Re-expression of miR-200s in claudin‐low mammary tumor cells alters cell shape and reduces proliferation and invasion potentially through modulating other miRNAs and SUZ12 regulated genes

Background MicroRNAs are a class of non-coding RNAs that regulate gene expression through binding to mRNAs and preventing their translation. One family of microRNAs known as the miR-200 family is an important regulator of epithelial identity. The miR-200 family consists of five members expressed in two distinct clusters; the miR-200c/141 cluster and the miR-200b/200a/429 cluster. We have found that murine and human mammary tumor cells with claudin-low characteristics are associated with very low levels of all five miR-200s. Methods To determine the impact of miR-200s on claudin-low mammary tumor cells, the miR-200c/141 cluster and the miR-200b/200a/429 cluster were stably re-expressed in murine (RJ423) and human (MDA-MB-231) claudin-low mammary tumor cells. Cell proliferation and migration were assessed using BrdU incorporation and transwell migration across Matrigel coated inserts, respectively. miRNA sequencing and RNA sequencing were performed to explore miRNAs and mRNAs regulated by miR-200 re-expression while Enrichr-based pathway analysis was utilized to identify cellular functions modified by miR-200s. Results Re-expression of the miR-200s in murine and human claudin-low mammary tumor cells partially restored an epithelial cell morphology and significantly inhibited proliferation and cell invasion in vitro. miRNA sequencing and mRNA sequencing revealed that re-expression of miR-200s altered the expression of other microRNAs and genes regulated by SUZ12 providing insight into the complexity of miR-200 function. SUZ12 is a member of the polycomb repressor complex 2 that suppresses gene expression through methylating histone H3 at lysine 27. Flow cytometry confirmed that re-expression of miR-200s increased histone H3 methylation at lysine 27. Conclusions Re-expression of miR-200s in claudin-low mammary tumor cells alters cell morphology and reduces proliferation and invasion, an effect potentially mediated by SUZ12-regulated genes and other microRNAs.


Crystal violet staining and transwell invasion assays
Crystal violet staining and transwell invasion assays were performed as described in [36] using 25,000 human  tumor cells and 30,000 murine tumor cells for the transwell assays. RNA extraction, taqman qRT-PCR for microRNA expression and qRT-PCR for gene expression RNA extraction, Taqman qRT-PCR and qRT-PCR for gene expression were performed as described in [37]. All miR-200 Taqman probes were obtained from Ther-moFisher Scientific (Waltham, MA). miR-200 expression for the murine mammary tumor cell lines was normalized to sno202 and sno234 while the miR-200 expression of the human breast cancer cell lines was normalized to RNU44 and RNU48. All gene primers were obtained from Bio-Rad Laboratories (Mississauga, ON); Cdh1 (qMmu-CID0005843), Hprt (qMmuCED0045738), Snai1 (qMmu-CID0024342), Snai2 (qMmuCED0046072), Twist1 (qMmuCED0004065), Twist2 (qMmuCID0009652) Vim (qMmuCID0005527), Zeb1 (qMmuCID0009095), and Zeb2 (qMmuCID0014662). Hprt was used as the housekeeping gene. miRNA sequencing miRNA sequencing libraries were generated using NEB Multiplex small RNA library Prep Set for Illumina and sequencing quality was determined using an Agilent 2100 Bioanalyzer. Libraries were sequenced using an Illumina NextSeq 500 instrument. The Q30 scores for all samples were above 93%. Reads were then 3′-adaptor trimmed and filtered ≤ 15 bp reads with cutadapt software (v1.14). Trimmed reads were aligned to the reference genome with bowtie software. miRNA expression levels were calculated using mirdeep2 (v0.0.8) and differentially expressed miRNAs were performed with edgeR (v3.18.1). Library preparation, sequencing and data analysis were performed by Arraystar Inc. (Rockville, MD). Four independent samples were sequenced.

RNA sequencing
RNA sequencing for one set of RJ423EV samples and the RJ423ba429 samples was performed at the Genome Quebec Innovation Centre at McGill University using the Illumina Hiseq 2500 v4 PE125 as previously described [37]. RNA sequencing for a second set of RJ423EV samples as well as RJ423-200c/141, MDAEV, MDA-200c/141 and MDA-200ba429 were performed by Arraystar Inc (Arraystar Inc., Rockville MD). All Fastq files were processed using Genialis software (Genialis Inc, Houston, TX) following the standard RNA-seq pipeline which uses BBDuk to remove adapters and trim reads, STAR to align the reads, and feature counts to generate gene level counts. RNA sequencing has been uploaded to GEO under accession number GSE150107. Note that our original data for RJ423EV and RJ423-200ba429 samples found at GSE113162 [36] were analyzed by the Genome Quebec Innovation Centre at McGill University and thus might differ from the data in this manuscript that was analyzed using Genialis software. Three independent samples were sequenced for the RJ423 variants and four independent samples were sequenced for the MDA-231 variants.

BrdU and H3K27me3 flow cytometry
For the murine cell lines, a FITC BrdU flow kit (BD Biosciences, San Jose, CA, cat #559,619) and for the human cells an APC BrdU flow kit (BD Biosciences, San Jose, CA, cat #552,598) were used following the manufacturer's protocol. The APC kit was required for the human cell lines as MDA-231EV cell lines express GFP. Briefly, cells were incubated with 1 mM BrdU in fully supplemented media for 45 min. Cells were then fixed, washed and analyzed on an Accuri C6 cytometer (BD Biosciences, San Jose, CA) using a flow rate of 35 μl/min and 20,000 events were collected. For the H3K27me3 analysis, cells were fixed in 4% paraformaldehyde for 15min and then permeabilized in 90% methanol for 10 min. Cells were then incubated with a 1:200 dilution of anti-H3K27me3 (Cell Signaling cat #9733S) or IgG KP isotype control (Cell Signaling cat #3900s) for 1 h followed by a 30 min incubation with the appropriate secondary antibody (Alexa Fluor 488 for the mouse cells and APC for the human cells). Cells were then resuspended in 5% 7-AAD solution for 5 min. Cells were analyzed on an Accuri C6 cytometer (BD Biosciences, San Jose, CA) using a flow rate of 35 μl/min and 20,000 events were collected.

KEGG pathway, gene ontology and ChEA
For RNA sequencing data Enhrichr [39,40] was used to identify KEGG and Gene Ontology Pathways affected as well as transcription factors implicated in regulating the differentially expressed genes (ENCODE and ChEA Consensus TFs from ChIP-X).

Statistics
Statistical significance was determined using an ANOVA followed by a Dunnett's test using GraphPad Prism 9 (San Diego, CA). For the MCF-7 and MDA-231 variants, all comparisons were made using MDA-231EV cells as the control and for the RJ345 and RJ423 variants, all comparisons were made using RJ423EV cells as the control.

miR-200 expression and the impact on mesenchymal gene expression and cell shape
As shown in Fig Since the miR-200 family has been implicated in regulating epithelial to mesenchymal transition (EMT) [19,[21][22][23][24][25], the expression of the epithelial gene Cdh1, and several mesenchymal genes were evaluated. The altered miR-200 expression in RJ423 and MDA-MB-231 cells did not significantly increase the expression of the epithelial gene, Cdh1 and had inconsistent effects on mesenchymal genes (Fig. 1b, d). Only Twist2 and Vim were significantly reduced by miR-200 expression in RJ423 cells (Fig. 1c) and only Snai1 was significantly reduced by miR-200 expression in MDA-MB-231 cells (Fig. 1d) however, the expression of Snai1 was extremely low in all the cell lines. Somewhat surprisingly, re-expression of miR-200c/141 in both RJ423 and MDA-MB-231 cells induced a significant increase in Snai2 expression (Fig. 1b, d). Zeb1, Zeb2 are predicted targets of all five miR-200s while Snai2 is a predicted target of miR-200b, miR-200c and miR-429 according to miRDB (mirdb.org). Zeb1 was reduced approximately twofold in both RJ423ba429 cells compared to RJ423EV cells and MDA-231c141 cells compared to MDA-231EV cells but these differences did not reach statistical significance. The mesenchymal genes evaluated in this study are regulated by other miRNAs in addition to the miR-200 family and several transcription factors and thus it is the balance of transcriptional induction and repression as well as post-transcriptional regulation that would determine gene expression levels.
During routine culture of these cells it was observed that RJ345 (Fig. 2a) and MCF-7 (Fig. 2f ) cells displayed a more rounded cell shape typically associated with epithelial cells while RJ423 (Fig. 2b) (Fig. 3b).
To determine the impact of miR-200s on migration, matrigel-coated, transwell assays were performed. In RJ423 cells, re-expression of either the miR-200c/141 cluster or the miR-200b/200a/429 cluster significantly reduced cell migration and invasion however, these cells still migrated and invaded more efficiently than the RJ345 cells ( Fig. 3c-g). Re-expression of the miR-200c/141 cluster but not the miR-200b/200a/429 cluster in MDA-MB-231 cells significantly reduced migration and invasion and re-expression of the miR-200c/141 cluster returned migration and invasion capacity to a level similar to MCF-7 cells (Fig. 3h-l).

Re-expression of miR-200s influence the expression of other miRNAs
miRNA sequencing was performed to confirm that the appropriate miR-200 family members had been upregulated in RJ423c141, RJ423ba429, MDA-231c141 and MDA-231ba429 cells and to determine whether reexpression of miR-200s influenced the expression of  Table 1. In each cell line, the miR-200s expressed by the transfected plasmid were found within the top 10 upregulated miRNAs confirming appropriate miR-200 expression. In contrast to the qPCR data (Fig. 1a, c)

Re-expression of miR-200s altered the expression of genes regulated by SUZ12
RNA sequencing was also performed to identify genes altered by miR-200s. Hierarchical clustering revealed that the cell types segregated into discrete clusters and RJ423c141 cells clustered more closely with RJ423EV cells than RJ423ba429 cells (Fig. 4a). The analysis of this data was complicated by the fact that the RNA sequencing for the RJ423c141 and RJ423ba429 cells was performed on different days. In each RNA sequencing run, the same RNA from RJ423EV cells was used. However, the hierarchical clustering reveals some batch effects from the different runs as the two sets of RJ423EV samples (RJ423EVB samples that were utilized with the RJ423ba49 cells and RJ423EVC samples that were utilized with the RJ423c141 cells) showed distinct RNA expression profiles (Fig. 4b). Using a of log2FC ≥ 1 and FDR ≤ 0.01, 494 genes were upregulated and 1002 genes were downregulated in RJ423c141 cells compared to RJ423EV cells and 186 (~ 12%) of these genes had predicted miR-200c or miR-141 binding sites (using miRDB, Fig. 4b). When RJ423ba429 cells were compared to RJ423EV cells, 1691 genes were upregulated and 1899 were downregulated and 437 (~ 12%) of these genes had predicted miR-200b, 200a or miR-429 binding sites (Fig. 4b). Of the genes differentially regulated in RJ423c141 vs. RJ423EV and RJ423ba429 vs. RJ423EV, 913 were shared and 583 and 2677 were unique to RJ423c141 and RJ423ba429 cells, respectively (Fig. 4b).
Since RJ423ba429 cells and MDA-231c141 cells display a change in cell morphology (Fig. 2), the genes unique to these two cell lines were further examined using Enrichr [40,41] to identify pathways associated with the unique genes in RJ423ba429 and MDA-231c141 cells (Fig. 5a, b). Although there were only a small number of pathways shared between RJ423ba429 and MDA-231c141 cells, the top ChEA category was the same in both cell lines, SUZ12 (Fig. 5c). The Encode and ChEA Consensus TFs from ChIP-X database from Enrichr lists 4439 predicted When comparing the RJ423ba429 unique gene profile to the MDA-231c141 unique gene profile, it was determined that 370 genes were shared between the unique gene profiles of the two cell lines (Fig. 6a). The KEGG mmu-miR-429-3p 9.7 3.6 × 10 −18 mmu-miR-9-3p − 3.9 9.9 × 10 −17 and Gene Ontology Pathways as well as the ChEA pathways of these shared genes are presented in Fig. 5b, c and suggest that altering miR-200 expression in these cells influences pathways such as hippo signaling, endothelial migration and metallopeptidase activity (Fig. 6d). The top ChEA pathways remained SUZ12 and 194 of the 370 shared genes (~ 52%) were predicted SUZ12 targets.

H3K27me3 levels are elevated in RJ423ba429 and MDA-231c141 cells
SUZ12 is a member of the polycomb repressor complex 2 (PRC2) which mono-, di-and tri-methylates lysine 27 on histone H3 [43][44][45][46][47][48][49][50]. To further investigate differences in SUZ12 activity, flow cytometry was used to quantify trimethylated histone H3 (H3K27me3). As shown in Fig. 7, RJ423ba429 cells had significantly higher levels of H3K27me3 than RJ423EV cells (Fig. 7a) and MDA-231c141 cells had significantly higher levels of H3K27me3 than MDA-231EV cells (Fig. 7b). The luminal breast cancer cell line, MCF-7, also had significantly elevated levels of H3K27me3 compared to the claudin-low breast cancer cell line MDA-231EV while RJ345 cells that have both luminal and basal-like characteristics, did not have higher levels of H3K27me3 compared to the claudin-low RJ423EV cells. Our observations that re-rexpression of miR-200s in claudin-low mammary tumor cells can inhibit proliferation and migration/invasion as well as altering cell shape is consistent with several published studies. These studies showed that re-expression of miR-200c in MDA-MB-231 cells significantly suppressed cell migration and promoted a more rounded cell morphology [54,55], while treatment of MDA-231 cells with a miR-200c mimic was shown to significantly inhibit transwell invasion [56][57][58][59] or proliferation [57,60]. Although our current study did not find that re-expression of the miR-200b/200a/429 cluster in MDA-MB-231 cells significantly impacted proliferation or invasion, other groups have shown that treating MDA-MB-231 cells with a miR-200a mimic [61], a miR-200b mimic [62][63][64][65], a miR-429 mimic [63], or overexpressing miR-200b [66], significantly reduced migration/invasion and treating MDA-MB-231 cells with a miR-200b or miR-429 mimic significantly reduced proliferation [63,64]. It should be noted that two studies found miR-200 effects that contracted our findings. The first study stably expressed either the miR-200b/200a/429 or miR-200c/141 clusters in MDA-MB-231 cells and found that both clusters reduced cell growth (the miR-200c/141 cluster was more effective than the miR-200b/200a/429 cluster) and both clusters promoted migration and invasion [67]. The second study showed that re-expression of miR-200c significantly promoted invasion in a transwell assay [68]. Studies on murine claudin-low mammary tumor cells are far rarer and the only study outside our lab on murine claudin-low mammary tumor cells was performed in vivo and found that re-expression of the miR-200c/141 cluster in p53 null mammary tumors reduced tumor growth and metastasis to the lungs [69].
RNA sequencing revealed a similar pattern with RJ423ba429 and MDA-231c141 cells expressing more differentially expressed mRNAs than RJ423c141 or MDA-231ba429 cells, respectively. It also revealed that the miR-200c/141 and miR-200b/200a/429 clusters targeted distinct mRNAs despite both clusters sharing identical seed sequences. Cluster specific mRNA (and miRNA) expression profiles were observed in both murine and human mammary tumor cell lines reducing the possibility of this being an artifact. Another interesting observation from the RNA sequencing study was that no more that 12% of the differentially expressed mRNAs in any of the cell lines were predicted targets of the miR-200 family. This is not completely unexpected given the large number of miRNAs also regulated by the miR-200 family as described above. However, this data suggests that studies focusing only on predicted mRNA targets may miss a majority of mRNAs regulated by their miRNA of interest .
To reduce the complexity, differentially expressed mRNAs unique to RJ423ba429 and MDA-231c141 cells (the two lines showing changes in cell shape) were further examined. In both cell lines, the top ChEA pathway identified was SUZ12. SUZ12 is a component of the Polycomb repressive complex 2 (PRC2) that mono-, di-and trimethylates lysine 27 of histone H3 (H3K27) [43][44][45][46][47][48][49][50]. Trimethylation of H3K27 promotes chromosome compaction preventing transcription factors from accessing transcription start sites in promoters [70][71][72]. miR-200b and miR-200c have been reported to target SUZ12 [73][74][75], downregulating its expression and leading to hypomethylation of H3K27 and loss of H3K27 trimethylation. However, re-expression of the miR-200b/200a/429 cluster in RJ423 cells and re-expression of the miR-200c/141 cluster in MDA-MB-231 cells resulted in a significant increase in H3K27me3 levels. In normal mammary epithelial cells, higher levels of H3K27me3 are observed in mature luminal cells compared to mammary stem cells and H3K27me3 levels have been observed to increase during mammary epithelial differentiation during pregnancy [76]. Therefore, re-expression of the miR-200b/200a/429 cluster in RJ423 cells and re-expression of miR-200c/141 cluster in MDA-MB-231 cells may reduce the stem cell features of these cells and drive them to a more differentiated phenotype. Evaluation of gene specific H3K27 methylation patterns are required to confirm this and identify the genes affected.

Conclusions
Understanding how miRNAs influence cellular function is already complex due to the large number of predicted targets and false positives in existing databases, however, our study demonstrates that miR-200s can regulate other microRNAs as well as epigenetic modulators thus adding additional complexity to this system. We have also shown that re-expression of miR-200s in claudin-low mammary tumor cells can revert claudin-low cells to a more epithelial phenotype provided the level of miR-200 re-expression is sufficiently high and this reversion is associated with alterations in SUZ12 regulated genes and reduced cell migration.