Maternal immune activation induces sustained changes in fetal microglia motility

Maternal infection or inflammation causes abnormalities in brain development associated with subsequent cognitive impairment and in an increased susceptibility to schizophrenia and autism spectrum disorders. Maternal immune activation (MIA) and increases in serum cytokine levels mediates this association via effects on the fetal brain, and microglia can respond to maternal immune status, but consensus on how microglia may respond is lacking and no-one has yet examined if microglial process motility is impaired. In this study we investigated how MIA induced at two different gestational ages affected microglial properties at different developmental stages. Immune activation in mid-pregnancy increased IL-6 expression in embryonic microglia, but failed to cause any marked changes in morphology either at E18 or postnatally. In contrast MIA, particularly when induced earlier (at E12), caused sustained alterations in the patterns of microglial process motility and behavioral deficits. Our research has identified an important microglial property that is altered by MIA and which may contribute to the underlying pathophysiological mechanisms linking maternal immune status to subsequent risks for cognitive disease.

www.nature.com/scientificreports/ study, we induced MIA by Poly(I:C) injection at two different gestational days (E12 and E15) and characterized the biochemical and morphological microglial phenotypes at different developmental stages in offspring, from fetal to young adult. Furthermore, we used two-photon imaging to quantify microglial process motility, as we were unaware of any prior studies that examined how MIA effects this parameter. Microglial process motility is a key physiological parameter linked to their immune-surveillant and neuronal homeostasis functions, and we hypothesized that MIA may alter this motility. We found that maternal inflammation caused by E12 and E15 Poly(I:C) -injection consistently lead to increased IL-6 expression in liver, placenta and in fetal microglia. Microglial morphological parameters were unchanged in fetal and postnatal brains by E12 and E15 Poly(I:C) -injection, with only subtle changes in other biochemical phenotypes. However, MIA induced changes in process motility in embryonic microglia, and changes persisted into the postnatal and adolescent stages, albeit in different directions. Our results add an important functional parameter to the range of persistent changes seen in microglia following MIA that adds support to microglial abnormalities contributing to the link between maternal inflammation and subsequent cognitive disorders in offspring.

Results
Pattern of maternal and fetal cytokine changes following Poly(I:C) injection at E12 or E15. To induce MIA, we injected intraperitoneal Poly(I:C) to dams on two gestational days, E12 or E15 (birth is typically at E21). Poly(I:C) is a synthetic analogue which mimics the viral double-stranded RNA that activates Toll-like receptor 3, the innate immune receptor which promotes antiviral innate immune response 20,21 . Control mice were injected with saline and the mRNA expression profile of pro-inflammatory molecules was measured by real time PCR (Fig. 1), in fetal microglia, and in the maternal liver, placenta and brain. IL-6, IL-1β, TNFα and IL-17a, and the cell surface markers CD68 and ICAM-1, have all been previously implicated as key pro-inflammatory molecules that may mediate the behavioral consequences of maternal infection in the offspring 22 . Injections at both gestational day E12 (E12 Poly(I:C) -injection) and E15 (E15 Poly(I:C) -injection) significantly increased expression of IL-6 in both the maternal liver and placenta, but not in the maternal brain ( Fig. 1b-d and Table 1).
The levels of all the other cytokines (IL-1β, TNFα, IL-17a) did not change in maternal tissues from E12 or E15 MIA. (Fig. 1b-d, Table 1). CD68 expression was significantly increased in the maternal brain following E15 Poly(I:C) -injection, while no difference was seen for ICAM-1 expression (supplementary Fig. S1, Table 1). To investigate how MIA influenced expression of these pro-inflammatory molecules in the fetal brain, we isolated microglia from E18 fetal cortices using magnetic-activated cell sorting (MACS) system with CD11b magnetic beads (Fig. 1e). Increased IL-6 expression was also observed in fetal microglia, after both E12 and E15 Poly(I:C) -injection (Fig. 1f). The cytokine profile depended on the timing of MIA: IL-1β increased in E15 Poly(I:C) -injected mice but not in E12 Poly(I:C) -injected mice, while TNFα was decreased in E12 Poly(I:C) -injected mice but was unchanged in E15 Poly(I:C) -injected mice. We did not detect any differences in the expression of IL-17a, CD68 and ICAM-1 ( Fig. 1f and Table 1).
Effects of maternal inflammation on microglial morphology. Microglia recognized to exist in different functional phenotypes that include surveillant modes in healthy brain, and reactive phenotypes in the inflamed/injured brain, with these phenotypes having different biochemical and morphological characteristics 23 . The different microglial cytokine expression patterns observed above are consistent with MIA inducing a change in microglial phenotype, as observed previously 16 . Therefore, we next examined if MIA also altered microglial morphology. We used fetal brains from CX 3 CR1-EGFP mice so as to visualize CX 3 CR1-+'ve cells with EGFP fluorescence, and injected dams with either saline or Poly(I:C) at E12 or E15. Fetal brains at E18 were fixed and microglia in cortical layers were quantified in coronal brain slices. As peripheral macrophages populate the brain surface 24 , we distinguished fluorescent cells in the dense superficial (0-50 μm), and more sparse deeper (200-300 μm) layers, and quantified soma areas, cell density, total process length, number of processes, and number of process branches-all typical morphological parameters that can change when adult microglia are activated (Fig. 2a-h) [25][26][27] . None of these parameters changed for neither the E12 or the E15 Poly(I:C) -injected mice (Fig. 2a-h and Table 2) as compared to control mice, for either superficial or deeper layers. We also similarly examined morphological parameters in microglia from brains of mice at 10 days postnatal (P10) (located at 100-200 μm below from the surface), following E12 and E15 Poly(I:C) -injection. Although the absolute microglial density increased with maturation, again there were no differences in any morphological parameters compared to age-matched control mice ( Fig. 2i-n). Finally, we isolated microglia from P10 cortices using MACS and examined gene expression profiles following E12 or E15 Poly(I:C) -injection. We focused on microglia-specific genes related to inflammation [28][29][30][31] and differentiation 32 (Fig. 2o, p and Table 3). There were no differences in the gene expression levels of CD68, ICAM-1, IL-17a or Sall1 following either E12 or E15 Poly(I:C) -injection. In contrast, TMEM119 expression was selectively increased in P10 microglia from E15 Poly(I:C) -injected mice whereas P2Y12R expression was significantly reduced in E12 Poly(I:C) -injected mice (Fig. 2o, p and Table 3). These results suggest that MIA is not associated with microglial morphological changes but can cause some subtle differences in the level of postnatal microglial differentiation, with this difference depending on the timing of MIA.
MIA increases motility of fetal microglial. The function of microglia in vivo depends critically on the motility of their processes, both in surveying brain parenchyma and in the directed migration preceding phagocytosis. Hence, to further characterize how MIA effects microglial phenotypes, we quantified process motility in brain slices acutely isolated from E12 and E15 Poly(I:C) -injected CX 3 CR1-EGFP mice using two-photon microscopy ( Fig. 3a-f). We targeted microglia 200-300 μm below the brain slice surface to minimize any effects of acute inflammation resulting from cells injured at the surface of the brain slices. The velocity of the tips of    www.nature.com/scientificreports/ microglial processes were analyzed by tracking their location every minute and quantifying the total distance moved over time, while tip trajectories were quantified using rose diagrams to represent the frequency of process migration angles, which we called "directionality" (see material and methods) and where a value of 1 equates to total movements being symmetrical. The combination of tip velocity and tip directionality reflect the microglial surveillance patterns in brain parenchyma ( Fig. 3a-d). Maternal injection of Poly(I:C) at both E12 and E15 increased the velocity of microglial tip movements (Fig. 3e, Table 4), while there were no detectable changes in directionality in either E12 or E15 Poly(I:C) -injected mice (Fig. 3f, Table 4). Finally, we examined how the MIA effects compared to effects of acute application of cytokines to control microglia. Microglial movements in control brain slices with and without addition of exogenous IL-6 were compared ( Fig. 3g-j). Acute application of IL-6 had the same effects as was seen with MIA, an increase in microglial process velocity but no change in directionality (Fig. 3i, j, Table 4).
Effects of MIA on microglial process motility are observed postnatally. An important question was whether MIA had sustained effects on microglia after birth. To test this hypothesis, we used two-photon in vivo imaging in CX 3 CR1-EGFP mice, at P10 and at P42 (Fig. 3k-r). For technical reasons, we imaged P10 mice immediately after the cranial window surgery, whereas P42 animals were imaged two weeks after the surgery. Again, we imaged microglia from between 100 μm to 200 μm beneath the brain surface, to minimize any effects of acute surface inflammation. In addition, we quantified the microglial morphology just after the cranial surgery (using in vivo imaging) and compared these parameters to that of in the fixed tissue (where there was no cranial surgery). The density of microglia, soma size, total process length, number of processes and number of branches were again not different in E12 control mice, E12 MIA mice, E15 control mice and E15 MIA mice ( Supplementary Fig. S2 a-l, Table 5). Nor were there differences between morphology in fixed and live tissues, indicating no detectable influence of the craniotomy operation ( Supplementary Fig. S2 a-l, Table 5). However, in contrast to the direction of the change observed in E18 slices, the velocity of microglial tip movements was lower in P10 animals with prior MIA, for both E12 and E15 Poly(I:C) -injections (Fig. 3m, Table 4). In E12 www.nature.com/scientificreports/ Poly(I:C) -injected mice, the directionality of microglia process movements was increased, suggesting a more repetitive or targeted movement pattern. This directionality change was not seen in E15 Poly(I:C) -injected mice (Fig. 3n, Table 4). In addition, the fluorescent variation in primary process of microglia was larger in E12 and E15 MIA mice compared with each control, suggesting some of the microglial properties change in microglia ( Supplementary Fig. S2 m, n). We next examined the effects of both E12 and E15 Poly(I:C) -injection on microglial process motility in late adolescence (P42). No differences in baseline tip velocity or directionality were observed ( Supplementary  Fig. S3, Table 6). We also examined the response to an acute inflammatory challenge in P42 mice, using lipopolysaccharide (LPS) injections (1 mg/kg, i.p.) which is known to transform microglia 33,34 . LPS injection caused a series of expected morphological changes to microglia in E12 saline -injected control mice; cell soma areas were significantly increased, total process length was shortened, and the number of branches was reduced [34][35][36] . The exact same pattern of changes was seen in E12, and E15 Poly(I:C) -injected mice ( Supplementary Fig. S4, Table 7). Acute LPS also caused a similar extent of increase in the microglial tip velocity across all three mice groups (Fig. 3q). However, the effects on directionality differed, in control and E15 Poly(I:C) -injected mice there was no change with LPS, but E12 Poly(I:C) -injected mice showed both higher velocity and directionality after LPS injection (Fig. 3q, r, Table 4), indicating a subtle difference in the characteristics of microglial processes motility following Poly(I:C) -injection at E12.
MIA induces sustained behavioral differences in late adolescent offspring. The sustained effects of MIA on microglial motility were subtle, and hence we wished to validate that E12 and E15 Poly(I:C) -injected mice could induce sustained behavioral deficits. In the open field behavior assay, we found reductions in the percent time spent in the center of the field in mice with E12 Poly(I:C) -injection, while no differences were seen for mice with the E15 Poly(I:C) -injections ( Fig. 4a, Table 8). The total distance traveled in the open field was also less in E12 Poly(I:C) -injected mice, but not in E15 Poly(I:C) -injected mice (Fig. 4b, Table 8). In a social interaction assay, a naive same-sex mouse was placed in the same cage as the unfamiliar mouse for 10 min. Mice sniff and follow the unfamiliar mouse as social behavior 37 , and we quantified these two aspects of social interaction of control and MIA (E12 and E15) mice. The sniffing time was decreased in both E12 and E15 Poly(I:C) -injected mice compared with that of the control mice ( Fig. 4c, Table 8). The time spent following the unfamiliar mouse was significantly decreased in the E12 Poly(I:C) -injected mice, but not in the E15 Poly(I:C) -injected mice (Fig. 4d, Table 8). Finally, we looked for correlation between the behavioral changes in each mouse, and the measurements of microglial directionality in E12 and E15 Poly(I:C) -injected mice both before and after LPS injection (Fig. 4e). In E12 Poly(I:C) -injected mice, the time spent following the unfamiliar mouse directly correlated with changes in directionality (Fig. 4f, bottom right panel, Table 8). There was no relationship between directionality and the percent of time in the center of the open field or with the distance traveled, or with the time spent sniffing (Fig. 4f, Table 8). In E15 Poly(I:C) -injected mice, there was no correlation between social behavior and microglial directionality (Fig. 4g, Table 8). Our result confirms that our MIA models induce sustained behavioral effects, particularly in E12 Poly(I:C) -injected mice, and that the extent of some of these behavioral assays correlates with the extent of changes in microglial directionality.

Discussion
In this study, we aimed to identify whether altered microglial properties may be associated with MIA. We induced MIA at two different times in gestation, E12 corresponding to the second trimester and E15, corresponding to third trimester. Maternal inflammation at different gestational periods results in different predispositions to human disease 38 and to different cognitive deficits in rodents. Offspring from E12 Poly(I:C) -injected rodents, for example, show deficits in latent inhibition and pre-pulse inhibition (PPI) 39 , while E15 Poly(I:C) -injected offspring have normal PPI 40 . The timing of the MIA can also affect microglial responses. E12 Poly(I:C) -injection increased postnatal microglial cytokine and chemokine expression in the offspring, but did not change microglial morphology and density 39,41,42 . In contrast, E15 Poly(I:C) -injection did cause morphological changes in microglia 40 . We first evaluated the cytokine response after our MIA models at mid and late gestation, focusing on expression levels of a series of six pro-inflammatory cytokines (IL-6, IL-1β, IL-17a, TNFα) or cell-surface molecules (CD68 and ICAM-1). These cytokines were selected to include those (IL-6, IL-1β, TNFα) that can be increased in the serum of patients with schizophrenia or autism spectrum disorders [43][44][45][46] . Both MIA models caused an inflammatory response in both maternal organs and in the fetal brain. IL-6 is a key mediator of MIA sequelae 12 and was increased in fetal microglia. Other inflammatory molecules IL-1β and TNFα showed differential responses-IL-1β in fetal microglia increased when MIA occurred at later gestation, while TNFα in fetal microglia decreased only with the earlier MIA. This could be due to a transient change in expression patterns, with IL-1β increasing 3 days after MIA and TNFα decreasing at 6 days after MIA, or it may reflect different responses of E12 and E15 microglia to MIA. A different response of E12 and E15 microglia is consistent with differentiation into a more mature phenotype beginning between these two ages 47,48 . Regardless, our results show that fetal microglia respond, at a biochemical level at least, to maternal inflammation at both mid (E12) and later (E15) gestation.
We next examined morphological changes. Immature microglia display an ameboid morphology during gestation that gradually becomes more ramified over postnatal development 49 . We also observed increases in total process length, and in the number of processes and branches between E18 and P10 (Table 9). However, at both ages, there was no difference in morphology between control and MIA mice, both for E12 and E15 Poly(I:C) -injections. This contrasts to the biochemical phenotypes, and our inability to detect any differences after MIA may partly relate to the already ameboid morphology of microglia at embryonic and early postnatal ages [50][51][52][53]  www.nature.com/scientificreports/ no change has been observed across different studies 42 . We here confirm the qualitative observations that late gestation Poly (I:C) induces no change in microglia morphology, and extend that to mid gestation (E12) MIA 54 . It appears that morphology is a poor measure of changes in microglia phenotype after MIA. In contrast, we show here that microglial process motility is a more sensitive measure of prior MIA. Some aspect of tip motility, either basal tip velocity or directionality, or the response to LPS, was altered by prior MIA. In developing and adult brains, microglial processes are constantly motile, sampling brain parenchyma and contacting soma, synapses, glia and extracellular spaces [55][56][57] . This is an important part of their physiological function-immune surveillance to enable rapid responses to brain damage 58 and also sculpting circuits during development via synapse formation or elimination in an activity-dependent fashion 13,14,[59][60][61][62] . Microglial surveillance may also be important for neural circuit plasticity in the healthy adult brain 63 . Thus, MIA induced changes in microglial tip motility, it may not only effect how synapses are formed and pruned during development but may have ongoing effects on immune-surveillance and neural circuit homeostasis. Just how the altered motility parameters may specifically affect microglial functions is speculative without further follow-up studies. The increase in tip velocity in E18 microglia is at a developmental stage when microglia are proliferating and migrating to the growing cortices and this motility may in part be mediated via trophic signals 48 . An increase in tip velocity may be predicted to enhance migration, although this doesn't seem to manifest as clear increases in cortical densities in latter postnatal brains 42 . At P10, we observed (for the E12 Poly(I:C) -injected mice), decreases in tip velocity and increases in directionality. At P10 microglia are actively interacting with neurons and potentially both forming and pruning synapses, as well as assisting with apoptosis. The decrease tip velocity may suggest brief pauses during brain surveillance, while the increase directionality may suggest more frequent back and forth movements to single targets. Together this may indicate increased microglial-neuron interactions, something that could ultimately impact postnatal synapse numbers and which could be more directly examined with future in vivo imaging. It was interesting that the directionality changes at P10 were only seen after E12 MIA, and the subsequent behavioral phenotypes were also more severe after E12 MIA. At P42, akin to adolescence, changes in basal microglial motility parameters after MIA were not observed, but different LPS-induced motility patterns occurred after prior MIA. LPS increased tip velocity in all mice cohorts, as seen previously 64 , but mice with E12 MIA also showed an increase in directionality after LPS. We don't believe this has been noted before, but may again reflect more repetitive interactions with synaptic elements. Although speculative, this may be associated with improved cognitive symptoms sometimes observed in patients with autism during fever and systemic inflammation 65,66 .
In the offspring of MIA mice, LPS injection can also reduce the extent of social behavioral deficits, and this may be due to activation of Toll-like receptor subtypes (TLR2, TLR4) on microglia and release of cytokines (IL-17) that can increase social behaviors 67 . We also observed impaired social behaviors in offspring of MIA mice, and particularly for E12 Poly(I:C) mice. Furthermore, in these E12 Poly(I:C) mice there did seem to be some weak relationship between directionality and time spent following the unfamiliar mouse (Fig. 4f). Mice with greater directionality after LPS spent longer following the unfamiliar mouse, suggesting an increased microglial tip process directionality may be associated with improved social behaviors. While this association is speculative, it does suggest it worthwhile to further investigate the physiological significance of directionality and its possible link to how inflammation can increase social behaviors in ASD model mice.  www.nature.com/scientificreports/ In conclusion, we report that changes in basal or induced microglial tip motility patterns can be a sensitive measure of neuro-immune alterations in the offspring of mothers with infection, and show here that such motility changes can be observed as early as E18 and are sustained through to adolescence (P42). This strengthens the ideas that changes in microglial functions may contribute to the link between maternal infection and subsequent cognitive defects, and susceptibility to schizophrenia, or developmental disorders in the offspring.

Methods
Animals. All experiment protocols were approved by the Animal Care and Use Committees of Kobe University Graduate School of Medicine and were conducted according to the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The animals in this study were given free access to food and water and housed in a 12 h light/dark cycle. In order to visualize microglial morphology and motility, we used CX 3 CR1-EGFP transgenic mice expressing enhanced green fluorescent protein (EGFP) under the control of the CX 3 CR1 promoter, specific for microglia and monocytes 68 . We procured heterozygous CX 3 CR1-EGFP fetuses after male CX 3 CR1-EGFP mice (homozygous) were crossed with female C57BL/6J mice 46 .  Cranial window surgery. We performed surgery for in vivo imaging on P10 71 and on 4-week-old 72 male mice. In P10 mice, the skull was exposed and cleaned under isoflurane (1%) anesthesia, and a custom-made head plate was firmly attached to the skull with dental cement (G-CEM ONE; GC, Tokyo, Japan). The head plate allowed us to securely attach the mouse to a stainless frame for both cranial window surgery and the subsequent two-photon imaging. After securing the head plate we performed a circular craniotomy (1.6 mm diameter) over the left somatosensory cortex (S1, centered at 1.0 mm posterior from bregma and 2.5 mm lateral from the midline). A 2 mm glass coverslip was placed over the exposed brain to form a cranial window. The edges of the cranial window were sealed with a combination of adhesive glue (Aron Alpha, Konishi, Osaka Japan) and dental adhesive resin cement (Super Bond; Sun Medical, Shiga, Japan). Imaging began immediately after surgery. For 4-week-old mice, following anesthesia with ketamine (74 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.), the skull was  www.nature.com/scientificreports/ exposed and cleaned, and a different and larger (custom-made) head plate was similarly attached. One day later, we performed the circular craniotomy (2 mm diameter) over the left primary sensory cortex (S1, centered at 1.5 mm posterior from bregma and 2.5 mm lateral from the midline) under isoflurane (1%) anesthesia. After the craniotomy, 2% (w/v) agarose L dissolved (Nippon Gene, Tokyo, Japan) in saline was applied, and a glass window comprising two coverslips (2 & 4.5 mm each; Matsunami Glass, Osaka, Japan) were placed over the brain surface with ultraviolet curable adhesive (NOR-61, Norland, New Jersey, USA). The edges of the cranial window were sealed with a combination of dental cement and dental adhesive resin cement (Super Bond; Sun Medical, Shiga, Japan). These 4-week mice were returned to their home cages and imaging experiments were performed two weeks later, at P42.
Acute slice imaging. To obtain E18 cortical slices, male and female fetal brains were quickly isolated and sliced coronally (500 μm) with a vibratome in cold, oxygenated cutting solution (4 °C, bubbled with 95% O 2 -5% CO 2 ). Immediately after sectioning, the slices were warmed in oxygenated recovery solution (  www.nature.com/scientificreports/  Image analysis. All images were analyzed using MATLAB software (MathWorks, Natick, MA) and ImageJ (National Institutes of Health, Bethesda, MD). Movies were corrected for focal plane displacements using both TurboReg 46 . The E18 microglia in the superficial and deep layers of the somatosensory cortex (0-50 μm and To quantify E18 and P10 microglial morphology and P42 microglial morphology before and after LPS administration, the reconstructed Z-stacked confocal images and the reconstructed 20 μm Z-stacked two-photon images were analyzed by using the ImageJ plug-in Simple Neurite Tracer and the segmented line tool in ImageJ respectively. In addition, to assess microglial motility based on the reconstructed Z-stacked two-photon movies (time interval 1 min), we used the ImageJ plug-in Manual Tracking. Using chemotaxis and Migration Tool 2.0 (IBIDI), we measured the velocity and directionality of microglial processes by tracking the process tip every 1minute 74,75 . The rose diagram is a circular histogram which displays the direction or angle that the microglial tip moves, relative to its starting point, and the frequency of each direction. Each direction angle was binned, with 9 bins per quadrant (10° each), and the area of each bin is the number of tip movements in that direction. To compare the directionality statistically, we calculated the coefficient of variation of each bin (every 10 degrees value). An averaged value of 1 for the total microglial tip movements represent symmetrical movements.
Behavioral tests. Animals were allowed to acclimatize to the behavioral testing room for at least 1 h before testing. 6-week-old juvenile male mice were placed in a 50 × 50 cm white plexiglas box open field and movements were recorded by a video and subsequently analyzed using Duo Mouse software (National Institute of Genetics, Shizuoka, Japan) and MATLAB (MathWorks, Natick, MA) software packages. Mice were tracked by camera (C270N HD WEBCAM, Logicool, Tokyo, Japan) and the total movement over 10 min in the open field quantified. The center of the open field was defined as a central 12.5 × 12.5 square (156.25 cm 2 , 6.25% of the total area) and the total time when mice were tracked as being inside this region quantified. For the socialization assays, a novel 6-week-old unfamiliar male was introduced into the test subject's arena into the diagonally opposite region for 10 min and the resulting socialization parameters were scored subjectively as one mouse walking closely behind the other, keeping pace, while nose-to-nose sniffing was defined as one mouse closely approaching the other and sniffing its nose. Non-social parameters were scored as arena exploration (walking around the arena, sniffing the walls or floor).    www.nature.com/scientificreports/

Data availability
All data were analyzed using GraphPad Prism 8 statistical software (GraphPad Software Inc., La Jolla, CA). All data are presented as means ± SD. Unpaired t-test, paired t-test, one-way ANOVA and Pearson's correlation test were used to test for statistical significance. Table 9. Comparison of morphological properties between E18 and P10 microglia from control (E12 and E15 saline -injected) and MIA (E12 and E15 Poly(I:C) -injected) mice.