A Gut Molecule Was Tied to Alzheimer’s. In People It’s a Pattern; the Cause Was Shown in Mice
This weekend’s headline said the gut makes a molecule that raises Alzheimer’s risk. We read the Wisconsin paper, its supplementary tables and its source data. The human evidence is an association in 1,196 blood samples, the cause-and-effect evidence is in male mice, and the 43 percent everyone repeated is not a count of people.
A June 2026 Nature Communications study from the University of Wisconsin-Madison linked imidazole propionate (ImP), a molecule gut bacteria make from the amino acid histidine, to Alzheimer's-related measures. In 1,196 cognitively unimpaired adults, each unit of blood ImP went with a cognitive composite score 0.38 to 0.43 standard deviations lower and with higher pTau-217 and neurofilament light; the authors say the cross-sectional design cannot show which came first. A one-variant genetic analysis estimated an odds ratio of 1.16 for Alzheimer's per standard deviation of ImP. The cause-and-effect experiments were in male mice, where ImP in drinking water increased amyloid plaques. The paper describes no clinical test and no way of lowering ImP in people.
Every couple of years, a group of people in Wisconsin give up about five hours to be measured. Blood is drawn, questionnaires are filled in, and a tester runs through tasks that include naming animals and listing words that begin with C, F and L. The Wisconsin Registry for Alzheimer’s Prevention has held these visits since late 2001. A 2018 review of the registry put the average age at enrolment at 54 and reported that 73 percent of the volunteers had a parent with Alzheimer’s dementia.
Blood from 1,196 people in that registry and the university’s Alzheimer’s Disease Research Center is behind a headline that travelled this weekend: that the gut may make a molecule that raises Alzheimer’s risk. The paper, from a University of Wisconsin-Madison team with Barbara Bendlin and Federico Rey as senior authors, appeared in Nature Communications on June 26; the university’s release followed on August 12, and ScienceDaily ran it on October 3. We read the full text, its supplementary tables, its source data and the public genetics database behind its most repeated number. The short answer: in people, higher blood levels of a bacterial compound called imidazole propionate (ImP) went with lower thinking-test scores and higher levels of two blood markers used in Alzheimer’s research. The cause-and-effect experiments were in male mice. And the 43 percent in the coverage is not a count of people in the study.
The molecule: What imidazole propionate is
ImP is what certain gut bacteria leave behind when they break down histidine, an amino acid, without oxygen. Histidine is converted to a compound called urocanate, and bacteria carrying an enzyme called urocanate reductase turn that into ImP, which can pass into the blood. Cutting out histidine is not a simple fix, the release notes: the body needs it, and it is in many common foods.
The molecule has a record. A 2018 paper in Cell from Fredrik Bäckhed’s group in Gothenburg found more of it in people with type 2 diabetes and reported that it impaired glucose tolerance in mice. A 2020 study of 1,958 Europeans found no link to how much histidine people ate, and more ImP in people with the low-diversity gut pattern called Bacteroides 2, the pattern in the Danish inflammation study we read in September. In 2025, a paper in Nature tied ImP to atherosclerosis in two human cohorts and in mice.
The people: What 1,196 blood samples showed
All 1,196 were classed as cognitively unimpaired; the paper’s flow diagram shows 3,643 people enrolled, these being the ones with blood-chemistry data. Their average age was 61, 69 percent were women and 94 percent were white. ImP was measured in fasting plasma on a commercial research platform that reports relative amounts, not concentrations, the supplementary methods say.
After adjusting for age, sex, body mass index and the APOE ε4 risk gene, each one-unit step up in ImP (the cohort spanned about two units on the paper’s log scale) went with a score 0.38 to 0.43 standard deviations lower across three versions of a composite of learning, recall and executive-function tests. Higher ImP also went with higher blood levels of pTau-217, a tau marker, and neurofilament light (NfL), a marker of nerve-cell damage. Among the 224 people who had given spinal fluid, NfL was the only marker that tracked with ImP; the amyloid and tau measures did not. Brain-glucose scans of 168 people showed no significant overall link.
| The evidence | What was found | What it cannot show |
|---|---|---|
| People, blood: 1,196 adults with no cognitive impairment, average age 61. ImP in fasting plasma, thinking tests, three blood markers | Per unit of ImP, a composite test score 0.38 to 0.43 standard deviations lower; pTau-217 and NfL higher; GFAP not significant | Which came first. The authors call the human design cross-sectional |
| People, spinal fluid (224) and brain-glucose scans (168) | Only NfL tracked with ImP; amyloid and tau measures did not; no overall link on scans | Much smaller groups than the blood analysis |
| Genetics: published data from 5,941 Finnish men and an Alzheimer’s study of 111,326 cases and 677,663 controls. One DNA variant, rs7969761 | T version: higher ImP, and an Alzheimer’s odds ratio of 1.02 per copy. Mendelian randomization: odds ratio 1.16 (95% confidence interval 1.05 to 1.29) per standard deviation of ImP | It rests on one variant; the gene and mechanism at that spot are not yet identified |
| Mice: males, 8 per group, two Alzheimer’s-model strains. About 800 micrograms of ImP a day in drinking water | Amyloid plaques about 75 against 45; reactive astrocytes 97 against 13; more tau phosphorylation | Whether the same happens in people, or in female mice |
Rey says in the release that people with the highest levels declined much faster. The paper compares the top and bottom quarters of ImP (223 and 242 people at the first visit), each tested up to three times. The high group started lower, and its scores fell more steeply with age, by an extra 0.011 to 0.018 standard deviations a year. The groups differed in ways the models adjusted for: 48 percent of the high group were men, against 9 percent of the low group. The abstract describes the human links as holding “both cross-sectionally and longitudinally”; the limitations section says “the cross-sectional design for the clinical associations precluded assessment of temporal or causal relationships in humans” and that the modelled trajectories “are based on statistical assumptions and require validation in longitudinal cohorts.”
In stool from 294 people in a companion study, the paper names 12 species carrying what looks like the gene for urocanate reductase; only one, Streptococcus pasteurianus, tracked with ImP in blood.
The gene: Where the 43 percent comes from
The release describes a genetic variation linked to higher ImP as “present in about 43% of the people in the study.” The paper’s genetic analysis uses published results from two other projects, not genotypes from the 1,196 volunteers. A 2022 study of Finnish men had tied one DNA variant, rs7969761, to blood ImP: the T version went with higher levels. The paper gives that version’s frequency as 43 percent among people of European ancestry. In the Finnish study’s public results browser, 0.43 is listed as the variant’s minor-allele frequency, and the allele count beside it puts the T version on about 57 percent of gene copies in the 5,941 men measured, which would make it the more common version. Either way, the number counts copies of a DNA letter among Finnish men, not people in Wisconsin.
The second project is a genetic study of Alzheimer’s with 111,326 cases, some counted because a parent had dementia, and 677,663 controls. There, each copy of the T version went with an odds ratio of 1.02, or 2 percent higher odds. Combining the two, a method called Mendelian randomization put the odds 16 percent higher for each standard-deviation rise in genetically predicted ImP. The paper presents this as genetic evidence of cause. It rests on one variant, and the authors write that the gene and mechanism at that spot still need to be identified. Since nobody in the blood analysis had dementia, the word “risk” in the headline rests here.
The mice: Where cause and effect was tested
The experiments that can show cause were done in mice. Two strains engineered to develop Alzheimer’s-like changes, one amyloid plaques and one tau pathology, drank water carrying about 800 micrograms of ImP a day until they were six months old. Every animal was male, eight per group. In the amyloid strain, treated mice averaged about 75 plaques in the brain region imaged against 45 in untreated littermates, by the paper’s source data. In the tau strain, the median count of reactive astrocytes (support cells that respond to injury) was 97 against 13. In ordinary mice, swallowed ImP turned up in brain tissue, and more of a tracer dye leaked through the blood-brain barrier.
The human evidence is a pattern in blood samples. The cause-and-effect evidence is in male mice. The headline joined the two.
The limits: What the paper says it cannot show yet
The authors list the gaps. Their human cohort “consisted primarily of cognitively unimpaired individuals, which may limit the generalizability” of the findings. The ImP values “do not provide absolute concentrations.” Only male mice were used, and ImP delivered in drinking water rather than made in the gut “may alter ImP kinetics and host responses.” The paper describes no clinical test for ImP and tried no way of lowering it in people. If memory changes in you or someone you love are on your mind, that is a conversation for a doctor.
Funding came from the Wisconsin Partnership Program, the National Institutes of Health and the U.S. Department of Agriculture, with part support from the Novo Nordisk Foundation and grants to individual authors. The competing-interests statement lists co-author Bäckhed as a co-founder and shareholder of Implexion Pharma AB and Roxbiosens Inc. who receives research funds from BioGaia AB and Novo Nordisk A/S; co-author Katharina Beck as an Implexion Pharma shareholder; Bendlin as having consulted for Cognito Therapeutics, New Amsterdam Pharma and Merry Life Biomedical; and co-authors Kaj Blennow and Henrik Zetterberg as advisers to numerous drug and diagnostics companies and co-founders of Brain Biomarker Solutions in Gothenburg AB.
Also this weekend: The other gut-brain headline
A second gut-brain story ran on ScienceDaily on October 4. The UCLA Health study, published in eBioMedicine on September 9, built a brain ageing index from resting-state brain scans of 1,462 healthy adults aged 18 to 65 in three cohorts. Older-looking brains went with poorer working memory and executive function and more depressive symptoms. The gut link comes from one cohort, where stool data were analysed for 182 of 344 people, and the authors write that a cross-sectional design cannot establish cause or timing. It was funded by the National Institutes of Health, and three of its 17 authors declare industry ties. The University of Wisconsin-Madison, UCLA Health and the papers’ authors had no involvement in this article. It is educational, describes what the cited papers, data files and releases say, and is not medical advice.