Can Low-Dose THC Raise Brain mTOR While Lowering Adipose mTOR? A Mouse Time-Course Study

Short answer: Yes—in a controlled experiment involving male mice, continuous low-dose Δ⁹-tetrahydrocannabinol (THC) was associated with higher brain mTOR phosphorylation early in treatment and lower mTOR phosphorylation in visceral adipose tissue later. Hippocampal metabolic changes peaked around day 14, while several plasma and adipose metabolic differences appeared later. The observations concern tissue signaling and metabolites in mice, not measured human blood THC, demonstrable cognitive improvement in the animals studied, individual impairment, or proof that THC prevents aging.

The primary source is Bilkei-Gorzo and colleagues’ 2024 original experimental paper, Bidirectional Effect of Long-Term Δ9-Tetrahydrocannabinol Treatment on mTOR Activity and Metabolome, published in ACS Pharmacology & Translational Science (DOI 10.1021/acsptsci.4c00002). Its major interpretive lesson is that the word THC effect is incomplete without specifying the tissue, biological endpoint, treatment duration, animal age, and measurement technique.

What was the scientific question?

The investigators asked whether a sustained, relatively low THC administration regimen alters the mechanistic target of rapamycin (mTOR) signaling system and the metabolome differently across brain and peripheral compartments. mTOR participates in nutrient sensing, protein synthesis and metabolic regulation. A single expectation that mTOR must rise or fall uniformly throughout the body would have obscured the experiment’s central observation: responses varied by tissue and time.

The authors had previously reported age-related cognitive and synaptic changes in other mouse experiments. The 2024 investigation tested possible biological correlates, including phosphorylated mTOR, synaptic-associated protein abundance, and small-molecule metabolites. Importantly, it did not directly repeat a full battery of memory, reaction-time or driving-related behavioral tests as primary outcomes. The present data therefore support a hypothesis about mechanisms; they do not independently establish that the treated mice learned better, or that a person would experience a beneficial cognitive effect.

Which animals received THC, and how was it administered?

The study used male C57BL/6J mice at approximately 18 months (the older cohort) and 4 months (the younger comparison cohort). The main experimental regimen administered 3 mg THC per kilogram of body weight per day by implanted subdermal osmotic minipump, compared with vehicle. The vehicle was an ethanol/Cremophor/saline mixture at a 4:4:2 volume ratio; vehicle controls were necessary to distinguish effects of the active treatment from the delivery system and its solvent formulation.

Older animals were sampled after 14 or 28 days of treatment and at day 42, following 14 days without the active pump-delivered regimen. The pumps were designed to operate for a maximum of 28 days. Young animals were sampled for the principal age comparison at day 28. The main metabolomic or validation experiments described eight animals per treatment group; a separate day-14 hippocampal protein cohort contained 14 animals in each treatment arm before the specific analyte comparisons. A separate dose-response experiment used older mice across vehicle and 0.3, 1 and 3 mg/kg/day conditions, described as 12 animals per group.

Those numbers should not be pooled into one fictitious trial size. Terminal tissue collection at different days involved different cohorts. A mouse’s hippocampus, plasma and fat are related biological compartments, not independent human participants. Metabolite features counted by mass spectrometry are analytes, not additional animals.

What was actually measured: mTOR proteins, metabolites, or THC concentrations?

The brain signaling endpoint was the proportion of phosphorylated mTOR relative to total mTOR, measured in tissue preparations using automated immunoassay-based Western blotting. The investigators separately quantified the synaptic-associated proteins synaptophysin and PSD-95. Those are tissue protein measurements. Their abundance does not itself count newly formed synapses or prove a cognitive outcome.

Broad, untargeted metabolomic profiling used ultrahigh-performance liquid chromatography coupled to tandem mass spectrometry (UPLC–MS/MS) across hippocampus, blood plasma, and visceral white adipose tissue. The resulting compound and pathway comparisons concern measured small-molecule features, followed by normalization and statistical testing. They are not equivalent to a validated forensic quantitative assay of parent THC or THC-COOH in the animals’ venous whole blood.

A separate targeted LC–MS/MS procedure did measure THC, but in an in-vitro minipump release experiment involving pump fluid collected under laboratory conditions. It assessed stability of delivery output, not an individual animal’s blood or brain THC concentration. Treating this in-vitro chromatographic result as proof of a particular in-vivo exposure concentration would be a serious specimen-category error.

What happened to brain mTOR signaling over time?

The cortex showed a statistically significant treatment-by-time interaction for relative mTOR phosphorylation: F(2,39) = 7.302, p = 0.002. The day-14 cortical comparison favored increased phosphorylation in THC-treated mice (Figure 1 reports p = 0.0112), but the separate day-28 and day-42 contrasts were not statistically significant (p = 0.3790 and 0.3325, respectively). That is a temporary early shift, not evidence of an unbroken increase throughout treatment.

A separate hippocampal cohort also showed increased phosphorylation on day 14: 47.30% higher in the treated animals, t(12) = 2.708, p = 0.0190. The raw outcome is phosphorylation relative to total mTOR, a signaling proxy. It is more precise to say increased measured mTOR phosphorylation than to imply that the investigators continuously monitored enzyme activity within each living neuron.

Crucially, the amount of total cortical mTOR protein did not show a significant main treatment effect, F(1,41) = 0.489, p = 0.489, or treatment-by-time interaction, F(2,41) = 0.491, p = 0.615. Phosphorylation and total protein abundance answer different questions. A nonsignificant protein comparison should not be recast as proof of exact biochemical equivalence.

Did synaptic protein abundance change on the same schedule?

Not exactly. In the cortex, the synaptic-associated protein synaptophysin was approximately 40.2% higher at day 28 in the THC-treated group (t(12) = 2.384, p = 0.0345). PSD-95 was approximately 42.2% higher at day 28 (t(15) = 2.530, p = 0.0231). At day 14, however, the corresponding comparisons were not significant: synaptophysin −14.53%, p = 0.3421, and PSD-95 +16.4%, p = 0.4874.

The staggered findings are biologically interesting: phosphorylation differences appeared earlier than the later protein-abundance differences. Yet the sequence alone cannot establish that early mTOR phosphorylation caused subsequent protein expression changes. Demonstrating mediation would require appropriate manipulations and measurements beyond temporal association. Nor should protein amounts be substituted for directly observed dendritic-spine counts or a behavioral memory score from this particular experiment.

What changed in hippocampal metabolites at day 14?

The researchers reported that 109 of 494 measured hippocampal metabolites (approximately 22%) were significantly increased after 14 days of THC treatment. That count is a tally of chemical features, not 109 animals, independent replicates or clinically proven therapeutic biomarkers. The broad early shift had largely diminished by day 28 and after withdrawal at day 42.

At day 14 the study identified higher metabolite-group levels associated with glycolysis (F(1,14) = 6.815, p = 0.0206), the pentose-phosphate pathway (F(1,14) = 6.951, p = 0.0195), and citric-acid-cycle/oxidative-phosphorylation-associated pathways (F(1,14) = 5.213, p = 0.0386). Long-chain fatty-acid (p = 0.0288) and polyunsaturated-fatty-acid (p = 0.0364) group comparisons also rose. Branched-chain amino-acid metabolites showed a treatment-group effect, F(1,14) = 6.207, p = 0.0259.

These findings concern relative molecular abundance in extracted hippocampal tissue. A greater metabolite signal does not by itself demonstrate increased real-time neuronal energy expenditure, improved learning, a higher brain THC concentration, or a beneficial clinical outcome. It supplies a plausible mechanistic pattern that must be assessed with the underlying design, assay and multiple-testing burden.

Evidence table: brain and fat measurements versus important null comparisons

How did plasma changes differ from the brain findings?

The blood specimens used for the metabolome were separated plasma, not forensic venous whole blood. At day 14, the plasma pattern featured increases mainly among lipid-related compounds. The investigators identified differences in fatty-acid amides (p = 0.0322), phosphatidylethanolamines (p = 0.0038), phosphatidylinositols (p = 0.0168), diacylglycerols (p = 0.0062), and ceramides (p = 0.0287). A purine metabolite grouping involving hypoxanthine/xanthine and inosine moved in the opposite direction, p = 0.0195.

By day 28, reductions predominated across many amino-acid-related plasma metabolite groups, including branched-chain amino acids (p = 0.0146), lysine (p = 0.0039), phenylalanine (p = 0.0009), and tryptophan (p = 0.0052). Yet polyunsaturated fatty acids remained elevated (F(1,14) = 16.52, p = 0.0012), and certain acyl-carnitines also differed significantly. It would therefore be inaccurate to say simply that THC lowered all plasma metabolites.

At day 42, 14 days after stopping administration, 11.85% of 675 measured plasma compounds were reduced and 2.37% increased according to the paper’s analytical classification. These percentages count chemical features; they are not effect sizes for an individual person’s THC metabolism or probabilities of impairment. Plasma amino-acid and phospholipid group differences remained statistically detectable in selected comparisons.

Were N-acetylated amino acids dose dependent?

The authors singled out plasma N-acetylated amino acids for targeted follow-up. Their combined day-28 treatment comparison in older mice was substantial, F(1,14) = 34.68, p < 0.0001, with lower concentrations in the treated group. In a separate older-mouse experiment involving vehicle and multiple THC doses, the reported dose effect was F(2,40) = 5.730, p = 0.0065.

This is a dose-related metabolite observation, not an independently validated dose-reconstruction method. The separate targeted plasma assay measured N-acetylated amino-acid concentrations, not THC exposure in human forensic blood. A metabolite pattern may respond to multiple physiological processes, ages, diets and collection conditions; without specificity, calibration against the intended forensic decision, and clinical outcome validation, the study cannot determine when a person used cannabis, the administered THC dose, or whether that person was impaired.

Did young and old mice respond in the same way?

No. In the day-28 comparison, the authors reported several opposite directions of plasma metabolite change in four-month versus eighteen-month-old mice. For N-acetylated amino acids the reported age-associated comparison was F(1,98) = 219.6, p < 0.0001. Other metabolite-group age effects were also statistically strong. Polyunsaturated fatty-acid changes were an important exception: the paper did not find a corresponding age effect for that grouping (F(1,210) = 0.254, p = 0.615).

These are differences between two specified ages of male laboratory mice, under implanted continuous delivery. They do not justify a human therapeutic age cutoff or a claim that a specific THC dose will improve cognition in older individuals but worsen cognition in younger ones. The age groups are biologically distinct and the experimental findings are mechanistic, not a validated human treatment schedule.

What did the visceral adipose tissue reveal?

In visceral white adipose tissue, the mTOR phosphorylation pattern moved in the opposite direction from the transient early brain increase. The overall treatment effect was F(1,34) = 15.64, p = 0.0004, with the individually significant lower phosphorylation comparison occurring at day 28 (p = 0.0185). The treatment-by-time interaction was not significant, F(2,34) = 0.648, p = 0.530. Figure 7’s day-14 and day-42 pairwise comparisons were also nonsignificant, p = 0.2414 and 0.1277.

The adipose metabolome displayed another temporal pattern. The authors reported little change at day 14, greater downward shifts in selected amino-acid and carbohydrate features at day 28, and a more complex mix after withdrawal at day 42. At day 28, 20.6% of amino-acid and 10.3% of carbohydrate metabolite features were lower, with none in those categories classified as higher. At day 42 the percentages and directions differed: 19.4% of amino-acid features and 34.5% of carbohydrate features were higher. The adipose results therefore cannot be flattened into a claim of universally suppressed metabolic activity.

Among individual grouped adipose outcomes, day-42 glycogen-related features decreased (F(1,14) = 17.51, p = 0.0009) while pentose metabolites increased (F(1,14) = 11.32, p = 0.0046). These observations are molecular endpoints in sampled fat. No tissue result here constitutes a validated chemical estimate of THC retained in fat or an empirically demonstrated duration of human intoxication.

Evidence table: biological matrix, timeline, and what the result cannot establish

Which negative findings are necessary to report?

A balanced interpretation should include the null comparisons, not merely a list of selected positive results. Body-weight trajectories did not differ materially between groups, although food intake exhibited a treatment-by-time interaction near the end of treatment, F(12,408) = 2.613, p = 0.0023. Home-cage activity showed no significant main treatment effect during the measured late treatment window (F(1,18) = 0.344, p = 0.5658) or late withdrawal window (F(1,16) = 1.009, p = 0.3302).

The authors also reported no significant treatment-related change in CB1 receptor protein expression in the tested cortex, hippocampus and adipose preparations. The in-vitro minipump release experiment did not identify significant differences among the sampled weekly output periods, p = 0.1769. These results help evaluate alternative explanations but cannot prove perfect equivalence: statistical power, timing, assay variability and the size of potentially undetected changes remain relevant.

Several day-specific brain and fat mTOR and synaptic-protein contrasts were nonsignificant despite positive differences at other times. A selective summary that suppresses those comparisons would substantially exaggerate certainty, especially because hundreds of molecular features and multiple pathways were screened.

Why do measured plasma metabolites not equal measured blood THC?

Forensic toxicology depends on defining the matrix–analyte–method combination. A mass-spectrometry result for N-acetylated amino acids in mouse plasma is physically measured chemistry, but the target is not parent THC. A measured tissue phosphoprotein is a biological response marker, not a drug level. An administered 3 mg/kg/day dose is a nominal treatment input, not a measured blood concentration. A separate in-vitro THC release assay is not interchangeable with in-vivo drug distribution.

These distinctions mirror the broader explanation in the published biological matrices guide. Direct chemical identity and method-specific calibration are addressed separately in the published confirmatory drug-testing guide. Time and route-dependent exposure interpretation is developed in the published pharmacokinetics guide.

Neither a nominal dose, an untargeted metabolomic profile, a phosphorylated protein ratio nor a mathematical exposure model can be assigned an invented human whole-blood THC concentration. The paper reported no individual driving tasks, observed field sobriety performance, crash data, or validated concentration–impairment classification. Even where an experiment shows a reproducible mechanism, translating it to a named person’s behavior requires different evidence.

What are the principal scientific limitations?

Species and sex: the core experiments involved male laboratory mice, not human patients or mixed-sex forensic case populations. Age: four-month and eighteen-month mice were examined at differing stages; the young age group was sampled at day 28 for the main comparison. Delivery: continuous subdermal pumping differs materially from human smoking, vapor inhalation, edible exposure and other clinical regimens. Specimens: tissue and plasma collections were terminal or separately sampled, not repeated brain biopsies on the same animals. Endpoint: the principal measurements were proteins and metabolic profiles, not new cognitive performance or actual life span.

Statistics and multiplicity: untargeted metabolomics assesses many correlated chemical features, and the authors applied combinations of ANOVA, Welch comparisons and false-discovery-oriented procedures depending on the analysis. A nominal pathway p-value cannot be treated as a universal clinical biomarker validation result. Compound-feature counts reflect the detection and classification pipeline. Technical normalization, imputation of missing values with minimum observed values and grouped pathway tests introduce analytical decisions whose potential influence deserves review.

Mechanism: a temporal sequence—earlier brain phosphoprotein change and later synaptic protein or plasma change—does not identify a proven single causal chain. Age-dependent opposite metabolite signals likewise do not establish a safe or effective human drug intervention. The authors’ antiaging hypothesis is a mechanistic interpretation requiring dedicated lifespan, functional and clinical studies, not a result demonstrated in people by this experiment.

How should forensic experts use this study without overstating it?

It is useful as evidence that THC-related biological response can depend on duration, organ system, organism age and the selected endpoint. It shows why a single broad characterization such as “THC stimulates metabolism” or “THC suppresses mTOR” is scientifically incomplete. It also supplies an example of how rigorous source review must separate molecular measurements from clinical and legal endpoints.

It is not a validated tool for reconstructing THC use time from a human blood sample, calculating a whole-blood concentration from plasma metabolites, diagnosing impairment, identifying an individual’s cognitive capacity, or inferring the cause of a particular incident. Those claims would require actual specimen testing under a validated forensic method plus independent behavioral and case-specific evidence where applicable. The appropriate scientific conclusion is narrower: sustained low-dose THC produced temporally and anatomically divergent signaling and metabolic profiles in the tested mouse cohorts.

Frequently asked scientific and forensic questions

1. Did the investigators directly measure THC in the treated mice’s blood?

No forensic quantitative mouse whole-blood or plasma THC result was reported as a study exposure outcome. The broad plasma assays concerned other metabolites. A separate LC–MS/MS parent-THC assay quantified release from osmotic pumps in vitro, which is not a blood measurement.

2. Does increased mTOR phosphorylation prove THC improved memory?

No. It is a protein-signaling observation. This paper did not independently establish memory improvements with a contemporaneous behavioral testing battery. Mechanistic plausibility is different from a demonstrated behavioral outcome.

3. Did brain and adipose mTOR both rise after THC treatment?

No. The cortex and hippocampus showed an early day-14 phosphorylation increase, while the visceral adipose comparison was lower, most clearly at day 28. The cortex day-28 and day-42 pairwise contrasts were nonsignificant; the adipose treatment-by-time interaction was also nonsignificant.

4. Were 494 metabolites measured in 494 mice?

No. The figure of 494 refers to assayed hippocampal chemical features, of which 109 were classified as increased at day 14. The main metabolomic comparisons described about eight mice per treatment group, not hundreds of independent animals.

5. Can lower plasma N-acetylated amino acids identify someone’s dose or last cannabis use?

Not on these data. The observed mouse metabolic dose-response was not validated as a specific human THC exposure clock or a validated forensic dosing algorithm. It did not establish the concentration, sensitivity, specificity or individual time-course needed for that purpose.

6. Did the experiment demonstrate that THC prevents aging in people?

No. The antiaging discussion is a mechanistic hypothesis informed partly by other research. This study measured mouse tissue signaling and metabolites; it did not measure human lifespan or establish a therapeutic indication or dose.

7. Can a forensic expert use these findings to declare a driver impaired or unimpaired?

No. No human driving, field sobriety, validated psychomotor impairment threshold, incident reconstruction or clinical outcome classification was tested. The study may inform background biological reasoning, but it cannot replace validated specimen analysis and person-specific evidence.

Primary research source

Bilkei-Gorzo A, Schurmann B, Schneider M, Kraemer M, Nidadavolu P, Beins EC, Müller CE, Dvir-Ginzberg M, Zimmer A. Bidirectional Effect of Long-Term Δ9-Tetrahydrocannabinol Treatment on mTOR Activity and Metabolome. ACS Pharmacology & Translational Science. 2024;7:2637–2649. doi:10.1021/acsptsci.4c00002. The analysis above is based on this complete original experimental article, including its methods, reported positive and null results, figures and discussion.