Pancreatic cancer is a lethal disease owing to its intrinsic and acquired resistance to therapeutic modalities. The altered balance between pro- and antiapoptosis signals within cancer cells is critical to therapeutic resistance. However, the molecular mechanisms underlying increased antiapoptosis signals remain poorly understood. In this study, we report that PRMT1 expression is increased in pancreatic cancer tissues and is associated with higher tumor grade, increased aggressiveness, and worse prognosis. PRMT1 overexpression increased arginine methylation of HSPs of 70 kDa (HSP70); this methylation enhanced HSP70 binding and stabilization of BCL2 mRNA through AU-rich elements in 3′-untranslated region and consequentially increased BCL2 protein expression and protected cancer cells from apoptosis induced by cellular stresses and therapeutics. RNA binding and regulation function of HSP70 was involved in pancreatic cancer drug resistance and was dependent on protein arginine methylation. These findings not only reveal a novel PRMT1–HSP70–BCL2 signaling axis that is crucial to pancreatic cancer cell survival and therapeutic resistance, but they also provide a proof of concept that targeted inhibition of this axis may represent a new therapeutic strategy.

Significance:

This study demonstrates that a PRMT1-mediated stabilization of BCL2 mRNA contributes to therapeutic resistance in pancreatic cancer and that targeting this pathway could overcome said resistance.

Pancreatic ductal adenocarcinoma (PDAC) is the third leading cause of cancer-related death in the United States and is expected to surpass colorectal cancer to become the second leading cause by 2030 (1). Most patients with PDAC are diagnosed with advanced-stage disease, which is unsuitable for curative surgery. Intrinsic and acquired resistance to therapeutic modalities is a critical challenge in clinical practice, causing existing therapeutic strategies to have little impact on overall survival (2). The altered balance between pro- and antiapoptosis signals within cancer cells is associated with therapeutic resistance, cancer progression, and metastasis, which account for most cancer-related mortality (3).

HSPs of 70 kDa (HSP70) constitute an evolutionarily conserved protein family and these proteins are ATP-dependent molecular chaperones (4). HSP70 is the primary member of this family, playing a critical role in promoting cell survival under various stress conditions, including cytotoxic chemotherapy (5). As a classic protein chaperone, HSP70 has a highly conserved domain structure to facilitate client protein folding. It binds directly to several important proteins involved in both intrinsic and extrinsic apoptosis pathways to inhibit cell death and provide survival advantages to tumor cells (6). In addition to its canonical function as a protein chaperone, HSP70 has RNA binding ability and can bind to and stabilize specific mRNA molecules containing AU-rich elements (ARE) in the 3′-untranslated region (3′-UTR; refs. 7–9). HSP70 RNA-binding activity is independent of its protein chaperone function (10). However, whether this novel nucleic acid binding activity of HSP70 regulates cell survival is unknown.

The function of HSP70 is regulated by multiple posttranslational modifications, such as phosphorylation, acetylation, malonylation, ubiquitination, and methylation (11, 12). HSP70 protein methylation was first discovered to occur on a conserved K561 residue, and trimethylation at this lysine residue altered its binding affinity to a client protein, α-synuclein (13). More recently, Gao and colleagues showed that CARM1/PRMT4-mediated methylation of R469 affected the interaction between HSP70 and TFIIH and therefore influenced RARβ2 gene activation (14). However, it remains unknown whether there are other methylation sites in HSP70 protein, whether HSP70 is methylated by other protein arginine methyltransferases (PRMT), and how arginine methylation influences HSP70 biological functions in cancer development.

Protein arginine methylation is catalyzed by a sequence-related family of nine PRMTs in humans (15). The classic substrate of PRMT is histone. Methylation of arginine residues in histone tails is a component of “histone code,” which is an important regulator of gene expression at the transcriptional level (16). Arginine methylation has also been found on many non-histone proteins localized in the cell nucleus, cytoplasm, and cell membrane (17–19). Arginine methylation of various proteins is involved in a growing list of biological activities, such as transcription, RNA processing, translation, DNA damage repair, signal transduction, cell cycle, and apoptosis, among others (15, 20, 21). Clinical and basic science studies indicate that dysregulation of PRMTs is involved in cancer development (22). However, the role of specific PRMT members and protein arginine methylation in PDAC initiation and progression is unknown. Such information is indispensable for the rational development of PRMT inhibitors to treat cancer.

In this study, we determined the impact of methylation of HSP70 on therapeutic resistance and underlying mechanisms in PDAC cells. We established that the novel PRMT1–HSP70–BCL2 signaling axis is crucial to PDAC cell survival and drug resistance. Targeted inhibition of this axis could represent a new therapeutic strategy for PDAC.

Details are available in Supplementary Information for cell lines and cell culture; plasmids and cloning procedures; antibodies, siRNAs, primers and compounds; shRNA construct and transfection; Western blot analysis; Alcian blue staining; cell scratch-wound healing assay; cell invasion assay; in vitro cytotoxicity and cell viability assay; immunoprecipitation assay; protein purification from bacterial or mammalian cells; immunofluorescence assay; in vitro methylation assay; protein sequence alignment; mass spectrometry analysis; generation of gene knockout (KO) cell lines; annexin V-propidium iodide assay; quantitative real-time PCR assay; ribonucleoprotein immunoprecipitation; mRNA decay assay; protein-nucleic acid binding assay; and luciferase assay.

Human tissue specimens and IHC analysis

PRMT1 expression was analyzed using human pancreatic, gastric, and colon cancer or mouse PDAC tissue samples. The human PDAC tissue microarray information is listed in Supplementary Table S1. The use of all tissue samples was approved by the Institutional Review Board of MD Anderson. Human or mouse tissue sections (4 μm thick) of formalin-fixed, paraffin-embedded tumor specimens were deparaffinized in xylene and rehydrated in graded alcohol. Standard IHC procedures were performed on the tissue microarray or tissue sections using the anti-PRMT1 (1:180; CST) antibody. A positive signal was indicated by a yellow-brown color staining. The percentage of positive signal area was scored as percentage score: <10% (0), 10% to 25% (1), 25% to 50% (2), 50% to 75% (3), and >75% (4). The intensity of staining was scored as intensity score: no staining (0), light brown (1), brown (2), and dark brown (3). The overall score is obtained by multiplying percentage score and intensity score. The overall score <4 is regarded as negative/weak expression, 4≤ overall score <8 is regarded as moderate expression and overall score ≥8 is regarded as strong expression. For each mouse sample, five randomly selected fields were scored and calculated to get the average staining score. All tissues sections were scored by two independent investigators prior not knowing the patient outcomes, and the mean values of two independent scores are presented.

Animal strains and tumor modeling experiments

Female athymic nude mice and C57BL/6 mice were purchased from Jackson Laboratory. The nude mice were used when they were 8 weeks old. All animals were maintained in an MD Anderson animal facility approved by the Association for Assessment and Accreditation of Laboratory Animal Care in accordance with the current regulations and standards.

For xenograft modeling, tumor cells (1 × 106) in 0.1 mL of Hank's balanced salt solution without calcium and magnesium were injected subcutaneously into the flanks of nude mice. The tumors' length and width were measured with a caliper twice per week. The tumor-bearing mice were euthanized when they became moribund or at indicated time points after inoculation, and the tumors were removed and weighed. The tumor volumes (mm3) were calculated using the following formula: width2 × length/2. The PRMT1 inhibitor DB75 was injected intraperitoneally into the mice twice per week for 3 weeks. The low dose was 5 mg/kg and the high dose was 20 mg/kg. Vehicle solution was used as a control.

Statistical analysis

Kaplan–Meier survival curves were calculated using the overall survival time for each human patient. The log-rank test was used to test for significant differences between the groups. All in vitro and in vivo experiments were performed in triplicate or as indicated. Statistical significance was determined by the Student t test (two-tailed unpaired or paired t test depending on the experiment) or one-way ANOVA. For tissue microarray analysis, associations between categorical variables were tested by Pearson correlation coefficient, two-tailed Chi-square test or Fisher exact test. All significance was defined as *, P < 0.05; **, P < 0.01; ***, P < 0.001, or nonsignificant. Data are represented as mean ± SD or mean ± SEM.

Aggressive PDAC expresses an elevated level of PRMT1 protein

To study the biological roles of PRMTs in PDAC, we first examined the expression of two representative PRMTs—PRMT1, the major type I PRMT; and PRMT5, the major type II PRMT—in a panel of human PDAC cell lines. Although the expression level of PRMT5 was similar among all tumor and normal cell lines, PRMT1 expression was higher in tumor cell lines than in normal human pancreatic cell line HPNE (Fig. 1A). The increased expression of PRMT1 in PDAC cells was further confirmed in paired normal pancreas and PDAC specimens from human patients (Fig. 1B). Therefore, we focused our study on the role of PRMT1 in PDAC carcinogenesis.

Figure 1.

PRMT1 is overexpressed in PDAC and is associated with clinicopathologic features of PDAC. A, Western blot analysis of PRMT1 and PRMT5 expression in a panel of human PDAC cell lines and an immortalized human pancreatic ductal cell line, HPNE. B, Western blot analysis of PRMT1 expression in paired human normal pancreatic tissue (N) and PDAC specimens (T). C and D, Human PDAC tissue microarray stained with a specific anti-PRMT1 antibody. PRMT1 expression in PDAC tumor tissues (bottom) and adjacent normal tissues (top) are shown in C. PRMT1 expression in human PDAC tumor tissues with differentiation grade I (top) and grade III (bottom) are shown in D. E, Statistical analysis of PRMT1 staining scores for tumor tissues and adjacent normal tissues. Error bars represent 25 and 75 percentiles. F and G, PRMT1 staining scores are categorized to three groups and compared between tumor and normal tissues in F or different tumor differentiation grades in G. P values were calculated by Chi-square test. H, Overall survival curve of PDAC patients with high (score ≥6, N = 28) or low (score <6, N = 50) expression of PRMT1 in tumor tissues. I and J, Overall survival curves and median overall survival of patient groups with different PRMT1 expression levels and tumor differentiation grades. Differentiation grade I and II are defined as good, while grade III is defined as poor differentiation. P value of overall log-rank test is shown in I.

Figure 1.

PRMT1 is overexpressed in PDAC and is associated with clinicopathologic features of PDAC. A, Western blot analysis of PRMT1 and PRMT5 expression in a panel of human PDAC cell lines and an immortalized human pancreatic ductal cell line, HPNE. B, Western blot analysis of PRMT1 expression in paired human normal pancreatic tissue (N) and PDAC specimens (T). C and D, Human PDAC tissue microarray stained with a specific anti-PRMT1 antibody. PRMT1 expression in PDAC tumor tissues (bottom) and adjacent normal tissues (top) are shown in C. PRMT1 expression in human PDAC tumor tissues with differentiation grade I (top) and grade III (bottom) are shown in D. E, Statistical analysis of PRMT1 staining scores for tumor tissues and adjacent normal tissues. Error bars represent 25 and 75 percentiles. F and G, PRMT1 staining scores are categorized to three groups and compared between tumor and normal tissues in F or different tumor differentiation grades in G. P values were calculated by Chi-square test. H, Overall survival curve of PDAC patients with high (score ≥6, N = 28) or low (score <6, N = 50) expression of PRMT1 in tumor tissues. I and J, Overall survival curves and median overall survival of patient groups with different PRMT1 expression levels and tumor differentiation grades. Differentiation grade I and II are defined as good, while grade III is defined as poor differentiation. P value of overall log-rank test is shown in I.

Close modal

To determine the role of PRMT1 in PDAC pathogenesis, we used a human PDAC tissue microarray (Supplementary Table S1). Higher expression of PRMT1 was observed in tumor tissues than in adjacent normal tissues (Fig. 1C, E, F). High expression of PRMT1 was associated with poor tumor differentiation status and reduced overall patient survival (Fig. 1D, G, H; Supplementary Table S1). Moreover, combined PRMT1 expression level and tumor differentiation grade predicted the prognosis of PDAC patients (Fig. 1I and J). PRMT1 expression is increased in metastasis-derived cells than their isogeneic parental cells (Supplementary Fig. S1A). Increased PRMT1 mRNA expression was also associated with reduced survival time and 5-year overall survival rates in several other human gastrointestinal cancers (Supplementary Fig. S1B). IHC staining also revealed that PRMT1 was overexpressed in these gastrointestinal cancer tissues (Supplementary Fig. S1C).

Progressive increase of PRMT1 expression indicates PDAC development and progression

PDAC arises from noninvasive precursor lesions, mostly pancreatic intraepithelial neoplasia (PanIN; ref. 23). In human PDAC specimens, PRMT1 expression was low in normal glandular cells, modest in PanIN lesions, and high in advanced PDAC tissues (Fig. 2A). To further investigate the expression change of PRMT1 at each step during PDAC carcinogenesis, we performed immunohistochemical staining of pancreatic tissues from two PDAC mouse models-KC mice (LSL-KrasG12D/+;Pdx-Cre) and KPC mice (LSL-KrasG12D/+;p53R172H/+;Pdx-Cre; refs. 24, 25)—at various stages. Positive staining of PRMT1 in both PanIN lesions and invasive PDAC lesions was observed, but no or very weak staining was seen in normal mouse pancreatic tissues (Supplementary Fig. S2A and S2B). Moreover, PRMT1 expression gradually increased as the disease developed from low-grade PanIN (PanIN 1A/B) to high-grade PanIN (PanIN 2/3) and finally invasive PDAC (Fig. 2B and C). To rule out the possibility that high PRMT1 expression is simply a marker of active cell dividing, we costained PRMT1 and Ki67 in mouse PanIN lesion. These two proteins showed different expression patterns (Supplementary Fig. S2C).

Figure 2.

Progressive increase of PRMT1 expression during PDAC progression from PanIN to invasive PDAC. A, IHC staining showing PRMT1 expression in human PDAC specimens at different stages. B, IHC staining showing PRMT1 expression in various pancreatic lesions from mouse PDAC models. C, Statistical analysis of the average staining scores for different groups (n = 8). D, Representative images (left) of Alcian blue/eosin staining of pancreatic tissues from mice treated with either PRMT1 inhibitor DB75 or PBS twice per week. Scale bars, 500 μm. Statistical analysis is shown in right panel (n = 8).

Figure 2.

Progressive increase of PRMT1 expression during PDAC progression from PanIN to invasive PDAC. A, IHC staining showing PRMT1 expression in human PDAC specimens at different stages. B, IHC staining showing PRMT1 expression in various pancreatic lesions from mouse PDAC models. C, Statistical analysis of the average staining scores for different groups (n = 8). D, Representative images (left) of Alcian blue/eosin staining of pancreatic tissues from mice treated with either PRMT1 inhibitor DB75 or PBS twice per week. Scale bars, 500 μm. Statistical analysis is shown in right panel (n = 8).

Close modal

To substantiate the critical role of PRMT1 in PDAC carcinogenesis, we treated KC mice at the age of 8 weeks, when PanINs begin to occur in mouse pancreas, with furamidine, a recently identified PRMT1 inhibitor (also known as DB75) for 6 weeks (26). Treatment with furamidine decreased the percentage of PanIN lesions, which were stained by either Alcian blue (Fig. 2D) or anti-CK19 antibody (Supplementary Fig. S2D). This result indicates that the PRMT1 inhibitor has some cancer prevention potential.

How PRMT1 is upregulated during cancer progression is not clearly understood. We analyzed the PRMT1 locus in a reported PDAC whole-exome sequencing database containing 105 cases (27) and found that about 64% of the tumor samples had gain or amplification of the PRMT1 gene (Supplementary Fig. S2E). These suggest that genomic instability and the resultant copy number alteration may contribute to PRMT1 overexpression. Moreover, beginning at the PanIN stage, stromal cells are activated, and proinflammatory cytokines are secreted to promote PDAC tumorigenesis (28). We found that treatment with TNFα could dose-dependently increase PRMT1 expression in cultured human and mouse PDAC cells (Supplementary Fig. S2F). Therefore, local inflammation may also contribute to elevated PRMT1 expression during pancreatic carcinogenesis.

PRMT1 is a protumorigenic protein in vitro and in vivo

To determine the effect of altered PRMT1 expression on PDAC growth, we knocked down PRMT1 in MDA28 cells with high PRMT1 expression and overexpressed PRMT1 in PANC-1 cells with low PRMT1 expression (Fig. 3A and B; left). Knockdown of PRMT1 decreased MDA28 cell proliferation, whereas overexpression of PRMT1 promoted PANC-1 cell proliferation (Fig. 3A and B; right). Knockdown of PRMT1 also attenuated the migration and invasion of PDAC cells, whereas ectopic expression of PRMT1 did the opposite (Supplementary Fig. S3A and S3B).

Figure 3.

Altered expression of PRMT1 affects PDAC cell proliferation and growth. A,In vitro cell proliferation of MDA28 cells transfected with siCtrl or siPRMT1. B,In vitro cell proliferation of PANC-1 cells transfected with pcDNA3 or pHA-PRMT1 plasmids. C, PRMT1 was knocked down in MDA28 cells, and the growth of tumor xenografts in nude mice was slower than those of mock-transfected cells. D and E, IC50 of DB75 and TC-E5003 in three PDAC cell lines. F, H7 mouse PDAC cells were injected subcutaneously into C57BL/6 mice to grow for 1 week. The mice then received treatment with vehicle or low-dose (5 mg/kg) and high-dose (20 mg/kg) DB75 via intraperitoneal injection twice per week. Tumors were photographed and weighed at day 21, and tumor volumes were monitored for 21 days. P value was determined by one-way ANOVA. *, P < 0.05; **, P < 0.01.

Figure 3.

Altered expression of PRMT1 affects PDAC cell proliferation and growth. A,In vitro cell proliferation of MDA28 cells transfected with siCtrl or siPRMT1. B,In vitro cell proliferation of PANC-1 cells transfected with pcDNA3 or pHA-PRMT1 plasmids. C, PRMT1 was knocked down in MDA28 cells, and the growth of tumor xenografts in nude mice was slower than those of mock-transfected cells. D and E, IC50 of DB75 and TC-E5003 in three PDAC cell lines. F, H7 mouse PDAC cells were injected subcutaneously into C57BL/6 mice to grow for 1 week. The mice then received treatment with vehicle or low-dose (5 mg/kg) and high-dose (20 mg/kg) DB75 via intraperitoneal injection twice per week. Tumors were photographed and weighed at day 21, and tumor volumes were monitored for 21 days. P value was determined by one-way ANOVA. *, P < 0.05; **, P < 0.01.

Close modal

To determine the role of PRMT1 in vivo, we used either PRMT1-specific shRNA or small-molecule inhibitors. Knockdown of PRMT1 inhibited xenograft tumor growth of MDA28 and PANC-1 cells in nude mice (Fig. 3C; Supplementary Fig S3C). DB75 and TC-E5003, two small molecular inhibitors of PRMT1, have very weak inhibition on other PRMT members (26, 29). In our analysis, both DB75 and TC-E5003 were cytotoxic to PDAC cells at micromolar levels in vitro (Fig. 3D and E; Supplementary Fig. S3D). These inhibitors decreased PRMT1- mediated protein arginine methylation in cultured cells (Supplementary Fig. S3E). In animal models, intraperitoneal administration of DB75 dose-dependently inhibited H7 and MDA28 subcutaneous tumor growth without obvious side effects (Fig. 3F; Supplementary Fig. S3F). Consistent with a previous report (26), treatment with DB75 only inhibited PRMT1 enzymatic activity rather than its expression level (Supplementary Fig. S3F). In addition, these inhibitors did not affect viability of stromal pancreatic stellate cells (Supplementary Fig. S3G), which demonstrates a good cellular selectivity of PRMT1 inhibitors. These results indicate that PRMT1 has protumorigenic functions and inhibition of PRMT1 suppresses tumor growth in vitro and in vivo.

PRMT1 interacts with HSP70 protein

PRMT1 is the predominant methyltransferase accounting for >85% of all arginine methylation in human cells (15). To identify PRMT1 substrates, methylation of which would affect its protumorigenic functions, we used HA-tagged PRMT1 as a bait and identified a group of PRMT1 binding partners using coupled immunopurification and mass spectrometry (Supplementary Table S2). Interestingly, major HSP70 family members were found to be novel PRMT1 binding partners, and this was confirmed by reciprocal immunoprecipitation assays (Fig. 4AC). The colocalization of PRMT1 and HSP70 was also demonstrated in cell culture. In mouse embryonic fibroblasts, HSP70 expression was very low under normal culture conditions, and it was dramatically induced and translocalized into nuclei after heat shock treatment (Supplementary Fig. S4A), which is consistent with previous report (30). In contrast, high HSP70 expression was mainly and constantly localized in the cell nuclei of PANC-1 and MiaPaCa-2 cells, even without heat shock treatment (Fig. 4D).

Figure 4.

PRMT1 colocalizes and interacts with HSP70 protein. A, HA-tagged PRMT1 and Myc-tagged HSP70 isoforms were coexpressed in 293T cells, and cell lysates were immunoprecipitated (IP) with anti-Myc antibody and immunoblotted (IB) with anti-HA antibody. B, Protein interaction between ectopically expressed HA-PRMT1 and endogenous HSP70 isoforms. C, Protein interaction between endogenous PRMT1 and HSP70 in L3.7 cells. HC, heavy chain. D, Immunofluorescence demonstrated colocalization of HSP70 and PRMT1 in PANC-1 cells. Merge I overlaid green and red fluorescence, while Merge II included all three fluorescence colors. E and F, Mapping the motif of HSP70 that interacts with PRMT1. Myc-tagged full-length (FL) or a series of deletion mutants of HSP70 and HA-tagged PRMT1 were transfected into 293T cells. Cell lysates were immunoprecipitated with anti-HA antibody and then immunoblotted with anti-Myc antibody. WCL, whole cell lysate.

Figure 4.

PRMT1 colocalizes and interacts with HSP70 protein. A, HA-tagged PRMT1 and Myc-tagged HSP70 isoforms were coexpressed in 293T cells, and cell lysates were immunoprecipitated (IP) with anti-Myc antibody and immunoblotted (IB) with anti-HA antibody. B, Protein interaction between ectopically expressed HA-PRMT1 and endogenous HSP70 isoforms. C, Protein interaction between endogenous PRMT1 and HSP70 in L3.7 cells. HC, heavy chain. D, Immunofluorescence demonstrated colocalization of HSP70 and PRMT1 in PANC-1 cells. Merge I overlaid green and red fluorescence, while Merge II included all three fluorescence colors. E and F, Mapping the motif of HSP70 that interacts with PRMT1. Myc-tagged full-length (FL) or a series of deletion mutants of HSP70 and HA-tagged PRMT1 were transfected into 293T cells. Cell lysates were immunoprecipitated with anti-HA antibody and then immunoblotted with anti-Myc antibody. WCL, whole cell lysate.

Close modal

To identify the regions of HSP70 and PRMT1 interaction, we generated a series of Myc-tagged deletion mutants of HSP70 and HA-tagged deletion mutants of PRMT1 according to their domain structure (Supplementary Fig. S4B and S4C). As shown in Fig. 4E, the amino acids M410-E460 of HSP70 were critical for the interaction. Deletion of either the N-terminal or C-terminal part of motif M410-E460 dramatically increased its binding affinity to PRMT1 (compare lane 2 and 4 to lane 7 in Fig. 4E). This could be due to the removal of obstructive motifs at either terminus, and exposure of the binding domain potentially made this motif more accessible for binding by PRMT1.

To further pinpoint the key binding motif, we subdivided M410-E460 into three fragments. Interestingly, we found that all 50 amino acids were necessary for the interaction, because binding affinity gradually increased when the carboxyl terminus extended from 430 to 441 and 460 (Fig. 4F). We speculate that there is a small core binding motif within domain 410 to 460, and that the proximal amino acids around this core motif are also supportive to the interaction. However, in a similar assay using deletion mutants of PRMT1, deletion of either the N-terminus or C-terminus of PRMT1 significantly decreased its protein stability (Supplementary Fig. S4D). This result indicates that the structural integrity of PRMT1 is important, but it prevented us from mapping the binding motif of PRMT1 with HSP70.

PRMT1 catalyzes methylation of HSP70

We next determined whether PRMT1 can methylate HSP70. A specific antibody against methylated arginine (ASYM24) was validated using PRMT1 knockdown and overexpression cell lysates (Supplementary Fig. S5A). Increased methylation of HSP70 was observed when PRMT1 was overexpressed in PDAC cells (Supplementary Fig. S5B). To identify methylation sites, we purified Myc-tagged HSP70 protein using immunoprecipitation (Supplementary Fig. S5C). The prominent band of HSP70 was excised and subjected to mass spectrometry, which led to the identification of R416 as a methylation site (Fig. 5A). This site is in the 410 to 460 binding motif. Protein sequence analysis revealed that there were only four arginine residues between amino acids 400 and 500 of HSP70, and R469 was previously reported to be a PRMT4 methylation site (14). Multiple sequence alignment showed that among the remaining three arginine residues, R416 and R447 were highly conserved among different isoforms of HSP70 (Supplementary Fig. S5D) and during evolution (Supplementary Fig. S5E). In contrast, R458 was not well conserved.

Figure 5.

PRMT1 catalyzes methylation of HSP70 at two arginine residues. A, Mass spectrometry analysis of arginine methylation sites of HSP70 protein. B,In vitro methylation assays using synthetic peptides and liquid scintillation counting. R3 is the positive control peptide containing three GAR repeats. Individual P value was determined between each experimental group and blank group. ***, P < 0.001; N.S., nonsignificant. C, In vitro methylation assays using purified GST-fusion proteins and autoradiography. 3RA is full-length HSP70 with 416, 447, and 458 triple R-to-A mutation. D, Top view of the complex obtained from flexible protein-peptide docking. PRMT1 is shown as magenta surface; HSP70 peptide is shown as gray line. E, Interaction between HSP70 residue R416 and PRMT1 residue E152 at the active site. HSP70 peptide is shown as cyan line and targeted R416 is highlighted in yellow. F, Interaction between HSP70 residue R447 and PRMT1 residue E161 at the active site. HSP70 peptide is shown as gray line and targeted R447 is shown as gray stick. G, HSP70 residue R458 failed to fit into the active site in the protein-peptide study. HSP70 peptide is shown as cyan line and targeted R458 is shown as cyan stick. In D–G, PRMT1 is shown in magenta, SAM is shown in brown, and critical residues of the PRMT1 active site are shown in green.

Figure 5.

PRMT1 catalyzes methylation of HSP70 at two arginine residues. A, Mass spectrometry analysis of arginine methylation sites of HSP70 protein. B,In vitro methylation assays using synthetic peptides and liquid scintillation counting. R3 is the positive control peptide containing three GAR repeats. Individual P value was determined between each experimental group and blank group. ***, P < 0.001; N.S., nonsignificant. C, In vitro methylation assays using purified GST-fusion proteins and autoradiography. 3RA is full-length HSP70 with 416, 447, and 458 triple R-to-A mutation. D, Top view of the complex obtained from flexible protein-peptide docking. PRMT1 is shown as magenta surface; HSP70 peptide is shown as gray line. E, Interaction between HSP70 residue R416 and PRMT1 residue E152 at the active site. HSP70 peptide is shown as cyan line and targeted R416 is highlighted in yellow. F, Interaction between HSP70 residue R447 and PRMT1 residue E161 at the active site. HSP70 peptide is shown as gray line and targeted R447 is shown as gray stick. G, HSP70 residue R458 failed to fit into the active site in the protein-peptide study. HSP70 peptide is shown as cyan line and targeted R458 is shown as cyan stick. In D–G, PRMT1 is shown in magenta, SAM is shown in brown, and critical residues of the PRMT1 active site are shown in green.

Close modal

To verify that these arginine residues are PRMT1 methylation sites, we synthesized peptides with lengths of 19 amino acids containing R416, R447, and R458, as well as the corresponding methylation-site mutant peptides, R416A, R447A, and R458A, for in vitro methylation assay (Supplementary Table S3). A well-established peptide containing three GAR repeats (R3) was used as a positive control (31). Results showed that R416 and R447, but not R458, can be methylated by PRMT1 when 3H-labeled S-adenosyl methionine (SAM) was added as the methyl group donor (Fig. 5B). Finally, purified GST-HSP70 fusion proteins from bacteria and HA-PRMT1 from 293T cells were used for the in vitro methylation assay. Methylation of HSP70 was measured using autoradiography. Again, R416 and R447, but not R458, were targets for methylation (Fig. 5C).

The above experiments suggest a direct interaction between HSP70 and PRMT1 and reveal two methylation sites of HSP70. To further explore the interaction mechanism that facilitates HSP70 methylation by PRMT1 from a structural point of view, we conducted molecular modeling studies to predict the binding poses of PRMT1–HSP70 complex. Because the crystal structure of human PRMT1 is not currently available, we first built the homology model of human PRMT1 in complex with the methyl group donor SAM, based on the crystal structure of rat PRMT1 (PDB ID: 1ORI), which has a very high sequence identity with the human counterpart (31). The initial structures of HSP70 peptides containing R416, R447, and R458 were taken from PDB structure 4PO2. Then flexible protein-peptide docking was performed using both Z-dock and Cluspro2.0 webserver (32, 33), and the best poses were kept for further studies. Targeted arginine residues R416 and R447 were clearly situated in a deep pocket formed by residues Glu152, Tyr156, and Glu161 of PRMT1 and the ligand SAM, between strand β4 and helix αD (Fig. 5D; ref. 31).

The guanidine groups of targeted arginine R416 and R447 formed salt bridges with E152 and E161, respectively. Also, the aromatic group of Tyr156 lay in parallel to the methylene groups of the targeted arginine (Fig. 5E and F). These interactions assisted the stable binding of targeted arginine residues to the enzymatic site, which is consistent with our in vitro methylation assay results showing that these two sites are methylated by PRMT1. In contrast, peptide containing R458 was unable to fit into the active site (Fig. 5G). Presumably, R458 is in the middle of a beta strand and has limited flexibility. Our data also suggest that the negative charges of E152 and E161 are critical for the binding of targeted arginine and catalysis, and in fact these two glutamic acids are highly conserved among the PRMT family.

Methylation of HSP70 protects PDAC cells from apoptosis

Given that HSP70 protects cancer cells under stress (34, 35), we investigated the effect of HSP70 methylation on cell survival under various types of cellular stress using cell lines expressing wild-type or methylation-deficient HSP70 proteins. HSP70-KO MiaPaCa-2 cells were generated and confirmed (Supplementary Fig. S6A). Six HSP70-null cell clones were selected to measure their sensitivity to gemcitabine. Loss of HSP70 sensitized MiaPaCa-2 cells to gemcitabine treatment (Supplementary Fig. S6B), and reconstitution of wild-type HSP70 promoted cell growth and rescued their resistance to gemcitabine (Fig. 6AC; Supplementary Fig. S6D). Consistently, deletion of HSP70 attenuated tumor growth in vivo (Supplementary Fig. S6E).

Figure 6.

PRMT1-mediated HSP70 methylation protects cancer cells against various stresses. A and B, Two HSP70-KO cell clones (KO-16 and KO-20) were transfected with pcDNA3.0 control plasmid or wild-type (WT) HSP70 expression plasmid. Cell growth inhibition induced by gemcitabine (Gem) was measured. C, HSP70-KO MiaPaCa-2 cells were transfected with wild-type (WT), single-methylation site mutant (R416A or R447A), or 416 and 447 double-site mutant (Mut) HSP70 plasmids. Cell growth inhibition of gemcitabine was measured. The sgCtrl cells were pooled cells of six control clones from Supplementary Fig. S6A and were used as the control group. P value was determined by one-way ANOVA. D and E, PRMT1 was overexpressed (D) or knocked down (E) in FG cells and cell growth inhibition of gemcitabine was measured. F, MiaPaCa-2 cells were treated with DB75 (5 μmol/L), gemcitabine (20 μmol/L), or a combination of both for 48 hours and the number of live cells was determined using the CCK-8 assay. G–I, HSP70-KO MiaPaCa-2 cells were transfected with pcDNA3.0 (P3.0), WT HSP70, or double-site mutant (Mut) HSP70 plasmids and then challenged with etoposide (G), H2O2 (H), or glucose deprivation (I) for 48 hours, and the number of live cells was determined using CCK-8 assay. J, HSP70 methylation reduced gemcitabine-induced apoptosis. HSP70-KO MiaPaCa-2 cells transfected with pcDNA3.0, WT HSP70, or mutant HSP70 were treated with gemcitabine or vehicle for 24 hours and apoptotic cells were then analyzed using Annexin V-propidium iodide (PI) staining and flow cytometry. The percentage of apoptotic cells is shown. K, HSP70 methylation inhibited stress-induced activation of caspase-9. After treatment with gemcitabine for 24 hours, the cleavage of caspase-9 was detected by Western blot analysis. ***, P < 0.001. All bar plot data are means ± SD.

Figure 6.

PRMT1-mediated HSP70 methylation protects cancer cells against various stresses. A and B, Two HSP70-KO cell clones (KO-16 and KO-20) were transfected with pcDNA3.0 control plasmid or wild-type (WT) HSP70 expression plasmid. Cell growth inhibition induced by gemcitabine (Gem) was measured. C, HSP70-KO MiaPaCa-2 cells were transfected with wild-type (WT), single-methylation site mutant (R416A or R447A), or 416 and 447 double-site mutant (Mut) HSP70 plasmids. Cell growth inhibition of gemcitabine was measured. The sgCtrl cells were pooled cells of six control clones from Supplementary Fig. S6A and were used as the control group. P value was determined by one-way ANOVA. D and E, PRMT1 was overexpressed (D) or knocked down (E) in FG cells and cell growth inhibition of gemcitabine was measured. F, MiaPaCa-2 cells were treated with DB75 (5 μmol/L), gemcitabine (20 μmol/L), or a combination of both for 48 hours and the number of live cells was determined using the CCK-8 assay. G–I, HSP70-KO MiaPaCa-2 cells were transfected with pcDNA3.0 (P3.0), WT HSP70, or double-site mutant (Mut) HSP70 plasmids and then challenged with etoposide (G), H2O2 (H), or glucose deprivation (I) for 48 hours, and the number of live cells was determined using CCK-8 assay. J, HSP70 methylation reduced gemcitabine-induced apoptosis. HSP70-KO MiaPaCa-2 cells transfected with pcDNA3.0, WT HSP70, or mutant HSP70 were treated with gemcitabine or vehicle for 24 hours and apoptotic cells were then analyzed using Annexin V-propidium iodide (PI) staining and flow cytometry. The percentage of apoptotic cells is shown. K, HSP70 methylation inhibited stress-induced activation of caspase-9. After treatment with gemcitabine for 24 hours, the cleavage of caspase-9 was detected by Western blot analysis. ***, P < 0.001. All bar plot data are means ± SD.

Close modal

To determine whether arginine methylation at specific sites affects HSP70 function in gemcitabine resistance, we transfected two single arginine-to-alanine mutants, R416A and R447A, as well as a double mutant (Mut), into HSP70-null cells. Wild-type HSP70 slightly promoted cell growth (Supplementary Fig. S6C) and also rescued cell drug resistance to a similar level to that of parental MiaPaCa-2 cells, whereas R416A and R447A mutants could only rescue about half of the drug resistance level, and the double-mutant protein had no ability to rescue drug resistance (Fig. 6C).

Because these two arginine residues are PRMT1 methylation sites, we manipulated PRMT1 expression in FG cells and assessed their resistance to gemcitabine. As expected, overexpression of PRMT1 increased gemcitabine resistance and knockdown of PRMT1 decreased gemcitabine resistance in FG cells (Fig. 6D and E). Consistently, DB75 also sensitized PDAC cells to gemcitabine-induced cell death (Fig. 6F). These findings indicate that PRMT1-mediated methylation of arginine 416 and 447 is essential for drug resistance function of HSP70 in cancer cells.

To determine whether HSP70 methylation renders PDAC cells resistant to other cellular stress conditions, we reconstituted tumor cells with either wild-type or mutant HSP70 and then exposed those cells to etoposide (a DNA-damaging reagent), hydrogen peroxide (to induce oxidative stress), or glucose deprivation (to mimic nutrient limitation). Wild-type HSP70 had a clearly better protective function against various stresses than did mutant HSP70 (Fig. 6GI). As expected, wild-type HSP70 protected cells from apoptosis, whereas the methylation-deficient HSP70 showed no protection on stress-induced apoptosis (Fig. 6J); and wild-type HSP70 inhibited the stress-induced cleavage of caspase-9, whereas this protective effect was abolished by R-to-A mutation of HSP70 (Fig. 6K).

Arginine methylation of HSP70 enhances its target mRNA-binding and -stabilizing ability

HSP70 is canonically known as a molecular chaperone that facilitates protein folding during protein synthesis and prevents the aggregation of proteins when cells are exposed to homeostatic challenges (5). However, HSP70 was later found to be able to bind AREs of certain mRNAs and regulate mRNA decay, although HSP70 does not have a typical nucleic acid binding domain (7, 9, 36). Interestingly, a large number of PRMT substrates are associated with RNA processing (15, 37), which prompted us to determine whether PRMT1-mediated HSP70 methylation is involved in its RNA binding and stabilization activity.

BCL2 is a major antiapoptotic protein, which is overexpressed in a variety of cancers and associated with resistance to chemotherapy (38, 39). BCL2 mRNA has a long 3′-UTR that contains several AREs (40, 41). We first altered HSP70 expression in MDA28 cells and found that knockdown of HSP70 markedly reduced BCL2 expression, whereas wild-type HSP70 increased BCL2 protein levels (Supplementary Fig. S7A). In addition, reconstitution of wild-type, but not mutant HSP70, increased BCL2 in HSP70-null cells (Fig. 7A).

Figure 7.

PRMT1-mediated methylation enhances HSP70 binding and stabilizing of BCL2 mRNA. A, Methylation-deficient HSP70 failed to induce antiapoptotic BCL2 protein. B, Wild-type (WT) and HSP70-KO MiaPaCa-2 cells (left) and HSP70-KO cells reconstituted with WT or mutant (Mut) HSP70 (right) were treated with gemcitabine (Gem). Expression of BCL2 was determined by Western blot analysis. C, PRMT1-mediated HSP70 methylation is required for BCL2 protein induction after treatment with gemcitabine. BCL2 protein was detected by Western blot analysis in HSP70-KO MiaPaCa-2 cells cotransfected with HSP70 and PRMT1 with or without gemcitabine. Myc, Myc-HSP70 or Myc-tagged HSP70 protein. D, PRMT1-mediated methylation of HSP70 increased BCL2 mRNA levels in cells treated with gemcitabine. BCL2 mRNA levels were determined by qRT-PCR in HSP70-knockout MiaPaCa-2 cells cotransfected with HSP70 and PRMT1 and treated with gemcitabine. E, mRNA time course decay assay. WT HSP70 stabilized BCL2 mRNA in cells treated with actinomycin D. Sample errors were obtained and regression analysis was performed as described in Materials and Methods. F–H, A ribonucleoprotein immunoprecipitation (RNP-IP) assay was conducted in HSP70-KO MiaPaCa-2 cells. mRNA for BCL2 (lanes 2, 6, and 10 for primer pair 1 and lanes 3, 7, and 11 for primer pair 2), VEGFA (lanes 4, 8, and 12), and GAPDH (lanes 5, 9, and 13) were determined by reverse-transcriptase PCR. Lane 1, 100-base pair DNA ladder. Equal loading and immunoprecipitation efficiency are shown in G and gel band quantification is shown in H. I, WT HSP70 protein had better binding ability to BCL2 mRNA than did methylation-deficient HSP70. The electrophoretic mobility shift assay was performed using biotin-labeled BCL2 RNA probe and different amounts of WT and mutant HSP70 protein. The open arrowhead indicates the position of unbound free probe and the solid arrowhead indicates bands corresponding to RNA–protein complexes. J, HSP70 stabilized mRNA through binding to the AU-rich elements in the 3′-UTR of BCL2. The first 557 base pairs of the 3′-UTR sequence of human BCL2 mRNA were cloned into a pMIR-REPORT vector. The recombinant vector was cotransfected with WT or mutant HSP70 for 36 hours, and luciferase activity was measured. **, P < 0.01; ***, P < 0.001; NS, nonsignificant. All bar plot data are means ± SD.

Figure 7.

PRMT1-mediated methylation enhances HSP70 binding and stabilizing of BCL2 mRNA. A, Methylation-deficient HSP70 failed to induce antiapoptotic BCL2 protein. B, Wild-type (WT) and HSP70-KO MiaPaCa-2 cells (left) and HSP70-KO cells reconstituted with WT or mutant (Mut) HSP70 (right) were treated with gemcitabine (Gem). Expression of BCL2 was determined by Western blot analysis. C, PRMT1-mediated HSP70 methylation is required for BCL2 protein induction after treatment with gemcitabine. BCL2 protein was detected by Western blot analysis in HSP70-KO MiaPaCa-2 cells cotransfected with HSP70 and PRMT1 with or without gemcitabine. Myc, Myc-HSP70 or Myc-tagged HSP70 protein. D, PRMT1-mediated methylation of HSP70 increased BCL2 mRNA levels in cells treated with gemcitabine. BCL2 mRNA levels were determined by qRT-PCR in HSP70-knockout MiaPaCa-2 cells cotransfected with HSP70 and PRMT1 and treated with gemcitabine. E, mRNA time course decay assay. WT HSP70 stabilized BCL2 mRNA in cells treated with actinomycin D. Sample errors were obtained and regression analysis was performed as described in Materials and Methods. F–H, A ribonucleoprotein immunoprecipitation (RNP-IP) assay was conducted in HSP70-KO MiaPaCa-2 cells. mRNA for BCL2 (lanes 2, 6, and 10 for primer pair 1 and lanes 3, 7, and 11 for primer pair 2), VEGFA (lanes 4, 8, and 12), and GAPDH (lanes 5, 9, and 13) were determined by reverse-transcriptase PCR. Lane 1, 100-base pair DNA ladder. Equal loading and immunoprecipitation efficiency are shown in G and gel band quantification is shown in H. I, WT HSP70 protein had better binding ability to BCL2 mRNA than did methylation-deficient HSP70. The electrophoretic mobility shift assay was performed using biotin-labeled BCL2 RNA probe and different amounts of WT and mutant HSP70 protein. The open arrowhead indicates the position of unbound free probe and the solid arrowhead indicates bands corresponding to RNA–protein complexes. J, HSP70 stabilized mRNA through binding to the AU-rich elements in the 3′-UTR of BCL2. The first 557 base pairs of the 3′-UTR sequence of human BCL2 mRNA were cloned into a pMIR-REPORT vector. The recombinant vector was cotransfected with WT or mutant HSP70 for 36 hours, and luciferase activity was measured. **, P < 0.01; ***, P < 0.001; NS, nonsignificant. All bar plot data are means ± SD.

Close modal

Next, we found that treatment with low-dose gemcitabine could dose-dependently induce BCL2 expression in PDAC cells (Supplementary Fig. S7C, top). Moreover, when wild-type and HSP70-depleted cells were treated with gemcitabine, BCL2 was dramatically induced in wild-type cells but only minimally elevated in HSP70-depleted cells (Fig. 7B, left). Re-expression of wild-type HSP70 could rescue the induction of BCL2 after treatment with gemcitabine (Fig. 7B, right). We further observed that HSP70-mediated induction of BCL2 was abolished by either R-to-A mutation of HSP70 or the expression of a dominant-negative PRMT1 that lacks methyltransferase activity (Fig. 7C).

Finally, we determined the effect of HSP70 methylation on BCL2 mRNA expression levels. A qRT-PCR assay revealed that transfection with wild-type HSP70 could increase BCL2 mRNA levels in PDAC cells (Supplementary Fig. S7B). HSP70 could increase the steady-state level of BCL2 mRNA after treatment with gemcitabine, and PRMT1 overexpression enhanced HSP70-mediated BCL2 mRNA induction (Fig. 7D). Interestingly, a selective BCL2 inhibitor, venetoclax, could also sensitize PDAC cells to gemcitabine (Supplementary Fig. S7C, bottom), which is similar to our result of PRMT1 inhibitor.

To determine whether HSP70 methylation influences BCL2 mRNA stability, we measured BCL2 mRNA levels in time course after treatment with actinomycin D to block new mRNA synthesis. In cells transfected with wild-type HSP70, BCL2 mRNA decayed slowly, with a half-life of 5.07 ± 0.36 hours, whereas the decay rate of BCL2 transcript was faster in cells transfected with mutant HSP70 (t1/2 = 3.02 ± 0.29 hours; Fig. 7E). To determine whether the decay rate difference was due to the altered mRNA binding ability of HSP70, we conducted a ribonucleoprotein immunoprecipitation coupled reverse transcriptase PCR assay. With equal amounts of input mRNA and protein, wild-type HSP70 could bind more BCL2 mRNA than could mutant HSP70 (Fig. 7FH; Supplementary Fig. S7E). VEGFA mRNA, a positive control in our study that was previously reported to be bound by HSP70 (8), was also more enriched in wild-type than mutant HSP70 precipitates.

Analyzing the sequence of BCL2 mRNA 3′-UTR revealed that the first 400 base pairs proximal to the stop codon were highly conserved between human and mouse, and there were three typical AREs within this region (Supplementary Fig. S7D). To test the direct binding of HSP70 with AREs of BCL2 mRNA, we performed an electrophoretic mobility shift assay using purified wild-type and mutant HSP70 proteins and biotin-labeled RNA probes. Wild-type HSP70 had a higher binding affinity with ARE-containing RNA probes than did mutant HSP70 (Fig. 7I). To confirm the binding sites within this sequence, we cloned a 557-base pair fragment containing three AREs in the 3′-UTR of BCL2 mRNA for a luciferase assay (Supplementary Fig. S7F). HSP70 increased luciferase activity when this fragment was used as the 3′-UTR of luciferase gene, whereas the increase was abolished when AREs were mutated (Fig. 7J; Supplementary Fig. S7G). Collectively, these experiments established that methylation of R416 and R447 residues of HSP70 promoted its binding and stabilization of BCL2 mRNA through the AREs in 3′-UTR.

In this study, we provided many lines of evidence that methylation of HSP70 impacts therapeutic resistance in PDAC and is critically important in human PDAC progression. We first demonstrated that expression of PRMT1 gradually elevated during pancreatic carcinogenesis, and that overexpression of PRMT1 was associated with poor prognosis in PDAC patients. The pro-tumorigenic role of PRMT1 was shown by our in vitro and in vivo studies. In addition, we identified HSP70 as a novel substrate of PRMT1, and our study depicts a model in which PRMT1-mediated methylation of HSP70 enhances its binding and stabilization of BCL2 mRNA under stress conditions. The increased expression of BCL2 protein protects cancer cells from apoptosis induced by various cellular stresses, particularly chemo therapeutics (Supplementary Fig. S8). Therefore, we have established that the novel PRMT1-HSP70-BCL2 signaling axis is crucial to PDAC cell survival and drug resistance. Targeted inhibition of this axis could represent a new therapeutic strategy for PDAC.

Upregulation of PRMT1 is reported in many types of human cancers (22). In our study, overexpression of PRMT1 was observed in both human and mouse PDAC, and more precisely, a stepwise upregulation of PRMT1 was observed during cancer progression. However, the mechanisms underlying PRMT1 upregulation in cancers are not understood. Our analysis of a public PDAC dataset revealed that PRMT1 gene amplification occurs in more than 60% of all cases. However, in The Cancer Genome Atlas dataset, the rates of copy number change and somatic mutations of PRMT1 gene are not very high in human PDAC patients. Moreover, deep sequencing of the cancer genome revealed that among all PRMTs, only PRMT8 is usually mutated (22). Therefore, genetic alteration may only be a minor reason for PRMT1 overexpression in cancers. In PDAC, the gradually elevated expression of PRMT1 is consistent with disease progression from low-grade PanIN to high-grade PanIN and finally PDAC. During this process, some driver mutations, such as KRAS, CDKN2A/p16, p53, and SMAD4, are accumulated (23). Thus, whether PRMT1 upregulation is a consequence of these genetic alterations warrants further study.

Additionally, our results showed that treatment with TNF-α induced PRMT1 expression in PDAC cells. Similar inflammation-induced upregulation of PRMT1 has been reported in a lung inflammation model (42, 43). Given that chronic pancreatitis is a well-documented risk factor associated with PDAC, inflammation may also contribute to PRMT1 upregulation during disease development.

A large number of PRMT1 substrates are nucleic acid binding proteins, including proteins involved in DNA damage repair and mRNA splicing (22). Our study demonstrates that HSP70, a classic chaperone protein, is a novel substrate of PRMT1. Several recent studies have suggested that post-translational modifications regulate the diverse HSP70 cellular functions mainly by changing HSP70 chaperoning substrates or co-factors (14, 44–46). In contrast, our findings demonstrate for the first time that PRMT1-mediated methylation of HSP70 regulates its mRNA binding and stabilization activity, which is a non-classic function of HSP70.

Binding of HSP70 to specific mRNA species was reported many years ago (7, 9, 36), but the detailed binding mechanism and its regulation are poorly studied. A recent study showed that the SBD domain of HSP70 is sufficient to bind and stabilize its target mRNAs, and RNA binding by HSP70 is independent of its peptide-binding activity (10). Interestingly, the two methylation sites discovered in our study, R416 and R447, are located in the SBD domain of HSP70, and a previous study showed that arginine methylation of an RNA-binding protein affects its RNA binding ability (47). Our results demonstrate for the first time that arginine methylation of R416 and R447 of HSP70 enhances its binding and stabilization of BCL-2 mRNA, which is then translated to an important antiapoptotic protein. Our findings expand knowledge about HSP70 function to a new realm of mRNA regulation.

Our study also demonstrates the pro-tumorigenic role of PRMT1 in PDAC, and that PRMT1-mediated HSP70 methylation enhances cancer cell resistance to therapeutic drugs, e.g., gemcitabine. Therefore, our study provides at least two rationales for the combination use of PRMT1 inhibitor and gemcitabine to treat PDAC. First, inhibition of PRMT1 enzymatic activity itself can reduce tumor growth. This is shown not only in our study but also in another study using the same inhibitor (DB75) to reduce leukemia cell growth (26). Second, inhibition of PRMT1 can potentiate the cytotoxicity of gemcitabine by reducing cancer cell resistance to therapeutic drugs via HSP70-BCL2 pathway. Therefore, inhibition of PRMT1 may alleviate cancer cell therapeutic resistance. However, there are still some issues that people should carefully consider when targeting PRMT1 for cancer treatment. The first is the cell survival/growth dependency on PRMT1 and possible adverse effects of PRMT1 inhibitors to normal cells. The second is the selectivity of PRMT1 inhibitors. The inhibitor tools used in our study may have polypharmacological effects beyond PRMT1 inhibition. Therefore, to develop PRMT1 inhibitors with improved selectivity is important, while targeting PRMT1 downstream components, e.g., BCL2, may be an alternative to overcome this limitation.

In conclusion, our study identified HSP70 as a novel substrate of PRMT1 and demonstrated that PRMT1-mediated arginine methylation of HSP70 is crucial for its mRNA-binding and -stabilizing activity. Our study not only reveals a novel PRMT1-HSP70-BCL2 regulatory pathway that plays an important role in PDAC development, progression, and therapeutic resistance, but also suggests a potential new therapeutic approach to be developed for the treatment of PDAC.

L. Wang reports grants from NIH during the conduct of the study. D. Wei reports grants from NIH during the conduct of the study and grants from Elsa U. Pardee Foundation outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

L. Wang: Conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft, writing-review and editing. Z. Jia: Data curation, formal analysis, supervision, investigation and methodology. D. Xie: Data curation, investigation and methodology. T. Zhao: Data curation, software, investigation, visualization and methodology. Z. Tan: Data curation, software and visualization. S. Zhang: Data curation, software, investigation, visualization and methodology. F. Kong: Data curation, investigation and methodology. D. Wei: Formal analysis, supervision, investigation and methodology. K. Xie: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.

This work is dedicated to Dr. Xie's beloved father, Yongming Xie, who fell sick in August 2018 and passed away on February 6, 2019, whereas Dr. Xie was prevented from visiting his father in September 2018 until too late. This work was supported in part by the NIH grants R01CA173322, R01CA195651, R01CA198090, and R01CA220236-01 (to K. Xie). We thank Erica Goodoff for editorial comments and the proteomic core facility for protein methylation analysis, which was supported in part by Cancer Prevention Research Institute of Texas (CPRIT) grant no. RP130397 and NIH grant no. 1S10OD012304-01. We thank Mark T. Bedford and Darren E. Richard for providing some PRMT1 constructs.

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.

1.
Rahib
L
,
Smith
BD
,
Aizenberg
R
,
Rosenzweig
AB
,
Fleshman
JM
,
Matrisian
LM
. 
Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States
.
Cancer Res
2014
;
74
:
2913
21
.
2.
Kamisawa
T
,
Wood
LD
,
Itoi
T
,
Takaori
K
. 
Pancreatic cancer
.
Lancet
2016
;
388
:
73
85
.
3.
Schniewind
B
,
Christgen
M
,
Kurdow
R
,
Haye
S
,
Kremer
B
,
Kalthoff
H
, et al
Resistance of pancreatic cancer to gemcitabine treatment is dependent on mitochondria-mediated apoptosis
.
Int J Cancer
2004
;
109
:
182
8
.
4.
Hageman
J
,
van Waarde
MA
,
Zylicz
A
,
Walerych
D
,
Kampinga
HH
. 
The diverse members of the mammalian HSP70 machine show distinct chaperone-like activities
.
Biochem J
2011
;
435
:
127
42
.
5.
Wu
J
,
Liu
T
,
Rios
Z
,
Mei
Q
,
Lin
X
,
Cao
S
. 
Heat shock proteins and cancer
.
Trends Pharmacol Sci
2017
;
38
:
226
56
.
6.
Murphy
ME
. 
The HSP70 family and cancer
.
Carcinogenesis
2013
;
34
:
1181
8
.
7.
Henics
T
,
Nagy
E
,
Oh
HJ
,
Csermely
P
,
von Gabain
A
,
Subjeck
JR
. 
Mammalian Hsp70 and Hsp110 proteins bind to RNA motifs involved in mRNA stability
.
J Biol Chem
1999
;
274
:
17318
24
.
8.
Kishor
A
,
Tandukar
B
,
Ly
YV
,
Toth
EA
,
Suarez
Y
,
Brewer
G
, et al
Hsp70 is a novel posttranscriptional regulator of gene expression that binds and stabilizes selected mRNAs containing AU-rich elements
.
Mol Cell Biol
2013
;
33
:
71
84
.
9.
Laroia
G
,
Cuesta
R
,
Brewer
G
,
Schneider
RJ
. 
Control of mRNA decay by heat shock-ubiquitin-proteasome pathway
.
Science
1999
;
284
:
499
502
.
10.
Kishor
A
,
White
EJF
,
Matsangos
AE
,
Yan
Z
,
Tandukar
B
,
Wilson
GM
. 
Hsp70′s RNA-binding and mRNA-stabilizing activities are independent of its protein chaperone functions
.
J Biol Chem
2017
;
292
:
14122
33
.
11.
Cloutier
P
,
Coulombe
B
. 
Regulation of molecular chaperones through post-translational modifications: decrypting the chaperone code
.
Biochim Biophys Acta
2013
;
1829
:
443
54
.
12.
Silver
JT
,
Noble
EG
. 
Regulation of survival gene hsp70
.
Cell Stress Chaperones
2012
;
17
:
1
9
.
13.
Jakobsson
ME
,
Moen
A
,
Bousset
L
,
Egge-Jacobsen
W
,
Kernstock
S
,
Melki
R
, et al
Identification and characterization of a novel human methyltransferase modulating Hsp70 protein function through lysine methylation
.
J Biol Chem
2013
;
288
:
27752
63
.
14.
Gao
WW
,
Xiao
RQ
,
Peng
BL
,
Xu
HT
,
Shen
HF
,
Huang
MF
, et al
Arginine methylation of HSP70 regulates retinoid acid-mediated RARbeta2 gene activation
.
Proc Natl Acad Sci U S A
2015
;
112
:
E3327
36
.
15.
Bedford
MT
,
Clarke
SG
. 
Protein arginine methylation in mammals: who, what, and why
.
Mol Cell
2009
;
33
:
1
13
.
16.
Bannister
AJ
,
Kouzarides
T
. 
Regulation of chromatin by histone modifications
.
Cell Res
2011
;
21
:
381
95
.
17.
Lee
YH
,
Stallcup
MR
. 
Minireview: protein arginine methylation of nonhistone proteins in transcriptional regulation
.
Mol Endocrinol
2009
;
23
:
425
33
.
18.
Liao
HW
,
Hsu
JM
,
Xia
W
,
Wang
HL
,
Wang
YN
,
Chang
WC
, et al
PRMT1-mediated methylation of the EGF receptor regulates signaling and cetuximab response
.
J Clin Invest
2015
;
125
:
4529
43
.
19.
Wei
H
,
Mundade
R
,
Lange
KC
,
Lu
T
. 
Protein arginine methylation of non-histone proteins and its role in diseases
.
Cell Cycle
2014
;
13
:
32
41
.
20.
Bedford
MT
,
Richard
S
. 
Arginine methylation an emerging regulator of protein function
.
Mol Cell
2005
;
18
:
263
72
.
21.
Lake
AN
,
Bedford
MT
. 
Protein methylation and DNA repair
.
Mutat Res
2007
;
618
:
91
101
.
22.
Yang
Y
,
Bedford
MT
. 
Protein arginine methyltransferases and cancer
.
Nat Rev Cancer
2013
;
13
:
37
50
.
23.
Ryan
DP
,
Hong
TS
,
Bardeesy
N
. 
Pancreatic adenocarcinoma
.
N Engl J Med
2014
;
371
:
1039
49
.
24.
Hingorani
SR
,
Wang
L
,
Multani
AS
,
Combs
C
,
Deramaudt
TB
,
Hruban
RH
, et al
Trp53R172H and KrasG12D cooperate to promote chromosomal instability and widely metastatic pancreatic ductal adenocarcinoma in mice
.
Cancer Cell
2005
;
7
:
469
83
.
25.
Tuveson
DA
,
Shaw
AT
,
Willis
NA
,
Silver
DP
,
Jackson
EL
,
Chang
S
, et al
Endogenous oncogenic K-ras(G12D) stimulates proliferation and widespread neoplastic and developmental defects
.
Cancer Cell
2004
;
5
:
375
87
.
26.
Yan
L
,
Yan
C
,
Qian
K
,
Su
H
,
Kofsky-Wofford
SA
,
Lee
WC
, et al
Diamidine compounds for selective inhibition of protein arginine methyltransferase 1
.
J Med Chem
2014
;
57
:
2611
22
.
27.
Witkiewicz
AK
,
McMillan
EA
,
Balaji
U
,
Baek
G
,
Lin
WC
,
Mansour
J
, et al
Whole-exome sequencing of pancreatic cancer defines genetic diversity and therapeutic targets
.
Nat Commun
2015
;
6
:
6744
.
28.
Helm
O
,
Held-Feindt
J
,
Grage-Griebenow
E
,
Reiling
N
,
Ungefroren
H
,
Vogel
I
, et al
Tumor-associated macrophages exhibit pro- and anti-inflammatory properties by which they impact on pancreatic tumorigenesis
.
Int J Cancer
2014
;
135
:
843
61
.
29.
Bissinger
EM
,
Heinke
R
,
Spannhoff
A
,
Eberlin
A
,
Metzger
E
,
Cura
V
, et al
Acyl derivatives of p-aminosulfonamides and dapsone as new inhibitors of the arginine methyltransferase hPRMT1
.
Bioorg Med Chem
2011
;
19
:
3717
31
.
30.
Welch
WJ
,
Feramisco
JR
. 
Nuclear and nucleolar localization of the 72,000-dalton heat shock protein in heat-shocked mammalian cells
.
J Biol Chem
1984
;
259
:
4501
13
.
31.
Zhang
X
,
Cheng
X
. 
Structure of the predominant protein arginine methyltransferase PRMT1 and analysis of its binding to substrate peptides
.
Structure
2003
;
11
:
509
20
.
32.
Kozakov
D
,
Hall
DR
,
Xia
B
,
Porter
KA
,
Padhorny
D
,
Yueh
C
, et al
The ClusPro web server for protein-protein docking
.
Nat Protoc
2017
;
12
:
255
78
.
33.
Pierce
BG
,
Wiehe
K
,
Hwang
H
,
Kim
BH
,
Vreven
T
,
Weng
Z
. 
ZDOCK server: interactive docking prediction of protein-protein complexes and symmetric multimers
.
Bioinformatics
2014
;
30
:
1771
3
.
34.
Aghdassi
A
,
Phillips
P
,
Dudeja
V
,
Dhaulakhandi
D
,
Sharif
R
,
Dawra
R
, et al
Heat shock protein 70 increases tumorigenicity and inhibits apoptosis in pancreatic adenocarcinoma
.
Cancer Res
2007
;
67
:
616
25
.
35.
Beere
HM
. 
Death versus survival: functional interaction between the apoptotic and stress-inducible heat shock protein pathways
.
J Clin Invest
2005
;
115
:
2633
9
.
36.
Wilson
GM
,
Sutphen
K
,
Bolikal
S
,
Chuang
KY
,
Brewer
G
. 
Thermodynamics and kinetics of Hsp70 association with A + U-rich mRNA-destabilizing sequences
.
J Biol Chem
2001
;
276
:
44450
6
.
37.
Pahlich
S
,
Zakaryan
RP
,
Gehring
H
. 
Protein arginine methylation: cellular functions and methods of analysis
.
Biochim Biophys Acta
2006
;
1764
:
1890
903
.
38.
Delbridge
AR
,
Strasser
A
. 
The BCL-2 protein family, BH3-mimetics and cancer therapy
.
Cell Death Differ
2015
;
22
:
1071
80
.
39.
Tomek
M
,
Akiyama
T
,
Dass
CR
. 
Role of Bcl-2 in tumour cell survival and implications for pharmacotherapy
.
J Pharm Pharmacol
2012
;
64
:
1695
702
.
40.
Bandyopadhyay
S
,
Sengupta
TK
,
Fernandes
DJ
,
Spicer
EK
. 
Taxol- and okadaic acid-induced destabilization of bcl-2 mRNA is associated with decreased binding of proteins to a bcl-2 instability element
.
Biochem Pharmacol
2003
;
66
:
1151
62
.
41.
Bevilacqua
A
,
Ceriani
MC
,
Schiavone
N
,
Nicolin
A
. 
BCL-2 regulation targeting the AU-rich domain
.
Nucleosides Nucleotides Nucleic Acids
2000
;
19
:
1273
9
.
42.
Liu
L
,
Sun
Q
,
Bao
R
,
Roth
M
,
Zhong
B
,
Lan
X
, et al
Specific regulation of PRMT1 expression by PIAS1 and RKIP in BEAS-2B epithelia cells and HFL-1 fibroblasts in lung inflammation
.
Sci Rep
2016
;
6
:
21810
.
43.
Sun
Q
,
Liu
L
,
Roth
M
,
Tian
J
,
He
Q
,
Zhong
B
, et al
PRMT1 upregulated by epithelial proinflammatory cytokines participates in COX2 expression in fibroblasts and chronic antigen-induced pulmonary inflammation
.
J Immunol
2015
;
195
:
298
306
.
44.
Cho
HS
,
Shimazu
T
,
Toyokawa
G
,
Daigo
Y
,
Maehara
Y
,
Hayami
S
, et al
Enhanced HSP70 lysine methylation promotes proliferation of cancer cells through activation of Aurora kinase B
.
Nat Commun
2012
;
3
:
1072
.
45.
Seo
JH
,
Park
JH
,
Lee
EJ
,
Vo
TT
,
Choi
H
,
Kim
JY
, et al
ARD1-mediated Hsp70 acetylation balances stress-induced protein refolding and degradation
.
Nat Commun
2016
;
7
:
12882
.
46.
Truman
AW
,
Kristjansdottir
K
,
Wolfgeher
D
,
Hasin
N
,
Polier
S
,
Zhang
H
, et al
CDK-dependent Hsp70 Phosphorylation controls G1 cyclin abundance and cell-cycle progression
.
Cell
2012
;
151
:
1308
18
.
47.
Denman
RB
. 
Methylation of the arginine-glycine-rich region in the fragile X mental retardation protein FMRP differentially affects RNA binding
.
Cell Mol Biol Lett
2002
;
7
:
877
83
.

Supplementary data