Background
Breast cancer (BC) is the most commonly diagnosed cancer and is the leading cause of cancer-related death in women. Different breast cancer subtypes have been described based on gene expression analysis and it is widely accepted that there are five distinct intrinsic molecular subtypes: luminal A, luminal B, HER2-enriched, normal-like and basal-like breast cancer [
1]. The majority of basal-like breast cancers are defined triple-negative breast cancers (TNBC), as they lack the expression of the estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2/neu), making them difficult to treat [
2,
3]. TNBC accounts for approximately 15% of invasive breast cancers and represents the most aggressive breast cancer subtype: indeed TNBC is a typically high-grade tumor, poorly differentiated, and associated with poor prognosis and molecular heterogeneity [
2‐
4]. Since TNBC does not respond to hormonal and target therapies, the only current therapeutic options are represented by tumor excision, radiation therapy and conventional chemotherapy. Increasing evidence points out that the microvascular density in TNBC is higher with respect to other breast cancers subtypes, thus highlighting that angiogenesis crucially supports TNBC progression. Therefore, the study of efficacious anti-angiogenic therapies in TNBC is critical [
5,
6].
One of the factors involved in TNBC aggressiveness is the High Mobility Group A1 (HMGA1), a member of non-histone chromatin proteins [
7]. HMGA1 is an architectural transcription factor which, by altering chromatin structure and interacting with transcription factors, can regulate the transcription of several genes [
8‐
10]. HMGA1 is defined as an oncofetal protein due to its expression pattern: indeed, it is highly expressed during embryogenesis, while its expression decreases or is absent in adults; it is then re-expressed in a variety of tumors, including breast cancer [
11,
12]. In this context, several works established that HMGA1 expression is correlated with high tumor grade and metastatization, resistance to therapies and poor prognosis [
7,
13‐
15]. Furthermore, a causal role of HMGA1 in breast cancer onset and progression has been demonstrated. In fact, HMGA1 over-expression in non-tumorigenic MCF-7 human breast epithelial cells leads to the acquisition of a transformed and aggressive phenotype [
16], whereas HMGA1 silencing in highly aggressive TNBC cell lines causes the reversion of the tumorigenic phenotype, as assessed both by in vitro and in vivo approaches [
13,
17]. We have previously reported that HMGA1 exerts its action by governing the transcription of gene networks fundamental in supporting TNBC aggressiveness [
13,
18‐
20]. In detail, HMGA1 induces the expression of epithelial to mesenchymal transition (EMT) - and stemness-associated genes and, through the regulation of CCNE2/CDK2 complex, it is able to modulate the Hippo pathway, finally regulating the localization and activity of YAP and conferring metastatic abilities to TNBC cells [
18]. In addition, HMGA1 regulates a gene network linked to secreted proteins. In fact, by inducing the expression of key components of the plasminogen activation system, such as PLAU and SERPINE1, HMGA1 is able to modulate the TNBC cell secretome stimulating TNBC cell migration in an autocrine way [
20]. However, little is known about how HMGA1 governs these cancer-related gene networks.
In order to identify HMGA1-molecular partners involved in developing TNBC aggressiveness, we performed RNA sequencing analysis (RNA-Seq) on the MDA-MB-231 TNBC cell line at 24 and 72 h after HMGA1 silencing, looking for transcription factors as putative upstream regulators of HMGA1-gene networks. We found Forkhead box M1 (FOXM1) transcription factor as a novel HMGA1-molecular partner. FOXM1 overexpression is detected in a variety of human cancers, where it drives the expression of critical genes involved in the regulation of different cancer hallmarks including high proliferation, invasion, drug resistance and angiogenesis; moreover, its overexpression is associated with poor clinical prognosis [
21,
22]. Intriguingly, FOXM1-associated pathway has been found to be the top up-regulated pathway in TNBC but not in other breast cancer subtypes, suggesting a crucial role of FOXM1 in TNBC [
23]. Our results indicate that HMGA1 and FOXM1 together regulate a common pro-tumorigenic TNBC gene network. Specifically, HMGA1 stabilizes FOXM1 in the nucleus preventing its degradation and increasing FOXM1-dependent transcriptional activity. Furthermore, we found that HMGA1 and FOXM1 cooperatively promote the tumor angiogenic process in in vitro and in vivo models. Our study thereby describes a new molecular mechanism fundamental in TNBC aggressiveness.
Materials and methods
Cell culture and treatments
MDA-MB-231, MDA-MB-157 and HEK293T cell lines were routinely grown in high glucose DMEM, with 10% tetracycline-free FBS, 2 mM L-Glutamine, 100 U/ml Penicillin and 100 μg/ml Streptomycin (Euroclone). Lipofectamine™ RNAiMAX reagent (Thermo Fisher Scientific) was used for transfection of 30 pmol of siRNA/35-mm dish, following the manufacturer instructions. For co-silencing experiments, 30 pmol of each specific siRNA was used up to a final amount of 60 pmol/condition. The cells were processed after 24, 48 and 72 h of silencing, depending on the specific experiment. siCTRL and siRNAs against HMGA1 and FOXM1 have been previously used [
13,
24]. Plasmid transfections were carried out using Lipofectamine 3000 (Invitrogen/ThermoFisher Scientific) for MDA-MB-231 cells following the manufacturer protocols, and the standard Calcium Phosphate transfection method for HEK293T cells. For the treatment with MG132 proteasome inhibitor (Sigma), 48 h after the siRNA transfection, MDA-MB-231 cells were treated with 10 μM of the proteasome inhibitor MG132 or DMSO, as negative control, for 6 h and then lysed in SDS sample buffer [62.5 mM Tris, pH 6.8; 2% SDS; 10% glycerol; 200 mM DTT, 1 mM Na
3VO
4, 5 mM NaF and mammalian protease inhibitor cocktail (PIC) (Sigma)] for Western blot analysis. For the cycloheximide (Sigma) treatment, 48 h after siRNA transfection, MDA-MB-231 cells were treated with 50 μM of cycloheximide and lysed at different time points (45, 90, 150 and 240 min) in SDS sample buffer for western blot analysis.
Plasmids construction
pEGFP-N1, pEGFP-N1 HMGA1a, pRL-CMV Renilla (Promega) and pGL4.11 (Promega) were already present in the laboratory. pEGFP-FOXM1 and pGL3-5BS (containing five repetitions of FOXM1 binding sites-TAAACA) were kind gifts of Dr. Muy Teck Teh (Department of Diagnostic and Oral Sciences, Blizard Institute, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK). The vectors used in this work named pGL4.10-VEGFprom (− 1000–1), pGL4.10-VEGFprom (− 1000–500), pGL4.10-VEGFprom (− 500–1), as they contain a portion of the VEGFA promoter that goes from − 1000 to − 1 bp, from − 1000 to − 500 bp and from − 500 to − 1 bp respectively, were a gift from David Mu (Addgene plasmid #66128, #66129, #66130). The deletion mutants pGL4.11-VEGFpromWT(− 388–1), pGL4.11-VEGFprom (− 338–1), pGL4.11-VEGFprom (− 172–1) and the pGL4.11-VEGFprom (− 104–1), containing a region of the VEGFA promoter spanning from − 388 to − 1 bp, from − 338 to − 1 bp, from − 172 to − 1 bp, and from − 104 to − 1 bp from the TSS respectively, were generated in the laboratory by amplifying the pGL4.10-VEGFprom (− 1000–1) with the forward primers 5′-GGGGTACCCCGGGGCGGGCCGGGGGCGGGGTCC-3′, 5′-GGGGTACCCCCTTTTTTTTTTAAAAGTCGGC-3′, 5′- GGGGTACCTGGAATTTGATATTCATTGATCCG -3′, 5′- GGGGTACCTGTATTGTTTCTC GTTTTAATTT- 3′ for the − 388, − 338, − 172 and − 104 to − 1 bp fragments respectively and the common reverse primer 5′- CCCAAGCTTAAAATCCACAGTGATTTGGGGAA - 3′. We then created a pGL4.11-VEGFpromMUT (− 388–1) mutated in two SP1 binding sites (GGGCGG ➔ GGAAGG) by amplifying the pGL4.10-VEGFprom (− 1000–1) with the forward primer 5′- GGGGTACCCCGGGAAGGGCCGGGGAAGGGG TCC -3′ (GC ➔ AA) and the reverse primer used in previous experiments. Subsequently, all the PCR products were cloned in KpnI and HindIII (Amersham Biosciences) restriction sites of pGL4.11 vector. All the plasmids generated in the laboratory were sequenced by Eurofins Genomics sequencing service.
Luciferase assay
HEK293T cells were plated at the density of 35*104 cells per 35-mm-diameter culture dish and processed 40 h after the Calcium phosphate transfection. Specifically, cells were transfected with 200 ng of the specific Luciferase reporter construct, 600 ng of pEGFP-N1 HMGA1a or/and 600 ng pEGFP-FOXM1 and 10 ng of pRL-CMV Renilla expression vector (Promega), as a normalizer for transfection efficiency. The luciferase assay was also carried out by transfecting 30 pmol of HMGA1 siRNA/35-mm dish followed, 24 h later, by the transfection of 200 ng of the reporter construct and 600 ng of pEGFP-FOXM1. The Dual-Luciferase® Reporter Assay System (Promega) was used for the luciferase reporter assay, following the manufacturer instructions. The measurements were carried out using the Berthold Lumat LB 9501 Tube Luminometer; two technical replicates were performed for each sample.
Immunoblotting
Cells were washed in ice-cold PBS and then lysed in SDS sample buffer, supplemented with proteases inhibitors, as reported before. Total lysates were separated by SDS-PAGE and then transferred to nitrocellulose membrane ∅ 0.2 μm (GE Healthcare, WhatmanTM). Western blot analyses were performed according to standard procedures, using the following antibodies: αHMGA1 [
13]; α-β-actin (A2066, Sigma); αFOXM1 (A301-533A-M, Bethyl Laboratories; D3F2B Cell Signaling Technology); αGFP (kind gift of L. Collavin, LNCIB, Trieste).
Co-immunoprecipitation
MDA-MB-231 cells protein extracts were prepared in an IP buffer containing 25 mM Tris/HCl pH 7.4, 0.5% v/v NP40 and 125 mM NaCl supplemented with 1 mM PMSF, 1 mM Na3VO4, 5 mM NaF and protease inhibitors cocktail (Sigma). For the Co-IP experiment, 250 μg of cell lysate was incubated with 1 μg of either α-HMGA1 or α-GFP (GTX113617 Genetex), as a negative control, in IP buffer O/N at 4 °C. Subsequently, pre-washed A/G proteins agarose beads (GE Healthcare) were blocked in PBS 0.1% Bovine serum albumin (BSA-Sigma) and 100 μl of beads in IP buffer were added to the solution of lysate/antibody and incubated 2 h at 4 °C. After the incubation, the beads were washed three times in IP buffer and proteins were eluted by boiling the beads for 4 min in SDS sample buffer and subjected to western blot analysis with the indicated antibodies.
Immunostaining
HEK293T, MDA-MB-231 and MDA-MB-157 cells were grown on glass coverslips and silenced for HMGA1 and/or FOXM1, as previously described. HEK293T cells were transfected with 1 μg of pEGFP-FOXM1 1 day post HMGA1-silencing, by Calcium phosphate method. Then cells were fixed in a solution of 4% PFA. After a permeabilization with 0.3% Triton/PBS and saturation in 5% BSA/PBS, cells were incubated with the following primary antibodies: αFOXM1 (Cell Signaling Technology); α-20S Proteasome β2 (MCP165 - sc-58410); α-20S Proteasome β5 (A10 - sc-393931). The α-rabbit and α-mouse Alexa 594 and 488 were used respectively. The images were visualized using a Nikon Eclipse e800 microscope and acquired using Nikon ACT-1 software, then analyzed by the ImageJ software analyser.
Migration assay
For the wound healing assay, MDA-MB-231 and MDA-MB-157 cells were seeded in antibiotics-free DMEM at a density of 2*105 cells/well in a 35-mm dish in biological triplicates and silenced for HMGA1 and/or FOXM1 expression, as described before. Cells were cultured to 90% confluence and then scraped with a 200-μl tip, and wound closure has been followed for 7 h. Two images for the same area were taken for each well and the wound areas were analyzed by ImageJ software.
Gene expression analysis
Total RNA from MDA-MB-231 cells was isolated following the manufacturer’s instructions of the TRIzol reagent (Thermo Fisher Scientific), subjected to DNase-I (Thermo Fisher Scientific) treatment and subsequently purified using phenol-chloroform. For quantitative RT-PCR (qRT-PCR), mRNA was reverse transcribed with Random primer by the Superscript III (Thermo Fisher Scientific), according to the manufacturer’s instructions. qRT-PCR was carried out with iQ™ SYBR Green Supermix (BIO-RAD); specific primers used are listed in Additional file
1: Table S1. The data obtained were analyzed with BIO-RAD CFX Manager software and the relative gene expression was calculated by the ΔΔCt method, using the GAPDH as a normalizer.
Preparation of cells for RNA-sequencing analysis
MDA-MB-231 cells were silenced for HMGA1 and the RNA was collected at 24 and 72 h after the silencing. Three biological replicates were made for each condition. The total RNA was then extracted and checked as described above. Then, an aliquot of RNA was reverse transcribed and the silencing of HMGA1 was confirmed by qRT-PCR.
RNA-sequencing analysis
Demultiplexed raw reads (fastq) generated from the Illumina HiSeq were checked using FASTQC tool (Version 0.11.3). All samples passed the quality standards. Then we aligned them to the reference genome (UCSC-hg19) using STAR [
25], version 2.0.1a using recommended options and thresholds. HTSeq-count (version 0.6.1) was used to generate gene counts. Starting from read counts, differential gene expression analysis was performed using EdgeR (version 1.10.1, R version: 3.2.3, [
26]) comparing the different time points using a quantile-adjusted conditional maximum likelihood (qCML) method. In order to identify the relationship between each sample and every other sample, the Euclidean distance between each pair of samples was calculated using the log-transformed values of the complete dataset. Average linkage clustering was then used to generate a sample-to-sample distance heatmap, via the cluster3 package (Cluster3:
http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm#ctv) [
27]. For statistical analyses the adjusted
p-values were generated via the Benjamini-Hochberg procedure. Finally, genes were selected as differentials with a cutoff of 0.5 for the log Fold change and 0.05 for the False Discovery Rate.
Functional analysis of differentially expressed genes
Differentially expressed genes were analyzed using GSEA [
28,
29] and Ingenuity Pathway Analysis (IPA, Ingenuity® Systems,
www.ingenuity.com) [
30]. The prediction of the transcription factors was obtained using the “upstream regulators” module (IPA suite). For every upstream regulator an overlap
p-value and a z-score were calculated: the
p-value indicates the significance based on the overlap between dataset genes and known targets regulated by the molecule, while the z-score is used to infer the possible activation (z-score ≥ 1.8) or inhibition (z-score ≤ − 1.8) of the molecule based on prior knowledge stored in the proprietary Ingenuity Knowledge Base. All statistical test and calculation have been performed in R [
31] environment. For the enrichment in protein localization, genes were annotated using Uniprot (
https://www.uniprot.org) and proteins were analyzed with David/Ease [
32,
33] interrogating the Geoterm Cellular Compartment gene ontology. For Fig.
6a) we selected the most significant terms in enrichment clusters (enrichment score > 3) including nuclear, microtubules, adherence/junction and secreted/exosome terms.
Preparation of the conditioned medium (CM) for angiogenic assays
MDA-MB-231 were seeded and silenced for HMGA1 and/or FOXM1 as described and 1 day before the supernatant collection the culture medium was substituted with serum-free DMEM. After 72 h of silencing, the CM was collected, centrifuged at 240Xg at 4 °C for 5 min to deposit cells debris and stored at − 80 °C. Next, the cells were washed in PBS 1X and lysed in SDS sample buffer as described. The HMGA1 and FOXM1 protein levels were checked by western blot analysis.
Endothelial cells proliferation analysis
Human umbilical vein endothelial cells (HUVEC) have been isolated from human umbilical cords [
34] and seeded in 96-well plate at the density of 5*10
3 cells/well. The cells were incubated with HMGA1, FOXM1 and HMGA1/FOXM1 conditioned medium and the control medium for 18 h. Normal Human Serum 10% was used as conditioned medium for the positive control experiments. After a washing step, the cells were fixed and permeabilized with Fix&Perm kit (Nordic-MUbio) for 15 min at room temperature in the dark. After two washing steps, the cells were incubated with the α-Ki-67 antibody (Dako), followed by the incubation with the secondary antibody α-mouse FITC (Dako). The cells were washed twice and lysed. Measurements were performed using Infinite200 (Tecan).
Transwell migration assay of endothelial cells
HUVEC cells were seeded in 200 μl of Endothelial serum-free medium (Invitrogen) at the density of 15*104 cells /well in the upper compartment of 8 μm pore 24-transwell plate, pre-coated with human Fibronectin (Sacco) in the lower face. Then HUVEC were incubated with 500 μl of HMGA1 and/or FOXM1 CMs, and with serum-free medium as a negative control or Normal Human Serum 10% (NHS) as a positive control in the lower compartment of the transwell. After 18 h of incubation, HUVEC were lysed with lysis buffer for Coulter and the number of migrated cells were counted with Coulter Counter BD.
HUVEC at the density of 5.5*104 were plated on wells precoated with Matrigel (12 μg/ml) (Becton Dickinson) and incubated for 18 h with CMs diluted 1:2, with 20 ng/ml VEGF as a positive control or with serum-free medium as a negative control. After a 4% paraformaldehyde fixation step and staining with Phalloidin-Alexa Fluor 546 (Invitrogen), the number of tubules was counted under a Leica AF6500 microscope using LAS software (Leica).
Preparation of cells for zebrafish injection
MDA-MB-231 cells were seeded at the density of 1.2*106 per 10-mm-diameter culture dish and silenced for HMGA1 and/or FOXM1 as described before. After 40 h of silencing the cells were counted and injected in the yolk sack of zebrafish embryos.
Zebrafish xenograft
Zebrafish were raised and maintained as previously described [
35]. Embryos were generated by natural pair-wise mating and were kept and handled for all experiments in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl
2, 0.33 nM MgSO
4) and PTU (1-phenyl 2-thiourea, 0.03 mg/ml) to reduce zebrafish skin pigmentation for microscope analysis. All experimental procedures were performed conforming to the ITA guidelines (Dgl 26/2014) in accordance with EU legislation (2010/63/UE); this protocol was approved by a committee of the Italian Health Ministry (cod. 04086.N.15Y). In this study, the Tg (fli1:EGFP) (y1) zebrafish embryos were used. This transgenic fluorescent strain expresses in the entire vasculature EGFP under the control of the fli1 promoter [
36]. Before all experimental procedures, animals were properly anesthetized by using 1:100 dilution of 4 mg/ml Tricaine (Sigma-Aldrich Co., St Louis, MO, USA). Two days post fertilization (2dpf), zebrafish embryos were dechorionated and microinjected with MDA-MB-231 siCTRL and siHMGA1/siFOXM1 cells or cellular medium alone (vehicle) as control experiment. Each cells suspension was stained by 2 μg/ml DiI (Sigma Aldrich) for 10 min at 37 °C, re-suspended in DMEM medium and kept on ice before injection. Microinjections were performed with the electronic FemtoJet microinjector (Eppendorf) using borosilicate glass micro-capillaries (20 mm O.D. Fifteen millimeters I.D.; Eppendorf). Approximately 500 cells were microinjected into the yolk of each embryo, which was then maintained in E3 medium/PTU for 1 h at 28 °C. Afterwards embryos were kept at 34 °C to allow tumor cells survival and growth as previously described [
37]. Twenty-four hours after microinjection, the embryos were observed with a fluorescence microscope (Leica DM 2000). During all the procedures living animals, properly anesthetized, were positioned in 1.5% methylcellulose (Sigma Aldrich Co). The images of the tumor masses were acquired in red signal (DiI staining) and then merged with the respective bright field image by using the Leica Application Suite X (LAS X) software. Instead, to evaluate the host angiogenic response to the injected tumor cells, vessels images were acquired in green (GFP) signal and then analyzed using the ImageJ software. For both analyses, 30 animals for each cell line condition, divided into two independent experiments, were considered. For gene expression, total RNA was extracted from at least 20 embryos for each experiment using Trizol Reagent (Invitrogen, Life Technologies, Milan, Italy). RNA concentration was determined by Nanoquant (Tecan). One microgram of total RNA was reverse transcribed using M-MLV reverse transcriptase (Life Technologies). Gene expression was analyzed by qRT-PCR (Step One Plus, Applied Biosystems) using the SYBR Green system (Life Technologies). The primer sequences are listed in Additional file
1: Table S1. The gene analysis was repeated three times and the values were normalized respect to medium treated animals.
Breast cancer dataset
Data from breast cancer patients have been obtained from the TCGA BRCA dataset (updated to September 2018) using the cgdsr package for R. Survival analysis has been performed using Kaplan Meier plotter [
38] with distant metastasis-free survival (DMFS) as a read-out on a cohort of 1746 breast cancer patients and relapse-free survival (RFS) as a read-out on cohorts of 3951 breast cancer patients or 255 TNBC patients.
Statistical analysis
Data were analyzed by a two-tailed Student’s t-test, and results were considered significant at a p-value < 0.05. The results are presented as the mean and standard deviation (±SD). Specifically, a p-value< 0.05 is indicated with *, a p-value< 0.01 with ** and a p-value< 0.001 with ***.
Discussion
HMGA1 is an architectural transcription factor, widely considered a master regulator of breast cancer progression [
7,
13,
17,
54]. Indeed, it has a causal role, both at early stages, bringing the mammary epithelial cells to acquire a malignant phenotype [
55], and during breast cancer progression, by promoting cellular migration and invasion capabilities, consequently leading to the metastatization event [
13,
17]. Thus, deepening the knowledge about the molecular mechanisms that HMGA1 controls will be paramount for the discovery of new targeted and efficacious therapies. With the aim to identify new molecular partners of HMGA1 in regulating breast cancer gene networks, we performed RNA-Seq analysis of TNBC cells depleted for HMGA1 expression. Here, we have discovered FOXM1 as a novel HMGA1-molecular partner, demonstrating that HMGA1 stabilizes FOXM1 in the nucleus preventing its degradation, increasing FOXM1-dependent transcriptional activity and potentiating its crucial role in tumour angiogenesis.
It is known that HMGA1 interacts with several transcription factors and guides their action on a high number of target genes involved in many cellular processes such as cell growth, proliferation, differentiation and cell death [
7]. We found that our bioinformatic analysis on RNA-Seq data from TNBC cells silenced for HMGA1 is very robust in discovering HMGA1-molecular partners. In fact, among the putative partners obtained by our analysis, we found RB1 and E2F, whose connection with HMGA family members has been previously described. Specifically, HMGA1 enhances E2F transcriptional activity by directly binding RB1, inhibiting its tumor suppressive activity [
40]. Our data indicate that HMGA1 could modulate the activity of these factors also in breast cancer progression, bringing to light common crucial HMGA1-oncogenic pathways in different cancer types. Moreover we found FOXM1, whose pathway is the top up-regulated in TNBC [
23], as HMGA1 partner. We demonstrate that HMGA1 and FOXM1 regulate a common gene network, characterized by factors with a clear role in cancer EMT, migration, and angiogenesis, key processes involved in conferring aggressiveness to TNBC. Among the genes, CCNE2, which has been connected to the migratory ability of tumoral cells [
56], has been proved to be under the transcriptional control of HMGA1, thus promoting the migratory and invasive abilities of breast cancer cells [
18]. Furthermore, HMGA1 has been found to regulate the EMT, a process necessary to cell migration, by impacting on the WNT-beta catenin pathway, known to contribute to metastatization [
57]; indeed, HMGA1 regulates LEF1, one of the actors of this pathway [
13]. A considerable amount of data sustains the involvement of FOXM1 in several features of breast cancer progression, such as the EMT, the migration and invasion abilities of tumor cells, in which FOXM1 controls the transcription of several metalloproteinases and the EMT inducer SNAI2 [
58]. Our results demonstrate that HMGA1 cooperates with FOXM1 in regulating the expression of a common gene network, enhancing the aggressiveness of TNBC cells, highlighting a dependence of breast tumor cells on HMGA1 and FOXM1 synergic action.
We observed that HMGA1 forms a complex with FOXM1 and improves FOXM1 protein stability and nuclear localization, while in the absence of HMGA1, FOXM1 is led to proteasomal degradation. These findings suggest that HMGA1, by stabilizing FOXM1, increases its transcriptional activity. This is consistent with literature data showing that FOXM1 is tightly regulated by its molecular partners that can increase FOXM1 transcriptional activity by promoting its stabilization and nuclear localization [
42,
59,
60], as for instance it has been shown for the interaction with MTDH in glioblastoma [
61].
Angiogenesis is one of the cancer hallmarks in which FOXM1 is mainly involved, inducing matrix metalloproteinase genes as well as VEGFA [
62]. As VEGFA is a growth factor essential for normal and pathological angiogenesis, its gene is finely regulated by a plethora of transcription factors, such as Sp1/Sp3, AP-2, Egr-1, STAT3 and HIF1, integrating multiple signals [
44]. In particular, it has been demonstrated that HMGA1 regulates VEGFA gene expression in diabetic retinopathy and, by interacting with HIF1, in 3T3 L1 adipocytes [
45,
63]. Intriguingly, in this study we found that HMGA1 regulates the transcription of VEGFA, acting through two independent ways: on one side, HMGA1 potentiates the transcriptional activity of FOXM1; on the other, HMGA1 acts through Sp1 in a different promoter region. Consistently, the observation that HMGA1 interacts with Sp1 is also reported in the control of insulin receptor (IR) gene transcription, in which HMGA1, by interacting with Sp1 and C/EBP beta, facilitates the binding of both factors to the IR promoter, synergistically activating IR transcription [
51]. Similarly, HMGA1 by interacting with FOXM1 could enhance its binding to the VEGFA promoter. Our results show that HMGA1, by preventing FOXM1 degradation, increases its level and therefore this could account for the increase of FOXM1 transcriptional activity. We cannot exclude that this result is attributable to one hypothesis or the other, or to the combination of both.
Our data clearly show that HMGA1 and FOXM1, in addition to regulating VEGFA, have a strong impact on tumor angiogenesis. Previous works performed in a rat model of cerebral ischemia and in HUVEC cells showed an involvement of HMGA1 in modulating angiogenic proteins, such as Angiopoietin-1, a factor fundamental in maintaining the tumor vascularization [
64,
65]. Our data demonstrate for the first time a direct involvement of HMGA1 in the process of tumor angiogenesis. Indeed, the supernatants of TNBC cells depleted for HMGA1 and FOXM1 abolished the ability of endothelial cells to organize in vessel-like structures. In line with this aspect, we observed that MDA-MB-231 cells injected in zebrafish embryos induced abnormalities in angiogenesis of the host and that this phenotype is abolished when HMGA1 and FOXM1 are depleted. These findings suggest that HMGA1 and FOXM1 have a leading role in guiding breast cancer cells to secrete pro-angiogenic factors. In accordance to this, we previously demonstrated that HMGA1 has a profound impact in the breast cancer cell secretome, inducing the release of a pool of pro-migratory proteins that act with an autocrine mechanism [
20]. This study provides an additional explanation of the mechanism of HMGA1 in promoting TNBC aggressiveness. Indeed, in breast cancer patients a profound molecular relation between HMGA1, FOXM1 and VEGFA has been further highlighted by TCGA analysis, which confirmed a strong enrichment of VEGFA in patients that overexpressed HMGA1 and FOXM1. Moreover, the co-expression of the three factors has been found to be a poor prognostic value of DMFS and RFS in breast cancer patients.
Acknowledgments
We thank Dr. Muy-Teck Teh (Queen Mary University of London, London, UK) for kindly providing the pEGFP-FOXM1 and pGL3-5BS plasmids. We thank Dr. David Mu (Eastern Virginia Medical School, Norfolk, USA) for kindly providing pGL4.10-VEGFprom(-1000 to -1), pGL4.10-VEGFprom (-1000 to -500) and pGL4.10-VEGFprom (-500 to -1).
We thank Professor Licio Collavin (Università degli Studi di Trieste, Trieste, Italy) for the α-GFP antibody. We are grateful to Gabriel Ruiz Romero, Giuseppe Dall’Agnese, and Letizia Fontana for technical assistance.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.