Vasopressin neurons are not ordinary nerve cells. They sit at the interface of the nervous and endocrine systems, translating patterns of electrical spikes into the secretion of a peptide hormone at axonal terminals located in the posterior pituitary, a considerable distance from their cell bodies. When osmoreceptors detect that the blood is becoming too concentrated, or that plasma volume is falling, synaptic inputs drive these neurons to fire in complex, dynamically heterogeneous patterns. The summed secretory output of thousands of such neurons generates the plasma vasopressin signal that acts on the kidneys to regulate water loss. The precision of this control is remarkable, and it depends on a supply chain that spans the length of the neuron.

The hormone itself is synthesised in the neuronal cell bodies, where the vasopressin gene is transcribed into mRNA and translated into peptide. The hormone is then packaged into vesicles and transported along the axons to large stores at the pituitary terminals. These stores are the system’s buffer, and they are what allow the animal to respond immediately to a challenge. Supported by activity-dependent upregulation of synthesis and transport, the stores can maintain an elevated secretion response for several days of sustained high osmolarity, a situation that experimenters reproduce in the laboratory through dehydration or salt loading. Yet the stores are not inexhaustible.

One of the most intriguing observations in this field is that, despite the upregulation of synthesis that accompanies prolonged challenge, the pituitary stores gradually decline during the sustained response. Once the challenge ends, the stores recover only slowly over a further extended period. This pattern raises a fundamental question of control engineering: how does the system know how much hormone to make? The stores themselves sit far away in the posterior pituitary, and there is no evidence of a signal travelling back from the stores to inform the cell bodies of their contents. The neurons appear to be managing a supply chain without inventory reports from the warehouse.

Previous simpler models offered a candidate solution. They explained the observed synthesis dynamics based on activity-dependent upregulation of transcription and mRNA content. In this scheme, the firing activity of the neuron itself is the signal. Sustained high activity drives increased transcription of the vasopressin gene, raising the level of mRNA in the cell body, which in turn raises the rate of peptide synthesis and transport toward the terminals. Because mRNA is relatively stable on long timescales, its abundance integrates neuronal activity over hours and days, smoothing out the rapid fluctuations of spiking into a slow, demand-tracking signal for production.

The new work takes this idea and embeds it in a far more realistic setting. Rather than treating the system as a set of abstract compartments, MacGregor built a detailed neuronal model that couples spiking, secretion, and synthesis within a single computational framework, simulating a complete neural system from physiological input to hormonal output. The model receives osmotic and volume-related inputs, converts them into spike patterns in populations of vasopressin neurons, links those spikes to secretion from the pituitary stores, and simultaneously links activity to mRNA transcription and the resulting synthesis and transport of new hormone. This allows the full dynamics of a prolonged osmotic challenge and the subsequent recovery period to be simulated end to end.

The simulations reproduce the characteristic experimental signature of the system: an immediate secretory response sustained by the existing stores, a gradual upregulation of mRNA and synthesis as activity persists, a slow decline of the stores despite increased production, and a slow restoration of the stores once the challenge is relieved. Crucially, the model achieves this balance without any feedback signal from the stores themselves. The coordination emerges from the timing properties of the molecular machinery. Spikes and intracellular calcium provide fast signals that govern secretion and short-term regulation, while the accumulation and decay of mRNA provide a slow signal that governs the rate of hormone manufacture. In effect, the mRNA pool remembers how hard the neuron has been working over the past days and sets production accordingly.

This division of labour across timescales is what the study identifies as the key organising principle of the system. The model suggests that mRNA acts as a long-timescale memory of neuronal activity, extending temporal integration beyond spike activity and intracellular calcium to coordinate synthesis with sustained demand. The implications reach beyond vasopressin. Many neuroendocrine and peptidergic systems face the same structural problem: hormone or peptide is released far from where it is made, stores are finite, and demand can persist for days. If activity-dependent transcription can serve as a decentralised demand signal in the vasopressin system, similar mechanisms may underpin supply-demand matching in other hypothalamic systems, such as those governing oxytocin release, feeding behaviour, or stress hormone regulation.

The computational approach also demonstrates the value of whole-system modelling in physiology. Isolated measurements of firing rates, mRNA levels, store sizes, or plasma hormone concentrations each capture only a slice of the dynamics, and it is difficult to infer from any one of them how the system achieves its balance. By coupling all of these processes in a single model calibrated against the known behaviour of the system during dehydration and salt loading, the study shows that a plausible, mechanistically grounded set of local rules is sufficient to reproduce the global performance of the neuroendocrine circuit. No central controller is required; the balance emerges from the interplay of fast electrical signals and slow gene-expression dynamics distributed across thousands of neurons.

For researchers studying fluid homeostasis, the model provides a testable framework. It makes concrete predictions about how mRNA levels, synthesis rates, and store contents should evolve under different patterns of osmotic challenge, and about how perturbing transcription or transport should alter the system’s ability to sustain secretion. More broadly, it illustrates how biological systems can solve control problems without explicit feedback loops, using the intrinsic timescales of their molecular components as memory. As the authors’ analysis shows, the humble vasopressin neuron, by letting its own activity write a slow record into its mRNA, has evolved an elegant answer to the problem of keeping a distant warehouse stocked while the orders keep flowing in.

Subject of Research: Computational modelling of activity-dependent vasopressin mRNA transcription and hormone store dynamics during prolonged osmotic challenge

Article Title: Modelling vasopressin synthesis and storage dynamics during prolonged osmotic challenge and recovery based on activity dependent upregulation of mRNA transcription

Article References: Modelling vasopressin synthesis and storage dynamics during prolonged osmotic challenge and recovery based on activity dependent upregulation of mRNA transcription. (n.d.). https://doi.org/10.1371/journal.pcbi.1014832

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014832

Keywords: vasopressin, hypothalamus, neuroendocrine, mRNA transcription, posterior pituitary, osmotic challenge, computational model, hormone secretion, PLOS Computational Biology, fluid homeostasis, activity-dependent regulation, supply-demand balance

The research, published in Physiological Reports by a team led by Takahiro Masuda and Masahide Yoshida of Jichi Medical University together with colleagues in Japan and the United States, examined whether the gene encoding SGLT2, known formally as Slc5a2, is switched on in the neurons that govern thirst and vasopressin synthesis. For years, physiologists have observed that SGLT2 inhibitors stimulate water intake and vasopressin secretion even when blood volume and plasma osmolality remain normal, which suggests that something beyond simple dehydration is driving these responses. Because the drugs can cross the blood-brain barrier and SGLT2 immunoreactivity had previously been reported in brain tissue, the team suspected that the transporter might be present in the very neurons that control drinking behavior and hormone release.

To find out, the researchers studied male Sprague-Dawley rats and combined two complementary techniques: droplet digital PCR, an ultra-sensitive method for quantifying gene transcripts, and in situ hybridization, which reveals exactly which cells express a given gene. Using droplet digital PCR on micro-punched tissue samples, they detected Sglt2 messenger RNA in four key regions: the organum vasculosum of the lamina terminalis (OVLT) and the subfornical organ (SFO), two sensory circumventricular organs that form the brain’s thirst center, and the supraoptic nucleus (SON) and paraventricular nucleus (PVN) of the hypothalamus, where vasopressin is synthesized. The relative abundance of Sglt2 transcripts in these regions, normalized to the housekeeping gene Polr2a, fell in the range of 0.02 to 0.04, a level comparable to that of Agtr1a, the angiotensin II receptor type 1a gene that is well established as a central regulator of drinking and vasopressin release.

Brain expression was, unsurprisingly, far lower than in the kidney, the body’s dominant site of SGLT2 expression. Depending on which of three internal control genes was used for normalization, Sglt2 expression in the paraventricular nucleus, the region with the strongest signal, was roughly 76 to 205 times lower than in kidney tissue. Yet the authors argue that low abundance does not mean functional irrelevance. The fact that Sglt2 transcripts appear at levels similar to those of Agtr1a, a gene whose central role in thirst regulation is beyond dispute, suggests that even modest neuronal expression could be physiologically meaningful, particularly if the neurons involved are the direct targets of circulating drug.

The anatomical story became far more compelling when the team turned to double fluorescent in situ hybridization, a technique that can visualize two genes simultaneously within single cells. In the OVLT and the SFO, Sglt2 messenger RNA was clearly present in neurons that also express Nos1, the gene encoding neuronal nitric oxide synthase. These NOS1-expressing neurons are known from landmark studies to be the excitatory, thirst-promoting cells of the circumventricular organs; when activated, they drive drinking behavior, and they send direct projections to the vasopressin-producing nuclei of the hypothalamus. Quantification showed that approximately 48.1 percent of NOS1-positive neurons in the OVLT and 60.3 percent in the SFO expressed Sglt2. In other words, the molecular target of SGLT2 inhibitors sits within roughly half of the very neurons that command an animal to drink.

The co-localization in the hypothalamus was even more striking. In the supraoptic nucleus, 80.8 percent of vasopressin-expressing neurons carried Sglt2 transcripts, and in the paraventricular nucleus the figure reached 84.0 percent. The researchers also examined oxytocin neurons, the other major neuroendocrine population in these nuclei, and found Sglt2 expression in 58.9 percent of oxytocin neurons in the SON and 65.7 percent in the PVN. Chromogenic in situ hybridization confirmed the pattern at the tissue level, with Sglt2 signals appearing not only in the four fluid-regulating regions but also in the choroid plexus, consistent with earlier reports in human and mouse brains that identified the transporter in choroid plexus epithelial cells and ependymal cells. Prior work had also located SGLT2 in brain pericytes and in neurons of the rostral ventrolateral medulla, a master controller of sympathetic nerve activity, but this study is the first to pin the transporter down at single-cell resolution within the thirst and vasopressin circuits themselves.

These findings provide an anatomical basis for a hypothesis that has been gathering force in the literature: that SGLT2 inhibitors act directly on the brain, not merely through the kidney. Two features of the circumventricular organs make this plausible. First, the OVLT and SFO sit outside a functional blood-brain barrier, meaning drugs circulating in the bloodstream can reach them directly. Second, SGLT2 inhibitors as a class have been shown to penetrate the blood-brain barrier, and central administration of the drug tofogliflozin has already been demonstrated to stimulate water intake in non-diabetic rats, just as peripheral administration of ipragliflozin does in both diabetic and non-diabetic animals. If SGLT2 inhibitors enhance the excitability of NOS1 thirst neurons in the OVLT and SFO, that could trigger drinking independently of any actual fluid deficit, which would explain the paradoxical observation that these drugs increase vasopressin secretion and copeptin levels, a stable clinical surrogate for vasopressin, even under normal osmotic conditions.

The authors go further, weaving in evidence from glucose-sensing physiology. In the SFO, a subset of so-called glucose-inhibited neurons, cells that become excited when intracellular glucose falls, shares properties with NOS1 neurons, and many glucose-inhibited neurons also respond to angiotensin II, a potent dipsogenic hormone. Roughly 27 percent of SFO neurons show this glucose-inhibited character, and the majority of them respond to angiotensin II. The team speculates that by reducing intracellular glucose uptake through SGLT2, the inhibitors might directly stimulate a NOS1-positive, glucose-sensing population, initiating drinking before osmotic diuresis has had time to deplete body fluids. Intriguingly, a parallel logic operates in the kidney: in macula densa cells, which behave in many ways like neurons, SGLT1-mediated glucose uptake supports NOS1 activity that modulates tubuloglomerular feedback. The authors suggest that SGLT-dependent glucose sensing linked to nitric oxide synthase may thus push systemic fluid balance in a negative direction in both organs, a conceptual echo of coupled hunger-and-thirst sensing described in the fly brain.

Important caveats remain, and the authors are careful about them. This study mapped gene expression; it did not directly test function. Whether blocking SGLT2 in these neurons actually changes their electrical activity, drinking behavior, or vasopressin release has not yet been demonstrated, and global SGLT2 knockout mice, which drink more and secrete more vasopressin than wild-type animals, cannot disentangle central from renal mechanisms because fluid balance disturbances confound the interpretation. Brain-region-specific knockout models and intracerebroventricular drug administration will be needed to establish causality, along with electrophysiological recordings and activity markers such as Fos in SGLT2-expressing neurons. It is also uncertain how much of the vasopressin response during SGLT2 inhibition reflects direct central action versus classical osmotic pathways, since clinical studies show copeptin remains strongly tied to plasma osmolality during treatment. Still, the finding that Sglt2 is expressed in the majority of vasopressin neurons and roughly half of thirst-promoting NOS1 neurons reframes how scientists think about these drugs. Beyond fluid balance, vasopressin and oxytocin neurons influence social behavior, stress responses, and learning and memory, raising the speculative but testable possibility that SGLT2 inhibitors touch neuroendocrine circuits far beyond hydration. If the central mechanisms are confirmed, they could deepen understanding of the brain-kidney axis and help explain why a drug class designed to squeeze glucose out of urine has turned out to be one of the most versatile cardiorenal therapies ever developed.

Subject of Research: SGLT2 gene expression in thirst-promoting NOS1 neurons and vasopressin-synthesizing neurons in the rat brain

Article Title: SGLT2 gene expression in thirst‐promoting NOS1 neurons and vasopressin‐synthesizing neurons in male rats

Article References: Masuda, T., Yoshida, M., Kim, Y. C., Vallon, V., Morishita, Y., Nagata, D., & Onaka, T. (2026). SGLT2 gene expression in thirst‐promoting NOS1 neurons and vasopressin‐synthesizing neurons in male rats. Physiological Reports, 14(17), Article e71100. https://doi.org/10.14814/phy2.71100

Image Credits: AI Generated

DOI: 10.14814/phy2.71100

Keywords: SGLT2, Slc5a2, thirst, vasopressin, NOS1 neurons, subfornical organ, organum vasculosum laminae terminalis, supraoptic nucleus, paraventricular nucleus, fluid homeostasis, in situ hybridization, cardiorenal