Introduction

The majority of breast tumors (70%) express estrogen receptor α (ERα) which is successfully targeted by adjuvant therapies that increase overall survival1. The current standard adjuvant treatments for patients with ERα+ breast cancer either inhibit ERα activity, e.g., tamoxifen (TAM) for premenopausal women, or block the conversion of androgens to estrogens by aromatase inhibitors (AIs), e.g., letrozole, in postmenopausal women2. Unfortunately, endocrine therapies are limited by the development of acquired endocrine resistance in ~30–40% of initially responsive patients that can occur up to 30 years after primary therapy3,4. A variety of mechanisms have been implicated in TAM-resistance (TAM-R)5,6, including altered microRNA (miRNA) and long noncoding RNA (lncRNA) expression7,8,9. Most miRNAs are transcribed, by RNA polymerase II, either as introns of host genes or as independent genes called primary (pri)-miRNAs10. Pri-miRNAs are processed by the DROSHA-DGCR8 microprocessor complex to precursor (pre)-miRNAs prior to nuclear export11. In the cytoplasm, the double stranded pre-miRNA is unwound by the DICER-TRBP complex to incorporate one strand of the miRNA (called the guide strand) into the RNA induced silencing (RISC) complex containing the catalytic Argonaut proteins, e.g., AGO212. By basepairing with nucleotides in the 3′UTR of target genes within RISC, miRNAs can act as either oncomiRs by reducing protein levels of tumor suppressors or as tumor suppressors by decreasing oncogenic proteins in breast tumors7. The processing of pri-miRNA transcripts is regulated in part by post-transcriptional modifications (PTMs) of pri-miRNA13.

Next-generation sequencing (NGS) and mass spectrometry identified N(6)-methyladenosine (m6A) as the most common modification of mRNA and lncRNAs14,15. m6A plays a role in pre-mRNA processing, alternative splicing, nuclear export, stability, and translation16,17 by acting as a ‘conformational marker’ that induces sequence-dependent outcomes in RNA remodeling18. A recent report also identified higher m6A in selected pri-miRNA sequences that corresponded with increased levels of the corresponding mature miRNA in MDA-MB-231 triple negative breast cancer (TNBC) cells13.

m6A methylation is added by the RNA methyltransferase complex (WTAP, METTL3, METTL14, VIRMA, and RBM15), removed by the dioxygenases FTO and ALKBH5, and recognized by a variety of ‘readers’, including YTHDF1, YTHDF2, and HNRNPA2/B119,20,21. METTL3 methylation of m6A on pri-miRNAs13 and RNA-dependent interaction of HNRNPA2/B1 with DGCR8, a component of the DROSHA complex, stimulate processing of selected pri-miRNA-m6A to precursor miRNA (pre-miRNA)22. HNRNPA2/B1 transcript expression is upregulated in breast tissue of postmenopausal parous women23, but its role in the protective effect of early pregnancy on postmenopausal ERα+ breast cancer is unknown24. HNRNPA2/B1 protein expression was higher in breast tumors compared to normal breast and knockdown of HNRNPA2/B1 inhibited the proliferation of MCF-7 and MDA-MB-231 breast cancer cells by causing S phase arrest and apoptosis25.

HNRNPA2 and HNRNPB1 are two splice isoforms transcribed from the same locus but are traditionally treated as a single protein26. HNRNPB1 is a lower abundance (~5%) N-terminal splice variant of the more highly expressed HNRNPA2 isoform and contains an additional 12 aa encoded by exon 227. HNRNPA2/B1 share the remaining protein structure including an RNA-binding domain containing two RNA recognition motifs (RRMs) separated by a 15 aa linker and a C-terminal Gly-rich, low complexity region with a prion-like domain (PrLD), RGG box, and Py-motif including M9 nuclear localization signal28. In addition to its recognition of m6A in pri-miRNA and role in RNA splicing and processing29, HNRNPA2/B1 is involved in DNA repair30 and genome stability31.

In MCF-7 ERα+ breast cancer cells, enhanced cross-linking immunoprecipitation (eCLIP) using antibodies specific to HNRNPB1 alone or HNRNPA2/B1 in combination identified 1,472 transcripts bound by both HNRNPB1 and HNRNPA2/B1, 899 transcripts uniquely bound by HNRNPB1, and 479 transcripts uniquely bound by HNRNPA2/B132. HNRNPB1 binding sites revealed a preference for 5′-AGGAAGG-3′ versus 5′-UGGGGA-3′ for HNRNPA2/B132. HNRNPA2/B1 binding peaks were primarily in chromatin samples, consistent with HNRNPA2/B1 binding to nascent transcripts32.

Here we identified HNRNPA2/B1 expression to be higher in LCC9 and LY2 endocrine-resistant cells compared to parental MCF-7 luminal A breast cancer cells. We used miRNA-seq to identify differences in miRNA transcripts in MCF-7 cells when HNRNPA2/B1 is overexpressed and evaluated the pathways and mRNA targets associated with each misregulated miRNA for relevance to breast cancer and endocrine resistance. Progesterone receptor (PR) action in breast cancer and TGFβ signaling via miRNA in breast cancer were identified as pathways downstream of the upregulated miRNAs, and TGFβ signaling via SMADs and activation of Notch signaling were identified as pathways downstream of the downregulated miRNAs. TGFβ signaling, response to estradiol, and cell-substrate adhesion were pathways associated with mRNA targets of the identified miRNAs. Accordingly, overexpression of HNRNPA2/B1 in MCF-7 cells reduced their sensitivity to 4-hydroxytamoxifen and fulvestrant, indicating that increased HNRNPA2/B1 plays a role in tamoxifen and fulvestrant resistant cell proliferation.

Results and Discussion

Expression of RNA writers, readers, and erasers in breast cancer cells

TAM/fulvestrant-resistant LCC9 breast cancer cells have higher levels of expression of diverse miRNAs compared with parental, TAM-sensitive MCF-7 cells33. To determine if there are differences in the expression of the genes encoding the readers, writers, and erasers of reversible m6A RNA modification19 between MCF-7 and LCC9 cells, we examined the steady state transcript levels of m6A writers (WTAP, METTL3, and METTL14), readers (YTHDF1, YTHDF2, YTHDF3, and HNRNPA2/B1) and erasers (FTO and ALKBH5) in RNA-seq data from our previous RNA-seq study, GEO accession number GSE8162034 (Fig. 1A). The expression of METTL3 and YTHDF1 transcripts was lower in LCC9 than MCF-7 cells whereas WTAP, FTO, ALKBH5, and HNRNPA2/B1 were higher in LCC9 than MCF-7 cells. The possible role of the expression of METTL3, YTHDF1, WTAP, FTO, and HNRNPA2/B1 transcripts in human breast tumors on overall survival was examined using the online tool Kaplan-Meier Plotter35. There was no association of overall survival (OS) for breast cancer patients based on primary tumor expression of METTL3, YTHDF1, or WTAP (Supplementary Fig. 1). Low expression of FTO was associated with lower OS (Supplementary Fig. 2A). However, higher FTO nuclear staining was reported in ER-/PR-/HER2+ breast tumors36. Patients with ER-/PR-/HER2+ breast tumors have ~40% lower disease-free survival compared to women with luminal A breast tumors37. HNRNPA2/B1 transcript expression was higher than any of the other genes examined in the m6A pathway (Fig. 1B). HNRNPA2/B1 protein expression was also ~2.6-fold higher in LCC9 and LY2 cells than MCF-7 cells (Fig. 1C,D, Supplementary Fig. 3). Kaplan-Meier (K-M) survival analysis showed that higher expression of HNRNPA2/B1 is associated with lower OS to ~150 months (Supplementary Fig. 2B). After ~220 months, the black line denoting high HNRNPA2B expression indicates reduced survival for those 3 patients in the K-M plot (Supplementary Fig. 2B). More data are needed to better understand whether low HNRNPA2B1 in the primary tumor predicts reduced OS after ~220 months. Thus, because of the high expression of HNRNPA2B1 at the transcript and protein levels in LCC9 endocrine-resistant cells, its association with lower survival, and its role in increasing pri-miRNA processing22, we selected HNRNPA2B1 for further study.

Figure 1
figure 1

Expression of the genes encoding the readers, writers, and erasers of reversible m6A RNA modification. (A,B) Data are from a previous RNA-seq experiment in MCF-7 and LCC9 cells (GEO GSE81620). Data are the average of three replicate experiments +/− SEM. with FPKM = fragments Per Kilobase of transcript per Million mapped reads. *P < 0.05 in a two-tailed student’s t test. (C) Representative western blot of HNRNPA2B1 protein expression in WCE from MCF-7 and LCC9 cells. The blot was stripped and reprobed for GAPDH. The numerical values are HNRNPA2B1/GAPDH in these blots. The full-length blot of GAPDH is shown in Supplementary Fig. 1C. (D) Summary of relative HNRNPA2B1 protein expression in LCC9 and LY2 cells compared to MCF-7 parental cells. P < 0.05, One way ANOVA followed by Tukey’s test.

Full size image

miRNA-seq analysis of HNRNPA2/B1-regulated miRNAs in MCF-7 cells

Based on our observation of higher HNRNPA2/B1 in LCC9 compared to MCF-7 cells, we hypothesized that the overexpression of HNRNPA2/B1 in LCC9 cells promotes processing of pri-miRNAs resulting in increased pre- and mature miRNAs that act on targets and pathways to promote endocrine resistance. We note that HNRNPA2/B1 upregulated miR-99a, miR-125b, and miR-149 in MDA-MB-231 TNBC cells22, and we reported higher levels of miR-125b and miR-149, but not miR-99a, in LY2 endocrine resistant breast cancer cells as compared to MCF-7 cells in an earlier study38. To evaluate the effect of increased HNRNPA2/B1 on mature miRNA expression in breast cancer, MCF-7 cells were transiently transfected with a control vector for 48 h or an expression vector for HNRNPA2/B1 for 48 or 72 h (Fig. 2A). A limitation of this analysis was that a 72 h control-transfected group was not included. We did not detect differences in control gene (GAPDH) expression between 48 and 72 h control-transfected samples (Supplementary Fig. 3E). However, complete RNA transcriptome analysis of the 72 h control-transfected MCF-7 cells would have been a better control for the 72 h HNRNPA2/B1-transfected cells.

Figure 2
figure 2

HNRNPA2B1 overexpression in MCF-7 cells. (A) The ΔCT values for HNRNPA2B1 normalized to 18 S of each of the six samples used for RNA se. MCF-7 cells were transfected with pCDNA3 control or pCDNA-3-HNRNPA2B1. Each point is the mean of triplicate determinations within one qPCR run of these samples. *p < 0.05, One way ANOVA followed by Tukey’s test. (B) Western blot for HNRNPA2B1 in MCF-7 cells control-transfected (C) and transfected with HNRNPA2B1 for 48 h. The blot was stripped and reprobed for GAPDH. Values are the HNRNPA2B1/GAPDH in this blot. The full-length blot of GAPDH is shown in Supplementary Fig. 1D. (C) Summary of relative HNRNPA2B1 protein expression in MCF-7 cells transfected for 48 h vs. control, n = mean ± std of 7 biological replicates. P < 0.0004, two-tailed student’s t-test. (D) The heat map represents the miRNAs having a fold-change of ±4. Yellow is upregulated and purple is downregulated (scale at top). Genes were clustered based on similar expression profiles.

Full size image

The transfection resulted in average ~5 fold increase in HNRNPA2/B1 protein expression (Fig. 2B,C). miRNA was isolated from six replicate experiments 48 or 72 h after HNRNPA2/B1 transfection for global changes in the miRNA transcriptome (miRome). Supplementary Table 1 shows a summary of the sequence analysis of the samples. A heatmap shows the relative consistency of miRNA expression changes in the replicate samples within each comparison and the changes between time after HNRNPA2/B1 transfection (Supplementary Fig. 4).

Three pairwise comparisons were evaluated: 48 h versus control, 72 h versus control, and 72 h versus 48 h. In total, 795 miRNAs were differentially expressed (p ≤ 0.05). 210 (110 up and 100 down) common to both time points, 236 (148 up and 88 down) uniquely at 48 h, and 349 (177 up and 172 down) uniquely at 72 h (Table 1). The identities and values of differentially expressed miRs are shown in Supplementary Tables 2–7 for all comparisons. Note that several miRs are listed twice, due to their coding from multiple gene locations. A heatmap for differentially expressed miRs passing a fold change (FC) threshold of ±4 (Log2FC ±2) in one or more of the comparisons is shown in Fig. 2D.

Table 1 Comparison of the number of differentially expressed miRNAs using a p-value cutoff of ≤0.05.
Full size table

miRNAs upregulated in HNRNPA2/B1-transfected MCF-7 cells

Based on previous reports that HNRNPA2B1 increases processing of pri-miRNA to pre-miRNA and mature miRNAs13,22, we hypothesized that HNRNPA2/B1 overexpression would increase levels of miRNAs regulated by m6A in the respective pri-miRNA. We focus only on the miRNAs whose expression was significantly increased in response to HNRNPA2B1 transfection (Fig. 3, Tables 2, 3). Figure 3 shows that 148 and 177 miRNAs were uniquely increased at 48 and 72 h after HNRNPA2B1 transfection while 110 miRNAs were increased at both time points.

Figure 3
figure 3

Venn diagram depicting the number of different and common miRNAs identified as upregulated after transient HNRNPA2B1 overexpression in MCF-7 cells after 48 or 72 h. MetaCore Enrichment by Pathway Maps analysis of DE miRNAs upregulated after 48 h (left, # 1–5) and 72 h (right, 1–6) (both versus control) and those identified in common at 48 and 72 h (below, #1–10).

Full size image
Table 2 Fourteen miRNAs were upregulated ≥2.0-fold by transient overexpression of HNRNPA2/B1 in MCF-7 cells at 48 and 72 h.
Full size table

Fourteen miRNAs were increased by ≥2.0-log fold at both 48 and 72 h (Table 2). Of the six miRNAs on which publications were found, four (miR-1266, miR-2861, miR-7107-5p, and miR-762) have oncogenic, endocrine- or chemo-resistance activities in breast cancer (Table 2). Sixty miRNAs were increased at 48 h, but not 72 h (Table 3). Of the twenty miRNAs with publications, seven had reported oncomiR functions and six had tumor suppressor functions. HNRNPA2/B1 transfection increased miR-222-5p, which is increased in TAM-R MCF-7 cells, and its role in TAM-R and targets, including ESR1 (ERα) and cell cycle genes, e.g., CDKN1B (P27/KIP1) have been reviewed7,39. While miR-595 has no established role in breast cancer, it has tumor promoter roles in papillary thyroid carcinoma (PTC)40 and human glioblastoma (GBM) cells41. However, miR-595 acts as a tumor suppressor in ovarian cancer42; thus, its role in breast cancer remains to be determined.

Table 3 Sixty miRNAs were upregulated ≥2.0-fold by transient overexpression of HNRNPA2/B1 in MCF-7 cells at 48 h, but not at 72 h.
Full size table

Fifty-one miRNAs were increased at 72 h, but were not increased at 48 h. Of the 17 miRNAs with published information relevant to cancer, 3 miRNAs (miR-4763-3p, miR-4787-5p, and miR-4800-3p) were reported to be higher in fulvestrant-resistant MCF-7 versus TAM-resistant MCF-7 cells43.

MetaCore analysis was performed on all three groups of upregulated miRNAs (common to 48 and 72 h, unique to 48 h, and unique to 72 h (Fig. 3). Pathways identified included “PR action in breast cancer: stimulation of metastasis” (involving downregulation of miR-29a-3p) and “TGFβ signaling via miRNA in breast cancer” (involving downregulation of miR-21-5p, miR-200a-3p, miR-200a-5p, miR-200c-3p, miR-200c-5p, miR-200b-3p, upregulation of miR-181a-5p) (Fig. 4). miR-200 family members are downregulated in breast cancer and in TAM-R breast cancer cell lines and tumors (reviewed in7,44). The decrease in miR-200 family members would be expected to relieve repression of ZEB1/2 leading to repression of E-cadherin and EMT, an indicator of breast cancer progression and metastasis45. The GO processes identified included “Cellular response to estrogen stimulus” (upregulation of miR-574-5p and miR-466) (Supplementary Fig. 6, Supplementary Table 2). Increased serum of miR-574-5p is a marker of breast cancer46.

Figure 4
figure 4

Venn diagram depicting the number of different and common miRNAs identified as downregulated after transient HNRNPA2B1 overexpression in MCF-7 cells after 48 or 72 h. MetaCore Enrichment by Pathway Maps analysis of DE downregulated miRNAs after 48 h and 72 h (both versus control) identified common and unique pathways putatively regulated by the DE miRNAs.

Full size image

miRNAs downregulated in HNRNPA2/B1-transfected MCF-7 cells

Unexpectedly, we identified 88, 172, and 100 miRNAs downregulated at 48 h, 72 h, and both time points, respectively (Supplementary Fig. 5). This is the first identification of miRNAs downregulated in response to HNRNPA2/B1 overexpression. Of course, this could be a direct or indirect effect. Another HNRNP family member, HNRNPA1, can either promote or inhibit pri-miRNA processing, resulting in increased mature miR-18a47 and reduced let-7a-1 in HeLa cells48. We did not detect any significant increase in miR-18a nor a decrease in let-7a-1 in HNRNPA2/B1-transfected MCF-7 cells, implying that these two HNRNPs have different targets in different cells.

We focused on those downregulated ≥2.0-log fold: 57 at 48 h, 130 at 72 h, and 18 in common (Fig. 4, Supplementary Table 7). Nine of the 18 common downregulated miRNAs had roles in breast cancer, including miR-221-3p and miR-222-3p (both target ESR1) and miR-515-5p and miR-516-5p, which are increased in TAM-R MCF-7 cells (Table 4). Of the 57 miRNAs downregulated at 48 h (Table 5), 26 have reported roles in breast cancer. Some have roles in TAM-R. let-7i and miR-489 are downregulated in TAM-resistant breast cancer cells and miR-101, miR-221, miR-222, and miR-515 are upregulated in TAM-resistant MCF-7 and other breast cancer cells7,39,49. Both miR-29a-3p and miR-29b-3p which reduce TAM-resistant MCF-7 cell (LCC9 and LY2 cell lines) proliferation, migration, and colony formation33 were downregulated by HNRNPA2/B1 overexpression.

Table 4 Eighteen miRNAs were downregulated by transient overexpression of HNRNPA2/B1 in MCF-7 cells at 48 and 72 h.
Full size table
Table 5 Fifty-seven miRNAs were downregulated by transient overexpression of HNRNPA2/B1 in MCF-7 cells at 48 h.
Full size table

MetaCore pathway analysis identified “TGFβ signaling via SMADs in breast cancer” for the common 18 downregulated miRNAs, as well as “PR action in breast cancer: stimulation of metastasis” in the 48 h and “Activation of Notch signaling in breast cancer” in the 72 h downregulated miRNAs (Fig. 4). MetaCore enrichment analysis by GO processes identified “cellular response to estrogen stimulus (miR-206)” and “response to estrogen” (also miR-206) (Supplementary Fig. 7). E2, and the ER-selective agonist PPT suppressed miR-206 expression in MCF-7 cells50.

Identification of experimentally validated gene targets of the miRNAs differentially expressed in MCF-7 cells transfected with HNRNPA2/B1

The differentially expressed miRNAs were searched against miRTarBase51 for experimentally validated gene targets. Table 6 shows the number of differentially expressed miRNAs and the number of validated gene targets for these miRNAs. Genes identified as targets of the DE miRNAs were used as an input into categoryCompare52 to identify enriched Gene Ontology Biological Processes (GO:BP)52.

Table 6 Identification of experimentally validated gene targets of the miRNAs differentially expressed in MCF-7 cells transfected with HNRNPA2/B1.
Full size table

Processes putatively regulated by the HNRNPA2B1-regulated miRNAs in MCF-7 cells include TGFβ signaling (Fig. 5), which is protective in normal breast epithelium but acts as a tumor promoter after genetic and epigenetic changes involved in breast tumorigenesis accrue45. TGFβ induces EMT in breast cancer cells in a pathway involving tumor suppressor miR-34 family members and we observed miR-34c-5p was downregulated by HNRNPA2/B1 overexpression at 48 h (Supplementary Table 3). Other processes identified as downstream of HNRNPA2B1-regulated miRNAs included response to estrogen/estradiol, stem cell population maintenance, Wnt signaling, regulation of cell junction organization, cellular response to steroid hormone stimulus, JNK/MAPK cascade, and nuclear transport (Fig. 6). Future studies will address which targets in these pathways are downstream of HNRNPA2B1-regulation of miRNA expression and their role in endocrine-resistance.

Figure 5
figure 5

Enriched GO:BP (Gene Ontology: Biological Processes) for genes targeted by differentially expressed miRNAs at the indicated times. mRNA targets identified in miRTarBase as validated targets for the differentially expressed miRNAs at each time point were compared to control or the indicated comparison for GO:BP analysis in categoryCompare.

Full size image
Figure 6
figure 6

Enriched GO:BP (Gene Ontology: Biological Processes) for genes targeted by differentially expressed miRNAs at the indicated times. mRNA targets identified in miRTarBase as validated targets for the differentially expressed miRNAs at each time point were compared to control or the indicated comparison for GO:BP analysis in categoryCompare.

Full size image

qPCR validation of HNRNPA2/B1-downregulated miRNA targets

We selected miR-29a-3p, miR-29b-3p, and miR-222-3p for validation by qPCR based on their roles in breast cancer and responses to antiestrogen therapies7,33,34,53,54,55,56. Because 48 h HNRNPA2B1-transfected MCF-7 cells showed decreased expression of each of these miRNAs (Tables 4 and 5), we expected each miRNA to be decreased in new transient transfection experiments. Indeed, miR-29a-3p, miR-29b-3p, and miR-222-3p transcript expression was reduced by 48 h of HNRNPA2B1 transfection in MCF-7 cells, whereas transfection with the parental expression vector pcDNA3.1 had no significant effect (Fig. 7).

Figure 7
figure 7

Regulation of miR-29a-3p, miR-29b-3p, and miR-222-3p by HNRNPA2B1. MCF-7 cells were either non-transfected (control), transfected with pcDNA3.1 parental vector, or an expression vector for HNRNPA2B1 for 48 h. qPCR for (A) miR-29a-3p; (B) miR-29b-3p, and (C) miR-222-3p. Each miRNA was normalized by RNU48. Values are relative expression normalized to control-transfected cells from 11 biological replicate experiments with multiple control and transfected wells in each experiment. Data were analyzed by two way ANOVA followed by Tukey’s post hoc test, *p < 0.05; **p < 0.01.

Full size image

Both miR-29a and miR-29b are considered tumor suppressors in breast cancer and their repression results in in cancer stem cell expansion in vitro 57. Progestins repress miR-29a and miR-29b in ER+/PR+ breast cancer cells58 and “PR action in breast cancer: stimulation of metastasis” was identified in MetaCore analysis. Patients with ductal carcinoma and elevated miR-29b levels had a significantly longer disease-free survival (DFS) and lower risk to relapse58. miR-29b-3p was downregulated in exosomes of patients with breast cancer recurrence, suggesting a role for miR-29b-3p in inhibition of breast cancer progression and recurrence59. Downregulation of miR-222-3p is associated with AI-resistance in long-term estrogen-deprived MCF-7 cells60. Since miR-222 represses TGFβ-stimulated breast cancer growth56, HNRNPA2B’s downregulation of miR-222-3p may facilitate TGFβ signaling as identified in the MetaCore analysis. Hence, downregulation of these three miRNAs by HNRNPA2B1 may be involved in development of a TAM-R phenotype and worse prognosis in vivo, although further experiments will be needed to determine the pathways and targets involved.

qPCR validation of HNRNPA2/B1-upregulated miRNA targets

Based on their upregulation on HNRNPA2B1-transfected MCF-7 cells, we performed qPCR to validate increases in miR-1266-5p, miR-1268a, and miR-671-3p in separately HNRNPA2B1-transfected MCF-7 cells (11 biological replicate experiments, Fig. 8). As expected, all three miRNAs were significantly higher in HNRNPA2B1-transfected MCF-7 cells.

Figure 8
figure 8

Regulation of miR-1266-5p, miR-1268a, and miR-671-3p by HNRNPA2B1. MCF-7 cells were either non-transfected (control), transfected with pcDNA3.1 parental vector (pcDNA), or an expression vector for HNRNPA2B1 for 48 h. qPCR for (A) miR-1266-5p; (B) miR-1268a, and (C) miR-671-3p. Each miRNA was normalized by RNU48. Values are relative expression normalized to control-transfected cells from 11 biological replicate experiments with multiple control and transfected wells in each experiment. Data were analyzed by two way ANOVA followed by Tukey’s post hoc test, *p < 0.01; **p < 0.001.

Full size image

Expression of miR-1266 was increased in breast tumors showing recurrence or metastasis after TAM treatment with Kaplan-Meir analysis showing that higher miR-1266 was associated with shorter OS and DFS61. This suggests a role for increased miR-1266 in TAM-resistant breast cancer progression. miR-1268a is upregulated in drug-resistant MDA-MB-231 cell lines62,63. We observed and increased in miR-1268a in HNRNPA2B1-transfected MCF-7 cells, but whether this increase is associated with endocrine-resistance in ERα+ breast cancer cells remains to be evaluated.

miR-671-5p was identified as a tumor suppressor in breast cancer, as its expression is lower in invasive breast tumors compared with normal breast. It directly targets FOXM1, and miR-671-5p overexpression inhibits the proliferation and invasion of MDA-MB-231, Hs578T, and T47D, but not MCF-7 cells in vitro 64. Likewise, miR-671-5p was downregulated in fulvestrant-resistant MCF-7 cells43. Thus, the increase in miR-671-5p with HNRNPA2B1 overexpression in MCF-7 cells might not have the same tumor suppressor properties. Because HNRNAP2B1 caused multiple changes in miRNA expression, it will be necessary to analyze combinations of miRNAs and examine both phenotypic and transcriptomic responses.

Transient overexpression of HNRNPA2/B1 reduces TAM and fulvestrant responses in MCF-7 cells

Since LCC9 TAM-resistant cells have higher HNRNPA2/B1 than parental, TAM-sensitive MCF-7 cells, we examined whether transient transfection of MCF-7 cells with HNRNPA2/B1 would impact cell viability in response to TAM or fulvestrant treatment. HNRNPA2B1 overexpression alone does not positively affect MCF-7 viability. We actually observed a 10–15% reduction in MCF-7 cell viability 24 and 48 h post- transfection (Fig. 9A). However, in response to 4-OHT or fulvestrant treatment, HNRNPA2/B1 overexpression was able to increase cell viability, indicating decreased sensitivity to ER antagonists (Fig. 9B). This response is similar to the increased viability of LCC9 cells treated with 4-OHT and fulvestrant33. These data suggest a role for increased HNRNPA2/B1 expression in endocrine-resistance in MCF-7 cells. Future experiments will be required to examine the precise pathways and the role of the altered miRNAs and their targets in this response. Additional future directions include examination of how HNRNPA2B1 overexpression in MCF-7 cells and knockdown in LCC9 cells affects phenotypes associated with TAM-resistance including cell invasion, migration, clonogenic survival, and examination of mRNA targets/proteins of the pathways identified, particularly TGFβ signaling.

Figure 9
figure 9

Effect of transient HNRNPA2B1 cells on MCF-7 cell viability. (A) Results are the Absorbance readings at 490 nm from an MTT assay in which 5,000 MCF-7 cells were plated in OPTI-MEM for 18 h prior to transfection with vector control (pcDNA cont) or HNRNPA2B1 for 24 or 48 h followed by an MTT assay. Each bar is the avg. ± SEM of 4 wells in one experiment. B) MCF-7 cells were transfected with the pcDNA control vector or HNRNPA2B1 for 24 and then treated with vehicle control (DMSO), 100 nM or 1 µM 4-OHT or 100 nM fulvestrant for 48 h followed by an MTT assay. The control was set to 1 for each transfection. Each bar is the avg. ± SEM of 4 wells in one experiment. *p < 0.05 vs. control in both A and B. In (B) **p < 0.05 vs. the same treatment between control vs. HNRNPA2B-transfected cells. Student’s 2-tailed t-test was used for analysis.

Full size image

Conclusions

The primary goal of this study was to identify the global impact of HNRNPA2/B1 overexpression on the miRNA transcriptome of luminal A MCF-7 breast cancer cells, based on the observation of higher HNRNPA2/B1 in LCC9 endocrine-resistant breast cancer cells. We report the comprehensive miRNA changes after 48 and 72 h of HNRNPA2/B1 transfection. A limitation of this study is that both 48 and 72 h HNRNPAB1-transfected cells were compared to control-transfected MCF-7 cells at 48 h. Time- and direction-specific regulated miRNAs were characterized using the MetaCore GO enrichment analysis algorithm, and PR action in breast cancer and TGFβ signaling via miRNA in breast cancer were identified as pathways downstream of the HNRNA2B1 miRome in MCF-7 cells. HNRNPA2B1-downregualtion of miR-29a-3p, miR-29b-3p, and miR-222-3p were confirmed by qPCR in separate experiments. Each of these miRNAs has established roles in breast cancer, including the PR action and TGFβ signaling pathways that were identified in MetaCore analysis. Transient overexpression of HNRNPA2B1 in MCF-7 cells abrogated the ability of 4-OHT and fulvestrant to reduce cell growth. These data support a role for increased HNRNPA2B1 in processes contributing to endocrine-resistance in breast cancer.

Methods

Cell culture and treatments

MCF-7 cells were purchased from American Type Tissue Collection (ATCC, Manassas, VA, USA) and were used within 9 passages from ATCC. MCF-7 cells were grown as described previously65 prior to transient transfection with pcDNA3.1+C-DYK or pcDNA3.1+C-DYK into which HNRNPA2/B1 was cloned (purchased from GenScript, Piscataway, NJ, USA) using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) and Opti-MEM® Reduced Serum Medium (Thermo Fisher Scientific, Carlsbad, CA, USA). The medium was changed from OPTI-MEM (Thermo Fisher Scientific) to Modified IMEM (Thermo Fisher) + 10% FBS six hours after transfection. For the 72 h transfected cells, the medium was replaced with fresh medium 48 h post transfection. A total of six biological replicates for each sample were analysed: control, HNRNPA2/B1-transfected for 48 h, and HNRNPA2/B1-transfected for 72 h.

For miRNA-seq

miRNA was isolated from six separate, biological replicate experiments for each sample group (Control, HNRNPA2/B1 48 h, HNRNPA2/B172 h) using the miRNeasy mini kit from Qiagen (Hilden, Germany) following the manufacturer’s directions. The integrity of the miRNA was confirmed using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Libraries were prepared using QIAseq miRNA Library Kit (Qiagen). cDNA samples were barcoded with QIAseq miRNA NGS ILM IP primers. Adaptor dimers were removed from amplified libraries using QIAseq miRNA NGS beads. The quantity and quality of the library were analyzed on an Agilent Bioanalyzer using the Agilent high sensitivity DNA kit. Pooled library samples were run on an Illumina miSeq to test quantity and quality using the miSeq Reagent Nano Kit V2 300 cycles (Illumina, Foster City, CA). Library and PhIX control (Illumina, Cat. No. FC-110-3001) were denatured and diluted using the standard normalization method to a final concentration of 6 pM. 300 µl of library and 300 µl of PhIX were combined and sequenced on Illumina MiSeq. Based on the miSeq run, the concentration of the libraries was corrected and re-pooled. Sequencing was performed on the University of Louisville Center for Genetics and Molecular Medicine’s (CGeMM) Illumina NextSeq 500 using the NextSeq 500/550 75 cycle High Output Kit v2 (FC-404-2002). Seventy-two single-end raw sequencing files (.fastq) representing three conditions with six biological replicates and four lanes per replicate were downloaded from Illumina’s BaseSpace onto the KBRIN server for analysis. Data were analyzed using miRDeep266 and edgR67.

The sequence reads were mapped to the human reference genome (hg19). Quality control (QC) of the raw sequence data was performed using FastQC (version 0.10.1)68. The FastQC results indicate quality trimming is not necessary since the minimum quality value for all samples is well above Q30 (1 in 1000 error rate). Preliminary adapter trimming was performed on each of the samples to remove the Qiagen 3′ Adapter sequence with Trimmomatric v0.3369. For all of the samples, a peak around 22 bp was found with a broader peak prior to 22 bp (data not shown). Further examination of the overrepresented full-length sequences show that many of these are from other ncRNA sequences. Furthermore, the mapping rate is similar among replicates of the same samples. Therefore, although the distributions differ, the resulting data appears to be consistent with previous miR reports70. The trimmed sequences were directly aligned to the human hg19 reference genome assembly using mirDeep266. Supplementary Table 1 indicates the number of raw reads, number of reads after trimming, and number of reads successfully aligned for each of the samples. The aligned sequences were then used as inputs into mirSeep2 along with the mirBase71 release 22 mature miRNA and miRNA hairpin sequences. The result is a file containing the number of reads mapping to each of the 2,822 human (hsa) miRs. After quantification, the resulting counts for each miR in each sample were combined into a reads matrix. Using the counts table resulting from the previous step, differentially expressed miRs were determined using edgeR67. Using a p-value cutoff of 0.05, the number of differentially expressed miRs in each comparison is shown in Table 1. A heatmap was constructed for differentially expressed miRs passing a FC threshold of ±4 (Log2FC ± 2) in one or more of the comparisons (Fig. 1). The resulting heatmap is shown with up-regulated genes (treatment vs. control) in red and down-regulated genes in green (Fig. 5). The differentially expressed miRs are shown in Tables 1–4, Supplementary Tables 2, 3 for all comparisons. Several miRs are listed twice, due to their coding from multiple gene locations. The raw data were uploaded in the Gene Expression Omnibus (GEO) database as GSE122634.

In silco identification of mRNA targets for miRNAs identified as HNRNPA2/B1-regulated in MCF-7 cells

Experimentally validated mRNA targets for human miRs were downloaded from miRTarBase release 6.172 from http://mirtarbase.mbc.nctu.edu.tw/php/download.php which contains 410,620 miRNA-mRNA interacting pairs. The differentially expressed miRs were then searched against miRTarBase for gene targets. Table 5 shows the number of differentially expressed miRs and the number of validated targets for these miRs. Genes identified as targets of the DEG miRNAs were used as an input into categoryCompare [13] to determine enriched Gene Ontology Biological Processes (GO:BP) and KEGG pathways52 (Supplementary Fig. 8).

In silico MetaCore network analysis

Pathway and network analysis of differentially expressed genes was performed in MetaCore version 6.27 (GeneGO, Thomson Reuters, New York, NY, USA).

RNA extraction and quantitative real-time PCR (qPCR)

RNA was extracted using the RNeasy Mini Kit (Qiagen, Gaithersburg, MD, USA). For miRNA analysis, RNA was isolated using miRNeasy Mini Kit RNA (Qiagen). RNA concentration and quality was assessed using a NanoDrop spectrophotometer (Thermo Scientific, Rockford, IL, USA). The TaqMan® MicroRNA Reverse Transcription Kit and the High Capacity cDNA Reverse Transcription Kit for RNA (both from ThermoFisher) were used to make cDNA for miRNA and mRNA, respectively. Quantitative real-time PCR (qPCR) HNRNPA2/B1 was performed using TaqMan assays (ThermoFisher). 18S rRNA (ThermoFisher) was used as normalizer. qPCR for miR-29a-3p, miR-29b-3p, and miR-222-3p used TaqMan assays and were normalized to RNU6B (ThermoFisher). qPCR was performed using an ABI Viia 7 Real-Time PCR system (LifeTechnologies) with each reaction run in triplicate. The comparative threshold cycle (CT) method was used to determine ΔCT, ΔΔCT, and fold-change relative to control73.

MTT assay

MCF-7 cells were transfected in 6-well plates for 24 h prior to counting and replating (5,000 cells/well) to 96-well plates in phenol red free IMEM supplemented with 5% charcoal-stripped fetal bovine serum (CSS-FBS, Atlanta Biologicals, Lawrenceville, GA, USA) and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA, USA). Cells were treated with vehicle control (DMS), 100 nM or 1 µM 4-OHT (4-hydroxytamoxifen, Sigma-Aldrich, St. Louis, MO, USA), or 100 nM fulvestrant (ICI 182,780; Tocris, Ellisville, MO, USA) for 48 h and cell viability quantified using CellTiter Aqueous One Solution Cell Proliferation Assay (Promega, Fitchburg, WI, USA).

Western blot

Whole cell extracts (WCE) were prepared in RIPA buffer (Sigma-Aldrich) with added phosphatase and complete protease inhibitors (Roche, Indianapolis, IN, USA). Protein concentrations were determined using the Bio-Rad DC protein assay (Bio-Rad, Hercules, CA, USA). 40 or 45 µg of WCE protein were electrophoresed on 10% SDS-PAGE gels and electroblotted on PVDF membranes (Bio-Rad) for western blotting with the following antibodies: HNRNPA2B1 (B1 epitope-specific32) antibody # 18941 from IBL America (Minneapolis, MN USA); GAPDH cat.# sc-365062 (Santa Cruz Biotechnology, Dallas, TX, USA); β-actin (cat. # A5316, Sigma-Aldrich). Bands were visualized using a Bio Rad ChemiDoc MP imager and quantified by UN-SCAN-IT Graph Digitizer Software 7.1 (Silk Scientific, Orem, UT, USA).

Statistics

Statistical analyses were performed using GraphPad Prism 5 (Graph Pad Software, Inc., San Diego, CA, USA). For data in which two samples were compared, Student’s two-tailed test was performed. For data in which more than two samples were compared, one way ANOVA followed by Tukey’s post hoc test was performed.