Why Folic Acid Protects a Developing Spine: The Vitamin a Connection

In 1991 the Medical Research Council stopped a trial early. Women who had already carried one pregnancy affected by a neural tube defect were given either folic acid or a placebo before trying to conceive again. Among those taking folic acid, recurrence fell by roughly 70 percent, a result large enough that handing anyone else a placebo became indefensible. What followed was fast by the standards of public health. Flour was fortified, supplements were recommended to women who might become pregnant, and rates of spina bifida and anencephaly dropped in country after country.

For the 35 years since, nobody has been able to say what the vitamin was actually doing inside an embryo.

The neural tube is the embryonic structure that becomes the brain and the spinal cord. It forms when a flat sheet of cells rolls up and seals along a seam, and in humans that seam closes about four weeks after conception, often before a woman knows she is pregnant. A failure to seal at the lower end produces spina bifida; a failure at the upper end produces anencephaly, which is fatal. By the time anyone could notice, the window in which anything might have been done has already shut.

The missing mechanism was not merely an intellectual annoyance. The 1991 trial established that folic acid works; it could not say what it works on. Folic acid prevents many neural tube defects but not all of them, and without knowing what it does there is no way to predict which pregnancies it will fail to protect. Nothing could be improved on, only repeated.

A paper published on July 30 in the Proceedings of the National Academy of Sciences proposes an answer, and the answer is a detour. Folic acid, on this account, does not do the protective work itself. It raises production of an enzyme, and that enzyme manufactures the signaling molecule of a different vitamin altogether: retinoic acid, the active form of vitamin A. Take the enzyme away and folic acid protects nothing.

The work comes from Abraham Fainsod's laboratory at the Hebrew University of Jerusalem, with José António Belo's group at NOVA Medical School in Lisbon. Their experimental animal is Xenopus, the African clawed frog, whose embryos develop outside the mother in a dish where a neural tube can be watched closing. To generate defects worth studying, the team knocked down pax3, a gene the tube needs in order to close. The resulting defects are rescuable by folic acid, which is the entire point of the model: the frog reproduces the human phenomenon in miniature and lets you interfere with it.

From there the argument runs in four steps, each of them an intervention rather than an observed correlation. Retinoic acid, or its chemical precursors, rescued the pax3 defects as well as folic acid did. Folic acid was shown to raise expression of aldh1l1, and with it the production of retinoic acid. Knocking out ALDH1L1 activity with CRISPR abolished folic acid's protective effect. And the human version of the enzyme, introduced into frog embryos that lacked their own, converted retinaldehyde into retinoic acid and restored neural tube closure.

That last result is the strongest thing in the paper and the easiest to overread. A human protein did the job. The embryos it did the job in were still frogs.

The cellular story underneath runs against intuition. A tube that fails to close, you would assume, has not grown enough. The paper describes the opposite. When retinoic acid signaling drops, the precursor cells of the neural plate over-proliferate and the neural tube expands pathologically instead of folding shut. ALDH1L1 is what allows folic acid to restore a normal rate of division. The vitamin is not fueling construction so much as keeping it in bounds.

A folate enzyme moonlighting on vitamin A sounds like a category error, and it is not. The day job of ALDH1L1 is 10-formyltetrahydrofolate dehydrogenase, a step in folate metabolism. But it belongs to the aldehyde dehydrogenase superfamily, which also contains ALDH1A1, ALDH1A2 and ALDH1A3, the canonical enzymes that produce retinoic acid in a developing embryo. Relatives in that family sharing a substrate is ordinary biochemistry rather than a surprise.

The study reports one further rescue. Folic acid also corrected neural tube defects caused by alcohol exposure in the same model, which widens the mechanism past a single genetic trigger.

One misreading of this paper would be actively dangerous. Nothing in it is a reason to take vitamin A in pregnancy. Retinoid excess is a documented human teratogen. High intake of preformed vitamin A in early pregnancy was linked to birth defects in cranial-neural-crest tissues in a 1995 New England Journal of Medicine study; fetal retinoid syndrome is a recognized clinical entity; and isotretinoin, the acne drug, carries some of the strictest pregnancy-prevention requirements attached to any prescription medicine, for exactly this reason. Vitamin A deficiency also causes birth defects, and the two patterns overlap. The pathway is dose-critical in both directions, which is precisely why discovering that retinoic acid does the protective work is not an argument for more of it.

Folic acid guidance is unchanged. This paper does not question it, revise it or propose a number. Its single forward-looking sentence is a hedge: the findings suggest opportunities to refine strategies for preventing neural tube defects. That is a prospect, not a recommendation.

The limits are worth listing. One interventional mechanism, in one model organism, from one group, with the independent corroboration that would normally follow still pending. No human, clinical or epidemiological data appear in the paper at all. Sample sizes and rescue percentages per condition are not in the abstract. And the result is not brand new: the same analysis has been publicly readable as a bioRxiv preprint since October 2025. What happened on July 30 was peer review clearing, which counts for something but is not a discovery date.

What the paper does offer is a claim shaped so that other people can attack it. If folic acid protects a neural tube by way of ALDH1L1 and retinoic acid, the same enzyme should matter in a mammal, and removing it from a mouse should break the same protection. That experiment is easy to describe. Someone will run it.