Novel mechanism for OSM-promoted extracellular matrix remodeling in breast cancer: LOXL2 upregulation and subsequent ECM alignment

Human breast cancer cell lines used in experiments were purchased from the American Type Culture Collection (ATCC; Manassas, VA). Human luminal A MCF7 and T47D [ER+, PR+, HER2−] and triple negative basal B MDA-MB-231 [ER−, PR−, HER2− were cultured in RPMI 1640 (Genesee Scientific; San Diego, CA), while BT474 [ER+, PR+, HER2+] and triple negative basal A MDA-MB-468 [ER−, PR−, HER2−] cells were cultured in DMEM (Genesee Scientific). Sk-Br-3 [ER−, PR−, HER2+] breast cancer cells were cultured using McCoy’s 5A media (ATCC). All cell media contained 10% v/v Fetal Clone III (Thermo Fisher; Waltham, MA) and 1% v/v penicillin/streptomycin (Genesee Scientific). Cells were cultivated in tissue culture treated T-75 flasks (Genesee Scientific) kept in a Model 3110 (Forma Scientific; Marietta, OH) incubator at 37 °C and 5% CO₂. Cells grown to ~ 75% confluence before plating for experiments. Cells were treated with recombinant human OSM, IL-6, leukemia inhibitory factor (LIF) (25 ng/mL), and/or interleukin 1β (IL-1β; 10 ng/mL) from Peprotech Inc. (Rocky Hill, NJ) at various time intervals depending on the experiment and highlighted in the figures.

The Cancer Genome Atlas (TCGA) RSEM counts associated with breast invasive carcinoma (BRCA), glioblastoma multiforme (GBM), prostate adenocarcinoma (PRAD), and ovarian cancer (OV) were downloaded from the Broad GDAC Firehose repository (https://​gdac.​broadinstitute.​org/​). Using python, RSEM count data was standardized to Z-score for comparison and outlier patients above and below 3 standard deviations were removed from the dataset. Genes were then plotted, and correlation was assessed by Pearson coefficient using the SciPy package [57]. The line of best fit was determined by linear regression using the Polyfit function in the SciPy package. Specific code used for the analysis is available upon request at GitHub.

The data associated with van de Vijver et al. [58] was downloaded and coded to ensure key results and figures from the data could be generated. Observed events were coded to be positive outcomes for metastasis, or identified as a death from cancer without metastasis, and a censored event to be any other outcome (by van de Vijver’s definitions). The survival function was censored at 10 years to reduce the influence of the few cases with far longer survival times. Survival plots were created with the Survminer (Kassambara et al., 2019) and Survival (Therneau, 2015) libraries in R. OSM and LOXL2 cut points were based on the Maximally Selected Rank Statistic [59], which algorithmically searches the data for optimal cut points.

RNA extraction from treated cell cultures was performed using RNA STAT-60 (Tel-Test, Inc.; Friendswood, TX) following the standardized protocol on Tel-Test’s website. Isolated RNA concentration and quality was analyzed using a Nano-Dropper2000 (Thermo Scientific; Waltham, MA) and the agarose bleach gel protocol [60], respectively. Synthesis of cDNA was prepared using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems; Foster City, CA) and 1 μg of sample mRNA. Combining SYBR Green MasterMix (Bio-Rad; Hercules, CA) with sample cDNA, each sample was run in duplicate, at a minimum, on a 96-well plate. Roche Light Cycler 98 and accompanying software was used to determine mRNA expression.

Cells were lysed using RIPA, 1% v/v Protease Inhibitor Cocktail (Sigma-Aldrich; St. Louis, MO), and 100× Halt™ Phosphatase Inhibitor (Thermo Fisher). Ten micrograms of total protein was loaded per well, as determined by Peirce™ BCA Protein Assay Kit (Thermo Fisher). The Chameleon® Duo Protein Ladder (LiCor Biosciences; Lincoln, NE) was used as a protein molecular weight marker. Proteins were separated using Tris-Glycine SDS-PAGE gels and transferred onto nitrocellulose membranes (Thermo Scientific). Subsequently, the membrane was thoroughly dried and rewetted with ddH2O. Five milliliters of REVERT™ Total Protein (LiCor Biosciences) stain was added before the REVERT wash solution. The rinsing and the blot was imaged using Odyssey CLx (LiCor Biosciences). The nitrocellulose membrane was blocked using Odyssey PBS Blocking Buffer (LiCor Biosciences) and incubated with the following primary antibodies in addition to 0.2% Tween20: LOX (1:500, Santa Cruz Biotech; Dallas, TX), LOXL1 (1:200, Santa Cruz Biotech), LOXL2 (1:1,000, Genetex; Irvine, CA), LOXL3 (1:200, Santa Cruz Biotech), LOXL4 (1:200, Santa Cruz Biotech), E-cadherin (1:500, Abcam; Cambridge, UK), Snail-1 (1:1,000, Cell Signaling Technology; Danvers, MA), and GAPDH (1:500, Santa Cruz Biotech). Membranes were further incubated using 800 channel fluorophore conjugated donkey secondary antibodies (1:15,000, LiCor Biosciences), or HRP conjugated secondary antibodies (1:10,000, Jackson ImmunoResearch Laboratories; West Grove, PA) prior to addition of ECL substrate (Thermo Fisher). The target proteins were then visualized using the Odyssey CLx and quantified using LiCor Image Studio software. Proteins were then normalized either against a REVERT total protein stain or GAPDH expression and compared against non-treated controls.

MCF7 and MDA-MB-468 cells are treated with OSM for 24 h before samples were lysed with RIPA. One microgram of Rapid PNGase F (Cell Signaling) enzyme was added to 10 μg of total protein from OSM-treated cell lysates. The assay was performed following the accompanying Rapid PNGase F protocol. LOXL2 proteins were visualized using the immunoblot techniques described above.

Using Qiagen’s protocol, the best LOXL2 knockdown in MCF7 cells came from the combination of 5 nM siLOXL2 #2 and 5 nM siLOXL2 #3 (Qiagen; Hilden, GER), called siLOXL2 (2/3), with 3 μL Transfection reagent (Qiagen) for 48 h. To make the control siControl, 5 nM of scrambled siRNA (Qaigen) with 3 μL transfection reagent was used. The optimal cell density for transfection was 125,000 MCF7 cells in a 6-well plate. The MCF7-shLOXL2 and MCF7-shCTRL cells used in this experiment have previously been published and characterized [43].

MCF7 cells were transfected with siCTRL or siLOXL2 (2/3) for 48 h then treated with OSM in serum and phenol red-free RPMI 1640 for 24 h. The conditioned media (CM) was collected and immediately centrifuged at 8000g for 10 min to remove cellular debris. Lysates were collected to confirm LOXL2 knockdown. Conditioned media (1.75 mL) from each sample was added to separate 3 K filter tubes (MilliporeSigma; Burlington, MA), which were centrifuged at 4000g in a swinging bucket centrifuge for 30 min. The rest of the assay was formulated using the recipe previously published with volumes adjusted to fit within a 96-well fluorescence compatible plate (Thermo Fisher) [61]. The plates were read every 30 s using a BioTek Mx plate reader with Ex/Em 490/540 and 10 nm bandwidth.

The LOXL2 ELISA (R&D Systems; Minneapolis, MN) was performed and analyzed according to the protocol provided by the manufacturer. Protein was concentrated from CM using acetone precipitation [62]. For each condition, one-part CM was mixed with four parts of 100% acetone chilled to − 80 °C, and placed back into a − 80 °C freezer overnight. The CM was centrifuged using 13,000g for 10 min at − 10 °C. The acetone was decanted and the protein precipitates were dried for 20 min before being reconstituted with 1× dilution reagent (R&D Systems) containing 0.2% Tween20 at 1/3 the original volume. The samples were sonicated prior to addition to ELISA plate for a 3-fold increase in concentration.

Immunofluorescence staining was carried out as previously described [43]. Briefly, cells were cultured in 8-well chamber glass slides, fixed for 5 min with 4% PFA containing 5% sucrose and 0.1% Triton X-100, and re-fixed for an additional 25 min with 4% PFA containing 5% sucrose. The cells were washed 10 min with PBS and an additional 10 min with PBS containing 0.05% Tween 20. Fixed cells were blocked with IF buffer (130 mM NaCl, 7 mM Na₂HPO₄, 3.5 mM NaH₂PO₄, 7.7 mM NaN₃, 0.1% BSA, 0.2% Triton X-100, 0.05% Tween 20) supplemented with 10% donkey serum for 1 h and incubated overnight at 4 °C with mouse monoclonal [HECD-1] to E-cadherin (1:500, Abcam). The cells were washed three times with PBS for 15 min each and incubated for 1 h with donkey anti-mouse conjugated to Alexa Fluor®647 (1:200, Molecular Probes; Eugene, OR), washed as above, and mounted with VECTASHIELD mounting medium with 4′, 6-diamidino-2-phenylindole (DAPI). For F-actin staining, cells were incubated overnight with Alexa Fluor®488 Phalloidin (1:40) (Molecular Probes), washed three times with PBS for 15 min each, and mounted with VECTASHIELD mounting medium with DAPI. Immunofluorescent images were captured by a Nikon A1R confocal microscope.

Nuclear/cytoplasmic fractionation assay was carried out as previously described [63]. Cells were washed twice with PBS, scraped, and collected on ice into 1.5-mL microcentrifuge tubes. Tubes were spun with table-top centrifuge, and supernatant was discarded. Fractionation was performed with 0.1% NP40 in PBS. Cell pellets were triturated 5× with ice-cold 0.1% NP40 in PBS (900 μl for 10 cm dish) using P1000 micropipette that was cut at its end. Aliquots of 300 μL of these samples were placed into fresh tubes (designated as “Total”). The remaining samples were centrifuged for 1 min 16,200g to pellet nuclei. Aliquots of 300 μL of the supernatant were collected into fresh tubes (designated as “Cyto”). In total, 100 μl of 4× Laemmli sample buffer was added immediately to Total and Cyto samples. Nuclei pellets were resuspended with ice-cold 0.1% NP40 in PBS (1 mL for 10 cm dish), re-pelleted, and resuspended with 180 μL of 1× Laemmli sample buffer (designated as “Nuc”). Total and Cyto samples were sonicated using microprobes at level 2, twice for 5 s.

Rat-tail collagen I (Corning; Corning, NY) was used to form a 1.5 mg/mL collagen I matrix in a 35-mm petri dish with a 14-mm imbedded coverslip. On ice, rat-tail collagen I and 10× RPMI 1640 media (Corning) was diluted 1:10 with 1× PBS and adjusted with 0.1 M NaOH to bring the final pH to 7.4. The MCF7 and MDA-MB-468 cells were seeded homogeneously in the matrix before adding 400 μL containing 100,000 cells to each petri dish-imbedded coverslip. The matrix solution was incubated 20 min at 37C and 5% CO₂. Phenol red-free RPMI 1640 media (Thermo Fisher – Gibco) was added to each sample along with OSM and either 500 μM pan-LOX inhibitor β-aminopropionitrile βAPN (Thermo Fisher) or 200 nM LOXL2/3-specific small molecule inhibitor PXS-5120A (Pharmaxis; New South Wales, AUS) for 48 h. Images were processed, and the area of the matrix was analyzed using ImageJ area measurement tools [64, 65].

Live cell imaging was performed using the Leica SP8 white light confocal microscope system with attached Peltier, which maintains cells at 37 °C and 5% CO₂. In total, 100,000 MCF7 cells were seeded into the same rat-tail collagen matrix described above for the Collagen I Contraction Assay. The samples were exposed to recombinant human OSM and 500 μM βAPN for 36 h prior to imaging. Collagen I fibers were visualized using reflectance mode confocal imaging [66, 67]. Cells were stained with membrane intercalating fluorescent dye Cell Tracker™ Red (Life Technologies; Carlsbad, CA) for 1-h pre-imaging at a 1:1,000 dilution in serum/phenol-free RPMI 1640. Each image consisted of 15 μm z-stacks split into 44 sections that were taken with four frames stitched together in a 2 × 2 format at × 63 magnification using a water immersion objective.

Live cell images of MCF7 cells in collagen I “pucks” were analyzed using CurveAlign4.0 software developed at the University of Wisconsin [68, 69]. Selected regions of interest (ROI) were analyzed—the areas between the seeded cells and radiating outward perpendicular to the MCF7 cells as illustrated in Supplemental Figure 4. ROIs were utilized because accurate whole image analysis was not possible due to the varying directions of collagen I fiber alignment. We used sum[(fiber dispersion coefficient) × (# of features (fibers) for each ROI)] / (total sampled features in image) to determine the average level of fiber dispersion for collagen I in each treatment group. For the fiber dispersion coefficient, 0 equals completely random fibers, 1 means all fibers are in alignment.

Statistical analysis was performed using Prism 8.0 software. All significant results were determined by various statistical methods including Student’s t test, One-way ANOVA, two-way ANOVA, and log rank test. Significance is denoted as n.s. (not significant), p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

To determine whether high co-expression of LOXL2 and OSM mRNA is associated with increased rates of IDC metastasis in patients, we created a distant metastasis-free survival plot using microarray data from the van de Vijver et al. patient study consisting of 295 IDC patients [58]. This database was utilized because the patient population selected for this study and the metadata for metastasis is well characterized. We compared low OSM/ low LOXL2 to low OSM/ high LOXL2, high OSM/ low LOXL2, and high OSM/ high LOXL2 mRNA expression in patients and found that higher levels of OSM and LOXL2 mRNA combined, led to significantly more metastatic events in a 10-year span (Fig. 1a). High expression of each individual gene also led to faster onset of metastasis, but high OSM and high LOXL2 co-expression had a greater significant impact on distant metastasis-free survival.

Our lab, and others, have previously analyzed Oncomine™ and other breast cancer patient databases demonstrating a correlation between reduced recurrence-free survival (RFS) of breast cancer patients with higher expression of LOXL2 [38, 41, 43] and reduced survival rates with higher expression of OSM/OSMR [70, 71]. Taken together, these results confirm that high OSM and LOXL2 co-expression is associated with an overall worse prognosis in IDC patients than high expression of either gene alone.

To assess the correlation between OSM signaling and LOXL2 expression in cancer patients, we analyzed the expression of LOXL2 mRNA in patients and compared it to OSMR mRNA expression. OSM is most often produced by neutrophils and macrophages found in the TME, and in turn, IDC cells, while having the capacity to secrete OSM, normally secrete none to very low levels [4, 5]. Therefore, comparing OSM mRNA expression from tumor samples directly against LOXL2 would not yield a highly relevant correlation. To assess OSM signaling, which occurs by OSM binding to the OSMR complex on tumor cells, we investigated the expression of the beta subunit of OSMR. Furthermore, increased levels of OSM in the TME promote the overexpression of OSMRβ mRNA in IDC cells [72‐74]. To assess the correlation between OSMR and LOXL2, we used RNA-Seq data from The Cancer Genome Atlas (TCGA) to assay transcriptional expression of biopsied patient samples in several cancer subtypes. Specifically, LOXL2 was compared against OSMR expression in glioblastoma, breast cancer, prostate cancer, and ovarian cancer. We observed a moderate to weak, yet significant, positive Pearson correlation of 0.263 (p = 6.97 × 10^-20) between OSMR expression and LOXL2 expression in breast cancer patients (Fig. 1b). There was also a weak to moderate positive, significant correlation between OSMR and LOXL2 expression in the other cancers investigated. To determine whether the correlation has a potential impact on gene expression we used a least squares linear regression to attain the line of best fit. Then we analyzed the slope and 95% confidence interval (CI) as illustrated in the table accompanying Fig. 1b, where a larger slope suggests a greater impact on gene expression. Each cancer analyzed had a positive slope above 0.2 and had a correlation between OSMR and LOXL2 mRNA expression. This data suggests that increasing OSMR mRNA is correlated with increasing LOXL2 transcripts in multiple forms of cancer; including breast cancer.

Next, we analyzed the breast cancer patient data for correlation between OSMR mRNA expression and other lysyl oxidase family members: LOX and LOXL1-4. We performed the same correlation analysis as above, but instead focused on OSMR and lysyl oxidase mRNA expression in breast cancer patients. We observed a significant, moderate to weak positive correlation (0.469 and 0.263) comparing OSMR expression to LOX and LOXL2 expression respectively, as determined by Pearson correlation analysis. While LOXL1/3/4 had significant Pearson correlation coefficients, the Pearson coefficients were considered very weak/negligible because they fall below 0.20 [75]. LOX and LOXL2 also generated a line of best fit with a positive slope above 0.2, when compared to OSMR expression, while the slopes for LOXL1/3/4 were approximately zero (Fig. 1c). These results suggest that increasing OSMR gene expression is correlated with increasing LOX and LOXL2 gene expression but not with LOXL1/3/4 expression. Further analysis of OSM mRNA expression in relation to lysyl oxidase mRNA expression again highlighted a significant correlation between OSM and all lysyl oxidases, except for LOXL4 (Supp. Figure 1). However, the Pearson correlation coefficient values for all the lysyl oxidases were below 0.25. The slopes for the lines of best fit for LOX and LOXL1-3 were also positive, but had large 95% CI. This data suggests that OSM gene expression is slightly correlated to increased LOX and LOXL1-3 expression. Taken together, these results confirm the positive correlation between OSM signaling and lysyl oxidase mRNA expression in breast cancer patients.

As the human breast cancer patient data demonstrated a correlation between the proinflammatory cytokine OSM and the collagen crosslinking enzyme LOXL2, we set out to determine whether OSM could promote the expression of LOXL2 at the transcriptional level. qRT-PCR was performed on three IDC cell lines with varying estrogen receptor (ER), progesterone receptor (PR), and ErbB2 (HER2) status: luminal A MCF7 (ER+ PR+ HER2−), basal A MDA-MB-468 (ER− PR− HER2−), and basal B MDA-MB-231 (ER− PR− HER2−). These cell lines were chosen because they represent increasing tumor cell aggressiveness and invasiveness, respectively [76], and they each express receptors for OSM/IL-6 cytokines [22, 77]. The cells were treated with recombinant human OSM (25 ng/mL), IL-6 (25 ng/mL), or both for 1, 2, 4, 12, 24, and 48 h and compared against untreated controls. Our cells were treated with cytokine concentrations designed to saturate the IDC cells, to mimic an actual TME, where cytokines have been shown to be present in high concentrations due to secretion by tumor-associated neutrophils, macrophages, and fibroblasts [70, 78‐80]. qRT-PCR analysis of MCF7 cells showed that OSM treatment induced a ~ 3-fold increase in LOXL2 mRNA, relative to the non-treated controls, at 12 and 24 h, whereas IL-6 treatment produced no significant change in LOXL2 mRNA (Fig. 1d). In MDA-MB-468 cells, OSM induced LOXL2 mRNA by ~ 2-fold at 24 h (Fig. 1e). In MDA-MB-231 cells, there was no significant change in expression of LOXL2 mRNA with any treatment groups (Fig. 1f). No effect on LOXL2 expression was somewhat expected since ER− MDA-MB-231 cells are already highly invasive, which can limit the impact of OSM signaling on promoting invasive potential [9]. MDA-MB-231 cells produce high levels of LOXL2, as highlighted in the following paragraph, limiting further induction by OSM. These results demonstrate that OSM signaling leads to an increase in LOXL2 mRNA expression in IDC cells.

To determine whether OSM-induced LOXL2 mRNA translates to the protein level, we performed immunoblot blot assays. Analysis was performed on MCF7, MDA-MB-231, MDA-MB-468, BT474 (ER+ PR+ HER2+), and Sk-Br-3 (ER− PR− HER2+) IDC cell lines treated with OSM, IL-6, LIF (all at 25 ng/mL), and IL-1β (10 ng/mL) for 24 h before cell lysates were collected and compared against untreated controls. Analysis of proinflammatory cytokines in the IL-6 family, as well as IL-1β, was performed to determine whether LOXL2 induction is unique to OSM signaling or has the potential to be broadly applicable to proinflammatory cytokines. In MCF7 cells, OSM induced a greater than 3.5-fold increase in LOXL2 protein expression, a ~ 2.5-fold increase with IL-1β treatment, but no change with the rest of IL-6-family cytokines (Fig. 2a). MDA-MB-468 cells showed a ~ 2-fold induction of LOXL2 protein expression with OSM treatment, a slight upregulation by IL-1β (not significant), and no change with IL-6 or LIF (Fig. 2b). BT474 cells showed the largest increase in LOXL2 expression with a ~ 10-fold increase, with LIF also promoting a ~ 3-fold induction (Fig. 2c). While with the Sk-Br-3 cells, OSM induced the expression of LOXL2 ~ 2-fold and IL-6 promoted a ~ 3-fold increase in LOXL2 protein (Fig. 2d). None of the cytokines induced a significant change in LOXL2 protein expression in highly invasive MDA-MB-231 cells that constitutively express very high levels of LOXL2 (Fig. 2e), and T47D (ER+ PR+ HER2−) cells did not express any LOXL2 protein before or after treatment (data not shown). The bands represent the 105 kDa LOXL2 protein, which we expect correlates to the secreted and enzymatically active form of LOXL2 that is glycosylated at the N593 and N627 amino acids [50]. LOXL2 protein induction was highest after 24 h, which was supported by In-Cell Western analysis (Supp. Figure 2). These results confirm that OSM signaling, and to a lesser extent IL-1β signaling, induce LOXL2 protein expression in IDC cells, while other IL-6-family cytokines do to a much lesser extent.

Expression of LOXL2 in breast cancer cells is positively correlated with the invasive potential of the breast tumor cells [34, 41]. Therefore, we wanted to determine and compare relative LOXL2 expression between the IDC cell lines treated with OSM. To compare relative expression, immunoblot analysis was performed on lysates from non-treated and OSM-treated cells after 24 h. We observed a significant stepwise increase in constitutive LOXL2 protein expression from the least (MCF7) to most (MDA-MB-231) aggressive cell line. OSM treatment promoted the expression of LOXL2, bridging the gap in LOXL2 expression between the cell lines (Fig. 2f). Our results suggest that OSM-induced LOXL2 protein expression may be correlated to the development of more aggressive invasive ductal carcinomas due to the incremental increase in LOXL2 with OSM exposure. Taken together these results confirm OSM-induced LOXL2 at the mRNA level leads to LOXL2 protein expression, which is correlated with increasing aggressiveness of IDC cells.

To characterize the effects of OSM signaling on the expression of the different family members of lysyl oxidase, we performed immunoblot assays using MCF7 and MDA-MB-468 cells that were treated for 24 h with OSM. LOX expression was analyzed with IL-6 and LIF treatments, in addition to OSM. Besides LOXL2, the only lysyl oxidase detectable by immunoblot analysis in MCF7 cells was LOXL1. OSM treatment, however, did not alter any of the other lysyl oxidase members (Fig. 2g). These results suggest that OSM induces only LOXL2 expression in the IDC cell lines. Based on these results, we chose to further focus on the OSM-LOXL2 axis in IDCs cells using MCF-7 cells as our model system. Though OSM also exclusively induced the expression of LOXL2 in MDA-MB-468 cells, these cells constitutively expressed LOX protein (Supp. Figure 3). This high endogenous expression of LOX may represent a confounding variable for functional analysis of LOXL2. Taken together, OSM signaling does not impact the expression of all lysyl oxidases but seems to be unique to LOXL2.

EMT has been widely implicated in regulating cell invasion and metastasis [81]. OSM signaling and LOXL2 nuclear localization have been implicated in promoting epithelial to mesenchymal transition in ductal carcinoma cells [17‐19, 43, 56, 82]. Indeed, MCF7 cells treated with OSM induced cytoplasmic localization of the epithelial marker E-cadherin (E-Cad) as depicted by immunofluorescence analysis (Fig. 3a). Given that OSM induces LOXL2 expression and the latter has been also implicated in promoting EMT, we next explored whether EMT induced by OSM is dependent on LOXL2 expression. To this end, MCF7 cells stably expressing shRNA targeting LOXL2 (MCF7-sh-LOXL2) and MCF7 cells stably expressing sh-Non-target (MCF7-sh-Non-target) were treated with OSM, and expression of E-cadherin and Snail, a transcription factor mediating EMT, were determined by immunoblot analysis. Our results demonstrate that knockdown of LOXL2 in MCF7 cells did not inhibit OSM-induced EMT, given that E-cadherin expression was slightly downregulated and Snail expression was upregulated upon OSM treatment in both MCF7-sh-Non-target and MCF7-sh-LOXL2 cells (Fig. 3b). Similarly, in MDA-MB-468, our results demonstrate that knockdown of LOXL2 in MDA-MB-468 cells did not inhibit OSM-induced EMT, given that E-cadherin expression did not change and Snail expression was upregulated with OSM treatment in both siCTRL and siLOXL2-exposed MDA-MB-468 cells (Fig. 3c).

Notably, we previously demonstrated that nuclear expression of LOXL2 is required to promote EMT in MCF7 cells [43]. Therefore, we determined the cellular localization of LOXL2 upon OSM induction. We envisioned that OSM induced only cytoplasmic expression of LOXL2, thus promoting EMT independent of LOXL2 expression. To this end, we performed nuclear and cytosolic fractionation on MCF7 cells treated with OSM, using GAPDH as a cytosolic marker and Snail as a nuclear marker (Fig. 3d). Indeed, OSM induced cytoplasmic expression of LOXL2 while the nuclear fraction did not contain any nuclear LOXL2. This data confirmed that OSM did not induce nuclear LOXL2 expression where it could promote EMT through the stabilization of Snail. Taken together, these results demonstrate that OSM promotes EMT independently of its induction of LOXL2 protein.

LOXL2, in addition to its cell autonomous activity, has been well studied for its extracellular activity on ECM proteins [38, 83]. Secreted LOXL2 promotes collagen I fiber crosslinking and affects matrix remodeling, which has been linked to increased metastatic capability in breast cancer [38‐42]. Interestingly, immunoblot analysis suggested that OSM-induced expression of the N-linked glycosylated form of the LOXL2 protein (105 kDa) which is secreted into the tumor microenvironment [50]. To confirm that OSM-induced the expression of glycosylated and enzymatically active LOXL2, we determined the N-linked glycosylation status of expressed LOXL2. MCF7 and MDA-MB-468 cell lysates treated with OSM were exposed to the N-linked deglycosylase enzyme, PNGase F, before immunoblot analysis was performed. OSM induced the expression of the 105 kDa LOXL2 protein, which was reduced in size to 87 kDa following the addition of PNGase F in both MCF7 (Fig. 4a) and MDA-MB-468 (Fig. 4b) cells. To confirm and quantify LOXL2 secretion, we performed an ELISA on MCF7 cells treated with OSM for 36 h. We observed a significant induction in LOXL2 protein secretion with OSM treatment averaging 877 pg/mL of LOXL2 in solution. In comparison to the non-treated samples that averaged 347.6 pg/mL of LOXL2, we observed an ~ 2.5-fold induction in LOXL2 secretion (Fig. 4c). Together, these results confirm that the LOXL2 protein expressed through OSM signaling in IDC cells is N-linked glycosylated and secreted.

In order to confirm that the LOXL2 protein induced by OSM is enzymatically active, we performed a lysyl oxidase activity assay on MCF7 cell conditioned media. MCF7 cells were transfected with siLOXL2 or siCTRL and treated with OSM for 24 h. In the siLOXL2 group, we observed that OSM-treated samples had significantly reduced lysyl oxidase activity compared to the siCTRL group (Fig. 4d). We saw a significant ~ 2-fold increase in lysyl oxidase activity with OSM treatment when comparing against non-treated controls in the siCTRL group. The accompanying immunoblot confirms the knockdown of LOXL2 protein in the MCF7 cell line. Further immunoblot analysis confirmed that there was no impact on LOXL1 protein expression (Supp. Figure 4). Based on these results, we conclude that OSM-induced LOXL2 is enzymatically active and accounts for all of the lysyl oxidase enzymatic activity present in MCF7 cell conditioned media.

To assess the effect of OSM-induced LOXL2 on crosslinking collagen I, we performed a collagen contraction assay. MCF7 and MDA-MB-468 cells were seeded into a 1.5 mg/mL rat-tail collagen I matrix, and the cells were treated for 48 h with OSM, a combination of OSM and the pan-LOX inhibitor βAPN (500 μM), or a combination of OSM and the LOXL2/3-specific inhibitor PXS-5120A (200 nM). βAPN, or β-aminopropionitrile, is a small molecule inhibitor (SMI) commonly used as a nonspecific inhibitor for lysyl oxidase proteins [64, 83]; PXS-5120A (PXS-S1A) is a LOXL2-specific inhibitor at a range of concentrations in the nanomolar range [84, 85]. OSM induced a ~ 2.5-fold increase in collagen I contraction in MCF7 cells, as compared to the non-treated control, while OSM-induced contraction was blocked by βAPN and PXS-5120A treatment (Fig. 5a). In MDA-MB-468 cells, we saw a similar ~ 2-fold increase; however, the total contraction was substantially reduced compared to MCF7 cells (Fig. 5b). These results demonstrate that OSM-induced LOXL2 increases collagen I contraction, and suggest that OSM promotes crosslinking through induced LOXL2, as collagen I contraction correlates to collagen crosslinking [64, 65].

To visualize collagen alignment, we performed Live-Cell confocal imaging on MCF7 cells treated with OSM and/or βAPN for 36 h in the collagen I matrix described above. Prior to imaging, the MCF7 cells were exposed to CellTracker Red (depicted in red), and the collagen I fibers were visualized by resonance scanning of the matrix (depicted in green). As seen visually, OSM-promoted collagen I alignment and inhibition of LOXL2 with βAPN reduced fiber alignment (Fig. 5c). Images were processed by ImageJ in order to highlight fiber density at its greatest intensity (Fig. 5d). CurveAlign4.0 software [68, 69] was used to quantify the alignment of collagen I fibers in-between, and tangential to, MCF7 cells by analyzing alignment in selected regions of interest (ROI) (Supp. Figure 5). Analysis of the fiber dispersion coefficient in the ROIs using CurveAlign4.0 showed a significant ~ 2-fold increase in fiber alignment with OSM treatment, as the closer the coefficient is to 1 the more alignment is present. The increase in alignment was significantly reversed with the inhibition of LOXL2, using βAPN (Fig. 5e). These results confirm that OSM-induced LOXL2 leads to collagen I fiber alignment in the ECM.

After determining that OSM-induced LOXL2 promotes a significant increase in collagen I fiber alignment, we wanted to examine whether OSM-induced LOXL2 in turn impacts invasion. Research suggests that an increase in collagen I fiber alignment leads to an increase in invasion of breast cancer cells [52, 53]. To assess the impact of OSM-induced LOXL2 on IDC cell invasive potential, we performed a 3D invasion assay on IDC cells. We utilized MCF7 cells that incorporate green fluorescent protein (GFP) for visualization using fluorescent microscopy. MCF7-GFP cells supplemented with β-estradiol (50 ng/mL) were seeded in a collagen I solution with a concentration of 1.5 mg/mL. The cell suspension was then added to wells of a 96-well plate containing a circular “cell-free zone” for cells to invade towards. Cells were treated with OSM, OSM with βAPN (500 μM), or OSM with PXS-5120A (200 nM) and compared against a non-treated control. Images were taken at day 0, as a control, and the experiment was run until day 5 (Fig. 6a). We analyzed the fluorescent images using ImageJ to determine the number of cells that entered the “cell-free zone” after day 5. The total number of MCF7-GFP cells that entered the “cell-free zone” with OSM treatment increased greater than 3-fold compared to the non-treated control, and cell invasion was in turn significantly decreased by LOXL2 inhibition using both βAPN (~ 3-fold) and PXS-5120A (~ 3-fold) SMIs (Fig. 6b). These results suggest that OSM-induced LOXL2 is critical to OSM-promoted invasion and may likely have impact on metastasis.

Taken together, these results demonstrate that OSM induces sufficient LOXL2 protein expression/secretion to promote remodeling and alignment of collagen I fibers in the ductal carcinoma tumor microenvironment. Due to the alignment of collagen I fibers, it is expected that OSM-induced LOXL2 will promote ductal carcinoma cell invasion and metastasis as tumor cells migrate along aligned collagen fibers [37, 53]. This was supported by our 3D invasion assay results, which showed that OSM induced an ~ 3-fold increase in the number of MCF7-GFP cells that invaded the “cell-free zone,” when compared to OSM treatment in conjunction with LOXL2 inhibition. Therefore, this research highlights a novel mechanism in ductal carcinoma tumor progression, independent of EMT. Further research is needed to confirm that OSM-induced LOXL2 extracellular matrix remodeling leads to a significant increase in metastasis.