Bacterial TLR2/6 Ligands Block Ciliogenesis, Derepress Hedgehog Signaling, and Expand the Neocortex

ABSTRACT Microbial components have a range of direct effects on the fetal brain. However, little is known about the cellular targets and molecular mechanisms that mediate these effects. Neural progenitor cells (NPCs) control the size and architecture of the brain and understanding the mechanisms regulating NPCs is crucial to understanding brain developmental disorders. We identify ventricular radial glia (vRG), the primary NPC, as the target of bacterial cell wall (BCW) generated during the antibiotic treatment of maternal pneumonia. BCW enhanced proliferative potential of vRGs by shortening the cell cycle and increasing self-renewal. Expanded vRGs propagated to increase neuronal output in all cortical layers. Remarkably, Toll-like receptor 2 (TLR2), which recognizes BCW, localized at the base of primary cilia in vRGs and the BCW-TLR2 interaction suppressed ciliogenesis leading to derepression of Hedgehog (HH) signaling and expansion of vRGs. We also show that TLR6 is an essential partner of TLR2 in this process. Surprisingly, TLR6 alone was required to set the number of cortical neurons under healthy conditions. These findings suggest that an endogenous signal from TLRs suppresses cortical expansion during normal development of the neocortex and that BCW antagonizes that signal through the TLR2/cilia/HH signaling axis changing brain structure and function.

ABSTRACT Microbial components have a range of direct effects on the fetal brain. However, little is known about the cellular targets and molecular mechanisms that mediate these effects. Neural progenitor cells (NPCs) control the size and architecture of the brain and understanding the mechanisms regulating NPCs is crucial to understanding brain developmental disorders. We identify ventricular radial glia (vRG), the primary NPC, as the target of bacterial cell wall (BCW) generated during the antibiotic treatment of maternal pneumonia. BCW enhanced proliferative potential of vRGs by shortening the cell cycle and increasing selfrenewal. Expanded vRGs propagated to increase neuronal output in all cortical layers. Remarkably, Toll-like receptor 2 (TLR2), which recognizes BCW, localized at the base of primary cilia in vRGs and the BCW-TLR2 interaction suppressed ciliogenesis leading to derepression of Hedgehog (HH) signaling and expansion of vRGs. We also show that TLR6 is an essential partner of TLR2 in this process. Surprisingly, TLR6 alone was required to set the number of cortical neurons under healthy conditions. These findings suggest that an endogenous signal from TLRs suppresses cortical expansion during normal development of the neocortex and that BCW antagonizes that signal through the TLR2/cilia/HH signaling axis changing brain structure and function. IMPORTANCE Fetal brain development in early gestation can be impacted by transplacental infection, altered metabolites from the maternal microbiome, or maternal immune activation. It is less well understood how maternal microbial subcomponents that cross the placenta, such as bacterial cell wall (BCW), directly interact with fetal neural progenitors and neurons and affect development. This scenario plays out in the clinic when BCW debris released during antibiotic therapy of maternal infection traffics to the fetal brain. This study identifies the direct interaction of BCW with TLR2/6 present on the primary cilium, the signaling hub on fetal neural progenitor cells (NPCs). NPCs control the size and architecture of the brain and understanding the mechanisms regulating NPCs is crucial to understanding brain developmental disorders. Within a window of vulnerability before the appearance of fetal immune cells, the BCW-TLR2/6 interaction results in the inhibition of ciliogenesis, derepression of Sonic Hedgehog signaling, excess proliferation of neural progenitors, and abnormal cortical architecture. In the first example of TLR signaling linked to Sonic Hedgehog, BCW/TLR2/6 appears to act during fetal brain morphogenesis to play a role in setting the total cell number in the neocortex.
M icrobial components have a range of effects on the brain, the breadth of which is incompletely understood. For instance, bacterial metabolites released from the microbiome are believed to program responses in the brain (1)(2)(3). Bacterial structural components such as bacterial cell wall (BCW), a polymerized carbohydrate and amino acid network (4), both stimulate brain inflammation during meningitis (5) and direct noninflammatory responses when released from the gut microbiome to circulate in serum (6)(7)(8)(9)(10). Such signals are believed to contribute to neurodegeneration and behavioral abnormalities (11,12). In the context of the maternal/fetal interface, bacterial metabolites from the microbiome can affect fetal development (1). However, little is known about how bacterial components released by antibacterial treatment of infection during pregnancy affect fetal brain development despite ample evidence that prenatal maternal bacterial infections increase risks for neurocognitive deficits (13)(14)(15). One proposed mechanism is the maternal immune activation (MIA), a syndrome whereby inflammation in the mother incites cytokines that relay an inflammatory signal across the placenta to immune cells in the fetus (16,17). However, this does not address events early in gestation before the fetal immune response is mature or the possibility of a direct interaction of microbial products with Toll-like receptors (TLRs) in fetal neural progenitor cells (NPCs). Along these lines, BCW, the major component of bacterial lytic debris, is recognized by TLR2 (5) present on NPCs even in the early stages of brain development when immune cells are not yet formed (18,19). We have previously shown that purified BCW can cross the placenta and traffic to the fetal brain, leading to an abnormal increase in cortical neuronal number in a TLR2-dependent fashion (20). Thus, microbial ligands appear to induce events affecting early fetal brain development independent of classical inflammation. Here, we investigate the cellular targets and molecular mechanisms by which BCW expands cortical neuronal numbers.
The primary NPCs are ventricular radial glia (vRGs; also called apical radial glia) that reside in the ventricular zone (VZ) lining the brain ventricle (21)(22)(23)(24). To populate the fetal cortex, vRGs can divide symmetrically to increase their pool or asymmetrically to produce one vRG (preserving their pool) and one progeny that differentiates into a neuron, an intermediate progenitor cell (IPC), or an outer radial glia (oRG; also called basal radial glia). IPCs and oRGs form the subventricular zone (SVZ), where they divide to produce neurons (22,(25)(26)(27)(28)(29). Young neurons migrate along the radial fibers of RGs to form the six-layered neocortex in an inside-out manner: late-born neurons migrate past earlyborn neurons to form a more superficial layer closer to the brain surface. Therefore, NPCs control the size and architecture of the brain and understanding the mechanisms regulating NPCs is crucial to understanding brain developmental disorders.
Using a model of pneumococcal pneumonia in pregnant mice followed by antibiotic treatment that floods the fetus with BCW while curing the mother, we describe that vRGs are the cellular target of BCW in the fetal brain. In a restricted window of vulnerability, BCW enhanced proliferative potential of vRGs, resulting in an expanded NPC pool that propagated through the entire period of neurogenesis, increasing neurons in all layers in the neocortex. Remarkably, TLR2 localized at the base of cilia in vRGs and the BCW-TLR2 interaction suppressed ciliogenesis, leading to derepression of Hedgehog (HH) signaling. This noninflammatory, noncanonical TLR2 signaling axis through HH reshaped brain architecture. We also show that TLR6 is an essential partner of TLR2 in this process. Further, TLR6 alone was required to set the normal number of cortical neurons, suggesting there exists an endogenous ligand that is antagonized by BCW.

RESULTS
BCW exposure during a restricted developmental window increases neurons in all cortical layers. Embryonic day 10 (E10) marks the initiation of neurogenesis in the mouse neocortex (23,24,30). We previously showed that injecting mothers with purified BCW at E10, but not E15, increases the number of cortical neurons at E16 and P10 (20). To align this observation with a clinically relevant scenario and further narrow the window of vulnerability before E15, we initiated maternal pneumonia by challenge with Streptococcus pneumoniae at E10 or E11, followed at 24 h by treatment with ampicillin (Amp), which generates a rapid release of BCW into the maternal bloodstream. BCW crosses the placenta and is detectable in the fetal brain from 6 h onward (i.e., at E11 or E12, respectively) ( Fig. 1a) (20). Treatment of the phosphate-buffered saline (PBS) group with Amp controlled for any antibiotic effects on the endogenous maternal microbiome. An increase in the number of fetal cortical neurons was seen after E10, but not E11, challenge, narrowing the period of vulnerability to direct effects of BCW on fetal neurons to before E11 ( Fig. 1a and b).
To define which layers of excitatory neurons were increased by BCW challenge, we labeled neurons with specific layer markers: TBR1 for layer VI, which arises early in development; CTIP2 (also known as BCL11B) for layer V, which arises next; and SATB2 for layers II to IV, which arise last in development (31)(32)(33). Bacterial challenge did not affect the layering of the cortex (Fig. 1c). However, the E10 challenge, but not E11, significantly increased the number of neurons in each layer ( Fig. 1c; see also Fig. S1 in the supplemental material) even though most SATB2 1 II-IV neurons arise at E14 to E17 and most TBR1 1 layer VI neurons arise at E11 to E13. The abnormal cortical architecture persisted after birth ( Fig. 1b; see also Fig. S1b). Since BCW is cleared from the fetal brain within 2 to 3 days (20), these findings suggest that an encounter to bacterial products at the onset of neurogenesis establishes a persistent dysregulation long after the exposure. Several lines of evidence distinguished the fetal NPC response from MIA. Expansion of the neocortex did not correlate with maternal serum cytokine levels since these were elevated after both E10 and E11 challenges (see Fig. S2a), despite fetal vulnerability only after E10 challenge. The response also occurred before substantial appearance of fetal microglia as indicated by equivalent, low levels of the microglial activation marker IBA1 at both E10 and E11 (see Fig. S2b). These features argued against MIA.
It has been suggested that fetal exposure to microbial products from maternal infection is linked to permanent disorders of behavior or neuropsychiatric diseases (13)(14)(15). Consistent with this, bacterium-challenged mice showed spatial recognition defects, repetitive behaviors, and sociability defects (see Fig. S3), indicating that abnormal brain development may lead to abnormal architecture and disordered postnatal behavior in humans who experienced maternal bacterial infection during early fetal development.
Rapid initial expansion of vRGs by shortening the cell cycle time. vRGs are the primary NPCs that produce neurons, either directly or through oRGs and IPCs. We hypothesized that an initial expansion of vRGs followed by increased production of oRGs and IPCs could create a progressive wave of increased neurons into all layers of the cortical plate. Consistent with our hypothesis, vRGs (PAX6 1 TBR2cells in the VZ) but not IPCs (TBR2 1 cells) were greatly expanded by 1 day after antibiotic treatment with E10, but not E11, challenge ( Fig. 2a and b), indicating that vRGs amplified themselves as primary targets of BCW.
We next investigated how BCW caused vRGs to expand. The division mode of NPCs is closely associated with cell cycle kinetics (34). In particular, vRGs undergoing selfamplifying divisions show a shorter cell cycle than vRGs producing differentiating progenies (35). Thus, we investigated if the expanded vRGs were characterized by a shortened cell cycle by using a double thymidine analogue labeling method with successive injections of 5-chloro-29-deoxyuridine (CldU) at 2 h (h) and 5-ethynyl-29-deoxyuridine (EdU) at 0.5 h before harvest (36,37). EdU 1 PAX6 1 TBR2cells represent vRGs in S phase at the time of harvest, whereas CldU 1 EdUcells represent vRGs that have left S phase and entered G 2 phase during the 1.5 h between CldU and EdU injections (Fig. 2c). The nuclei of vRGs show interkinetic migration (37): nuclei undergo S phase at the upper part of the VZ, move down to the ventricular surface during G 2 phase, and divide at the ventricular surface. Consistent with interkinetic nuclear migration of vRGs, the nuclei of vRGs in S phase (EdU 1 ) were concentrated in the upper part of the VZ (Fig. 2c). Remarkably, many CldU 1 EdU -vRG nuclei were present in the lower VZ in BCW-challenged but not control E10 embryos, suggesting that more cells exited S phase and entered G 2 phase during the 1.5 h interval in E10-challenged than control embryos (Fig. 2c). Accordingly, E10 challenge, but not E11, dramatically increased the number of CldU 1 EdUcells compared to control (Fig. 2d, left) indicating an increased exit from S phase. However, the total number of vRGs in S phase (EdU 1 cells) did not change (Fig. 2d, right), indicating that as exit from S phase increased so did the number of vRGs entering S phase. Although the number of vRGs increased after E10 challenge (Fig. 1b), the number of vRGs in S phase (EdU 1 vRGs) did not increase, indicating the proportion of vRGs in S phase was decreased in E10 challenged embryos. Since the length of a cell cycle is proportional to the number of the cells in that cell cycle these findings are indicative of a shortened S phase. Indeed, challenge at E10, but not E11, significantly shortened the lengths of S phase and the cell cycle ( Initial expansion of vRGs leads to expanded NPCs in later stages. The increase in neuron number in all cortical layers suggests that the early expansion of vRGs was maintained and propagated into downstream cell lineages throughout the neurogenic period. To test this possibility, we quantified the number of vRGs, oRGs, and IPCs (as defined in Fig. 3) at E16, a late stage of cortical neurogenesis. All 3 cell types were increased solely in the E10 challenged group, consistent with a time-restricted vulnerability of vRGs to BCW ( Fig. 3a and b). The increase of oRGs and IPCs indicates that the initial expansion of vRGs resulted in the subsequent expansion of downstream NPCs. These results suggest that E10 challenge expanded all NPC types by E16.
To understand how expanded vRGs increased NPCs in later developmental stages, we labeled proliferating NPCs by pulses of EdU at E12, E13, or E14 and quantified EdU 1 FIG 2 S. pneumoniae challenge expands vRGs and shortens vRG cell cycle. (a and b) S. pneumoniae challenge at E10 but not E11 expands vRGs. We challenged dams at E10 or E11 (blue, PBS; red, S. pneumoniae), treated them with Amp 24 h later, and quantified RGs (PAX6 1 TBR2cells; green) and IPCs (PAX6 -TBR2 1 cells; purple) 48 h later (E12 or E13, respectively). Two sections were quantified from each embryo, and bars represent the means 6 the SEM of the number of embryos indicated in each bar. (c to e) Cell cycle kinetics of RGs in the 48 h after challenge at E10 or E11. Blue, Amp; red, S. pneumoniae 1 Amp. We injected CldU and EdU at 2 and 0.5 h before harvest, respectively. Note the many CldU 1 EdUcells in the lower VZ (c) in embryos that were challenged with bacteria at E10 and treated with Amp. The graphs show RGs that had left S phase (CldU 1 EdU -; panel d, left), total RGs in S phase (EdU 1 ; panel d, right), RGs in cell cycle (Ki67 1 ; panel e, left), and the relative lengths of S phase and the cell cycle of RGs (panel e, middle and right; control Amp alone set at 100%). The data are shown as means 6 the SEM. P values were determined by unpaired two-tailed t tests. Spn, S. pneumoniae. (Continued on next page) Maternal Bacteria Redirect Fetal Neurodevelopment mBio cells in the VZ and SVZ 24 h later ( Fig. 3c and d). EdU injection labeled more RGs and IPCs at all stages in the challenged group than in control with the highest increase of EdU 1 vRGs at E12 while EdU 1 IPCs showed strong increases at E13 and E14 injection times (Fig. 3e). These results suggest that more vRGs expanded at E12 in the challenged group than in control and that increased vRGs subsequently produced more IPCs at E13 and E14. This supports a model that E10 challenge resulted in a wave of expansion of both RGs and IPCs over time in their expected developmental sequence, with vRGs impacted first, followed by IPCs. Next, we determined whether the expanded NPCs were sustained into later stages through self-renewal. We injected dams with EdU at E13 and with BrdU at 1.5 h before harvest at E14 (Fig. 3f) to identify progenitors that proliferated at E13 (EdU 1 ), remained as progenitors, and proliferated again at E14 (BrdU 1 ). The proportion of self-renewed progenitors (EdU 1 BrdU 1 ) was significantly increased in E10 challenged group. Consistently, more EdU 1 cells were positive for Ki67, a proliferation marker. These findings showed that the initial expansion of vRGs at E12 propagated to the expansion of all three NPC types at later stages in E10 challenged embryos through increased proliferative capacity and self-renewal.
BCW components require TLR2/6 to expand neurons. BCW interacts with TLR2 (5). TLR2 is expressed by both NPCs and embryonic neurons and TLR2 expression is maintained at a similar level during the challenge period (18,19). The present study identified NPCs (vRGs) as the specific target of BCW with subsequent impact on all types of cortical neurons. TLR2 functions as a heterodimer with either TLR1 or TLR6. The impact of either TLR2 signaling combination in fetal brain development or NPC proliferation is unknown. To identify which TLR2 heterodimer mediates BCW effects in NPCs, we challenged pregnant mice bearing wild-type (WT), Tlr1 2/2 , Tlr2 2/2 , or Tlr6 2/2 embryos with S. pneumoniae or PBS at E10, treated with Amp, and examined the brains at E16. Neuronal layering was normal in all mutant embryos (Fig. 4a). However, S. pneumoniae challenge increased the number of cells in the cortical plate in WT and Tlr1 2/2 embryos but not in Tlr2 2/2 and Tlr6 2/2 embryos (Fig. 4b), demonstrating that the TLR2/6 axis is required for BCW to increase cortical neurogenesis. This result could be ascribed to the genotype of the fetus, and not the mother, as shown by the comparison of Tlr2 2/2 versus Tlr2 1/fetuses carried by Tlr2 1/mothers. Neurons increased in Tlr2 1/but not Tlr2 2/2 fetal brains (Fig. 4b, gray bars).
Loss of TLR signaling affects normal brain architecture. We next tested whether loss of TLR signaling affected brain structure in unchallenged controls. Remarkably, the number of cortical neurons was significantly greater in the untreated Tlr6 2/2 embryos compared with WT (Fig. 4b). TUNEL staining showed few, if any, apoptotic cells, in both WT and Tlr6 2/2 untreated brain sections, ruling out that less apoptosis in Tlr6 2/2 brains accounted for increased neurons (see Fig. S4). On the other hand, more untreated Tlr6 2/2 NPCs cultured in vitro expressed a proliferation marker, Ki67, than untreated WT NPCs (Fig. 4c), suggesting that augmented NPC proliferation underlies increased production of neurons. Further work will be necessary to reveal how endogenous TLR6 signaling limits the number of cortical neurons in healthy WT animals. Nonetheless, the increase in the baseline neuronal number in unchallenged Tlr6 2/2 mice establishes the requirement for an unknown endogenous TLR6 axis signal to set the number of neurons during normal neurodevelopment and that TLR ligands, such as BCW, modulate that signal. Together, our results indicate that TLR6 suppresses cortical neuronal expansion under healthy conditions and that pathogenic TLR ligands reverse such suppression.
TLR2/6 signaling derepresses Hedgehog signaling. To understand the molecular mechanism for BCW-induced expansion of vRGs, we used spatial transcriptomics to compare gene expression in E12 brains from dams challenged at E10 with S. pneumoniae or  ) brains. (f) NPC self-renewal was assessed by labeling with EdU at E13 (green arrow), followed by a BrdU pulse (pink arrow) at 1.5 h before harvest at E14. Self-renewed cells (EdU 1 BrdU 1 ) were expressed as the percentage of total EdU 1 cells (left). The samples were also stained for the proliferation marker Ki67 as a function of EdU 1 (right). P value was determined by an unpaired two-tailed t test. Spn, S. pneumoniae.
PBS and treated at E11 with Amp. Comparisons focused on the VZ, where vRGs reside and are marked by expression of PAX6. Several signaling pathways, including HH, mTOR, and PI3K/AKT, were significantly enriched within the S. pneumoniae-challenged VZ, while tumor necrosis factor and the inflammatory response was not ( Fig. 5a; see also Fig. S5 and Table S1). Enrichment of HH and PI3K/AKT/MTORC1 signaling was notable because both induce the expansion of NPCs and the growth of the neocortex (38)(39)(40)(41)(42), which were the consequences of BCW challenge at E10. Activation of PI3K/AKT/MTORC1 signaling in vitro and in the brains of challenged mice depended on TLR2 (see Fig. S6). Reverse transcription-quantitative PCR (RT-qPCR) confirmed that multiple genes known to be upregulated by activation of HH signaling in the embryonic neocortex (43,44) were upregulated upon BCW challenge but not in Tlr2 2/2 dams (Fig. 5b). The levels of Gli1, the well-known readout of strong HH signaling were not upregulated. This result was in fact FIG 4 S. pneumoniae challenge requires TLR2/6 to expand cortical neurons. (a and b) Effect of Tlr1, -2, or -6 deficiency on cortical expansion. Dams deficient in Tlr1, -2, or -6 were challenged with S. pneumoniae or PBS at E10, treated with Amp at E11, and harvested at E16. In addition, heterozygous Tlr2 1/parents were crossed, mothers were challenged as described above and embryonic brains were analyzed in accordance with fetal genotype (gray bars). TBR1, CTIP2, and SATB2 staining shows normal layering. Scale bar, 100 mm. consistent with the previous finding that the loss of GLI3, which suppresses the expression of HH target genes, increases the expression of HH target genes but not Gli1 in the developing neocortex at this stage in the neocortex (44). This suggests that Gli1 expression may require much stronger activation of HH signaling than that achieved by GLI3 loss or S. pneumoniae. On the other hand, Fgf15, a target of HH signaling in the developing cortex, which is upregulated by GLI3 loss and expands vRGs by shortening their cell cycle (44), was upregulated in the E10 challenged group and remained upregulated between E12 and E16 (Fig. 5c).
To understand how BCW enhanced HH signaling, we examined GLI3, a transcription factor whose activity is directly regulated by HH signaling (44,45). In the absence of active HH signaling, full-length GLI3 (GLI3FL) is proteolytically cleaved to become GLI3 repressor (GLI3R). GLI3R suppresses the expression of HH target genes in vRGs and its loss strongly increases the expression of HH target genes, including Fgf15 (44). Consistent with the enrichment of HH signaling signatures, including increased Fgf15, the GLI3FL activator levels relative to GLI3R levels were significantly increased in E10-challenged fetal brains in a TLR2 and TLR6 dependent manner (Fig. 5d and e). Remarkably, treatment with a GLI inhibitor (GANT61), as well as a TLR2/6 inhibitor (Git27), significantly blocked BCW-induced proliferation of NPCs in vitro (Fig. 5f). These findings collectively suggest that BCW acting through TLR2/6 expands NPCs at least in part by increasing the ratio of GLI3 activator to repressor, resulting in the derepression of HH target gene expression.
BCW suppresses ciliogenesis via TLR2/6. GLI3FL must shuttle through primary cilia to be proteolytically cleaved to become GLI3R (46,47). Accordingly, the loss of cilia in NPCs leads to the loss of GLI3R and derepression of HH target genes, including Fgf15 (44). Remarkably, BCW challenge at E10 significantly decreased the number of vRGs with primary cilia (Fig. 6a). Moreover, this effect was lost in Tlr2 2/2 or Tlr6 2/2 mice (Fig. 6b). TLR2 localized to primary cilia in vRGs (Fig. 6c). To further test whether BCW decreased cilia through TLR2/6, we treated NPC cultures with BCW in the presence or absence of TLR2/6 inhibitor Git27 (Fig. 6d). Consistent with in vivo observations, BCW decreased the number of NPCs with cilia. This decrease was blocked by the TLR2/6 inhibitor Git27. Moreover, BCW failed to decrease cilia in NPCs lacking TLR2. Also consistent with in vivo observations, BCW altered cilia and proliferation of NPCs cultured from E11 but not E12 embryos (Fig. 6e). Together, these findings suggest that BCW-TLR2/6 signaling from cilia suppresses the formation of cilia and GLI3R, leading to the derepression of HH target genes and the expansion of NPCs.

DISCUSSION
The maternal fetal interface monitors a constant, undercurrent dialog between the fetus and the mother, including responses to microbial metabolites released from the maternal microbiota or inflammatory molecules that characterize MIA (1-3). Here, we investigate the cellular targets and molecular mechanisms of effects of microbial products that cross the placenta and interact directly with fetal cortical NPCs. MIA is known to affect fetal brain development (16,17). Several lines of evidence argue against this as a mechanism for the NPC expansion observed by BCW challenge. NPC expansion occurred at E10 but not E11 challenge despite similar maternal serum cytokine profiles at both time points. Further, the fetal TLR genotype determined the NPC expansion response regardless of the TLR genotype of the mother, which may affect MIA. Finally, FIG 5 Legend (Continued) embryos exposed to Amp (blue) or S. pneumoniae 1 Amp challenge (red). (c, left) Violin plot of Fgf15 expression obtained from control (Amp) and treated (S. pneumoniae 1 Amp) E12 VZ as assessed by spatial transcriptomics. (Right) Fgf15 expression levels compared by RT-qPCR. Challenge occurred at E10 and samples were harvested at the indicated day. Values represent three to four dams and three to five embryo lysates per dam. The data are shown as means 6 the SEM. P values were determined by an unpaired two-tailed t test. (d and e) WT, Tlr2 2/2 , and Tlr6 2/2 dams were challenged at the indicated day with PBS (blue) or S. pneumoniae (red) and treated with Amp 24 h later, and brains were harvested after another 24 h, as indicated. Brain lysates were assessed by Western blotting for full-length GLI3 activator form (GLI3FL) and GLI3R (processed repressor) (d) (Fig. S8). (e) Ratio of protein levels of GLI3FL to GLI3R. We used two to three dams per condition with two to four embryos/dam. P values for all panels were determined by unpaired, twotailed t tests. Gels were spliced for labeling purposes and to remove unwanted unnecessary background. (f) NPCs harvested at E11 were plated in vitro and pretreated with or without inhibitors of GLI (GANT61, 5 mM) or TLR2/6 (Git27, 10 mg/mL) for 2 h, followed by BCW (MOI of 0.1) in the presence of inhibitor. Proliferation was assessed by Ki67 1 staining. Spn, S. pneumoniae.  The present study examined the consequences of the release of bacterial lytic products in the maternal bloodstream in a context relevant to clinical maternal/fetal care as initiated by b-lactam antibiotic therapy to cure pregnant mice with pneumococcal pneumonia. In this model, a burst of BCW release occurs over 6 to 24 h after antibiotic administration as bacteria rapidly die, creating a pulse of bacterial debris that accumulates in the fetal brain. S. pneumoniae challenge at E10 (but not E11), followed by Amp treatment to create BCW, increased the number of all excitatory neuronal types of the neocortex, indicating a wide impact of bacterial components released in the mother on fetal brain architecture. These structural changes were accompanied by cognitive and behavioral abnormalities indicating that bacterial infection and treatment in the early stages of pregnancy could have long-lasting effects on brain structure and function.

Maternal Bacteria Redirect Fetal Neurodevelopment mBio
BCW increased the number of excitatory neurons in all cortical layers although they arise at different times spanning E11.5 to E17.5 (48). An effect initiated at E10, but spanning many days thereafter, suggested that the consequences of BCW challenge originated in the pool of NPCs and were reflected in all cells born afterward. Indeed, all NPCs (vRGs, oRGs, and IPCs) underwent expansion. We suggest that vRGs are the primary target of BCW because only PAX6 1 TBR2 -NPCs were expanded initially and immediately after the challenge. It is unclear why only vRGs respond to BCW. BCW might be present in cerebrospinal fluid, to which only vRGs are exposed. vRG expansion later progressed as a wave in time through the generation of later cell types.
Before producing neurons, vRGs divide symmetrically to amplify themselves. As they switch to divide asymmetrically to produce neurons, the cell cycle lengthens (35,49). Our data suggest that signals from bacterial debris such as BCW expand the initial pool of vRGs at the early stage of cortical development by delaying the lengthening of their cell cycle and increasing self-amplifying divisions. This initial expansion of vRGs was propagated to the expanded pool of IPCs and oRGs that were also maintained by increased self-renewal. It is notable that BCW expanded both IPCs and oRGs, a feature thought to underlie the development and evolution of the large and folded neocortex in higher mammals, including humans (24,(50)(51)(52)(53). This emphasizes that bacterial components could have profound effects on features of human cortical development.
Activation of TLRs in the postnatal brain by pathogens is well-known to induce an innate immune response. During bacterial meningitis, BCW interacts with TLR2/1 to initiate inflammation and neuronal damage (5,54,55). TLRs are expressed in the fetus as early as E10; however, it has remained unclear whether BCW/TLRs might act as morphogens in mammalian neurodevelopment independent of inflammation, as seen in Drosophila (56,57). Our data show that activation of TLR2 and TLR6, but not TLR1, increased fetal neurogenesis and altered brain morphology in the absence of inflammation. Thus, by switching partners in the heterodimer at different time points of development, TLR2 participates in either fetal neurogenesis (TLR2/6) or postnatal inflammation (TLR2/1) in mammals.
In addition to the participation of TLRs in responses to BCW in challenged fetuses, unchallenged Tlr6 2/2 mice showed a significantly increased neuronal number compared to unchallenged WT mice. Fittingly, TLR6 loss increased the number of proliferating NPCs. This suggests a requirement for an endogenous TLR6 axis signal to limit the number of neurons during normal neurodevelopment and that TLR ligands, such as BCW, modulate that signal. Together, our results might indicate that TLR2 and TLR6 Maternal Bacteria Redirect Fetal Neurodevelopment mBio suppress cortical neuronal expansion under healthy conditions and that pathogenic TLR ligands reverse such suppression. Further studies will be necessary to test this hypothesis. Spatial transcriptomics of control versus challenged fetal brains revealed an enrichment in genes associated with several signaling pathways. PI3K/AKT signaling was enhanced consistent with our previous findings that it promotes NPC proliferation downstream of TLR2 in vitro (20). In contrast, HH signaling has not been previously connected to TLRs. Our data suggest that BCW activated HH signaling indirectly via suppressing cilia and GLI3R formation. Supporting this, a previous study showed that TLR2 promotes the disassembly of cilia in cultured cells (58). We revealed that TLR2 localized to cilia in vRGs and that BCW challenge decreased the number of vRGs with cilia in a TLR2/6-dependent manner in vitro and in vivo. Thus, TLR2 appears to signal to disassemble the primary cilium. Since vRGs form the wall of the cerebral ventricle and project cilia into the cerebrospinal fluid, the association of TLRs with cilia may promote their function as sentinels. Moreover, as shown for the genetic ablation of cilia in vRGs at early stages of corticogenesis (44), the decrease in ciliated vRGs in BCW-challenged embryos was accompanied by increased activator to repressor ratio of GLI3 and derepression of HH target genes, including Fgf15, which shortens the cell cycle and promotes proliferation of vRGs. Importantly, genetic removal of cilia or GLI3 by E12.5 but not by E14.5 expands NPCs and the cortex (43,44,59). Notably, E10 to E12 is a critical developmental window when vRGs switch from symmetric self-amplifying divisions to asymmetric neurogenic divisions (60). This switch in division modes is critical to determining the number of neurons and the size of the cortex because it determines the number of founding vRGs that subsequently generate all the excitatory neurons in the neocortex directly or through secondary progenitors (IPCs and oRGs). S. pneumoniae challenge at E10 and ampicillin treatment at E11 released BCW during E11 to E12, exactly overlapping with the time when vRGs switch division modes. Importantly, challenge at E10 induced changes similar to the loss of GLI3 by E12, while challenge at E11 did not because the release of BCW occurred after the critical period when vRGs switch division modes. These findings highlight the importance of developmental timing in NPC responses to regulatory factors and provide a possible explanation for the timedependent effects of S. pneumoniae challenges.
We, therefore, propose a novel neurodevelopmental pathway that is driven by direct targeting of NPCs by BCW and connects TLRs to morphogenesis. TLR2, newly found at the base of primary cilia of vRGs, modulates the HH/GLI3/FGF15 pathway to expand NPCs leading to permanent changes in the trajectory of development of neocortical architecture and neurocognitive function. This process is independent of TLR2/1-induced inflammation and neuronal death that characterizes BCW-induced responses from the innate immune system after birth. Importantly, in Tlr6 2/2 control embryos without BCW challenge, the loss of TLR6 increased the number of cortical neurons, suggesting that an endogenous TLR signal regulates normal neurodevelopment in the embryo and that BCW released from maternal infection antagonizes that signal.
An important next question is whether BCW affects the development of interneurons and the ratio of inhibitory interneurons to excitatory neurons. Does BCW affect the proliferation of interneuron progenitors in the ventral telencephalon? If BCW selectively expands excitatory neurons, does that result in a lower-than-normal density of interneurons or an increased survival of interneurons and normal density of interneurons in response to expanded excitatory neurons? Of note, HH signaling plays critical roles in specifying and maintaining molecularly distinct NPC types that produce diverse classes of interneurons (61,62). Does BCW alter interneuron fates by affecting HH signaling in interneuron progenitors? A lower-than-normal density of interneurons, perhaps with differential effects on interneuron classes, might result in excitatory/inhibitory imbalance contributing to neurocognitive deficits associated with prenatal maternal bacterial infections in humans and in our mouse model.
Timed pregnancies were dated as follows: mating pairs were together for 24 h and then separated. Dams were checked for vaginal plugs at 48 h and were palpated every other day. When positive by palpation (ca. E6 to E8), mice were anesthetized and scanned using 40 MHz center frequency transducer (MX550D) following each uterine horn to count the number of embryos.
Dams were challenged intratracheally at E10 or E11 with either PBS (controls) or 10 6 CFU S. pneumoniae strain TIGR4X in PBS. Twenty-four hours after challenge, blood titers were obtained to ensure consistent bacterial load (mean, 10 5 CFU/mL; for continued inclusion in the study challenged dams required $10 4 CFU/mL and controls required no colonies at detection threshold of 10 3 CFU/mL), and the dams were treated with ampicillin (100 mg/kg) intraperitoneally twice daily until embryos were harvested for analysis. A control of challenge alone with no antibiotic treatment was not performed as the mice would succumb to infection.
Neural progenitor cell culture. Embryonic brains were harvested between E11 and E13 as indicated, and the cortical hemispheres were separated from the ganglionic eminences and the meninges. The tissue was processed using the Brainbits protocol, and cells were plated at a density of 400,000 cells/well on poly-D-lysine-coated plates or coverslips in complete neurobasal media. At 24 h after seeding, the cells were washed and incubated for 24 h with purified BCW (at a multiplicity of infection [MOI] of 0.1, prepared as described previously [20]). In some cases, cells were pretreated for 2 h prior to BCW challenge with TLR2/6 antagonist Git27 (10 mg/mL), TLR2/1 antagonist Cu-CPT22 (1 mM), or GLI inhibitor GANT61 (5 mM). For analysis, cells were either lysed as described for Western blotting or imaged for proliferation by Ki67 staining or presence of cilia (cilia, ARL13B, green; nucleus, DAPI, blue).
Immunohistochemistry. (i) Frozen sections. Embryonic brains were fixed in 4% paraformaldehyde overnight, cryoprotected in 30% sucrose at 4°C, frozen in M1 embedding matrix (Thermo Scientific, 1310), and cut sagittally to a thickness of 12 mm. For staining, sections were placed in boiling antigen retrieval buffer (10 mM sodium citrate [pH 6.0]) for 10 min, rinsed in PBS, and placed in blocking solution (5% donkey serum, 0.1% Triton X-100 in PBS) for 45 min. Sections were incubated with primary antibodies in blocking solution overnight at 4°C, rinsed three times in PBS, incubated with secondary antibody at room temperature for 2 h, rinsed three times in PBS, stained with of Hoechst dye for 10 min, and mounted in Prolong diamond antifade mounting media (Invitrogen, P36961).
(ii) Paraffin sections. Embryonic brains were harvested as described above, fixed in 10% formalin solution, embedded in paraffin, and sectioned as 4-mm-thick sections.
(iii) Staining procedures. Nissl (Sigma, C5042), TUNEL (EMD Millipore, S7101), Iba1 (CST 17198), and EdU (Invitrogen, C10632) staining was performed according to the manufacturer's instructions. For BrdU costaining, primary and secondary antibody treatment was performed first, followed by denaturing in 2 N HCl for 30 min at 37°C. Sections were rinsed in 0.1 M borate solution (pH 8.5) three times for 5 min each time, followed by PBS washes. Primary antibody solution containing anti-BrdU antibody was applied to the sections and placed in 4°C overnight, followed by the fluorescence staining method described above. A Tyramide Super-Boost kit (Invitrogen, B40922) was used for triple staining for TLR2, gamma tubulin, and ARL13B according to the manufacturer's protocol. The primary and secondary antibodies used for immunohistochemistry are listed in Table 2.
Microscopy. (i) Stereology. Both frozen and paraffin-embedded sagittal sections were stained with cresyl violet, and the entire cortical plate was contoured and counted by using StereoInvestigator with a Cavalieri estimator and an optical fractionator probe (MBF Biosciences) (a representative zone is shown  Fig. S7a in the supplemental material). Both right and left sides of the cortex were measured and the average of three sections per embryo was calculated.
(ii) Fluorescence image acquisition and analysis. Images were acquired on a Nikon C2 (Nikon NIS elements software) or a 3i Marianas system (Slidebook software) using a 40Â or 63Â objective. For frozen sections of 12-mm thickness, a z-stack of 13 optical sections at a step size of 0.34 mm and with a tiling of multiple images were combined for quantification analysis. For paraffin sections, a two-dimensional montage of a tiled image of the cortical plate region of the right hemispheres was captured using a 40Â oil objective on the Marianas. Using Imaris Â64 (Bitplane) image analysis software, one to three sagittal sections of the entire cortex from each E16 embryo were counted in a rectangular column of cortex at the level of the choroid plexus roof plate evaginated telencephalic vesicle, as indicated by the arrow in Fig. S7b. For consistent quantification of progenitor cells in slices, the VZ was defined as the area lining the ventricle and containing dense PAX6 1 TBR2 2 nuclei up to the area where cells uniformly express TBR2 (RGs and newborn IPs express PAX6; IPs express TBR2); the SVZ as the second cell-dense area containing uniformly TBR2 1 cells above the VZ and below a cell-sparse area. Continuous stretches of TBR2 1 cells form the boundary between the VZ and SVZ. We defined oRGs as PAX6 1 TBR2 2 cells above the VZ. Counting was done by at least two blinded individuals to ensure unbiased analysis.
(iii) Cell cycle analysis. We injected CldU at 2 h before harvest and EdU at 0.5 h before harvest. The vRGs that have left S phase during the 1.5-h interval between the CldU and EdU injections will be marked as CldU 1 EdU 2 (L cells = CldU 1 EdU 2 Pax6 1 Tbr2 2 ). The ratio of the length of a period of the cell cycle to that of another period is equal to the ratio of the number of cells in each period. Thus, we determined the length of S phase (Ts) from the equation 1.  Maternal Bacteria Redirect Fetal Neurodevelopment mBio Embryonic brain lysis. Individual embryonic brain cortices lysed in 250 mL ice-cold RIPA buffer (Sigma, R0278) with protease inhibitors (CST 5872S). The tissue was homogenized and vortexed for 2 h at 4°C. Lysates were centrifuged to collect debris and supernatant was harvested.
Western blot analysis. The concentration of total protein in each sample was determined by BCA (Thermo Scientific), and equal amounts were loaded onto a 4 to 12% SDS-PAGE. The antibodies used for Western blot analysis are listed in Table 2.
Behavioral analysis. Pups born to mothers treated at E10 with S. pneumoniae plus ampicillin or with ampicillin alone were assessed for cognitive function and repetitive behaviors at 2, 4, and 6 months of age. Gender equal numbers of pups from at least three litters (n . 12/group) per assay were tested. All assays were recorded by video and quantified by a blinded human observer. Working memory, repetitive behavior, and sociability were assessed as previously described (20,63).
Spatial transcriptomics. Data were generated and analyzed under the direction of Jeremy Crawford who was blinded to the experimental groups. Whole heads from E12 embryos were used for analysis with a Visium spatial transcriptomics kit (10X Genomics, PN-1000184). Immediately after dissection, samples were frozen in isopentane chilled by liquid nitrogen, embedded in OCT (Sakura, 4583), and then stored at 280°C. The, 10-mm sagittal sections were obtained beginning at the periphery to the midline according to 10X Genomics-suggested practices. Tissue permeabilization was optimized to 12 min. A single section (representative zone shown in Fig. S7c) was then obtained from each of four mice per condition, and two sections were placed on each Visium slide (i.e., one section from each condition). Sections were imaged on a Nikon Eclipse Ni-E scope, and slide fiducials and tissue coverage areas were manually annotated using Loupe Browser (10X Genomics). Libraries were prepared according to manufacturer recommendations and sequenced on an Illumina NovaSeq platform with a sequencing configuration of 28-10-10-120 (R1-i7-i5-R2) at .200 million clusters per library. Data were analyzed using SpaceRanger (v1.2.1; 10X Genomics) using the corresponding mm10 reference, with the reverse read trimmed to 90 bp.
Downstream analyses were performed using the STUtility (v0.1.0) (64) and Seurat (v4.0.3) (65) packages in the R environment (v4.1.0), with each library independently processed using SCTransform (regressing out the percentage of mitochondrial expression per spot) and then merged into a single object. Capture areas overlaying putative ventricular zone regions were identified by anatomical recognition based on the prenatal mouse brain atlas (66) and confirmed by the abundance of foxg1 and pax6 overlapping expression in these sites, and analyses focused on these regions. Differentially expressed genes (DEGs) between conditions were determined using the Seurat FindMarkers function with default parameters, with P values adjusted for multiple comparisons using Bonferroni correction. Gene set enrichment analysis (GSEA) (67) was performed using GSEAPreranked on a DEG list obtained using FindMarkers with min.pct set to 0.01 and logfc.threshold set to 0, with ranks ordered by average log 2 -fold change; for this analysis, Hallmarks (v7.4) (68) gene sets were queried using the Mouse MSigDB symbol remapping chip (v7.0) with a classic enrichment statistic. To ensure that the conclusions drawn from spatial transcriptomics experiments were rigorous, experiments included multiple, biologically independent sample replicates for each point of comparison. All statistical analyses are corrected for type 1 error (false discovery rate) by adjusting P values using the Benjamini-Hochberg method (69). All data and accompanying documentation are archived on servers both in the laboratory and in the St. Jude Children's Research Hospital cloud, which is backed up daily.
Statistics. We used Prism version 6.05 (GraphPad) for statistical analysis. Detailed statistical parameters are in each figure legend. All male and female embryos were analyzed. Groups contained at least two to three pregnant mice that yielded at least four to six embryos/dam, a cohort size giving a power of 0.95 to detect a 0.5-log change in cell number at a P value of 0.01. The treatment assignment, challenge status, and endpoint analysis were blinded to investigators.
Data analysis was performed to achieve robust and unbiased results. Unless otherwise specified, comparisons of phenotypes were analyzed by calculating the mean and standard deviation. P values were calculated using a two-tailed, unpaired Student t test for two-group comparisons or using two-tailed t test with Welch's correction or a Mann-Whitney test in Prism version 6.05 (GraphPad). In instances where multiple experimental conditions are compared to a single control group, statistical significance was tested using one-way analysis of variance (ANOVA), followed by Bonferroni's multiple-comparison test. A P value of ,0.05 was considered significant.
Data availability. Spatial transcriptomics data, including images, slide fiducials, filtered SpaceRanger outputs, and corresponding raw sequencing data are available via GEO accession GSE229699.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.