Transcription factor Runx2 controls the development and migration of plasmacytoid dendritic cells

To identify additional regulators of pDC development, we searched for transcription factors whose expression is directly activated by E2-2 in pDCs. The expression of RUNX2 has been previously detected by microarrays in human pDC-derived lymphomas (Dijkman et al., 2007), including the human pDC cell line CAL-1 (Cisse et al., 2008). Chromatin immunoprecipitation/deep sequencing (ChIP-Seq) analysis of E2-2 target genes in CAL-1 cells showed prominent E2-2 binding to the RUNX2 locus. In addition to several major binding peaks in the intron and 3′ region, the binding of E2-2 to RUNX2 promoter was also detected (Fig. 1 A). After doxycycline (Dox)-inducible knockdown of E2-2 by short hairpin RNA (shRNA), we observed a reduction in RUNX2 expression in CAL-1 (unpublished data) as well as in another human pDC cell line, Gen2.2 (Fig. 1 B). The reduction was comparable to that of TCF4 (the gene encoding E2-2) itself; in contrast, another direct E2-2 target SLA2 (Ghosh et al., 2010) did not decrease in these conditions. These results suggest that RUNX2 is directly activated by E2-2 in human pDCs.

In the mouse, preferential expression of Runx2 in sorted pDCs compared with cDCs has been noticed in microarray-based expression studies (Crozat et al., 2010; Satpathy et al., 2011; Miller et al., 2012). To analyze Runx2 expression in the entire hematopoietic system at the single-cell level, we used Runx2 allele with the LacZ reporter replacing the first coding exon (Otto et al., 1997). The analysis of LacZ expression in Runx2^LacZ/+ heterozygous (Het) mice by flow cytometry showed a specific LacZ⁺ population in the BM (Fig. 1 C) and peripheral lymphoid organs (not depicted). Almost all LacZ⁺ cells showed the Bst2⁺ B220⁺ phenotype characteristic of pDCs (Fig. 1 C). Conversely, only pDCs but no other lineages (including cDCs or B cells) harbored a LacZ⁺ population in the BM, spleen, and LN (Fig. 1, D and E). The partial expression of LacZ in pDCs likely reflects the variegation of the reporter allele, which includes the heterologous drug selection cassette. Indeed, the expression of the wild-type Runx2 allele was comparable in LacZ⁺ and LacZ⁻ populations (not depicted), and Runx2^LacZ/LacZ pDCs showed a more uniform LacZ expression (Fig. 1 E). Collectively, these data demonstrate that within the hematopoietic system, Runx2 is expressed specifically in the pDC lineage.

To determine the functional significance of Runx2 expression in pDCs, we analyzed homozygous Runx2^LacZ/LacZ KO cells. Because KO mice die perinatally due to failed ossification, we reconstituted irradiated CD45.1⁺ recipient mice with KO or control WT fetal liver cells. Flt3L-supplemented cultures of BM cells from the resulting chimeric mice showed normal development of KO pDCs (unpublished data). Within the BM, the fraction of KO donor-derived CD45.2⁺ Bst2⁺ B220⁺ pDCs was slightly but significantly increased compared with WT-derived pDCs (Fig. 2, A and B). In contrast, the frequencies of KO pDCs were reduced 4–5-fold in the peripheral blood, spleen, and LNs (Fig. 2, A and B). A comparable decrease was also observed in the very small fraction of pDCs in the thymus (Fig. 2 B). On the other hand, no significant pDC decrease was detected in the Peyer’s patches, which support a divergent IFN-dependent pathway of pDC development (Li et al., 2011). Thus, all populations of BM-derived peripheral pDCs were decreased in KO chimeras, whereas the pDC population that may develop in situ was less affected. The decrease was restricted to pDCs, as other hematopoietic populations were detected at normal percentages in the spleen and LN (Fig. 2 C). Furthermore, serum IFN-α production in response to TLR9 ligand unmethylated CpG oligonucleotide (CpG) was severely impaired in mice reconstituted with KO cells (Fig. 2 D). Because CpG-induced IFN-α production is completely dependent on pDCs (Cisse et al., 2008), these data confirm the loss of peripheral pDCs after Runx2 deletion.

Next, we generated competitive chimeras in which WT or KO fetal liver cells were mixed at a 1:1 ratio with CD45.1⁺ total BM cells and transplanted into irradiated CD45.1⁺ recipients. The fractions of donor-derived CD45.2⁺ pDCs in the BM were comparable between recipients of WT and KO cells; however, the fraction of KO pDCs was strongly reduced in the spleen (Fig. 2 E). We also determined the ratios of donor-derived CD45.2⁺ to competitor-derived CD45.1⁺ cells in each immune lineage, and normalized the ratios from the spleen and LN to those in the BM. We found that the ratio of donor to competitor pDCs was decreased >6-fold in the spleen and LN of KO cell recipients (Fig. 2 F). This defect was specific to pDCs, because other KO hematopoietic lineages including B and T cells and granulocytes were detected at normal frequencies. These results confirm that the loss of peripheral pDCs after Runx2 deletion is cell intrinsic, consistent with the pDC-specific expression of Runx2.

No consistent increase in the fraction of Annexin V⁺ pDCs was detected in the KO BM (unpublished data), arguing against pDC apoptosis as a primary cause of the phenotype. However, increased pDC fraction in the KO BM suggested retention of mature pDCs in the BM. Ly49Q is a pDC-specific receptor expressed on the majority of peripheral pDCs and on the subset of BM pDCs that corresponds to fully mature cells (Kamogawa-Schifter et al., 2005; Omatsu et al., 2005). We found that the fraction of Runx2-deficient Ly49Q⁺ pDCs was significantly increased in the BM but drastically reduced in the periphery (Fig. 3, A and B). Indeed, the ratio of Ly49Q⁺ to Ly49Q⁻ cells was significantly increased (∼1.5-fold) among KO pDCs in the BM, but reduced 5–6-fold in the periphery (Fig. 3 C). The residual Ly49Q⁻ pDCs in the periphery of KO chimeras likely arise from immature Ly49Q⁻ BM pDCs, which may migrate from the BM with low efficiency by a Runx2-independent mechanism. This is consistent with the documented migration of immature Ccr9⁻ pDCs to the periphery, where they can give rise to mature pDCs (Schlitzer et al., 2011). Collectively, these data suggest that Runx2-deficient mature pDCs accumulate in the BM and fail to migrate into the periphery, causing the overall depletion of pDCs.

To explore the mechanism of Runx2 activity in pDCs, we performed genome-wide expression analysis of donor-derived WT or KO pDCs sorted from the BM of recipient mice (Supplemental Dataset 1). Pairwise comparison of expression profiles revealed small subsets of genes whose expression differed >2-fold between WT and KO pDCs (Fig. 4 A). Several of these differentially expressed genes encoded surface markers, allowing validation by single-cell analysis. In particular, the expression of CD11c was reduced, whereas the expression of CD8α and c-Kit was increased on KO pDCs (Fig. 4 B). We further examined the genes that are enriched in the pDC lineage compared with cDCs (the “pDC signature” genes; Elpek et al., 2011). Whereas the majority of these signature genes were unchanged, a small set of genes (36 out of 872, 4.1%) were reduced in KO pDCs (Fig. 4 C). Thus, Runx2 regulates a distinct subset of the pDC-specific gene expression program.

The genes down-regulated in Runx2-deficient pDCs included Ccr2 and Ccr5, two related chemokine receptors that are preferentially expressed in pDCs compared with cDCs (Elpek et al., 2011). Because Ccr2 and Ccr5 are known regulators of immune cell migration, we focused on their potential role in pDCs downstream of Runx2. Reduced expression of both Ccr2 (∼2-fold) and Ccr5 (∼5-fold) transcripts in KO pDCs was confirmed by qRT-PCR (Fig. 4 D). We next measured the ability of pDCs to migrate in vitro to chemokines MCP-1 (Ccr2 ligand) as well as MIP-1α and RANTES (Ccr5 ligands) in a trans-well assay. The percentage of migrating cells in response to each of these chemokines was reduced in Runx2-deficient KO pDCs (Fig. 4 E). Thus, Runx2 facilitates the functional expression of Ccr2 and Ccr5 in the pDCs.

To explore the role of Ccr2 and Ccr5 in the steady-state migration of pDCs, we tested their expression at the surface of pDCs. Ccr2 was expressed only in a fraction of pDCs in the BM and periphery, whereas Ccr5 was expressed prominently in nearly all pDCs (Fig. 5 A). The expression of both receptors was up-regulated in the mature Ly49Q⁺ pDCs compared with immature Ly49Q⁻ pDCs in the BM (Fig. 5 B). The surface levels of both Ccr2 and Ccr5 were strongly reduced in Runx2-deficient pDCs, particularly in the Ly49Q⁺ subset (Fig. 5 C). These results suggest that the expression of Ccr2 and Ccr5 is induced on pDCs in a Runx2-dependent manner during their maturation in the BM.

Ccr2 and Ccr5 are encoded by closely adjacent homologous genes and may show genetic redundancy. We found that the fraction of Ccr2⁺ cells was increased in Ccr5^−/− pDCs (Fig. 5 D), suggesting that Ccr5 deficiency causes a compensatory up-regulation of Ccr2. In contrast, the expression of Ccr5 was not changed in Ccr2^−/− animals. The heterogeneous Ccr2 expression on pDCs argues against a primary role of this receptor in pDC migration. Indeed, the analysis of pDC populations in competitively reconstituted recipients of Ccr2^−/− BM (Yona et al., 2013) did not reveal any cell-intrinsic defects (data not shown). On the other hand, the prominent homogeneous expression of Ccr5 and the apparent compensatory changes in Ccr5^−/− pDCs suggested a role of Ccr5 in pDC migration.

The analysis of Ccr5^−/− animals revealed normal pDC numbers in the BM and spleen, but a twofold decrease in the number of pDCs in the LN (unpublished data). To confirm the cell-intrinsic role of Ccr5 in pDC migration, we performed competitive reconstitution by mixing equal numbers of Ccr5^−/− (CD45.2⁺) and WT (CD45.1⁺) BM cells and transplanting them into irradiated WT recipients (CD45.1⁺). To reduce the likely compensatory effects of Ccr2, we also transplanted the same BM mixture into the recipients deficient in MCP-1 (Ccl2), one of the two major Ccr2 ligands (Jia et al., 2008; Tsou et al., 2007). B cells do not express Ccr5 (unpublished data), and the fraction of Ccr5^−/− B cells was similar between the BM, spleen, and LN in all recipients. In contrast, the fraction of total Ccr5^−/− pDCs was increased in the BM of WT recipients (Fig. 5 E). When normalized by B cells, the fraction of Ccr5-deficient mature Ly49Q⁺ pDCs was significantly higher in the BM compared with LN in all recipients (Fig. 5 F). Notably, MCP-1–deficient recipients showed a significant progressive reduction of Ly49Q⁺ pDCs in the spleen and LN compared with the BM (Fig. 5 F). These data suggest that Ccr5 contributes to the migration of pDCs from the BM to the spleen and especially to the LN, and its function can be partially substituted by Ccr2.

Our data demonstrate that, within the murine hematopoietic system, Runx2 is expressed exclusively in the pDC lineage. Consistent with its activation by the pDC master regulator E2-2, Runx2 is dispensable for pDC lineage commitment but regulates late steps of pDC differentiation. This includes the acquisition of pDC-related surface markers (such as the expression of DC marker CD11c) and migration from the BM into the periphery. The documented expression of human RUNX2 in normal and malignant pDCs (Dijkman et al., 2007; Crozat et al., 2010) suggests that its function in this lineage is evolutionarily conserved. Our results reveal the complexity of the pDC-specific transcriptional network, whose members govern successive differentiation steps, including commitment and maintenance (E2-2), maturation within the BM (SpiB) and migration to the periphery (Runx2). In addition, they establish Runx2 as one of the few regulators of immune system development that are highly lineage-specific in their expression and function. Finally, these results show that the entire Runx protein family, including Runx2, has important functions in immune system development.

Unlike many cell types, including T and B lymphocytes and cDCs, pDCs develop to full functional maturity in the BM. Thus, the migration from the BM to the periphery represents a distinct checkpoint of pDC development, and here we identify Runx2 as its key regulator. The function of Runx2 is likely mediated through multiple target genes, which collectively facilitate pDC migration. For instance, Runx2-deficient pDCs showed low levels of CD11c, an integrin whose function is unknown but likely involves migration control. In addition, our results implicate chemokine receptor Ccr5 as a target of Runx2 that controls pDC migration into the periphery. Ccr5 facilitates the migration of multiple cell types, including activated pDCs into LN (Diacovo et al., 2005); however, its expression and role in the steady-state pDCs have not been explored previously. The function of Ccr5 in pDCs appears to be partially redundant with Ccr2, whose expression is increased on Ccr5-deficient pDCs. The combined loss of Ccr5 and one Ccr2 ligand (MCP-1) revealed a cell-intrinsic role of Ccr5 in pDC migration from the BM to the spleen and particularly to the LN. However, MCP-1 is partially redundant with MCP-3 (Tsou et al., 2007; Jia et al., 2008), thus the complete loss of Ccr5 and Ccr2 signaling is likely to cause stronger impairment of pDC migration. Although the Ccr2/Ccr5 double-deficient animals are currently unavailable, this possibility could be tested in the future. Importantly, our results suggest that pDC migration may be affected by pharmacological inhibition of Ccr5 and/or Ccr2. This possibility and its implications for pDC function should be considered in designing and interpreting the applications of Ccr5/Ccr2 inhibitors.

All experiments were performed according to the investigator’s protocol approved by the Institutional Animal Care and Use Committee of Columbia University. Runx2^LacZ mice (Otto et al., 1997; provided by G. Karsenty and S. Kousteny, Columbia University, New York, NY) were backcrossed to C57BL/6 background for >10 generations. Animals deficient for Ccr5, Ccr2, and Ccl2 (C57BL/6 background) were obtained from The Jackson Laboratory. For hematopoietic reconstitution, Runx2^LacZ/+ mice were intercrossed and embryos isolated on day 14.5 post-coitum and genotyped. Fetal liver cells (2 × 10⁶) from Runx2^LacZ/LacZ or littermate control Runx2^+/+ embryos were injected i.v. into lethally irradiated C57BL/6 mice congenic for CD45.1 (B6.SJL; Taconic). Donor reconstitution reached >90% in all recipients. For competitive reconstitution, 10⁶ fetal liver cells were mixed with 10⁶ BM cells from B6.SJL mice and transplanted as above. For the analysis of Ccr5 function, BM cells from Ccr5^−/− and B6.SJL mice were mixed 1:1 and transplanted (2 × 10⁶/recipient) into irradiated B6.SJL (CD45.1⁺) or Ccl2^−/− (CD45.2⁺) mice. The fraction of Ccr5^−/− (CD45.2⁺) cells was similar in both recipient groups, suggesting a nearly complete reconstitution by donor cells. All recipients were analyzed at least 2 mo after reconstitution. For in vivo IFN-I response, mice were injected i.v. with 5 µg CpG (ODN 2216; InvivoGen) complexed with DOTAP (Roche), and serum IFN-α production was measured by ELISA (PBL Interferon Source).

Flow cytometric analysis and sorting were performed as previously described (Cisse et al., 2008). The expression of LacZ was measured with the fluorescent LacZ substrate fluorescein di-β-d-galactopyranoside using the FluoReporter LacZ Flow Cytometry kit (Invitrogen). The expression of Ccr2 and Ccr5 was detected with mAb MC-21 and MC-68, respectively (Mack et al., 2001). The cells were stained with these mAb (rat IgG2b) or an isotype control, followed by biotin-conjugated mouse anti-rat IgG2b (BD Biosciences); blocked in 10% rat serum; and then stained with PerCP-Cy5.5-conjugated streptavidin and fluorophore-conjugated mAb to other surface markers.

For the migration assay, total BM cells from recipients of WT or KO fetal livers were depleted of lineage-positive cells (CD3, Ter119, CD19, NK1.1, and Gr1) by magnetic cell separation (Miltenyi Biotec). Media containing 10 µg/ml of recombinant MCP-1, MIP1-α, or RANTES (PeproTech) was added to the well and 2 × 10⁵ enriched cells were placed in the trans-well insert with 5-µm pore size (Corning). After 3-h incubation at 37°C, cells that had passed through the trans-well insert were collected and the number of pDCs was measured by flow cytometry.

ChIP-Seq analysis and shRNA-mediated knockdown of E2-2 in human pDC cell lines CAL-1 (Maeda et al., 2005) and Gen2.2 (Chaperot et al., 2006) will be described in a separate publication. In brief, ChIP-Seq in CAL-1 cells was performed using a newly developed rabbit mAb to E2-2, and E2-2-bound or total input chromatin were amplified and sequenced on a HiSeq 2000 instrument (Illumina). Sequencing reads were aligned to the human genome (UCSC hg19), and binding peaks were visualized using DNANexus genome browser (DnaNexus, Inc). For inducible knockdown, CAL-1 and Gen2.2 were retrofitted with ecotropic retrovirus receptor and tetracycline repressor for Dox-inducible retroviral expression of shRNAs (Ngo et al., 2006). Cells were transduced with pRSMX vector-based constructs expressing E2-2–specific shRNA under tetracycline-inducible promoter, selected with puromycin and split into parallel cultures with or without 0.5 mg/ml Dox.

For microarray analysis, BM cells from recipients of WT or KO fetal livers were pooled and depleted of lineage-positive cells using magnetic cell separation. Donor-derived pDCs (CD45.2⁺ CD11b⁻ B220⁺ Bst2⁺) were sorted directly into TRIzol reagent (Invitrogen), and total RNA was reverse transcribed, amplified, labeled, and hybridized to Mouse Genome 1.0 ST arrays (Affymetrix). The results were gcRMA-normalized by the manufacturer’s software and filtered by matching to the probeset used by ImmGen consortium (Crozat et al., 2010; Satpathy et al., 2011; Miller et al., 2012). Pairwise analysis was performed using the NIA Array software. RNA from a separately sorted pDC sample was reverse transcribed and assayed by qRT-PCR for the indicated genes as previously described (Cisse et al., 2008). All microarray data are deposited in the NCBI’s GEO database under the accession no. GSE49859.

All p-values were calculated by unpaired, two-tailed Student’s t test.

A supplemental dataset contains the microarray probes.

We thank D. Rashkovan and S. Jung for analyzing Ccr2^−/− chimeric mice, M. Mack for antibodies to Ccr2 and Ccr5, G. Karsenty and S. Kousteny for the Runx2^LacZ strain, J. Plumas and L. Chaperot for Gen2.2 cells, D. Ganguly for control qPCR, and Y. Goltsev for helpful discussions.

This work was supported by the New York State Department of Health grant N09G-22, National Institutes of Health grant AI072571, and Leona M. and Harry B. Helmsley Charitable Trust (B. Reizis), Irvington Institute Fellowship of the Cancer Research Institute (V. Sisirak), and American Society of Hematology (H.S. Ghosh).

The authors declare no competing financial interests.