Hofbauer Cells Spread Listeria monocytogenes among Placental Cells and Undergo Pro-Inflammatory Reprogramming while Retaining Production of Tolerogenic Factors

ABSTRACT Pregnant women are highly susceptible to infection by the bacterial pathogen Listeria monocytogenes, leading to miscarriage, premature birth, and neonatal infection. L. monocytogenes is thought to breach the placental barrier by infecting trophoblasts at the maternal/fetal interface. However, the fate of L. monocytogenes within chorionic villi and how infection reaches the fetus are unsettled. Hofbauer cells (HBCs) are fetal placental macrophages and the only leukocytes residing in healthy chorionic villi, forming a last immune barrier protecting fetal blood from infection. Little is known about the HBCs’ antimicrobial responses to pathogens. Here, we studied L. monocytogenes interaction with human primary HBCs. Remarkably, despite their M2 anti-inflammatory phenotype at basal state, HBCs phagocytose and kill non-pathogenic bacteria like Listeria innocua and display low susceptibility to infection by L. monocytogenes. However, L. monocytogenes can exploit HBCs to spread to surrounding placental cells. Transcriptomic analyses by RNA sequencing revealed that HBCs undergo pro-inflammatory reprogramming upon L. monocytogenes infection, similarly to macrophages stimulated by the potent M1-polarizing agents lipopolysaccharide (LPS)/interferon gamma (IFN-γ). Infected HBCs also express pro-inflammatory chemokines known to promote placental infiltration by maternal leukocytes. However, HBCs maintain the expression of a collection of tolerogenic genes and secretion of tolerogenic cytokines, consistent with their tissue homeostatic role in prevention of fetal rejection. In conclusion, we propose a previously unrecognized model in which HBCs promote the spreading of L. monocytogenes among placental cells and transition to a pro-inflammatory state likely to favor innate immune responses, while maintaining the expression of tolerogenic factors known to prevent maternal anti-fetal adaptive immunity.

growth factor beta (TGF-b) (3,(24)(25)(26). Therefore, based on their fetal origin and M2polarized phenotype, HBCs are expected to be highly permissive to L. monocytogenes infection. It is unknown whether infection can repolarize HBCs toward an M1-like proinflammatory and antimicrobial phenotype. The anti-inflammatory environment of the placenta is critical during the second and third trimesters of pregnancy, and tilting the balance toward an excessive inflammatory state can lead to adverse pregnancy outcomes (5). Whether HBCs are involved in detrimental placental inflammation by an M1 repolarization, which could contribute to villitis observed in listeriosis cases, is unsettled (27,28). HBCs were proposed to maintain an M2 phenotype despite pro-inflammatory environmental cues, such as stimulation with the M1-polarizing agents interferon gamma (IFN-g)/lipopolysaccharide (LPS) (24)(25)(26)(29)(30)(31)(32). In this work, we measured the HBCs' susceptibility to L. monocytogenes infection and studied the HBCs' transcriptome, cytokine production, and expression of surface markers to gain detailed information about their responses to L. monocytogenes. As controls, we exposed HBCs to the potent M1-polarizing agents (IFN-g/LPS) and used the well-characterized human macrophage model THP-1 cells (M0 [untreated] and M1 [IFN-g/LPS-polarized]) (33)(34)(35)(36). We found that HBCs are able to kill non-pathogenic Listeria innocua and display low susceptibility to L. monocytogenes infection. However, L. monocytogenes cells still replicate in the HBCs and can exploit these cells to spread to other placental/fetal cells. HBCs are plastic; upon infection, they undergo an M1-like repolarization accompanied by the production of numerous chemokines, with the potential to elicit maternal leukocyte infiltration of chorionic villi. However, remarkably, the production of tolerogenic factors remained sustained.

RESULTS
Hofbauer cell isolation and surface marker characterization. HBCs were isolated from healthy term placentas, and their purity was assessed by fluorescence microscopy after 72 h of culture, corresponding to the time point of all infections performed in this article. The pan-leukocyte CD45, M2 macrophage CD163, fetal fibroblast CD90, and trophoblast cytokeratin 7 markers were labeled with fluorescent antibodies (Fig. S1A in the supplemental material) (26,37). We observed that CD45 1 cells were highly vacuolated, which is a characteristic of HBCs (Fig. S1B) (37). As expected, 98% of CD45 1 cells were also CD163 1 , since a minor proportion of HBCs is known to be CD163 low or negative ( Fig. S1C) (26,37). HBC purity, measured as the percentage of CD45 1 cells, was 92.58% 6 1.28% (mean 6 standard error of the mean [SEM]) (n = 7 placentas), and at least 90 Â 10 6 HBCs were recovered per placenta, similar to previous reports (26,37). Trophoblasts and fibroblasts were the only cell contaminants, as the sum of the three cell populations (CD45 1 , CD90 1 , and C7 1 ) reached 100% (Fig. S1C). Among these cells, only fibroblasts proliferated, as confirmed by co-labeling cell markers and the Ki67 marker (data not shown), with a doubling time of 17 h. Back calculation indicated that HBC purity was .97% at the time of their isolation, consistent with the purity achieved when negative selection was used (37). Labeling of HBCs with M1-specific (CD80) and M2-specific (CD163, CD14, CD206, and CD209) markers confirmed their M2 phenotype, as more than 90% of HBCs were CD80 2 CD163 1 CD14 1 CD206 1 (Fig. S1D).
Hofbauer cells control the non-pathogenic bacterial species L. innocua and display low susceptibility to L. monocytogenes infection. We analyzed whether HBCs could phagocytose L. monocytogenes and control infection in their basal M2 state, and whether their antimicrobial activity could be enhanced upon exposure to potent M1polarizing agents, as shown using non-placental macrophages (23). HBCs were stimulated for 24 h with the prototypical M1-polarizing cocktail IFN-g/LPS (stimulated) or not (untreated). For comparison with a well-characterized human macrophage model, we used phorbol myristate acetate (PMA)-differentiated THP-1 cells, either untreated (M0 phenotype [M0-THP-1 cells]) or stimulated by IFN-g/LPS for 24 h (M1 phenotype [M1-THP-1 cells]) (33,34,38,39). HBCs and THP-1 cells were infected with wild-type (WT) L. monocytogenes and an isogenic listeriolysin O (LLO)-deficient mutant (Dhly) that fails to escape from the phagosome and replicate in macrophages (40)(41)(42). The phagocytic efficiencies were measured by enumeration of fluorescently labeled intracellular bacteria (microscopy) and viable intracellular bacteria (CFU) ( Fig. 2A and B). Untreated HBCs were 2-fold less phagocytic than M0-THP-1 cells ( Fig. 2A and B). However, stimulated HBCs maintained their phagocytic efficiency, unlike M1-THP-1 cells, whose phagocytic efficiency dropped significantly, by 10-fold ( Fig. 2A and B). We next measured the number of viable intracellular bacteria (CFU) at different times post-infection. As previously reported, M1-THP-1 cells controlled L. monocytogenes better than M0-THP-1 cells For fluorescence microscopy, cells were labeled with DAPI and Abs against CD45 (N CD45 ), extracellular bacteria (N e ), and total bacteria (N t ). Fifteen random fields of view (;700 HBCs per condition) were acquired. Phagocytosis was calculated as the number of intracellular L. monocytogenes per cell with the formula (N t 2 N e )/N CD45 6 SEM. n = 3 placentas: *, P , 0.05. For CFU counts, n = 3 placentas: *, P , 0.05; **, P # 0.0043. All P values for A and B were calculated by two-way analysis of variance and Sidak's multiple comparisons tests. (C and D) CFU counts from THP-1 cells (C) or HBCs (D) at the indicated time points on the figure. Data are expressed as the number of CFUs per cell 6 SEM. n = 3 for each experiment whose results are shown in panels A to D, data obtained from a total of 5 placentas: *, P , 0.02; **, P , 0.006; ***, P , 0.0009; ****, P , 0.0001 (two-way analysis of variance and Tukey's multiplecomparison tests). (E) The fold increases of CFUs/cell from the experiments whose results are shown in panels C and D were calculated relative to the CFUs/cell 6 SEM (n = 3) at 1 h of infection for HBCs and THP-1 cells. *, P # 0.041 (ratio paired Student's t test). (F) Stimulated or untreated HBCs ( Fig. 2C) (34)(35)(36). Similarly, stimulated HBCs controlled infection better than untreated HBCs, and remarkably, there was no net increase in viable intracellular bacteria 5 to 7.5 h post-infection in untreated or stimulated HBCs (Fig. 2D). To compare HBCs' and THP-1 cells' ability to control phagocytosed L. monocytogenes, the fold changes in viable intracellular bacteria over time were calculated relative to the counts at 1 h of infection to correct for differences in phagocytosis (Fig. 2E). We found that untreated and stimulated HBCs were significantly less susceptible to L. monocytogenes infection than THP-1 cells.
We next analyzed whether HBCs were able to kill the nonpathogenic species Listeria innocua. We found that HBCs were able to kill ;40% and ;50% of phagocytosed L. innocua cells 5 h and 10 h post-phagocytosis, respectively (Fig. 2F).
To determine if stimulated HBCs, similar to stimulated non-placental macrophages, could confine L. monocytogenes in the phagosome better than their non-stimulated counterparts, we measured the percentage of intracellular L. monocytogenes present in the endolysosomal compartments (LAMP-1 1 ) 2 h post-infection by fluorescence microscopy (Fig. 3A). As a complementary approach, we measured the percentage of intracellular L. monocytogenes cells that recruited F-actin once they reached the cytosol (Fig. 3A). Consistent with increased L. monocytogenes phagosomal containment in stimulated HBCs, the proportion of LAMP-1 1 intracellular bacteria increased significantly, whereas the F-actin 1 intracellular bacteria decreased significantly in a similar proportion (Fig. 3B).
To determine whether HBC stimulation, in addition to improving the containment of L. monocytogenes in LAMP-1 1 vacuoles, could also affect the bacterial cytosolic doubling time, cells were infected with pactA-RFP L. monocytogenes cells, which express red fluorescent protein (RFP) once they reach the host cell cytosol ( Fig. 3C and 4A) (10). The cytosolic doubling time was calculated based on the enumeration of RFP-positive (RFP 1 ) bacteria by live-cell fluorescence microscopy 4 to 6 h post-infection (Fig. 3C). The doubling time in untreated HBCs increased significantly, from 83.59 6 7.70 min to 104.06 6 10.08 min (mean 6 SEM) in IFN-g-/LPS-stimulated cells ( Fig. 3C; Movies S1 and S2).
Collectively, the data show that in their basal state, HBCs can exert antimicrobial activities and display low susceptibility to L. monocytogenes infection compared to an established macrophage model, yet they fail to fully control L. monocytogenes cytosolic growth. Furthermore, HBCs display functional plasticity when exposed to prototypical M1-polarizing agents, as evidenced by significantly increased phagosomal containment and increased L. monocytogenes cytosolic doubling time.
Hofbauer cells can spread L. monocytogenes to other placental and fetal cells. The virulence factor ActA, which promotes L. monocytogenes cell-to-cell spread via Factin-based intracellular motility, is required for placental/fetal infection (12,(43)(44)(45). We tested whether HBCs could act as transmission vessels to other placental cells in an ActA-dependent fashion. Live-cell imaging of HBCs infected with pactA-RFP L. monocytogenes revealed that cytosolic bacteria were motile and formed extracellular protrusions, which are the structures supporting cell-to-cell spread ( Fig. 4A; Movies S1 and S2). We then measured the percentages of L. monocytogenes-infected HBCs by fluorescence microscopy 0.5 to 24 h post-infection (Fig. 4B). Infected cells were defined as cells with one or more intracellular L. monocytogenes. There was a significant increase in the percentages of infected (stimulated and untreated) HBCs over time, which was a manifestation of L. monocytogenes intercellular spreading, because gentamicin in the culture medium killed extracellular bacteria (Fig. 4C). Consistent with better control of L. monocytogenes by stimulated HBCs (Fig. 2), the results in Fig. 4C show that at time points 5, 7.5, 10, and 24 h, the percentages of infected IFN-g-/LPS-stimulated HBCs were significantly decreased compared to the percentages of infected untreated HBCs. Next, we assessed whether infected HBCs could spread L. monocytogenes to other placental and fetal cells. HBCs were infected with WT or isogenic DactA L. monocytogenes and were detached and co-cultured overnight on a confluent monolayer of primary human trophoblasts (PHTs; isolated from the same placenta) or primary human umbilical vein endothelial cells (HUVECs; commercially available) in the presence of gentamicin. The data showed that WT L. monocytogenes, but not the DactA mutant, could spread from infected HBCs to PHTs and HUVECs, regardless of the HBCs' stimulation state ( Fig. 5A and B; Fig. S2). The numbers of infectious foci per cell surface area were similar in PHTs and HUVECs (Fig. 5B). However, the average size of the infectious foci was larger in HUVECs than in PHTs. Also, untreated HBCs appeared to spread infection more efficiently than stimulated HBCs, as characterized by both decreased numbers and sizes of foci in HUVECs and PHTs. This is likely due to the higher intracellular bacterial load in untreated HBCs than in stimulated HBCs (Fig. 2). We tested whether this could also result from more efficient L. monocytogenes motility in untreated HBCs, but the speed of pactA-RFP L. monocytogenes cells, measured by live-cell imaging, showed that this was not the case (Fig. 5C). However, there was a significantly higher percentage of non-motile bacteria in stimulated HBCs than in untreated HBCs, consistent with less efficient bacterial spread by stimulated HBCs (Fig. 5D). monocytogenes-infected cells were fixed with PFA 2 h post-infection and labeled with anti-L. monocytogenes fluorescent Abs to label total and extracellular L. monocytogenes. Intracellular L. monocytogenes cells were color-coded in red. LAMP-1 was labeled with antihuman LAMP-1 Abs and secondary Alexa Fluor 488-conjugated Abs (green), and in a second set of samples, F-actin was labeled with Alexa Fluor 488-conjugated phalloidin (green). The levels of LAMP-1-positive (LAMP-1 1 ) bacteria were 11.3% and 20% in untreated and stimulated HBCs, respectively. Scale bar represents 5 mm. (B) The percentages of LAMP-1 1 and F-actin 1 intracellular bacteria were counted from at least 19 planes (corresponding to a minimum of 500 bacteria) per experimental condition and were expressed as fold change 6 SEM in comparison to the results for untreated HBCs. n = 3 placentas: *, P # 0.0368 (paired Student's t test) (C) Bacterial division time was calculated based on the number of RFP 1 L. monocytogenes cells 4 to 6 h post-infection. n = 3 placentas, different from those used in the experiment whose results are shown in panels A and B: *, P = 0.0280 (paired Student's t test).
L. monocytogenes-infected Hofbauer cells undergo pro-inflammatory reprogramming, including the upregulation of cytokines and chemokines known to recruit leukocytes into chorionic villi. To gain information about HBCs' plasticity and responses to L. monocytogenes infection, we performed RNA sequencing (RNA-seq) and measured an array of 20 cytokines released by infected HBCs. RNA-seq analysis of HBCs infected for 5 h by WT L. monocytogenes in comparison to control noninfected cells revealed that among the 12,593 genes (expression levels above 2 counts per million [CPM] in at least half of the samples), 1,913 (15.2%) were significantly differentially expressed (DE), including 1,037 (8.2%) that were upregulated and 876 (7.0%) that were downregulated, as illustrated by the volcano plot and heatmaps in Fig. 6A and B. monocytogenes. Cells were infected for up to 24 h in the presence of gentamicin. At the indicated time points on the figure, cells were fixed with PFA and labeled using Abs directed against CD45 (green), against extracellular L. monocytogenes, and against total bacteria following cell permeabilization and nuclei were labeled with DAPI. (A) HBCs were infected with pactA-RFP L. monocytogenes for 4 h. Cells were placed on the atmosphere-controlled microscope, and images were recorded for 2 h. The representative picture was taken 5 h post-infection (from Movie S1) and is an overlay of phase-contrast and green color-coded fluorescent L. monocytogenes images. Arrows indicate L. monocytogenes cells with comet tails. Scale bar represents 30 mm. (B) Representative images of infected untreated (UT) and stimulated HBCs in which intracellular L. monocytogenes cells are color-coded in red (selected from 3 independent experiments corresponding to 3 placentas). Scale bar represents 30 mm. (C) Fluorescence images were randomly acquired (;700 cells per condition), and the average percentages of infected HBCs (with at least 1 intracellular L. monocytogenes) 6 SEM were counted over time. n = 3, the same placentas as used in the experiment whose results are shown in panel B: *, P , 0.0395; **, P # 0.0085; ****, P , 0.0001. The asterisks on the gray bars (stimulated cells) indicate significant differences in comparison to the corresponding time points in untreated HBCs. *, P , 0096. Mixed-effect model where the main effect and interaction of time and treatment were the independent variables, and Tukey's multiple-comparison test was used for all P values generated for the data in panel C. The top 50 DE genes in 5-h-infected HBCs predominantly included genes encoding cytokines, chemokines, and their receptors and regulators (26 out of 50 genes) (Fig. 6B). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses (DAVID 6.8) of the DE genes revealed that infection led to the upregulation of numerous pro-inflammatory pathways ( Fig. 6C and D). These included signaling through pattern recognition receptors (Toll-like receptor 2 [TLR2]/MyD88, RIG-I/cGAS, NOD1/RIPK2, and NLRP3/caspase-1) that converged toward IFN regulatory factor 7 (IRF7) and NF-k B activation (among other transcription factors) and the downstream production of numerous pro-inflammatory mediators, consistent with sensing and responding to extracellular and intracellular Gram-positive bacteria. In accordance with this finding, there was a significant upregulation of immune signaling pathways, including phosphoinositide 3-kinase (PI3K), extracellular signal-regulated kinases 1 and 2 (ERK1/2), mitogen-activated protein kinase (MAPK), Notch, JAK-STAT, and NF-k B, the biosynthetic pathways for a set of cytokines (IFN-a/b, IL-1 a/b, IFN-g, IL-6, tumor necrosis factor alpha [TNF-a], IL-12, and IL-8), and multiple chemokines, as well as the autocrine responses to some of these mediators ( Fig. 6C and D). Importantly, the pro-inflammatory remodeling observed at the transcriptional level was confirmed by significant production of the corresponding cytokines (IFNa/b, IL-1 a/b, IFN-g, IL-6, TNF-a, IL-12, and IL-8) by L. monocytogenes-infected HBCs (Fig. 7). The production of IL-18 and IL-1b reflected that L. monocytogenes infection activated caspase-1, in accordance with the significant upregulation of NLRP3 and CASP1 transcripts and associated pathways ( Fig. 6) (46). Interestingly, genes coding for T-cell chemokines (CXCL9, CXCL10, CXCL11, CCL3L1, CCL5, CCL1, CCL22, and CCL24) and neutrophil and monocyte/macrophage chemokines (CXCL8 [IL-8], CXCL1, CXCL2, CXCL6, CCL3, CCL4, CCL2, CCL7, and CCL8) were all upregulated (Table S1) (47), as were antimicrobial response pathways and antigen presentation by major histocompatibility complex class I (MHC-I) (Fig. 6). RNA-seq analysis comparing infected HBCs at 5 and 24 h post-infection showed that there were only minor variations in the upregulated pathways 24 h postinfection ( Fig. S3 and Table S2). Surprisingly, although many genes were downregulated 5 h post-infection, there was no significant GO or KEGG pathway associated with these genes. Downregulated pathways reached statistical significance (EASE score [a modified  Table S2). These pathways included phagosome formation and maturation, antigen presentation by MHC-II, metabolism (tryptophan, fatty acid, and pyruvate metabolism), transcription regulation, including epigenetic control, and translation (ribosome biogenesis).
In parallel experiments, we performed RNA-seq analysis of HBCs exposed to IFN-g/LPS for 24 h in comparison to control untreated cells. Data showed that 983 (7.8%) genes were significantly induced and 1,183 (9.4%) were significantly repressed ( Fig. 8A and B). The top 50 significantly DE protein-coding genes were largely dominated by immune genes (Fig. 8B). Gene Ontology analysis of DE genes revealed that the top upregulated pathways included IFN-g and LPS signaling, confirming HBC responsiveness to these agents (Fig. 8C). Other upregulated pathways reflected the production of and responses to pro-inflammatory cytokines, immune cell trafficking, and immune defense pathways. These included critical immune signaling, such presence of gentamicin. HBCs and PHTs were isolated from the same placenta, and HUVECs were commercially purchased. Cells were fixed with PFA, permeabilized, and labeled with fluorescent Abs against L. monocytogenes (Lm, red), E-cadherin (PHTs, green), and CD45 (HBCs, yellow) and with phalloidin (F-actin in HUVECs, green) and DAPI (nuclei, blue). Scale bar represents 30 mm. Images of the corresponding experiment involving stimulated HBCs are shown in  as MAPK and ERK1/ERK2 cascades, NF-k B, and key pro-inflammatory cytokines IFN-a/b, TNFa, IL-1a/b, IL-6, IL-8, and IL-12. Confirming these pathways, IFN-a/b, TNF-a, IL-1a/b, IFN-g, IL-6, IL-8, and IL-12 cytokine production were markedly increased (Fig. 9). Numerous chemokine genes were highly upregulated, including CXCL9, -10, -11, and -8 ( Fig. 8A and B; Table S1). At the protein level, CXCL8 (IL-8) was the most abundantly produced cytokine at the basal level (in our array), and its production increased ;30-fold upon stimulation. Other major upregulated pathways included antigen processing for MHC-I presentation, with upregulation of classical (HLA-A, -B, and -C) and non-classical (HLA-E, -F, and -G) HLA-I genes. Conversely, tissue-specific homeostatic pathways (mostly associated with the M2 phenotype) were downregulated, including extracellular matrix organization, angiogenesis, and wound healing (Fig. 8C).
In conclusion, HBCs can undergo pro-inflammatory reprogramming in response to L. monocytogenes infection and to exposure to potent M1-polarizing agents, as characterized by the upregulation of pro-inflammatory and antimicrobial transcriptional pathways, consistent with the increased production of pro-inflammatory cytokines and chemokines and enhanced L. monocytogenes infection control ( Fig. 2 and 4).
M1/M2 gene signature and cytokine analyses support that L. monocytogenes infection repolarizes HBCs toward an M1 phenotype. The transcriptional remodeling of infected or IFN-g-/LPS-stimulated HBCs reflected an M1 polarization. To further establish whether HBCs could repolarize, we next analyzed DE M1 and M2 signature genes 5 h and 24 h after L. monocytogenes infection and compared the results with 24h IFN-g-/LPS-treated HBCs. Based on the literature, we generated a list of 75 M1 and 80 M2 gene markers, which encode transcription factors, cytokines/chemokines, immune receptors, immune transducers, and antimicrobial and pro-and anti-inflammatory molecules (Table S3). HBC infection or stimulation with IFN-g/LPS for 24 h led to the predominant upregulation of M1 genes, while M2 genes were downregulated (Table S3). Strikingly, M1-and M2-specific transcription regulators were upregulated and downregulated, respectively, in infected HBCs and IFN-g-/LPS-stimulated HBCs, similar to a previous report of M1-polarized blood-derived human macrophages (  Table S3). Interestingly, a feature of HBCs infected by L. monocytogenes for 24 h but not observed in IFN-g-/LPS-stimulated HBCs was the downregulation of 13 MHC-II genes (M1 markers), as previously reported as an L. monocytogenes virulence mechanism in non-placental macrophages (Table S3) (49,50). Our data further support that this is due to the downregulation of the MHC-II transcriptional coactivator CIITA (Table S3) (51).
Azari et al. Another indication of macrophage repolarization is their metabolic shift, with M1 macrophages relying more on glycolysis and accumulating itaconate, whereas M2 macrophages rely on oxidative phosphorylation and have an intact tricarboxylic acid (TCA) cycle (55). Two major metabolic genes involved in M1 polarization were significantly upregulated 5 and 24 h post-infection, the PFKFB3 gene, which stimulates glycolysis, and the ACOD1 gene, involved in itaconate production ( Fig. 6A; Table S4) (55,56). In parallel, genes supporting fatty acid synthesis and oxidation, the FASN (57) and HADH (58) gene, respectively, were downregulated ( Fig. 6D; Table S4). Fatty acids are required for the TCA cycle in M2 macrophages (55,59). Other downregulated genes involved in the TCA cycle were the genes coding for pyruvate carboxylase (PC), an enzyme involved in the production of oxaloacetate from pyruvate, and lactate dehydrogenase (LDHD) and ME1, both involved in the production of pyruvate (Table S4) (60). Downregulation of these genes and pathways in L. monocytogenes-infected HBCs suggests a metabolic shift toward an M1-like phenotype by disruption of M2-specific metabolic pathways. In comparison, only ACOD1 was strongly induced in IFN-g-/LPS-stimulated HBCs (Table S4).
As an alternative approach to assess macrophage repolarization, we measured the surface expression of canonical M1/M2 markers by quantitative fluorescence microscopy (Fig. S4). This approach was favored over flow cytometry, which would have altered surface markers while detaching these strongly adherent cells. We found similar results for infected and IFN-g-/LPS-stimulated HBCs: CD163 (M2 marker) decreased only slightly, CD86 (M1 and M2b markers) increased slightly, and the expression of CD80 (M1 marker) and CD14/CD206/CD209 (M2 markers) remained minimally affected (Fig. S4).
In conclusion, both transcriptional reprogramming and cytokine profiling support an M1 repolarization of HBCs upon infection by L. monocytogenes and stimulation by IFNg/LPS. Surprisingly, the expression of some M2 surface markers decreased only slightly and the CD80 M1 surface marker did not increase upon either infection or IFN-g/LPS stimulation.

DISCUSSION
Although previous studies have been devoted to establishing how L. monocytogenes can breach the placental barrier, no studies addressed the fate of L. monocytogenes within the placenta nor its mechanism of transmission to the fetus. In particular, whether or not the fetal, placenta-resident macrophages, Hofbauer cells, can control L. monocytogenes infection was unknown. In the present study, we showed that, at the basal state, HBCs killed a large proportion of phagocytosed non-pathogenic bacterial species, such as L. innocua, and displayed relatively low susceptibility to infection by L.  Table S3. monocytogenes. However, L monocytogenes could exploit HBCs to maintain an intracellular niche and infect surrounding placental cells. Furthermore, the data support that HBCs underwent pro-inflammatory reprograming toward an M1 phenotype upon L. monocytogenes infection, leading to the production of major pro-inflammatory cytokines. However, HBCs upregulated the expression of key tolerogenic genes and maintained the production of tolerogenic cytokines, in accordance with their role in preventing the activation of maternal anti-fetal adaptive immune responses.
Common placental complications of infectious and non-infectious etiologies, including villitis, chorioamnionitis, miscarriage, and preterm delivery, are associated with abnormal placental inflammation, and the contribution of HBCs to inflammation has been debated (3,17,88,89). Some studies support that HBCs maintain an M2 antiinflammatory profile in cases of gestational diabetes mellitus (GDM) and villitis (24,25,29). Other studies support a pro-inflammatory role for HBCs in GDM, villitis of unknown etiology (VUE), and during infection (30,(90)(91)(92)(93). A recent study proposed that, despite stimulation by IFN-g/LPS for 96 h, HBCs resisted M1 repolarization and displayed an M2b phenotype (29). Our data strongly support the plasticity of infected or IFN-g-/LPSstimulated HBCs, as characterized by their transcriptional reprogramming toward an M1 pro-inflammatory phenotype, accompanied by the production of pro-inflammatory cytokines and chemokines. In accordance with HBCs' plasticity are the findings of a previous study that compared the DNA methylomes of HBCs, decidual macrophages (placental macrophages of maternal origin), and fetal and maternal blood monocytes  (94). That study found that HBCs' M1 and immune-response-related genes were hypermethylated, whereas M2 genes were hypo-methylated, consistent with the HBC's antiinflammatory and homeostatic functions at basal levels. Of interest, DNA methyl transferases DNMT3A and DNMT3B (which promote de novo methylation) were highly expressed by HBCs in comparison to the levels of maternal monocyte/macrophages, suggesting that HBCs are prone to epigenetic remodeling (94). Our data support that HBCs can undergo M1-like reprogramming. Indeed, numerous M2 transcripts, including the transcription factor (TF) MAF, which assembles enhancers for sustained M2 gene transcription, were downregulated in stimulated and in infected HBCs, similarly to M1-polarized adult blood-derived macrophages (48). Reciprocally, numerous M1 genes were upregulated, including the transcription factor IRF1, which assembles enhancers for M1 gene transcription (48). In addition, the transcriptome of L. monocytogenes-infected HBCs reflected increased glycolysis and itaconate production (an M1 signature), whereas several M2 metabolic pathways involving fatty acids and pyruvate were downregulated (Fig. 6). Alternatively, fatty acids are required for bacterial membrane synthesis (95); therefore, downregulation of fatty acid synthesis could be a strategy used by HBCs to limit bacterial proliferation. Gene Ontology and KEGG pathway analyses remarkably recapitulated the detection of extracellular and intracellular Gram-positive organisms by pattern recognition receptors that leads to transcriptional activation of numerous pro-inflammatory cytokines and chemokines and their downstream signaling pathways ( Fig. 6C and D). The cytokine array of L. monocytogenes-infected HBCs confirmed the production of many of these pro-inflammatory cytokines (IL-8, TNF-a, IL-6, IL-23, IFN-a/b, IFN-g, IL-12, IL-17A, and IL-1a), including the inflammasome-/caspase-1-dependent cytokines IL-1b and IL-18 ( Fig. 7) (46).
L. monocytogenes-infected HBCs undergo unique M1-like reprogramming, which could be described as M1t (tolerogenic), for the maintenance of tolerogenic factors associated with fetal tolerance. Some tolerogenic properties were also observed in IFN-g-/LPS-stimulated HBCs, suggesting that HBCs may be programmed to maintain fetal tolerance despite various pro-inflammatory stimulations (Table 1). HBCs maintained or increased the expression of genes encoding chemokines, surface receptors, and secreted factors that favor the recruitment and development of regulatory T cells (Table 1), and they failed to express the co-stimulatory molecule CD80 (Fig. S4). Thus, we propose that HBCs are programmed to respond to L. monocytogenes infection by promoting an environment favorable to the innate immune responses to infection, while attempting to prevent deleterious adaptive anti-fetal immunity, at least during the first 24 h of exposure to the pathogen. It should also be considered that L. monocytogenes may induce some of the tolerogenic responses as a virulence mechanism, as illustrated by the downregulation of the MHC-II gene transcriptional coactivator CIITA and numerous MHC-II-coding genes (49). The tolerogenic properties of HBCs may constitute the HBCs' Achilles heel during infection, as they can potentially lead to antibacterial T cell response suppression and may facilitate the establishment of L. monocytogenes in the placenta and subsequent spread to the fetus.
The literature reports that different types of macrophages display a wide range of susceptibility to L. monocytogenes infection. Murine M2-polarized liver-resident Kupffer cells are highly susceptible to and rapidly die from L. monocytogenes infection (96). We found that pro-survival and apoptotic pathways were both upregulated in infected HBCs. However, pro-survival pathways likely dominated, since within 24 h of infection, there were no apoptotic nuclei (DAPI [49,6-diamidino-2-phenylindole] labeling) and no detectable cell detachment or death of infected HBCs, unlike THP-1 cells, which detached massively 24 h post-infection (data not shown). Human blood-derived macrophages and murine peritoneal macrophages can kill L. monocytogenes, and more so after IFN-g, TNF-a, or granulocyte-macrophage colony-stimulating factor (GM-CSF) stimulation (97,98). Stimulation of murine bone marrow-derived macrophages by IFNg/LPS enhances L. monocytogenes control while decreasing phagocytic efficiency but does not induce net bacterial killing (31, 99, 100). We found that HBCs restrict L. monocytogenes infection, although not completely, as a proportion of cytosolic L. monocytogenes cells remain viable, motile, and able to divide (Fig. 3C and 5C and D). However, the net number of viable intracellular bacteria did not increase between 5 and 7.5 h post-infection (Fig. 2D), supporting that intracellular killing of L. monocytogenes compensated for its cytosolic multiplication. Consistent with the notion that HBCs display some antimicrobial properties and are able to kill phagocytosed bacteria, we showed that a significant proportion (;40%) of phagocytosed L. innocua cells (non-pathogenic) were killed as early as 5 h post-infection (Fig. 2F), in accordance with a recent study showing that first-trimester HBCs were bactericidal against nonpathogenic bacterial species (101). We found that ;11% of intracellular L. monocytogenes cells remained trapped in LAMP-1-positive vacuoles 2 h post-infection, which increased to 20% in stimulated HBCs (Fig. 3A and B), consistent with the notion that HBCs may successfully kill a small proportion of phagosomal L. monocytogenes. We also found that in stimulated HBCs, the bacterial cytosolic division time increased while the proportion of motile bacteria decreased 4 to 6 h post-infection ( Fig. 3C and 5D). These data suggest a model in which non-motile bacteria might be targeted by autophagy, since motility is critical to escape from autophagy (102). The M1-like pro-inflammatory reprogramming of HBCs observed 5 h post-infection is consistent with the fact that the net number of viable intracellular bacteria stopped increasing between 5 and 7.5 h post-infection, indicating the activation of some antimicrobial mechanisms. Together, these data indicate that HBCs display some antimicrobial properties at the phagosomal and cytosolic levels at basal state and have the intrinsic ability to further increase their activities if properly stimulated. However, these antimicrobial functions are not sufficient to avert L. monocytogenes intracellular survival.
Macrophages have been shown to spread L. monocytogenes infection to other cell types, including endothelial cells (41,103,104), and L. monocytogenes cell-to-cell spread is required for infection of the placental/fetal unit (11,12,45). Based on their presence at the placental/fetal interface, their phagocytic ability, and their motility, HBCs are the best candidates for fetal transmission of L. monocytogenes. In support of this, we found that HBCs transfer L. monocytogenes cells to other primary placental cells, including fetal endothelial cells (Fig. 5A and B; Fig. S2). Furthermore, HBCs may play a general role in spreading infection to the fetus, as recently proposed for the Zika virus (105)(106)(107)(108). During infection, tissue-resident macrophages typically release chemokines to attract inflammatory monocytes, neutrophils, and other leukocytes for subsequent clearance of the pathogens (96,109). We observed that L. monocytogenes-infected HBCs upregulate numerous transcripts for chemokines that attract neutrophils, monocytes/macrophages, and T cells, which was confirmed at the protein level with IL-8. The pro-inflammatory profile of infected HBCs is in accordance with the villitis observed in animal and human listeriosis cases (11,27,28). Villitis can lead to infiltration of maternal leukocytes, as evidenced by trafficking of CD8 1 cytotoxic T cells in villitis of unknown etiology (VUE) (92,110). VUE is also accompanied by increased levels of CXCL9, -10, and -11 (CXCR3 ligands) in HBCs and can lead to fetal death and other complications (111). Similarly, we observed induction of gene expression for these chemokines in L. monocytogenes-infected HBCs. Therefore, it is likely that if infection of chorionic villi is not resolved, exacerbation of inflammation by HBCs and infiltration of maternal leukocytes may contribute to poor outcomes, such as abortion and preterm birth, observed in listeriosis cases in pregnancy (6).

MATERIALS AND METHODS
Bacterial strains and culture. Wild-type (WT) L. monocytogenes strain 10403S and its isogenic Dhly and DactA mutants were from Daniel Portnoy (University of California, Berkeley). The isogenic red fluorescent L. monocytogenes (pactA-RFP) was from Anna Bakardjiev (University of California, San Francisco) (10). L. innocua (strain BUG 499) was from Pascale Cossart (Pasteur Institute, Paris). Bacteria were grown overnight in antibiotic-free brain heart infusion (BHI) broth in a shaking incubator at 37°C until reaching an optical density at 600 nm (OD 600 ) of 0.7 to 0.8, and were diluted to the indicated multiplicity of infection (MOI) in serum-/antibiotic-free cell culture medium as shown under the corresponding figures.
HBC isolation and characterization. Placentas (a total of 18 placentas) were obtained from healthy, singleton Cesarean section 38 The following morning, cells were passed through a 40-mm sieve, centrifuged, and suspended at 10 6 /ml in isolation RPMI. HBCs were isolated based on their ability to attach to the cell culture plate for 30 min at 37°C, whereas non-adherent chorionic cells were removed by washes. The cells were cultured with Dulbecco's modified Eagle's medium/nutrient mixture F-12 (1:1) (DMEM/F-12 GlutaMax; Gibco) containing 10% FBSH, 50 U/ml penicillin, and 50 mg/ml streptomycin (Invitrogen) (complete DMEM/F-12). All HBCs were cultured for 48 h and then exposed to IFN-g/LPS for 24 h (in parallel, untreated HBCs were cultured for 24 h in the absence of IFN-g/LPS). Subsequently, the stimulated and untreated cells were infected, so that all infection experiments described in this article were carried out 72 h after plating of HBCs (29,37). For purity assessment, untreated cells were fixed Trophoblast, HUVEC, and THP-1 cell culture. Primary human trophoblasts (PHTs) were isolated (from the same placenta as HBCs for each experiment) as previously described (112) and cultured in complete DMEM/F-12. Human umbilical vein endothelial cells (HUVECs) were from ScienCell Research Laboratories and cultured as previously described (113). THP-1 cells (ATCC TIB-202; authenticated by ATCC's short tandem repeat profiling) were cultured in RPMI 1640 GlutaMax containing 10% FBSH, 50 U/ml penicillin, and 50 mg/ml streptomycin (complete RPMI) and differentiated into adherent cells using 160 nM phorbol myristate acetate (PMA) 48 h prior to infection (36). THP-1 cells (treated with PMA for 24 h) and HBCs (cultured for 48 h) were treated for 24 h with lipopolysaccharide (LPS from Escherichia coli O55:B5; Sigma-Aldrich) and IFN-g (recombinant; R&D systems) at 100 ng/ml and 100 U/ml (5 ng/ml), respectively (31,36).
Gentamicin survival assay. HBCs (0.6 Â 10 6 cells/well) and THP-1 cells (0.5 Â 10 6 cells/well) were plated, in duplicate, in 24-well tissue culture plates (Falcon) and infected at 1 , MOI , 2 for L. monocytogenes and L. innocua (for all infections in this study unless otherwise indicated in the figure legends). Cells were infected for 30 min, followed by washes and treatment with 15 mg/ml gentamicin for the indicated times shown on the figures. For bacterial enumeration, cells were washed with PBS and lysed in water/0.2% Triton X-100. Cell lysates were diluted and plated on BHI/agar plates. Data were expressed as the number of CFU per cell 6 SEM (n = 3 placentas or 3 independent THP-1 cell experiments).
Measuring macrophage infection by fluorescence microscopy. Infected HBCs were fixed in 4% PFA at the indicated time points shown on the figures and labeled with DAPI and mouse anti-CD45 Abs and secondary Abs, and extracellular bacteria versus total bacteria were labeled as previously described (113). A minimum of 15 fluorescence images were randomly acquired in each experimental condition in duplicates (more than 700 cells per condition). To calculate phagocytosis, the numbers of total bacteria (N t ), extracellular bacteria (N e ), and CD45 1 cells (N CD45 ) were enumerated. Entry was expressed as the number of intracellular bacteria per cell as follows: (N t 2 N e )/N CD45 6 SEM (n = 3 placentas). To calculate the percentages of infected HBCs, the number of infected CD45 1 cells (N CD45i ) and the total number of CD45-positive cells (N n ) were enumerated, and the percentage of infected cells was measured at each time point as follows: (N CD45i /N n ) Â 100 6 SEM (n = 3, the same placentas as were used for entry).
LAMP-1 and F-actin labeling. HBCs were stimulated or not with IFN-g/LPS for 24 h and washed and infected with WT L. monocytogenes. Two hours post-infection, cells were fixed with 4% PFA and labeled with A488-conjugated phalloidin (Molecular Probes) or anti-LAMP-1 Ab (clone H4A3; Developmental Studies Hybridoma Bank) for 1 h at RT. Extracellular and total bacteria were labeled with anti-L. monocytogenes antibodies (113). Fluorescence images were acquired, and the percentages of LAMP-1 1 and Factin 1 intracellular L. monocytogenes cells were counted from at least 19 planes (corresponding to a minimum of 500 bacteria) per experimental condition and are expressed as fold change in comparison to untreated HBCs 6 SEM (n = 3 placentas).
Cell-to-cell-spreading assay. HUVECs (8 Â 10 3 cells/well) and trophoblasts (7.5 Â 10 5 cells/well) were cultured in duplicate in 24-well plates to reach 90 to 100% confluence after 72 h of culture. HBCs (2 Â 10 6 cells/well) were seeded in 6-well tissue culture plates for 72 h. HBCs were infected with WT or DactA L. monocytogenes at an MOI of 10 for 1 h, washed with DMEM/F-12 GlutaMax, and incubated in infection DMEM/F-12 (DMEM/F-12 containing 15 mg/ml gentamicin). Five hours post-infection, HBCs were detached using Cellstripper reagent containing 15 mg/ml gentamicin according to the manufacturer's protocol. Five-hundred-microliter amounts of HBCs (3 Â 10 4 ) in the medium of the target cell type were transferred onto each well of trophoblasts or HUVECs at similar concentrations per cell surface area to correct for the difference in the sizes of each cell type. The plates were centrifuged at 360 Â g for 3 min at RT and were incubated at 37°C for 1 h. Cells were washed and incubated with FBSH-and gentamicin-containing medium at 37°C for 18 h. After fixation with 4% PFA, HBCs were labeled with anti-CD14 antibodies conjugated to Alexa Fluor 647 (clone 63D3; Biolegend), HUVECs with Alexa Fluor 488 phalloidin (Molecular Probes), and trophoblasts with anti-E-cadherin Ab (clone HECD-1; abcam) (secondary antibodies were conjugated to Alexa Fluor 488). Following permeabilization with 0.2% Triton X-100, bacteria were labeled with anti-L. monocytogenes antibodies and secondary anti-rabbit antibody Alexa Fluor 568. Nuclei were labeled with DAPI. A minimum of 20 phase-contrast and fluorescence images (22,550 and 840 nuclei for trophoblasts and HUVECs, respectively) were randomly acquired with the 20Â objective from two wells per condition. For trophoblasts and HUVECs, the number of the infectious foci, the area of the infectious foci (A i ), and the total cell surface area (A t ) in each plane were measured using the Metamorph analysis software (Molecular Devices). The results were averaged, and the area covered by foci was expressed as follows: Quantitative fluorescence imaging of HBC surface markers. For quantification of fluorescence intensities, 10 sets of phase-contrast and fluorescence images were randomly acquired for each experimental condition, using the 40Â objective (at least 500 cells per experimental condition). The fluorescence background was subtracted, and a threshold was applied to only measure the cell-associated fluorescence. The number of nuclei (N n ) and the integrated fluorescence intensity (I f ) were automatically measured using Metamorph software. Fluorescence intensity per cell was calculated using the formula (I f /N n ) in each plane, and the results were averaged. Data were shown as average fluorescence intensity (AFI)/nucleus 6 SEM ( Fig. S4 shows a representative experiment of n = 3 independent experiments from 3 placentas). The antibodies were against CD80 (clone 2D10; Biolegend), CD86 (clone IT2.2; Biolegend), CD163 (clone GHI/61; Thermo Fisher Scientific), CD206 (clone 15-2; Biolegend), CD209 (clone 9E9A8; Biolegend), and CD14 (clone 63D3; Biolegend).
Live-cell imaging. HBCs were seeded (3 Â 10 6 ) in glass bottom dishes (P35G-1.5-10-C; MatTek). Cells were infected with pactA-RFP bacteria (MOI of 5) for 30 min, washed, and incubated in infection DMEM/ F-12. Four hours post-infection, cell culture medium was replaced with DMEM/F-12 without phenol red (Gibco) containing 25 mM HEPES and 15 mg/ml gentamicin (cell-imaging medium). Three movies were acquired, each corresponding to a different placenta (n = 3 placentas; these 3 placentas were used for all live-cell-imaging experiments explained in this section), with an average of 100 macrophages in each plane per placenta. Z series of fluorescence images were acquired every 1 min for 120 min using a 40Â oil objective. The percentages of non-motile bacteria were enumerated by counting bacteria that remained in a perimeter of 2 mm for 10 min (after flattening the images using the Z Projection function in MetaMorph). A total of nine 10-min segments from the beginning, middle, and end of the 2-h movies were analyzed, the percentage of non-motile bacteria in each segment was enumerated, and the results were averaged. To measure bacterial speed, images were flattened, and bacterial trajectories were tracked using the Multi-Line function of the MetaMorph analysis software. The speeds (s) of at least 20 individual bacteria per experiment were measured as s = d/t, where d is the distance traveled in mm and t the elapsed time in min. To calculate the bacterial division time, phase-contrast and fluorescence images were acquired from 10 different memorized positions, and images were reacquired 2 h later at the same positions (n = 3 individual experiments from the above-mentioned 3 placentas), 10 planes per experiment. The numbers of red-fluorescent cytosolic bacteria before (B) and after (b) and the elapsed times (t) were measured, and the division (generation) time (G) was expressed as follows: (114). Minimums of 129 and 576 bacteria (per experiment) were counted before and after infection, respectively.
Explant immunofluorescence. Placental explants were obtained and cultured as described previously (115). The villi were sectioned and labeled according to the protocol by Ganesan et al. (116), with primary antibodies (rabbit anti-C7 Ab, mouse anti-CD14 Ab, and rabbit anti-vimentin Ab [clone SP20; abcam] and mouse anti-CD34 Ab [clone QBEND/10; Thermo Fisher]) followed by incubation with secondary Alexa Fluor antibodies. The nuclei were labeled with DAPI.
Cytokine array to determine the inflammatory profile of HBCs. Cell culture supernatants were collected in duplicates and centrifuged for 20 min at 4°C at 14,000 rpm. Meso Scale Discovery (MSD) cytokine arrays were performed in duplicate using supernatants diluted 1:2 with the cell culture medium and using the Meso QuickPlex SQ 120 device. Each cytokine concentration was established based on standard curves obtained from the calibrators provided.
RNA sequencing, generation of differentially expressed gene lists, and data analysis. Cells obtained from 3 placentas (the same ones used for cytokine arrays) were lysed in TRIzol reagent (Invitrogen). The PureLink RNA extraction kit with on-column DNase treatment (Invitrogen) was used to isolate high-quality RNA (extracted separately for each of n = 3 experiments). Input RNA quality and quantity was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) and Qubit fluorometer (Thermo Fisher Scientific), respectively. The RNA integrity number values were greater than 7 and the RNA concentration was greater than 100 ng/ml for all samples. Total RNA-seq libraries were generated with the NEBNext ultra II directional RNA library prep kit for Illumina (catalog number E7760L; NEB) and NEBNext rRNA depletion kit with sample purification beads (catalog number E6350; NEB), with an input amount of 200 ng total RNA per sample. Libraries were pooled and sequenced on an Illumina NovaSeq SP flowcell in paired-end 150-bp format (Illumina, San Diego, CA) to a read yield of between 70 and 80 million reads (equivalent to 35 to 40 million clusters). Raw fastq was aligned to human reference genome GRCh38 with hisat2 version 2.1.0 (117). Alignment QC was assessed by using RSeQC. Gene-wise counts were generated with featureCounts from the subread package version1.5.1 for genes annotated by Ensembl GRCh38.92 (118). An average of 46.6 million pairs of reads were sequenced per sample over two lanes with a 96.2% average overall alignment rate. An average of 89.0% of reads were uniquely mapped, and 76.1% of mapped reads aligned to known exonic regions (coding sequences [CDSs] and untranslated regions [UTRs]). Ribosomal contamination was 0.44%. Counts from separate sequencing lanes for the same sample were summed. Count normalization and differential expression were performed with edgeR using a negative binomial distribution with a generalized linear model (GLM) with a paired design (119). Genes were tested if at least half of the samples had an expression of 2 CPM (counts per million). Genes with an absolute Log 2 -fold change (Log 2 FC) of .1 with a false discovery rate (FDR) of ,0.05 were considered significant and are referred to as differentially expressed (DE) genes. R packages ComplexHeatmap and EnhancedVolcano were used for the heatmaps and volcano plots, respectively. The Gene Ontology (GO) analyses were performed using DAVID 6.8 (120,121). Upregulated and downregulated DE genes were uploaded separately as induced and repressed gene lists, respectively. The program uses a list of DE genes and does not use Log 2 FC or FDR. For the biological processes (BP), the GOTERM_BP_DIRECT terms were used by setting the EASE score at 0.05 and selecting the pathways with an FDR of ,0.05. The same criteria were used to select for the KEGG pathways. The pathways were manually grouped and color-coded based on their functional categories. The Venn diagrams were drawn using the online tool at http://bioinformatics.psb.ugent .be/webtools/Venn/.
Statistical analyses. Summary statistics, i.e., mean values 6 SEM, are shown in the figures. All data were the average values of at least 3 independent experiments, corresponding to at least 3 different placentas. All graphs and statistical analyses were generated using GraphPad Prism 7 and 8. For independent data, Student's t tests were used for analysis of two-group comparisons. Analysis of variance (ANOVA) or a mixed-effect model was used for analysis of multiple-group comparisons. Tukey's or Sidak's method was used to control for multiple comparisons when needed. A P value of ,0.05 after adjustment for multiple comparisons was considered statistically significant. P values and statistical methods used for each experiment are indicated in the corresponding figure legends.
Data availability. The RNA-seq raw data and the Log 2 transcripts per million (TPM) table are available under GEO accession number GSE174689.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
Research reported in this publication was supported by the Institute of Allergy and Infectious Diseases of the National Institutes of Health under award numbers R21AI105588, R03AI149371, R01AI107250, and R01AI157205 and by CTSA grant number UL1TR001070. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
We acknowledge the precious guidance received from John M. Robinson (The Ohio State University School of Medicine) and Yoel Sadovski (University of Pittsburgh) for trophoblast and villous tissue isolation and Seth Guller's laboratory (Yale School of Medicine) for Hofbauer cell isolation. All our thanks to Bella Cho for drawing the cartoon presented in Fig. 1. We are thankful to Pearlly Yan (OSU) for her guidance on RNA sequencing processes and analyses. We are grateful to Jerzy Kulski (UWA Medical School, University of Western Australia) for the insightful information about human MHC-II genes and their regulation of expression. Finally, we thank Hannah Cackovic, Sarah Buck-Davis, Devra Mast, and Jessica Russo for their participation in obtaining consent from participants and collecting biological specimens.