An analysis of mice found that gut bacterial communities differed between the methamphetamine exposure phase and the abstinence phase, while their overall structure also changed across the timeline. The result describes a shifting association between the drug-use phase and the microbiome, not a demonstrated biological mechanism.
The experiment randomly assigned mice to methamphetamine, saline or control groups, with eight animals in each group. The methamphetamine group received 1 mg/kg over 60 days, followed by 60 days of abstinence. Feces were collected separately from each animal at planned points during the 120-day experiment, with about 50 mg collected at each time point.
Different phases, different communities
Sequencing of 159 fecal samples produced 4,306,249 high-quality bacterial reads. After the data were rarefied to 4,285 reads per sample, the analysis identified 2,758 OTUs, spanning 14 bacterial phyla and 99 genera. Researchers grouped sequences at 97% similarity and assigned taxonomy with the Greengenes 13_8 database before calculating diversity and community-structure measures.
Shannon diversity, a broad measure of how varied and evenly represented a community is, did not differ significantly among the experimental groups. It was lower during the addiction phase than during abstinence across all groups, with a Mann-Whitney U statistic of 1,792 and a reported P value below 0.001.
At the genus level, methamphetamine-exposed mice had less Enterococcus and more Bifidobacterium and Turicibacter than controls during addiction. The abstinence phase showed partial recovery, but its composition remained distinct from the control groups. Dehalobacterium was reported as the genus that distinguished the two phases.
To compare the overall makeup of the communities, the researchers used beta diversity, which looks at how samples differ from one another. Bray-Curtis PCoA separated addiction and abstinence samples within the methamphetamine group. Its first two axes accounted for 42.0% and 17.4% of the variation, respectively; addiction samples were more dispersed, and ANOSIM found a significant difference between phases, with a reported P value below 0.05. These results describe community structure, not direct microbial functions.
The timeline was not a straight line back
Across seven time points, overall Shannon diversity did not differ significantly, with a Kruskal-Wallis statistic of 5.2871 and a P value of 0.51. Adjacent-time comparisons did show temporary fluctuations, however, and addiction day 60 was not significantly different from abstinence day 1.
Time-series analysis also identified shifts in Verrucomicrobia and Actinobacteria, plus reproducible temporal patterns in Butyricimonas, Candidatus Arthromitus, Enterococcus and Turicibacter.
The community structure also changed over time. In the temporal Bray-Curtis PCoA, the first two axes explained 40.15% and 29.51% of variation. PERMANOVA found significant effects of addiction and abstinence time points, with a reported P value below 0.05. Within abstinence, day 30 and day 60 were not significantly different, while day 1 differed from the other abstinence points.
Prediction came with clear limits
Machine learning produced an exploratory phase classifier from the 15 most important genera. Its out-of-bag error, the model's internal error estimate, was 0.129 for addiction samples and 0.348 for abstinence samples, with an overall error of 0.222. Arthrobacter was the most important genus, followed by Butyricimonas. The uneven errors suggest that the model recognized addiction samples more reliably than abstinence samples, but it was evaluated in the same dataset and was not externally validated.
A separate random-forest regression tried to estimate the time point from genus-level relative abundance. It used a sample-level split, with 70% of samples for training and 30% for testing. On the test data, the addiction model had an R-squared of 0.7818, with a mean absolute error of 9.26 days and a root mean square error of 10.40 days. The abstinence model had an R-squared of 0.4592, with corresponding errors of 13.34 and 15.74 days. The model therefore tracked the addiction timeline more closely than the abstinence timeline, but the analysis was exploratory.
The findings do not show what caused the microbial shifts. The authors say that 16S rRNA sequencing and relatively sparse sampling during abstinence limited the taxonomic and temporal resolution. The study did not directly assess microbial metabolites, inflammatory responses or gut-brain functional endpoints, and the reported signatures still require validation in independent animals and external cohorts.
The work therefore remains a map of time-dependent microbiome associations in a mouse model, not a validated prediction tool. Further studies would need to test whether the patterns hold at the animal level and outside the original cohort, then connect them to measured biological functions.
Paper data and sources
Original title: Longitudinal dynamics of the gut microbiome during methamphetamine addiction and abstinence
Authors: Wang L, Wang J, Zhang Y et al.
Journal/Repository: Frontiers in microbiology
Status: Peer-reviewed
First online: 2026-10-08
DOI: Not available
Original paper · Full text