ETO family protein Mtg16 regulates the balance of dendritic cell subsets by repressing Id2

Given the prominent expression of ETO protein Mtg16 in DCs, we analyzed the DC compartment of Mtg16^−/− mice. Because Mtg16-null pDCs have aberrant surface phenotype (see below), they were defined using a combination of surface markers and broad gates to allow for marker level variation (Fig. S1). We further defined pDCs using two pDC-specific markers, Bst2 (which was reduced on Mtg16-null pDCs) or SiglecH (which was unaffected). Either definition revealed a several-fold decrease of pDC population in the BM and spleen of Mtg16^−/− mice (Fig. 1, A and B). Furthermore, Mtg16-null pDCs showed reduced levels of many characteristic surface markers, including Bst2, B220, Ly6c, and MHC class II, and a nearly complete absence of CD4 and Ccr9 (Fig. 1 C). Conversely, Mtg16-null pDCs expressed myeloid marker CD11b, which is normally absent from the pDC lineage (Fig. 1 D). The induction of CD11b was prominent (approaching the levels of monocytes and CD11b⁺ cDCs) and specific for pDCs versus other CD11b⁻ cells (Fig. 1 D). Thus, Mtg16-deficient pDCs are both reduced in numbers and have an aberrant surface phenotype, as described in several models including Ikaros^L/L (Allman et al., 2006), E2-2^+/− (Cisse et al., 2008) and Runx2^−/− (Sawai et al., 2013) mice.

Consistent with the reduction of pDCs, the production of IFN-α in response to Tlr9 ligand unmethylated CpG oligonucleotides (CpG) was reduced in cultures of Mtg16^−/− BM and spleen cells (Fig. 1 E). A minor (∼1.5-fold) but significant reduction was also observed in sorted Mtg16-null pDCs, suggesting an additional functional defect (Fig. 1 E). Similarly, intracellular staining showed an ∼1.5-fold reduction of IFN-α–producing cells within the pDC population (not depicted). We then tested CpG-induced production of serum IFN-α, which serves as a specific readout of pDC activity in vivo (Reizis et al., 2011a). Strikingly, Mtg16^−/− mice failed to produce IFN-α after CpG challenge, although they showed efficient IFN-α response to non-pDC–specific ligand poly-I:C (Fig. 1 F). The complete lack of CpG-induced IFN-α response likely reflects the reduced number and IFN-producing capacity of Mtg16-null pDCs, as well as additional defects of their in vivo function such as impaired migration (Asselin-Paturel et al., 2005). Indeed, impaired IFN production by pDCs in vivo but not in vitro has been recently documented in very young animals (Belnoue et al., 2013). Collectively, the loss of Mtg16 results in reduced pDC development, aberrant differentiation, and impaired pDC-dependent immune response.

To test whether the function of Mtg16 in pDCs is cell intrinsic, we analyzed mice that were competitively reconstituted with WT and Mtg16^−/− BM. Quantitative analysis of pDC development was precluded by the competitive disadvantage of Mtg16^−/− HSCs (Fischer et al., 2012) and rapid disappearance of their progeny in chimeras reconstituted at 50:50% ratio (not depicted). We therefore reconstituted CD45.1 recipients with 95% CD45.2 donor and 5% CD45.1 competitor BM. Control BM faithfully reconstituted to ∼90–95% at all time points, whereas Mtg16-null donor contribution was evident at 5 wk but became barely detectable by 8 wk (Fig. 2 A and not depicted). At 5 wk, Mtg16^−/− BM yielded ∼70% chimerism in B cells and monocytes, but only ∼40% in pDCs (Fig. 2 A). Furthermore, the resulting Mtg16-null pDCs showed the same reduction of B220 and expression of CD11b (Fig. 2 B). This aberrant phenotype was observed in Mtg16-null but not control or competitor pDCs and was specific for pDCs versus B cells (Fig. 2 B). To confirm the pDC-intrinsic activity of Mtg16 independently of the HSC defect, we established Flt3L-supplemented cultures of total WT or Mtg16^−/− BM cells mixed 50:50% with CD45.1 competitor BM. We observed an approximately threefold reduction in the fraction of Mtg16-null pDCs (Fig. 2 C), which had the same aberrant phenotype as ex vivo pDCs (Fig. 2 D). Finally, we sorted Flt3⁺ DC progenitors from control and Mtg16^−/− BM and cultured them with Flt3L. Although MHC II⁺ cDCs were produced in both cultures, the MHC II⁻ SiglecH⁺ pDC population was absent from Mtg16-null cultures (Fig. 2 E). Altogether, these data suggest a cell-intrinsic requirement for Mtg16 in pDC development and differentiation.

Noncompetitive cultures of Mtg16^−/− BM showed the expected reduction of pDCs fraction; in addition, we observed a threefold increase in CD11b⁺ CD24⁺ cells that may correspond to CD8⁺ cDC-like cells (Naik et al., 2005) or their precursors (Fig. 3 A). We therefore tested whether impaired pDC development in Mtg16^−/− mice is associated with a shift toward the alternative cDC cell fate. The analysis of splenic cDC compartment revealed an increased fraction of CD8⁺ cDCs, reflecting an approximately twofold increase in their absolute numbers (Fig. 3 B). Staining for additional CD8⁺ cDC markers Dec205 and CD24 confirmed the increased fraction of CD8⁺ cDCs in the spleen (Fig. 3 C).

Besides the canonical pre-DC–derived CD8⁺ cDCs, the CD8⁺ DC population includes a subset that is related to pDCs and requires E2-2 for development (Bar-On et al., 2010). We therefore crossed Mtg16^−/− mice to the Cx3cr1-GFP reporter, which distinguishes the two subsets as GFP⁻ and GFP^hi, respectively. The CD8⁺ cDC population in Mtg16^−/− mice consisted exclusively of the GFP⁻ subset, whereas GFP^hi CD8⁺ cDCs were virtually absent (Fig. 3 D). Thus, the loss of Mtg16 blocks the development of “pDC-like” but not of the canonical CD8⁺ cDCs. We also noticed that CD11b⁺ cDCs showed a different pattern of Cx3cr1-GFP expression, which normally defines a Notch2-dependent GFP^lo subset and a GFP^hi subset that resembles monocytes (Lewis et al., 2011). Whereas the proportions of these two populations were unchanged in Mtg16^−/− spleens, the GFP^lo subset had lower levels of GFP, and a novel GFP⁻ population was apparent (Fig. 3 D). Although the significance of these changes is unclear, they may indicate enhanced differentiation of Notch2-dependent DCs from pre-DCs.

We asked whether the enhanced differentiation of cDCs at the expense of pDCs originates in DC progenitors. Using the Cx3cr1-GFP reporter to define early DC progenitors (Fig. S2), we found that MPs and monocyte/DC progenitors were reduced in Mtg16^−/− BM, whereas CDPs were unchanged (Fig. 4 A). In contrast, the population of MHC II⁻ Flt3⁺ Sirpα^lo CD11c^lo pre-DCs as originally defined by Naik et al. (2006) and Liu et al. (2009) was increased in the BM and spleen (Fig. 4 B). A recent analysis of pre-DCs in the BM revealed their substantial heterogeneity, such as the expression of pDC marker SiglecH and the retention of pDC potential in a fraction of pre-DCs (Satpathy et al., 2012a). We used pDC marker Bst2 to rigorously exclude pDCs from the pre-DC population, which was then further resolved using SiglecH and an early pDC marker Ccr9. Similar to Satpathy et al. (2012a), we found that WT BM pre-DCs contained a large fraction of Ccr9⁻ SiglecH⁺ cells, which could give rise to both cDCs and pDCs in Flt3L-supplemented culture (Fig. S2). This pre-DC subset was virtually absent in the WT spleen (Fig. S2), consistent with the terminal cDC commitment of splenic pre-DCs. Notably, the SiglecH⁺ subset comprised the majority of pre-DCs in Mtg16^−/− BM, and its fraction and number were strongly increased in Mtg16^−/− spleen (Fig. 4, C and D). These data suggest that Mtg16 deficiency promotes the development of SiglecH⁺ pre-DCs and possibly their exit to the periphery, where they are likely to undergo differentiation toward cDCs.

To explore the mechanism of Mtg16 activity in pDC development, we compared genome-wide expression profiles of pDCs from the BM of Mtg16^−/− and control mice (Dataset S1). We found that Id2 was among the top differentially expressed genes, ranking as the #7 up-regulated gene in Mtg16-deficient pDCs (Fig. 5 A and Dataset S1). A nearly 10-fold higher level of Id2 expression in the bulk Mtg16^−/− pDC population was confirmed by quantitative RT-PCR (qRT-PCR; Fig. 5 B). To analyze the expression of Id2 at the single-cell level, we used a reporter strain in which a cell surface marker (human CD5 [hCD5]) had been inserted into the 3′ UTR of the Id2 locus (Jones-Mason et al., 2012). In control reporter mice, pDCs were hCD5 negative, whereas Mtg16^−/− pDCs uniformly expressed hCD5 (Fig. 5 C).

To analyze the cell type specificity of Id2 induction, we first established the pattern of Id2-hCD5 expression. Consistent with the essential role of Id2 in CD8⁺ cDCs and NK cells, the Id2-hCD5 reporter was expressed at the highest level in these cell types (Fig. 5 D). It was also expressed in other lineages including T cells and monocytes, but was absent from pDCs and B cells (Fig. 5 D). In contrast to the induction of hCD5 in Mtg16-null pDCs, other cell types did not show consistent changes in hCD5 expression (Fig. 5 E). Reflecting the loss of pDC-related CD8⁺ cDC population (Fig. 3 D), the CD5^lo CD8⁺ cDCs corresponding to this population were absent from Mtg16^−/− spleens (Fig. 5 E). Thus, the loss of Mtg16 causes the aberrant induction of Id2 transcription specifically in pDCs.

We analyzed the stage of Id2 induction during DC development in Mtg16^−/− animals. No hCD5 expression was detected in WT BM progenitors including long-term HSCs, Flt3⁺ multipotent progenitors, MPs, and DC progenitors including canonical (CD115⁺) and putative pDC-biased (CD115⁻) CDPs (Fig. 6 A). Furthermore, no induction of Id2 was observed in any of these early progenitors in Mtg16^−/− BM. In contrast, Mtg16^−/− pre-DCs (Lin⁻ MHC II⁻ B220⁻ Sirpa^lo Flt3⁺ CD11c^lo) showed an induction of Id2 expression, both in the BM and particularly in the spleen (Fig. 6 B). The strong induction of Id2 was also detected by qRT-PCR in the Ccr9⁻ SiglecH⁺ fraction of pre-DCs from the BM and spleen of Mtg16-deficient mice (Fig. 6 C). Conversely, the expression of E2-2 was decreased approximately twofold, consistent with impaired pDC potential. Thus, the loss of Mtg16 leads to precocious induction of Id2 in pre-DCs, which is consistent with the enhanced differentiation of the CD8⁺ cDC subset.

We analyzed the mechanism of Mtg16 activity in pDCs using human pDC lymphoma cell lines, which recapitulate the expression program and function of pDCs (Chaperot et al., 2006; Karrich et al., 2012). Because ETO proteins are known to associate with E proteins, we tested whether MTG16 interacts with E2-2, the predominant E protein in pDCs. Both MTG16 and E2-2 were expressed in the human pDC cell line Gen2.2 but not in T cell lymphoma MOLT-4 (Fig. 7 A). Immunoprecipitation of Mtg16 from Gen2.2 cells revealed its association with endogenous E2-2 (Fig. 7 A). E proteins including HEB and E2-2 (Sepp et al., 2011) comprise two isoforms, of which only the long isoform is capable of binding ETO proteins (Guo et al., 2009). Only the long isoform of E2-2 was found associated with endogenous MTG16, further confirming the binding specificity (Fig. 7 A).

To analyze the binding of Mtg16 to chromatin in pDCs, we performed parallel chromatin immunoprecipitation (ChIP) followed by sequencing (ChIP-Seq) of MTG16 and E2-2 in the human pDC cell line CAL-1. The promoter of MTG16 (CBFA2T3) itself is a target of binding by Mtg16 (Soler et al., 2010) and indeed was among the top MTG16-binding targets in CAL-1 (Fig. 7 B and Dataset S2). Consistent with the lack of direct DNA binding, MTG16 ChIP yielded lower chromatin enrichment and fewer binding peaks than E2-2 ChIP (Dataset S2). Strikingly, however, 2,050 out of 2,332 MTG16-bound genes (88.8%) were also occupied by E2-2, and 1,692 (82.5%) of those common genes harbored exactly coinciding peaks of MTG16 and E2-2 binding (Dataset S1). The coincident binding of MTG16 and E2-2 was observed on all top targets of MTG16 such as CBFA2T3 (Fig. 7 A). It was also observed on pDC-enriched targets of E2-2 (Ghosh et al., 2010), including PTCRA, TLR9, and PACSIN1 (Fig. 7 C), suggesting that MTG16 is a general cofactor rather than a specific corepressor of E2-2.

To analyze sequence specificity of DNA binding, we used a biophysically motivated, position-specific affinity matrix (PSAM) approach that is equal or superior to other approaches for motif discovery in ChIP data (Weirauch et al., 2013). The analysis of E2-2 peaks using the PSAM-based algorithm MatrixREDUCE revealed a single major sequence motif (CAgcTG) consistent with the canonical E box (Fig. 7 D). The same sequence also represented the top motif in MTG16-bound ChIP peaks in CAL-1 cells, but not in mouse erythroleukemia cells (not depicted; Soler et al., 2010). To further test the overlapping binding preferences of E2-2 and MTG16 in CAL-1, we used the defined E2-2 motif to predict the binding of MTG16 to DNA. The prediction ability of the motif can be quantified by the regression coefficient (“slope”) of the underlying linear model. We found that the E2-2 motif was highly predictive not only of E2-2 but also of MTG16 binding (p-value <10⁻¹⁶) in CAL-1 cells (Fig. 7 E). These data demonstrate that MTG16 recapitulates the binding specificity of E2-2, suggesting that it is recruited to DNA predominantly through E2-2 in pDCs.

The ChIP-Seq analysis revealed strong peaks for both MTG16 and E2-2 enrichment at ∼105 and ∼135 kb 5′ of ID2 (Fig. 7 F). This region has no adjacent genes other than ID2 but was recently shown to contain multiple enhancers, including those specific for cDCs or monocytes (Andersson et al., 2014). We also observed a minor MTG16 peak coinciding with a prominent E2-2 peak within ∼5 kb of the ID2 promoter region. The binding of the ID2 promoter-proximal regions by MTG16 and E2-2 was confirmed by qPCR analysis of unamplified ChIP material (Fig. 7 G). Although the function of the distal or proximal MTG16–E2-2–binding sites in ID2 repression remains to be clarified, these data reveal the recruitment of both proteins to the ID2 locus in pDCs. Together with the reported role of E2-2 in Id2 repression (Ghosh et al., 2010), they suggest that the MTG16–E2-2 complex binds to and represses the ID2 locus in pDCs.

To confirm that de-repression of Id2 contributes to the impaired pDC development in Mtg16-deficient animals, we tested whether it could be rescued by Id2 deficiency. Because Id2-deficient mice die perinatally on C57BL/6 background, we intercrossed Mtg16^−/− Id2^+/− animals to obtain Mtg16^−/− Id2^−/− or Mtg16^−/− Id2^+/+ littermate embryos. Intercrosses of WT animals produced control Mtg16^+/+ embryos. Hematopoietic reconstitution of adult recipients with Mtg16^−/− fetal livers (FLs) was inefficient (not depicted), reflecting the defective function of Mtg16-null HSCs (Fischer et al., 2012). We therefore analyzed pDC development in the embryo, which commences early during definitive hematopoiesis (Dakic et al., 2004). The pDC population could be detected with specific markers in the FLs of WT embryos (Fig. 8 A). The fraction of pDCs was reduced approximately twofold in Mtg16^−/− FLs but increased several-fold in Mtg16^−/− FLs that were also deficient for Id2 (Fig. 8, A and B). The changes in phenotypic pDC population were reflected in CpG-induced IFN-α secretion by cultured FL cells, which was significantly increased in Mtg16^−/− Id2^−/− FL (Fig. 8 C). Furthermore, the reduced Bst2 levels on Mtg16^−/− pDCs were increased by Id2 deletion (Fig. 8 D), although the levels of other markers were not affected consistently (not depicted). Thus, concomitant loss of Id2 facilitates the development and function of Mtg16-deficient pDCs, suggesting that Id2 induction contributes to their abnormal development.

Our analysis of Mtg16-deficient mice revealed reduced numbers, aberrant surface phenotype, and impaired function of pDCs, collectively resulting in a failure of pDC-dependent IFN response in vivo. In contrast, Mtg16 deletion caused an expansion of cDC progenitors (pre-DCs) and of the canonical CD8⁺ cDC subset. Unlike other known regulators of DC development (Satpathy et al., 2012b), the loss of Mtg16 changed DC lineage allocation and enhanced one subset development at the expense of another. These data provide strong genetic evidence for a common developmental pathway of DC development and for the affiliation of pDCs with the DC lineage despite their many lymphoid features (Reizis et al., 2011b). Our data suggest that Mtg16 acts as a molecular “rheostat” that facilitates pDC development and restricts the CD8⁺ cDC population, whose expansion may be detrimental for antibacterial immunity (Sathaliyawala et al., 2010).

The development of pDCs requires E protein E2-2, whereas its antagonist Id2 is expressed in cDCs and is required for CD8⁺ cDC development. Notably, Id2 is expressed in most immune cell types but is specifically excluded from pDCs and B cells, the two lineages with high E protein activity conferred by E2-2 and E2a, respectively. Thus, the development of pDCs involves both the induction of E2-2 and the repression of Id2, although the mechanisms of either event remained unclear. We found that Mtg16 deletion caused de-repression of Id2 in pDCs, effectively reducing the net E protein activity in these cells. Indeed, many features of pDCs in Mtg16^−/− mice resembled those in E2-2^+/− mice (Cisse et al., 2008), including reduced pDC numbers, decreased levels of certain pDC markers (Ccr9, CD4), and impaired IFN production capacity. In addition, Mtg16^−/− mice lacked the Cx3cr1-GFP⁺ CD8⁺ cDC population, whose development depends on E2-2 (Bar-On et al., 2010). Defective pDC development in Mtg16^−/− embryos was enhanced by Id2 deletion, suggesting a genetic interaction between Mtg16 and Id2. Similarly, Id2 was induced in pre-DCs, likely facilitating their differentiation toward the Id2-dependent CD8⁺ subset. Collectively, these results suggest that Mtg16 controls subset allocation during DC development in part by repressing Id2 expression and thereby enhancing E protein activity.

The roles of Mtg16 in facilitating pDC development and restricting cDC development might be explained either by independent activities of Mtg16 in committed pDC and cDC progenitors or by a rheostat function in a single common progenitor. Although the former possibility cannot be excluded, our data favor the latter scenario. In particular, the SiglecH⁺ subset of pre-DCs (Satpathy et al., 2012a) may represent a transition stage between CDPs and fully committed pre-DCs, and as such retain both pDC and cDC potential (at least in vitro). Importantly, this pre-DC subset showed the induction of Id2 in the Mtg16-deficient BM and elevated cell numbers in Mtg16-deficient spleen. It is therefore likely that pre-DCs with elevated Id2 levels undergo precocious cDC commitment, exit into the periphery, expand, and differentiate into CD8⁺ cDCs. It has been demonstrated that DC potentials can be predetermined at early stages of hematopoiesis (Naik et al., 2013), emphasizing the progressive restriction of lineage potentials in DC progenitors. Our results suggest that Mtg16 represents a cell-intrinsic factor that restricts the cDC potential at late CDP/early pre-DC stage in the BM.

ETO proteins in general and Mtg16 in particular can interact with multiple transcription factors including E proteins (Zhang et al., 2004), Notch receptor intracellular domains (Hunt et al., 2011), and a GATA/SCL/LMO-containing complex (Soler et al., 2010). Our ChIP data suggest that in pDCs, Mtg16 is recruited to chromatin primarily through the E protein E2-2. Notably, Mtg16 was recruited even to the genes activated by E2-2, despite the proposed role of ETO proteins as inhibitors of E protein–mediated transcriptional activation (Zhang et al., 2004; Guo et al., 2009). Therefore, ETO proteins may represent general components of chromatin-bound E protein complexes, whereas additional factors determine the effect of these complexes on target gene activation or repression. Importantly, Mtg16 and E2-2 are jointly bound to the ID2 locus in pDCs, and the deletion of Mtg16 (this study) and of E2-2 (Ghosh et al., 2010) causes the induction of Id2 in these cells. Although the induction of the Id gene expression by an E protein has been described in invertebrates (Bhattacharya and Baker, 2011), the repression has not been observed previously. Thus, the DC lineage commitment involves a positive feedback mechanism whereby E proteins directly repress the transcription of their inhibitor Id2 through ETO protein corepressors.

Collectively, our results suggest a model whereby ETO protein Mtg16 regulates pDC versus cDC lineage commitment. Uncommitted DC progenitors express low levels of both E2-2 and Id2, with high-level Id2 expression repressed by Mtg16 in complex with E2-2 and possibly other transcription factors. Stochastic or induced up-regulation of E2-2 may facilitate Mtg16-mediated repression of Id2, enhancement of net E protein activity, and pDC commitment. Conversely, up-regulation of Id2 would inhibit E protein activity and thereby dislodge Mtg16 from the Id2 locus, leading to further up-regulation of Id2 and differentiation along the cDC pathway. Because the deletion of E2-2 causes Id2 up-regulation and spontaneous differentiation into cDCs (Ghosh et al., 2010), the cDC lineage likely represents the “default” pathway of DC differentiation. This scenario describes a mechanism of pDC versus cDC commitment driven by mutually antagonistic transcription factors, a common mechanism of lineage specification in development.

All animal experiments were performed according to the investigator’s protocol approved by the Institutional Animal Care and Use Committee of Columbia University. Mtg16^+/− mice (Chyla et al., 2008) on pure C57BL/6 background were intercrossed to generate homo- or heterozygous progeny. Age-matched WT C57BL/6 mice bred in the same animal facility were used as controls; where indicated, heterozygous Mtg16^+/− mice were used as littermate controls. The Id2-hCD5 reporter allele (Jones-Mason et al., 2012; provided by Y. Zhuang, Duke University, Durham, NC) and Cx3cr1-GFP reporter allele (Jung et al., 2000) on C57BL/6 background were crossed to Mtg16^+/− mice and used as heterozygotes. The Id2-null allele was created by replacing the first exon of Id2 with a LacZ marker and selection cassette, resulting in a functional null allele as judged by the expected absence of lymph nodes and CD8⁺ cDCs in homozygous mice. The resulting Id2^+/− mice were backcrossed to C57BL/6.

For the analysis of Mtg16^−/− Id2^−/− embryos, timed matings were performed between Mtg16^+/− Id2^+/− and Mtg16^−/− Id2^+/− animals, and FLs were isolated for genotyping and analysis on days 17.5–18.5 postcoitum. Matings between WT animals were used to generate control embryos. For competitive hematopoietic reconstitution, BM cells from Mtg16^−/− or control Mtg16^+/− mice were mixed with BM cells from C57BL/6 mice congenic for CD45.1 (B6.SJL) at 95:5% ratio and injected i.v. into lethally irradiated B6.SJL recipient mice. For IFN induction in vivo, mice were injected i.v. with 5 mg CpG type A oligonucleotide (ODN 2216; InvivoGen) complexed with DOTAP (Roche; 30 ml DOTAP/150 ml total volume) or with 0.3 mg poly-I:C (GE Healthcare). Serum IFN before injection or 6 (for CpG) or 12 h (for poly-I:C) after injection was measured by ELISA using anti–mouse IFN-α antibodies (PBL Interferon Source).

Single-cell suspensions were stained for multicolor analysis with the indicated fluorochrome-conjugated antibodies (BD or eBioscience). The samples were acquired on an LSRII flow cytometer or sorted on FACSAria flow sorter (BD) and analyzed with FlowJo software (Tree Star). In all experiments, pDCs were defined by a combination of at least two pDC-enriched markers (CD11c, B220, and/or Ly6C) with at least one pDC-specific marker (Bst2 and/or SiglecH). Lineage-negative cells were defined as CD3⁻, CD19⁻, NK1.1⁻, Ter119⁻, Ly6G⁻, B220⁻. For pDC development in vitro, total BM cells (2 × 10⁶/ml) were cultured for 7 d with 100 ng/ml recombinant human Flt3L (PeproTech). For competitive BM cultures, donor BM cells were mixed 1:1 with B6.SJL BM cells and cultured as above. For the culture of DC progenitors, Lin⁻ Sca-1⁻ c-Kit^+/lo Flt3⁺ cells were sorted from pooled BM and cultured with Flt3L as above. To measure IFN production, total cell suspensions or sorted pDCs were incubated for 18 h in the presence of 2 µM CpG type A, and IFN-α in the supernatants was measured by ELISA.

For microarray analysis, BM cells from WT control or Mtg16^−/− mice were pooled and enriched by negative selection of lineage (CD11b, CD19, DX5, TER119)-negative cells using magnetic microbeads (Miltenyi Biotec). The pDCs (Lin⁻ CD11c^int 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 analyzed using the NIA Array software. For qRT-PCR analysis, RNA from a separately sorted pDC sample was reverse transcribed and assayed by SYBR Green-based real-time PCR. The expression of the Id2 was normalized to that of β-actin and expressed relative to the WT sample via the ΔΔC_T method.

Human pDC lymphoma cell lines CAL-1 (Maeda et al., 2005) and Gen2.2 (Chaperot et al., 2006) and a control human T cell lymphoma MOLT-4 were maintained in complete RPMI medium with 10% FCS under regular tissue culture conditions. For coimmunoprecipitation, Gen2.2 and MOLT-4 cells were lysed using RIPA buffer, and ∼1 mg of whole cell lysate was subjected to immunoprecipitation with anti-Mtg16 polyclonal antibody (C-20; Santa Cruz Biotechnology, Inc.) or IgG control using protein G–agarose beads. The resulting samples were analyzed by Western blotting using anti-Mtg16 antibody or anti–E2-2 monoclonal antibody (clone 367.2; Santa Cruz Biotechnology, Inc.).

For ChIP, CAL-1 cells were cross-linked with formaldehyde for 10 min, sonicated to yield chromatin fragments of ∼200 bp, and subjected to immunoprecipitation with anti-Mtg16 polyclonal antibody or IgG control. E2-2 was immunoprecipitated using a newly developed rabbit monoclonal antibody that recognizes human E2-2 but not related E protiens E2a or HEB. After cross-link reversal, the isolated chromatin was analyzed by qPCR for the indicated ∼5-kb region of ID2. The enrichment was normalized to 10% of total chromatin input of the respective sample and expressed as fold enrichment relative to control IgG sample.

Isolated chromatin samples were submitted to the Yale Center for Genome Analysis for amplification, library construction, and sequencing. The samples were sequenced on a single-end version 3 Illumina flow cell on a HiSeq 2000. Sequencing reads were aligned to the human genome (UCSC hg19), and the enrichment peaks were identified, separated, and annotated using the default parameters of MACS 1.4.0 software. Sequence peaks were visualized using DNAnexus genome browser. For motif discovery, we used the BayesPeak algorithm (Spyrou et al., 2009) with a window size set to 100 bp (approximately half of the mean DNA fragment length) to identify 71,326 nonoverlapping peaks with a posterior probability (PP) score of >0.1 in the ChIP-Seq profile for E2-2. MatrixREDUCE (Foat et al., 2006) was used to fit a PSAM (“motif”) that optimally explained the variation in PP score across all peaks from the underlying sequence.

Statistical significance was estimated with an unpaired, two-tailed Student’s t test and is indicated as follows: *, P ≤ 0.05; **, P ≤ 0.005; ***, P ≤ 0.0005.

The ChIP-Seq and microarray data have been deposited in the Gene Expression Omnibus (GEO) database under the accession no. GSE43963.

Fig. S1 shows a representative gating strategy to define pDCs in the BM of Mtg16-deficient mice. Fig. S2 characterizes DC progenitors. Dataset S1, included as a separate Excel file, shows genome-wide expression analysis of Mtg16^−/− pDCs. Dataset S2, included as a separate Excel file, shows ChIP-Seq results for MTG16 and E2-2 in CAL-1 cells.

We thank Y. Zhuang for the Id2 reporter strain, G. Della Gatta, A. Ferrando, and J. Overton for help with ChIP-Seq experiments, and A. Singhee and S. Weisberg for help with ChIP-Seq analysis.

This work was supported by National Institutes of Health (NIH) grant AI072571 and New York State Dept. of Health grant N09G-22 (to B. Reizis), an American Society of Hematology Postdoctoral Fellowship (to H.S. Ghosh), NIH grants AI101251 and AR044535 and the Dana Foundation Neuroimmunology grant (to K. Liu), and NIH grants HG003008 and CA121852 (to H.J. Bussemaker).

The authors declare no competing financial interests.