Cell-cycle regulated transcription associates with DNA replication timing in yeast and human
© Fraser; licensee BioMed Central Ltd. 2013
Received: 7 June 2013
Accepted: 7 October 2013
Published: 10 December 2013
Eukaryotic DNA replication follows a specific temporal program, with some genomic regions consistently replicating earlier than others, yet what determines this program is largely unknown. Highly transcribed regions have been observed to replicate in early S-phase in all plant and animal species studied to date, but this relationship is thought to be absent from both budding yeast and fission yeast. No association between cell-cycle regulated transcription and replication timing has been reported for any species.
Here I show that in budding yeast, fission yeast, and human, the genes most highly transcribed during S-phase replicate early, whereas those repressed in S-phase replicate late. Transcription during other cell-cycle phases shows either the opposite correlation with replication timing, or no relation. The relationship is strongest near late-firing origins of replication, which is not consistent with a previously proposed model—that replication timing may affect transcription—and instead suggests a potential mechanism involving the recruitment of limiting replication initiation factors during S-phase.
These results suggest that S-phase transcription may be an important determinant of DNA replication timing across eukaryotes, which may explain the well-established association between transcription and replication timing.
The timing of DNA replication during S-phase of the cell cycle plays an important role in genome integrity, the mutational spectrum, and a wide range of human diseases . Despite many recent advances in our ability to measure the time of replication (Trep) across entire genomes [2–7], our understanding of what regulates this timing remains far from complete [1, 8–11]. The time at which origins of replication (ORIs) fire is thought to be determined in M-phase  or G1 [13, 14], at which point factors such as Cdc45 and Sld3 bind to ORIs that will fire early in the following S-phase [15, 16]. These and several other proteins critical for replication initiation are present at copy-numbers lower than the number of ORIs [17–19], and their over-expression advances Trep for many late-firing ORIs in both budding and fission yeast [12, 17–20], suggesting that their re-use may be a key step in regulating ORI firing time. However, what determines the relative affinities of different ORIs for these limiting factors - and hence their temporal order of initiation - is largely unknown .
Among the strongest correlates (and potential determinants) of Trep in metazoans are transcriptional activity and chromatin state. Although transcriptionally active euchromatin has been known to replicate earlier than repressive heterochromatin for over 50 years [11, 21], the reason - and even the direction of causation - has remained elusive. The two major models [8, 11], not mutually exclusive, are that 1) the euchromatic chromatin structure is more permissive both to transcription and to DNA replication initiation, or 2) Trep itself affects chromatin structure and transcription as a result of changes in the nuclear milieu during S-phase. The former is most directly supported by experiments altering ORI firing time via manipulation of histone modifications [8–10, 18, 22–24], whereas the latter is supported by differences in chromatin and transcription of DNA templates injected into cells during either early or late S-phase [8, 9, 25, 26].
Measuring Trep genome-wide in the budding yeast Saccharomyces cerevisiae (Sc), Raghuraman et al.  reported a surprising lack of association between transcription and Trep (with the exception of the eight histone genes, which are highly transcribed in S phase and are replicated early). However, this analysis only involved clusters of co-expressed genes, and did not actually compare the highest- versus lowest-expressed genes. Nevertheless, it has been widely interpreted in the literature as indicating the absence of any association, and many authors have speculated as to why budding yeast lacks this relationship [5, 8–11]. Similarly, the fission yeast Schizosaccharomyces pombe (Sp) is thought to lack any association between transcription and replication timing , though again no systematic comparison has been reported.
To more systematically visualize these patterns, I calculated the correlation between the expression levels of all cell cycle-regulated genes measured in synchronized cultures [27, 28] with their Trep, separately for each expression data time-point (see Materials and methods). Plotting these correlation coefficients as a function of the time at which the expression data were sampled, I found a striking relationship: both the strength and direction of the correlation oscillate as a function of cell-cycle stage (Figure 1B). In these plots, positive r values represent time-points at which up-regulated genes tend to be replicated late in S phase; negative r values indicate times when up-regulated genes are replicated early. Consistent with the results in Figure 1A, in both species of yeast, genes highly expressed in G2 phase are replicated early, while those expressed in late M/G1 are replicated late. The oscillation is observed regardless of the method used to achieve cell-cycle synchronization (Additional file 1: Figures S1 and S2).
To further characterize this relationship, I plotted a moving average of Trep for the cell cycle-regulated genes in each species, ordered by their time of maximal expression. If expression in certain cell-cycle phases correlates with early or late replication, this should be reflected by troughs or peaks in such a plot. Again in both species a similar trend emerged: Trep reaches a maximum for genes expressed in G1, and a minimum for those expressed in G2 (Figure 1C; Additional file 1: Figure S3), consistent with the correlation analysis (Figure 1B). The strong conservation of this pattern was surprising, considering how much the regulation of DNA replication has diverged in the hundreds of millions of years separating these two yeast lineages .
Although the strongest association between high mRNA levels and early replication was observed for G2-phase expression levels, it is important to note that this does not imply these genes are maximally transcribed in G2. Rather, one would expect maximal transcription to occur in the time leading up to the maximal transcript level, that is, in S phase. Indeed, plotting mRNA levels for G2-upregulated genes (those with early Trep in Figure 1C), it is clear that their transcript levels show the greatest increase - likely reflecting active transcription - in S phase (Additional file 1: Figure S4A). Likewise, genes with late Trep show the opposite pattern: maximal decrease in mRNA levels during S phase (Additional file 1: Figure S4B).
Another factor that may influence the relationship between S-phase transcription and replication timing is a gene’s distance from the nearest ORI. Under the model where chromatin affects both transcription and Trep, the strongest association would be expected for genes near ORIs, whereas if instead Trep affects a gene’s level of S-phase transcription, the relationship should be independent of distance to the nearest ORI . Separating genes into two classes, ORI-proximal or ORI-distal, the ORI-proximal class showed far stronger oscillations (Figure 3C; ORI distance cutoffs, chosen to result in approximately equal-sized lists, were 5 kb from the nearest ORI in Sc and 10 kb in Sp, due to the higher density of known ORIs in Sc; results from equal distance cutoffs are shown in Additional file 1: Figure S5). Because ORI-proximal genes tend to be replicated earlier than ORI-distal genes, this result could not be an indirect effect of the stronger association for late Trep genes, as it acts in the opposite direction. This result suggests that the relationship is unlikely to be caused by an effect of Trep on S-phase transcription, which is one of the two major classes of models that have been proposed to explain the transcription/Trep association [8, 25, 26].
To put into perspective the strength of the relationship between Trep and cell cycle-regulated gene expression in human, I compared it to the well-established association between Trep and average (asynchronous) gene expression. The latter provides a useful benchmark because it is regarded as a strong relationship that has been observed in numerous studies across diverse metazoans [5, 6, 8, 9]. To facilitate a direct comparison with the results in Figure 4, I used the same Trep data  for the same genes, but replaced the cell cycle-synchronized gene expression data  with high-coverage RNA-seq data from asynchronous HeLa cells . The correlation between asynchronous expression and Trep was r = -0.16 for late Trep genes (the genes represented by the red line in Figure 4B) and r = -0.15 for ORI-proximal genes (represented by the blue line in Figure 4C). In both cases, the asynchronous data explained less than a third of the variance in Trep that is explained by S-phase transcription (see Materials and methods). Differing quality of the two gene expression data sets [30, 31] could contribute to this difference; however, because RNA-seq is of far higher precision than spotted cDNA microarrays , any difference would likely underestimate the strength of the cell-cycle oscillations (Figure 4). These results suggest that the relationship between Trep and S-phase transcription in human is substantially stronger than the well-established association with asynchronous expression.
These results suggest that 1) S-phase transcription is associated with DNA replication timing in budding yeast, fission yeast, and human; 2) the association is strongest for genomic regions near ORIs, excluding the causal model in which Trep affects transcription [8, 9, 25, 26]; 3) it is also strongest for regions replicated in late S phase, implying that early-firing ORIs are not affected by this relationship; and 4) this association explains at least three times more of the variability in Trep than the well-known association with (asynchronous) gene expression in human.
Although the replication of these patterns across three species (and across multiple data sets within species; Additional file 1: Figures S1 and S2) lends confidence to their robustness, several caveats should be considered. First, gene expression was represented by transcript abundances, which is a function of both transcription and mRNA decay; therefore, the correlations reported here may underestimate the relationship between transcription and Trep. This prediction can be tested once rates of transcription have been measured throughout the cell cycle. Second, data quality is critical in any analysis; poor-quality data can reduce, or entirely mask, a real relationship. However, in most analyses reported here this is not a major concern, because it could only make the current results conservative (one exception to this is the ORI-proximal versus distal analyses (Figures 3C and 4C): if Trep was measured more accurately near ORIs, this would lead to stronger ORI-proximal correlations, although additional analysis suggests this is not the case (see Materials and methods)). Third, correlation does not imply causation. Although the evidence does not support a model where Trep affects transcription (Figures 3C and 4C), I cannot determine whether transcription itself is affecting Trep, or whether unobserved (latent) factors may be involved. With this caveat in mind, I believe there is still sufficient evidence to propose a testable model to account for these data.
The proposed mechanism likely acts in concert with other factors determining Trep, and thus is not inconsistent with evidence supporting these other factors. For example, although the determination of early versus late-firing ORIs is completed during M/G1 [12–14], S-phase transcription may still influence firing time specifically at late-firing ORIs (Figure 5).
Future work integrating these results with other (non-mutually-exclusive) mechanisms affecting Trep - for example, Forkhead transcription factors  and subnuclear positioning [8, 29, 34, 35] - may lead to a unified framework for understanding the causes, and consequences, of the temporal program of DNA replication across eukaryotes.
Materials and methods
Genome-wide Trep values were downloaded for all three species [2, 4, 6], and mapped onto genes by linear interpolation to the gene’s midpoint. Asynchronous yeast expression levels (used in Figure 2) were taken from [36, 37], using the poly-A data for Sc and the median of wild-type replicates for Sp. Asynchronous HeLa RNA-seq data were from the ENCODE project . Identities of cell cycle-regulated genes, their expression levels, and the cell-cycle phase of each expression time-point were acquired from [27, 28, 30]. All cell-cycle expression data were measured as mRNA levels relative to asynchronous levels of each gene, as opposed to absolute mRNA abundances that can be measured by RNA-seq; therefore, these expression levels represent the relative induction or repression of each gene throughout the cell cycle. The order of maximum expression levels was obtained from  for Sp and  for Sc. ORI locations were downloaded from ORIdb  for both yeasts (using only 'confirmed' or 'likely' ORIs), and from  for human (see below).
Yeast data analysis
All correlations were Pearson’s (significance cutoffs given in each figure legend). Trep moving averages (Figure 1C) were calculated for windows of 100 genes for Sc and 60 genes for Sp (due to the smaller number of cell cycle-regulated genes in Sp). For Figures 1A and 3A, the G2 expression data were represented by the 42 minute time-point for Sc and 135 minute time-point for Sp; for Figure 1A, Sc M/G1 was represented by the 70 minute time-point, and Sp G1 was represented by the 225 minute time-point. For Figure 3B, the early/late S-phase cutoff was chosen at halfway through S phase of each Trep data set (39.6 minutes after release from hydroxyurea arrest in Sp, and 26.8 minutes after release from cdc7 arrest in Sc). The cutoff for ORI-proximal versus ORI-distal (5 kb from each gene’s 5' end in Sc and 10 kb in Sp) was chosen in each yeast to result in gene lists of approximately equal size.
P-values in Figure 2 were calculated with a two-tailed Student’s t-test. Because the Sc expression levels were calculated as a ratio of mRNA/genomic DNA from asynchronous cells , they represent the number of mRNAs per DNA copy, and thus account for the fact that genes with early Trep spend a greater portion of the cell cycle with two copies. Although the Sp expression data  do not account for this, correcting for the effect by subtracting a fraction of each expression level proportional to the time each gene spends with two copies had only a minimal effect.
All code and data are available at .
Human data analysis
Human ORIs were defined as Orc1 binding sites  located within 1 mb of early-replicating peaks in the HeLa Trep profile, which indicate active ORIs (this window size was necessitated by the low resolution of the Trep profile) . The early/late Trep cutoff was the first 50% of S phase and the ORI-proximal/distal cutoff was 10 kb from each gene’s 5' end. Due to the higher number of expression data points per cell cycle in human (approximately 15 in human versus approximately 9 for both yeasts), a two-point moving average was used for plotting human correlation coefficients.
To compare asynchronous expression versus S-phase transcription in HeLa cells, I compared high-coverage RNA-seq data from HeLa cells  with Trep for the same genes analyzed in Figure 4B,C. The fraction of variance in Trep explained by the expression data is simply the r2 value from the Pearson’s correlation. Comparing these values for the asynchronous data with the strongest G2-phase (used to represent S-phase transcription, as described above) correlations, among the late-replicating genes (represented by the red line in Figure 4B) 2.7% of the variance in Trep was explained by the asynchronous data, versus 8.1% for S-phase transcription. Likewise for ORI-proximal genes (represented by the blue line in Figure 4C), the asynchronous data explained 2.3% of the variance in Trep, versus 7.6% for S-phase transcription.
To determine whether Trep is measured with greater accuracy near ORIs, I compared the Trep data used in Figure 4 with an independent Trep data set from HeLa cells . Restricting the analysis to the cell cycle-regulated genes analyzed in Figure 4C, I found that ORI-distal genes actually showed better agreement between Trep data sets than did ORI-proximal genes (r = 0.59 and 0.46, respectively). This implies that, if anything, Trep is measured less accurately in ORI-proximal regions, which would lead to an underestimate of the strength of the oscillating correlation (blue line in Figure 4C).
Origin of replication
Time of replication.
I would like to thank M Botchan, N Cozzarelli, A Donaldson, D Gilbert, M Kobor, and J Rine for helpful advice. This work was supported by NIH grant 1R01GM097171-01A1. HBF is an Alfred P Sloan Fellow and a Pew Scholar in the Biomedical Sciences.
- Watanabe Y, Maekawa M: Spatiotemporal regulation of DNA replication in the human genome and its association with genomic instability and disease. Curr Med Chem. 2010, 17: 222-View ArticlePubMedGoogle Scholar
- Raghuraman MK, Winzeler EA, Collingwood D, Hunt S, Wodicka L, Conway A, Lockhart DJ, Davis RW, Brewer BJ, Fangman WL: Replication dynamics of the yeast genome. Science. 2001, 294: 115-View ArticlePubMedGoogle Scholar
- Farkash-Amar S, Lipson D, Polten A, Goren A, Helmstetter C, Yakhini Z, Simon I: Global organization of replication time zones of the mouse genome. Genome Res. 2008, 18: 1562-View ArticlePubMedPubMed CentralGoogle Scholar
- Eshaghi M, Karuturi RK, Li J, Chu Z, Liu ET, Liu J: Global profiling of DNA replication timing and efficiency reveals that efficient replication/firing occurs late during S-Phase in S. pombe. PLoS ONE. 2007, 2: e722-View ArticlePubMedPubMed CentralGoogle Scholar
- Schübeler D, Scalzo D, Kooperberg C, van Steensel B, Delrow J, Groudine M: Genome-wide DNA replication profile for Drosophila melanogaster: a link between transcription and replication timing. Nat Genet. 2002, 32: 438-View ArticlePubMedGoogle Scholar
- Dellino GI, Cittaro D, Piccioni R, Luzi L, Banfi S, Segalla S, Cesaroni M, Mendoza-Maldonado R, Giacca M, Pelicci PG: Genome-wide mapping of human DNA-replication origins: Levels of transcription at ORC1 sites regulate origin selection and replication timing. Genome Res. 2013, 23: 1-View ArticlePubMedPubMed CentralGoogle Scholar
- Hiratani I, Ryba T, Itoh M, Yokochi T, Schwaiger M, Chang CW, Lyou Y, Townes TM, Schübeler D, Gilbert DM: Global reorganization of replication domains during embryonic stem cell differentiation. PLoS Biol. 2008, 6: e245-View ArticlePubMedPubMed CentralGoogle Scholar
- Hiratani I, Takebayashi S, Lu J, Gilbert DM: Replication timing and transcriptional control: beyond cause and effect - part II. Curr Opin Genet Dev. 2009, 19: 142-View ArticlePubMedPubMed CentralGoogle Scholar
- Donaldson AD: Shaping time: chromatin structure and the DNA replication programme. Trends Genet. 2005, 21: 444-View ArticlePubMedGoogle Scholar
- McCune HJ, Donaldson AD: DNA replication: telling time with microarrays. Genome Biol. 2003, 4: 204-View ArticlePubMedPubMed CentralGoogle Scholar
- Gilbert DM: Replication timing and transcriptional control: beyond cause and effect. Curr Opin Cell Biol. 2002, 14: 377-View ArticlePubMedGoogle Scholar
- Wu PYJ, Nurse P: Establishing the program of origin firing during S phase in fission yeast. Cell. 2009, 136: 852-View ArticlePubMedPubMed CentralGoogle Scholar
- Raghuraman MK, Brewer BJ, Fangman WL: Cell cycle–dependent establishment of a late replication program. Science. 1997, 276: 806-View ArticlePubMedGoogle Scholar
- Dimitrova DS, Gilbert DM: The spatial position and replication timing of chromosomal domains are both established in early G1 phase. Mol Cell. 1999, 4: 983-View ArticlePubMedGoogle Scholar
- Kamimura Y, Tak YS, Sugino A, Araki H: Sld3, which interacts with Cdc45 (Sld4), functions for chromosomal DNA replication in Saccharomyces cerevisiae. EMBO J. 2001, 20: 2097-View ArticlePubMedPubMed CentralGoogle Scholar
- Aparicio OM, Stout AM, Bell SP: Differential assembly of Cdc45p and DNA polymerases at early and late origins of DNA replication. Proc Natl Acad Sci U S A. 1999, 96: 9130-View ArticlePubMedPubMed CentralGoogle Scholar
- Tanaka S, Nakato R, Katou Y, Shirahige K, Araki H: Origin association of Sld3, Sld7, and Cdc45 proteins is a key step for determination of origin-firing timing. Curr Biol. 2011, 21: 2055-View ArticlePubMedGoogle Scholar
- Mantiero D, Mackenzie A, Donaldson A, Zegerman P: Limiting replication initiation factors execute the temporal programme of origin firing in budding yeast. EMBO J. 2011, 30: 4805-View ArticlePubMedPubMed CentralGoogle Scholar
- Douglas ME, Diffley JFX: Replication timing: the early bird catches the worm. Curr Biol. 2012, 22: R81-View ArticlePubMedGoogle Scholar
- Patel PK, Kommajosyula N, Rosebrock A, Bensimon A, Leatherwood J, Bechhoefer J, Rhind N: The Hsk1(Cdc7) replication kinase regulates origin efficiency. Mol Biol Cell. 2008, 19: 5550-View ArticlePubMedPubMed CentralGoogle Scholar
- Lima de Faria A: Incorporation of tritiated thymidine into meiotic chromosomes. Science. 1959, 130: 503-View ArticlePubMedGoogle Scholar
- Vogelauer M, Rubbi L, Lucas I, Brewer BJ, Grunstein M: Histone acetylation regulates the time of replication origin firing. Mol Cell. 2002, 10: 1223-View ArticlePubMedGoogle Scholar
- Goren A, Tabib A, Hecht M, Cedar H: DNA replication timing of the human beta-globin domain is controlled by histone modification at the origin. Genes Dev. 2008, 22: 1319-View ArticlePubMedPubMed CentralGoogle Scholar
- Knott SR, Viggiani CJ, Tavaré S, Aparicio OM: Genome-wide replication profiles indicate an expansive role for Rpd3L in regulating replication initiation timing or efficiency, and reveal genomic loci of Rpd3 function in Saccharomyces cerevisiae. Genes Dev. 2009, 23: 1077-1090.View ArticlePubMedPubMed CentralGoogle Scholar
- Lande-Diner L, Zhang J, Cedar H: Shifts in replication timing actively affect histone acetylation during nucleosome reassembly. Mol Cell. 2009, 34: 767-View ArticlePubMedPubMed CentralGoogle Scholar
- Zhang J, Xu F, Hashimshony T, Keshet I, Cedar H: Establishment of transcriptional competence in early and late S phase. Nature. 2002, 420: 198-View ArticlePubMedGoogle Scholar
- Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen MB, Brown PO, Botstein D, Futcher B: Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization. Mol Biol Cell. 1998, 9: 3273-View ArticlePubMedPubMed CentralGoogle Scholar
- Rustici G, Mata J, Kivinen K, Lió P, Penkett CJ, Burns G, Hayles J, Brazma A, Nurse P, Bähler J: Periodic gene expression program of the fission yeast cell cycle. Nat Genet. 2004, 36: 809-View ArticlePubMedGoogle Scholar
- Natsume T, Tanaka TU: Spatial regulation and organization of DNA replication within the nucleus. Chromosome Res. 2010, 18: 7-View ArticlePubMedGoogle Scholar
- Whitfield ML, Sherlock G, Saldanha AJ, Murray JI, Ball CA, Alexander KE, Matese JC, Perou CM, Hurt MM, Brown PO, Botstein D: Identification of genes periodically expressed in the human cell cycle and their expression in tumors. Mol Biol Cell. 2002, 13: 1977-View ArticlePubMedPubMed CentralGoogle Scholar
- Djebali S, Davis CA, Merkel A, Dobin A, Lassmann T, Mortazavi A, Tanzer A, Lagarde J, Lin W, Schlesinger F, Xue C, Marinov GK, Khatun J, Williams BA, Zaleski C, Rozowsky J, Röder M, Kokocinski F, Abdelhamid RF, Alioto T, Antoshechkin I, Baer MT, Bar NS, Batut P, Bell K, Bell I, Chakrabortty S, Chen X, Chrast J, Curado J, et al: Landscape of transcription in human cells. Nature. 2012, 489: 101-108.View ArticlePubMedPubMed CentralGoogle Scholar
- Wang Z, Gerstein M, Snyder M: RNA-Seq: a revolutionary tool for transcriptomics. Nat Rev Genet. 2009, 10: 57-63.View ArticlePubMedPubMed CentralGoogle Scholar
- Knott SR, Peace JM, Ostrow AZ, Gan Y, Rex AE, Viggiani CJ, Tavaré S, Aparicio OM: Forkhead transcription factors establish origin timing and long-range clustering in S. cerevisiae. Cell. 2012, 148: 99-111.View ArticlePubMedPubMed CentralGoogle Scholar
- Yaffe E, Farkash-Amar S, Polten A, Yakhini Z, Tanay A, Simon I: Comparative analysis of DNA replication timing reveals conserved large-scale chromosomal architecture. PLoS Genet. 2010, 6: e1001011-View ArticlePubMedPubMed CentralGoogle Scholar
- Ryba T, Hiratani I, Lu J, Itoh M, Kulik M, Zhang J, Schulz TC, Robins AJ, Dalton S, Gilbert DM: Evolutionarily conserved replication timing profiles predict long-range chromatin interactions and distinguish closely related cell types. Genome Res. 2010, 20: 761-770.View ArticlePubMedPubMed CentralGoogle Scholar
- Wells ML, Huang W, Li L, Gerrish KE, Fargo DC, Ozsolak F, Blackshear PJ: Posttranscriptional regulation of cell-cell interaction protein-encoding transcripts by Zfs1p in Schizosaccharomyces pombe. Mol Cell Biol. 2012, 32: 4206-View ArticlePubMedPubMed CentralGoogle Scholar
- Wang Y, Liu CL, Storey JD, Tibshirani RJ, Herschlag D, Brown PO: Precision and functional specificity in mRNA decay. Proc Natl Acad Sci U S A. 2002, 99: 5860-View ArticlePubMedPubMed CentralGoogle Scholar
- Gauthier NP, Jensen LJ, Wernersson R, Brunak S, Jensen TS: Cyclebase.org: version 2.0, an updated comprehensive, multi-species repository of cell cycle experiments and derived analysis results. Nucleic Acids Res. 2010, 38: D699-View ArticlePubMedGoogle Scholar
- Siow CC, Nieduszynska SR, Müller CA, Nieduszynski CA: OriDB, the DNA replication origin database updated and extended. Nucleic Acids Res. 2012, 40: D682-View ArticlePubMedGoogle Scholar
- txn-rep.zip. [http://www.stanford.edu/group/fraserlab/txn-rep]
- Chen CL, Rappailles A, Duquenne L, Huvet M, Guilbaud G, Farinelli L, Audit B, D'Aubenton-Carafa Y, Arneodo A, Hyrien O, Thermes C: Impact of replication timing on non-CpG and CpG substitution rates in mammalian genomes. Genome Res. 2010, 20: 447-457.View ArticlePubMedPubMed CentralGoogle 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 cited.