Genetic inhibition of the atypical kinase Wee1 selectively drives apoptosis of p53 inactive tumor cells
© Pappano et al.; licensee BioMed Central Ltd. 2014
Received: 20 January 2014
Accepted: 30 May 2014
Published: 13 June 2014
Tumorigenesis is the result of genomic or epigenomic insults and subsequent loss of the proper mechanisms to respond to these alterations leading to unscheduled growth. Tumors arising from these mutations often have altered cell cycles that offer proliferative advantages and lead to the accumulation of additional mutations that can lead to more aggressive phenotypes. Nevertheless, tumor cells must still adhere to the basic tenets of the cell cycle program to ensure their survival by DNA duplication, chromosomal segregation and cytokinesis. The atypical tyrosine kinase Wee1 plays a key role in regulating the cell cycle at the DNA synthesis and mitotic checkpoints via phosphorylation and subsequent inactivation of cyclin-dependent kinases (CDKs) in both healthy and tumorigenic cells.
To assess the role of Wee1 in tumor cell proliferation we performed small interfering RNA (siRNA) experiments in a panel of diverse cell lines derived from various tissue origins. We also tested the hypothesis that any potential effects would be as a result of the kinase activity of Wee1 by siRNA rescue studies with wild-type or kinase-dead versions of Wee1.
We find that, in general, cells with wild-type p53 activity are not susceptible to loss of Wee1 protein via siRNA. However, Wee1 siRNA treatment in tumor cells with an inherent loss of p53 activity results in a deregulated cell cycle that causes simultaneous DNA synthesis and premature mitosis and that these effects are kinase dependent. These cumulative effects lead to potent inhibition of cellular proliferation and ultimately caspase-dependent apoptosis in the absence of co-treatment with cytotoxic agents.
These results suggest that, while Wee1 acts as a tumor suppressor in the context of normal cell growth and its functional loss can be compensated by p53-dependent DNA damage repairing mechanisms, specific inhibition of Wee1 has deleterious effects on the proliferation and survival of p53 inactive tumors. In total, targeting the atypical kinase Wee1 with an siRNA-based therapeutic or a selective ATP competitive small molecule inhibitor would be a feasible approach to targeting p53 inactive tumors in the clinic.
KeywordsWee1 p53 Apoptosis CDK1 CDK2 DNA damage
Proper maintenance of the cell cycle is essential for the development and homeostasis of all living organisms. Neoplasms arising within these organisms also rely on a coordinated cell cycle to facilitate their rapid growth rate in response to external sources of nutrients and signaling stimuli. Progression through the cell cycle in both naïve and tumor tissue is monitored at checkpoints that sense possible defects in DNA synthesis and chromosomal segregation. Activation of these checkpoints results in cell cycle arrest and allows cells to rectify any negative perturbations that may be transmitted to their resulting daughter cells. Tumor cells with defective checkpoints rely more heavily on other checkpoints within the cell cycle to ensure their survival. Thus, deregulation of cell cycle control can have catastrophic results and therefore has led to a concerted effort towards generation of novel therapeutic agents targeting the cell cycle in tumors .
Transition through the cell cycle is dependent on the cyclin-dependent kinase (CDK) family of regulatory proteins. CDK activity is tightly monitored through several complex mechanisms including modulation of CDK stability by binding partner cyclins and CDK inhibitors [2, 3]. However, only a subset of CDK-cyclin complexes are directly involved in progression of the cell cycle through these respective checkpoints. The active CDK2-Cyclin E complex is essential to drive the G1/S transition of the cell cycle after the restriction checkpoint and the CDK1-Cyclin B complex (also known as the M-phase-promoting-factor) is the master regulator that initiates the G2/M transition after the mitotic checkpoint . In addition to their association with cyclins and CDK inhibitors, CDK1 and CDK2 activity are modulated both negatively and positively by phosphorylation and de-phosphorylation events . Wee1 is an atypical tyrosine kinase that most closely resembles serine/threonine kinases in both sequence and structure  and acts directly upon CDK1 and CDK2. Wee1 phosphorylation of tyrosine 15 (Y15) of CDK1 and CDK2 results in CDK inactivation and inhibition of S-phase and mitotic entry [6, 7]. Wee1 kinase activity therefore serves as a master regulator of cell cycle checkpoints by inactivating the CDK1-Cyclin B and CDK2-Cyclin E complexes until it can be assured that genomic integrity will be maintained and that the appropriate genetic information will be passed on to daughter cells.
Previous reports of Wee1 inhibition by small molecule kinase inhibitors and siRNA demonstrate that loss of Wee1 activity sensitizes p53 inactive cells to DNA damaging agents and radiosensitization [8–12]. We hypothesized that loss of Wee1 in the absence of cytotoxics should be able to affect tumor cell proliferation because all metazoan cells, including cancer cells, rely on at least partially functioning checkpoints to insure their survival. To elucidate the roles of Wee1 in cancer cell cycle progression we have utilized siRNA knockdown and rescue. We find that loss of Wee1 kinase activity results in dramatic cell cycle events including simultaneous mitosis and DNA synthesis that ultimately lead to apoptosis in a sub-set of p53 deficient cells. The anti-proliferative and apoptotic effects of Wee1 siRNA treatment can be circumvented by expression of a wild-type Wee1 rescue construct but not a kinase-defective version in affected target cells. This work provides new insights into the development of cancer therapeutics, suggesting that a small molecule inhibitor of Wee1 kinase should be efficacious against a large number of p53 inactive solid tumors as a single agent and provide a safe therapeutic window in the p53 wild-type tissues of patients.
All cell lines were obtained from ATCC (Manassas, VA, U.S.A.). All cell lines were grown according to manufacturer’s conditions in the presence of fetal bovine serum (Invitrogen, Carlsbad, CA, U.S.A.).
siRNA oligos and transfection
All siRNAs were obtained from the ON-TARGET plus collection purchased from Dharmacon (Lafayette, CO, U.S.A). The sense strand sequences of the Wee1 siRNA oligos employed in the study are as follows: #5 5′-AAUAGAACAUCUCGACUUA-3′, #6 5′-AAUAUGAAGUCCCGGUAUA-3′, #7 5′-GAUCAUAUGCUUAUACAGA-3′, #8 5′-CGACAGACUCCUCAAGUGA-3′. The negative control siRNA oligos: NT1 and NT2 product numbers D-001810-01 and D-001810-02. The Ran oligo targeting sequence was 5′- AGAAGAAUCUUCAGUACUAUU-3′. Transfections of siRNA duplexes were performed using RNAiMAX reagent (Invitrogen).
Cell proliferation and caspase assays
Viable cell numbers were measured using CellTiter-Glo reagent (Promega, Madison, WI, U.S.A.) or CyQuant (Invitrogen). Caspase-Glo 3/7 assay (Promega) was used to measure cellular caspase-3/7 activity.
Antibodies and Western blot analysis
Antibodies used included mouse anti-Wee1, mouse anti-P53 (Santa Cruz Biotechnology, Santa Cruz, CA, U.S.A.), rabbit anti-phospho-CDK1 (Y15), rabbit anti-phospho-H2AX (S139), mouse anti-phospho-Histone H3 (S10), rabbit anti-Histone H3 (Cell Signaling Technology, Danvers, MA, U.S.A.), mouse anti-CDK1 (Millipore, Billerica, MA, U.S.A.), mouse anti-P21 (BD Biosciences, San Diego, CA, U.S.A.), and mouse anti-Actin (Sigma, St. Louis, MO, U.S.A.). Western blots were performed as previously described .
DNA cloning and Engineered cell lines
The human Wee1 coding sequence was amplified using standard PCR protocols and cloned into the pLVX-puro lentiviral construct (Clontech, Mountain View, CA, U.S.A.). Kinase-altered Wee1 (K328R) and siRNA resistant constructs were introduced using the QuikChange site directed mutagenesis kit (Stratgene, Santa Clara, CA, U.S.A.). Cell lines were generated immediately after infection by mass selection in 2 μg/ml puromycin (Clontech).
siRNA rescue experiment
Engineered NCI-H1299 cell lines were grown in DMEM containing 10% fetal bovine serum and 2 μg/ml puromycin. Cell lysate samples were collected 48 hr after transfection for Western blot analysis. Five days after siRNA transfection, microscopic cell images were taken and both floating and attached cells were then collected by trypsinization and counted for viable cell number using a Vi-CELL cell viability analyzer (Beckman Coulter, Brea, CA, U.S.A.).
Time-lapse light microscopy movies
NCI-H1299 and A549 cells were plated and transfected 24 hours later with either control siRNA (NT1) or Wee1 siRNA (#8). Plates were placed in an incubator and images were captured every 30 minutes using IncuCyte™ High Definition (HD) Imaging (Essen Bioscience, Ann Arbor, MI, U.S.A). Movies were extracted for the indicated time frames (3 days of siRNA treatment) at a rate of 10 frames per second.
Analysis of cell cycle by flow cytometry
At varied times after transfection with 5nM siRNA, adherent cells were trypsinized and combined with any floating cells present, then washed with cold PBS. Cells were then stained with 0.5 ml PBS containing 50 μg/ml propidium iodide (PI), 0.1 mg/ml RNase A, 0.1% BSA, and 0.1% Triton-X100 for 20 min and cell cycle distribution was analyzed using a BD LSR-II flow cytometer (BD Biosciences, San Jose, CA, U.S.A.).
Double thymidine block
H1299 cells were treated with 2 mM thymidine (Sigma) and blocked for 20 hrs. Cells were then released by washing with PBS and feeding with regular culture medium. Four hours after the first release, cells were transfected with 5nM siRNA and incubated for another 4.5 hrs before the second treatment of 2 mM thymidine. Cells were blocked with thymidine for another 14.5 hrs and released again into regular medium. At the second release (t = 0), cells should be synchronized at G1/S phase boundary of the cell cycle. Cell samples were collected at varied time points for cell cycle and Western blot analysis as described above.
Click-iT EdU flow cytometry assay
NCI-H1299 cells were plated in 6-well plate and transfected with 5 nM siRNA as described above. After 24 or 48 hr incubation, cells were labeled with 10 μM EdU (Invitrogen) for 30 min. Both floating and attached cells were then collected for fixation, Click-iT reaction, and cell cycle staining following manufacturer’s protocol. Cell samples were analyzed for DNA content and EdU level using BD LSR-II flow cytometer.
Wee1 siRNA treatment results in potent inhibition of proliferation in a subset of human solid tumor cell lines
Wee1 siRNA treatment is on-target and its effects are kinase activity dependent
Using these experimental methods we could also determine if the siRNA effects were a result of loss of a scaffolding function or kinase activity of Wee1. To answer this question, we generated a siRNA-resistant/kinase-altered version of Wee1 in which altering the critical active site lysine of Wee1 to an arginine residue (K328R) results in a decrease of Wee1 kinase activity . NCI-H1299 cells over-expressing the siRNA-resistant/kinase-altered version of Wee1 were not rescued to a similar extent as the kinase-active version (Figure 3f–o). The siRNA-resistant/kinase altered Wee1 construct does produce a slight amount of phenotypic rescue (~35%) for both siRNA #6 and #8, but this can be explained by one of two likely possibilities: the K328R change may not produce a fully kinase-dead Wee1 or the over-expressed siRNA-resistant Wee1 constructs may cause some degree of siRNA protection to the endogenous Wee1 mRNA present in the cells. These effects can be seen on the molecular level when relative levels of phospho- to total-CDK (p/t CDK) are taken into account. The level of p/t CDK is essentially absent in the vector control NCI-H1299 cells (Figure 3e), restored in the siRNA-resistant cells (Figure 3j) and partial when the siRNA-resistant/kinase altered Wee1 is over-expressed (Figure 3o). These experimental results imply that the Wee1 siRNAs are on-target and that the loss of kinase activity and the subsequent downstream signaling effects are responsible for the dramatic effects of Wee1 siRNA treatment on cellular proliferation seen in a specific subset of solid tumor cell lines.
Wee1 knockdown effects are stronger in tumor cell lines with recognized inactive p53 status
Wee1 siRNA effects on cell line panel
Direct evidence for the importance of p53 status in response to loss of Wee1 activity
Loss of Wee1 activity results in caspase-dependent apoptosis in p53-null cells
Loss of Wee1 activity has dramatic cell cycle effects in a subset of tumor cell lines
Loss of Wee1 activity results in mistimed cell cycle events
Wee1 knockdown results in incorrect timing of DNA replication origin firing
Wee1 is a master regulator of cellular integrity and its activity keeps normal cells from proceeding through the S-phase and G2/M checkpoints before their genomic content is adequately prepared to do so [6, 7]. We found Wee1 to be over-expressed in a number of solid primary tumors (data not shown, http://htp://www.oncomine.com) and this has recently been demonstrated in primary glioblastomas . These data suggest that Wee1 may play a role in some cancer cells to maintain a level of genetic stability to enable the growth advantage conferred by their new cancer karyotype. We hypothesized that cancer cells would be susceptible to loss of Wee1 kinase activity based on the premise that all metazoan cells, including those in neoplasms, rely on the correct timing of cell cycle events for their survival and proliferation. Our studies indicate that this hypothesis is correct in a subset of solid tumor cell lines with inactive p53. Upon loss of Wee1 kinase activity resulting from specific siRNA knockdown, these cells undergo DNA synthesis while attempting simultaneous mitosis, ultimately leading to mitotic catastrophe and caspase-dependent apoptosis.
p53 is a short-lived transcription factor that acts as a tumor suppressor to detect and eliminate incipient cancerous cells. p53 activity results from a complex network involving the DNA damage response, transcriptional activation and post-translational modifications and results in regulation of hundreds of genes . Neutralization of p53 function is a major hallmark of tumor cells and this loss can be the result of direct somatic mutation, deletion, proteasomal degradation or sequestration to achieve a pathologic survival advantage . One such protein that mediates degradation of p53 that we have utilized in this study is the human papillomavirus protein E6. In addition to the numerous methods to hinder p53 activity mentioned above, analyzing p53 function in cells is further complicated by the genetic redundancy inherent to cells in the form of the other members of the p53 gene family, TP63 and TP73 . As a result, loss of p53 activity is difficult to categorize by simple direct genotyping methods of TP53. This can be best witnessed within our cell line panel data in response to Wee1 knockdown. While there certainly appears to be a trend towards TP53 status, there are exceptions to this rule including the U87MG and MCF-7 WT p53 cell lines. To attempt to avoid some of the inherent complexities of the p53 tumor suppressor pathway, we focused our study on cell lines that have well characterized p53 activity in the literature, including the p53 null NCI-H1299 and p53 WT A549 cell lines, as well as the isogenic RKO pair of cell lines. Therefore, we can only conclude from our studies that, in general, there is a susceptibility of p53 inactive cells to respond to Wee1 knockdown.
Previous studies have linked p53 status to cellular sensitivity to loss of Wee1 by siRNA or small molecule inhibitors but always in the presence of cytotoxic agents [8, 10–12, 14]. The proposed hypothesis for this mechanism is that cells with inactive p53 have a defective G1/S checkpoint and therefore can be sensitized to inhibition of the G2/M checkpoint in combination with DNA-damaging agents . However, our results demonstrate that specific loss of Wee1 kinase activity results in inhibition of proliferation and apoptosis in a subset of solid tumor cell lines in the absence of co-treatment with cytotoxic agents. Therefore, a highly specific small molecule inhibitor targeting only the cell cycle suppressor Wee1 should kill p53 inactive tumor cells in the absence of cytotoxic treatment. It should be noted that, while p53 status is difficult to ascertain due to its complexity, most normal tissues would have a functioning p53 tumor suppressor pathway.
Our findings reveal that Wee1 inhibition affects both the G1/S and G2/M transitions through the cell cycle. Wee1 is capable of inactivating both the CDK2-CyclinE and CDK2-CyclinB complexes [6, 7] and the Y15 specific antibody used in this study and others is identical in both CDK1 and CDK2. Therefore, loss of Y15 phosphorylation seen after Wee1 siRNA knockdown may correspond to either or both of these cyclin dependent kinases. We have presented data consistent with the hallmarks of activation of both of these kinases including phospho-Histone H3 upregulation (CDK1) and DNA synthesis marked both by DNA content > 4 N as well as continued incorporation of EdU into cells with higher DNA content (CDK2). It should be noted that we did not observe similar effects upon siRNA knockdown of the other members of the DNA damage response including CHK1/2 and the related CDK regulatory threonine kinase Myt1 (data not shown). It appears therefore that Wee1 acts as the major regulatory kinase in the control of the G1/S and G2/M checkpoints. This is likely because of its direct proximity in the DNA damage response to the CDKs and the importance of the Y15 site over the T14 site phosphorylated by Myt1.
In this study we demonstrate that loss of Wee1 protein results in dramatic cell cycle events, including mitotic catastrophe and mistimed DNA synthesis, ultimately resulting in caspase-dependent apoptosis in a subset of p53 inactive cells. We have achieved this by using siRNA knockdown and provided evidence that this knockdown is specific and kinase dependent through siRNA rescue studies. One compelling finding of this study that will require further analyses is how the subset of p53 active cells are able to accommodate the loss of Wee1 activity and proceed through the cell cycle and proliferate at an equivalent rate to the no siRNA and non-targeting siRNA controls. We postulate loss of Wee1 results in activation of the DNA damage responses and that p53 inactive cells cannot rebalance CDK activities and ultimately have cell cycle mistiming and die via apoptosis. In conclusion, we believe that these data have important ramifications for the treatment of malignancies, and a renewed effort to identify highly specific small molecule inhibitor of Wee1 could provide an unmet clinical need in the treatment of solid tumors.
Small interfering RNA
Non-small cell lung
The authors would like to acknowledge fellow AbbVie employees Julie L. Wilsbacher and Darren C. Phillips for their suggestions and experimental advice and Joel D. Leverson for careful review of the manuscript.
- Malumbres M, Barbacid M: Cell cycle, CDKs and cancer: a changing paradigm. Nat Rev Cancer. 2009, 9 (3): 153-166. 10.1038/nrc2602.View ArticlePubMed
- Hunter T, Pines J: Cyclins and cancer. II: Cyclin D and CDK inhibitors come of age. Cell. 1994, 79 (4): 573-582. 10.1016/0092-8674(94)90543-6.View ArticlePubMed
- Sherr CJ: Cancer cell cycles. Science. 1996, 274 (5293): 1672-1677. 10.1126/science.274.5293.1672.View ArticlePubMed
- Hochegger H, Takeda S, Hunt T: Cyclin-dependent kinases and cell-cycle transitions: does one fit all?. Nat Rev. 2008, 9 (11): 910-916. 10.1038/nrm2510.View Article
- Squire CJ, Dickson JM, Ivanovic I, Baker EN: Structure and inhibition of the human cell cycle checkpoint kinase, Wee1A kinase: an atypical tyrosine kinase with a key role in CDK1 regulation. Structure. 2005, 13 (4): 541-550. 10.1016/j.str.2004.12.017.View ArticlePubMed
- Coleman TR, Dunphy WG: Cdc2 regulatory factors. Curr Opin Cell Biol. 1994, 6 (6): 877-882. 10.1016/0955-0674(94)90060-4.View ArticlePubMed
- Watanabe N, Broome M, Hunter T: Regulation of the human WEE1Hu CDK tyrosine 15-kinase during the cell cycle. Embo J. 1995, 14 (9): 1878-1891.PubMed CentralPubMed
- Hirai H, Iwasawa Y, Okada M, Arai T, Nishibata T, Kobayashi M, Kimura T, Kaneko N, Ohtani J, Yamanaka K, Itadani H, Takahashi-Suzuki I, Fukasawa K, Oki H, Nambu T, Jiang J, Sakai T, Arakawa H, Sakamoto T, Sagara T, Yoshizumi T, Mizuarai S, Kotani H: Small-molecule inhibition of Wee1 kinase by MK-1775 selectively sensitizes p53-deficient tumor cells to DNA-damaging agents. Mol Cancer Ther. 2009, 8 (11): 2992-3000. 10.1158/1535-7163.MCT-09-0463.View ArticlePubMed
- Li J, Wang Y, Sun Y, Lawrence TS: Wild-type TP53 inhibits G(2)-phase checkpoint abrogation and radiosensitization induced by PD0166285, a WEE1 kinase inhibitor. Radiat Res. 2002, 157 (3): 322-330. 10.1667/0033-7587(2002)157[0322:WTTIGP]2.0.CO;2.View ArticlePubMed
- Mir SE, De Witt Hamer PC, Krawczyk PM, Balaj L, Claes A, Niers JM, Van Tilborg AA, Zwinderman AH, Geerts D, Kaspers GJ, Peter Vandertop W, Cloos J, Tannous BA, Wesseling P, Aten JA, Noske DP, Van Noorden CJ, Wurdinger T: In silico analysis of kinase expression identifies WEE1 as a gatekeeper against mitotic catastrophe in glioblastoma. Cancer Cell. 2010, 18 (3): 244-257. 10.1016/j.ccr.2010.08.011.PubMed CentralView ArticlePubMed
- Wang Y, Decker SJ, Sebolt-Leopold J: Knockdown of Chk1, Wee1 and Myt1 by RNA interference abrogates G2 checkpoint and induces apoptosis. Cancer Biol Ther. 2004, 3 (3): 305-313. 10.4161/cbt.3.3.697.View ArticlePubMed
- Wang Y, Li J, Booher RN, Kraker A, Lawrence T, Leopold WR, Sun Y: Radiosensitization of p53 mutant cells by PD0166285, a novel G(2) checkpoint abrogator. Cancer Res. 2001, 61 (22): 8211-8217.PubMed
- Pappano WN, Jung PM, Meulbroek JA, Wang YC, Hubbard RD, Zhang Q, Grudzien MM, Soni NB, Johnson EF, Sheppard GS, Donawho C, Buchanan FG, Davidsen SK, Bell RL, Wang J: Reversal of oncogene transformation and suppression of tumor growth by the novel IGF1R kinase inhibitor A-928605. BMC Cancer. 2009, 9: 314-10.1186/1471-2407-9-314.PubMed CentralView ArticlePubMed
- Beck H, Nahse V, Larsen MS, Groth P, Clancy T, Lees M, Jorgensen M, Helleday T, Syljuasen RG, Sorensen CS: Regulators of cyclin-dependent kinases are crucial for maintaining genome integrity in S phase. J Cell Biol. 2010, 188 (5): 629-638. 10.1083/jcb.200905059.PubMed CentralView ArticlePubMed
- Morgan-Lappe SE, Tucker LA, Huang X, Zhang Q, Sarthy AV, Zakula D, Vernetti L, Schurdak M, Wang J, Fesik SW: Identification of Ras-related nuclear protein, targeting protein for xenopus kinesin-like protein 2, and stearoyl-CoA desaturase 1 as promising cancer targets from an RNAi-based screen. Cancer Res. 2007, 67 (9): 4390-4398. 10.1158/0008-5472.CAN-06-4132.View ArticlePubMed
- Watanabe N, Arai H, Nishihara Y, Taniguchi M, Watanabe N, Hunter T, Osada H: M-phase kinases induce phospho-dependent ubiquitination of somatic Wee1 by SCFbeta-TrCP. Proc Natl Acad Sci U S A. 2004, 101 (13): 4419-4424. 10.1073/pnas.0307700101.PubMed CentralView ArticlePubMed
- The p53 Mutation handbook 2.0. http://p53.free.fr,
- Kessis TD, Slebos RJ, Nelson WG, Kastan MB, Plunkett BS, Han SM, Lorincz AT, Hedrick L, Cho KR: Human papillomavirus 16 E6 expression disrupts the p53-mediated cellular response to DNA damage. Proc Natl Acad Sci U S A. 1993, 90 (9): 3988-3992. 10.1073/pnas.90.9.3988.PubMed CentralView ArticlePubMed
- Cramer LP, Mitchison TJ: Investigation of the mechanism of retraction of the cell margin and rearward flow of nodules during mitotic cell rounding. Mol Biol Cell. 1997, 8 (1): 109-119. 10.1091/mbc.8.1.109.PubMed CentralView ArticlePubMed
- Meek DW: Tumour suppression by p53: a role for the DNA damage response?. Nat Rev Cancer. 2009, 9 (10): 714-723.PubMed
- Brosh R, Rotter V: When mutants gain new powers: news from the mutant p53 field. Nat Rev Cancer. 2009, 9 (10): 701-713.PubMed
- Kaelin WG: The emerging p53 gene family. J Natl Cancer Inst. 1999, 91 (7): 594-598. 10.1093/jnci/91.7.594.View ArticlePubMed
- Anderson HJ, Andersen RJ, Roberge M: Inhibitors of the G2 DNA damage checkpoint and their potential for cancer therapy. Prog Cell Cycle Res. 2003, 5: 423-430.PubMed
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2407/14/430/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/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.