Ikaros cooperates with Notch activation and antagonizes TGFβ signaling to promote pDC development

Plasmacytoid and conventional dendritic cells (pDCs and cDCs) arise from monocyte and dendritic progenitors (MDPs) and common dendritic progenitors (CDPs) through gene expression changes that remain partially understood. Here we show that the Ikaros transcription factor is required for DC development at multiple stages. Ikaros cooperates with Notch pathway activation to maintain the homeostasis of MDPs and CDPs. Ikaros then antagonizes TGFβ function to promote pDC differentiation from CDPs. Strikingly, Ikaros-deficient CDPs and pDCs express a cDC-like transcriptional signature that is correlated with TGFβ activation, suggesting that Ikaros is an upstream negative regulator of the TGFβ pathway and a repressor of cDC-lineage genes in pDCs. Almost all of these phenotypes can be rescued by short-term in vitro treatment with γ-secretase inhibitors, which affects both TGFβ-dependent and -independent pathways, but is Notch-independent. We conclude that Ikaros is a crucial differentiation factor in early dendritic progenitors that is required for pDC identity.

Deficiency of the Ikaros zinc finger DNA-binding protein and tumor suppressor, encoded by the Ikzf1 gene, is associated with profoundly impaired DC development. Mice homozygous for a dominant-negative (dn) Ikzf1 mutation lack all cDCs, while animals with a null mutation predominantly lack cDC2s [13]. In contrast, mice carrying the hypomorphic Ik L/L mutation show a selective block in bone marrow (BM) pDC development, leading to an absence of peripheral pDCs, although cDCs appear normal [14]. These studies highlight the sensitivity of the DC lineages to Ikaros levels, where pDC development requires more Ikaros function than cDCs. In man, patients with germline IKZF1 mutations also exhibit reduced pDC, but not cDC numbers, indicating a conserved role for Ikaros in DC development [15]. Interestingly, IKZF1 deletions are associated with blastic plasmacytoid dendritic cell neoplasms (BPDCN), a malignancy of pDC precursors with poor prognosis [16][17][18]. Thus Ikaros is required for DC development, but little is known about its molecular mechanisms.
Here we show that Ikaros deficiency leads to multiple defects in pDC and cDC development. In particular, Ikaros is required for CDP development, where it antagonizes TGFβ function to promote pDC differentiation. We further show that Ikaros cooperates with Notch pathway activation to support the homeostasis of DC progenitors. Lastly, we show that a transient incubation of bone marrow cells with γ-secretase inhibitors rescues pDC development from WT and Ikaros-deficient BM progenitors, revealing a potentially novel way to enhance pDC function.

Ikaros is required for CDP differentiation
To determine how pDC differentiation is affected by Ikaros deficiency, we evaluated DC progenitor populations in Ik L/L mice. Ik L/L cells express functional Ikaros proteins at~10% of WT levels, and although Ik L/L mice die from Notch-dependent T cell leukemias at 4-6 months of age, the animals studied (6-8 weeks of age) showed no signs of transformation (normal CD4/CD8 profiles, T cell receptor chain usage, Notch pathway activation) [19][20][21].
Successive stages of DC development were analyzed, which included BM Lin -Sca1 -CD135 + cells, containing CD117 hi CD115 + MDPs and CD117 lo CD115 + CDPs, as well as the more INSERM, CNRS, University of Strasbourg and the  ANR-10-LABX-0030-INRT grant. JM received predoctoral fellowships from the Ministry of  Technology and Research and Fondation ARC; CS  a master fellowship from the Fondation pour la  Recherche Médicale (FRM-DEA20140630557) and GC a pre-doctoral fellowship from the IGBMC PhD programme (ANR-10-LABX-0030-INRT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

The Notch pathway is activated in Ikaros-deficient pDCs
We previously observed in a genome-wide study that genes associated with the Notch pathway (eg. Hes1, Ptcra, Uaca) are upregulated in the BM pDCs of Ik L/L mice [14]. Higher Hes1 and Ptcra mRNA levels were confirmed by RT-qPCR (Fig 2A). To determine if Ikaros deficiency results in Notch activation throughout pDC development, we crossed Ik L/L mice with animals carrying a Hes1-GFP knock-in (KI) reporter [23]. Total BM cells from Ik +/+ (WT) and Ik L/L Hes1-GFP KI mice contained similar frequencies of GFP + cells (mostly CD19 + B cells) (Fig 2B  and 2C). In contrast, GFP + cells were nearly absent in WT BM pDCs, but they were present in a fraction of Ik L/L pDCs (7-35%). Ik L/L GFP + pDCs were mostly SiglecH + CCR9 lo , suggesting an immature phenotype ( Fig 2D) [24,25]. CCR9 lo pDCs from WT Hes1-GFP KI mice did not express GFP. These data indicated that the Hes1 locus, and perhaps the Notch pathway, are ectopically activated during pDC development in the mutant mice.

γ-secretase inhibitors rescue Ik L/L CDP differentiation in vitro and in vivo
To determine if ectopic Notch activation interferes with pDC differentiation in Ik L/L mice, we first blocked Notch signaling in Flt3L-supplemented cultures of total BM cells, using a γ-secretase inhibitor (GSI, Compound E) [26,27]. As γ-secretase is required to cleave and activate ligand-bound Notch receptors, GSIs are potent inhibitors of Notch function. In the absence of GSI (DMSO), WT cultures generated robust numbers of CD11c + CD137 + CD11b -pDCs over an 8-day period, while Ik L/L cultures did not ( Fig 3A). Strikingly, GSI treatment significantly enhanced WT pDC differentiation, and rescued pDC development in the Ik L/L cultures to levels of WT cells. This occurred early, as GSI treatment at day 0 was both necessary and sufficient to rescue Ik L/L pDC development (Fig 3B and 3C). Similar results were obtained with other GSI compounds (DAPT and MRK003). In addition, early GSI treatment resulted in an increase in total cell numbers (Fig 3D), which correlated with an expansion of immature CD11ccells, particularly in the Ik L/L cultures ( Fig 3B). The pDCs produced in the GSI-treated cultures were more immature, and expressed low levels of CCR9 and Ly49Q ( Fig 3E); B220 levels, however, remained unchanged after GSI treatment. Importantly, the GSI-rescued WT and Ik L/L pDCs expressed mRNA for Ifna following TLR9 stimulation in vitro with CpG ODN 1885 (Fig 3F), suggesting functionality. Because GSI treatment at day 0 of culture was sufficient to induce differentiation, GSI was added only once at the onset of culture in subsequent experiments.
To identify the DC progenitor cells sensitive to GSI, we co-cultured WT and Ik L/L total BM cells, purified Lin -Sca1cells, MDPs or CDPs (all CD45.2 + ), with CD45.1 + supporting WT BM cells, in the presence of GSI and Flt3L, for 8 days (Fig 4A). The ability of the different CD45.2 + populations to give rise to pDCs was evaluated. GSI treatment consistently increased pDC differentiation from Ik L/L CDPs (Fig 4B). On the contrary, GSI did not affect WT MDPs (2 out of 3 experiments) and CDPs, even though it enhanced pDC development from total WT BM cells. We also analyzed Lin -Sca1 -CD117 lo CD135 + CD115cells ("CD115 -CDPs") in these assays, as they were reported to contain pDC-specific precursors [28], even though they existed in similar numbers in WT and Ik L/L BM (S1A and S1B Fig); GSI did not affect the pDC production from either WT or Ik L/L CD115 -CDPs (S1C and S1D Fig), and these cells were not studied further. These results therefore suggested that Ikaros negatively regulates a γ-secretasesensitive pathway mainly in (CD115 + ) CDPs.
To determine if transient GSI treatment rescues Ik L/L pDC development in vivo, we adoptively transferred GSI-treated BM cells into recipient mice. WT and Ik L/L BM cells (CD45.2 + ) were cultured with Flt3L and GSI for 2 days, and then transplanted into irradiated hosts (CD45.1 + CD45.2 + ) along with CD45.1 + supporting WT BM cells. BM and spleen cells were analyzed 9 days later for CD45.2 + pDCs (Fig 4C and 4D). In the BM, Ik L/L cells generated few CD11c + CD137 + CD11b -pDCs, regardless of GSI treatment (S2 Fig). However, in the spleen, GSI-treated Ik L/L cells generated CD11c + CD137 + CD11b -pDCs while the DMSO-treated cells did not (Fig 4C and 4D). WT cells generated slightly more pDCs after GSI treatment compared with DMSO. Importantly, the CD45.1 + supporting cells produced similar frequencies of pDCs in all conditions, indicating that GSI treatment enhanced Ik L/L and WT pDC differentiation in a cell-intrinsic manner.
Collectively, our results indicated that γ-secretase inhibitors rescue Ikaros-deficient pDC development in vitro and in vivo.

GSI promotes CDP differentiation via Notch-independent pathways
Because γ-secretase inhibitors affect other pathways in addition to Notch, we tested the role of Notch activation in pDC development by genetic means. Ik L/L mice were crossed with animals carrying a floxed null allele for Rbpj (Rbpj f/f ), the Notch transcriptional mediator, and the R26-CreERT2 transgene [29,30]. Ik L/L Rbpj +/+ Cre + (Ik L/L RBPJ WT) and Ik L/L Rbpj f/f Cre + (Ik L/L RBPJ KO) mice, along with control animals, were treated with tamoxifen for 5 days to delete Rbpj, and analyzed 5 days after the last injection. Deletion was confirmed by Western BM cells from the tamoxifen-treated mice were cultured with Flt3L for 8 days, in the presence or absence of GSI, and cell expansion and pDC development were studied ( Fig  5A and 5B). In these experiments, we reasoned that if GSI rescues pDC development by inhibiting Notch signaling, then (i) Rbpj inactivation should mimic the effects of GSI, and (ii) GSI should not have additional effects when Rbpj is deleted.
When cell numbers were evaluated, we observed that the samples treated with GSI contained significantly higher numbers of cells, regardless of RBPJ and/or Ikaros status (Fig 5B). This suggested that GSI inhibits the function of pathways other than Notch. Likewise, when pDC development was evaluated (Fig 5A and 5B), Rbpj deletion by itself did not enhance the differentiation of Ik WT DMSO-treated cells, indicating that Notch activation is not required to limit pDC development when Ikaros is present. Further, in Ik WT cells, GSI treatment enhanced pDC differentiation in both RBPJ WT and KO conditions, suggesting that GSI enhances pDC differentiation in the absence of Notch. Interestingly, when similar experiments were performed in Ik L/L conditions, GSI treatment rescued pDC development in the RBPJ WT samples, as expected, but no rescue was observed when both RBPJ and Ikaros were mutated. GSI nevertheless still increased total cell numbers in the cultures from the RBPJ-Ikaros double mutant cells, indicating that its effects on pDC differentiation and cell expansion are separable.
To determine why GSI treatment did not rescue pDC development in Ik L/L RBPJ KO BM cultures, we analyzed the BM DC progenitor populations of tamoxifen-treated Ik L/L RBPJ KO mice and littermate controls (Fig 5C and 5D). Specifically, we evaluated the CDP population in the double mutant mice, as GSI rescues Ik L/L CDP differentiation. These experiments revealed that MDPs and CDPs were barely detectable in most of the Ik L/L RBPJ KO BM samples (4 out of 5), while the BM from single mutant mice contained easily recognizable MDP and CDP cells. These results indicated that the GSI target population is absent in the Ik L/L RBPJ KO BM, and suggested that Ikaros and Notch activation cooperate to generate or maintain MDP and CDP cells in the BM.
Collectively, our results demonstrate that GSI treatment inhibits a Notch-independent pathway important for CDP development.

The TGFβ pathway is activated in Ikaros-deficient CDPs
To further investigate the molecular pathways targeted by Ikaros and γ-secretase in CDPs, we studied the transcriptome profiles of WT and Ik L/L MDPs and CDPs, cultured in the presence or absence of GSI. We used a protocol similar to the one above, and co-cultured CD45.2 + WT or Ik L/L MDPs, and CDPs, with supporting CD45.1 + WT BM cells. CD45.2 + cells were purified after 24h, and their transcriptomes were analyzed by microarray.
In the vehicle-treated samples, 963 genes were differentially expressed >1.5-fold between Ik L/L CDPs and all WT populations (Fig 6A). Approximately 30% of these genes were deregulated in both Ik L/L MDPs and CDPs (clusters III and IV), and 70% were deregulated only in the Ik L/L CDPs (clusters I and II). To determine how these genes are expressed during WT DC development, we compared their levels of expression in Ik L/L CDPs with those in WT progenitors and mature DC populations, as reported by the ImmGen Compendium (GSE15907), using unsupervised clustering [7]. Interestingly, this revealed that, among the genes up-regulated in Ik L/L CDPs (Fig 6B), the large majority (>70%) were related to mature cDC genes, and not pDCs. The remainder of the genes were DC progenitor-related. In contrast, among the genes down-regulated in Ik L/L CDPs (S4A Fig), most were related to DC progenitor (CMP, MDP, CDP) genes. Further, gene set enrichment analyses (GSEA) indicated that both the up-and down-regulated genes in the Ik L/L CDPs correlated with those normally expressed in WT cDCs (S4B Fig). Thus, Ikaros is required to repress the premature expression of cDC-associated genes in CDPs. We then asked if the cDC transcriptional hallmarks that characterize the Ik L/L CDPs were also retained in the BM pDCs from Ik L/L mice. Indeed, GSEA analysis showed that genes up-or down-regulated in Ik L/L pDCs (transcriptome data from [14]) were also up-or down-regulated in mature cDCs (S4C Fig), thereby confirming our hypothesis.
Among the genes deregulated in Ik L/L CDPs, only 70 were differentially expressed between GSI and DMSO treated samples (Fig 6C, S1 Table). To identify the potential upstream pathways involved in the regulation of their expression in DC progenitors, we performed Ingenuity Pathway Analysis (Fig 6D). This revealed Ikaros to be a significant probable regulator, which validated our approach, and showed the importance of Ikaros in CDPs. The top candidate, however, was TGFβ1, which was interesting because TGFβ1 was previously reported to skew CDP differentiation towards the cDC lineage at the expense of pDCs [9]. We therefore asked if the deregulated genes found in Ik L/L CDPs were enriched for TGFβ-associated genes, by GSEA. These results showed a strong and direct correlation between the genes down-regulated in Ik L/L CDPs and those down-regulated by TGFβ1 signaling (Fig 6E) [9]. Conversely, the genes up-regulated in Ik L/L CDPs were up-regulated by TGFβ1 activation (Fig 6F). Thus, Ikaros expression is correlated with reduced TGFβ1 signaling in CDP cells.
To determine if Ikaros directly regulates the TGFβ pathway, we investigated its capacity to bind TGFβ target genes. The low number of CDPs in WT mice did not allow us to directly investigate Ikaros binding in these cells. We therefore compared Ikaros binding to chromatin from 2 unrelated precursor cell types (pre-B cells and DN3 thymocytes) [31,32], because conserved binding might indicate that Ikaros regulates similar elements across hematopoietic cell types. These analyses showed strong and conserved Ikaros binding to several TGFβ target genes implicated in DC differentiation (eg. Axl, Irf1, Irf4, Nfkb2, Nfkbie, Rel, Relb) (Fig 6G and  S4 Fig), and suggested that Ikaros may directly regulate the expression of TGFβ target genes in CDPs.

Inhibition of TGFβ signaling rescues Ikaros-deficient pDC development
To determine if the TGFβ pathway is activated in Ik L/L CDPs, we studied the mRNA expression of genes encoding upstream components of this pathway. Although the level of transcripts encoding the type I and type II TGFβ receptors, and the downstream SMAD proteins, were similar between WT and Ik L/L MDPs and CDPs, we found that the mRNA levels of Eng, encoding the type III TGFβ receptor Endoglin, was higher in Ik L/L MDPs (2x) and CDPs (2.8x), regardless of GSI treatment (Fig 6A). Endoglin (CD105), is an auxiliary receptor for the TGFβ receptor complex, which has been shown to positively modulate TGFβ signaling [33]. Higher CD105 expression was also detected on Ik L/L MDPs, CDPs and pDCs (Fig 7A and 7B). In contrast, CD105 levels were stable in other BM populations, including CD11c + CD317 -cDCs (Fig 7B), indicating that Endoglin expression is specifically increased in Ik L/L pDCs and DC progenitors. In addition, we observed that Ikaros bound to the Eng locus in pre-B and DN3 cells, suggesting that it is an Ikaros target gene (S5A Fig). Lastly, we analyzed the functional consequence of TGFβ inhibition on pDC development. WT and Ik L/L BM cells were cultured with Flt3L for 8 days, in the presence or absence of a TGFβR1 inhibitor (SB431542), and/or GSI. SB431542 treatment alone did not affect total cell numbers (S5B Fig), but increased pDC numbers in both WT and Ik L/L cultures (Fig 7C and  7D). In contrast, GSI treatment alone increased both total cell numbers and pDC numbers. The combination of SB431542 and GSI gave similar total and pDC numbers, compared with GSI alone. These results suggested that TGFβ inhibition promotes pDC differentiation in Ik L/L CDPs.

Discussion
Here we identify Ikaros as a promoter of early DC development. We show that Ikaros cooperates with Notch signaling to enhance the emergence and/or survival of MDPs and CDPs in the BM (Fig 7E). We also show that Ikaros is required to promote CDP differentiation and cell fate specification towards the pDC and cDC lineages, in large part by correctly regulating the expression of DC-specific target genes, and secondly, by antagonizing TGFβ function. These results indicate that the general absence of mature DCs in Ikaros null mice [13,34], as well as the selective absence of pDCs in Ikaros hypomorphic animals [14], are due at least in part to CDP defects.
Our results suggest that Ikaros antagonizes a TGFβ-dependent gene expression program in CDPs. TGFβ was previously reported to skew CDP differentiation towards the cDC lineage at the expense of pDCs, in part because it induces the expression of Id2, which inhibits the master pDC regulator E2.2 [9,[35][36][37]. We show that Ikaros-deficient CDPs display a premature cDC gene expression signature, indicating that Ikaros represses the expression of mature cDC-associated genes in DC progenitors. In addition, Ikaros-deficient BM pDCs also display a cDC signature, suggesting that the mutant CDPs that commit to the pDC lineage continue to express a promiscuous cDC gene expression program. Neither Id2 nor E2.2 are affected at the mRNA level in Ikaros-deficient CDPs and pDCs, suggesting that Ikaros promotes CDP differentiation independently of E2.2.
How Ikaros antagonizes TGFβ function remains only partially understood. Certain TGFβ target genes are enriched among the genes deregulated by GSI in CDPs. Furthermore, Ikarosdeficient CDPs ectopically express high levels of endoglin which can potentiate TGFβ signaling [38]. Because no other TGFβ receptors or downstream SMAD factors are deregulated in these cells, endoglin upregulation probably plays an important role in activating the TGFβ pathway in the mutant DC progenitors. Interestingly, betaglycan, a type III TGFβ receptor closely related to endoglin in structure and function, is a substrate of γ-secretase, and GSI inhibits TGFβ2-mediated reporter gene expression via betaglycan inactivation in HepG2 cells [39]. γsecretase cleavage of type III TGFβ receptors may therefore inhibit TGFβ receptor signaling in Ikaros-deficient cells. If true, this suggests that Ikaros may be a novel upstream regulator of TGFβ signaling.
In addition to its role in CDP differentiation, Ikaros is also required for MDP and CDP homeostasis. Observed only in compound mutants deficient for Ikaros and RBPJ where both populations are absent, our results demonstrate that Ikaros cooperates with Notch activation to maintain DC progenitor survival and/or expansion. Notch signaling by itself was previously found to promote DC development in vitro via up-regulation of the Frizzled family Wnt receptors [10], but the basis for its cooperation with Ikaros remains to be elucidated. We have reported that Ikaros antagonizes Notch function in T cells, and interacts directly with the activated Notch1 protein to control a set of common target genes [40]. Whether Ikaros and Notch regulate common genes in DC progenitors remains to be investigated. Other studies have suggested that the Notch and TGFβ pathways interact to regulate common genes. Indeed, Hes1 is a common target of both pathways, because it is transcriptionally regulated by the Notch receptor intracellular domain or by Smad3 following TGFβ signaling [41]. In Ik L/L cells, however, Hes1 up-regulation was observed in pDCs but not in the more immature dendritic progenitors, suggesting that Hes1 is differentially regulated by Notch and TGFβ activation in these populations.
Finally, our results with γ-secretase inhibitors are unexpected and intriguing, and indicate that these compounds can be exploited to enhance and rescue WT and Ikaros-deficient DC development in vitro, though the effects are stronger in the mutant cells. We showed that transient GSI treatment promotes the generation of CD11ccells, probably the upstream precursors of MDPs and CDPs, and pDC differentiation from CDPs. These actions suggest that GSI molecules might be considered as a potential treatment to enhance pDC function during certain viral infections, like chronic HIV or hepatitis C virus. Conversely, it will be important to test if GSI molecules might have a differentiating effect on BPDCN cancers, a rare and fatal leukemia with few options for treatment [42].

Microarray analysis
Transcriptome analyses were performed with Affymetrix Gene ST 1.0 arrays. Unsupervised hierarchical clustering and K-means clustering were performed using Cluster 3. GSEA was performed using the GSEA 2.0 software [44,45]. Microarray data are available in the GEO databank (GSE114108).  Fig 6A, using Immgen transcriptome data for DC progenitors and mature subsets (GSE15907). (B) GSEA enrichment plots of genes up-or down-regulated in Ik L/L CDPs compared with WT (clusters II and IV, and clusters I and III in Fig 6A, respectively). (C) GSEA enrichment plots of genes specifically up-or downregulated in Ik L/L pDCs compared with WT (FC>2; p 0,05) [14]. In (B) and (C), the ranked gene list corresponds to the differential gene expression between WT cDCs and pDCs (Immgen GSE15907). NES: normalized enrichment score; FDR: false discovery rate.