Does Prenatal THC Change Muscle Genes Without Demonstrable Fibrosis? An 18-Offspring Primate Pilot

Short answer: In a small controlled rhesus-macaque experiment, prolonged maternal oral Δ9-tetrahydrocannabinol (THC) exposure was associated with differences in offspring skeletal-muscle gene expression, especially immune and inflammatory signaling. The investigators reported 13 differentially expressed genes in fetal muscle and 33 in infant muscle. Yet their separate tissue-staining analysis found no statistically significant difference in the percentage of rectus femoris collagen between exposed and control groups (reported p > 0.05). Altered transcription did not establish muscle fibrosis, clinically important weakness, or a THC-caused injury in an individual child.

The original peer-reviewed primary research is Moellmer and colleagues, published July 31, 2024, in PLOS ONE: doi:10.1371/journal.pone.0306868. The distinction matters when scientific findings are offered in litigation: a tissue-specific molecular signal is a measured laboratory outcome, while an alleged functional developmental injury is a materially different proposition requiring its own evidence.

What question did this original experiment actually test?

Researchers asked whether chronic maternal exposure to THC, beginning before conception and continuing through pregnancy, altered the offspring’s rectus femoris skeletal muscle at fetal or infant developmental stages. They assessed RNA expression using a targeted neuroinflammatory gene panel and separately examined stained tissue for collagen. The experiment was designed to examine early biological mechanisms rather than diagnose a condition in pregnant patients or children.

That scope is narrower than the phrase “prenatal cannabis harm” may suggest. The investigators did not administer THC to human pregnant volunteers, measure human fetal muscle, determine whether particular children had been injured, or assess the offspring’s walking, strength, fine motor skills, exercise tolerance, or future orthopedic function. Nor did they use an individual forensic toxicology specimen to reconstruct a maternal THC dose. The tested setting was a preclinical animal experiment with a controlled oral exposure schedule.

How many animals were studied, and when were the tissues obtained?

The researchers enrolled 18 adult female rhesus macaques. Their offspring were assigned to two distinct developmental endpoint cohorts: 10 fetal offspring and eight infant offspring. Within each endpoint cohort, the adult females were allocated to THC-exposed or control conditions. These are not 18 fetal animals plus another 18 infants and not 36 independent offspring. A prenatal animal that was analyzed at the fetal endpoint was not later evaluated as an infant in this design.

The fetal offspring were delivered by cesarean section for collection at gestational day 155. The infant offspring were born at term, approximately gestational day 165, and their tissue was collected 180 days after delivery. The experimental procedures involved necropsy and collection of rectus femoris muscle. Thus, the relevant measured biological matrix was solid skeletal-muscle tissue, preserved as formalin-fixed, paraffin-embedded sections. It was not blood, plasma, serum, urine, oral fluid, or a breath alcohol sample.

One infant-cohort specimen did not satisfy the stated NanoString quality-control criteria and was excluded from gene-expression analysis. Consequently, an unqualified denominator of eight infant molecular profiles would be misleading. The report does not provide a person-by-person validated prediction model or a prevalence estimate for human fetal disease.

What THC dose and exposure route were controlled?

The exposed adult females consumed research-grade THC in edible cookies. The administered daily dose was escalated before timed breeding: 0.07 mg/kg/day in weeks 0–3, 0.14 mg/kg/day in weeks 4–6, 0.29 mg/kg/day in weeks 7–9, and 0.36 mg/kg/day in weeks 10–12. The investigators then maintained 0.36 mg/kg/day through the offspring endpoint. This is an administered oral mass per maternal body weight per day, not a measured THC concentration in fetal tissue.

The original study did not establish a corresponding individual fetal whole-blood THC concentration for each gene or collagen result. Maternal oral dosing does not supply a validated conversion to forensic venous blood THC, urinary cannabinoid metabolites, or a child’s internal tissue concentration. The absence of directly paired chemical exposure measurements also limits any quantitative dose-response reconstruction across individual offspring.

What methods measured molecular change and collagen?

RNA was isolated from paraffin-embedded rectus femoris tissue and tested on a NanoString nCounter 770-gene neuroinflammatory panel, including 13 internal reference genes. A generalized linear model assuming a negative-binomial distribution compared expression between control and exposed groups. The authors used a reported fold-change criterion of at least 1.5 in magnitude and gene-level p-value of 0.05 or less to identify differentially expressed genes. Those are screening decisions within a specified gene panel, not an unbiased entire-genome survey.

RNA quality was an important analytical constraint. The reported RNA integrity numbers were 1.1–2.5, and DV200 percentages were 12–63%, reflecting challenges inherent in fixed-tissue RNA. The assay was selected because NanoString can interrogate transcripts in that setting; its ability to yield counts does not remove sampling and degradation concerns. Gene-level numerical p-values should not be silently converted into clinical diagnostic probabilities.

For histology, investigators applied Masson-Goldner trichrome staining to muscle sections and analyzed segmented stained area using image-processing methods. The outcome was the proportion of stained tissue classified as collagen in the sampled section, compared by unpaired t-test. That measurement differs from transcript counts, actual muscle tensile strength, measured walking behavior, a physician’s fibrosis diagnosis, and microscopic proof of clinical myopathy.

What did the fetal and infant gene panels find?

At the fetal endpoint, the study identified 13 differentially expressed genes: six increased and seven decreased in the exposed-versus-control analysis. Examples reported in original Table 1 include HMOX1 (fold change +2.60, p = 0.05), P2RY12 (−2.48, p = 0.04), BCL2L11 (+2.13, p = 0.04), SLC6A1 (−1.83, p = 0.00522), and PCNA (−1.81, p = 0.01). These values describe comparative RNA expression, not fold-increases in muscle damage or probability of a fetal injury.

At the infant endpoint, 33 genes were reported as differentially expressed: 25 increased and eight decreased. Among Table 2 examples were CCL2 (+14.86, p = 0.02), PTX3 (+9.89, p = 0.02), SERPINE1 (+7.83, p = 0.02), HMOX1 (+3.96, p = 0.02), APOE (−3.22, p = 0.01), and IGF1 (−1.76, p = 0.04). A larger relative fold-change in one transcript is not the same as a quantified protein change, a measured tissue defect, or an individual infant’s clinical prognosis.

The different counts across endpoints do not by themselves prove an escalating biological injury over time. The fetal and infant cohorts involved different animals, different ages and tissue contexts, a small sample, and one excluded infant molecular specimen. Endpoint differences are informative hypotheses but are not within-child longitudinal trajectories.

Evidence table 1: what was directly observed?

What did inflammation-pathway analysis add?

After identifying differentially expressed genes, the authors used STRING and Enrichr to ask whether particular biological processes were overrepresented. In fetal muscle, four upregulated genes—VEGFA, BAD, BCL2L11, and RELN—were associated with PI3K-Akt signaling. In infant muscle, the authors identified ten upregulated genes in an inflammatory-response grouping, including MYD88, RELB, CSF1, CCL2, CCR2, TLR2, FOS, HMOX1, TNFSF4, and PTX3. Reported enrichment strengths were 1.31 for fetal PI3K-Akt and 1.16 for infant inflammatory response.

The source also described enrichment for interleukin-related signaling, including IL-4/IL-13, IL-10, and IL-24 pathways. Its pathway analysis applied Benjamini-Hochberg multiple-testing correction, with reported false-discovery rates of 0.05 or less for the highlighted STRING enrichment results. These pathway labels are computational interpretations of gene sets: the researchers did not directly establish circulating cytokine concentrations, activated immune-cell counts, inflammatory lesions, or histologically diagnosed tissue injury from those pathway annotations alone.

Inflammatory signaling can include regulatory and counter-regulatory processes. An increase in a gene associated with a pathway may reflect altered response or adaptation without identifying a specific pathological direction. The research supports further mechanistic investigation, but the names of biological pathways are not substitutes for a clinically validated phenotype.

Evidence table 2: representative gene results and their limits

Positive and negative signs represent the direction of reported comparative expression. These are relative expression values, not concentrations in ng/mL or mass-based THC measurements. The table reproduces only selected source findings; it cannot be used to infer a diagnostic error rate or reconstruct an individual offspring’s health.

Was fibrosis detected in stained muscle?

No statistically significant collagen-area difference was reported. Investigators examined fetal and infant rectus femoris sections after Masson-Goldner trichrome staining and found comparable percentages of collagen area between THC-exposed and control tissues, with the study reporting p > 0.05. The original publication did not provide a single precise numerical p-value for that comparison in the results prose, so a more exact test result should not be invented.

The correct inference is narrow: under the sample size, section selection, segmentation method, and statistical analysis used, the study did not detect a group difference in this chosen collagen proxy. A non-significant result does not statistically prove strict equivalence, zero microscopic change everywhere, or long-term freedom from disease. Conversely, transcript enrichment does not override the negative histological result and establish observed fibrosis. Both sets of evidence must be presented together.

Why do molecular and tissue findings sometimes point in different directions?

RNA abundance is an intermediate biological signal. Translation into protein depends on regulation, cellular composition, stability, and tissue context. Tissue remodeling then requires additional biological events before it becomes visible as increased collagen area. A gene may change without detectable net tissue remodeling, and a histological assay may lack power for small, patchy, or temporally different effects. Therefore an apparently provocative molecular pathway and a negative histology comparison need not be logically contradictory.

This experiment did not directly quantify the corresponding pathway proteins or prospectively measure locomotor performance. The fetal and infant groups were cross-sectional endpoints, and paraffin-embedded material adds analytical limits. An expert assessment should avoid treating several mechanistic steps as though each had been directly observed. The evidentiary hierarchy remains: administered exposure; tissue transcript count; computational pathway enrichment; measured collagen area; and any independently assessed functional consequence.

Can these results be used as an individual maternal or infant THC exposure test?

No. The study used an experimentally controlled oral maternal exposure, not a validated clinical or forensic assay for timing THC use from an offspring’s gene panel. It did not establish cutoffs for HMOX1, CCL2, or any other transcript that distinguish exposed from unexposed humans with known sensitivity and specificity. A urine THC-COOH finding, a positive maternal oral-fluid result, a prenatal self-report, and a muscle transcript profile are materially different observations.

For a case involving biological evidence, consult principles governing biological matrices in toxicology and the contemporaneous specimen-collection record. Testing methods, actual analytes, collection times, storage, and quality assurance must be established separately. One cannot use a controlled animal dosing regimen as a conversion table for a human maternal blood concentration or retroactively attribute a child’s muscle condition to a specific cannabis exposure.

What would be needed to support developmental-muscle causation in a real case?

A claimant-specific analysis would need to identify the alleged physical condition, objective clinical findings, its timing, competing congenital or acquired explanations, reliable exposure data, and evidence that links the particular exposure to the diagnosed outcome. Temporal sequence alone does not close that inferential gap. Animal-model findings can contribute biological plausibility without proving the exposure, pathology, magnitude, or causal mechanism in an individual child.

The study’s own endpoints provide an example of why evidentiary levels matter. The researchers found a difference in selected RNA transcripts, interpreted associated pathway sets, but did not detect collagen-area separation or measure movement and muscle force. Translating the pathway result into “prenatal THC caused congenital fibrosis” would add at least two unmeasured propositions: that altered transcription caused persistent pathological remodeling, and that such remodeling produced an actual functional injury in the individual at issue.

Forensic and medical experts should also account for species differences, oral-dose schedule, exposure timing, sample-size limitations, endpoint selection, fixed-tissue RNA preservation, and multiple gene comparisons. Similarities between rhesus macaques and humans support modeling relevance; they do not validate a human person-specific diagnostic classification.

How should the study be described in scientific testimony or research writing?

A defensible summary identifies the model, the administered exposure, the measured skeletal-muscle tissue, both the positive molecular results and the negative collagen result. It should not conflate “fetal” and “infant” cohorts, imply the same animal was followed longitudinally through both tissue collections, or mislabel a gene’s fold-change as a change in blood THC level. The original authors regarded their work as a pilot investigation that establishes feasibility and identifies hypotheses for further study.

For related research methods and interpretive guidance, the site’s forensic toxicology evidence library organizes discussions of assays, specimens, and evidentiary limits. This primary paper’s precise contribution is to show how a laboratory-detectable transcript difference can coexist with a non-significant tissue-level outcome. Neither result should be ignored to favor a predetermined legal narrative.

Frequently asked questions

1. Did prenatal THC produce muscle fibrosis in this study?

The investigators reported no significant difference in collagen-stained area between exposed and control rectus femoris samples (p > 0.05). This is not proof of strict equivalence or of clinically normal muscle function. Histologic fibrosis was not demonstrated by their chosen quantitative collagen outcome.

2. Does 33 differentially expressed genes mean 33 injured infant muscles?

No. Thirty-three refers to distinct transcripts meeting the authors’ statistical and fold-change rules in the infant group comparison. It is not a count of injured animals, damaged fibers, or diagnosed conditions. One infant RNA specimen was excluded for assay quality.

3. Was THC measured in fetal or infant blood?

The key exposure was maternal administered oral THC, whereas the reported outcomes were RNA signals and staining in offspring rectus femoris tissue. The paper did not supply a paired forensic whole-blood THC concentration for each gene or collagen result.

4. Did the researchers test whether offspring could walk or had muscle weakness?

No performance endpoint of that kind was reported for the study’s primary tissue analyses. The work therefore cannot establish actual childhood motor impairment merely from pathway enrichment or differential expression.

5. Why did the fetus and infant cohorts have different gene counts?

They represent different developmental stages, different animals, tissue contexts, and partly different analyzable sample sets. The fetal comparison yielded 13 reported transcripts and infant comparison 33; the counts alone do not prove cumulative worsening within an animal over time.

6. Are low p-values evidence that a particular person’s illness was caused by THC?

No. A gene-level p-value characterizes a model-based group comparison under assumptions; it does not give the probability of individual causation. Establishing an individual exposure and injury requires separate scientifically valid evidence.

7. What is the most important limitation for courtroom use?

The study is a small rhesus-macaque pilot with molecular and collagen outcomes, not a validated human diagnostic or dose-response investigation. It can inform mechanistic plausibility while leaving human individual injury, fetal THC concentrations, long-term motor function, and event-specific causation unresolved.

Bottom line

Moellmer and colleagues showed that chronic experimentally administered maternal THC was associated with selected fetal and infant rhesus-macaque muscle RNA changes—13 and 33 genes, respectively—with enriched signaling related to inflammation and cellular regulation. They did not show a significant difference in the measured collagen-stained area or directly demonstrate a clinical muscle disorder. Scientific interpretation and expert testimony must preserve that distinction between molecular signal, tissue phenotype, and individual causation. Original study: Moellmer et al., PLOS ONE (2024), doi:10.1371/journal.pone.0306868.