Mice with neuron-specific deletion of Wnk1 have polyuria and relative hypotonic urine compared with WT mice. WNK1 kinase is ubiquitously expressed. Mice with global deletion of the Wnk1 gene are embryonic lethal due to angiogenesis defects (16, 17). We have generated viable adult mice with global Wnk1 deletion rescued by expression of constitutive-active OSR1 in endothelia but not in the kidney and brain. We found that these mice had relatively higher urine output and lower urine osmolality compared with WT littermates, which persists during water restriction (data not shown), suggesting that these mice have diabetes insipidus (DI). To investigate whether the defects are due to loss of WNK1 in the central nervous system or the kidney (i.e., central vs. nephrogenic DI), we generated mice with neuron-specific conditional KO (cKO) of Wnk1 using synapsin1-Cre (Wnk1fl/fl; syn1-Cre) (Figure 1A). In normal WT mice, WNK1 protein abundance in brain regions including the OVLT was lower than in the kidney, yet with significant expression (Figure 1, B and C). In neuronal specific Wnk1-cKO mice, WNK1 was markedly reduced in brain regions but not in the kidney. Immunofluorescent staining confirmed WNK1 expression in neurons in brain regions including the OVLT, SFO, and cerebral cortex of WT mice, and conditional deletion of Wnk1 markedly reduced the expression (Figure 1, D–F, and Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/JCI164222DS1).

Neuron-specific conditional KO of Wnk1 markedly reduces WNK1 in brain regions, including OVLT. (A) Genotyping of Wnk1-cKO mice mediated by neuron-specific Syn1-Cre. Genomic tail-clip DNA was used for analysis. Lane 1, Wnk1fl/+;Syn1-Cre; lane 2, Wnk1fl/fl; lane 3, Wnk1fl/fl;Syn1-Cre. PCR is shown to detect WT vs. Wnk1fl/fl locus (exon 2 and neo cassette are floxed). PCR forward primer F is located at exon 2. Reverse primers R1 and R2 are located at intron 2 and neo cassette, respectively. Note that Syn1-Cre is only active in neurons, so unexcised Wnk1fl/fl locus is detected in tail-clip DNA. With large size neo cassette in the floxed locus, the F/R1 primer set does not amplify under the condition of PCR reaction. Additionally, PCR is shown to detect Syn1-Cre using Cre-specific primers. (B) Representative Western blot of WNK1 protein in WT and cKO brain regions shown relative to the kidney. Hippo, hippocampus. Cortex, cerebral cortex. (C) Quantitation (mean ± SEM) of 4 separate experiments, as shown in B. One WT and cKO mouse for each experiment. WNK1 was normalized to Gapdh and compared with the WT kidney (set as “1”). *P < 0.05, #P < 0.01, KO vs. WT by unpaired t test. (D) Atlas of brain section for immunofluorescent staining, as in E and F. OVLT (also known as vascular-organ-of-lamina-terminalis [VOLT]) is marked by a red line. 3V, third ventricle; MnPO, median preoptic nucleus; MPA, medial preoptic area. (E and F) Immunofluorescent staining of WNK1 in OVLT neurons colocalized with neuronal marker β-3 tubulin in WT (E) and cKO (F) mice. Scale bar: 100 μm.

Autophosphorylation of WNK1 at S382 within the kinase domain reflects WNK1 kinase activation (9, 18–20). We found that water restriction (WR) increased the abundance of S382 phospho-WNK1 (Figure 2A, inset, and Supplemental Figure 2), supporting the role of WNK1 as an osmosensor. Balance studies revealed that water intake and urine output were significantly higher in cKO mice compared with control littermates (control; Wnk1fl/fl or Wnk1fl/+ without Cre) under free access to water (ad libitum) (Figure 2, A and B). While plasma osmolality was not significantly different during ad libitum, urine osmolality was lower in Wnk1-deleted mice compared with control mice (Figure 2, C and D). Polyuria with relative urine hypotonicity in Wnk1-cKO mice versus control mice persisted during water restriction (Figure 2, B and D). Plasma osmolality became significantly higher in cKO mice compared with control mice during water restriction (Figure 2C). These findings that polyuria and relative hypotonic urine persist in water restriction indicate DI, not polydipsia, as the underlying cause. Higher water intake in cKO mice is a compensatory response to polyuria (see Figure 3 below for effect of Wnk1 deletion on osmolality-induced thirst).

Wnk1-cKO mice exhibit partial central diabetes insipidus with impaired AVP and copeptin release in response to water restriction. (A) Water intake, (B) urine volume, (C) plasma osmolality, (D) urine osmolality, (E) plasma AVP level, and (F) copeptin level of control (Ctrl) and cKO mice at either ad libitum water intake or after 24-hour water restriction (WR). The inset in A shows Western blotting analysis of abundance of total and phospho-WNK1 (p-WNK1) using antibody against total WNK1 and against S382 phospho-WNK1. Arrowheads indicate molecular size 250 kDa. Lysates from WT OVLT tissue at ad libitum water intake and after 24-hour water restriction were immunoprecipitated by anti-WNK1 antibody and probed by anti-WNK1 and anti-p-WNK1 antibody. Representative of 4 separate experiments. Each experiment consists of 1 mouse ad libitum and 1 mouse on water restriction. For statistical analysis was performed with 2-way repeated ANOVA with Šidák post hoc analysis; for statistical analysis of the inset in A, unpaired 2-tailed t test was performed. For bar graphs, n = 6–8 mice, as indicated in scatter plots.

Wnk1 deletion impairs hyperosmolality-induced AVP release but not osmotic thirst. (A) Plasma osmolality, (B) [Na+], (C) relative p-WNK1/WNK1 ratio in OVLT, (D) plasma AVP, (E) urine volume, (F) urine osmolality, and (G) water intake in WT and Wnk1-cKO mice after mannitol or vehicle injection. Urine volume and water intake were measured 120 minutes after injection. Other measurements were taken 30 minutes after injection in separate mice from those in which urine and water intake were measured. The inset in C is representative of 3 experiments. Each experiment consists of 1 mouse injected with vehicle and 1 mouse injected with mannitol. Statistical analysis in A, B, and D was performed with 2-way repeated ANOVA with Šidák post hoc analysis; otherwise, unpaired t test was used. For bar graphs in A, B, D–G, n = 5 mice for each experimental condition, as indicated in scatter plot.

Neuronal deletion of Wnk1 impairs hypertonicity-stimulated release of AVP. To further support the central DI phenotypes of neuronal Wnk1 deletion, plasma levels of AVP and copeptin were measured. The half-life of circulating endogenous AVP is several minutes. Copeptin is the inactive N-terminal fragment of pre-pro-AVP, which is more stable in circulation and believed be a more reliable measurement of AVP release. As shown, basal levels of AVP and copeptin were not significantly different between cKO and control mice (Figure 2, E and F). Water restriction stimulated release of AVP and copeptin in control mice, and the increases were blunted in Wnk1-cKO mice. Thus, WNK1 is involved in hypertonicity-induced AVP release. Additional molecules or pathways besides WNK1 may be involved in regulating AVP release at least at the basal state (see Discussion and below). Overall, the central DI phenotypes of Wnk1-cKO mice is partial. With ad libitum water access, cKO mice appear grossly normal with indistinguishable activity level and apparent normal growth curve and body size compared with control mice (data not shown).

Neuronal deletion of Wnk1 does not impair hypertonicity-stimulated thirst. Hypertonicity stimulates thirst as well as release of AVP. Whether the two processes are mediated by the same molecular mechanism is unknown (21). The finding that cKO mice have higher water intake (than control mice) to compensate for polyuria (Figure 2A) suggests that Wnk1 deletion does not affect hypertonicity-induced thirst. Here, we used intraperitoneal mannitol injection to further examine the role of WNK1 kinase in osmolality-induced AVP release and thirst response. Mannitol injection raised plasma osmolality in both WT and cKO mice in 30 minutes (Figure 3A). The increases were significantly higher in cKO mice compared with WT mice (see below). Plasma [Na+] was decreased in mannitol-injected mice due to dilution by water extraction from cells (Figure 3B). Hypertonicity induced WNK1 phosphorylation in OVLT neurons in 30 minutes (Figure 3C) and increased plasma AVP levels in WT but not in Wnk1-cKO mice (Figure 3D). Osmolality and water load from mannitol injection induced significant urine output within 120 minutes. Consistent with the finding that Wnk1 deletion impairs AVP release, mannitol-induced urine volume was significantly higher and urine osmolality lower in cKO mice compared with that in WT mice (Figure 3, E and F). The relative hypotonic urine also explains higher plasma osmolality in cKO mice (Figure 3A). Water intake in response to hypertonicity was significantly higher in cKO mice than in WT mice (Figure 3G). Polyuria and hypotonic urine caused by defective AVP release likely account for higher plasma osmolality and thus higher water intake in cKO versus WT mice. Overall, these results using mannitol injection agree with the results shown in Figure 2 by using water restriction to increase plasma tonicity. They support the notion that Wnk1 deletion impairs AVP release but not osmolality-induced thirst.

Hypertonicity stimulates K+ current-mediated membrane potential oscillation in OVLT neurons involving WNK1 kinase. To further examined the role of WNK1 kinase in osmosensory neurons, we isolated OVLT neurons for whole-cell current-clamp recording. Current-clamp recording was performed with bath containing 140 mM NaCl and 5 mM KCl; pipette containing 130 mM K-acetate; and injection of 600 pA currents over 500 milliseconds (Figure 4A). Injecting 600 pA currents depolarized freshly isolated OVLT neurons to approximately +150 mV and elicited membrane potential oscillation that decayed rapidly in the baseline condition (Figure 4B). Recording from the same neuron after incubation with an additional 5 mM NaCl (i.e., bath contains 145 NaCl) for 3–5 minutes resulted in marked enhancement in membrane potential oscillation (Figure 4B). Superimposition further illustrated the difference between baseline and 5 mM NaCl. Rerecording after washout of 5 mM NaCl revealed that membrane potentials were restored to the baseline without oscillation (Figure 4B). As previously reported (22, 23), OVLT neurons are heterogeneous, not every neuron responded to hypertonic challenge (Figure 4C represents an example of a nonresponsive neuron). The percentage of NaCl-responsive neurons was used as a readout. As shown, in the control condition (labeled vehicle, i.e., DMSO), 13 of 19 neurons recorded (68%) exhibited hypertonicity-induced membrane potential oscillation (Figure 4D). In contrast, preincubation of isolated neurons with a pan-WNK inhibitor, WNK463 (10 μM for 3 hours) (24), significantly reduced the percentage of neurons (22%, 2 of 9) that responded with oscillation. In further support for the role of WNK1, hypertonicity-induced membrane potential oscillation was impaired in neurons isolated from neuron-specific Wnk1-cKO mice (Figure 4E). We also found that 10 mM mannitol (on top of 140 NaCl in bath) exerted the same effect (Supplemental Figure 3), indicating that it is activated by osmolality, not selectively by Na+. Thus, hypertonicity activates membrane channels in OVLT neurons through WNK1 cascade.

Hypertonicity induces membrane potential oscillation in freshly isolated OVLT neurons mediated by WNK1. (A) Ruptured whole-cell current-clamp recording for membrane potentials. Pipette and bath solution are indicated. (B and C) Membrane potentials of freshly isolated OVLT neurons at baseline, after incubation with 5 mM NaCl for 3 minutes and 5 minutes after washout of 5 mM NaCl hypertonicity. 600 pA currents were injected to depolarize membrane potential from the resting potential –55 mV to +150 mV. B and C represent examples of NaCl-responsive and nonresponsive neurons, respectively. (D) Treatment with pan-WNK kinase inhibitor (WNK463). Green and cyan bars indicate responsive (R) and nonresponsive (NR), respectively. WNK463 treatment significantly decreased the percentage distribution of responsive neurons vs. vehicle (Veh) treatment. P < 0.01, WNK463 vs. Veh, by 2-tailed Fisher’s exact test. (E) Wnk1-cKO eliminated NaCl responsiveness. P < 0.01, cKO vs. WT, by 2-tailed Fisher’s exact test. In D and E, OVLT neurons were isolated form 4–5 mice for vehicle-treated, WNK463-treated, WT, and cKO groups.

Release of AVP is also stimulated by hypovolemia through peripheral baroreceptor (25, 26). Hypovolemia also stimulates thirst through angiotensin II generated centrally and peripherally (27, 28). Whether baroreceptor-mediated AVP release and angiotensin II–stimulated thirst converge on same osmosensory neurons that utilize WNK1 as the molecular sensor is unknown. The 24-hour water restriction employed in our experiment increases plasma osmolality by approximately 2%–3% (~8 mOsm/kg increase on baseline 310 mOsm/kg, Figure 2C). An estimated more than 8%–10% reduction in central volume is required to stimulate AVP and angiotensin II release (25–28). Thus, baroreceptor-mediated AVP release is not likely involved in our experiment. Nonetheless, we asked whether angiotensin II could activate membrane potential oscillation in isolated OVLT neurons and found that it did not whereas the positive control 5 mM NaCl did (Supplemental Figure 4).

We further characterized the biophysical basis of membrane potential oscillation. OVLT neurons contain voltage-activated Na+ and K+ channels, mediating the membrane depolarization and repolarization phase of an AP, respectively. The membrane oscillation that we observed here was not an AP, as the repolarization did not reach below threshold membrane potential. The threshold membrane potential for opening Kv channels is more positive to the equilibrium potential for K+ (EK). Thus, we reasoned that the membrane potential oscillation is due to membrane potential alternating between depolarization from injecting positive currents and hyperpolarization from K+ efflux passing through open voltage-activated Kv channels. To validate the hypothesis, we replaced pipette and bath K+ with nonpermeant Cs+ and found that it eliminated membrane potential oscillation (Supplemental Figure 5A). The high activation threshold and delayed inactivation kinetics suggest that high-threshold Kv channels, such as Kv2 and Kv3 channels, are involved (29, 30). We then used tetraethylammonium (TEA) to distinguish between Kv2 and Kv3 channels. While all K+ channels are susceptible to blockade by millimolar concentrations of TEA applied intracellularly, extracellular TEA selectively blocks Kv3 channels with IC50 of approximately 300 μM (30). We found that 3 mM TEA completely blocked hypertonicity-induced responses in OVLT neurons (Supplemental Figure 5B), supporting that it is due to K+ efflux through Kv3 channels (see below for further identification of Kv3’s channel).

Kinase activity of WNK1 is involved in hypertonicity stimulation of K+ currents in OVLT neurons. Additional support for the notion that kinase activity of WNK1 is involved comes from the following results of studies substituting Mg-ATP in the pipette with ATP-free solution or ATP analogs. As shown in neurons recorded with pipette containing 2 mM Mg-ATP, 66% of neurons responded to 5 mM NaCl hypertonic challenge (Figure 5A). In ATP-free pipette solution, 0% of neurons responded (Figure 5B). When substituted with the nonhydrolyzable AMP-PNP, only 10% of neurons responded (Figure 5C). Incomplete suppression by AMP-PNP is likely due to residual ATP in the pipette. ATPγS is an ATP analog that functions as a substrate for protein kinase but not for protein phosphatase (31). Interestingly, when substituted with ATPγS, 42% of neurons responded to hypertonic challenge (Figure 5D). Among the 10 neurons that responded, 7 of 10 did not exhibit washout (Figure 5E shows an example without washout). Thus, hypertonicity-induced Kv3-mediated membrane potential oscillation response is due to ATP- and protein kinase–mediated phosphorylation. Dephosphorylation by protein phosphatase(s) underlies the fast recovery when hypertonic stimulation is removed. Overall, these results strongly support the hypothesis that phosphorylation by WNK1 kinase and dephosphorylation by a phosphatase(s) are important for the on/off effect of detecting hypertonic stress.

Effects of removal of intracellular ATP or substitution by ATP analogs on hypertonicity-induced membrane potential oscillation. Whole-cell patch-clamp recordings were performed as in Figure 3, with the exception that ATP in the pipette was removed or replaced as indicated. [Mg2+] was kept constant. (A) Control experiments with 2 mM ATP in the patch pipette. (B) Zero ATP in the patch pipette. (C) Patch pipette contained 2 mM AMP-PNP. (D) Patch pipette contained 2 mM ATPγS. (E) With ATPγS in the pipette, in 7 of 10 cells that responded to hypertonicity stimulation, membrane potential oscillation persisted after hypertonic NaCl was washed out. Shown is a representative example of persistent oscillation after washout. Note that ATPγS is a substrate for kinase but not for phosphatase due to thio-linkage between sulfur and oxygen atom. Pie charts in A–D show distribution of responsive and nonresponsive neurons. B and C are statistically significantly different from A, P < 0.05 by 2-tailed Fisher’s exact test. OVLT neurons were isolated from 4–6 mice for each experimental setting.

WNK1 in PVN-projecting CVO neurons is involved in hyperosmolality regulation of AVP release and water homeostasis. Osmosensory neurons in CVOs project to AVP-producing magnocellular neurosecretory neurons in the PVN and SON. Here, we used retrograde neuronal tracing to further investigate the hypothesis that WNK1 in CVO neurons is involved with osmolality detection. Subsets of OVLT neurons (Figure 6A) and SFO neurons (data not shown) displayed strong EGFP fluorescence (rather than tdTomato red fluorescence at baseline) following injection of retrograde adeno-associated virus–Cre recombinase (AAVrg-Cre) into the PVN of tdTomato-EGFP (mT/mG) reporter mice. The results validate the approach to reach CVO neurons by injecting retrograde AAV into the PVN.

Deletion of Wnk1 in PVN-projecting OVLT neurons is responsible for the partial CDI phenotype. (A) Injection of AAV-retro-Cre virus into PVN of tdTomato-EGFP reporter mice resulted in green fluorescence in neurons of OVLT nuclei, which otherwise exhibited tomato red fluorescence. Scale bar: 200 μm. (B) PVN injection of AAV-retro-Cre virus into Wnk1fl/fl mice resulted in deletion of Wnk1 in OVLT compared with control experiments with injection of AAV-retro-Cre virus into PVN of WT mice. Scale bar: 100 μm. (C) Urine volume, (D) urine osmolality, and (E) plasma osmolality of Wnk1fl/fl mice before and after injection with AAV-retro-Cre virus during at libitum and after water restriction (WR). (F) Urine volume, (G) urine osmolality, and (H) plasma osmolality of WT mice before and after injection with AAV-retro-Cre virus. Data shown are mean ± SEM from before injection (labeled retro-AAV –) and after injection (labeled retro-AAV +). Statistical analysis by 2-way repeated ANOVA with Šidák post hoc analysis. n = 4–6 mice as indicated by scatter plots.

Next, we studied Wnk1fl/fl mice in which the PVN was injected with retrograde AAVrg-Cre to delete Wnk1 in CVOs. WT mice were used as controls; they were injected with PVN with the same retrograde AAVrg-Cre. Immunofluorescent staining confirmed that WNK1 was markedly reduced in OVLT neurons in Wnk1fl/fl mice (but not in WT mice) in which the PVN was injected the retrograde AAVrg-Cre (Figure 6B). Water intake (data not shown), urine output, and urine osmolality were measured before and after retrograde AAV-Cre injection during ad libitum water access and water restriction. Supplemental Figure 6 shows the experimental protocol and timeline. Figure 6, C–E, show results from experiments with Wnk1fl/fl mice; Figure 6, F–H, shows result from experiments with control mice. In experimental mice with ad libitum water access, 10 days after injection urine output was significantly increased and urine osmolality significantly decreased versus before injection (Figure 6, C and D). Water restriction studies were performed to determine whether water homeostasis defects in mice received AAVrg-Cre injection in the PVN are due to DI or untoward effect of polydipsia from PVN injection. Polyuria and relative hypotonic urine persisted during water restriction (Figure 6, C and D), indicating that the effect is due to DI. Plasma osmolality was not different before and after retrograde AAVrg at ad libitum, but the difference became apparent with water restriction (Figure 6E). For control mice, there were no differences in these parameters before and after injection (Figure 6, F–H). The effect was due to deletion of Wnk1 in CVOs not in the PVN, as injecting nonretrograde AAV-Cre into the PVNs of Wnk1fl/fl mice did not induce DI phenotypes, whereas simultaneously performed experiments injecting the retrograde AAVrg-Cre as the positive control exerted effects (Supplemental Figure 7, D–F versus A–C). Of note, the abundance of WNK1 expression in the PVN was approximately 25% relative to that in the OVLT (Supplemental Figure 8).

Circulating plasma levels of copeptin were measured in AAVrg-injected mice as well as control mice at day 10 after injection, first at ad libitum and then after 24-hour water restriction (see experimental protocol in Supplemental Figure 6). Deleting WNK1 in CVOs by injecting a retrograde AAVrg-Cre virus in the PVNs of Wnk1fl/fl mice abolished hypertonicity-induced copeptin release, as evidenced by the lack of differences in the circulating levels between ad libitum and water restriction (Figure 7A). As a positive control, circulating levels of copeptin were increased by water restriction (vs. ad libitum) in the control WT mice with similar retrograde virus injection (Figure 7B). In separate groups of mice, we isolated OVLT neurons for whole-cell current-clamp recording from mice in which Wnk1 was deleted by injection with a retrograde AAVrg-Cre virus. As is shown in Figure 4E for neurons isolated from syn1-Cre–mediated cKO mice, hypertonicity-stimulated membrane potential oscillation was eliminated in mice in which Wnk1 was deleted in the OVLT by PVN retrograde AAV virus injection (0 of 11 neurons responded, Figure 7C). For control, hypertonicity stimulated membrane potential oscillation in 5 of 10 OVLT neurons isolated from WT mice injected with retrograde AAV virus injection (Figure 7D).

Deletion of Wnk1 in PVN-projecting OVLT neurons eliminates hypertonicity-induced membrane potential oscillation and blunts copeptin release in response to water restriction. (A and B) Cooopeptin release in Wnk1fl/fl and control WT mice with PVN injected with AAV-retro-Cre virus. Statistical comparison was made by paired t test between ad libitum and WR. (C and D) In separate groups of experimental (Wnk1fl/fl) and control (WT) mice, OVLT neurons were isolated for recording of membrane potential oscillation. Pie charts show distribution of neurons that exhibit membrane potential oscillation responsive and nonresponsive to HTS (5 mM NaCl). P < 0.01 between pie chart in C and D by 2-tailed Fisher’s exact test. In A and B, n = 5 Wnk1fl/fl and WT mice per experiment, as indicated in scatter plots.

Activation of Kv3.1 underlies hypertonicity-stimulated AVP release and water homeostasis. The results of high activation threshold and slow inactivation kinetics and inhibition by the extracellular TEA in whole-cell current-clamp recording of freshly isolated neurons suggest that channels of the Kv3 family are involved. Quantitative real-time PCR analysis revealed Kv3.1b was the predominant subtype in OVLT neurons (Supplemental Figure 9). We knocked down Kv3.1b in OVLT neurons by direct injection of AAV virus encoding shRNA against Kv3.1 gene. Successful knockdown of Kv3.1b in OVLT was evident by marked reduction of Kv3.1b protein in the OVLT but not in adjacent brain region (Figure 8, A and B, and Supplemental Figure 10). Compared with mice injected with control scrambled RNA, mice injected with anti-Kv3.1 shRNA developed polyuria and relative hypotonic urine during ad libitum water intake, which persisted during water restriction (Figure 8, C and D). Plasma osmolality was not different between mice with control scrambled RNA and with Kv3.1 shRNA during ad libitum water intake, but plasma osmolality became significantly different after water restriction (Figure 8E). As expected, water restriction increased circulating copeptin levels in control mice (Figure 8F). Knocking down Kv3.1b abolished the WR-induced increase, supporting its role in mediating hypertonicity stimulation of AVP release. For phenotype comparison, we also performed experiments by direct injection of an AAV virus carrying shRNA against Wnk1 gene. Mice with knockdown of WNK1 in the OVLT exhibited partial DI phenotypes, as in mice with knockdown of Kv3.1b (Supplemental Figure 11), supporting the notion that WNK1 and Kv3.1 act in the same pathway.

Knockdown of Kv3.1 by shRNA in OVLT causes partial central diabetes insipidus and impairs copeptin release in response to water restriction. (A) OVLT tissues from mice with direct injection scrambled RNA (Ctrl) or shRNA against Kv3.1 were probed by antibody against Kv3.1b. Note that the Kv3.1 shRNA targets both alternatively spliced Kv3.1a and Kv3.1b isoforms. (B) Mean ± SEM of Kv3.1b protein abundance from 3 separate experiments as shown in A (data from each experiment is the average of triplicate samples). Statistical analysis by unpaired t test. (C) Urine volume, (D) urine osmolality, (E) plasma osmolality, and (F) copeptin levels of mice injected with control scrambled RNA (labeled –) or shRNA against Kv3.1b (labeled +) into OVLT and at either ad libitum water intake or after 24-hour water restriction (WR). Unpaired t test for comparison between control scrambled RNA and Kv3.1 shRNA in B. In C–F, n = 5 mice per group injected with control scrambled or with Kv3.1b shRNA as indicated in scatter plots. Statistical analysis was performed with by 2-way repeated ANOVA with Šidák post hoc analysis.

Activation of WNK1 in CVOs stimulates AVP release with reduction in urine output and increases in urine osmolality. Chloride ion (Cl–) binding to the activity center of WNK kinases inhibits catalytic activities, and mutation of Cl–-binding amino acids in WNKs activates their kinase activity (32). Mice carrying knockin allele of Cl–-insensitive WNK4 have phenotypes that are characteristic WNK4 gain of function (33). We generated conditional knockin (cKI) mice carrying a floxed allele of conditionally activatable Cl–-insensitive WNK1 (Supplemental Figure 12). The CKI allele carried a reverse-oriented exon 3 with nucleotides coding for double L369 and L371 to phenylalanine mutation (L369F/L371F) that reorientates and becomes active upon Cre-mediated excision.

Mice heterozygous for Cl–-insensitive Wnk1-knockin allele and control WT mice were injected with an AAV-Cre virus in the OVLT. As shown, the activity of WNK1 in OVLT neurons in knockin mice was enhanced, as demonstrated by increased abundance of phospho-OSR1/SPAK (Figure 9A). Plasma AVP levels were increased in cKI mice after AAV-Cre injection compared with those before (Figure 9B). Along with the increases in AVP levels, mice exhibited reduction in urine volume as well as an increase in urine osmolality (Figure 9, C and D). WT mice that received AAV-Cre injection in the OVLT did not exhibit changes in plasma AVP levels or urine volume and urine osmolality (Figure 9, E–G). Thus, an increase in WNK1 kinase activity leads to AVP release.

Activation of WNK1 in OVLT increases AVP release. WT mice or mice heterozygous for GOF Cl–-insensitive Wnk1-knockin (Wnk1-KI) allele received AAV-Cre virus injection in OVLT. (A) Relative abundance of phospho-OSR/SPAK (p-OSR/SPAK) in KI mice before (–) and after (+) injection, as measured by Western blotting analysis of OVLT using antibody against S373-phospho-SPAK/S325-phospho-OSR1. The inset shows representative Western blotting of 3 separate experiments. Each experiment consists of 3 replicates of WT and 3 Wnk1-KI mice. Each data point in the bar graph represents the average of 3 replicates. Statistical analysis by unpaired t test. (B) Plasma AVP level, (C) urine volume, (D) urine osmolality in heterozygous Wnk1-KI mice in which OVLT was injected with AAV-Cre virus, (E) plasma AVP level, (F) urine volume, and (G) urine osmolality of WT mice in which OVLT was injected with AAV-Cre virus. In B–G, n = 5 mice, as indicated in line plots. Statistical analysis by paired t test.

Inhibition of WNK1 by pan-WNK inhibitors or genetic deletion of Wnk1 blunts hypertonicity stimulation of AP firing in OVLT neurons. Our studies thus far have examined the role of WNK1 in water homeostasis and hypertonicity stimulation of AVP release at the whole-animal level as well as in freshly isolated OVLT neurons. To solidify the conclusion, we further addressed the involvement of the WNK1 using ex vivo recordings of spontaneous APs on the OVLT-containing brain slices. PVN-projecting OVLT neurons were identified with the retrogradely expressed fluorescence signals (Figure 10, A–C). Spontaneous APs could be detected when the OVLT neurons were synaptically isolated with bath application of synaptic blockers (see Methods) and current clamped at approximately –50 ± 5 mV (Figure 10D). Hypertonic challenge (Δ[NaCl] = 5 mM) increased spontaneous AP generation by approximately 33% (15 of 46 cells) in the vehicle-treated neurons (Figure 10, D–F). Preincubation of pan-WNK inhibitor WNK463 (10 μM) for 3 hours significantly reduced the percentage of neurons that responded to hypertonicity (Figure 10, G–I, 1 of 17 cells, 6%; P = 0.048, 2-tailed Fisher’s exact test comparing Figure 10, F and I). In addition to pharmacological inhibition of WNK1, we carried out studies in mice in which Wnk1 in the OVLT was deleted by injecting a retrograde Cre-expressing AAV in the PVN (Figure 11, A–C). Approximately 36% (8 of 22 cells) of control WT neurons increased AP firing in response to hypertonicity stimulation (Figure 11, D–F), while only 8% (2 of 24 cells) of the Wnk1-cKO neurons responded to hypertonicity stimulation (Figure 11, G–I; P = 0.032, 2-tailed Fisher’s exact test comparing Figure 11, F and I). These data support the notion that WNK1 function is crucial for the osmolality sensing and stimulation of AP generation in PVN-projecting OVLT neurons.

Pharmacological inhibition of WNK1 abolishes hypertonicity-induced spike generation in PVN-projecting OVLT neurons. (A) Schematic of the retrograde tracer CTB-594 (Alexa Fluor 594–conjugated recombinant cholera toxin subunit B) injection at the PVN for labeling of PVN-projecting OVLT cells. (B) Representative coronal section of the mouse brain injected with CTB-594 at the PVN region. Scale bar: 1 mm. (C) Overlay of epifluorescence and IR-DIC images showing CTB-expressing neurons in the OVLT region. Scale bar: 10 μm. A recording pipette attached to a CTB-expressing cell is illustrated. (D) Top: Representative traces of the spontaneous firing recorded from a NaCl-responsive (R; cyan trace) neuron and a NaCl-nonresponsive (NR; red trace) neuron. Slices were incubated in the vehicle-containing solution before recording. Bottom: Histogram of z score from the representative NaCl-R and NaCl-NR cells. (E) Distribution of the z score change (Δz score) in response to 5 mM NaCl stimulation of all recorded cells in the vehicle group. The dashed line indicates 0.5. (F) Pie chart showing distribution of NaCl-R (Δz score > 0.5) and NaCl-NR (Δz score < 0.5) PVN-projecting OVLT neurons in the vehicle group. (G) Top: Representative traces of the spontaneous firing recorded from a NaCl-R neuron and a NaCl-NR neuron. Slices were incubated in the WNK463-containing solution before recording. Bottom: Histogram of z score from the representative NaCl-R and NaCl-NR cells. (H) Distribution of the Δz score in response to 5 mM NaCl stimulation of all recorded cells in the WNK463 group. The dashed line indicates 0.5. (I) Pie chart showing distribution of NaCl-R and NaCl-NR PVN-projecting OVLT neurons in the WNK463 group. *P = 0.048, between F and I, 2-tailed Fisher’s exact test. The vehicle group consists of recordings of 46 cells from 34 mice. WNK463 consists of 17 cells from 9 mice.

Wnk1 deletion reduces hypertonicity-induced spike generation in PVN-projecting OVLT neurons. (A) Schematic of the virus-mediated KO of Wnk1 in PVN-projecting OVLT neurons via injection of Cre-expressing retrograde virus at the PVN region. (B) Representative coronal section of the mouse brain injected with Cre-expressing virus at the PVN region. Scale bar: 1 mm. (C) Overlay of epifluorescence and IR-DIC images showing Cre-expressing neurons in the OVLT region. Scale bar: 10 μm. A recording pipette attached to a Cre-expressing cell was illustrated. (D) Top: Representative traces of spontaneous firing recorded from a NaCl-R neuron (R; cyan trace) and a NaCl-NR neuron (NR; red trace) in the WT mice. Bottom: Histogram of z score from the representative NaCl-R and NaCl-NR cells. (E) Distribution of the Δz score in response to 5 mM NaCl stimulation of all recorded neurons in WT mice. The dashed line indicates 0.5. (F) Pie chart showing distribution of NaCl-R and NaCl-NR PVN-projecting OVLT neurons in WT mice. (G) Top: Representative traces of spontaneous firing recorded from a NaCl-R neuron and a NaCl-NR neuron in the Wnk1–conditional KO (cKO) mice. Bottom: Histogram of z score from the representative NaCl-R and NaCl-NR cells. (H) Distribution of the Δz score in response to 5 mM NaCl stimulation of all recorded neurons in Wnk1-cKO mice. The dashed line indicates 0.5. (I) Pie chart showing distribution of NaCl-R and NaCl-NR PVN-projecting OVLT neurons in Wnk1-cKO mice. *P = 0.032, between F and I, 2-tailed Fisher’s exact test. The WT group consists of recordings of 22 cells from 13 mice; the cKO group consists of 24 cells from 11 mice.

We also confirmed that the stimulatory effect on AP firing is due to hypertonicity, rather than triggered by increases in [Na+]. We found that increasing extracellular osmolality via bath application of additional 10 mM mannitol increased AP generation in approximately 45% of OVLT neurons (Supplemental Figure 13). Furthermore, we demonstrated that Kv3 underlies the hypertonicity-induced AP generation: TEA (at 3 mM, which preferentially blocks Kv3) prevented hypertonicity-induced AP generation in OVLT neurons (Supplemental Figure 14). Overall, our results provide compelling support for the notion that activation of Kv3.1 is important for WNK1-mediated regulation of APs in OVLT neurons in response to hypertonicity.