There is a particular kind of waiting room that has no chairs. It belongs to the people who learn, often after a relative's diagnosis or a research study invitation, that they carry a variant in GBA1, the most common genetic risk factor for Parkinson's disease, one that raises risk up to thirtyfold. The number sounds like a sentence. It is not. By the age of 80, only about 20 percent of carriers develop Parkinson's, a figure the authors of a new study put at the centre of their work. Something other than the variant decides. For decades nobody could say what. Now a team led by Elisa Menozzi, with Anthony Schapira and Stanislav Dusko Ehrlich at UCL Queen Square Institute of Neurology, has gone looking in an unglamorous place: the stool sample.

Writing in Nature Medicine (volume 32, pages 2096 to 2106, published online 20 April 2026), the group reports faecal metagenomic sequencing and detailed clinical assessment from 464 people recruited in the UK and Italy: 271 with Parkinson's, 43 carriers of GBA1 variants who had no Parkinson's diagnosis, and 150 healthy controls, more than half of whom were the partners of the patients, a deliberate choice to hold kitchen, water supply and household habits roughly constant. Comparing patients with controls, they found 176 bacterial species that differed, 93 more abundant and 83 less. The species enriched in Parkinson's included Bifidobacterium longum and B. dentium, Streptococcus mutans and Lactobacillus paragasseri; among the most depleted were butyrate producers, Roseburia intestinalis, R. inulinivorans and an unclassified Faecalibacterium, the quiet fermenters that turn dietary fibre into a fuel the gut lining runs on.

The striking part is what happened when the researchers looked at the carriers who were not ill. Rather than asking only which species differed significantly, they measured coherence: whether a species moved in the same direction, in the same people, group after group, using the effect size known as Cliff's delta. Of the 176 species altered in Parkinson's, 142 (81 enriched, 61 depleted) were shifting the same way in the carriers without symptoms, but less far. Roughly a quarter of the gut microbiome, by the authors' estimate, sat in an intermediate state: not the controls' pattern, not the patients', something in between. Within the patient group, those with GBA1 variants (109) and those without (162) looked alike, which argues the pattern tracks the disease process rather than the gene itself.

The clinical correlations are where a family will feel the story. Carriers whose microbiome alterations sat above the group median had more depression, more anxiety and more autonomic trouble than those below it, and all ten carriers who met the Movement Disorder Society research criteria for prodromal Parkinson's fell on the altered side, a split unlikely to be chance. The team then distilled the data into a 16 species tally, the PD Microbiome Score-16: ten species counted as present and up, six as absent and down, giving a range from minus 6 to plus 10. Healthy controls at the top of that range, 27 people, about 18 percent of the control group, had worse depression, anxiety, autonomic symptoms and constipation than the 21 controls at the bottom, along with poorer diet scores and more frequent appendectomy history. Similar microbiome alterations appeared in three independent cohorts from the United States, Korea and Turkey, totalling 638 patients and 319 controls.

Diet ran through all of it, inversely. Using a self reported dietary quality score, the researchers found that the more altered a person's microbiome, the lower the diet score, a trend across the whole cohort and strongest in the patients. Ehrlich put it to Alzforum in the flattest terms available: "In PD patients, clearly, the better their diet, the less the microbiome is altered." That is an association measured at one moment in time, in people whose disease may already have changed how and what they eat, and it is not advice. But it points at the one variable in this story that is neither genetic nor irreversible, and the authors are explicit that whether nutrition can modify the trajectory is a question for trials, not for inference.

Which brings the honest caveat into the frame. This is a cross-sectional study: one stool sample, one clinic visit, no follow-up yet. Nobody in it has been watched over years to see who converts. The carrier group is 43 people, small for a group carrying the study's whole argument. Parkinson's medication, slowed gut transit and constipation itself all reshape gut bacteria, and a snapshot cannot separate cause from consequence. What the design does buy, and what earlier case-control work could not, is a middle group defined by genetics rather than by hindsight: people who are not patients, who share a known risk, and whose guts already look part of the way there.