Selective down-regulation of SMAD3 in cDCs

To examine the expression of SMAD2 and SMAD3 in cDCs and their progenitors, we isolated LMPPs as Lin⁻Sca-1⁺CD34⁺CD117⁺CD135⁺ cells; the DC progenitor cells: MDPs as Lin⁻CD117^hiCD135⁺CD115⁺Sca-1^-, and CD115⁺ CDPs as Lin⁻CD117^intCD135⁺CD115⁺CD127^- (Fogg et al, 2006; Onai & Ohteki, 2016), SiglecH⁻Ly6C⁻/SiglecH⁻Ly6C⁺ CD11c⁺MHCII⁻CD135⁺CD172α⁻ pre-DCs (Schlitzer et al, 2015), and differentiated cDCs: BM CD11b⁺CD11c⁺ cDCs and splenic CD11b⁺CD11c⁺ cDCs of C57BL/6 mice. We also examined their expression patterns in BM-derived DCs (BMDCs): GM-CSF plus IL-4–induced BMDCs that yield CD11b⁺ cDCs and FLT3L-induced BMDCs that yield both cDCs and pDCs (Yamaguchi et al, 1997; Waskow et al, 2008). SMAD2 mRNA was expressed in all examined cells, whereas SMAD3 mRNA was reduced to almost undetectable levels in CD115⁺ CDPs, SiglecH⁻ pre-DCs, CD11b⁺CD11c⁺ cDCs, and CD11c^hi BMDCs (Fig 1A). Immunoblotting confirmed that the SMAD2 protein (60 kD) is kept expressed, whereas the SMAD3 protein (52 kD) expressed in whole BM was reduced to an undetectable level in GM-CSF plus IL-4–induced BMDCs (Fig S1A). Immunocytochemistry determined by the proximity ligation assay (PLA; Söderberg et al, 2006) showed that the SMAD3 protein was not detected in CD115⁺ CDPs, SiglecH⁻ pre-DCs, and CD11c⁺ cells, whereas the SMAD2 protein was expressed in all examined subsets (Fig 1B and C). cDCs are further classified into type 1 cDCs (cDC1s) and type 2 cDCs (cDC2s) with distinct features and functions (Murphy et al, 2016; Ginhoux et al, 2022; Zhang et al, 2023). We confirmed that sorted CD8a⁺ (cDC1s) or CD11b⁺ (cDC2s) cells out of CD11c⁺ cells were both negative for SMAD3 (Fig 1B and C).

The expression of SMADs is generally ubiquitous and constitutive in normal cells (Brown et al, 2007; Miyazawa et al, 2024). We confirmed that SMAD2 and SMAD3 proteins were expressed in macrophages, naïve and activated CD4⁺ T cells (Fig S1B).

We screened the expression patterns of SMAD2 and SMAD3 in immune cells using representative open public data sources (ImmGen consortium: Heng et al, 2008; Yoshida et al, 2019; UCSC Cell Atlas: Domínguez Conde et al, 2022). Microarray (Fig S2A) and RNA-seq (Fig S2B) of murine immune cells and their precursor cells, as well as human single-cell RNA-seq (Fig S2C), show that the expression of SMAD3 mRNA is specifically down-regulated in mouse and human cDC subsets, whereas SMAD2 mRNA is expressed in all immune cells.

Our data validated by the open public data sources show that SMAD3 is selectively and specifically down-regulated in cDCs and its progenitors: CD115⁺ CDPs and SiglecH⁻ pre-DCs.

SMAD3 deficiency enhances cDC differentiation between the MDP and CD115⁺ CDP in vivo

We next examined the roles of SMAD3 in cDC differentiation in vivo using Smad3^−/− mice. Immunophenotyping was performed using flow cytometry according to the published gating procedures (Liu et al, 2020; Fig S3A). The numbers of cDC subsets: CD11c⁺, CD11b⁺CD11c⁺, and CD8⁺CD11c⁺ cells in BM, spleens, and superficial and mesenteric lymph nodes of Smad3^−/− mice were significantly increased compared with those of littermate control Smad3^+/+ mice (Fig 2A).

We next traced upstream cDC progenitors. A significantly increased CD115⁺ CDP (Lin⁻CD117^int) along with a decreased MDP (Lin⁻CD117^hi) was observed in the BM of Smad3^−/− mice compared with that of the littermate control (Fig 2B). DNGR-1 (encoded by the Clec9a gene and known as CLEC9A and CD370)–positive CDPs are cDC-restricted (Schraml et al, 2013; Cabeza-Cabrerizo et al, 2021). CX3CR1 is expressed on the MDP and cDC-P (Liu et al, 2009). Expressions of Cx3cr1 and Dngr1 mRNA in Lin⁻CD115⁺ BM cells and the proportion of CX3CR1⁺CD370⁺Lin⁻CD117^intCD135⁺CD115⁺ BM cells were significantly increased in Smad3^−/− mice compared with the littermate control (Fig 2C). SiglecH⁻Ly6C⁺ and SiglecH⁻Ly6C⁻ pre-DCs with a cDC potential (Schlitzer et al, 2015) were significantly increased in the BM of Smad3^−/− mice (Fig 2D). In support of our findings, ImmGen consortium data demonstrate down-regulation of SMAD3 mRNA in the BM CDP in addition to CD11b⁺ and CD8⁺ DCs (Fig S2A).

We confirmed that the proportions of the haematopoietic progenitor cells detected as c-Kit⁺Lin⁻Sca-1⁺ (KLS) or CD34⁺ KLS cells, LMPPs, common myeloid progenitor as Lin⁻Sca-1⁻CD16/32⁻CD34⁺CD117⁺, and granulocyte–macrophage progenitor as Lin⁻Sca-1⁻CD16/32⁺CD34⁺CD117⁺ were unaltered in the BM of 8-wk-old Smad3^−/− mice compared with littermate control Smad3^+/+ mice bred in the specific pathogen-free environment before the onset of any signs of inflammation (Yang et al, 1999; Yoon et al, 2015) (Fig S3B).

Immunophenotyping of DC progenitor subsets of Smad3-deficient mice suggests that SMAD3 deficiency facilitates cDC differentiation at the developmental stage between the MDP and CD115⁺ CDP in the steady state.

SMAD3-mediated TGF-β signalling inhibits cDC differentiation

We examined the direct effects of SMAD3 on cDC differentiation using FLT3L-induced BMDCs and GM-CSF plus IL-4–induced BMDCs (Xu et al, 2007) transfected with either SMAD3 DNA or SMAD3 siRNA 4 h before culture. Expression levels of FLAG-tagged SMAD3 were confirmed by flow cytometry (Fig S4A) and immunocytochemistry using the PLA (Fig S4B). SMAD3 mRNAs in the transfected BMDCs were confirmed by quantitative RT–PCR (RT–qPCR; Fig S4C). Flow cytometry data showed that the forced overexpression of SMAD3 resulted in a significantly decreased CD115⁺ CDP in both FLT3L-induced or GM-CSF plus IL-4–induced BMDCs (Fig 3A, upper contour plots, and Fig S4D). In contrast, knockdown of SMAD3 resulted in a significantly increased CD115⁺ CDP in both FLT3L-induced or GM-CSF plus IL-4–induced BMDCs (Fig 3A, lower contour plots, and Fig S4D) in consistent with the findings in Smad3-deficient mice (Fig 2B). The overexpression of SMAD3 reduced MHCII⁺CD11c⁺, CD11b⁺CD11c⁺, and CD24⁺CD11c⁺ cells (Naik et al, 2005) (Fig 3B, upper contour plots, and Fig S4F), whereas knockdown of SMAD3 increased MHCII⁺CD11c⁺, CD11b⁺CD11c⁺, and CD24⁺CD11c⁺ cells (Fig 3B, lower contour plots, and Fig S4F) with the gating procedures (Fig S4E). Transfection of pcDNA or control siRNA reduced the proportions of the cells highly positive for MHCII and CD11c; however, net percentages of MHCII⁺CD11c⁺ were not altered by transfection (Fig 3B and D).

SMAD3-knocked-down BMDCs developed significantly more dendrite formation, which is the most characteristic morphological feature of the cDC (Steinman & Cohn, 1973), whereas SMAD3-overexpressed BMDCs did not develop dendrite formation (Fig 3C). SMAD2 and SMAD3 are the R-SMADs shared by TGF-β and activin among TGF-β superfamily cytokines (Batlle & Massague, 2019). Activin A (10 ng/ml) showed no effect on FLT3L or GM-CSF plus IL-4–induced BMDC differentiation (Fig 3D, lowest contour plots, and Fig S4G). A high concentration of TGF-β1 (5 ng/ml) completely blocked FLT3L or GM-CSF plus IL-4–induced Smad3^+/+ BMDC differentiation, which was abolished in Smad3^−/− BMDCs (Fig 3D, middle contour plots, and Fig S4G), indicating that the potent inhibitory effect of high-dose TGF-β on DC differentiation is SMAD3-dependent. These data show that SMAD3-mediated TGF-β signalling inhibits cDC differentiation.

SMAD3-mediated TGF-β signalling down-regulates cDC-related genes

To identify the target genes of SMAD3 to inhibit cDC differentiation, we screened essential cytokines, their signalling molecules, and the transcription factors for cDC differentiation: Flt3, Csf2ra, Pu.1, Gfi1, Irf2, Irf4, Irf8, Id2, Batf3, and RelB (Zhang et al, 2023), as well as TGF-β1, TGF-β2, and TGF-β3 in FLT3L-induced BMDCs and GM-CSF plus IL-4–induced BMDCs transfected with either SMAD3 DNA (black circles) or control pcDNA (white circles) (Figs 4A and B and S5). CD11b⁺CD11c⁺ cells were sorted from FLT3L-induced and GM-CSF plus IL-4–induced BMDCs before cell lysis. We found that the overexpression of SMAD3 significantly suppressed the mRNA expression of Flt3, Id2, and Irf4 in FLT3L-induced BMDCs (Fig 4A) and suppressed the mRNA expression of Flt3 and Irf4 in GM-CSF plus IL-4–induced BMDCs (Fig 4B). SMAD3 did not affect the Id2 mRNA expression in GM-CSF plus IL-4–induced BMDCs (Fig S5), which are comprised of both cDCs and monocyte-derived macrophages (Helft et al, 2015).

We examined the requirement of SMAD3 for TGF-β1 to suppress these identified cDC-related genes: Flt3, Id2, and Irf4 (Takagi et al, 2011). TGF-β1 suppressed the mRNA expression of these cDC-related genes in a dose-dependent manner in both FLT3L-induced and GM-CSF plus IL-4–induced Smad3^+/+ BMDCs, which was abolished in Smad3^−/− BMDCs (Fig 4C and D).

To confirm the effect of SMAD3 on the expression of the identified cDC-related genes in vivo, we compared the mRNA expression levels of Flt3, Id2, and Irf4 in cDCs and their progenitor cells in the BM and spleens of Smad3^−/− mice (black bars) and the littermate control Smad3^+/+ mice (white bars) (Fig 4E). Smad3^−/− CD11b⁺ cDCs and SiglecH⁻Ly6C⁺ pre-cDCs expressed significantly higher levels of Flt3, Id2, and Irf4. Smad3^−/− CD115⁺ CDPs expressed higher levels of Id2 and Irf4. Smad3^−/− SiglecH⁻Ly6C⁻ pre-DCs and CD8⁺ cDCs expressed higher levels of Flt3 and Id2 compared with the Smad3^+/+ littermate control (Fig 4E).

These data indicate that SMAD3-mediated TGF-β signalling inhibits cDC differentiation by repressing cDC-related genes such as Flt3, Id2, and Irf4.

STAT3 and c-SKI repress the transcription of SMAD3 for cDC differentiation

We next investigated the mechanisms whereby SMAD3 is down-regulated for cDC differentiation. The ligand of CD115, M-CSF, induces STAT3 activation in macrophages (Novak et al, 1995). STAT3 as the signalling molecule of FLT3L and GM-CSF (Onai et al, 2006; Li & Watowich, 2013; Wan et al, 2013) is essential for FLT3L-responsive DC progenitor proliferation (Laouar et al, 2003). Therefore, we examined the effect of STAT3 on SMAD3 expression in BMDCs transfected with STAT3 or control pcDNA and siSTAT3 or control siRNA. Expression levels of STAT3 mRNA in the transfected BMDCs were confirmed by RT–qPCR (Fig S6A). The overexpression of STAT3 suppressed, whereas knockdown of STAT3 up-regulated the expression of Smad3 mRNA in both FLT3L-induced and GM-CSF plus IL-4–induced BMDCs (Fig 5A).

We examined whether and how STAT3 regulates the Smad3 gene promoter activity using the Smad3 gene promoter luciferase reporter construct spanning 2 kb upstream of the first exons of the Smad3 gene transfected in FLT3L-induced or GM-CSF plus IL-4–induced BMDCs. SMAD2, SMAD3, and SMAD4 synergistically induced the Smad3 promoter activity (Fig 5B, white bars), which was suppressed by STAT3 in both FLT3L-induced or GM-CSF plus IL-4–induced BMDCs (Fig 5B, light grey bars). To identify a corepressor of STAT3, we screened the representative transcriptional repressors of R-SMADs: SKI/SnoN and TGIF (Deheuninck & Luo, 2009; Heldin & Moustakas, 2012; Batlle & Massague, 2019). Among them, c-SKI showed the synergy with STAT3 to repress the Smad3 promoter activity (Fig S6B). The overexpression of c-SKI exerted the repressive effect on the Smad3 promoter activity, which was strengthened in synergy with STAT3 (Fig 5B, thick grey and black bars). Knockdown of c-SKI by siRNA completely abolished the repressive effect of STAT3 on the SMAD2/3/4-induced Smad3 promoter activation (Fig 5C, black bars). In contrast, c-SKI alone retained the repressive effect on the SMAD2/3-induced Smad3 promoter activation when STAT3 was knocked down, although the repressive effects of c-SKI and STAT3 were more effective in synergy (Fig 5D).

We performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) with CD11b⁺ FLT3L-induced BMDCs to identify the enriched loci for DNA-binding STAT3 and c-SKI in association with the histone modification status within the Smad3 coding and flanking regions (Fig 5E). ChIP-seq showed that the whole Smad3 coding and flanking regions were epigenetically inactive, being highly enriched in trimethylated histone H3 lysine 27 (H3K27me3), which is associated with repression by polycomb group complexes. Transcriptionally active epigenetic marks such as acetylated lysine 23 (H3K23Ac) and trimethylated histone H3 lysine 4 (H3K4me3) were minor, but there were some bivalent sites enriched in both H3K27me3 and H3K4me3. Bivalent chromatin modification state balances the expression or repression of the important regulatory genes during cell differentiation (Macrae et al, 2023). We found that STAT3 and c-SKI bound to the bivalent sites within the proximal region of the Smad3 gene. We then analysed the de novo enriched motifs using HOMER and searched the JASPAR CORE database. A canonical DNA-binding motif of STAT3 is TTCnnnGAA (Seidel et al, 1995). A reported DNA-binding motif of c-SKI is GTCTAGAC in chicken embryo fibroblasts (Nicol & Stavnezer, 1998). A de novo motif analysis showed that the STAT3-binding motif contained half-sequences: TTCC (Hutchins et al, 2013) (Fig 5F), and the c-SKI–binding motif contained the GTCTAG element (Fig 5G) in CD11b⁺ FLT3L-induced BMDCs.

ChIP–qPCR for the proximal promoter region of the Smad3 gene validated the results of ChIP-seq, showing that STAT3 and c-SKI bound to the same sites with SMAD2 in CD11b⁺ FLT3L-induced BMDCs (−1,196 to −1,003 and −220 to −28) (Fig 5H), which were epigenetically inactive with H3K27me3 (Fig 5I).

These data indicate that STAT3 in synergy with c-SKI represses canonical SMAD pathway–induced transcription of the Smad3 gene for cDC differentiation.

c-SKI is required for STAT3 to interact with SMAD2 in cDCs

As a consequence of repression of SMAD3 by STAT3 and c-SKI, SMAD2 is the remained R-SMAD in cDC precursors and cDCs. We sought to confirm the physiological interactions among STAT3, c-SKI, and SMAD2 in MDPs and cDC precursors sorted from BM, cDCs sorted from BM and spleens, and FLT3L-induced or GM-CSF plus IL-4–induced BMDCs using the PLA. The PLA showed the close proximity between c-SKI and STAT3 and the close proximity between c-SKI and SMAD2 in CD115⁺ CDPs, SiglecH⁻ pre-DCs, CD11c^hi BMDCs, and CD11b⁺CD11c⁺ splenic cDCs, which was not observed in MDPs (Fig 6A). Knockdown of c-SKI by siRNA abolished the interaction between SMAD2 and STAT3 (Fig 6B), whereas knockdown of STAT3 had no effect on the interaction between c-SKI and SMAD2 in CD11c^hi BMDCs (Fig 6C). We found that SMAD2 was C-terminally phosphorylated, indicating the presence of TGF-β ligand–bound activated type I receptor signalling. Taken together, c-SKI is required for STAT3 to interact with C-terminally phosphorylated SMAD2 in cDCs.

Interaction of phosphorylated STAT3 with c-SKI is essential for repression of SMAD3 in cDCs

SMAD2 and SMAD3 interact with the N-terminal region of SKI, whereas SMAD4 interacts with the SAND-like domain of SKI, which blocks the ability of the SMAD complexes to activate transcription of TGF-β target genes (Akiyoshi et al, 1999; Suzuki et al, 2004; Takeda et al, 2004; Batlle & Massague, 2019). STAT3 is phosphorylated at the C-terminal tyrosine residue (Y705) and serine residue (S727) upon stimulation with cytokines, protein tyrosine kinase receptors, or intracellular protein tyrosine kinases (Hillmer et al, 2016). Therefore, we investigated the mechanisms whereby c-SKI and STAT3 repress transcription of the Smad3 gene using the Smad3 gene promoter luciferase reporter assay with various combinations of mutants of c-SKI and STAT3 transfected in CD11b⁺ FLT3L-induced or GM-CSF plus IL-4–induced BMDCs.

A mutant of c-SKI that does not interact with SMAD2/3 (Δ2/3) failed to repress the Smad3 promoter activity, whereas a mutant of c-SKI that does not interact with SMAD4 (W274E) (Wu et al, 2002; Nagata et al, 2006) retained the repressive effect on the Smad3 promoter activity (Fig 7A). Inactive mutants of STAT3 at Y705 and S727 residues, Y705F and S727A, respectively, abolished the repressive effect on SMAD2/3-induced Smad3 promoter activation in both CD11b⁺ FLT3L-induced and GM-CSF plus IL-4–induced BMDCs (Fig 7B).

We next examined how binding capacity of c-SKI with SMADs and phosphorylation status of STAT3 affect cDC differentiation using CD11b⁺ FLT3L-induced or GM-CSF plus IL-4–induced BMDCs transfected with various combinations of mutants of c-SKI and STAT3. Expression levels of FLAG-tagged c-SKI and mutants were confirmed by flow cytometry (Fig S6C) and immunocytochemistry using the PLA (Fig S6D). The overexpression of full-length c-SKI enhanced, whereas knockdown of c-Ski by siRNA suppressed, cDC differentiation (Fig 7C). Transfection of Δ2/3 failed to induce cDC differentiation, whereas transfection of W274E enhanced cDC differentiation, as well as full-length c-SKI (Fig 7C). In the same manner as c-Ski, the overexpression of STAT3 enhanced, whereas knockdown of STAT3 by siRNA suppressed cDC differentiation (Fig 7D). The enhancing effect of STAT3 on cDC differentiation was abolished by mutations of Y705F and S727A.

These data demonstrate that phosphorylated STAT3 at Y705 and S727 interacts with c-SKI and that SMAD2, but not SMAD4, is required to repress transcription of the Smad3 gene for cDC differentiation.

Mice

Age-matched female Smad3^{+/+, −/−} mice (8 wk, six mice/genotype) were maintained and used for experiments according to the ethical guidelines for animal experiments and the safety guidelines for gene manipulation experiments at the University of Tsukuba, Japan, Tokyo Medical University, Japan, and Konkuk University, Korea, under approved animal study protocols. No statistical method was used to predetermine sample sizes.

Generation of mouse BMDCs

BM cells were flushed from femurs and tibias. BM cells counted using a Countess 3 automated cell counter (AMQAX2000; Thermo Fisher Scientific) were cultured in RPMI 1640 (SH30255.01; GE Healthcare), supplemented with 10% heat-inactivated FBS (12483-020; Gibco), 50 μM 2-mercaptoethanol (M314A; Sigma-Aldrich), 1 mM sodium pyruvate (S8636; Sigma-Aldrich), and 1% penicillin/streptomycin (SV30010; GE Healthcare), containing mouse GM-CSF (20 ng/ml; 315-03; PeproTech) plus IL-4 (20 ng/ml; 214-14; PeproTech) or FLT3L (200 ng/ml; 250-31L; PeproTech) at 1 × 10⁶ cells/ml at 37°C with or without TGF-β1 (0.1, 0.2, 0.5, 1.0, and 5.0 ng/ml; 7666-MB; R&D Systems) or activin A (10 ng/ml; 120-14; PeproTech) for 7–8 d. For time course analyses (Fig 4A and B), BMDCs were cultured for 1–8 d.

Flow cytometry

Mouse BMDCs, BM, spleens, and superficial and mesenteric lymph node cells were incubated on ice for 30 min with optimal concentrations of anti-mouse CD16/CD32 antibody (clone 2.4G2), biotin-conjugated mouse lineage panel (559142; BD Pharmingen), the fluorochrome-conjugated antibodies against anti-biotin streptavidin, anti-mouse CD34 (clone RAM34), anti-mouse CD127 (clone SB/199), anti-mouse CD117 (c-Kit) (clone 2B8), anti-mouse Ly-6A/E (Sca-1) (clone D7), anti-mouse CD115 (c-fms) (clone AFS98), anti-mouse CD135 (FLT3) (clone A2F10), anti-mouse CD370 (clone Clec9A), anti-mouse CX3CR1 (clone SA011F11), anti-mouse Ly6C (clone HK1.4), anti-mouse MHC Class II (clone I-A/I-E) (clone M5/1 114.15.2), anti-mouse CD11c (clone N418), anti-mouse CD8a (clone 53-6.7), anti-mouse CD24 (clone M1/69), anti-mouse CD11b (clone M1/70), anti-mouse CD172α (clone P84) and anti-mouse SiglecH (clone 551), and anti-tag FLAG (clone 8H8L17). All antibodies were purchased from BD Pharmingen, eBioscience, Invitrogen, or BioLegend. Dead cells were excluded by 7AAD (559925; BD Pharmingen). The samples were acquired by LSRFortessa (BD Bioscience) and analysed by FlowJo V10 (BD Bioscience). Gating procedures for flow cytometry data were performed according to the published protocol (Liu et al, 2020).

Cell isolation

Spleens and LNs were chopped and digested with type III collagenase (LS004182; Worthington Biomedical Corporation) and DNase I (11284932001; Roche) in RPMI 1640 containing 10% FBS for 30 min (spleens and LNs) at 37°C as described previously (Inaba et al, 2009). EDTA (5 mM; T4049; Sigma-Aldrich) was added for the final 5 min. Cell numbers were counted using a Countess 3 automated cell counter (Thermo Fisher Scientific). CD8⁺CD11c⁺ cells (130-091-169; Miltenyi Biotec), CD11b⁺CD11c⁺ cells (130-049-601, 130-108-338; Miltenyi Biotec), CD115⁺ Lin⁻ BM cells (130-096-354; Miltenyi Biotec), CD11b⁺ cells, and CD4⁺ T cells (130-049-601, 130-104-454; Miltenyi Biotec) were enriched using the MACS magnetic sorting system (Miltenyi Biotec). Purity was confirmed as >85%. To isolate LMPP, MDP, and CD115⁺/CD115⁻ CDP cells, Lin⁻ cells were presorted from mouse BM using the mouse lineage cell depletion kit (130-090-858; Miltenyi Biotec) and MACS system (Miltenyi Biotec). Lin⁻ cells were stained with fluorochrome-conjugated anti-mouse CD117 (clone 2B8), Sca-1 (clone D7), anti-mouse CD34 (clone RAM34), CD127 (clone SB/199), CD115 (clone AFS98), and CD135 (clone A2F10) antibodies for cell sorting using FACSAria III (BD Bioscience). To isolate pre-DCs, BM cells were stained with fluorochrome-conjugated anti-mouse CD11c (clone N418), MHC Class II (clone I-A/I-E) (clone M5/1 114.15.2), CD135 (clone A2F10), CD172α (clone P84), SiglecH (clone 551), and Ly6C (clone HK1.4) antibodies for cell sorting using FACSAria III. CD11b⁺CD11c⁺ cells were sorted from FLT3L-induced BMDCs and GM-CSF plus IL-4–induced BMDCs before cell lysis. All antibodies were purchased from BD Pharmingen, eBioscience, or BioLegend.

Quantitative RT–PCR

The total RNA was extracted using TRIzol according to the manufacturer’s instructions (15596-026; Invitrogen). The RNA was reverse-transcribed with a cDNA RT kit (18080200; Invitrogen). The mouse cDNA was quantitated by SYBR Green (4368708; Applied Biosystems) using ABI 7900 (Applied Biosystems). The following primers are listed in the table (Table S1). The relative mRNA levels to GAPDH were calculated by the comparative Ct method. Each experiment was performed in triplicate.

Western blotting

Cells lysed in RIPA buffer were electrophoresed on 10% SDS–polyacrylamide gel, transferred to nitrocellulose membrane (Millipore), and probed with antibodies against anti-mouse SMAD2 (clone D43B4; Cell Signaling Technology), anti-mouse phospho-SMAD2 (S465/467) (clone E8F3R; Cell Signaling Technology), anti-mouse SMAD3 (clone EP568Y; Abcam), phospho-SMAD3 (S423/425) (clone EP823Y; Abcam), and β-actin (clone N-21; Santa Cruz Biotechnology). Blots were visualized using an ECL kit (GE Healthcare). Three independent experiments were performed.

Immunocytochemistry

Lin⁻CD117^hiCD135⁻CD115⁻CD127⁻, Lin⁻CD117⁺Sca1⁺CD34⁺CD135⁺, Lin⁻CD117^hiCD135⁺CD115⁺CD127⁻Sca-1⁻, Lin⁻CD117^intCD135⁺CD115⁺CD127⁻, CD11c⁺MHCII⁻CD135⁺ CD172α⁻SiglecH^-Ly6C⁻, CD11c⁺MHCII⁻CD135⁺CD172α⁻SiglecH^-Ly6C⁺, CD8⁺CD11c⁺, CD11b⁺CD11c⁺ BM cells, CD8⁺CD11c⁺, CD11b⁺CD11c⁺ splenic cells, and FLT3L or GM-CSF plus IL-4–induced BMDCs were fixed with 4% PFA in phosphate-buffered saline for the May–Grunwald/Giemsa staining or PLA (Söderberg et al, 2006). For the PLA, fixed cells were permeabilized by 0.1% Triton X-100 in PBS, and stained with rabbit antibodies against SMAD2, SMAD3, phospho-SMAD2 (S465/467), phospho-SMAD3 (S423/425) (Cell Signaling Technology), STAT3 (clone C-20; Santa Cruz Biotechnology), c-SKI (clone 6D763; Santa Cruz Biotechnology), and FLAG (clone 8H8L17; Invitrogen). Subsequently, they were reacted with the Duolink In Situ PLA probe anti-rabbit PLUS (DUO92002; Sigma-Aldrich) and PLA probe anti-rabbit MINUS (DUO92005; Sigma-Aldrich), or PLA probe anti-mouse PLUS (DUO92001; Sigma-Aldrich) and PLA probe anti-mouse MINUS (DUO92004, Sigma-Aldrich), and signals were detected using In Situ Detection Reagents Red (DUO92008; Sigma-Aldrich) according to the manufacturer’s fluorescence instructions. CD11c, CD11b, or CD8 cells were stained with Alexa Fluor 488 (green). The nucleus was stained with DAPI (blue). Slides were observed using a confocal microscope, LSM700 and LSM900, with a 374-nm axial resolution at 550 nm (Carl Zeiss). Signals in each field (n = 10 fields) were quantified using BlobFinder software V3.2 (Uppsala University). Five independent experiments were performed.

Transfection

BM cells were transfected with 200 nM of a non-targeting siRNA (D-001810-01-05; Dharmacon), SMAD3 siRNA (L040706-00-0005; Dharmacon), STAT3 siRNA (L-040794-01-0005; Dharmacon), c-SKI siRNA (L-042265-01-0005; Dharmacon), empty pcDNA3 STAT3, STAT3 Y705F, S727A (submitted by J. Darnell; Addgene), FLAG-tagged SMAD2, FLAG-tagged SMAD3, FLAG-tagged SMAD4, FLAG-tagged c-SKI, FLAG-tagged c-SKI (Δ2/3 or W274E), FLAG-tagged TGIF, and FLAG-tagged SnoN in nucleofection buffer for 4 h before FLT3L-induced or GM-CSF plus IL-4–induced BMDC culture using Nucleofector I/II/2b (program Y-001), 4D-Nucleofector (program DK-100), and Amaxa Nucleofector kits for mouse dendritic cells (VPA-1011 and V4XP-4012; Lonza) with ∼60–75% transfection efficiency according to the manufacturer’s protocol. Knockdown efficiencies by siRNA were confirmed by RT–qPCR. Transfection efficiencies of FLAG-tagged plasmids were confirmed by detection of FLA using flow cytometry (Figs S4A and S6C) and immunocytochemistry by the PLA (Figs S4B and S6D), with ∼55–60% and 75–80% of transfection efficiencies, respectively. Each experiment was performed in triplicate, and three independent experiments were performed.

Luciferase assay

The 2,000-bp promoter region of SMAD3 was generated by PCR from genomic DNA using primers (Table S2). The product was verified by sequencing and was subcloned into a pGL3 basic firefly luciferase construct (E1751; Promega) using MluI and XhoI sites. The promoter constructs in various combinations with STAT3, STAT3 Y705F, S727A, FLAG-tagged SMADs, FLAG-tagged c-SKI, c-SKI (Δ2/3 or W274E), FLAG-tagged TGIF, FLAG-tagged SnoN, p300, empty pcDNA3, or non-targeting siRNA, STAT3 siRNA, c-SKI siRNA (Dharmacon RNA Technologies) plasmid were transfected in FLT3L-induced or GM-CSF plus IL-4–induced BMDCs with a control TK-pRL Renilla luciferase control reporter plasmid (E2231; Promega). Five to seven days after transfection, BMDCs were lysed by Dual-Luciferase Reporter Assay Kit (E1910; Promega), and the lysate was measured by a luminometer (Berthold Technologies). Each experiment was performed in triplicate.

Chromatin immunoprecipitation

Chromatin was prepared from BMDCs. Immunoprecipitation was performed with antibodies against SMAD2 (Cell Signaling Technology), SMAD3 (Abcam), c-SKI, STAT3 (Santa Cruz Biotechnology), trimethyl histone H3K4 (clone C42D8), trimethyl histone H3K27 (C36B11; Cell Signaling Technology), and acetyl histone H3K23 (17-10112; Millipore) using Pierce Magnetic ChIP Kit (26157; Thermo Fisher Scientific) according to the manufacturer’s protocol. Immunoprecipitated DNA released from the cross-linked proteins was quantitated by ABI 7900 (Applied Biosystems) using the primers (Table S3) and was normalized to input DNA. Each experiment was performed in triplicate.

ChIP sequencing

The construction of library was performed using NEBNext Ultra DNA Library Prep Kit for Illumina (New England Biolabs) according to the manufacturer’s instructions. Briefly, the chipped DNA was ligated with adaptors. After purification, PCR was done with adaptor-ligated DNA and index primer for multiplexing sequencing. The library was purified using magnetic beads to remove all reaction components. The size of the library was assessed by Agilent 2100 Bioanalyzer (Agilent Technologies, Amstelveen). High-throughput sequencing was performed as paired-end 100-bp sequencing using NovaSeq 6000 (Illumina, Inc.). The results were analysed by Integrated Genome Browser 10.0.0 (bioviz.org). De novo motifs were identified from the STAT3 and c-SKI ChIP-seq–binding sites using HOMER de novo motif analysis.

Statistics

Data were analysed using a two-tailed unpaired t test. A P-value < 0.05 was considered to indicate statistical significance.