Nephronectin promotes breast cancer brain metastatic colonization via its integrin-binding domains

This study demonstrates a role for the extracellular matrix protein nephronectin (NPNT) in promoting experimental breast cancer brain metastasis, possibly through enhanced binding to- and migration through brain endothelial cells. With the introduction of more targeted breast cancer treatments, a prolonged survival has resulted during the last decade. Consequently, an increased number of patients develop metastasis in the brain, a challenging organ to treat. We recently reported that NPNT was highly expressed in primary breast cancer and associated with unfavourable prognosis. The current study addresses our hypothesis that NPNT promotes brain metastases through its integrin-binding motifs. SAGE-sequencing revealed that NPNT was significantly up-regulated in human breast cancer tissue compared to pair-matched normal breast tissue. Human brain metastatic breast cancers expressed both NPNT and its receptor, integrin α8β1. Using an open access repository; BreastMark, we found a correlation between high NPNT mRNA levels and poor prognosis for patients with the luminal B subtype. The 66cl4 mouse cell line was used for expression of wild-type and mutant NPNT, which is unable to bind α8β1. Using an in vivo model of brain metastatic colonization, 66cl4-NPNT cells showed an increased ability to form metastatic lesions compared to cells with mutant NPNT, possibly through reduced endothelial adhesion and transmigration.


NPNT transcript is up-regulated in human BC and linked to poor prognosis. To investigate
whether NPNT expression was dysregulated in BC, we performed deep sequencing of human tumour tissue and adjacent non-cancerous breast tissue from 22 patients. Pooled RNA from tumour tissue was compared to pooled RNA from pair-matched normal tissue. A total of 1,323 genes were differentially expressed: 613 up-regulated and 710 down-regulated. The exhaustive list of genes is presented in Supplementary Table S1. NPNT was upregulated 3.63 fold (p = 2.58E−12) in tumour tissue compared to normal breast tissue, indicating a dysregulation of NPNT gene expression in BC. No change was detected in the levels of integrin subunits α8, αV, β1, β3 or β5. To verify the results from the deep sequencing, RT-qPCR was performed on RNA extracted from each tumour sample and the pair-matched normal tissue using NPNT specific primers. An increase in NPNT expression was observed in 21 out of 22 samples (Fig. 1a). Using the Pam50 classifier 20 and the BreastMark database 21 , no association was found between high NPNT expression and survival, hazard ratio 0.8678 (0.7235-1.041) (Fig. 1b). There was however, a significant association between elevated NPNT levels and poor outcome in patients with the luminal B subtype, hazard ratio 1.46 (1.072-1.989) (Fig. 1c).
Resections from five patients with BC brain metastasis were analysed by RNA Scope in situ hybridization (ISH) for the presence of NPNT mRNA. Positive and negative controls can be viewed in Supplementary file 1 and Supplementary Fig. S1. As a proof of concept, NPNT mRNA was detected in all of the samples (Fig. 1d), with three samples showing areas of clear positive staining and two showing weak focal staining. For the sample in Fig. 1d, showing strong staining, the surrounding tumour stroma was negative for NPNT. IHC analyses demonstrated that NPNT protein expression corresponded to the NPNT mRNA expression (Fig. 1e, f, Supplementary Fig. S1). www.nature.com/scientificreports/ The presence of the NPNT-binding integrin subunit α8 (Itgα8) was analysed by IHC in the human BC brain metastases. Itgα8 was present in the same areas that showed NPNT expression (Fig. 1g). Positive staining for the α8 subunit indicates the presence of the α8β1 receptor. These results show that the brain metastatic BC cells express both NPNT and integrin α8β1.
NPNT is upregulated in mouse models of BC. To assess whether mouse models of BC could recapitulate the NPNT expression observed in human BC tissue, we analysed NPNT protein levels by IHC on tissue sections from either MMTV-PyMT mice, a mouse model that closely mimics the development of human BC 22 , wild-type female FVB mice, or pregnant FVB mice. Normal mammary glands from FVB mice showed no NPNT staining whereas glands from pregnant FVB mice showed weak staining. MMTV-PyMT BC tissue was clearly positive (Fig. 2a). Three human BC cell lines with different receptor status were used to create xenograft tumours in mice: MCF-7 (luminal A), SK-BR-3 (HER2+) and BT474 (luminal B) 23 . NPNT was expressed in all tumours (Fig. 2b, c), with highest expression in the BT474 tumours (Fig. 2c).
The expression and distribution of Npnt and the proposed NPNT-interacting integrins in the brain were retrieved from the Mouse Allen Brain Atlas 24 (2004 Allen Institute for Brain Science. Allen Mouse Brain Atlas. Available from: https ://mouse .brain -map.org/), an open database generated by the Allen Institute (https ://www. allen insti tute.org/). ISH showed generally low Npnt expression in the mouse brain, with stronger expression along the granule cell layer of the dentate gyrus, an area known to contain stem/progenitor cells 25 ( Supplementary  Fig. S2). Staining for integrin subunit α8 mRNA was mostly negative in adult mouse brain tissues (results not shown). The RGD-binding integrins αVβ3 and αVβ5 are expressed by many endothelial cells in the body and are potential receptors for NPNT 10 . The mRNA of the αV subunit was distributed evenly throughout the brain tissue ( Supplementary Fig. S2), with strong expression in the pyramidal layer of the piriform area of the cerebral cortex ( Supplementary Fig. S2). Taken together, these results show that NPNT is expressed in different mouse models of BC, and only in restricted regions of the brain.
NPNT co-localizes with integrin α8β1 on the cell surface of mouse BC cells. The mouse BC cell line 66cl4 was chosen for further studies due to low endogenous expression of NPNT, expression of integrin α8β1 ( Supplementary Fig. S3), and a weak metastatic potential in vivo, preferentially metastasising to the lungs 18,19 . This cell line was therefore an optimal choice for over-expression of either wild type NPNT (66cl4-NPNT), or NPNT where the two known integrin-binding sites (RGD and LFEIFEIER-enhancer site) were mutated (66cl4-RGE-AIA). As a control 66cl4 cells were stably transfected with an empty vector (66cl4-EV) 19 . More detailed information and expression data can be viewed in Supplementary file 1 and Supplementary  Fig. S3. To test whether NPNT could locate to the cell surface of 66cl4 cells, we analysed the distribution of NPNT, integrin αVβ3 and integrin subunit α8 by immunofluorescence (Fig. 3). The α8 subunit is believed to exclusively heterodimerize with the β1 integrin subunit 26,27 , hence the staining represents α8β1. The control cells, 66cl4-EV, showed no expression of NPNT, and focal distribution of α8β1 (Fig. 3a). On 66cl4-NPNT cells NPNT co-localized with α8β1 (Fig. 3b). 66cl4-NPNT cells showed no expression of integrin αVβ3 (Fig. 3c). When recombinant mouse NPNT (rmNPNT) was exogenously supplied to 66cl4-EV cells, co-localization of integrin α8β1 and rmNPNT was observed (Fig. 3d). However, in cells expressing NPNT mutated in the integrin binding sites, all co-localization with α8β1 was lost (Fig. 3e). Controls showed no autofluorescence and no unspesific binding of the secondary antibody (Fig. 3f). Taken together, these results demonstrate that extracellular NPNT can co-localize with integrin α8β1. npnt and its integrin-binding sites enhance the rate of brain metastatic colonization in mice. As a model system for BC brain metastasis, the mCherry expressing 66cl4 cells were used to explore the role of NPNT and integrin interaction during the metastatic process. To assess the importance of NPNT in establishing BC brain metastasis in vivo, a mouse model for experimental brain metastasis was used 28 . The 66cl4-EV, -NPNT and -RGE-AIA cells were injected into the carotid artery to assess whether cells were able to establish brain metastases. As a control, three mice were injected with buffer only. Metastatic cells were identified using antibodies towards the overexpressed mCherry protein 18 . Controls showed that the mCherry antibody was specific (Fig. 4a). All cell lines established brain metastases within seven days, where the metastatic lesions presented as four different phenotypes (Fig. 4b): single cells (I), cells surrounding vessels/vessel co-option (II), vessel outgrowth (III) and established metastatic tumour (IV). Compared to the 66cl4-EV cells, the 66cl4-NPNT the probability for disease specific survival based on high/low (median cut-off) expression levels of NPNT mRNA in BC (all subtypes) and related to months after diagnosis. Data was collected using the publicly available database BreastMark. N = 876, number of events = 466. (c) Kaplan-Meier survival plots showing the probability for disease specific survival based on high/low expression levels of NPNT mRNA (median cut-off) in luminal B subtype of BC and related to months after diagnosis. Data was collected using the publicly available BreastMark database. p < 0.05 was considered statistically significant. N = 323, number of events = 188. (d, e) Tissue sections of BC brain metastasis (N = 5) were analysed by RNA Scope ISH for the presence of NPNT mRNA. Each brown dot indicates the presence of NPNT mRNA. (b) Representative image of metastasis with high levels of NPNT mRNA. The tumour stroma was negative; T = tumour, S = stroma. Counterstained with haematoxylin. (e-g) Tissue sections of the same region from one BC brain metastasis showing positive brown staining for NPNT mRNA (e) NPNT protein (f) and integrin α8β1 (Itgα8) protein (g) in the cancer cells. Counterstained with haematoxylin.
Scientific RepoRtS | (2020) 10:12237 | https://doi.org/10.1038/s41598-020-69242-1 www.nature.com/scientificreports/ cells on average established more of lesion type I + II (Fig. 4c). Mutating the RGD and EIE integrin-binding sites drastically reduced the amount of lesion I + II. More advanced lesions (lesion III and IV) were less common, but similarly as for the smaller lesions, 66cl4-NPNT cells established 50 lesions in total, while 66cl4-EV cells established 35 lesions. The 66cl4-RGE-AIA cells established only 12 lesions in total. Buffer injected controls had no lesions. Taken together, these results indicate that NPNT promotes BC brain metastasis in an integrindependent manner.
NPNT does not increase the mean vessel density (MVD). NPNT was recently linked to angiogenesis 17 .
To investigate whether 66cl4 cells expressing NPNT influenced angiogenesis in our brain metastasis model, brain tissue sections were IHC stained for the blood vessel marker, CD31 ( Supplementary Fig. S4). We found no differences in mean vessel density (MVD) and mean vessel size (MVS) between mice injected with the different 66cl4 clones ( Supplementary Fig. S4). However, the MVS tended to increase in the brains of all the cancer cell injected mice compared to the buffer-injected mice ( Supplementary Fig. S4), but the difference was not statistically significant. Transvascular pillars form during intussusceptive angiogenesis; a process more common  www.nature.com/scientificreports/ in brain metastasis where vessels increase in size 4,29 . Structures resembling transvascular pillars were observed ( Supplementary Fig. S4) which could explain the increased MVS.
NPNT triggers intracellular signalling in brain endothelial cells. To assess whether NPNT could influence intracellular signalling in brain endothelial cells as a part of the metastatic process, bEND.3 cells were seeded onto an rmNPNT coated surface. A protein of approx. 70 kDa showed an increase in tyrosine phosphorylation after three hours of incubation (Fig. 5a). The phosphorylation increased with time and was most prominent after 24 h. As the bEND.3 cells do not express integrin α8β1 ( Supplementary Fig. S3), NPNT must therefore bind to another unidentified receptor. To assess which signalling pathways were involved, the cells were analysed using Proteome Profiler (Fig. 5b). As a non-adhesion control, bEND.3 cells were seeded on pluroniccoated wells. The results showed that phosphorylation of extracellular signal-regulated kinase 1 (ERK1) and ERK2 (T202/Y204, T185/Y187) were strongly up-regulated (Fig. 5b, green box), and phosphorylation of AMPactivated protein kinase (AMPKα1; T183) was strongly down-regulated (Fig. 5b, red box). Also heat-shock protein 60 (HSP60) was strongly upregulated (Fig. 5b, green box). Some proteins showed more inconsistent results, with no change in one run and up/down regulations in the other. These are also marked in Fig. 5b and included mitogen stimulated kinase 1 (MSK1) and MSK2 (S376/S360), CREB (S133), signal transducer and activator of transcription 2 (STAT2; Y689) and STAT3 (Y705). Taken together, these results show that in response to rmNPNT, several signalling pathways are activated in the endothelial cells, as shown on the tyrosine phosphorylation on Western blots and on the Proteome Profiler. This suggests that the endothelial cells are equipped with currently unidentified NPNT receptors. Fig. S3) and possibly other integrins, the 66cl4 cells adhere strongly to recombinant NPNT ( Supplementary Fig. S3), as we have also shown previously 19 . To assess the role of the integrin-binding motifs in the adhesion process, we produced and purified wild-type recombinant mouse NPNT (WT-rmNPNT) and NPNT where both the RGD-and EIE-motifs were mutated (mutant-rmNPNT). 66cl4-EV cells seeded on protein-coated wells showed increased ability to adhere to WT-compared to mutant-rmNPNT (Fig. 5c). Additionally, adhesion to WT-rmNPNT was significantly reduced in the presence of RGD peptide (Fig. 5d), as also shown previously when coating with a commercially available rmNPNT 19 . This indicates the involvement of both the RGD-and EIE-integrin-binding motifs in the adhesion process. We next assessed the role of NPNT in the adhesion of 66cl4 cells to primary mouse brain endothelial cells (MBECs). For this purpose, a well-documented in vitro attachment model was used 30,31 . As shown in Fig. 5e, 66cl4-NPNT cells showed increased ability to attach to the MBEC monolayers compared to the 66cl4-EV and 66cl4-RGD-AIA cells. When the 66cl4 cells were seeded onto the mouse brain endothelial cell line bEND.3 no significant differences were observed (Fig. 5f). This could be explained by the lack of integrin α8β1 expression by the bEND.3 cells (Supplementary Fig. S3).

the integrin-binding sites of npnt are important for adhesion and transmigration through endothelial cells. Through expression of integrin α8β1 (Supplementary
To analyse whether NPNT expression in BC cells could have an effect on their ability to migrate through MBEC monolayers, we used an in vitro model of transmigration 30,31 . As shown in Fig. 5g, 66cl4-NPNT cells had an increased ability to migrate through the MBEC monolayers compared to the 66cl4-EV and 66cl4-RGD-AIA cells. Taken together, these results show that the integrin-binding sites of NPNT are not only important for binding to primary endothelial cells, but also for the transmigratory process, implicating a functional role of NPNT in metastasis.

Discussion
Today, better treatment options for BC patients have improved the outcome and prolonged patient survival. As a consequence, the incidence of BC brain metastases have increased 32,33 . Hence, BC brain metastasis is a growing clinical challenge that needs to be addressed. Our functional studies demonstrate a role for NPNT in establishing BC brain metastases.
In a panel of mouse BC cell lines, NPNT was previously identified as one of the genes associated with increased metastatic propensity. NPNT knock-down also resulted in diminished spontaneous metastasis to the lungs, liver and spine in a mouse BC model 18 . During kidney development in mice, NPNT is known to bind integrin α8β1, triggering epithelial cells of the ureteric bud to invade the metanephric mesenchyme 10,13 . NPNT has also been linked to stem cell differentiation 15,16 and is expressed in areas of the mouse brain known to harbour stem/progenitor cells (Supplementary Fig. S2). NPNT, and its potential downstream effects, is therefore intriguing with regards to metastasis. More recently, we reported the involvement of the NPNT integrin-binding motifs in enhancing BC lung metastasis in mice 19 . NPNT was also found to promote anchorage independent growth and survival, providing cancer cells with a growth benefit 19,34 . These results could indicate a functional role for NPNT in metastasis. Other ECM proteins, such as tenacin C (TNC), have also been reported to play an important role in the early steps of metastatic seeding of the lungs. However, TNC had no effect on the seeding of BC cells in the brain, suggesting that other ECM proteins could have a role in brain metastasis 35 . www.nature.com/scientificreports/ We postulated a role for NPNT in establishing BC brain metastasis, an organ that shows selectivity towards metastatic cells, and where as much as 30% of all metastases originate from BC . With regards to Paget's "seed and soil" hypothesis , the brain may therefore be a very particular kind of soil. In our study, we found that 21 out of 22 patients had up-regulated expression of NPNT mRNA in primary human BC compared to pair-matched non-cancerous breast tissue. In addition, high levels of NPNT transcript were linked to poor prognosis for the luminal B subtype, and both NPNT and integrin α8β1 were found to be expressed in human BC brain metastases. When wild-type NPNT was overexpressed in mouse BC cells, an increased number of metastatic lesions was observed in the brain when cells were injected into the carotid artery (Fig. 4). This is in line with our previous results showing that NPNT enhances metastasis to the lungs via its integrin-binding motifs 19 . Mutating the integrin-binding motifs of NPNT stongly reduced the cell's ability to establish metastatic lesions in the brain (Fig. 4), and lungs 19 . Although in vivo studies resulted in the same tendencies as previously published 19 , we now observed that the double mutant (66cl4-RGE-AIA) established fewer metastatic lesions than the control cell line (66cl4-EV). This finding was unexpected. This could reflect a difference in methodology between the two studies, but could also be a tissue specific finding. Though more work would be needed, it is tempting to speculate the the double mutant somehow sequesters cell surface receptors and functions as an inhibitor of brain metastasis. The fact that NPNT only resulted in a prognostic value for the luminal B subtype, although up-regulated in most breast cancers, could also indicate that the combination of the molecular profile of the BC cells and NPNT play a part in the metastatic process. In the current study, we additionally show that NPNT is involved in adhesion to and migration through the brain endothelial cells (Fig. 5e, g). These results are intriguing as they indicate that a cell surface localisation of NPNT is important during the metastatic process. NPNT contains five EGFlike domains, a linker region with two integrin-binding motifs, and a MAM-domain 9,10 . Three of the EGF-like repeats are calcium-binding, and such EGF-repeats are known to be involved in protein-protein interactions 36 . The EGF-like repeats of NPNT are also reported to bind chondroitin sulfate E (CS-E), which is expressed in brain, kidney, cartilage and hair follicles 37 . The EGF-like repeats may also possibly bind EGFR 15,17 . The NPNT linker region can potentially bind all RGD-recognising integrins, where NPNT is reported to bind integrins αVβ3, αVβ5, αVβ6, α4β7, α5β1 and α8β1 10 . MAM-domains are evolutionarily conserved and thought to convey an adhesive function 38 . The MAM domain of NPNT was reported to bind heparan sulphate proteoglycans (HSPGs)/heparin 37 and the basement membrane proteins QBRICK, Fras1 and Frem2 39 . The MAM domain is also postulated to be involved in NPNT dimerization and tetramerization 11 . Dimers or tetramers of NPNT might bind several different receptors at once, offering a potential mechanism of action. Further studies are needed to reveal the exact receptors involved in NPNT-binding in brain endothelial cells. Interestingly, both NPNT and α8β1 were expressed in the human BC brain metastases in a similar pattern and intensity, indicating that these proteins might be employed also by human BC cells to enhance their ability to grow in the brain microenvironment. Others have reported that expression of α8β1 in BC cells increases their migration 40 , and expression of NPNT increases metastasis 18 .
Exosomes are secreted small extracellular vesicles that can facilitate bidirectional cell-cell communication, but are also implicated in creating a pre-metastatic niche. Cargo from BC-derived exosomes has been found in lungs, liver, bone and brain, whilst colorectal cancer-derived exosomes primarly home to the liver 41 . Interestingly, in our recent publications we show that 66cl4 cells concentrates truncated versions of NPNT into exosomes 19,42 , which could potentially create a favourable microenvironment in the brain during metastasis.
While exosomes could be important in NPNT transmission and thus metastasis to the brain, the signalling casades introduced by NPNT in the brain are largely unknown. When brain endothelial cells were seeded on rmNPNT, phosphorylation of ERK1/2 was strongly up-regulated and phosphorylation of AMPK1α was strongly down-regulated (Fig. 5b). AMPKs are guardians of cellular energy and are "switched on" when cells are under metabolic stress 43 or when released from ECM anchorage, which leads to increased autophagy through regulation of mammalian target of rapamycin complex 1 (mTORC1) 44 . Based on our experiments, NPNT seems to counteract this metabolic stress signal by down-regulating the phosphorylation of AMPKα1. When phosphorylated, active AMPK inhibits cell growth and proliferation and promotes cell polarity through phosphorylation of p53, and pharmacological activation of AMPK inhibits cancer cell growth. AMPK is therefore suggested to be an "energy checkpoint" that delays progress through the cell cycle if energy is low 43 . Active AMPK is also involved in maintaining cell polarity, especially in epithelial cells 43 . Metformin is a widely used pharmaceutical was analysed by real-time cell analysis using the xCELLigence system. The experiments were performed at least twice with similar results and with two technical replicates per run. Impedance created by adhering cells gave the arbitrary "cell index" value that is proportional to the amount of adhered cells. Adhesion was recorded every 15 min for 12 h. Wells were coated with either 3% BSA as a negative control or 10 μg purified wild-type rmNPNT (Wild-type) or rmNPNT mutated in the integrin-binding sites (Mutant). Error bars show the pooled standard deviation. (d) Wells were coated with 10 μg purified wild-type rmNPNT. Adhesion of 66cl4-EV cells in the presence of RGD-blocking peptide or scrambled negative control peptide. p = 0.019. (e) 66cl4 cells were seeded onto a confluent monolayer of MBECs. After three hours, non-adherent cells were washed off and adherent cells were counted. Fold differences between the cells is shown in the graph (N = 2, n = 3). p = 0.013. (f) 66cl4 cells were seeded onto a confluent monolayer of bEND.3 cells. After three hours, non-adherent cells were washed off and adherent cells were counted. Fold differences between the cells is shown in the graph (N = 3, n = 3). p = NS. (g) 66cl4 cells were seeded onto a confluent monolayer of MBECs. Transmigration of 66cl4 cells was followed by imaging every 5 min for 24 h. Every cell that transmigrated through the monolayer was recorded and fold differences between the cells is presented in the graph (N = 3, n = 3), p = 0.044. Scientific RepoRtS | (2020) 10:12237 | https://doi.org/10.1038/s41598-020-69242-1 www.nature.com/scientificreports/ for diabetes type 2 and it activates AMPK. Interestingly, an epidemiological study on metformin use showed a significant reduction in the incidence of different types of cancer 45 .

conclusions
In conclusion, we demonstrate an involvement of NPNT in promoting brain metastasis. Our analyses show that NPNT is overexpressed in primary BC, and also present in BC brain metastasis, where it is located in the same areas as its receptor, integrin α8β1. Furthermore, NPNT is also linked to poor prognosis for the luminal B subtype. We also show that the integrin-binding motifs of NPNT are important for BC cell adhesion to-and migration through brain endothelial cells. Additionally, NPNT triggered several intracellular signalling pathways in the endothelial cells known to be involved in proliferation, differentiation, growth and development. Mouse BC cells overexpressing NPNT showed increased ability to establish brain metastases in vivo, an ability that was lost when the integrin-binding motifs were mutated. This indicates an important role of the integrin-binding motifs in establishing brain metastasis with possible implications for prognosis. Breast cancer brain metastasis is an increasing clinical challenge that needs to be met, where NPNT and its binding partners could represent novel drug targets, inhibiting brain metastasis in high risk patients before it even occurs. Reverse transcriptase quantitative PCR (RT-qPCR). RT-qPCR was performed as previously Experimental brain metastasis model. BALB/c mice were bred in-house and eight-ten week old females were selected for intracarotid injections, as previously described 48 . Experimental brain metastases were established by injecting 66cl4 (1.0 × 10 6 ) cells into the right carotid artery of 8-10 week old female BALB/c mice (N = 5/group). Control mice were either subjected to the same procedure, but with only buffer injected (N = 3), or were not injected (N = 2). On day 7, mice were anaesthesized and transcardially perfused with PBS followed by 4% paraformaldehyde (PFA). Brains were removed and additionally immersion fixed in PFA overnight at 4 °C. Further details of the procedure can be found in Supplementary file 1.  Primary mouse brain endothelial cells (MBECs) for adhesion and transmigration experiments. Primary MBECs were isolated from female BALB/c mice (bred in house) according to optimized and established protocols as previously described 49,50 . A detailed isolation procedure is described in Supplementary file 1. Adhesion and transmigration experiments were performed as previously described 30 . In short, isolated MBECs were cultured to confluence. The 66cl4 cells were seeded onto the MBEC monolayer and either incubated 3 h for adhesion or 24 h for transmigration experiments. The experiment was repeated three times with three replicate wells per cell line.