Gestational iron deficiency affects the ratio between interneuron subtypes in the postnatal cerebral cortex in mice

The high prevalence of gestational iron deficiency (gID) in non-anemic pregnant women remains a major concern (Auerbach et al., 2021) due to its potential to affect brain development in the growing fetus and newborn offspring (Georgieff, 2020). Studies in humans and animal models have shown that gID during the first trimester is associated with many intellectual disabilities (Wiegersma et al., 2019) that are associated with an imbalance in excitation and inhibition (Rivero et al., 2015; Bollmann et al., 2015; Lener et al., 2016) such as Autism Spectrum Disorders (ASD) (Schmidt et al., 2014; Pizzarelli and Cherubini, 2011; Dickinson et al., 2016), Attention Deficit Hyperactivity Disorder (ADHD) (Doom et al., 2015), bipolar/unipolar disorder and anxiety disorders (Chen et al., 2013; Yang et al., 2021).

While many studies of gID in animal models have focused on the hippocampus (i.e. Liu et al., 2021; Tran et al., 2012; Rao et al., 2013, 2011; McEchron and Paronish, 2005), our own studies in mice showed that gID was associated with a highly blunted response to excitatory stimuli that likely involves cortical neurons (Rudy and Mayer-Proschel, 2017). Interestingly, the changes we saw in response to excitatory stimuli within gID animals occurred despite postnatal iron supplementation, suggesting an impact of gID on embryonic neural progenitor cells and their progeny.

To gain insight into the cellular components that might contribute to a shift in the excitatory/inhibitory (E/I) balance in gID offspring, we focused on GABAergic interneurons in the cerebral cortex, a brain region where even subtle changes in the E/I balance are associated with changes in behavior (Yizhar et al., 2011).

GABAergic inhibitory interneurons that populate the cerebral cortex represent 20-30% of neurons in the adult brain, and are responsible for controlling, both temporally and spatially, the amount of excitatory and inhibitory inputs that individual neurons receive (Marín, 2012). Thus, GABAergic interneurons have an effect on the E/I balance that is disproportionately large compared with their relatively small cell number in the brain. GABAergic interneurons originate from two regions of the subpallium, the ganglionic eminence and the preoptic area. The ganglionic eminence is a transient portion of the subpallium found during development that can be further subdivided into three major progenitor domains: the lateral ganglionic eminence (LGE), medial ganglionic eminence (MGE) and the caudal ganglionic eminence (CGE), which are all separate from the preoptic area (POA) (Flames et al., 2007).

The MGE gives rise to the majority of the GABAergic interneurons that will populate the cortex (Welagen and Anderson, 2011) and is dependent on the morphogen sonic hedgehog (Shh) (Xu et al., 2010; Hernandez-Miranda et al., 2010) and on the induction of the transcription factor Nkx2.1. Nkx2.1 is transiently expressed in the MGE where it is required for induction of other factors (i.e. Lhx6) that affect tangential and radial migration as well as the differentiation of progenitor cells into parvalbumin- (PV) and somatostatin- (SST) expressing interneurons (Sandberg et al., 2016). Regulation of Nkx2.1 itself is dependent on appropriate Shh levels, with high levels of Shh signaling favoring the generation of SST-expressing interneurons at the expense of PV-expressing interneurons and low levels of Shh favoring the generation of PV at the expense of SST (Tyson et al., 2015; Kelsom and Lu, 2013; Xu et al., 2010). Thus, the ratio between these two subtypes of Nkx2.1-derived interneurons is dependent on the intensity of Shh signaling (Sansom and Livesey, 2009; Briscoe, 2009; Riobo and Manning, 2007; Ruiz i Altaba et al., 2002).

Inhibitory interneurons are also specified in the CGE (Wonders and Anderson, 2006; Butt et al., 2008; Hebert and Fishell, 2008). This specification is insensitive to Shh (Xu et al., 2010) and generates, among others, calretinin- (CR) expressing interneurons (Wonders and Anderson, 2006). In addition, there is a subset of interneurons which co-expresses CR and SST. These CR⁺/SST⁺ cells are derived from a small subset of progenitors that reside at the border between the MGE and CGE (Sousa et al., 2009), are preferentially born at early ages and give rise to inhibitory interneurons that are functionally distinct from SST⁺/CR⁻ interneurons (Xu et al., 2006). Irrespective of whether progenitors arise in the MGE or the CGE, their fate is determined before exiting the ganglionic eminence (Spatazza et al., 2017; Wonders and Anderson, 2006) and their prospective lineage markers are retained (Hsieh and Baraban, 2017; Inan et al., 2013)

Although the fate of the cells is relatively stable, the number of cells that will eventually populate the cerebral cortex is guided by migration (Brandao and Romcy-Pereira, 2015; Marín and Rubenstein, 2001) and programmed cell death (Southwell et al., 2012; Wong et al., 2018), which occurs between postnatal day (P) 0 and P14 and eliminates nearly half of the migrating GABAergic progenitor cells. Migration and elimination of GABAergic progenitors is mostly complete by P14, when progenitors begin expressing mature interneuron markers PV, SST or CR. Given the well characterized pathways which generate and specify cortical interneurons throughout development, we tested whether gID could impact the generation of these cells.

We used our previously described nutritional mouse models (Rudy and Mayer-Proschel, 2017), in which we randomly assigned dams to control groups (animals receiving iron-normal diet before and throughout gestation) or iron-deficient groups (animals receiving iron-deficient diet before and throughout gestation until the first week post-partum). After 7 days post-partum, the lactating dams and offspring in the iron-deficient cohort were switched to an iron-normal diet until the time of sacrifice (designated as gID group).

To determine the time period of embryonic brain iron depletion and subsequent iron tissue repletion in the cerebral cortex of embryos and offspring, we used Inductively Coupled Plasma Mass Spectrometry (ICP-MS) to measure iron levels at embryonic day (E) 14, P7, P14 and at P42 (a time point at which we identified a change in the E/I balance in previous work using the same dietary model; Rudy and Mayer-Proschel, 2017). As shown in Fig. 1A, gID E14 brains have significantly lower tissue iron levels than brains harvested from E14 control embryos (6.4 µg/g versus 11 µg/g, P<0.0001). Cortical iron levels remained significantly lower in the gID offspring compared with controls at P7 (3.3 µg/g versus 4.3 µg/g, P=0.04) and P14 (2.6 µg/g in gID versus 4.4 µg/g in controls, P<0.001). However, by P42 cortical iron levels were no longer significantly different in the gID offspring compared with those of control brains (4.2 µg/g versus 5.2 µg/g, P=0.3).

As iron deficiency affects growth, we monitored the crown-rump length of embryos exposed to iron deficiency and found no difference in crown rump length of gID embryos compared with controls (P>0.06) (Fig. 1B). Weight was, however, impacted by gID during postnatal development despite iron supplementation to the dam. At P14 gID pups weighed ∼27% less than controls (6.0 g gID versus 9.0 g in controls, P=0.0007). At P21 (the time of weaning), P28, P35, P42 and P100, the weight of gID animals remained 25-35% lower than controls (P<0.001) (Fig. 1C). The significantly lower weights of gID animals persisted until at least P100 and were not driven by sex or litter sizes, as both male and female gID animals weighed significantly less than respective males and females from the control group and litters were not significantly different in the two groups.

To test our hypothesis that changes in the E/I balance in the gID groups stem from changes in interneuron progenitor cells during embryonic development (that contribute to the E/I balance), we labeled sections with anti-Nkx2.1 and the proliferation marker Ki67 and quantified the number of proliferating and non-proliferating Nkx2.1⁺ cells in the MGE, the preoptic area and regions outside the MGE (Fig. 2A-E). We observed a significant increase in the number of proliferating Nkx2.1⁺ cells (Nkx2.1⁺/Ki67⁺) in the MGE of gID brain compared with controls (Fig. 2C), but did not see changes in the number of Nkx2.1⁺/Ki67⁺ cells in the POA (Fig. 2D), nor did we observe a difference in the number of non-proliferating cells (Nkx2.1⁺/Ki67⁻) (Fig. 2E) outside the MGE in gID animals compared with controls.

As proliferation of Nkx2.1⁺ cells is driven by the morphogen Shh we tested whether gID could affect Shh signaling and thus the early embryonic specification of interneurons. We used the Gli1-lacZ reporter mouse (Garcia-Marques et al., 2010; Ahn and Joyner, 2004), which allows for analysis of canonical Shh signaling. Shh induces gene expression through Gli transcription factors which confer specific cell identities (Sansom and Livesey, 2009) including interneuron specification in the MGE (Briscoe, 2009; Riobo and Manning, 2007; Ruiz i Altaba et al., 2002). lacZ expression in heterozygous Gli1 reporter mouse embryos is indistinguishable from endogenous Gli1 mRNA expression and a useful proxy for Shh signaling (Bai et al., 2002). Furthermore, increased Gli1 signaling is associated with increased expression of Nkx2.1 and proliferation of Nkx2.1⁺ progenitor cells (Palma and Altaba, 2004; Welagen and Anderson, 2011; Marín and Rubenstein, 2001; Pleasure et al., 2000). As shown in Fig. 2F, gID embryos showed expanded expression of Gli1 in the MGE with a significant increase in lacZ expression specific to the rostral regions of the MGE (Fig. 2G), suggesting that the increased proliferation of Nkx2.1⁺ progenitor cells could be driven by changes in gID-associated Shh signaling.

Proliferating Nkx2.1⁺ progenitor cells (cells fated to generate GABAergic interneurons; Sousa and Fishell, 2010; Xu et al., 2005; Sussel et al., 1999; Komada et al., 2008; Wines-Samuelson et al., 2005) in the MGE contribute to the number of GABAergic progenitors which reach the cortex, although their final number is controlled by programmed cell death that peaks at around P7 to ensure appropriate number of mature inhibitory interneurons (Finlay and Slattery, 1983). As gID leads to suboptimal iron levels in brains until P42, and iron is an important metabolic component in many cell processes, we next tested whether the increase in proliferating progenitor cells in embryonic brains could trigger changes in programmed cell death and skew the number of more mature neurons at later time points. As shown in Fig. 2H, and quantified in Fig. 2I, we did not observe a significant difference in the number of cleaved caspase⁺ cells observed in gID brains compared with controls at the peak time of cleaved caspase 3-mediated cell death.

To test whether the increase in embryonic Nkx2.1⁺ cells has a consequence for the composition of interneuron subtypes at later stages, we quantified the number of major subtypes of GABAergic interneurons across medial and lateral regions in the cerebral cortex of P14 gID and control brains. Two of the three major subtypes of GABAergic interneurons, PV⁺ and SST⁺ predominantly originate from the MGE (Marín et al., 2000; Nery et al., 2002), whereas most of the remaining subtypes including CR⁺ interneurons, as well as SST⁺/CR⁺ double positive cells, originate outside the MGE domain or at the border of MGE/CGE (Miyoshi, 2019; Mi et al., 2018). We quantified the total number of SST⁺, PV⁺, CR⁺ and SST/CR⁺ in medial (circles) and lateral (squares) regions of the cortex with each data point representing the average of commensurate sections from the right and left cortical hemisphere in each animal. In the brains of gID animals we found a trend towards more SST⁺ and a significant decrease in PV⁺ cells in medial (P=0.006) and lateral (P<0.0001) regions (as defined in Materials and Methods) (Fig. 3A,B). The number of CR⁺ and SST⁺/CR⁺ was comparable across regions and dietary groups and raw cell counts in our control brains were overall consistent with findings from others (e.g. Meyer et al., 2011; Clem and Barth, 2006). To account for the smaller size of gID animals at this age (see Fig. 1C), we calculated the percentage of each interneuron type from all DAPI labeled cells (Fig. 3C) and found in gID cortices a 28% (P=0.02) increase in SST⁺ interneurons in medial regions in conjunction with a 33% decrease in PV⁺ interneurons in the same region (P=0.0001). We also saw a trend towards an increased percentage of SST⁺ cells in more lateral regions (P=0.06) that coincided with a significant decrease in the percentage of PV⁺ cells (P=0.003) in the same region in gID brains. The percentage of CR⁺ and CR⁺/SST⁺ cells from all cells were not significantly different in gID brains compared with controls (P>0.1). When we further examined the percentage of SST⁺ and PV⁺ interneurons at six distinct distances from the midline, we found that gID brains had an increased percentage of SST⁺ interneurons and a decreased number of PV⁺ interneurons at every distance from the midline measured. A repeated measure mixed model analysis of this data confirmed that the percentage of SST⁺ interneurons was significantly increased in gID brains across all distances from midline (P=0.03), whereas the percentage of PV⁺ interneurons was significantly decreased (P=0.02) (Fig. 3D,E). It also confirmed that in both gID and control brains there is a significant correlation between interneuron count and distance from the midline. As we saw an overall decrease in PV⁺ cells, we further examined the anterior region of the cortex and found a significant loss of PV⁺ interneurons from the more superficial layers 2/3 (P=0.04) with no changes in the percentage of PV⁺ cells in the 4/5 and 6a deeper layers (Fig. 3F). In our previously published research (Rudy and Mayer-Proschel, 2017), we reported a significantly higher seizure threshold in the gID animals compared with controls irrespective of sex; however, females were more sensitive to Class V seizures than males. Based on these observations, we prioritized female animals in our analysis of postnatal animals.

Exposure of the developing brain to gID resulted in cellular changes in progenitor cell pools that are known to give rise to interneurons in the adult brain. To determine whether the cellular changes we observed in iron-depleted P14 brains persisted into adulthood when iron levels were restored, we quantified PV⁺, SST⁺, CR⁺ and CR⁺/SST⁺ cells in the cerebral cortex at P100. For these experiments we again labeled and counted cells across the entire cerebral cortex as shown by a representative image in Fig. 4A. We found that the raw cell counts for SST⁺ cells still trended towards higher number in gID brains (Fig. 4B). When we calculated the percentage of SST⁺ cells of all DAPI⁺ cells in the outlined regions of the cerebral cortex we found that, compared with control brains, gID offspring had a significantly higher percentage of SST⁺ cells in the medial region (42%, P=0.03) and a trending increase in the lateral region (P=0.06) (Fig. 4C). The overall distribution of SST⁺ cells as measured by percentage of labeled cells from distance to midline was not affected by gID, indicating that the surplus SST⁺ cells in gID are not the result of a site-specific accumulation (Fig. 4D). The number of PV⁺, CR⁺ and CR⁺/SST⁺ cells as well as the percentage of these cells of all DAPI labeled cells was comparable in gID versus control brains in both lateral as well as medial regions (P>0.9).

Our analysis was motivated by our previous discovery of functional changes in excitation and inhibition in the brains of animals that were exposed to gID (Rudy and Mayer-Proschel, 2017). Our nutritional model of gID renders embryonic brain tissue iron deficient as early as E14, and iron supplementation to the dams 1 week after birth prevents severe iron deficiency anemia (Rudy and Mayer-Proschel, 2017). In this model, brain iron levels remain low at P7 and P14 as seen by others (Hubbard et al., 2013) and are restored by P42. Restoration of brain iron levels through postnatal supplementation is not associated with normalization of body weight, an observation seen by others and ascribed to being, at least in part, a result of persistent differences in body fat (Kwik-Uribe et al., 2000). The dietary protocol we use mimics a human relevant situation where gID without severe anemia is present during gestation and is associated with low birth weight (Ribot et al., 2012). As in human populations, we exposed the newborn offspring (via dam) to iron supplementation, which restores brain iron levels by P42. Despite this, gID offspring at this age showed a persistent blunted seizure response to an excitatory stimulus (Rudy and Mayer-Proschel, 2017), suggesting a disrupted E/I balance.

Inhibitory interneurons represent contributors in establishing the E/I balance of the adult cerebral cortex and, once specified, represent a relatively stable cell population. In fact, MGE-derived interneuron progenitor cells migrate and integrate into neuronal networks even after transplantation into adult tissue (Wichterle et al., 1999). The MGE is the birthplace of interneurons and is defined by expression of Nkx2.1⁺. Shh signaling both induces Nkx2.1 and drives proliferation of Nkx2.1⁺ cells in the MGE (reviewed in i.e. Llorca and Deogracias, 2022)

Our analysis shows an increase in proliferating Nkx2.1⁺ cells in the MGE at E14 in gID brains compared with controls. This is preceded by an increase in Gli1 signaling in the MGE of gID embryos at E12, suggesting that suboptimal iron levels affect Shh signaling and downstream cell proliferation. Although the mechanism is under investigation and outside the scope of this study, we suggest that low iron levels affect lipid homeostasis which in turn impacts lipid modifications that are essential for proper Shh signaling (Grover et al., 2011; Ducuing and Querenet, 2013; Creanga et al., 2012; Palm et al., 2013; Ryan and Chiang, 2012). Changes in lipids modulate the function of Shh by influencing not only its distribution (Lou et al., 2005; Johnson et al., 2002; Digilio et al., 2003; Cooper et al., 2003; Yoshida and Wada, 2005; Dietschy and Turley, 2004; Huang et al., 2007) but also signaling itself (Briscoe, 2009; Riobo and Manning, 2007; Ruiz i Altaba et al., 2002). Iron affects lipid homeostasis through several mechanisms including through its role as a crucial cofactor for members of the cytochrome P450 superfamily (Correia et al., 2011; Dhur et al., 1989; Kamei et al., 2013). Although the iron-lipid dynamic in the embryonic brain is not known, it is likely that changes in lipid homeostasis affect Shh signaling at various levels and contribute to the gID-associated changes in the Shh signaling-dependent embryonic MGE.

The changes in specific embryonic progenitor cells that can give rise to interneurons allow for prediction of changes in the progeny of these cells. As nearly 70% of adult cortical GABAergic interneurons originate from MGE-specified Nkx2.1⁺ cells (Harwell et al., 2015) and generate PV⁺ and SST⁺ cells, changes in the number of Nkx2.1⁺ cells could affect the number of these more mature cells later in development. In contrast, a smaller subpopulation of GABAergic progenitors originates from the CGE and populates the superficial layers of the cortex (Miyoshi and Fishell, 2011) mainly giving rise to CR⁺ cells. These cells would less likely be affected by changes in the Nkx2.1⁺ cell pool. When we analyzed the composition of these interneurons at P14, we found increased numbers of SST⁺ interneurons together with a decrease in PV⁺ interneurons in gID. Although our observations at this stage are correlative, increased Shh signaling in the MGE has been shown to lead to production of SST⁺ interneurons at the expense of PV⁺ interneurons (Xu et al., 2005; Sussel et al., 1999). Whether this change is based on a fate switch or due to changes in proliferation remains unclear. We also did not observe changes in SST⁺/CR⁺ cells. These cells arise in the Nkx6.2 domain in the dorsal MGE where Shh signaling is already enhanced relative to the ventral MGE (Wonders et al., 2008; Yu et al., 2009). A further modest increase in signaling may not lead to further changes in these cells. While we focused on Nkx2.1⁺ progenitor population, it is likely that the gID-induced increase in proliferation of Nkx2.1⁺ cells affects downstream events that impact the balance of SST⁺ and PV⁺ neurons.

gID may also drive an imbalance of interneuron cell populations independently or in addition to changes of Shh signaling. For example, Nkx2.1 has been shown to restrict the expression of the orphan nuclear receptors Coup-TF1/2 in the MGE to a small subregion that is thought to bias progenitor cell toward specification to SST⁺ cells as well as cholinergic and pallidal neurons (Hu et al., 2017). Loss of Coup-TF1/2 affects apoptosis of MGE progenitors and is associated with precocious neurogenesis and increased proliferation (Hu et al., 2017). The increased proliferation of Nkx2.1 progenitor cells could thus be an initial Shh driven event that is exacerbated by further restriction of the Coup-TF1/2 domain. Shh-independent events, like non-canonical Wnt signaling, could also contribute to the impairment we identified in the MGE in gID embryos (McKenzie et al., 2019). Genetic loss of the non-canonical Wnt receptor Ryk in the MGE leads to the majority of both PV⁺ and SST⁺ cells failing to acquire subtype identity. This phenotype is, however, different from the upregulation of SST and concurrent downregulation of PV⁺ cells which is predicted by increased Shh signaling (Xu et al., 2005) and which we observed in the cellular composition in our gID brains. We also considered changes in programmed cell death as a possible consequence of gID that could ultimately affect the number and identity of interneurons that stably integrate into the mature cortex (Priya et al., 2018; Wong et al., 2018; Mancia Leon et al., 2020; Carriere et al., 2020). Although we cannot rule out low-level changes in cleaved caspase 3-mediated cell death over a prolonged time period, or shifts in cell death at other time points not tested here, we did not see significant changes of caspase⁺ cells in gID cortices at P7, the peak of programmed cell death. It is also possible that gID affects overall cell migration patterns, although the number of CR⁺ interneurons which migrate from caudal regions of the ganglionic eminence were unaffected by gID (Jimenez et al., 2002) and the number of postmitotic Nkx2.1⁺ cells outside the defined region of the embryonic MGE was not affected in gID brains, making it less likely that global migratory pathways are disrupted by gID.

The changes we identified in gID P14 brains occurred in the context of decreased tissue iron levels in the offspring that have not yet been normalized by the postnatal iron supplementation. We thus wanted to know whether resolution of low iron levels would also result in normalization of the cellular changes. We found partial normalization at P100, when the number and percentage of PV⁺ cells were no longer decreased, but the percentage of SST⁺ cells remained increased compared with controls. The ‘normalization’ of the number of PV⁺-expressing interneurons was surprising given that PV⁺-expressing interneurons are thought to be relatively stable in both morphology and distribution after P18 and continuing through P90 (Inan and Anderson, 2014). The same stability over time, however, has not been reported in the expression of the PV protein itself, which is known to be activity dependent (Patz et al., 2004). It is therefore possible that the ‘normalization’ of PV⁺ cell numbers represents a compensatory increase in the activity (and therefore immunofluorescent staining intensity) of the remaining PV-expressing interneurons. As our counting was based on blinded analysis of immunofluorescence images, it is possible that increased expression of PV protein in gID animals due to changes in activity could lead to an inflation of cell counts in gID animals. The use of a suitable reporter mouse line in future experiments could help rule out this possibility.

Although it is not yet clear what mechanisms underly the changes in interneurons we identified, it is tempting to speculate that they contribute to the blunted seizure response we saw previously in gID animals at P42 (Rudy and Mayer-Proschel, 2017). SST⁺ cells are the second most abundant type of interneuron in the cortex (Barinka and Druga, 2010; Wonders and Anderson, 2006). Release of SST can regulate the morphology and function of surrounding excitatory synapses and can lower excitatory synaptic transmission (Hou and Yu, 2013). Interestingly, higher concentrations of SST in the brain are known to correlate with decreased seizure intensity and an increased time to seizure onset, whereas loss of SST is known to have the opposite effect (Liguz-Lecznar et al., 2016). Thus, an increase in the total percentage of SST⁺ interneurons, as observed in our gID brains at P14 and P100, is likely to alter the overall E/I balance within the cortex at later time points. The increase in SST⁺ cells that was associated with a decrease in PV⁺ cells, especially in layer 2/3, is of special interest in light of data showing an association of gID with ASD in humans (Sidrak et al., 2014; Schmidt et al., 2014). Chandelier cells are a specific subtype of PV⁺ interneuron which resides preferentially in the upper portion of layer 2/3 and preferentially synapse with nearby (locally projecting) excitatory pyramidal cells at the axon initial segment (Woodruff et al., 2009). These cells are functionally distinct from the fast-spiking basket cells (Miyamae et al., 2017) and studies in ASD patients and genetic animal models of autism have revealed a specific reduction in PV⁺ chandelier cells that reside in layer 2/3 of the prefrontal cortex, while the number of basket cells was not severely affected (Ariza et al., 2016; Hashemi et al., 2017; Bozzi et al., 2018; Juarez and Martinez Cerdeno, 2022; Hsieh and Baraban, 2017). We have not yet resolved whether the loss of PV⁺ cells in gID brains represent chandelier cells, but such an additional analysis seems warranted and may provide additional insight into the cellular targets that connect gID with developmental disorders like ASD.

In summary, our analysis provides a link between gestational iron deficiency and specific cellular changes in the early embryo that affect the composition of adult cell populations crucial for establishing an E/I balance in the cortex. We furthermore identified gID-associated Shh signaling changes as a target for future studies.

All protocols were approved by the University Committee on Animal Resources at the University of Rochester. Mice were euthanized in accordance with American Veterinary Medical Association Guidelines for the Euthanasia of Animals.

Male and female Swiss Webster mice (age 8 weeks), strain code 024, were purchased from Charles River Laboratories. Animal diets were purchased from Envigo (formerly Harlan Teklad) custom research diets. Iron-normal (IN) diet contained 240 µg Fe/g of food (TD.05656). Iron-deficient (ID) diet contained 2-6 µg Fe/g of food (TD.80396). Diets were identical except for iron content. Animals were randomly assigned to the IN or gID cohorts with at least three separated dams/cohort. Dams were kept in pairs until mating to avoid isolation stress. Litters were not culled and pups were selected from three different litters/experimental cohort. Animals had free access to food and distilled water, while being housed on Bed-o'Cob ¼” corn cob bedding at 23°C, 43% relative humidity and kept on a 12 h light/12 h dark cycle. For timed matings, one male was placed into a cage with four females. The females were checked for the presence of a semen plug every 8 h (at 5pm, 1am and 9am). Successfully plugged females were marked before being placed into a separate cage with a non-mated random female. Pregnant females were euthanized at the appropriate time of gestation and embryos were harvested. Crown-rump length and uterine position was recorded for each embryo and matched as closely as possible between groups. For embryonic tissue, whole brains from E14 embryos were collected and processed.

Gli1-lacZ reporter mice, which carry a β-galactosidase knock in mutation that replaces the N terminal of Gli1 with the lacZ gene, were purchased as heterozygous males (JAX lab stock number 008211/background Swiss Webster) (Bai et al., 2002) and bred to wild-type Swiss Webster females that were either exposed to an IN or ID diet as described. Embryos were dissected as described and processed for analysis. For lacZ analysis, brains were fixed in 4% paraformaldehyde at 4°C for 30 min, cryoprotected in 30% sucrose overnight at 4°C and embedded in OCT. Then 10μm sections were generated and β-galactosidase activity was analyzed by incubation in X-gal solution at 37°C for 8 h.

For postnatal tissue harvest animals were perfused with heparin to prevent the contamination of blood-derived antigens and the contribution of blood to measured tissue iron levels in micro-dissected cortical brain samples. All analyses were conducted across 3-4 independent litters. Whole brains were micro-dissected for the cerebral cortex in Leibovitz's L-15 with L-glutamine (Thermo Fisher Scientific, MT10045CV). To ensure metal measurements reflected cortical metal levels only, cerebral cortices were micro-dissected to exclude nearby white matter and meninges. The cerebral cortex (both hemispheres for P7 animals, the left hemisphere for P14 and older animals) was then put in pre-weighed 1.7 ml polystyrene microcentrifuge tubes on ice (Laboratory Products Sales, L211511). Tissue-containing tubes were then weighed and processed for analysis

In collaboration with the University of Rochester Elemental Analysis Facility, cortical tissue iron levels were measured via ICP-MS using a NexION 2000C ICP Mass Spectrometer (PerkinElmer, NB150045). All cortices were hydrolyzed in their microcentrifuge tubes in ultrapure 67-70% nitric acid (VWR International, 87003-658) in a heat block at 100°C for 1 h. Before being run, samples were brought up to a final volume of 10 ml in UltraPure de-ionized water and metal levels per sample were normalized to their corresponding wet weights (µg/g) upon initial harvest before being frozen at −80°C for storage.

Whole brains were used for analysis of embryonic tissues and sections of cortex were used for postnatal tissue analysis. Postnatal brains were dissected from the skull in ice cold PBS, fixed in 4% paraformaldehyde for 24 h, cryoprotected in 30% sucrose for 24-48 h, cut starting at the midsagittal plane and frozen in OCT cryopreservation medium (Sakura, Tissue-Tek OCT Compound). Right and left hemispheres were frozen in the same block and sectioned at 10 μm thickness using a Leica CM3050S cryotome.

Immunofluorescence was conducted using a standard protocol performing antigen retrieval for all antibodies with a 10 mM sodium citrate solution at pH 6.0 and 100°C for 30 min. Primary antibodies were as follows: calretinin (MilliporeSigma, ab5054) at 1:200 dilution for 72 h; parvalbumin (Abcam, AB11427) at 1:1000 dilution for 72 h; somatostatin (Chemicon, ab354) at 1:120 dilution for 72 h; Ki67 (BD Pharminogen, 550609) at 1:50 dilution for 24 h; Nkx2.1 (Abcam, ab76013) at 1:400 dilution for 24 h; and cleaved caspase 3 (MilliporeSigma, PC679) at 1:250 dilution for 24 h. For cleaved caspase 3 staining, anti-mouse IgG (H&L) AlexaFluor488 (Invitrogen, A-11001) was used at 1:1000 dilution without a primary antibody in order to identify non-specific labeling and blood vessel-associated cell death. After immunofluorescent labeling, sections were imaged using a Leica Microsystems TCS SP5 confocal microscope (20× objective) or a Nikon A1R confocal microscope (20× and 40× objective) (for cleaved caspase 3) using consistent settings between samples.

All image analysis was performed using the free open-source FIJI software (Schindelin et al., 2012). The Allen Mouse Brain Atlas (100042147; https://mouse.brain-map.org/experiment/thumbnails/100042147?image_type=atlas) was used to define the boundaries of the cortex, which was matched for each image, traced by hand and extracted. All steps were performed blinded and using consistent settings between conditions.

To obtain cell counts, images were first converted to binary format. For Nkx2.1/Ki67 analysis, a 1 px median filter was applied to both channels to reduce background noise in the analysis. For all images, the watershed function was used to separate touching cells. The analyze particles function was then used to count objects in each image. DAPI nuclei counts were used to estimate the total number of cells in each image. After subtracting the no primary control channel, cleaved caspase 3 counts were performed by hand using the cell counter function in FIJI.

The right and left hemispheres were imaged and cells labeled for SST, PV and CR were counted in FIJI software at 0.5 mm, 0.8 mm and 1.1 mm (medial); and 1.4 mm, 1.7 mm and 2.0 mm (lateral) distance from the longitudinal fissure. For each distance, nuclei counts were performed as above for the right and left hemispheres and averaged to give a total cell count between the hemispheres at each distance from the midline. SST/PV/CR⁺ cells were counted manually using the cell counter function in FIJI. Cortical layers were approximated using morphological characteristics as shown in the Allen Brain Atlas. The cortex was divided morphologically into layer 2/3, 4/5 and 6a and SST, PV and CR counts were binned by layer for analysis of each region.

Morphological characteristics were used to match serial sections through each embryonic brain. To calculate the number of Nkx2.1⁺/Ki67⁺ cells, images were thresholded using the colocalization thresholding function of FIJI, which generates a new channel based on the intensities of both parent channels. Regions of interest were then drawn by hand that included the entirety of each co-labeled region. For each section, there were two clustered regions of Nkx2.1⁺/Ki67⁺ cells per hemisphere (the MGE and POA). Cell counts were then performed as previously described on the colocalized channel in each region of interest. The average cell number for the two combined hemispheres was then reported per image. To calculate the number of Nkx2.1⁺/Ki67⁻ cells, the Nkx2.1⁺ cells detected as Ki67⁺ were removed and the total remaining Nkx2.1⁺ were then counted. As there was no observable boundary between Nkx2.1⁺/Ki67⁻ cells from the MGE and POA, these cells were counted per hemisphere and the average value reported.

Analysis was conducted using GraphPad Prism 9.4.0. We used two-way ANOVA and post hoc Holm-Sidak multiple comparison or unpaired one-tailed or two-tailed Student t-tests where appropriate. QQ plot was used to confirm normal distribution. Definition of statistical significance: P<0.05 = 95% confidence interval. The specific test used for each dataset is described in the figure legends.

We thank the Elemental Analysis Facility of the Environment Health Science Center at the University of Rochester under directorship of Dr Matthew Rand for guidance and Thomas Scrimale for technical assistance in the iron analysis shown in Fig. 1A. We thank Mary Wines-Samuelson and Chris Proschel for their feedback and critique of this work. We also thank current and past members of the Mayer-Proschel lab for helpful scientific discussions.