Polycomb-like Protein 3 Induces Proliferation and Drug Resistance in Multiple Myeloma and Is Regulated by miRNA-15a

Multiple myeloma is the second most prevalent hematologic malignancy in the world (1). The outcomes of newly diagnosed patients with myeloma have greatly improved in recent decades because the introduction of novel agents (such as proteasome inhibitors and immunomodulatory drugs) administered in combination with autologous stem cell transplantation (2). Nevertheless, many patients continue to relapse; and after prolonged salvage treatment, the disease becomes multidrug resistant and leads to refractory multiple myeloma and eventual death (3, 4).

Multiple myeloma is a heterogeneous disease, and treatment benefits are not uniform, even in patients that harbor similar cytogenetic abnormalities. Accumulating evidence indicates that dysfunctional epigenetic changes—including in DNA methylation, histone modification, and noncoding RNA (such as miRNA) dysregulation—are also involved in the development and drug resistance of myeloma (4–8).

Polycomb-like protein 3 (PCL3) also called PHF19 is a polycomb-like (PCL) protein. PCL proteins are polycomb repressive complex 2 (PRC2)-associated factors that form subcomplexes with PRC2 core components (9) and modulate the enzymatic activity of PRC2. PCL proteins are important components of the PRC2 complex, are major H3K27 methyltransferases, and are required for recruiting the PRC2 complex to target genes; thus, they play important regulatory roles in PcG protein–mediated transcriptional repression (10). In normal tissues, PHF19 serves a role in transcriptional regulation of gene expression in embryonic stem cell renewal and differentiation (11–13). Recently, PHF19 was found to be highly expressed in a variety of cancers, including melanoma, hepatocellular carcinoma, glioblastoma, ovarian carcinoma, and multiple myeloma (14–18), and the expression of PHF19 has been found to be correlated with overall tumor progression (19). The role(s) of PHF19 in tumor (including multiple myeloma) pathogenesis and drug resistance, however, needs to be further elucidated.

Here we demonstrated that PHF19 was highly expressed in multiple myeloma, especially in relapsed or refractory myeloma cells, and that was associated with inferior outcomes among patients with multiple myeloma. PHF19 overexpression promotes multiple myeloma cell growth and drug resistance. Knocking down PHF19 expression with use of short hairpin RNA (shRNA) increased the sensitivity of multiple myeloma cells to common chemotherapeutic reagents and enhanced the apoptosis of the multiple myeloma cells. Our study also documented the pivotal role of PHF19 in promoting phosphorylation-mediated inactivation of EZH2, a core protein of the PRC2 complex, by activating the AKT pathway. PHF19 thus engendered a repressed H3K27me3 level and consequently upregulated expression of several prosurvival targets involved in multiple myeloma cell growth and drug resistance. In addition, we also showed that PHF19 was a direct target of miR-15a, loss of miR-15a led to overexpression of PHF19 in multiple myeloma cells. These findings suggest that the miR-15a/PHF19/pho-EZH2 axis acts as a key mediator of multiple myeloma proliferation and drug resistance.

We determined levels of PHF19 mRNA in patients with myeloma, with use of Affymetrix U133 Plus 2.0 microarrays, as described previously (20). Results are available in the NIH Gene Expression Omnibus (GEO) under accession numbers GSE2658 and GSE31161. Microarray data on monoclonal gammopathy of undetermined significance (MGUS) and normal plasma cells are available under accession number GSE5900. Statistical analyses of microarray results relied on GCOS1.1 software (Affymetrix), and included log-rank tests for univariate association with disease-related survival.

All human multiple myeloma cell lines were grown under Mycoplasma-free conditions and maintained in complete culture medium (RPMI1640 supplemented with 10% FBS and 100 IU/mL penicillin) in tissue culture flasks, at 37°C in a 5% CO₂ humidified incubator. HEK293T and MCF-7 cells were cultured in DMEM with 10% FBS and 100 IU/mL penicillin. All the cells were maintained in our lab and shown to be Mycoplasma-free via PCR; they were used for experiments within 8 passages after thawing. Identities of the cell lines were confirmed by short tandem repeat testing.

Bone marrow mononuclear cells (BMMC) were obtained from healthy donors and patients with multiple myeloma who were newly diagnosed at the Department of Lymphoma and Myeloma, Institute of Hematology and Blood Disease Hospital, Chinese Academy of Medical Sciences, Peking Union Medical College (Tianjin, China). BMMCs were isolated from BM via density gradient centrifugation using ficoll-hypaque (GE Healthcare). Multiple myeloma cells and normal plasma cells were purified from BM aspirates, using anti-CD138 microbeads (Miltenyi Biotec). All studies with human samples were done under the approval of the School Review Board, in accordance with the Declaration of Helsinki. Good clinical practice guidelines were followed, and written informed consents were obtained from healthy donors and patients with multiple myeloma to process research on their samples.

The following primary antibodies were used for Western blotting: anti-PHF19 (catalog no. 77271), anti-EZH2 (catalog no. 5246), anti-Mcl-1(catalog no. 94296), anti-Bcl-xL (catalog no. 2764), anti-phosphorylated PDK1 (S241; catalog no. 3061), anti-pan AKT, anti-phosphorylated AKT (Ser473), anti-phosphorylated AKT (T308), anti-β actin, anti-GAPDH from Cell Signaling Technology; anti-phospho-EZH2 (Ser21; catalog no. IHC-00388) from Bethyl Laboratories Inc.; and anti-HIF-1α (catalog no. 36169) from Abcam. Secondary antibodies, including horseradish peroxidase–conjugated anti-mouse and anti-rabbit antibodies were from Cell Signaling Technology.

The following drugs (their sources) were used in this study: bortezomib (Velcade, Janssen Pharmaceuticals), melphalan (L-PAM), epirubicin hydrochloride (Pharmorubicin; Pfizer); doxycycline and LY294002 were obtained from Sigma-Aldrich (MilliporeSigma) and dissolved in water or dimethyl sulfoxide, respectively, at appropriate concentrations. The final concentration of dimethyl sulfoxide in the culture medium was <0.1%, and did not affect drug effects and cell growth per se.

Primers for PHF19 cDNA clones (forward, 5′-GTG TCT AGA ATG CTG GTC TTG GTA ATC CGT GG-3′; reverse, 5′-GTG GGA TCC TCA GTA AGG GGT GGT CCC TTC CCA C-3′) were purchased from Beijing Genomics Institute (BGI, China). PHF19 shRNAs were: sh-1, 5′-CCT CAA GTC CTC TAT CAC CAA-3′; sh-2, 5′-GCG GCT GCC TCG TGA CTT TCG AAG ATA ATA GTG AAG CCA CAG ATG TAT TAT CTT CGA AAG TCA CGA GGC AGC TGC-3′; and sh-3, 5′-GCG GAA GGA CAT ACA GCA TGC CGG TGT TTA GTG AAG CCA CAG ATG TAA ACA CCG GCA TGC TGT ATG TCC TTC TGC-3′.

The full-length coding sequence for PHF19 was inserted into the pCDH-EF1-MCS-GFP vector. PHF19 was knocked down by shRNA were designed with use of an RNA interference platform provided by the Broad Institute. The oligomers were cloned into the pTRIPZ vector. The PMIRH15a-PA-1 vector was purchased from System Biosciences (SBI). Lentivirus was packaged in HEK293T cells, harvested after 48 hours, and concentrated 10-fold by the Lenti-X Concentrator (Takara Bio USA, Inc.). A total of 1 × 10⁶ cells were plated in a 12-well plate. Lentivirus was added to the cells along with 8 μg/mL polybrene, and stable cells were selected after 72 hours according to the manufacturer's instruction. The expression of target proteins was examined by Western blot analysis.

Whole multiple myeloma cells were lysed with RIPA lysis buffer; proteins (30 μg) were separated with SDS-PAGE, transferred onto a polyvinylidene fluoride membrane, blocked with 5% nonfat milk in Tris-buffered saline solution containing 0.1% Tween 20 (TBS-T) at room temperature for 1 hour, and then incubated overnight at 4°C with primary antibody. Protein bands were visualized with Super Signal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific). Co-immunoprecipitations (Co-IP) were carried out with the Pierce Direct Magnetic IP/Co-IP Kit (Thermo Fisher Scientific), using antibodies as above. Normal mAb IgG from Cell Signaling Technology was used as an isotype control.

Cell proliferation assays were carried out with use of a cell counting kit-8 (CCK8; Dojindo Laboratories) assay in 96-well plates and counted once a day for 5 days. Equal number of multiple myeloma cells was seeded in 96-well plates (5,000 cells/well). Cells were incubated in RPMI1640 medium at 37°C in a 5% CO₂ humidified atmosphere. After incubation, 10-μL CCK8 was added to a plate each day and plates were incubated at 37°C for another 2 to 4 hours. Finally, absorbance was determined at 450 nm using a microplate reader SpectraMax 3 (Molecular Devices). Apoptosis was quantified with use of the Annexin V: PE and Annexin V: FITC Apoptosis Detection Kits (BD Biosciences), following the manufacturer's protocol. Fluorescence was measured on an LSR II cytometer (BD Biosciences) and data analyzed using FlowJo Software v10.0 (Tree Star).

RNA was extracted using the Qiagen RNeasy Kit. Whole RNA (500 ng) was subjected to library preparation, with use of the BGI Library prep for Illumina kit. Paired-end reads were mapped to the human genome (hg19) using Tophat, with only unique mapped reads with fewer than two mismatches used for downstream analysis. Genes were assembled using Cufflinks. Normalized transcript abundance was computed using Cufflinks and expressed as FPKM (fragments per kilobase of transcripts per million mapped reads). Gene-level FPKM values were computed by summing the FPKM values of their corresponding transcripts. A gene is expressed if the FPKM is greater or equal to 1 in at least one sample. Edge R was adopted to calculate the P values of differential gene expression using the normalized gene-level read counts. For calling differentially expressed genes, we used the FDR to correct the P values, with a cutoff of 0.05 as significantly differentially expressed genes. To evaluate the biological functional relevance of genes, we employed the gene ontology enrichment analysis on groups of genes, using the hypergeometric test via the Gene Ontology. P values were corrected through Benjamini and Hochberg (BH) test with a cutoff value of 0.05.

A total of 1 × 10⁶ ARP1 cells expressing either normal (ARP1 EV) or high (ARP1 OE) levels of PHF19 were injected subcutaneously into the right flank of nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice (n = 4/group). After 7 days, mice were treated with bortezomib (1 mg/kg) or PBS, twice weekly for 3 weeks. On day 28, mice were executed to harvest and examine the tumors. Tumor volumes (measured by caliper) were calculated by the formula: length × width² × 0.52. And cells with PHF19-inducible knockdown of ARP1 (PHF19 shRNA) were injected under the skin of the abdominal flank of NOD/SCID mice (n = 4). Depletion of PHF19 expression was induced with doxycycline (2 mg/mL). All procedures involving use of live animals were approved by the Institutional Animal Care and Use Committee of the Institute of Hematology, Chinese Academy of Medical Sciences (Tianjin, China).

Total RNA yields were obtained with the RNeasy Mini Kit (QIAGEN) according to manufacturer's instructions, and first-strand cDNA synthesis reactions were performed with the Transcriptor First Strand cDNA Synthesis Kit (Roche). qRT-PCR reactions were performed in triplicate on an ABI 7500 Real-Time PCR System (Applied Biosystems). Data were analyzed according to the ΔΔC_t method. Primers were purchased from BGI.

To confirm that PHF19 is the target of miR-15a, we used a luciferase reporter assay. PHF19 3′UTR wild-type (wt) and mutated (mut) recombined with SV40-Luc-MCS plasmids were obtained from Genechem corporation in China. HEK293T cells were cultured and planted with 3 × 10⁴ cells per well in 96-well plates. After 24 hours, plasmids were transfected to these cells, with use of lipofectamine 2000 (Invitrogen). PHF19 3′UTR wt or mut plasmids (1,000 ng) were cotransfected with miR-15a or empty vector (EV). The phRL-CMV plasmids (Promega) that constitutively express Renilla luciferase were included in all transfections. Forty-eight hours after transfection, cells were analyzed by the Dual-Luciferase Reporter (DLR) Assay (Promega), and the ratio of Firefly luciferase activity to Renilla luciferase activity was calculated.

In accordance with PHF19 expression, survival curves were plotted using the Kaplan–Meier method. The two-tailed Student t test was used to compare two groups. One-way ANOVA was used to assess more than two groups. GraphPad Prism (version 6.01, GraphPad software Inc.) was utilized, and P ≤ 0.05 was considered statistically significant. Image J software was utilized for the statistical analysis of these protein expression levels.

We compared the expression of PHF19 through gene expression profiling (GEP) in plasma cells from healthy donors [normal plasma cells (NPC), n = 22], individuals with MGUS (n = 44), individuals with smoldering multiple myeloma (SMM, n = 12), and patients with newly diagnosed multiple myeloma (NDMM, n = 351) from the Total Therapy 2 (TT2) dataset. When an Affymetrix signal of 1,500 for PHF19 was set at as the cutoff, plasma cells in 96 of 351 (27.3%) patients with multiple myeloma exhibited an obviously high level of PHF19 expression compared with normal plasma cells; plasma cells from individuals in the MGUS and SMM groups did not show high PHF19 expression levels (Fig. 1A). Expression of PHF19 was significantly higher in relapsed and high-risk myeloma subgroups (Fig. 1B and C). In sum, these results strongly suggest that high levels of PHF19 are strongly correlated with an aggressive multiple myeloma phenotype.

Subsequently, we correlated PHF19 expression with multiple myeloma patient outcomes, determined by the P value and HR at the best expression signal cutoff, using the R package. The 21% (74/351) patients of multiple myeloma with elevated PHF19 expression had inferior event-free survival (EFS; Fig. 1D) and overall survival (OS; Fig. 1E) in the TT2 clinical trial. Similar results were obtained in an independent IFM cohort that included 119 patients with multiple myeloma (Supplementary Fig. S1A). These findings strongly suggest that PHF19 is a high-risk gene and that PHF19 expression positively correlates with drug resistance and poor prognosis in multiple myeloma.

The association between the level of PHF19 expression and the high risk and poor survival of patients with multiple myeloma described above strongly suggests that PHF19 is a gene related to myeloma progression. However, the biological roles of PHF19 in myeloma cell formation and development have not been fully characterized. To address this gap, we overexpressed full-length PHF19 in ARP1 and OCI-My5 multiple myeloma cell lines via lentivirus-mediated transfection and observed the alterations in cancer-related behavior in these two cell lines. Western blots verified that PHF19 was significantly increased in these cells (Fig. 2A). PHF19 over-expression cells (PHF19 OE) cells demonstrated significantly enhanced cell proliferation compared with those transfected with an EV (Fig. 2B). PHF19 OE also induced resistance to anti-multiple myeloma treatments. Twenty-four hours after treatment with bortezomib, epirubicin, or melphalan in designated concentration, flow cytometry revealed a decrease in multiple myeloma cell apoptosis in PHF19 OE groups compared with the EV groups (Fig. 2C and D).

Next, we examined the growth inhibitory effect of the drugs in PHF19 OE cells. OCI-My5 cells were cultured for 48 hours in the presence of a series of diluted concentrations of bortezomib, epirubicin, or melphalan. Enhanced PHF19 expression decreased cell sensitivity significantly after 48 hours in a dose-dependent manner, with an increased the half maximal inhibitory concentration (IC₅₀) in PHF19 OE cells (Supplementary Fig. S2A–S2C). These results suggest that ectopic expression of PHF19 promotes multiple drug resistance in multiple myeloma cells. We then tested the effects of PHF19 OE in a xenograft mouse model. ARP1 EV or PHF19 OE cells were injected into the right flank of NOD/SCID mice. After 28 days, the mice were euthanized to measure tumor burden. We found that mice engrafted with ARP1 PHF19 OE cells had significantly larger tumor volumes than the control mice which were engrafted with ARP1 EV cells (P = 0.0079; Fig. 2E and F). Both in vitro and in vivo data demonstrated that overexpression of PHF19 enhances the tumorigenic capacity of multiple myeloma.

To determine whether multiple myeloma cell survival relies on PHF19, we also knocked down PHF19 in ARP1 and OCI-My5 multiple myeloma cell lines, via a doxycycline-inducible shRNA lentivirus delivery system. Upon induction with 2 μg/mL doxycycline, PHF19 protein expression was remarkably downregulated (Fig. 3A). Cell proliferation assays indicated that, compared with controls, PHF19 depletion significantly suppressed cell growth in ARP1 (day 6, P < 0.001) as well as OCI-My5 (day 6, P < 0.001) cell lines (Fig. 3B). We next evaluated whether PHF19 knockdown promotes drug sensitivity in multiple myeloma cells. PHF19-knockdown cells were treated for 24 hours with bortezomib, epirubicin, or melphalan at indicated doses. As expected, the knockdown of PHF19 increased cell apoptosis in both ARP1 and OCI-My5 cell lines in the presence of doxycycline (Fig. 3C and D). We then used CCK8 assays to examine the growth inhibitory effect of the drugs in PHF19-knockdown cells. OCI-My5 cells were cultured for 48 hours in the presence of a series of diluted concentrations of bortezomib, epirubicin, or melphalan. The knockdown of PHF19 expression increased cell sensitivity significantly after 48 hours, in a dose-dependent manner, with a reduced IC₅₀ in PHF19-knockdown cells (Supplementary Fig. S2A–S2C). The in vivo study also showed that tumor volumes in the PHF19 shRNA knockdown group induced by doxycycline was significantly smaller than those in the control mice (P = 0.04; Fig. 3E and F). In sum, these results further support that PHF19 promotes myeloma growth and induces resistance to chemotherapy.

PHF19 is a Tudor domain–containing protein that plays an important role in recruiting the PRC2 complex to CpG islands (21). EZH2 is the enzymatic subunit of PRC2 that methylates lysine 27 of histone H3 (H3K27), to promote transcriptional silencing (10–12). Our co-IP analysis indicated that PHF19 binds directly to EZH2 in ARP1, KMS11, and HEK293T cell lines (Fig. 4A; Supplementary Fig. S3). Further study demonstrated that high levels of PHF19 suppressed bimethylation and trimethylation of H3K27 (Supplementary Fig. S4). Correspondingly, the knockdown of PHF19 by shRNA decreased EZH2 phosphorylation and restored H3K27 methylation (Fig. 4B and C; Supplementary Fig. S4). However, according to our RNA sequencing (RNA-seq) analysis, expression of three lysine demethylases (KDM7A, KDM6A, and KDM6B), which induces the demethylation of H3K27, was not elevated in PHF19 OE cells (data not shown). The level of EZH2 phosphorylation at Ser21 was enhanced significantly in PHF19 OE cells (Fig. 4B). When EZH2 is phosphorylated (Ser21), its transmethylase function is inactivated, causing H3K27 demethylation, which then upregulates the expression of downstream genes. Furthermore, high levels of PHF19 were associated with increased pho-Ser21 EZH2 in primary multiple myeloma cells (Supplementary Fig. S4). Because phosphorylation-mediated inactivation of EZH2 leads to drug resistance in multiple myeloma (22), these data strongly indicate that PHF19 directly affects the enzymatic activity of EZH2 methyltransferase and exerts a drug-resistance function through the phosphorylation of EZH2. We also found that the drug-resistance genes of multiple myeloma (Bcl-xL, Mcl-1, and HIF-1α) were significantly upregulated by PHF19 (Fig. 4D), while they were significantly downregulated in cells that were depleted of PHF19 (Fig. 4E).

To establish which pathways are involved in PHF19-mediated multiple myeloma cell survival and drug resistance, we performed RNA-seq on ARP1 cells with aberrant PHF19 expression. Distinct clusters of ARP1 PHF19 OE samples were distinguished from the EV samples through a principal component analysis (PCA; Supplementary Fig. S5A). The heatmap revealed that 1,426 genes were significantly differentially expressed between ARP1 PHF19 OE and EV cells (FDR < 0.05). Multiple myeloma pathogenesis genes (CCL3, FOS, JUN, KLF, and RELB) were significantly upregulated by PHF19 overexpression (Supplementary Fig. S5B). The UBB, UBA52, UBE2L6, and UBXN6 genes, which control unfolded protein degradation through ubiquitination and the proteasomes system, were also upregulated in PHF19 OE cells (Supplementary Fig. S5C). Gene set enrichment analysis (GSEA) showed that differential expression of genes in PHF19 OE cells was enriched in signal pathways, such as the NF-κB, hypoxia, and p53 pathways, which play pivotal roles in multiple myeloma cell proliferation. We also used RNA-seq to profile gene expression in shRNA-PHF19 and control ARP1 cells, and we identified 1,505 differentially expressed genes (Supplementary Fig. S5D). GSEA also indicated that the MYC, NF-κB, and KRAS signaling pathways were enriched in shRNA-PHF19 multiple myeloma cells (Supplementary Fig. S5E). Further analysis indicated that 224 genes were overlapped between the two RNA-seq studies (Fig. 5A and B), and the GSEA hallmark gene analysis indicated that these genes were involved in multiple signaling pathways (such as the NF-κB, EMT transition, hypoxia, and PI3K/AKT/mTOR pathways) that were implicated in cell growth and tumorigenesis (Fig. 5C). These data further support our hypothesis that PHF19 promotes myeloma cell survival through multiple signaling pathways. These RNA-seq data have been submitted to the GEO database (GSE128406) of the NCBI (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE128406).

Cha and colleagues reported that AKT-mediated phosphorylation of EZH2 suppresses methylation of H3K27 (23), and our RNA-seq data strongly indicated that the PI3K/AKT pathway was activated in PHF19 OE cells (Fig. 5D). We therefore investigated the relationship between AKT activation and phosphorylation of EZH2, and we found significantly increased levels of phosphorylation of AKT at Ser-473, phosphorylation of PDK1 at Ser-241, and phosphorylation of EZH2 at Ser-21 in PHF19 OE cells. These findings showed a correlation between the activation of AKT pathways and EZH2 phosphorylation (Fig. 6A). To further verify whether EZH2 phosphorylation is regulated by AKT in this context, we treated the PHF19 OE ARP1 and OCI-My5 cells with LY294002, a commonly used broad-spectrum inhibitor of PI3K/AKT. Western blot analysis showed that pharmacologic inhibition of the PI3K/AKT pathway significantly reduced EZH2 phosphorylation at Ser21 in both ARP1 and OCI-My5 cells (Fig. 6B), similar results were also found in PHF19 OE JJN3 cells (Supplementary Fig. S6) These results showed that PHF19 induced EZH2 phosphorylation by PI3K/AKT pathway.

Complex genomic events underlie multiple myeloma development. Gains of 9, 15, or 19 are common clonal events observed in multiple myeloma (24). The PHF19 gene is located on chromosome 9q33.2, and multiple myeloma cells commonly display DNA amplification in this region. To determine the mechanism that underlies the increase of PHF19 in myeloma cells, we first investigated whether PHF19 is caused by the DNA amplification of chromosome 9q; however, we found a negative correlation between those two (Fig. 7A), indicating that high levels of PHF19 are not induced by DNA amplification. We then hypothesized that a miRNA-based posttranscriptional mechanism was likely involved in PHF19 regulation. On the basis of our report that downregulated miR-15a in multiple myeloma cells is correlated with drug resistance and poor survival of patients with multiple myeloma (25–27), we examined the expression of miR-15a and PHF19 in patients with NDMM (n = 27). Our data revealed that miR-15a expression was negatively correlated with PHF19 (r = −0.651, P = 0.0005; Fig. 7B and C). We also hypothesized that miR-15a was directly correlated with PHF19, which may explain the overexpression of PHF19 in multiple myeloma. Our results confirmed that overexpression of miR-15a via transient transfection of a mimic oligo (PMIRH 15a-PA-1) significantly downregulated PHF19 expression at the mRNA level as well as in protein levels (Fig. 7D and E; Supplementary Fig. S7). To establish a direct connection between miR-15a and PHF19, we undertook a bioinformatics analysis to test the possibility that the seed sequence of miR-15a binds with the 3′UTR of PHF19, as shown in Fig. 7F, followed by a dual luciferase reporter assay. We also found that overexpression of miR-15a reduced the activity of the luciferase reporter gene fused to the wt PHF19 3′UTR by a significant 52.5%, compared with the control group in HEK293T cells, while a deletion of 6 nucleotides in the miR-15a binding site in the 3′UTR of PHF19 induced no such change (Fig. 7G). These data underscore that PHF19 is one of the direct targets of miR-15a. On the basis of these findings, we have developed a working model, shown in Supplementary Fig. S8.

Wang and colleagues first described a human homolog of the Drosophila polycomb-like protein PHF19 (PCL3) in 2004 (28). Since then, PHF19 has been ascribed putative roles in the memory of cellular identity based on establishing and faithfully maintaining transcription states at the chromatin level. PHF19 is markedly overexpressed in many types of cancers and is highly correlated with tumor progression (14–18). Recently, new evidence has emerged that PHF19, a gene downstream of miR-155, participates in CD8⁺ T-cell biology (29). Here we have demonstrated through GEP analysis that PHF19 is significantly upregulated in samples from patients with multiple myeloma, especially in the relapsed and high-risk myeloma samples. Survival analysis clearly indicates that patients with multiple myeloma with elevated PHF19 have shorter survival and inferior prognoses in two different clinical cohorts: TT2 and IFM. Collectively, these clinical data suggest that the PHF19 gene plays a pivotal role in drug resistance and disease progression in multiple myeloma.

Drug resistance is a major obstacle limiting the effectiveness of multiple myeloma treatment and significantly contributing to the cause of relapse in patients with the disease. Elucidating the mechanism that underlies drug resistance and, of equal importance, developing a corresponding treatment, is critical for the clinical management of multiple myeloma. Here, we have identified the high-risk gene PHF19, which induces drug resistance and shorter survival of patients, our findings are consistent with the report of Zhou and colleagues (30, 31), which found that PHF19 is one of 56 drug resistance genes. We have also used in vitro and in vivo approaches to further demonstrate that high levels of PHF19 promote proliferation and multidrug resistance in multiple myeloma cells. Our RNA-seq results have uncovered that many transcription factors and oncogenes are downstream targets of PHF19 in promoting the cell growth and drug resistance of multiple myeloma cells (Supplementary Fig. S5). We document that JUN, JUNB, JUND, FOSB, KLF, and RELB, which play pivotal roles in the pathogenesis and progression of multiple myeloma (32, 33), are significantly upregulated in PHF19 OE cells. In addition, we report that ubiquitin genes, including UBB, UBA52, UBE2L6, and UBXN6, are upregulated in PHF19 OE cells: given that high levels of ubiquitin genes facilitate protein degradation via the ubiquitin-proteasome pathway, this upregulation likely contributes to the cell growth and drug resistance induced by PHF19 overexpression. Our data also indicate that CCL3, which is critical for the development of multiple myeloma bone disease, is significantly increased in multiple myeloma cells that express high levels of PHF19, indicating that PHF19 overexpression likely plays a role in bone disease related to multiple myeloma. Finally, the GSEA shows that critical signaling pathways, such as NF-κB, hypoxia, MYC-known to be involved in the pathogenesis of multiple myeloma, are dysregulated by PHF19.

PHF19 is a PRC2-associated factor that forms subcomplexes with PRC2 core component EZH2, which is a histone methyltransferase (10) that participates in the proliferation and differentiation of cells. Our data reveal that PHF19 promotes the phosphorylation-related inactivation of EZH2 and subsequently decreases histone H3K27 methylation, leading to upregulation of HIF-1α, Bcl-xL, and Mcl-1, and promoting the proliferation and drug resistance of multiple myeloma cells. These data are consistent with previous reports by other groups that phosphorylation-mediated EZH2 inactivation promotes CAM-DR in multiple myeloma (22).

PHF19 has no kinase enzyme domain and it promotes EZH2 phosphorylation. The underlying mechanism is still unclear. We hypothesize that upstream proteins of the AKT signaling pathway are directly linked to PHF19 (34). We performed immunoprecipitation for PHF19, followed by mass spectrometry, and detected the insulin receptor substrate 4 (IRS4), which is a poorly studied member of the IRS family. IRS4 activates the PI3K/AKT pathway in breast cancer and induces mammary tumorigenesis and confers resistance to HER2-targeted therapy (35). Another interesting molecule listed was receptor for activated protein C kinase 1 (RACK1), shown to be a multifaceted scaffolding protein involved in multiple biological events via interaction with different partners, including cell migration, and angiogenesis. RACK1 is an EphB3-binding protein and mediates the assembly of a ternary signal complex that comprises protein phosphatase 2A, AKT and itself in response to EphB3 activation, thereby leading to reduced AKT phosphorylation and subsequent inhibition of cell migration in non–small cell lung cancer (36). Further experiments are needed to verify our hypotheses. On the basis of our RNA-seq and Western blot analysis, the PI3K/AKT pathway is significantly activated in PHF19 OE multiple myeloma cells. Depression of AKT activation by the inhibitor LY294002 notably decreases the phosphorylation of EZH2 and phosphorylation-related inactivation of EZH2 downregulates of the level of methylation of H3K27 (a transcriptionally repressive marker), which in turn activates expression of downstream targets. Our results thus establish that the phosphorylation-related inactivation of EZH2 plays a pivotal role in mediating PHF19-induced multiple myeloma cell proliferation and drug resistance.

Recent reports show that PHF19 is highly expressed in various forms of cancer, including multiple myeloma (16–19); however, the molecular mechanism involved in the upregulation of PHF19 is poorly elucidated. Because the human PHF19 gene is located on chromosome 9q33.2, which is commonly amplified in multiple myeloma cells, we initially looked for a correlation between the PHF19 gene and chromosome 9q33 amplification; however, we did not find that the level of expression of PHF19 is correlated with 9q amplification in primary patient samples. In keeping with this, results from use of cancer profiling arrays reported in the literature, PHF19 is upregulated in other kinds of tumors without amplification of chromosome 9q (28, 37). Therefore, we concluded that other mechanisms underlie PHF19 regulation in multiple myeloma. Of note, no activation mutations, gene amplifications, or a hypomethylated promoter of PHF19 has been reported in tumor cells, suggesting that increased accumulation of PHF19 results from changes in posttranscriptional regulation. Our previous studies demonstrate that aberrant expression of miRNAs serves a pivotal role in the pathogenesis of multiple myeloma (38, 39). For example, miR-15a is significantly downregulated in multiple myeloma cells, promoting cell proliferation, and drug resistance (25–27). Here we used bioinformatics analysis and a luciferase report assay to identify PHF19 as a novel target of miR-15a. Our data validate that the seed sequence of miR-15a directly binds to the 3′UTR of PHF19 and suppresses PHF19 mRNA transcription and expression.

In sum, this study reveals that the gene for the PRC2 complex member polycomb-like protein PHF19 is overexpressed in patients with high-risk myeloma, and that this overexpression correlates with drug resistance and inferior outcome of patients with multiple myeloma. PHF19 promotes the phosphorylation-related inactivation of EZH2 by activating PI3K/AKT pathways that mediate PHF19-induced proliferation and drug resistance of multiple myeloma cells. PHF19 causes demethylation of histone H3K27 and promotes expression of HIF-1a, Bcl-xL, and Mcl-1, thereby inducing multiple myeloma cell proliferation and conferring drug resistance. We also show that PHF19-induced inactivation of EZH2 by phosphorylation is a novel epigenetic mechanism involved in the promotion of multiple myeloma cell proliferation and drug resistance. Our findings implicate the miR-15a/PHF19/pho-EZH2 epigenetic axis as a critical participant in multiple myeloma tumorigenesis and suggest that it is a promising target for multiple myeloma treatment.

K.C. Anderson is a consultant for Bristol-Myers Squibb, Millennium-Takeda, Sanofiaventis, Janssen, and Gilead. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Yu, C. Du, W. Sui, F. Zhan, M. Hao

Development of methodology: T. Yu, C. Du, Z. Yu, M. Hao

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Yu, C. Du, W. Sui, L. Liu, Z. Li, J. Xu, X. Wei, W. Zhou, S. Deng, D. Zou, G. An, K.C. Anderson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Yu, C. Du, X. Ma, W. Sui, L. Zhao, Z. Li, G. An, Y.-T. Tai, K.C. Anderson, F. Zhan, M. Hao

Writing, review, and/or revision of the manuscript: T. Yu, C. Du, L. Zhao, G. An, Y.-T. Tai, G. Tricot, K.C. Anderson, L. Qiu, F. Zhan, M. Hao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Qiu

Study supervision: L. Qiu, M. Hao

The authors acknowledge the patients and physicians who participated in sample collections. We thank Drs. Sonal Jhaveri-Schneider (Postdoc and Graduate Student Affairs Office, Dana-Farber Cancer Institute) and Jie Ni (South Eastern Sydney Local Health District) for editing some drafts of the manuscript. We thank Dr. Teru Hideshima, Kenneth Wen, and Dr. Wenjuan Yang (Dana-Farber Cancer Institute) for their assistance and constructive advice on this project. This work was supported by Natural Science Foundation of China (81570181 and 81400174, to M. Hao; 81630007 and 81920108006, to L. Qiu); Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences CAMS-2016-I2M-3-031, CAMS-2017-I2M-1-005, and CAMS 2017-I2M-1-015 (to M. Hao and L. Qiu); Tianjin Science and Technology Supporting Program (17JCYBJC27900, to M. Hao) and the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (2018RC320012, to M. Hao).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.