D-Serine in Schizophrenia Research: An Honest Review

Schizophrenia is where D-Serine has been studied most seriously in humans, and it is also where the honest framing matters most. The logic is compelling: schizophrenia appears to involve underactive NMDA receptors, D-Serine is the co-agonist that helps those receptors open, and people with schizophrenia tend to have lower D-Serine. So topping it up should help — and small trials since 1998 have shown a real but modest and inconsistent signal, mostly on the hard-to-treat negative symptoms and cognition. This page walks through that evidence without hype. The bottom line up front: D-Serine is investigational. It is not an approved treatment for schizophrenia, results are mixed, and no one should self-treat a serious psychiatric illness with an experimental amino acid.

Table of Contents

  1. The NMDA-Hypofunction Hypothesis
  2. Evidence That D-Serine Is Low in Schizophrenia
  3. The Landmark 1998 Add-On Trial
  4. What Later Randomized Trials Actually Showed
  5. Why Negative Symptoms and Cognition Are the Targets
  6. The Genetics: DAOA/G30 and DAAO
  7. The DAAO-Inhibitor Alternative
  8. Related Co-Agonist-Site Approaches
  9. The Honest Bottom Line
  10. Key Research Papers
  11. External Resources
  12. Connections
  13. Featured Videos

The NMDA-Hypofunction Hypothesis

For much of the twentieth century, schizophrenia was framed almost entirely around dopamine, because the drugs that reduced psychosis all blocked dopamine receptors. But dopamine blockade treats positive symptoms (hallucinations, delusions) far better than it treats negative symptoms (blunted emotion, social withdrawal, poverty of speech) or the cognitive deficits that most limit people's lives. Something else had to be going on.

A major clue came from NMDA-receptor antagonists such as ketamine and phencyclidine (PCP). In healthy people, these drugs produce a strikingly schizophrenia-like state: not just perceptual distortions but also negative-symptom-like flattening and cognitive impairment. That observation gave rise to the NMDA-hypofunction hypothesis: that a shortfall in NMDA-receptor signaling contributes to schizophrenia, particularly to the negative and cognitive dimensions that dopamine drugs miss. Balu and Coyle's review lays out how the "glycine modulatory site" — the co-agonist site D-Serine occupies — became a leading target on the strength of this idea (Curr Opin Pharmacol 2015; PMID 25540902).

If NMDA receptors are underactive, one obvious lever is the co-agonist site. You cannot safely flood the brain with glutamate — that risks excitotoxicity — but nudging the co-agonist site with D-Serine might enhance NMDA signaling more gently and selectively. That is the entire rationale for the D-Serine trials that followed.

Back to Table of Contents

Evidence That D-Serine Is Low in Schizophrenia

The hypothesis gained weight when researchers measured D-Serine directly. Hashimoto and colleagues reported that serum D-Serine levels, and the ratio of D-Serine to total serine, were significantly lower in patients with schizophrenia than in controls (Arch Gen Psychiatry 2003; PMID 12796220). This was consistent with the idea that reduced co-agonist availability could be one contributor to NMDA hypofunction in the illness.

It is important to read this carefully. A correlation between low D-Serine and schizophrenia does not prove that low D-Serine causes the illness, nor that raising it will treat it. The abnormality could be a cause, a consequence, or a marker of something else entirely (for example, altered activity of the degrading enzyme DAAO). Still, the finding was replicated in several cohorts and, combined with the genetics discussed below, made D-Serine a serious candidate rather than a long shot.

Back to Table of Contents

The Landmark 1998 Add-On Trial

The pivotal early human test came from Tsai and colleagues, who added D-Serine (about 30 mg/kg/day) to patients' existing antipsychotic medication in a placebo-controlled, double-blind design (Biol Psychiatry 1998; PMID 9836012). The results were encouraging: patients receiving D-Serine as an add-on showed significant improvements in positive symptoms, negative symptoms, and cognitive measures compared with placebo, and the compound was well tolerated over the six-week study.

This trial is rightly cited as a landmark because it was the first solid demonstration that targeting the co-agonist site with D-Serine could produce measurable clinical benefit as an adjunct — on top of standard treatment, not replacing it. It converted the NMDA-hypofunction idea from theory into a testable therapeutic strategy. But a single positive trial, however clean, is a starting point, not a conclusion. The obvious next step was replication in larger and more varied populations — and that is where the story became more complicated.

Back to Table of Contents

What Later Randomized Trials Actually Showed

Subsequent trials produced a genuinely mixed picture — the honest heart of this page.

Pooled analyses of glutamatergic add-on agents (D-Serine, glycine, sarcosine, D-cycloserine) have generally concluded that there is a small average benefit for negative and total symptoms, but with meaningful heterogeneity between studies and a real risk that publication bias inflates the apparent effect. The most recent methodological work explicitly frames the central unresolved issue as one of dose and target engagement — that many trials may simply not have delivered enough D-Serine to the brain to test the hypothesis fairly (Sehatpour et al., Biol Psychiatry 2025; PMID 39218136). That is a reason for continued careful research, not a reason to conclude the drug "works."

Back to Table of Contents

Why Negative Symptoms and Cognition Are the Targets

The reason researchers keep returning to D-Serine despite mixed results is that it aims at the part of schizophrenia that current medications treat worst. Antipsychotics are reasonably effective against positive symptoms but do little for the negative symptoms — social withdrawal, reduced motivation, emotional flattening — and little for the cognitive impairments in memory, attention, and processing speed that most strongly predict whether someone can work, study, and live independently.

Because the NMDA co-agonist site is mechanistically tied to plasticity and learning, it is a rational target for exactly these treatment-resistant domains. Even a modest, reliable improvement in negative symptoms or cognition would be clinically valuable, because there is so little else on offer for them. This is why a "small effect" is not automatically an uninteresting effect in this context — but it must be a real, replicated small effect, delivered safely, and that bar has not yet been cleared for regulatory approval.

Back to Table of Contents

The Genetics: DAOA/G30 and DAAO

Human genetics gave the D-Serine hypothesis independent support. Genes controlling D-Serine metabolism have repeatedly surfaced in schizophrenia association studies. The gene DAO encodes D-amino acid oxidase, the enzyme that degrades D-Serine, and DAOA/G30 (originally named "D-amino acid oxidase activator," G72) sits at a locus linked to the illness. Variants that would be expected to increase DAAO activity — and thus lower D-Serine — have been associated with elevated risk (Shinkai et al., Neuromolecular Med 2007; PMID 17627036; Corvin et al., Am J Med Genet B 2007; PMID 17492767).

Animal genetics reinforced the theme from the synthesis side. Balu and colleagues showed that mice engineered to lack serine racemase — and therefore unable to make normal amounts of D-Serine — display a convergence of schizophrenia-relevant abnormalities, tying multiple risk pathways back to the D-Serine system (PNAS 2013; PMID 23729812). The convergence of human genetics, biochemistry, and animal models on this one pathway is the strongest part of the D-Serine story — stronger, honestly, than the clinical-trial results themselves.

Back to Table of Contents

The DAAO-Inhibitor Alternative

If the problem in some patients is that D-Serine is degraded too quickly by DAAO, an elegant alternative to administering D-Serine directly is to block DAAO and let the brain's own D-Serine rise. The most-studied clinical example is sodium benzoate (a common food preservative that happens to inhibit DAAO). Lin and colleagues reported that adding sodium benzoate to clozapine improved symptoms in treatment-resistant schizophrenia (Biol Psychiatry 2018; PMID 29397899).

The DAAO-inhibitor strategy is attractive partly for safety reasons: raising endogenous D-Serine within physiological control loops may avoid the very high exogenous doses that drive nephrotoxicity in animals. Interestingly, sodium benzoate has itself been shown to attenuate D-serine-induced kidney injury in rats (see the safety page), which reflects the same DAAO biology from the toxicology side. Even so, sodium benzoate is likewise investigational for psychiatric use and is not a validated, approved treatment.

Back to Table of Contents

D-Serine is one of several molecules that target the NMDA co-agonist site, and the family's collective track record is part of the honest picture:

The pattern across this whole class is consistent: a strong mechanistic rationale, encouraging small studies, and repeated difficulty converting that into robust, approvable clinical benefit. D-Serine fits that pattern.

Back to Table of Contents

The Honest Bottom Line

Putting it together plainly:

Back to Table of Contents

Key Research Papers

  1. Tsai G, Yang P, Chung LC, et al. (1998). D-serine added to antipsychotics for the treatment of schizophrenia. Biol Psychiatry. — PMID 9836012
  2. Hashimoto K, Fukushima T, Shimizu E, et al. (2003). Decreased serum levels of D-serine in patients with schizophrenia. Arch Gen Psychiatry. — PMID 12796220
  3. Heresco-Levy U, Javitt DC, Ebstein R, et al. (2005). D-serine efficacy as add-on pharmacotherapy to risperidone and olanzapine for treatment-refractory schizophrenia. Biol Psychiatry. — PMID 15780844
  4. Kantrowitz JT, Malhotra AK, Cornblatt B, et al. (2010). High dose D-serine in the treatment of schizophrenia. Schizophr Res. — PMID 20541910
  5. Balu DT, Coyle JT (2015). The NMDA receptor 'glycine modulatory site' in schizophrenia: D-serine, glycine, and beyond. Curr Opin Pharmacol. — PMID 25540902
  6. Balu DT, Li Y, Puhl MD, et al. (2013). Multiple risk pathways for schizophrenia converge in serine racemase knockout mice. Proc Natl Acad Sci USA. — PMID 23729812
  7. Lin CH, Lin CH, Chang YC, et al. (2018). Sodium Benzoate, a D-Amino Acid Oxidase Inhibitor, Added to Clozapine for the Treatment of Schizophrenia. Biol Psychiatry. — PMID 29397899
  8. Shinkai T, De Luca V, Hwang R, et al. (2007). Association analyses of the DAOA/G30 and D-amino-acid oxidase genes in schizophrenia. Neuromolecular Med. — PMID 17627036
  9. Corvin A, McGhee KA, Murphy K, et al. (2007). Evidence for association and epistasis at the DAOA/G30 and D-amino acid oxidase loci in schizophrenia. Am J Med Genet B. — PMID 17492767
  10. Kuppili PP, Nebhinani N, Gupta N (2021). Efficacy of adjunctive D-Cycloserine for the treatment of schizophrenia: a systematic review and meta-analysis. J Neural Transm. — PMID 33439362
  11. Sehatpour P, Javitt DC, Kantrowitz JT (2025). Finding the Right Dose: NMDA Receptor-Modulating Treatments for Cognition and Plasticity. Biol Psychiatry. — PMID 39218136

PubMed Topic Searches

Back to Table of Contents

External Resources

Back to Table of Contents

Connections

Back to Table of Contents