Soy isoflavones wiped out anti-PD1 response in mice despite higher SCFAs

If you take a soy or isoflavone supplement and you are on immune checkpoint immunotherapy, or about to start it, this one deserves ten minutes. A new mouse study found that adding isoflavones to the diet wiped out the tumor-shrinking effect of anti-PD1 therapy, and it did so in animals whose guts were producing more short-chain fatty acids, not fewer. That is a problem for one of the tidiest stories in microbiome oncology, which holds that fiber helps immunotherapy because fiber raises short-chain fatty acids. Here the marker went up and the drug stopped working.

Key Takeaways

• Mice fed soy isoflavones lost their response to anti-PD1 immunotherapy, while mice on a high-fiber diet without isoflavones responded well (p = 0.023).
• The isoflavone-fed mice had the highest fecal short-chain fatty acid levels in the study, which undercuts the assumption that more short-chain fatty acids means a better immunotherapy response.
• Pure genistein, the main soy isoflavone, reproduced the effect on its own in an experiment the team ran twice.
• This is a mouse study. Nobody has tested it in people, and it does not show that eating soy foods harms immunotherapy in humans.

Isoflavones erased the anti-PD1 response in breast tumor models

Mice fed diets containing soy isoflavones stopped responding to anti-PD1 immunotherapy, even though those same mice had the highest gut short-chain fatty acid levels in the experiment. That is the central result of a study published in Frontiers in Immunology by Fabia de Oliveira Andrade, Leena Hilakivi-Clarke and colleagues at Georgetown University and the University of Minnesota.

The team fed four-week-old female mice diets differing in two things: microbiota-accessible carbohydrates (fermentable fibers gut bacteria can actually eat, abbreviated MACs) and isoflavones (soy compounds close enough to estrogen in shape to switch on estrogen receptors). The high-fiber diet behaved as the field predicts. Anti-PD1, a checkpoint inhibitor antibody that releases a brake on tumor-killing T cells, significantly shrank triple-negative breast tumors in those mice, p = 0.023. The same diet with isoflavones added produced no significant tumor reduction at all.

The short-chain fatty acid data are what make this awkward. Short-chain fatty acids are the anti-inflammatory byproducts bacteria release when they ferment fiber. The isoflavone diet did not blunt them; it raised them. Fecal acetic, propionic and butyric acid were all significantly higher in the isoflavone group than in the plain high-fiber group (p = 0.001, 0.02 and 0.02), and all four measured acids exceeded the low-fiber group. The mice with the best fermentation profile were the ones the drug failed in.

Why soy supplement users on immunotherapy are the group to watch

Anyone taking concentrated soy isoflavone supplements during checkpoint immunotherapy for breast cancer is the population this study points at most directly. Checkpoint inhibitors work by blocking the proteins tumors use to send an "off" signal to T cells, and they are now standard care in triple-negative breast cancer, where options have historically been thin.

Isoflavone supplements are commonly taken for menopausal symptoms, which are themselves common in women treated for breast cancer. The National Center for Complementary and Integrative Health notes that isoflavones are structurally similar to estrogen and that soy in supplement form has not been shown to reduce breast cancer risk. The diets here contained 125 to 275 parts per million isoflavones, and one arm used 500 ppm of purified genistein, well above what a mixed diet delivers.

The broader group affected is everyone told to eat more fiber to improve immunotherapy odds. That advice is not overturned here, but its proposed mechanism took a real hit.

Two tumor models, two independent replicates, one isolated compound

The study's credibility rests on isolating isoflavones from fiber, then reproducing the effect in a second cancer model. In the first experiment, mice received allografts of E0771 triple-negative breast cancer cells, with 33, 33 and 32 mice on the low-fiber, high-fiber and high-fiber-plus-isoflavone diets and 8 to 9 per treatment group. A second experiment used single allografts with 9 to 10 mice per group; there anti-PD1 worked on both the low-fiber (p = 0.0009) and high-fiber (p = 0.0182) diets and again did nothing in the isoflavone group.

The cleanest test was the third. Rather than switching whole diets, the researchers added purified genistein to the low-fiber diet, changing one variable. Anti-PD1 worked in the controls and was significantly impaired once genistein was present, p = 0.0202 in the first run and p < 0.0001 when the experiment was repeated. The team then ran a biologically different tumor type: estrogen-receptor-positive mammary tumors initiated with DMBA. There, isoflavones did not merely block the benefit. Anti-PD1 actively promoted tumor growth (p = 0.003).

The immune readout points the same way. Anti-PD1 reduced exhausted CD8 T cells (killer T cells worn down by chronic stimulation) in high-fiber mice but increased them in isoflavone-fed mice, on two independent exhaustion markers (TIM3+, p = 0.013 versus p = 0.002; KLRG1+, p = 0.044 versus p = 0.01).

What this changes for someone managing their own treatment

Nothing here justifies changing a cancer treatment plan, but it does justify a specific conversation. If you take a concentrated soy isoflavone or genistein supplement and you are receiving or scheduled for checkpoint immunotherapy, tell your oncologist the exact product and dose. Supplements are frequently left off medication lists, and this gives a concrete reason to disclose this one.

The distinction between whole soy foods and isolated high-dose supplements is doing a lot of work here, and the study cannot resolve it. Tofu, edamame and soy milk deliver isoflavones at far lower concentrations than a capsule. This paper tested rodent chow formulated to high isoflavone levels and a purified compound, not a serving of tofu.

For readers who track microbiome markers, one more note: a good short-chain fatty acid result on a stool test is not, on this evidence, a reliable proxy for immunotherapy readiness. Here it pointed the wrong way.

The gap between mouse diets and a human plate

The largest limitation is the obvious one: these are mice, and no patient was studied. Mouse tumor models respond to checkpoint blockade differently than human cancers do, and the isoflavone exposures were high and continuous. The authors frame the human question as open, writing that whether patients with advanced estrogen-receptor-positive breast cancer respond better to anti-PD1 on low-estrogen diets "should be studied."

The short-chain fatty acid measurements carry their own caveat. The authors note that despite highly significant differences, the number of biological replicates was low and those profiles "should be interpreted as exploratory findings that warrant confirmation in future, larger-scale validation studies." The work also spanned two institutions with different animal housing, flagged as a source of baseline microbiome variation.

The counterpoint literature is also real. A 2020 JAMA Network Open study of 52 patients found higher fecal short-chain fatty acids associated with longer progression-free survival on nivolumab or pembrolizumab, including acetic acid (HR 0.29) and propionic acid (HR 0.08). This mouse study does not refute that association. It shows the relationship is not simple, and that one dietary compound can raise the metabolite while sinking the response.

Blocking estrogen receptor alpha as the proposed rescue

The researchers' proposed fix is to block the estrogen receptor rather than change the diet. Combining tamoxifen, a drug that blocks estrogen receptor alpha, with anti-PD1 in isoflavone-fed mice cut tumor growth significantly versus control, tamoxifen alone or anti-PD1 alone (p < 0.0001). In the estrogen-receptor-positive model on a low-fiber diet, the combination beat control (p = 0.008), anti-PD1 alone (p = 0.019) and tamoxifen alone (p = 0.008). Gene expression analysis pointed to Th17 differentiation as the most altered immune pathway, with 123 differentially expressed genes at a false discovery rate below 0.05.

That is a testable clinical hypothesis, not a protocol. The work was funded by NIH grant R21CA256428, an exploratory mechanism, and the authors declared no commercial or financial conflicts.

Common Questions

Does this mean I should stop eating tofu and soy milk? No. This study tested rodent diets formulated to 125 to 275 ppm isoflavones and purified genistein at 500 ppm, not ordinary soy foods, and it tested mice rather than people. There is no finding here about dietary soy intake in humans on immunotherapy.

Does fiber still help immunotherapy? In this study the high-fiber diet without isoflavones was the condition where anti-PD1 worked best, so fiber itself looks fine. What failed was the assumption that the short-chain fatty acid level is the thing to optimize.

Why did isoflavones raise short-chain fatty acids but hurt the response? The study does not fully answer this, but it implicates estrogen receptor alpha signaling rather than fermentation. Isoflavones activate that receptor, blocking it with tamoxifen restored anti-PD1 activity, and anti-PD1 increased rather than decreased exhausted CD8 T cells in isoflavone-fed mice.

Should I mention supplements to my oncology team? Yes, and list the specific product and dose. Concentrated isoflavone supplements are the exact exposure this study modeled, and supplements are commonly omitted from medication histories.

References

Fabia de Oliveira Andrade, Kerrie B. Bouker, Melike Ozgul-Onal, et al. (2026). Isoflavones impair anti-PD1 efficacy in breast cancer, regardless of dietary fiber or fecal short-chain fatty acid levels. Frontiers in Immunology. https://doi.org/10.3389/fimmu.2026.1835466