Dendritic cell fate is determined by BCL11A

Confirmation That Bcl11a-Deficient Fetal Liver and Spleen Is Devoid of pDC.

In agreement with previous analyses of conventional Bcl11a^−/− knockout (KO) mice by Liu et al. (5), we observed an absence of B220⁺ cells in liver and spleen isolated from fetal KO mice (Fig. S1A). The loss of B220⁺ cells appeared dose-dependent because Bcl11a^+/− heterozygotes contained B220⁺ populations intermediate between Bcl11a^+/+ wild-type and Bcl11a^−/− homozygous-deficient siblings. In addition to the previously documented loss of B220⁺ B lymphocytes in these mice, we conjectured that B220⁺ pDCs might also be absent from these tissues.

To test the role of Bcl11a in murine pDC development, we isolated hematopoietic cells from the fetal liver and spleen of Bcl11a^+/+, Bcl11a^+/−, and Bcl11a^−/− E18 embryos. Examination of fetal organs was mandatory because constitutive homozygous deficiency of Bcl11a is neonatal lethal. In mouse, pDCs are B220⁺ Cd11c⁺ Cd11b^–, and express the specific marker Bst2/PDCA1/CD317. In n = 3 independent experiments, pDCs were missing from Bcl11a^−/− KO mice. Collectively, the missing cells were Cd11c⁺ CD11b^– Cd19^– RB6-8C5⁺ B220⁺ PDCA1⁺. In one experiment, pDC were reduced 25-fold in the Bcl11a KO(^−/−) compared with the wild-type littermate control (^+/+) (Fig. S1B).

We could not detect Siglec-H transcripts in Bcl11a^−/− fetal liver precursors in vivo (Fig. S1C) or ex vivo (Fig. 1C), whereas Siglec-H expression was detectable in heterozygous and wild-type controls. Though immature pDCs (PDCA1^–) have been reported to be negative for Siglec-H transcripts, mature pDCs (PDCA1⁺) are routinely positive for expression of this gene (18). In light of this observation, we could not exclude the possibility that pDC progenitors were blockaded at an intermediate stage of differentiation; this could result from cell-extrinsic Bcl11a-dependent effects modulated by accessory cells or the stromal milieu in general.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

In vitro lack of pDC development in cultures of Bcl11a-deficient fetal liver or adult bone marrow. (A–D) Fetal liver cells from Bcl11a^−/−, Bcl11a^+/−, and Bcl11a^+/+ mice were cultured in vitro with Flt3-L to induce the growth of cDC and pDC from progenitor cells. (A) Day 3 of culture (Upper) resulted in efficient production of cDCs (Cd11b⁺ Cd11c⁺) among Bcl11a^+/+ and Bcl11a^+/− genotypes, but a near absence of pDCs (Cd11b^– Cd11c⁺) in the Bcl11a^−/− homozygous knockout. Day 10 cultures (Lower) continued to exhibit this marked reduction in pDC generation (>10-fold compared with wild type). Results depicted are representative of n = 3 independent experiments. (B) Additional markers (B220, Ly6c) were used to track the outgrowth of cells in additional experiments (n = 3), confirming as above that only a minute portion of pDC-like cells (Cd11b^– Cd11c⁺) could be detected in Bcl11a^−/− cultures. (C) The pDC transcript Siglec-H was undetectable by RT-PCR in day 3 Flt3-L culture of Bcl11a^−/− fetal liver progenitors. (D) Bcl11a^−/− fetal liver culture (Flt3-L 7 d) failed to mount a measurable IFN-α response to type A CpG oligonucleotide (ODN D19) after 24 h of stimulation. (E–J) Adult bone marrow cells from pI:pC-treated Bcl11a^F/FMx1-creYFP⁺ conditional knockout (cKO) mice or littermates (8 wk old) were cultured in vitro with Flt3-L to induce the growth of cDC and pDC from progenitor cells. (E) Flow cytometry of day 10 Flt3-L culture exhibited no pDC production from YFP⁺ (Bcl11a-deleted) bone marrow cells, whereas as shown in F the production of cDC was unimpeded and equivalent to control (n = 3 independent experiments). (G and H) Cultures of GM-CSF–supplemented cells (day 7) produced statistically similar frequencies of cDCs in both YFP⁺ (Bcl11a-deleted) and wild-type bone marrow. (I) Total cell numbers in Flt3-L cultures were drastically reduced (day 10; P = 0.008; Student t test), whereas total cells in GM-CSF stimulations were reduced only modestly (day 7; P = 0.04). (J) cDC-to-pDC ratios in Flt3-L–supplemented cultures were significantly different in YFP⁺ Bcl11a-deleted cells vs. total bone marrow cells (207.6 ± 87.7 vs. 3.05 ± 0.63, respectively, P = 0.018).

Bcl11a Is Required for Cytokine-Induced pDC Differentiation from Fetal Liver Progenitors in Vitro.

The capacity of Bcl11a-deficient fetal liver cells to differentiate into pDCs was further investigated. Bcl11a^−/− fetal liver cells were cultured in defined cytokine-supplemented media to reveal whether pDC precursors, which might persist in vivo, could be differentiated ex vivo. The development of pDCs and cDCs depends crucially on the hematopoietic cytokines Flt3-ligand (Flt3L) and GM-CSF in both human and mouse (10). Flt3L alone induces the growth of cDC and pDC in vitro from progenitor cells, and moreover, these are the only cell populations generated in Flt3L-supplemented bone marrow cultures (33). In contrast, GM-CSF exclusively promotes the differentiation of cDC from early hematopoietic progenitors and induces a single population of Cd11c⁺ Cd11b⁺ B220^– cDCs, but not Cd11c⁺ Cd11b^– B220⁺ pDCs, in culture (33).

In agreement with the observations made in vivo using KO mice, the ex vivo differentiation of Bcl11a^−/− hematopoietic fetal-liver precursors failed to generate prototypical Cd11c⁺ Cd11b^– Ly6c⁺ B220⁺ pDCs in Flt3L-supplemented cultures (Fig. 1 A and B). Bcl11a^−/− KO cultures consisted almost entirely of Cd11c^+/− Cd11b⁺ B220^– myeloid cells and cDCs at all three time-points examined (days 3, 7, and 10). pDC and pDC-like cells that persisted were on average reduced >10-fold compared with cultures derived from littermate controls. In contrast to the profound impedance of pDC differentiation, Bcl11a^−/− cDCs in Flt3L cultures appeared to differentiate fully, although clearly reduced in their relative percentage (Fig. 1 A and B). As reported in a recent study (9), cDC generation from Bcl11a^−/− KO fetal precursors in control experiments using GM-CSF is abundant and unperturbed. The expression of PDCA1 was low among all Flt3L-derived pDCs, as previously documented (34), so the additional surface-marker Ly6c was included along with Cd11b, Cd11c, and B220 to track the outgrowth of cells using flow cytometry analysis. These cellular phenotypic profiles were confirmed in three independent experiments. Moreover, similar to RT-PCR of KO whole fetal liver, Siglec-H transcripts were undetectable among fetal liver precursors cultured in Flt3L for 3 d (Fig. 1C), underscoring the absence of this pDC-specific molecular program. The lack of a functional pDC phenotype in Bcl11a^−/− mice was also supported by the lack of IFN-α production subsequent to stimulation of cultured cells with the type A CpG oligodeoxynucleotide D19, as determined by ELISA (n = 1) of culture supernatants (Fig. 1D).

Bcl11a Is Required for Cytokine-Induced pDC Differentiation from Adult Bone Marrow in Vitro.

To determine whether the defect in fetal pDC development observed among Bcl11a-deficient conventional knockout mice is retained in adults, we constructed a floxed (F) allele of Bcl11a (Fig. S2 and SI Text). Our strategy results in deletion of the first exon, which is included in all previously characterized Bcl11a transcripts in mouse and man (2, 4, 35, 36). We crossed these mice to transgenic mice in which Cre recombinase is driven by the promoter of the Mx-1 gene (37). The Mx1-cre transgene is expressed in hematopoietic stem cells (HSCs) and in all downstream lineages after induction with type I IFN or poly(I):poly(C) (pI:pC), allowing reliable deletion throughout the entire adult hematopoietic compartment. These mice were also crossed to a floxed STOP-YFP reporter mouse to positively identify the cells in which Mx1-Cre recombinase is activated.

Young adult (5 wk) Bcl11a^F/FMx1-CreYFP⁺ mice underwent pI:pC induction, and bone marrow was collected at 8 wk for Flt3L and GM-CSF–supplemented cultures. In agreement with the observations made in vivo and ex vivo using KO mice, YFP⁺ bone marrow precursors failed to generate Cd11c⁺Cd11b^– YFP⁺ pDCs in Flt3L-supplemented cultures (Fig. 1E). To confirm phenotype, ∼75% of the Cd11c⁺CD11b^– pDCs in the control cultures were weakly positive for both PDCA-1 and Siglec-H. Interestingly, Bcl11a^F/FMx1-CreYFP⁺ cultures consisted of substantial numbers of Cd11c^+/−Cd11b⁺ MHC II^hi YFP⁺ cDCs at day 10 (Fig. 1F), indicating a difference in cDC development between fetal and adult precursors. The cDC to pDC ratio was significantly altered in YFP⁺ cells compared with total bone marrow in Flt3L supplemented cultures (Fig. 1J; 207.6 ± 87.7 to 3.05 ± 0.63, respectively, P = 0.018). In GM-CSF–supplemented cultures, cDCs differentiated normally and abundantly from YFP⁺ bone marrow (Fig. 1 G and H). These cellular phenotypic profiles were confirmed in triplicate and in n = 3 independent experiments. Though total cell numbers were greatly reduced in Flt3L-supplemented cultures [15.8 × 10⁶ vs. 3.43 × 10⁶ in control and conditional KO (cKO), respectively], there was also mild reduction of total cells in GM-CSF–supplemented cultures (51.1 × 10⁶ vs. 35.7 × 10⁶ in control and cKO, respectively; Fig. 1I), indicating either a slight reduction in cDC development efficiency or possibly a reduction in cell survival.

Deletion of Bcl11a in Hematopoietic Precursors Results in a Robust Loss of pDCs but Spares Lin⁻ Sca-1⁺ c-kit⁺ Progenitors, T cells, and cDCs.

The floxed (F) allele of Bcl11a was crossed to transgenic mice in which Cre recombinase is driven by the promoter of the Vav gene (38). The iVav-cre transgene is expressed in HSCs and in all downstream lineages (39), allowing reliable deletion throughout the entire embryonic and adult hematopoietic compartment.

Young adult (6 wk) Bcl11a^F/FVav-Cre mice consistently displayed a radical loss of B cells and pDCs in bone marrow, with a relative 10- to 50-fold reduction compared with control littermates (n = 5 independent experiments) as determined by PDCA1, CD11c, and B220 staining (Fig. 2A). The typical loss of B cells in any given experiment was observed to be proportional to the loss of pDCs, and the approximate B:pDC ratio of 10:1 typically observed in controls was also maintained in Bcl11a^F/FVav-Cre knockout mice irrespective of the degree of cellular deletion in the bone marrow. An examination of the periphery resulted in a similar pairwise reduction of B cells and pDCs in the spleen (Fig. 2A); however, here the reduction was not as pronounced as in bone marrow. Importantly, splenic cDCs (B220^– CD11c^high CD11b⁺ MHCII^high) were unaffected by Bcl11a deletion (Fig. 2B), as were CD8⁺ or CD8⁻ cDC subsets (CD11c⁺ CD8⁺ CD11b⁻ MHCII⁺ B220⁻; CD11c⁺ CD8⁺ CD11b⁻ MHCII^high B220⁻, respectively; Fig. 2C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Conditional elimination of Bcl11a in vivo leads to loss of pDC cellularity and function without perturbation of cDCs. (A) pDC are depleted in Bcl11a conditionally deleted adult mice. Homozygous Bcl11a cKO mice bearing loxP-flanked (F) targeted alleles (Bcl11a^F/F) were intercrossed with the Vav-Cre deleter C57/BL6 strain (Vav-Cre⁺ F/F), which expresses Cre recombinase in HSC. Dot plot of B220⁺ PDCA1⁺ cells (Left) or Cd11c^int PDCA⁺ cells (Right) shows a near complete loss of pDCs in the bone marrow of Vav-Cre⁺ F/F mutants compared with littermate controls (10-fold to 50-fold loss; n = 5 independent experiments). The simultaneous loss of B220⁺ PDCA1^– B cells served as an internal control for Cre-mediated deletion efficiency. A similar loss of pDCs was confirmed in the spleen (Center), although the fold reduction (∼fivefold) was not as pronounced as in bone marrow. (B) The generation of cDC (B220^– MHC II^hi CD11c^hi CD11b⁺) was unaffected by Vav-Cre–mediated knockout of Bcl11a. (C) The frequency of CD8⁺ and CD8⁻ cDC subsets in the spleen were statistically indistinguishable between Vav-Cre⁺ F/F and controls. (D) Although fewer overall LSK progenitors were detected in Vav-Cre⁺ F/F mice, both MDP and CDP progenitor populations persisted within the Flt3⁺ compartment in cKO mice at percentages similar to controls. MDP (Lin^– Flt3⁺ Sca-1⁻ CD115⁺ c-kit^hi); CDP (Lin^– Flt3⁺ Sca-1⁻ CD115⁺ c-kit^lo). (E) pDC are also depleted in Bcl11a^F/FMx1-Cre mice following induction of Cre-mediated deletion in HSC by treatment with pI:pC, but CDP and cDC are spared, analogous to results obtained with Vav-Cre deletion. (F) Hematopoietic lineages other than B and pDC were statistically unaffected in Vav-Cre⁺ F/F mice relative to Vav-Cre^– controls across independent experiments (n = 3; four per genotype) plotted as mean ± SEM (*P < 0.05; **P < 0.001; Student t test). (G) Vav-Cre⁺ F/F mice are significantly impaired in type I IFN response to viral infection with ∼1 × 10⁷ pfu of human HSV strain 17 i.v. by tail-vein injection (n = 1; four per genotype) as measured by ELISA (232 ± 44.4 to 518 ± 144 pg/mL; data as means ± SEM; P < 0.05, Mann–Whitney, two-tailed test).

To determine whether dendritic cell precursors were impaired, bone marrow was stained for various lineage markers, including those that discriminate common myeloid progenitors (CMPs; Lin⁻ Flt3⁺ Sca-1⁻ c-kit^high CD115⁺) and CDPs (Lin⁻ Flt3⁺ Sca-1⁻ c-kit^lo CD115⁺; Fig. 2D). Consistent with a recent report (6), we observed an ∼twofold reduction of Lin⁻ Sca-1⁺ c-kit⁺ cells (LSK; Fig. 3D) in Bcl11a^F/FVav-Cre bone marrow. However, neither macrophage-DC progenitor (MDP) nor CDP progenitor populations were substantially impaired in their relative percentages (Fig. 3D). Further, statistically similar relative percentages of cDCs (CD11c⁺ CD11b⁺ B220⁻), granulocytes (Gr-1⁺ CD11b⁺), macrophages (CD11b⁺ F4/80⁺), and T cells (CD3ε⁺ CD4CD8 DP and SP) were observed in bone marrow and spleen of cKO and control littermate mice (Fig. 2E). These observations are consistent with the original report (5) of conventional embryonic deletion of Bcl11a wherein B cells were selectively abolished yet T cells as well as myeloid and erythroid cells were preserved.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

ChIP-seq analysis of genome-wide BCL11A DNA binding in vivo. (A) Peak mapping along the loci of multiple pDC-related genes. ChIP-seq peaks from the CAL-1 human pDC (black, peak score ≥10) were determined and aligned with ENCODE Consortium ChIP-seq data for the human B lymphoblastoid cell line GM12878 (red, peak score ≥10). ID4 and the T-cell–specific paralogue, BCL11B, were chosen as negative controls because of their similarity to BCL11A and other IDs that contained highly correlated peaks. CAL-1 ChiP-seq resulted in smaller and fewer peaks than the ENCODE data (Fig. S3). (B) A pie chart representation of the distribution of BCL11A binding sites in six different genomic regions. Core promoters are within ±2 kb from the transcriptional start site (TSS); upstream is from 2 to 20 kb upstream from the TSS; and intergenic is a region not included as a promoter, upstream region, intron, or exon. The CAL-1 ChIP-seq peak distribution shows a striking similarity to that of the ENCODE Consortium’s GM12878 cell line. (C) Overlap of BCL11A target genes between the two cell lines. A target gene was defined by a binding site occurring within 50 kb upstream through the intron of that gene. Analysis of the false discovery rate and associated Q-values were performed using Benjamini–Hochberg statistics. (D) De novo motif analysis from BCL11A ChIP-seq in CAL-1 cells. Alignment of the top 500 and top 1,000 CAL-1 ChIP-seq peaks show strong similarity to the top-ranked BCL11A ChIP-seq motif found in GM12878 cells, an EICE consensus site. (E) EICE motifs comprise the top-ranked BCL11A binding motifs in the GM12878 cell line. EICE is a previously identified composite DNA binding site for the Ets factor PU.1/SPI1 and for the IFN regulatory factor 4 (IRF4) that mediates cooperative binding of these factors to DNA. MEME analysis was used to calculate the expectation E-value for the occurrence of this motif within the GM12878 coincident peak sequences of BCL11A, PU.1/SPI1, and IRF4.

We confirmed these observations in the aforementioned Bcl11a^F/FMx1-Cre adults following pI:pC-induced deletion (37). Here, too, B cells and pDCs were significantly reduced with little alteration of other cell populations (Fig. 2F). Except for the peculiar observation that T-cell absolute numbers are reduced exclusively among c-kit⁺ precursor T cells when tamoxifen-inducible CreERT2 is used to induce a global deletion of Bcl11a exon 4 (6), our results are consistent with all prior data derived from conventional embryonic Bcl11a-deleted mice wherein approximately normal T-cell development occurs. Furthermore the Mx1-Cre data eliminate the possibility that this difference with our results derived from sparing of T cells of embryonic origin.

To investigate the ability of Bcl11a^F/FVav-cre mice to produce type I IFN upon viral infection, we analyzed peripheral blood mononucleated cells of Cre-positive and Cre-negative littermates to assure expected phenotypes before delivering 1 × 10⁷ pfu of human HSV strain 17 i.v. by tail-vein injection. As shown in Fig. 2G, serum levels of IFN-α were markedly reduced 12 h postinjection in Bcl11a^F/FVav-Cre mice compared with Cre-negative littermate controls (232 ± 88.9 to 518 ± 288 pg/mL, respectively). Thus, Bcl11a-deficient mice have a reduced capacity to generate this critical pDC mediator of the innate immune response.

Genome-Wide BCL11A Target Genes Include Previously Established Regulators of pDC Biology.

ChIP-seq of the human CAL-1 cell line was used to identify genome-wide binding sites of BCL11A in pDCs. We found that BCL11A is recruited to the loci of several previously established pDC factors, including SPI1, SPIB, E2-2/TCF4, IRF4, MTG16, and ID2, and, interestingly, to its own proximal promoter region (Fig. 3A). The occupancy pattern of BCL11A in CAL-1 cells (within putative proximal promoters and intronic or intergenic loci) bore a striking resemblance to its binding distribution in the human EBV-transformed lymphoblastoid cell line (LCL) GM12878, based on published ENCODE Consortium data (www.factorbook.org/; Fig. 3B). Additionally, the top-ranked DNA binding motifs for the top CAL-1 and GM12878 binding sites were identical (Fig. 3 D and E and Fig. S4). Our data further suggest that some targets are bound in a cell-context dependent fashion (Fig. 3C). For example, whereas BCL11A was recruited to the same location in the ID3 locus in both the B and pDC cell lines, its recruitment to the ID2 locus was differential across the two lines. Ongoing studies of these and other cell-contextual patterns of BCL11A genomic binding are being confirmed in additional B-cell lines.

Evidence for a BCL11A-Regulated ID/E–Protein Interplay in Control of pDC Development.

Recent findings have linked Id2 and Id3 regulation to differentiation of cDCs and pDCs via transcriptional inhibition of E2-2/Tcf4 (24, 26, 40). Among our CAL-1 genome-wide target genes we observed strong BCL11A binding to ID3 (Fig. 3A). However, ID2 peaks were weak and did not align with the relatively strong ID2 peaks observed in GM12878 B cells (Fig. 3A). Using ChIP followed by endpoint PCR, we confirmed (n = 3) robust recruitment of BCL11A to the proximal promoter of ID3 in chromatin prepared from the CAL-1 cell line, but not from an amplicon 10 kB upstream or from a cell line (Hodgkin lymphoma L428) negative for BCL11A expression (Fig. 4B). Additionally, the proximal promoter of E2-2/TCF4 was also efficiently PCR amplified (n = 2) from CAL-1 chromatin (Fig. 4B). Within the ChIP regions of ID3 and E2-2 (base pairs –679 and –3762 upstream of their respective transcriptional start sites) reside evolutionarily conserved sequences (Fig. S4) that match the BCL11A binding consensus of Fig. 3D. In ID3 this putative BCL11A binding site is positioned <100 bp downstream of a MspI/HpaII restriction site whose differential methylation has been correlated with myeloid vs. lymphoid lineage choice (23) (Fig. 4C). Consistent with its absence in the CAL-1 ChIP-seq target list, we found no evidence for BCL11A recruitment to ID2. However, we confirmed binding of BCL11A to its own locus (Fig. 4B), raising the possibility that BCL11A is autoregulatory.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

BCL11A regulates ID2, ID3, MTG16, and E2-2/TCF4 in human B and pDC cell lines. (A) BCL11A isoforms (XS or XL) retrovirally overexpressed in Raji and BJAB Burkitt’s lymphoma B-cell lines altered transcription (average threefold) of the HLH proteins ID2 and ID3 compared with mock-transduced control cells. Multiple (n > 4) independent experiments tracked by multiple probe elements spotted per LymphoChip cDNA microarray. (B) Inducible shRNA silencing of BCL11A down-regulates transcription of ID3 and E2-2. (Left) CAL-1 pDC cell line stably transduced with inducible shRNA targeted to exon 2 of Bcl11a under the control of a doxycycline (DOX)-inducible H1 promoter containing TETR binding sites (59). RT-PCR with optimal cycling conditions in the CAL-1 pDC cell line revealed strong knockdown of BCL11A transcripts beginning 6 h postinduction and continuing for >24 h. ID3 and E2-2 transcripts were correspondingly reduced throughout the 24-h experiment, whereas ID2 was consistently, although transiently, induced; the irrelevant target gene CD37 was unaffected. GAPDH amplification shows equivalence of mRNA amplification and loading. (Right) At 24 h, shRNA-mediated BCL11A knockdown resulted in ID3 transcriptional inhibition >80%, and E2-2 of 60%, when normalized to the GAPDH housekeeping control gene. Mean ± SD depicted; n ≥ 3 experiments. (C) BCL11A is recruited to 5′ regulatory regions of ID3, E2-2, and itself. ChIP from CAL-1 human pDC cells using anti-BCL11A or control IgG antibodies was analyzed by endpoint PCR. Primers annealing 10 kb upstream of the ID3 TSS failed to amplify and thus served as a negative control. (D) Overexpression of BCL11A (XL) up-regulates ID3 promoter-driven reporter transcription in CAL1 pDCs and Raji B cells. (Upper) Schematic of the ID3 proximal promoter indicating the species-conserved EICE-like BCL11A DNA binding motif (Fig. S4); the location of BLIMP-1 and E-box motifs; and the HpaII methylation site. An 820-bp fragment spanning this region (indicated by arrows) that was consistently PCR amplified in ChIP experiments (indicated by wedges) was inserted upstream of luciferase in the pGL2 luciferase vector. (Lower) Dual luciferase activity was determined following transient transfection with increasing DNA levels (in nanograms) of the reporter into CAL1 human pDC cells or Raji B cells. Mean relative luciferase activity ± SEM obtained depicted; n = 2 independent experiments read in triplicate. (E) Knockdown of BCL11A results in MTG16 knockdown 24 h post-DOX induction in CAL-1 pDC cells.

As an initial test of function, we overexpressed the major BCL11A isoform (XL) or a smaller, less well-characterized (4) isoform (XS) as retroviruses in Raji B cells. We observed that both BCL11A-XL and -XS strongly up-regulated ID3 (average of 2.8- to 3.0-fold, respectively; Fig. 4A), whereas ID2 was uniformly repressed (2.6- to 3.0-fold) by BCL11A-XL and -XS. Next, inducible short-hairpin RNA interference in CAL-1 pDC cells was used to specifically target a region (exon 2) common to all BCL11A isoforms. We observed a robust knockdown of BCL11A transcripts as well as a corresponding reduction in both ID3 and E2-2 transcripts at select time points (Fig. 4C). Knockdown of BCL11A was nearly absolute by the end of the first 24 h of shRNA induction. Assayed at the 24-h time point and normalized to the housekeeping control gene GAPDH, shRNA-mediated BCL11A knockdown resulted in an ID3 transcriptional inhibition of over 80%, and E2-2 of 60%, whereas the irrelevant target CD37 was broadly unchanged. Direct ID3 regulation was confirmed by BCL11A up-regulation of ID3 promoter-driven luciferase activity in CAL1 and Raji (two- to 10-fold; Fig. 4D) transient transfections. Measurement at the earliest time point (6 h) showed, as expected, down-regulation of the direct E2-2 target gene, ID2 (19). However, at subsequent time points (12 h, 24 h), ID2 consistently and gradually returned to baseline levels, suggesting an indirect affect by BCL11A knockdown. As a potential mechanism, the ETO family corepressor MTG16, an established (41) negative regulator of ID2 (41) and a BCL11A target gene (Fig. 3A), was virtually depleted upon BCL11A shRNA knockdown (Fig. 4E). Cell viability was similar among experimental and control groups at all time points. These results collectively support a model for ID/E–protein interplay regulated by BCL11A through its direct transcriptional activation of the pDC-essential gene E2-2 and the E-protein antagonist ID3, while indirectly suppressing ID2 expression through direct activation of MTG16.