Retroviral insertional mutagenesis identifies the del(5q) genes, CXXC5, TIFAB and ETF1, as well as the Wnt pathway, as potential targets in del(5q) myeloid neoplasms

An interstitial deletion of the long arm of chromosome 5, del(5q), is a recurring abnormality in myeloid disorders, including myelodysplastic syndromes (MDS), de novo acute myeloid leukemia (AML), and therapy-related myeloid neoplasms (t-MN) comprising therapy-related MDS and AML (t-MDS/t-AML).1 In t-MN, a del(5q) occurs in approximately 40% of patients and is associated with prior therapy with alkylating agents, a complex karyotype, TP53 mutations, a strong propensity to progress to t-AML, and a poor outcome.32 We previously identified two haploinsufficient tumor suppressor genes on 5q, the early growth response gene, EGR1 [(5q31.2, deleted in all t-MN with a del(5q)] and the adenomatous polyposis coli gene, APC [5q22.2, deleted in >95% of t-MNs with a del(5q)], and showed that heterozygous loss of Egr1 and Apc in mice promote the pathogenesis of MDS/AML, in co-operation with knockdown of Trp53 and/or alkylating agent therapy.4 EGR1 is a member of the WT-1 family of transcription factors and is a transcriptional regulator of many tumor suppressor genes, including TP53, CDKN1A/p21 and TGFB.5 The APC protein acts as a tumor suppressor and is a negative regulator of the WNT signaling pathway.6 Adding to the difficulty of identifying additional genes on 5q that may be contributing to the pathogenesis of MDS/AML, most patients have large deletions encompassing over 70 Mb (5q14-q33). The complex cytogenetic abnormalities and recurrent somatic mutations, such as ASXL1, NF1, TET2 and TP53 associated with advanced MDS and AML with a del(5q), has provided some insight; however, the complete genetic profile, and consequences thereof, are not yet known.73 Whether prior cytotoxic therapy induces mutations and/or promotes expansion of pre-existing genetic mutations in t-MNs is also poorly understood. We used a mutagenesis approach in mice, heterozygous for Egr1 (Egr1) to identify genes and signaling pathways that predispose to developing del(5q) myeloid neoplasms (MNs).

Forward genetic screens with the recombinant retrovirus MOL4070LTR create a sensitized background for the development of MNs, and have been shown to be a powerful approach for the identification of secondary cooperating mutations.8 Untreated WT and Egr1 mice do not develop disease up to two years of age;5 however, MOL4070LTR-treated WT and Egr1 mice developed T-cell or myeloid neoplasms with a median survival of 480 and 430 days (d), respectively (P=0.150) (Figure 1A). Egr1 mice developed transplantable MNs with a shorter latency (389 d vs. 474 d; P=0.049) and at a higher overall frequency than WT littermate controls (Figure 1B and Online Supplementary Figure S1). There was also a significant increase in myeloid versus no disease in Egr1 mice (P<0.001), but not WT mice (P=0.36) (Figure 1C), indicating that loss of one allele of Egr1 shifts the disease spectrum to favor the development of MNs. A Fisher’s exact test indicated that the higher frequency of MNs observed in Egr1 mice compared to WT showed a trend towards significance (P=0.052). The morphological and phenotypic features of the diseases were similar in the Egr1 and the WT mice (Figure 1D).

Figure 1.Egr1 haploinsufficiency accelerates MOL4070LTR-induced myeloid disease. (A and B) Kaplan-Meier survival curves of WT (n=61) and Egr1^+/− (n=77) neonatal mice injected with MOL4070LTR retrovirus. (A) Overall survival and (B) survival of mice with myeloid neoplasms. Median survival of all mice or mice that specifically developed myeloid disease was 430 and 389 days for Egr1^+/− and 480 and 474 days for WT mice. Myeloid neoplasms were identified by splenomegaly, leukocytosis, anemia, thrombocytopenia, and enrichment of myeloid cells in hematopoietic and non-hematopoietic tissue. (C) Distribution of the disease phenotypes observed in MOL4070LTR-treated mice. In Egr1^+/− mice, there is an increase in neoplastic diseases concomitant with an increase in myeloid versus lymphoid neoplasms. A trend towards a significant increase in myeloid disease in Egr1^+/− mice compared to WT was observed (P=0.052 by one-tailed Fisher’s exact test). (D) Complete blood counts (CBC) and phenotype analysis of WT and Egr1^+/− mice treated with MOL4070LTR retrovirus, when the mice developed a myeloid neoplasm. ”No disease” mice were Egr1^+/− mice that were not treated with retrovirus and remain healthy. WBC: white blood cells; RBC: red blood cells; Hb: hemoglobin; PLT: platelets. Spleen size (g) and flow cytometric analysis showing granulocytes (Gr1⁺, CD11b/Mac1⁺), KIT⁺ myeloid cells, and immature erythroblasts (CD71⁺, Ter119⁺) in the spleen. Significant differences using pairwise 2-sample Student’s t-tests (Control vs. WT; Control vs. Egr1^+/−) are shown.

We mapped the common retroviral integration sites (CISs) in MNs from WT and Egr1 mice using bar-coded splinkerette PCR and Illumina high-throughput sequencing. TAPDANCE software, a fully automated process to identify and annotate CISs,9 identified 159 and 365 CISs (P<0.05) in myeloid neoplasms from WT (n=29) mice or Egr1 mice (n=46), respectively. Whereas none of the CISs were significantly associated with the WT genotype, 5 CISs showed significant associations with the Egr1genotype (Figure 2A). None of the genes implicated in cell growth and cancer within the common insertion sites (Gfi1, Ccdc21, Cd47, Cd52, Cxxc5, Evi5, Psd2, Tmem173 and Ubxn11) displayed a significant change in expression in myeloid neoplasms due to retroviral integration (Online Supplementary Figure S2A) except for Evi1 (Mecom: MDS1 and EVI1 complex locus), which showed a significant increase (Online Supplementary Figure S2B). We showed that self-renewal of Evi1-expressing progenitors was enhanced by loss of Egr1 in vitro; however, there was no apparent co-operation between Evi1 overexpression and Egr1 haploinsufficiency leading to hematopoietic neoplasms using an in vivo mouse model (Online Supplementary Figure S2C and D).

Figure 2.Co-occurring CISs target human del(5q) genes. (A) Common insertion site (CIS) analysis was performed on mice that developed myeloid neoplasms. List of CISs found statistically to occur more frequently in Egr1^+/− than WT mice. Some mice had multiple insertions for the same CIS. The CIS on chromosome 18 is syntenic to human 5q31.2, and includes Dnajc18, Cxxc5, Tmem173, Psd2, and Ecscr. (B) Human 5q31.1–q31.2 is syntenic to mouse chromosomes 13 and 18. Fisher’s exact test with multiple testing correction revealed that CISs on chromosome 18 (event 1) and on chromosome 13 (event 2) co-occur more frequently than predicted by chance (P<0.0001). The second CIS (event 2) was found with equal frequency in Egr1^+/− and WT mice. (C) Expression of candidate del(5q) genes was examined by Q-PCR in Egr1^+/− mice that had proviral integrations proximal to candidate gene (CIS YES) versus those that did not (CIS NO). (D) Fold expression of genes was measured by comparing gene expression in MNs with del(5q) compared to non-del(5q). Expression in 2 independent cohorts for the University of Chicago (UC-1; GSE39991 and UC-2; SRA061655) as well as the TCGA database were examined. UC-1: 10 del(5q) and 28 non-del(5q); UC-2: 12 del(5q) and 16 non-del(5q); TCGA: 18 del(5q) and 161 non-del(5q). Evidence for haploinsufficient expression in del(5q) therapy-related myeloid neoplasm patients patients was observed for H2AFY, TIFAB, TMEM173, and CXXC5.

Identifying which genes on 5q play a critical role in the development of MDS and AML continues to be a major challenge. Interestingly, one CIS significantly associated with MNs in Egr1+/− mice mapped to a region of mouse chromosome 18 that is syntenic to human 5q31.2 and proximal to the DNAJC18, ECSCR, TMEM173, CXXC5 and PSD2 genes (Figure 2B). Of relevance to MNs with a del(5q), this CIS co-occurred with a CIS on chromosome 13 (syntenic to 5q31.1 and proximal to the TIFAB, and H2AFY genes) (P=0.001). In Gr1CD11b myeloid bone marrow cells isolated from Egr1 mice that had proviral integrations proximal to the 5q genes (CIS YES), compared to mice that did not (CIS NO), Tifab expression was slightly higher and Cxxc5 lower (though this was not significant), raising the possibility that aberrant expression of these genes may be important (Figure 2C). In leukemia samples from t-MN patients with a del(5q), CXXC5, TMEM173, TIFAB, and H2AFY showed approximately 50% lower expression relative to non-del(5q) samples, consistent with haploinsufficiency (Figure 2D). Recent studies suggest that TIFAB (TRAF-interacting protein with forkhead-associated domain, family member B) and CXXC5 (CXXC finger protein 5) may be important genes for the pathogenesis of myeloid disease. Deletion of Tifab contributes to an MDS-like phenotype in mice by changing the dynamic range of innate immune pathway activation.10 CXXC5, also known as RINF, is a candidate tumor suppressor in AML, since it inhibits leukemia cell proliferation and Wnt signaling.1211 Taken together, TIFAB and CXXC5 warrant strong consideration for future studies of myeloid neoplasms associated with a del(5q).

Traditional CIS analysis reveals a potentially important gene (e.g. Evi1) only if insertions proximal to a genomic region occur in multiple mice more frequently than could be expected by chance. It does not, however, take into account proviral insertions, proximal to multiple different genes within the same signaling pathway or ontology term, in multiple mice. We applied the Genomic Regions Enrichment of Annotations Tool (GREAT) to evaluate whether Egr1 mice with myeloid disease had proviral integrations proximal to multiple genes within the same signaling pathway or gene ontology.13 GREAT analysis revealed the twenty most significant Molecular Signature Database (MSigDB) pathway and gene ontology molecular terms that were enriched in WT and Egr1 mice that developed MNs (Online Supplementary Tables S1 and S2). As expected, several pathway terms were shared in both WT and Egr1 mice; however, others were unique (Figure 3A). Shared pathways included an enrichment of cancer signatures, such as AML and CML, as well as MAPK, JAK-STAT and cytokine signaling.

Figure 3.GREAT analysis of proviral insertions in Egr1+/− mice reveals deregulation of the Wnt signaling pathway. (A) GREAT was used to search for enrichment of proviral insertions near genes that target common signaling pathways, biological terms and shared transcription factor binding sites. Venn diagram illustrates the terms, assigned by GREAT analysis, which were enriched in WT, Egr1^+/− or both mouse genotypes. Interestingly, an enrichment of proviral insertions near Wnt signaling pathway genes (right) was observed in Egr1^+/− mice. (B) Expression of Lef1 and Tcf7 (major mediators of Wnt signaling) was assessed in spleen cells isolated from mice that developed myeloid disease after exposure to MOL4070LTR virus. Lef1 and Tcf7 are expressed at significantly higher levels in Egr1^+/− mice compared to WT mice consistent with activation of the Wnt/β-catenin pathway in Egr1^+/− mice. (C) The MSigDB predicted promoter motifs ontology set contains genes that share a transcription factor binding site in their promoters. Venn diagram illustrates that genes with promoter motifs that bind EGR transcription factors were specifically targeted in WT mice suggesting that a disruption of EGR-regulated genes, through proviral insertion, promotes myeloid disease. EGR1 target genes (confirmed by EGR1 ChIP analysis in K562 cells) with significant changes in gene expression of del(5q) versus non-del(5q) MNs is shown on the left. Deregulation of these genes, possibly through EGR1 haploinsufficiency, may play an essential role in the malignant transformation process observed in patients with a del(5q). Of note, ETF1 maps to 5q31.2 immediately distal to EGR1 within the commonly deleted segment in MNs.

Biological processes and pathways enriched only in Egr1 mice included hematopoiesis, PIP₃ signaling, and Wnt/β-catenin signaling, as well as protein kinase and cytokine binding. We previously observed that concurrent haploinsufficiency of Egr1 and Apc, a key regulator of the Wnt/β catenin pathway, co-operate in the pathogenesis of MDS and AML. In this regard, we observed an increase in expression of two major transcriptional mediators of Wnt/β catenin signaling: Lef1 (P<0.0001) and Tcf7 (P=0.01014). This is consistent with Wnt activation in MOL4070LTR Egr1 mice (Figure 3B). LEF1 (Lymphoid Enhancer-Binding Factor 1) was also very highly expressed (2–15-fold) in del(5q) t-MN patients (University of Chicago series 1 and 2) and del(5q) AML patients in the TCGA database, and we recently identified significant changes in the expression of Wnt pathway genes, consistent with activation, in del(5q) patients.6

EGR1, a transcriptional regulator expressed at haploinsufficient levels in del(5q) t-MN patients, is involved in the homeostasis of hematopoietic stem cells, in myeloid differentiation, and in neoplastic transformation.144 We hypothesize that dosage-dependent deregulation of hematopoietic transcriptional programs contributes to the development of myeloid disease; however, the affected Egr1 targets genes are still not known. Interestingly, there was a significant enrichment of provirally-targeted genes that contain binding sites for at least one or more EGR family members (EGR1, EGR2, EGR3, EGR4) in WT, but not Egr1 mice (Figure 3C and Online Supplementary Tables S3 and S4). Of these 40 potential EGR target genes, we identified 7 genes that were confirmed EGR1 targets by chromatin immunoprecipitation in K562 cells, and were differentially expressed in del(5q) versus non-del(5q) MNs. These include EGR1, ETF1 and ITPR1 [down-regulated in del(5q) neoplasms] and BCL6, KLF3, LEF1 and NRGN [up-regulated in del(5q) neoplasms]. These genes warrant consideration as potential EGR1 target genes that are deregulated during leukemogenesis in t-MN del(5q) patients. Notably, ETF1 (Eukaryotic Translation Termination Factor 1) has previously been suggested to act as a haploinsufficient tumor suppressor protein in MDS/AML patients with a del(5q), possibly through production of potentially oncogenic, aberrant proteins.15 In addition, ETF1 (5q31) and LEF1 (WNT target gene) expression may be deregulated in del(5q) patients due to EGR1 haploinsufficiency and/or loss of one copy of ETF1 and APC (5q22.2), respectively.

Patients with high-risk MDS, AML or t-MN with a del(5q) present with a very complex genetic profile and have a poor response to current therapies and a poor prognosis. In this study, we used our previously described mouse model of t-MN, with loss of the critical del(5q) haploinsufficient myeloid tumor suppressor gene, EGR1. Haploinsufficiency of Egr1, does not cause phenotypic abnormalities in mice under basal conditions; however, a reduction in Egr1 levels creates a sensitized background that co-operates with secondary mutations to give rise to myeloid malignancies. Moreover, our findings provide evidence that deregulation of genes at 5q31.1–5q31.2 (with TIFAB, CXXC5 and ETF1 being the most likely candidates) and activation of specific signaling pathways (with Wnt/β-catenin pathway the most likely aberrant pathway) are critical alterations that promote the development of MNs in co-operation with EGR1 haploinsufficiency. Notably, our studies also identify potential EGR1 target genes that may be deregulated during the progression towards t-MN, with ETF1 and LEF1 warranting further investigation.

Understanding the pathobiology of MN with a del(5q) has been challenging due to the large size of deletion and the ensuing haploinsufficient loss of multiple genes, and the additional genetic alterations that co-operate with 5q loss. This study is a step towards defining the critical pathways at play in MNs with a del(5q), which may inform the development of effective therapeutic strategies for this genetic subset of patients.