Postnatal iron supplementation fails to fully rescue brain metal and transcriptional defects caused by nutrition-based gestational iron deficiency in a mouse model

Abstract

Gestational iron deficiency (GID) is associated with long-term cognitive and behavioral impairments in offspring, but the effectiveness of postnatal iron supplementation in restoring brain metal homeostasis and developmental programming remains unclear.

Objective:

To determine how the timing of postnatal iron supplementation influences brain metal homeostasis and long-term gene expression following GID.

Methods:

A nutrition-based mouse model included three groups: nutritional iron-normal controls (NIN; 240 mg Fe/kg diet), GID offspring repleted with iron at birth (P0GID), and at postnatal day 7 (P7GID). Dams received an iron-deficient diet (2.2 mg Fe/kg) before and during gestation. Iron (Fe), copper (Cu), zinc (Zn), calcium (Ca), manganese (Mn), and magnesium (Mg) concentrations were measured in blood, cerebral cortex, and hippocampus at postnatal days (P)7, P14, and P40 using Inductively Coupled Plasma Mass Spectrometry. Cohorts contained 7–19 offspring animals from at least three litters. Metal concentrations were analyzed using linear models controlling for age, sex, litter size, treatment, and treatment-by-age interactions. Spatial transcriptomics, cell-type deconvolution, differential expression, and gene set enrichment analyses were performed in P40 male P7GID and NIN brains (n=3/group).

Results:

GID caused persistent, region-specific disruptions in brain metal homeostasis that were not fully corrected by postnatal iron supplementation. At P40, cortical Mg was significantly reduced, while hippocampal Fe, Cu, Mn, and Ca were significantly elevated relative to NIN controls (all p<0.05). Developmental trajectories of Fe, Zn, Mn, and Ca differed significantly between treatment groups. Blood metal levels poorly reflected brain metal status in GID, with significant correlations for Fe observed only at P7 and P14 in the P7GID cortex (p<0.03). Transcriptomic analyses revealed persistent dysregulation of iron-responsive, myelination-related, and neurodevelopmental pathways.

Conclusions:

GID induces lasting alterations in brain metal balance and developmental gene-expression programs that are not fully rescued by postnatal iron supplementation, highlighting gestation as a critical window for iron-dependent neurodevelopment.

Introduction

Normal Fe homeostasis is vital for the healthy development of the central nervous system (CNS)(1) and maternal Fe stores are essential for providing all embryonic and fetal Fe(2–8). It is acknowledged that the consequences of Iron deficiency (ID) in newborns are significant, as studies in humans and animal models have found a strong association with impaired cognitive function (9–26). A prospective study involving pregnant women and their offspring found that maternal ID in the first trimester was associated with a higher risk of the offspring developing autism spectrum disorders (ASD)(24). Another study showed an increased risk of schizophrenia in offspring born to Fe deficient mothers (27), while East et al (28, 29) recently found that GID and early postnatal ID is associated with Attention Deficit Syndrome, general impaired cognition, and poor executive control in adulthood.

Gestational and neonatal ID are not rare occurrences. Many epidemiological studies suggest that 80% of women have insufficient Fe reserves to adequately supply a growing fetus and 40% of women worldwide have depleted Fe stores at the start of pregnancy(30–36). While these numbers were often ascribed to populations who have substantial food insecurities and inadequate access to Fe rich food sources, a recent multicenter study across high resource counties (Ireland, UK, Sweden, Netherlands(37)) found that 4 of 5 women were iron deficient by the third trimester when using the ferritin of <30-μg/L threshold and still 1 in 2 women were deficient in the third trimester when using a ferritin <15 μg/L threshold, that is related to compromised fetal Fe accretion.

In light of the high incidence of GID and neonatal ID, Fe supplementation has been regarded as a logical intervention. However, its efficacy remains uncertain and the consequence of Fe supplementation on the distribution of other critical metals in the brain of offspring remains unknown. This knowledge gap is highlighted by a recent study that shows significant heterogeneity in the regional distribution of divalent metals in the adult mouse brain even at baseline (38), suggesting region-specific needs and handling of divalent metals. Changes in Fe, one of the most abundant divalent metals in the brain, are thus likely to affect the distribution of other metals that often use similar transport mechanisms. While the degree of metal changes in the brain under GID and subsequent supplementation are largely unknown, it is also not clear to what extent blood metal levels can provide insight into the efficacy of Fe supplementation for restoring region specific Fe depletions in the brain as well as associated brain levels changes of divalent metals.

Animal studies of postnatal Fe supplementation to iron deficient pups showed effective normalization of hematological parameters and restoration of whole brain Fe levels, but the status of other metals was not analyzed and learning deficits were still present (39–41). Our own studies demonstrated persistent functional and cellular impairments in the offspring born to GID, non-anemic dams despite postnatal Fe supplementation and normalization of whole brain Fe levels in the offspring (42, 43).

Epidemiological studies in humans also question the efficacy of Fe supplementation and raise concerns of a potential risk of postnatal Fe overload in the offspring. The limited evidence showing clear benefits have led the U.S. Preventive Services Task Force (USPSTF) to conclude that current data are insufficient to confirm health benefits of gestational Fe supplementation for mothers or infants (44). The recommendation for postnatal Fe supplementation to infants is also a matter of uncertainty and its benefit or risk seem to depend on the degree of deficiency that can range from marginal Fe depletion to severe Fe deficiency anemia (IDA)(45).

Given the strong evidence that gestational and neonatal ID negatively affect brain development and the limited understanding of how Fe supplementation influences brain metal distribution, we developed a mouse model combining GID with postnatal Fe supplementation. In this study, GID refers to a state in which dams are iron deficient but remain non-anemic, while embryonic brains become iron deficient by mid-gestation and offspring develop IDA, allowing us to isolate the effects of prenatal brain ID from maternal IDA. Postnatal Fe supplementation was achieved by switching dams to an Fe sufficient died at birth (P0GID) or 7 days postpartum (P7GID). Our goal was to track the levels of Fe and other metals in blood and across multiple brain regions in the offspring and across time to assess consequences to the neural development in the offspring at a time when blood iron markers suggest restoration of iron levels and an absence of ID (P40). Using Inductively Coupled Plasma Mass Spectrometry (ICP-MS) and spatial transcriptomics, we identified molecular targets sensitive to GID that may contribute to long-term neurological impairments that persist despite normalization of blood Fe levels.

Material and Methods

Nutritional Mouse Model, Cohorts and Study Design

All protocols were pre-approved by the University Committee on Animal Resources at the University of Rochester (Animal Welfare Assurance Number: D16–00188) and all animal exposures, handling, and sample microdissections were conducted by the same individual (JC). Ten-week-old virgin Swiss Webster mouse dams were obtained from Charles River Laboratories (Wilmington, MA, USA) and consistently housed under controlled temperature and humidity conditions on a 12 h light/dark cycle. Upon delivery, all dams were fed ad libitum standard 5010 rodent diet (LabDiet, St. Louis, MO, USA; 0001326) and allowed to acclimate for two weeks to vivarium housing. All animals were then switched from the vivarium diet (which contains 180–200 mg/kg Fe) to the custom-formulated Fe sufficient diet (approximately 200mg/kg Fe), to establish a common baseline and to allow acclimatization to the new diet. The number of animals used for this study was informed by our previous work on GID using the same mouse strain and dietary regimen. After two weeks on the custom Fe sufficient diet, 24 virgin 3–5 month old dams were randomly assigned to three experimental cohorts: An iron sufficient cohort (NIN) and two gestational iron deficient (GID) cohorts (2–4 mg/kg Fe) with different supplementation times (P0GID and P7GID). Both GID cohorts require postnatal supplementation to avoid the development of severe IDA and death of the offspring post weaning. In the NIN cohort dams and offspring received a Fe sufficient diet throughout the experiments while P0GID and P7GID cohorts received a Fe deficient diet (2–4mg/kg Fe) 2 weeks prior to mating and throughout gestation and were switched to an Fe sufficient diet at birth or 7days postpartum respectively. For mating all dams were paired with a male Swiss Webster mouse similar in size and age. Pregnant dams were checked for pups daily and once born, pups were denoted as postnatal day (P0) and litter sizes were recorded. To minimize maternal stress and capture the natural intra-litter heterogeneity in Fe responses, litters (average size of 8 pups) were not culled. Amongst the 3–8 independent litters used within this study per treatment group and timepoint, data were sampled using a minimum of 1 pup/sex/litter whenever possible. 3–5 pups from each individual litter were selected for ICP-MS analysis. All animals were humanely euthanized in accordance with the AVMA Guidelines for the Euthanasia of Animals, and body weights were measured once per week at pup ages corresponding to postnatal P7, P14, and P40. If a litter contributed to both P7 and P14 timepoints, pup removals at P7 never encompassed more than 44% of the original litter size and all pups were weaned by P21.

Diet composition:

Diets were purchased from Envigo (presently Inotiv, Inc. and formerly Harlan Teklad, Indianapolis, IN, USA) (Supplemental Table S1A) and differed only by Fe content, ensuring that offspring outcomes reflect reduced dietary Fe during gestation rather than other nutritional variables. Both diets contain 35g/kg Mineral mix (TD.81062) thus keeping trace element levels consistent in both diets. The Fe-sufficient (NIN) diet, TD.05656 contains, as per supplier, approximately 240 mg/kg Fe added in form of Ferric Citrate while the Fe-deficient (ID) (TD.80396) contains, as per supplier, 2.2 mg/kg Fe. Prior to feeding we confirmed the values of Iron (Fe), Copper (Cu), Zinc (Zn), Calcium (Ca), Manganese (Mn) and Magnesium (Mg) in the diet and found good agreement to the suppliers stated levels (Supplemental Table S1B). In total 125 animals were used for ICP-MS and 24 dams for generating offspring.

Whole Blood Collections and analysis.

We quantified metals in whole blood rather than serum to reflect total circulating metals, including both hemoglobin-bound and plasma iron pools, which is appropriate for assessing overall metal burden across development but differs from standard indices of bioavailable iron. Peripheral whole blood in PND7 pups was essentially trunk blood collected post decapitation. In exposed dams and offspring older than PND14, blood was collected utilizing a 25-gauge syringe (Fisher Scientific; 22–257-136) for intravenous terminal blood draws at the inferior venae cava. Blood samples designated for hematological analysis were collected with a heparin-coated (Sigma-Aldrich; H3149) syringes. A HESKA-HemaTrue^® Veterinary Hematology Analyzer (Heska Corporation; 5600) was utilized to determine hematocrit and hemoglobin values. Blood samples designated for metal measurements were collected with uncoated syringes and collected in pre-weighed 1.7mL polystyrene RNase, DNase, Pyrogen-free microcentrifuge tube (Laboratory Products Sales, Inc (LPS); L211511). Blood collected was immediately placed on ice upon harvest, tubes were weighed again before storing at −80°C.

Cortical and Hippocampal Brain Tissue Micro-Dissections.

The cerebral cortex and hippocampus of each pup were carefully micro-dissected from the whole brain in Leibovitz’s L-15 with L-glutamine (Fisher Scientific, Waltham, MA, USA). To ensure that metal measurements were tissue-specific and not contaminated with blood, pups older than P14 were perfused with heparin/1X PBS (Sigma-Aldrich, St. Louis, MO, USA) prior to micro-dissections. Additionally, cerebral cortices were micro-dissected to exclude nearby white matter (i.e., corpus callosum) and cortical meninges were removed. Each individual brain region was placed into pre-weighed 1.7 mL polystyrene RNase, DNase, Pyrogen-free microcentrifuge tubes on ice (Laboratory Products Sales, Inc. (LPS), Rochester, NY, USA), weighed, and then stored at −80 °C until metal analysis.

Inductively Coupled Plasma Mass Spectrometry (ICP-MS)

As previously described(42), and in collaboration with the University of Rochester Elemental Analysis Facility, an individual blinded to sample identity conducted multi-element analysis using a PerkinElmer NexION 2000 ICP Mass Spectrometer (PerkinElmer, Waltham, MA, USA) operating at sensitivities in the parts per billion (ppb) and parts per trillion (ppt) ranges. Briefly, samples were hydrolyzed in their collection tubes in ultrapure (for trace metal analysis) 67–70% nitric acid (VWR International, Radnor, PA, USA) in a heat block at 100 °C for 1 h and brought up to a final volume of 10 mL in UltraPure de-ionized water. After accounting for dilution factors and normalizing each sample’s detected levels by their respective wet weight collected prior to storage, metal levels herein were reported either as a net concentration (μg/dL for blood and μg/g for brain tissue) or relative mean concentration.

Spatial Transcriptomic Profiling

Spatial transcriptomic profiling was performed using NanoString GeoMx Digital Spatial Profiler (DSP) with the whole transcriptomic panel. We used 6 male animals in total (3 NIN, 3 GID P7), each animal contributing one tissue section, from which multiple Regions of Interest (ROIs) were selected, including superior and inferior colliculi, posterior medial thalamus, hippocampus, and posterior and anterior cortices. Each region is thus represented by 3 animals per group, with one ROI per animal. Briefly, the left hemispheres of P40 mouse brains perfused with heparinized PBS, were immediately embedded in OCT, and cryosectioned sagittally at 10 μm thickness onto Superfrost Plus slides under RNase-free conditions, with sections kept below 0 °C during collection and stored at −80 °C until processing. For GeoMx preparation, fresh-frozen sections were fixed overnight in 10% neutral buffered formalin, washed in PBS, baked at 60 °C, and sequentially dehydrated through graded ethanol before air-drying. Target retrieval was performed at 90 °C in Tris-EDTA buffer following a brief DEPC-H₂O pre-equilibration. RNA targets were exposed using Proteinase K at 37 °C, followed by postfixation in 10% NBF and PBS washes. Slides were then hybridized overnight at 37 °C with Buffer R–based RNA probe solution. The next day, sections underwent saline sodium citrate (SSC) and formamide-SSC washes before morphology marker staining with AF647-conjugated GFAP (Cell Signaling Technology, 3657S, Clone CA5), AF488-conjugated NeuN (Cell Signaling Technology, 54761S), and SYTO 83 nuclear stain in Buffer W, followed by overnight incubation at 4 °C. After final SSC washes, slides were imaged on GeoMx DSP. Following ROI selection, the UV-photocleaved oligonucleotide tags were eluted into individual wells and processed for next-generation sequencing. Each ROI eluate was subjected to library preparation with Illumina-compatible adapters and unique sample barcodes were added through PCR amplification. Amplified libraries were bead-purified, quantified, and pooled in equimolar amounts prior to sequencing. The final pooled library was then loaded onto an Illumina platform, where paired-end sequencing was performed to generate digital counts corresponding to each probe’s oligonucleotide tag. Resulting FASTQ files were processed using the GeoMx NGS analysis pipeline.

GeoMx Data Processing and Analysis.

Raw data were processed using the GeoMxTools R package (RRID:SCR_023424). Samples were quality controlled based on sequencing metrics and area. Outlier probes were removed prior to aggregation at the gene level. To define minimum expression threshold per segment, the limit of quantification (LOQ) was calculated using negative control probes and genes detected in less than 10% of the segments were excluded from the analysis. Expression data were normalized using upper quartile (Q3) normalization for downstream analysis.

Cell Type Deconvolution:

GeoMx generates bulk-like transcriptomic profiles per ROI, where expression is aggregated across all cells within a region. To recover cell-type level information, each ROI’s expression profile is computationally deconvolved using the SpatialDecon R package (RRID:SCR_026836) against a reference derived from a comprehensive single-cell RNA sequencing of mouse brains(46). Iron metabolism related genes were excluded from the markers to prevent confounding from treatment effects. Full ROIs were used for all selected regions except for the hippocampus where GFAP-guided segmentation was applied. However, as GFAP segments captured markedly smaller aeras than corresponding non-GFAP segments, deconvolution was performed in the non-GFAP segmented ROI to reflect the broader cellular composition of the hippocampus. Deconvolution was performed on Q3-normalized GeoMx data, resulting in one composition vector summing to one per ROI per animal. Estimated cell type proportions not significantly different from zero were set to zero, and proportions were renormalized to sum to one. Treatment associated differences in cellular composition were assessed using Dirichlet regression. A regional deregulation score was calculated for each brain region by first computing a composite score per cell type that combines effect size (absolute difference in mean proportions between treatment and control groups) with statistical confidence (absolute Z-value from Dirichlet regression), then averaging this composite score across all cell types for each brain region.

Differential Gene Expression Analysis:

Differential gene expression analysis was performed for each brain region comparing P7GID and control samples. Linear mixed-effects models were fit to the log₂-transformed Q3-normalized expression data using the mixedModelDE function (GeomxTools package) with treatment groups as a fixed effects and slide as a random effect.

Gene Set Enrichment Analysis (GSEA):

GSEA was performed using the fgsea R package (RRID:SCR_020938) to identify biological pathways dysregulated between P7GID and NIN across brain regions. Gene sets were obtained from the Molecular Signatures Database (MSigDB), including Gene Ontology Biological Processes, KEGG Medicus and Reactome pathways. Additionally, high priority autism susceptibility genes that had more than ten published reports from the SFARI database were included in the analysis as a custom gene set labeled as Neurodevelopment_ASD. For each brain region, genes were ranked by the product of their fold-change estimate and −log₁₀(p-value) from differential expression analysis, then used as input to GSEA.

To reduce redundancy across pathways, significantly enriched pathways (FDR < 0.05) from an initial analysis were grouped into 40 broader biological categories. Gene sets aggregated under these categories were then used for a second round of enrichment analysis. To visualize genes driving enrichment, leading edge genes were extracted from GSEA results and heatmaps were generated for each significant pathway-region combination. Separately, gene sets related to metal homeostasis (38, 47) were analyzed via GSEA as independent categories to assess treatment effects on myelination and metal related transcriptional programs.

Statistical Analysis

Measured continuous metal concentrations ≤ 0 were imputed by the smallest observed positive concentration value for that metal. The age of each mouse at the time of sacrifice was grouped into 3 categories: P7 (5.5–9.5 days), P14 (12.5–15.5 days), or P40 (39.5–47.5 days). Groups contained the following number of animals: P6 (NIN:18, P0GID:18, P6GID:12), P14 (NIN:11, P0GID:10, P6GID:17), P40 (NIN:13, P0GID:14, P6GID:12). Pups were derived from at least 3 different dams. The sample mean and standard deviation (SD) of the concentration of each metal (Cu, Zn, Ca, Fe, Mn, Mg) was tabulated for each combination of tissue type (blood, cerebral cortex, hippocampus), Fe treatment group (NIN, GID, P7GID), and age group (P7, P14, P40 in Supplementary Tables S2A–S2C, respectively). Log-link Gamma generalized linear models (GLM) were used to model the average concentration of each metal within each tissue type as a function of Fe treatment group, age group, litter size (integer, treated continuously), sex (male, female), and the interaction between Fe treatment group and age. Since the interaction was typically significant (4-df F-test p<0.05), the data was further stratified by age or Fe treatment group. Specifically, for each age group we fit a separate GLM to estimate the Fe treatment group effects within that age group, and for each Fe treatment group we fit a separate GLM to estimate the age group effects within that Fe treatment group, with all models adjusted for litter size and sex. GLM results are summarized in Supplementary Tables S3A–S3C and graphically presented by plotting the estimated relative mean concentration of each metal as a function of the 3 Fe treatment groups (relative to NIN in Figure 3), or by age (relative to P7 in Figure 4), for each of the 6 metals within each of the 9 combinations of tissue type within a given age or Fe treatment group. All 95% confidence intervals (CI) and p-values were likelihood-based, and a two-sided nominal 0.05 level of significance was used for all hypothesis tests, with no adjustment for multiple comparisons. Statistically significant differences from baseline in the relative mean concentration of each metal were indicated by open circles (p≤0.05) in the graphical summaries of the GLM results, while non-significant differences were denoted by closed circles (p>0.05); significant differences between the two non-baseline groups were indicated with an asterisk*. Significant interactions between diet and age imply that the effects of diet differ by age and the effects of age differ by diet. The correlation of the log-transformed concentration of each metal in the blood vs brain (cerebral cortex, hippocampus) was estimated and tested (Supplementary Table S4). All computations were performed in R. Distributions of maternal and offspring hematocrits by Fe treatment group and age were compared using ANOVA with post hoc Tukey correction for multiple comparisons using GraphPad Prism 10.

Figure 3. Impact of timed iron supplementation on tissue-specific metal levels.

Values present GLM-estimated relative mean metal concentrations from 10–12 animals/group normalized to NIN controls of each age groups. Thus, values reflect modeled deviations rather than raw concentrations. Panels display GLM-estimated relative means of blood (A, cerebral cortex (B) and hippocampus (C) metals measured in P0GID and P7GID pups are plotted relative to NIN within each age. Open symbols: statistically significance (p<0.05 (◯)), close symbols: and non-significant changes (p >0.05 (⚫)) relative to NIN. Significant differences between P0GID and P7GID are indicated in the metal panels list for each tissue with an asterisk*.

Figure 4. Trajectories of offspring brain metals across time.

Values represent GLM-estimated relative mean metal concentrations from 10–18 animals/group, normalized to P7 within each diet group. GLM-estimated relative means concentrations of blood (A), cerebral cortex (B) and hippocampus (C) metals at P14 and P40 are plotted relative to the mean levels in the P7 pups within each diet group. Statistically significant (≤ 0.05) and non-significant (>0.05) changes are indicated with open symbols (◯) and closed symbols (⚫) respectively.

Results

Supplementation at birth does not fully normalize brain iron and other divalent metal levels

To determine the regional impact of Fe supplementation on metal level in blood and brain of the offspring, we used our well-defined nutrition-based ID model where virgin Swiss Webster dams were randomly assigned to three experimental cohorts: An iron sufficient cohort (NIN) and two GID cohorts with different supplementation time points (P0GID and P7GID) (Figure 1). The NIN cohort refers to dams and offspring that received a Fe sufficient diet (240μg/g Fe) throughout the experiments. For the P0GID cohort dams received a Fe deficient diet (2μg/g) 2 weeks prior to mating (to deplete Fe stores) and throughout gestation until day of birth (P0) when dams (and pups) are switched to the Fe sufficient diet thus limiting the window of consumption of the ID diet to pregnancy. P7GID refers to the cohort where dams received a Fe deficient diet 2 weeks prior to mating, throughout gestation and until 7 days post partum (P7). At this time dams (and pups) are switched to the Fe sufficient diet thus generating an early postnatal ID in the pups by delaying supplementation. At weaning all pups remained on the Fe normal diet.

Figure 1. Schematic of dietary paradigms.

NIN (nutritional iron normal), P0GID (gestationally iron deficient, switched to iron normal diet at birth), P7GID (gestationally iron deficient, switched to iron normal diet 7 days after birth), P0-P40 (postnatal day 7 to day 40). Abbreviations: Cu (copper), Zn (zinc), Ca, (calcium), Fe (iron), Mn (manganese), Mg (magnesium). Created with BioRender.com

To define the normal baseline and ID associated maternal blood values, we analyzed hematocrits (Htc) and Hemoglobin (Hgb) (Figure 2A, C) at 7 days post-partum and at the time of weaning (P21) and found a significant decrease at day 7 postpartum in the P7GID dams (no supplementation yet) compared to NIN dams (p=0.015). However, Hematocrits (30±6.5%) and Hemoglobin values (10±1.5%) were consistent with our previous findings(48) and indicating normal to borderline postpartum range. As expected, Htc and Hgb values in the P0GID dams at postpartum day 7 were not significantly reduced (p=0.059, p=0.3 respectively) suggesting normal maternal blood parameters. At weaning (P21), maternal blood values were indistinguishable from NIN dams (p>0.5). Analysis of offspring Hct (Fig. 2B) and Hbg (Fig 2C) showed significantly decreased levels in P0GID and P7GID pups compared to NIN pups at P7 (p=0.004 and p<0.0001, respectively). Values from P7GID pups were also significantly decreased compared to P0GID pups (p<0.001) suggesting IDA. By P14, Htc and Hbg values in the P0GID group were still lower than NIN but did not reach significance (p=0.3). P7GID pups continued to show significantly reduced Htc (p=0.01) and Hgb levels (p=0.04). By P40, Htc and Hgb levels normalized across all groups (p>0.7).

Figure 2. Impact of supplementation timing on hematocrits.

Maternal (A, C) and offspring (B, D) Hematocrits and Hemoglobin levels, respectively, were analyzed at different times post-partum. All values graphed are represented as Mean ± SD. Minimum of 1 animal per sex across a minimum of three independent litters within each treatment group. Number of dams =4–8, number of offspring n=P7:16–19, P14:7–15, P40:10–17. Statistical significance (*p<0.05) was determined by ANOVA and Post hoc analysis with Tukey correction for multiple comparisons.

To determine the impact of GID on tissue metal levels in offspring we analyzed blood and brain metal levels at P7, P14 and P40 (notably, values measured at P7 in the P7GID cohorts, represent pups that were not yet exposed to a Fe sufficient diet, thus effectively being still iron deficient). The means of each age and diet group are generated from 10–18 pups derived from at least 3 independent litters. Concentrations of Fe, Cu, Zn, Ca, Mn and magnesium Mg were measured in all diet groups with mean ± standard deviation (SD) in offspring blood, cerebral cortical and hippocampal tissue at P7, P14 and P40 (Supplemental Tables S2A–C, respectively). We used whole blood metal measurements as serum iron is highly dynamic and under circadian and hepcidin-mediated regulation in mice, limiting its reliability as a steady-state biomarker of iron status (49, 50). In contrast, murine genetic models demonstrate that iron-restricted erythropoiesis and elevated erythrocyte zinc protoporphyrin can occur despite preserved circulating iron parameters, highlighting the value of whole blood–based functional measures (51, 52).

To facilitate comparisons, metal concentrations were first analyzed as a function of diet using NIN as reference adjusted for litter size and sex and stratified by age group and tissue. Results of tests of interaction between diet and age (most, but not all significant) are shown in Supplemental Table S3A (blood), S3B (cerebral cortex) and S3C (hippocampus) along with 95% confidence intervals and p-values for the relative means. The log-link gamma GLM accounts for higher variances at larger concentrations (heteroscedasticity), ensures that all predicted mean concentrations are non-negative, and estimates relative mean concentrations adjusted for litter size and sex. Figure 3A–C shows the GLM-based estimated mean relative to that in NIN pups.

As shown in Figure 3 (A) blood, (B) cerebral cortex and (C) hippocampus) and summarized in Table 1, we found significantly lower blood and brain Fe levels in P7 pups in P0GID and P7GID groups compared to the NIN group, confirming ID.

Table 1.

Summary of metal changes across diet groups and ages

Age [days]	Region	Diet	Fe	Cu	Zn	Mn	Ca	Mg
7	Blood	P0GID	↓	↑	→	↑	↑	→
7	Blood	P7GID	↓	↑	→	→	→	↓
7	Cortex	P0GID	↓	↑	→	↑	↑	↑
7	Cortex	P7GID	↓	→	→	↑	→	→
7	Hippocampus	P0GID	↓	→	→	↑	↑	↑
7	Hippocampus	P7GID	↓	→	→	↑	→	→
14	Blood	P0GID	→	→	→	→	↑	↑
14	Blood	P7GID	↓	→	→	→	↑	↑
14	Cortex	P0GID	↓	→	→	↑	→	→
14	Cortex	P7GID	↓	→	→	↑	→	↓
14	Hippocampus	P0GID	→	→	→	↑	→	→
14	Hippocampus	P7GID	→	→	→	↑	↓	↓
40	Blood	P0GID	→	→	↑	↑	→	→
40	Blood	P7GID	→	→	→	→	→	↓
40	Cortex	P0GID	→	→	→	→	→	↓
40	Cortex	P7GID	→	→	→	→	→	↓
40	Hippocampus	P0GID	→	↑	→	↑	↑	→
40	Hippocampus	P7GID	↑	↑	→	↑	↑	→

Arrows indicate direction of changes in GID cohorts relative to NIN values:

(↓) statistically significant decrease, (→) insufficient evidence of a difference, (↑) Statistically significant increase relative to Iron normal (NIN) controls. P0GID= gestational iron deficient followed supplementation at birth via dam, P7GID= gestational iron deficient followed supplementation at postnatal day 7 via dam.

We also found that blood Fe depletion was associated with an elevation in blood Cu and Ca level but only in the P0GID cohort. In the brain, elevation of Cu was only seen in the P0GID cortex, while Mn levels were increased irrespective of region of supplementation timing. The P0GID cohorts showed additional elevation of Ca and Mg that were not present in the P7GID cohorts at P7, suggesting a transient over- compensation of specific metals when Fe is supplied early.

By P14, we saw the first indication of repletion with Blood Fe levels in the P0GID cohort being no longer significantly different relative to the NIN cohorts. The P7GID cohort still showed significantly decreased Fe levels, consistent with Htc and Hbg values. Blood Cu, Zn, and Mn levels did not differ from NIN in either supplementation cohort, while Ca and Mg levels were elevated regardless of supplementation timing. Brain metal levels did not reflect changes in blood. Most notably, Fe levels in the cerebral cortex remained significantly decreased in both supplementation cohorts. This decrease in Fe in the cortex was associated with increased Mn levels. In contrast, hippocampal Fe level were no longer different relative to NIN levels irrespective of supplementation timing but Mn levels remained significantly elevated. In contrast to brain, blood Mn levels showed no difference relative to NIN levels at P14 while brain did not show the elevations of Ca we saw in blood.

By P40, blood metal levels were largely comparable to NIN levels except for a sustained elevation of Zn and Mn in the P0GID cohorts and a reduction in Mg in the P7GID group. In the brain, Fe, Cu, Zn, Ca and Mn levels in the cortex were comparable relative to NIN levels in both the P0GID and P7GID cohort and only Mg levels remained decreased across both diet groups. In contrast, the hippocampus showed sustained elevation of Cu, Mn and Ca irrespective of supplementation timing. Interestingly, the P7GID cohort showed an elevation of Fe that is not present in the P0GID group.

Overall blood metal levels did not consistently reflect brain metal levels across time points and supplementation cohorts. To further understand the impact of supplementation, we compared metal levels between the P0GID and P7GID groups (marked with * asterisk in the metal legend Fig 3A–C) and found timing specific differences in most of the metals at P7. Specifically in blood, all metals, except Cu, were significantly different between P0GID and P7GID pups. In the cerebral cortex Cu, Ca, Mn and Mg levels varied depending on timing of supplementation. In contrast, in the hippocampus Cu levels did not differ between diet groups. As pups aged, the effects of supplementation timing became less pronounced and followed different patterns in blood and brain.

By P14, blood Ca and Mn were still significantly different between diet groups while in the brain only Mg and Ca levels responded to timing of supplementation differently.

By P40, blood and brain metal levels in the P7GID cohorts were largely similar relative to those in the P0GID cohorts, with the only difference being P7GID specific decreased Mg levels in the cerebral cortex.

Taken together, Fe levels remained decreased until P14 in the cerebral cortex in both supplementation groups but normalized to NIN levels in the hippocampal tissue suggesting a more effective and/or faster supplementation. Manganese levels remained generally elevated but were not always correlated with Fe levels. Ca, Cu and Mg levels were most affected in the P0GID relative to NIN cohorts in young animals, while diet specific decreases in older tissues were less apparent. Zn levels were, for the most part, not altered by age, region or diet. Notably, by P40, brain metal levels were still not comparable to NIN levels especially in the hippocampus despite full normalization of blood Fe levels.

GID significantly impacts the trajectory of metals in the brain of offspring that is not normalized by postnatal supplementation

Our analysis thus far uncovered region and diet specific changes in metal levels across different supplementation groups relative to levels found in NIN brains. However, these analyses did not directly allow a comparison across time points within the specific supplementation cohort. Thus, we next modeled the mean concentration of each metal as a function of age, using P7 as reference, adjusted for litter size and sex, stratified by diet and tissue (Supplemental Tables S3A–C). In blood, the normal developmental trajectory (NIN) showed decreased Cu, Mg, and Ca by P14 relative to P7, with Fe, Zn, and Mn remaining unchanged (Fig. 4A).

This pattern was largely preserved in the P0GID cohort, except that Mn levels declined. In contrast, the P7GID cohort showed significant increases in Fe, Ca, and Mn at P14 relative to P7 confirming an impact of ID. Brain metal trajectories differed markedly from those in blood, even under NIN conditions, highlighting the limited utility of blood measurements as a proxy for brain metal homeostasis. This was especially apparent in P40 animals were significant changes remain in the trajectory of metals in the brain especially under conditions of delayed supplementation (Table 2). In the cerebral cortex of NIN animals (Figure 4B and Tabel 2 gray highlighted), Fe, Cu, Zn, and Mn levels increased by P40 relative to P7, while Ca and Mg declined. In the hippocampus (Figure 4C and Tabel 2 blue highlighted), Fe levels declined over time in contrast to cortex. Despite the regional opposing differences in Fe levels, both brain regions showed increased Mn and decreased Ca and Mg by P40 relative to P7. These normal brain metal trajectories were disrupted by GID, regardless of supplementation timing. In the P0GID cohort, Mn failed to increase with age in the cerebral cortex and hippocampus remaining unchanged at P40 relative to P7. In the hippocampus specifically, P0GID animals showed elevated Cu levels at P40 but no decrease in Fe levels and no increase in Mn levels that marked the normal trajectory in NIN tissue. The P7GID cohorts exhibited yet another distinct trajectory. In the P40 cerebral cortex, Cu, Zn, and Mn increased similarly to NIN, but Fe failed to increase (despite being supplied in the diet), and Ca failed to decrease by P40. In the hippocampus, Fe, Zn and Mn levels were aberrantly regulated at P40 relative to P7, a pattern not observed in either NIN or P0GID animals. Mg was the only metal that showed a normal trajectory in blood and brain in both P0GID and P7GID at P40. All other metals displayed region- and diet-specific deviation from the trajectories of metal levels seen in NIN brains from P7 to P40.

Table 2.

Summary of trajectory of statistically significant metal changes

Tissue	Diet	Trajectory of metal changes at P40 relative to P7
Metals		Fe	Cu	Zn	Mn	Ca	Mg
Blood	NIN	↑	↓	→	↓	↓	↓
Blood	P0GID	↑	↓	→	↓	↓	↓
Blood	P7GID	↑	↓	→	→	↓	↓
Cerebral	NIN	↑	↑	↑	↑	↓	↓
Cerebral	P0GID	↑	↑	→	→	↓	↓
Cerebral	P7GID	→	↑	↑	↑	→	↓
Hippocampus	NIN	↓	→	→	↑	↓	↓
Hippocampus	P0GID	→	↑	→	→	↓	↓
Hippocampus	P7GID	↑	→	↑	→	→	↓

Arrows indicate direction of significant change in P40 relative to P7 in the prospective tissue and dietary group: decrease (↓), no change (→), or increase (↑). NIN=Iron normal condition, P0GID= gestational iron deficient followed supplementation at birth via dam, P7GID= gestational iron deficient followed supplementation at postnatal day 7 via dam, P7, P40=postnatal day 7 and 40

Blood metal levels are a poor surrogate to predict brain metal status

The altered trajectories of metals in blood and brain suggest that blood metal levels may not accurately predict brain metal levels, especially in light of the observed heterogeneity between subregions. To quantify the possible predictive value of blood metal concentrations for those in the brain, we conducted a correlation analysis focusing on Fe as well as Mn which is known to interact with Fe (53) and has been shown to accumulation under ID (54–57). Data was stratified by brain tissue (cerebral cortex, hippocampus), age, and diet (Supplemental Table S4). We found that at P7 and P14 in NIN animals, there is insufficient evidence of a significant blood-brain Fe correlation. Blood Mn only significantly correlates with hippocampal Mn (p=0.026) at P7. In contrast, by P40 in NIN brains we found a significant blood-brain correlation in Fe levels in both cortical (p=0.003) and hippocampal (p=0.001) regions. However, the correlation of blood-brain Mn was regions specific with blood Mn only correlating with hippocampal Mn (p=0.025). In P0GID brains there was insufficient evidence of a blood-brain correlation of Fe or Mn at P7. At P14, blood Fe significantly correlates with cortical Fe (p=0.031) but not hippocampal Fe, and there is insufficient evidence of any correlation between blood Mn and brain Mn. In contrast, at P40 there was a significant Fe blood-brain correlation (p<0.006) and a significant correlation between blood Mn and hippocampal Mn (p=0.034). This correlation was not apparent in the cerebral cortex. Interesting, correlations between blood and brain Fe and Mn levels were largely lost in the P7GID cohorts, with only blood Fe levels correlating positively with cortical Fe levels at P7 (p=0.023) and negatively at P14 (p=0.032).

Taken together, blood–brain correlations, were transient, region-specific, and highly dependent on dietary history, indicating that blood metal measurements (at least in respect to Fe or Mn) cannot reliably infer brain metal status even under non-iron normal conditions.

Brains exposed to GID and early postnatal ID show region specific changes in the transcriptional landscape

To determine the potential impact of GID and its metal changes on cells and transcriptional programs, we conducted a spatial transcriptomic analysis focusing on P7GID offspring as this group includes an early ID and exhibited the most persistent and widespread disruptions in brain metal homeostasis at P40, providing a stringent test of long-term transcriptional consequences. Unlike our bulk analysis, where we only performed a detailed metal analysis in microdissected cortex and hippocampus tissues (Figure 3–4), we included in our spatial transcriptomics hindbrain regions (inferior and superior colliculus) as well as regions of the thalamus. The ventral hindbrain has been described in previous work as being highly vulnerable to ID with some mouse strains losing up to 40% of Fe (58). The superior (SC) and inferior colliculus (IC) specifically represent signaling centers that convert visual, auditory, and somatosensory inputs into motor functions. We have found previously that GID impairs axonal maturation in the auditory nerve, causing altered conduction velocity via changes in axonal diameter and neurofilament regulation (59). This observation might suggest that auditory midbrain centers such as the IC could be vulnerable to ID. We also included the thalamus as a region of interest (ROI) based on a study that found increased Fe levels in the thalamus being associated with depressive symptoms suggesting vulnerability of this region to Fe disruptions (60). In addition, Quantitative Susceptibility Mapping (QSM) of adult human brains identified the thalamus as a region of elevated magnetic susceptibility, indicative of higher paramagnetic Fe content (ferritin-bound Fe)(47). Shown in Figure 5A is a representative map of selected ROIs in P7GID and NIN brains (data from one tissue section from three different animals/cohort contributed to the analysis). Differentially expressed genes across brain regions in P40 P7GID compared to NIN mice revealed changes of different magnitudes across all the ROIs analyzed (Supplemental Figure S1)

Figure 5. P7GID brains show region specific alterations in cellular composition and neural gene expression programs at P40.

(A). Shown is a representative sagittal brain section (P40 male) used for GeoMx analysis with regions of interest (ROIs) indicted.. Scale bar is equivalent to 1mm. IC = Inferior colliculus, SC = superior colliculus, HC = Hippocampus, PCx = Posterior cortex, TH = Thalamus, ACx = Anterior Cortex. (B) Analysis was conducted from 3 NIN, 3 P7GID animals, with each animal contributing one tissue section. Bar plot shows regional deregulation scores quantifying overall cellular composition changes between P7GID and Normal (NIN) groups. (C) Plot shows number of differentially expressed genes (FDR < 0.05, linear mixed-effects model) per brain region comparing P7GID to Normal samples. (D) Bubble chart shows mean proportions for cell types estimated from GeoMx spatial transcriptomic data across brain regions for Normal (NIN) and P7GID mice at P40. Bubble size represents mean proportion. (E) Dot plot of GSEA results show pathways significantly deregulated (FDR < 0.05) in more than half of the brain regions examined. Dot size indicates statistical significance (-log10 adjusted p-value) and color indicates normalized enrichment score (NES) with red representing enrichment and blue representing depletion.

Transcriptomic data was utilized for deconvolution of cell types within the ROIs to characterize changes in cellular composition associated with delayed Fe supplementation and pathway enrichment analysis to identify biological processes dysregulated in response to regional alterations in metal homeostasis. Differential gene expression analysis showed that the superior colliculus displayed the largest deregulation score and differential gene expression changes (Figure 5B, C), followed by posterior cortex and hippocampus. GSEA analysis revealed pathways that were significantly deregulated (FDR < 0.05) in more than half of the brain regions examined (Figure 5E). Cell type deconvolution revealed changes in cellular landscape in multiple ROIs and changes in the mean proportions for cell types estimated from GeoMx spatial transcriptomic data across brain regions for Normal (NIN) and P7GID mice at P40 (Figure. 5D). Interestingly, these pathways showed region-specific directionality. Particularly striking is the regional dysregulation of genes associated with neurodevelopment and Autism Spectrum Disorder (ASD- Figure 5E, purple bar) that are downregulated in the superior and inferior colliculi, cortex anterior, hippocampus and thalamus, yet upregulated in the posterior cortex. Consistent with the deregulation of these specific genes, genes associated with synaptic transmission and organization as well as learning/memory showed a similar pattern of dysregulation.

In contrast, genes associated with myelination and mitochondrial function (Figure 5E, black bar) showed the opposite pattern, with a specific downregulation in posterior cortex but upregulation in inferior and superior colliculi as well as the thalamus associated regions. This opposing pattern of gene dysregulation is also extended to genes associated with developmental patterning and dendritic development suggesting compensatory and/or maladaptive region-specific response to GID and postnatal ID. P7GID brains show region specific disruptions in myelination gene expression and prolonged changes in specific metal transport pathways

While neuronal pathways showed a high degree of deregulation, we also found upregulation of transcripts associated with oligodendrocytes (OLs) in the SC, (Figure 6A) and the hippocampus (Figure S2) as well as down- regulation of OLs in the posterior cortex in P7GID compared to NIN (Figure 6A). Since the cerebral cortex and hippocampus showed different levels of Fe, and gliogenesis is known to be highly regulated by Fe levels (47, 61, 62), we analyzed myelin related transcripts that are affected by metal changes. A recent study used combined Quantitative susceptibility mapping (QSM), which represents a MRI technique to measure Fe content in the brain, with spatial expression, and identified myelination genes that correlate with QSM (47). Analysis of these QSM correlated myelin genes in our samples showed significant dysregulation of myelin related transcripts across all brain regions, with depletion in the posterior cortex and enrichment in all the other regions tested (Figure 6A). Notably, estimated oligodendrocyte abundances in these regions correlate with enrichment of myelin programs suggesting that delayed supplementation in GID has lasting consequences in myelination programs and altered oligodendrocyte development.

Figure 6. P7GID brains show altered expression of genes involved in myelination and metal homeostasis.

(A) GSEA dot plot shows myelination and metal homeostasis gene sets that are significantly deregulated (FDR < 0.05) in at least one of the brain regions examined. Dot size indicates statistical significance (−log10 p-value) and the color indicates normalized enrichment score (NES) with red representing enrichment and blue representing depletion. (B) Heatmap displays z-scaled expression of Iron associated in superior colliculus and (C) posterior cortex and (D) hippocampus iron normal animals (Normal) versus GID animals that received supplementation at P7, GIDP7). Bars highlight Transferrin receptor (Tfrc) and Ferritin heavy chain 1 (Fth1).

As changes in oligodendrocytes and myelin related transcripts seem to correlate with Fe content (38, 47), we tested whether the disruption in myelin related genes is correlated with expression of genes involved in metal homeostasis programs in our ROIs. As shown in Figure 6A, genes involved in Fe homeostasis in general showed still a slight enrichment in the inferior colliculus (FDR = 0.046) and while we saw no overall significant changes in the other regions, specific Fe associated transcripts showed persistent changes across our samples in P7GID compared to NIN, exemplified by changes in Transferrin receptor (Tfrc) and Ferritin heavy chain 1 (Fth1) (Figure 6B–D, highlighted bar).

Given the co-regulation of Fe and Mn, we also analyzed Mn transporters. We found no consistent differential regulation in Slc11a2, (also referred to as DMT1), which mediates the shared transport of Mn and Fe allowing both metals to compete for uptake through the same cellular pathway(63, 64). Slc39a8 (high-affinity Mn importer) failed to clear the gene-level detection threshold due to the negative control probes while Slc30a10 (efflux transporter) cleared detection but was not significantly differentially expressed between the groups in any of the regions tested (data not shown). Cu homeostasis associated transcripts also showed no significant enrichment in any region or condition tested (data not shown) while Zn associated transcripts were depleted in the anterior cortex and inferior colliculus only (FDR < 0.05) (Figure 6A).

GID is associated with deregulation of genes associated with neurodevelopmental and Autism Spectrum Disorders (ASD)

The significant opposing changes in genes associated with neurodevelopment and ASD (Fig 5E, purple bar) in SC and posterior cortex prompted us to a more detailed analysis. Specifically, in the SC estimated Di- and mesencephalon inhibitory and excitatory neurons were significantly reduced (Dirichlet regression, p < 0.001) while estimated oligodendrocytes and hindbrain neurons were significantly increased (Dirichlet regression, p < 0.05 and p < 0.001 respectively) in P7GID brains compared to NIN brain (Figure 7A). In contrast, in the posterior cortex telencephalic projection neurons significantly increased (p<0.01) in P7GID brain compared to controls while oligodendrocytes, hindbrain neurons and astrocytes were decreased (Figure 7B). The SC also showed the greatest depletion of neurodevelopmental and ASD-associated genes, with 153 leading-edge genes (i.e., the core genes driving the GSEA enrichment signal), out of 385 pathway genes displaying a predominantly downregulated pattern in P7GID compared to controls (Figure S3A). In contrast, the posterior cortex showed the greatest enrichment, with 121 of these genes displaying an upregulated pattern in P7GID compared to controls (Figure S3B).

Figure 7. Superior colliculus and posterior cortex show opposing directionality of neurodevelopmental gene changes in P40 P7GID brain.

Stacked bar plot shows estimated cell type proportions in (A) superior colliculus (B) and posterior cortex across individual samples. Normal refers to NIN, GID P7 refers to P7GID cohorts (each bar represents one ROI from one animal, n=3 animals per group)

Discussion:

In this study we focused on the impact of gestational iron deficiency (GID; P0GID) and combined gestational plus early postnatal iron deficiency (P7GID) on offspring brain development. We chose this model based on the observation of significant consequences resulting from fetal and early neonatal Fe underloading (28, 45, 65–67). Limiting exposure to the gestational period and a brief postnatal window also reflects clinical standard-of-care, where iron supplementation is provided to iron-deficient pregnant individuals and/or newborns to prevent severe IDA(68, 69). We intentionally avoided a cross-fostering approach that, while having the advantage of more clearly defining the onset of supplementation, this practice is associated with stress in the offspring leading to increased body weight, altered blood sugar metabolism (70) and altered serotonin neuronal functions in the prefrontal cortex and hippocampus (71).

Postnatal Fe supplementation fails to normalize metal distribution

Our model addresses the knowledge gap of the consequences of Fe supplementation to brain metal homeostasis and neural development in the offspring that experienced ID during gestation and early postnatal life. Fe supplementation to iron deficient human neonates has been effective in restoring blood Fe levels and preventing progression to Fe deficiency anemia (IDA), but efficacy of Fe supplementation on neurological functions is still debated (5, 40, 72–89). Supplementation at birth to the dams in our model did not have an impact on offspring Fe level in blood and cerebral cortex until at least P14 while the hippocampus showed Fe level comparable to control brains at P14 and even increased levels at P40, suggesting compensatory responses driven by Fe uptake under ID.

In examining other metals, we observed unexpected patterns for manganese (Mn) at P7 and P40. Iron status is a major regulator of Mn absorption, distribution, and circulating levels(90, 91) and both human and animal studies consistently report elevated blood Mn levels in ID and IDA across age groups (53–55, 92–98). However, we did not detect elevated blood Mn levels in dams (Supplementary figure S4) or P7GID pups despite presence of ID and decreased Fe levels. In contrast, Mn levels were significantly elevated in the P0GID cohort despite 7 days of iron supplementation. This was an unexpected finding as iron repletion typically suppresses Mn uptake via shared transport pathways (99–104). As anticipated, Mn levels declined at P14 when Fe levels normalized but rose again at P40 despite normal Fe level. This pattern may reflect differences in the deficiency duration, nutritional sources (milk versus chow) and timing of supplementation. While both cohorts experienced GID, P7GID animals remained deficient postnatally, potentially enabling tighter homeostatic regulation of Mn. In contrast, P0GID animals were supplemented from birth during a developmental period when metal transport systems are highly active but not yet fully regulated. As a result, iron repletion may not have immediately suppressed shared transport pathways, allowing continued Mn uptake. Early repletion may also have caused a transient overshoot in divalent metal absorption at P7. Additionally, the transition from milk to chow (P14 to P40) introduces dietary variability that may further influence Mn levels.

In the brain, the Mn-Fe relationship was more predictable: at P7, decreased Fe was accompanied by increased Mn, and by P40, normalization of Fe coincided with normalization of Mn. Unexpectedly, however, iron supplementation was relatively ineffective at restoring brain Fe. We do not know whether regional Mn redistribution might have been driven by transporter expression as Slc39a8 (a high-affinity Mn importer) fell below detection thresholds, and Slc30a10 (an efflux transporter) was not differentially expressed across regions. These findings suggest that Mn regulation in the brain may occur primarily at the level of transporter function rather than transcript abundance. Post-translational controlled mechanisms like trafficking, localization, and protein stability (105), can modulate Mn flux without corresponding changes in mRNA levels (106). Furthermore, redundancy among Mn transport systems, particularly in the brain, may enable compensatory shifts in Mn distribution even when individual transporter expression remains unchanged (107). Therefore, transcript-level analysis alone may underestimate changes in Mn handling under GID and delayed iron supplementation.

The differential regulation of Fe and Mn in blood versus brain was underscored by their poor correlation across compartments, highlighting the limited ability of blood measures to predict brain metal status—particularly in the context of GID and early postnatal iron deficiency. Blood might therefore not represent a reliable biomarker for brain metal status. Ongoing efforts to identify more predictive biomarkers emphasize integrating measures of iron stores, transport, cellular demand, and regulatory signaling. (108–110).

Mn was not the only metal that showed persistent and biphasic changes in the brain alongside discordant blood patterns. At P40, both P0GID and P7GID hippocampi showed elevated Ca and Cu relative to NIN, while blood Zn was elevated only in P0GID. Prior studies indicate that iron deficiency alters calcium and phosphate transport (111) and disrupts developmental expression of calcium-dependent neurotrophic factors and signaling proteins critical for synaptogenesis (112, 113). Whether elevated hippocampal Ca in P7GID has functional consequences for synapse formation warrants further investigation. Our analysis of the trajectory of metal levels across time revealed a high ability of adaptation to disrupted blood metal level but a less effective ability to maintain a normal trajectory in the brain, suggesting that early disruptions in metal homeostasis produce lasting alterations in brain “metal landscapes,” reflecting compensatory and/or maladaptive responses.

Iron associated genes in P40 P7GID offspring show changes that signal region-specific compensatory and maladaptive responses

To probe functional consequences of the altered metal landscape, we performed spatial transcriptomic analysis in P40 offspring and identified persistent, metal transport–related gene expression changes.

In the hippocampus, reduced transferrin receptor (Tfrc) and a trend toward increased ferritin heavy chain (Fth1) in P7GID animals aligned with ICP-MS data showing elevated Fe at P40, suggesting a sustained adaptive response—enhancing Fe uptake while increasing storage to limit ferroptotic risk. This resilience is consistent with prior studies demonstrating recovery of hippocampal structure and function following delayed iron supplementation (114). In contrast, the superior and inferior colliculus and the anterior cortex, showed increased expression of Tfrc and Fth1 in P7GID offspring, consistent with enhanced Fe sequestration to buffer excess intracellular iron. Notably, Fth1 can reinforce Hypoxia-inducible factor 1-(HIF1) signaling thus sustaining hypoxic gene expression (115). While transient activation of HIF may be protective, chronic signaling is linked to mitochondrial dysfunction (116), synaptic impairment, and deficits in plasticity and memory formation (117). Consistent with this, we observed enrichment of pathways related to mitochondrial function, respiration, synaptic organization, and learning and memory.

The posterior cortex displayed a distinct pattern of regulation, with increased Tfrc and decreased Fth1 expression in P7GID regions relative to NIN. This signature is characteristic of ID, reflecting depletion of stored Fe and compensatory upregulation of uptake pathways Tfrc (118). Although bulk ICP-MS of the cortex did not indicate Fe deficiency at P40, the spatial transcriptomic data suggests localized cellular iron deficiency within subregions. Future studies will integrate spatial transcriptomics with laser ablation ICP-MS to directly co-register metal levels and gene expression.

A novel finding was the reduced expression of Zn associated genes in the inferior colliculus and the anterior cortex (Fig 6A). This decrease might be of functional relevance as Zn is loaded into presynaptic glutamatergic vesicles and released into the synaptic cleft along with glutamate during synaptic transmission(119, 120). Reduced Zn-related gene expression (especially Slc30a3, which has been associated with neuropsychiatric phenotypes (121)) could impair synaptic modulation leading to reduced sensory precision, and impaired cortical gain control. Synaptic Zn has also been shown to potentiate AMPA receptor function and shape frequency tuning in the auditory cortex(122). Whether similar mechanisms operate in midbrain auditory structures such as the inferior colliculus remain likely but unresolved. Interestingly, our previous work demonstrated that GID was associated with a disruption of auditory nerve development including axon structure and neurofilament phosphorylation of (59). However, the role of Zn associated transcripts in these processes has not been established.

The impact of GID and postnatal supplementation is associated with a region-specific increase rather than decrease of oligodendroglia cell types

Cell-type deconvolution of regional transcriptomic data revealed increased myelinating oligodendrocytes (OL) in the hippocampus and superior colliculus (SC) of P7GID animals relative to controls (Fig 6A). This was unexpected, as most studies report reduced OL abundance in iron deficiency (61, 123–125). However, prior work largely focused on white matter and different dietary models. We propose that early iron supplementation may preserve or expand OL populations in the hippocampus, especially given the high baseline OL density in CA1 during early development and its vulnerability to iron fluctuations (126). This interpretation is supported by evidence that gray matter iron depletion precedes white matter loss (127), which could lead to preferential supplementation of this region. Our ICP-MS analysis also showed increased Fe levels in the hippocampus in the P7GID cohorts, which coincides with an increased proportion of OLs (Figure S2).

Increased OL signatures in the SC represent a novel finding with potential functional implications, given the precise myelination required for topographic mapping underlying sensory processing; disruption could impair adaptive sensory integration and motor responses (128). In contrast to the hippocampus and SC, the posterior cortex exhibited reduced OL proportions, potentially reflecting increased OL death, reduced progenitor pools, or impaired maturation. These changes align with prior reports linking pre- and early postnatal iron deficiency to disrupted glial proliferation and oligodendrocyte development (61, 123, 129–131), as well as altered astrocyte (123, 132) and microglial function (125).

The superior colliculus and posterior cortex are highly sensitive to GID and exhibit neurodevelopmental disorder (NDD)-associated phenotypes later in life

A surprising and unexpected finding was the persistent remodeling of cellular and transcriptional programs linked to neurodevelopmental disorders (NDDs) in the superior colliculus (SC) and posterior cortex of P7GID animals, nearly five weeks after iron supplementation (P40). Notably, these regions showed significant enrichment of Simons Foundation Autism Research Initiative (SFARI) curated Neurodevelopmental Disorders (NDD)-associated genes, including those linked to autism spectrum disorder (Fig. 5E, Fig. S3).

The SC plays a major role in processing sensory information through GABAergic inhibitory interneurons that form microcircuits for refining sensory maps and directing attention(133). Disruptions in this region have been correlated with ASD (134) and Attention-Deficit/Hyperactivity Disorder (ADHD) (135–137) supported by studies in congenital blind children who show a more than 30 times higher prevalence of presenting with ASD symptoms (138, 139). We now provide evidence that this region is highly sensitive to GID and significant transcriptional and cellular changes remain even after Fe supplementation. Our model of GID also provides an insight into the window of vulnerability. While somatosensory and visual circuits form embryonically, their segregation into unimodal pathways occurs postnatally(140). Because GID induces ID as early as E14(43), it may disrupt this segregation, leading to imbalanced sensory circuit identity. Consistent with this, deconvolution analysis revealed altered proportions of excitatory and inhibitory neurons in the P7GID SC (Fig. 7A), potentially impairing sensory channeling.

The posterior cortex, which provides visual association input to the SC and supports visuomotor integration, also showed marked transcriptomic dysregulation with enrichment of NDD-associated genes (Fig. S3). These changes suggest an increased proportion of telencephalon-projecting excitatory neurons, which regulate laminar organization and interneuron recruitment. Overrepresentation of these populations could shift inhibitory tone and disrupt cortical tuning and computation(141). Together, these findings align with epidemiological evidence linking GID to increased NDD risk (7, 66, 142) and, for the first time, identify the SC and posterior cortex as particularly vulnerable to gestational and early postnatal iron deficiency.

Limitations

This study has several limitations.

(i) Our sufficient iron diet (NIN 240mg/kg Fe) is based on the generally used rodent vivarium diets which can be considered “iron-high” relative to physiological needs and is optimized for colony management. While a purified diet with more physiological Fe levels allows tighter control of micronutrient composition and is highly valuable for mechanistic studies, our decision to use standard chow Fe levels as the control was based on the following considerations: (1) Maintaining the vivarium diet Fe levels in the control group preserves baseline physiological adaptation and avoids introducing a two-step adjustment which could independently alter metabolism, microbiota composition, or immune tone. (2) The relatively high iron content in NIN establishes a robust iron-replete state with adequate tissue stores. This provides a large dynamic range for detecting hematologic and molecular consequences of iron restriction, which was the focus of our study. (3) Anchoring our comparison to the vivarium chow Fe conditions allow our study to be compared to other studies that induce gestational stressors, including environmental Pb (all Pb data in rodents use normal vivarium chow) or maternal infections.

(ii) Our dietary model isolates gestational iron deficiency without severe maternal anemia, and thus does not encompass the full clinical spectrum of iron deficiency severity observed in human pregnancies. Extrapolation to more severe or prolonged deficiency should be made cautiously.

(iii) ICP-MS is highly sensitive and allows detection of trace metals as elemental ions this method does not provide information regarding speciation (i.e Fe²⁺ vs Fe³⁺), spatial information within a tissue or sub-cellular resolution. Future studies are planned to integrate regional metal imaging with spatial gene expression.

(iv) Our study does not include analysis of maternal milk composition that could influence neonatal metal exposure. Given that the relationship between milk metal levels in pups is not strictly proportional and is tissue-specific, we focus on pup tissue metal levels that reflect a combination of maternal supply and tissue-specific regulatory mechanisms, rather than a direct one-to-one relationship with milk composition.

(v) Our study does not provide specific information regarding iron uptake into the mouse brain that occurs primarily at the blood–brain barrier via transferrin receptor–mediated endocytosis, followed by intracellular release of iron and transport into the brain parenchyma through DMT1 and ferroportin (143–146). We recognize that iron deficiency induces compensatory mechanisms to increase iron delivery to the brain, including enhanced transferrin-mediated uptake, increased vesicular trafficking at the endothelial interface, and recruitment of alternative carriers such as H-ferritin(147–149). However, our data suggests that despite such adaptations and early postnatal supplementation, regional brain iron levels remain altered, indicating that increased transport at the blood brain barrier cannot fully overcome the longer-term consequence of GID.

(vi) Spatial transcriptomic analyses focused on the P7GID cohort, which exhibited the most persistent metal dysregulation; although this allowed a stringent assessment of long-term developmental consequences, it does not exclude additional, potentially distinct transcriptional effects in earlier-or later supplemented cohorts. In addition, while all bulk data controlled for sex as a variable, our transcriptomic study was limited to P40 brains from males.

(vii) Our data revealed robust alterations that support our previous finding of an impaired balance of excitation and inhibition in P40 offspring exposed to the same dietary regimens. However, we have not tested whether the specific transcriptional changes are causative for these functional impairments, which was beyond the scope of this work.

Abbreviations:

GID

Gestational Iron Deficiency

ICP-MS

Inductively Coupled Plasma Mass Spectrometry

ID

Iron deficiency

NIN

Iron-Sufficient Control Diet Group (commonly referring to a nutritionally adequate standard diet)

P

Postnatal Day (used to indicate age in days after birth)

P0GID

Gestational Iron Deficiency group with iron repletion starting at Postnatal Day 0 (birth)

P7GID

Gestational Iron Deficiency group with iron repletion starting at Postnatal Day 7

Data Availability:

Data described in the manuscript, code book, and analytic code will be made available upon request pending approval and in accordance with NIH guidelines and requirements