TMEM63B channel is the osmosensor required for thirst drive of interoceptive neurons.

Thirst plays a vital role in the regulation of body fluid homeostasis and if deregulated can be life-threatening. Interoceptive neurons in the subfornical organ (SFO) are intrinsically osmosensitive and their activation by hyperosmolarity is necessary and sufficient for generating thirst. However, the primary molecules sensing systemic osmolarity in these neurons remain elusive. Here we show that the mechanosensitive TMEM63B cation channel is the osmosensor required for the interoceptive neurons to drive thirst. TMEM63B channel is highly expressed in the excitatory SFO thirst neurons. TMEM63B deletion in these neurons impaired hyperosmolarity-induced drinking behavior, while re-expressing TMEM63B in SFO restored water appetite in TMEM63B-deficient mice. Remarkably, hyperosmolarity activates TMEM63B channels, leading to depolarization and increased firing rate of the interoceptive neurons, which drives drinking behavior. Furthermore, TMEM63B deletion did not affect sensitivities of the SFO neurons to angiotensin II or hypoosmolarity, suggesting that TMEM63B plays a specialized role in detecting hyperosmolarity in SFO neurons. Thus, our results reveal a critical osmosensor molecule for the generation of thirst perception.

A splicing-dependent ER retention signal regulates surface expression of the mechanosensitive TMEM63B cation channel.

TMEM63B is a mechanosensitive cation channel activated by hypoosmotic stress and mechanic stimulation. We recently reported a brain-specific alternative splicing of exon 4 in TMEM63B. The short variant lacking exon 4, which constitutes the major isoform in the brain, exhibits enhanced responses to hypoosmotic stimulation compared to the long isoform containing exon 4. However, the mechanisms affecting this differential response are unclear. Here, we showed that the short isoform exhibited stronger cell surface expression compared to the long variant. Using mutagenesis screening of the coding sequence of exon 4, we identified an RXR-type endoplasmic reticulum (ER) retention signal (RER). We found that this motif was responsible for binding to the COPI retrieval vesicles, such that the longer TMEM63B isoforms were more likely to be retrotranslocated to the ER than the short isoforms. In addition, we demonstrated long TMEM63Bs could form heterodimers with short isoforms and reduce their surface expression. Taken together, our findings revealed an ER retention signal in the alternative splicing domain of TMEM63B that regulates the surface expression of TMEM63B protein and channel function.