Relationship of micro-RNA, mRNA and eIF Expression in Tamoxifen-Adapted MCF-7 Breast Cancer Cells: Impact of miR-1972 on Gene Expression, Proliferation and Migration

Background: Tamoxifen-adapted MCF-7-Tam cells represent an in-vitro model for acquired tamoxifen resistance, which is still a problem in clinics. We here investigated the correlation of microRNA-, mRNA- and eukaryotic initiation factors (eIFs) expression in this model. Methods: MicroRNA- and gene expression were analyzed by nCounter and qRT-PCR technology; eIFs by Western blotting. Protein translation mode was determined using a reporter gene assay. Cells were transfected with a miR-1972-mimic. Results: miR-181b-5p,-3p and miR-455-5p were up-, miR-375, and miR-1972 down-regulated and are significant in survival analysis. About 5% of the predicted target genes were significantly altered. Pathway enrichment analysis suggested a contribution of the FoxO1 pathway. The ratio of polio-IRES driven to cap-dependent protein translation shifted towards cap-dependent initiation. Protein expression of eIF2A, -4G, -4H and -6 decreased, whereas eIF3H was higher in MCF-7-Tam. Significant correlations between tamoxifen-regulated miRNAs and eIFs were found in representative breast cancer cell lines. Transfection with a miR-1972-mimic reverses tamoxifen-induced expression for a subset of genes and increased proliferation in MCF-7, but reduced proliferation in MCF-7-Tam, especially in the presence of 4OH-tamoxifen. Migration was inhibited in MCF-7-Tam cells. Translation mode remained unaffected. Conclusions: miR-1972 contributes to the orchestration of gene-expression and physiological consequences of tamoxifen adaption.


Introduction
Breast cancer still represents the most common neoplasia in women worldwide, showing increasing incidence. Although this disease has an overall good prognosis, certain subtypes are still challenging and there is a need for the identification of additional therapy target molecules, as well as predictive biomarkers [1].

miRNA Extraction and Analysis
miRNA was extracted using the Nuceospin miRNA kit, as described in the manufacturer's protocol (Macherey-Nagel, Düren, Germany). Normal cDNA was synthesized as previously described using oligo dT primers, as well as random hexamer primers [32]. microRNA was determined using the Nanostring nCounter microRNA v2 human mi-croRNA panel according to the manufacturer's recommendations. Data were analyzed and normalized using the nSolver 4.0 software (Nanostring, Seattle, WA, USA).

Gene Expression Analysis
For gene expression analysis, RT-qPCR, as well as nCounter analysis using the elements-chemistry and the 48-gene tag set for tamoxifen regulated genes, as well as the pam50 gene panel [34], was used as previously described [32].

Transfection with miRNA Mimics and Plasmids
For miRNA-1972, the miRidian micro RNA mimic and micro RNA mimic negative control (Horizon, Perkin-Elmer, Cambridge, UK) was used. Cells were grown to about 30% confluency in 6-well plates and the mimic (10 nM) transfected using Dharmafect2 transfection reagent, as recommended by the manufacturer (Horizon, Perkin-Elmer, Cambridge, UK). For determination of the polio-IRES/cap-translation ratio, we followed the procedure published in Vo et al.'s work in 2019 [35].

Proliferation Assays
For the proliferation tests, the cells were transfected at about 1/3 confluency in a 6-well plate. The next day, cells were detached, counted and 5000 cells seeded into each well of a 96 well plate. The next day, and every 24 h for a further four days, resazurin (10 µg/mL) was added and fluorescence was determined as soon as a color change became visible. Fluorescence was measured in a BMG-Labtech Clariostar reader (BMG-Labtech, Ortenberg, Germany), with excitation at 544 nm and emission set to 590 nm. All values were corrected for the results at day 1 and expressed relative to the control treatment.

Scratch Assays
For scratch assays, transfected cells were transferred to a 24 well plate and grown to confluency. Then, the cells were subjected to serum starvation for 24 h. A scratch was applied using a 10 µL pipette tip, the cells were subsequently washed with fresh, serum-free medium and the scratches photographed at a defined position with an inverted, phase contrast microscope using 4× and 10× objectives (Nikon TL-100), equipped with a Nikon camera system, every 24 h for 3 days. Scratches were measured at 3 positions in the microphotographs using the ImageJ software and the difference towards day 1 was calculated.

Target Prediction and Enrichment Analysis
Presumed miRNA targets were downloaded from Targetscan 7.2 [37] for each micro RNA, regardless of context++ score. These lists were compared to the list of 702 of the most regulated genes (at least 2-fold regulation and p adj < 0.01) described in Porsch et al.'s work [32]. Venny 2.0 [38] was used to compare these lists and the consensus list was then submitted for enrichment analysis to the EnrichR website [39]. Here, enrichment for KEGG, reactome and GEO kinase perturbations were evaluated for miR-1972, -181-5p, -213, -375, and -455. Adjusted p-values (p adj < 0.05) were assessed for determining significant enrichments. Network analysis was carried out using the Genemania website [40,41]

Statistics
All statistical calculations were performed using SPSS (IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY, USA: IBM Corp.). For determination of statistical significance, either Student's t-test or ANOVA with either Tamhane-T2 or LSD post-hoc analysis, depending on the presence of equal variances, was used.

miRNAs Regulated by Tamoxifen Adaption
We first screened for 4-OH-tamoxifen-regulated microRNAs after 12 weeks of 4OH tamoxifen treatment using the nCounter microRNA panel on RNA from the three Tamadapted cell-lines described in Porsch et al.'s work, 2019. This panel comprises about 798 miRNAs, which are detected without any further amplification techniques (Supplemental Table S1). We were able to detect the expression of 169 miRNAs. From these, the expression of five miRNAs was found to be significantly changed in the 4-OHtamoxifen-adapted MCF-7 cell lines (Table 2). These miRNAs were analyzed further by qRT-PCR in the same cell lines to obtain further proof for the nCounter results (Table 2). This analysis was extended to selected breast cancer cell lines representing major breast cancer subtypes ( Figure 1). These cell lines were the triple negative lines MDA-MB-231 and -468, HS578T and UACC3199; the latter is also genome-hyper-methylated, and therefore BRCA defective [42]. SK-BR-3 represents an HER2/NEU over-expressing tumor and T47D and MCF-7 are ERand PR-positive luminal A-cell lines.
The results of qRT-PCR-based determination of the miRNAs in MCF-7 and three independently obtained MCF-7-TamR cell lines correlated well with the nCounter data; however, quantitative differences were present. Compared to earlier studies, we obtained similar data as published for miR-375 [18] miR-181b [20] and miR-455 [43]. In contrast to these studies, we found additional Tam-regulated miRNAs, such as miR-1972 and miR-181d (Table 2). When comparing the microRNA expression of MCF-7 with the other cell lines, most of the microRNAs exhibited a similar pattern of expression. miR-375 showed the largest differences between the cell lines. Especially, the TNBC cell lines HS578T and MDA-MB-231 contained very low amounts of this microRNA, whereas T-47D showed a slightly higher abundance compared to the other luminal-A MCF-7 cells line. miR-455 was not detectable in the TNBC cell line MDA-MB-468. The results of qRT-PCR-based determination of the miRNAs in MCF-7 and three independently obtained MCF-7-TamR cell lines correlated well with the nCounter data; however, quantitative differences were present. Compared to earlier studies, we obtained similar data as published for miR-375 [18] miR-181b [20] and miR-455 [43]. In contrast to these studies, we found additional Tam-regulated miRNAs, such as miR-1972 and miR-181d (Table 2). When comparing the microRNA expression of MCF-7 with the other cell lines, most of the microRNAs exhibited a similar pattern of expression. miR-375 showed the largest differences between the cell lines. Especially, the TNBC cell lines HS578T and MDA-MB-231 contained very low amounts of this microRNA, whereas T-47D showed a slightly higher abundance compared to the other luminal-A MCF-7 cells line. miR-455 was not detectable in the TNBC cell line MDA-MB-468.  Table S2.

Prognostic Impact of Tamoxifen-Regulated micro RNAs
Next, we investigated whether these microRNAs have an impact on breast cancer survival using the KM-plotter miRNA tool [44] using the METABRIC data [45]. For miR-1972, no reliable data were present in this dataset (Győrffy Balázs, personal communication 2020). However, using the pan-cancer dataset, a result could be obtained but should be taken with care. Indeed, all tamoxifen-regulated miRNAs had significant impact on overall survival (Table 3). A significant impact was found for all cases and persisted when we restricted the analysis to ER+ and cases that received endocrine therapy. Nevertheless, there was no correlation between up or downregulation by 4OH-tamoxifen and HR. Table 3. Kaplan-Meier overall survival analysis for significantly 4OH-tamoxifen-regulated mi-croRNAs using the KM-plotter website [44] and the METABRIC [45] and pan-cancer (miR-1972) dataset. Hazard ratio (HR) for high expression of the microRNA, as well as log-rank p, is shown. * Figure 1. Abundance of tamoxifen-regulated miRNAs in breast cancer cell lines as determined by qRT-PCR, relative to the housekeeping gene rpl-13. Log 2 of the relative expression towards MCF-7 is shown (log 2 Fc). Error bars indicate standard error (n = 3-6). Results of the ANOVA test including post-hoc analysis are presented in Supplemental Table S2.

Prognostic Impact of Tamoxifen-Regulated micro RNAs
Next, we investigated whether these microRNAs have an impact on breast cancer survival using the KM-plotter miRNA tool [44] using the METABRIC data [45]. For miR-1972, no reliable data were present in this dataset (Győrffy Balázs, personal communication 2020). However, using the pan-cancer dataset, a result could be obtained but should be taken with care. Indeed, all tamoxifen-regulated miRNAs had significant impact on overall survival (Table 3). A significant impact was found for all cases and persisted when we restricted the analysis to ER+ and cases that received endocrine therapy. Nevertheless, there was no correlation between up or downregulation by 4OH-tamoxifen and HR. Table 3. Kaplan-Meier overall survival analysis for significantly 4OH-tamoxifen-regulated microR-NAs using the KM-plotter website [44] and the METABRIC [45] and pan-cancer (miR-1972) dataset. Hazard ratio (HR) for high expression of the microRNA, as well as log-rank p, is shown. * Please note that results for miR-1972 might be unreliable, as the median expression was very low. n.a.: not available.

Evaluation of MicroRNA-mRNA Correlations and Enrichment Analysis
We then combined the results of the microRNA analysis with the mRNA results published earlier [32]. First, we determined the overlap between the lists of the predicted miRNA target genes and the list of 702 mRNAs that were regulated at least by a factor 2 with a p adj < 0.01. The number of Targetscan-predicted genes ranged from 4772 (miR-1972) to 258 (miR-455). An overlap of the lists was found for 4% of the predicted targets on average ( Table 4). Each of the consensus lists and a merged list were then further analyzed by enrichment analysis for KEGG, REACTOME and GEO kinase perturbations on the EnrichR website [39]. Whereas no single miRNA resulted in significantly enriched pathways, the combination of all regulated 225 target genes provided significant enrichment results. These enriched pathways included leishmania and human leukemia virus 1 infections, the Th1-, -Th2 differentiation signalling and morphine addiction (Supplemental Table S3A). The GEO kinase perturbation data suggested 15 kinases, as influenced by the tamoxifenregulated miRNAs. In the pathway enrichment analysis, these kinases were associated with major cancer pathways, such as PI3K-AKT-, ErbB-, interleukin or FoxO-signalling (Supplemental Table S3B). We further investigated whether these kinases might form an interaction network using the Genemania website. Indeed, all 15 kinases could be included into a single network ( Figure 2) and a further 20 proteins were added by the Genemania algorithm.

A miR-1972 Mimic Effects Gene Expression
We then focused further on the miR-1972, as its regulation by tamoxifen has not been described before. Targetscan analysis [37] suggested a comparatively high number of target genes and about 3.6% of these genes were indeed regulated in the tamoxifen adaption time course experiment published earlier [32]. We, therefore, decided to transfect MCF-7 as well as MCF-7-Tam cells with a miR-1972 mimic to explore the effects on gene expression further. The mimic transfection should especially counteract the tamoxifen-mediated downregulation of this miRNA in tamoxifen-adapted cells. For the analysis, we used the established nCounter technique for the tamoxifen gene-set developed earlier [32] and also the pam50 gene set, as established for the prosigna test [46][47][48]. Indeed, in both gene sets, significant gene expression changes could be identified. In case of the tam-gene set, 12 genes were significantly regulated by miR-1972 mimic transfection ( Figure 3, Table S4) in the TamR-cell line. Here, we were especially interested in genes where miR-1972 mimic transfection resulted in a more "MCF-7-like" expression. This was the case for 9 of these genes ( Figure 4). MCF-7-Tam showed significant alterations in 29 genes of the pam 50 gene set, compared to the MCF-7 cell-line. Furthermore, transfection with the miR-1972 mimic resulted in 11 changes in MCF-7 and 14 in MCF-7-TamR cells, with 5 of these genes altered in both cell lines. With respect to the genes that are assigned to an intrinsic subtype [49], MCF-7-Tam cells showed changes in all subtypes (Table S4). These were three genes in "luminal A", nine genes in "luminal B", three in "normal like", seven in "Her2 enriched" and five in the "basal" subtype. The miR-1972 mimic transfection caused at least a partial reversal of the tamoxifen adaptation effect for eight of these genes ( Figure 4).

Protein Translation Initiation and eIF-Expression in Breast Cancer Cell Lines
In addition to the pathway enrichment analysis, we were particularly interested in investigating whether tamoxifen adaption would change the mode of protein translation. Indeed, we recently observed that tamoxifen-adapted MCF-7 cell lines exhibited significant changes in the ratio of polio-IRES to cap-mediated translation. In reporter gene assays, the ratio of polio-IRES-driven firefly luciferase to cap-driven renilla-luciferase dropped in Tam-adapted cells by 57% (p < 0.01). We, therefore, hypothesized that this result could be mediated by differential expression of eIFs, which might be caused at least partly by the differential expression of tamoxifen-regulated micro RNAs. In our mRNA expression studies, several significant differences for the eIFs were evident, but expression differences expressed as log 2 Fc were all below ±1 (Table 5). Interaction network of micro-RNA associated kinases as obtained from "Genemania" [40,41]. A total of 15 kinases associated with the tamoxifen-regulated micro RNAs (Table 4) were used for this analysis (shaded circles) and a further 20 genes were added by the Genemania algorithm.

A miR-1972 Mimic Effects Gene Expression
We then focused further on the miR-1972, as its regulation by tamoxifen has not been described before. Targetscan analysis [37] suggested a comparatively high number of target genes and about 3.6% of these genes were indeed regulated in the tamoxifen adaption time course experiment published earlier [32]. We, therefore, decided to transfect MCF-7 as well as MCF-7-Tam cells with a miR-1972 mimic to explore the effects on gene expression further. The mimic transfection should especially counteract the tamoxifen-mediated downregulation of this miRNA in tamoxifen-adapted cells. For the analysis, we used the established nCounter technique for the tamoxifen gene-set developed earlier [32] and also the pam50 gene set, as established for the prosigna test [46][47][48]. Indeed, in both gene sets, Figure 2. Interaction network of micro-RNA associated kinases as obtained from "Genemania" [40,41]. A total of 15 kinases associated with the tamoxifen-regulated micro RNAs (Table 4) were used for this analysis (shaded circles) and a further 20 genes were added by the Genemania algorithm.  (Table S4). These were three genes in "luminal A", nine genes in "luminal B", three in "normal like", seven in "Her2 enriched" and five in the "basal" subtype. The miR-1972 mimic transfection caused at least a partial reversal of the tamoxifen adaptation effect for eight of these genes ( Figure 4).      We then further investigated the protein abundance for the selected eIFs under the influence of 4OH-tamoxifen by Western blotting. We obtained defined Western blot signals ( Figures 5 and S1) for twelve eIFs (eIF2A, eIF2α, -3A, -3D, -3H, -4A1, -4B, 4E, -4G, -4H, -4EBP1, and -6) and two phosphorylated eIFs (phospho-eIF2α and phospho-eIF4E-BP1). From these, eIF2A, -3H, -4H, and -6 turned out to be significantly regulated and the ratio of phosphorylated eIF4EBP1 to eIF4BP-1 showed a statistical trend. When compared to the mRNA data (Table 5), eIF2A, -4A1 showed consistent results, whereas eIF4B showed no significant alteration in the Western blots.  Supplemental  Table S5. ** p < 0.01; * p < 0.5; + p < 0.1.
In the next step, we analyzed the eIF expression in the cell lines that were investigated for micro-RNA expression ( Figure 6). Here, eIF2A, eIF3H, eIF4A1, eIF4G, eIF4E, and eIF6 showed the most significant differences between the cell lines (Table S5).

Correlation Analysis for eIFs and microRNAs in Breast Cancer Cell Lines
As we detected differences for 4OH-Tam-regulated miRNAs as well as for eIFs, we tested for correlations between miRNA and eIF abundance for the cell lines included into this study. The results were clustered and are shown in Figure 7. Several highly significant correlations were found, suggesting a functional or regulatory relationship between some miRNAs and eIFs. Three major clusters could be identified by this analysis. The first cluster contains peIF4EBP1 and its ratio to eIF4EBP1.
The second group contains miR-375 and -1972 together with eIF1, -2A, -4B, -4H and eIF4EBP1. The third cluster contains the miR-181 family members, as well as several eIFs. In this group, most of the highly significant correlations can be observed especially between eIF3A, eIF3H, eIF4G and the eIF2α-phosphorylation ratio.
However, when comparing these data with miRNA Targetscan predictions, only one consistent correlation was found (Table 6). miR-375 and miR-1972 are both predicted to interact with eIF4H. However, the context score for miR-1972 in particular was very low. Biomolecules 2022, 12, x FOR PEER REVIEW 12 of 21  Table S5.

Correlation Analysis for eIFs and microRNAs in Breast Cancer Cell Lines
As we detected differences for 4OH-Tam-regulated miRNAs as well as for eIFs, we tested for correlations between miRNA and eIF abundance for the cell lines included into this study. The results were clustered and are shown in Figure 7. Several highly significant  Table S5. eIF4EBP1. The third cluster contains the miR-181 family members, as well as several eIFs. In this group, most of the highly significant correlations can be observed especially between eIF3A, eIF3H, eIF4G and the eIF2α-phosphorylation ratio.
However, when comparing these data with miRNA Targetscan predictions, only one consistent correlation was found (Table 6). miR-375 and miR-1972 are both predicted to interact with eIF4H. However, the context score for miR-1972 in particular was very low.

Figure 7.
Correlation matrix of eIF-protein-and miRNA-abundance in the cell lines. A total of 8 cell lines and 12 eIFs, as well as 2 phosphorylation ratios and 6 miRNAs, were included in this analysis. eIF-and miRNA-abundance was expressed relative to MCF-7 and the Pearson correlation factor and significance determined. The correlation factors were then used for this cluster analysis using the CIMminer on-line tool [51]. Major clusters are numbered from 1 to 3. Significant correlations are indicated by *: p < 0.05 and **: p < 0.01.  Figure 7. Correlation matrix of eIF-protein-and miRNA-abundance in the cell lines. A total of 8 cell lines and 12 eIFs, as well as 2 phosphorylation ratios and 6 miRNAs, were included in this analysis. eIF-and miRNA-abundance was expressed relative to MCF-7 and the Pearson correlation factor and significance determined. The correlation factors were then used for this cluster analysis using the CIMminer on-line tool [51]. Major clusters are numbered from 1 to 3. Significant correlations are indicated by *: p < 0.05 and **: p < 0.01.

Transfections Using the miR-1972 Mimic
As we have observed several interesting correlations between Tam-regulated micro RNAs and eIFs, we intended to further support these results by transfection of MCF-7 and MCF-7-TamR cells with the miR-1972 mimic and a control RNA. Again, we determined the abundance of eIFs by Western blotting. However, here only minor changes were observed ( Figure S1). Only for eIF4A1 was a downregulation found in MCF-7 (p < 0.1) and an upregulation found for MCF-7-Tam (p < 0.05). Interestingly, eIF4A1 was downregulated by tamoxifen adaption (Figure 5). eIF6 was upregulated in MCF-7-Tam cells (p < 0.1). The ratio of the polio-IRES to cap-mediated protein translation also remained unchanged upon transfection with the 1972 mimic (data not shown).

Effect of the miR1972-Mimic on Proliferation and Migration
As miR-1972 might be a factor for the aggressiveness of tumor cells, we further investigated its impact on proliferation and migration of MCF-7 and MCF-7-Tam cells. In proliferation assays, miR1972-transfected MCF-7 cells showed a significantly higher resorufin signal than control transfections, whereas the signal remained nearly unchanged in MCF-7-TamR cells (Figure 8). We also repeated these experiments in medium supplemented with 4OH-tamoxifen (10 nM); however, here MCF-7 showed only statistically insignificant effects but the negative effect on MCF-7-TamR cells became larger and statistically significant. Nevertheless, none of these differences did exceed a log 2 Fc value of ± 1.

Transfections Using the miR-1972 Mimic
As we have observed several interesting correlations between Tam-regulated micro RNAs and eIFs, we intended to further support these results by transfection of MCF-7 and MCF-7-TamR cells with the miR-1972 mimic and a control RNA. Again, we determined the abundance of eIFs by Western blotting. However, here only minor changes were observed ( Figure S1). Only for eIF4A1 was a downregulation found in MCF-7 (p < 0.1) and an upregulation found for MCF-7-Tam (p < 0.05). Interestingly, eIF4A1 was downregulated by tamoxifen adaption (Figure 5). eIF6 was upregulated in MCF-7-Tam cells (p < 0.1). The ratio of the polio-IRES to cap-mediated protein translation also remained unchanged upon transfection with the 1972 mimic (data not shown).

Effect of the miR1972-Mimic on Proliferation and Migration
As miR-1972 might be a factor for the aggressiveness of tumor cells, we further investigated its impact on proliferation and migration of MCF-7 and MCF-7-Tam cells. In proliferation assays, miR1972-transfected MCF-7 cells showed a significantly higher resorufin signal than control transfections, whereas the signal remained nearly unchanged in MCF-7-TamR cells (Figure 8). We also repeated these experiments in medium supplemented with 4OH-tamoxifen (10 nM); however, here MCF-7 showed only statistically insignificant effects but the negative effect on MCF-7-TamR cells became larger and statistically significant. Nevertheless, none of these differences did exceed a log2Fc value of ± 1. Figure 8. Impact of transfection with the miR-1972 mimic on viability/proliferation of MCF-7-and MCF-7-Tam cells in normal medium and medium containing 4OH-tamoxifen (10 nM). Cell Figure 8. Impact of transfection with the miR-1972 mimic on viability/proliferation of MCF-7and MCF-7-Tam cells in normal medium and medium containing 4OH-tamoxifen (10 nM). Cell viability/proliferation was determined using the resazurin assay. The fluorescence signal was normalized to the signal after seeding and shown relative to control transfections as log 2 Fc. * indicates significant differences to the control transfection determined by ANOVA and Tamhane-T2 post-hoc test (p < 0.05). The experiments were performed three times with 3 to 4 replicas each.
In scratch experiments, MCF-7 cells did migrate faster than the MCF-7-TamR cells. In addition, MCF-7 cells did not show a statistically significant change in migration, whereas MCF-7-TamR cells exhibited an even reduced scratch closure (Figure 9). viability/proliferation was determined using the resazurin assay. The fluorescence signal was nor-malized to the signal after seeding and shown relative to control transfections as log2Fc. * indicates significant differences to the control transfection determined by ANOVA and Tamhane-T2 post-hoc test (p < 0.05). The experiments were performed three times with 3 to 4 replicas each.
In scratch experiments, MCF-7 cells did migrate faster than the MCF-7-TamR cells. In addition, MCF-7 cells did not show a statistically significant change in migration, whereas MCF-7-TamR cells exhibited an even reduced scratch closure (Figure 9).

Discussion
We recently reported on the changes in gene expression within 12 weeks of 4OHtamoxifen treatment of the MCF-7 cell line, with a focus on long non coding RNAs [32]. We here analyzed gene expression in these cells further, now concentrating on micro RNAs, their relationship to the tamoxifen-regulated mRNAs and eukaryotic initiation factors. As expected from the literature, we identified tamoxifen-regulated micro RNAs, some of them already known, but especially the tamoxifen-mediated downregulation of miR-1972 represented a new finding.
The survival data obtained from the METABRIC dataset showed that all miRNAs identified as tamoxifen regulated had a prognostic impact, especially when endocrine therapy was provided (Table 3). However, data on miR-1972 might not be reliable, as the median expression was very low (Győrffy Balázs, personal communication 2020).
As the first step for an integrative analysis of mRNA and miRNA expression data, we determined whether the predicted target genes were significantly regulated by tamoxifen. The expression of about 4% of the predicted targets was indeed changed.
An enrichment analysis using these genes identified several pathways associated with infections and immune responses. In these pathways, the NF-kB-inhibitor α (NFKBIA) represented one of the major hits. This is consistent with the data suggesting that NF-kB plays an important role in acquired tamoxifen resistance [52][53][54][55][56]. The result for the pathway leading to morphine addiction seems hard to explain; however, several effects of tamoxifen on morphine responses and vice versa have been published. For example, glucuronidation of morphine and tamoxifen occurs by the same enzymes in mice, which could lead to reduced tamoxifen efficiency [57]. In mice, estrogens and tamoxifen also modified methadone responses [58]. Evidently, this cross talk should be further evaluated.
We also found several protein kinases associated with the miRNA co-regulated genes in the GEO kinase perturbation data. Most of these kinases have a well-known impact on breast cancer prognosis. These include EGFR [59], CDK8/19 [60], FGFR3 [61] and ROCK2 [62]. However, not much literature exists on breast cancer and SNRK, which should lead

Discussion
We recently reported on the changes in gene expression within 12 weeks of 4OHtamoxifen treatment of the MCF-7 cell line, with a focus on long non coding RNAs [32]. We here analyzed gene expression in these cells further, now concentrating on micro RNAs, their relationship to the tamoxifen-regulated mRNAs and eukaryotic initiation factors. As expected from the literature, we identified tamoxifen-regulated micro RNAs, some of them already known, but especially the tamoxifen-mediated downregulation of miR-1972 represented a new finding.
The survival data obtained from the METABRIC dataset showed that all miRNAs identified as tamoxifen regulated had a prognostic impact, especially when endocrine therapy was provided (Table 3). However, data on miR-1972 might not be reliable, as the median expression was very low (Győrffy Balázs, personal communication 2020).
As the first step for an integrative analysis of mRNA and miRNA expression data, we determined whether the predicted target genes were significantly regulated by tamoxifen. The expression of about 4% of the predicted targets was indeed changed.
An enrichment analysis using these genes identified several pathways associated with infections and immune responses. In these pathways, the NF-kB-inhibitor α (NFKBIA) represented one of the major hits. This is consistent with the data suggesting that NF-kB plays an important role in acquired tamoxifen resistance [52][53][54][55][56]. The result for the pathway leading to morphine addiction seems hard to explain; however, several effects of tamoxifen on morphine responses and vice versa have been published. For example, glucuronidation of morphine and tamoxifen occurs by the same enzymes in mice, which could lead to reduced tamoxifen efficiency [57]. In mice, estrogens and tamoxifen also modified methadone responses [58]. Evidently, this cross talk should be further evaluated.
We also found several protein kinases associated with the miRNA co-regulated genes in the GEO kinase perturbation data. Most of these kinases have a well-known impact on breast cancer prognosis. These include EGFR [59], CDK8/19 [60], FGFR3 [61] and ROCK2 [62]. However, not much literature exists on breast cancer and SNRK, which should lead to further investigations. By testing for interactions using the Genemania website, all 15 kinases could be placed into a single network, which represented several cancer relevant pathways. Taken together, the results of this computational analysis are in line with earlier studies, showing that tamoxifen resistance is established by a regulatory network of micro-RNAs and signaling pathways [43,63].
As the regulation of hsa-mir-1972 seemed new to this research topic, we then focused on this miRNA by manipulating its expression by transfection with a miR-1972 mimic. We expected that this approach would provide information about whether miR-1972 is involved in the establishment of tamoxifen gene expression. Indeed, restoration of miR-1972 expression in MCF-7-Tam cells influenced the parameters that were associated with the tamoxifen adaption process. This included effects on proliferation, migration and especially changes in the expression of tamoxifen-regulated genes and genes included in the prognostically important pam50 panel.
Interestingly, miR-1972 transfection reversed the expression of a subset of tamoxifenregulated genes into the direction of MCF-7 expression levels. We suggest that this holds for the participation of this micro RNA in the establishment of the tamoxifen-induced gene expression pattern. Furthermore, transfection with the miR-1972 mimic induced a decrease in the proliferation/vitality of tamoxifen-adapted cells, especially when 4OH-tamoxifen was present. This might be interpreted as the restoration of tamoxifen sensitivity; however, a more detailed analysis would be needed to prove this point further, especially as the effect was rather low. Migration was even further reduced in MCF-7-Tam cells, which is in line with the reduction in vitality by miR-1972 mimic transfection. In MCF-7, however, the miR-1972 mimic caused a small increase in proliferation, which was reduced in the presence of 4OH-tamoxifen. We assume that this might be caused by either the different biology of the parental cell-line or by off-target effects, caused by an unphysiologically high intracellular mimic concentration.
For miR-375, an impact on tamoxifen sensitivity has already been shown [18] and we found this micro RNA in the same eIF/miRNA cluster as miR-1972. This suggests a similar function of these two micro RNAs in breast cancer biology.
miR-1972 has rarely been investigated in breast cancer. However, this miRNA seems overexpressed in cancer tissue compared to normal tissue [64] and was, therefore, included into the dbDEMC database among the top50 breast cancer-related miRNA candidates [65]. In a comparison of 2D and 3D cell culture, however, miR-1972 was not regulated [66]. Furthermore, in an evaluation of circulating miRNAs of cancer patients and healthy donors, miR-1972 was also not conspicuous [67]. However, most interestingly, Hoppe et al. reported an upregulation of this miRNA in aromatase inhibitor (AI)-resistant MCF-7-derived cell lines [68]. This seems to be in contrast to our data but may argue for a different mechanism leading to AI resistance.
For other cancer entities, miR-1972 seems more important; especially for papillary thyroid carcinoma [69], osteosarcoma [70], ovarian cancer [71] and chronic myeloid leukemia [63]. In several of these studies, sponging of miR-1972 by non-coding RNAs has been proposed as a molecular mechanism. In osteosarcoma, differentiation antagonizing non-protein coding RNA (DANCR) decoys miR-1972 [70] but this RNA was apparently not regulated in our tamoxifen cell model. Other miR-1972-sponging linc-RNAs, such as linc00588 [72] or lnc01207 [73], were also not regulated in our cell model. Regarding gynaecological cancers, linc01125 [71] sponged miR-1972 in ovarian cancer, but again, this linc-RNA was not regulated in our tamoxifen cell model. However, such interactions might not be necessary, as miR-1972 is already down-regulated by tamoxifen itself. This might be different for dedicator of cytokinesis 9-antisense RNA2 (DOCK9-AS2) that sponges microRNA-1972 (miR-1972), leading to the upregulation of catenin β1 (CTNNB1) in thyroid cancer [69]. DOCK-AS2 was indeed moderately up-regulated in MCF-7 by 4OH tamoxifen (log 2 Fc = 0.5, p adj = 0.007); however, CTNNB1 was not. Interestingly, in thyroid cancer, this DOCK2-AS2 sponging was correlated with WNT-signalling, a pathway that is also supposed to be involved in tamoxifen resistance [56,74,75]. Nevertheless, possible sponging mechanisms on miR-1972 require further analysis.
A second focus of this investigation was based upon the hypothesis that miRNAregulated eIFs contribute to tamoxifen resistance. We have indeed found that the ratio of polio-IRES to cap-mediated translation was altered in MCF-7-TamR cells. This was accompanied by changes in the abundance of eIF2A, eIF3H, eIF4H, eiF4G and eIF6. In addition, the phosphorylation ratio of eIF2α and eIF4EBP1 was changed. This fits well with an altered start site selection in MCF-7-TamR cells. Disappointingly, miR-1972 had no effect on the polio-IRES/cap ratio. However, for eIF4A1, a modest upregulation in MCF-7 and downregulation in MCF-7-Tam-cells was observed. Additionally, eIF6 was upregulated by the miR-1972-mimic transfection in MCF-7-Tam cells. This initiation factor is not known to contribute to start site selection; however, it seems to be associated with the stress response via phosphorylation by GSK3 [76]. A miR-1972 regulation might, therefore, also reflect a stress response of the cells, which might result from the reduced proliferation/vitality caused by miR-1972 mimic transfection. Again, we propose that miR-1972 contributes to the tamoxifen effects on translation, but is not a master regulator.

Conclusions
The miRNA-1972 seems, in part, responsible for gene expression and physiological changes resulting from long-term exposure to 4OH-tamoxifen. The impact of this microRNA on the prognosis of breast cancer needs, however, further evaluation.