CRLF2 overexpression results in reduced B cell differentiation and upregulated E2F signaling in the Dp16 mouse model of Down syndrome

Abstract

Children with Down syndrome (DS) are 10-fold more likely to develop B-cell acute lymphoblastic leukemia (B-ALL), with a higher frequency of rearrangements resulting in overexpression of cytokine receptor-like factor 2 (CRLF2). Here, we investigated the impact of CRLF2 overexpression on B-cell progenitor proliferation, immunophenotype, and gene expression profile in the Dp(16)1Yey (Dp16) mouse model of DS compared to wild-type (WT) mice. CRLF2 overexpression enhanced immature B-lymphoid colony development and increased the proportion of less differentiated pre-pro-B cells, with a greater effect in Dp16 versus WT. In CRLF2-rearranged (CRLF2-R) B-ALL patient samples, cells with higher CRLF2 expression exhibited a less differentiated B-cell immunophenotype. CRLF2 overexpression resulted in a gene expression signature associated with E2F signaling in both Dp16 B-progenitors and in DS-ALL patient samples, and PI3K/mTOR and pan-CDK inhibitors which reduce E2F-mediated signaling demonstrated cytotoxicity in CRLF2-R B-ALL cell lines and patient samples. CRLF2 overexpression alone in Dp16 stem and progenitor cells did not result in leukemic transformation in recipient mice. Thus, CRLF2 overexpression results in reduced B cell differentiation and enhanced E2F signaling in Dp16 B-progenitor cells and DS-ALL patient samples. These findings suggest a functional basis for the high frequency of CRLF2-R in DS-ALL as well as a potential therapeutically targetable pathway.

GRAPHICAL ABSTRACT

Introduction

Children with Down syndrome (DS) have a 10-fold increased risk of developing B cell acute lymphoblastic leukemia (B-ALL), and have poorer outcomes due to both increased relapse and treatment-related mortality [1, 2]. The spectrum of cytogenetic alterations is very different in DS compared to non-DS ALL. Approximately half of DS-ALL cases have cytokine receptor-like factor 2 rearrangements (CRLF2-R), compared to only 5-10% of non-DS ALL cases, with half of these also having Janus Kinase 2 (JAK2)-activating point mutations [3-8].

We sought to identify functional effects to explain the increased frequency of CRLF2-R ALL in individuals with DS, and identify upregulated signaling pathways in this subtype, which could be targeted therapeutically. Alternative therapeutic approaches offer benefit in patients with DS, due to their increased vulnerability to chemotherapeutic toxicity [1, 9, 10]. We utilized the Dp(16)1Yey (Dp16) mouse model of DS, which has triplication of ~115 genes with orthologues on human chromosome 21 (Hsa21), including the Down syndrome critical region (DSCR) [11]. Since none of the existing mouse models of DS spontaneously develop ALL, we evaluated ex vivo effects of CRLF2 overexpression in Dp16 versus WT hematopoietic stem and progenitor cells (HSPCs), including colony formation, lymphoid differentiation, and gene expression. We also transplanted CRLF2-overexpressing Dp16 or WT cells into recipient mice. See Supplementary Data for Methods.

Results and Discussion

Utilizing the Dp16 mouse model of DS, we compared effects of CRLF2 overexpression in Dp16 versus WT BM cells. We transduced Dp16 and WT BM HSPCs with a CRLF2-GFP or control GFP vector, and confirmed 70-95% of cells expressed GFP or hCRLF2 and GFP by flow cytometry prior to experiments (Supplemental Figure 1A). CRLF2 overexpression increased B-lymphoid colony production in both genetic backgrounds. Colonies persisted for only one replating (data not shown), indicating CRLF2 overexpression was insufficient for transformation. Colony production was lower, but the fold increase in colony formation with CRLF2 overexpression was significantly greater, in the Dp16 background (mean fold change 10.8 vs 6.4 p = 0.004, Figure 1A). These findings are consistent with prior studies reporting reduced B-lymphoid progenitors in DS models, including trisomy 21 fetal liver cells [12] and the Ts65Dn [13, 14] and Ts1Rhr [13] mouse models of DS.

Figure 1. CRLF2 overexpression enhances B-lymphoid colony formation and reduces B cell maturation, with a greater fold change in the Dp16 background.

(A) CRLF2 overexpression in BM HSPCs from both WT and Dp16 mice increased B-lymphoid colony growth compared to GFP control. Histogram depicts colony counts from six samples (two technical replicates across three independent experiments, Student’s t-test, *p<0.01, **p<0.001). Error bars indicate standard deviation. (B) CRLF2 overexpression in WT and Dp16 HSPCs grown in B cell-differentiation conditions resulted in an increased percentage of the less-differentiated pre-pro-B cells. This effect was greater in the Dp16 versus WT genetic background. Histogram shows percentage of pre-pro-B cells in the GFP⁺ gate of each experimental group, with mean values from three independent experiments with two biological replicates each (Student’s t-test, *p<0.05, **p<0.001). Error bars indicate standard deviation.

To characterize the effects of CRLF2 overexpression on B-lymphoid differentiation in the Dp16 and WT backgrounds, we performed Hardy fraction flow cytometric analysis [15] on CRLF2-GFP and GFP-transduced HSPC-enriched BM following one week co-culture on OP9 cells, which promote B cell differentiation [16]. Overexpression of CRLF2 significantly increased the percentage of cells in the less differentiated Hardy Fraction A (pre-pro-B) (mean of 3.9% vs 1.7% in Dp16, p=0.0002; and 2.8% vs 1.6% in WT, p=0.0002). The effect of CRLF2 overexpression was significantly greater in Dp16 versus WT cells (3.9% vs 2.8%, p=0.012) (Figure 1B).

We further evaluated the effects of CRLF2 overexpression on B-lineage differentiation by examining CRLF2-R B-ALL cases that exhibited both a less differentiated pro-B (CD34⁺/CD19⁺) population and a more differentiated pre-B (CD34⁻/CD19⁺) population. We identified two cases with these features, one DS-ALL (839) and one non-DS ALL (105127-R) (Supplemental Table 1). B-ALL cells with the highest CRLF2 staining demonstrated a significantly higher proportion of pro-B cells (78.2% versus 51.4%, p=0.024, Figure 2). We also assessed the correlation of CRLF2 and CD34 staining in non-DS CRLF2-R ALL patient samples, which only had CD34⁺ pro-B cells. We observed a significant (p<0.0001) positive correlation between CRLF2 and CD34 in each B-ALL patient sample. We observed the strongest correlation (by r value) between CRLF2 and CD34 in the DS-ALL patient sample (839, Supplemental Figure 2). Thus, higher levels of CRLF2 were associated with reduced B progenitor differentiation, similar to that observed in Dp16 and WT HSPCs.

Figure 2. CRLF2 overexpression is associated with reduced B cell differentiation in B-ALL patient samples.

(A) In two CRLF2-overexpressing B-ALL patient samples (DS-ALL 839 and non-DS ALL 105127-R) demonstrating both pro-B and pre-B populations, the high-CRLF2 mean fluorescence intensity (MFI) quartile demonstrated a significantly higher mean proportion of pro-B cells (p=0.024). (B) Representative flow plots showing CRLF2 overexpression correlates with a less differentiated immunophenotype in B-ALL patient samples 839 (DS) and 105127-R (non-DS).

These effects of CRLF2 overexpression in reducing B cell differentiation are concordant with prior studies [3, 4]. Our study is unique in demonstrating a significantly greater effect in the Dp16 versus WT background. The greater effect of CRLF2 overexpression in Dp16 versus WT HSPCs suggests that CRLF2 rearrangements provide a greater competitive advantage in the DS background, which may explain why this lesion occurs in a greater proportion of DS-ALL cases.

To further evaluate CRLF2 overexpression in the Dp16 and WT backgrounds, we compared transcriptional signatures associated with CRLF2 overexpression in sorted GFP⁺ Dp16 and WT HSPCs after one week of OP9 co-culture, which promotes B cell differentiation [16, 17]. We performed RNA sequencing and gene set enrichment analysis (GSEA) using the Hallmark gene expression sets [18] for these comparisons: (1) Dp16 CRLF2 versus Dp16 GFP; (2) Dp16 CRLF2 versus WT CRLF2; and (3) WT CRLF2 versus WT GFP. We compared three independent replicates per group, and confirmed CRLF2 overexpression (Supplemental Figure 1B). We found that CRLF2-mediated decreases in B cell differentiation in the Dp16 background are associated with upregulated E2F signaling (Figure 3). This was the only gene set with a familywise error rate p value <0.05 (Supplemental Table 2). To assess the relevance to human DS-ALL, we performed GSEA using publicly-available data from three cohorts containing CRLF2-overexpressing and CRLF2-WT DS-ALL patient samples. We observed significant upregulation of E2F signaling in two of three cohorts (GSE20910 and GSE17459-BFM) [7, 19], providing further support for the potential relevance of the E2F pathway as a therapeutic target for CRLF2-overexpressing DS-ALL (Supplemental Figure 3 and Supplemental Table 3).

Figure 3. Dp16 CRLF2 cells demonstrate enrichment for E2F targets.

Gene set enrichment analysis plots show upregulation of E2F targets in Dp16 CRLF2 cells compared to (A) Dp16 GFP cells and (B) WT CRLF2 cells. Normalized enrichment score (NES), false discovery rate (FDR) q-value, and familywise error rate (FWER) p-value are also displayed.

The majority of the leading edge genes from the E2F gene set are associated with DNA synthesis and cell cycle progression, and have potential as targets for leukemia therapies, including Wee1 [20] and Rrm2 [21] (Supplemental Table 4). We did not observe significant enrichment for Hallmark gene expression sets in WT cells overexpressing CRLF2, suggesting that CRLF2 confers a selective advantage in the DS background. In addition, a high percentage of DS-ALL [22] and iAMP21 B-ALL [23] cases have RB1 deletions, suggesting that higher E2F activity may be selected for with trisomy 21, via CRLF2 overexpression and/or RB1 deletion. Studies are needed to decipher the region of triplication contributing to this effect.

Targeting E2F directly is difficult, due to multiple E2F isoforms and limited potency of pan-E2F inhibitors [24, 25]. E2F activation involves signaling inputs from PI3K/mTOR and CDK activity [26-28]. We reasoned that inhibiting these pathways may be cytotoxic in CRLF2-overexpressing DS-ALL blasts. We tested a PI3K/mTOR inhibitor AZD2014 and a pan-CDK inhibitor AT7519, which directly reduce expression of E2F-transcribed genes [26, 29], in seven non-DS B-ALL cell lines and nine PDX-expanded B-ALL patient samples (Supplemental Table 1).

Both compounds demonstrated nanomolar-range cytotoxicity in all B-ALL cell lines, irrespective of CRLF2 mutation status. In the PDX-expanded primary samples, the compounds demonstrated nanomolar-range cytotoxicity in CRLF2-overexpressing cases, but demonstrated inconsistent potency in the CRLF2/JAK2 WT cases, and no clear specificity for CRLF2-overexpressing DS-ALL cases (Supplemental Figure 4). These experiments demonstrate preclinical utility of compounds that reduce E2F signaling in CRLF2-R DS and non-DS ALL.

Finally, we transplanted CRLF2-overexpressing Dp16 or WT HSPCs into recipient mice. We confirmed 70-95% transduction efficiency of GFP and CRLF2-GFP in Dp16 and WT HSPC cultures before transplantation. Mice engrafted with 2-20% GFP⁺ cells up to day 100 post-transplant, and the stained GFP⁺ cells of most recipients were typically B220⁺ B cells (Supplemental Figure 5), alongside GFP⁺ myeloid or T cells. However, engraftment was undetectable by day 140 post-transplant. Average engraftment was higher at every time point, and sustained for longer, in mice receiving WT versus Dp16 cells, suggesting an inherent reduction in repopulating capacity of Dp16 cells. No mice transplanted with CRLF2-overexpressing Dp16 or WT HSPCs developed leukemia, with at least 7 mice per group monitored for 6 months, which included a minimum of one month after engraftment was undetectable. Interestingly, B220⁺ CRLF2-overexpressing Dp16 cells displayed reduced staining intensity in most engrafted recipient mice relative to other groups and to B220⁺ GFP⁻ cells within the same mice. Since B220 expression increases as B cell differentiate [15], CRLF2-overexpressing Dp16 cells may have demonstrated particularly dim B220 staining due to reduced B cell differentiation relative to the other conditions. As a control, we transplanted mice with HSPCs overexpressing NRAS^G12D. Mice transplanted with NRAS^G12D-overexpressing Dp16 or WT HSPCs developed rapid disease, with 9/24 WT mice and 10/25 Dp16 mice succumbing to CD4⁺CD8⁺ T-ALL with median latencies of 92 and 101 days (data not shown). The only current DS-ALL mouse model requires CRLF2 overexpression, JAK2^R683G, and Pax5^+/− [13]. However, it utilizes the Ts1Rhr mouse model of DS, which only has 31 triplicated Hsa21 orthologues. We expected the extra triplicated Hsa21 orthologues in Dp16 mice might contribute to B-ALL with only CRLF2 overexpression, but additional alterations are likely required.

Our Dp16 ex vivo differentiation model demonstrates the proliferative advantages of CRLF2 overexpression in Dp16 versus WT cells, which may contribute to the increased susceptibility to CRLF2-R ALL in children with DS. The upregulated E2F signaling may provide opportunities for therapeutic intervention. This model will be useful to test cytotoxicity of targeted therapies and combination regimens, and to investigate the leukemogenic effects of other clinically relevant oncogenes. Ultimately, insights from these studies may improve efficacy and reduce toxicity of therapy for CRLF2-R ALL, particularly in children with DS.

Supplementary Material

1

Supplemental Figure 1. Confirmation of CRLF2 overexpression in transduced cells. (A)CRLF2 overexpression in transduced BM HSPCs from both WT and Dp16 mice was confirmed by flow cytometry prior to downstream experiments. Representative results are shown. (B) hCRLF2 overexpression in the independent biological replicates used for RNA-Seq analysis was confirmed (hCRLF2 transcripts per million (TPM) shown, Student’s t-test, *p<0.005).

2

Supplemental Figure 2. CD34 and CRLF2 staining intensity is positively correlated for five independent B-ALL samples. Five PDX-expanded CRLF2-R DS- or non-DS ALL patient samples were assessed for CD34 (PE) and CRLF2 (BV605) surface staining intensity. We observed a significant (p<0.0001) positive correlation between CD34 and CRLF2 for each of the five samples. Best-fit trendlines are shown, r values were calculated by Pearson correlation.

3

Supplemental Figure 3. CRLF2-overexpressing DS-ALL samples display upregulation of E2F signaling. Gene set enrichment analysis plots show upregulation of E2F targets in CRLF2-overexpressing compared to CRLF2-WT DS-ALL patient samples in patient samples from (A) GSE20910 and (B) GSE17459-BFM. Normalized enrichment score (NES), false discovery rate (FDR) q-value, and familywise error rate (FWER) p-value are also displayed.

4

Supplemental Figure 4. B-ALL cell lines and DS-ALL and non-DS ALL patient samples are sensitive to AZD2014 and AT7519. CRLF2-R (solid circles) or CRLF2-WT (open circles) B-ALL cell lines were incubated for 72 hours with different doses of (A) AZD2014 or (B) AT7519. CRLF2-R (solid shapes) or CRLF2-WT (open shapes) DS-ALL (triangles) and non DS-ALL (circles) patient samples were incubated for 72 hours with different doses of (C) AZD2014 or (D) AT7519. Viability was measured by ATP assay, with three replicates per data point.

5

Supplemental Figure 5. The majority of engrafted GFP⁺ Dp16 or WT HSPCs are B220⁺ B cells. Recipient mice were bled every 2-4 weeks, and cells were stained with B, T, or myeloid cell markers. In the majority of engrafted mice, most of the stained GFP⁺ cells are B220⁺ B cells, with some myeloid and T cells. Representative examples demonstrate percentages of B220⁺ B cells. B220⁺ B cells, stained with two colors to differentiate between B220⁺ cells and myeloid or T cells, are displayed in the upper right gate. See Supplemental Data for methods details.

Highlights.

• High CRLF2 reduces B cell differentiation in Down syndrome mouse progenitor cells

• High CRLF2 is associated with less B-cell differentiation in human leukemia cells

• High CRLF2 is associated with E2F signaling in Down syndrome mouse and human cells

• Inhibitors of E2F signaling are cytotoxic in CRLF2+ acute lymphoblastic leukemia

• High CRLF2 in Dp16 cells is not sufficient to generate leukemia in recipient mice

Data Availability Statement

The datasets generated and/or analyzed during the current study are available in the European Nucleotide Archive repository: https://www.ebi.ac.uk/ena/browser/view/PRJEB43591.