Deletion of Cd151 reduces mammary tumorigenesis in the MMTV/PyMT mouse model
Tetraspanins are transmembrane proteins that serve as scaffolds for multiprotein complexes containing, for example, integrins, growth factor receptors and matrix metalloproteases, and modify their functions in cell adhesion, migration and transmembrane signaling. CD151 is part of the tetraspanin family and it forms tight complexes with β1 and β4 integrins, both of which have been shown to be required for tumorigenesis and/or metastasis in transgenic mouse models of breast cancer. High levels of the tetraspanin CD151 have been linked to poor patient outcome in several human cancers including breast cancer. In addition, CD151 has been implicated as a promoter of tumor angiogenesis and metastasis in various model systems.
Here we investigated the effect of Cd151 deletion on mammary tumorigenesis by crossing Cd151-deficient mice with a spontaneously metastasising transgenic model of breast cancer induced by the polyoma middle T antigen (PyMT) driven by the murine mammary tumor virus promoter (MMTV).
Cd151 deletion did not affect the normal development and differentiation of the mammary gland. While there was a trend towards delayed tumor onset in Cd151−/− PyMT mice compared to Cd151+/+ PyMT littermate controls, this result was only approaching significance (Log-rank test P-value =0.0536). Interestingly, Cd151 deletion resulted in significantly reduced numbers and size of primary tumors but did not appear to affect the number or size of metastases in the MMTV/PyMT mice. Intriguingly, no differences in the expression of markers of cell proliferation, apoptosis and blood vessel density was observed in the primary tumors.
The findings from this study provide additional evidence that CD151 acts to enhance tumor formation initiated by a range of oncogenes and strongly support its relevance as a potential therapeutic target to delay breast cancer progression.
KeywordsTetraspanin CD151 Breast Cancer Metastasis
Neutral buffered formalin
Polyoma middle T antigen
Mouse mammary tumor virus.
Breast cancer is the most commonly diagnosed cancer among women and despite some major advances in diagnosis and treatment, it remains the second leading cause of cancer death in women worldwide. Similarly to other cancers, some of the major challenges in the treatment of breast cancer reside in the lack of response or development of resistance to existing therapies and the devastating consequences of metastasis. Better prognostic markers as well as new targeted treatments that could be used alone, or most likely in combination with existing therapies are needed to improve patient outcomes.
The tetraspanin CD151 is part of the tetraspanin family of transmembrane proteins, which consists of 33 members in humans. These proteins serve as scaffolds for multiprotein complexes (called TEMs or Tetraspanin-Enriched Microdomains) where they associate with molecules such as integrins, growth factor receptors and matrix metalloproteases, modifying their functions in various cellular processes. CD151 forms tight complexes with the laminin binding integrins (α3β1, α6β1 and α6β4) , modulates their signaling and contributes to integrin mediated cell adhesion and motility.
Interestingly, β1-integrin and β4-integrin- heterodimers (reviewed in ) are both expressed by breast epithelial cells (mostly as α3β1 and α6β4) and have been shown to be required for tumorigenesis and metastasis in the MMTV/PyMT and MMTV/Neu (rat homolog of ErbB2) mouse models of breast cancer [3, 4], respectively. CD151 expression itself has been associated with poor patient outcome in several malignancies, including cancers of the breast [5, 6], prostate , lung , and kidney , whereas it has also been found to correlate with improved survival in endometrial cancer . Notably in breast cancer, elevated expression of CD151 correlates with lymph node invasion and poor overall survival of patients with invasive ductal carcinoma [5, 6]. In accordance with its association with poor prognostic in patients, CD151 has been implicated as a promoter of tumor angiogenesis and/or metastasis in vitro in human breast cancer cell lines and in several in vivo model systems including xenografts [5, 11], matrigel plug and tumor implantation experiments , as well as experimental metastasis models [13, 14]. In addition, a recent study showed that CD151 plays a role in mammary cell proliferation, suggesting the involvement of CD151 in tumor cell growth .
Altogether these data strongly indicate a role for CD151 in tumor growth and metastasis, suggesting that it could be used as a target molecule for the design of new breast cancer therapies. However, when we started this work, the possible direct cause-effect relationship between CD151 expression and breast tumor onset/progression and metastasis had never been tested. In order to address this question, we studied the effect of Cd151 deletion on de novo breast tumorigenesis and spontaneous metastasis in the very well characterized MMTV/PyMT transgenic breast cancer mouse model . In this model, the polyoma middle T oncogene is expressed under the transcriptional control of the mouse mammary tumor virus promoter. The mammary tumors that develop in MMTV/PyMT female mice recapitulate the histological stages of human breast cancer from premalignant lesions to invasive carcinoma  and they also display activation of the same signaling pathways that act downstream of the ErbB2 oncogene and are often activated in breast cancer, such as c-Src, PI3K and Ras .
It is interesting to note that a study addressing the impact of Cd151 deletion in another mouse model of breast cancer (the ErbB2 model) has recently been published by Deng and colleagues . The results from both studies will be compared in the discussion.
Here we show that Cd151–null mice develop smaller and fewer PyMT mammary tumors than their age matched controls. Our results suggest that Cd151 deletion impairs tumor initiation and/or tumor growth in the MMTV/PyMT model, while an apparent effect on tumor metastasis could be attributable to larger tumor burden in Cd151 +/+ mice.
Animal maintenance was in accordance with the Animal Care and Ethics Committee at the Australian BioResources specific pathogen free (SPF) animal breeding facility (Moss Vale, New South Wales). All animal monitoring and experiments were approved by the Animal Care and Ethics Committee at the University of Newcastle. In the tumorigenesis experiments, we used the well-characterized FVB/N (FVB) MMTV/PyMT mouse line (MT#634) carrying a mouse mammary tumor virus promoter-driven polyoma middle T transgene . A pure FVB genetic background is most commonly used for mouse tumorigenesis experiments because of its permissiveness to spontaneous tumor development. Cd151−/− mice are grossly healthy on the C57Bl/6 (B6) background  but they develop a severe kidney disease on a pure FVB background . Hence FVB Cd151−/− mice could not be used for the tumorigenesis experiments. Since F1 hybrid FVB x B6 Cd151−/− mice were healthy and did not present any sign of kidney disease onset (monitored for the appearance of proteinuria over a 12 months period, our unpublished data), we conducted the experiments on this hybrid background. This allowed us to keep all experimental animals on a mixed but yet homogenous (50% B6 and 50% FVB) genetic background. FVB Cd151+/− mice were produced by backcross for 10 generations from the original B6 Cd151−/− and maintained as heterozygotes. Heterozygous B6 Cd151+/− females were crossed with FVB Cd151+/− males carrying the MMTV/PyMT transgene (PyMT Cd151+/− males, see Additional file 1: Figure S1 for breeding details) in order to generate the experimental F1 animals. Genotyping was performed as previously described . Experimental and control littermates were co-housed throughout the experiments, in a temperature controlled facility with a 12-h light: dark cycle.
Animal monitoring and tissue collection
Beginning at weaning (3 weeks of age), female mice were palpated twice weekly for the onset of mammary tumors. For each mouse, tumor palpation was performed in each of the ten mammary glands, in a genotype-blinded fashion. At 15–16 weeks of age, female mice were euthanased by CO2 inhalation, and all the tumors were dissected and weighed. For each experimental mouse, half of the biggest tumor was fixed in 10% neutral buffered formalin (NBF) for paraffin embedding, one quarter was snap frozen in liquid nitrogen, and the last quarter was snap frozen in OCT compound. At the time of dissection and after excision of the tumors, the lungs were exposed and inflated via tracheal injection of 1 ml of 10% NBF in order to inflate and fix the lung lobes. Lungs were then excised and further fixed in 10% NBF for at least 24 hours before paraffin embedding.
Whole mount analysis
Whole mount analysis was performed by spreading inguinal #4 mammary glands onto poly-lysine slides followed by overnight fixation in 10% NBF, defatting in acetone and overnight staining in carmine alum (0.2% carmine and 0.5% aluminium sulphate) as previously described . The stained glands were then dehydrated in a graded ethanol series, incubated in xylene for 1 hour and stored in methyl salicylate.
Tumor histology and immunohistochemistry
Tumor histology/stage was assessed on the largest tumor for each mouse using 5 μm paraffin sections stained with hematoxylin and eosin. Immunohistochemistry (IHC) was performed on 5 μm paraffin sections using a peroxidase VECTASTAIN ABC elite kit and DAB peroxidase substrate kit as per the manufacturer’s recommendations (Vector Laboratories, Burlingame, CA). The antibodies and dilutions used for IHC were rabbit monoclonal anti-Ki67 at 1:200 (Neomarkers, Kalamazoo, MI, USA) and polyclonal rabbit anti-cleaved caspase 3 at 1:800 (Cell Signaling Technology, Danvers, MA, USA).
Immunofluorescence labelings were performed on 5 μm frozen sections as previously described . Primary antibodies and dilutions used for immunofluorescence labeling were rabbit anti-CD151 (LAI-2) at 1:500 ; rat anti-CD31 (BD Biosciences, Franklin Lakes, NJ, USA) at 1:200; rabbit anti-α3 integrin (a kind gift from Dr. Fiona Watt, Wellcome Trust Centre for Stem Cell Research, Cambridge, UK) at 1:1000; rat anti-β1 integrin (BD Biosciences) at 1:200; rat anti-β4 integrin (BD Biosciences) at 1:200; rat anti- α6 integrin (Chemicon, Temecula, CA, USA) at 1:200. The double labeling where two rabbit primary antibodies were used (CD151 and α3 integrin) was performed sequentially following established methods as described previously .
Quantitation of tumor cell proliferation, apoptosis and vascularization
The largest tumor for each mouse was used to quantitate proliferation, apoptosis and vascularization in a genotype-blinded manner. Tumor cell proliferation was assessed by immunohistochemistry (as described above) for the commonly used Ki67 nuclear marker. Ki67 stained slides were scanned in a digital format using the Aperio™ digital pathology system (Aperio) and 200× magnification snapshots of digital images were generated using Imagescope. Quantitation of proliferation (expressed as percent Ki67 positive nuclear area per total nuclear area) was then performed on the 200× digital images using ImmunoRatio, a publicly available web-based application . The extent of apoptosis in the tumors was quantitated on Aperio images of cleaved-caspase 3 IHC stained slides, using the positive pixel count algorithm at 200× magnification (5 fixed size (300 μm × 300 μm) images were used for each tumor section).
To estimate blood vessel density, CD31 immunofluorescence labeling was performed and Image J was used to quantitate the proportion of CD31 positive area inside the tumors. At least five fluorescent microscope pictures (100× magnification) per tumor were used in the CD31 analysis.
Analysis of lung metastasis
All the lung lobes were dissected and processed for paraffin embedding. Five microns paraffin sections were stained with hematoxylin and eosin and slides were scanned in a digital format using the Aperio™ digital pathology system (Aperio). The lung area per section was measured using Scanscope and the metastatic burden (mm2 of metatastases/cm2 lung) was calculated for each animal, using the data from 3 sections at least 100 μm apart, as previously described .
Statistical analysis was conducted using Prism 6 (Graphpad software). Kaplan-Meier survival curves were analysed with the log-rank test. Metastasis distribution was assessed with a contingency table and Chi-square test. All the other data sets were submitted to the Shapiro-Wilk normality test and depending on the result of this test, parametric or non-parametric comparison tests were performed. Specific tests used for each data set are mentioned in the figure legends. In all tests, P-values <0.05 were considered significant.
Deletion of Cd151does not interfere with mammary gland development and differentiation
Expression pattern of CD151 in the normal mammary gland and PyMT mammary tumors
Loss of Cd151significantly decreases mammary tumor multiplicity and growth
To avoid the kidney phenotype that has been described in FVB Cd151−/− mice , we chose to investigate the effects of Cd151 deletion on mammary tumorigenesis on a F1 (FVB×B6) background (see explanations in Methods). Consistent with previous reports showing increased PyMT tumor latency on the B6 background [24, 25], the onset of mammary tumors was significantly delayed by about 20 days on the (FVBxB6) F1 background in comparison to the pure FVB background (Additional file 2: Figure S2).
Effect of Cd151deletion on development of lung metastases
In this study, we found that deletion of Cd151 significantly impaired tumor development and progression in the murine MMTV/PyMT breast tumorigenesis model. These results are consistent with previous work investigating CD151’s function in other breast tumor models, suggesting that regardless of the tumor initiating oncogene, CD151 enhances tumor initiation and subsequent progression. Whilst the increase in tumor latency in Cd151−/− mice was only approaching significance, there was significant decrease in the number of tumors per mouse, suggesting that Cd151 deletion might impair PyMT tumor initiation. This could be determined in future experiments by assessing the early stages of PyMT tumorigenesis (from the age of weaning) and whether the extent of hyperplasia is different between genotypes before a palpable tumor can be detected. The decreased number of detectable tumors could also reflect a pronounced defect in tumor growth. Indeed, the size of the mammary tumors was also reduced in Cd151−/− mice, as shown by the total tumor burden per mouse (expressed as total tumor weight per mouse), as well as the individual tumor weight (represented as the weight of the largest tumor for each mouse). Intriguingly however we did not find a difference in tumor vascularisation or tumor cell proliferation and apoptosis that could explain the decreased tumor size in Cd151 −/− mice. In conclusion, the absence of effect on tumor proliferation and apoptosis reinforces the hypothesis stated above that Cd151 deletion may act predominantly by delaying PyMT tumor initiation, as suggested by the trend towards delayed tumor onset and the decreased number of tumors per mouse.
In recent years, Sadej and colleagues have reported a decrease in tumor growth associated with CD151 knock-down in a subcutaneous xenograft model . Similar to our result, CD151 knock-down did not have an effect on overall vessel density in the xenografts. Interestingly however the authors of that study described a pronounced decrease in the dense angiogenic network typically observed at the subcutaneous border of this type of tumors. We were not able to similarly assess the tumor border in our model system due to the nature of the de novo MMTV/PyMT tumors and because rich vascular networks do not develop at the periphery of tumors in this model.
Tumor progression in the MMTV/PyMT model has been very well characterized; it follows several well-defined stages from hyperplasia to non-invasive adenoma and finally invasive carcinoma . Loss of the myoepithelial cell layer surrounding the hyperplastic luminal epithelial structures is an important step in the progression to invasive carcinoma in these tumors [26, 17]. Moreover, mammary myoepithelial cells have been referred to as ‘natural tumor suppressors’ because of their capacity to block tumor cell growth and invasion [27, 28]. Because CD151 is primarily expressed in the mammary myoepithelial cells and the PyMT tumors arise from luminal epithelial cells in the mammary gland where CD151 has reduced expression, it is tempting to speculate that Cd151 deletion could affect tumor development indirectly through the microenvironment, by modifying the phenotype of the myoepithelial cells. However CD151 is also expressed, to a lesser extent, at the basolateral membrane of the luminal epithelium. It would be interesting to evaluate separately the specific effect of Cd151 deletion in the stroma (tumor vasculature, immune cells, myoepithelial cells) and the luminal epithelium/tumor cells respectively, using a conditional knock-out model. In addition, one could study the tumor cell autonomous versus microenvironment effect by performing orthotopic tumor grafts.
A recent study  investigated the impact of Cd151 deletion on tumor onset, growth and metastasis of another well characterized mouse mammary tumor model, the MMTV/Neu model (overexpressing multiple copies of wild-type Neu, the rat homolog of ErbB2). In their study, Deng and colleagues observed a significant delay in tumor onset as well as a decreased number of metastases per animal in the Cd151−/− and Cd151+/− groups compared with Cd151+/+ and they also suggest that the effects of CD151 in the MMTV/Neu model are largely mediated through α6β4 integrin. Together our data shows that Cd151 deletion significantly impairs tumor occurrence and growth. On assessment of lung metastasis, we obtained different results from Deng and colleagues. The median incidence, number and size of lung metastases did not vary with the Cd151 genotype in our study, and the higher metastatic burden in a proportion of wild-type mice was associated with increased primary tumors. These results suggest that ablation of Cd151 does not directly affect metastasis in the PyMT model. Future studies using experimental metastasis models will be required to elucidate the contradictory effects on metastasis in the MMTV/Neu and MMTV/PyMT models.
In addition, two other studies have assessed the role of CD151 on de novo tumorigenesis in mouse cancer models other than breast. Firstly, similarly to our results in the PyMT breast cancer model, Takeda and colleagues have demonstrated that CD151 promotes tumor incidence and multiplicity in a skin carcinogenesis model . Secondly, recent work in our laboratory has shown that deletion of Cd151 reduces spontaneous metastasis of prostate tumors in the TRAMP model . Altogether, the literature and the data we report here identify CD151 as an enhancer of de novo tumorigenesis and/or spontaneous metastasis across a broad range of cancer types.
Interestingly, the functional role of CD151 in tumorigenesis and metastasis has been mainly linked to its association with the laminin receptors in the literature. These adhesion receptors include mostly α3β1, α6β1 and α6β4 integrins in the mammary epithelium. Laminin binding integrins have a well-documented role in malignant cell processes regulating crucial functions such as cell proliferation and invasion. β1-integrins represent the most predominantly expressed integrins in the mammary epithelium and their direct involvement in mammary tumorigenesis has been extensively demonstrated using mouse models of breast cancer. For example, it has been shown that β1-integrin is absolutely required for PyMT mammary tumor initiation and progression [4, 30]. Similarly, several studies have demonstrated that ablation of crucial signaling molecules downstream of β1-integrin such as focal adhesion kinase (FAK) also dramatically decreased cancer cell proliferation [31, 32, 33]. In contrast to the total block in PyMT induced tumorigenesis, mammary epithelial deletion of β1 integrin in an MMTV/Neu (activated ErbB2) mouse model did not totally prevent tumor development. Although it significantly delayed tumor onset, all of the mice developed tumors that were histologically identical to control mice . The major effect of β1 deletion in the context of activated ErbB2 mammary tumorigenesis was the decrease in incidence of lung metastasis as well as decreased metastatic burden. There is to our knowledge no published report on the role of β4-integrin in the MMTV/PyMT model but it has been shown however that β4-integrin collaborates with ErbB2 to promote mammary tumorigenesis [3, 16] in the MMTV/Neu mouse model.
In summary, the results of this study show that CD151 enhances mammary tumor initiation and progression in the MMTV/PyMT mouse model but no direct effect on metastasis was demonstrated. The effects of Cd151 deletion in the MMTV/PyMT model are likely to be mediated by a combination of both β1- and β4-integrins but this remains to be tested. In conclusion, our findings strongly support the relevance of CD151 as a potential therapeutic target to delay breast cancer progression.
The authors thank Celeste Harrison and Matthew Bowman for technical assistance. This work was supported by an early career Fellowship from the Cancer Institute NSW (06/ECF/1/22) and a Gladys M. Brawn Memorial post-doctoral Fellowship to S.R. L.K.A was supported by a Principal Research Fellowship from the National Health and Medical Research Council of Australia.
- 4.White DE, Kurpios NA, Zuo D, Hassell JA, Blaess S, Mueller U, Muller WJ: Targeted disruption of beta1-integrin in a transgenic mouse model of human breast cancer reveals an essential role in mammary tumor induction. Cancer Cell. 2004, 6 (2): 159-170. 10.1016/j.ccr.2004.06.025.CrossRefPubMedGoogle Scholar
- 5.Sadej R, Romanska H, Baldwin G, Gkirtzimanaki K, Novitskaya V, Filer AD, Krcova Z, Kusinska R, Ehrmann J, Buckley CD, Kordek R, Potemski P, Eliopoulos AG, Lalani e-N, Berditchevski F: CD151 regulates tumorigenesis by modulating the communication between tumor cells and endothelium. Mol Cancer Res. 2009, 7 (6): 787-798. 10.1158/1541-7786.MCR-08-0574.CrossRefPubMedGoogle Scholar
- 10.Voss MA, Gordon N, Maloney S, Ganesan R, Ludeman L, McCarthy K, Gornall R, Schaller G, Wei W, Berditchevski F, Sundar S: Tetraspanin CD151 is a novel prognostic marker in poor outcome endometrial cancer. Br J Cancer. 2011, 104 (10): 1611-1618. 10.1038/bjc.2011.80.CrossRefPubMedPubMedCentralGoogle Scholar
- 11.Yang XH, Richardson AL, Torres-Arzayus MI, Zhou P, Sharma C, Kazarov AR, Andzelm MM, Strominger JL, Brown M, Hemler ME: CD151 accelerates breast cancer by regulating alpha 6 integrin function, signaling, and molecular organization. Cancer Res. 2008, 68 (9): 3204-3213. 10.1158/0008-5472.CAN-07-2949.CrossRefPubMedPubMedCentralGoogle Scholar
- 13.Zijlstra A, Lewis J, Degryse B, Stuhlmann H, Quigley JP: The inhibition of tumor cell intravasation and subsequent metastasis via regulation of in vivo tumor cell motility by the tetraspanin CD151. Cancer Cell. 2008, 13 (3): 221-234. 10.1016/j.ccr.2008.01.031.CrossRefPubMedPubMedCentralGoogle Scholar
- 15.Novitskaya V, Romanska H, Dawoud M, Jones JL, Berditchevski F: Tetraspanin CD151 regulates growth of mammary epithelial cells in three-dimensional extracellular matrix: implication for mammary ductal carcinoma in situ. Cancer Res. 2010, 70 (11): 4698-4708. 10.1158/0008-5472.CAN-09-4330.CrossRefPubMedPubMedCentralGoogle Scholar
- 17.Lin EY, Jones JG, Li P, Zhu L, Whitney KD, Muller WJ, Pollard JW: Progression to malignancy in the polyoma middle T oncoprotein mouse breast cancer model provides a reliable model for human diseases. Am J Pathol. 2003, 163 (5): 2113-2126. 10.1016/S0002-9440(10)63568-7.CrossRefPubMedPubMedCentralGoogle Scholar
- 19.Deng X, Li Q, Hoff J, Novak M, Yang H, Jin H, Erfani SF, Sharma C, Zhou P, Rabinovitz I, Sonnenberg A, Yi Y, Zhou P, Stipp CS, Kaetzel DM, Hemler ME, Yang XH: Integrin-associated CD151 drives ErbB2-evoked mammary tumor onset and metastasis. Neoplasia. 2012, 14 (8): 678-689.CrossRefPubMedPubMedCentralGoogle Scholar
- 20.Wright MD, Geary SM, Fitter S, Moseley GW, Lau LM, Sheng KC, Apostolopoulos V, Stanley EG, Jackson DE, Ashman LK: Characterization of mice lacking the tetraspanin superfamily member CD151. Mol Cell Biol. 2004, 24 (13): 5978-5988. 10.1128/MCB.24.13.5978-5988.2004.CrossRefPubMedPubMedCentralGoogle Scholar
- 23.Tuominen VJ, Ruotoistenmaki S, Viitanen A, Jumppanen M, Isola J: ImmunoRatio: a publicly available web application for quantitative image analysis of estrogen receptor (ER), progesterone receptor (PR), and Ki-67. Breast Cancer Res. 2010, 12 (4): R56-10.1186/bcr2615.CrossRefPubMedPubMedCentralGoogle Scholar
- 24.Davie SA, Maglione JE, Manner CK, Young D, Cardiff RD, MacLeod CL, Ellies LG: Effects of FVB/NJ and C57Bl/6 J strain backgrounds on mammary tumor phenotype in inducible nitric oxide synthase deficient mice. Transgenic Res. 2007, 16 (2): 193-201. 10.1007/s11248-006-9056-9.CrossRefPubMedPubMedCentralGoogle Scholar
- 31.Lahlou H, Sanguin-Gendreau V, Zuo D, Cardiff RD, McLean GW, Frame MC, Muller WJ: Mammary epithelial-specific disruption of the focal adhesion kinase blocks mammary tumor progression. Proc Natl Acad Sci U S A. 2007, 104 (51): 20302-20307. 10.1073/pnas.0710091104.CrossRefPubMedPubMedCentralGoogle Scholar
- 32.Provenzano PP, Inman DR, Eliceiri KW, Beggs HE, Keely PJ: Mammary epithelial-specific disruption of focal adhesion kinase retards tumor formation and metastasis in a transgenic mouse model of human breast cancer. Am J Pathol. 2008, 173 (5): 1551-1565. 10.2353/ajpath.2008.080308.CrossRefPubMedPubMedCentralGoogle Scholar
- 34.Huck L, Pontier SM, Zuo DM, Muller WJ: Beta1-integrin is dispensable for the induction of ErbB2 mammary tumors but plays a critical role in the metastatic phase of tumor progression. Proc Natl Acad Sci U S A. 2010, 107 (35): 15559-15564. 10.1073/pnas.1003034107.CrossRefPubMedPubMedCentralGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2407/14/509/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. 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.