Pre-T-cell receptor (TCR) signal transduction is required for developing thymocytes to differentiate from CD4−CD8− double-negative (DN) cell to CD4+CD8+ double-positive (DP) cell. Notch signalling is required for T-cell fate specification and must be maintained throughout β-selection, but inappropriate Notch activation in DN4 and DP cells is oncogenic. Here, we show that pre-TCR signalling leads to increased expression of the transcriptional repressor Bcl6 and that Bcl6 is required for differentiation to DP. Conditional deletion of Bcl6 from thymocytes reduced pre-TCR-induced differentiation to DP cells, disrupted expansion and enrichment of intracellular TCRβ+ cells within the DN population and increased DN4 cell death. Deletion also increased Notch1 activation and Notch-mediated transcription in the DP population. Thus, Bcl6 is required in thymocyte development for efficient differentiation from DN3 to DP and to attenuate Notch1 activation in DP cells. Given the importance of inappropriate NOTCH1 signalling in T-cell acute lymphoblastic leukaemia (T-ALL), and the involvement of BCL6 in other types of leukaemia, this study is important to our understanding of T-ALL.
αβT cells develop in the thymus, which provides an essential environment for T-cell fate specification and differentiation (Koch and Radtke, 2011; Hosokawa and Rothenberg, 2018). CD4−CD8− double-negative (DN) cells differentiate via an immature CD8+ single-positive (ISP) to become CD4+CD8+ double-positive (DP) cells, which give rise to CD4 or CD8 T cells. The DN population can be subdivided by ordered expression of CD44 and CD25 as follows: CD44+CD25− (DN1), CD44+CD25+ (DN2), CD44−CD25+ (DN3) and CD44−CD25− (DN4) (Hayday and Pennington, 2007; Shah and Zúñiga-Pflücker, 2014).
Notch1 activation promotes T-cell fate in bone marrow (BM)-derived precursors entering the thymus, and mice deficient in Notch1 exhibit failed T-cell lineage commitment, whereas forced expression of Notch1 leads to generation of T cells in BM (Radtke et al., 2010; Hosokawa and Rothenberg, 2018). Binding of Notch1 on lymphoid precursors with its ligand Dll4 leads to its cleavage and proteolytic release of activated intracellular Notch1 (Notch intracellular domain; NICD), which travels to the nucleus to bind Rbpj and coactivate transcription of target genes, including Hes1 and Notch3 (Reizis and Leder, 2002; Bellavia et al., 2007b). Notch signalling is maintained during β-selection (Ciofani and Zúñiga-Pflücker, 2005; Hosokawa and Rothenberg, 2018), but following pre-TCR signalling, Notch genes are downregulated as Id3 rises and supresses E2A-mediated transcription (Allman et al., 2001; Yashiro-Ohtani et al., 2009). Inappropriate Notch activation after β-selection is oncogenic and contributes to T-ALL (Tzoneva and Ferrando, 2012; Pelullo et al., 2014).
The pre-TCR complex is essential for αβT cell development to stop further recombination of the TCRβ loci, induce expansion and differentiation from DN3 to DP, and rescue developing T cells from apoptosis. Several other signalling pathways are involved in this process, including IL7, Hedgehog and Wnt signalling (Staal et al., 2001; Goux et al., 2005; Outram et al., 2009; Rowbotham et al., 2009; Shah and Zúñiga-Pflücker, 2014; Boudil et al., 2015; Sahni et al., 2015).
Multiple studies showed that Bcl6 is a master regulator of the T follicular helper cell lineage and germinal centre B-cell fate (Ye et al., 1997; Yu et al., 2009). Bcl6 is also expressed in developing lymphocytes and protects pre-B cells from apoptosis during immunoglobulin light chain rearrangement; however, in thymocytes, IL7 signalling during β-selection represses Bcl6 expression (Heng et al., 2008; Duy et al., 2010; Boudil et al., 2015). We used conditional deletion of Bcl6 from thymocytes to show that Bcl6 promotes pre-TCR-induced differentiation to DP cell and attenuates Notch1 activation. Thus, this study is important to our understanding of T-ALL, given the significance of dysregulated NOTCH1 in this disease.
Bcl6 promotes DN to DP transition in foetal thymocytes
We previously employed a modelling approach that used transcript expression and degradation measurements to identify transcriptional targets of pre-TCR signalling, and identified Bcl6 as a candidate gene that showed increased transcription in the first 24 h after the pre-TCR signal (Sahni et al., 2015). Bcl6 expression in whole-genome transcription datasets confirmed that Bcl6 is upregulated between DN3 and DP populations (Fig. S1A,B; Heng et al., 2008).
Therefore, to investigate the role of Bcl6 in thymocyte differentiation after pre-TCR signalling, we examined thymocyte development in conditional knockout mice in which floxed Bcl6 (Bcl6f/f) alleles were deleted from thymocytes, from the DN2 stage onwards, by Cre under control of the lck-promotor (Bcl6f/flck-Cre+, referred to as Bcl6coKO). In these mice, LoxP sites were inserted into the Bcl6 gene locus, flanking exons 7–9 encoding the zinc finger domain of Bcl6 (Hollister et al., 2013).
In embryos, αβT cell development first occurs in a synchronised wave facilitating investigation of the rate of differentiation. We compared T cell development in the mouse thymus from Bcl6coKO and control (Cre−) littermates between embryonic day (E)15.5 and E17.5, the stage at which TCRβ expression first occurs and the pre-TCR first signals for differentiation to DP cell. On E15.5, the Bcl6coKO thymus contained more thymocytes than the control, and although there were no differences in the distribution of the DN subsets, the proportion of DN3 and DN4 cells that expressed intracellular (ic) TCRβ was higher in Bcl6coKO than in the control (Fig. 1A-D). In contrast, on E16.5, the day on which the DP population first emerges, the Bcl6coKO thymus contained fewer cells than the control (Fig. 1E), and there was a reduction in the DP population from 12.8% in control to 4.85% in Bcl6coKO. The CD8+ ISP population was reduced from 16.6% in the control to 12.0% in the Bcl6coKO thymus, whereas the proportion of DN cells increased (Fig. 1F). Examination of the DN subsets revealed a small increase in the proportion of DN3 cells and a decrease in DN4 cells (Fig. 1G), and the percentage of cells that expressed icTCRβ was lower in DN3 and DN4 subsets in Bcl6coKO compared with control (Fig. 1H). We did not detect a significant difference in the proportion of CD3+ icTCRβ− cells, which represent the γδ T-cell population, between Bcl6coKO and control (Fig. S2A,B). To assess cell cycle status, we compared intracellular cyclin B1 staining between Bcl6coKO and control. The proportion of cells that stained positive for intracellular cyclin B1 was high in all populations, reflecting the rapid increase in cell numbers in the foetal thymus during late gestation, but we did not observe significant differences between genotypes in DN3, DN4, ISP and DP populations (Fig. S2C,D).
Expression of CD2 was lower in Bcl6coKO compared with control E16.5 thymus (Fig. 1I). Comparison of the ratios of DN:ISP and ISP:DP showed an increase in both ratios in Bcl6coKO thymus, indicating that conditional deletion of Bcl6 affected the rate of transition from DN to ISP and rate of transition from ISP to DP (Fig. 1J). The reduced transition to DP was present in E17.5 Bcl6coKO thymus compared with control, which contained fewer thymocytes and a reduced proportion of DP cells and increase in DN cells (Fig. 2A-C).
Bcl6 promotes differentiation to DP in a Rag-independent manner
Although Bcl6 expression was upregulated after pre-TCR signal transduction, Bcl6 RNA was present in DN3 cells before pre-TCR signal transduction (Fig. S1A,B) (Heng et al., 2008; Sahni et al., 2015), so it is possible that differentiation to DP cells is less efficient in the Bcl6coKO thymus because of an earlier influence of Bcl6 on TCRβ gene rearrangement or expression. To test this, we crossed Bcl6coKO mice onto a Rag1−/− background (Rag1−/−Bcl6f/flck-Cre+, referred to as RagKOBcl6coKO; Rag1−/−Bcl6f/flck-Cre−, referred to as RagKOBcl6control) and induced differentiation by anti-CD3 treatment in foetal thymus organ cultures (FTOC) (Levelt et al., 1993). The transcriptional changes caused by anti-CD3 treatment of Rag-deficient FTOC mirror those that occur in defined thymocyte populations as they differentiate from DN3 to DP (Sahni et al., 2015). This experimental system thus enabled us to investigate the rate of pre-TCR-induced differentiation from DN3 to DP independently of the requirement for TCRβ rearrangement in a normal three-dimensional thymus environment in which all other developmental cues, including IL7 and Dll4, were present at physiological levels. After anti-CD3 treatment, RagKOBcl6coKO thymocytes differentiated less efficiently than RagKOBcl6control thymocytes, indicating that the role of Bcl6 in differentiation to DP is Rag-independent. On days 2 and 3 after anti-CD3 treatment, there was a reduction in the DP population in RagKOBcl6coKO compared with RagKOBcl6control FTOC and the ratio of ISP:DP was increased (Fig. 2D-F).
Transcriptional targets of Bcl6 activity at the DN to DP transition
During β-selection, IL7 signalling has been shown to reduce Bcl6 expression in TCRβ+ DN3 and DN4 thymocytes (Boudil et al., 2015). However, Bcl6 rises after pre-TCR signalling for differentiation to DP (Fig. S1A,B) (Heng et al., 2008; Sahni et al., 2015). Although β-selection is absolutely dependent on the pre-TCR, it also involves coordinated signalling from IL7 and Notch pathways. Therefore, to investigate the function of Bcl6 in differentiation to a DP cell, we assessed the transcriptional response to anti-CD3 treatment in thymocytes from RagKOBcl6coKO and RagKOBcl6control, so that developing thymocytes were induced to differentiate in a synchronised manner by pre-TCR signalling, but all other external signals (such as IL7) remained constant and were at normal levels. We carried out RNA sequencing (RNA-seq) on fluorescence-activated cell sorting (FACS)-sorted thymocyte populations from RagKOBcl6control and RagKOBcl6coKO in the absence of pre-TCR signalling and after pre-TCR signal transduction for induction of differentiation to DP by anti-CD3 treatment. As expected, Bcl6 expression (expressed as reads per kilobase of transcript per million mapped reads; RPKM) in control cultures increased after pre-TCR signal transduction and was ∼sevenfold higher in sorted ISP cells compared with the CD25+ DN population and ∼sixfold higher again in the DP population (Fig. 2G). To assess the efficiency of deletion in Bcl6coKO we also assessed Bcl6 exons 7-9, which encode the zinc finger domain of Bcl6 and are floxed in this model (Hollister et al., 2013). Mean normalised reads of exons 7-9 were low in control and RagKOBcl6coKO in the absence of pre-TCR signalling, but expression increased in control cultures after pre-TCR signalling in ISP and DP populations. By contrast, in the RagKOBcl6coKO, levels remained low in ISP and were ∼2.5-fold lower in the RagKOBcl6coKO DP dataset than in the control (Fig. 2H).
We then analysed the RNA-seq datasets from FACS-sorted CD25+ DN cells in the absence of anti-CD3 treatment to investigate the impact of Bcl6 deficiency on DN cells before pre-TCR signal transduction (Fig. 2I,J). We used principal component analysis (PCA) to explore the variability in datasets in an unbiased way. PCA separates datasets by genotype on PC1, which accounted for 62.2% of variability. Because Bcl6 is a transcriptional repressor, we identified genes that were more highly expressed in the Bcl6coKO thymus than in controls. These genes contributed strongly to PC1 and so would potentially normally be repressed by Bcl6 (Fig. 2I). They included known targets of transcriptional repression by Bcl6 in other tissues, such as Stat1, Ifitm1, Ifitm3, Bcl2, Itm2b and Itgae (Shaffer et al., 2000; Ci et al., 2009; Liu et al., 2016). Interestingly, the PCA also highlighted genes associated with the Notch signalling pathway and Notch target genes, such as the canonical Notch target gene Hes1 and genes encoding components of the pathway, including Adam10, Dlk1, Dtx3, Dtx3l, Nedd4l, Maml2, Agfg1 and Mfng (LaFoya et al., 2016; Kovall et al., 2017), which were more highly expressed when Bcl6 was deleted. Differentially expressed genes (DEG) (P<0.05) were identified using Ebayes statistics, and to distinguish DEG that are important for the function of Bcl6, we intersected the 3000 genes that contributed most to PC1 with the 1500 most significant DEG. We identified 1419 genes, shown clustered as a heat map in Fig. 2J. These DEG included Notch-associated genes (Adam10, Dtx3 and Dtx3l), known Bcl6-targets (Bcl2 and Ifitm1) and components of the Wnt and Hedgehog signalling pathways (Dixdc1 and Gli1).
Next, to investigate the impact of Bcl6 deletion after pre-TCR signal transduction, we analysed the RNA-seq datasets from FACS-sorted CD8ISP and DP populations from anti-CD3-treated RagKOBcl6control and RagKOBcl6coKO FTOC after 2 days in culture. The RagKOBcl6coKO thymocytes differentiated more slowly than their RagKOBcl6control counterparts on anti-CD3 treatment (Fig. 2D-F), so we used canonical correspondence analysis (CCA) to compare their overall pattern of transcription to the transcriptional changes that occurred during pre-TCR-induced thymocyte differentiation from DN3 to DP (Fig. 3A,B). We generated a scale from the transcriptome of undifferentiated DN3 cells to the transcriptome of thymocytes at 21 h after initiation of pre-TCR signal induction (Sahni et al., 2015) and plotted our datasets against this scale. The CCA showed that the four DP datasets had a more differentiated pattern of transcription than the ISP populations, but for each population the control datasets had a more mature transcriptional signature than their Bcl6coKO counterparts (Fig. 3A). We then created a scale from the transcriptome of ISP and DP cells from Immgen (Heng et al., 2008), and used CCA to plot our datasets against this scale. For both populations, control datasets showed a more mature transcriptional signature than their Bcl6coKO counterparts (Fig. 3B). These CCA confirmed that, in the absence of Bcl6, pre-TCR-induced differentiation is severely impaired, as not only were the sizes of the DP and ISP populations reduced but the transcriptome of those cells that had differentiated was less mature than that of their control counterparts.
In keeping with the action of IL7 signalling to reduce Bcl6 expression during β-selection (Boudil et al., 2015) and the rise in Bcl6 expression following pre-TCR signal transduction, expression of Il7r declined between CD25+ DN, ISP and DP, and was not significantly different between any FACS-sorted Bcl6coKO and control populations (Fig. 3C). By contrast, expression of Bcl2, a known Bcl6 target gene in other cell types, was significantly higher in all Bcl6coKO populations compared with their control counterparts, and declined as cells differentiated from CD25+ to ISP to DP (Fig. 3D).
PCA on ISP and DP datasets separated them by developmental stage on PC1, accounting for 43.3% variability, and by genotype on PC3, accounting for 9.4% of variability (Fig. 3E). Then, to highlight genes in the ISP population important for the differences between control and Bcl6coKO, we intersected the 4000 genes that contributed most to PC3 with the 1800 significant DEG (P<0.05) between control and Bcl6coKO ISP datasets to identify 1712 genes, shown clustered as a heat map (Fig. 3F). The intersection highlighted genes that are Bcl6 targets in other tissues (Ifitm3, Bcl2 and Ifitm2), and Notch-associated genes (e.g. Adam12, Myc and Il2ra), including the signature Notch target Heyl.
To investigate the influence of Bcl6 on the DP population, we intersected the 4000 genes that contributed most to PC3 with 3100 significant DEG (P<0.05) between control and Bcl6coKO DP datasets, to identify 2992 genes, shown clustered on a heat map (Fig. 3G). The intersection included maturation genes, expressed at lower levels in Bcl6coKO thymocytes than in controls (Cd4, Cd8a, Tcf7 and Lat), and genes involved in Notch signalling, which were expressed at higher levels in Bcl6coKO, suggesting increased Notch-mediated transcription.
Conditional deletion of Bcl6 increases Notch-mediated transcription
We next compared expression of Notch-associated DEG from DP datasets and found increased expression of many genes upregulated by Notch signalling in Bcl6coKO thymocytes compared with controls, and decreased expression of genes that are downregulated by Notch activation in thymocytes (Arenzana et al., 2015; Chen et al., 2019) (Fig. 4A). Consequently, to test on a wider set of genes whether conditional deletion of Bcl6 leads to an overall increase in Notch-mediated transcription, we carried out CCA to compare our DP datasets with the transcriptome of control thymocytes and those with enhanced Notch1-mediated transcription (Arenzana et al., 2015). The Bcl6coKO datasets scored higher on the scale of control to active Notch signalling than control datasets, confirming that conditional deletion of Bcl6 leads to greater Notch-mediated transcription in DP cells (Fig. 4B).
To investigate whether Bcl6 directly represses Notch1, Notch3 or Rbpj, we examined their expression levels in the RNA-seq datasets. In fact, expression of Notch1 was modestly but significantly lower in the Bcl6coKO DN3 population than in controls (P<0.05) and, as expected, Notch1 and Notch3 were downregulated following pre-TCR signal transduction whereas Rbpj expression remained constant (Fig. 4C). None of these genes were more highly expressed in the Bcl6coKO ISP or DP populations compared with their control counterparts, indicating that the increase in Notch-mediated transcription observed in the Bcl6coKO DP cells was not because Bcl6 acts directly to repress Notch1, Notch3 or Rbpj.
Bcl6 inhibits Notch activation
As several DEG highlighted by PCA encode molecules that regulate Notch signal transduction at the protein level by facilitating cleavage and processing of Notch1 to generate the NICD, we hypothesised that Bcl6 attenuates Notch signalling by reducing formation of NICD. To test this, we compared the quantity of NICD in control and Bcl6coKO thymocytes by western blot (Fig. 4D). Bcl6coKO thymocytes contained ∼2.7-fold more NICD than controls, but levels of actin were equivalent. We then treated E16.5 Bcl6coKO and control FTOC for 2 days with Notch inhibitor (γ-secretase inhibitor) DAPT or its vehicle (DMSO) to confirm that the increased presence of NICD in Bcl6coKO thymocytes was attributable to increased processing of full-length Notch1. As expected, DMSO-treated Bcl6coKO FTOC contained more NICD than DMSO-treated control FTOC, and DAPT treatment reduced the quantity of NICD in FTOC from both genotypes by ∼twofold (Fig. 4E).
To test whether Bcl6 promotes pre-TCR-induced differentiation to DP cells in part by attenuation of Notch activation, we induced differentiation of RagKOBcl6coKO FTOC with anti-CD3 treatment in the presence of DAPT or vehicle (DMSO). DAPT treatment reduced NICD by ∼twofold (Fig. 4F) and also increased the rate of differentiation compared with control cultures, with an increase in the percentage of DP cells and the DP:ISP ratio, and a reduction in the CD8 ISP population (Fig. 4G,H).
Bcl6 is required for survival of DN4 cells
Conditional deletion of Bcl6 on E16.5 led to a significant increase in icTCRβ− DN4 cells, which have been shown to die by apoptosis (Falk et al., 2001; Hager-Theodorides et al., 2007). We measured apoptosis by annexinV staining in thymocyte populations during embryonic development on E15.5 and E16.5. We detected no difference in the proportion of annexinV+ cells between Bcl6coKO and control DN3 populations or in E15.5 DN4 populations (Fig. 5A,B). However, apoptosis was increased on E16.5 in the Bcl6coKO DN4 population compared with control, indicating that Bcl6 promotes thymocyte survival after the pre-TCR-dependent transition (Fig. 5B). We also found increased annexinV staining in the foetal DP population (Fig. 5C). Consistent with this, our RNA-seq datasets showed significantly higher expression of pro-apoptotic genes in Bcl6coKO DP cells compared with controls, whereas expression of several anti-apoptotic genes was significantly lower (Fig. 5D).
Bcl6 in adult T cell development
In adult mice, the Bcl6coKO thymus contained fewer cells than control littermate thymus, and although the proportion of DP and SP populations were not different, the proportion of DN cells was modestly decreased (Fig. 6A-C). To assess cell cycle status, we measured intracellular cyclin B1 expression, and found no significant differences in the proportion of positive cells between Bcl6coKO and control in DN3, DN4, ISP or DP populations (Fig. S3A,B). In contrast, intracellular expression of Bcl2 was significantly higher in the Bcl6coKO DN3 and DP populations compared with controls (Fig. 6D,E). Within the DN population, the proportion of DN3 cells was increased and DN4 cells decreased (Fig. 6F), suggesting that the reduction in the overall proportion of DN cells was a result of loss of DN4 cells. Indeed, the number of DN3 cells was not different between Bcl6coKO thymus and control thymus, whereas the number of DN4 cells was significantly lower (Fig. 6F). This suggested that, as in the embryo, the adult thymus showed decreased pre-TCR-dependent differentiation, but that in adults it led to only a modest reduction in thymocyte number. In adult thymus, T-cell development has reached steady state, with accumulation of the DP population and feedback mechanisms controlling the rate of differentiation to maintain production and size of the DP pool; thus, it is difficult to detect changes in the rate of differentiation, which are evident in synchronised foetal thymocyte differentiation (Outram et al., 2009; Rowbotham et al., 2009). We therefore synchronised T-cell development in adults in vivo by hydrocortisone treatment to deplete thymocytes, allowing measurement of recovery of the DP population (Rowbotham et al., 2009). Three days after hydrocortisone treatment, the Bcl6coKO thymus contained fewer DP cells and a lower ratio of DP:DN than controls, confirming that Bcl6 promotes the transition from DN to DP cells in the recovering adult thymus before it has reached steady state (Fig. 6G).
In adult Bcl6coKO DN3 and DN4 populations, icTCRβ expression was reduced compared with controls (Fig. 7A,B); however, we did not observe a significant difference in the proportion of cell surface CD3+icTCRβ− DN cells (which represent the γδ T-cell population) between Bcl6coKO and control thymi (Fig. S3C,D). The reduction in the DN4 population and increase in proportion of icTCRβ− DN4 cells could be caused by failure of expansion of icTCRβ+ cells, by increased cell death of DN4 cells or by a combination of both. After pre-TCR signalling, DN3b (icTCRβ+) cells rapidly upregulate cell-surface transferrin receptor (CD71) to undergo a burst of proliferation. The percentage of CD71+ cells was decreased in Bcl6coKO DN3 cells compared with controls (Fig. 7C,D). Gating on DN3b enriched for CD71 expression, and in a representative experiment 55% of control DN3b cells were CD71+ compared with 20.5% in Bcl6coKO cells. As expected, cell surface CD71 expression was lower within the DN3a (icTCRβ−) population, with fewer CD71+ cells in Bcl6coKO than control. We detected no significant differences in the proportion of CD71+ cells in the DN4 populations (Fig. 7E), whereas the proportion of CD71+ cells was reduced in Bcl6coKO CD8+ ISPs compared with controls (Fig. 7F). These data indicated that Bcl6 deficiency impacts on thymocyte development immediately after initiation of pre-TCR signal transduction and leads to a reduced proliferative burst of icTCRβ+ DN3 cells, resulting in less efficient enrichment of icTCRβ+ cells within the DN4 population.
DN4 cells that fail to express icTCRβ upregulate CD69 and die by apoptosis (Falk et al., 2001). The Bcl6coKO DN4 population contained a higher proportion of annexinV+ and CD69+ cells than controls, confirming increased DN4 apoptosis in the absence of Bcl6 (Fig. 7G,H). Quantitative RT-PCR analysis of FACS-sorted DN4 cells showed that expression of the survival gene Bcl2l1 was lower in adult Bcl6coKO DN4 cells than in controls in two independent experiments (Fig. 7I). Taken together, these data indicate that Bcl6 is important in adult thymus for efficient enrichment and expansion of icTCRβ+ DN cells and survival of DN4 cells.
Here, we have identified two novel important functions for Bcl6 during T cell development. We show that Bcl6 is required for pre-TCR-induced differentiation from DN3 cell to DP cell and for attenuation of Notch1 activation.
Conditional knockout of Bcl6 from foetal thymocytes resulted in reduced differentiation from DN to ISP and from ISP to DP cell, and increased cell death in the DN4 population. The requirement for Bcl6 at this transition was independent of Rag, but Bcl6 deficiency resulted in dysregulated β-selection with an increase in TCRβ− DN4 cells and increased cell death within the DN4 population in foetus and adult. RNA-seq of ISP and DP populations showed that not only was pre-TCR-induced differentiation less efficient in the Bcl6coKO thymocytes, but also that the ISP and DP cells were less mature than their wild-type counterparts, confirming the requirement for Bcl6 in differentiation to DP cells.
Increased IL7 signalling to above physiological levels has also been shown to inhibit differentiation to DP cell. Furthermore, as IL7 signalling leads to downregulation of Bcl6, our study suggests that low Bcl6 expression may contribute to the arrest induced by increased IL7 signalling (Yu et al., 2004; Hong et al., 2012; Boudil et al., 2015).
Our study also indicates that Bcl6 functions to limit Notch activation in developing thymocytes, but Notch1 and Notch3 (itself a Notch1 target) were not differentially expressed between Bcl6coKO and control in the ISP and DP datasets. We therefore hypothesised that, during thymocyte development, Bcl6 represses a set of genes that are involved in Notch processing and activation in order to attenuate the generation of NICD. Western blotting confirmed the increased presence of NICD in Bcl6coKO thymocytes and that NICD concentrations were sensitive to γ-secretase inhibition. Although transcription of Notch1 and Notch3 are downregulated following pre-TCR signalling, Notch1 and Notch3 protein are still present in differentiating thymocytes after β-selection, but Notch signalling ceases during differentiation to DP cell (Allman et al., 2001; Hosokawa and Rothenberg, 2018). Inappropriate Notch activation in DN4 and DP populations causes oncogenesis and dysregulated T-cell development, so regulation of Notch activation is essential after β-selection. Several transcriptional mechanisms to limit Notch signalling have been described, including rapid downregulation of Notch1 and Notch3 transcription, and activation of ikaros (Ikzf1) transcription by Notch3. Ikaros then competes to silence Notch target genes such that, in the absence of ikaros activity, increased Notch1 activation can arrest thymocyte differentiation at the DN3 stage (Bellavia et al., 2002; Bellavia et al., 2007a; Yashiro-Ohtani et al., 2009; Geimer Le Lay et al., 2014; Arenzana et al., 2015). Our study identifies an additional mechanism to attenuate Notch activation at the protein level, as we showed that Bcl6 represses expression of molecules required for Notch activation, thereby reducing production of NICD. Overall, this model suggests that Bcl6 acts to dampen Notch signalling and safeguard against inappropriate Notch activation, as its deficiency led to upregulation of only a subset of Notch target genes. Activated Notch1 has been shown to regulate the PI3K-AKT pathway and promote cell survival (Gutierrez and Look, 2007; Wong et al., 2012) but, despite the increase in NICD, thymocyte cell death was increased in the absence of Bcl6 and we did not observe malignant transformation of thymocytes in our colony of Bcl6coKO mice.
During Xenopus development, Bcl6 also restricts Notch signalling, inhibiting transcription of a subset of Notch target genes to achieve cell-type-appropriate gene expression for left-right asymmetry (Sakano et al., 2010).
In different experimental systems, increased levels of Notch1 activation and transcription have been described to promote differentiation beyond the DN3 stage in the absence of TCRβ chain expression and/or to arrest thymocyte development at the DN3 stage (Michie et al., 2007; Dudley et al., 2009; Arenzana et al., 2015), whereas we show that Notch inhibition can recover differentiation to DP in Bcl6coKO FTOC. In the future, it will be important to investigate the extent to which the increase in NICD contributes to dysregulated development at the DN3 to DN4 stages and the partial arrest in differentiation observed when Bcl6 is conditionally deleted.
Our RNA-seq indicated that Bcl6 also influences the expression of thousands of genes, including many known regulators of T-cell development. For example, conditional Bcl6 deficiency increased expression of the transcription factors Klf2, Runx1 and Gli1 and the Wnt pathway component Dixdc1, all important in T-cell development (Staal et al., 2001; Woolf et al., 2003; Carlson et al., 2006; Drakopoulou et al., 2010). Several DEG, which are subjects of transcriptional regulation by Bcl6 in other cell types, have also been described as transcriptional targets of Notch and are involved in malignant transformation in leukaemias including T-ALL (Myc, Il2ra, Ccnd2, Ccne1, Bcl2, Igf1r) (Reizis and Leder, 2002; Weng et al., 2006; Shin et al., 2008; Rao et al., 2009; Medyouf et al., 2011; Ferreira et al., 2012; Witkowski et al., 2015), suggesting that Bcl6-Notch crosstalk is significant in T-ALL.
In summary, our study demonstrates the importance of Bcl6 in T-cell development in the thymus at the transition from DN cell to DP cell. We show that Bcl6 is required for attenuation of NICD and for pre-TCR-induced differentiation to DP cell, and that Bcl6 promotes enrichment of icTCRβ+ DN cells and cell survival after β-selection.
MATERIALS AND METHODS
Western blots were as described (Barbarulo et al., 2011), using anti-cleaved-Notch1 (clone Val1744, Cell Signaling Technology) detected by X-Ray Film Processor (Protec); and anti-actin (clone AC-15, Sigma-Aldrich) visualised by the UVItec Gel-Documentation system and UVIband image software. Quantification of bands was carried out using ImageJ software; NICD levels were normalised to actin levels.
Cell suspensions were prepared and stained as described (Hager-Theodorides et al., 2009) using directly conjugated antibodies (see Table S1) from eBioscience (San Diego, USA) and BioLegend, (San Diego, USA), acquired on C6Accuri (BD Biosciences) or Cytoflex (BeckmanCoulter) and analysed using Flowjo10.6 (TreeStar, USA). Intracellular staining was as described (Papaioannou et al., 2019). Apoptosis was measured using annexinV-FITC (eBioscience) and annexin binding buffer (BioLegend, San Diego, USA) as described (Shah et al., 2004). To analyse immature thymocyte subsets and exclude mature SP and γδ T-cell populations, we gated out cell-surface CD3+ cells from ISP and DN populations in experiments shown in Fig 1F, Fig 2B and Fig 6D,F.
Quantitative reverse transcriptase PCR
DN4 (CD4−CD8−CD25−CD44−CD3−) cells were FACS-sorted on FACSAriaIII, the RNA extracted and quantitative RT-PCR carried out as described (Yanez et al., 2019), using primers from Quantitec (Qiagen, Netherlands). RNA levels were relative to Hprt.
RNA-seq by UCL Genomics on Illumina NextSeq500 was as described (Solanki et al., 2017). Data are available (GSE152944). Datasets were processed and standardised using Bioconductor package DESeq2 to generate normalised estimates of transcript abundance. DEG were determined using moderated Ebayes t-statistic (P<0.05) from the limma package in Bioconductor. Principal component analysis used CRAN package ade4. Canonical correspondence analysis (Ono et al., 2014) used the CCA function of CRAN package vegan as previously described (Solanki et al., 2018). To represent environmental variables of interest, the 2000 most significant DEG (lowest P-values, calculated by moderated eBayes adjusted for false positives) between respective starting and ending precursor populations were used. In Fig. 3A, the scale was generated from the 2000 most significant DEG between time 0 h and time 21 h for time-course transcriptome datasets (Array-express E-MTAB-308) following pre-TCR signal transduction (Sahni et al., 2015). In Fig. 3B, the scale was generated using the 2000 most significant DEG between ISP and DP Immgen transcriptome datasets (GSE15907) (Heng et al., 2008). In Fig. 4B, the scale was generated using the 2000 most significant DEG between datasets from thymocytes with normal to high levels of Notch signalling (GSE67572) (Arenzana et al., 2015). Heat maps in Fig. 4A and Fig.5D were generated using the CRAN package Pheatmap and RColorBrewer: rows were centred, unit variance scaling was applied to rows and rows were clustered using the Pearson correlation distance and average linkage.
To determine expression levels of different exons of Bcl6, we used the Python scripts dexseq_prepare_annotation.py from the Bioconductor DEXSeq package to prepare the genome annotation and then dexseq_count.py to generate counts of exons using the .bam alignment files as input. The exon count files were then inputted into DEXSeq, which generated the normalised number of reads (expression) of each exon (Anders et al., 2012; Reyes et al., 2013).
Unpaired two-tailed Student's t-tests were used; *P≤0.05, **P≤0.01 and ***P≤0.001. To allow comparison between litters, relative values for each genotype or treatment were calculated by dividing by the mean value for controls from same litter. Data represent at least three experiments.
Research at UCL Great Ormond Street Institute of Child Health is supported by the NIHR Biomedical Research Centre at Great Ormond Street Hospital and UCL.
Conceptualization: A.S., D.C.Y., H.S., T.C.; Formal analysis: A.S., D.C.Y., J.R., H.S., T.C.; Investigation: A.S., D.C.Y., C.-I..L., A.B., S.R., H.S., T.C.; Data curation: A.S., J.R., T.C.; Writing - original draft: A.S., T.C.; Writing - review & editing: A.S., D.C.Y., C.-I.L., J.R., S.R., H.S.; Supervision: T.C.; Project administration: T.C.; Funding acquisition: T.C.
This work was funded by the Medical Research Council (MR/P000843/1 and MR/5037764/1) and Great Ormond Street Hospital Charity (GOSHCC). A.B. was supported by a fellowship from Institut Pasteur Cenci Bolognetti Foundation; H.S. by a Child Health Research Appeal Trust studentship; J.R. by a studentship from the Biotechnology and Biological Sciences Research Council London Interdisciplinary Biosciences Consortium (LiDO); and A.S. by a studentship from Great Ormond Street Hospital Charity and the Medical Research Council.
RNA-seq data are available at Gene Expression Omnibus under accession number GSE152944.
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.192203.reviewer-comments.pdf
The authors declare no competing or financial interests.