|Home | About | Journals | Submit | Contact Us | Français|
The extent to which the DNA relaxation activities of eukaryotic topoisomerases (topo I and topo II) are redundant during gene transcription is unclear. Although both enzymes can often substitute for each other in vivo, studies in vitro had revealed that the DNA cross-inversion mechanism of topo II relaxes chromatin more efficiently than the DNA strand-rotation mechanism of topo I. Here, we show that the inactivation of topo II in budding yeast produces an abrupt decrease of virtually all polyA+ RNA transcripts of length above ~3kb, irrespective of their function. This reduction is not related to transcription initiation but to the stall of RNA polymerase II (Pol II) during elongation. This reduction does not occur in topo I mutants; and it is not avoided by overproducing yeast topo I or bacterial topo I, which relaxes (−) DNA supercoils. It is rescued by catalytically active topo II or a GyrBA enzyme, which relaxes (+) DNA supercoils. These findings demonstrate that DNA relaxation activities of topo I and topo II are not interchangeable in vivo. Apparently, only topo II relaxes efficiently the (+) DNA supercoils that stall the advancement of Pol II in long genes. A mechanistic model is proposed.
During transcriptional elongation, the DNA double helix rotates axially relative to the RNA polymerase. At the chromosomal level, this rotation is hindered by the viscous drag of chromatin fibers and by interactions that anchor DNA and the transcription ensembles to subnuclear structures (1). Consequently, (+) torsional stress overwinds and positively supercoils the DNA ahead of the transcribing RNA polymerase; and (−) torsional stress unwinds and negatively supercoils the DNA behind it (2). In eukaryotic cells, DNA torsional stress is relaxed by topoisomerase I (topo I) and topoisomerase II (topo II) (3,4). Topo I uses the Type-1B mechanism, in which the enzyme temporarily cleaves one strand of the duplex allowing the DNA to rotate around the uncleaved strand (5). Topo II uses the Type-2 mechanism, in which the enzyme removes DNA supercoil crossings by transporting one segment of duplex DNA through a transient double-strand break in another segment (5).
Numerous studies had indicated that DNA relaxation activities of topo I and topo II can substitute each other to support DNA transcription. In particular, relaxation of (+) torsional stress by either topo I or topo II is sufficient to allow the progression of transcribing RNA polymerases (6–8). Accordingly, yeast Δtop1 single mutants are viable and show no major defects in gene expression (9–11). Similarly, inactivation of topo II does not preclude RNA synthesis in Saccharomyces cerevisiae (11), in Schizosaccharomyces pombe (12) or in human cells (13). Only when both topoisomerases are defective in yeast Δtop1 top2-ts double mutants, global RNA synthesis is reduced (9,14,15).
Other studies, however, have suggested that the DNA relaxation mechanisms of topo I and topo II may not be interchangeable. Yeast minichromosomes with a high (+) DNA torsional stress are efficiently relaxed by topo II but not by topo I (16). Likewise, a recent study indicated that topo I and topo II are prone to relax, respectively, the (−) and (+) torsional constrains produced by the high transcription rate of yeast rDNA genes (17). These distinct preferences of topo I and II had been explained in terms of how torsional stress affects twist (double helical winding) and writhe (supercoiling) deformations of chromatinized DNA (18–20). Chromatin under (+) torsional stress would mainly deform DNA by writhe, which obstructs the strand-rotation mechanism of topo I but facilitates the DNA cross-inversion mechanism of topo II (16). Conversely, chromatin under (−) torsional stress would promote mainly DNA untwisting, which is efficiently relaxed by topo I (17).
Despite the above observations, however, there is no clear evidence that topo I and topo II play distinctive DNA relaxation roles during general transcription (i.e. in Pol II transcribed genes). One reason is that, in addition to transcriptional elongation, topoisomerases are implicated in the regulation of gene expression at multiple stages. Topo I controls transcription initiation in a subset of yeast genes (10) and facilitates nucleosome disassembly at gene promoters (12). Topo II modulates chromosome architecture and long-range chromatin structure (21); and mainly interacts with gene promoter regions in yeast (11), where it binds nucleosome free DNA and acts redundantly with topo I to enhance recruitment of Pol II (15). Hence, the molecular mechanisms which lead to alterations of DNA transcription on inhibition of cellular topoisomerases are complex; and ascertaining in which steps topo I and topo II are redundant or play specific roles is difficult. Here, we show that the sole inactivation of topo II in budding yeast produces an abrupt decrease of all Pol II transcripts of length above ~3kb, which is caused by a stalling of Pol II during elongation. Our experiments indicate that this length dependent effect is consequent to the unique capability of topo II, but not topo I, to relax the (+) supercoiled chromatin that accumulates after transcribing a critical DNA length. These findings demonstrate that the DNA relaxation activities of topo I and topo II are not redundant during Pol II transcription in vivo. We propose a model to explain why normal production of long gene transcripts requires the topo II mechanism.
Saccharomyces cerevisiae strains JCW25 (TOP1 TOP2), JCW26 (TOP1 top2–4) and JCW28 (Δtop1 top2–4) are derivatives of FY251 and had been previously described (22). Yeast strains were grown at 26°C to logarithmic phase in YPD medium containing 2% glucose. At time 0min, aliquots of the cultures were taken. The rest of the cultures were placed at 37°C and samples were taken after 120min. Cells were pelleted and immediately stored at −80°C.
Yeast cells were washed twice by centrifugation (5000g at 4°C) in MilliQ water. Total RNA was extracted with the RiboPure Yeast kit (Ambion, Austin, TX, USA) and treated with DNase I (F. Hoffmann-La Roche, Basel, Switzerland) to remove contaminating genomic DNA. Resulting RNA was quantified by spectrophotometry in a NanoDrop ND-1000 (NanoDrop Technologies, Wilmintong, DE, USA) and its integrity checked by gel electrophoresis. Purified RNA was aliquoted and kept at −80°C.
Fifteen micrograms of total RNA were used for cDNA synthesis and labeling with Cy3-dUTP and Cy5-dUTP fluorescent nucleotides, following indirect labeling protocol (CyScribe post-labeling kit, GE-Healthcare, New York, NY, USA). Labeling efficiency was evaluated by measuring Cy3 or Cy5 absorbance in a Nanodrop Spectrophotometer. Microarrays encompassing the complete set of 6306 ORFs coded by the S. cerevisiae genome, printed with a duplicated set of synthetic oligonucleotides (70-mer, Yeast Genome Oligo Set, OPERON, Cologne, Germany), were provided by the Genomics Unit of the Scientific Park of Madrid (Spain). Microarray prehybridization was performed in 5× SSC (SSC: 150mM NaCl, 15mM Na-citrate, pH 7.0), 0.1% SDS, 1% BSA at 42°C for 45min. (Fluka, Sigma-Aldrich, Buchs SG, Switzerland). Labeled cDNA was dried in a vacuum trap and used as a probe after resuspension in 110µl of hybridization solution (50% Formamide, 5× SSC, 0.1% SDS, 100µg/ml salmon sperm from Invitrogen, Carlsbad, CA, USA). Hybridization and washing were performed in a Lucidea Slide Pro System (GE Healthcare, Uppsala, Sweden). Arrays were scanned with a GenePix 4000B fluorescence scanner and analyzed by Genepix® Pro 6.0 software (Axon Instruments, MDS Analytical Technologies, Toronto, Canada). Data was filtered according to spot quality. Only those spots with intensities at least twice the background signal and with at least 75% of pixels with intensities above background plus 2 SDs were selected for further calculations. Following these criteria, over 50% of spots in each array were usually found suitable for further analysis. Genomic data sets for transcript abundance analyses are stored in the GEO databases (http://www.ncbi.nlm.nih.gov/geo) with accession number GSE18242.
Microarray data were analyzed by comparing pairs of a mutant versus reference TOP1 TOP2 strain. Results for each gene were given as a ratio of pixel intensities (ratio of medians of the mutant strain divided by the reference strain). Ratios were normalized within the Genepix® Pro 6.0 software. The experimental design provided for each condition (0, 120min) up to six determinations for each gene (three biological replicates and two replicated spots). Those genes for which a minimum of 12 (out of 18) data values passed the microarray quality standards were considered for statistical analyses. Data were calculated as binary logarithms (log2) of fluorescence ratios. Significant changes of expression values between the starting point and 120min after the temperature shift were determined by one-way ANOVA (P<10−3).
An aliquot of RNA used in the microarray experiments was reserved for qRT-PCR follow-up studies. First strand cDNA was synthesized from 2µg of total DNAseI-treated RNA in a 20-µl reaction volume using Omniscript RT Kit (Qiagen, Valencia, CA, USA) following manufacture’s instructions. qRT-PCR reactions were conducted in triplicate using the SYBR Green PCR Master Mix (Applied Biosystems) and the ABI-PRISM 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Gene-specific primers were designed using Primer Express software (Applied Biosystems). Amplified fragments were confirmed by sequencing in a 3730 DNA Analyzer (Applied Biosystems) and comparison with the published genomic data at SGD. Real time PCR conditions included an initial denaturation step at 95°C for 10min, followed by 40 cycles of a two-step amplification protocol: denaturation at 95°C for 15s and annealing/extension at 60°C for 1min.
Yeast transcription run-on, array hybridization and data analyses were done as previously described (23) with minor modifications. Around 5×108 cells were washed in cold water and permeabilized with 0.5% N-lauryl sarcosine sodium sulfate. In vivo transcription was conducted by incubating cells during 5min at 30°C in 250ml of buffer T (20mM Tris–HCl pH 7.7, 200mM KCl, 50mM MgCl2, 2mM DTT) plus 4mM each of CTP, ATP, GTP and 12ml of 33P-UTP (3000Ci/mmol, 10mCi/ml). In vivo labeled RNA was purified and hybridized (0.2–2×107dpm/5ml) on array membranes. Genomic DNA was random priming labeled and hybridized on the array membranes of the same lot. Genomic data sets for transcription run-on analyses are stored in the GEO databases (http://www.ncbi.nlm.nih.gov/geo) with accession number GSE16673.
ChIP experiments for RNA polymerase II were performed as previously described (24) with minor modifications. Briefly, 50ml of yeast culture were collected at OD600 0.5. Crosslinking was performed by adding formaldehyde to 1% to the culture and by incubating at room temperature for 15min. Then, 2.5ml 2.5M glycine were added and the culture was incubated for 5min. Cells were then harvested and washed four times with 25ml Tris–saline buffer (150mM NaCl 20mM Tris–HCl pH 7.5) at 4°C. Cell breakage was performed in 300ml of lysis buffer (see the above reference) with glass beads, and cell extracts were sonicated in a Bioruptor sonicator (Diagenode) for 30min in 30s on/30s off cycles (chromatin is sheared into an average size of 300bp). Immunoprecipitation was performed with magnetic beads coated with pan anti-IgG antibodies (Dynal), which were incubated before with anti-RNA polymerase II monoclonal antibody 8WG16. Finally, qRT-PCR were performed to quantify immunoprecipitation using a dilution of 1:1500 for the input samples and another of 1:10 for the immunoprecipitated samples. The primers used are listed in Supplementary Table S3.
Topoisomerase expression plasmids YEpTOP2-GAL1, pRK-G1T1 and YEptopA-GAL1, respectively, carry the genes to produce S. cerevisiae topo II, S. cerevisiae topo I and Escherichia coli topo I under the galactose inducible Gal1 promoter (25). Plasmid YEpTOP2 (Y782F)-GAL1 is a derivate of YEpTOP2-GAL1 and induces the expression of a catalytically inactive topo II (26). Plasmid pSTS77 induces the expression of a GyrB-GyrA fusion protein that is catalytically active in S. cerevisiae (27). Control plasmid was Yep24 (NEB). Extrachromosomal expression of topoisomerases was induced during 12h by growing the yeast transformants in selective media containing 1.5% glucose 0.5% galactose for TOP2 genes, 1% glucose 1% galactose for the TOP1 gene and 2% galactose for the TopA gene and the GyrB-GyrB fusion protein.
Total RNA was extracted from isogenic TOP2 and top2-ts S. cerevisiae strains, before and after shifting the cells to the non-permissive temperature, to inactivate topo II in the mutant strain (37°C). Microarray comparison of polyA+ RNA levels between both strains indicated that topo II inactivation results in relatively minor changes in the yeast transcriptome. After 120min at 37°C, the relative abundance of 176 transcripts increased and 195 decreased to significant levels (Supplementary Table S1). Gene ontology analysis showed that a large number (29%) of increased transcripts corresponded to genes of ribonucleoproteins involved in ribosome biogenesis, whereas a higher proportion of reduced transcripts (39%) corresponded to genes involved in oxidative metabolism in the mitochondrion (Figure 1A and Supplementary Table S1). Independently from the functional effects observed after the inactivation of topo II, we noticed that nearly all transcripts >3kb (about 500, representing nearly 9% of the yeast genome) were reduced irrespective of their function (Figure 1A, red diamonds). This drop in abundance of long transcripts was unrelated to temperature changes, as the same transcripts appear essentially unchanged after both 25 to 37°C and 37 to 25°C thermal shifts (28). The decrease of long transcripts abundance was also unrelated to the expression level of their corresponding genes. Both highly and poorly transcribed genes were similarly influenced by the length dependent effect of topo II inactivation (Figure 1B, red diamonds).
A detailed analysis of transcript level variation according to the gene size revealed that average abundance values do not change for transcripts <2.5kb, an incipient reduction appears for sizes >2.5kb and an abrupt decrease occurs when transcripts are >3.5kb (Figure 2A). This length dependence effect was validated by qRT-PCR. Transcript variation of 16 genes of different sizes (Supplementary Table S2) showed a good correlation with the microarray data (Figure 2B); thus confirming that topo II inactivation selectively decreases the production of long gene transcripts.
We next asked if a similar dependence between transcript abundance and their physical length occurred in other topoisomerase mutants. We examined the transcriptome alterations reported for Δtop1 single mutants (10,29) and for Δtop1 top2-ts double mutants (15,29). The alterations of transcript levels in Δtop1 single mutants were clearly independent of gene size (Figure 2C). In the case of Δtop1 top2-ts double mutant, simultaneous inactivation of topo I and topo II produced a global decrease of transcripts abundance relative to the control cells, in agreement with that observed in previous studies (9,14,15) (Figure 2D). However, this general reduction did not show significant dependence on the transcript length as compared to the selective reduction of long transcripts produced on inactivation of topo II in the top2-ts single mutant (Figure 2E).
Next, we asked if the reduction of long transcripts abundance caused by topo II inactivation was a consequence of a decrease in their transcription initiation, which is exercised mainly at the RNA polymerase recruitment step and can be determined by transcription run-on (23,30). Therefore, we conducted a genomic transcription run-on (GRO) to examine the variation of transcription rates produced in the top2-ts relative to TOP2 cells. After shifting the yeast cultures to the non-permissive temperature (37°C), yeast cells were permeabilized in a cold sarkosyl buffer that disrupts most chromatin proteins but not the transcribing RNA polymerases. Radioactive UTP was subsequently added to the cells and became incorporated into nascent RNA molecules, which can elongate up to 300nt in these conditions (23,30). The resulting radiolabeled RNA was extracted and analyzed by hybridization on genomic macro-arrays to provide a measure of the transcription rate of each gene at the moment of cell sampling. The GRO results clearly indicated that genome-wide alterations of transcription initiation produced by the inactivation of topo II are independent of gene size (Figure 3). Therefore, the decrease of long transcript abundance caused by topo II deficiency was not attributable to a failure of RNA polymerase recruitment and initiation of transcription.
Following the above, we examined if the reduction of long transcripts abundance was caused by an impairment of transcription elongation. To do so, we carried out chromatin immunoprecipitation (ChIP) to examine the intragenic distribution of RNA polymerase II (Pol II) in a subset of genes ranging between 1 and 8kb in length (Supplementary Table S3). Pol II distribution in the top2-ts mutant relative to TOP2 strain after shifting the cells to 37°C produced a similar transcript length dependent pattern in all cases (Figure 4). Irrespective of gene size, inactivation of topo II did not significantly change Pol II allocation up to a 2-kb distance from the transcription start site (TSS). However, as seen in long genes, Pol II markedly accumulated in the intragenic segments between 2 and 3kb from the TSS; and became depleted in the distal segment beyond 4kb from the TSS. These results were consistent with the above GRO experiments indicating that topo II inactivation does not alter transcription initiation of long genes; and clearly indicated that the selective reduction of their transcripts was due to an impairment of the progression of Pol II after transcribing between 2 and 3kb from the TSS.
Finally, we examined if the reduction of long transcript abundance observed in the top2-ts single mutant was solely due to the inactivation of topo II and dependent on its DNA relaxation activity (Figure 5). To do this, we transformed the TOP2 and the top2-ts yeast strains with plasmids Yep24 (control); YEpTOP2-GAL1, which allows ectopic expression of yeast TOP2 gene under the inducible Gal1 promoter; or its derivative YEpTOP2 (Y782F)-GAL1, which induces the expression of a mutated topo II enzyme that lacks the active-site tyrosine required for catalytic activity (26). On induction of the ectopic TOP2 gene, the reduction of long transcript abundance did not occur in the top2-ts cells when shifted to 37°C (Figure 5B). However, the reduction persisted when we induced the topo II mutant instead, thus confirming the requirement of catalytically active topo II as the only cause of the length dependent effect (Figure 5C).
We then conducted analogous experiments to examine if the reduction of long transcripts caused by the lack of topo II activity could be compensated by the ectopic expression of other topoisomerases. We transformed the TOP2 and the top2-ts mutant strains with plasmid pRK-G1T1 to overproduce S. cerevisiae topo I; with plasmid YEptopA-GAL1 to produce E. coli topo I (23) and with plasmid pSTS77 to produce an E. coli GyrB-GyrA fusion protein (27). As illustrated in Figure 5, the yeast topo I uses a DNA strand-rotation mechanism to relax (+) and (−) torsional stress (3,5). The bacterial topo I uses a DNA single-strand-passage mechanism on unwound regions of the duplex and relaxes (−) torsional stress when it is produced in yeast cells (25). The GyrB-GyrA fusion protein (GyrBA) uses a DNA cross-inversion mechanism like topo II and relaxes (+) torsional stress when it is produced in yeast (27). These compensation experiments showed that the ectopic production of yeast topo I (Figure 5D) and bacterial topo I (Figure 5E) in the top2-ts cells did not prevent the reduction of long transcript abundance on inactivation of topo II. However, the expression of the GyrBA enzyme was able to compensate the absence of topo II activity (Figure 5F). Therefore, the incapacity to relax (+) supercoils by topo II, but not topo I, was the more likely cause of the stall of Pol II and consequent reduction of long gene transcripts.
Whether the DNA relaxation activities of eukaryotic topo I and topo II are redundant or have distinctive roles during genome transactions is a long-standing question. In this study, we show that the sole inactivation of topo II in S. cerevisiae produces a decrease of virtually all Pol II transcripts of length above ~3kb, irrespective of their function. This selective reduction is not due to an alteration of transcription initiation, but to an impairment of the RNA polymerase during elongation. Therefore, proper production of this subset of transcripts demands a specific role of topo II, which cannot be compensated for by topo I.
Any dependence of transcript abundance on transcript length can be reasoned by the alleged role of topoisomerases in removing the DNA torsional stress generated during transcription elongation (3,4). In vitro studies have shown that the failure to relax (+) torsional stress stalls the progression of transcribing polymerases (6,7). The same has been concluded from recent in vivo work, which showed that convergent transcription is impaired under topo I and topo II inactivation as a function of supercoiling accumulation (8). Hence, a dependence on transcript length is expected in cells deficient in topoisomerase activity, since the longer the transcribed distance the more (+) torsional stress will be accumulated. Strikingly, we do not observe such length dependence in Δtop1 top2-ts double mutants of S. cerevisiae. Inactivation of both topoisomerases causes instead a general reduction of transcript abundance, which probably reflects a global down-regulation of transcription by alterations in both the initiation and elongation steps, irrespective of the gene size (12,15,29). Our observation is consistent with a recently reported intragenic distribution of Pol II in Δtop1 top2-ts double mutants of S. cerevisiae, which showed no significant length dependence alterations relative to wild-type strains (15). Similar results have been found for Δtop1 top2-ts double mutants of S. pombe; in which only a slight accumulation of Pol II was observed in coding regions of genes (12).
Contrasting with the above, the transcript length dependence uncovered here is prominent and occurs specifically in top2-ts single mutants. No such dependence is observed in Δtop1 single mutants. The inactivation of topo II does not affect the transcription frequency of long genes. It produces a stall in the advancement of Pol II after transcribing beyond 2kb, which suitably explains the selective decrease of long transcript abundance. The transcript reduction is not rescued by overproducing yeast topo I or bacterial topo I, which relaxes (−) DNA supercoils. It is rescued by catalytically active topo II or a GyrBA enzyme, which relaxes (+) DNA supercoils. Therefore, the more likely cause of the stall of Pol II and subsequent reduction of long transcripts is the incapacity of the top2 cells to relax the (+) DNA supercoils accumulated during transcription elongation. This finding substantiates the biological relevance of what we had observed in previous in vitro studies: (+) supercoiled chromatin is efficiently relaxed by the DNA cross-inversion mechanism of topo II, but not by the DNA strand-rotation mechanism of topo I (16).
An intriguing aspect of the transcript length dependence exposed here is that, rather than being gradual, it has a sharp inflection point at ~3kb. Up to this critical length, (+) torsional stress can be either relaxed by topo I alone or it is not high enough to delay the normal progression of Pol II. Beyond this length, topo I is no longer efficient and topo II is then required to avoid the stalling of Pol II. We believe that this inflection reveals how increasing levels of (+) torsional stress affect the conformation of intracellular chromatin. In this regard, in vitro studies had shown that torque generated ahead of RNA polymerases is enough to partially unwrap nucleosomal DNA and flip from (−) to (+) the entry and exit DNA crossing of the nucleosomes (18–20). This plasticity allows eukaryotic chromatin to buffer moderate levels of (+) torsional stress without shortening or buckling supercoils. However, when this buffering capacity is surpassed, chromatin enters into a supercoiling regime (18–20). At this stage, chromatin tightly condenses and (+) torsional stress may enforce also a flip of the global left-handed chirality of nucleosomes (18). Remarkably, we had previously showed that chromatin is efficiently relaxed by topo II but not by topo I when its DNA superhelical density is >0.04 (16), a value in which chromatin has entered the supercoiling regime (18–20). With these premises, the abrupt reduction of long transcript production reported here is likely to reflect the transition of intracellular chromatin from the buffering stage to the supercoiling stage during transcription elongation, as illustrated by the model in Figure 6. This transition probably occurs during normal transcription, since the generation rate of torsional stress in front of RNA polymerases is higher than the relaxation rate provided by cellular topoisomerases (19,31). Accordingly, we propose that topo I activity alone and the buffering capacity of yeast chromatin are sufficient to support the normal production of transcripts up to ~3kb in length. After this point, increasing levels of (+) torsional stress drive chromatin into the supercoiling regime downstream of the transcription complex. Now, topo I is no longer efficient and the relaxation of DNA by topo II activity becomes essential to support the normal progression of the RNA polymerase.
Our conclusions on the requirement of topo II activity during Pol II elongation are comparable to those inferred for Pol I transcription of yeast rDNA genes. Former studies indicated that inactivation of either topo I or topo II stimulates Pol I transcription initiation; whereas inactivation of both enzymes affected the elongation of long, but not short, Pol I transcripts (14). More recent visualization analyses revealed that effective elongation by Pol I depends mainly on topo II activity (17). Therefore, although Pol I and Pol II markedly differ in the structure and dynamics of their transcribing ensembles, topo II may be similarly required for the removal of (+) DNA supercoils accumulated ahead of both RNA polymerases. The biological relevance of this activity is probably masked by the other essential roles of topo II during chromosome replication and segregation (4,21). Since topo II inactivation decreases the abundance of long transcripts but does not abolish them, incipient or minor phenotypes are possibly undetected along the other major or lethal effects of topo II.
In prokaryotes, DNA topoisomerases have a clear division of roles to relax DNA during transcriptional elongation. Bacterial topo I specifically relieves (−) torsional stress, whereas DNA gyrase removes (+) supercoils (3,4). In eukaryotes, however, since topo I and topo II are able to relax both (+) and (−) torsional stress, a general belief is that both enzymes can substitute for each other in these tasks. Our study demonstrates that this is not the case. Topo II has a specific role in removing (+) supercoils accumulated during transcriptional elongation, a function that cannot be compensated by cellular topo I. These findings imply also that biological and pharmacological effects of topoisomerase inhibitors could be dependent on local levels of DNA supercoiling and, thus, on the transcript length of neighbouring genes.
Supplementary Data are available at NAR Online: Supplementary Tables 1–3.
Plan Nacional I + D + I of Spain [AGL2000-0133 and BIO2005-00840 to B.P.; BFU2008-00366 and BFU2011-23851 to J.R.]. Funding for open access charge: Plan Nacional I+D+I of Spain [BFU2011-23851 to J.R.].
Conflict of interest statement. None declared.
We thank Drs Chaysavanh Manichanh, Xavier Fernández, Belén Martínez and Toni Celia for technical advice and discussion.