Maternal inflammation significantly impacts cortical interneuron development in a subtype-specific manner

Severe infections during pregnancy are one of the major risk factors for cognitive brain impairment in offspring. It has been suggested that maternal inflammation leads to dysfunction of cortical GABAergic interneurons that in turn underlies cognitive impairment of the affected offspring. However, the evidence comes largely from studies of adult or mature brain and how impairment of inhibitory circuits arises upon maternal inflammation is unknown. Here we show that maternal inflammation affects multiple steps of cortical GABAergic interneuron development, i.e. proliferation of precursor cells, migration and positioning of neuroblasts as well as neuronal maturation. Importantly, the development of distinct subtypes of cortical GABAergic interneurons was discretely impaired as a result of maternal inflammation. This translated into a reduction in cell numbers and redistribution across cortical regions and layers. Furthermore, vulnerability of GABAergic interneuron subtypes was associated with varying impact of maternal inflammation on interneuron precursor pools that depends on the stage of brain development. Thus, differential effect of maternal inflammation on GABAergic interneuron subtypes and the time of insult might be key factors contributing to etiology of cognitive impairment in maternal inflammation-affected offspring.


Introduction
Development of the fetal brain is highly influenced by the maternal environment. A multitude of factors such as maternal nutrition, stress, hormonal imbalance as well as the maternal immune status play key roles in shaping normal brain development 1 . Maternal inflammation is known to increase the risk for severe psychiatric disorders including schizophrenia, bipolar disorder, intellectual disability, anxiety, autism spectrum disorders (ASDs) and cerebral palsy 2,3 with high societal costs 4 . Although the exact mechanism of adverse neurodevelopment upon maternal inflammation is not known, data from animal experiments and clinical observations suggest that the cytokine-related pro-inflammatory response in the mother contributes to disordered development of the fetal brain and predisposes the offspring to additional stressors during postnatal maturation of the brain [5][6][7] .
Two of the three classes of GABAergic interneurons are generated in the medial ganglionic eminence (MGE), i.e., the parvalbumin (PV)-and somatostatin (SST)-expressing interneuron types, whereas the third class is generated in the caudal ganglionic eminence (CGE) and gives rise to a highly heterogeneous group of neurons that express serotonin receptor 3A (5HT3AR) and consists of vasoactive intestinal polypeptide (VIP), neuropeptide Y (NPY), reelin-positive and a few other subtypes [8][9][10][11] . In mice, cortical GABAergic interneurons are produced during middle-late gestation (E9.5-E18.5) 12,13 , and migrate tangentially from the GEs to the cortical plate, where they start migrating radially into prospective cortical layers 14,15 . Subsequently, a significant proportion of interneurons undergoes programmed cell death 16,17 , while the surviving cortical interneurons mature over two to three months in mice 9 .
Excitation-inhibition imbalance in the brain has been proposed as a major factor underlying the behavioral outcomes and cognitive decline associated with various neurodevelopmental disorders, including those associated with maternal inflammation [18][19][20][21] . Indeed, several studies show changes in activity and distribution of GABAergic interneurons in the cortex and hippocampus of the adult mouse brain upon maternal inflammation [22][23][24][25] . Such abnormalities are thought to occur due to 4 perturbed brain development during early fetal or juvenile periods. Importantly, the development of neuronal circuits in the brain continues until late adolescence, up to 60-70 days postnatally in rodents and 20-25 years in humans 26 . Despite being acute, maternal inflammation can have longlasting effects on several critical developmental processes such as precursor cell proliferation, neuronal migration and differentiation during embryogenesis, as well as postnatal neuronal maturation including neuronal survival, dendritic, synaptic and axonal pruning and synaptogenesis.
Some studies have addressed the developmental impairment of principal cortical neurons upon maternal inflammation 6,27,28 , but surprisingly little is known regarding how development and maturation of cortical GABAergic interneurons are affected, and more importantly how those defects of GABAergic interneurons that are observed in adult [22][23][24][25]29 arise during brain development. Abnormal development of cortical inhibitory circuits will have a significant impact on animal behavior leading to phenotypes resembling human psychiatric disorders as has been shown for a number of genetic mouse models [30][31][32] . Furthermore, as cortical GABAergic interneurons represent a diverse class of neurons with more than twenty subtypes 33,34 , various subtypes of GABAergic interneurons might be differentially affected by maternal inflammation contributing to the complex behavioral abnormalities of the offspring.
To study the impairment of cortical interneuron development due to maternal inflammation, we utilized a mouse model of maternal inflammation that involved injecting polyriboinosinicpolyribocytidilic acid (poly I:C) at gestational day 9.5 (GD9.5). Poly I:C is a synthetic polynucleotide that mimics viral double stranded (ds) RNA and is known to cause an acute inflammatory response by activating pro-inflammatory cytokines including IL-6 35,36 . Following this, we undertook a detailed analysis of embryos and pups exposed to maternal inflammation. We found poly I:C to have an acute effect on GABAergic interneuron precursor proliferation with follow-up effects on the migration, positioning and maturation of GAD+ neuroblasts. The effect of maternal inflammation was interneuron subtype-specific and demonstrated differential vulnerability of interneuron subtypes to mother-derived insults.

Animal breeding and genotyping
All animal experiments were conducted in accordance with the guidelines of the National Animal Ethic Committee of Denmark. C57BL/6J (Janvier Labs), PV Cre (017320, Jackson Labs), ROSA-tdTomato (007905, Jackson Labs) and GAD67-EGFP (Gad1 tm1.1Tama ) 37 mice were used in this study, maintained in IVC cages and provided food and water ad libitum. Heterozygous GAD67-EGFP mice were bred with wildtype mice. Date of the vaginal plug was treated as GD0.5, and embryos and pups were timed accordingly. GD is used here to refer to the maternal stage of pregnancy while E is to indicate the age of the embryo. EGFP+ animals were identified by PCR using the following primers (in 5'-3' orientation) that amplified a 345 bp region:  Acoustic startle reaction (ASR) and pre-pulse inhibition (PPI) were tested at 5 weeks of age as pre-pulse + startle trials (pre-pulses of 72, 74, 78, and 86 dB(A)); 5 trials with only background noise). Tube movements were averaged over 100 ms following onset of the startle stimulus (AVG).
The five AVGs for each pre-pulse intensity were averaged and used to calculate PPI, which was expressed as percent reduction in averaged the pre-pulse AVGs compared to the average of the 10 middle startle trials: %PPI = [1-(Pre-pulse + pulse/Pulse)] x 100.

Perfusion, Sectioning and Immunohistochemistry
All steps were performed at ambient temperature unless otherwise noted.
Postnatal mice were anaesthetized with a combination of xylazine and ketamine injected i.p. This was followed by transcardial perfusion initially with cold PBS to flush out blood and then with cold 4% paraformaldehyde (PFA). Brains were then removed and post-fixed in 4% PFA overnight at 4°C before being stored in PBS with 0.01% Sodium Azide.

Image acquisition and analysis
Images were acquired using a confocal microscope (Leica SP8, Leica Microsystems) and analyzed using ImageJ (NIH) and Imaris (Bitplane AG). After correcting for brightness and contrast, figures were prepared using Adobe Illustrator (Adobe Inc). Graph preparation and statistical analysis were carried out using Prism 7.0 (GraphPad).

Maternal inflammation model in transgenic mice that labels all cortical GABAergic interneurons
In order to reveal how the development of cortical GABAergic interneurons is affected by maternal inflammation, we utilized the GAD67-EGFP knock-in transgenic mouse line that labels all GABAergic interneurons in the cortex by EGFP 37

Maternal inflammation impacts early stages of cortical GABAergic interneuron development
While the effect of maternal inflammation on offspring behavior has been well documented, there is little information about how maternal inflammation affects the development of inhibitory circuits in the cortex that might underlie abnormal animal behavior. We therefore investigated the effect of To account for the effect on distribution of GAD+ neuroblasts caused by acute maternal inflammation we analyzed the directionality of neuroblast migration in the IZ and cortical plate at E17.5. While GAD+ neuroblasts in the cortex of control offspring predominantly migrated towards the pia, their direction in inflammation-affected cortices was significantly different with a greater fraction migrating laterally (Figure 2f).

Maternal inflammation impairs positioning of GABAergic interneurons in the developing cortex
As interneuron migration continues until the second postnatal week, we chose to analyze whether these initially observed differences continue to affect the positioning of interneurons in the postnatal cortex. We hence studied several postnatal stages from P3 to P30 to understand how maternal inflammation can lead to a disordered arrangement of GABAergic interneurons in the mature cortex.
At P3 and P6, the difference in GAD+ cell density had amplified to cover the entire length of the cortical plate with the exception of the lowermost bins (bins 9-10, Figure 3a,b). Furthermore, the magnitude of the difference was consistently greater in the prospective somatosensory cortex than the motor cortex. Interestingly, the relative distribution of GAD+ neuroblasts across bins was similar in pups exposed to PBS or poly I:C (Suppl Figure 2a,b). This suggests that the decrease in GAD+ neuroblasts is equally distributed across the whole length of the cortex. By P9, while the reduction in density persisted in the somatosensory cortex, much lower difference could be observed in the motor cortex ( Figure 3c) indicating a differential effect of maternal inflammation on neuroblasts destined for distinct cortical regions.
Cortical GABAergic interneurons are overproduced during embryogenesis, and half of them subsequently undergo apoptosis within the second postnatal week 16,17 . We therefore investigated whether an increase in programmed cell death of cortical interneurons due to maternal 14 inflammation could explain in part the reduction of their number in the cortex of maternal inflammation-affected mice. To this end, we counted activated caspase-3 and EGFP-expressing cells in several sections covering the motor and somatosensory cortices. At both P3 and P6, we observed a similar number of activated caspase-3+ (Casp3+) as well as EGFP+ Casp3+ cells in control pups and pups exposed to maternal inflammation (Figure 3d,e). Thus, activation of the maternal immune system during early gestation did not augment apoptosis in the developing cortices of pups exposed to maternal inflammation.
To follow up the effect of maternal inflammation during postnatal brain maturation, we further studied the density and positioning of interneurons at P15. Similar to P9, the density of GAD+ interneurons in S1 was severely reduced (Figure 3f,g). Analysis of the layer-wise distribution of interneurons showed an equal effect across the whole cortical length in the somatosensory cortex ( Figure 3h). In contrast, in the motor cortex, despite a small decrease in the density of GAD+ interneurons in the layer 2/3, no statistically significant difference was observed (Suppl. Figure 2c).
We hence show that an acute induction of inflammation in pregnant dams impairs migration and final positioning of GAD+ neuroblasts with the effect being more pronounced in the somatosensory than in the motor cortex.

Differential effect of maternal inflammation on distinct subtypes of cortical GABAergic interneurons
While functional impairment of cortical GABAergic interneurons has been proposed to be one of the major factor contributing to imbalance between excitation and inhibition in patients with schizophrenia and other psychiatric disorders, most of the attention has been directed towards dysfunction of PV+ GABAergic interneurons 25,49,50 . However, cortical GABAergic interneurons are highly heterogeneous, and PV+ interneurons represent only one class of interneurons 8 . Therefore, the differential impact of maternal immune activation on development and maturation of various interneuron subtypes remains to be elucidated.
To this end, we studied the effect of maternal inflammation on the organization of the three largest interneuron populations, i.e. those that express PV, SST and VIP (Figure 4a,b, Suppl. Figure 2d,e).
In the somatosensory cortex, the previously observed decrease in the density of GAD+ interneurons stemmed from a decrease in both PV+ and SST+ interneurons (Figure 4a-d).
Interestingly, there was a clear layer-specific effect of maternal inflammation in the distribution of PV+ and SST+ interneurons. Thus, the decrease in number of PV+ interneurons was restricted to layer 4 at P15 (Figure 4c), whereas the reduction of SST+ interneurons was more pronounced across all layers, albeit statistically significant only in layers 2-4 at P15 (Figure 4d). There was no effect of maternal inflammation on the distribution of CGE derived VIP+ interneurons at P15 (Suppl. Figure 2d,e). At P60, the density of PV and SST was not statistically significant between maternal inflammation-affected and control animals even though an appreciable difference could be seen in the case of L4 SST+ neurons (Figure 4e,f). This is likely caused by the non-uniform expansion of cortical areas during synaptogenesis and gliogenesis 51 . In the motor cortex, there was no significant effect on distribution of interneuron subtypes at P15 and P60 (data not shown).
In addition to the reduced interneuron density, a reduction in PV expression in the cortex and hippocampus has been described in several studies for adult mice 21 . While we confirmed that maternal poly I:C exposure led to a similar reduction in PV expression in both M1 and S1 cortical regions at P60 (Figure 4i were assessed for intrinsic electrophysiological properties using whole-cell patch-clamp recordings of fluorescent neurons in acute brain slices (Figure 4k-n). The resting membrane potential in poly I:C-exposed animals was found to be significantly hyperpolarized as compared to controls ( Figure   4k,l). In addition, the firing frequency of PV+ interneurons was reduced in poly I:C-exposed animals ( Figure 4n) reflecting an overall slower spiking rate due to maternal immune activation.
To investigate if the changes in electrophysiological properties are accompanied by anatomical changes, we analyzed the morphology of biocytin-filled PV+ interneurons (Figure 4o). There was no significant effect of maternal inflammation on the length of dendritic filaments (Figure 4p) and branch depth (Figure 4q). However, Sholl analysis of the reconstructed interneurons showed a significant reduction in overall dendritic tree complexity of the maternal inflammation-affected PV+ interneurons ( Figure 4r). These results indicate that acute maternal immune response affects morphology and electrophysiology of GABAergic interneurons.

Timing of maternal inflammation affects discreet pools of interneuron progenitors
One of the most remarkable effects of maternal inflammation on development of the GABAergic interneurons was the decrease in the density of EGFP+ neuroblasts as early as E14.5 when maternal immune activation was induced at GD9.5 (Figure 2b,c). As cortical GABAergic interneurons are generated by precursor cells mainly in the MGE and CGE between E9.5 and 18.5 12,13 , acute maternal inflammation might affect the generation of GABAergic interneurons. To investigate this, GAD67-EGFP pregnant mice were injected with poly I:C at GD9.5 and proliferating cells were labelled using a 2-hour pulse of the thymidine analog, 5-bromo-2'-deoxyuridine (BrdU), at GD10.5 or GD14.5. By co-labeling interneuron precursors for BrdU along with Nkx2-1 (MGEspecific marker 41,54 ) or COUP-TFII (CGE-specific marker 55 ), we sought to ascertain differences in proliferation of interneuron precursors due to maternal inflammation. At E10.5, we found a ~25% decrease in the percentage of Nkx2.1+ precursors that co-labelled with BrdU in embryos exposed to maternal inflammation in comparison to unexposed mice (Figure 5a As with the previous analysis, a 2-hour pulse of BrdU was given before collecting the embryonic brains for analysis at E14.5. We observed a ~20% decrease in COUP-TFII+ progenitors that colabelled with BrdU, similar to the effect on Nkx2.1+ progenitors at E9.5-E10.5 (Figure 5f). Thus, progenitors of GABAergic interneurons exhibit differential vulnerability that depends on stage of embryonic brain development.

Discussion
Despite significant evidence showing that maternal immune activation during pregnancy leads to cognitive dysfunction in the offspring that might be triggered by excitation-inhibition imbalance, little is known how the development of inhibitory system in the brain is affected. Here we showed that maternal immune activation results in multiple "hits" on development of GABAergic neurons, thus affecting proliferation of precursors, migration and positioning of neuroblasts as well as their maturation.
Genetic and environmental factors that affect cortical development provide important insights into the cellular and molecular basis of neurodevelopmental disorders. Available epidemiological and clinical findings 56 point towards a link between maternal infections during gestation and increased risk of developing a mental disorder due to impaired cortical development. While the precise mechanism of action is elusive, an increase in maternal cytokine levels is thought to lead to the transmission of maternal inflammation to the fetal brain 3 . However, the developmental changes occurring upon maternal inflammation in the fetal brain and during postnatal brain maturation have been understudied owing to ethical and technical challenges in humans. Thus, rodent models of maternal inflammation provide an excellent alternative to study the effects of an acute inflammatory insult on brain development. Accordingly, poly I:C and lipopolysaccharide (LPS) have emerged as two popular molecules that mimic viral and bacterial infections respectively and robustly activate the maternal immune system. In this study, we mimicked maternal viral infection by injecting poly I:C at GD9.5 and followed the development of cortical GABAergic neurons in the offspring brains from E10.5. Strikingly, in spite of an acute immune response, we revealed that the effect was both immediate, i.e. affecting proliferation of precursors of interneurons, and longlasting, i.e. affecting migration of neuroblasts and maturation of cortical GABAergic neurons during late embryonic or early postnatal brain development.
One of our major findings is that the effect of maternal inflammation has a differential effect on GABAergic interneuron subtypes, highlighting subtype-specific vulnerability of neurons to maternally-derived stimuli. Hitherto, studies have revealed a convergent effect of genetic and environmental schizophrenia-related insults on PV+ interneurons 21,25 . However, while PV+ interneurons are crucial in synchronizing spike timing via gamma-oscillations, suppression of their activity alone has been found insufficient to reproduce schizophrenia-like phenotype in genetic or environmental mouse models of schizophrenia 57 . We show here a differential impact of maternal inflammation on interneuron subtypes and cortical regions that goes beyond PV+ interneurons.
GABAergic interneurons in the somatosensory cortex were more affected in comparison to their counterparts in the motor cortex. Furthermore, the impact on PV+ interneurons was localized specifically to layer 4 and could be traced as early as P15, suggesting a specific developmental impairment. Importantly, we found a significant impact of maternal inflammation on SST+ interneurons, which were affected not only in layer 4 as PV+ interneurons but also in layer 2/3 despite both subtypes being derived from MGE-derived Nkx2.1+ progenitors. The overlap in layer 4 is of significance as SST+ interneurons in this layer target mainly PV+ fast-spiking interneurons unlike in layer 2/3 where pyramidal neurons are the primary targets 58 . Selective ablation of SST+ interneurons during development has been shown to arrest the maturation of PV+ interneurons in layer 5/6 suggesting an early role for SST+ interneurons in PV+ interneuron maturation 59 .
The effect of maternal inflammation at GD9.5 on MGE-derived PV+ and SST+ interneurons in the offspring can be observed as early as E10.5, due to decreased proliferation of MGE-derived Nkx2.1+ progenitors. However, this effect was acute with proliferation returning to wild type levels by E14.5. This immediate and short-lived effect of maternal inflammation is in line with previous reports on cortical progenitors in an LPS model 28,60 . Contrary to the effect on MGE-derived subtypes, we found no change in CGE-derived VIP+ interneurons in animals exposed to maternal inflammation at E9.5, and neither CGE-derived COUP-TFII+ progenitors nor VIP+ interneurons in the mature cortex were affected. This is ostensibly due to a later 'birthdate' of CGE-derived interneurons between E12.5-18. addition to the impact on precursor proliferation, described above, we show that migration of interneurons into the dorsal cortex was impaired as early as E14.5. The effect of maternal inflammation on the distribution of cortical GABAergic interneurons was maintained at E17.5 but could only be seen in the superficial cortical bins. This could be due to a greater proportion of GAD+ neuroblasts in maternal inflammation-exposed embryos migrating laterally and further work using time-lapse microscopy will be important to clarify the mechanism of this impairment.
Likewise, while analysis of early postnatal stages showed a decreased density of GAD+ neuroblasts in regions of the developing cortex that correspond to both motor and somatosensory cortices, by P15 the effect persisted only in the somatosensory cortex. This suggests a differential regional impairment of cortical interneurons. Despite this, perinatal reduction in the number of GAD+ neuroblasts in the motor cortex might still affect the maturation of early cortical circuits and have functional outcomes that could not be measured in this study.
In addition to the differences in proliferation and migration, the decrease in interneuron numbers can also be attributed to programmed cell death. Previous studies have shown that GAD+ neuroblasts undergo apoptosis during the first two postnatal weeks with a peak between P6-P9 16,17 . However, at both P3 and P6, we observed comparable numbers of GAD+ neuroblasts expressing activated caspase-3 suggesting that maternal inflammation does not affect interneuron cell death. GABAergic interneurons across all developmental processes. Such data will provide mechanistic insight into etiology of human neurodevelopmental disorders and will identify therapeutical window during brain development when impairment GABAergic circuitry can be corrected.