SAP domain-dependent Mkl1 signaling stimulates proliferation and cell migration by induction of a distinct gene set indicative of poor prognosis in breast cancer patients

Abstract

Background: The main cause of death of breast cancer patients is not the primary tumor itself but the metastatic disease. Identifying breast cancer-specific signatures for metastasis and learning more about the nature of the genes involved in the metastatic process would 1) improve our understanding of the mechanisms of cancer progression and 2) reveal new therapeutic targets. Previous studies showed that the transcriptional regulator megakaryoblastic leukemia-1 (Mkl1) induces tenascin-C expression in normal and transformed mammary epithelial cells. Tenascin-C is known to be expressed in metastatic niches, is highly induced in cancer stroma and promotes breast cancer metastasis to the lung.

Methods: Using HC11 mammary epithelial cells overexpressing different Mkl1 constructs, we devised a subtractive transcript profiling screen to identify the mechanism by which Mkl1 induces a gene set co-regulated with tenascin-C. We performed computational analysis of the Mkl1 target genes and used cell biological experiments to confirm the effect of these gene products on cell behavior. To analyze whether this gene set is prognostic of accelerated cancer progression in human patients, we used the bioinformatics tool GOBO that allowed us to investigate a large breast tumor data set linked to patient data.

Results: We discovered a breast cancer-specific set of genes including tenascin-C, which is regulated by Mkl1 in a SAP domain-dependent, serum response factor-independent manner and is strongly implicated in cell proliferation, cell motility and cancer. Downregulation of this set of transcripts by overexpression of Mkl1 lacking the SAP domain inhibited cell growth and cell migration. Many of these genes are direct Mkl1 targets since their promoter-reporter constructs were induced by Mkl1 in a SAP domain-dependent manner. Transcripts, most strongly reduced in the absence of the SAP domain were mechanoresponsive. Finally, expression of this gene set is associated with high-proliferative poor-outcome classes in human breast cancer and a strongly reduced survival rate for patients independent of tumor grade.

Conclusions: This study highlights a crucial role for the transcriptional regulator Mkl1 and its SAP domain during breast cancer progression. We identified a novel gene set that correlates with bad prognosis and thus may help in deciding the rigor of therapy.

Figure 1

Screen for SAP-dependent Mkl1 target genes and their implication in cancer. (A) Scatter plot and (B) Venn diagram representing classification of Mkl1 target genes into three groups: SRF-dependent/SAP-independent (blue), SRF-dependent/SAP-dependent (red) and SRF-independent/SAP-dependent (green). The scatter plot (A) represents the log fold change (logFC) in gene expression in HC11-∆SAP versus HC11-FL control cells (x-axis; ∆SAP vs. FL) and between HC11-mutB1 versus HC11-FL control cells (y-axis; mutB1 vs. FL). Each dot represents a probeset, and the one for tenascin-C is highlighted (Tnc). The vertical and horizontal lines in the chart denote the 2-fold change cutoff (logFC = -1). The Venn diagram (B) represents the number of probesets for transcripts, which are more than 2-fold reduced in either HC11-mutB1 or HC11-ΔSAP cells when compared to HC11-FL control cells. Boxes below the Venn diagram indicate the cell strains that have reduced levels of the respective transcripts. (C, D) Functional analysis for the three Mkl1-regulated gene sets performed using the IPA software. The high-level functional (C) and disease (D) categories are displayed along the x-axis of each bar chart. The y-axis displays the –log of the P-value determined by right-tailed Fisher’s exact test. The P-value is a measure of the likelihood that the association between a set of genes in each dataset and a related function or disease is due to random association. The grey vertical line denotes the cutoff for significance (P = 0.05; -logP = 1.3).

Figure 2

SRF-independent/SAP-dependent transcripts represent direct Mkl1 target genes requiring the SAP domain of Mkl1 to induce transcription from their proximal promoter. (A) The indicated promoter constructs that contained at least 500 bp upstream of the transcription start site (TSS) and were linked to the secreted alkaline phosphatase (SEAP) reporter gene, were cotransfected in HC11 cells together with an inactive Mkl1 devoid of the transactivation domain [13] or the FL-Mkl1 construct. SEAP activity is expressed as fold induction above the level obtained with the inactive Mkl1. In addition to Tnc, for 8 out of the 12 new promoters tested, induction greater than 2-fold (indicated by the red line) was obtained. Values are means ± SEM from three to seven independent experiments. (B) HC11 cells were cotransfected with the indicated promoter constructs that were either > 500 bp or shortened to 200 bp upstream of the TSS, and with vectors encoding the indicated mutant Mkl1 constructs. SEAP activity is normalized and expressed as in (A). Means ± SEM from at least three independent experiments and significant differences between either mutB1- and ΔSAP-Mkl1-transactivated promoter constructs or between the longer and shorter promoter constructs transactivated by mutB1-Mkl1, ***P < 0.001, **P < 0.01, *P < 0.05 are shown.

Figure 3

The different HC11 cell strains proliferate at different rates and show distinct migration behaviors. (A) Immunoblot with mAb65F13 of Mkl1 proteins in whole-cell extracts from the empty vector-, FL-, mutB1- or ΔSAP-transfected HC11 strains. Anti-Gapdh served as loading control. Endogenous Mkl1 protein was below the detection limit in empty vector cells. (B) SAP-dependent proliferation of HC11 mammary epithelial cells. Proliferation rates of the four HC11 cell strains were assessed by BrdU incorporation into newly synthesized DNA immediately after plating (0 h) as well as at 24, 48, 72 and 96 h. Means ± SD from three independent experiments and significant differences to the HC11-ΔSAP cells, ***P < 0.001, **P < 0.01, *P < 0.05 are shown. (C) SAP-dependent migration of HC11 mammary epithelial cells. Cell migration of the four HC11 strains was evaluated by Transwell migration assay using filters with 8 μm pore size. Quantification of the cell migration was measured by the area on the lower side of the filter covered with cells. Above the bar graph, a photo of fixed and stained cells seeded in parallel in a 24-well plate is shown as a seeding control, and representative photos of fixed and stained cells of each of the cell strains that have migrated to the lower side of the filter, are shown below (bar, 200 μm). Data and statistical significance are expressed as in (B).

Figure 4

SAP-dependent Mkl1 target genes are mechanoresponsive. (A) Effect of static (20%) and cyclic (15%, 0.3 Hz) strain on Tnc and c-fos mRNA levels. HC11 cells were cultured on either growth factor reduced matrigel matrix- or fibronectin-coated silicone membranes in 0.03% serum-containing medium for 24 h before applying static or cyclic strain for 1 h. Cells cultured under the same conditions and not exposed to mechanical stimulation were used as a resting control. The two types of coating gave identical results under the indicated experimental conditions. Total RNA was extracted and qRT-PCR was performed for Tnc and c-fos mRNA levels. Values normalized to Gapdh are expressed relative to the values of resting cells. Data represent means ± SD from three independent experiments. Significant differences to the resting control, ***P < 0.001, **P < 0.01, *P < 0.05. (B) SAP-dependent genes respond to cyclic strain. HC11 cells were stretched and mRNA analyses were performed as described in (A). Data and statistical significance are expressed as in (A).

Figure 5

SAP-dependent Mkl1 target genes are associated with typical high-proliferative poor outcome classes in breast cancer. The expression levels for the SRF-dependent/SAP-independent (A) and SRF-independent/SAP-dependent (B) gene sets are analyzed across the 1881-sample breast cancer data set stratified according to PAM50 subtypes (left panels), estrogen receptor (ER)-status (middle panels) and histological grade (right panels), and represented by box plots using the GOBO bioinformatics tool. The number of tumors in each breast cancer subtype and the significant difference in gene expression (P-value calculated using ANOVA) between them are shown above the box plots.

Figure 6

The SRF-independent/SAP-dependent genes represent a bad prognostic signature for breast cancer patients. Tumors in the 1881-sample breast cancer data set were stratified into three quantiles, low, intermediate and high, based on SRF-dependent/SAP-independent (A) or SRF-independent/SAP-dependent (B) gene expression. (A) Kaplan-Meier survival analysis using distant metastasis free survival (DMFS) as endpoint and 10-year censoring for all tumors (n = 1379; left panels), or in the subgroups of lymph node (LN)-negative (n = 1111; middle panels) and untreated tumors (n = 821; right panels) was performed using the GOBO bioinformatics tool, interrogating the group of SRF-dependent/SAP-independent target genes. P-value is calculated using log-rank test. (B) Kaplan-Meier survival analysis for tumors with expression of SRF-independent/SAP-dependent Mkl1 target genes was performed as in (A). Association with clinical outcome was assessed in the subgroups of ER-positive (n = 856; left panel), Grade 1 (n = 141, middle panel) and Grade 2 (n = 446; right panel) tumors in addition to all tumors and the subgroups used in (A). P-value is calculated using log-rank test.

Figure 7

Elevated expression of SAP-dependent Mkl1 target genes is a poor prognosis factor in breast cancer independent of histological grade. Multivariate analysis supporting the Kaplan-Maier survival analysis (shown in Figure 6) for the SRF-dependent/SAP-independent (A) and SRF-independent/SAP-dependent (B) gene sets, was performed using the GOBO bioinformatics tool. The analysis was executed for all tumors (n = 1379) and in the subgroups of LN-negative (n = 1111) and untreated tumors (n = 821) (A, B), as well as in the subgroups of ER-positive (n = 856), Grade 1 (n = 141) and Grade 2 (n = 446) tumors (B), using LN-status, ER-status, and stratified histological grade (histological grade 1 and 2 vs. 3) as covariates, DMFS as endpoint and 10-year censoring. The hazard ratio and the 95% confidence interval (CI) are plotted for each of these covariates. Specified covariates may be omitted in certain comparisons, e.g. ER-status is omitted when analyzing ER-positive tumors only, or when not all of the investigated cases have clinical follow-up or clinical information for the specified covariate.

Figure 8

Schematic representation of the Mkl1 action in breast cancer. A circular Mkl1 model is depicted with four of its domains: RPEL, actin binding motifs with RPxxxEL core consensus; B1, basic domain involved in SRF-binding; SAP, homology domain found in the nuclear proteins SAF-A/B, Acinus, PIAS; TAD, transactivation domain. Serum response factor is drawn as a red shape and putative unidentified DNA-binding proteins as white shape with a question mark. Mkl1 exerts two distinct modes of action: one of them is through the B1 domain required for serum response factor (SRF)-binding activity and induction of SRF/Mkl1 target gene expression; the other one is strongly dependent on the SAP domain and triggers the expression of a specific set of pro-proliferative and pro-migratory genes that we called SAP-dependent Mkl1 target genes. High expression of SRF-dependent genes is associated with good clinical outcome for breast cancer patients, whereas elevated expression of SAP-dependent targets correlates with poor prognosis and indicates a significant role for these genes in breast cancer progression.