Background
Turnover of the extracellular matrix (ECM) is a critical step in various aspects of tumor cell biology, e.g. in orchestrating breast cancer cell differentiation driving malignancy and metastasis [
1,
2]. Inter-α-trypsin inhibitory (ITI) proteins comprise a family of secreted serine protease inhibitors found in both the ECM and in the blood circulation [
3]. ITIs are composed of a light chain, also called Bikunin, and different homologous heavy chains (i.e. ITIHs). ITIHs are covalently linked to Bikunin and thereby form a structural and functionally unique protein with a plasma protease inhibitory activity [
4]. Beyond this the biological function of ITI heavy chains remains largely unknown. Trimming of precursor ITIH proteins at a conserved cleavage site unmasks a C-terminal amino acid [
4], which is involved in hyaluronic acid (HA) binding [
5]. Owing to that ITI heavy chains were originally referred to as serum-derived HA associated proteins (SHAPs) [
6], implicating a wide spectrum of biological activities. HA that is the major proteoglycan of the ECM interacts with a large number of HA-binding proteins (HABPs) [
4] like HA-receptors CD44 and RHAMM [
7,
8]. Unlike all other described HABPs, ITI heavy chains are covalently linked with HA [
3], whose complexation generate stable “cable-like structures” supporting ECM integrity. In 1994 Chen and colleagues showed that ITI heavy chains are involved in organizing and controlling of the cumulus-oocyte expansion [
9]. In carcinogenesis of various tumor entities, accumulating studies propose a tumor suppressive role of ITI heavy chains mediated by their ECM-stabilizing activity [
10‐
12]. ITIH1 and ITIH3, for instance, have been demonstrated to cause clear retardation of lung metastasis in vivo [
12] thereby suggesting an important role of ITI heavy chains in repressing malignant diseases independently of Bikunin.
In 2004 we identified ITIH5 as the fifth heavy chain member of the ITI family [
13]. ITIH5 contains all structural features found in ITIH1-3, including distinct functional domains (VIT and vWA) and the conserved cleavage site. Nevertheless, its expression pattern differs from that of other heavy chains, i.e. ITIH5 is abundantly expressed in the placenta and moderately expressed in various organs such as the mammary gland [
13] indicating a local, tissue-specific function. ITIH5 dysfunction has been shown to contribute to inflammatory skin diseases [
14] and obesity, thus potentially acting as regulator of human metabolism [
15]. In tumor development, downregulation of ITIH5 caused by aberrant DNA hypermethylation has been reported in breast cancer [
16,
17], bladder cancer [
18], colon cancer [
19], gastric cancer [
20] and lung cancer [
21]. Based on an integrated genomic and transcriptomic approach Wu and colleagues recently demonstrated rare somatic
ITIH5 gene mutations in lung cancer whose frequency increased up to 6% in corresponding metastases [
22]. Loss of ITIH5 expression in breast and bladder cancer has been associated with clinical parameters of malignant progression and metastasis [
16,
18,
23] predicting poor prognosis in both entities. These findings strengthen a putative role of ITIH5 as a tumor suppressor in various tumor types, but mechanisms of its function have not been described so far.
In the present study we give clear evidence that the ECM modulator ITIH5 is involved in controlling breast cancer cell migration and colonization in vitro and in vivo. Moreover, ITIH5 drives an epigenetic reprogramming that reverses the aggressive phenotype of basal-like MDA-MB-231 cancer cells to an epithelial-like phenotype involving re-expression of the well-known tumor suppressor gene DAPK1.
Discussion
Previously, we revealed that loss of ITIH5 expression caused by aberrant promoter hypermethylation is associated with poor prognosis and clinical correlates of metastasis in breast cancer [
16,
23]. In the current study, ITIH5 downregulation was abundantly found in distant metastases and intrinsic subtypes associated with poor prognosis, i.e. luminal B, HER2-enriched and basal-like breast cancer. ITIH5 loss predicted shorter overall survival of patients with non-metastatic tumors proposing a prominent role of ITIH5 especially in tumors which tend to metastasize early and whose disease management and personalized therapy is still insufficient. To give insight into ITIH5 biology going beyond the assumed role as a prognostic biomarker in breast carcinomas, we established two different stable gain-of-function models, i.e. weak-aggressive T47D and metastatic MDA-MB-231 single-cell clones overexpressing full-length ITIH5. In both cell lines ITIH5 mediated suppression of colony and cell growth while only in luminal-type T47D cells ITIH5-triggered increased programmed cell death. However, this is consistent with our recent finding in luminal-like RT112 bladder cancer cells due to ITIH5 re-expression [
18]. These data indicate that ITIH5 may control mechanisms to reduce cancer cell growth independently of a given tumor subtype or entity similar to the described function of ITIH1-3 by stabilizing ECM integrity [
9,
45,
46].
In MDA-MB-231 breast cancer cells ITIH5 induced a phenotypic switch, which to our knowledge has not yet been reported for any member of the ITI protein family before. Originally metastatic cancer cells underwent an epigenetic shift driven by ITIH5 that cause a distinct signature of expressed genes. Among others, re-expression of known tumor suppressor genes such as
DAPK1 [
42] was clearly demonstrated. As a consequence, forced ITIH5 expression led to a remarkable low-aggressive phenotype causing a reduction of lung colonies in vivo. As metastases were almost exclusively found in lungs of mice injected with cancer cells lacking ITIH5 expression, impaired tumor initiation capabilities could be suggested, a feature mainly attributed to CSC.
Mechanistically, ITIH5 expression was associated with regulation of genes involved in categories of cell adhesion and cell differentiation. Matrix adhesion of ΔpBK-ITIH5 cells was significantly enhanced on physiologically coated substrates, mimicking the basement membrane (BM). ITIH5 also altered the composition of such specialized ECM structures as the BM constituent collagen type IV was identified being upregulated. According to this, profound changes in expression of integrin cell surface receptors were demonstrated that are known to bind to the BM being involved in controlling cell adhesion and migration [
34,
47]. Because of their outside-in-signaling capacity, integrins function not only as regulators of cell adhesion but also as sensors of their extracellular environment regulating downstream signaling [
48] and it is likely that they have completely different effects on behavior of cancer cells, depending on which integrin receptors and ligands are exposed [
49]. Alterations in the profile of integrin expression as identified in ITIH5 clones have been reported to cause dramatic shifts in modes of cell migration [
34]. In particular the balance between β1, a putative metastasis suppressor in human cancer [
50], and β3 integrin is thought to play a critical role [
51]. Interestingly, increased β3 integrin was observed due to ITIH5 re-expression in MDA-MB-231 cells. Nevertheless, β1 integrin, which is almost not expressed in mock clones, is even stronger induced in ITIH5 clones so that the balance between β3 and β1 integrin was clearly shifted towards β1. While β3 integrin has been reported being associated with Rac1 activation, β1 integrin regulates in particular RhoA activity [
34]. This notion is important because Rac1 facilitates F-actin polymerization and locally decreases cell-membrane tension that lead to lamellipodia formation during the first step of cell migration. Its activity is blocked by RhoA GTPases in the second phase of cell migration regulating actomyosin contractility [
52].
Already in 2005, Danen et al. reported that integrin αVβ3 promotes directional cell migration in the absence of integrin α5β1 being characterized by a single large lamellipodium and lower RhoA activity [
53,
54] as also obvious in mock control cells. In turn, α5β1 is particularly efficient at promoting later phases of cell spreading by supporting strong RhoA-mediated contractility and random migration. In our ΔpBK-ITIH5 model we showed that ITIH5-expressing MDA-MB-231 cells were not able to disseminate from neighboring cells moving as single-cells directional into the wounded area. As a consequence ITIH5-expressing significantly higher contractile cell forces compared to their mock clones. This result is in good agreement with the simultaneous upregulation of active RhoA-GTPases in ITIH5 clones, which are known to mediate matrix adhesion-dependent cell forces via Rho/Rock signaling cascades [
55] giving a mechanistic explanation for the high-adhesive, well-differentiated phenotype. These findings were associated with clustering of ΔpBK-ITIH5 cells and with reduced polarization into a distinct protrusive front and a retracting rear end. Truong et al. have recently reported that functional inhibition of β1 integrin converted the migratory behavior of human triple-negative breast cancer (TNBC) cells from collective to single-cell movement facilitating lung colonization in vivo [
56]. Moreover, β1 integrin promotes an epithelial phenotype in those TNBC cells by restoring, for instance, E-cadherin expression in a TGF-β dependent manner. Hence, upregulation of desmosomal components like DSP and DSC2 linking neighboring cells may contribute to tightly organized colony structures of ITIH5-expressing MDA-MB-231 cells impairing mesenchymal single-cell migration.
It is astonishing that expression of a single ECM factor in vitro, i.e. ITIH5, can effect hyper- or hypomethylation of more than 1500 CpG sites in metastatic cancer cells. The term “epigenetic reprogramming” is commonly used to describe profound alterations in the epigenetic makeup (e.g. [
57,
58])—and therefore appears to be justified in this context. Addressing the question why those DNA regions showed differences in DNA methylation, we focused on mechanisms known to be involved in regulating DNA methylation dynamics. So far increasing evidence suggest that histone modifications, namely H3K27Me3 and H3K4Me3, and associated PcG and trithorax-group (trxG) proteins are not only critical for changes in gene expression upon embryonal stem (ES) cell differentiation [
59], but also for development of cancer (stem) cells [
60‐
63]. Cross talk between histone methylation marks and DNA methylation is thought to regulate DNA methylation dynamics via recruiting proteins like DNA methyltransferases (DNMTs) [
64]. In agreement with that, GSEA analysis revealed highly significant enrichment of genes harboring targets of the Polycomb protein SUZ12. By correlating corresponding CpG positions with histone modification marks as described by Ku et al. [
39], 214 promoters were identified that have been previously reported being marked by either H3K4Me3 and/or H3K27Me3 in ES cells and have changed their DNA methylation status in ITIH5 clones. Importantly, genes associated with both H3K27Me3 alone and a combined, i.e. with a potentially bivalent H3K4Me3 and H3K27Me3 status, were significantly overrepresented. Thus, enrichment of promoter regions associated with dynamics in H3 methylation could indeed contribute to the epigenetic shift allowing distinct DNA demethylation patterns as observed for the DAPK1 5’UTR sequence close to the TSS.
DAPK1 is a calmodulin-regulated and cytoskeleton associated serine/threonine kinase [
65,
66]. Accumulating evidence suggest that DAPK1 plays an important role in tumor suppression. Epigenetic silencing of
DAPK1 has been demonstrated to correlate with higher risk for recurrence and metastasis in various tumor entities [
42]. DAPK1 is a pro-apoptotic factor (e.g. [
67]) that abrogates matrix survival signals by inside-out inactivation of β1 integrin impairing the p53-apoptosis pathway [
68]. Aside of its apoptotic function Kuo and colleagues postulated an apoptosis-independent mechanism of DAPK1, i.e. uncoupling of stress fibers and focal adhesions by modulation of integrin adhesion [
43]. This study fits to our observation that the cytoskeleton was re-organized in DAPK1-expressing ΔpBK-ITIH5 cells. It has been shown that DAPK1 mediates a disruption of the cell polarity by blocking the Rho-GTPases cdc42 in MDA-MB-231 cells leading to inhibition of cell migration in a wound healing assay [
44]. Consistent with that, knockdown of DAPK1 had restored motile capacities, at least in part, of ITIH5-expressing MDA-MB-231 cells, indicating involvement of DAPK1 in the RhoA-β1-integrin-mediated signaling axis. A cartoon summarizing these finding is illustrated in Fig.
9e.
Underlying mechanisms of the epigenetic shift induced by ITIH5 in basal-type breast cancer cells and the putative role of specific ECM components and receptors appear complex, and must be addressed in future studies. As luminal T47D cells already grow in epithelial-like clusters, it makes sense that ITIH5 did not trigger a similar effect in those already well-differentiated tumor cells. Beyond that different settings of cell-surface receptors might explain a responsibility for ITIH5-mediated functions such as HA-crosslinking in dependence of a given background. For instance, MDA-MB-231 cells highly express CD44, a known HA-receptor facilitating metastatic CSC-like features [
69], whereas T47D has been previously characterized as CD44
low [
70]. Since Mina Bissell postulated a profound impact of the ECM and regulatory proteins on cell differentiation [
1] already in 1982 [
71], it is by now well described that epigenetic gene expression control such as chromatin remodeling [
2,
72] can be orchestrated by signals from the cellular microenvironment. Biomechanical cues as modified by ITIH5 are thought to contribute to global internal organization of nuclei [
73,
74] controlling chromatin structure [
36]. Irrespective of that our data underline the complex but fundamental effects of the ECM and its constituents on cell phenotypes and differentiation in the context of malignant progression.
Methods
Animals
Female BALB/cnu/nu mice were purchased from Charles River Laboratories International (Wilmington, MA). All animal procedures and experiments were conducted in accordance with the German federal law regarding the protection of animals. The respective protocols were approved by the administration of the “Landesamt für Umwelt, Natur und Verbraucherschutz” (LANUV, Recklinghausen, Germany - AZ 87-51.04.2010.A226). For the care of laboratory animals, Guide for the Care and Use of Laboratory Animals (National Institutes of Health publication 86-23, 1985 revision) was followed.
TCGA data set
Data from breast cancer, normal and metastatic tissues were used from The Cancer Genome Atlas (TCGA) [
25], comprising overall patients’ data of an independent platform: Gene expression IlluminaHiSeq (n = 1215). The data of this study can be explored using the cBio Cancer Genomics Portal (
http://cbioportal.org).
Cell lines and reagents
Breast cancer cell lines T47D and MDA-MB-231 were obtained from the American Type Culture Collection (ATCC, Manassas, VA), which assures molecular authentication of cell] lines [
77], and was resuscitated before using in experiments. Otherwise cell lines were authenticated, within 12 months of being used in the study and were cultured as described previously [
78] and regularly tested for mycoplasma infection using the PCR-based Venor
® GeM Mycoplasma Detection Kit (Minerva Biolabs, Berlin, Germany).
Transfection and single-cell cloning of T47D and MDA-MB-231 cells
Transfection of both T47D and MDA-MB-231 cells with ITIH5-pBK-CMV expression vector, containing the full-length human ITIH5 cDNA derived from normal breast tissue, was performed as recently described [
16]. Single-cell clones were selected by limited dilution under geneticin (G418) pressure (T47D: 400 μg/ml; MDA-MB-231: 1000 μg/ml).
RNA interference of DAPK1
Human ΔpBK-ITIH5 and mock clones were transfected with HiPerfect transfection reagent (Qiagen) applying two siRNA sequences directed against DAPK1 alone (#1: Hs_DAPK1_6, Cat. No. SI02223781, 5’-CGGCTATTACTCTGTGGCCAA -3’ and #2: Hs_DAPK1_6, Cat. No. SI02223774, 5’- AAGCATGTAATGTTAATGTTA.-3’ (20 nM each)), or in combination of both according to the manufacturer’s instructions. Cells were treated every 48 h with siRNA sequences to ensure sufficient DAPK1 knockdown. Commercial non-silencing control siRNA (nc siRNA) (5’-AATGCTGACTCAAAGCTCTG-3’) served as negative control. Knockdown was verified by RT-PCR and western blot analysis after 48, 96 and 144 h. Functional studies were started immediately after 48 h siRNA treatment.
Nucleic acid extraction and reverse transcription PCR
Total cellular RNA from cultured cells and tumor nodules of mice lungs (samples pooled for test group) was prepared by using TRIzol reagent (Invitrogen). cDNA was synthesized using the reverse transcription system (Promega, Madison, WI) as previously described [
16].
Real-time PCR
cDNAs were amplified by real-time PCR using SYBR-Green PCR mix (Bio-Rad Laboratories, Munich, Germany) performed in an iCycler IQ5 (Bio-Rad Laboratories) and quantified by the comparative C
T method calculating relative expression values as previously described [
79]. All used primers spanned at least one intron, and are listed in Additional file
5.
In vitro demethylation
Whole-genome demethylation of human stable MDA-MB-231 clones was performed as recently published [
80]. In brief, demethylation agent 5-aza-2’-deoxycytidine (DAC) was added to a final concentration of 5 μM on days 1, 2 and 3. On day 3 cells were additionally treated with 300 nM trichostatin A (TSA) (Sigma-Aldrich). Cells were harvested on day 4 for RNA and DNA extraction.
Bisulfite-modification and methylation-specific PCR (MSP)
Bisulfite conversion and MSP reaction conditions of in vitro derived DNA was performed as specified previously [
81]. For used
DAPK1 MSP primers and cycle conditions see Additional file
6.
Pyrosequencing
Pyrosequencing of 14 CpG sites within the
DAPK1 5’UTR region was performed by using the PyroMark PCR Kit (Qiagen) for initial fragment amplification. The PyroMark96 ID device and the PyroGoldSQA reagent Kit (Qiagen) were used as previously described [
18]. The DAPK1 assay was designed by using the Pyromark Assay Design Software (Qiagen) and all primers are listed in Additional file
7.
GTPases pulldown
Activation of both Rac1 and RhoA was measured by using the
Active Rac1 Detection Kit (#8815, Cell Signaling, Danvers, MA, USA) and the
Active Rho Detection Kit (#8820, Cell Signaling) respectively, according to the manufacturer’s instructions. In brief, single-cell ΔpBK-ITIH5 and mock clones were cultured in G418 containing growth medium for 48 h. Subsequent to the cell lysis, 550 μg of total cell protein lysate for each clone was mixed with 20 μg of GST-PAK1-PBD capturing (active) RAC1-GTP or GST-Rhotekin-RBD for RhoA. Glutathione matrix-immobilized Rac1-GTP or Rho-GTP was eluted in SDS sample buffer supplemented with DTT. After heat denaturation (5 min, 95 °C) Rac1 and RhoA proteins were detected by western blot analysis using specific antibodies (see Additional file
8). Total cellular RAC1 or RhoA protein was determined for each sample and used for normalization.
Western blot
Western blot analysis was performed as previously described [
82] but slightly modified as following: Proteins were extracted in RIPA lysis buffer, then separated in 4–12% Bis-Tris gels (Invitrogen Life Technologies, Darmstadt, Germany) under reducing (50 mM DTT) conditions using MES-SDS running buffer and electroblotted onto nitrocellulose membranes (0.2 μm). Commercial primary antibodies used are listed in Additional file
8. The generated anti-ITIH5 antibody was previously characterized [
18]. Equal protein loading was monitored by using β-actin specific antibody.
Immunofluorescence
MDA-MB-231-ITIH5 ΔpBK-ITIH5 and mock clones (3 × 104 cells/well) were plated onto 12 mm round glass coverslips. After 24 h incubation, cells were fixed with 4% paraformaldehyde (PFA) and 0.5% Triton X-100 in cytoskeleton buffer (10 mM PIPES, 150 mM NaCl, 5 mM EGTA, 5 mM glucose, and 5 mM MgCl2, pH 7.0) for 10 min at room temperature. Afterwards, cells were gently washed twice with PBS and post-fixed with 4% PFA for 10 min at room temperature. Subsequently, cells were washed thrice with cytoskeleton buffer. For vinculin labeling, cells were incubated with the monoclonal antibody hVIN-1 (Sigma-Aldrich, Deisenheim, Germany) for 30 min at room temperature followed by Alexa 488-conjugated goat anti-mouse IgG (Molecular Probes, Eugene, OR). The actin cytoskeleton was labelled with Alexa 594-conjugated phalloidin (Molecular Probes). Coverslips were mounted in Prolong (Molecular Probes). Specimens were observed using an Axiovert 200 microscope (Zeiss, Jena, Germany) equipped with a Plan-Apochromat 100×/1.40 NA oil immersion objective in combination with 1.6× or 2.5× optovar optics. Images were recorded with a cooled, back-illuminated CCD camera (Cascade, Photometrics, Tucson, AZ) driven by IPLab Spectrum software (Scanalytics Inc., Rockville, MD).
Scanning electron microscopy
Cells were fixed in 3% glutaraldehyde (in 0.1 M Soerensen’s phosphate buffer [13 mM NaH2PO4 × H2O; 87 mM Na2HPO4 × 2H2O; pH 7.4]) for at least 1 h, then rinsed in 0.1 M Soerensen’s phosphate buffer. Next, cells were dehydrated in a graded ethanol series (30, 50, 70, 90, 3% × 100%) and critical-point-dried in carbon dioxide (CPD 010, Balzers Union, FL). The dried samples were fixed on SEM stubs and sputter-coated with gold (SCD 030, Balzers Union), then analyzed with an ESEM XL 30 FEG (FEI Philips, Eindhoven, Netherlands) in high vacuum mode at an accelerating voltage of 10 kV .
Cell attachment assay
Cell adhesion experiments were carried out as previously described [
79] with minor modifications: Six-well plates were coated with HA (100 μg/ml; Sigma-Aldrich) or Matrigel™ (10 μg/ml; Sigma-Aldrich) and cells (5 × 10
5 cells/well) were incubated to adhere on surface for 30 min at 37 °C. Attached cells were fixed with 70% ethanol for 10 min and stained with 0.1% crystal violet. After 20 min cells were exhaustively washed with water and dried overnight. The dye was dissolved in 0.002% Triton X-100 in 100% isopropanol and carried over into a 96-well plate to measure the optical density at 590 nm using an ELISA reader (SpectraMax 340; Molecular Devices; CA).
Fabrication of silicone rubber substrates
Substrate preparation and characterization of elastomer material properties (Young’s modulus and Poisson’s ratio) were performed as previously described [
83]. In brief, cross-linked elastomeric silicone rubber was used (Sylgard 184, Dow Corning), which is supplied as a two-component kit consisting of base and cross-linker oil. Both components were mixed at a ratio of 1:50 and mixed with 5% (v/v) yellow-green fluorescent nanobeads (0.2 μm diameter, FluoSpheres, Invitrogen). This pre-polymer mixture was applied onto a micro-structured silicon dioxide mold containing 500 nm high microdots with an edge length of 2.5 μm and a lattice constant of 3.5 μm, to generate a regular bead layer within the elastomeric substrate. The polymer layer was then covered by a glass coverslip. A defined layer thickness of 80 μm was produced by putting spacers between the silicon surface and the coverslip. Pre-polymer mixtures were heat cross-linked (60 °C) overnight and finally displayed a Poisson’s ratio of 0.5 and a Young’s modulus of 15 kPa. For cell culture, the silicon mold and spacer were removed and glass bottom covered elastomer substrates were glued to a 3.5 cm Petri dishes with 1.5 cm holes.
Traction force microscopy and cell force retrieval
Live cell analyses were performed at 37 °C and 5% CO
2 (cell incubator XL2, Carl Zeiss, Germany) using an inverted confocal laser scanning microscope (cLSM710, Carl Zeiss, Germany), utilizing a 40× EC Plan-Neofluar oil immersion objective (PH3, NA = 1.3, Carl Zeiss, Germany). Images were taken using the imaging software ZEN 2.1, Carl Zeiss Germany). Confocal micrographs of the cells (phase contrast) and of yellow-green fluorescent beads were taken using an argon ion laser (488 nm) with a transmitted light detector and a 490–530 nm bandpass filter, respectively. Cells were seeded onto fibronectin-coated (20 μg/cm
2) TFM substrates 48 h before measurement. Only well-adhered cells were analyzed. Traction forces applied by a single cell to an elastic substrate of defined stiffness cause deformations fields that were visualized by tracking fluorescent marker beads in the substrate. From the displacement of these particles cell forces were calculated. Substrate deformation was captured in the presence of cells and substrate relaxation was obtained after cell elimination by trypsinization. Cell area force fields (AFF) were retrieved from vector displacement fields (DVF) determined by correlating the nanobead displacement in the deformed and the relaxed, cell-free elastomer. MatLab-based algorithms were used for data processing as previously described [
29,
84].
XTT cell proliferation assay
The XTT proliferation assay (Roche Diagnostics, Mannheim, Germany) was used and performed as previously described [
16].
Apoptosis assay
Activity of the effector caspases 3 and 7 in ITIH5 and mock single-cell clones was analyzed by using the Apo-One® Homogeneous Caspase-3/7 Assay (Promega, Mannheim, Germany) according to the manufacturer’s instructions. Briefly, cells (1.5 × 104) were seeded in 96-cell culture wells and incubated overnight (20% O2, 5% CO2, 37 °C). Afterwards, staurosporine (1 μM, Sigma-Aldrich, Deisenhofen, Germany) was applied to induce apoptosis. Fluorescence intensity was quantified by using an ELISA plate reader (excitation: λ = 485 nm; emission: λ = 577 nm).
Colony formation assays were performed as previously described [
79]. In vitro motility was analyzed performing a monolayer scratch wound assay as previously specified [
85].
MDA-MB-231 cells (3 × 106) of the ITIH5 test set (ΔpBK-ITIH5 clones) or the control set (ΔpBK-mock clones) were intravenously inoculated into the lateral tail vein of 7 week old female Balb/cnu/nu mice. After 50 days, mice were μCT scanned, and then sacrificed. Lungs were harvested, photographed with the Discovery V12 stereomicroscope (Zeiss), analyzed with DISKUS software package (Königswinter, Germany), formalin-fixed (10%) and paraffin-embedded. H&E-stained sections from each lung tissue as well as a further slide sectioned at 30 μm increments in the vertical plane were examined by a pathologist in a blinded manner to quantify the number of micro-metastases.
In vivo micro-computed tomography
Whole-body scans of mice were performed using non-invasive μCT. A gantry-based dual energy micro-computed TomoScope 30s Duo (CT Imaging, Erlangen, Germany) was used. Matched pairs of mice (
n = 7 each) were scanned 50 days after tumor cell injection and anaesthetized using a 1.5% isoflurane inhalation narcosis. Mice were scanned both natively and after intravenous application of eXIA™160 (Binitio Biomedical, Ottawa, Canada), an iodine-based and radiopaque blood pool contrast agent. Injected dose of 0.1 ml/20 g body weight was used [
86]. Images were reconstructed using a Feldkamp type reconstruction (CT-Imaging, Erlangen, Germany) generating a voxel size of 70 × 70 × 70 μm
3. Subsequently, images were analyzed using Amide [
87]. 3D architecture was visualized using Imalytics Preclinical software [
88].
Gene expression profiling
Gene expression profiling of the ITIH5 test set (three independent MDA-MB-231 ΔpBK-ITIH5 clones) and the control set (three independent MDA-MB-231 ΔpBK-mock clones) was carried out by the IZKF Chip-Facility (Interdisciplinary Centre for Clinical Research Aachen within the Medical faculty of the RWTH Aachen University) using the Affymetrix 1.0 ST gene array (Affymetrix, Santa Clara, CA).
Profiling of stably transfected MDA-MB-231 breast cancer cells was performed using BRB-ArrayTools developed by Dr. Richard Simon and BRB-ArrayTools Development Team version 4.3.0 – Beta. In order to identify the significantly regulated candidate genes the
class comparison evaluation was used [
89], which met the following criteria: Significantly (
p < 0.05) differentially expressed with a minimal change in expression by 3-fold. Exact permutation
p-values for significant genes were computed based on 35 available permutations. Genes were excluded when less than 20% of expression data had at least a 1.5-fold change in either direction from gene’s median value. Gene Ontology (GO) categories were determined by applying a gene set comparison analysis that is similar to the gene set enrichment analysis described by Subramanian et al. [
90]. Tests used to find significant gene sets were: LS/KS permutation test (to find gene sets which have more genes differentially expressed among the phenotype classes than expected by chance). Over-represented GO lists were considered significant when the threshold of determining significant gene sets is equal or below 0.005 (LS/KS permutation test).
DNA methylation profiling
DNA methylation profiles were analyzed in three independent MDA-MB-321 ΔpBK-ITIH5, two mock clones and WT by using the HumanMethylation450 Beadchip technology (Illumina, San Diego, USA). Hybridization of bisulfite converted DNA (200 ng) and initial data evaluation was performed by the DKFZ Gene Core Facility (Heidelberg, Germany).
Limma-
T-test statistics was calculated in
R [
91] to select for CpG sites with significant differences in DNA methylation (adjusted
p value <0.05 and 20% differential DNA methylation level between both test groups). Cluster analysis of the CpG sites was performed with the “pheatmap package” for R using complete linkage and Euclidean distance [
92]. The Gene Ontology analysis was performed using the
GOrilla software tool to visualize GO terms of target (1511 GpG sites) and background list (all analyzed CpG sites) [
93]. Overlap of significantly hyper- and hypomethylated CpG sites between ΔpBK-ITIH5 and ΔpBK-mock clones with gene set data bases was performed using a public gene set enrichment analysis platform (GSEA;
http://www.broadinstitute.org/gsea/index.jsp) [
90,
94]. The probes / CpG sites of the HumanMethylation450 BeadChip were furthermore annotated with previously published data on the presence of two histone H3 modifications (H3K4Me3 and H3K27Me3) close to a transcription start site in embryonic stem cells [
39]. We used the information on the probed location (GRC36 reference) provided by the manufacturer (HumanMethylation450 v1.2 Manifest File). A promoter region that contained at least one probed CpG site with a significant difference in DNA methylation level was called deregulated (Additional file
3). The subsequent analysis was limited to the 12,564 (69%) regions with a minimum of 5 probed CpG sites to reduce the bias introduced by a low coverage. Methylation β-values of multiple significant different methylated CpG sites were averaged after transformation to M-values.
Statistics
Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software Inc., La Jolla, CA) and SPSS 20.0 (SPSS, Chicago, IL). Differences were considered statistically significant if the two sided p-values were equal or below 5% (≤0.05). To compare two or more groups the Mann-Whitney or Kruskal-Wallis test was used, respectively. Overall survival (OS) was measured from surgery until death and was censored for patients alive at the last follow-up using the univariate log-rank tests.
Acknowledgement
The excellent technical assistance of Roswitha Davtalab, Hiltrud Königs and Sonja von Serényi is thankfully acknowledged.
This paper is dedicated to the memory of our wonderful colleague Dr. Jürgen Veeck, who recently passed away in his personal fight against cancer.