G1 checkpoint establishment in vivo during embryonic liver development
© Wang et al.; licensee BioMed Central Ltd. 2014
Received: 9 December 2013
Accepted: 9 May 2014
Published: 19 May 2014
The DNA damage-mediated cell cycle checkpoint is an essential mechanism in the DNA damage response (DDR). During embryonic development, the characteristics of cell cycle and DNA damage checkpoint evolve from an extremely short G1 cell phase and lacking G1 checkpoint to lengthening G1 phase and the establishment of the G1 checkpoint. However, the regulatory mechanisms governing these transitions are not well understood. In this study, pregnant mice were exposed to ionizing radiation (IR) to induce DNA damage at different embryonic stages; the kinetics and mechanisms of the establishment of DNA damage-mediated G1 checkpoint in embryonic liver were investigated.
We found that the G2 cell cycle arrest was the first response to DNA damage in early developmental stages. Starting at E13.5/E15.5, IR mediated inhibition of the G1 to S phase transition became evident. Concomitantly, IR induced the robust expression of p21 and suppressed Cdk2/cyclin E activity, which might involve in the initiation of G1 checkpoint. The established G1 cell cycle checkpoint, in combination with an enhanced DNA repair capacity at E15.5, displayed biologically protective effects of repairing DNA double-strand breaks (DSBs) and reducing apoptosis in the short term as well as reducing chromosome deletion and breakage in the long term.
Our study is the first to demonstrate the establishment of the DNA damage-mediated G1 cell cycle checkpoint in liver cells during embryogenesis and its in vivo biological effects during embryonic liver development.
Endogenous and exogenous DNA damage are life-long threats to the health of an organism and can limit the survival and regenerative potential of both embryonic stem cells (ESCs) and adult stem cells (ASCs). Any genetic alterations in the progenitor cells can compromise the genomic stability and functionality of entire cell lineages [1–3]. Moreover, damage to cellular DNA can be the most important initiating factor in the development of cancer. The DNA damage response (DDR) network has developed to sense and respond to DNA damage and is critical for the maintenance of genetic integrity. The DDR is a complex network that involves the control of cell cycle arrest, the activation of DNA repair machinery, the induction of apoptosis, and the regulation of telomere length. Within this network, the activation of DNA damage checkpoints plays a central role in DDR signaling to ensure the correct scan of the entire genome before DNA replication (S phase) and cell division (M phase) [2, 3].
Although the DNA damage-mediated checkpoint is critical in DDR signaling, many of the regulatory components that govern this signaling pathway, specifically in cells at the embryonic stage and during developmental processes, are not known. The cell cycle and the DNA damage checkpoint change over time, from an extremely short G1 cell cycle to maintain pluripotency, to lengthening the G1 phase during differentiation [4–7]; from lacking a G1 checkpoint [8–10] to the establishment of a G1 cell cycle arrest. For example, both murine and human ES cells, as well as human embryonic carcinoma (EC) cells, are defective in the G1 checkpoint after DNA damage [8–10]. Differentiated EC cells show an increased G1 cell population but lack a G1 checkpoint, even though DDR protein activation appears to be normal . Thus, the developmental stages and circumstances under which the G1 cell cycle checkpoint is required remain unclear. Moreover, most studies regarding DDR checkpoints and DNA repair in human and murine stem cells have been performed in vitro with ESC lines and compared to mouse embryonic fibroblasts (MEFs) or genetic knockout MEFs, an approach which reflects only a short period of embryonic development.
For stem or progenitor cells, it is necessary to evolve effective and non-mutagenic DNA repair capacities to avoid passing mutations on to subsequent generations and initiating cancers [11, 1]. In both humans and mice, ES cells have been shown to be more capable of repairing DNA damage than their differentiated derivatives . Recent studies suggest that the kinetics of DNA repair are different in the hematopoietic stem cells (HSCs) and progenitor cells of human versus mouse; whereas murine HSCs display faster repair kinetics, human HSCs are less capable of DNA repair and are more pro-apoptotic [1, 11, 12]. Therefore, different cell lineages, as well as different species of stem and progenitor cells, have different DDR and repair capacities. DNA double-strand breaks (DSBs), which arise during DNA replication or following exposure to ionizing radiation (IR), are considered the most harmful lesions. The principal mechanisms of DSB repair in mammalian cells include nonhomologous end-joining (NHEJ) and homologous recombination repair (HR). HR ensures accurate DSB repair, while NHEJ repair is rapid and efficient but error-prone [13, 14].
In the past, MEF cells, derived from various genetic knockout mice, have been used for cell cycle related studies in vitro whereas cell cycle studies in vivo at embryonic stages have been performed by in situ assays. There have been no detailed investigations of DDR kinetics, including checkpoints and DNA damage repair, at different embryonic developmental stages during organ development by using live cells. In this study, we investigated when (at which embryonic stage) and how the DNA damage-mediated G1 checkpoint is established during in vivo embryonic liver development and associated DNA damage repair pathways.
Mouse strains and embryos
ICR mice (CD-1, Harlan UK Ltd, UK) were provided and maintained by the Laboratory Animal Unit of the University of Hong Kong and used for all experiments. Embryos at different stages, including E11.5, E13.5, E15.5, and E17.5, were obtained from pregnant ICR mice. Post-natal mice at P0, P7, P14, P21, and P56 were also used. H&E stained mouse liver tissue structures from embryonic stage to adult were shown in Additional file 1: Figure S1. This study was approved by The Committee on the Use of Live Animals of the University of Hong Kong (CULATR 1623–08).
Ionizing radiation (IR)
Pregnant mice were subjected to 4–6 Gy of IR (Gammacell 3000, MDS Nordion, Germany) at defined embryonic stages. At 0, 6, 16, and 24 hours after IR, pregnant mice were sacrificed, and embryonic livers were dissected for cell cycle analysis and other experiments. P0 to P56 mice were also subjected to 2 Gy of IR, and the liver cells were isolated at multiple time points.
Isolation of fetal or adult liver cells
Fetal livers were dissected out from mouse embryos (E11.5 liver had to be dissected out under a dissection microscope), minced, and digested with collagenase-V (100 units/ml, Sigma-Aldrich, St. Louis, MO, USA) for 10 minutes at 37°C. The tissue was then filtered through a 40 μm nylon mesh to remove debris. The cells were collected by centrifugation (500 g for 5 minutes) at 4°C. Isolated single liver cells were fixed with cold 80% ethanol and kept at -20°C for cell cycle analysis. The same procedure was used to isolate adult liver cells. For cell cycle analysis, a pool of 3–5 of E11.5 embryonic livers and 2–3 of E13.5 or E15.5 fetal livers was collected.
Tissue specimens and nuclear protein fractions
The liver tissue was frozen in liquid nitrogen immediately after harvest for the generation of protein lysates. For nuclear protein extraction, 20 g of fresh fetal liver was homogenized thoroughly on ice and centrifuged. The pellet was re-suspended in buffer B (5 mM HEPES, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT, 26% (v/v) glycerol, pH 7.9) and 300 mM NaCl for 20 minutes at 4°C. After centrifugation (24000 g for 20 minutes at 4°C), the supernatant containing the nuclear protein was kept at -70°C prior to performing the DNA repair assays. For immunohistochemistry, liver tissue was fixed in 4% paraformaldehyde overnight and then embedded in paraffin blocks. For E11.5 to E15.5, the whole embryos were fixed; for E17.5 to P56, the dissected livers were fixed.
For analysis of the cell cycle, the nuclei were stained with propidium iodide (PI; Sigma) and analyzed with a Cytomics FC 500 (Beckman Coulter, Indianapolis, IN, USA). Fetal liver cells were also stained with an anti-albumin antibody (R&D Systems, Minneapolis, MN, USA) at different embryonic stages to establish a threshold by which albumin-positive populations were gated for the analysis of DNA content. Starting at E11.5, over 80% of the isolated embryonic liver cells expressed albumin (Additional file 1: Figure S2). The percentage of each cell cycle population was calculated with ModFit v3.1 software (Verity Software House, Topsham, ME, USA).
Antibodies, Western Blots (WB), and Immunoprecipitation (IP)
Antibodies against cyclins A, E, and B1 and those against Cdk1, p21, RAD51, and Ligase IV were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Antibodies against cyclin D1, phospho-Cdk2 (Thr160), and γ-H2AX were purchased from Cell Signaling Technology (Beverly, MA, USA). An anti-Cdk2 antibody was kindly provided by Prof. K Yamashita (Kanazawa University, Japan). For WB, 10–40 μg of total tissue protein lysate was loaded and separated on 10% or 12% acrylamide gels and transferred to PVDF membranes, which were then incubated with primary and secondary antibodies. Protein expression was revealed with enhanced chemi lumescent (ECL) reagents (Amersham, GE healthcare, UK). For IP, total cell lysates (100–200 μg) were incubated with 1–2 μl of primary antibody followed by incubation with 30 μl of protein G sepharose (Amersham Biosciences, Sweden). After 3 to 4 washes, the immune complexes were dissolved in SDS sample buffer for WB. For most IP assays, rabbit-derived antibodies were used to avoid cross-reaction.
For antigen unmasking, deparaffinized sections were boiled for 10 minutes. After blocking with 3% H2O2 for 10 minutes and 1% BSA for 1 hour, the sections were incubated with an anti-γH2AX antibody overnight at 4°C. Polymer horseradishperoxidase-conjugated secondary antibodies and DAB + Chromogen (Dako North America, Carpinteria, CA, USA) were used to visualize the signal.
In vitroNHEJ assay
A pUC19 plasmid was cut with the restriction enzyme PvuII overnight at 37°C and followed by alkaline phosphatase (New England BioLabs, Ipswich, MA, USA) treatment for 1 hour to prevent self-ligation of the fragment. The linearized plasmid was identified by agarose gel electrophoresis and purified by a Gel purification kit (Qiagen). 50 ng of the linearized DNA was used as a substrate and incubated with fetal liver nuclear protein (10 μg) in T4 ligase buffer and 1 mM dNTPs for 2 hours at 14°C. After this reaction, the DNA was de-proteinized by purification using a PCR purification kit (Qiagen). The amount of end-joined DNA products was measured by quantitative real-time PCR with primer set A for a loading control, resulting in the amplification of a 151 bp DNA product at a 143 bp distance from a PvuII site. Primer set B flanked the joining junction for the amplification of the joined DNA fragments. The relative NHEJ activity was calculated as the ratio of the end-joined products normalized to the loading control products. Primer sets A and B are shown in Additional file 1: Table S1.
In situcell death detection
Apoptotic liver cells on embryonic liver sections were determined by using in situ cell death detection fluorescein kit (Roche Applied Science, IN, USA) following manufacturer’s instruction.
G-banding and spectral karyotyping (SKY) analysis
Pregnant mice were exposed to 0.5 Gy IR at embryonic stages 11.5 and 15.5. The embryos surviving this low dose of IR developed normally. The offspring were allowed to grow to adulthood (7 weeks old), and the liver cells were then isolated from the mice and cultured. Chromosome G-banding and SKY analysis was performed by the Van Andel Institute (Grand Rapids, MI, USA) to examine chromosome breakage and rearrangement.
Cell cycle shifting at different developmental stages
Cell cycle drivers at different developmental stages
Next, we examined the expression of cell cycle drivers (cyclins and Cdks) in liver cells at different developmental stages. For the G1 and S phases, cyclin E and Cdk2, but not cyclin A, showed consistently high levels of expression at all developmental stages (Figure 1C). Cyclin D1 was expressed at low levels during most stages and was enhanced at P14 to P56, although the G1 cell phase was dominant at these developmental time points (Figure 1A, B, C). For the G2/M phase, cyclin B1 was consistently expressed at high levels during all stages and was further enhanced after the embryonic stage, whereas Cdk1 was down-regulated in the adult liver (Figure 1C). When the Cdk inhibitors were measured, p27 expression seemed to be in accordance with the increased G1 population, while p21 was not detectable (Figure 1D). Taken together, we observed that only Cdk1 and cyclin A were down-regulated during embryonic liver development.
Establishment of the G1 checkpoint in E13.5/E15.5 embryonic liver cells
Cdk2 and p21 became regulated in response to IR in E13.5/E15.5 embryonic liver cells
Regulation of the IR-mediated G1 checkpoint through the down-regulation of the Cdk2/cyclin E complex in E15.5 embryonic liver cells
The interaction of Cdks and cyclins in response to IR in liver cells was compared at E11.5 and E15.5 by co-IP. The expression of the Cdk2/cyclin E complex was significantly reduced after IR in E15.5, but not E11.5, liver cells (Figure 3D). This was confirmed by two different co-IP approaches (Figure 3D). Further, Cdk2 phosphorylation at Thr 160, which was necessary for the activation of Cdk2 complexes, was reduced after IR at E15.5 (Figure 3D). The interactions between Cdk1 and the cyclins A and B1 and between Cdk2 and the cyclins A and B1 remained the same before and after IR in both E11.5 and E15.5 liver cells (Figure 3E). These results suggested that reduced Cdk2 activity by the down regulation of phosphorylation of Cdk2 (Thr160) and the Cdk2/cyclin E complex was required for DNA damage-mediated G1 cell cycle arrest in E15.5 cells, and that this down regulation may be regulated by p21 (Figure 3F).
Enhanced DNA damage repair capacity in E15.5 compared to E11.5 embryonic liver cells
Short-term biological effects
Statistical comparison of early apoptosis in embryonic liver sections between E11.5 and E15.5 after IR
Early apoptosis (%)
24.8% ± 1.7
6.7% ± 0.4
Long-term biological effects
G-banding & spectral karyotyping (SKY) analysis in adult liver cells born from mouse exposing to IR (0.5 Gy) at E11.5 or E15.5 embryonic stage
IR at E11.5
IR at E15.5
Overall abnormal chromosomes
Chromosome breakage (%)
Chromosome rearrangement (%)
Chromosome deletion among breakage (%)
In this study, we characterized the establishment of the G1 checkpoint and the associated DSB repair, as well as the biological effects of these processes, in the murine embryonic liver at different stages of development. It is the first investigation using live embryonic liver tissue to investigate the DNA damage-mediated checkpoint and repair during embryonic development.
ES cells lack DNA damage-mediated G1 checkpoints in culture [8–10] and employ S and G2 cell cycle arrest as protective mechanisms [9, 10]. When ES cells differentiated and entered into lineage and organ development, G2 cell cycle arrest was still the earliest response to DNA damage (Figure 2A, B, C, E)  during the early developmental stages of embryonic liver cells (E11.5 to E15.5). Beginning at E13.5/E15.5, the embryonic liver cells started to establish a G1 DNA damage checkpoint, where G1 cell cycle arrest replaced transient G2 arrest for longer periods of time following DNA damage (Figure 2A, B, C, D, E) . It is unclear why the G1 checkpoint is established so much later than the G2 checkpoint. A requirement for rapid proliferation results in the cells being mostly in the S and G2/M cell cycles, with a short G1 phase, during the early developmental stage, which might also equip cells with S and G2/M cell cycle checkpoint mechanisms in response to DNA damage. During the differentiation of ES cells and embryo development, the G1 cell cycle length and G1 population both increase [4–7, 17–19], as does the G1 checkpoint machinery (Figures 1A, 2A, B, C, D) . More importantly, cell cycle arrest provides time for the repair of DNA damage. The G2 arrest could prevent apoptosis, while the G1 arrest is important to prevent damaged DNA from entering into the S phase. Thus, by comparison between E11.5 and E15.5, our study demonstrates that the development of the G1 checkpoint play a critical role in long-term genome stability (Table 2) (Figure 5A, B), although aberrant chromosomes still exist in the adult liver cells, and protection is, therefore, not complete (Table 2).
γH2AX positive cells in E11.5 livers remained high throughout longer time course after IR (Figure 4C), the relative larger population of E11.5 liver cells in S phase could be a cause; as it has shown that DSB foci developed during the course of the UV-induced replication arrest . However, a minority of the UV induced foci showed DSB γH2AX colocalizing with 53BP1, whereas the majority was pan-nuclear γH2AX, a pre-apoptotic signal in the S phase . In our experimental setting: (a) IR did not induce replication arrest (Figure 2); (b) after a short G2 arrest, E11.5 liver cells returned to a basal level but not an S phase arrest (Figure 2); (c) the fraction of γH2AX positive cells was not significantly different between E11.5 and E15.5 liver cells under the condition without IR (p = 0.349) (Figure 4C), whereas S phase population of both were high (over 50%) (Figure 2A, C). Moreover, rapid NHEJ repair is not limited to specific cell phase but occurs throughout all cell cycle phases . Thus, high levels of γH2AX foci in E11.5 liver cells were likely due to their DNA repair capability not being competent enough, as well as lacking G1 arrest checkpoint control (Figures 2 and 4). In accordance, early apoptosis was significantly higher in E11.5 compared to E15.5 liver cells (p = 0.005) (Figure 4D, Table 1), which is a cellular consequence of absence of G1 checkpoint as well as retaining high level of un-repaired DNA damage in E11.5.
The initiation of the DNA damage-mediated G1 cell cycle arrest occurs at embryonic stage E13.5/E15.5 in embryonic liver cells, and this process is regulated by the p21-mediated down-regulation of the Cdk2/cyclin E complex. The G1 checkpoint, in combination with DNA repair, plays a biological role in repairing cellular DSBs and in preventing early apoptosis in the short term and reducing chromosome abnormalities in the long term.
This study is supported by Research Grants Council of Hong Kong to XQ Wang (GRF HKU 778809).
- Milyavsky M, Gan OI, Trottier M, Komosa M, Tabach O, Notta F, Lechman E, Hermans KG, Eppert K, Konovalova Z, Omatsky O, Domany E, Meyn MS, Dick JE: A distinctive DNA damage response in human hematopoietic stem cells reveals an apoptosis-independent role for p53 in self-renewal. Cell Stem Cell. 2010, 7: 186-197. 10.1016/j.stem.2010.05.016.View ArticlePubMedGoogle Scholar
- Fraosina G: The bright and the dark sides of DNA repair in stem cells. J Biomed Biotechnol. 2010, 2010: 845396-Google Scholar
- Poehlmann A, Roessner A: Importance of DNA damage checkpoints in the pathogenesis of human cancers. Pathol Res Pract. 2010, 206: 591-601. 10.1016/j.prp.2010.06.006.View ArticlePubMedGoogle Scholar
- Orford KW, Scadden DT: Deconstructing stem cell self-renewal: genetic insights into cell cycle regulation. Nat Rev Genet. 2008, 9: 115-128. 10.1038/nrg2269.View ArticlePubMedGoogle Scholar
- Singh AM, Dalton S: The cell cycle and Myc intersect with mechanisms that regulate pluripotency and reprogramming. Cell Stem Cell. 2009, 5: 141-149. 10.1016/j.stem.2009.07.003.View ArticlePubMedGoogle Scholar
- Lange C, Calegari F: Cdks and cyclins link G1 length and differentiation of embryonic, neural and hematopoietic stem cells. Cell Cycle. 2010, 9: 1893-1900. 10.4161/cc.9.10.11598.View ArticlePubMedGoogle Scholar
- Lange C, Huttner WB, Calegari F: Cdk4/cyclinD1 overexpression in neural stem cells shortens G1, delays neurogenesis, and promotes the generation and expansion of basal progenitors. Cell Stem Cell. 2009, 5: 320-331. 10.1016/j.stem.2009.05.026.View ArticlePubMedGoogle Scholar
- Aladjem MI, Spike BT, Rodewald LW, Hope TJ, Klemm M, Jaenisch R, Wahl GM: ES cells do not activate p53-dependent stress responses and undergo p53-independent apoptosis in response to DNA damage. Curr Biol. 1998, 8: 145-155. 10.1016/S0960-9822(98)70061-2.View ArticlePubMedGoogle Scholar
- Wang X, Lui VC, Poon RT, Lu P, Poon RY: DNA damage mediated s and g(2) checkpoints in human embryonal carcinoma cells. Stem Cells. 2009, 27: 568-576. 10.1634/stemcells.2008-0690.View ArticlePubMedGoogle Scholar
- Momcilović O, Choi S, Varum S, Bakkenist C, Schatten G, Navara C: Ionizing radiation induces ataxia telangiectasia mutated-dependent checkpoint signaling and G(2) but not G(1) cell cycle arrest in pluripotent human embryonic stem cells. Stem Cells. 2009, 27: 1822-1835. 10.1002/stem.123.View ArticlePubMedGoogle Scholar
- Mohrin M, Bourke E, Alexander D, Warr MR, Barry-Holson K, Le Beau MM, Morrison CG, Passegué E: Hematopoietic stem cell quiescence promotes error-prone DNA repair and mutagenesis. Cell Stem Cell. 2010, 7: 174-185. 10.1016/j.stem.2010.06.014.View ArticlePubMedGoogle Scholar
- Seita J, Rossi DJ, Weissman IL: Differential DNA damage response in stem and progenitor cells. Cell Stem Cell. 2010, 7: 145-147. 10.1016/j.stem.2010.07.006.View ArticlePubMedGoogle Scholar
- Price BD, D’Andrea AD: Chromatin remodeling at DNA double-strand breaks. Cell. 2013, 152: 1344-1354. 10.1016/j.cell.2013.02.011.View ArticlePubMedGoogle Scholar
- Pallis AG, Karamouzis MV: DNA repair pathways and their implication in cancer treatment. Cancer Metastasis Rev. 2010, 29: 667-685.View ArticleGoogle Scholar
- Serrano L, Liang L, Chang Y, Deng L, Maulion C, Nguyen S, Tischfield JA: Homologous recombination conserves DNA sequence integrity throughout the cell cycle in embryonic stem cells. Stem Cells Dev. 2011, 20: 863-874.View ArticleGoogle Scholar
- Rogakou EP, Pilch DR, Orr AH, Ivanova VS, Bonner WM: DNA double-stranded breaks induce histone H2AX phosphorylation on serine 139. J Biol Chem. 1998, 273: 5858-5868. 10.1074/jbc.273.10.5858.View ArticlePubMedGoogle Scholar
- White J, Stead E, Faast R, Conn S, Cartwright P, Dalton S: Developmental activation of the Rb-E2F pathway and establishment of cell cycle-regulated cyclin-dependent kinase activity during embryonic stem cell differentiation. Mol Biol Cell. 2005, 16: 2018-2027. 10.1091/mbc.E04-12-1056.View ArticlePubMedGoogle Scholar
- Dalton S: Exposing hidden dimensions of embryonic stem cell cycle control. Cell Stem Cell. 2008, 4: 9-10.View ArticleGoogle Scholar
- Wang Y, Baskerville S, Shenoy A, Babiarz JE, Baehner L, Blelloch R: Embryonic stem cell-specific microRNAs regulate the G1-S transition and promote rapid proliferation. Nat Genet. 2008, 40: 1478-1483. 10.1038/ng.250.View ArticlePubMedGoogle Scholar
- Salomoni P, Calegari F: Cell cycle control of mammalian neural stem cells: putting a speed limit on G1. Trends Cell Biol. 2010, 20: 233-243. 10.1016/j.tcb.2010.01.006.View ArticlePubMedGoogle Scholar
- Calegari F, Huttner WB: An inhibition of cyclin dependent kinases that lengthens, but does not arrest, neuroepithelial cell cycle induces premature neurogenesis. J Cell Sci. 2003, 116: 4947-4955. 10.1242/jcs.00825.View ArticlePubMedGoogle Scholar
- Lim S, Kaldis P: Loss of Cdk2 and Cdk4 induces a switch from proliferation to differentiation in neural stem cells. Stem Cells. 2012, 30: 1509-1520. 10.1002/stem.1114.View ArticlePubMedGoogle Scholar
- Limoli CL, Giedzinski E, Bonner WM, Cleaver JE: UV-induced replication arrest inthe xeroderma pigmentosum variant leads to DNA double-strand breaks, gamma-H2AX formation, and Mre11 relocalization. Proc Natl Acad Sci U S A. 2002, 99: 233-238. 10.1073/pnas.231611798.View ArticlePubMedGoogle Scholar
- de Feraudy S, Revet I, Bezrookove V, Feeney L, Cleaver JE: A minority of foci or pan-nuclear apoptotic staining of gammaH2AX in the S phase after UV damage contain DNA double-strand breaks. Proc Natl Acad Sci U S A. 2010, 107: 6870-6875. 10.1073/pnas.1002175107.View ArticlePubMedGoogle Scholar
- Rothkamm K, Krüger I, Thompson LH, Löbrich M: Pathways of DNA double-strand break repair during the mammalian cell cycle. Mol Cell Biol. 2003, 23: 5706-5715. 10.1128/MCB.23.16.5706-5715.2003.View ArticlePubMedGoogle Scholar
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/2.0), 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.