Dimethyl fumarate therapy in a Western diet pregnancy model: a pilot study in guinea pigs

Abstract

Background

Maternal Western diet (WD) consumption during pregnancy is linked to adverse gestational and offspring metabolic outcomes. Dimethyl fumarate (DMF), a Nrf2 activator with antioxidant and anti-inflammatory properties, is a proposed therapeutic candidate, but its safety and efficacy during pregnancy remain unexplored. This pilot study aimed to establish a guinea pig model of gestational WD consumption and evaluate the feasibility of DMF treatment on maternal, placental, and fetal outcomes.

Methods

Pregnant guinea pigs consuming a modified WD were randomized to receive either DMF or vehicle treatment beginning at gestational day 19. Maternal weight, urinalysis, blood glucose, and ketones were monitored weekly. Dams were sacrificed near-term, and placental and fetal biometric measures were recorded. The placenta was evaluated for a panel of antioxidant and angiogenic candidate genes via RT-qPCR, and maternal and fetal livers were assessed for triglyceride content, histopathology, and transcriptomic changes via RNA-sequencing.

Results

DMF did not alter maternal weight, blood glucose, urinalysis, nor fetal or placental gross outcomes. Hepatic triglyceride concentrations were unaffected in both dams and fetuses. WD consumption resulted in maternal hepatic glycogen accumulation and mild vacuolar hepatopathy without fibrosis, while fetal livers exhibited microvesicular hepatopathy and abundant extramedullary hematopoiesis without fibrosis, irrespective of treatment. Placental RT-qPCR revealed that exposure to DMF led to a decrease in transcription of Sod1 (p = 0.046). For Hmgcr (p = 0.071) and Nqo1 (p = 0.067), numerical decreases were observed, although these did not meet the threshold for statistical significance. RNA-sequencing identified 38 differentially expressed genes (DEG) in maternal livers including changes in some immune-related transcripts. Fetal hepatic RNA-sequencing identified 51 DEG, distinctly different from maternal DEG, indicating an activation of interleukin 12 production. Distinct differences between fetal and maternal liver transcriptomic profiles were evident, with over 9,000 DEG (65% of annotated transcripts). Dimethyl fumarate conferred no measurable benefit on maternal or fetal hepatopathy and had a paradoxical effect in lowering placental Sod1 gene expression, but, importantly, no overt harm was appreciated. These findings establish the feasibility of a gestational guinea pig WD model and gestational DMF therapy and provide initial insights into maternal-fetal responses to DMF, most notably highlighting divergent hepatic transcriptome between dams and offspring. Future research should assess whether DMF or related Nrf2 activators can be developed as pregnancy therapeutics, explore interactions between antioxidants and placental angiogenesis, and determine long-term offspring outcomes after exposure to WD and DMF during gestational development.

Introduction

Child health and development appear to be directly affected by the health status of the mother. The global prevalence of maternal obesity and gestational diabetes mellitus (GDM) continues to increase and is associated with adverse pregnancy outcomes (APO) with transgenerational consequences, including putting offspring at higher risk of cardiovascular disease, obesity, and poorer cognitive performance (Hunt & Schuller, 2007; Godfrey et al., 2017; Strauss, 2021; Kim & Ayabe, 2023; Almutairi, Alsaykhan & Almatrood, 2024). Consumption of a Western diet (WD), characterized by the dietary intake of ultra-processed food and drinks that are high in sugar and fat, is highly associated with the development of metabolic diseases like obesity and diabetes (Quan et al., 2021; Elsakr et al., 2021; Machairiotis et al., 2021; García-Montero et al., 2023). With the increasing incidence of metabolic disease among women of reproductive age in the Western world (Vohra et al., 2024), there is a need to develop therapeutic strategies to mitigate the negative effects of consuming a WD during pregnancy.

Gestational metabolic diseases subject the developing fetus to stressors and insults that lead to long-term detrimental effects, such as programming for cardiovascular disease and diabetes (Entringer, 2013; Wesolowski et al., 2017; Tong & Kalish, 2021; Reichetzeder, 2021; Hill & Hill, 2024). The placenta serves as the critical point of influence where maternal health or disease can shape fetal development, aligning with David Barker’s hypothesis for developmental origins of health and disease (Barker & Osmond, 1986; Wadhwa et al., 2009; Almond & Currie, 2011). Targeting this interface, particularly the placenta, offers a promising opportunity to intervene and protect fetal outcomes and long-term health. It is crucial to understand the mechanisms that incite placental injury and how the placental response to stress and inflammation impacts the fetus in pregnant women consuming a WD. Gestational WD consumption is associated with redox imbalance and inflammatory signaling, both of which are implicated in APO and the epigenetic programming of metabolic diseases (Frias et al., 2011; Abbasi, Bakshimoghaddam & Alizadeh, 2021; Perry, Stephanou & Rayman, 2022; Nash et al., 2023a, 2023b). Thus, targeting systemic oxidative stress and inflammation during gestation is a promising strategy to mitigate placental dysfunction and improve offspring outcomes.

Dimethyl fumarate (DMF) is an FDA-approved drug with potent antioxidant and anti-inflammatory effects primarily through the activation of the Nrf2 pathway and inhibition of NF-κB signaling (Bresciani et al., 2023; Kourakis et al., 2025). Nrf2, a transcription factor more formally known as Nfe2l2, is a master regulator of cellular antioxidant responses, controlling the transcription of genes involved in detoxification, redox homeostasis, and cytoprotection (Rushworth & MacEwan, 2011; Vomund et al., 2017; He, Ru & Wen, 2020). Activation of Nrf2 in the placenta may enhance its resilience to oxidative insults associated with maternal stress, supporting nutrient transport and vascular integrity. Simultaneously, suppression of NF-κB may reduce inflammation in the placenta. Together, these dual actions of DMF offer a compelling approach to ameliorate placental stress, prevent maladaptive fetal programming, and reduce the long-term risk of metabolic disease in offspring exposed to WD in utero.

The guinea pig (Cavia porcellus) is an excellent non-primate model for placental physiology due to key similarities with humans in morphology, implantation, development, steroidogenesis, and function (Mess, 2007; Morrison et al., 2018; Aguilera et al., 2022; Canizo, Zhao & Petropoulos, 2025). Unlike the rat or mouse, the guinea pig possesses a hemochorial placenta—structurally and functionally analogous to that of humans—with deep trophoblast invasion and extensive remodeling of maternal spiral arteries (Thompson et al., 2016; Aguilera et al., 2022). Its extended gestation and precocial fetal development further enhance translational relevance, allowing for a more accurate modeling of human fetal adaptations to suboptimal intrauterine environments. As such, the guinea pig offers a powerful platform to examine how WD-induced metabolic disturbances during pregnancy influence placental function and long-term offspring health.

This study aims to establish a guinea pig model for human gestational WD consumption, with the secondary aim of testing the feasibility of DMF treatment in gestational guinea pigs. This pilot study explored the feasibility of using a custom, modified WD diet and its effect on maternal blood glucose, hepatopathy, and the ability to conceive and sustain pregnancy. To our knowledge, it is the first study to characterize fetal hepatic and placental transcriptomic responses following maternal DMF exposure. We hypothesized that the administration of DMF in the context of gestational WD consumption would be tolerated by both dams and fetuses without resulting in adverse pregnancy outcomes. To evaluate this, we examined gross fetal and placental outcomes, characterized the hepatic transcriptome of dams and fetuses, and assessed the expression of key antioxidant and inflammatory genes in placental tissue.

Materials and Methods

Study design

This was a randomized, controlled study conducted in pregnant Hartley guinea pigs. A total of seven animals were used, including one boar used exclusively for breeding purposes and six nulliparous sows weighing 507 ± 13 g (mean ± SD) at acquisition which served as experimental units. Pregnancy was confirmed by transabdominal ultrasound at gestational day 19 (GD19), which aligns with the reported timepoint for reliable, early ultrasonographic pregnancy detection in guinea pigs (approximately GD21) (Wilson et al., 2021). Following confirmation of pregnancy at GD19, pregnant sows were assigned to one of two treatment groups: vehicle (n = 3) or dimethyl fumarate (DMF) (n = 3). Control animal received a daily oral gavage of the vehicle while treated animals received DMF suspended in methyl cellulose. All animals were fed a Western diet prior to and throughout pregnancy, no chow-fed control group was included. This design was chosen for this pilot study to specifically evaluate the effect and feasibility of DMF treatment in pregnant sows consuming a Western diet. There were two to four fetuses per litter for a total of nineteen fetuses (n = 10 from control dams) (n = 9 from DMF-treated dams).

The experimental unit for maternal outcomes was the individual dam. For placental and fetal outcomes, the litter was considered the experimental unit, with offspring nested within dam; this non-independence was accounted for statistically by including dam as a random effect in a mixed-effects model. No animals were excluded from analysis.

Sample size determination

The sample size for this exploratory, mechanistic pilot study was determined based on feasibility and ethical considerations, with the goal of minimizing animal use while generating preliminary data to inform a larger, adequately powered study. No a priori power calculation was performed.

Rearing & husbandry

All methods were approved by the Oregon State University Institutional Animal Care and Use Committee (IACUC) under protocol #2024-0452. All animals were housed at the Oregon State University Laboratory Animal Resources Center (LARC) under a 14:10 light-dark cycle (lights on at 06:30, lights off at 20:30) with ambient conditions maintained at 21 °C and 30% humidity. One boar and six nulliparous, 6-week old female Hartley guinea pig sows were acquired from Elm Hill labs (Chelmsford, MA, USA). Sows were randomly pair-housed while the boar was housed individually unless being used for breeding. During the 2-week acclimation period, dams were inspected daily and were weighed twice weekly. All animals had ad libitum access to shelters, enrichment, chow, hay, and water, although chow was withheld during breeding periods.

Diet

The boar was fed a standard guinea pig chow diet throughout the experimental window. The females began on the standard chow-based diet during acclimation and until baseline urine samples could be collected, after which the WD was introduced as 10% of the ration. Over the course of a week, the percentage of WD was gradually increased in 20% increments until WD fully replaced the chow. Animal weights were monitored closely during this period. Full acceptance of the WD was confirmed after an 8-h residual test during the light phase, a timing chosen based on the diurnal feeding behavior of guinea pigs (Elfers, Armbrecht & Mazzuoli-Weber, 2021). In this test, measured amounts of WD and regular chow were placed in separate crocks in each cage and consumption was determined by weighing the remaining food at the end of the interval. Animals consistently consumed the WD but left the chow, after which the diet consisted entirely of WD.

The modified WD (40% kcal from fat, 15% from sucrose, and 0.35% from cholesterol) was provided by Inotiv (CAT #TD.240402; Teklad Custom Diets, Madison, WI, USA). The diet was modified to contain less sucrose than traditional WD models to mitigate the risk of animals developing gestational diabetes. Diet components were provided by Inotiv and are summarized in Table 1, and energy provided in Table 2.

Table 1. Western diet ingredients in g/Kg as fed.

Formula	g/Kg
Casein	90.0
Isolated soy protein	85.0
L-Methionine	3.5
L-Arginine HCl	3.0
Corn starch	214.6
Maltodextrin	89.3
Sucrose	150.0
Cellulose	89.3
Guar gum	22.3
Anhydrous milkfat	178.5
Corn oil	8.93
Cholesterol	3.03
Vitamin mix (Teklad 40060)	10.0
Vitamin C (L-ascorbyl-2-polyphosphate)	0.75
Folic acid	0.01
Mineral mix (Teklad 98057)	18.0
Calcium phosphate, dibasic	18.0
Calcium carbonate	7.0
Potassium citrate, monohydrate	3.0
Potassium chloride	2.0
Sodium chloride	2.0
Magnesium oxide	1.5
Ferric citrate	0.2
Magnesium sulfate, monohydrate	0.1
TBHQ (antioxidant)	0.04

Table 2. Western diet macronutrient composition.

	% by weight	% kcal
Protein	15.8	15.6
CHO	43.3	42.6
Fat	18.9	41.9

Note:

Macronutrient (% by weight) and energy composition (% kcal) in WD as fed.

Blood glucose and ketone measurement

The ear was briefly warmed by rubbing and palpated to promote blood flow. The guinea pig was restrained in the weigh box with the lid removed, allowing secure but gentle immobilization in a familiar position to minimize stress-induced hyperglycemia. A viable capillary was visually identified for sampling alternating ears between events to minimize tissue trauma. Once selected, the capillary was lanced with one swift prick of a Microlet lancet (CAT#10000096-002; Ascensia Diabetes Care, Parsippany, NJ, USA). Once a blood droplet formed, it was loaded onto a compatible, fresh test strip attached to the blood ketone meter (CAT# 860001641752; KetoBM, Sheridan, WY, USA) and then into the glucometer (CAT# 301939763012; Ascensia Diabetes Care, Parsippany, NJ, USA). Reported values were immediately recorded. The lanced ear was cleaned and pressure applied until hemostasis was achieved.

Urine collection

Sows were placed in plastic individual guinea pig cages without bedding, with access to WD and water starting at 07:00 for up to 6 h once per week for urine collection. If the initial collection attempt was unsuccessful, a second attempt was made the following day. If the second attempt was unsuccessful, no additional urine collection attempts were made until the following week, only the ear was pricked for immediate evaluation of blood glucose and ketones to screen for gestational diabetes and ketosis, which were considered humane endpoints. During urine collection, sows were observed every 30 min for urine voiding and urine was collected immediately after observation. If fecal contamination was observed, the urine was not collected and the animal was transferred to a clean cage. Feed and water contamination was mitigated by providing a shallow ration in the crocks to avoid spillage. Once a sample of urine was observed on the clean surface of the plastic cage, the animal was immediately removed and placed back into the housing cage. The urine was aspirated using a transfer pipette and a small amount of urine was immediately applied to a urinalysis reagent strip (CAT# UO31-101; ACON Laboratories, San Diego, CA, USA) and the observations recorded (leukocytes, nitrite, urobilinogen, protein, pH, blood, specific gravity, ketone, bilirubin, and glucose). The remaining urine was transferred into 1.7 mL tubes which were stored at −80 °C. If after 6 h, the sow did not void, she was placed briefly with her pen-mate in the plastic guinea pig cage and closely observed for urinary expulsion; it was learned that social stimulation sometimes facilitated urine voiding.

Breeding

The breeding protocol used in this study was based on previously published methods for timed mating in guinea pigs, specifically those described by Wilson et al. (2021) by trained personnel to minimize stress and discomfort. Sow vaginal membranes were observed twice daily until evidence of membrane rupture was identified at which point the WD was removed from the cage, and the male was introduced into the cage. During breeding windows, the chow was withheld, and animals had ad libitum access to hay and water only. The breeding window began when vaginal membranes were observed to be ruptured and extended for two nights and 3 days (72 h). Every 12 h, the male was removed from the cage and offered regular chow while the females were offered the WD diet for 2 h. Following the mealtime, the WD was removed from the cage and the male reintroduced. At no point did the male have access to the WD.

Beginning 14 days after the vaginal membranes were first observed to be ruptured, the sow’s vaginal membranes were re-examined for rupture twice daily for 2 days. If they remained intact within 16 days after the initial breeding window, it was presumed the initial breeding was successful and the sow would undergo ultrasound confirmation of pregnancy daily between days 18 and 22 after initial exposure to the boar. An Esoate MyLab30 VET portable ultrasound was used for transabdominal ultrasound exams. Guinea pigs were gently restrained in dorsal recumbency, the abdomen was shaved, and conductive gel was applied to the abdomen. Sonographic exams did not exceed 5 min to minimize animal distress and discomfort. Gestational day 19 (GD19) was determined on the basis of the first day when gestational sacs could be visualized via ultrasound. If a sow did not conceive after initial exposure to the boar, the sow would be observed and bred using the previously described method.

Dosage and administration of treatment

After pregnancy could be confirmed by ultrasound (GD19), the animals began receiving the treatment (DMF suspended in 1% methylcellulose) or vehicle (1% methylcellulose) once daily at 07:00 until study termination. Because of the very low solubility of DMF in water (Kai et al., 2013), a viscous 1% solution of methylcellulose was used as a suspension vehicle for delivery. Methylcellulose is commonly used as an inert vehicle for oral gavage in rodent studies (Ikeda et al., 2024; Pitarokoili et al., 2015). Within each randomly assigned pair in the cage, one animal was randomly allocated to the treatment group and the other to the control group; cage location and dosing order were not controlled. The dose of DMF derived from the human equivalent dose (HED) (240 mg; 4 mg/kg BW) using a guinea pig conversion factor of 4.6 (Nair & Jacob, 2016), which resulted in 18.4 mg/kg BW. This pilot study used a single dose for translational relevance; pharmacokinetic or pharmacodynamic data in guinea pigs is currently lacking.

DMF (#A10402-22; ThermoFisher, Waltham, MA, USA) was suspended in 1% food-grade methylcellulose at 1:1 ratio (w/w). For dosing, 0.037 mL of this suspension per kg body weight was drawn for each animal, providing the calculated 18.4 mg/kg dose of DMF. This small volume of DMF suspension was then topped up with additional 1% methylcellulose to reach a final oral gavage volume of 0.5 mL per animal. Control animals received 0.5 mL of 1% methylcellulose without DMF. Animals were briefly restrained in a standard manner to administer the oral gavage which was done by the same handler to minimize variability.

Dam termination & tissue collection

Between GD52 and GD57, sows were anesthetized with isoflurane and euthanized via pentobarbital overdose following intracardial heparinization under ultrasound guidance. The sow was placed into an induction chamber with 5% isoflurane vapor and 1 L/min O₂ until surgical plane anesthesia was achieved, as confirmed by the absence of palpebral and pedal withdrawal reflexes. Once this was achieved, the animals was immediately placed in dorsal recumbency, the thorax clipped and conductive gel applied. Ultrasound guidance was used to guide a 20 g × 1.5-inch hypodermic needle into the heart. When blood could be drawn into the syringe, five hundred USP porcine heparin was injected intracardially to prevent post-mortem coagulation and facilitate high-quality tissue collection for histology and RNA-seq applications. Following heparinization, the dam was euthanized with pentobarbital overdose (VINV-CIII-0015; Dechra Vet Products, Boston, MA, USA). An approximately six-centimeter midline incision was made beginning near the pubic symphysis and extending to the diaphragm of the dam, the uterine horns were manually removed, and the left lateral lobe of the liver was removed. A 3-g portion of liver was placed in 4% formalin and a small piece was placed in a cryovial and flash frozen in liquid nitrogen.

Fetal and placenta measurements and tissue collection

Fetuses were removed from the uterine horns immediately following maternal euthanasia, and a clamp was applied to the umbilical cord prior to severing it. The fetus was weighed (without the clamp) on a calibrated digital gram scale, and crown-rump length and biparietal length were measured using calibrated digital calipers. Measurements were obtained using consistent anatomical landmarks and techniques by the same investigator for each fetus. Tissue collection was performed as rapidly as possible and in a consistent manner across all animals to minimize ischemia time. Following the extraction of the fetus, each placenta was dissected from the uterine horns and weighed on a digital gram scale, and the diameter was measured using calibrated digital calipers. The same instruments were used throughout the study to ensure internal consistency of measurements. Sex was determined during fetal necropsy by visualizing internal sex organs. The left lateral liver lobe, and a portion of the placenta were placed immediately in 4% formalin. A portion of the placenta and the left medial liver lobe were placed in individual cryovials and immediately flash-frozen in liquid nitrogen. Formalin-fixed and frozen tissues were stored in temperature-controlled environments until batch processing, which was performed after collection of all tissues. Due to staggered euthanasia time points, storage duration varied across samples. Formalin fixed-tissues were stored in fixative at controlled room temperature and processed within 3–6 weeks of preservation, while frozen tissues were stored at −80 °C and processed within 5–8 weeks of preservation.

Hepatic triglyceride quantification

Triglycerides in fetal and maternal hepatic tissue were measured using a commercial kit following the manufacturer instructions with some modifications (CAT# 10010303; Cayman Chemical, Ann Arbor, MI, USA). The protocol was modified in the following ways: tissue was weighed on a digital scale and homogenized using a Bullet Blender Next Advance (Laboratory Instruments, Santa Clara, CA, USA) for approximately 5 min in 1.5 mL of the diluted NP40 substitute assay reagent with 1 mM EDTA; after centrifugation and transfer, the supernatant was frozen at −80 °C; to conduct the assay, the supernatant was sonicated at 37 °C until homogenous and diluted 10-fold in NP40 substitute assay reagent. Final concentration was normalized by total sample mass.

Hepatic RNA-sequencing

The RNA isolation from frozen fetal and maternal hepatic tissue were carried out using the RNeasy Mini Kit (CAT# 74101; Qiagen, Germantown, MD, USA) with some modifications to optimize RNA quality including surface decontamination with RNase Zap (Cat# AM9780; Thermo Scientific, Waltham, MA, USA). Thirty milligrams of frozen tissue was weighed out and immediately submerged in 1.2 mL ice-cold TRIzol^® Reagent (Cat# 15596018; Ambion, Carlsbad, CA, USA). Tissue was homogenized using a Bullet Blender Next Advance (Laboratory Instruments, Santa Clara, CA, USA) for 1-min increments and rested on ice for 1 min in between until no pieces of tissue could be visualized. Then 240 μL of molecular biology grade chloroform (Cat# J67241.AP; Thermo Scientific, Waltham, MA, USA) was added to the tube. The tubes were briefly vortexed and then centrifuged at 13,000× g for 15 min at 4 °C. The upper phase was removed from the tube and mixed with an equal part of 70% ethanol before being transferred to a spin column and following the manufacturer’s instructions for the rest of the RNA isolation process. Quality of RNA was evaluated by Bioanalyzer 2100; RNA quantification ranged from 319–1,900 ng/μL prior to dilution; RNA integrity number (RIN) for submitted samples was >7.7 (8.8 ± 0.42).

The RNA was sent to the Center for Quantitative Life Sciences (CQLS; RRID: SCR_018373) at Oregon State University for sequencing using the Illumina NextSeq2000. Prior to sequencing, samples were processed using 3′ mRNA library prep (Lexogen QuantSeq), and accurate quantification was performed using qPCR on a pool of the samples and quantified using Qubit. The sequencing was performed as 100 bp single-end. The reads were analyzed by the CQLS by trimming using fastp (https://github.com/OpenGene/fastp), rRNA removal using SortMeRNA (https://github.com/sortmerna/sortmerna), alignment using STAR (https://github.com/alexdobin/STAR), and quantification using Salmon (https://github.com/COMBINE-lab/salmon). The final annotated dataset for the Cavia porcellus (Feb 2008 Broad/cavPor3) with raw count was used in RNAseqChef (Etoh & Nakao, 2023). In the RNAseqChef software, datasets for the dams and the pups were uploaded separately as raw counts + metadata. For treatment effects, pair-wise DEG analysis was conducted using EBSeq2, which is recommended over DEseq for a small number of samples (n = 3 per group) (Etoh & Nakao, 2023). For the comparison between maternal and fetal hepatic transcriptome (n = 6 per group), DEseq was employed. Differentially expressed genes were identified using a Benjamini-Hochberg false discovery rate (FDR) threshold of <0.05, with additional inclusion criteria of 0-fold change and a base mean ≥4. Enrichment analysis was performed in RNAseqChef, indicating Mus musculus as a species (as it has a better annotation than guinea pig), and results for the KEGG pathways and Gene Ontology biological process were generated.

All sequencing data generated in this study are publicly available in the NCBI Gene Expression Omnibus (GEO) under accession number GSE312339.

Histology and staining

Formalin-fixed tissues were embedded in paraffin and sectioned (4 µm) by the Oregon State University Veterinary Diagnostic Laboratory. Adjacent tissues were stained with hematoxylin (Cat# 6765010; Epredia, Kalamazoo, MI, USA) and eosin (Cat# 676608; Epredia, Kalamazoo, MI, USA) or periodic acid Schiff (PAS) (CAT# FX2109; Cancer Diagnostics Inc., Durham, NC, USA), or Sirius Red. Hematoxylin and eosin (H&E) and PAS staining were performed according to laboratory-specific standard operating procedures, and Sirius Red (SR) staining was informed by the protocol provided by Dr. John A. Kiernan and the University of Western Ontario, Canada (Kiernan), but Weigert’s counterstaining was excluded. These protocols are included in File S1.

RTq-PCR of placental target genes

RNA isolation from frozen placentas was carried out using the RNeasy Mini Kit according to the manufacturer’s instructions; surfaces were decontaminated with RNase Zap (Cat# AM9780; Thermo Scientific, Waltham, MA, USA). Placental tissue was sampled without stratification by anatomical region prior to downstream analysis. Complementary DNA (cDNA) was synthesized from total RNA using the Takara EcoDry Premix (Oligo dT) system according to the manufacturer’s instructions (Cat# 639543; Takara Bio, Mountain View, CA, USA) and stored at −20 °C. Quality of RNA was evaluated by Bioanalyzer 2100; RNA quantification ranged from 563–2,788 ng/μL prior to dilution; RNA integrity number (RIN) for submitted samples was 6.5 ± 1.3. Six technical replicates per biological sample were included.

Primers for target genes were designed from NCBI RefSeq mCavPor4.1 Cavia porcellus reference genome (University of Maryland, 2023) and manufactured by Integrated DNA Technologies (IDT, Coralville, IA, USA). RTq-PCR was conducted according to a previously established protocol (Rosa et al., 2018). Five reference genes were tested using geNorm (Vandesompele et al., 2002) as previously described (Rosa et al., 2018): Actb, Gapdh, Rps9, Rps15, and Uxt. geNorm indicated that only Rps9, Rps15, and Uxt were adequate as reference genes as they had an M-value <1.5. The normalization factor was calculated using those three genes with a V-value of 0.39. Primer pairs for target and reference genes are provided in Table 3.

Table 3. RTqPCR primer sequences.

Symbol	Gene name	Accession #	Primer sequence
Actb	Actin beta	XM_005003653.3	F–ACCCAGATCATGTTCGAGACTTTC
			R–TCACCGGAGTCCATCACGAT
Cat	Catalase	NM_001439566.1	F–GCAACTACCCCTCCTGGACTT
			R–TGAGGCCAAATCTTGGTGAGA
Gapdh	Glyceraldeyde-3-phosphate dehydrogenase	NM_001172951.1	F–ATGATTCTACCCACGGCAAGTT
			R–GGATCTCGCTCCTGGAAGATG
Gst	Glutathione S transferase	NM_001172839.1	F–GGCTGCAGCGGGAGTAGA
			R–CAGAAGCCCATCATTTTTTAACTTG
Hif1a	Hypoxia inducible factor 1 alpha	XM_013155206.2	F–GGTGGATATGTCTGGGTTGAAAC
			R–ATACCACTCACAACATAATTTACACACACA
Hmgcr	3-hydroxy-methylglutaryl-CoA reductase	XM_023564436.1	F–TTGCCAGAATAGTGTGTGGTACTGT
			R–CTTTGACCTGTTGTGAATCATGTG
Hmox1	Heme oxygenase 1	XM_023561275.2	F–CCAGCAGATCTGGTTATGGAGTT
			R–CCTGGGTGTGCACCTCCTT
Nfe2l1	Nuclear factor erythroid derived 2,-like factor 1	XM_023563871	F–GAAGCAGTGCCTAGTGAGAACGA
			R–GCTTCCATGGCCTGCATCT
Nfe2l2	Nuclear factor erythroid derived 2,-like factor 2	XM_013142514.2	F–TCCCAGGTTTCTTCACCTACATTT
			R–GCCTGATTAGGTGCAGTGAAGAC
Nfe2l3	Nuclear factor erythroid derived 2,-like factor 3	XM_003467935.4	F–GCAGCAGCACAATAATGAGAATAAG
			R–TCCATTTAGGTGTCTCTCATTTCTAGAT
Nqo1	NAD(P)H quinone dehydrogenase 1	NM_001172986.1	F–CCGAGTCTGCTCTAGCATATAAAGAA
			R–AAGTGGGAACTGAAATATCAGAAGGT
Rps9	Ribosomal protein S9	XM_005001732.3	F–GCCGGTGGCAAGAAGCT
			R–AAACCGTATTCGCCGATCAG
Rps15	Ribosomal protein S15	XM_003460968.4	F–GGCGGTCCTTTCCGAAGA
			R–TCAGCTGCTCGTAGGACATGTC
Sod1	Superoxide dismutase 1	XM_003467248.4	F–CACTTCGAGCAGAAGGCAAAC
			R–TATCGCCAAACTCATGGACATG
Uxt	Ubiquitously expressed transcript	XM_005000244.3	F–CATTGAGCGACTCCAGGAAACT
			R–GCGTGAAGTATCAGAGACCACTGT
Vegfa	Vascular endothelial growth factor A	XM_013156845.1	F–GCAAGGCAAGAAAAAAAATCAGTT
			R–GCTTTCTCCGCTCTGAGCAA

Note:

Exon-exon junction underlined.

Statistical analysis

Except for the RNAseq data, the statistical analyses were conducted in R (version 2024.12.1+563). Data were analyzed using linear mixed-effects models to evaluate the effect of treatment on maternal, placental, and fetal outcomes using the lmer function from the lme4 package in R (Bates et al., 2015). Models were fit using restricted maximum likelihood (REML). Treatment (vehicle vs. dimethyl fumarate) was specified as a fixed effect. Gestational day at sacrifice and litter size were evaluated as biologically relevant covariates a priori and included where appropriate to avoid overparameterization given the sample size. To account for the non-independence of offspring within the same dam, a random intercept for dam was included. Denominator degrees of freedom for fixed effects were estimated using the Satterthwaite approximation as implemented in the lmerTest package (Kuznetsova, Brockhoff & Christensen, 2017). Significance threshold established at p < 0.05; p-values between 0.05 and 0.10 are reported for descriptive purposes. Model assumptions were evaluated through visual inspection of residuals. Adjustments for multiple comparisons were not performed and should be considered when interpreting results.

Blinding

Treatment allocation was not blinded during treatment administration, sample collection, or data analysis. The pathological evaluation of tissue slides was conducted in a blinded and independent manner.

Results

Breeding

Conception was achieved the first time for five out of six sows (83% conception rate). The sow that did not conceive belonged to the control group and she conceived on the following cycle.

Gestational weight

Dimethyl fumarate did not influence maternal weight gain during gestation adjusted for litter size (Fig. 1). Dimethyl fumarate had no effect on gestational weight compared to controls (p = 0.66) and there was no treatment by time interaction (p = 0.58).

Figure 1. Maternal body weight.

Beginning 1 week prior to mating (week-1) through gestational week 7 for vehicle (VEH) and dimetyl fumarate (DMF) treated dams adjusted for litter size. Error bars indicate standard error of the mean.

Urinalysis, blood glucose, and ketones

Data is provided in File S2. Throughout the study, the urinary data varied slightly or were negative or below the detectable limit. Blood glucose was also checked weekly. Blood glucose was within acceptable limits for guinea pigs (126 mg/dL ± 27). There was no treatment or time effect on guinea pig urinalysis or blood glucose (File S3). The concentration of blood ketones was also not affected by treatment or time or their interaction (File S3). The tests demonstrated poor reproducibility within the same animal, and thus data should be taken with caution.

Fetal gross parameters

Analysis indicated no significant difference in fetal crown-rump length, body weight, or biparietal diameter between groups (Fig. 2). A summary of fixed effects is provided in Table 4.

Figure 2. Fetal gross parameters at term.

(A) Fetal weight in grams, (B) crown-rump length in millimeters, and (C) biparietal diameter in millimeters measured at termination represented as group means of vehicle (VEH) and dimethyl fumarate (DMF) treated dams. Error bars indicate the standard error of the mean.

Table 4. Table of fixed effects.

Outcome	Fixed effect	Estimate	SEM	df	p-value
Fetal length (mm)	Treatment a	0.15	5.45	3.2	0.982
	Gestational days a	2.52	1.15	3.8	0.096
	Litter size	−2.33	3.23	4.1	0.510
Fetal mass (g)	Treatment a	7.91	7.8	3.1	0.386
	Gestational days a	4.19	1.9	3.7	0.099
	Litter size	−6.41	4.72	4.2	0.243
Biparietal diameter (mm)	Treatment a	0.01	0.64	4.6	0.990
	Gestational days a	0.40	0.12	3.3	0.039
	Litter size	−0.48	0.44	4.8	0.321
Placenta mass (g)	Treatment a	0.42	0.60	2.9	0.539
	Gestational days	0.07	0.23	3.7	0.772
	Litter size a	−0.72	0.31	4.3	0.076
Placenta diameter (mm)	Treatment	−0.44	1.59	4.1	0.794
	Gestational days	0.20	0.43	3.5	0.665
	Litter size	0.11	0.90	4.5	0.914
Fetal hepatic TG (mg/g)	Treatment	12.3	6.18	3.6	0.127
	Gestational days	3.45	1.62	2.8	0.130
	Litter size	−3.63	4.61	4.0	0.475
Dam hepatic TG (mg/g)	Treatment a	−3.31	2.77	4.6	0.248
	Gestational days a	1.51	0.57	4.7	0.017
	Litter size	1.04	1.54	5.7	0.508

Note:

Fixed effect, estimate, SEM, df, and p-values for gross parameters and hepatic triglyceride levels. Matching superscripts within each parameter indicate which continuous effects were included as covariates for treatment effect. Estimates are a result of dimethyl fumarate vs. vehicle treatment (DMF vs. VEH).

Placental gross parameters

Placental mass and diameter were similar between vehicle- and DMF-treated groups at term, with no apparent treatment-related differences (Fig. 3). A summary of fixed effects is provided in Table 4.

Figure 3. Placental gross parameters at term in vehicle (VEH) and dimethyl fumarate (DMF) treated dams.

(A) Placental mass in grams and (B) placental diameter in millimeters of vehicle and DMF-treated dams. Error bars indicate the standard error of the mean.

Fetal and maternal hepatic triglycerides

Fetal and maternal hepatic triglyceride concentrations were unaffected by DMF treatment compared with vehicle (Fig. 4). A summary of fixed effects is provided in Table 4.

Figure 4. Maternal and fetal liver triglycerides.

(A) Fetal and (B) maternal hepatic triglyceride levels in vehicle (VEH) and dimethyl fumarate (DMF) treated dams adjusted by litter size and gestational age. Error bars indicate the standard error of the mean.

RTq-PCR

Treatment with DMF resulted in a significant decrease in placental Sod1 transcription (p = 0.046), a numerical decreases in Hmgcr transcription (p = 0.071) and Nqo1 transcription (p = 0.067) (Fig. 5). No other genes were affected by DMF treatment in the placenta.

Figure 5. RTq-PCR gene expression for 10 key genes in placentas.

Vehicle (n = 5) and DMF (n = 4) treated fetuses adjusted by litter size and gestational age: (A) catalase; (B) heme oxygenase 1; (C) 3-hydroxy-3-methylglutaryl-CoA reductase (0.071); (D) hypoxia-inducible factor-1; (E) NAD(P)H: quinone oxidoreductase 1 (p = 0.067); (F) NFE2 like BZIP transcription factor 1 (Nrf1); (G) NFE2 like BZIP transcription factor 2 (Nrf2); (H) NFE2 like BZIP transcription factor 3 (Nrf3); (I) superoxide dismutase 1 (p = 0.046); (J) vascular endothelial growth factor. Error bars indicate standard error of the mean, significance denoted with * when p < 0.05.

Maternal hepatic histology

All dams demonstrated minimal-mild midzonal vacuolar hepatopathy. Histopathology revealed excessive glycogen storage, ~10% binucleated hepatocytes, ~10% hepatocytes with eccentrically displaced nuclei showing pyknosis or karyorrhexis, and macrovascular vacuolization. In addition, ~40–75% of hepatocytes contained eosinophilic homogenous intracytoplasmic material which displaced nuclei. Maternal livers demonstrated diffuse PAS staining (magenta color) (Fig. 6), which was abolished by amylase digestion (Fig. 7), consistent with glycogen accumulation. Occasional punctate magenta foci remained following amylase digestion (Fig. 7), likely representing diastase-resistant PAS-positive structures such as glycoproteins in membranes or sinusoidal basement membranes rather than glycogen. Liver sections were negative for Sirius Red staining, indicating the absence of collagen deposition and fibrosis (File S4).

Figure 6. Representative 20× images of PAS stained sow liver sections.

Top row (A–C): vehicle treated animals (1, 3, 5, respectively). Bottom row (D–F): dimethyl fumarate treated animals (2, 4, 6, respectively).

Figure 7. Representative 20× images of PAS stained, amylase digested sow liver sections.

Top row (A–C): vehicle treated animals (1, 3, 5, respectively). Bottom row (D–F): dimethyl fumarate treated animals (2, 4, 6, respectively). Open arrows indicate hepatocytes with vacuolar hepatopathy and triangles indicate bi-nucleated hepatocytes.

Fetal hepatic histology

Abundant, coalescing clusters of hematopoietic cells were present throughout the lobule, most prominently in the periacinar region. Rare binucleated hepatocytes and minimal nuclear atypia (karyomegaly, anisokaryosis) were noted. All fetuses demonstrated microvesicular hepatopathy. All liver sections displayed abundant coalescing clusters of extramedullary hematopoietic cells. Fetal livers demonstrated minimal granular PAS staining within the lysosome (Fig. 8) which was abolished by amylase digestion (Fig. 9), consistent with glycogen accumulation. Further, sections were negative for Sirius Red staining, indicating the absence of collagen deposition and fibrosis (File S4).

Figure 8. Representative 20× images of PAS stained fetal liver sections.

Top row (A–C): fetuses of vehicle treated dams (1D, 3A, 5A, respectively). Bottom row (D–F): fetuses of dimethyl fumarate treated dams (2A, 4B, 6A, respectively).

Figure 9. Representative 20× images of PAS stained, amylase digested fetal liver sections.

Top row (A–C): fetuses of vehicle treated dams (1D, 3A, 5A, respectively). Bottom row (D-F): fetuses of dimethyl fumarate treated dams (2A, 4B, 6A, respectively). Open arrows indicate hepatocytes with microvesicular hepatopathy, stars indicate extramedullary hematopoietic cell clusters, and triangles indicate bi-nucleated hepatocytes.

Maternal liver RNA-sequencing

A total of 38 DEG were identified, of which 11 were upregulated and 27 were downregulated in the DMF vs. VEH treatment group with FDR ≤ 0.05 (Fig. 10A). KEGG enrichment analysis identified eight significantly enriched pathways among the DEG. A total of five pathways were upregulated and three pathways downregulated by DMF treatment. Upregulated pathways showed significant enrichment based on adjusted p-value but lacked a calculated q-value due to small gene set size; these are indicated in grey in the KEGG dot plot (Fig. 10B). Gene ontology biological process analysis (GO:BP) identified eight significant enriched terms, of which the top five are shown in the dot plot (Fig. 10C). All data, plus the results of KEGG and GO:BP enrichment analyses, are provided in File S5. GO:BP enrichment analysis indicated apparent downregulation of the complement and coagulation cascade and upregulation of the glutathione metabolism and P450-related pathways via Mgst3.

Figure 10. Analysis of differentially expressed genes in maternal liver of VEH- and DMF-treated dams.

(A) Heatmap of differentially expressed genes in the liver of dams between dimethyl fumarate (DMF) and vehicle-treated (CTR) groups. Rows represent the 38 DEGs (11 upregulated in red, 27 downregulated in blue; FDR < 0.05, base mean ≥ 4), and columns represent individual samples. Expression values are variance-stabilized and row-scaled (Z-score) for visualization. Both genes and samples are hierarchically clustered, showing distinct expression patterns between groups. (B) Results of KEGG analysis with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR. (C) Results of Gene ontology biological process with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR.

Fetal Liver RNA-sequencing

A total of 51 DEG were identified, of which 38 were upregulated and 13 were downregulated in fetuses from DMF vs. fetuses from VEH dams with FDR ≤ 0.05 (Fig. 11A). The analysis identified three significantly enriched KEGG pathways among the DEG (all driven solely by Ptprf) (Fig. 11B). GO:BP analysis identified two significantly enriched terms, related to interleukin 12 production (Fig. 11C). All data, including the results of the KEGG and GO:BP enrichment analyses, are provided in File S6. GO:BP analysis of the fetal liver highlighted upregulation of pathways related to interleukin-12 (IL-12) production, driven by Ido1, Irf8, and Laptm5. KEGG analysis identified downregulation of pathways related to adherens junctions, cell adhesion molecules, and insulin signaling, though each was represented by a single gene (Ptprf).

Figure 11. Analysis of differentially expressed genes in fetal liver of VEH- and DMF-treated dams.

(A) Heatmap of differentially expressed genes in the liver of fetuses from dimethyl fumarate (DMF) and vehicle-treated (CTR) dams. Rows represent the 51 DEGs (38 upregulated in red, 13 downregulated in blue; FDR < 0.05, base mean ≥ 4), and columns represent individual samples. Expression values are variance-stabilized and row-scaled (Z-score) for visualization. Both genes and samples are hierarchically clustered, showing distinct expression patterns between groups. (B) Results of KEGG analysis with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR. (C) Results of Gene ontology biological process with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR.

Hepatic transcriptomic difference between dams and pups

No DEG were commonly affected by DMF in dams and fetuses with no overlap observed in either KEGG or GO:BP enrichment analyses (Fig. 12; File S7). Over 9,000 genes were differentially expressed between maternal and fetal liver (65% of annotated measured mRNA) (File S7). The MSigDB Hallmark analysis revealed DEG more expressed in dams vs. fetal liver are associated with terms related to bile and lipid metabolism, and DEG more expressed in fetal vs. dams liver are associated with terms related to heme metabolism and DNA replication (Fig. 12B). The KEGG analysis revealed enrichment of pathways related to protein metabolism and peroxisome in the DEG more expressed in maternal vs. fetal liver and enrichment of pathways related to DNA replication and cell cycle in DEG more expressed in fetal vs. maternal liver (Fig. 12C). The GO:BP analysis revealed enrichment of terms related to protein and lipid metabolism in DEG more expressed in maternal vs. fetal liver and terms related to DNA replication and RNA splicing in DEG more expressed in fetal vs. maternal liver (Fig. 12D).

Figure 12. Analysis of differentially expressed genes in fetal liver as compared to maternal liver.

(A) Heatmap of differentially expressed genes between the liver of dams and fetuses (a.). Rows represent the 9,584 DEGs (4,725 upregulated in dams vs. fetus in red, 4,859 downregulated in dams vs. fetuses in blue; FDR < 0.05, base mean ≥ 4), and columns represent individual samples. Expression values are variance-stabilized and row-scaled (Z-score) for visualization. Both genes and samples are hierarchically clustered, showing distinct expression patterns between groups. (B) MSigDB Hallmark analysis with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR. (C) Results of KEGG analysis with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR. (D) Results of Gene ontology biological process with q value and GeneRatio in DEG down- or up-regulated in DMF vs. CTR.

Discussion

Dimethyl fumarate uniquely influences maternal and fetal hepatic transcriptomes but does not ameliorate hepatopathy

Maternal and fetal liver phenotype

Hepatic triglyceride (TG) accumulation did not reach thresholds previously proposed for metabolic-associated steatotic liver disease (MASLD) (previously referred to as NAFLD) in humans (>55.6 mg/g) (Szczepaniak et al., 2005); however, these criteria have not been validated in guinea pigs and are therefore presented for reference only. Consistent with this, histological assessment did not reveal evidence of overt lipid accumulation. PAS staining with diastase confirmed that vacuolar structures were glycogen rather than lipid. Together, these findings indicate that under the conditions of this study, the Western-style diet was not associated with measurable hepatic steatosis after ~10 weeks of feeding, most of that time during pregnancy. This may reflect the limited duration of exposure, as previous studies have reported steatosis after ≥16 weeks of feeding a WD when hay was provided ad libitum (Tveden-Nyborg et al., 2016; Ipsen et al., 2016). Notably, those studies used similar dietary compositions and ad libitum hay, though with differences in fat sources. In a study where hay was withheld, diets containing 15–60% energy from fat from vegetable oil with 0.25% cholesterol achieved a steatotic model within 12 weeks in non-pregnant guinea pigs (Torres-Gonzalez et al., 2007; deOgburn et al., 2012). Likewise, Sinclair et al. (2018) reported maternal steatosis and increased fetal hepatic accumulation in guinea pigs exposed to a WD from weaning through gestation without hay (Thompson et al., 2014).

While maternal steatosis was not observed in the present study, marked glycogen accumulation was noted. Vacuolated hepatocytes, a feature of hepatic stress, were observed in DMF and VEH-treated groups. Glycogen-type vacuolar hepatopathy in pregnancy is not well described in the literature, and glycogen accumulation is not a normal finding in late-gestation, healthy guinea pigs (Dauprat et al., 1984); however in the absence of a chow-fed control group, these findings should be interpreted as relative to the experimental conditions rather than definitively abnormal. Nonetheless, DMF treatment did not appear to ameliorate these histological features in dams or fetuses. Indications of hepatopathy and displaced nuclei are associated with hepatic stress or metabolic dysregulation (Franco-Mahecha & Carrasco, 2021; Kim et al., 2022; Soon & Torbenson, 2023). Livers were positive for displaced nuclei, consistent with vacuolar hepatopathy and moderate to marked glycogen accumulation as previously described (Cortright et al., 2014; Khoury et al., 2018), though their functional significance in this context remains unclear. Still, excessive glycogen accumulation should be interpreted with caution; while unusual, no negative effects were observed, and without a negative control, it is unclear if this is a normal finding in late-gestation guinea pigs.

Similar to dams, hepatic TG content in fetuses indicated the absence of overt steatosis. Histologically, fetal livers displayed a mild to moderate diffuse microvesicular hepatopathy, distinct from the more prominent glycogen-associated vacuolization observed in dams. The significance of microvesicular hepatopathy in the fetus, especially in the absence of increased TG content remains descriptive as no functional or protein-level assessments were performed. PAS staining and amylase digestion confirmed only minimal periacinar glycogen accumulation. This limited glycogen storage may be physiologic, as glycogen in fetal guinea pigs rapidly accumulates between GD60–67 (Dauprat et al., 1984) whereas the animals in this study were terminated earlier (~GD57).

Interpretation of lysosomal glycogen is limited by the absence of a negative control and by the scarcity of literature (Zirin, Nieuwenhuis & Perrimon, 2013). The abundance of hematopoietic cells is a normal feature, reflecting the role of the fetal liver as a primary site of hematopoiesis (Ernst et al., 1993; Hashimoto et al., 2012). Most importantly, DMF treatment did not result in observable alterations to the fetal liver as all fetuses displayed similar phenotypes.

Maternal hepatic transcripts

The transcriptome of the maternal liver offers valuable insights into the systemic effects of DMF, particularly in the context of a WD. As the central hub of nutrient metabolism and antioxidant production, the liver is uniquely positioned to reflect metabolic adaptations during pregnancy. Interestingly, upregulation of P450-related pathways was directly contradictory to previous research, which demonstrates that DMF metabolism bypasses cytochrome P450 (Aubets et al., 2019). Pathway analysis indicated this signal as driven solely by Mgst3, a glutathione S-transferase also annotated to the P450 pathway. As such, these results do not support broad induction of P450 metabolism but instead may reflect changes in glutathione-associated pathways, consistent with an antioxidant response. While these findings are directionally consistent with activation of cytoprotective pathways, target engagement of the Nrf2 axis was not directly assessed and no functional or protein-level measurements were performed. Therefore it is imperative that these transcriptomic changes are interpreted descriptively and not confirmatory of Nrf2 activation.

Fetal hepatic transcripts

Interestingly, there was no overlap between fetal and maternal hepatic DEG in response to DMF treatment. While increased IL-12 signaling could appear paradoxical given DMF’s known immunosuppressive action (Lehmann et al., 2007), Ido1 is a key immunoregulatory enzyme involved in tryptophan catabolism and immune tolerance, particularly in pregnancy (Pallotta et al., 2022). Prior studies report DMF inhibits Ido1 activity in IFN-γ-stimulated human PBMCs (Eminel et al., 2017), and suppresses IL-12 (Al-Jaderi & Maghazachi, 2016), making our observed enrichment of IL-12 signaling in fetal liver contradictory. This discrepancy may reflect the unique immune landscape of the fetal liver as a hematopoietic organ. In this context, DMF-induced upregulation of Ido1, Irf8, and Laptm5 may reflect developmental regulation of hematopoiesis and immune tolerance (Gu et al., 2010; Xia et al., 2020; Newman et al., 2021; Zhang et al., 2024), or changes in cell populations.

Divergent transcriptomes highlight functional specialization of fetal and maternal liver

Transcriptomic comparison between maternal and fetal livers revealed striking differences with >9,000 DEG without any overlap between DEG by DMF treatment. This highlights that maternal and fetal livers are transcriptionally distinct and functionally divergent, as described previously (Kapitulnik et al., 1987; Ring et al., 1999; Li et al., 2012; Bonder et al., 2014).

Maternal liver was enriched with terms related to protein and lipid metabolism and peroxisomal activity, consistent with the role of the mature liver in nutrient processing and metabolism. In contrast, fetal liver was dominated by DNA replication, RNA splicing, cell cycle, and heme metabolism—hallmarks of rapid cellular proliferation and differentiation. Upregulation of ribosome pathways in fetuses relative to dams indicates high translational activity necessary to sustain hepatic proliferation, whereas the maternal liver prioritizes metabolic and catabolic functions.

These distinctions are critical for interpreting maternal exposure effects on fetal hepatic development. To our knowledge, our study is the first to compare maternal and fetal guinea pig liver transcriptomes via RNAseq. Because guinea pigs serve as a valuable model for translational metabolic research (Morrison et al., 2018), understanding these developmental differences is crucial. Differences in liver maturity may obscure the manifestation of hepatic responses observed in maternal liver, which may be revealed after the liver reaches maturity, consistent with the developmental origins of disease hypothesis (Brumbaugh & Friedman, 2014; Hoffman et al., 2021). Longitudinal studies are therefore needed to capture the transition to hepatic maturity and fully define fetal programming.

Placental morphology and gene expression is largely unaffected by dimethyl fumarate

Given reports that DMF is anti-angiogenic via VEGFR2 suppression (Ocaña et al., 2023), we assessed its impact on placental development. No changes in placental mass or diameter were observed. Expression of genes related to oxidative stress (Cat, Hmox1, Nqo1, and Sod1), transcriptional regulation (Nfe2l1, Nfe2l2, and Nfe2l3), cholesterol metabolism (Hmgcr), and angiogenesis (Hif1a and Vegfa) were largely unchanged.

Despite prior evidence that DMF activates Nrf1 (Hayashi et al., 2017; Piel et al., 2024) and Nrf2 (Manai & Amadio, 2022; Izumi & Koyama, 2024) we found no activation of target genes Nqo1, Sod1, and Hmox1 (He, Ru & Wen, 2020). Instead, Sod1 decreased, and transcription of Hmgcr and Nqo1 had numerical decreases, while the transcription of Hmox1 and Cat was unaffected, contrary to the expected effects (Singh et al., 2019; Dai et al., 2020; Tonev & Momchilova, 2023; Timpani et al., 2023). While DMF is a well-known activator of Nrf2, its effect on Nrf2 activation remains inconsistent and may depend on tissue type, species, dose, and duration of treatment (Hammer et al., 2018; Akino et al., 2019; Manai & Amadio, 2022; Timpani et al., 2023).

Because Nqo1 and Sod1 are canonical Nrf2 target genes (He, Ru & Wen, 2020), the lack of upregulation observed suggests DMF did not activate the Nrf2 pathway in the placenta under present experimental conditions. These observations may reflect tissue-specific Nrf2 responses (Brennan et al., 2016) and the unique environment of the placenta which favor regulated oxidative stress to promote a pro-angiogenic physiology and may explain the blunted antioxidant response observed here (Chang, Wakeland & Parast, 2018; Muralimanoharan, Kwak & Mendelson, 2018).

Nrf3 has been shown to upregulate genes in the SREBP2-dependent mevalonate pathway, including Hmgcr (Waku et al., 2021; Waku & Kobayashi, 2021). A numerical decrease in Hmgcr, a key regulator of metabolism and a known Nrf3-inducible gene (Waku & Kobayashi, 2021), was observed; however, this finding was not statistically significant and therefore only suggestive, not confirmatory, of altered Nrf3 activity.

Overall, these findings indicate that DMF does not activate Nrf2 in the placenta under these conditions, highlighting context-specific effects. Given the small sample size and lack of a negative control for WD, these results should be interpreted cautiously and warrant further investigation.

Dimethyl fumarate does not affect fetal development or other gestational parameters

Both VEH and DMF-treated dams exhibited similar weight gain trajectories over the study period, with progressive increases consistent with gestation (Czaja, 1983). Values for all gross fetal measures were consistent with expected ranges for late-gestation guinea pig fetuses (Sparks et al., 1985; Kaufmann & Benirschke, 2004; Candia et al., 2023).

Maternal treatment with 18.4 mg/kg/day of DMF did not have any effect on fetal developmental outcomes. This is consistent with a previous study involving rats, 25 mg/kg/day DMF (equivalent to the 18 mg/kg/day used in our study, when accounting for the species conversion factors) (Nair & Jacob, 2016) did not affect fetal weight (Gold et al., 2015). A human meta-analysis also demonstrated that DMF treatment in pregnant mothers who have multiple sclerosis did not negatively impact birth weight compared to the untreated population (Borriello & Ianniello, 2022; Hellwig et al., 2024). Together, these findings support that DMF treatment in pregnant guinea pigs at the human-equivalent dose does not alter maternal or fetal gestational parameters, which remain consistent with normal gestational progression.

Model of western diet consumption in pregnancy

A central aim of this study was to establish a WD model of pregnancy in guinea pigs that reflects human patterns of metabolic dysfunction without inducing overt maternal toxicity. The diet (45% fat, 15% sucrose, and 0.35% cholesterol, with ad libitum hay) was designed to model low-grade metabolic stress rather than severe obesity or insulin resistance.

Although the diet did not induce hepatic triglyceride accumulation exceeding the diagnostic thresholds for MASLD, consistent hepatocellular glycogen accumulation and vacuolar hepatopathy indicate that metabolic stress was achieved. These findings suggest that early hepatic adaptation in both dams and fetuses, providing a valuable window to study metabolic dysregulation prior to overt disease (Teasdale et al., 2024; Kshirsagar & Sharman, 2023).

From a translational perspective this model may better reflect human diet-induced metabolic risk in pregnancy (Fouda et al., 2022). Because the guinea pig exhibits several metabolic features that more closely resemble those of primates than other rodents (Castañeda-Gutiérrez et al., 2011; deOgburn, Murillo & Luz Fernandez, 2016; Morrison et al., 2018; Skat-Rørdam et al., 2021) and its placenta shares many physiological characteristics with human placenta (Mess, 2007; Thompson et al., 2016), gestational exposure to WD represents a relevant system to study maternal-fetal metabolic interactions and programming under WD conditions. Future refinement of the model, such as prolonged dietary exposure or formulaic changes, may further drive the phenotype toward overt steatosis and enable investigation of progressive metabolic dysfunction; though inclusion of a negative control group is absolutely necessary to fully validate the model.

Strengths and limitations

Robust fetal and maternal hepatic RNA-seq provided a systems-level view of DMF’s impact. The use of the guinea pig model, for its similarities in metabolism and placentation to humans, added clinical relevance compared to more commonly-used rodent models.

As with any pilot and feasibility study, this research was not without limitations. The most prominent limitation is the small sample size (n = 3 experimental units), which restricts statistical power and generalizability. The statistical modeling approach was intentionally simplified to avoid over-parameterization, which may limit the ability to fully account for potential covariates and interaction effects; therefore, results, particularly those approaching statistical significance, should be interpreted with caution. Methodological constraints also limited the interpretation including the absence of a true negative control group and a relatively short duration of WD feeding (~10 weeks), which may have been insufficient to induce a full metabolic phenotype. Future work could consider withholding hay or extending diet exposure to >16 weeks.

The WD used in this study contained 0.04 mg/kg of tert-butylhydroquinone (tBHQ), a preservative that is also known to activate Nrf2 (Zagorski et al., 2013). Although estimated exposure (~2.16 mg/kg BW/day) was substantially lower (15-fold lower) than therapeutic doses reported in prior studies (Wang et al., 2014; Ye et al., 2016), we cannot fully exclue the possibility of low-level Nrf2 activation. Future studies may consider alternative diets without this additive. Finally, placental RNA quality precluded RNA-seq analysis, restricting assessment to a small panel of candidate genes using RTqPCR and reducing the ability to capture broader transcriptomic effects. Collectively, these limitations emphasize that findings should be interpreted as preliminary and used to inform future, more robust, study design.

Conclusions

This pilot study sought to establish a guinea pig model for gestational WD consumption and to evaluate the feasibility of DMF treatment during pregnancy. We hypothesized that DMF would mitigate WD-associated oxidative stress and inflammation, thereby protecting the fetus; however, our findings did not support this hypothesis. Dimethyl fumarate conferred no measurable benefit on maternal or fetal hepatopathy and did not improve placental antioxidant gene expression within the parameters assessed but was likewise not associated with overt toxicity or harm. Treatment with DMF had a unique and paradoxical enrichment of IL-12 signaling in the fetal liver, which warrants further investigation. These findings suggest that DMF may act differently on maternal and fetal tissues. Results of fetal necropsies reinforce the safety profile of DMF therapy at human equivalent dosing in pregnancy, indicating no evidence of teratogenic or anti-angiogenic effects.

Overall, this study demonstrates the feasibility of applying WD and DMF therapy in pregnant guinea pigs and provides a foundation for larger, more conclusive studies. Future work should improve upon this study by including larger samples sizes, appropriate controls, expanded DMF dosages, extended dietary exposure, improved placental RNA preservation, and longitudinal metabolic assessments. To maximize translational relevance, future studies should model pre-existing maternal metabolic dysfunction prior to pregnancy and evaluate how antioxidant therapy intersects with placental function and offspring outcomes.

Funding Statement

This work was supported by the Charles E. and Clara Marie Eckelman Fellowship. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Additional Information and Declarations

Competing Interests

Kristin M. Patton is the owner of Four Star Quarter Horses.

Author Contributions

Daniella Hasan conceived and designed the experiments, performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the article, and approved the final draft.

Michelle Kutzler performed the experiments, authored or reviewed drafts of the article, and approved the final draft.

Kristin M. Patton performed the experiments, analyzed the data, authored or reviewed drafts of the article, and approved the final draft.

Massimo Bionaz conceived and designed the experiments, performed the experiments, analyzed the data, authored or reviewed drafts of the article, and approved the final draft.

Animal Ethics

The following information was supplied relating to ethical approvals (i.e., approving body and any reference numbers):

The Oregon State University IACUC approved the study (Protocol #2024-0452).

Microarray Data Deposition

The following information was supplied regarding the deposition of microarray data:

The data is available at GenBank: GSE312339.

Data Availability

The following information was supplied regarding data availability:

The data is available at Figshare:

- Hasan, Daniella; Bionaz, Massimo; Kutzler, Michelle; Patton, Kristin (2025). Supplemental Files. figshare. Dataset. https://doi.org/10.6084/m9.figshare.30571475.v3

The gene expression data is available at GenBank: GSE312339.