The team began by mapping what happens to the serine synthesis pathway after cerebral ischemia. Using a mouse model of distal middle cerebral artery occlusion, they found that PHGDH, the rate-limiting enzyme that converts the glycolytic intermediate 3-phosphoglycerate into serine, was markedly elevated in the peri-infarct cortex at 7 and 14 days after stroke. Its downstream partners, PSAT1 and PSPH, rose in parallel, and liquid chromatography-mass spectrometry confirmed a 65 percent increase in serine content in the recovering tissue. The enzyme’s activity, in other words, surges precisely during the delayed window when new vessels begin to form.

Cell-type analysis pinpointed the source. Roughly 77 percent of GFAP-positive astrocytes in the ischemic cortex expressed PHGDH, compared with about 20 percent of CD31-positive endothelial cells, with only scattered signal in neurons, microglia, and oligodendrocytes. Confocal microscopy with three-dimensional reconstruction showed PHGDH-positive astrocytic processes wrapping around and ensheathing peri-infarct blood vessels, placing the enzyme’s product, serine, in direct physical contact with the growing vasculature. Hypoxia experiments in cultured astrocytes and endothelial cells reproduced the effect, suggesting that low oxygen itself triggers the metabolic shift.

To test whether astrocytic PHGDH was truly necessary, the researchers deleted the enzyme selectively in astrocytes by injecting an adeno-associated virus carrying Cre recombinase under the astrocyte-specific GfaABC1D promoter into the cortex of floxed Phgdh mice. The deletion cut PHGDH expression in the ischemic cortex by 66 percent, eliminated most PHGDH-positive astrocytes, and reduced serine concentrations in the injured tissue. The vascular consequences were striking: vascular area and length in the peri-infarct cortex fell by 29 and 22 percent respectively, and the number of proliferating BrdU-positive endothelial cells dropped sharply, along with markers of endothelial tip cells and proangiogenic factors such as VEGFA, VEGFR2, and CD105.

The vessels that did form were also of poorer quality. New vessels after stroke are often leaky, and the tracer experiments made the deterioration vivid. Alexa Fluor 555-cadaverine, a small fluorescent molecule, accumulated far more extensively in the brain parenchyma of PHGDH-deleted mice, and in vivo multiphoton microscopy showed increased leakage of 40-kilodalton FITC-dextran and endogenous IgG from cerebral vessels. Pericyte coverage of capillaries diminished, and the tight junction proteins ZO-1 and claudin-5, which seal the endothelial barrier, were substantially reduced in isolated microvessels. Functional vessel formation, measured by co-labeling BrdU with tomato lectin, fell by 33 percent, and laser speckle imaging confirmed reduced cerebral blood flow. The mice fared worse on behavioral tests of forelimb force and beam walking, linking the vascular failure to lasting neurological deficits.

The flip side proved equally persuasive. When the team overexpressed PHGDH specifically in astrocytes of wild-type mice using a GfaABC1D-driven AAV vector, serine levels in the ischemic cortex rose, endothelial proliferation increased, vascular area and length expanded, and vascular leakage declined. Pericyte coverage improved, perfused capillary length grew, regional cerebral blood flow recovered more fully, and the animals showed significantly better long-term neurological function. Boosting a single metabolic enzyme in support cells, it appeared, could tilt the entire poststroke repair program toward regeneration.

The researchers then traced how astrocyte-derived serine actually reaches the vasculature. Their candidate was SLC38A2, a system A amino acid transporter expressed in brain endothelial cells with a preference for neutral amino acids such as serine. PHGDH overexpression raised SLC38A2 levels in isolated brain microvessels, and when the team silenced the transporter specifically in endothelial cells with a blood-brain barrier-crossing AAV-BR1 shRNA vector, serine content in cortical microvessels fell by 55 percent. Levels of glycine, glutamine, and alanine were unaffected, arguing that the effect was specific to serine transport. Knocking down SLC38A2 also abolished the gains in vascular area, pericyte coverage, and barrier integrity that PHGDH overexpression had produced, identifying the transporter as the essential conduit between astrocytic metabolism and endothelial remodeling.

Downstream of serine delivery, the study implicated the mechanistic target of rapamycin complex 1, or mTORC1, a nutrient-sensing signaling hub long known to drive cell growth and angiogenesis. Astrocytic PHGDH deletion suppressed the phosphorylation of S6 and 4EBP1, canonical mTORC1 readouts, in isolated brain microvessels, while PHGDH overexpression enhanced their activation. Endothelial SLC38A2 silencing likewise inactivated mTORC1 in PHGDH-overexpressing mice. Most tellingly, rapamycin, a pharmacological mTORC1 inhibitor, reduced vascular area and length, worsened tracer leakage and IgG extravasation, and shrank the length of perfused capillaries in stroked animals, phenocopying the loss of PHGDH itself.

The findings reframe astrocytes as metabolic suppliers of poststroke angiogenesis, extending earlier work showing that serine synthesis supports endothelial proliferation and that PHGDH variants in humans disrupt retinal vascular integrity. They also fit a broader picture in which amino acid metabolism, from aspartate to serine, fuels the mTORC1-dependent translation program that growing endothelial cells require. In this scheme, astrocytes act as a metabolic bridge: sensing ischemia, upregulating PHGDH, exporting serine through endothelial SLC38A2, and thereby licensing mTORC1 activation inside the vessel wall.

The authors are careful about the limits of the work. The signals that induce PHGDH after stroke remain unidentified, the contribution of endothelial PHGDH to the same process was not dissected, and serine’s effects on neurons, oligodendrocytes, and other glial lineages await study. Whether other amino acid transporters share the serine delivery role, and whether serine supplementation could help patients, are open questions that will require clinical investigation. Still, the study offers a concrete and druggable axis, PHGDH, SLC38A2, and mTORC1, that connects glial metabolism to vascular repair, and it suggests that boosting this pathway during the subacute phase after ischemic stroke might one day help the brain rebuild its own blood supply.

Subject of Research: The role of astrocytic PHGDH-mediated serine synthesis in poststroke vascular remodeling and recovery after ischemic stroke

Article Title: PHGDH-derived serine is essential for vascular remodeling after stroke

Article References: Liu, Z., Liu, C., Du, P., Geng, X., Wang, S., Wang, Z., Huang, Y., Chang, L., Wang, Y.-C., Fan, W., & Zhao, B.-Q. (2026). PHGDH-derived serine is essential for vascular remodeling after stroke. Journal of Advanced Research. https://doi.org/10.1016/j.jare.2026.09.004

Image Credits: AI Generated

DOI: 10.1016/j.jare.2026.09.004

Keywords: ischemic stroke, PHGDH, serine synthesis, astrocytes, angiogenesis, vascular remodeling, SLC38A2, mTORC1, blood-brain barrier, cerebral blood flow, neurovascular unit, stroke recovery

For decades, L-serine was classified as a non-essential amino acid, meaning scientists assumed the brain could synthesize sufficient quantities internally through the phosphorylated pathway. Experiments in adult mice supported this view: deleting Phgdh, the gene encoding a key enzyme in the brain’s de novo serine synthesis, sharply reduces brain L-serine levels, indicating that local production dominates in mature animals. Yet the new work, led by Akshay K. Tiwari and Herman Wolosker of the Technion Israel Institute of Technology together with colleagues in Israel, the United States, and Switzerland, demonstrates that this picture is incomplete, particularly during early life. The developing brain, it seems, depends heavily on a supply line from the blood.

The researchers began by testing what happens when dietary serine disappears. Three-week-old mice were placed on a diet deficient in both serine and glycine, the latter restricted to prevent the two amino acids from replenishing one another through biochemical interconversion. After just ten days, high-performance liquid chromatography revealed that hippocampal L-serine had fallen by 45 percent and D-serine by 60 percent, with similar drops in the cerebral cortex. Serum serine and glycine also declined, confirming that the diet had genuinely restricted the amino acids reaching the brain. Because D-serine is manufactured inside the brain from L-serine by the enzyme serine racemase, the result made one thing clear: dietary L-serine is a major determinant of brain D-serine levels, and by extension, of the signaling molecules that calibrate NMDA receptor activity.

The metabolic consequences extended far beyond serine itself. Targeted metabolomics showed widespread reprogramming in the cortex, liver, and kidneys, with reductions in the serine precursor phosphoserine, pyruvate, and alpha-ketoglutarate, and increases in metabolites such as acetyl-CoA, NADPH, and sedoheptulose 7-phosphate. Perhaps most concerning was the fate of sphingolipids, the fatty molecules that build cellular membranes. When serine becomes scarce, the enzyme serine palmitoyltransferase substitutes alanine in its condensation reaction, producing 1-deoxysphinganine and, ultimately, a family of atypical lipids called 1-deoxysphingolipids. These compounds are cytotoxic and are known culprits in hereditary sensory and autonomic neuropathy type 1 and diabetic neuropathy. In the serine-deprived mice, the brain alanine-to-serine ratio doubled, and sphingolipidomic analysis revealed a striking accumulation of 1-deoxydihydroceramide species in the hippocampus. Gene expression shifted as well, with circadian rhythm genes such as Bmal1, Cry1, Per2, and Dbp downregulated, and several endothelium-enriched genes, including Cldn5 and Mfsd2a, altered alongside them.

Having established that blood-borne serine matters, the team set out to find which transporter ferries it across the blood-brain barrier. Two endothelial serine transporters, SLC38A5 and SLC1A4, were already known, but deleting them reduces brain L-serine by only 15 to 20 percent and 10 percent respectively, leaving a gap too large to explain the dietary effects. Mining RNA sequencing datasets and a human brain microvessel proteomics database, the researchers identified SLC38A2, also called SNAT2, as a strong candidate. This broad-spectrum neutral amino acid transporter of the system A family was known to move serine, alanine, glutamine, and other substrates in vitro, but its physiological role at the blood-brain barrier had never been defined. To settle the question, the team created mice in which Slc38a2 was selectively deleted from endothelial cells by crossing floxed animals with Tie2-Cre lines, then confirmed the deletion in purified brain microvessels by western blot.

The knockout animals told a remarkably selective story. Untargeted metabolomics of the cortex and hippocampus of eleven-day-old pups revealed that total serine was the only metabolite consistently reduced, with its precursor phosphoserine dropping in parallel; no other SLC38A2 substrate was measurably affected. HPLC analysis confirmed that both hippocampal L-serine and D-serine declined in the knockout pups, while serum amino acid levels remained normal, pointing to a brain-specific effect. When the researchers injected radiolabeled L-serine directly into the heart of the animals and measured its extraction into brain tissue ten seconds later, uptake in the knockout pups fell by 45 percent, while kidney uptake was unchanged. Fluorescein permeability assays confirmed that the barrier itself remained intact, ruling out nonspecific leakage.

Immunoelectron microscopy added a fascinating structural dimension. Gold particles labeling SLC38A2 decorated both the luminal membrane, facing the blood, and the abluminal membrane, facing the brain tissue, of endothelial cells in the neocortex, along with some intracellular signal. This dual placement means the transporter can, in principle, mediate flux in either direction depending on concentration gradients, and the isotope tracing experiments suggested exactly that. Whereas brain uptake of radiolabeled L-serine fell in the knockouts, the extraction of L-glutamine into the brain more than doubled, implying that glutamine transport through SLC38A2 normally runs in the opposite direction, from brain to blood. Stable isotope tracing with doubly labeled serine and uniformly labeled glucose confirmed that the brain’s internal serine synthesis from glucose was not compensating for the loss, and that glycine production was largely spared, with a cortical increase in glycine synthesis from glucose suggesting an adaptive mechanism to maintain glycine levels.

The developmental consequences were visible in the finest architectural details of the brain. Electron microscopic analysis of the dentate gyrus, a region undergoing intense synaptogenesis in the second week of life, revealed that asymmetric, predominantly glutamatergic synapses were selectively vulnerable in the knockout pups. The presynaptic vesicle clusters of these excitatory synapses were 25 percent smaller than in wild-type littermates, with the size distribution shifted toward smaller values, and the postsynaptic density area was similarly reduced. Symmetric, largely GABAergic synapses were unaffected, as was the total number of synapses of either type. Notably, markers of neurons and astrocytes, including NeuN, PSD-95, and GFAP, were unchanged, and bulk RNA sequencing detected only a handful of differentially expressed genes, indicating that the primary defect is metabolic rather than a broad developmental failure. Behavioral testing showed that negative geotaxis, grip strength, and limb suspension reflexes were all preserved, suggesting the pups are resilient to these subtle synaptic changes.

The story grew more nuanced as the mice matured. In juvenile and adult knockout animals, L-serine levels recovered, but D-serine remained persistently low, indicating that luminal SLC38A2 continues to import serine even when circulating concentrations fall after weaning. Meanwhile, glutamine and proline accumulated dramatically in the hippocampi of older knockout mice, implying that in adults, abluminal SLC38A2 normally exports these amino acids from the brain into the blood. Because glutamine and proline both serve as vehicles for ammonia disposal, the authors propose that SLC38A2 helps the aging brain rid itself of excess nitrogen, a function that becomes increasingly important as brain glutamine synthetase activity rises several-fold after the neonatal period. The finding echoes earlier work showing that other blood-brain barrier serine transporter deletions, including SLC38A5 and SLC1A4, also produce excitatory synaptic abnormalities, though the SLC38A2 phenotype proved milder, hinting that additional serine transporters partially compensate.

The therapeutic implications are already taking shape. On one side, blocking luminal SLC38A2 could dampen D-serine synthesis in conditions marked by NMDA receptor overactivation, such as epilepsy and traumatic brain injury, where excess astrocytic D-serine contributes to synaptic damage. On the other, the transporter may represent a metabolic weak point in brain cancer: glioblastoma growth depends on dietary L-serine crossing the blood-brain barrier, and SLC38A2 is overexpressed in these tumors, raising the prospect that selective inhibitors could starve the malignancy of an essential building block. Intriguingly, SLC38A2 expression is reduced in the dorsolateral prefrontal cortex of people with schizophrenia, a disorder closely tied to NMDA receptor hypofunction, opening a possible link between vascular amino acid transport and psychiatric disease. The authors caution that SLC38A2 is a contributor to brain serine import rather than necessarily the sole or dominant transporter, and that future work combining deletions of multiple endothelial serine transporters will be needed to complete the picture. Still, the study rewrites a fundamental rule of neurochemistry: the developing brain is not a self-sufficient serine factory, but an organ that relies on its vascular gatekeepers, and on the diet itself, to build the synapses that make thought possible.

Subject of Research: Role of the endothelial blood-brain barrier transporter SLC38A2 in brain L-serine and D-serine homeostasis and synapse development

Article Title: Endothelial SLC38A2 controls brain D-serine homeostasis and synapse development

Article References: Tiwari, A. K., Lahkar, A., Sajrawi, C., Hussein, Y., Majcher, A., Bonelli, S., Agranovich, B., Abramovich, I., Stern, S., Chaudhry, F. A., Hornemann, T., Radzishevsky, I., & Wolosker, H. (2026). Endothelial SLC38A2 controls brain D-serine homeostasis and synapse development. iScience, 29(10), Article 117487. https://doi.org/10.1016/j.isci.2026.117487

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117487

Keywords: SLC38A2, blood-brain barrier, L-serine, D-serine, NMDA receptors, synapse development, amino acid transport, sphingolipids, metabolomics, glioblastoma, endothelial cells, brain development