AGE‐modified basement membrane cooperates with Endo180 to promote epithelial cell invasiveness and decrease prostate cancer survival

Biomechanical strain imposed by age‐related thickening of the basal lamina and augmented tissue stiffness in the prostate gland coincides with increased cancer risk. Here we hypothesized that the structural alterations in the basal lamina associated with age can induce mechanotransduction pathways in prostate epithelial cells (PECs) to promote invasiveness and cancer progression. To demonstrate this, we developed a 3D model of PEC acini in which thickening and stiffening of basal lamina matrix was induced by advanced glycation end‐product (AGE)‐dependent non‐enzymatic crosslinking of its major components, collagen IV and laminin. We used this model to demonstrate that antibody targeted blockade of CTLD2, the second of eight C‐type lectin‐like domains in Endo180 (CD280, CLEC13E, KIAA0709, MRC2, TEM9, uPARAP) that can recognize glycosylated collagens, reversed actinomyosin‐based contractility [myosin‐light chain‐2 (MLC2) phosphorylation], loss of cell polarity, loss of cell–cell junctions, luminal infiltration and basal invasion induced by AGE‐modified basal lamina matrix in PEC acini. Our in vitro results were concordant with luminal occlusion of acini in the prostate glands of adult Endo180ΔEx2–6/ΔEx2–6 mice, with constitutively exposed CTLD2 and decreased survival of men with early (non‐invasive) prostate cancer with high epithelial Endo180 expression and levels of AGE. These findings indicate that AGE‐dependent modification of the basal lamina induces invasive behaviour in non‐transformed PECs via a molecular mechanism linked to cancer progression. This study provides a rationale for targeting CTLD2 in Endo180 in prostate cancer and other pathologies in which increased basal lamina thickness and tissue stiffness are driving factors. Copyright © 2014 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.


Introduction
Structural changes in the basal lamina during ageing include altered matrix composition, organization and progressive thickening [1,2]. The extent of these alterations is organ-specific and exacerbated by metabolic or endocrine disorders [3][4][5]. Age-related reorganization of the basement membrane in prostate gland acini was reported four decades ago and indicated that its thickness can increase from 400-700 Å in younger (4-6 months) mice to 3000 Å in older (29-36 months) mice [3]. Given that progressive prostate enlargement and stiffening occurs during the common age-related

M Rodriguez-Teja et al
Endo180 is a multiple C-type lectin-like domain (CTLD) receptor that binds to extracellular collagens via its fibronectin type II (FNII) domain [7] and glycosylated forms of collagen via its second CTLD, CTLD2 [8]. Endo180 is a highly sensitive and specific serum marker of metastatic disease in breast cancer [9], and its expression correlates with Gleason score and overall survival in prostate cancer [10,11]. Endo180-dependent prostate cancer progression involves its functional switch from suppressor to promoter of epithelial-mesenchymal transition (EMT) upon its disassociation from CD147 (EMMPRIN, Basigin) and relocalization from the plasma membrane to endosomes in PEC acini [11]. Given that endosomal Endo180 enhances myosin-light chain-2 (MLC2) phosphorylation and cell migration [12], we postulated that the basal lamina stiffness triggers Endo180-dependent PEC invasiveness in the ageing prostate gland.

Analysis of mouse tissue
Wild-type and Endo180 ΔEx2 -6/ΔEx2 -6 mice were genotyped for the Endo180 allele by Southern blot analysis [17,18]. Prostate glands were excised and fixed in 4% w/v paraformaldehyde prior to processing. Sections were deparaffinized in 100% v/v xylene (30 min, twice) and hydrated in graded ethanol solutions (100%, 95%, 70% v/v) and then double-distilled water. Antigen retrieval was performed using 20 μg/ml Proteinase K diluted in TE buffer at 37 ∘ C (30 min), boiling in 1 mM EDTA, pH 8 (10 min), washing in PBS (10 min, three times) and inactivation of endogenous peroxidase in 3% v/v hydrogen peroxidase (10 min). After washing in PBS (5 min, three times), the samples were incubated (1 h) in blocking solution (PBS +20% w/v FBS). Sections were incubated with anti-collagen IV (R1041, Acris Antibodies) or anti-laminin (sc-5583, Santa Cruz, 1/50 in blocking solution) at 4 ∘ C (16 h), washed in PBS (5 min, three times), incubated with AlexaFluor-488 goat anti-rabbit IgG (H + L) diluted in blocking solution (1 h), washed in PBS (5 min, three times), counterstained with DAPI, washed with PBS (5 min, twice) and mounted in Vectashield ® . Images were acquired using a Zeiss S100 Axiovert epifluorescent/brightfield microscope. Densitometric analysis was performed using ImageJ algorithms to calculate the ratio of immunostained area (pixels) divided by the average intensity of background (pixels). To analyse luminal obstruction, the following assumption was made: one luminal epithelial cell layer = hollow; two luminal epithelial cell layers = partial obstruction; and more than two luminal epithelial cell layers = complete obstruction. Collagen IV and laminin intensities and luminal obstruction were calculated from sections of anterior and ventral prostates obtained from five wild-type and six Endo180 ΔEx2-6/ΔEx2-6 adult (age >18 months) mice.

Analysis of human tissue
Tissue biopsies on the NCLPC1 and NCLPC4 prostate cancer TMAs used in this study were obtained with consent and approved by the Research Ethics Committee of Newcastle University Medical School. IHC was carried out using a Bond III fully automated staining system and Bond Polymer Refine Detection (Leica Microsystems). Antigen retrieval was carried out using Epitope retrieval, pH 6 (100 ∘ C, 30 min). 1/2500 rabbit anti-AGE pAb (ab23722, Abcam; performed at Imperial College London) or 1/50 anti-Endo180 (39.10; performed at Newcastle University) were applied to tissue sections (30 min) and 1× diaminobenzidine was used as a chromogen. The sections were counterstained with Mayer's haematoxylin (MHS1, Sigma-Aldrich) and mounted in Faramount aqueous medium (S3025, Dako). Immunostaining was quantified as percentage positive epithelial cells (0-100) × staining intensity (0-3), giving possible scores of 0-300. Two cores were evaluated from each tumour and the mean calculated. If one core was unavailable, the value for the remaining core was used. A cut-off score of 30 defined positive versus negative tumours was used. This immunohistochemical scoring system is in routine use in the Departments of Medicine and Surgery and Cancer at Imperial College London [19].

Results
Advanced glycation end-product-dependent, non-enzymatic crosslinking of the basal lamina as a model for its structural alteration during ageing To simulate increased thickness and stiffness of the basal lamina associated with ageing, glycolaldehyde was used to non-enzymatically crosslink rBM matrix by promoting Schiff base adduct formation, Amadori rearrangement and subsequent advanced glycation end product (AGE) generation. AGE generation was determined by the measurement of fluorophores in glycolaldehyde-treated rBM [13]. Fluorophore production associated with AGE generation was effectively blocked by the Schiff base reducing agent, sodium cyanoborohydride, and the reactive aldehyde quencher, aminoguanidine (see supplementary material, Figure  S1A). The formation of high molecular weight rBM protein fragments that were not able to penetrate SDS-PAGE gel (see supplementary material, Figure  S1B) and structural rearrangement of the rBM major matrix components, collagen IV and laminin, into dense bundles (see supplementary material, Figure S1C, D), confirmed that glycolaldehyde treatment induced the formation of crosslinks. Rheometric analysis confirmed that there was a significant increase in the elastic moduli of rBM following its treatment with glycolaldehyde for 6 h (175 ± 90 Pa) or 14 h (322 ± 160 Pa), compared to its treatment with PBS (120 ± 55 Pa) (see supplementary material, Figure S1E). Herein, the following nomenclature describes the different biomechanical properties of rBM after these three pretreatments: 'native' (14 h PBS); 'semi-stiff' (6 h glycolaldehyde); and 'stiff' (14 h glycolaldehyde). Importantly, the 3.2 ± 2.0-fold increase in elastic modulus after 14 h pretreatment with glycolaldehyde was concordant with the 2.5-to 3.4-fold increase in stiffness observed in malignant versus normal prostate tissue (see supplementary material, Figure  S1E, Table S1) [20][21][22][23][24][25].

AGE-modified basal lamina induces PEC invasiveness
To ascertain whether basal lamina stiffness can promote tumourigenic behaviour in normal prostate epithelia, we compared three-dimensional (3D) acinar structures generated by non-transformed PECs (RWPE1 cells) in native, semi-stiff and stiff rBM. AGE generation in stiff rBM was detected by immunostaining of pentosidine, an indicator of collagen IV crosslinking [26] and a biomarker for tissue stiffness [27], and was associated with increased thickness of the collagen IV and laminin layer that assembled around PEC acini ( Figure 1A). The marked depletion of collagen IV and laminin at points of PEC protrusion from the basal surface of the acini, together with mislocalization of markers of cell-cell adhesion (E-cadherin) and apical polarity (GM130; early endosomal antigen-1, EEA1) induced by glycated and crosslinked rBM ( Figure 1A), suggests that a stiff microenvironment contributes to the induction of their invasive behaviour. The fact that stiffened rBM did not promote polyADP ribose polymerase (PARP) cleavage (see supplementary material, Figure  S2) confirmed that the levels of AGE generation and stiffness used in our model had no effect on cell viability. In keeping with these findings, acinar morphogenesis proceeded at similar rates on native and stiff rBM (see supplementary material, Figure S3). Comparison of PEC acini in native and stiff rBM ( Figure 2A) revealed a significant transition towards an irregular (polygonal) shape ( Figure 2B), accompanied by a 0.8-fold reduction in size ( Figure 2C) and a 0.4-fold decrease in luminal:total area ratio ( Figure 2D) on the latter. The respective 1.8-fold and 2.8-fold increases in the number of acini with PECs visibly protruding from their basal surface into semi-stiff and stiff rBM ( Figure 2E) confirmed that high AGE content and stiffened basal lamina induces invasive behaviour in non-transformed PECs. These findings were validated in an alternative experimental system, in which ribose was used to non-enzymatically crosslink rBM and shown to induce similar changes in PEC acinar morphology (see supplementary material, Figure S4A-D). The promotion of PEC invasion from the basal surface of acini, breakage of the basal lamina at points of PEC protrusion, reduction in luminal area and loss of epithelial cell polarity facilitated by AGE-induced stiffness of the basal lamina suggests that this biophysical change can trigger invasive behaviour in the prostate gland.

AGE-modified basal lamina induces Endo180-dependent contractile signalling and PEC invasiveness
In relation to the induction of PEC invasiveness by AGE-dependent basal lamina stiffness, we tested the hypothesis that CTLD2, which directly binds to glycosylated collagens, including basement membrane collagen IV [8], is a mechanotransducer that directs the response of PECs to this environmental change ( Figure 1B). In support of this theory, immunostaining with a mouse anti-human Endo180 monoclonal antibody (mAb), A5/158, that recognizes CTLD2 ( Figure 1B) [15], revealed that endogenously expressed Endo180 is strongly localized at the basal surface of PEC acini in native rBM ( Figure 1C), where it is able to bind to extracellular glycated collagens. Immunostaining with a second mouse anti-human Endo180 mAb, 39.10, detected Endo180 exclusively in endosomes ( Figure 1C), due to its epitope CTLD4 ( Figure 1B) being masked by its interaction with CD147 at the basal surface of PEC acini [11]. Interestingly, levels of endosomal Endo180 were visibly increased in PECs protruding from the basal surface of acini into stiff rBM ( Figure 1C, arrow). Moreover, targeted blockade of CTLD2 by A5/158 mAb decreased the transition of PEC acini towards a polygonal shape (0.6-fold), decreased the basal protrusion of PECs (0.7-fold) and increased acinar size (2.2-fold) in comparison to IgG control treatment ( Figure 3A-D). The anti-invasive effects of A5/158 mAb treatment were recapitulated in our alternative ribose-based model (see supplementary material, Figure S4E). We next considered whether the pro-invasive signals activated by stiff rBM in our 3D PEC acinar model involved the pro-migratory, actinomyosin-based, cell contractility signals generated by Endo180-containing endosomes in two dimensional (2D) culture models [12]. To explore this, pMLC2, total MLC2 and Endo180 levels were measured using immunoblot analysis of lysates generated from PEC acini cultured in native and stiff rBM in the presence or absence of control IgG or A5/158 mAb ( Figure 4A). Stiff rBM induced a 1.3-fold increase in the pMLC2:MLC2 ratio, which was reduced to 0.3-fold by A5/158 mAb treatment (a decrease of ∼77%) without changing Endo180 expression ( Figure 4B). The A5/158-sensitive (CTLD2-dependent) phosphorylation of MLC2 associated with increased rBM stiffness directly correlated with a decrease in E-cadherin and an increase in β 1 -integrin ( Figure 4B). The stiffness-induced changes in E-cadherin and β 1 -integrin levels were restored in PEC acini by A5/158 mAb treatment (E-cadherin from 0.6-fold to 1.0-fold; β 1 -integrin from 1.6-fold to 0.8-fold) ( Figure 4B). These findings suggest that CTLD2-dependent actinomyosin-based contractility signals generated in PECs, at stiff basal lamina interfaces, drive their increased invasion.

Endo180 modulates its extracellular matrix remodelling partners in AGE-modified basal lamina
Given its role in collagen remodelling, we considered whether cooperative protease partners of Endo180, co-expressed on invasive PECs in human prostate tumours [10], were under the regulatory control of Endo180 under stiff conditions. Accordingly, A5/158 mAb treatment reversed the up-regulation of membrane-type 1 matrix metalloproteinase (MT1-MMP; from 1.3-fold to 0.8-fold) and urokinase plasminogen activator (uPA; from 1.2-fold to 0.5-fold) in stiff rBM (see supplementary material, Figure S5A). These findings point towards a modulatory role for Endo180 in directing pericellular proteolysis during PEC invasion through a stiffened basal lamina. However, A5/158 mAb did not inhibit collagen uptake by PECs (see supplementary material, Figure S5B), indicating that Endo180-dependent clearance of collagen degradation products is not necessary for PEC invasiveness.

Prostate acini in Endo180 mutant mice have luminal obstruction and reduced BM matrix
Our in vitro findings support the hypothesis that CTLD2 binding to modified collagen IV [8] in a stiffened basal lamina can promote PEC invasiveness (Figure 2A). To establish whether CTLD2 in Endo180 contributes to remodelling of prostate epithelia in vivo, we investigated the consequence of a targeted genetic modification in the MRC2 gene that results in CTLD2 being constitutively exposed in a truncated form of the transcribed receptor ( Figure 5A) [17,18]. This truncation in Endo180 simulates the 'open' conformation of the receptor that is predicted to exist in the acidic environment of endosomes ( Figure 5A) [28][29][30]. Interestingly, respective reductions were observed in levels of the core basal lamina matrix components, collagen IV and laminin, in the ventral (0.7-fold and 0.6-fold) and anterior (0.8-fold and 0.7-fold) regions of the prostate glands of Endo180 ΔEx2-6/ΔEx2-6 mice compared to their wild-type counterparts ( Figure 5B-D). Moreover, the modified basal lamina in mutant mice, with exposed CTLD2, was accompanied by an increase in complete cellular obstruction of acinar lumen in the ventral (1.7-fold) and anterior (1.9-fold) prostate, when compared to wild-type animals ( Figure 5E-G), which is indicative of lesion development.
Endo180 combined with AGE decreases prostate cancer survival AGE accumulation in human tissue is a hallmark of several age-related diseases, including multiple cancer types (reviewed in [31]). The interaction of AGE with its receptor, RAGE, is linked to the promotion of prostate cancer cell growth and invasion [32,33]; and the expression of RAGE and its ligand high-mobility group box 1 (HMBG1) has been associated with poor overall survival in a cohort of 58 patients with clinical stage III and IV prostate cancer [34]. Our in vitro findings suggest that prostate cancer has the potential to progress when the basal lamina becomes stiffened by AGE exposure and  triggers Endo180-dependent mechanotransduction and invasion of PECs.
Immunostaining of two University of Newcastle tissue microarrays (NCLPC1 and NCLPC-4; see supplementary material, Table S2) revealed high levels of epithelial AGE in 55% (68/123) of prostate tumours ( Figure 6A). Kaplan-Meier analysis revealed that high epithelial AGE content per se was not prognostic ( Figure 6B), whereas high epithelial AGE accumulation combined with an Endo180-positive, but not an Endo180-negative, status ( Figure 6A) was predictive of a decrease in overall survival ( Figure 6C). In the subset of men with prostate tumours with high epithelial AGE content, 62% (31/50) deaths were recorded after 5 years in the Endo180-positive group, which was significantly higher (p <0.016) than the 28% (5/18) deaths recorded in the Endo180-negative subset ( Figure 6C; see also supplementary material, Table S3). Endo180 had less impact on the 5-year survival rate of men with AGE-negative prostate tumours, with 69% (29/42) deaths in the Endo180-positive subset and 54% (7/13) deaths in the Endo180-negative subset (p = 0.4) ( Figure 6D; see also supplementary material, Table  S3). This compelling evidence links the expression of Endo180 in epithelial cells, in an environment associated with increased non-enzymatic crosslinking and stiffening of the basal lamina, to the switch from indolent to aggressive prostate cancer and reduced survival.

Discussion
Here we show that the progressive thickening and stiffening of the basal lamina associated with normal ageing and metabolic disease [1][2][3][4][5][6] can induce normal PEC invasiveness. The study introduces a novel 3D model of PEC acini grown in rBM matrix that has been subject to non-enzymatic crosslinking by AGE, recapitulating basal lamina thickening in the ageing mouse prostate, and the increased levels of tissue stiffness observed in prostate cancer (see supplementary material, Figure  S1E, Table S1) [20][21][22][23][24][25]. The findings are supported by in vivo and clinical evidence that link Endo180 dysfunction to basal lamina degradation, accompanied by luminal obstruction of prostate acini and the cooperation of Endo180 with AGE in the promotion of prostate cancer progression.
In line with our previous studies, where Endo180 has been shown to play a role in EMT and prostate cancer progression [10,11], and the spatiotemporal activation of actinomyosin-based contractile signals and cell migration [12], we propose a model in which two functional CTLDs in Endo180, CTLD2 and CTLD4, are responsible for directly modulating PEC invasiveness (see supplementary material, Figure S6). We believe that the molecular basis of this regulation involves a switch between the closed and open conformation of Endo180's ectodomain [28][29][30]. We speculate that, under normal tissue conditions, Endo180 adopts a closed conformation and maintains acinar homeostasis via the interaction of its CTLD4 with CD147 at the plasma membrane and its FNII domain with collagen IV at the basal lamina interface; whereas in stiff microenvironments it is proposed that Endo180 adopts an open conformation that allows it to sense crosslinked collagen fibres, promoting actinomyosin-based contractility and driving PEC invasion through the basal lamina. Given that antibody blockade of CTLD2 in Endo180 can partially block the from wild-type mice and Endo180 ΔEx2-6/ΔEx2-6 mice; TOPRO-3 nuclear stain (blue), collagen IV (green) and laminin (red); scale bar = 50 μm. (C) Ratio of relative staining:background intensity (mean ± SD) for collagen IV (black bars) or laminin (grey bars) in ventral prostates of wild-type and Endo180 ΔEx2-6/ΔEx2-6 mice. (D). Analysis in (C) conducted on anterior prostates. (E) Epifluorescent images of ventral prostate sections from wild-type and Endo180 ΔEx2-6/ΔEx2-6 mice; TOPRO-3 nuclear stain (blue), collagen IV (green); extent of luminal obstruction (hollow, partial and complete luminal obstruction); scale bar = 50 μm. Graphs show extent of luminal obstruction (%, mean ± SD for 50 acini) in ventral (F) and anterior (G) prostate glands of five wild-type mice (black bars) and six Endo180 ΔEx2-6/ΔEx2-6 mice (grey bars).
invasive phenotype induced by the stiff basal lamina, it is plausible that this approach could be of therapeutic benefit in prostatic and other disease pathologies in which basal lamina thickness and stiffness is identified as a driving factor (see supplementary material, Figure S7).
Loss of collagen IV and laminin in the prostate gland of mutant Endo180 ΔEx2-6/ΔEx2-6 mice indicates that the homeostasis of basal lamina turnover is defective in these animals. The mechanisms underpinning this phenomenon could involve activation of proteases that degrade the basal lamina or defective production and organization of its matrix components. The first mechanism is supported by the finding that Endo180 forms an EMT-suppressor complex with the extracellular matrix  Table S3 (see supplementary material) shows patient deaths (censored) after 3, 5, 7 and 10 years. metalloproteinase inducer CD147 [11], and its targeted blockade results in down-regulation of MT1-MMP and uPA (see supplementary material, Figure S5), which both drive basal lamina matrix degradation [35][36][37][38]. The latter mechanism is supported from the finding in a different cellular context, whereby Endo180 orchestrates collagen deposition by primary human osteoblasts [16].
The in vitro systems used to model early tumourigenesis or tumour progression in response to tissue stiffness used constitutively active Ras transformation in a single Madine Derby canine kidney epithelial cell, or ErbB2 activation in a single MCF10A mammary gland epithelial cell, localized in corresponding non-transformed epithelial layers [39][40][41]. Although conceptually intriguing, these models do not explain

SUPPLEMENTARY MATERIAL ON THE INTERNET
The following supplementary material may be found in the online version of this article:  Table S1. Stiffness of malignant compared to normal human prostate tissue Table S2. Patient characteristics of the NCLPC1/4 prostate cancer tissue microarray Table S3. Patient deaths associated with levels of Endo180 and AGE