A single enzyme produced by a common gut bacterium can push a mouse's behavior in opposite directions, depending on which standard anxiety test the animal faces. The finding, posted as a preprint on bioRxiv in early 2025, illustrates a persistent methodological headache in behavioral neuroscience: the same intervention can look anxiolytic in one assay and anxiogenic in another, leaving researchers to wrestle with what the animal actually feels.

The enzyme, beta-glucuronidase from Bacteroides fragilis, is not a drug candidate. It is a metabolic workhorse that cleaves glucuronidated compounds in the intestine, freeing molecules that can enter the bloodstream and, potentially, the brain. When the research team gave purified enzyme to mice via oral gavage for 14 days, they saw a roughly 30 percent increase in time spent in the open arms of the elevated plus maze — a classic sign of reduced anxiety. But in the open field test, the same mice spent about 20 percent less time in the center of the arena, a pattern that typically signals heightened anxiety.

Two assays, two conclusions. Which one, if either, reflects a real change in the mouse's emotional state? The answer matters not only for this enzyme but for the broader field of gut-brain axis research, where single behavioral tests are often used to claim that a microbe or metabolite alters anxiety. This study, led by a team at the University of California, San Francisco, and including collaborators from the University of Cork, suggests that without a battery of tests and careful controls for locomotion, such claims may be on shaky ground.

The Two Assays: What They Measure and What They Miss

The elevated plus maze (EPM) and the open field test (OFT) are among the most widely used behavioral assays in rodent anxiety research. Both rely on the rodent's natural aversion to open, brightly lit spaces — a trait that is thought to parallel aspects of human anxiety. In the EPM, the apparatus consists of two open arms and two enclosed arms raised above the floor. Rodents typically spend more time in the enclosed arms; an increase in open-arm time is interpreted as reduced anxiety. The OFT, by contrast, is a simple square or circular arena. Rodents normally prefer to stay near the walls (thigmotaxis); more time in the center is taken as lower anxiety.

Despite their superficial similarity, the two tests capture different aspects of behavior. The EPM specifically measures approach-avoidance conflict — the animal must decide between the safety of the enclosed arms and the potential exploration of open arms. The OFT, on the other hand, is more sensitive to general locomotor activity and exploratory drive. A mouse that is simply more active might spend more time in the open arms of the EPM and also more time in the center of the OFT — but that is not what the researchers observed. Instead, the enzyme increased open-arm time while decreasing center time, a dissociation that cannot be explained by a simple change in overall activity.

To disentangle these effects, the team analyzed locomotion separately. They found that the enzyme-treated mice showed a modest increase in total distance traveled in both assays, but the pattern of movement differed. In the EPM, the increased locomotion was concentrated in the open arms, whereas in the OFT, the mice moved more along the walls and avoided the center. This suggests that the enzyme's effect on anxiety-like behavior is context-dependent, possibly mediated by differences in the sensory or motivational features of the two tests.

One key difference is the presence of enclosed arms in the EPM, which provide a clear safe haven. The OFT lacks such a compartment; the walls are the only refuge, and the center is uniformly exposed. This difference may engage distinct neural circuits. For instance, the EPM is more sensitive to drugs that act on the GABA-A receptor, like benzodiazepines, which robustly increase open-arm time. In contrast, the OFT is more responsive to serotonergic manipulations. A meta-analysis of over 200 studies found that benzodiazepines increase open-arm time by roughly 50–80 percent but have smaller effects on OFT center time, often around 20–30 percent, and sometimes even decrease it at higher doses due to sedation. This suggests that the two tests are not interchangeable and that the enzyme's effect may resemble a drug that differentially modulates these systems.

Another factor is the duration of the test. The EPM is typically a 5-minute trial, while the OFT often lasts 10–30 minutes. The enzyme's effect on locomotion might habituate over time, leading to different outcomes. The researchers recorded behavior in 5-minute bins but did not report time-course data in the preprint. Future analyses could reveal whether the dissociation grows or shrinks with prolonged exposure.

Why the Same Enzyme Produces Opposite Results

One plausible explanation is that beta-glucuronidase alters the metabolism of endogenous compounds that affect both anxiety and locomotion. The enzyme is known to reactivate glucuronidated metabolites, including neurotransmitters and steroid hormones, that have been inactivated by the liver. By cleaving the glucuronic acid group, the enzyme can increase the concentration of active molecules in the gut and, potentially, in the brain. Among the candidates are catecholamines like dopamine and norepinephrine, which influence both mood and movement.

Another possibility is that the enzyme acts indirectly through the gut-brain axis, for example by modulating the immune system or the vagus nerve. The gut microbiome is known to produce a wide range of neuroactive compounds, and beta-glucuronidase could be a key regulator of their bioavailability. The researchers measured several metabolites in the blood and brain, but the preprint does not report a clear mechanism. Instead, the study highlights the complexity of interpreting behavioral data when the underlying biology is not fully understood.

The contradictory results also underscore a well-known but often ignored principle in behavioral neuroscience: no single assay is a pure measure of anxiety. Each test is influenced by confounding variables such as locomotion, exploration, stress reactivity, and even circadian timing. The EPM and OFT are both sensitive to changes in anxiety, but they are not interchangeable. In fact, a meta-analysis of published studies found that the correlation between open-arm time in the EPM and center time in the OFT is only around 0.3 to 0.4, meaning that they share less than 20 percent of their variance. That leaves plenty of room for dissociations like the one seen here.

To further complicate matters, the enzyme might affect not only anxiety but also other behavioral domains such as risk assessment, decision-making, or memory. For example, the EPM involves a conflict between exploration and fear, which also engages cognitive processes. The OFT may be more sensitive to changes in general arousal. Without additional tests, it is difficult to pinpoint which domain is primarily affected. The researchers could have included a test of locomotion independent of anxiety, such as the rotarod or a simple open field with no center-wall distinction, to separate motor effects from emotional ones. They did measure total distance, but that is a coarse metric.

A counter-argument to the enzyme's direct effect is that the behavioral changes could be due to peripheral discomfort rather than anxiety. The enzyme might cause gastrointestinal distress, leading to reduced center time in the OFT due to malaise, while the EPM's enclosed arms provide a comfortable resting place. However, the researchers monitored body weight and food intake and found no differences, suggesting the mice were not overtly sick. Still, subtle visceral sensations cannot be ruled out.

Lessons for Gut-Brain Axis Research

The gut-brain axis has become a hot topic in neuroscience, with hundreds of papers each year claiming that a particular probiotic, prebiotic, or microbial metabolite alters anxiety or depression in rodents. But many of these studies rely on a single behavioral test, often the OFT or the EPM, without controlling for locomotion or using a second confirmatory assay. The present study suggests that such conclusions may be premature. If a single enzyme can produce opposite results in two standard tests, then a positive result in one test should be interpreted with caution.

The researchers themselves are careful not to overclaim. They note that the effect size is moderate and that the behavioral changes are within the range seen in other studies. They also point out that the enzyme was given at a relatively high dose, and it is unclear whether similar effects would occur with lower doses or with the endogenous levels produced by the gut microbiome. Nevertheless, the study serves as a reminder that the gut-brain axis is a complex system, and behavioral readouts are only as reliable as the assays used to measure them.

To address these issues, some researchers have advocated for a "behavioral battery" approach, where multiple tests are used in the same animals to capture different dimensions of anxiety-like behavior. For example, the EPM, OFT, light-dark box, and novelty-suppressed feeding test each probe different aspects of approach-avoidance conflict, and their combined results can provide a more comprehensive picture. In the present study, the team used only two tests, but they plan to add more in future work, including the light-dark box and the marble burying test, which measures a different form of anxiety-related behavior.

Another approach is to use factor analysis to extract latent variables from multiple tests. For instance, a 2023 study by researchers at the University of Zurich combined data from the EPM, OFT, light-dark box, and elevated zero maze in over 500 mice. They identified three factors: anxiety-like behavior, exploration, and locomotion. The enzyme in the current study might load differently on these factors, explaining the dissociation. The authors did not perform such analysis, but it could be a valuable addition in future work.

The field could also benefit from standardized protocols. A 2022 survey of 100 gut-brain studies found that 40 percent used the OFT only, 25 percent used the EPM only, and only 15 percent used both. Moreover, the parameters — arena size, lighting, acclimation time, test duration — varied widely. This heterogeneity makes cross-study comparisons difficult and may contribute to the reproducibility crisis. The present study used standard protocols: a 5-minute EPM and a 30-minute OFT, with illumination around 100–150 lux. But even small changes, such as lighting level, can alter results. For example, brighter light (above 200 lux) increases avoidance of open areas in both tests, potentially masking drug effects. The researchers should report lighting precisely to aid replication.

Trade-offs and Limitations

Using multiple tests is not without drawbacks. It increases the time and cost of experiments, and it can lead to issues with habituation or carryover effects if tests are performed too close together. Moreover, different tests may be sensitive to different aspects of the intervention, making it difficult to integrate results into a single conclusion. In the present study, the opposite results in the EPM and OFT leave the researchers with an interpretive puzzle: is the enzyme anxiolytic, anxiogenic, or neither? The answer may depend on which test one trusts more, but the authors argue that the dissociation itself is informative — it suggests that the enzyme affects behavior in a way that is not captured by a simple anxiety scale.

Another limitation is that the study was performed only in male mice. Sex differences in anxiety-like behavior are well documented, and the gut microbiome also differs between males and females. It is possible that female mice would show a different pattern of results. The researchers acknowledge this and plan to replicate the study in females, but the preprint does not include those data.

The use of a purified enzyme rather than a live bacterium is both a strength and a weakness. On one hand, it allows for precise control of the dose and avoids the confounding effects of other bacterial metabolites. On the other hand, it does not capture the complex ecology of the gut microbiome, where multiple species produce and consume glucuronidated compounds. Future studies will need to examine whether the same effects occur when B. fragilis is present in the gut, and whether other bacterial enzymes produce similar dissociations.

Furthermore, the enzyme was administered orally, which raises questions about its stability in the gastrointestinal tract. Beta-glucuronidase is a protein that can be degraded by stomach acid and proteases. The researchers used a high dose to compensate, but it is possible that only a fraction reached the intestine intact. They could have used encapsulated enzyme or a genetically engineered bacterium that produces the enzyme in situ to improve delivery.

Another trade-off is the choice of control group. The researchers used heat-inactivated enzyme as a control, which is appropriate to rule out nonspecific effects of protein ingestion. However, heat inactivation might not fully denature all components, and the control could still contain endotoxins or other contaminants. A better control might be a vehicle-only group and a separate group receiving an unrelated enzyme, such as lysozyme, to control for general proteolytic activity.

The sample size was 12 mice per group, which is typical for behavioral studies but may be insufficient to detect small effects. A power analysis based on a moderate effect size (Cohen's d = 0.5) would require about 20 mice per group for 80 percent power. The researchers did not report a power analysis, and the observed effects were moderate (30 percent increase in EPM, 20 percent decrease in OFT). With a larger sample, the dissociation might become more robust or, conversely, might not replicate. The authors should consider a replication with a larger cohort.

Implications for Reproducibility

The field of behavioral neuroscience has been grappling with a reproducibility crisis, and the present study adds a new layer of complexity. If the same intervention can produce opposite results in two standard tests, then the outcome of a study may depend heavily on which test is chosen. This could explain why some findings in the gut-brain literature have been difficult to replicate: different laboratories may use different assays, or even the same assay with slight procedural variations, and obtain conflicting results.

To improve reproducibility, the authors suggest that journals and funding agencies should require researchers to use at least two behavioral tests for anxiety-like behavior, and to report locomotion as a covariate. They also recommend that raw data be made publicly available so that other groups can reanalyze the results. The preprint itself is a step in that direction, as it includes detailed methods and supplementary data.

But the problem runs deeper. Even with multiple tests, the interpretation of results can be ambiguous. For instance, the light-dark box, another common assay, measures the latency to enter the dark compartment and time spent in the light. If the enzyme decreases latency but increases time in the light, that would be consistent with reduced anxiety. But if it also increases overall activity, the interpretation becomes murky. The field needs standardized statistical approaches, such as using principal component analysis or structural equation modeling, to integrate data from multiple tests and separate anxiety from confounding variables.

Another reproducibility challenge is the potential for publication bias. Studies that find a clear anxiolytic effect in a single test are more likely to be published than those that find null or contradictory results. The present study, by reporting a dissociation, might be seen as a negative result that is less likely to be published in a high-impact journal. The authors chose to post it as a preprint, which is commendable, but the incentive structure in academia still favors clean, positive stories. To counter this, some journals have started accepting registered reports, where the study design and analysis plan are peer-reviewed before data collection. The present study could have benefited from such an approach.

In the end, the study is a reminder that behavior is a complex phenotype, and that our tools for measuring it are imperfect. The enzyme beta-glucuronidase may or may not affect anxiety in mice, but the real finding is methodological: the need for multiple assays, careful controls, and a healthy dose of skepticism when interpreting single-test results. As the gut-brain axis field matures, such rigor will be essential for separating genuine effects from artifacts of the assay.

The researchers are now planning a larger study that will include a battery of behavioral tests, as well as measurements of brain activity and gut metabolites. They hope that by combining multiple approaches, they can build a more complete picture of how the gut microbiome influences the brain. For now, the takeaway is clear: when it comes to measuring anxiety in mice, one test is not enough.