The E–Id protein axis modulates the activities of the PI3K–AKT–mTORC1–Hif1a and c-myc/p19Arf pathways to suppress innate variant TFH cell development, thymocyte expansion, and lymphomagenesis

Miyazaki et al. show that Id2 and Id3 suppress the development and expansion of innate variant TFH cells by acting upstream of the Hif1a/Foxo/AKT/mTORC1 pathway as well as the c-myc/p19Arf module. Mice depleted for Id2 and Id3 expression developed colitis and αβ T-cell lymphomas, and the transcription signatures of Id2- and Id3-depleted lymphomas revealed similarities to genetic deficiencies associated with Burkitt lymphoma.

rise to CD4 + CD8 + double-positive (DP) cells (Carpenter and Bosselut 2010). Upon reaching the DP compartment, thymocytes exit the cell cycle, initiate TCRa locus rearrangement, and undergo positive and negative selection (Singer et al. 2008;Kreslavsky et al. 2010). The lymphoid populations can be segregated into adaptive or innate immune cells. The selection process permits the developmental progression of a selected group of adaptive T-lineage cells that have acquired a TCR with moderate affinity for major histocompatibility complex (MHC) class II (CD4 single positive [CD4SP]) or class I (CD8SP) associated with self-antigens (Klein et al. 2014). On the other hand, innate T-lineage cells are selected by CD1 for invariant natural killer T (iNKT) cells and by MHCrelated protein MR1 for mucosal-associated invariant T lymphocytes (MAIT) (Bendelac et al. 2007;Gold and Lewinsohn 2013) Adaptive B and T cells express an enormously diverse antigen receptor repertoire. They maintain a naïve lymphoid cell state until they encounter invading pathogens, upon which they expand and differentiate into effector cells. Innate lymphoid cells (ILCs) carry germline-encoded receptors or express a limited antigen receptor repertoire and have the potential to rapidly induce cytokine expression (Cerutti et al. 2013;Verykokakis et al. 2014). The innate lymphoid system is comprised of multiple cell types, including natural killer (NK), lymphoid tissue inducer (LTi), type 2 ILCs, and innate-like B and T cells (Diefenbach et al. 2014). The innate-like B-and T-cell compartment consists of marginal zone B cells, iNKT cells, MAIT cells, and subsets of gd T cells. ILCs act primarily by modulating the activities of adaptive immune cells.
It is now established that a large majority of developmental trajectories in the thymus involve regulation by members of the helix-loop-helix (HLH) family (Rothenberg 2014). These include E proteins as well as Id proteins. Four E proteins have been identified and characterized. They include E12, E47, HEB, and E2-2. E12 and E47 are encoded by the E2A locus and are generated by differential splicing. E protein DNA-binding activity is regulated by the Id gene products, named Id1-4. Id proteins contain an HLH dimerization domain but lack the basic DNA-binding region. They function predominantly by antagonizing the DNA-binding activities of E proteins (Benezra et al. 1990;Lazorchak et al. 2005;Miyazaki et al. 2014).
E protein levels are abundant in T-cell progenitors, where they activate TCRb V(D)J locus rearrangement and induce the expression of genes encoding for members of the Notch and pre-TCR signaling cascade (Ikawa et al. 2006;Agata et al. 2007). E47 expression declines in resting DP cells and decreases further upon maturating beyond the TCR checkpoint (Engel et al. 2001;Miyazaki et al. 2011). High E2A expression prevents developmental progression, whereas decreasing E2A and HEB levels promote positive selection (Rivera et al. 2000;Jones and Zhuang 2007). HEB acts in the DP compartment to promote the development of NKT cells (D'Cruz et al. 2010). The traversal of these checkpoints in response to pre-TCR, gd TCR, and ab TCR signals is facilitated by the induction of Id3 expression. Beyond the pre-TCR, gd TCR, and ab TCR checkpoints, Id3 expression is required to maintain the naïve state (Verykokakis et al. 2010(Verykokakis et al. , 2014Miyazaki et al. 2011;Li et al. 2013). The combined activities of Id2 and Id3 are required to promote efficient developmental progression of CD8SP cells (Jones-Mason et al. 2012).
Here we examined how Id2 and Id3 act mechanistically beyond the TCR checkpoint to orchestrate T-cell fate. We found that the activation of Id3 and Id2 expression in DP cells is sequential and that Id2 and Id3 suppressed the development and expansion of innate variant follicular helper T (T FH )-like cells acting in turn to promote the ectopic development of germinal center (GC) B cells. The innate T FH -like cells carried a highly restricted antigen receptor repertoire indicative of a self-renewing population. We identified a genetic network involving the Id-E protein, AKT-FOXO-mTOR, and Myc-p19Arf modules, which orchestrate a self-renewal-specific program of gene expression. Finally, mice depleted for Id2 and Id3 in T cells developed colitis as well as T-cell lymphoma. Collectively, these data point to a regulatory circuitry that underpins the mechanism by which Id2 and Id3 act to antagonize an innate variant T FH -specific program of gene expression, maintain thymocyte quiescence, and suppress the development of lymphoma.

Expression patterns of Id2 and Id3 in positively selected thymocytes
Previous studies have demonstrated that Id3 expression is induced at the pre-TCR checkpoint and further elevated during the positive selection process, whereas Id2 expression is low in positively selected DP cells but elevated in CD4SP or CD8SP cells Engel et al. 2001;Miyazaki et al. 2011;Jones-Mason et al. 2012). To examine in greater detail how Id2 and Id3 expression is regulated during positive selection, we used Id2-YFP and Id3-GFP reporter mice (Yang et al. 2011a). Positively selected cells, identified as CD5 + CD69 À or CD69 + TCRb À DP cells, expressed higher levels of Id3 but did not display significant levels of Id2-YFP (Fig. 1A). Id2 expression was only detectable in TCRb + DP cells (Fig. 1A). The majority of mature CD62L + CD4SP or CD8SP cells displayed abundant levels of Id2 and Id3 expression (Fig. 1A). Collectively, these data indicate that the induction of Id2 and Id3 expression during positive selection is sequential: Id3 expression is activated by TCR signaling in positively selected cells, whereas Id2 expression is induced at a later stage by a separate pathway, which remains to be revealed.
Development of T FH -like cells and GC formation in primary and peripheral lymphoid organs derived from Id2 fl/fl Id3 fl/fl IL7R Cre mice Figure 1. Development of CXCR5 + PD-1 + ab T cells and IgG1 class-switched B cells in thymi derived from Id2 fl/fl Id3 fl/fl IL7R Cre mice. (A) Flow cytometric analysis of CD69 versus CD5 expression and CD69 versus TCRb expression gated on CD4 + CD8 + DP cells (top left); CD69 versus CD62L expression gated on CD4SP (CD4 + CD8 À TCRb + ) cells (bottom left); and GFP versus YFP expression gated on CD5 À CD69 À , CD5 + CD69 À , CD69 + TCRb À , CD69 + TCRb + (DP), and CD69 + CD62L À and CD69 À CD62L + (CD4SP) cells (right). Numbers in quadrants indicate percentages of cells in each compartment. Data are representative of two independent experiments. (B, top left) Flow cytometric analysis of CD4 versus CD8 expression in total thymocytes derived from 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre and Id2 fl/fl Id3 fl/À control mice. (Top right) The graph shows the number of total thymocytes of 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre and Id2 fl/fl Id3 fl/À control mice. The bottom panels show the expression of CXCR5 and PD-1 (middle) and CXCR5 and ICOS (bottom) gated on CD4SP (CD4 + CD8 À TCRb + ) cells. The absolute number and frequency of gated cells are shown in adjacent panels. Numbers in plots refer to CXCR5 + PD-1 + and CXCR5 + ICOS + cells. Data represent the mean 6 SD from two independent experiments analyzing four 5-wk-old mice. (C) The top graphs show the percentage and absolute number of B cells (B220 + CD19 + ) in thymi derived from 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre and littermate control mice. Flow cytometric analysis of IgG1 versus IgD expression and CD38 versus Fas expression gated on the B220 + CD19 + population. The bottom right panel shows IgG1 and IgD expression gated on the CD38 À Fas hi cells. Data represent the mean 6 SD from two independent experiments analyzing four 5-wk-old mice. (D) The absolute number of CXCR5 + PD-1 + CD4T cells and frequency of GC (Fas + GL7 + ) B cells and IgG1 class-switched (IgG1 + IgD À ) B cells in spleens derived from 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre mice. Data represent the mean 6 SD from two independent experiments analyzing four 5-wk-old mice. (E) Representative immunofluorescence staining with anti-B220 antibody and PNA in spleens derived from 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre mice. Arrows indicate PNA + GCs. (F) Representative ELISPOT wells using thymocytes (top) and the frequency of IgM-or IgG1-secreting cells in bone marrow (BM), spleen (Spl), and thymus (Thy) derived from 6-mo-old Id2 fl/fl Id3 fl/fl IL7R Cre and littermate control mice (middle and bottom) are shown. (E) Data are representative of two independent experiments with three mice each. (*) P < 0.05; (**) P < 0.01 (Student's t-test). et al. 2012). To evaluate the roles of Id2 and Id3 throughout thymocyte development, we used Id2 fl/fl Id3 fl/fl IL7R Cre mice to ablate Id2 and Id3 expression in common lymphoid progenitors (CLPs) (Schlenner et al. 2010;Jones-Mason et al. 2012;Niola et al. 2012). Consistent with previous observations, the CD8SP compartment was virtually absent in Id2 fl/fl Id3 fl/fl IL7R Cre mice ( Fig. 1B; Jones-Mason et al. 2012). Additionally, Id2 fl/fl Id3 fl/fl IL7R Cre mice displayed an increase in the total number of DN cells that expressed TCRb (Supplemental Fig. 1A). Since a previous study showed aberrant development of T FH -like cells in thymi derived from Id3-null mutant mice, we examined Id2 fl/fl Id3 fl/fl IL7R Cre mice for CXCR5, PD-1, and ICOS expression (Miyazaki et al. 2011). We found that a large fraction of CD4SP and DN TCRb + as well as iNKT cells expressed CXCR5, PD-1, and ICOS in thymi derived from 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre mice ( Fig. 1B; Supplemental Fig. 1A,B). CXCR5 + PD-1 + T FH -like cells were detectable in Id3 À/À , Id3 fl/À CD4 Cre , and Id3 fl/fl IL7R Cre mice but not in Id2 fl/fl IL7R Cre mice (Supplemental Fig. 1C,D). The T FH -like populations were accompanied by increased numbers of B220 + CD19 + thymic B cells, Fas hi CD38 À GC B cells, and IgG1 class-switched cells (Fig. 1C). We found that the segregation of cortical and medullary regions was completely abolished, and B220 + cells were observed throughout the thymi derived from 5-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre mice (Supplemental Fig. 1E). In addition, we noted spontaneous GCs in the spleens derived from Id2 fl/fl Id3 fl/fl IL7R Cre mice, accompanied by fewer naïve CD4 T cells and an increased number of T FH cells (Fig. 1D,E; Supplemental Fig. 2A,B). The elevated number of T FH cells was associated with the development of GC B cells, IgG1 class-switched B cells, plasma blasts, and plasma cells ( Fig. 1D; Supplemental Fig. 2C). Furthermore, substantial numbers of IgM-and IgG1-secreting cells were detected in thymi and spleens, but not in the bone marrow, derived from 4-to 6-mo-old Id2 fl/fl Id3 fl/fl IL7R Cre mice (Fig. 1F). Taken together, these data indicate that Id2 and Id3 suppress the development and/or selection of T FH -like cells and GC B cells in primary and peripheral lymphoid organs.

Development of innate T FH -like cells in Id2 fl/fl Id3 fl/fl IL7R Cre mice
To examine in greater detail the phenotypes associated with the development of T FH -like cells, CD4SP cells were analyzed for the expression of markers associated with maturation and migration. In line with previous studies, we found that TCRb hi DP and CD4SP thymocytes displayed aberrant CCR7, CXCR4, CD62L, and CD69 expression in Id2 fl/fl Id3 fl/fl IL7R Cre mice ( Fig. 2A; Jones-Mason et al. 2012). The level of CD44 expression displayed by Id2 fl/fl Id3 fl/fl IL7R Cre CD4SP thymocytes was distinct from that observed for splenic CD4 + T cells, suggesting that the CD4SP cells in the thymus were not derived from recirculating CD4 + T cells (Fig. 2B). Previous studies have demonstrated that a fraction of CD4SP cells derived from Id3 À/À thymi aberrantly expressed PLZF or Eomes (Verykokakis et al. 2010(Verykokakis et al. , 2013Miyazaki et al. 2011;Li et al. 2013;D'Cruz et al. 2014). The fraction of thymocytes expressing PLZF was increased in the thymi and spleens derived from Id2 fl/fl Id3 fl/fl IL7R Cre mice, while Eomes expressors were mostly lacking (Fig. 2C). Since PLZF is the transcription factor that is expressed in innate T cells such as iNKT cells, we examined whether the innate T FH -like population was distinct from that of iNKT cells. We found that only a small fraction of CXCR5 + CD4SP cells represented iNKT cells and that the majority of CXCR5 + CD4SP cells expressed PLZF (Fig. 2D,E). Furthermore, we found significantly increased numbers of IL-4-but not IFN-g-producing CD4SP cells (Fig. 2F). Taken together, these data indicate that depletion of Id2 and Id3 expression at an early developmental stage results in the development of an innate T FH -like population in the thymus.
Rapidly expanding innate T FH -like cells in Id2 fl/fl Id3 fl/fl IL7R Cre mice To obtain further insight into the mechanism that underpins the roles of Id2 and Id3 in thymocyte development, Id2 fl/fl Id3 fl/fl IL7R Cre thymi were examined for abnormalities upon aging. We found that at 2 wk after birth, thymi derived from Id2 fl/fl Id3 fl/fl IL7R Cre mice displayed a complete block in positive selection, with a pronounced defect during the CD69 + TCRb À -to-CD69 + TCRb hi transition (Fig. 3A). For comparison, we analyzed DP cells derived from TCRa À/À mice for CD5 or CD69 expression. We found that DP cells derived from TCRa À/À thymi did not display either CD69 or CD5 expression (Supplemental Fig. 3A). These data suggest that DP thymocytes that have received a TCR signal required Id2 and Id3 expression to fully differentiate into the TCRb hi stage. However, unlike the complete block in positive selection observed in 2-wkold mice, we found that in adult (4-to 8-wk-old) mice, CD4SP cells in Id2 fl/fl Id3 fl/fl IL7R Cre mice readily were detectable and continued to expand (  Fig. 3C). Annexin V expression was not affected in these populations (Supplemental Fig. 3D). We also observed a substantial increase in cell size across the CD4SP but not in the DP compartment in Id2 fl/fl Id3 fl/fl IL7R Cre thymocytes (Fig.  3E). Over time, the majority of thymocytes expressed CXCR5 and PD-1 associated with large cell size (Supplemental Fig. 3E). Sections of thymi derived from 6-mo-old Id2 fl/fl Id3 fl/fl IL7R Cre mice showed effacement of the cortico-medullary junction due to proliferation of neoplastic cells with large, hyperchromatic, irregular nuclei and scant cytoplasm (Fig. 3F).
To determine whether the expanding population was derived from a small fraction of selected thymocytes, the shows CCR7 and CXCR4 expression in CD69 À TCRb À and TCRb + DP cells presented as mean fluorescence intensity (MFI). Flow cytometric analysis of CCR7 versus CD62L, CD127 versus CXCR4, and CD69 versus CD62L gated on CD4SP cells derived from 6-wkold Id2 fl/fl Id3 fl/fl IL7R Cre and littermate control mice. Numbers adjacent to the outlined areas or in quadrants indicate the percentage of cells in the population that was examined. (Bottom right) The graph shows the CD62L, CCR7, CXCR4, and CD127 expression in CD4SP cells, presented as MFI. Data are representative of one experiment with 6-wk-old mice (mean 6 SD; n = 3). (B) Flow cytometric analysis of CD44 expression in CD4SP thymocytes (blue) and splenic CD4 T cells (red) derived from 6-to 8-wk-old  Cold Spring Harbor Laboratory Press on July 19, 2018 -Published by genesdev.cshlp.org Downloaded from TCR repertoires were examined using RNA sequencing (RNA-seq). To this end, RNA was isolated from CD4SP cells from Id3 +/À , Id3 À/À , and Id2 fl/fl Id3 fl/fl IL7R Cre thymi as well as wild-type T FH cells. We found that Vb and Va transcripts associated with wild-type T FH cells and CD4SP thymocytes isolated from Id3 +/À and Id3 À/À thymi displayed a similar polyclonal Vb or Va repertoire (Fig. 3G,H). On the other hand, CD4SP cells derived from Id2 fl/fl Id3 fl/fl IL7R Cre mice were enriched for TRBV5, TRBV13-2, and TRBV13-3 as well as distally located TRAV5-1 and TRAV6-1 transcripts (Fig. 3G,H). We also observed a substantial enrichment for transcripts encoding 39 Ja elements (Supplemental Fig. 4A). Va and Vb gene usage observed in Id2 fl/fl Id3 fl/fl IL7R Cre CD4SP cells, however, was not invariant, as described for iNKT or MAIT cells, but rather differed between the various Id2 fl/fl Id3 fl/fl IL7R Cre mice that were examined (Supplemental Fig. 4B-D; Le Bourhis et al. 2011). To exclude the possibility that the oligclonal TCRa and TCRb repertoires were caused by aberrant TCR recombination at the DP stage, we examined the CD69 À TCRb À DP cells from Id2 fl/fl Id3 fl/fl IL7R Cre mice for TCRb and TCRa rearrangements. No obvious differences were detected for TCRb and TCRa loci in Id2 fl/fl Id3 fl/fl IL7R Cre DP cells as compared with control cells (Supplemental Fig. 5A,B). Collectively, these observations indicate that Id2 and Id3 suppress the oligoclonal expansion of innate variant T FH -like populations.

Id2 and Id3 suppress the expansion of innate variant T FH -like cells beyond the TCR checkpoint
To deplete Id2 and Id3 expression in T-lineage cells, Id2 fl/fl Id3 fl/fl CD4-Cre mice were generated. Similar to as described above for Id2 fl/fl Id3 fl/fl IL7R Cre mice, at 2 wk after birth, positive selection was severely impaired in Id2 fl/fl Id3 fl/fl CD4 Cre thymi (Fig. 4A). However, 7-wk-old Id2 fl/fl Id3 fl/fl CD4 Cre mice displayed CD4SP, DN TCRb hi , and iNKT populations that expressed high levels of CXCR5, PD-1, ICOS, and PLZF expression ( Fig. 4B; Supplemental Fig. 6A,B). Again, these populations were associated with elevated numbers of thymic B cells as well as IgG1 class-switched GC B cells (Supplemental Fig. 6C).
Since recent studies have demonstrated a critical role for Id2 and Id3 in regulatory T (T reg ) cells, it remained possible that the innate T FH -like population developed because of systemic inflammatory conditions (Miyazaki et al. 2014). To exclude this possibility and investigate the role of Id2 and Id3 in thymocyte development beyond the TCR checkpoint, Id2 fl/fl Id3 fl/fl dLck Cre mice were generated (Zhang et al. 2005). We found that Cremediated deletion in dLck Cre mice was initiated beyond the CD69 + TCRb + DP stage ( Fig. 4C; Supplemental  Fig. 7A). We found that a substantial fraction of peripheral T reg cells did not exhibit Cre activity, indicative of T reg cells expressing wild-type levels of Id2 and Id3 in dLck Cre mice (Supplemental Fig. 7A). As observed for Id2 fl/fl Id3 fl/fl Il7R Cre and Id2 fl/fl Id3 fl/fl CD4 Cre mice, in thymi derived from Id2 fl/fl Id3 fl/fl dLck Cre mice, we found increased percentages of CD4SP cells and iNKT cells that expressed CXCR5, PD-1, and PLZF and a variant but limited TCRVa and TCRVb repertoire ( Fig. 4D; Supplemental Fig. 7B-D).
Again, we noted an increase in the percentages of thymic B cells and a significant relationship between the proportions of thymic B cells and CXCR5 + PD-1 + CD4SP cells (R 2 = 0.51893) (Fig. 4E). Notably, we detected transcripts initiated from the first exon of the Id2 and Id3 genes in sorted CXCR5 À PD-1 À CD4SP cells but not in sorted CXCR5 + PD-1 + CD4SP cells from identical Id2 fl/fl Id3 fl/fl dLck Cre mice, consistent with the notion that the aberrant activation of CXCR5 and PD-1 expression was caused by depletion of Id2 and Id3 expression (Supplemental Figure 7E). Taken together, these data indicate that Id2 and Id3 expression is required to suppress the development and expansion of an innate variant T FH -like population beyond the TCR checkpoint.

Gene expression signature of innate variant T FH -like cells
To determine how Id2 and Id3 suppressed the oligoclonal expansion of an innate variant T FH -like population, CD4SP cells derived from control and Id2 fl/fl Id3 fl/fl IL7R Cre thymi were isolated and analyzed by RNA-seq. Twothousand-three-hundred-seventy-nine genes were differentially expressed (greater than twofold, P < 0.05; 1291 up-regulated; 1088 down-regulated) in Id-depleted CD4SP cells (Fig. 5A). Prominent among those genes were Cxcr5, Ccr10, Wnt10a, Gzma, Myb, Cebpb, Tgfbr3, and Smad7 (Fig. 5A). Gene ontology (GO) analysis revealed that a large fraction of differentially expressed transcripts encoded for proteins associated with metabolism, cytokine production, RNA metabolism, T-cell activation, and cell cycle progression (Fig. 5B). Furthermore, Kyoto Encyclopedia of Genes and Genomics (KEGG) pathway analysis revealed p53 and genes associated with the PI3K-AKT, MAPK, and Rap1 pathways as well as genes involved in the suppression of inflammatory bowel disease as being affected by depletion of Id2 and Id3 expression (Fig. 5B). Next, we compared the transcription signatures of Id2/Id3depleted CD4SP thymocytes with Id3 À/À CD4SP and Graphs show the frequency of CD4SP cells in total thymocytes and CD5 + CD69 À , CD69 + TCRb, À and CD69 + TCRb + cells gated on the DP compartment. Data are representative of one experiment with 2-wk-old mice (mean 6 SD; n = 5 biological replicates). (B) Flow cytometric analysis of CD4 versus CD8 expression in total thymocytes and CXCR5 versus PD-1 expression gated on CD4SP cells (CD4 + CD8 À TCRb + CD1d-tet À ) or DN;TCRb + cells (CD4 À CD8 À TCRb + CD1d-tet À ) derived from 7-wk-old Id2 fl/fl Id3 fl/fl CD4 Cre or littermate control mice. Data were obtained from two independent experiments. (C) Representative flow cytometric analysis of YFP expression gated on each compartment derived from Rosa YFP/YFP (red) or Rosa YFP/YFP dLck Cre (blue) mice. Numbers above the lines indicate percentages of YFP-expressing cells. Data were derived from two independent experiments. (D) Flow cytometric analysis of CD4 versus CD8 expression (left) and TCRb versus CD1d-tet expression (middle right) in total thymocytes and CXCR5 versus PD-1 expression gated on CD4SP (CD4 + CD8 À TCRb + CD1d-tet À ) cells (middle left) or iNKT (CD1d-tet + TCRb + ) cells (right) derived from 10-wk-old Id2 fl/fl Id3 fl/fl dLck Cre or littermate control mice. Data were derived from four independent experiments. (E) Graphs show the frequencies of thymic B (B220 + CD19 + ) cells and the correlation between thymic B cells and CXCR5 + PD-1 + CD4SP cells. Data were derived from seven independent experiments. Ten-week-old to 16-wk-old mice were analyzed. Cold Spring Harbor Laboratory Press on July 19, 2018 -Published by genesdev.cshlp.org Downloaded from Figure 5. Transcription signature of innate T FH -like cells. (A) Volcano plot of RNA-seq analysis of mRNA expression in sorted CD4SP thymocytes derived from 8-wk-old littermate control or Id2 fl/fl Id3 fl/ll IL7R cre mice. Red dots represent genes up-regulated in Id2 fl/fl Id3 fl/fl IL7R Cre CD4SP cells (more than twofold, P < 0.05, calculated by raw count value). Blue dots represent genes downregulated in Id2 fl/fl Id3 fl/fl IL7R Cre CD4SP cells (more than twofold, P < 0.05). Gray dots represent genes whose expression was not significantly altered. (B) Clusters of genes whose expression was modulated in Id2-and Id3-depleted CD4SP thymocytes were identified using GO terms (top) and KEGG pathways (bottom). P-values are shown. (C) RNA-seq data of mRNA expression isolated from CXCR5 + CD44 hi CD4T cells from SRBC-immunized wild-type spleen (T FH ) and CD4SP thymocytes derived from 8-wk-old Id2 fl/fl Id3 fl/fl IL7R Cre (Id2 À/À Id3 À/À CD4SP), Id3 À/À (Id3 À/À CD4SP), and control thymi. Expression was normalized to control CD4SP cells. The Venn diagram shows quantification of genes significantly changed in T FH cells, Id3 À/À CD4SP cells, and Id2 fl/fl Id3 fl/ll IL7R cre CD4SP cells compared with control CD4SP cells. (D) Heat maps are presented to visualize differences in gene expression patterns for T FH cells, CD4SP cells sorted from Id3 À/À thymocytes, and CD4SP cells sorted from IL7R-Cre;Id2F/F;Id3F/F mice as compared with CD4SP control thymocytes. Bars above the heat maps indicate similarities between the different samples presented in each of the lanes. Differentially expressed genes are shown only for those that displayed greater than twofold differences (P < 0.05) as compared with control CD4SP thymocytes. The heat maps represent different ontology groups, including genes encoding for factors that regulate transcription (left), signal transduction (middle), and cell cycle (bottom right). (Top right) The heat map displays the KEGG pathway for chemokines, cytokines, and associated receptors.
Cold Spring Harbor Laboratory Press on July 19, 2018 -Published by genesdev.cshlp.org Downloaded from peripheral T FH cells. We found that 924 genes showed overlap between T FH and Id2-and Id3-deficient CD4SP cells as compared with wild-type CD4SP thymocytes. Two-hundred-ninety-nine genes showed overlap between CD4SP cells isolated from Id3 À/À and Id2 fl/fl Id3 fl/fl IL7R Cre thymi, whereas the expression of 1362 genes was modulated only in innate T FH -like cells when compared with control CD4SP cells ( Fig. 5C; Supplemental Fig. 9A). Transcription signatures derived from Id3 À/À CD4SP cells showed an intermediate pattern when comparing T FH and CD4SP cells derived from Id2 fl/fl Id3 fl/fl IL7R Cre thymi (Fig.  5C). The expression patterns of FceR1 and IL-17rb in control and mutant thymocytes were validated using flow cytometry (Supplemental Fig. 9B). Notably, genes associated with T FH cell function were altered across the three populations, including Maf, Batf, Foxo1, Foxp1, Sh2d1a, Cxcr5, Il4, and Il21 but not Bcl6 and Ascl2 (Fig. 5D; Ma et al. 2012;Wang et al. 2014;Xiao et al. 2014). As predicted, the expression levels of genes associated with an innatespecific program of gene expression were also elevated, including Zbtb16 (Plzf), Myb, and Egr2 (Kovalovsky et al. 2008;Hu et al. 2010;Seiler et al. 2012). Finally, consistent with an expanding population of Id2-and Id3-depleted CD4SP thymocytes, the expression patterns of genes associated with cell cycle regulation were affected (Fig.  5D). Taken together, these observations indicate that gene expression patterns of Id-depleted CD4SP thymocytes overlap, albeit partially, with that of T FH and innate T-lineage cells.

A genetic circuitry links Id2 and Id3 and the AKT-FOXO-mTORC1 axis
In previous studies, we had identified a spectrum of target genes regulated by the Id-E protein module in T-lineage cells, including Cxcr5, Il10, Hif1a, Ikzf3, Myb, Il10ra, E2f2, and Bmp7 (Miyazaki et al. 2014). Here we found additional target genes that are directly regulated by the E-Id protein axis, including Foxo1, Foxo3, Foxp1, Slamf6, Il7r, Il6ra, Cxcr4, Tgfbr3, Wnt10a, Cdc25b, Gadd45b, Bcl2l1, Bcl2l11, and Rps6ka2 (Supplemental Fig. 9C; data not shown). To examine for molecular pathways or modules associated with depletion of Id2 and Id3 expression, transcription signatures derived from Id-depleted CD4SP thymocytes and E protein occupancy were linked into a common framework using Cytoscape software (Fig.  6A). Using this approach, we identified an ensemble of E protein targets-including Foxo1, Foxo3 and Foxp1associated with the maintenance of a naïve and quiescent state. Additionally, a cluster of target genes was identified closely linked with an innate-like transcription signature, including Myb, Plzf, Egr2, and Sox13 ( Fig. 6A; Melichar et al. 2007;Kerdiles et al. 2009;Feng et al. 2011;Hedrick et al. 2012). Finally, we found that clusters of genes associated with the PI3K-AKT/FoxO/mTOR, NFkB/ TNF, MAPK, Rap1, cytokine/chemokine, and p53/cell cycle pathways were targeted by the E and Id protein module (Fig. 6B).
To validate these findings, we examined activation of the mTOR and AKT pathways in developing thymocytes using flow cytometry. We found significant increases in phospho-S6 and phospho-4E-BP1 levels in Id2-and Id3-depleted PD-1 + CD4SP cells derived from Id2 fl/fl Id3 fl/fl dLck Cre thymi as compared with control thymi (Fig. 6C). As predicted, phospho-S6 and phospho-4E-BP1 levels were not altered in PD-1 À CD4SP or DP cells derived from Id2 fl/fl Id3 fl/fl dLck Cre versus control mice (Fig. 6C). To determine whether the E-Id protein axis also regulates the activity of the mTORC pathway to control the expansion of innate gd T cells, Id3 À/À Vg1.1 + cells were examined for the expression of phospho-S6 and phospho-4E-BP1. Indeed, Id3 À/À Vg1.1 + cells displayed significantly increased levels of phopho-S6 and phospho-4E-BP1 (Supplemental Fig. 10). In contrast, Id3 À/À Vg2 + cells were not associated with elevated levels of phopho-S6 and phospho-4E-BP1 (Supplemental Fig. 10). Next, we examined the phosphorylation status of AKT (pAKT) in sorted control and Id-depleted CD4SP thymocytes. We found that resting control CD4SP cells lacked detectable levels of pAKT (Fig. 6D, left panel). However, activating TCR signaling readily elevated levels of pAKT (Fig. 6D, middle and right panels). Notably, Id2-and Id3depleted CD4SP cells exhibited pAKT expression in the absence of TCR stimulation, and TCR-mediated signaling resulted in even higher levels of pAKT (Fig. 6D, bottom). Taken together, these data indicate that the E-Id protein axis modulates the PI3K-AKT-FOXO-mTORC1 pathway at multiple steps.

Id2 and Id3 suppress the development of ab T-cell lymphomas
Previous studies have demonstrated that the development of Burkitt lymphoma is closely associated with mutations across the Id3 HLH region, that forced expression of Id2 in a murine model of BCR-ABL interferes with the development of chronic myeloid leukemia, and that Id3-deficient mice develop gd T-cell lymphomas (Ko et al. 2008;Li et al. 2010;Schmitz et al. 2012). To determine whether the combined loss of Id2 and Id3 expression in T-lineage cells leads to the development of lymphoma, aged Id2 fl/fl Id3 fl/fl IL7R Cre mice were monitored for signs of distress. We found that a substantial fraction of 6-to 8mo-old Id2 fl/fl Id3 fl/fl IL7R Cre mice exhibited ruffled fur, hunched posture, and rectal prolapse (data not shown). The majority of Id2 fl/fl Id3 fl/fl IL7R Cre or Id2 fl/fl Id3 fl/fl dLck Cre mice died within 1 yr (Fig. 7A). Histological analysis showed moderate to severe colitis in 6-to 8-moold Id2 fl/fl Id3 fl/fl IL7R Cre , Id2 fl/fl Id3 fl/fl CD4 Cre , and Id2 fl/fl Id3 fl/fl dLck Cre mice ( Fig. 7B; Supplemental Fig. 11A). In addition to colitis, we noted splenomegaly and large subcutaneous and/or mesenteric lymph nodes in 21 out of 41 8-to 14-mo-old Id2 fl/fl Id3 fl/fl IL7R Cre , Id2 fl/fl Id3 fl/fl CD4 Cre , and Id2 fl/fl Id3 fl/fl dLck Cre mice (Fig. 7C). Four out of 21 Id2 fl/fl Id3 fl/fl IL7R Cre , Id2 fl/fl Id3 fl/fl CD4 Cre , and Id2 fl/fl Id3 fl/fl dLck Cre mice also developed thymic lymphoma ( Fig. 7C; Supplemental Fig. 11B). Histopathological analysis of the peripheral lymphoid organs displayed a disorganized architecture, such as blurring of the demarcation between red and white pulp in the spleen, with Figure 6. Id2, Id3, and the FOXO/mTOR axis. (A) Regulatory network that connects the activity of E proteins and an ensemble of transcriptional regulators in Id2-and Id3-depleted CD4SP thymocytes into a common framework. E2A ChIP-seq (chromatin immunoprecipitation [ChIP] combined with deep sequencing) data were derived from E2A-deficeint T-cell lymphoma cells transduced with E47 (Lin et al. 2010). The widths of the connectors reflect relative peak scores of E2A occupancy. Colors reflect higher (red) or lower (blue) gene expression levels in Id2 fl/fl Id3 fl/fl IL7R Cre CD4SP cells as compared with control CD4SP thymocytes. (B) Networks that link distinct pathways, as defined by KEGG analysis, to E2A occupancy. (C) Flow cytometric analysis of phosphorylation of S6 (pS6) and 4E-BP1 (p4E-BP1) in DP and CD4SP cells for the indicated populations derived from control and Id2 fl/fl Id3 fl/fl dLck Cre mice. The graph shows the level of phosphorylated S6 and 4E-BP1, presented as MFI. Data are representative of three independent experiments (mean 6 SD; n = 3 biological replicates). (D) Flow cytometric analysis of phosphorylation of AKT. Sorted CD4SP cells (CD4 + CD8 À TCRb + CD1d-tet À ) derived from the Id2 fl/fl Id3 fl/fl IL7R Cre thymus were stimulated with plate-coated anti-CD3e (2 mg/mL) and anti-CD28 (5 mg/mL) antibodies and analyzed at the indicated time points for pAKT expression. (*) P < 0.05; (**) P < 0.01 (Student's t-test). The Venn diagram shows overlaps among T FH cells, Id3 À/À CD4SP cells, Id2 À/À Id3 À/À CD4SP cells, and Id2 À/À Id3 À/À lymphoma cells as compared with control CD4SP cells, as shown in Figure 5C. (G) Heat map for selected significantly differentially expressed genes that are differentially expressed in Id2-and Id3-depleted CD4SP thymocytes as well as lymphomas. Clusters of genes associated with cell cycle and tumor suppression are indicated. (H) Visualized RNA-seq data across the Cdkn2a locus, presented in reads per million reads aligned (RPM). Arrows indicate transcription start site of p19Arf and p16Ink4a gene and direction of transcription.

420
GENES & DEVELOPMENT Cold Spring Harbor Laboratory Press on July 19, 2018 -Published by genesdev.cshlp.org Downloaded from infiltrating monomorphic lymphoid cells containing large polymorphic nuclei and scant cytoplasm ( Fig. 7D; Supplemental Fig. 11C). The infiltrate in the lungs derived from Id2 fl/fl Id3 fl/fl CD4 Cre mice was composed of neoplastic nucleated cells (Fig. 7E). We note that lymphocyte infiltration was not caused by inflammatory disease, since there was no evidence of elevated mucin levels in the bronchial epithelium (lack of periodic acid Schiff [PAS]positive goblet cells) (Fig. 7E). The lymphomas that developed in lymph nodes and spleens were mainly composed of TCRb + CD3 + T-lineage cells associated with a CD8 low CD4 À , CD8 À CD4 À or CD8 À CD4 + phenotype (Supplemental Fig. 11E). As expected, a large fraction of these cells showed CXCR5 and PD-1 expression (Supplemental Fig. 11E). Finally, we found that lymphoma cells were highly malignant, as evidenced by metastasis across the liver (17 out of 21), kidney (three out of 10), and lung (10 out of 21) tissues (Fig. 7C,E; Supplemental Fig. 11B,D). Collectively, these observations indicate that depletion of Id2 and Id3 in T-lineage cells leads to increased levels of morbidity caused by the development of lymphoma and/or inflammatory disease.

The c-Myc and p19Arf axis and the development of Id2-and Id3-deficient lymphomas
To gain insight into the mechanism that underpins the development of lymphoma in Id2-and Id3-deficient mice, transcription signatures were analyzed from a set of lymphomas using RNA-seq. We found that Id2-and Id3-deficient lymphomas revealed transcription profiles that overlapped with those derived from Id2-and Id3deficient CD4SP cells (Fig. 7F). Specifically, ;2100 genes were differentially expressed in Id2-and Id3-deficient lymphomas versus Id2-and Id3-depleted CD4SP cells (Supplemental Fig. 12A,B). The majority of isolated lymphomas expressed a limited TCR repertoire, arguing against polyclonal expansion (Supplemental Fig. 12C,D). We found that the differences in transcription profiles between lymphomas and CD4SP cells were closely associated with genes involved in metabolism, proliferation, and the immune response (Supplemental Fig. 12E). As predicted, we found substantial decreased levels of Foxo1 and Foxo3 as well as an elevated abundance of c-Myc and Hif1a (Fig. 7G). Conspicuous among the spectrum of aberrantly expressed genes was the Cdkn2a locus. Notably, Cdkn2a transcript abundance was high in Id2 fl/fl Id3 fl/fl IL7R Cre CD4SP cells but low in lymphoma cells (Fig. 7G). The Cdkn2a locus encodes for two tumor suppressors: p16Ink4a and p19Arf. p19Arf is activated by c-Myc and has been demonstrated to suppress lymphomagenesis (Kamijo et al. 1997;Zindy et al. 2003). We found that p19Arf but not p16Ink4a transcript abundance was elevated in the self-renewing innate variant T FH -like population (Fig. 7H). In contrast, p19Arf transcript levels were virtually undetectable in Id-deficient lymphomas (Fig. 7H). Taken together, these data point to a regulatory circuitry that maintains thymocyte quiescence and provide a mechanism involving c-myc and p19Arf that underpins the development of T-cell lymphoma in Id2-and Id3-deficient T cells.

Discussion
Previous studies have demonstrated that positive selection is enforced by the E and Id protein module (Bain et al. 1999;Jones and Zhuang 2007;Jones-Mason et al. 2012).
Here we examined how the Id proteins orchestrate thymocyte selection. Based on this analysis, we propose two key steps that involve Id2 and Id3 during the positive selection process. The first step involves the induction of Id3 expression by TCR-mediated signaling. The second step involves the induction of Id2 expression in cells that have already received a TCR signal. We demonstrate here that Id2-and Id3-depleted thymocytes were not able to pass this second step of selection, with the exception of a slowly expanding population of cells that was characterized by an innate-like transcription signature. Innate T cells, including iNKT, T-CD4T, and MAIT cells, are selected by a mechanism that is distinct from that of conventional CD4SP thymocytes (Li et al. 2005;Lee et al. 2010;Alonzo and Sant'Angelo 2011;Gold and Lewinsohn 2013). Specifically, iNKT and MAIT cells are MHC class I-like-restricted, involving innate TCR signaling. Additionally, iNKT and T-CD4 T cells are selected by DP thymocytes rather than by thymic epithelial cells. Hence, we propose that differences in TCR-mediated signalingconventional or innate-mediated TCR signaling-affect the duration and/or periodicity of Id3 and/or Id2 transcription and that CD4 T cells selected by MHC class II in thymic epithelial cells but not innate CD4 T cells require Id2 and Id3 expression. Differences in the strength, periodicity, and timing of Id2 and Id3 expression would then instruct progenitors to commit to either the adaptive or innate T-cell lineage.
We found that Id2-and Id3-depleted positively selected thymocytes express a T FH -like program of gene expression. Particularly intriguing was the decline of Foxo1 and Foxp1 abundance in Id-depleted CD4SP thymocytes. Foxo1 and Foxp1 are well-characterized suppressors of the T FH cell fate (Hedrick et al. 2012;Wang et al. 2014;Xiao et al. 2014). These data bring into question how Id proteins and Foxo1 and Foxp1 are linked. We found E2Abound sites across regulatory regions associated with the Foxo1 and Foxp1 loci. Hence, we suggest that in CD4SP thymocytes, the E proteins act as transcriptional repressors that interfere with Foxo1 and Foxp1 transcription. High levels of Id2 and Id3 antagonize the DNA-binding activity of E proteins, relieving Foxo1 and Foxp1 from E2A/HEB-mediated repression. Elevated levels of Foxo1 and Foxp1 in turn would prevent the premature activation of a T FH lineage-specific program of gene expression (Supplemental Fig. 13A).
A notable feature of our findings is the development of an innate variant T FH cell population in mice depleted for Id2 and Id3 expression. Are these cells genuine T FH cells? We noted overlap but also significant differences in transcription signatures upon comparing T FH and Iddepleted innate T FH cells. Notably, Bcl6 and Ascl2 expression, closely associated with a T FH -specific program of gene expression (Crotty 2014), were not modulated, at least transcriptionally, in Id-depleted CD4SP thymocytes. Despite these differences, we found that Iddepleted innate variant T FH cells secreted high levels of IL-4, required interaction with MHC class I-like molecules, and their presence was closely associated with the spontaneous development of GCs. Hence, we suggest that these cells represent a distinct subset of T FH cells (Supplemental Fig. 13B).
A key aspect of the findings reported here involves a slowly expanding population of innate T FH -like cells. How do Id2 and Id3 regulate thymocyte quiescence? We found that multiple pathways known to maintain lymphocyte quiescence were affected by depletion of Id2 and Id3 expression. Prominent among these was the FOXO-mTOR module. The roles for the FoxO proteins in maintaining thymocyte quiescence and acting as tumor suppressors are well documented (Paik et al. 2007;Kerdiles et al. 2009;Hedrick et al. 2012). Similarly, spontaneous activation of the mTORC1 pathway in Tsc1-deficient T cells leads to an increase in cell size and loss of quiescence, phenotypes that are equivalent to that described here for Id2-and Id3-depleted CD4SP thymocytes (Yang et al. 2011b). As mentioned previously, the entire ensemble of Foxo loci, including Foxo1, Foxo3, and Foxo4, appears to be controlled by the E-Id protein module. Likewise, the AKT-FOXO-mTOR pathway was activated in Id2-and Id3-depleted CD4SP thymocytes as well as in Id3-deficient innate gd T cells. Multiple levels of regulation by the E-Id axis appear to be involved here: (1) modulation of Foxo transcript levels, (2) activation of the AKT pathway plausibly mediated by changes in the expression of an ensemble of PTPN phosphatases that are modulated upon depletion of Id2 and Id3, and (3) activation of the mTORC pathway in part by induction of Rps6ka2 expression. We suggest that this form of regulation is not restricted to innate variant T FH and a subset of Vg1.1 gd T cells. Rather, we propose that the regulation of the PI3K-FOXO-mTOR pathway by the E-Id axis is a general mechanism that underpins the homeostasis of lymphoid progenitors and self-renewing committed B and T lymphoid cells.
Finally, we found that c-Myc and p19Arf levels were elevated in Id-depleted CD4SP thymocytes. p19Arf is a well-known tumor suppressor that is regulated by c-Myc (Lowe and Sherr 2003). Previous studies have demonstrated that c-myc expression is directly regulated by E proteins (Schwartz et al. 2006). Hence, we suggest that loss of Id expression leads to elevated c-Myc abundance, which in turn leads to the induction of p19Arf expression. We speculate that cellular expansion upon depletion of Id2 and Id3 expression is attenuated by the induction of p19Arf expression. This then would lead to a population that slowly self-renews. How do lymphomas develop in Id2-and Id3-depleted thymocytes? We speculate that, through mechanisms yet to be determined, p19Arf expression is inactivated in a single progenitor, releasing the brakes and ultimately leading to the development of a monoclonal ab T-cell lymphoma (Supplemental Fig. 13C; von Boehmer 2004).
The mechanisms that underpin the development of ab T-cell lymphomas in Id2-and Id3-depleted mice overlap with those observed in human Burkitt lymphoma and a murine model of Burkitt lymphoma (Sander et al. 2012;Schmitz et a. 2012). It is now established that the development of Burkitt lymphoma is closely associated with high levels of c-Myc expression and mutations localized across the HLH region of Id3 (Love et al. 2012;Schmitz et al. 2012). Likewise, we found that Id2-and Id3-depleted murine T-cell lymphomas expressed relatively high levels of c-Myc expression. There are also similarities between the two sets of lymphomas as they relate to the PI3K pathway. Burkitt lymphomas are associated with increased PI3K signaling and display an activated AKT pathway (Sander et al. 2012;Schmitz et al. 2012). We found that Id-deficient CD4 T cells displayed decreased abundance of Foxo1/3 expression as well as activated AKT and mTORC1 pathway. Upon examining transcription signatures in human T-cell lymphomas, we found that changes in Id2, Foxo1, and Foxo3 abundance were significantly associated with the development of human T-cell lymphoma, (Supplemental Fig. 14;Piccaluga et al. 2007). Finally, since a very high fraction of aged mice display symptoms of inflammatory disease, it is conceivable that chronic inflammation contributes to the development of T-cell lymphoma in Id2-and Id3-deficient mice similarly to as described for viral infections associated with the development of human lymphomas.

Materials and methods
Mice C57BL/6, Id3 f/f , Id2 f/f , Id3 À/À , Id2 YFP/+ , Id3 GFP/+ , IL7Ra Cre , CD4 Cre , dLck Cre , and ROSA YFP/YFP mice were bred and housed in specific pathogen-free conditions in accordance with the Institutional Animal Care and Use Guidelines of the University of California at San Diego.

RNA-seq analysis
Total RNA and the library preparations were described previously (Miyazaki et al. 2014). The strand-specific RNA-seq libraries were sequenced with a HiSeq 2500 sequencer (Illumina). Alignment and trimming of reads were performed using the OSA algorithm against the mm10 murine genome reference in Arraystudio (Omicsoft) (Hu et al. 2012). RNA transcripts were quantified using RSEM methods (http://deweylab.biostat.wisc. edu/rsem) as implemented in Arraystudio (Omicsoft). Abundance values (counts) were normalized and compared with calculated P-values using DESeq (http://www-huber.embl.de/users/anders/ DESeq). Genes whose abundance values were <10 in all samples were removed. MultiExperiment Viewer software was used to generate heat maps and for hierarchical clustering. GO analyses and visualization files were generated using HOMER (http:// biowhat.ucsd.edu/homer), and read pile-ups were visualized using the University of California at Santa Cruz Genome Browser.

Histology
Tissues were fixed in 4% paraformaldehyde (Electron Microscopy Sciences). Fixed tissues were embedded in paraffin and sliced, followed by haematoxylin and eosin (H&E) staining.

Statistical analyses
P-values were calculated with the two-tailed Student's test for two-group comparison, as applicable, with Microsoft Excel software.

Data access
RNA-seq data have been deposited at Gene Expression Omnibus under accession number GSE64779.