Although Cl- is the most abundant anion in living cells, chloride currents and their functional significance had been understudied until the CLC family of chloride channels and CFTR (cystic fibrosis transmembrane conductance regulator) were cloned and their dysfunctions were linked to human diseases 1-4. In addition to those on the cell surface, Cl- channels have long been proposed to exist in the intracellular membrane-bound organelles 3,5. However, the previously postulated intracellular Cl- channels, like CLCAs (chloride channel Ca2+-activated) and CLICs (chloride intracellular channels), are now considered not likely to function as anion channels 6,7. Therefore, the molecular identities and functions of organellar anion channels, including those in the SR/ER, remain largely unknown.
As the major internal Ca2+ store, Ca2+ release from SR/ER is mediated by two cation channels, RyRs (ryanodine receptors) and IP3Rs (inositol 1,4,5-trisphosphate receptors) 8-10. During the release, SR/ER membrane is charged upon the Ca2+ efflux, that hinders the continued Ca2+ release. Previously reported TRICs (TRimeric Intracellular Cation channels) acts as counter-ion channel to balance the loss of positive charges from the SR/ER as a result of the release 11. In addition to cations, anions have also been proposed to function as counter-ion for the release, and various Cl- channel activities have been long demonstrated in microsome preparations 12-17. A previous study using mouse forward genetics revealed that loss of CLCC1 (Chloride Channel CLIC Like 1), an ER-resident protein 18,19, leads to ER stress and neurodegeneration 19. However, despite the name, CLCC1 has little sequence similarity with CLIC family members or any known ion channels. In addition, question remains whether the recorded chloride currents in microsome prepared from the CLCC1 overexpressing cells were actually mediated by CLCC1 20,21. Therefore, further evidence is needed to know if CLCC1 functions as an anion channel.
Here, we demonstrate that CLCC1 is a pore-forming component of an ER anion channel by incorporating purified CLCC1 into lipid bilayer. Depletion of CLCC1 reduces ER Ca2+ release, probably through a counter-ion mechanism, but increases steady-state [Cl-]ER and [K+]ER. We identified CLCC1 rare variants in a Chinese ALS cohort and the disease-associated non-synonymous mutations impair CLCC1 channel conductance and promote misfolded protein accumulation in the mutation knockin mouse brain and spinal cord. Conditional removal of Clcc1 in ChAT-positive motor neuron cell-autonomously leads to ubiquitin-positive and mislocalized TDP-43, a pathological hallmark of ALS, and motor neuron loss. Therefore, we argue that misregulation of ER ion homeostasis maintained by an ER anion channel underlies ER unfolded protein response (UPR) and etiology of neurodegenerative diseases.
CLCC1 forms homomultimer in the ER membrane
Based on its primary sequence, CLCC1 shares little sequence similarity with any known ion channel but is predicted to contain three transmembrane segments (TMs) and a N-terminal signal peptide (Fig. 1a). We generated antibodies against the N- and C-termini of CLCC1 (Extended Data Fig. 1a). Using the C-terminal antibody, we confirm that as suggested by a previous report 18,19 CLCC1 is predominantly ER-localized, as demonstrated by its co-localization with CALNEXIN, an ER-resident protein (Extended Data Fig. 1b).
To understand how CLCC1 functions in the ER, we treated human 293FT cells with disuccinimidyl suberate (DSS), a crosslinker with a spacer length of 11.4 Å. The C-terminal antibody detected high molecular weight complexes in a DSS dosage-dependent manner from whole cell lysate. From the complex sizes, we speculated that CLCC1 forms homomultimers (Fig. 1b), which was supported by co-immunoprecipitations of differentially tagged CLCC1 co-expressed in the same cells (Extended Data Fig. 2a) and of exogenous tagged CLCC1 with endogenous CLCC1 (Extended Data Fig. 2b). Consistent with our cell culture data (Fig. 1b) and a previous report 20, our native gel experiments suggested complex formations (~180kD and ~360kD) of purified full-length CLCC1 and that (~360kD, the major band) of endogenous CLCC1 (Extended Data Fig. 2c). In addition, the purified full-length mouse CLCC1 (mCLCC1) gave a major high molecular weight peak by chromatographic column separation (Fig. 1c). Taken together, our data suggest that CLCC1 forms homomultimer in the ER membrane.
CLCC1 is a pore-forming component of an anion channel
Incorporation of the purified full-length mCLCC1 (Fig. 1c) into planar lipid bilayer resulted in frequent inward currents at 0 mV (-2.2 ± 0.1 pA) in asymmetric KCl solutions (In/Ex, 150/15 mM) and the currents became outward at 90 mV (1.6 ± 0.1 pA) (Fig. 1d). As a negative control, the protein purification buffer without protein gave rise to no current (Fig. 1d). Based on the fit of the current-voltage relationship, the reversal potentials were determined to be 56.8 mV (In/Ex, 150/15 mM KCl) and -60.3 mV (In/Ex, 15/150 mM KCl), which are close to the calculated values for Cl- by Nernst equation, and the slope conductance was 39.9 ± 1.0 pS (mean ± SEM). The permeability ratio of P(Cl-) to P(K+) is about 100 to 1 and similar results were obtained by using asymmetric NaCl solutions (Fig. 1d). Consistent with the single channel results, the reversal potential obtained from studying macroscopic currents was 61.9 mV in 15/150 mM KCl (In/Ex) (Fig. 1e), further supporting the anion selectivity. Both square step-like multiple-channel currents (Extended Data Fig. 3) and a large-scale current (80-90 pA) (Fig. 1e) suggested that CLCC1 is capable to mediate multi-channel currents.
Next, we examined CLCC1 channel permeability to various anions, including Br-, NO3-, and F-, by adding 150 mM KCl in cis (In) and equal electric charges of KBr, KNO3, or KF in the trans (Ex) chamber (Fig. 1f). The relative permeabilities of these anions to Cl- were 17.14 (PBr/PCl), 0.22 (PNO3/PCl), and 0.18 (PF/PCl), respectively, indicating a sequence of the CLCC1 anion selectivity of PBr > PCl > PNO3 > PF, which is similar to that of CFTR 22,23. In these experiments, no cation permeation was detected. Collectively, our results demonstrate that CLCC1 is a pore-forming component of an anion channel.
ER membrane topology of CLCC1 and its inhibition by luminal calcium
To examine CLCC1 topology in the ER membrane, we treated microsomes prepared from wildtype mouse cerebella and livers 24 with trypsin and analyzed the remaining CLCC1 fragments with our N- and C-terminal antibodies. In the absence of Triton X-100, the N-terminus and the first and second loops of CLCC1 and an ER lumen resident protein Bip were protected from trypsinization, but the C-terminus of CLCC1 was not (Fig. 1g and Extended Data Fig. 1c). As expected for membrane enclosure, the protection was disrupted by Triton X-100, suggesting that CLCC1 N-terminus and the second loop reside in ER lumen while C-terminus faces cytoplasm.
Interestingly, when we applied MTSET (methanethiosulfonate-ethyltrimethylammonium) 25, a membrane-impermeant thiol reagent that modifies cysteine residues, in trans but not to the cis side of the chamber we applied the purified CLCC1, the CLCC1 currents were suppressed (Fig. 1h), suggesting that a specific orientation of CLCC1 in the bilayer is responsible for the current. Based on the topology (Fig. 1g), cysteine residues are located in both the cytoplasm and ER lumen sides of CLCC1 and C350 lies at the end of TM3 (Fig. 1i). Protein alignment among different species revealed that C350 is in a consecutive row of four residues (FCYG), although it is less conserved than the other three surrounding resides (Fig. 1j). Instead of FCYG in Homo sapiens and Mus musculus, FFYG appears in Xenopus tropicalis, which prompted us to mutate C350 to F. C350F mCLCC1 is expressed and its chromatographic behavior is similar to wildtype mCLCC1 (Extended Data Fig. 4a). Importantly, C350F restored the CLCC1 currents even when MTSET was applied in trans side (Fig. 1h), suggesting that MTSET acts on C350 to modify the channel activity and the trans side is the CLCC1 cytoplasm side in the reconstructed lipid bilayer. Application of DIDS (4,4'-Diisothiocyano-2,2'-stilbenedisulfonic acid), a chloride transporter/channel blocker 4,26, significantly inhibited CLCC1 channel activity (Extended Data Fig. 4b and 4c). Consistent with MTSET acting on C350, C350F largely restores the DIDS inhibition on channel open probability (Po).
Because ER luminal Ca2+ is much higher than cytoplasm, we then asked whether Ca2+ is able to differentially regulate CLCC1 channel activity from ER luminal or cytoplasmic side. Application of Ca2+ in cis/ER lumen side blocked the CLCC1 channel activity, which could be partially rescued by addition of equal molar EGTA, a Ca2+ chelating agent (Fig. 1k and 1l). However, the same application in trans/cytoplasm side had no effect on the channel activity. Therefore, we conclude that, at least in our reconstructed lipid bilayer setting, high concentration of Ca2+ at the ER lumen side inhibits CLCC1 channel activity.
D25/D181 are the key residues responsible for Ca2+-dependent inhibition on the channel open probability
To identify key residues at lumen side responsible for the Ca2+ inhibition, we aligned the protein sequences across different species (Fig. 1m). Several conserved negatively charged residues at N-terminus drew our attention, including D25, D152, D153, E175, D176, and D181. Because CLCC1 D25E mutation has been associated with retinitis pigmentosa 21, we took D25E into account. In addition, we generated D152R/D153R, E175R/D176R, D181R, and D25E/D181R mutant CLCC1 for Ca2+-binding affinity and channel activity measurements.
Because C-terminus of CLCC1 resides at cytoplasmic/trans side, we truncated the C-terminus (1-365) and recorded a sound current that could be inhibited by Ca2+ (10mM) application at cis side (Fig. 1n). In addition, all mutant CLCC1 yielded reasonable currents as wildtype (1-365) did at 0 mV in absence of Ca2+ (Fig. 1o). However, the mutant CLCC1, including D25E, D181R, and D25E/D181R but not D152R/D153R, largely restored the currents in presence of 10mM Ca2+ at cis side (Fig. 1n), suggesting that D25 and D181 are the key residues for the calcium inhibition on CLCC1 channel activity. Our Ca2+ dose response curve for the inhibition and the calculated IC50 (mM) (for wildtype, 1.09 ± 0.15; D25E, 6.08 ± 1.62; D181R, 8.51 ± 1.04; D25E/D181R, 19.20 ± 4.24) also supported the notion that D25 and D181 are the key residues (Fig. 1p). The IC50 for D152R/D153R (1.06 ± 0.19) and E175R/D176R (2.83 ± 0.58) suggested that these negatively charged residues are either not involved in or less important than D25 and D181 for the inhibition.
To determine the Ca2+-binding affinity of the mutant CLCC1, we performed surface plasmon resonance (SPR)-based binding assay (Fig. 1q and 1r). Compared to that of the wildtype (1-365, 7.99 ± 0.62 mM), the KD values of the mutant CLCC1, including D25E (10.47 ± 0.27 mM), E175R/D176R (11.29 ± 0.29 mM), D181R (12.14 ± 0.29 mM), and D25E/D181R (22.94 ± 4.10 mM), were significantly increased, suggesting that these mutant CLCC1 significantly impair Ca2+-binding affinity. Consistent with the D25E/D181R mutant with the strongest effect on relieving the Ca2+ inhibition, it also affects the most of Ca2+-binding affinity (Fig. 1s). Therefore, we conclude that D25/D181 are the key residues responsible for Ca2+-dependent inhibition of the channel open probability.
CLCC1 maintains steady-state [Cl-]ER, [K+]ER, and ER morphology
To examine whether CLCC1 is involved in regulation of [Cl-]ER, we employed a previously optimized YFP Cl- sensor that responds to Cl- concentration change with super sensitivity and photostability 27. To create a ratiometric ER Cl- sensor, we built a signal sequence, a DsRed internal control, and an ER retention motif into the Cl- sensor, which we named RaMorideER (Fig. 2a). ER localization of RaMorideER was confirmed by its colocalization with ER-resident protein CALNEXIN (Fig. 2b). The ratio between YFP to Ds-Red signals responded correspondingly when extracellular [Cl-] ([Cl-]Extra) was switched from 140mM to 100 or 0 mM (Extended Data Fig. 5a and 5b).
Consistent with the essential role of CLCC1 in vivo, we failed to generate a CLCC1 KO 293FT cell line by Crispr/Cas9. Instead, we knocked down CLCC1 with two individual shRNAs (H3 and H4) (Fig. 2c-2e). Although the two shRNAs had different CLCC1 knockdown efficiencies (for H3, 22.5 ± 0.6% of scrambled control; for H4, 45.25 ± 2.1% of scrambled control), both of them significantly increased steady-state [Cl-]ER to a similar extent in comparison with scrambled shRNA control (Fig. 2f and 2g).
Given that the influx of K+ into ER lumen acts as a counter-ion mechanism during the internal Ca2+ release 11, we next measured steady-state [K+]ER upon knockdown of CLCC1. To achieve it, we employed a newly reported K+ sensor (doi.org/10.1101/2021.10.07.463355) and generated a ratiometric ER K+ sensor (RaMssiumER) (Fig. 2h), as we did for RaMorideER. The RaMssiumER responded well to a series of titrated [K+] solutions in vitro (Fig. 2h) and a K+ ionophore, valinomycin (Extended Data Fig. 5c). With the RaMssiumER probe, we detected significantly higher steady-state [K+]ER in H3 and H4 groups, compared to the scrambled control (Fig. 2i).
The concentration of electrically charged osmolytes, such as Cl-, inside a cell or intracellular membrane-bound organelle governs the volume of the compartment 5,28. Therefore, we asked whether depletion of CLCC1 changes ER volume. To this end, we collected 293FT cells expressing scrambled control or CLCC1 shRNAs and applied for transmission electron microscopy (TEM) (Fig. 2j and 2k). Enlarged and stubby ER morphology was documented in cells expressing the individual CLCC1 shRNA. In contrast, ribosome-bound and tubule-like ER was shown in the scrambled controls. In order to quantitatively reflect ER morphology, we measured ER width in these three groups of cells. ER width in the two individual CLCC1 shRNA groups was significantly increased as compared to scrambled shRNA control. Therefore, we conclude that CLCC1 loss alters steady-state ER ion homeostasis and leads to enlarged ER.
CLCC1 facilitates internal Ca2+ release
ER-localized ion channels have been proposed to control ER Ca2+ mobilization through a counter-ion mechanism 13,14,16,17. We then asked whether as an ER chloride channel CLCC1 is involved in regulation of ER Ca2+ release. Knockdown of CLCC1 by the two individual shRNAs markedly reduced internal Ca2+ release induced by ATP (adenosine triphosphate) (Fig. 3a), which triggers ER Ca2+ release by generating IP3 that activates IP3Rs 10. Compared to mock control and scrambled shRNA, knockdown of CLCC1 by the two individual shRNAs not only significantly reduced the amplitude, but also the rate (as reflected by the increase in time-to-peak), of ATP-induced Ca2+ release (Fig. 3b and 3c). Although the two shRNAs had different CLCC1 knockdown efficiencies, they impaired the ATP-induced Ca2+ amplitude and rate to a similar extent.
Analysis of the Ca2+ release dynamics in individual cells revealed that CLCC1 knockdown impaired ATP-induced Ca2+ oscillation (Extended Data Fig. 6a and 6b). Whereas less than 4.2 ± 0.6% of cells exhibited one ATP-induced Ca2+ spike in scrambled shRNA group, the proportion was significantly more in the CLCC1 knockdown groups (31.5 ± 2.3% for H3; 36.5 ± 4.9% for H4). However, the Ca2+ spike number greater than or equal to 3 was reduced in the CLCC1 knockdown groups compared to that of scrambled control (Extended Data Fig. 6b). The impairment of ATP-induced Ca2+ release seems not to be caused by shRNA off-target effects, because the reexpression of full-length (WT) mCLCC1 restored the release damaged by H3 shRNA alone (Extended Data Fig. 7a and 7b). In contrast, expression of mutant mCLCC1 lacking the ER lumen resident 2nd loop (Δ2nd loop) did not, suggesting that the 2nd loop of CLCC1 is crucial for its functions.
Next, we asked whether CLCC1 regulates internal Ca2+ release through RyRs, the predominant intracellular Ca2+ channels expressed in cardiomyocytes 10 . To this end, we stimulated cardiomyocytes cultured from wildtype (+/+) and NM2453 mutant (NM/NM) mice with caffeine, an agonist for RyR-mediated Ca2+ release. RyR-mediated Ca2+ release was significantly reduced in NM/NM cardiomyocytes as compared to +/+ controls (Extended Data Fig. 6c), demonstrating that CLCC1 facilitates ER Ca2+ efflux through regulation of the release process per se rather than regulation of a particular type of Ca2+ release channels.
CLCC1 dosage is crucial for maintenance of steady-state [Ca2+]ER level
To examine whether the impaired Ca2+ release upon CLCC1 knockdown results from a reduction in ER Ca2+ load, we depleted the ER Ca2+ store with cyclopiazonic acid (CPA), an inhibitor of sarco/endoplasmic reticulum Ca2+-ATPase (SERCA) 29. Knockdown of CLCC1 by H3 but not H4 shRNA significantly reduced CPA-sensitive cytosolic Ca2+ rise (Fig. 3g and 3h), suggesting that impairment of ER Ca2+ content depends on CLCC1 dosage as H3 has higher knockdown efficiency than H4 shRNA (Fig. 2e).
Given that depletion of CLCC1 increases ER volume (Fig. 2j and 2k), we next asked whether [Ca2+]ER is also impaired. We employed a previously reported low affinity Ca2+ probe, ER-GCaMP6-210 30, which correctly responded to CPA-induced internal Ca2+ depletion and follow-up ionomycin-mediated extracellular Ca2+ replenish (Extended Data Fig. 8a). Compared to mock and scrambled shRNA controls, knockdown of CLCC1 by both H3 and H4 shRNAs significantly decreased steady-state [Ca2+]ER level in cells expressing ER-GCaMP6-210 (Fig. 3g and 3h). The impairment caused by H3 shRNA was more severe than that by H4 shRNA, suggesting that depletion of CLCC1 decreases steady-state [Ca2+]ER level in a dosage-dependent manner.
To measure steady-state [Ca2+]ER in the diseased neurons, we cultured cerebellar granule neurons (CGNs) from wildtype (+/+) and NM/NM mutant mice (Fig. 3i and 3j). We employed previously described low affinity Ca2+ probe, ER-GCaMP6-150 30, and generated a ratiometric probe for [Ca2+]ER measurement by fusing ER-GCaMP6-150 with DsRed, which responded well to CPA-induced internal Ca2+ depletion (Extended Data Fig. 8b). After infecting the cultured CGNs with the probe, we detected significant decrease of [Ca2+]ER in the mutant CGNs compared to that of wildtype.
In summary (Fig. 3k and 3l), the knockdown of CLCC1 dosage-dependently increases steady-state [K+]ER and ER volume and decreases steady-state [Ca2+]ER. Instead, the knockdown increases steady-state [Cl-]ER and reduces internal Ca2+ release in a CLCC1 dosage-independent manner, suggesting that steady-state [Cl-]ER and internal Ca2+ release are primarily affected by CLCC1 knockdown.
A conserved lysine (K298) is responsible for PIP2 facilitation of CLCC1 channel activity
As a necessary cofactor of many ion channels, PIP2, an acidic phospholipid of the cell membrane, has been implicated in the regulation of ion channel functions, including of intercellular cation channels 31-33. To examine whether PIP2 affects CLCC1 channel activity, we included 2% PIP2 in the planar phospholipid bilayer. Interestingly, PIP2 significantly increased the slope conductance (80.1 ± 2.5 pS) and the open probability (Po) of wildtype mCLCC1 (Fig. 4a and 4b). Given that PIP2 regulates ion channels by binding to certain positively charged residues in the channel protein 32,33, we looked for positively charged residue(s) in CLCC1 and a positively charged lysine (K298) drew our attention (Fig. 4c). It lies in a consecutive row of six conserved residues-VPPTKA in the 2nd loop, which is required for CLCC1 facilitation of internal Ca2+ release (Extended Data Fig. 7). In addition, K298 is downstream of two proline residues, which usually present strong conformational rigidity, and lies at the beginning of a predicted alpha-helix.
We expressed and purified K298A mutant mCLCC1 and incorporated it into the lipid bilayer in the absence of PIP2, the mutant protein exhibited single channel activity with a slope conductance of 31.8 ± 0.7 pS, slightly lower than that of wildtype mCLCC1 (39.9 ± 1.0 pS) (Fig. 4a, 4d, and 4f). The Po at 0 mV did not differ from that of wildtype mCLCC1 (Fig. 4a and 4f). Next, we mutated K298 to the negatively charged residue glutamate (K298E). Like K298A, K298E also has little effect on the channel activity in absence of PIP2 (Fig. 4f). However, unlike wildtype mCLCC1 responsible to PIP2 (Fig. 4b and 4e), both K298A and K298E mutants abolished the responses, in terms of conductance and Po (Fig. 4e and 4f). Therefore, we conclude that PIP2 facilitates CLCC1 channel activity and a conserved K298 in the 2nd loop is responsible for the facilitation.
K298 is crucial for CLCC1 regulation of internal Ca2+ release
If K298 is functionally important for CLCC1 channel activity, we wondered whether K298 is equally important for internal Ca2+ release. To examine this, we employed a lentiviral inducible system to stably express wildtype and K298A mutant mCLCC1 in 293FT cells in a controllable manner (Extended Data Fig. 9a). Expression of exogenous mCLCC1 proteins was induced after application of doxycycline (Dox) (Extended Data Fig. 9b). Both the exogenous wildtype and K298A mutant mCLCC1 interacted with the endogenous hCLCC1 (Extended Data Fig. 9c), as shown by co-immunoprecipitation, supporting complex formation by exogenous mCLCC1 and endogenous hCLCC1 (Extended Data Fig. 2b). Induction of wildtype mCLCC1 did not alter the amplitude and rate of ATP-induced Ca2+ release (Fig. 4g-4i). However, expression of K298A mutant mCLCC1 significantly suppressed such activities, as shown by the reduction in both the amplitude and rate when compared to un-induced (minus Dox) cells or cells induced to express wildtype mCLCC1. In addition, induction of K298A mutant mCLCC1 expression, but not wildtype mCLCC1, decreased the number of ATP-induced Ca2+ oscillation (Extended Data Fig. 9d and 9e). These findings are all similar to that found in CLCC1-knockdown cells (Fig. 3a-3c and Extended Data Fig. 6a and 6b), suggesting a dominant-negative effect of the mutant protein in CLCC1 channel function. Taken together, our findings reveal that a conserved K298 in the 2nd loop is functionally important for CLCC1 to regulate the internal Ca2+ release.
K298A mutation promotes motor neuron loss and enlarges ER volume in vivo
To examine the in vivo effect of the conserved K298 residue, as it is critical for PIP2 facilitation on CLCC1 channel activity and internal Ca2+ release, we generated K298A knock-in mouse (Extended Data Fig. 10a and 10b). Although expression of K298A mutant mRNA and protein was confirmed by Sanger sequencing and mass spectrometry (Extended Data Fig. 10c-10e), the expression level of K298A mutant protein was as low as that of the NM2453 allele (Fig. 4j). Like Clcc1 KO (Extended Data Fig. 11a and 11b), we failed to produce mouse homozygous for K298A (Extended Data Fig. 11c), indicating that K298 is a key residue for its essential function in vivo.
Compound heterozygotes with the NM2453 and K298A mutations (NM/K298A) were viable but displayed severe body weight loss, hind leg weakness, trunk shaking, tail flagging, abnormal gaits, and ataxia phenotypes as early as 3 months of age (Extended Data Movie 1). The onset (~3 months of age) of these phenotypes is much earlier than that shown in the NM/NM mice (> 12 month of age) 19, but slower than that of KO/NM mice (Extended Data Movie 2), indicating that K298A is a partial loss-of-function allele. Like NM/NM mice, the NM/K298A compound heterozygotes displayed ER stress (Fig. 4k) and neuron degeneration in cerebellar granule neurons (Extended Data Fig. 10f). ER stress was also evidenced in hippocampal granule neurons in the compound heterozygotes but not in NM/NM mice 19 (Extended Data Fig. 10g). The severe motor impairment and hind leg muscle weakness prompted us to examine motor neuron pathologies in these compound heterozygotes mice. Indeed, ubiquitin-positive inclusions in ChAT-positive motor neurons and their number loss, two key ALS pathologies, were evidenced in the mutant spinal cords (Fig. 4l and 4m).
As knockdown of CLCC1 impairs ER ion homeostasis and leads to ER swelling (Fig. 2), we next asked whether dysfunction of CLCC1 impairs ER morphology in vivo. To this end, we examined the cerebella from wildtype and K298A/NM mice by TEM. Instead of ribosome-bound and tubule-like ER morphologies observed in wildtype cerebellar granule neurons, the mutant neurons harbored enlarged, stubby, and less ribosome-bound ER (Fig. 4n). Indeed, the ER width of mutant granule neurons was significantly increased compared to that of wildtype (Fig. 4o). Taken together, our findings demonstrate that disruption of channel function by the K298A promotes ER stress and motor neuron loss and enlarges ER volume in the diseased neuron in vivo.
Rare genetic variants in CLCC1 found in a Chinese ALS cohort
As lower motor neuron loss and its ubiquitin-positive inclusion are the key pathological features shown in ALS 34, we next asked whether the dysfunction of CLCC1 is relevant to the motor neuron diseases. To this end, we performed whole exome sequencing in a Chinese cohort (670 sporadic ALS patients and 1910 controls) and identified 8 rare variants in CLCC1 in the patients, including 6 nonsynonymous and 2 stopgain mutations (Fig. 5a, Extended Data Fig. 12, and Extended Data Table 1). Among the mutations, the S263R and W267R mutations have not been found in the public databases nor in our controls (Extended Data Table 1). No mutations in known ALS-causing genes were detected in the patients carrying S263R or W267R mutation. Notably, two geographically and genetically unrelated patients with similar clinical phenotypes shared the same S263R mutation (Extended Data Table 1 and Table 2). Both mutations change Ser and Try to Arg, suggesting that they perturb local steric hindrance and surface potential. A burden analysis 35 was further carried out and revealed that CLCC1 is associated with ALS (p = 1.51×10-6, with OR = 5.72), reaching suggestive significance (Fig. 5b).
ALS-associated rare variants impair the channel activity and promote ER stress and protein misfolding in vivo
Evolutionarily, CLCC1 orthologues appear in vertebrate bot not invertebrate and S263 and W267 are conserved across species (Fig. 5c). In addition, three-amino acid apart between S263 and W267 in a predict alpha helix suggests these two residues are structural proximity and functional synergy in channel function and activity. To examine whether S263R and W267R alter CLCC1 channel activity, we incorporated purified human wildtype (hWT), S263R, or W267R mutant CLCC1 proteins into the lipid bilayer. The slope conductance of both S263R and W267R was significantly lower than that of hWT in absence and presence of PIP2, respectively (Fig. 5d-5f). However, Po of S263R and W267R at 0 mV did not differ from that of hWT (Fig. 5g). Like hWT, S263R and W267R mutant CLCC1 were in response to PIP2, since PIP2 significantly increased the slope conductance and Po of S263R and W267R (Fig. 5f and 5g). To examine whether the ALS-associated rare variants affect [Cl-]ER, we expressed hWT, M29T, S263R, or W267R mutant CLCC1 in 293FT cells expressing RaMorideER (Fig. 5h). Compared to hWT, S263R or W267R but not M29T mutant CLCC1 significantly increased steady-state [Cl-]ER, supporting that S263R and W267R are functionally damaging mutations. Indeed, both S263R and W267R also significantly reduced the internal Ca2+ release induced by ATP (Fig. 5i).
To examine the biological consequence of S263R and W267R in vivo, we generated S263R and W267R knock-in mouse lines (Extended Data Fig. 13a and 13b). Mice heterozygous for S263R and W267R were viable and fertile, and no obvious ER stress and protein misfolding was disclosed in S263R heterozygous mutant (S263R/+) cerebella (Fig. 5j) at about one month of age. However, Bip upregulation and ubiquitin-positive misfolded protein accumulation were documented in the cerebella of S263R/NM and W267R/NM mice, indicating that the ALS-associated rare variants promote ER stress and protein misfolding in vivo. We failed to harvest S263R/KO and W267R/KO pup, suggesting that the rare variants are functionally damaging in vivo, which is independent of NM2453 allele (Extended Data Fig. 13c and 13d).
Mice homozygous for S263R and W267R are viable and no obvious ER stress and protein misfolding was documented in cerebellum, spinal cord, and thalamus. However, when we treated W267R/W267R mutant mice with a subdose of tunicamycin (3 mg/kg B.W.) 36, we observed Bip upregulation and misfolded protein accumulation in the mutant thalamus and spinal cord but not in that of the wildtype (Fig. 5k, 5l, and Extended Data Fig. 14), supporting the idea that the mutant neuron is more vulnerable to the ER stress challenge than the wildtype. Therefore, we concluded that two rare nonsynonymous mutations found in ALS, S263R and W267R, impair CLCC1 channel function, and promote ER stress and protein misfolding in vivo.
Dosage-dependent effect of CLCC1 in disease severity and cell-autonomous effect of CLCC1 in motor neuron loss and TDP-43 pathology
As K298A mutant CLCC1 protein is not as stable as wildtype CLCC1 (Fig. 4j, Extended Data Fig. 10d and 10e), the ALS-associated rare variants decrease CLCC1 expression in cerebellum (Fig. 6a and 6b). Indeed, the S263R and W267R mutant CLCC1 undergo K48-specific ubiquitination in brain (Fig. 6c), a major signal for target protein degradation by proteasome 37. However, the K298A, S263R, and W267R mutant CLCC1 are as stable as wildtype CLCC1 in a heterologous system, but the C-terminus is required for CLCC1 stability (Extended Data Fig. 9a-9c and Extended Data Fig. 15).
We then summarized CLCC1 expression levels and disease severity from the animals carrying one or two of the 5 Clcc1 alleles, including the previously reported NM2453 (NM) and 4 alleles generated in this study (KO, K298A, S263R, and W267R alleles) (Fig. 6d). Because the NM allele is a hypomorphic allele 19, it does produce wildtype CLCC1 protein but reduces the expression to ~5.5% of wildtype allele (Fig. 6d). The KO/NM mice are viable and displayed similar phenotypes as shown in K298A/NM mice, but the phenotype onset (1 month of age) is earlier than that (3 months of age) of K298A/NM mice, indicating that K298A is a partial loss-of-function allele (Extended Data Movie 2 and Fig. 6d). Although S263R and W267R impair CLCC1 expression to ~13.5% and ~19% of wildtype allele (50%), respectively, which are much higher than that (~3%) of K298A and that (~5.5%) of NM alleles (Fig. 6d).
In the K298A/+ colony, we were surprised to find that ~10% (20/210, K298A/+*) animals appeared to exhibit severe phenotypes as early as postnatal 90.9 ± 5.5 days (Extended Data Fig. 16a and 16b and Extended Data Movie 3), reminiscent of the phenotypes shown in NM/K298A (Fig. 4 and Extended Data Movie 1). Curved spine (kyphosis) and muscular atrophy were also evidenced in K298A/+* but not in wildtype and K298A/+ mice (Extended Data Fig. 16c and 16d). Because dosage of CLCC1 is critical for the mutant phenotypes, we examined CLCC1 expression in various tissues in these K298A/+* animals. As expected, CLCC1 expression level was significantly decreased in these tissues compared to that of wildtype and K298A/+ animals (Extended Data Fig. 17a and 17b). The decreased CLCC1 expression seems not to be explained by the decreased Clcc1 mRNA (Extended Data Fig. 17c).
Three major downstream pathways are involved in ER unfolded protein response, including PERK-eIF2α-ATF3, ATF6, and IRE1α-XBP1 38. For PERK-eIF2α-ATF4 pathway, we detected upregulation of phospho-eIF2α and ATF4 in mutant cerebella, including the KO/NM, S263R/S263R, and W267R/W267R genotypes. For ATF6 and IRE1α-XBP1 pathways, we analyzed RNA-seq data from +/+ and NM/NM cerebella. The known ATF6 downstream genes, including Hsp90b1, Hspa5, Herpud1, and Ppp1r15a, were all significantly upregulated in the NM/NM cerebella compared to that of wildtype (Extended Data Fig. 18a); however, the XBP1 downstream genes, including Dnajb9, Prdm1, Syvn1, and Edem1, kept no change between the two genotypes. These data suggested that dysfunction of CLCC1 leads to ER unfolded protein response mainly through PERK-eIF2α-ATF4 and ATF6 pathways. In addition, activation of proteasome-mediated protein degradation was evidenced by upregulation of genes involved in proteasome functions and K48-specific ubiquitination in the NM/NM and K298A/NM cerebella, respectively (Extended Data Fig. 19).
To gain insight into cell-autonomous or non-cell-autonomous effect of Clcc1 loss-of-function in motor neuron degeneration, we generated Clcc1 floxed (fl) mouse (Fig. 6e) and crossed it to ChAT-Cre mouse 39, to knockout Clcc1 in ChAT-positive motor neuron in spinal cord. ER stress was evidenced by upregulation of both Bip and ERp72 in ChAT-positive motor neurons in ChAT-Cre/+;fl/fl but not ChAT-Cre/+;fl/+ spinal cords (Fig. 6f and 6g). Misfolded protein accumulation was also evidenced by upregulation of ubiquitin in these Clcc1 conditional KO neurons (Fig. 6f and 6g). Compared to nucleus-localized TDP-43 in ChAT-Cre/+;fl/+ motor neurons, cytoplasm-mislocalized and ubiquitin-positive TDP-43 (Fig. 6h), one of the pathological hallmarks of ALS 40-42, were documented in the conditional KO neurons. Indeed, all the ChAT-Cre/+;fl/fl animals died before P30 (Fig. 6i) with significant loss of motor neurons (Fig. 6j). Therefore, we conclude that the effect of Clcc1 loss-of-function in motor neuron loss is cell-autonomous.