Glyphosate at the 'Safe' Dose Makes Rats Anxious

Glyphosate is the most widely used herbicide on Earth. It’s in Roundup, in your tomatoes, in wheat, in more than 80% of American urine samples in biomonitoring studies. The U.S. Environmental Protection Agency set a «safe» chronic daily intake: 2 mg per kilogram of body weight. In a new study from the University of Puerto Rico, researchers gave rats exactly that dose — not ten times more, not a hundred — precisely the EPA reference level. After sixteen weeks, those rats weren’t burning up or collapsing. They were just starting to fear things that weren’t scary.

What Actually Changed in the Rats’ Behavior

Illustration. Source: Unsplash

Demetrio Sierra-Mercado’s team split adult male rats into two groups. One received glyphosate in drinking water for all sixteen weeks at a concentration matching the EPA-approved human chronic reference dose, adjusted to rat body weight. The other drank filtered water. Everything else matched: food, light, temperature, handling.

Behavioral tests came in waves. At week 4, there was no difference: in the «open field» — a standard exploration test — both groups behaved identically. By week 10, the more sensitive elevated plus maze — a cross-shaped maze with two open arms and two walled arms — showed glyphosate rats avoiding the open arms. They hugged the walls. That’s a textbook anxiety signal.

By week 14 the gap deepened. In a novel object test — an unfamiliar item placed in the corner of the cage — the control rats behaved as rodents do: approached, sniffed, nibbled, investigated. The glyphosate rats froze in place. They looked at the new object as if it might eat them. That freezing response is another well-established anxiety marker.

Then week 16, the crescendo. The researchers played a neutral sound tone that had never been paired with pain or danger. Control rats barely reacted. Glyphosate rats froze much more often. But when the same tone was later paired with a mild electric foot shock (an actual threat), both groups reacted identically. On real threats, glyphosate rats behaved normally. On neutral signals, they behaved as if every sound were a warning.

This is a subtle but critical distinction. It’s not generalized fearfulness or panic-of-everything. It’s hypervigilance — the state in which the brain starts classifying harmless signals as dangerous. In clinical psychology, hypervigilance is a core symptom of generalized anxiety disorder and PTSD.

How a Field Herbicide Reached the Brain

The link between a chemical sprayed on crops and neural activity inside a rat’s brain isn’t obvious. The researchers found an intermediary: gut bacteria. Specifically one group.

Lactobacillus — a genus of bacteria that lives in healthy guts and produces lactic acid. It plays a huge role in maintaining homeostasis: protecting against pathogens, regulating immunity, and — crucially for this story — contributing to the synthesis of serotonin precursors. Serotonin is one of the main neurotransmitters governing mood and anxiety.

In glyphosate-exposed rats, Lactobacillus levels dropped sharply. Why? Biochemistry that toxicologists have known about for years. Glyphosate binds manganese — a trace mineral essential for many enzymes. Without manganese, Lactobacillus can’t function normally, because its metabolism depends heavily on it. Glyphosate, in effect, strips away the bacterium’s key nutrient and evicts it from the gut ecosystem.

The gut–brain axis is a two-way communication system between the digestive tract and the central nervous system. It operates through nerve pathways (primarily the vagus nerve), immune signals, and chemical messengers produced by gut bacteria. Up to 90% of the body’s serotonin is synthesized in the gut, not in the brain.

When Lactobacillus drops, serotonin precursor production suffers. And serotonin is one of the main chemical anchors that keeps anxiety in check. Its imbalance is a known risk factor for anxiety disorders in humans. The picture builds: glyphosate → Lactobacillus loss → weakened serotonin signaling → downstream brain changes.

One Brain Region That Decides What’s Dangerous

Next, the team looked inside the brains of the rats and asked: where exactly did the change happen? They checked three candidates: the amygdala, the prefrontal cortex, and the bed nucleus of the stria terminalis — BNST for short.

The amygdala is the classic «fear station, ” responsible for acute and immediate responses: see a predator, scream, run. It wasn’t changed. The prefrontal cortex, which acts as a brake on that fear, was also unchanged. But the BNST — a region most people haven’t heard of — showed significantly elevated cellular activity in the glyphosate rats.

The BNST (bed nucleus of the stria terminalis) is a relatively small structure in the forebrain, anatomically and functionally linked to the amygdala but specialized for something different: not acute «fight or flight, ” but prolonged, diffuse threat anticipation. The BNST is active when it’s unclear where danger might come from, when the brain needs to stay on guard without a concrete trigger.

That distinction matters. The amygdala is about panic at the sight of a snake. The BNST is about worry that a snake is probably somewhere around. Clinical anxiety disorders — generalized anxiety, social anxiety, PTSD — are linked specifically to chronic BNST overactivation, not to the amygdala. This study is the first to show that glyphosate hits the BNST specifically, not the fear system as a whole.

Why the BNST is vulnerable to changes in gut microbiota also comes back to serotonin. The BNST is densely innervated by serotonergic fibers, and shifts in serotonin precursor availability can significantly change its activity.

What This Means for Humans

The question every reader asks. And here it’s important to separate what’s actually been shown from what’s being extrapolated.

Shown: rats receiving glyphosate at the EPA-”allowed» dose for 16 weeks develop behavioral and neurochemical changes consistent with heightened anxiety, through the gut–brain axis. Independent research, including the NIEHS group’s 2023 study, sees the same vector: glyphosate at low doses increases rodent anxiety. This isn’t a one-off.

Not shown: that the same happens in humans. The «mg/kg» conversion from rat to human isn’t direct — metabolism, lifespan, dietary habits all differ. But two facts make the question sharp. First, real human exposure to glyphosate is well below the EPA limit (most estimates put it at 0.001–0.1 mg/kg, 20 to 2,000 times lower than the experimental dose). That reduces risk. Second, biomonitoring shows glyphosate is present in the urine of most people in industrialized countries, meaning exposure isn’t zero — it’s regular and chronic. Even a very small but constant dose across large populations can produce meaningful shifts that are hard to catch in individual testing.

What should you do? Don’t panic and don’t throw out the bread. But don’t dismiss this either. The paper is a case for more serious human epidemiology, for revisiting the «safe» threshold (which was calculated based on liver and kidney toxicity, not on neurobehavioral effects via the microbiome), and for more caution about herbicides in agriculture.

Critical Analysis: Limitations Worth Remembering

The paper is published in Frontiers in Toxicology, a peer-reviewed journal with a moderate impact factor. Not Science or Nature, but not a shady preprint either. The methodology is simple and transparent.

The authors themselves list the main caveats. First — male rats only. The female hormonal system interacts with the microbiome and anxiety differently; results may not replicate in females, or may replicate differently. This is a common problem in neuroscience literature, but it’s becoming less tolerable.

Second — gut microbiome assessed only in fecal samples, and only at the end of the experiment. Fecal bacteria aren’t the whole gut: a significant portion of microbes live in the mucosal layer and on walls, where fecal sampling misses them. And a single endpoint measurement can’t tell us whether microbiome changes preceded or followed behavioral changes.

Third — no direct extrapolation to human dosage. The paper doesn’t model how much glyphosate a real city resident actually gets from their diet. That’s a job for other studies.

And finally — it’s important to separate this paper from the Samsel-Seneff controversy. Stephanie Seneff and Anthony Samsel published articles in the 2010s attributing causal roles to glyphosate in a huge list of diseases, from autism to Alzheimer’s. Most of their arguments came under heavy criticism (see Mesnage & Antoniou’s 2017 review), and the scientific community largely rejected their methodology. The Sierra-Mercado paper is an entirely different story: methodologically careful, with a control group, specific neurobiological measurements, and cautious conclusions. Don’t conflate them.

Horizon

The next steps are obvious. Repeat with females. Add multiple doses below the EPA limit to find where the effect disappears. Run parallel human studies — measure glyphosate in blood or urine and correlate with anxiety metrics, mental health, inflammatory markers. EU regulators are already reconsidering glyphosate license renewal, and data on neurobehavioral effects via the microbiome are exactly what they should factor into the decision. A dose set decades ago based on different endpoints stops being «safe» once we start measuring different things.

For the reader, the usable takeaway is cleaner than the headlines suggest. Not «glyphosate is driving you crazy.» Rather: «the dose the EPA calls safe changes the wiring of a rat’s threat-detection system through its microbiome, and that’s a reason to keep asking questions.»