EPCR promotes breast cancer progression by altering SPOCK1/testican 1-mediated 3D growth
- 1.3k Downloads
Activated protein C/endothelial protein C receptor (APC/EPCR) axis is physiologically involved in anticoagulant and cytoprotective activities in endothelial cells. Emerging evidence indicates that EPCR also plays a role in breast stemness and human tumorigenesis. Yet, its contribution to breast cancer progression and metastasis has not been elucidated.
Transcriptomic status of EPCR was examined in a cohort of 286 breast cancer patients. Cell growth kinetics was evaluated in control and EPCR and SPARC/osteonectin, Cwcv, and kazal-like domains proteoglycan (SPOCK1/testican 1) silenced breast cancer cells in 2D, 3D, and in co-culture conditions. Orthotopic tumor growth and lung and osseous metastases were evaluated in several human and murine xenograft breast cancer models. Tumor-stroma interactions were further studied in vivo by immunohistochemistry and flow cytometry. An EPCR-induced gene signature was identified by microarray analysis.
Analysis of a cohort of breast cancer patients revealed an association of high EPCR levels with adverse clinical outcome. Interestingly, EPCR knockdown did not affect cell growth kinetics in 2D but significantly reduced cell growth in 3D cultures. Using several human and murine xenograft breast cancer models, we showed that EPCR silencing reduced primary tumor growth and secondary outgrowths at metastatic sites, including the skeleton and the lungs. Interestingly, these effects were independent of APC ligand stimulation in vitro and in vivo. Transcriptomic analysis of EPCR-silenced tumors unveiled an effect mediated by matricellular secreted proteoglycan SPOCK1/testican 1. Interestingly, SPOCK1 silencing suppressed in vitro 3D growth. Moreover, SPOCK1 ablation severely decreased orthotopic tumor growth and reduced bone metastatic osteolytic tumors. High SPOCK1 levels were also associated with poor clinical outcome in a subset breast cancer patients. Our results suggest that EPCR through SPOCK1 confers a cell growth advantage in 3D promoting breast tumorigenesis and metastasis.
EPCR represents a clinically relevant factor associated with poor outcome and a novel vulnerability to develop combination therapies for breast cancer patients.
KeywordsMatricellular Metastasis Microenvironment Sphere cultures Extracellular matrix
Activated protein C/endothelial protein C receptor
Mammary stem cells
Protease-activated receptor 1
Erythrocyte membrane protein 1
Polyoma middle T
SPARC/osteonectin, Cwcv and kazal-like domains proteoglycan
Endothelial protein C receptor (EPCR) is an endothelial type 1 transmembrane receptor that enhances the activation of protein C (PC) by the thrombin (IIa)-thrombomodulin (TM) complex . EPCR-dissociated activated protein C (APC) negatively regulates the coagulation process, while EPCR-bound APC induces cytoprotective signaling through the proteolytic cleavage of protease-activated receptor 1 (PAR1), leading to anti-inflammatory and anti-apoptotic responses .
Recently, research in EPCR has gained considerable momentum by the identification of new EPCR ligands . An EPCR domain distinct from the APC binding site was shown to interact with a specific T cell antigen receptor with potential implications in immunosurveillance of tumors . EPCR was also identified as the endothelial receptor for some subtypes of the erythrocyte membrane protein 1 (PfEMP1) on the surface of the parasite Plasmodium falciparum, mediating its sequestration in the blood vessels during severe malaria . FVII/FVIIa has been shown to bind EPCR with a similar affinity as PC/APC , whereas the binding of FX/FXa to EPCR remains an open question .
Recently, EPCR has been identified as a marker of multipotent mouse mammary stem cells (MaSCs). These EPCR+ cells (accounting for 3–7% of basal cells) exhibited a mesenchymal phenotype and enhanced colony-forming abilities . EPCR was also shown to be necessary for cell organization and growth of human mammary epithelial cells in 3D cultures .
In cancer, aberrant expression of EPCR is detected in tumors of different origin including the lung , breast , ovarian , colon , glioblastoma , mesothelioma , and leukemia . In lung tumorigenesis, APC/EPCR drives an anti-apoptotic program that endows cancer cells with increased survival ability, enhancing their metastatic activity to the skeleton and adrenal glands . Moreover, high expression levels of this single gene at the primary site in early stage lung cancer patients predict the risk of adverse clinical progression [9, 15].
In breast cancer patients, tumor cells often disseminate to target sites including the skeleton, lungs, brain, and lymph nodes . This event represents a frequent complication associated with a 5-year survival rate ~25.9%. Recent findings have unveiled novel markers in the primary tumor that predict the development of metastasis to target organs such as the skeleton . High EPCR levels have been associated with poor disease progression in the polyoma middle T (PyMT) breast cancer model, closely similar to the luminal B type in humans . Moreover, EPCR+ sorted MDA-MB-231 human breast cancer cells showed stem cell-like properties and enhanced tumor-initiating activity, an effect inhibited by APC-EPCR blocking antibodies . In contrast, overexpression of EPCR in MDA-MB-231 cells resulted in reduced final tumor volumes in a xenograft model despite favoring tumor growth at initial stages . The effect of EPCR at different stages of tumor progression remains poorly defined.
In this study, we addressed the functional role of EPCR in primary and metastatic tumor growth in breast cancer using several human and murine xenograft models. We found that EPCR silencing impaired orthotopic tumor growth and metastatic activity to the skeleton and lungs. Moreover, high EPCR expression levels associated with a poor clinical outcome in a cohort of breast cancer patients. Furthermore, we showed that EPCR effects in tumor progression were APC independent and were partially mediated by a novel mechanism involving SPOCK1. Thus, these findings unveil a novel mechanism mediated by EPCR in tumorigenesis and metastasis of breast cancer with potential clinical impact on the therapeutic management of breast cancer patients.
Cell lines and reagents
One thousand eight hundred thirty-three human breast cancer cell line was a kind gift from Dr. Massagué (Memorial Sloan-Kettering Cancer Center, NY, USA) . ANV5 murine breast cancer cell line was previously described [21, 22]. APC (Xigris®) was purchased from Eli Lilly (Indianapolis, IN, USA). Anti-EPCR antibodies RCR252 and RCR1 were kindly provided by Dr. Fukudome (Saga Medical School, Japan) while 1489 was kindly gifted by Dr. Esmon (Oklahoma Medical Research Foundation, Oklahoma City, USA). F(ab´)2 fractions of the RCR252 antibody were obtained as previously detailed . shRNAs cloned into PLKO.1-puro vector and the empty vector were obtained from Mission® (Sigma-Aldrich).
Cell proliferation assay
Cell proliferation was assessed using CellTiter 96® AQueous One Solution Cell Proliferation Assay (MTS), according to manufacturer’s recommendations (Promega). All absorbance values were normalized with the absorbance values from day 0 (5 h after seeding cells).
Cell cycle analysis
Cell cycle analysis was carried out with Click-iT® EdU Flow Cytometry Assay Kit (Invitrogen). Cells were maintained in culture for 24 or 48 h before adding 10 μM EdU for 2 h. Next, cells were harvested, fixed in formaldehyde (Click-iT® fixative), permeabilized in 1X Click-iT® saponin-based permeabilization and wash reagent, and incubated with the Click-iT® reaction cocktail for 30 min at room temperature in the dark. After a washing step, cells were incubated with 0.2 μg/μl RNase A (Sigma-Aldrich) for 1 h at room temperature, in the dark. 7AAD was added to the tubes 10 min before the acquisition of cells in a FACSCanto II cytometer (BD Biosciences). Data were analyzed using FlowJo® software v9.3.
Annexin-V flow cytometry assay
Cells were seeded into 24-well plates and cultured for 24 h. Next, cells were incubated with 2 μM staurosporine for 1 h or serum-starved overnight before the addition of 50 nM APC for 4 h followed by 2 μM staurosporine for 1 h next day. After staurosporine treatment, cells were harvested, resuspended in annexin-binding buffer (10 mM HEPES, 140 mM NaCl, and 2.5 mM CaCl2, pH 7.4) and incubated with Alexa Fluor 647-conjugated annexin-V and 7AAD (BD Biosciences) for 15 min at room temperature, in the dark. Cells were acquired in a FACSCanto II cytometer (BD Biosciences) and analyzed using FlowJo® software v9.3.
Cell culture in 3D
Culture media was mixed at 1:1 ratio with Growth Factor Reduced Matrigel (BD Biosciences). One hundred microliters of the mix were added to each well of a 96-well plate and incubated at 37 °C for 30 min. Five hundred (1833, BT-549, ANV5, MCF10A) or 1000 (MDA-MB-231) cells in medium with 10% matrigel were added on top of the coating and maintained in culture for 8–10 days. Medium with 10% matrigel was replaced at day 4–5. Pictures of the spheres were taken at day 8–10 at ×4 magnification using an inverted microscope (Leica) and analyzed using Fiji software .
In vivo experiments
Athymic nude mice (Foxn1 nu ) were purchased from Harlan (Barcelona, Spain) and maintained under specific pathogen-free conditions. Five- to six-week-old mice were used for all experiments. RAG-2−/− mice were bred at the in-house Animal Core Facility and used for the intratibial experiment. For the orthotopic injection, 50 μl containing 500,000 cells resuspended in Growth Factor Reduced Matrigel (BD Biosciences) mixed with PBS at 1:1 ratio were directly injected into the fourth mammary fat pads of mice (2 tumors per mouse). In the second orthotopic experiment, cells were injected resuspended in 20 μl of PBS without matrigel. Tumor growth was monitored regularly using a digital caliper and tumor volume was calculated as follows: π × length × width2/6. For intracardiac injection, 105 cells in 100 μl of PBS were inoculated into the left cardiac ventricle, using a 29G needle syringe . For intratibial injection, 15,000 cells in 5 μl of PBS were injected into the tibia’s bone marrow through the femoro-tibial cartilage using a Hamilton syringe . For intravenous injection, 100,000 cells in 100 μl of PBS were injected through the tail vein of mice. For BLI, animals were anesthetized and inoculated with 50 μl of 15 mg/ml D-luciferin (Promega). Images were taken during 1 min with a PhotonIMAGER™ imaging system (Biospace Lab) and analyzed using M3Vision software (Biospace Lab). Photon flux was calculated by using a region of interest (ROI) or by delineating the mouse for whole-body bioluminescence quantification. All bioluminescence signals were normalized with values from day 0, except for the metastasis experiment with RCR252 treatment. Radiographic and micro-computed tomography (Micro-CT) analyses were performed as described elsewhere .
RNA was extracted from snap-frozen mammary tumors and hybridized to Human Gene ST 2.0 microarrays (Affymetrix). Data were normalized with RMA (Robust Multi-Array Average) approach. Low expression probes were removed by filtering those that did not exceed a level of expression of 32 in at least one of the samples for each condition. Differentially expressed genes were identified using LIMMA (linear models for microarray data) method .
Statistical analysis was performed using SPSS v15.0. When data exhibited homoscedasticity, pairwise Student’s t test and Mann–Whitney U test were used for normally and non-normally distributed variables, respectively. When data exhibited heterocedasticity, Welch and Median tests were used for normally and non-normally distributed variables, respectively. ANOVA and posterior Bonferroni tests were used for multiple comparisons of normally distributed variables. Kruskal–Wallis and posterior Bonferroni adjusted-Mann–Whitney U tests were used for multiple comparisons of non-normally distributed variables. Statistical significance was defined as significant (p < 0.05, *), very significant (p <0.01, **) and highly significant (p < 0.001, ***). Other additional methods are included in the Additional file 1.
High EPCR expression in breast tumors correlates with poor clinical outcome
EPCR silencing impairs breast tumorigenesis and spontaneous metastases
To explore the relevance of these findings in vivo, we performed an orthotopic experiment using the highly metastatic subpopulation 1833, as outlined in Fig. 2f. Remarkably, EPCR silencing significantly reduced primary tumor growth after the injection of shControl, shEPCR#1, or shEPCR#2 1833 cells into the fourth mammary fat pads of athymic nude mice, in two independent experiments (Fig. 2g). Consistently, time until resection of tumors at 300 mm3 was significantly longer for EPCR-silenced groups, showing that control tumors maintained higher proliferation rates over the course of the experiment (Fig. 2h). Of note, several tumors in EPCR-silenced groups did not reach the size established for tumor resection by the end of the experimental period (Fig. 2h). In addition, BLI performed after tumor resection showed that the number of mice with metastasis and the number of metastatic events were lower in EPCR-silenced groups (Fig. 2i). Importantly, EPCR inhibition was confirmed by immunohistochemistry in resected primary tumors (Fig. 2j). Similarly, in another xenograft model of murine ANV5 cells, EPCR silencing reduced primary tumor growth after orthotopic injection of shControl and shEPCR#3 cells into athymic nude mice. Evaluation of spontaneous metastases in this model was limited by the highly frequent local recurrence after tumor resection (Additional file 3: Figure S2F, G).
Analysis of tumors in the orthotopic model of 1833 cells, either size-matched tumors resected at different time points (Additional file 4: Figure S3) or tumors of different size resected at the same time point (Additional file 5: Figure S4) revealed a slight increase in apoptosis (cleaved caspase-3) and/or necrosis and a lower proliferation rate, assessed by Ki67 staining, in EPCR-silenced tumors. Of note, we did not observe relevant changes in angiogenesis and immune infiltration patterns of tumors (Additional file 5: Figure S4 and Additional file 6: Figure S5). Taken together, these data indicate that EPCR contributes to primary tumor growth and the development of spontaneous metastases in breast cancer.
EPCR silencing reduces metastasis to the bone and lungs
To further explore the function of EPCR in bone colonization, shControl, shEPCR#1, and shEPCR#2 1833 cells were injected into the tibiae of immunocompromised mice. Bone colonization was analyzed by BLI, X-rays, and histological analysis (Fig. 3h). Tumors developed in all tibiae in shControl and shEPCR#2 mice, while two tibiae remained tumor-free in shEPCR#1 group. Differences in BLI became very relevant at advanced experimental time points (Fig. 3i, l). In addition, histological and X-ray analyses revealed reduced tumor burden and osteolytic bone areas in EPCR-knockdown groups at the end of the experiment (Fig. 3j–l). Of note, the magnitude of the effect of EPCR silencing revealed by different techniques (BLI vs. histology and X-rays) differs, an event probably related to the fact that X-ray analysis does not detect extraosseous tumor grown through the cortical bone on the periosteal surface (Fig. 3l, bottom panel). Similarly, extracortical tumor cells that contribute to tumor burden were lost during histological processing, whereas these cells contribute to BLI. Thus, these results indicate that EPCR promotes metastatic tumor growth in the osseous compartment.
Next, we evaluated the prometastatic activity of EPCR in an intra-tail injection model. For this purpose, we injected shControl and shEPCR#3 ANV5 cells intravenously into athymic nude mice and analyzed lung metastases at the end of the experiment (Additional file 7: Figure S6A). EPCR knockdown was able to block the development of lung metastases, assessed by BLI (Additional file 7: Fig. S6B, D) and tumor area quantification in H&E-stained lung sections (Additional file 7: Figure S6C, D).
APC does not mediate effects of EPCR silencing in vitro and in vivo
Next, we explored the mechanistic insights of EPCR function in tumor growth and metastasis. First, we analyzed whether the main known ligand of EPCR, APC, could signal and mediate cellular functions to favor tumor progression in MDA-MB-231, 1833, BT-549, and ANV5 cells. Stimulation of cells with APC did not affect their proliferation, cell cycle progression, and resistance to basal and induced apoptosis (Additional file 8: Figure S7). Accordingly, treatment with the F(ab)2´fraction of RCR252 antibody, which blocks APC binding to human EPCR (Additional file 9: Figure S8A), did not reduce bone metastasis of 1833 cells inoculated into the left cardiac ventricle of athymic nude mice (Additional file 9: Figure S8B–F).
Identification of SPOCK1 as a mediator of EPCR effects
Next, we tested the effects of SPOCK1 in vitro, by silencing SPOCK1 expression levels with shRNAs in MDA-MB-231, 1833, and BT549 human cell lines (Additional file 11: Figure S10A). Interestingly, SPOCK1 silencing did not affect cell growth kinetics in 2D cultures (Additional file 11: Figure S10B, C), but significantly reduced the number of spheres in 3D matrigel cultures in all cell lines (Fig. 4e). Conversely, ectopic expression of EPCR and SPOCK1 in non-tumorigenic MCF10A mammary cells significantly increased the number of spheres in 3D cultures (Fig. 4f). These data indicate that EPCR or SPOCK1 overexpression confers a growth advantage in 3D cultures in a non-tumorigenic mammary cell line, but per se EPCR or SPOCK1 are not sufficient to confer a tumorigenic phenotype requiring an oncogenic background. Taken together, these findings support the role of SPOCK1 mediating EPCR effects and suggest that EPCR could promote 3D growth of breast cancer cells by altering tumor-matrix interactions by modulating SPOCK1.
SPOCK1 silencing impairs breast tumorigenesis and metastases
Finally, we evaluated the bone metastatic activity of control and SPOCK1-silenced 1833 cells after intracardiac inoculation into athymic nude mice (Fig. 5e). All mice in the shControl and shSPOCK1#1 groups developed bone metastases, but only 3 mice in the shSPOCK1#2 group. Moreover, the number of bones with metastases was significantly lower in both SPOCK1-silenced groups compared to shControl (Fig. 5f). Consistently, BLI and H&E staining revealed a lower tumor burden in SPOCK1-silenced groups, associated with a lower osteolytic area (Fig. 5g–j). These results indicate that SPOCK1 silencing recapitulates the effects observed by EPCR silencing in vivo and further support the role of SPOCK1 as an effector of EPCR.
In this work, we unveiled a novel molecular mechanism of EPCR contributing to breast cancer progression favoring tumor growth and metastatic activity to target organs. EPCR endowed cells with advantageous growth in 3D, an effect partially mediated by the extracellular matrix proteoglycan SPOCK1. These cell functions were correlated with increased metastatic risk and poor clinical outcome in breast cancer patients. Importantly, this association was relevant in all the molecular subtypes, except luminal A, indicating that EPCR could be useful as a potential prognostic marker in these patient subsets.
Previous studies identified EPCR as a marker of human breast cancer stem cells with enhanced tumor-initiating and growth abilities in immunodeficient mice . In addition, EPCR deficiency attenuated spontaneous tumor growth in the PyMT murine breast cancer model . In agreement with these findings, we showed that EPCR silencing impaired orthotopic tumor growth of highly metastatic 1833 cells. In this model, differences in tumor size between EPCR+ and EPCR− tumors became more relevant at advanced experimental time points. In contrast in another study, although EPCR overexpression increased initial orthotopic growth of MDA-MB-231 cells, it resulted in smaller final tumor volumes , a finding possibly related to EPCR loss in evolving tumors. Thus, EPCR could display different roles at different stages of breast cancer progression such as initiation, maintenance, and target organ colonization. Future experiments will help to characterize its role in each of these stages in different histological subtypes.
Besides its role in tumorigenesis, EPCR also displayed a marked prometastatic activity to target organs, events that cooperatively support its contribution to prognosis. The consistent results obtained in both metastatic models indicate that EPCR confers a functional advantage at late stages of the metastatic process. Moreover, differences in metastatic tumor burden became more relevant at advanced experimental time points, indicating an effect more pronounced during the colonization of target organs, as evidenced by the overt osseous colonization observed in the intratibial model.
In contrast with previous findings in lung cancer , EPCR did not markedly contribute to tumor cell survival in the circulation and engraftment in secondary sites. The prominent effect in breast cancer during colonization was associated with its role in 3D growth and based on the low number of tumor nodules in shEPCR mice in both models (the bone and lung) of experimental metastasis, EPCR may also modulate metastatic tumor re-initiation at the target organ.
Tumors are organ-like structures composed of tumor cells and stromal cells embedded in a complex ECM within the tumor microenvironment . Components of the ECM such as tenascin C have been shown to promote breast cancer progression and metastasis [30, 31, 32]. In the same line, our study identified SPOCK1, a secreted matricellular protein as a markedly downregulated gene in EPCR-silenced tumors. SPOCK1 belongs to the Ca2+-binding proteoglycan family which includes SPARC, a well-studied tumor-associated component involved in regulating adhesion, matrix cellular interactions, and cell growth [33, 34]. Recently, SPOCK1 has been shown to promote epithelial-mesenchymal transition (EMT) and metastasis in other tumors including lung and gallbladder cancer and hepatocarcinoma [35, 36, 37]. Interestingly, high SPOCK1 expression levels were associated with adverse clinical outcome in the same subsets of breast cancer patients predicted by EPCR levels. Therefore, EPCR could promote tumor growth in vivo, in part, by modulating tumor-matrix interactions through SPOCK1 favoring an advantageous 3D growth of tumorigenic cells. Indeed, SPOCK1 silencing in breast cancer cells impaired the number of 3D spheres and primary and metastatic tumor growth, an effect that phenocopied EPCR silencing. Accordingly, EPCR has been required for cell organization and growth of mammary epithelial cells in 3D cultures  In agreement with these findings, EPCR/SPOCK1 axis activation in non-tumorigenic mammalian cells increased the number of spheres grown in 3D matrigel cultures. However, it was not sufficient to confer a tumorigenic phenotype.
A surprising finding of our study was the lack of effects mediated by APC, despite the fact that anti-EPCR blocking antibodies (1535) reduced orthotopic growth of MDA-MB-231 cells in previous studies . Although, we did not specifically address APC/EPCR effects in orthotopic tumors, we explored its contribution in vitro and during the development of bone metastases. EPCR-blocking antibodies in this model could not reduce the metastatic activity of 1833 cells, suggesting that EPCR triggered APC-independent effects. In this experiment, we used the F(ab´)2 fractions of the anti-EPCR blocking antibody to avoid any interference of the activated complement system, whereas Schaffner et al.  employed whole-body antibodies. Furthermore, the use of the same strategy of F(ab´)2 fractions showed a significant effect on a model of lung cancer metastasis underscoring the validity of this approach . Complementary to this view, other ligands different than APC binding to different regions of EPCR in each tumor type or accessible in specific microenvironments could account for these differences. Based on these findings, future experiments should address other mechanisms that could be mediated by EPCR in different tumor types and metastatic sites.
In summary, our study unveils a novel role of EPCR as a clinically relevant factor in breast cancer, which promotes primary tumor growth and metastatic activities in target organs. Unexpectedly, EPCR modulates tumor cell-ECM interactions involved in 3D growth required for tumor progression and metastasis, in part by upregulating SPOCK1. These findings underscore a novel role of EPCR as a novel prognostic factor and a potential therapeutic target in a subset of breast cancer patients.
We are grateful to Carmen Berasain for helpful discussions. We thank members of the Morphology, Genomics and Bioinformatics and Animal Core Facilities for their helpful assistance, especially, L. Guembe, D. Corbacho, and M. Ariz. We thank Dr. Esmon and Dr. Fukudome for anti-EPCR antibodies.
N.P. holds a FPU fellowship from the Spanish Ministry of Education. This work was supported by “UTE project” from the Foundation for Applied Medical Research and a grant from the Spanish Ministry of Economy and Competitiveness to F.L. (SAF2012-40056).
Availability of data and materials
The datasets supporting the conclusions of this article are included within the article and its additional files. Microarray datasets are included in https://www.unav.es/archivos-dianas-moleculares.
FL and NP designed the study, discussed the results, and wrote the manuscript. NP, CZ, SMC, and HM performed in vitro and in vivo experiments and discussed the results. RRG and MS provided some reagents and discussed the results. ME and ASA directed some critical experiments and discussed the results. SV discussed the results and provided expertise for some experiments. EG interpreted and performed expression arrays and bioinformatics analysis. JH performed critical experiments, provided reagents, and discussed the results. FL coordinated all tasks and edited the manuscript. All authors participated in drafting the article and discussing the results. All authors read and approved the final manuscript.
Authors declare that there are no competing interests.
Consent for publication
All animal procedures (protocol 161–14) were approved by the CEEA (Ethical Committee for Animal Experimental Research).
- 17.Pavlovic M, Arnal-Estape A, Rojo F, et al. Enhanced MAF oncogene expression and breast cancer bone metastasis. J Natl Cancer Inst. 2015;107(12):djv256. doi: 10.1093/jnci/djv256.
- 27.Patino-Garcia A, Zalacain M, Folio C, et al. Profiling of chemonaive osteosarcoma and paired-normal cells identifies EBF2 as a mediator of osteoprotegerin inhibition to tumor necrosis factor-related apoptosis-inducing ligand-induced apoptosis. Clin Cancer Res. 2009;15:5082–91.CrossRefPubMedGoogle Scholar
- 36.Shu YJ, Weng H, Ye YY, et al. SPOCK1 as a potential cancer prognostic marker promotes the proliferation and metastasis of gallbladder cancer cells by activating the PI3K/AKT pathway. Molecular cancer 2015; 14: 12-014-0276-y.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.