Key Points
PU.1 and IRF4 are required for PD-L2 expression in dendritic cells.
PU.1 and IRF4 transactivate the Pdcd1lg2 gene via direct binding to an EICE sequence.
PU.1 is involved in the p300-mediated histone acetylation of the Pdcd1lg2 gene.
Visual Abstract
Abstract
PD-L2, which has been identified as a PD-1 ligand, is specifically expressed in dendritic cells (DCs) and macrophages. The transcription factors that determine the cell type-specific expression of PD-L2 are largely unknown, although PD-1 and its ligands, which have been shown to play important roles in T cell suppression, have been vigorously analyzed in the field of cancer immunology. To reveal the mechanism by which Pdcd1lg2 gene expression is regulated, we focused on DCs, which play key roles in innate and acquired immunity. The knockdown of the hematopoietic cell–specific transcription factors PU.1 and IRF4 decreased PD-L2 expression in GM-CSF–induced mouse bone marrow–derived DCs. Chromatin immunoprecipitation assays, luciferase assays, and electrophoretic mobility shift assays demonstrated that PU.1 and IRF4 bound directly to the Pdcd1lg2 gene via an Ets-IRF composite element sequence and coordinately transactivated the Pdcd1lg2 gene. Furthermore, PU.1 knockdown reduced the histone acetylation of the Pdcd1lg2 gene. The knockdown of the typical histone acetyltransferase p300, which has been reported to interact with PU.1, decreased the expression and H3K27 acetylation of the Pdcd1lg2 gene. GM-CSF stimulation upregulated the Pdcd1lg2 gene expression, which was accompanied by an increase in PU.1 binding and histone acetylation in Flt3L-generated mouse bone marrow–derived DCs. The involvement of PU.1, IRF4, and p300 were also observed in mouse splenic DCs. Overall, these results indicate that PU.1 positively regulates Pdcd1lg2 gene expression as a transactivator and an epigenetic regulator in DCs.
Introduction
The Pdcd1lg2 gene encodes PD-L2 (CD273, B7-DC), which has been identified as the second ligand of programmed cell death-1 (PD-1). Although numerous studies have demonstrated the anticancer effects of checkpoint inhibitors targeting PD-1 and/or PD-L1 (another ligand of PD-1), the biological significance of PD-L2 and the relationship between PD-L2 and its target molecules is still unclear. Recently, it was reported that PD-L2 also binds to RGMb (DORAGON) (1, 2), and the binding of PD-L2 to RGMb or PD-1 suppresses T cell activation (3), induces Th1 responses (4), and regulates Ab production by B-1 cells (5). Furthermore, the involvement of PD-L2 in malaria and asthma was also demonstrated in studies using mouse models (6, 7). These findings suggest the possibility that PD-L2, which contributes to several events in adaptive immunity, may be a therapeutic target for immune-related diseases.
When PD-L2 and PD-L1 are compared, their gene expression profiles are shown to be strikingly different. PD-L1 is ubiquitously expressed, whereas the expression of PD-L2 is highly restricted to APCs, especially monocytes (8). In previous reports of the transcriptional regulation of the Pdcd1lg2 gene, it was revealed that IL-4 and IL-13 transactivate the Pdcd1lg2 gene via STAT6 in activated macrophages and dendritic cells (DCs) (9, 10). Because STAT6 is a ubiquitous molecule, the mechanism by which the Pdcd1lg2 gene is expressed in a monocyte-specific manner is still unknown. Therefore, it is necessary to identify the transcription factor(s) involved in cell type-specific Pdcd1lg2 gene expression.
The hematopoietic cell–specific transcription factor PU.1, which is encoded by the Spi1 gene, belongs to the Ets-family and is essential for gene expression and the development of lymphoid and myeloid cells. A Spi1 gene knockout mouse exhibited severe immunodeficiency, including the incomplete development and production of DCs, macrophages, B cells, T cells, NK cells, and neutrophils (11–16). PU.1 transactivates its target genes not only as a monomer that binds to Ets motifs but also as a dimer with IRF4 or IRF8 that binds to Ets-IRF composite element (EICE) sequences. Previously, studies, including ours, showed that PU.1 plays a critical role in the expression of several specific genes important for the function of DCs (17–22). Based on these observations, we hypothesized that PU.1 is involved in the cell type-specific expression of the Pdcd1lg2 gene and investigated the molecular mechanism by which PU.1 regulates the Pdcd1lg2 gene in the current study.
Epigenetic regulation is involved in the determination of cell type-specific gene expression. In particular, conformational changes in chromatin resulting from histone acetylation mediated by histone acetyltransferase (HAT) and histone deacetylation mediated by histone deacetylase, resulting in transactivation and suppression, respectively, are often observed in various cell types. Although PU.1 is known to interact with p300 and GCN5, which are typical HATs (23, 24), whether PU.1 affects histone acetylation in the Pdcd1lg2 gene in DCs has not been clarified so far.
In the current study, we identified PU.1 as a transcription factor that regulates the Pdcd1lg2 gene via cis-enhancing elements in a cooperative manner with IRF4 in DCs, and we also found that PU.1 modulates the p300-mediated histone acetylation of the Pdcd1lg2 gene.
Materials and Methods
Mice and cells
Bone marrow–derived DCs (BMDCs) were generated from BALB/c mice purchased from Japan SLC (Hamamatsu, Japan) as previously described (17, 19). CD11c+ DCs were isolated by using a MACS separation system with anti-mouse CD11c MicroBeads and an autoMACS (all from Miltenyi Biotech). All animal experiments were performed in accordance with the approved guidelines of the Institutional Review Board of Tokyo University of Science, Tokyo, Japan. The human embryonic kidney cell line HEK293T was maintained as previously described (21).
Knockdown by small interfering RNA introduction
Small interfering RNAs (siRNAs) for mouse Spi1 (Spi1-MSS257676), Irf4 (MSS205500, MSS205501), Irf8 (MSS236847), Ep300 (MSS220766), and control siRNA (Stealth RNAi siRNA Negative Universal Control) from Invitrogen was introduced into DCs with Nucleofector 2b (Lonza) using an Amaxa Mouse Dendritic Cell Nucleofector Kit (Lonza), as previously described (21).
Quantitative RT-PCR
The purification of total RNA and the synthesis of cDNA were performed using a ReliaPrep RNA Cell Miniprep System (Promega) and ReverTra Ace qPCR RT Master Mix (Toyobo, Osaka, Japan), respectively. The mRNA levels of Pdcd1lg2, Spi1, Irf4, Ep300 and Gapdh were determined with a StepOne Real-Time PCR System (Applied Biosystems) using THUNDERBIRD Probe qPCR Mix (Toyobo) and the TaqMan Gene Expression Assays (Mm00488142_m1 for mouse Spi1 and 4352329E for rodent Gapdh; Applied Biosystems); mRNA levels were also determined using THUNDERBIRD SYBR qPCR Mix (Toyobo) with the following synthesized oligonucleotides as primer sets: mouse Pdcd1lg2 (forward primer, 5′-GAACCTGAGCTTACAACTTCATCCT-3′; reverse primer, 5′-ACGTCTACGGTGTACACTTCTTTAGG-3′), mouse Irf4 (forward primer, 5′-CCCCATTGAGCCAAGCATAA-3′; reverse primer, 5′-GCAGCCGGCAGTCTGAGA-3′), and mouse Ep300 (forward primer, 5′-CCAAGCGCCTGCAAGAA-3′; reverse primer, 5′-TATCCTTGTAGTCATGGACAATACGTT-3′).
Flow cytometry
Flow cytometry for the detection of cell surface MHC class II (MHCII), CD11b, CD11c, and PDCA-1 was performed as described in our previous report (21, 22). An FITC-conjugated anti-mouse PD-L2 Ab (MIH37; Miltenyi Biotech) was used to stain the BMDCs and Flt3L-BMDCs. The intracellular staining was performed as previously described (25).
Western blotting analysis
Western blotting analyses using anti-PU.1 Ab (D-19; Santa Cruz Biotechnology), anti-IRF4 Ab (M-17; Santa Cruz Biotechnology), and β-actin (AC-15; Sigma-Aldrich) were performed as previously described (22).
Overexpression of PU.1 and IRF4
Expression plasmids, PU.1 (26), pIRES2-AcGFP1-Myc-IRF4 (generated as follows), and its control plasmid pIRES2-AcGFP1 (Clontech Laboratories), were introduced into BMDCs by an electroporation with Nucleofector 2b. GFP+ cells were judged to be the plasmid-carrying cells.
Mouse IRF4 cDNA isolated from pCR3.1-IRF4 (19) was inserted into the EcoRI/KpnI site of pCMV-Myc-N (Clontech Laboratories) to generate pCMV-Myc-IRF4, and the BglII/SalI–digested Myc-IRF4 cDNA fragment from pCMV-Myc-IRF4 was inserted into pIRES2-AcGFP1 to obtain pIRES2-AcGFP-Myc-IRF4.
Chromatin immunoprecipitation assay
Chromatin immunoprecipitation (ChIP) assays were performed as previously described (22). Anti-PU.1 Ab (D-19), anti-IRF4 Ab (M-17) (Santa Cruz Biotechnology), and anti-acetyl histone H3(K27) Ab (MABI0309; MEDICAL & BIOLOGICAL LABORATORIES) were used. Goat IgG (02-6202; Invitrogen) and mouse IgG1 (02-6100; Invitrogen) were used as isotype control Abs. The amount of precipitated DNA was determined by quantitative PCR using an Applied Biosystems StepOne Real-Time PCR System. The nucleotide sequences of the primer sets used for the PCR of acetyl histone H3(K27) were as follows: mouse Pdcd1lg2 gene CNS0 (forward primer, 5′-GGCAGCTGACAAGAACAATGG-3′; reverse primer, 5′-AGGCTCTCTCAAGCCGCTTA-3′), CNS1 (forward primer, 5′-TGTGGGAGGCAGGAGGAA-3′; reverse primer, 5′-CTTGGACCTTCAAACCAATGG-3′), CNS2 (forward primer, 5′-GTCCTAATGACTCCATCCCTAAGC-3′; reverse primer, 5′-ACACCTGTAGGACATTGCTGACA-3′), CNS3 (forward primer, 5′-CGCAGAGTGGATTTGAAACAAA-3′; reverse primer, 5′-CAGGGAGAAAAGTGACTAAATCAGAA-3′), and CNS4 (forward primer, 5′-AGAAACCTGGAGGCAGAAGCT-3′; reverse primer, 5′-CAGCGGGCCAGTGAACAG-3′). The nucleotide sequences of the primer sets used for the PCR of PU.1 and IRF4 were as follows: CNS2 forward primer, 5′-CGGCCTTACTTCCTAATTTCAAAC-3′; reverse primer, 5′-CTCTGCCCCCATCACATAGTG-3′) and CNS3 (forward primer, 5′-TGACAGCCAGCCCTGAGAA-3′; reverse primer, 5′-TCACAGAGGAAATGAAACCACTTG-3′). The primer sets for CNS0, CNS1, and CNS4 were same as those used for acetyl histone H3(K27).
Electrophoretic mobility shift assay
The probe DNA and transcription factor proteins were prepared as previously described (17, 21). The fluorescence of the electrophoresis gels was detected using a Typhoon FLA 7000 image analyzer (GE Healthcare).
Statistical analysis
A two-tailed Student t test was used to perform the statistical analysis. The p values <0.05 were considered to be significant.
Results
Involvement of PU.1 and IRF4 in the expression of PD-L2 in BMDCs
To evaluate the involvement of PU.1 and IRF4 in PD-L2 expression in BMDCs, we introduced Spi1 siRNA or Irf4 siRNA into BMDCs. First, we confirmed that the transfection of siRNA for Spi1 and Irf4 significantly decreased the mRNA levels of Spi1 and Irf4, respectively, and markedly reduced the amount of the target proteins in BMDCs (Fig. 1A, 1B). In these experimental conditions, the Pdcd1lg2 mRNA levels were significantly decreased by the knockdown of PU.1 and IRF4 (Fig. 1A, 1B). In contrast, the knockdown of IRF8, another typical partner of PU.1, did not affect the Pdcd1lg2 mRNA level (Fig. 1C), whereas the mRNA level of Il12b, which is a positive control gene that is cooperatively transactivated by PU.1 and IRF8 (27), was significantly downregulated by the Irf8 siRNA transfection. We also performed flow cytometry to define the cell surface expression level of PD-L2 and found that PD-L2 expression was mainly detected on MHCIIhigh/CD11bint cells (so-called DC-like cells) rather than MHCIIint/CD11bhigh cells (macrophage-like cells) (siNega in Fig. 1D, 1E). When PU.1 was knocked down, the number of DC-like cells and the PD-L2 expression level on DC-like cells were markedly decreased (Fig. 1D, bottom). A similar tendency was observed in IRF4 knockdown cells (Fig. 1E bottom) but not in IRF8 knockdown cells (Fig. 1F). Furthermore, the intracellular staining of permeabilized cells showed that the cellular PD-L2 protein levels in the total BMDCs were markedly and slightly reduced by the knockdown of PU.1 and IRF4, respectively (Fig. 1G). The above knockdown experiments demonstrate that suppression of PU.1 and IRF4 downregulated the expression of PD-L2 in BMDCs. Then, to evaluate the effect of the excess amount of PU.1 and IRF4 on PD-L2 expression, we overexpressed PU.1 or IRF4 in BMDCs. As shown in Fig. 1H, the cellular PD-L2 level was apparently upregulated in IRF4-overexpressed cells. These results indicate that PU.1 and IRF4 are involved in the transcription and subsequent protein expression of PD-L2 on BMDCs and the development of the PD-L2–expressing DC population.
Effects of PU.1 and IRF4 knockdown by siRNAs on the mRNA levels and cell surface expression levels in BMDCs. The mRNA expression of Pdcd1lg2 and the target gene in Spi1 siRNA-treated (A), Irf4 siRNA-treated (B), or Irf8 siRNA-treated (C) BMDCs at 48 h after siRNA transfection. The data represent the mean ± SD of three independent assays performed in triplicate samples. The protein levels of PU.1 and IRF4 in siRNA-transfected cells at 48 h after transfection were determined by Western blotting analysis (A and B). *p < 0.05. Surface expression level of PD-L2 in DC-like cells and macrophage-like cells in Spi1 siRNA- (D), Irf4 siRNA- (E), or Irf8 siRNA-treated (F) BMDCs at 48 h after transfection. Shadow represents isotype control; line represents anti-PD-L2 Ab. Intracellular expression level of PD-L2 in the total cells in siRNA-treated BMDCs (G) and expression plasmid-transfected BMDCs (H). A representative result of three independent experiments is shown (D–H).
Binding of PU.1 and IRF4 to the Pdcd1lg2 gene in DCs
To clarify whether PU.1 and IRF4 directly bind to the Pdcd1lg2 gene, we used publicly available ChIP-sequencing (ChIP-seq) data and found four highly H3K27-acetylated loci in BMDCs (termed CNS1, 2, 3, and 4) (Fig. 2A). Interestingly, stronger binding signals of PU.1 and IRF4 were also detected at the sites corresponding to the H3K27 signals, whereas such strong signals were not observed in the promoter region (CNS0), suggesting that PU.1 and IRF4 specifically bind to CNS1, 2, 3, and 4 rather than the promoter region of the Pdcd1lg2 gene. When we searched for motifs recognizable by PU.1 and IRF4, we found that a typical EICE sequence, which is the binding site for the PU.1/IRF4 heterodimer (28), is located in CNS1, 3, and 4 (Fig. 2B). To confirm that PU.1 and IRF4 bind to these EICE sequences in CNS1, 3, and 4, we performed ChIP assays using BMDCs. Quantitative PCR targeting the Pdcd1lg2 gene showed that PU.1 significantly binds to all CNS regions, and the amount of PU.1 binding to CNS1 and CNS3 was relatively higher than that found in other regions (Fig. 2C), whereas such binding was not detected in a control region (−1.5 kb of the Gata3) that does not contain binding sites for PU.1 and/or IRF4 (29). Although the amount of DNA immunoprecipitated with anti-IRF4 Ab was small, which was likely because of the low affinity of the Ab used, significant binding of IRF4 to CNS1 and CNS3 was detected (Fig. 2D). Based on these results, we demonstrated that both PU.1 and IRF4 bind to the Pdcd1lg2 gene around CNS1 and CNS3, which carry typical EICE sequences, and only PU.1 but not IRF4 binds to the Pdcd1lg2 gene via Ets motifs in other CNSs in PD-L2–expressing DCs.
Binding of PU.1 and IRF4 to the Pdcd1lg2 gene in BMDCs. ChIP-seq profiles showing PU.1 and IRF4 occupancies and histone modification in the Pdcd1lg2 gene (A). The following data were obtained from the Gene Expression Omnibus Accession Viewer (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE36099): PU.1, GCM881099; IRF4, GCM881147; H3K27Ac, GSM881080. The EICE sequence, which is the binding site of PU.1 and IRF4, is found in CNS1, 3, and 4 (B). The amount of genomic DNA in whole BMDCs immunoprecipitated by ChIP assays using anti-PU.1 Ab (C) and anti-IRF4 Ab (D) was determined by quantitative PCR targeting the Pdcd1lg2 gene. The data represent the mean ± SD of three independent assays performed in triplicate. *p < 0.05.
PU.1 and IRF4 transactivate the Pdcd1lg2 gene through direct binding to an EICE sequence in intron 2
To confirm that the PU.1/IRF4 heterodimer directly binds to EICE sequences in the Pdcd1lg2 gene, we performed an electrophoretic mobility shift assay (EMSA) using double-stranded oligo DNAs with the Pdcd1lg2 gene sequence containing EICE motifs as probes (Fig. 3A). When the probe DNA was mixed with PU.1 protein, the specific shifted bands appeared for all three probes (lanes 2, 8, and 14), suggesting that PU.1 bound to all tested DNA sequences. The addition of IRF4 to the mixture of PU.1 and probe DNA decreased the band intensity of the PU.1/probe complex and in turn caused the appearance of new bands, which likely represented the PU.1/IRF4/probe complex, at higher molecular weights (lanes 3, 9, and 15). Each complex composition was confirmed by the disappearance of specific bands and/or the appearance of the supershifted bands in the presence of anti-PU.1 Ab and anti-IRF4 Ab (lanes 5, 6, 11, 12, 17, and 18). Similar band shifts were observed in an EMSA using a well-characterized EICE sequence in the λ1B gene (30), as shown in Fig. 3B. When the IRF4 binding sequence in the CNS3 was mutated, the band intensity of PU.1/IRF4/probe was dramatically reduced (Fig. 3C). In Fig. 3A, the band intensity ratio of the PU.1/IRF4/CNS3 probe versus PU.1/CNS3 in lane 9 was highest compared with that of CNS4 (lane 15), indicating that the PU.1/IRF4 heterodimer prefers to bind to the EICE in CNS3 rather than the EICE in CNS4.
Transactivation of the Pdcd1lg2 gene by PU.1 and IRF4 via the EICE sequence. EMSA profiles obtained using fluorescein-labeled dsDNA corresponding to CNS1, CNS3, or CNS4 (A) using a probe corresponding to well-characterized EICE sequence in the λ1B gene (B) and using a mutant probe lacking the IRF4 binding motif in the CNS3 sequence (C). Relative luciferase activity in cells transfected with the pGL-4.10 vector, pGL4.10_CNS3_WT vector, pGL-4.10_CNS3_mut vector, or pGL4.10_Aldh1a2_EICE (22) (D). Relative luciferase activity is shown as the ratio of luciferase activity in the transfected cells to that observed in mock-transfected cells, and the data represent the mean ± SD of three independent assays performed in triplicate. *p < 0.05. The amount of PU.1 binding to the chromosomal DNA in siRNA-introduced BMDCs (E). The data represent the mean ± SD of three independent assays performed in triplicate. *p < 0.05.
To evaluate the effects of PU.1 and IRF4 on the transcriptional activity of the Pdcd1lg2 promoter and to clarify the involvement of CNS3-EICE as a cis-enhancing element in the Pdcd1lg2 gene, we performed a luciferase assay with coexpression plasmids for PU.1 and IRF4 (Fig. 3D). Luciferase activity driven by the CNS3 region of the Pdcd1lg2 gene was markedly increased by the coexpression of PU.1 and IRF4, as observed in a reporter plasmid carrying a typical EICE motif that was previously revealed to be transactivated by PU.1 and IRF4 (22). The transactivation effect of IRF4, which significantly upregulated the luciferase activity driven by CNS3, may be nonspecific because a similar tendency was observed for the promoterless reporter plasmid. When the EICE sequence in CNS3 was mutated, the synergistic effect of PU.1 and IRF4 on transactivation activity was not observed, suggesting that PU.1 and IRF4 cooperatively transactivate the Pdcd1lg2 gene through the EICE in CNS3. In the luciferase assays using reporter plasmids carrying the CNS1 region and the CNS4 region, cooperative transactivation by PU.1 and IRF4 was not observed (data not shown).
PU.1 bound to the CNS3 probe even when the IRF4 binding motif was mutated (Fig. 3C), probably because the remaining Ets motif in the mutated probe can be bound with monomeric PU.1. To clarify whether PU.1 binds to CNS3 by itself in living cells, we evaluated the effect of IRF4 knockdown on the binding degree of PU.1 to the EICE in CNS3 by a ChIP assay. As shown in Fig. 3E, the significant amount of PU.1 binding to CNS3 was detected in IRF4 knockdown BMDCs. In contrast, the coexpression of PU.1 without IRF4 did not increase the luciferase activity driven by the CNS3 (CNS3WT with PU.1 in Fig. 3D). These results suggest that both PU.1 and IRF4 are required for the EICE-mediated transactivation, although PU.1 can bind to the EICE without IRF4.
Overall, we demonstrated that PU.1 and IRF4 transactivate the Pdcd1lg2 gene through direct binding to the EICE sequence in CNS3.
Involvement of the acetylation of the H3K27 histone in the CNS region in PD-L2 expression in BMDCs
As shown in Fig. 4A, K27 residues of histone H3 in CNS regions are highly acetylated in DCs compared with those in another hematopoietic cell, mast cells, suggesting a relationship between the degree of the acetylation of CNS regions in the Pdcd1lg2 gene and the expression level of PD-L2. Recently, we found that PU.1 modulates the histone acetylation of the Ciita gene in DCs (17). In the current study, to clarify the involvement of PU.1 in H3K27 acetylation of the Pdcd1lg2 gene, we performed ChIP assays with anti-acetyl histone H3 (K27) Ab in BMDCs treated with Spi1 siRNA. It was confirmed that H3K27 residues in CNS1, CNS0, CNS2, and CNS3 were significantly acetylated in control siRNA-transfected DCs (siNega in Fig. 4B). When PU.1 siRNA was introduced into DCs, the H3K27 acetylation levels of the Pdcd1lg2 gene were drastically decreased (siSpi1 in Fig. 4B). In contrast, IRF4 knockdown did not affect the acetylation levels (data not shown).
Involvement of H3K27 acetylation in Pdcd1lg2 gene expression in BMDCs. ChIP-seq profiles showing the histone modification of the Pdcd1lg2 gene (A). DC represents GSM881080; mast cell represents GSM1329816. ChIP assay of Spi1 siRNA- (B) or Ep300 siRNA-treated (D) BMDCs using anti-acetyl H3K27 Ab (closed bars) or its control (open bars) in (B and D). The mRNA expression of Pdcd1lg2 and the target gene in Ep300 siRNA-treated BMDCs (closed bars) and control siRNA-treated BMDCs (open bars) in (C). The data represent the mean ± SD of three independent assays performed in triplicate [(C and D), and CNS0–3 in (B)]. CNS4 in (B) is data of single experiment performed in triplicate. *p < 0.05. n.s., not significant.
It was previously reported that PU.1 physically interacts with the coactivators p300 and GCN5, which possess a HAT activity in B cells and Th9 cells, respectively (23, 24). To reveal the involvement of p300 and GCN5 in the histone acetylation of the Pdcd1lg2 gene in DCs, we performed the knockdown of p300 and GCN5 by introducing siRNAs into BMDCs. The Ep300 mRNA level was reduced to ∼50% of that in control cells by the transfection of Ep300 siRNA into DCs in which the Pdcd1lg2 mRNA level was significantly decreased (Fig. 4C). Under this experimental condition, the H3K27 acetylation status in all CNS regions was reduced to the background level (Fig. 4D). In contrast, the knockdown of GCN5 did not affect the Pdcd1lg2 mRNA level (data not shown). These results suggest that PU.1 and p300 are involved in acetylation of the H3K27 in CNS3 in the Pdcd1lg2 gene in DCs.
Pdcd1lg2 gene expression under GM-CSF stimulation
The above-mentioned results suggested that PU.1 is crucial for the expression of PD-L2 in GM-CSF–induced BMDCs. In our recent study, it was revealed that GM-CSF stimulation upregulated the expression of IRF4 and the recruitment of PU.1 to chromosomal DNA in Flt3L-derived BMDCs (22). Therefore, we hypothesized that GM-CSF stimulation induces or upregulates the PD-L2 expression in Flt3L-BMDCs. Quantitative PCR showed that the Pdcd1lg2 mRNA level in whole Flt3L-BMDCs was significantly increased by GM-CSF stimulation (Fig. 5A). By using flow cytometry analysis, we found that conventional DCs (cDCs; CD11c+/PDCA-1−) expressed PD-L2 at a low level in steady-state condition, and PD-L2 expression was increased following GM-CSF stimulation (Fig. 5B). PD-L2 expression was also detected in GM-CSF–stimulated plasmacytoid DCs (CD11c+/PDCA-1+) that did not express PD-L2 in a nonstimulated state (Fig. 5B). A ChIP assay in a CD11c+ population of Flt3L-BMDCs revealed that GM-CSF stimulation increased the amount of PU.1 protein binding to CNS regions in the Pdcd1lg2 gene (Fig. 5C). Furthermore, GM-CSF stimulation tended to increase the H3K27 acetylation level of CNS regions in Flt3L-BMDCs (Fig. 5D). These results suggested that GM-CSF stimulation increased the expression of PD-L2 in Flt3L-BMDCs, probably because of the acceleration of PU.1 binding to the Pdcd1lg2 gene and the increase in the histone acetylation level in the Pdcd1lg2 gene.
The effect of GM-CSF on Flt3L-BMDCs. Flt3L-BMDCs generated from bone marrow cells by culture with 100 ng/ml of Flt3L for 6 d were stimulated by 10 ng/ml GM-CSF for 24 h. CD11c+ cells purified by microbeads are used for ChIP assays shown in (C) and (D). The mRNA expression of Pdcd1lg2 in GM-CSF–stimulated cells or nonstimulated cells (whole cells) (A). Surface expression level of PD-L2 in cDCs and plasmacytoid DCs (B). Shadow represents isotype control; line represents anti–PD-L2 Ab. ChIP assay of GM-CSF–stimulated cells or nonstimulated cells using anti-PU.1 Ab or its control (C) and anti-acetyl H3K27 Ab or its control (D). The data represent the mean ± SD of three independent assays performed in triplicate (A, C, and D). A representative result of three independent experiments is shown (B). *p < 0.05.
Involvement of PU.1, IRF4, and p300 in the Pdcd1lg2 gene expression in DCs ex vivo
The experiments using GM-CSF– or Flt3L-induced BMDCs indicated the involvement of PU.1, IRF4, and p300 in Pdcd1lg2 gene expression in DCs. To further confirm the contribution of these transcriptional regulators to the expression of the Pdcd1lg2 gene in natural DCs ex vivo, we used DCs purified from mouse spleen. We introduced Spi1, Irf4, or Ep300 siRNA into CD11c+ splenic DCs and quantified the mRNA levels in harvested cells after 2-d cultivation in the presence of GM-CSF. As shown in Fig. 6, the knockdown of PU.1, IRF4, and p300 significantly decreased Pdcd1lg2 mRNA levels. Therefore, the in vitro finding that PU.1, IRF4, and p300 are involved in Pdcd1lg2 gene expression in DCs was supported by the ex vivo results.
Involvement of PU.1, IRF4, and p300 in the expression of the Pdcd1lg2 gene in DCs ex vivo. CD11c+ DCs were collected from spleen using microbeads. The mRNA levels of the Pdcd1lg2 gene in Spi1 siRNA- (A), Irf4 siRNA- (B) or Ep300 siRNA-treated (C) cells. The data represent the mean ± SD of three independent assays performed in triplicate. *p < 0.05.
Discussion
The immune checkpoint molecule PD-1 and its ligand PD-L1 have been paid increasing attention in the field of cancer immunity. PD-L2, a second ligand of PD-1, has recently shown to be involved not only in cancer but also in allergy, infection, and humoral immunity (5–7), which raises the possibility that PD-L2 may be a therapeutic target for various immune-related diseases. The most definitive difference between PD-L1 and PD-L2 is their expression profiles: whereas the expression of PD-L1 is ubiquitous and IFN-γ inducible, PD-L2 is mainly detected in APCs, and the factors that induce PD-L2 expression are IL-4, IL-13, and GM-CSF (8). Previously, STAT6 was identified in studies analyzing the mechanism by which Pdcd1lg2 gene expression is induced by IL-4/IL-13 in activated macrophages and DCs (9, 10). However, considering that STAT6 is ubiquitous transcription factor, the transcriptional regulatory molecules that determine the cell type-specific expression of the Pdcd1lg2 gene remained unclear. In the current study, we indicated that the hematopoietic cell–specific transcription factor PU.1 plays a critical role in Pdcd1lg2 gene expression as a transactivator with a partner molecule IRF4 and as an epigenetic regulator with p300, which can explain the mechanism by which GM-CSF stimulation upregulates the expression of PD-L2 in a cell type-specific manner.
First, we demonstrated the involvement of PU.1 and IRF4 in Pdcd1lg2 gene expression by using Spi1 or Irf4 siRNA-treated BMDCs (Fig. 1A, 1B). As previously reported, we found that GM-CSF–generated BMDCs are heterogeneous and comprise two populations: a CD11bint/MHCIIhigh population (characterized as DC-like cells) and a CD11bhigh/MHCIIint population (macrophage-like cells) (31, 32). Flow cytometry analysis showed that PD-L2 is expressed in DC-like CD11bint/MHCIIhigh cells but not in macrophage-like cells and that PU.1 knockdown and IRF4 knockdown reduced the cell surface expression level of PD-L2 (Fig. 1). These results indicated that PU.1 and IRF4 are positive regulators of Pdcd1lg2 gene expression in BMDCs. Furthermore, the knockdown of PU.1 and IRF4 decreased the number of CD11bint/MHCIIhigh cells (Fig. 1D, 1E), suggesting that PU.1 and IRF4 are involved in the development and/or maintenance of a so-called DC-like population in BMDCs. The cell surface expression level of PD-L2 in DC-like cells and macrophage-like cells is very different (Fig. 1D–F). The expression level of PU.1 is similar between DC-like cells and macrophage-like cells, whereas the IRF4 expression level in DC-like cells is higher than that in macrophage-like cells [(32), not shown in our data]. We found that the Gm-csfr gene was highly expressed in DC-like cells (data not shown). Considering that GM-CSF stimulation increases the activity of PU.1 (33), including its transactivation activity and/or degree of binding to chromosomal DNA accompanied by a high amount of IRF4, it may cause the difference in PD-L2 expression level between DC-like cells and macrophage-like cells. Actually, overexpression of IRF4 rather than PU.1 upregulated the PD-L2 expression (Fig. 1H). The observation in the overexpression experiment may support our hypothesis that a high amount of IRF4 in cells in which PU.1 exhibits enough activity is required for the expression of PD-L2.
PU.1 binds to the EICE sequence as a heterodimer with IRF4 (28). Among the EICE sequences located in CNS1, 3, and 4, the EICE in CNS3 was the most preferred sequence for the binding of PU.1/IRF4 in vitro (Fig. 3A). In contrast to the binding of PU.1, which was observed in all analyzed CNS regions in living cells (Fig. 2C), IRF4 binding was detected in CNS1 and CNS3 (Fig. 2D). A luciferase assay showed that PU.1 and IRF4 exhibited transactivation activity via the EICE in CNS3 in a synergistic manner (Fig. 3D), whereas the cooperative transactivation by PU.1 and IRF4 was not observed in luciferase assays using reporter plasmids carrying the EICE sequence from CNS1 or CNS4 (data not shown). Based on these results, we conclude that PU.1 and IRF4 transactivated the Pdcd1lg2 gene via the EICE sequence in CNS3. During the binding of IRF4 to the CNS1 region (Fig. 2D), IRF4 may form a complex with transcription factor(s) other than PU.1, as IRF4 cannot bind to dsDNA by itself and requires other molecule(s) to do so.
Because H3K27 in all analyzed CNS regions was highly acetylated in DCs but not in mast cells in silico (Fig. 4A), H3K27 acetylation in CNS regions was also expected to contribute to Pdcd1lg2 gene expression in DCs. We confirmed that the knockdown of p300 decreased Pdcd1lg2 gene expression (Fig. 4C). As expected, the knockdown experiments showed that both PU.1 and p300 are involved in the H3K27 acetylation of CNS regions (Fig. 4B, 4D). Based on these results, we speculate that p300, which is recruited to chromatin through a physical interaction with PU.1 in CNSs, enhances H3K27 acetylation in CNS regions, resulting in the transcription of the Pdcd1lg2 gene. Although there has been a report that GCN5 recruited by PU.1 contributes to IL-9 expression in Th9 cells via histone acetylation (23), the involvement of GCN5 in Pdcd1lg2 transcription was not observed in our experimental conditions. In a previous study, p300 enhanced the transactivation activity of PU.1 through the acetylation of lysine residues in the transactivation domain of PU.1 (24). Although we did not analyze the effect of p300 on PU.1 modification, we showed that p300 and PU.1 are required for H3K27 acetylation and the transcription of the Pdcd1lg2 gene, which, to our knowledge, is a novel finding.
It is well known that GM-CSF stimulation induces the differentiation and characteristic gene expression of DCs. We previously found that GM-CSF stimulation increases the expression of PU.1 and the subsequent recruitment of PU.1 to chromosomal DNA (20, 22) and the IRF4 expression level in DCs was markedly upregulated by GM-CSF stimulation (22). In this study, we found that the amount of PU.1 binding in CNS regions and the level of H3K27 acetylation in CNS regions were increased by GM-CSF stimulation in Flt3L-induced DCs, which was accompanied by the strong upregulation of the Pdcd1lg2 mRNA level and an apparent increase in PD-L2 protein, especially in the cDC population (Fig. 5). Based on these observations, we conclude that GM-CSF stimulation induces Pdcd1lg2 gene expression through enhancing the expression and/or function of PU.1 and IRF4. Finally, via experiments using freshly isolated splenic DCs, we confirmed that PU.1, IRF4, and p300 govern Pdcd1lg2 gene expression in DCs ex vivo (Fig. 6).
In addition to DCs, macrophages also express PD-L2, especially under GM-CSF stimulation (8). The role of PU.1 in PD-L2 expression in macrophages was not evaluated in the current study because of the nondetectable level of PD-L2 in macrophage-like cells (Fig. 1D, 1E). However, we propose the possibility that PU.1 plays a key role in PD-L2 expression in macrophages, as PU.1 is critical for gene expression in macrophages as well. Further studies using PD-L2–expressing macrophages will be needed to confirm the involvement of PU.1 and IRF4 in PD-L2 expression in macrophages.
PD-L2 has an impact on the regulation of helper T cells. PD-L2 in DCs or activated macrophages suppresses Th2 responses and induces Th1 responses (9, 10). Compared with CD8α+ DCs, CD8α−DCs favors the Ag presentation to CD4+ T cells (34). CD8α− DCs may express PD-L2 under the influence of PU.1 because IRF4 is critical for the development of CD8α− cDCs (35). In the current study, we demonstrated the role of PU.1 in the cell type-specific expression of PD-L2 using GM-CSF–induced BMDCs, Flt3L-induced BMDCs, and purified CD11c+ cells from spleen. Further analyses focused on PU.1 will clarify the regulatory mechanisms involved in PD-L2 expression in various subtypes of DCs and macrophages.
Disclosures
The authors have no financial conflicts of interest.
Acknowledgments
We are grateful to the members of the Laboratory of Molecular and Cellular Immunology, Department of Biological Science and Technology, Tokyo University of Science for constructive discussions and technical support.
Footnotes
This work was supported by Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for Scientific Research (B) 20H02939 (to C.N.), JSPS Grant-in-Aid for Scientific Research (C) 19K05884 (to T.Y.) and 19K08920 (to K.K.), the Ministry of Education, Culture, Sports, Science and Technology-Supported Program for the Strategic Research Foundation at Private Universities (Translational Research Center, Tokyo University of Science), the Tokyo University of Science Grant for President’s Research Promotion (to C.N.), the Tojuro Iijima Foundation for Food Science and Technology (to C.N. and T.Y.), and the Takeda Science Foundation (to C.N.).
Abbreviations used in this article:
- BMDC
- bone marrow–derived DC
- cDC
- conventional DC
- ChIP
- chromatin immunoprecipitation
- ChIP-seq
- ChIP-sequencing
- DC
- dendritic cell
- EICE
- Ets-IRF composite element
- EMSA
- electrophoretic mobility shift assay
- HAT
- histone acetyltransferase
- MHCII
- MHC class II
- PD-1
- programmed cell death-1
- siRNA
- small interfering RNA.
- Received August 21, 2019.
- Accepted June 1, 2020.
- Copyright © 2020 by The American Association of Immunologists, Inc.