Bi-directional modulation of KV7 channels through KVS

To explore the potential modulation of KV7-mediated currents by KVS, we initiated our analyses by characterising K+ currents in Chinese hamster ovary (CHO) cells transiently co-expressing KV7.2 channel subunits together with different KVS in whole-cell patch-clamp recordings. Depolarising voltage steps (−100 mV to + 60 mV) induced slowly activating outwardly rectifying K+ currents that were completely absent in non-transfected CHO cells and in cells expressing only KVS (not shown). Strikingly, voltage-dependent currents were significantly reduced in cells co-expressing KV7.2 with either KV5.1, KV8.2, KV9.1 or KV9.2 subunits compared to cells expressing KV7.2 alone (Fig. 1A, B, D). Also, in cells co-expressing KV7.2 with either KV5.1, KV8.2, KV9.1 or KV9.2 we observed an alteration of biophysical characteristics in that the steady-state voltage-dependence of whole-cell currents was shifted to significantly more depolarised values as compared to cells expressing KV7.2 only. The magnitude of this shift was in the range of about + 10 mV for Vh values (Fig. 1F, H). The voltage sensitivity, as determined by the slope of the voltage dependence, was not altered by co-expression of KVS. The smaller current amplitudes and altered voltage-dependence were accompanied by significantly more depolarised resting potentials in the cells co-expressing KV7.2 with either KV5.1, KV8.2, KV9.1 or KV9.2 (not shown). In contrast, steady-state current amplitudes were significantly increased in cells co-expressing KV6.1 and KV8.1 channels without however affecting voltage dependence or membrane potentials compared to cells expressing only KV7.2 subunits (Fig. 1A, B, D). Co-expression of KV6.3, KV6.4 or KV9.3 did not affect the characteristics of currents through KV7.2 channels (Fig. 1A, D, H).

Fig. 1

Electrophysiological properties of CHO cells expressing KV7 channels are altered by co-expression of KVS. A Voltage stimulus protocol (left) and representative patch-clamp recordings from CHO cells expressing KV7.2 alone (black), and together with KV8.1 (orange) or KV8.2 (green). B, C Voltage-dependent current densities in cells expressing KV7 alone, and together with KV8.1 or KV8.2. D, E Summary statistics for steady-state current densities at + 20 mV obtained in recordings as shown in (B, C) as well as for similar recordings performed on cells expressing KV7 together with other KVS. F–I Normalized tail currents (F, G) and half-maximal activation voltage (H, I) as deduced from Boltzmann fits (solid lines in F, G) for the same cells as show in B–E. Recordings from cells expressing KV7.2 are shown in B, D, F and H while those expressing KV7.4 are shown in C, E, G and I

To assess whether KVS similarly modulated the properties of currents through other KV7 channels, we performed analogous experiments in CHO cells expressing either KV7.4 subunits alone or together with KVS. Notably, the effects of co-expression of KVS on KV7.4 subunits were very similar to those observed for KV7.2 channels, except that no effect was observed for KV5.1 and shifts in Vh were substantially more pronounced (Fig. 1C, E, G, I).

KV8 and KV7 subunits exist in multi-protein complexes in living cells

We then turned our attention to investigating how KVS subunits modulates KV7 channels at the molecular level, focusing on KV8.1 and KV8.2 subunits and analysed whether KV7 and KV8 subunits were present in the same protein complex and in close proximity to each other. We hence carried out BioID and co-immunoprecipitation experiments in HEK293 cells. For BioID experiments, we transiently overexpressed BioID2-HA-tagged KVS (KV8.1 or KV8.2) and flag-tagged KV7 subunits (Fig. 2). To ensure that the BioID2 fusion did not affect the expression and subcellular localisation of the KVS subunits, we performed immunostainings on transfected HeLa cells with antibodies against the HA tag at the C-terminal end of KV8.1, KV8.2 BioID2 fusion proteins. This revealed the expected intracellular localisation with a signal pattern typical of the endoplasmic reticulum (ER) within the cells. (Fig. 2B; Supplementary Fig. 6).

Fig. 2

KV7 and KVS assemble in close proximity. A BioID assay methodology. The KVS (8.1/8.2) was fused to the promiscuous form (BirA*) of the bacterial biotin ligase BirA and co-expressed with flag-tagged KV7.2 subunits in HEK293 cells. Upon the addition of biotin, proximal proteins (grey) were biotinylated within a labeling radius of ∼10 nm, whereas the distal proteins (green), remained unlabelled. B Representative confocal images of HEK293 cells transiently transfected with BioID2-HA-tagged KV8.1 (first row), BioID2-HA tagged KV8.2 (second row), BioID2-HA alone (third row). Cells were stained with either HA antibody (red, first column) or Alexa Fluor™ 488 streptavidin conjugate (green, second column). Scale bar, 10 μm. C Following biotin labelling, cells were lysed and biotinylated proteins were then purified using streptavidin beads and identified by western blot analysis. Immunoblotting (IB) was performed by using an anti-flag antibody to detect the flag-labelled KV7.2 subunits and an Alexa Fluor™ 488 streptavidin conjugate (SA) to detect the biotinylated proteins. Note the absence of flag-tagged KV7.2 subunits in the AP lane with BioID2-HA alone (indicated by red arrow). However, they are present in both the AP lanes (indicated by green arrows). Overlaying the signals obtained from anti-flag and streptavidin staining revealed a noticeable proximity of these distinct bands (merge). Abbreviations used: Inp (input), Sup (supernatant), AP (affinity purified). Similar results were obtained in n = 3 transfections

Biotinylation of endogenous proteins in cells expressing HA-BioID2-KvS (8.1 and 8.2) with exogenous biotin strongly stimulated a wide range of endogenous proteins on western blots probed with Alexa Fluor™ 488 streptavidin conjugate (data not shown). This indicates that the BioID2 moiety was adequately exposed in the KVS fusion construct, allowing for efficient biotinylation. Our next step was to test whether the Kv7 subunits are in close proximity to the KVS subunits. If they are indeed in close proximity, they should be biotinylated and then precipitated with streptavidin beads. To achieve this, we transiently expressed BioID2 tagged KVS (8.1 or 8.2) and flag-tagged KV7.2 constructs in HEK293 cells in the presence of 50 µM biotin for 24 h and then lysed the cells using a radio-immunoprecipitation assay (RIPA) lysis buffer. HEK293 cells transfected with BioID2-HA alone, processed in parallel, were used as negative controls. As shown in Fig. 2C (anti-flag staining, green arrows), the flag-tagged KV7.2 subunits were robustly precipitated using streptavidin beads, strongly suggesting that the KV7.2 subunits are indeed in close proximity (10 nm distance) to the KVS (8.1 and 8.2) subunits and are therefore biotinylated, whereas no precipitation occurred when BioID2 alone was used as a negative control (red arrow). To complement these data with an independent approach, we performed a proximity ligation assay (PLA) on cells transfected with myc-tagged KVS and flag-tagged KV7. Indeed, PLA signals were only observed in cells co-expressing both, KVS and KV7 (Supplementary Fig. 2).

We then investigated whether the flag-tagged KV7.2 was located in the same complex as the KVS subunits by carrying out co-immunoprecipitation experiments. The protein complexes were precipitated with anti-flag M2 beads, and the subunits in the precipitates were detected in western blots with antibodies directed against myc or flag tags, respectively (Fig. 3A, B). These results demonstrated that the KV7.2 and KVS channels were not only located in close proximity, but also exist in a single multi-protein complex. As a negative control, co-immunoprecipitation experiments were carried out using cell lysates devoid of flag-tagged KV7 subunits (Supplementary Fig. 3), demonstrating the specificity of co-immunoprecipitations.

Fig. 3

KV7 and KVS assemble into a protein complex. Co-immunoprecipitation experiments were performed using myc-tagged KV8.1 (left) and KV8.2 (right) subunits in combination with flag-tagged KV7.2 (A) and KV7.4 (B) channel subunits from HEK293 cell lysates using monoclonal anti-flag M2 conjugated agarose beads. The precipitate was then subjected to western blot analysis using monoclonal anti-flag M2 flag antibody and anti-myc antibodies respectively. Note the presence of both KV7 and KVS subunits in the precipitated fraction (indicated by black arrows). Asterisks indicate heavy chains detected by secondary antibodies. Abbreviations used: IP (immunoprecipitation), IB (immunoblotting), Sup (supernatant). Similar results were obtained in n = 3 transfections

KV8 modulate KV7 currents by affecting both membrane trafficking and biophysical properties

We then sought to determine whether KVS modulated KV7 currents through mechanisms similar to those involved in the regulation of KV2 channels. To this end, we measured the plasma membrane expression of both KV7.2 and KV7.4 channels containing an extracellular HA tag, in Xenopus laevis oocytes using a luminometric assay [34]. This method involves the oxidation of luminol by horseradish peroxidase (HRP) in conjunction with antibodies that can detect HA-tagged KV7 channels located on the oocyte surface. The intensity of light emitted directly correlates with the membrane expression of KV7 channels (Fig. 4A). In these experiments, co-expression of KV8.1 significantly increased the membrane expression of both KV7.2 and KV7.4 channels, whereas co-expression of KV8.2 slightly decreased the membrane expression of both channels (Fig. 4B, C), fully in line with the observed effects on current density (c.f., Fig. 1). Taken together, these data indicated that KV8.1 enhances KV7 currents by increasing their membrane expression, whereas co-expression of KV8.2 may reduce current amplitudes, at least in part, by attenuating surface expression of KV7 channels. Therefore, we conclude that KVS modulates KV7 currents by altering their membrane abundance, similar to the mechanisms reported for KV2.1 channels, but, unlike KV2 channels, in a bidirectional manner [5, 8].

Fig. 4

KV7 and KVS form complexes with altered membrane trafficking. A Surface expression of HA-tagged KV7 channels measured with a luminometric technique in Xenopus oocytes. Mean surface expression of B HA-tagged KV7.2 channels C HA-tagged KV7.4 channels (measured in relative light units [RLUs]) in Xenopus oocytes after injection of HA-tagged KV7 cRNA either alone or together with KV8.1 or KV8.2. Uninjected oocytes were used as negative control

We then set out to gain insight into whether KVS subunits modulate KV7 channels via heterotetramerization into the same channel complex or by close interaction, e.g., as associated but independent channel entities or as some sort of ancillary beta subunit (Fig. 5A). To determine whether all isoforms contribute to the channel pore (i.e. heterotetramerization), we generated channels containing mutations in the GYG pore motif (GYG/AAA exchange for KV8.1 and GYG/GYS exchange for KV7.2) that render these variants inactive. When co-expressed with wild-type subunits, such variants are known to attenuate whole cell currents via dominant-negative effects [26]. Assuming fully stochastic co-assembly of equally available subunits, co-expression of wild-type and mutant subunits is predicted to reduce whole cell current amplitudes to 1/16 (compared to cells expressing only wild-type subunits), leaving intact a minimal number of channels containing only wild-type subunits (illustrated in Fig. 5A, first row). We performed patch-clamp recordings on CHO cells expressing wild-type and mutant isoforms taking current densities as measure for potential co-assembly of the subunits. In these experiments we used co-transfected KV8.1 and KV7.2 variants at a ratio of 4:1 (KV8.1:KV7.2) to ensure an excess of KV8.1. Co-expression of wild-type KV8.1 with wild-type KV7.2 significantly increased whole cell current amplitudes compared to cells expressing only KV7.2 channels, as we had observed before (Fig. 5B). When we co-expressed pore-mutated KV7.2(GYS) with wild-type KV8.1, current amplitudes were reduced to virtually zero. These experiments demonstrated that KV8.1 subunits alone were not able to form an independent functional pore (Fig. 5B).

Fig. 5

KV7 and KVS might form heteromers with a single conducting pore. A Imaginable modes of interaction of KV7 with KVS, consequences of expression of pore mutant subunits and expected membrane conductance. *) The percentage of current reduction depends on the probability of inclusion of the pore-mutant subunit into a heterotetramer. Detailed explanation is given in the main text. B Patch-clamp recordings from CHO cells co-transfected with either combination of wild-type (GYG) and pore-mutated (AAA, GYS, resp.) KV8.1 and KV7.2 or with KV7.2 alone. Top panel shows stimulus protocol (black) and exemplary recordings. Bottom panel shows summary statistics for steady-state current densities at + 20 mV. C Patch-clamp recordings from CHO analogue to those shown in B, but examining the Kv8.1 pore-mutants GYR and W388G. For the experiments shown in B and C KV7.2 and KV8.1 plasmids were co-transfected at a mass ratio of 1:4. Where no KV8.1 was included into the transfection this was substituted by eGFP. D Surface quantification measurements from Xenopus oocytes similar to those shown in Fig. 4B, but using KV8.1 pore-mutants AAA and GYR variants. E Patch-clamp recordings from CHO analogue to those shown in B but performed using a 1:1 mass ratio for transfection. Electrophysiological data are shown as relative current densities normalized to the mean of current densities observed when transfecting wild-type KV7.2 with wild type KV8.1

When wild type KV7.2 was transfected together with pore-mutated KV8.1(AAA), current densities were reduced to approximately 50% of those observed when measuring cells transfected with both wild type constructs, resulting in current densities minimally smaller than those observed in cells transfected with KV7.2 alone (Fig. 5B, note that amount of cDNA encoding KV7.2 was kept constant). These results suggested that functional pore regions of KV8.1 are essential for KVS-dependent modulation of KV7 channels. Noteworthy, the (only) 50% current reduction observed upon KV8.1(AAA) co-transfection may indicate that during co-assembly KV7.2 homotetramer formation is more likely to occur than heterotetramer formation. In particular, the 50% reduction observed herein would be expected if the probability for KV7.2 homomers was approximately fivefold higher than that for KV7.2/KV8.1 heteromers. Yet, our observations clearly rule out alternative modes of interaction: If KV7.2 and KV8.1 co-existed as independent channels (Fig. 5A, second row), a substantial current should have been observable upon co-expression of pore-mutated KV7.2(GYS); If KV8.1was a beta-subunit to KV7.2 (Fig. 5A, third row), mutating the “pore” sequence of KV8.1 would not be expected to have any effect at all.

Notably, the GYG to AAA is a relatively aggressive way of disrupting a KV-channel’s pore, potentially causing the protein to misfold and become unstable. If this were the case, there would be no complex formation between KV7.2 channels and KV8.1(AAA) subunits and there would be no increased surface expression of KV7.2, which could explain the 50% reduction in current, seen with KV8.1(AAA) co-transfection. To rule this out, we first tested other, more conservative variants known to render KVS non, conducting (Fig. 5C). These were G399R, corresponding to the second G in the GYG motif, and W388G, located in the S4 segment. These variants were deduced from KV8.2, where they are naturally occurring pathogenic variants that have been demonstrated to render KV2.1-KV8.2 heterotetramers nonconducting without affecting surface expression [40]. We found that G399R and W388G reduced current amplitudes to 54.05 (± 4.96) % and 44.13 (± 20.27) %, respectively (p < 0.05 for both), and thus even minimally stronger than the AAA mutant. We next tested surface expression for the AAA and G399R variants using Xenopus laevis oocyte system and found that both the variants behave similar to the wild-type KV8.1, indicating no misfolding or stability issues (Fig. 5D). Finally, in CHO cells, we confirmed that also the 4:1 (KV8.1:KV7.2) transfection ratio utilized in the experiments from Fig. 5B, C did not affect our observations. Indeed, results observed with a 1:1 transfection ratio were similar to those obtained with a 4:1 ratio: KV7.2/KV8.1(AAA) co-transfection reduced current amplitudes to 49.4 (± 10) % as compared to co-transfection of wild-type channels (p < 0.05, Fig. 5E).

Taken together, these data suggest that KV7 and KV8.1 subunits can likely heteromerise into functional channels with a slight preference for KV7 homomers.

KV7 and KVS are expressed in the same organs and cells

To investigate whether an interaction between KV7 and KVS might be physiologically relevant, we used reverse transcription quantitative polymerase chain reaction (RT-qPCR) to examine the expression of KV8 in three tissues, where the role of KV7 channels is well established [46]. Indeed, we found that in the hippocampus and dorsal root ganglia, the abundance of Kcnv1 (KV8.1) mRNA was in the same order of magnitude as that of Kcnq2 (KV7.2) mRNA. In both these neural tissues, also Kcnv2 (KV8.2) mRNA was detected, though at much lower levels (Fig. 6A). By contrast, in heart, expression levels of Kcnv2 (KV8.2) were comparable to those of Kcnq1 (KV7.1), while Kcnv1 mRNA was not detected (Fig. 6B).

Fig. 6

KV7 and KVS co-express in various tissues in a cell-specific manner. A, B Reverse transcription quantitative PCR for KV7 and KVS transcripts from neuronal (A) and cardiac (B) tissue. For each condition 8 biological replicates with two technical replicates were included into the analysis. Shown are expression levels normalized to the mean expression level of KV7.2 for Hippocampus and Dorsal Root Ganglia (DRG) (A) and KV7.1 for Heart (B), respectively. C Expression dot plots of KV8.1, KV7 and KV2 genes, for comparison, in Hippocampus. Shown are only clusters with KV8.1 expression. The complete dot plot is given in Supplementary Fig. 5. D–G RNAScope of KV8.1 in the mouse hippocampus. D Low-magnification overview. Scale bar: 500 μm. E–G High-magnification confocal micrographs of CA 1 (E), CA 2 (F) regions and Dentate Gyrus (G), respectively. Scale bar: 100 μm. H Correlation coefficients for the single-cell expression of KV7 and KV2 with KV8.1 or KV8.2, respectively. Shown are Pearson’s correlation coefficients for log2 + 1-transformed transcript counts from individual cells as observed in three publicly available singe-cell RNA sequencing datasets. Included into the analysis were only cell types found to express the respective KVS

To further evaluate whether KV7 and KVS transcripts merely coexist in the same tissue, or are actually transcribed in the same cells we analysed publicly available single-cell RNA sequencing (scRNAseq) datasets from the hippocampus [36], dorsal root ganglia (DRG) [13] and the heart [42]. Consistent with our observations in the qPCR experiments we found high read counts for Kcnq2 and Kcnv1 in hippocampal and DRG cells, whereas Kcnv2 was barely detected. In the dataset from cardiac cells, in turn, Kcnq1 and Kcnv2 were predominant, whereas Kv8.1/Kcnv1 was not found (Supplement material file 1, Table S2).

In the hippocampus, Kcnv1 was encountered in seven clusters, with the highest transcript counts being observed in the clusters representing CA1-3 pyramidal neurons, dentate hilus and dentate principal neurons (Fig. 6C). To see if the observed transcriptomic expression pattern would also be supported by methods that provide spatially encoded information, we employed single molecule fluorescence RNA in-situ hybridization (“RNAScope”). Indeed, using RNAScope we found Kcnv1 expression patterns closely matching the observations made using scRNAseq (Fig. 6D–G). By far the highest Kcnv1 signal was obtained from the CA2 region, followed by the dentate gyrus and the CA1 region. In scRNAseq, both KV7.2/Kcnq2 and KV2.1/Kcnb1 were both found in each of these cell types /clusters (Fig. 6H, Supplementary Fig. 5A). We argued that if KV8.1 would interact with KV7.2 rather than a KV2 in hippocampal neurons, one would expect Kcnv1 expression levels to correlate stronger with those of Kcnq2 than with those of Kcnb1 or Kcnb2 at the single cell level. Indeed, Pearson’s correlation coefficient (PCC) for Kcnv1 with Kcnq2 was higher than the correlation of Kcnv1 with 99 [98–99] % of all other transcripts encountered, and in particular higher than that with Kcnb1 (Fig. 6H, left, Supplement material file 1, Table S3).

In dorsal root ganglia, Kcnv1 and Kcnq2 overlapped in two clusters (Supplementary Fig. 5B). In stark contrast to the hippocampus, Kcnv1 expression levels in DRG were more strongly correlated with those of KV2 channel genes rather than with those of KV7 channel genes (Fig. 6H, middle, Supplement material file 1, Table S3). In the heart, Kcnv2 was found exclusively in the cluster representing cardiac myocytes, as was Kcnq1 (Supplementary Fig. 5 C). Here, the PCC for Kcnv2 with Kcnq1 was higher than the correlation of Kcnv2 with 0.99 [0.98–1.00] % of all transcripts encountered, and, again, particularly higher than that with Kcnb1 (Fig. 6H, right, Supplement material file 1, Table S3). Taken together, these transcriptomic analyses suggest that KV7 and KV8 mRNA expression is strongly correlated within individual cells in hippocampus and heart, whereas this was not the case in the DRG.