Unexpected suppression of tumorigenesis by c-MYC via TFAP4-dependent restriction of stemness in B lymphocytes

Lymphocytes are one of the most rapidly proliferating cell types in the postnatal organism.^1-6 Extensive proliferative burst is essential to form diverse antigen receptor repertoires and facilitate requisite antigen-specific immune responses. In addition to demands for accurate DNA replication during extensive cell division, B lymphocytes are vulnerable to genomic instability as a result of DNA editing by recombination-activating gene (RAG) or activation-induced cytidine deaminase that primarily targets antigen receptor loci but also attacks other sites throughout the genome,⁷ potentially rendering lymphocytes highly cancer prone. However, cancers derived from developing lymphocytes are rarer than would be expected from their inherent risk, and their incidence is lower than that of common epithelial cancers (eg, 1.7 cases of acute lymphoblastic leukemia [ALL] per 100 000 population annually vs 38-127 cases of colon, lung, and breast cancer per 100 000 population annually ).⁸ Accordingly, we hypothesized that lymphocytes are protected from transformation by putative tumor-suppressive programs that are tightly linked to their proliferative program. Such programs may preserve the genomic integrity or restrict specific gene-expression programs that would enhance transformation.

The transcription factor c-MYC is a strong oncogenic driver; activating mutations and translocations that deregulate its expression are hallmark features of numerous tumors in humans and mice.^9-11 c-MYC is also essential for the proliferation of normal lymphocytes through control of the cell cycle and metabolism.^12-15 If lymphocytes activate putative tumor suppressors tied to their proliferation, c-MYC itself may directly induce such tumor suppressor genes as a feedback pathway. Therefore, it is conceivable that perturbation of any such pathways might accelerate the development of lymphoid malignancies induced by aberrant c-MYC expression, although it remains unknown whether such a c-MYC–induced tumor-suppressive mechanism exists.

In this study, we profiled direct c-MYC target genes that are mutated in primary human lymphoid cancers and identified the transcription factor TFAP4 (also known as AP4) as a c-MYC–inducible tumor suppressor in B lymphocytes. Recurrent mutations of TFAP4 are found predominantly in its DNA binding domain in human cancers, including ALL, and a notable proportion of cases of Burkitt lymphoma (BL). In c-MYC–driven mouse B-cell tumor models, TFAP4 functions as a c-MYC–induced haploinsufficient tumor suppressor. Single allele deletion of Tfap4 or specific deletion of a c-MYC binding site in the locus shortened the latency of the MYC transgene-driven tumorigenesis, with frequent loss of heterozygosity (LOH). Mechanistically, c-MYC drives the proliferation of B-cell progenitors, as well as promotes their differentiation, thus restricting the self-renewing capacity of B cells in undifferentiated states through TFAP4-dependent repression of the stemness factor Erg. Finally, reduced expression of TFAP4, despite high c-MYC expression, is associated with poor prognosis of a subset of patients with B-cell ALL (B-ALL). These results collectively show that the c-MYC–TFAP4 axis has dual roles in supporting the clonal expansion of normal lymphocytes^16,17 and paradoxically, suppresses the transformation of proliferating B cells.

Tfap4^−/−,^18,Tfap4^F/F,^16,Erg^F/F,^19,Cd79a-icre,²⁰ and Myc-GFP²¹ mice were described previously and were backcrossed to the C57BL/6N background for >10 generations. Eμ-Myc,^22,Ighg1-cre,^23,Rosa26^CAG-MYC,²⁴ and Rosa26^Pik3ca*25  mice were from The Jackson Laboratory. C57BL/6N and B6-Ly5.1 mice were from Charles River. All mouse experiments were conducted according to the protocol approved by the Washington University Animal Studies Committee.

Eμ-Myc mice were monitored weekly for palpable tumor development and body weight loss. Mice were euthanized when tumor sizes exceeded 20 mm in diameter or mice exhibited >20% body weight loss or paralysis. For transplantation of retrovirally transduced B cells, CD19⁺ bone marrow (BM) cells, isolated using anti-CD19 MicroBeads (Miltenyi Biotec) from 6- to 10-week-old mice, were infected with viral supernatants, followed by IV transfer of 1 × 10⁶ cells into sublethally irradiated (5.5 Gy) mice. For in vivo persistence analysis, transduced cells were transferred into congenic mice at an ∼1:1 ratio of Tfap4^+/−, Tfap4^+/−Erg^+/−, or Erg^+/− cells to wild-type (WT) cells.

Gene expression data from BL samples²⁶ were downloaded from the European Genome-Phenome Archive (EGAS00001003778). Data were analyzed using Galaxy as follows: counts per gene for each sample were obtained by running featureCounts on GRCh38-aligned BAM files and normalized using DESeq2. The normalized data were imported into the Phantasus application²⁷ for further analysis. After log₂ transformation and quantile normalization, ENTREZ-annotated genes were filtered for mean normalized counts >2 across samples. Data for B-ALL samples were downloaded from TARGET phase II, and gene expression was analyzed in Phantasus. Differentially expressed genes (DEGs) in both patient cohorts were analyzed using a limma tool, followed by gene set enrichment analysis (GSEA).

Survival data were compared using the log-rank test with adjustment for multiple comparisons when necessary. P values were calculated using an unpaired 2-tailed Student t test for 2-group comparisons and 1-way analysis of variance (ANOVA) for multigroup comparisons with the Tukey post hoc test or Kruskal-Wallis test with Dunn’s multiple comparison test. Human microarray data from previously published studies were compared using Mann-Whitney’s U test for 2-group comparisons and the Kruskal-Wallis test for multigroup comparisons. All statistical analyses were done using Prism. Adjusted values of P < .05 were considered significant.

To determine whether c-MYC engages tumor-suppressive programs in B lymphocytes, we initially defined c-MYC–dependent genes in developing B cells between Eμ-Myc and littermate WT B220⁺ immunoglobulin M (IgM)⁻ pro/pre-B cells. We combined this analysis with gene expression profiles of mouse germinal center B (GCB) cells¹³ and identified 107 genes that were commonly dependent on c-MYC in pro/pre-B and GCB cells (Figure 1A-C). The candidate genes were prioritized based on known roles in gene regulation, direct c-MYC targets,^16,17,28 and mutation frequencies in human hematopoietic malignancies using the COSMIC (Catalogue of Somatic Mutations In Cancer) database. Tfap4, encoding a bHLH transcription factor, was identified as a gene meeting all of the criteria (Figure 1D). We confirmed that TFAP4 expression was increased in B220⁺ IgM⁻ BM cells from Eμ-Myc mice compared with those from littermate WT mice (Figure 1E).

Somatic mutations of TFAP4 were found in lymphoid and other cancers registered in COSMIC, TCGA (The Cancer Genome Atlas), and PeCan (Pediatric Cancer Genomic Data Portal). Notably, 5 of 6 PeCan cases with TFAP4 mutations were found in lymphoid tumors (3 BL, 1 B-ALL, and 1 T-cell ALL [T-ALL]). Although TFAP4 mutations in these databases were rare (288/37 420 unique samples in COSMIC, 65/10 202 in TCGA, and 6/2578 in PeCan), analyses identified missense TFAP4 mutations in 12 of 101 cases in the BL data set:²⁶ 6 of 32 endemic BLs and 6 of 69 sporadic and HIV⁺ cases (Figure 1F; supplemental Figure 1A-B, available on the Blood Web site). Across these data sets, the majority of recurrent somatic mutations (≥4 independent cases), including 4 of 6 cases in PeCan lymphoid tumors and 3 additional BL cases, were mapped to the DNA binding domain (basic region)²⁹ (Figure 1G-H). A single amino acid substitution in the DNA binding residue was sufficient to compromise its transactivating potential, because expression of CD25/Il2ra, a direct TFAP4 target gene, was reduced in Tfap4^−/− CD8 T cells retrovirally complemented with the R50W or R60H variant compared with WT-TFAP4 or the R129W variant harboring a mutation outside of the basic region (Figure 1I; supplemental Figure 1C-F). These results suggest that the somatic mutations found in the basic region impair TFAP4 function and that the dysfunction of TFAP4 contributes to oncogenesis in lymphoid malignancies.

Because Tfap4 is a c-MYC–inducible gene^16,17,30 and many tumors associated with TFAP4 mutations express high levels of c-MYC, we next determined whether TFAP4 modulates c-MYC–induced tumorigenesis in B cells. We first used the Eμ-Myc mouse model, in which Myc is expressed under the control of the Igh μ enhancer, mimicking the t(8;14) IGH-MYC translocation.²² In this model, c-MYC is overexpressed in developing B cells starting at the pro-B–cell stage, leading to B-cell malignancy and nearly all mice dying within a year (Figure 2A). In striking contrast, haploinsufficiency of Tfap4 in Eμ-Myc mice accelerated tumorigenesis, with all mice dying by day 92, whereas homozygous Tfap4 deletion further accelerated tumorigenesis (Figure 2A). In 40% of tumors in Eμ-Myc mice, B220⁺ cells exhibited an immature phenotype, expressing Rag1 and VpreB1 (Figure 2B; supplemental Figure 2A-B), consistent with previous studies.^31,32 In contrast, 93% of Eμ-Myc Tfap4^+/− mice formed tumors with the immature phenotype (Figure 2B). The skewing of tumors toward the immature phenotype suggests that pro/pre-B cells with elevated c-MYC, but reduced TFAP4, become transformed with shorter latency after c-MYC is turned on in pro-B cells. The numbers of developing B cells, their short-term proliferation, and annexin V binding at the population level were comparable between Tfap4-haploinsufficient and Tfap4-proficient Eμ-Myc mice before tumor formation (supplemental Figure 2C-F). No survival difference was observed between Tfap4^+/+ and Tfap4^+/− Eμ-Myc pro/pre-B cells ex vivo (supplemental Figure 2G-H).

To determine whether TFAP4-mediated supression of Eμ-Myc tumors is cell-intrinsic, we deleted 1 or 2 Tfap4^F alleles in pro-B cells using Cd79a^icre, recapitulating the accelerated tumorigenesis and immunophenotype observed in germline Tfap4-deficient mice (Figure 2C; supplemental Figure 2I). We also tested whether TFAP4 functions as a suppressor of c-MYC–induced B cell malignancies in additional mouse models, because almost all B cells express aberrant MYC throughout their development in Eμ-Myc mice. First, we retrovirally expressed c-MYC in CD19⁺ BM cells from Tfap4^−/− or Tfap4^+/+ mice, followed by transfer to sublethally irradiated mice. In this model, we typically achieve transduction of ∼10% of pro/pre-B cells (data not shown), and tumor formation by c-MYC–transduced pre-B cells is dependent on a defect in a tumor suppressor.^33,34 In contrast to control mice receiving c-MYC–transduced Tfap4^+/+ CD19⁺ cells, the majority of mice receiving c-MYC–transduced Tfap4^−/− CD19⁺ BM cells died from pre-B–cell tumors (Figure 2D; supplemental Figure 2J). We also observed accelerated B-ALL disease progression induced by ectopic expression of p190 bcr-abl, which indirectly upregulates Myc, in CD19⁺ BM cells from Tfap4^−/− mice compared with those from Tfap4^+/+ mice (Figure 2E; supplemental Figure 2K). In addition to these immature B-cell malignancies, we observed accelerated death in the absence of TFAP4 in a mature B-cell lymphoma model induced by c-MYC and a constitutively active phosphatidylinositol 3-kinase²⁴ (Figure 2F). These data establish a cell-intrinsic requirement for TFAP4 in the suppression of MYC-mediated B-cell cancers in multiple mouse models.

c-MYC bound 2 consensus sites within the H3K27ac-marked intronic region at the Tfap4 locus in Eμ-Myc B220⁺ BM cells (supplemental Figure 3A). To determine whether direct activation of Tfap4 by c-MYC binding to this enhancer is necessary for tumor suppression, we deleted a 1-kb intronic region containing the 2 consensus sites using CRISPR/Cas9 (supplemental Figure 3A). Heterozygous deletion of the 1-kb region in Eμ-Myc mice recapitulated accelerated tumorigenesis (supplemental Figure 3B-C). This might be caused by the loss of other transcription factors binding at the enhancer or indirectly by altered expression of the neighboring genes. However, a more specific single-allele deletion of 1 MYC binding site also shortened survival of these mice compared with Eμ-Myc mice harboring WT Tfap4 loci (supplemental Figure 3B), whereas expression of 2 genes around the Tfap4 locus, or Runx1 located on the same chromosome, was unaffected (supplemental Figure 3C). Thus, direct activation of Tfap4 by c-MYC is essential for tumor suppression.

Tfap4 messenger RNA (mRNA) expression was comparable between pro/pre-B cells before transformation and pro/pre-B tumor cells in Eμ-Myc Tfap4^+/+ mice (Figure 3A). However, approximately one third of pro/pre-B tumors from Eμ-Myc Tfap4^+/− mice expressed substantially lower Tfap4 mRNA compared with pretransformed pro/pre-B cells (Figure 3A, dashed rectangle). This reduced Tfap4 expression was caused by LOH in these tumors (Figure 3B). A genomic region encompassing exons 2 through 4, deleted in the Tfap4^– allele, was detectable in all tumors from Eμ-Myc Tfap4^+/+ mice and a subset of tumors from Eμ-Myc Tfap4^+/− mice. By contrast, this genomic region was lost in tumor cells with low Tfap4 expression from Eμ-Myc Tfap4^+/− mice (Figure 3B, dashed rectangle). LOH of Tfap4 was also confirmed by loss of mCherry (mC) fluorescence in Eμ-Myc tumors harboring 1 null and 1 functional TFAP4-mC fusion knock-in allele¹⁷ (Tfap4^mC/−) (Figure 3C-D). These results indicate that, although TFAP4 downregulation is unnecessary for transformation, rare clones undergoing LOH at the Tfap4 locus predominantly contribute to tumorigenesis.

To gain insights into the molecular mechanisms of c-MYC–TFAP4–dependent tumor suppression, we profiled the gene expression of pretransformed B220⁺ IgM⁻ pro/pre-B cells, validated by unbiased expression of Ighv transcripts (supplemental Figure 4A), from Eμ-Myc Tfap4^+/− mice and control Eμ-Myc Tfap4^+/+ mice at 3 to 4 weeks of age. Because of rapid tumor development, analysis of pretransformed Eμ-Myc Tfap4^−/− pro/pre-B cells was excluded. Gross transcriptomic differences determined by principal component analysis and the expression of genes related to apoptosis, proliferation, and senescence were not detected between Eμ-Myc Tfap4^+/− cells and Eμ-Myc Tfap4^+/+ cells (supplemental Figure 4B-D). At the gene level, 29 genes were differentially expressed by ≥1.8-fold. Among these, 20 genes were directly bound by TFAP4, as confirmed by chromatin immunoprecipitation sequencing (current study), and 9 of the 20 genes were shared TFAP4 and MYC targets²⁸ in Eμ-Myc B cells (Figure 4A-C). Despite the lack of clear genotype-associated clustering (supplemental Figure 4E), an expression pattern of 17 of these 29 genes was preserved in Tfap4-deficient tumors, suggesting that their expression is maintained, or selected for, during tumorigenesis (supplemental Figure 4F).

To further examine gene expression program modules that were altered specifically in Tfap4-deficient pre-B tumors compared with Tfap4-proficient pre-B tumors, we conducted GSEA. Among curated gene sets in the Molecular Signature Database (https://www.gsea-msigdb.org/gsea/msigdb/),^35-37 genes related to MYC targets and E2F targets were significantly enriched in Tfap4-deficient/haploinsufficient tumors (Figure 4D-E). Gene sets associated with MYC and E2F targets were also significantly enriched in human BL samples harboring somatic TFAP4 mutations (Figure 4D-E). These findings suggest that TFAP4 functions as a negative feedback factor that restricts proliferative states of tumorigenic cells associated with high c-MYC and E2F activity. Unsupervised clustering of DEGs in pre-B–cell tumors with distinct Tfap4 genotypes revealed 2 clusters specifically associated with reduced Tfap4 expression (Figure 4F, clusters II/III). These clusters were enriched for genes associated with cancer/p53-related pathways (supplemental Figure 4H). In addition, a gene signature associated with an upregulated KRAS signaling pathway, which has been associated with the Eμ-Myc model,^38-40 was enriched specifically in Eμ-Myc Tfap4^+/− tumors (Figure 4F, cluster VIII; supplemental Figure 4H). Consistently, whole-exome sequencing and targeted sequencing of the tumor samples, which were mostly clonal or oligoclonal (supplemental Figure 5A), revealed frequent Kras mutations in Eμ-Myc Tfap4^+/− tumors, despite similar overall mutational burdens (supplemental Figure 5B-D). Overall, transcriptomic and genetic data suggest that TFAP4-dependent tumor suppression is mediated by the regulation of persistent engagement of cell proliferation via specific gene pathways rather than protection against global genomic instability.

TFAP4 supports the continued proliferation of normal lymphocytes by supplementing an early decay of c-MYC by maintaining the expression of numerous c-MYC target genes, but it is dispensable when c-MYC is expressed.^16,17 However, during tumorigenesis, our data showed that TFAP4 counteracts c-MYC–mediated transformation of developing B cells, likely through control of its unique targets. Among the unique TFAP4 targets in Figure 4B, Erg, encoding an ETS-family transcription factor, has been implicated in hematopoietic malignancies. Erg is also necessary to maintain the stemness of hematopoietic stem cells (HSCs).⁴¹ ERG overexpression in mouse hematopoietic progenitor cells causes lymphoid and megakaryocytic cancers.^42-45 In humans, although impaired ERG function has been associated with a small subset of B-ALL specifically harboring a DUX4 rearrangement,^45,46ERG was overexpressed in a majority of B-ALL cases compared with other tumor types and normal BM cells across multiple studies^47-51 (supplemental Figure 6A-E). Despite the limited sample number with gene expression data available, a human T-ALL case with a TFAP4^E57K substitution and an LMO2-LYL1 translocation in PeCan expressed ERG at a level higher than the 95% confidence interval of ERG expression in other T-ALL samples with an LMO2-LYL1 or LMO1-LMO2 translocation (supplemental Figure 6F-G), suggesting that elevated ERG is associated with functionally defective TFAP4.

In Eμ-Myc Tfap4^+/− pretransformed pro/pre-B cells, Erg expression was elevated by 3.8-fold compared with controls (Figure 5A). Erg was further upregulated in pro/pre-B tumors lacking TFAP4 (Figure 5B). TFAP4 bound an intronic region of the Erg locus, colocalizing with H3K27ac-marked accessible chromatin in Eμ-Myc B cells, whereas c-MYC binding was not detected in the genomic region encompassing the Erg locus (Figure 5C). These results suggest that Erg is a direct TFAP4 target. Erg is coexpressed with Myc and Tfap4 in pro-B cells before the Igh checkpoint at the population level but is downregulated in proliferating large pre-B cells as they differentiate into small pre-B cells (Figure 5D). However, within c-kit⁺ pro-B cells, expression of c-MYC protein correlates inversely with that of Erg mRNA, with Erg expression enriched in c-MYC⁻ pro-B cells and significantly downregulated as cells upregulate c-MYC during the transition to large pre-B cells (Figure 5E). Therefore, we hypothesized that Erg must be regulated in c-MYC⁺ B cells in a TFAP4-dependent manner to restrict the risk of tumorigenesis, because c-MYC and ERG are required for B-cell development but are also oncogenic. Accordingly, Erg was elevated in Tfap4^−/− vs Tfap4^+/+ c-MYC⁺ pro-B cells (Figure 5F). Furthermore, TFAP4-mediated Erg suppression was even more apparent in Eμ-Myc mice, because Erg was completely suppressed in Eμ-Myc vs WT B220⁺ IgM⁻ cells (Figure 5G). Erg suppression was partially reversed and completely lost in Eμ-Myc Tfap4^+/− and Eμ-Myc Tfap4^−/− pretransformed pro/pre-B cells, respectively (Figure 5G). These results indicate that TFAP4 prevents simultaneous expression of the 2 proto-oncogenes, c-MYC and ERG, in differentiating B cells.

To directly determine whether Erg derepression accelerates tumorigenesis, we deleted one Erg allele in Eμ-Myc Tfap4^+/− mice, because complete loss of Erg arrests B-cell development.⁵² Erg haploinsufficiency significantly delayed tumor onset and extended the survival of Eμ-Myc Tfap4^+/− mice (Figure 5H). Furthermore, B-cell tumors that developed in Erg^+/− mice were skewed toward the mature phenotype lacking expression of Rag1 and Vpreb1 (Figure 5I), indicating prolonged latency to transformation after MYC was turned on. These results indicate that TFAP4 suppresses the MYC-mediated transformation of B-cell precursors, at least in part, through the regulation of Erg.

ERG maintains the undifferentiated state of HSCs by inhibiting c-MYC targets, whereas c-MYC antagonizes self-renewal in favor of differentiation.^41,53 Because our data demonstrated that c-MYC suppresses Erg through TFAP4, we hypothesized that TFAP4 prevents B-cell progenitors from maintaining their undifferentiated state or stemness while rapidly proliferating, thus restricting transformation. To test this, we adoptively transferred a mixture of Tfap4^+/+ and Tfap4^+/− CD19⁺ cells transduced with c-MYC marked by distinct reporters, with comparable infection efficiency across genotypes, into sublethally irradiated congenic mice and assessed their expansion and differentiation (Figure 6A; supplemental Figure 6H). Three weeks after transfer, Tfap4^+/− cells became dominant among the total transduced donor-derived B220⁺ cells (Figure 6B-C). In addition to superior expansion and persistence of MYC⁺Tfap4^+/− cells, frequencies of IgM⁺ cells were significantly lower in Tfap4^+/− cells compared with Tfap4^+/+ cells (Figure 6D). These results imply that Tfap4^+/− pro/pre-B cells maintain their undifferentiated state during c-MYC–driven proliferation as a result of high Erg expression. Indeed, reducing Erg expression to 50% in Tfap4^+/− pro/pre-B cells diminished the expansion of overall donor-derived cells, primarily as a result of the reduced expansion of Tfap4^+/− cells relative to Tfap4^+/+ cells. Reduction of Erg in Tfap4^+/+ cells did not have any additional effect on the expansion of transferred cells (Figure 6C,E). Also, reducing Erg expression rescued the differentiation of transferred Tfap4^+/− MYC⁺ pro/pre-B cells, as determined by the frequencies of IgM⁺ cells (Figure 6F).

So far, our data have demonstrated that TFAP4 is mutated in human B-cell malignancies, and reduced TFAP4 functionality predisposes proliferating B cells to transformation in preclinical models. To gain insights into the role of TFAP4 in c-MYC–driven B-cell malignancies in humans, we reanalyzed the expression of c-MYC and TFAP4 in human pediatric B-ALL samples from the TARGET Phase II study (Figure 7A). Comparing patients with the top 25% and bottom 25% TFAP4/MYC ratios did not reveal a difference in the overall survival (data not shown). However, when we specifically compared samples with high MYC expression (z-score > 2), overall survival of patients with high TFAP4 expression was better than that of patients with low TFAP4 expression (Figure 7B-C). There was no difference in relative ERG expression between the 2 groups, possibly reflecting the selection of ERG-high clones during disease progression. Alternatively, the result also suggests an additional ERG-independent control of ALL progression by TFAP4 or ERG regulation by other oncogenic pathways (Figure 7D). These results suggest that upregulation of TFAP4 by c-MYC is a decisive factor for predisposition to B-cell malignancies, as well as is associated with the prognosis for patients with B-ALL.

Our study provides compelling evidence that c-MYC engages a tumor-suppressive program in B cells by inducing its direct downstream factor TFAP4. Coding mutations of TFAP4 prevail in lymphoid malignancies, particularly in a significant proportion of cases of BL. Our mouse models reveal substantial impacts on tumor onset by deletion or reduced expression of Tfap4 in developing B cells and mature B cells in the face of aberrant Myc expression. The tumor suppression is mediated, at least in part, by restricting the stemness factor ERG in immature B cells, because the reduced expression of Erg was sufficient to rescue developing B cells from aberrant expansion and accelerated transformation.

During hematopoiesis, HSCs give rise to various mature subsets through defined pathways. Because several proto-oncogenes are essential for normal hematopoiesis, their expression levels and timing must be tightly regulated to restrict the risk of leukemogenesis. Nonetheless, clonal hematopoiesis carrying oncogenic translocations is found in healthy individuals at a higher incidence than that of corresponding leukemias, including the t(8;14) translocation modeled by Eμ-Myc mice.^54-56 Therefore, appropriate engagement of tumor suppression is essential to prevent these clones from transformation. HSCs and undifferentiated progenitors generate greater numbers of mature progenies through proliferation that is tightly linked to differentiation and at the cost of their self-renewing potential. Thus, perturbation of normal differentiation could result in a decoupling of the loss of self-renewal capacity and proliferation, which has been associated with hematopoietic malignancies.^57-63 In HSCs, the antagonistic interplay between c-MYC and ERG is essential for balancing self-renewal and differentiation.⁴¹ ERG is essential for self-renewal of undifferentiated HSCs through suppression of c-MYC target genes; accordingly, loss of ERG closely recapitulates c-MYC overexpression in HSCs, resulting in depletion of undifferentiated HSCs.^41,53 We demonstrate that the balance between MYC and ERG controls self-renewal and differentiation of B-cell progenitors. Insufficient TFAP4 induction and subsequent Erg persistence selectively impair the differentiation program by c-MYC, allowing c-MYC–expressing cells to continue proliferation with an undifferentiated phenotype and eventually form tumors. Although a subset of B-ALL is characterized by reduced ERG activity that is likely caused by an independent mechanism, elevated ERG was noted in a majority of cases of B-ALL that also express elevated c-MYC.

The frequency of TFAP4 mutations is rare in human B-ALL. However, within MYC-high B-ALL cases, the level of TFAP4 expression is associated with patient prognosis. Because increased TFAP4 expression is associated with poor prognosis in some cancers,^64-66 lymphocytes may have established unique gene expression circuitry that converts the c-MYC–TFAP4 axis from a proproliferative/tumorigenesis module into a tumor-suppressive module by restricting their distinct targets. Similar counterregulation of MYC-driven proliferation may also be involved in the suppression of lymphomagenesis in GCB cells. GCB cells retain high proliferative potential similar to pro/pre-B cells, and TFAP4 may analogously function as a tumor suppressor by tightly coupling the expression of MYC and the restriction of continued proliferation. Thus, the induction of TFAP4 by MYC impacts the predisposition of B-cell malignancies, as well as likely contributes to disease progression. A therapeutic benefit for B-cell malignancies marked by high MYC expression and low TFAP4 could be achieved by inhibiting SCF ubiquitin ligase, leading to general cell cycle inhibition and specific stabilization of TFAP4, thus increasing the level of TFAP4 and balancing MYC and TFAP4 levels.^67,68 We uncover that perturbation of the c-MYC–initiated and TFAP4-mediated feedback mechanism dramatically accelerates tumorigenesis.

The authors thank Vivek Durai, Sunnie Hsiung, Tenzin Yangdon, Jason Walker, and Mike White for technical support; Barry Sleckman for MYC-GFP mice; Kenneth Murphy, Eugene Oltz, Jacqueline Payton, and Dan Littman for discussions and critical reading of the manuscript; the Maxim Artyomov laboratory for the development of the RNA sequencing analysis tool Phantasus; and the Genome Technology Access Center (Department of Genetics, Washington University School of Medicine) for help with genomic analyses.

This work was supported by National Institutes of Health (NIH), National Institute of Allergy and Infectious Diseases grants R56 AI114593-01A1 (T.E.) and R01 AI130152-01A1 (T.E.), and NIH, National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) grant R01 AR046000 (M.I.); Leukemia and Lymphoma Society Scholar Award 1349-18 (T.E.); a Siteman Investment Program Research Development Award (T.E.); and a Hu and Zeng Predoctoral Scholarship (E.T.). The Washington University Rheumatic Diseases Research Resource-based Center is supported by NIH, NIAMS grant P30 AR073752. The Genome Technology Access Center is supported, in part, by NIH, National Cancer Institute Cancer Center Support Grant P30CA91842 to the Siteman Cancer Center, and by the Institute of Clinical and Translational Sciences/Clinical and Translational Science Award (UL1TR000448) from the National Center for Research Resources, a component of the NIH, and the NIH Roadmap for Medical Research.

Contribution: T.E. conceived the project; E.T., C.C., and T.E. designed experiments and interpreted the results; E.T., Y.T., C.C., Y.X., M.H., C.F., S.R., and T.E. performed experiments; G.S.C. helped with the analysis of whole-exome sequencing data; M.I. provided Erg-flox mice; and E.T. and T.E. wrote the manuscript with editorial comments from all of the other authors.

Conflict-of-interest disclosure: T.E. and C.C. have filed a provisional US patent application to propose the use of stabilized TFAP4 to improve T-cell function (Application Number 16343951 “AP4 and Methods of Promoting T Cell Activation”). The remaining authors declare no competing financial interests.

Correspondence: Takeshi Egawa, Washington University in St Louis, 660 South Euclid Ave, Campus Box 8118, St. Louis, MO 63110; e-mail: egawat@wustl.edu.