MAPK/ERK2 phosphorylates ERG at serine 283 in leukemic cells and promotes stem cell signatures and cell proliferation

Immunoprecipitation

Whole-cell extracts from >3 × 10⁷ MOLT4 cells were prepared using lysis buffer with the following contents: 20 mM HEPES-NaOH (pH 7.8), 300 mM NaCl, 1 mM EDTA, 20% glycerol, 0.5% NP-40, protease inhibitor cocktails (Roche, Basel, Switzerland), 10 mM NaF and 1 mM Na₃VO₄ (1 ml lysis buffer per 1 × 10⁷ cells). Cleared lysate was incubated overnight with anti-ERG antibody (sc-354x, Santa Cruz, Dallas, TX, USA) at 4 °C (1 μl antibody per 1 × 10⁷ cells) and adsorbed onto protein G Dynabeads (Invitrogen, Carlsbad, CA, USA) for 3 h at 4 °C (20 μl Dynabead suspension per 1 × 10⁷ cells) the next day. The beads were washed with low detergent lysis buffer (20 mM HEPES-NaOH (pH 7.8), 300 mM NaCl, 1 mM EDTA, 20% glycerol and 0.1% NP-40) and the precipitated proteins were subjected to gel electrophoresis or in-solution digestion.

Enzymatic digestion and phosphoenrichment

Immunoprecipitated ERG was digested with chymotrypsin (Roche)/trypsin (Promega, Madison, WI, USA) and enriched for phosphopeptides for phosphosite identification (see Supplementary Methods).

Mass spectrometry

Peptide mixtures were separated and analyzed using the Fourier Transform mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). Mascot software (version 2.3.2; Matrix Science, London, UK) with the Swissprot database search algorithm was used for protein identification. Search parameters included three missed cleavages from trypsin proteolysis. Modifications selected were set as fixed carbamidomethylation of cysteins, variable oxidation of methionine and phosphorylation of serine, threonine and tyrosine residues. All peptides with high-quality tandem mass spectra were manually interpreted and confirmed. The sequences and phosphorylation sites of all identified phosphopeptides were carefully assigned accordingly.

Generation of ERG-pS283 specific antiserum

Mouse monoclonal phospho-specific antibody for ERG pS283 was produced by ProMab Biotechnologies, Inc. (Richmond, CA, USA) using the artificially synthesized phosphopeptide PTPQSKAAQP(p)*S*PSTVPKTEDQ, S283 highlighted by asterisk. The specificity of this antiserum for phosphorylated ERG is shown in Supplementary Figure S1.

Viral vectors and transduction

The open reading frame of human ERG3 (NM_182918.3, corresponding to transcript variant 1, mRNA) was subcloned into the pMIG+ retroviral vector that encodes the green fluorescence protein (GFP) downstream of an internal ribosomal entry site.²² Phoenix A cells were transiently transfected to generate virus-containing medium. Five rounds of retroviral transduction were performed over 3 days as previously described.¹³ See Supplementary Methods for details.

Site-directed mutagenesis

Various ERG constructs with point mutations were generated using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s instructions. The pMIG⁺ ERG plasmid was used as a backbone vector to generate mutant ERG constructs with alanine or aspartate mutations at S283. The mutation was subcloned into pMIG+ HA-ERG for tagged constructs. The mutagenesis primer sets are listed in Supplementary Table S1.

Cytokine-driven expansion culture

Transduced cells were expanded in cytokine-driven culture by plating 4 × 10⁴ viable cells per ml in Iscove’s modified Dulbecco’s medium supplemented with 10% low endotoxin fetal bovine serum (16000-044, Thermo Fisher Scientific), 2 mM L-glutamine, 50 μg/ml gentamycin, human stem cell factor (50 ng/ml), thrombopoietin (10 ng/ml), Flt3 ligand (50 ng/ml) and interleukin-6 (10 ng/ml). Cells were passaged weekly, and either returned to cytokine-driven culture or used for colony-forming unit assays. See Supplementary Methods for CD34+ cell enrichment and transduction.

Colony-forming unit assay

Clonogenic progenitors were quantified in triplicate cultures of 1% methylcellulose supplemented with 20 ng/ml stem cell factor, 20 ng/ml interleukin-3, 20 ng/ml interleukin-6, 6 μg/ml erythropoietin and 20 ng/ml granulocyte-colony stimulating factor.²³ The optimal seeding density of HSPCs was determined empirically to yield 20–200 colonies per 35 mm plate. KG1 and ME1 were plated at 3000 cells/plate and 25 000 cells per plate, respectively. Cultures were incubated at 37 °C, 5% CO₂ for 14–16 days and then total and GFP+ colonies were counted.

RNA sequencing

Single-stranded TruSeq complementary DNA libraries were generated and sequenced using the Illumina HiSeq2000 analyzer (BGI, Hong Kong). The sequencing reads were aligned to the human genome (hg19) using the software package TopHat.²⁴ Transcript expression was quantified with the software suite HOMER²⁵ using standard parameters and DeSeq²⁶ to identify differentially expressed genes. Data have been deposited in Gene Expression Omnibus as GSE53665.

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (v6.05, La Jolla, CA, USA), SAS software (v9.4, Cary, NC, USA) and R software (v3.0.2, R Foundation for Statistical Computing, Vienna, Austria). A P-value of <0.05 was considered significant. See Supplementary Methods for details.

Additional methods

For detailed descriptions of the protein degradation assay, immunostaining, fluorescence polarization assay, chromatin immunoprecipitation, quantitative real-time PCR, flow cytometry and fluorescence-activated cell sorting and transactivation assay, see the previous studies^6,13,27–30 and the Supplementary Methods.

ERG S283 is phosphorylated in human leukemic cells

To evaluate the role of phosphorylation on ERG activity, we first set out to identify sites of phosphorylation in endogenous ERG in MOLT4 T-ALL cells. To this end, we applied liquid chromatography coupled tandem mass spectrometry on immunoprecipitated ERG and identified five phosphopeptides corresponding to five unique phosphorylation sites: S55, S88, S103, S222 and S283 (Figure 1a and Supplementary Figure S2). The identity of ERG was confirmed by liquid chromatography coupled tandem mass spectrometry with over 60% sequence coverage from the in-gel trypsin digested sample and 20% sequence coverage from the in-gel chymotrypsin digested sample (Supplementary Figure S3). Apart from the detection of S55 that required chymotrypsin digestion, the other four serine residues were identified by trypsin digestion. Tandem mass spectra were manually curated to confirm phosphorylation (Figure 1b and Supplementary Figure S4).

To directly compare ERG phosphorylation in different leukemic cell types, we used a SILAC (stable isotope labeling of amino acids in cell culture)-based approach.³¹ MOLT4 cells cultured in media containing L-Arginine.HCl (¹³C₆, 99%) and L-Lysine.2HCl (¹³C₆, 99%) for six consecutive passages were ‘heavy’ labeled and their lysates were combined with lysates from KG1 AML cells cultured in normal media (‘light’). Phosphoproteomic analysis was also performed using primary mobilized CD34+ human HSPCs. ERG S88 was robustly phosphorylated in all three cell types, but ERG S283 phosphorylation (pS283), although abundant in T-ALL and AML cells, was minimal or undetectable in primary HSPCs (Figure 1c and Supplementary Figure S5). ERG phosphorylation at S103 and S222 was at or below the threshold for detection in KG1 and primary HSPCs.

We also extracted phosphoproteomic data from publicly available data sets to evaluate ERG phosphorylation in normal tissues (Figure 1d).^18–20 Interrogation of phosphorylation at S55 requires chymotrypsin digestion and was not always performed in these studies. Following the curation of these data sets and comparison with ours, a striking feature was the abundance of ERG pS283 in leukemic cells in contrast with either low or undetectable phosphorylation in normal human HSPCs and normal mouse tissues or in prostate cancer cells expressing the TMPRSS2-ERG fusion protein (Figure 1d).^18–20

To investigate whether ERG pS283 was also present in primary leukemic cells, we generated a monoclonal antibody that specifically recognized this modification but not the unmodified protein (see Methods; Supplementary Figure S1) and screened a panel of AML and T-ALL patient-derived xenografts.^32–35 Most T-ALL and AMLs had robust ERG S283 phosphorylation (Figure 1e). The ETP (early T-cell precursor) ALLs in particular had high ERG S283 phosphorylation that was minimal in healthy CD34+ HSPCs despite high total ERG expression (Figure 1e).

Overexpression of an ERG S283 phosphomimetic promotes maintenance and expansion of CD34+ HSPCs in cytokine-driven cultures

Following the identification of ERG pS283 and its relative abundance in leukemic cells, we set out to identify the biological impact of this modification on hematopoiesis. ERG overexpression in CD34+ cord blood (CB) cells results in the expansion of the progenitor pool, and a gene expression signature that shares features with normal HSCs, high ERG AMLs, ETP ALLs and leukemic stem cell signatures associated with poor clinical outcome.¹³ Given the frequency and relative abundance of pS283 in leukemic cell lines and in primary leukemic cells in contrast with CD34+ HSPCs, we investigated the impact of overexpressing phosphomimetic and phosphodefective S283 mutants of ERG in CD34+ CB cells in cytokine-driven cultures (Figure 2 and Table 1). CB cells were transduced with retroviral vectors expressing either the empty vector (pMIG+), wild-type (WT) ERG, ERG S283A (phosphodefective mutant) or ERG S283D (phosphomimetic mutant). Transduced cells were cultured in media supplemented with stem cell factor, thrombopoietin, Flt3 ligand and interleukin-6 with weekly passaging and harvesting for clonogenic assays, flow cytometry and molecular analysis (Supplementary Figure S6). Quantitative real-time PCR analysis of CB cells from three independent transductions confirmed that ERG expression levels in all three high ERG conditions were substantially elevated in comparison with the pMIG+ cells at each time point (Supplementary Figure S7). Transduced cells were identified by the expression of GFP. The proportion of CB cells transduced with ERG WT and mutant ERG progressively increased in culture in contrast with those transduced with the empty vector (pMIG+) that progressively decreased (Figure 2a). These data were consistent with previous observations that ERG overexpression conferred HSPCs with a competitive advantage in cytokine-driven cultures.¹³

Overexpression of the ERG S283D mutant was associated with a significant increase in the expansion of the transduced primitive population (GFP+ CD34+) (Figure 2b and Table 1). Although there were inherent differences between CB samples, the cumulative expansion of CD34+ progenitors in ERG S283D-transduced cells was greater than the expansion of CB cells transduced with ERG WT in three independent experiments (P<0.001, Table 1). As expected, transduced ERG WT was not phosphorylated at S283 in GFP+CD34+ cells (Supplementary Figure S8). Along with the expansion of CD34+ cells, overexpression of ERG S283D in CB cells also resulted in a cumulative increase in the numbers of CD34− cells consistent with the absence of a gross defect in progenitor differentiation (Figure 2c and Table 1, P<0.0001). Clonogenic assays are indicative of functional hematopoietic progenitors. Consistent with the increased numbers of CD34+ progenitors, there was an increase in the numbers of colony-forming units generated by cells transduced with ERG S283D compared with those transduced by ERG WT and ERG S283A that in turn was significantly greater than cells transduced with the pMIG+ vector (Figure 2d and Table 1, P <0.01).

To characterize the impact of high ERG S283D expression on the transcriptome of CD34+ HSPCs, we performed RNA sequencing on cell fractions sorted at week 0 following five rounds of transduction over 3 days. RNA sequencing was performed using two independent CB transductions (experiments 2 and 3). Gene set enrichment analysis showed that genes that were high in HSCs (relative to granulocyte–macrophage progenitors) were enriched in ERG 283D-transduced CB cells (Figure 2ei). The ERG +85 SCE is active in leukemic cells and we have previously shown that a signature derived from the activity of this enhancer is an independent risk factor for poor overall survival in AML.²⁷ ERG S283D-transduced CB cells were also enriched for genes in the +85 SCE (Figure 2eii) as well as gene sets associated with oncogenic processes (Supplementary Figure S9). Another notable observation was the enrichment of genes associated with RAS (Figure 2fi) and MEK (Figure 2fii) signaling in ERG S283D-transduced cells compared with ERG WT-transduced cells.

ERG S283 phosphorylation is mediated by the MAPK/ERK2 pathway

Given the abundant phosphorylation of ERG S283 in leukemic cells and the observation that ERG S283 phosphomimetic promotes a stem cell-like gene expression signature and expansion of CD34+ progenitors, we set out to identify the upstream pathways that mediate this modification in leukemic cells.

Human ERG3, the major hematopoietic isoform, contains a putative MAPK docking site (FxFP) encoded by exon 12 between the PNT and ETS domains (Figure 3a).^21,36 This motif is also known as a DEF domain and facilitates ERK1/2 docking in preference to c-Jun N-terminal kinase or p38.³⁷ Given this and the observation that MEK/ERK activity is reinforced by ERG S283 phosphomimetic in CD34+ CB cells, we considered the possibility of a recursive loop and evaluated the ability of ERK1 and ERK2 to phosphorylate ERG S283 in in vitro kinase reactions using purified WT ERG followed by mass spectrometry (Supplementary Figure S10). ERK2 but not ERK1 successfully phosphorylated ERG S283. These reactions were repeated using purified WT ERG or an exon 12 deletion mutant of ERG (ERG-ΔEx12). Consistent with loss of its binding domain, the percent phosphorylation on S283 was significantly reduced in ERG-ΔEx12 (Figure 3b). Myelin basic protein (1.5 nmol per reaction) was included in each reaction to serve as an internal indicator of kinase efficiency (Supplementary Figure S11). ERK2 did not phosphorylate ERG S88, suggesting that the abundant phosphorylation of this residue in hematopoietic cells is likely mediated by another kinase (Figure 3b). S222, on the other hand, can be phosphorylated by ERK2 even in the absence of the docking domain on exon 12 (Figure 3b).

Phorbol 12-myristate 13-acetate (PMA), a direct protein kinase C activator, also activates MEK/ERK and increases phosphorylation on ERK substrates.³⁸ PMA also mediates phosphorylation of non-ERK substrates, but unlike the phosphorylation of ERK substrates, the former cannot be attenuated by pretreatment with ERK inhibitors. To investigate the in vivo role of ERK on ERG S283 phosphorylation, serum-starved MOLT4 cells were treated with an MEK/ERK inhibitor (U0126) and/or PMA (Figure 3c). The experiment was performed using a SILAC-based method, where cells grown in normal media were treated with U0126 ± PMA and phosphorylation levels analyzed by mass spectrometry simultaneously with lysates of vehicle control-treated cells that were grown in heavy media. MEK/ERK inhibition (U0126) reduced S283 phosphorylation in MOLT4 cells (Figure 3c). PMA treatment increased S283 phosphorylation but pretreatment with U0126 attenuated this increase. These data were consistent with pS283 in MOLT4 cells being predominantly an ERK-mediated modification. On the other hand, S222 phosphorylation was not exclusively an ERK-mediated modification in that although U0126 reduced the degree of phosphorylation at this residue, it did not attenuate PMA-mediated phosphorylation (Figure 3c). Consistent with the in vitro assays in Figure 3b, S88 phosphorylation was largely independent of ERK (Figure 3c).

To evaluate whether ERG pS283 in primary leukemic cells was also mediated by MEK/ERK, we used two primary ETP ALL xenografts with high baseline pS283. Variations in ERG pS283 levels were determined using the custom anti-ERG pS283 antibody (Figure 3d). As with MOLT4 cells, U0126 reduced baseline pS283 levels in ETP ALL cells and the PMA-mediated increase in phosphorylation was attenuated by U0126 (Figure 3d). The variations in pERK levels with U0126 ± PMA serve as internal controls for target kinase inhibition and activation (Figure 3d). Taken together, these data were supportive of ERG pS283 in primary leukemic cells also being mediated by MEK/ERK.

To evaluate the relationship between pERK2 levels and ERG pS283, we screened a panel of patient-derived ETP ALL, AML xenografts and healthy CD34+ HSPCs (Figure 3e and Supplementary Figure S12). With some exceptions (for example, ETP1), the levels of pERK2 and ERG pS283 in individual leukemic samples were concordant and consistent across a second set of spleens from these xenografts (Supplementary Figure S13). Healthy CD34+ HSPCs, which have minimal level of pS283, had no discernible pERK2 (Figure 3e). However, MEK/ERK inhibition per se was not cytotoxic to these ETP ALL xenografts with high ERG pS283 (Supplementary Figure S14). Given the enhanced proliferation of CD34+ CB cells transduced with ERG S283D, we hypothesized that inhibiting MAPK/ERK signaling would have a greater impact on the clonogenicity of leukemic cells with high ERG pS283. To this end, we performed clonogenic assays using KG1 and ME1 AML cells in the presence of a MEK inhibitor, U0126. Although MEK inhibition affected colony-forming unit formation in both healthy CD34+ and leukemic cells, the relative impact was significantly higher on leukemic cells (see Materials and methods; Figure 3f).

The ERG S283 phosphomimetic is enriched on leukemia-associated enhancers

To assess how pS283 affected ERG activity, we evaluated protein stability, nuclear localization and DNA-binding properties of ERG S283 phosphomutants and that of WT ERG. Protein stability was measured by transfecting HEK293T cells with WT and mutant ERG and performing western blots on cell lysates collected at various intervals following the inhibition of de novo protein synthesis with cycloheximide. The protein half-lives were comparable and calculated at 38.5, 38.5 and 36.5 h for ERG WT, ERG S283A and ERG S283D, respectively (Figure 4a and Supplementary Figure S15).

ERG is a transcription factor and to explore whether pS283 of ERG alters its nuclear localization, we transduced MOLT4 T-ALL cells with WT ERG or the S283 mutants carrying an hemagglutinin tag to discriminate recombinant ERG from endogenous protein. Immunofluorescence images of WT and variant ERG showed little difference with all forms able to translocate into the nuclei of MOLT4 cells (Figure 4bi). This was further supported by immunoblots of the nuclear and the cytoplasmic fractions of the transduced cells (Figure 4bii).

Given the efficient nuclear localization of all forms, we evaluated possible differences in in vitro and in vivo DNA binding of ERG WT and S283D mutant. Glutathione S-transferase-tagged full-length ERG was bound in vitro to a DNA segment (5′-AGGACCGGAAGTAACT-3′) containing the consensus ETS motif (5′-GGAA/T-3′) in a fluorescence polarization assay. Comparable binding affinity of ERG WT (K_D = 23 nM) and S283D (K_D = 22 nM) was observed (Figure 4c, n = 2). In vivo DNA binding was assessed by performing chromatin immunoprecipitation assays in MOLT4 cells transduced with hemagglutinin-tagged WT and S283 mutant ERG (Figure 4di). An ETS deletion mutant of ERG (ERG-ΔETS), which is incapable of binding DNA, was used as a control. Chromatin immunoprecipitation was performed with an anti-hemagglutinin antibody and enrichment of transduced ERG forms was evaluated by quantitative real-time PCR at regions of the ERG +85 and HHEX +1 SCEs at which we have previously shown ERG binding in primary ALL.^6,39 The activity of the +85 SCE is linked to both higher endogenous ERG expression and poor prognosis.^6,27 There was greater enrichment of the phosphomimetic ERG S283D mutant at these enhancers in vivo compared with ERG WT or the phosphodefective S283A mutant (Figure 4di). To investigate the impact of differential in vivo binding of these transduced ERG forms at the ERG +85 enhancer on endogenous ERG expression, we measured endogenous transcripts using allele specific quantitative PCR (see Materials and methods). Occupancy of the ERG + 85 enhancer by S283D resulted in significantly higher levels of endogenous ERG transcripts than either WT ERG or S283A (Figure 4dii). A similar trend was observed in relation to binding of transduced ERG forms at HHEX+1 and levels of endogenous HHEX transcripts.

To evaluate the impact of MEK inhibition on ERG enrichment at stem cell enhancers in leukemic cells,²⁸ we treated KG-1 AML cells with either U0126 or dimethyl sulfoxide for 30 min, which was sufficient to ablate pERK2 and reduce ERG pS283 levels (Supplementary Figure S16), and performed quantitative real-time PCR–chromatin immunoprecipitation. There was significant reduction of endogenous ERG at the ERG+85, HHEX+1 and LMO2-23 stem cell enhancers (Figure 4ei) and concomitant reductions in endogeneous gene expression (Figure 4eii).

Taken together, our results suggest a model in which cooperative interaction between RAS/MAPK signaling and pS283 ERG contributes to the malignant phenotype of high ERG leukemias (Figure 5).