In 2021, the German Research Foundation (DFG) quietly updated its animal nutrition guidelines. The change was narrow: mouse diets used in DFG-funded studies should, where possible, be purified-ingredient chow rather than traditional grain-based formulas. The goal was to reduce an uncontrolled source of variation in microbiome experiments. But a recent survey of 22 DFG-funded mouse microbiome papers published between 2021 and 2025 found that 14 of them had switched diets after the policy was announced. The other eight stayed on grain-based chow. The result is a field split in two, with data sets that are increasingly difficult to compare.

A Single Policy Change Reshaped Gut Microbiome Research

The DFG's new standard was not a ban on grain-based diets. It was a recommendation, but one with teeth: grant applicants who chose grain-based chow had to justify the choice in their proposals. Reviewers, sensitized to reproducibility concerns, often pushed back. Over time, the recommendation became a de facto requirement for many labs. The policy's architects argued that purified diets—made from refined ingredients like casein, cornstarch, and soybean oil—would eliminate batch-to-batch variation in plant fiber content that could confound microbiome results.

In practice, the shift was dramatic. Among the 22 papers surveyed, the 14 that adopted purified chow reported different baseline microbial compositions than comparable studies still using grain-based feed. For example, levels of Lactobacillus and Clostridium clusters shifted in ways that could alter the interpretation of experimental treatments. Two of the studies explicitly noted that their results might not replicate under the other diet, a caveat that undermines the very reproducibility the policy was meant to strengthen.

The survey, conducted by a team at the University of Freiburg, was not designed to assign blame. The authors described the situation as an “unintended fragmentation” of the research landscape. One co-author, microbiologist Elena Voss, told this site that the DFG's intention was sound, but the outcome shows how a well-meaning policy can have ripple effects that no one anticipated.

Why Grain-Based Chow Was the Default for Decades

Grain-based chow—made from wheat, corn, oats, and other plant materials—has been the standard laboratory mouse diet for almost a century. It is cheap, nutritionally complete, and mimics the complex fiber profile that mouse gut microbes evolved to digest. Mice in the wild eat seeds and plant matter, and their microbiomes have adapted to ferment those fibers into short-chain fatty acids that support gut health and immune function.

Purified diets, by contrast, use isolated nutrients: casein as a protein source, cornstarch as carbohydrate, and cellulose as a fiber placeholder. They are more consistent across batches—a bag of purified chow from 2022 should be chemically identical to one from 2023—but they lack the diversity of indigestible plant compounds that shape the microbial community. Researchers who stayed with grain-based chow argued that the natural diet was more ecologically relevant. “If you want to understand how the microbiome works in a living animal, you need to feed it something like what it would eat,” said animal physiologist Markus Brandt of the Max Planck Institute for Evolutionary Biology, who was not involved in the survey.

Cost also played a role. Grain-based chow is roughly half the price of purified diets per kilogram, a nontrivial difference for labs housing hundreds of mice over multiyear studies. But the DFG policy did not include additional funding for diet costs, so some labs absorbed the expense or reduced animal numbers. Others simply switched and hoped the effect on their data would be small.

The DFG's Rationale: Statistical Power vs. Realism

The DFG's decision was rooted in the reproducibility crisis that has haunted preclinical research for over a decade. In microbiome studies, uncontrolled variation from diet can swamp the signal from an experimental treatment, forcing researchers to use larger sample sizes to detect effects. Purified diets reduce that uncontrolled variation, meaning a study with fewer animals can still reach statistical significance. For an agency that funds hundreds of projects, the appeal of needing fewer animals per study is clear: it stretches grant dollars further and reduces the number of animals used in research, an ethical consideration.

But critics argue that reducing variance by simplifying the diet does not make the results more robust—it makes them more brittle. A treatment effect seen in mice fed purified chow may disappear when the animals are returned to a grain-based diet, which is closer to what they would encounter in nature or in other labs. “You’re optimizing for statistical significance, not for biological generalizability,” said computational biologist Yuki Tanaka of the University of Tokyo, who has written about the trade-off. “The DFG policy may produce cleaner data, but cleaner data is not always better data.”

The debate echoes earlier controversies in neuroscience and pharmacology, where standardized housing and handling protocols improved within-lab reproducibility but made cross-lab comparisons harder. In those fields, the solution was not a single mandate but a set of community-agreed reporting standards. The DFG's approach was more top-down, and the survey suggests it may have been too blunt an instrument.

Systematic Survey Reveals Extent of Diet Shift

The survey, published as a preprint in early 2026, examined 22 DFG-funded mouse microbiome studies that appeared in peer-reviewed journals between 2021 and 2025. The research team identified the diet used in each study by reading the methods sections and, when necessary, contacting the authors. Of the 22 studies, 14 used purified chow, 7 used grain-based chow, and 1 used a mix. Before 2021, the same group of labs had used grain-based chow in roughly 70 percent of their microbiome work, according to the authors' informal records.

The impact on microbial community structure was not subtle. In the 14 switched studies, the relative abundance of Lactobacillus species dropped by an average of about 40 percent compared with earlier grain-based work from the same labs. Clostridium clusters XIVa and IV, which are important for butyrate production, also shifted. The authors of the survey did not claim that one diet was better, only that the two diets produced different baselines. “If you don't know which diet a study used, you cannot meaningfully compare its results to another study,” Voss said.

The survey has not yet been formally peer-reviewed, but several independent researchers contacted for this story said its findings align with their own unpublished observations. “I’d been noticing that my own data looked different after we switched diets,” said Brandt, who uses grain-based chow. “The survey confirms that it's a field-wide issue, not just my lab.”

Lost Comparability Threatens Meta-Analysis Efforts

Meta-analysis—the statistical combination of results from multiple studies—is a cornerstone of evidence-based science. In microbiome research, public databases like the European Nucleotide Archive hold thousands of 16S rRNA gene sequences from mouse experiments, many of which are used to build reference maps of the healthy gut microbiome. But if those sequences come from animals fed different diets, the maps may be unreliable. A treatment effect that appears significant when all studies are pooled may be driven by diet differences rather than the treatment itself.

Some research groups have begun to retroactively flag diet as a variable. A team at the University of Copenhagen recently re-analyzed a set of 15 published mouse microbiome data sets and found that diet explained roughly 12 percent of the variation in microbial beta-diversity, comparable to the effect of a typical drug treatment. The team's leader, bioinformatician Sofie Lindberg, said the finding was “sobering” because it suggests that many published comparisons may be confounded.

Funding agencies rarely adjust budgets for the added cost of maintaining two diet groups, which is one way to preserve comparability. A lab that wants to run a study on purified chow while also keeping a grain-based reference cohort must either cut other expenses or seek supplementary grants. Few have done so. The DFG has not announced any plans to revise its policy, but the survey's authors hope the agency will at least acknowledge the fragmentation it has caused.

Concrete Example: A Study on Antibiotic-Induced Dysbiosis

To illustrate the practical impact, consider a 2023 DFG-funded study that investigated how antibiotics alter the mouse gut microbiome. The researchers used purified chow and reported a dramatic drop in microbial diversity after treatment. However, a previous study from the same lab, using grain-based chow, had shown only a moderate decline. The discrepancy was initially attributed to different antibiotic regimens, but the diet change likely played a role. The purified chow, with its simpler fiber profile, may have left the microbiome more vulnerable to disruption. This example underscores how diet can interact with experimental interventions, complicating interpretation.

Another example comes from a 2024 study on probiotics. Researchers feeding purified chow observed that a Lactobacillus supplement colonized the gut more effectively than in earlier grain-based work. But they noted that the effect might be an artifact of the diet: the purified chow lacked the complex carbohydrates that native microbes compete for, giving the probiotic an advantage. Without a grain-based control group, the generalizability of the finding remains uncertain.

Trade-Offs and Counter-Arguments

Proponents of purified diets counter that the improved reproducibility within a lab is a genuine benefit. For mechanistic studies—where the goal is to dissect a molecular pathway—reducing noise is paramount. A purified diet allows researchers to isolate the effect of a single gene or drug without the confounding influence of variable plant fibers. “If you're testing a drug target, you want to control everything you can,” said pharmacologist Anna Richter of the University of Munich, who uses purified chow in her cancer research. “The microbiome is just one more variable to control.”

However, even within a lab, the switch to purified chow may introduce new sources of variation. Some batches of purified diets have been found to contain trace contaminants or to differ in micronutrient levels, though manufacturers claim tight quality control. Moreover, the long-term health effects of purified diets on mice are not fully understood. Some studies suggest that mice on purified chow develop higher rates of obesity and metabolic syndrome, which could themselves alter the microbiome. Thus, the pursuit of standardization may inadvertently create new confounders.

Another counter-argument is that the field of microbiome research is still young, and diversity in methods can be a strength. Different diets may reveal different aspects of microbial ecology, much like different model organisms reveal different biological principles. A single standard could stifle discovery. “We don't want to put all mice on the same diet and then find out we've been studying a laboratory artifact for 20 years,” warned ecologist David Chen of Stanford University, who studies wild mice.

Practical Takeaways for Preclinical Microbiome Studies

Researchers interviewed for this story offered several concrete recommendations. First, diet composition should be reported in full detail, including the manufacturer, product number, and ingredient list. Many journals already require this, but enforcement is inconsistent. Second, studies should justify their diet choice in relation to the research question: if the goal is to model a human disease that involves a high-fiber diet, grain-based chow may be more appropriate; if the goal is to test a precise molecular mechanism, purified chow may be better.

Third, meta-analysts should filter data sets by diet type explicitly, rather than assuming that all mouse chow is equivalent. Some databases already allow users to tag studies by diet, but the practice is not universal. Fourth, grant agencies should anticipate that methodological mandates can have unintended consequences. A policy that seems sensible on paper may create more problems than it solves if it is imposed without community input.

Finally, several researchers argued for community-level standards rather than top-down mandates. A consortium of microbiome labs could agree on a set of reference diets—one grain-based, one purified—that all members use for baseline measurements, allowing cross-study comparisons while preserving each lab's freedom to choose the diet that best fits its experimental question. Such a system would be voluntary, but it might achieve the DFG's goal of reducing uncontrolled variation without fragmenting the literature.

The Freiburg survey is a reminder that even well-intentioned policies can reshape a field in ways that are hard to reverse. As Voss put it, “We are not saying the DFG was wrong. We are saying that when you change one variable in a complex system, you should expect the system to respond in ways you didn't predict.” The microbiome, it turns out, is not the only thing that responds to a change in diet.