent higher power densities at high frequency bands (>8 Hz) compared with WT. f, frequency. B: summary of EEG power density as a function of frequency (f) for WT and Kcnq5 LOF mice. Significant enhancement of power densities for frequencies >8 Hz for Kcnq5 LOF mice compared with WT, consistent with increased epileptiform activity. Plots displayed as mean ± SE (envelopes); N = 7. C: example of EEGs from a Kcnq5 LOF mouse exhibiting epileptiform spikes (asterisks), polyspikes, and generalized high-amplitude oscillations and polyspikes. D: EEGs recorded at baseline (24°C) and during thermal-induced seizure (40°C) from a Kcnq5 LOF mouse. E: summary of thermal-induced seizure susceptibility for WT and Kcnq5 LOF mice. Seizure susceptibility plotted as fraction of animals that seized as a function of core body temperature. Approximately 50% of Kcnq5 LOF animals exhibited a thermal-induced seizure at 40.5°C compared with 0% of WT animals. N = 7 (combined data from 4 Q5d1 and 3 Q5d2 animals).