LOX-mediated collagen crosslinking and tissue stiffening promote focal adhesions and tumor progression in vivo

We next asked if LOX-mediated collagen crosslinking could stiffen the breast and promote invasion of a premalignant lesion. We used Ha-ras human MCF10AT MECs which upon injection into mice develop into premalignant tissues (Hu et al., 2008). We conditioned inguinal mammary fat pads of NOD/SCID mice, surgically-cleared of their epithelium, with control fibroblasts or those expressing elevated LOX (Fig 2A; Supp Fig 2 & 3). Rheological measurement revealed that the mammary glands conditioned with LOX expressing fibroblasts were stiffer (Fig 2B), picrosirius red staining showed they had more fibrillar collagen (Fig 2C & D), and SHG imaging revealed they had more linearized collagen (Fig 2C & E). Consistent with ECM stiffening, resident fibroblasts in the epithelial-cleared LOX-treated glands showed more FAK^pY397 and ^p130Cas immunostaining, indicative of increased focal adhesions and mechanosignaling in the stromal cells (Fig 2C; bottom).

Open in a separate window

Figure 2

Collagen remodeling and tissue stiffening promote focal adhesions and tumor invasion

A. Experimental design. B. Scatter plot of breast rheology following LOX (FB.LOX) or control (FB.ctrl) fibroblast conditioning. C. Top two panels: Photomicrographs of tissues described in B stained with Picrosirius Red and hematoxylin viewed under parallel and orthogonal polarizing filters. Bar 50μm. Third panel: SHG images of tissues from fibroblast-conditioned mammary glands. Bar 25μm. Fourth and fifth panels: Confocal images of fibroblasts residing within fibroblast-conditioned mammary glands stained for active FAK (FAK^pY397; red; fourth panel), ^p130Cas (red; fifth panel) and DAPI (nuclei; blue). Bar 50μm. D. Scatter plot of fibrillar collagen in tissue shown in C. E. Scatter plot of collagen linearity measured by Curvature Ratio from SHG images shown in C. F. Rheology of mammary glands with injected Ha-ras MCF10AT MECs. G. Lesion burden for mouse cohorts. H. Top: Photographs of mammary glands three weeks after injection with Ha-ras MCF10AT organoids. Middle: H&E stained tissue of MECs from lesions shown in H. Bottom: Confocal images of MECs in tissue shown in H stained for β1 integrin (red), active FAK (FAK^pY397; green) and DAPI (nuclei; blue; insert). Bar 50μm. Values in E and G are Mean ± SEM of several glands. * p≤0.05.

LOX pre-conditioning and stiffening of the mammary gland promoted the growth and invasion of Ha-ras premalignant mammary organoids injected into these tissues as revealed by a significant increase in lesion size and tumors that lacked margins (Figs 2F, G & H). MECs in the LOX pre-conditioned glands also had more focal adhesions, as revealed by elevated FAK^pY397 (Fig 2H) and ^p130Cas (not shown). To rule out any direct effect of LOX activity on MEC behavior, we also treated a cohort of animals in parallel (i.e. at time of MEC injection) with β–aminopropionitrile (BAPN), a natural and irreversible inhibitor of lysyl oxidase activity (Kagan and Li, 2003). Because BAPN treatment had no effect on tumor growth, invasion and focal adhesions (Supp Fig 4, 5, & 6) we concluded that it was LOX pre-conditioning of the stroma and gland stiffness that promoted mammary tumor progression, and not any direct effect of LOX on the mammary epithelium. These findings demonstrate how ECM crosslinking and stiffening can induce focal adhesions and promote the growth and invasion of an oncogene-initiated mammary epithelium in vivo.

Inhibiting LOX-mediated collagen crosslinking decreases fibrosis and reduces focal adhesions to inhibit breast tumor progression in vivo

We next asked if reducing LOX-dependent collagen crosslinking could temper tissue fibrosis and decrease focal adhesions, and whether this would inhibit breast tumor progression. We inhibited LOX activity using the cell soluble BAPN (Lucero and Kagan, 2006) or a LOX-specific function-blocking polyclonal which only inhibits extracellular LOX activity (Erler et al., 2006) and assayed collagen crosslinking, tissue fibrosis, focal adhesions and breast tumor development.

Inhibition of LOX activity was initiated in five month old parous animals, when LOX levels were already increased in the stromal cells but were non-detectable in the mammary epithelium (Fig 1H). Treatment was continued for one month after which the animals were allowed to recover for another month and then sacrificed (Fig 3A). At time of sacrifice, the non-treated MMTV-Neu mice had high levels of active LOX in the serum whereas BAPN or LOX inhibitory polyclonal-treated animals had reduced levels (Fig 3B). Mice with reduced LOX activity showed significant decreases in LOX-mediated collagen crosslinks, both reducible (DHLNL and HLNL) and nonreducible (pyridinoline) (Fig 3C & D), and the collagen fibrils adjacent to the epithelial lesions were less linear (Fig 3E & F). Consistent with the possibility that inhibiting collagen crosslinking prevented tissue fibrosis, LOX inhibition reduced fibrillar collagen (Fig 3G & H) and had reduced focal adhesions, as indicated by negligible FAK^pY397 (Fig 3I).

Open in a separate window

Figure 3

Inhibiting collagen crosslinking tempers tissue fibrosis and reduces focal adhesions

A. Experimental design. B. LOX enzymatic activity in serum of untreated (Control) compared to BAPN (+BAPN) or LOX inhibitory antibody-treated (+LOX-Ab) Neu mice. Values: Mean ± SEM. C. Scatter plots of reducible collagen crosslinks in LOX-inhibited (LOX-Inhib) and untreated (Control) Neu breasts. D. Scatter plots of pyridinoline in LOX-inhibited (LOX-Inhib) and untreated (Control) Neu breasts. E. SHG images of mammary glands from BAPN treated animals as described above. Bar 25μm. F. Collagen curvature in SHG images from D. (Supp Methods). Mean ± SEM 3-5 regions/4 glands/condition. G. Photomicrographs of sections from control and BAPN treated (LOX-Inhib) Neu mammary glands stained with Picrosirius Red and hematoxylin viewed under parallel or orthogonal polarizing filters. Bar 50μm. H. Fibrillar collagen in control and LOX inhibited glands from images shown in F. Mean ± SEM of 4-6 images/4-8 glands/condition. I. Confocal images of sections from control and BAPN treated Neu glands stained for active FAK (FAK^pY397; red) and DAPI (nuclei; blue). Bar 50μm.* p ≤0.05, ** p≤0.01, *** p≤0.001.

Inhibiting LOX activity also increased tumor latency (Fig 4A) and decreased tumor incidence (Fig 4B), despite ErbB2 activity (Fig 4C). Moreover, the palpable lesions formed in the LOX-inhibited animals were smaller (Fig 4D) and less proliferative (Fig 4E). LOX inhibited mammary glands also stained positively for cytokeratin 14 along the basal periphery of the ducts, implying retention of their myoepithelium (Fig 4H & I). H&E sections revealed that a significantly greater proportion of the lesions in the LOX-inhibited animals were hyperplastic or premalignant (hyperplastic alveolar nodules, HAN, or mammary intraepithelial neoplasia, MIN) and the tumors that did develop were mostly low grade as opposed to the high grade carcinomas typically observed in the untreated Neu mice (Fig 4G & F). These results show how inhibiting LOX activity reduces collagen crosslinking, tempers tissue fibrosis, decreases focal adhesions, and impedes tumor progression to reduce breast tumor incidence.

Open in a separate window

Figure 4

Inhibiting crosslinking reduces tissue fibrosis and tumor incidence and enhances tumor latency

A. Tumor latency. B. Tumor incidence without (Control) or with LOX inhibition (LOX-Inhib). C. Confocal images of activate ErbB2 (ErbB2^pY1248; red) in tissues from FVB non transgenic littermates, Neu (MMTV-Neu) and LOX inhibited mice (MMTV-Neu +LOX-Inhib). Bar 50μm. D. Tumor size in control and LOX inhibited mice. E. Confocal images of sections from LOX inhibited and control glands stained for PCNA (red) and DAPI (nuclei; blue). Bar 50μm. F. Grading of tumor lesions from mammary glands of control (n=14) and BAPN treated (LOX-Inhib, n=13) animals. H. H&E stained tissue showing typical Normal Glandular structure, Hyperplastic alveolar nodules (HAN) and mammary intraepithelial neoplasia (MIN) premalignant lesions and Grades I, II and III ductal tumors. I. Confocal images of breast tissue stained for cytokeratin 14 (red) and DAPI (nuclei; blue) in control and LOX inhibited tissue. Bar 50μm. I. Percent cytokeratin 14 positive glands detected in breast from control and LOX inhibited animals. Values in A, B, D, F & I; Mean ± SEM of 4-6 measurements/4-12 tissue sections/group. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001.

Collagen crosslinking and ECM stiffening promote focal adhesions and drive invasion of oncogenically-initiated mammary tissues in culture

We next assessed the effect of non-specifically inducing collagen crosslinking and stiffening on MEC invasion in the absence of fibroblasts, immune cells and other stromal and systemic cellular and soluble factors (Fig 5A experimental schemata). We explored the effect of nonspecific collagen crosslinking and stiffening on tumor invasion by adding nonmetabolizable L-ribose to the media of nonmalignant, MCF10A acini embedded in a 3D soft collagen I/reconstituted basement membrane gel (Col/rBM gel).

Open in a separate window

Figure 5

Collagen crosslinking and ECM stiffening promote focal adhesions and invasion of ErbB2 mammary colonies

A. Experimental design. B. Elastic modulus of ribose crosslinked (+Ribose) and untreated (Control) collagen gels. C. Cross sectional area of MEC colonies in +Ribose and Control gels. D. Confocal images of MCF10A colonies stained for β1 integrin (green), activate FAK (FAK^pY397; red) and DAPI (nuclei; blue) in Control and +Ribose gels. Bar 20μm. E. Top panels: Confocal images of MEC colonies expressing the ErbB2 chimera stained for β catenin (green), β4 integrin (red) and DAPI (nuclei; blue) in Control or +Ribose gels with (+ErbB2) or without ErbB2 activity. Bottom panels: SHG images of collagen fibrils and confocal images of eGFP expressing MEC colonies as described above. Bar 20μm. Yellow arrows identify collagen bundles surrounding the colony periphery and white arrows indicate an invading MEC. F. Percent colony invasion shown in E. G. Phase contrast images of MEC colonies expressing the ErbB2 chimera in Control or +Ribose gels, with active ErbB2 (+ErbB2); co expressing a doxycycline inducible FRNK (right panel) or treated with β1 integrin function blocking antibody (+β1 block). H. Percent colony invasion shown in G. Values in B, C, F & H; Mean ± SEM from 3-5 experiments and/or 50-100 colonies in 3 experiments. ** p ≤ 0.01, *** p ≤ 0.001.

To determine whether MEC invasion required oncogenic signaling we utilized MCF10As expressing an ErbB2 chimera (ErbB2.chim; called HER-2; the human homologue of Neu) consisting of the extracellular and transmembrane domains of low-affinity nerve growth factor receptor (p75NGFR), and the cytoplasmic kinase domain of ErbB2 linked to the synthetic ligand-binding domain from the FK506-binding protein (FKBP). In these MECs addition of the synthetic bivalent FKBP ligand, AP1510, drives homodimerization of ErbB2, activation of the kinase domain and initiation of ErbB2 signaling (Muthuswamy et al., 2001). Chimeric ErbB2 signaling drives proliferation and luminal filling but fails to induce MEC invasion in rBM (Fig 5E). To study effects of homo- (ErbB2/ErbB2) and heterodimer (ErbB1/ErbB2) formation we also engineered MECs in which the wild type ErbB2 could be overexpressed through addition of doxycycline (ErbB2.TetOn; Supp Fig 7 & 8).

In the absence of ErbB2 activity MECs in collagen/rBM gels assembled acini, as illustrated by basally-oriented β4 integrin and cell-cell localized beta catenin (Fig 5E). Glycation-mediated crosslinking stiffened the collagen (Fig 5B), increased colony size (Fig 5C), and disrupted tissue organization, as revealed by diffusely localized beta catenin and the appearance of random cells within the lumens (Fig 5E). MECs within the ribose crosslinked collagen gels also had elevated levels and co-localized β1 integrin and FAK^pY397, indicative of tension-induced focal adhesions (Fig 5D). Nevertheless, in the absence of ErbB2 activity, ECM stiffening did not drive MEC invasion (Fig 5E & F).

Addition of either AP1510 (1μM) or doxycycline (0.2μg/mL) to the mammary acini, induced and activated (ErbB2.TetOn) or directly activated ErbB2 (ErbB2.chim) (Supp Fig 9), promoted cell growth (not shown), drove luminal filling, and destabilized cell-cell junctions, as revealed by diffusely localized beta catenin (Fig 5E). Sustained ErbB2 activation did not drive MEC invasion, as indicated by colony integrity and the retention of basally-localized β4 integrin (Muthuswamy et al., 2001); Fig 5E & F). Yet, when ErbB2 was activated in colonies in the crosslinked, stiffened gels, colony architecture disintegrated, as revealed by the absence of detectable beta catenin and disorganized β4 integrin staining, and MECs invaded into the gels (Fig 5E & F). SHG imaging revealed that ErbB2 activation and ECM stiffening was accompanied by the appearance of prominent collagen bundles surrounding the colonies, and showed that single MECs invaded on fibrils that extended perpendicularly into the crosslinked, stiffened gels following ErbB2 activation (Fig 5E; Supp Fig 9). These findings show how collagen crosslinking and ECM stiffening, per se, cooperate with oncogenes such as ErbB2 to promote the invasive behavior of a mammary epithelium.

Interestingly, confocal imaging revealed elevated ^p130Cas and FAK^pY397 staining that co-localized with β1 integrin (colocalization analysis of FAK^pY397 and β1 integrin; Pearson's r = 0.78; ^p130Cas and β1 integrin, Pearson's r = 0.89) in the stiffer breast tissue of the Neu mice (Fig 6A; see insert) and reduced levels of FAK^pY397 following LOX inhibition (Fig 3H). Moreover, inhibiting β1 integrin activity, using the function blocking antibody AIIB2, or reducing integrin signaling by expressing an inducible FRNK (FRNK.TetOn; Supp Fig 10) prevented the force-mediated invasion of the ErbB2 activated mammary organoids (Fig 5G & H). These findings show how ECM crosslinking and stiffness cooperate with an oncogene to promote breast cell invasion and implicate integrin signaling in force-dependent tumor progression.

Open in a separate window

Figure 6

β1 integrin clustering promotes focal adhesions and drives invasion of a premalignant mammary epithelium in culture and in vivo

A. Combined and split (inserts) confocal images of tissue sections from MMTV-Neu breast stained for DAPI (nuclei; blue) and p130Cas (red; top), or β1 integrin (green; bottom) and, activate FAK (FAK^pY397; red). Bar 20μm. (colocalization analysis of the malignant tumors: FAK^pY397 and β1 integrin, Pearson's r = 0.78; ^p130Cas and β1 integrin, Pearson's r = 0.89) B. β1 integrin constructs used for the studies shown in C-G. C. Confocal images of MCF10A MEC rBM colonies expressing the β1 integrin wild type (β1(WT)) or clustering mutant (β1(V737N)) stained for β4 integrin (red; top), active FAK (FAK^pY397; red; bottom) and DAPI (nuclei; blue). D. Phase contrast images of Ha-ras MCF10AT MEC colonies expressing the β1 integrin wild type, glycan wedge or integrin cluster mutant in rBM. Bar 50μm. E. Percent invasion of the colonies shown in D F. Lesion size formed by Ha-ras MCF10AT MECs expressing the clustering β1 integrin mutation (β1(V737N)) and wild type integrin (β1(WT)). G. Top panels: Photomicrographs of H&E stained sections of tumors formed by Ha-ras MECs expressing the clustering β1 integrin mutation (β1(V737N)) and wild type integrin (β1(WT)). Bottom panels: Confocal images of tissues stained for active FAK (FAK^pY397; red) and DAPI (nuclei; blue). Bar 50μm. Values in E & F; Mean ± SEM; 12-50 measurements/3 experiments *** p≤ 0.001.

Integrin clustering promotes focal adhesions to drive invasion of a Ha-ras mammary epithelium in vitro and in vivo

To determine if integrin signaling would be sufficient to induce breast tumor invasion we expressed a series of integrin constructs in nonmalignant MCF10A and Ha-ras premalignant MCF10AT MECs and assayed for invasive behavior in rBM and in vivo (Fig 6B).

Expression of the V737N integrin, which recapitulates tension-dependent integrin clustering, promoted focal adhesions, indicated by elevated FAK^pY397 (Fig 6C) and disrupted the integrity of mammary colonies in rBM, revealed by luminal filling and altered β4 integrin localization (Fig 6C; top). In contrast, neither expression of a wild type (Fig 6C) nor a constitutively active β1 integrin (not shown; (Paszek et al., 2005)) induced focal adhesions and disrupted mammary tissue architecture. Moreover, while the V737N β1 integrin mutant failed to promote invasion of the MCF10A MECs (Fig 6C), the V737N integrin induced the invasion of the Ha-ras premalignant MCF10AT colonies in rBM (Fig 6D & E). Upon injection into nude mice the V737N integrin not only promoted focal adhesions, and increased lesion size (Fig 6F), but also induced the invasive behavior of the Ha-ras MCF10AT organoids, as revealed by the loss of epithelial lesion margins and increased FAK^pY397(see arrows in Fig 6G). These findings implicate tension-dependent integrin clustering and focal adhesions in breast tumor invasion. However, because integrin clustering could only promote invasion in Ha-ras premalignant MECs, they also illustrate how integrin-mediated mechanotransduction cooperates with oncogenic signaling to drive malignancy.

Cell manipulations

MEC and fibroblast cell lines were cultured as described (Paszek et al., 2005). Collagen crosslinking was induced by addition of 15mM L-ribose to the culture medium (Girton et al., 1999). rBM-conjugated PA gels with calibrated stiffness were prepared as described (Johnson et al., 2007).

Vector Constructs and gene expression

Full length human ErbB2 (K. Ignatoski) was cloned into the pRet puro Tet IRES EGFP tetracycline-inducible vector. The β1 integrin wild type, β1 integrin glycan wedge constitutively active and β1 integrin clustering mutant V737N constructs and the preparation of virus and cell infection and selection have been described (Paszek et al., 2005). Four myc-tags were added to the C-terminus of full length LOX and cloned into the pLV puro TetO₇mCMV tetracycline-inducible lentiviral vector and expressed biscistronically with EGFP.

Elastic modulus measurements

Mammary glands and gels were assayed for materials properties by unconfined compression and rheometrical analysis (Paszek et al., 2005).

Mice and Treatments

FVB-TgN MMTV-Neu, NOD/SCID, and BalbC nu/nu mice (Jackson Laboratory) were maintained in accordance with University of Pennsylvania and University of California IACUC guidelines. For LOX inhibition studies animals were treated with BAPN (3mg/kg; Spectrum) in the drinking water (4-8 mice/group; 4 studies) or a LOX function-blocking polyclonal (3mg/kg; OpenBiosystems, D8746) injected intraperitoneally twice per week (3-4 mice/; 1 study). Mice were sacrificed at 7-71/2 months of age at which time tail vein blood was collected. Lesions were detected by palpation (~3mm diameter) and tumor volume was assessed using calipers. At sacrifice mammary glands were excised, imaged and mechanically tested or snap frozen or paraformaldehyde fixed.

Tumor grading

Tumor grading was performed blinded on H&E stained sections from untreated (n=14) or BAPN treated (n=13) #2-3 mammary glands. Premalignant lesions were defined as hyperplastic alveolar hyperplasia (HAN) or Mammary Intraepithelial Neoplasia (MIN)-like foci. Grade I lesion were well defined as homogenous carcinomas, Grade II contained areas with strong necrosis, stromal reaction and/or red blood cells outside of tumor blood vessels and Grade III also contained necrosis and nuclear pleiomorphy. Adjacent tissue was defined as mammary tissue surrounding a tumor that was cut away from the tumor prior to subsequent analysis. A small set of anti-LOX treated mice (n=4) was compared with their controls (n=3), and showed the same tendency to less progressed lesions as was observed in the BAPN treated animals.

Xenograft manipulations

NOD/SCID mice (n=24) were used for mammary fat pad transplantation studies (Kuperwasser et al., 2004). Briefly, the rudimentary inguinal epithelium was removed from three week old anesthetized female mice and 5 × 10⁵ NIH 3T3 WT or LOX-expressing fibroblasts (2.5 × 10⁵ were treated with 4Gy irradiation 24 hours prior to injection) were injected into the left and right mammary glands. Two weeks following fibroblast injection, 1×10⁶ DCIS.com MCF10AT MEC 4 day old rBM-generated, proliferating organoids (16-20 cells/organoid) were suspended in Dulbeccos modified PBS and injected into the pre-conditioned fat pads. All mice were sacrificed at 8 weeks of age (3 weeks following MEC injection and 5 weeks after fibroblast injection), lesion volume was assessed by caliper measurement, photographed In Situ, and dissected gland fragments were imaged and mechanically-tested or snap frozen, or fixed and paraffin embedded.

Subcutaneous studies

Proliferating rBM-generated organoids (16-20 cells/organoid) from each experimental condition were injected subcutaneously into the rear flanks of BalbC Nu/Nu mice (6-8 mice/group; 5-10×10⁶) and tumor formation was monitored for three weeks and samples were assessed as above at study termination.

Collagen crosslinking

Mammary gland collagen was reduced with standardized NaB³H₄, hydrolyzed with 6N HCl and subjected to amino acid and collagen crosslink analyses as described (Yamauchi and Shiiba, 2008). The reducible crosslinks, dehydro-dihydroxylysinonorleucine (deH-DHLNL) and deH-hydroxylysinonorleucine (deH-HLNL), their ketoamines, were identified as their reduced forms, i.e. DHLNL and HLNL. The crosslinks analyzed (reducible and nonreducible) were quantified as moles/mole of collagen based on the hydroxyproline value of 300 residues per collagen molecule. LOX activity was measured in cell supernatant and plasma as described (Erler et al., 2006).

Immunostaining and Imaging

Immunofluorescence and imaging of 3D cultures and tissues was as described (Paszek et al., 2005). Picrosirius red analysis was achieved using paraffin sections of mammary glands stained with 0.1% Picrosirius Red (Direct Red 80; Sigma) and counterstained with Weigert's Hematoxylin to reveal fibrillar collagen. Sections were serially imaged using an Olympus IX81 fluorescence microscope fitted with an analyzer (U-ANT) and polarizer (U-POT; Olympus) oriented parallel and orthogonal to each other and quantified using minimal thresholding. Two-photon second harmonics imaging was performed on a Prairie Technology Ultima System attached to an Olympus BX-51 with a water immersion objective and samples were quantified by calculating the linearity of multiple collagen fibrils (see Appendix for further details).

Immunoblot analysis

Cells were lysed in RIPA or Laemmli buffer and assayed by immunoblotting (Johnson et al., 2007).

We thank K. Ignatoski for the ErbB2 construct, MP Marinkovich for the BM165 antibody, P. Mrass for 2 photon microscopy guidance and R.G. Wells for mentoring of KRL. The work was supported by NIH grants R01-CA078731 to VMW, R01-CA057621 to Z. Werb, and T32HL00795404 to KRL, and grants DOD W81XWH-05-1-330, DOE A107165 and CIRM RS1-00449 to VMW.