4T1 and B16 cells were cultured in DMEM (Gibco) with 10% FBS. HCmel12 cells were cultured in DMEM with 20% FBS. Cancer-associated fibroblasts (CAFs) was cultured in DMEM with 10% FBS and 1% Insulin-Transferrin-Selenium (100X, Gibco).

The generation of nidogen-1 CRISPR knockout cell lines was performed according to published protocols (8). sgRNA for CRISPR was designed using the DESKGEN cloud platform (9). Two sgRNA oligos 5’- CACCGCTGCCATCTGAATAATGAA -3’ and 5’- AAACTTCATTATTCAGATGGCAGC -3’ targeting nidogen-1 were cloned into the vector SpCas9-2A-Puro V2.0 8. After transfection, single cell isolation was performed. Knockout clones were confirmed by Sanger sequencing of the genomic DNA.

All animal experiments were approved and conducted according to the regulations of The Animal Experiments Inspectorate, Danish Veterinary and Food Administration, with license numbers 2012-15-2934–00222 and 2017-15-0201-01265, as well as regulations from the Department of Experimental Medicine, University of Copenhagen. We obtained the nidogen-1 knockout mouse as a generous gift from Prof. Thomas Krieg at the University Hospital of Cologne (10). The αSMA::RPF transgenic mouse was published previously (11). The PyMT-Col1α::YFP transgenic mice were generated and maintained by the Erez Lab (Raz Y. and Cohen N. et al., under revision). Other wildtype mice were ordered from Taconic Biosciences.

For the tumor-fat pad junction staining experiments, 1,000,000 4T1 H2B GFP cells were injected into the mammary fat pad of 10-week old BALB/c mice. Primary tumors together with mammary fat pads were resected 10 days after injection. For αSMA+ stromal cell and 4T1 FACS sorting experiments, 1,000,000 4T1 H2B GFP cells were injected into the mammary fat pad of 16-week old αSMA::RPF transgenic mice. Primary tumors together with mammary fat pads were resected 8 days after injection. For melanoma orthotopic models, 200,000 HCmel12 cells were injected subcutaneously into 10-18-week old mice. Humane endpoint was defined when primary tumor reached 10mm diameter. For melanoma experimental metastasis models, 200,000 HCmel12 cells were injected into the lateral tail veins of 9-11-week old mice. Mice were euthanized 4 weeks after injection.

The isolation of normal mammary gland fibroblasts (NMFs) and cancer-associated fibroblasts (CAFs) from the PyMT model, and healthy lung fibroblasts and metastatic lung fibroblasts from the BALB/c model were performed as previously described (12). This reference (12) published in the Journal of Visualized Experiments (JoVE) is a detailed step-by-step protocol methodology paper (with videos and graphs).

Cell exclusion was performed using the CD326 (EpCAM) antibody (clone G8.8, eBioscience) and the CD45 antibody (clone 30-F11, eBioscience). PDGFRα antibody used was clone APA5 (eBioscience). The isolation of GFP+ and RFP+ cells from the αSMA::RPF model was performed using the gentleMACS Dissociator following Miltenyi Biotec’s instructions. We have also included detailed control data for tumor cells control ( Supplementary Figure 6 ), dsRed channel control ( Supplementary Figure 7 ), and GFP channel control ( Supplementary Figure 8 ). In addition, detailed view of the FACS experiment in Figure 1C is also included ( Supplementary Figure 9 ).

RNA from sorted cells was isolated using the RNeasy Micro Kit (Qiagen), followed by cDNA synthesis using the iScript cDNA Synthesis Kit (Bio-Rad). qPCR was performed using the LightCycler 480 SYBR system (Roche). Nidogen-1 oligos were 5’-GTATCCCCCTCCCTGGAACT-3’ and 5’-TCGCTCATGGCGATGATACC-3’. Col1α oligos were 5’-TGTGTTCCCTACTCAGCCGTCT-3’ and 5’-CTCGCTTCCGTACTCGAACG-3’. αSMA oligos were 5’-AGCCAGTCGCTGTCAGGAA-3’ and 5’-CGAAGCCGGCCTTACAGA-3’.

For the qPCR analysis in Figure 1 , we first calculate the mean CT from raw CT, and then deducted b-actin value which gives the dCT value. And then we calculated the 2^-dCT*100 value. In the end, we normalize CAF in vitro = 1.

For qPCR analysis in Figure 2 , mGUS was used as a housekeeping gene. We first calculate the mean CT from raw CT, and then deducted mGUS value which gives the dCT value. And then we calculated the 2^-dCT*100 value. Then we normalized NMF or healthy lung fibroblast as 1 and normalized the rest values accordingly.

Different housekeeping genes were used because these two experiments were performed in two different labs (Erler Lab in University of Copenhagen and Erez Lab in Tel Aviv University). We acknowledge that in an ideal situation the same housekeeping gene should be used within the same study.

1,000,000 cells were seeded in a 10-cm dish in their normal growth medium. The day after, normal medium was switch to serum free medium. 24hr after, conditioned medium (CM) was collected, filtered through a 0.45 μm filter, and concentrated with Vivaspin 6 centrifugal concentrators (Sartorius). RIPA buffer was used to prepare cell lysates. Protein concentration was measured and adjusted using Bradford assay. For each condition, equal amount protein was loaded for western blot. Antibody against nidogen-1 was AP02274SU-N from Acris. Quantification of western blot results was performed using the Plot Lanes function of the Fiji software. In Supplementary Figures 2A, B we provided the full films of the western blots with protein size ladders labeled on 2b.

The left lung lobes were fixed in 10% neutral buffered formalin solution (Sigma) overnight at room temperature before stored in 70% ethanol at 4 °C. Lung lobes were then embedded in paraffin and sectioned with maximized surface area at 3 μm thickness. 6-step sections were performed, with 500 μm distance between step 3 and 4, and 100 μm distance between other steps. Haematoxylin and eosin (H&E) staining were performed to score lung metastatic tumors. For each mouse, the final lung metastasis foci number is the average number of the 6 sections scored.

For the tumor-fat pad junction staining, samples were preserved in 30% sucrose at 4°C overnight before embedding with OCT and snap frozen using dry ice. Samples were stored at -80°C. Tissue blocks were then cryosectioned at 6 μm thickness.

For immunofluorescent staining, frozen sections were first incubated at 37°C for 10 min, then fixed in 4% paraformaldehyde solution. Slides were then permeabilize using TBST (0.05% tween-20) followed by blocking with 5% normal serum from the host of the secondary antibody. Primary antibodies against nidogen-1 (1:100, Acris, AP02274SU-N), collagen IV (1:100, Millipore, AB756P) and αSMA (1:100, Sigma, C6198) was incubated overnight at 4 °C. Appropriate secondary antibodies (Alexa Fluor) were incubated for 1 hr at room temperature.

RNA in situ hybridization was performed using RNAscope 2.5 HD Detection Reagents-RED system following manufacturer’s protocol (Doc. Num. 320534 and 322350-USM). Nidogen-1 probe number is REF482541.

Transmission Electron Microscopy (TEM) was performed at the Oulu EM Core Facility. Lung samples were fixed in 0.1 M phosphate buffer containing 1% glutaraldehyde and 4% formaldehyde. Samples were then post-fixed in 1% osmium tetroxide, dehydrated in acetone and embedded using the LX-112 kit (Ladd Research). Thin sections were cut at 70nm using the Ultracut UCT ultramicrotome (Leica), stained in uranyl acetate and lead citrate and examined in the Tecnai G2 Spirit TEM microscope (FEI). Images were captured by the Quemesa camera and analyzed using the iTEM software (Olympus).

We previously identified nidogen-1 as strongly downregulated in metastases of the lungs compared to healthy lung tissue, using mass spectrometry analysis of decellularized organs obtained from the 4T1 syngeneic mouse model of breast cancer (7). It has previously been shown that nidogen-1 is expressed in mammary mesenchymal and myoepithelial cells, but not in epithelial cells (13). In addition, most fibroblast cell lines express nidogen-1 (13). We therefore initially focused on breast cancer and examined the nidogen-1 expression pattern in the 4T1 mouse model of breast cancer.

We injected 4T1 breast cancer cells expressing histone H2B-GFP into the mammary fat pad of syngeneic BALB/c mice and stained the border region of the 4T1 tumors and healthy fat pad with both nidogen-1 and collagen IV antibodies to visualize the BM ( Figure 1A ). Both nidogen-1 and collagen IV were present at the tumor region and the healthy fat pad region. The morphology of the stained structures suggests that both nidogen-1 and collagen IV are located at the BM surrounding vessels and mammary ducts ( Figure 1A ). However, nidogen-1 staining was stronger in the healthy fat pad region as compared to the tumor region. Since this is a staining of one single slide containing border regions of both tumor and healthy fat pad tissues and both regions on the same slide were treated with the same antibodies, incubation conditions, microscopy settings, we could reach the conclusion that nidogen-1 staining was stronger in the healthy fat pad region as compared to the tumor region.

We have also performed CD31 and nidogen-1 co-localization immunofluorescent staining in 4T1 primary tumor samples ( Supplementary Figure 3 ), and the co-localization results supports nidogen-1 is located at the BM surrounding vessels.

In addition, we provided deeper characterization of nidogen-1 from higher magnification immunofluorescent staining images showing detailed views of nidogen-1 in primary 4T1 tumor in fat pad area ( Supplementary Figure 4 ) and in lung metastasis of 4T1 tumor area ( Supplementary Figure 5 ).

To investigate which cells were producing nidogen-1, we used RNAScope technology to visualize nidogen-1 mRNAs in the border region of the 4T1 tumor and the healthy fat pad. We found that the majority of nidogen-1 mRNA (pink dot) is present in the healthy fat pad region and potentially associate with the expression pattern of αSMA ( Figure 1B ). These findings suggested that nidogen-1 is expressed by αSMA expressing cells and not the 4T1 cancer cells.

To confirm that nidogen-1 is expressed by αSMA positive cells and not 4T1 cancer cells, we injected H2B-GFP labelled 4T1 cells into αSMA::RPF transgenic mice (11). We then resected the primary tumor, performed single cell isolation, and used FACS to sort out GFP + 4T1 cells and RFP+ cells (containing fibroblasts, pericytes and myoepithelial cells) ( Figure 1C ). To quantify the mRNA levels of nidogen-1 in the ex vivo sorted out cells, we performed qPCR analysis. We also compared the nidogen-1 expression to in vitro cultured 4T1 and CAFs, which are major αSMA expressing cells present in tumors. The results indicated that nidogen-1 was expressed in αSMA+ cells and confirmed that 4T1 cells do not express nidogen-1 in tumors ( Figures 1C, D ).

To further investigate these findings at the protein level, we performed western blot analysis of nidogen-1 protein expression in 4T1 breast cancer cells and CAFs. In order to test the generalizability of the findings, we also included two melanoma cell lines, HCmel12 and B16F10. Analysis of the results confirmed that nidogen-1 is expressed in CAFs but not in 4T1, HCmel12, and B16F10 cancer cells ( Figure 1E ).

In order to confirm the specificity of the nidogen-1 antibody (Acris, AP02274SU-N), we utilized CRISPR to generate CAF nidogen-1 knockout cells. Western blot analysis showed an absence of nidogen-1 expression in the knockout cells ( Figure 1E ), validating our findings. Taken together, these findings suggest that nidogen-1 is highly expressed in CAFs but not in cancer cells.

We previously identified nidogen-1 as being downregulated in 4T1 breast tumors compared with healthy mammary fat pad (7). We therefore investigated whether CAFs express less nidogen-1 than normal fibroblasts. We examined this using two different mouse models of breast cancer ( Figures 2A, B ).

The first model was based on the PyMT transgenic breast cancer model, which develops spontaneous tumors. These mice were manipulated to express YFP when the collagen α1 gene is expressed, which predominantly occurs in CAFs (22). In this first model, we compared Col1α+ fibroblasts from healthy Col1α::YFP transgenic mice with fibroblasts from tumor bearing Col1α::YFP transgenic MMTV-PyMT mice ( Figure 2A ). We FACS sorted out YFP+ cells and performed qPCR analysis. We found that CAFs purified from the PyMT mice express 50% lower levels of nidogen-1 compared to fibroblasts isolated from normal mammary glands ( Figure 2A ). Our data also confirm αSMA as a marker for CAFs (14).

We then focused our attention to the metastatic site and used tail vein injection of cells to study the colonization and outgrowth of metastases, and isolate CAFs for analysis of nidogen-1 expression. In this second model, we performed tail vein injection of 4T1 cells into BALB/c mice and compared fibroblasts from healthy and metastatic lungs. Here, we FACS sorted PDGFRα+ fibroblasts and performed qPCR analysis. PDGFRα has been shown to be broadly expressed in normal and neoplastic fibroblasts (15). We have also shown that 4T1 cells do not express PDGFRα ( Supplementary Figure 10 ). We found that CAFs from metastatic lungs express 80% lower levels of nidogen-1 compared to healthy lung fibroblasts ( Figure 2B ). These data suggest that fibroblasts in the primary tumor and in metastatic lung foci express less nidogen-1 as compared to the healthy tissue. The data is consistent with our previous mass spectrometry findings that nidogen-1 is slightly downregulated in the primary breast cancer while substantially reduced in the metastatic lungs (7).

In order to test the clinical relevance of our findings, we probed human cancer data sets available from the Oncomine database (16). We found that in both the Curtis Breast dataset (17) and the Karnoub Breast dataset (18), nidogen-1 is downregulated in breast tumors compared to healthy breast tissues ( Figures 2C, D ). Taken together, these findings suggest that nidogen-1 expression is reduced in CAFs in breast tumors compared with healthy tissue.

It has been reported that nidogen-1 knockout mice do not have increased lung metastasis compared to nidogen-1 wildtype or heterozygous mice in a B16 melanoma tail vein injection model (19). To investigate this further, we obtained nidogen-1 knockout mice and switched to a melanoma model given this previous report and the strain of the mice (not syngeneic with the breast cancer models). We investigated the role of stromal-derived nidogen-1 in melanoma lung metastasis using wildtype and nodogen-1 knockout mice. Given that B16 cells do not form sufficient spontaneous metastases from orthotopic tumors, we used HCmel12 cells that naturally lack nidogen-1 expression, in orthotopic and tail vein metastasis models in both wildtype and nidogen-1 knockout syngeneic mice ( Figure 3 ).

In the orthotopic model, we did not observe any difference in primary tumor growth rate between wildtype and nidogen-1 knockout mice ( Figure 3A ). However, when we analyzed spontaneous lung metastasis in these mice, 7 out of 11 (64%) wildtype mice and 10 out of 12 (83%) of nidogen-1 knockout mice developed detectable lung metastases. Although not statistically significant, there was a trend that nidogen-1 knockout mice display more lung metastases (p = 0.28, chi-square test). Mice having lung metastases did not show any difference in the numbers of metastatic foci when comparing wildtype and nidogen-1 knockout mice ( Figure 3B ).

We then performed tail vein injection of HCmel12 into both wildtype and nidogen-1 knockout mice. 8 out of 10 wildtype mice (80%) and 7 out of 9 nidogen-1 knockout mice (78%) developed detectable lung metastases. When quantifying the number of lung metastatic lesions in mice developing lung metastasis, we found that although there was no statistically significant difference in the size of metastatic foci ( Figure 3C ), there was a significant increase in the number of metastatic foci in the nidogen-1 knockout mice compared with wildtype mice ( Figure 3D ). These findings suggest a role for stromal-derived nidogen-1 in supporting metastatic colonization.

The BM is known to act as a critical barrier against cancer progression. We therefore hypothesized that loss of nidogen-1 may cause BM defects in the lungs, which compromise BM barrier function, thereby increasing the ability of cancer cells to extravasate and colonize the lungs. In order to test this hypothesis, we performed electron microscopy (EM) to check the lung alveoli structures in both wildtype and nidogen-1 knockout mice.

Previous studies showed that loss of nidogen-1 causes thinning and discontinuity of the BM in brain capillaries (20), however loss of nidogen-1 does not cause BM defects in the kidneys and muscles (10).

Analysis of the EM data revealed defects in the lung alveoli of nidogen-1 knockout mice ( Figure 4 ). First, we observed fragmented endothelium in nidogen-1 knockout mice. While lung alveoli in wildtype mice had continuous and smooth endothelium, nidogen-1 knockout mice exhibited fragmented endothelium with obvious gaps. In addition, the BM underlying the endothelium was fragmented and poorly defined in some capillaries ( Figure 4A , arrows highlight fragmented endothelium with gaps, and poorly defined BM).

In addition, we also observed structural differences in alveolar interstitium between wildtype and nidogen-1 knockout mice. In the wildtype mouse lung, alveolar interstitium had well-defined BMs, extracellular matrix and fibroblasts, and exhibited lower amounts of fibrillar collagen and elastin. In comparison, the lungs of nidogen-1 knockout mice had enlarged interstitium with increased amounts of fibrillar collagen and elastin ( Figure 4B , arrow heads and stars highlight increased amounts of fibrillar collagen and elastin). Moreover, we observed that in some cases red blood cells were found in the alveolar lumen instead of in capillaries in the nidogen-1 knockout mice. This suggests that the lung vessels of the nidogen-1 knockout mice are leakier.