Histone deacetylase 9 deficiency exaggerates uterine M2 macrophage polarization

Abstract The maternal‐foetal interface is an immune‐privileged site where the semi‐allogeneic embryo is protected from attacks by the maternal immune system. Uterine macrophages are key players in establishing and maintaining pregnancy, and the dysregulation of the M1‐M2 subpopulation balance causes abortion. We separated two distinct mouse uterine macrophage subpopulations during early pregnancy, CD45+F4/80+CD206− M1‐like (M1) and CD45+F4/80+CD206+ M2‐like (M2) cells. The M1 preponderance was significantly exaggerated at 6 hours after lipopolysaccharide (LPS) treatment, and adoptive transfer of M2 macrophages partially rescued LPS‐induced abortion. RNA sequencing analysis of mouse uterine M2 versus M1 revealed 1837 differentially expressed genes (DEGs), among which 629 was up‐regulated and 1208 was down‐regulated. Histone deacetylase 9 (Hdac9) was one of the DEGs and validated to be significantly up‐regulated in uterine M2 as compared with M1. Remarkably, this differential expression profile between M1 and M2 was also evident in primary splenic macrophages and in vitro polarized murine peritoneal, bone marrow–derived and RAW 264.7 macrophages. In Hdac9/HDAC9 knockout RAW 264.7 and human THP‐1–derived macrophages, the expression of M1 differentiation markers was unchanged or decreased whereas M2 markers were increased compared with the wild‐type cells, and these effects were unrelated to compromised proliferation. Furthermore, Hdac9/HDAC9 ablation significantly enhanced the phagocytosis of fluorescent microspheres in M2 Raw 264.7 cells yet decreased the capacity of THP‐1‐derived M1 macrophages. The above results demonstrate that Hdac9/HDAC9 deficiency exaggerates M2 macrophage polarization in mouse and human macrophages, which may provide clues for our understanding of the epigenetic regulation on macrophage M1/M2 polarization in maternal‐foetal tolerance.


| INTRODUC TI ON
The maternal-foetal interface is a privileged site for co-ordinating the process of immune tolerance to protect the genetically foreign semi-allogeneic embryo from attack by the maternal immune system. The immune cells that reside in the decidua, which surrounds the placenta and conceptus, are highly specialized and important for the establishment of immune microenvironment during pregnancy.
Macrophages are the second most abundant immune cell population in the pregnant uterus (known as decidua), comprising approximately 20% of the total decidual leucocyte population. 1 and/or IL13, represent two ends of a functional differentiation spectrum. In response to infectious pathogens, macrophages undergo M1 activation and induce Th1 immunity, secreting pro-inflammatory cytokines, nitric oxide (NO), reactive oxygen species, proteolytic enzymes, etc On the other hand, macrophages also have the ability to control immune responses, producing anti-inflammatory cytokines such as IL10 and factors promoting tissue remodelling. 2,3 Yet, it is also increasingly recognized that other polarization phenotypes of macrophages also exist in a tissue-specific manner, due to their high diversity, plasticity and heterogeneity. For example, M2 macrophages can be subdivided into M2a, M2b, M2c and M2d, and functions of macrophages in inflammation versus immune regulation are not completely attributable to one or the other subset. 1 Similarly, dMϕs are differentiated or polarized into two distinct subpopulations, M1 and M2. 4 Appropriately and timely regulated M1/M2 polarization has been considered a key player in establishing and maintaining pregnancy during different phases of gestation, and the dysregulation of this balance is correlated with recurrent spontaneous abortion (RSA) in humans. 5,6 They are polarized towards the predominant M1 phenotype during the embryo implantation window, switch to a balanced M1/M2 profile during placenta development and uterine vasculature remodelling, and shift towards M2 polarization for pregnancy maintenance. 7 Different dMϕ subsets with distinct functional properties have been identified by flow cytometric studies. In the first-trimester human decidua, for example, dMϕs can be discriminated into CD14 + ICAM-3 + and CD14 + ICAM-3 − subsets. The CD14 + ICAM-3 − population expresses high levels of CD163, CD206, CD209 and NRP-1 and low levels of CD11c, displaying more pronounced M2 phenotype. In contrast, the CD14 + ICAM-3 + population is CD163-, CD206-, CD209-and NRP-1negative and expresses high levels of CD11c. 8 Thus, expression of ICAM-3 and CD11c correlates well with each other. 8 Another report shows that two distinct subsets of CD14 + CD11c hi and CD14 + CD11c lo express genes associated with inflammation and extracellular matrix formation, respectively; however, these two subpopulations secret both pro-inflammatory and anti-inflammatory cytokines, therefore do not fit the conventional M1/M2 profile. 9 With respect to mouse uterine macrophages, two abundant populations of F4/80 + MHCII hi and F4/80 + MHCII lo which differentially express M1 and M2 markers 10 have been identified.
Given that dMϕs contribute to the balance that establishes maternal-foetal tolerance, it is important to understand how their tolerogenic phenotypes are induced in a juxtacrine or paracrine manner. Human placenta-derived M-CSF and IL10 can induce dMϕ to differentiate towards homeostatic CD14 + CD163 + CD206 + CD209 + M2 phenotype, producing IL10 and CCL18. 11 Soluble human leucocyte antigen G5 (sHLAG5) secreted by trophoblasts reduces the expression of M1 marker CD86 and increases the expression of M2 marker CD163 and the phagocytic activity. 12 Human trophoblast and decidual stromal cells-secreted RANKL (nuclear factor-κB ligand) polarizes dMϕ towards M2 phenotype via activating AKT/ STAT6 (signal transducer and activator of transcription 6) signalling, and depletion of RANKL results in abnormal phenotypes of dMϕ in vivo and increased foetal loss rates in mice. 13 In light of these important findings, the mechanisms of dMϕ differentiation in the local homeostatic microenvironment at the maternal-foetal interface remain to be further explored. plus an open chromatin conformation, respectively. 14 Among these epigenetic modifications, histone acetylation is mediated by histone acetyltransferases, which use acetyl-CoA to modify ε-amino group of lysine residues on histone proteins, resulting in the elimination of positive charges of lysines and a relaxed chromatin conformation and mostly associated with increased transcriptional activity. On the other hand, histone deacetylases (HDACs) remove the acetyl group from the lysine residues in the N-terminal tails of nucleosomal core histones, mainly resulting in a more compact chromatin conformation and repression of gene transcription. To date, in humans and mice, four classes of HDAC have been characterized, including class I (HDAC1, 2, 3, 8), class IIa (HDAC4, 5,7,9), class IIb (HDAC6, 10) and class IV (HDAC11). 15 The class I members show homology to yeast RPD3 and normally localized in the nucleus. The class II HDACs share similarity with yeast HDA1 and have C-terminal nuclear export signal, with the class IIa possessing myocyte enhancer factor (MEF)2-interacting domain at the N-terminus, and the class IIb possessing tandem deacetylase domains. The class IV is unique in that it only has deacetylase domain. Accumulating evidence suggests the involvement of Hdacs in driving macrophage differentiation. For instance, in Hdac3 knockout (KO) macrophages, almost half of the inflammatory genes, such as Ifnb1-and Stat1-dependent genes, fail to be activated in response to LPS, and inflammatory response is ameliorated. 16,17 Furthermore, Hdac3 −/− macrophages show an IL4induced M2 phenotype and are hypersensitive to IL4 stimulation, due to the release of deacetylation at regulatory loci of many IL4regulated genes. 17 Despite the epigenetic regulatory mechanisms in macrophage polarization, information regarding dMϕs in the local environment at the maternal-foetal interface is limited.
In the present study, we have separated uterine

| Mice
Specific pathogen-free (SPF) inbred BALB/c mice at 8-10 weeks were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (VRL, Beijing, China). The mice were bred and maintained in a temperature-and humidity-controlled room with a constant photoperiod (light/dark = 12:12 hours). The protocols for animal studies were approved by the Committee on the Ethics of Animal Experiments of the Institute of Zoology, Chinese Academy of Sciences. Pregnancy was achieved by caging female mice with a male mouse at a 2:1 ratio, and the day when a copulatory plug was found was referred to as gestational day 1 (gd1).

| Establishment of a low-dose LPS-induced mouse abortion model
Mice on gd6 were injected intraperitoneally (ip) with LPS (L3024, Sigma-Aldrich) at 0.5, 1.2, 1.5, 2 or 4 μg reconstituted in 100 μL of phosphate-buffered saline (PBS) per mouse (n = 3 in each group). At 24 hours after LPS administration, mice were killed by cervical dislocation and abortion was evaluated. The status of the foetuses was considered as death (abnormally shaped or haemorrhagic sacs) or resorption (small or pale sacs with no discernible foetus). The minimum dose of LPS that caused over 80% of abortion rate was determined.
Subsequently, mice on gd6 were ip injected with the minimum dose of LPS and killed at 6, 12, 16, 18, 21, 24 and 30 hours thereafter.
The time point at which over 80% of abortion rate was observed was determined as the most optimal time for uterine tissue sampling.

| Isolation of mouse primary splenic and uterine macrophage cells
Primary splenic and uterine macrophage cells were freshly isolated from BALB/c mice. Briefly, spleens or uteri were removed from ab-
The libraries were sequenced on an Illumina HiSeq 2500 platform as paired-end 150 bp reads at Annoroad Gene Technology (Beijing, China; http://www.annor oad.com). Each step was strictly in accordance with transcriptome sequencing criteria.
After filtering low-quality reads and those containing adapters with Trimmomatic, HISAT2 was used for building the genome index, and clean data were then mapped to the reference genome with default parameters. The reference genomes and the annotation file were downloaded from ENSEMBL database (http://www. ensem bl.org/index.html). Read count for each gene in each sample was counted by HTSeq, and the abundance of each transcript in each sample was defined by FPKM (fragments per kilobase per million mapped reads). Sequencing results of two biologically repeated transcriptomes were synthetically analysed. DESeq2 was used for differential gene expression analysis. Genes with |log2FC| ≥ 1 and q < 0.05 were identified as differentially expressed genes (DEGs).
Hierarchical clustering was applied to cluster DEGs. Genes were clustered together by different distance calculated by log2FPKM of each gene. The enrichment of genes in gene ontology (GO) terms compared with the background genes and of DEGs in Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway between the two uterine macrophage subsets was implemented by Fisher's exact test, in which P-value was adjusted by multiple comparisons as qvalue. GO terms or KEGG terms with q < 0.05 were considered to be significantly enriched.

| Isolation of murine peritoneal macrophages (PM) and bone marrow-derived macrophages (BMM), RAW 264.7 cell culture and M1/M2 polarization
The BALB/c mice were killed by cervical dislocation and shortly immersed in 70% ethanol for sterilization. After injecting 15 mL PBS into the peritoneal cavity, PM cells were collected, washed with icecold PBS and cultured in DMEM complete medium for 4 hours in a humidified 37°C and 5% CO 2 incubator to obtain M0 cells. 18 The bone marrow cells were flushed from femurs and tibia of BALB/c mice, collected and cultured with DMEM complete medium for 3 hours. The non-adherent cells were collected and treated with 50 ng/mL of M-CSF for 5 days. Medium was changed every 2 or 3 days to generate mature BMM at the M0 state.

| RNA isolation and real-time quantitative reverse transcription PCR (qRT-PCR) analysis
Total RNA was extracted using TRIzol reagent, and reverse transcription was performed with 1 µg RNA and M-MLV reverse transcriptase (28025-021, Thermo Fisher Scientific) according to the manufacturer's instructions. cDNA was subjected to quantitative real-time PCR using primers listed in Table S1   Guide RNA sequences (Table S2)  Homozygous KO clones were confirmed by genomic PCR genotyping using primers spanning the deleted exon followed by DNA sequencing of PCR products directly or after T-vector cloning, by qRT-PCR using primers (Table S2) within or spanning the deleted exon, and/or by immunoblotting analysis. Three or four independent vector control or KO single-cell clones were analysed for each experiment. were visualized with a Gene Gnome XRQ Chemiluminescence detector (Syngene) and GAPDH was used as a protein loading control.

| MTT assay
Cells were seeded in 96-well plates at an optimal concentration and cultured for the indicated time periods. 15 μL 5 mg/mL of MTT reagent (M8180, Solarbio, Beijing, China) was added into each well, followed by an additional 4 hours of incubation at 37 C. The formazan blue product that formed in the cells was dissolved by adding 150 μL of dimethyl sulphoxide (DMSO). The optical density was measured at 570 nm. Data are expressed as means ± SEM of four replicate measurements. The OD values were determined at 540 nm using a microplate reader, and data are expressed as means ± SEM of five replicates.

| Phagocytosis assay
Cells were incubated with carboxylate-modified fluorescent latex beads with a mean diameter of 2 μm (L3030; Sigma-Aldrich, 1:400 dilution) for 4 hours, as described previously. 20 After incubation, the supernatant was discarded and the cells were trypsinized and washed three times with ice-cold PBS. Cells were fixed with 4% formaldehyde, and the percentage of intracellular fluorescent beads was analysed on a BD FACSCalibur flow cytometer.

| Statistical analysis
All experiments were repeated as indicated in the figure legends, and n indicates the number of independent biological repeats. Data are presented as the means ± SEM. Student's t test was used to evaluate the differences between two groups, and analysis of variance (ANOVA) was performed when more than two groups were compared. *P < .05; **P < .01. For all statistical tests, P < .05 was considered statistically significant.

| The dynamic balance of M1 and M2 uterine macrophages during mouse early pregnancy was disrupted after LPS treatment
As mentioned, mouse dMϕs are predominantly M1 phenotype during embryo implantation and switched to a balanced M1/M2 profile afterwards. We initiated to choose gd6, the end of the pu-  Figure 1A). Next, we used a low-dose LPS-induced abortion model (0.5 μg per mouse) 21,22 to investigate the dynamic changes of M1 and M2 macrophages after LPS treatment for 1, 3, 6, 12 and 24 hours. The results showed an overall increasing tendency of M1 cells, a decreasing trend of M2 cells and an increase in M1/M2 ratio, and significant differences between LPS-injected mice and the control group were observed at 6 hours and afterwards ( Figure 1B,C, and Figure S1). This indicates that M1 macrophages are the predominant phenotype in the decidua following treatment with LPS, and this imbalance occurs before abortion.

| The adoptive transfer of M2 macrophages can effectively rescue LPS-induced abortion
In order to study the abortion mechanism caused by M1/M2 im-  Figure 2D). In parallel, tail intravenous injection of the same amount of M1 or M2 macrophages had no effects in normal pregnant mice, presumably because the dose of macrophages was insufficient to reverse or to change pregnancy outcome ( Figure 2E).  Figure 4A). However, up-regulation of these marker genes was not satisfactory upon treatment with IL4 only, inconsistent with the results from previous studies. 24,25 In THP-1 cells, expression profiles of M2 markers did not show any obvious difference either between single and double cytokines, or between different induction time ( Figure 4B). Therefore, in the following studies, M1-and M2-type macrophage polarization in RAW 264.7 and THP-1 cells was established by stimulation with 100 ng/mL LPS + 20 ng/mL IFNG, and with 20 ng/mL IL4 + 20 ng/mL IL13, respectively, for 24 hours.

| RNA-seq analysis of DEGs in mouse uterine M1 versus M2 on Day 6 of pregnancy (gd6)
We next adopted bulk RNA-Seq to investigate DEGs between M1 and M2 macrophage subpopulations in mouse uterus on gd6. Figure   S2A shows the FACS-sorted M1 (CD45 + F4/80 + CD206 − ) and M2 macrophages (CD45 + F4/80 + CD206 + ) from gd6 mouse uteri. Two biological replicates were performed in both M1 (CD206-negative, CD206N) and M2 groups (CD206-positive, CD206P). The total reads and mapping efficiency of each sample are shown in Table S3, and F I G U R E 1 Flow cytometry analysis of uterus M1 and M2 macrophages and dynamic changes of two macrophage subsets after treatment with LPS. A, Gating strategy used to identify M1-like (CD45 + F4/80 + CD206 − cells) and M2-like (CD45 + F4/80 + CD206 + cells) macrophages in uterine and splenic tissues from BALB/c mouse on gd6. Total leucocytes (CD45 + cells) were gated within the leucocyte gate using forward light scatter (FSC) versus side scatter (SSC). B, The proportions of the two macrophage subsets at 1, 3, 6, 12 and 24 h after LPS (0.5 µg/100 µL PBS) ip injection on gd6 by flow cytometry analysis. C, Statistical analysis of the percentages of M1 or M2 cells, and ratio of M1/M2 macrophages at 1, 3, 6, 12 and 24 h following LPS or PBS treatment (experimental data shown in panel B). Data are shown as line graphs (means ± SEM. **P < .01, *P < .05, n = 3) the count of detected genes in each sample is illustrated in Figure   S2B. We identified 1837 DEGs, among which 629 was up-regulated in M2 versus M1 and 1208 was down-regulated ( Figure S2B (Table S6)

| Hdac9/HDAC9 KO exaggerates expression of M2 marker genes in both Raw 264.7 cells and THP-1derived macrophage cells
We next performed qRT-PCR to determine the expression level of M1-and M2-type macrophage polarization markers within F I G U R E 2 Effects of adoptive transfer of M1 or M2 cells on low-dose LPS-induced abortion or normal pregnancy in BALB/c mice. A, Different doses of LPS (0.5, 1.2, 1.5, 2 or 4 μg) were ip injected on gd6. The abortion was observed at 24 h post-LPS treatment. B, Each mouse was given 0.5 μg LPS on gd6, and the abortion was analysed at 6, 12, 16, 18, 21, 24 and 30 h after LPS administration. C, Statistical analysis of the number of aborted embryos at different time points. D, Effect of adoptive transfer of M1 or M2 macrophage on early abortion induced by LPS. Representative phenotypes are shown. 1 × 10 6 M1 and M2 macrophages cells were transferred on gd3 via tail vein injection, and 0.5 µg LPS was administrated on gd6. A same volume of PBS was given to control groups. Uteri were collected 21 h after LPS or PBS injection. E, Representative phenotypes of adoptive transfer of M1 or M2 cells into normal pregnant mice. NORM, normal pregnancy; NORM_M1, normal pregnancy with adoptive M1 transfer; NORM_M2, normal pregnancy with adoptive M2 transfer. In (D and E), the statistical analysis of the number of aborted embryos was shown below (data were expressed as means ± SEM. **P < .01). Arrows indicate the site of abortion Hdac9/HDAC9 KO Raw 264.7 or THP-1-derived macrophages. In Raw 264.7 cells, the Hdac9 KO macrophages expressed significantly higher levels of M2 markers Cd206 and Pparg, while the levels of M1 markers Cd86, Tnfα and Il6 remained unchanged ( Figure 6A).
Interestingly, HDAC9-deficient THP-1-derived macrophages expressed significantly higher levels of M2 markers CD206 and CD209, whereas expressing lower levels of M1 markers TNFα, IL6 and CXCL10 ( Figure 6B). Therefore, these observations implicated that  Figure 6C) and that no effect was observed in THP-1 cells ( Figure 6E). Therefore, the exaggeration of the expression of M2 markers after Hdac9 ablation was unlikely caused by a decreased proliferation or viability potential at the M0 state.

| Hdac9 KO had no effect on LPS/IFNGinduced M1-type pro-inflammatory NO release, but increased the phagocytosis capacity of M2-type macrophage in RAW 264.7 cells
We further performed NO assay and phagocytosis assay to detect functional phenotypes in Hdac9 KO clones distinguishable from wild-type (WT; empty vector) controls. There was no significant change in NO release after Hdac9 KO compared with WT controls ( Figure 6D). NO production in THP-1-differentiated macrophages was undetectable.
Phagocytosis plays a crucial role in macrophage-mediated host defence, which leads to internalization and distraction of pathogens.
Raw 264.7 and THP-1 cells have been demonstrated to be capable of phagocytosing red fluorescent latex beads of 2 μmol/L size by fluorescence measurements. [26][27][28][29] To determine whether Hdac9 KO affected the phagocytosis of M2-polarized macrophages, we examined the internalization of red fluorescent beads by FACS ( Figure 7A,C,D) and real-time live-cell confocal imaging ( Figure 7B, Supporting Video).
The data showed that M2 cells after Hdac9 deletion had higher capacity of uptaking latex beads in RAW 264.7 macrophages ( Figure 7A).
Although THP-1-derived M2 macrophages exhibited similar particle uptake rates between WT and KO cell lines ( Figure 7C), HDAC9 KO M1 macrophages showed decreased phagocytic capacity for latex beads ( Figure 7D). These results further support the tendency towards M2 differentiation after Hdac9/HDAC9 ablation.

| D ISCUSS I ON
F4/80, a well-known surface marker of mouse macrophage, has also been known as a marker for discriminating or FACS sorting macrophages in mouse uterus. 10,30,31 Two subsets of uterine Mϕs have been recognized, F4/80 + MHCII − and F4/80 + MHCII + . 30 The former subsets are defined as undifferentiated macrophages dependent on ovarian steroid hormones for maintenance, of which 70%-80% express CD11b, but hardly express class A scavenger receptor, macrosialin or sialoadhesin. The latter subsets are defined as mature macrophages, half of which express CD11b, class A scavenger receptor, macrosialin and sialoadhesin. 30 Additional studies also show that subpopulations of mouse dMϕs, defined by and Cd206, M2 phenotype markers associated with angiogenesis, tissue remodelling and repair, whereas MHCII hi Mϕs show a trend towards higher expression of M1 markers. 10 Although some studies have also proposed that F4/80 and CD11b are used together to FACS-sort other tissue-resident macrophages 32 and bone marrowderived monocytes/macrophages, 33,34 CD11b is more often referred as the monocyte marker, especially in the decidua. 35 Based on these literatures, we used CD45 and F4/80 as the surface sorting marker to separate macrophages.
CD206 has been characterized as the marker of tissue M2 macrophages in both humans and mice, and M1 macrophages are devoid of it. 36 In mice, aortic macrophage population has been divided into two distinct subpopulations, F4/80 + CD206 − iNOS + (M1) and F4/80 + CD206 + iNOS − (M2). 37 Mouse M1 and M2 BMMs are defined as F4/80 + CD11b + CD206 − iNOS + and F4/80 + CD11b + CD206 + iNOS − , respectively, after stimulation by polymer wear particles used in severe end-stage arthritis. 34 In addition, F4/80 + CD11c + CD206 − M1 and F4/80 + CD11c − CD206 + M2 cells reside in mouse epididymal fat tissue. 38 Yet, in the uterus or the decidua, although it has been recognized that human dMϕs express more CD206 as compared with peripheral blood macrophages, 39 there is no report on whether CD206 can be used as the surface sorting marker to discriminate M1 and M2. In this study, we firstly utilized antibodies against the surface epitope of CD206 and FACS-sorted two distinct macrophage subpopulations, CD45 + F4/80 + CD206 − and CD45 + F4/80 + CD206 + , and defined them as M1 and M2 cells, respectively. Indeed, F4/80 + CD206 − (M1) and F4/80 + CD206 + (M2) have also been characterized in mouse pancreas 40 and infracted myocardium. 41 It has been reported that adoptive transfer of Tregs, [42][43][44] CD25+Foxp3+ NK cells 45 and Tim-3+ peripheral NK cells 46 can protect against foetal loss in abortion-prone mice. Adoptive transfer of F I G U R E 3 Establishment of in vitro M1/M2 polarization models in mouse PM, BMM, RAW 264.7 cell line and human THP-1 monocytederived macrophages. A, Cells were classically activated to M1 condition with 100 ng/mL LPS + 20 ng/mL IFNG for 24 h and alternatively activated to M2 condition with 20 ng/mL IL4 + IL13 for 24 h, respectively. Morphological changes were recorded by light microscopy. For THP-1 monocytes, cells were firstly differentiated into macrophages by 24-or 48- RANK+ Mϕs, which are supposed to be more adaptive to M2 phenotype than RANK − Mϕs, on gd5, reverses embryo absorption in pregnant C57BL/6 mice with macrophage depletion. 13 Our data proved that LPS-induced abortion was associated with the predominant M1 phenotype and that adoptive transfer of M2 cells partially rescued low-dose LPS-induced abortion in mice. As abortion is a shift in the immunological response from Th2 to Th1 domination, our results correlate with lines of evidence that adoptive transfer of M2 macrophages with immunosuppressive properties is an effective treatment of chronic pro-inflammatory conditions in rodents. 47,48 As aforementioned, Hdacs mainly perform repressive function for gene transcription. Our results showed that Notably, Class IIa HDAC inhibitors attenuate inflammation in mouse and human macrophages, 49,53 stabilize atherosclerotic plaques in mice and limit the expression of inflammatory factors IL-1β and IL-6 in monocytes from patients with atherosclerosis, a chronic arterial inflammatory condition. 53 Therefore, it is reasonable to consider that high-level expression of HDAC9, a class IIa HDAC member, in M2 is involved in suppressing key genes that drive M2 differentiation.
Hdac9 KO mice develop age-dependent cardiac hypertrophy and heart failure, 54 polydactyly, 55 and are resistant to colitis, 56 obesity and glucose intolerance during high-fat feeding. 57 Hdac9-deficient mice also exhibit Th2 polarization in effector T cells via increased accumulation of H3, H3K9Ac, H3K14Ac and/or H3K18Ac at the promoters of Il4, Roquin and Pparg accompanied by increased expression of these genes in spleen and kidney. 58  and HDAC8 knockdown suppresses M2 marker genes via activating ERK signalling pathway in THP-1-derived macrophages. 61 HDAC2 is down-regulated in the peripheral blood monocytes/macrophages from patients with gestational diabetes mellitus, which is characterized by high serum levels of pro-inflammatory cytokines, and HDAC2 inhibition aggravates the secretion of pro-inflammatory cytokines in the monocytes/macrophages. 62 The effects of suppression of these two class I HDAC members on M1/M2 marker gene expression are not consistent with our results, suggesting distinct functions of different HDAC members in the local immunological environment at the maternal-foetal interface. We speculate that HDAC9 mainly functions as an epigenetic brake in uterine M2 macrophages by suppressing the expression of M2 marker genes, for instance, CD206, CD209 and PPARG, and loss of HDAC9 thereby releases the brake and exaggerates M2 phenotype. It also remains possible that deacetylation of non-histone proteins is mechanistically involved. In this perspective, our future studies will be aimed at defining the key HDAC9 target genes, including HDAC9-marked enhancers/promotors and HDAC9 partners, during macrophage differentiation via high-throughput omics approaches.
Accumulating evidence also reveals the function of HDAC9 in other inflammatory responses. HDAC9 deficiency alleviates the release of iNOS, cyclooxygenase-2 (COX-2), IL-1β, IL-6, TNFα and IL-18 and suppresses inflammation in mouse brain via inactivating IkBa/ NF-kB and MAPKs signalling pathways. 63 Prevented inflammation or decreased cytokine production by HDAC9 silencing or inhibition is also evident in splenocytes/kidneys due to PPARG overexpression, 58 as well as in colitis via increasing Foxp3 + T regulatory cell function 56 in mice. On the other hand, inflammation or LPS can drive the F I G U R E 6 Effect of Hdac9/HDAC9 ablation on the expression of M1/M2 markers, proliferation and NO production of macrophages. A, In Raw 264.7 cells, compared with WT controls, expression of M1 markers Cd86, Tnfα and Il6 mRNA remained unchanged and M2 markers Cd206 and Pparg were significantly increased in Hdac9 KO cells. B, In THP-1-derived macrophages, compared with WT controls, expression of M1 markers TNFα, IL6, CXCL10 were significantly down-regulated, and M2 markers CD206 and CD209 were significantly up-regulated in KO cells. n = 3 biological replicates per group. *P < .05. C, MTT assay of WT and Hdac9 KO Raw 264.7 cells at 24, 48, 72 and 96 h of proliferation. D, Effect of Hdac9 KO on LPS/IFNG-induced NO production analysed by the Griess reaction assay in RAW 264.7 cells. E, MTT assay of WT and HDAC9 KO THP-1 cells at 24, 48, 72 and 96 h of proliferation. Values are means of five replicates (n = 5) ± SEM. *P < .05 F I G U R E 7 Effect of Hdac9/HDAC9 deletion on the phagocytic capacity of RAW 264.7 and THP-1-derived macrophages by flow cytometry analysis. A, Phagocytosis assay of M2-type RAW 264.7 cells. B, Real-time live-cell confocal recording of the process of polarized M2 RAW 264.7 cell phagocytosis of latex beads. C, Phagocytosis assay of M2-type THP-1-derived macrophages. D, Phagocytosis assay of M1-type THP-1-derived macrophages. The lower panels in A, C and D show the statistical summary of the percentage of phagocytic cells. **P < .01, *P < .05 expression of HDAC9, triggering the activation of NF-kB-dependent inflammatory cytokines in microglial cells. 63 HDAC9 is also upregulated after ischaemic brain injury which is associated with exacerbating inflammation in rats. 64 Of note, genome-wide association study identifies a variant in HDAC9 associated with large vessel stroke, which would be consistent with the association with accelerating atherosclerosis. 65 Overall, HDAC9 is an important epigenetic inflammatory mediator regulating the inflammatory gene expression programme, and targeting HDAC9 could be an effective strategy for ameliorating inflammation. Besides, pharmacologic HDAC inhibitors have considerable therapeutic benefits as anti-inflammatory and immunosuppressive drugs in treating cancer, infectious and immunological diseases, etc 15,66 Based on available literatures, it is perplexing to understand the diverse HDAC functions. Our findings provide insights into Hdac9 as an epigenetic factor in maintaining immune homeostasis at the maternal-foetal interface. It is therefore tempting to speculate the opportunities of using HDAC9 inhibitors as potential therapeutic strategies of inflammation-related abortion.

ACK N OWLED G EM ENT
We are grateful to Prof. Jingpian Peng for his valuable guidance and great supports for this project and for allowing us to use the data generated in his laboratory as the foundation of our studies. We

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data relevant to the study are included in the article. The data that support the findings of this study are available from the corresponding authors upon request.