A simple model for the influence of meiotic conversion tracts on GC content.

Abstract

A strong correlation between GC content and recombination rate is observed in many eukaryotes, which is thought to be due to conversion events linked to the repair of meiotic double-strand breaks. In several organisms, the length of conversion tracts has been shown to decrease exponentially with increasing distance from the sites of meiotic double-strand breaks. I show here that this behavior leads to a simple analytical model for the evolution and the equilibrium state of the GC content of sequences devoid of meiotic double-strand break sites. In the yeast Saccharomyces cerevisiae, meiotic double-strand breaks are practically excluded from protein-coding sequences. A good fit was observed between the predictions of the model and the variations of the average GC content of the third codon position (GC3) of S. cerevisiae genes. Moreover, recombination parameters that can be extracted by fitting the data to the model coincide with experimentally determined values. These results thus indicate that meiotic recombination plays an important part in determining the fluctuations of GC content in yeast coding sequences. The model also accounted for the different patterns of GC variations observed in the genes of Candida species that exhibit a variety of sexual lifestyles, and hence a wide range of meiotic recombination rates. Finally, the variations of the average GC3 content of human and chicken coding sequences could also be fitted by the model. These results suggest the existence of a widespread pattern of GC variation in eukaryotic genes due to meiotic recombination, which would imply the generality of two features of meiotic recombination: its association with GC-biased gene conversion and the quasi-exclusion of meiotic double-strand breaks from coding sequences. Moreover, the model points out to specific constraints on protein fragments encoded by exon terminal sequences, which are the most affected by the GC bias.

Full text

Page 1

A Simple Model for the Influence of Meiotic Conversion
Tracts on GC Content
Marie-Claude Marsolier-Kergoat*
Institut de Biologie et de Technologies de Saclay, CEA/Saclay, Gif-sur-Yvette, France
Abstract
A strong correlation between GC content and recombination rate is observed in many eukaryotes, which is thought to be
due to conversion events linked to the repair of meiotic double-strand breaks. In several organisms, the length of
conversion tracts has been shown to decrease exponentially with increasing distance from the sites of meiotic double-
strand breaks. I show here that this behavior leads to a simple analytical model for the evolution and the equilibrium state
of the GC content of sequences devoid of meiotic double-strand break sites. In the yeast Saccharomyces cerevisiae, meiotic
double-strand breaks are practically excluded from protein-coding sequences. A good fit was observed between the
predictions of the model and the variations of the average GC content of the third codon position (GC3) of S. cerevisiae
genes. Moreover, recombination parameters that can be extracted by fitting the data to the model coincide with
experimentally determined values. These results thus indicate that meiotic recombination plays an important part in
determining the fluctuations of GC content in yeast coding sequences. The model also accounted for the different patterns
of GC variations observed in the genes of Candida species that exhibit a variety of sexual lifestyles, and hence a wide range
of meiotic recombination rates. Finally, the variations of the average GC3 content of human and chicken coding sequences
could also be fitted by the model. These results suggest the existence of a widespread pattern of GC variation in eukaryotic
genes due to meiotic recombination, which would imply the generality of two features of meiotic recombination: its
association with GC-biased gene conversion and the quasi-exclusion of meiotic double-strand breaks from coding
sequences. Moreover, the model points out to specific constraints on protein fragments encoded by exon terminal
sequences, which are the most affected by the GC bias.
Citation: Marsolier-Kergoat M-C (2011) A Simple Model for the Influence of Meiotic Conversion Tracts on GC Content. PLoS ONE 6(1): e16109. doi:10.1371/
journal.pone.0016109
Editor: Geraldine Butler, University College Dublin, Ireland
Received August 19, 2010; Accepted December 10, 2010; Published January 13, 2011
Copyright: ? 2011 Marie-Claude Marsolier-Kergoat. This is an open-access article distributed under the terms of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants from the ARC (Association pour la Recherche sur le Cancer, fixed subvention # 4917; http://www.arc-cancer.net/)
and from the ANR (Agence Nationale de la Recherche, ANR-07-BLAN-0091 - CSD 8; http://www.agence-nationale-recherche.fr/). The funders had no role in study
design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: mcmk@cea.fr
Introduction
Almost ubiquitous among eukaryotic organisms is a correlation
between GC content and meiotic recombination rates [1–5].
Whereas the causality relationships are debated in many cases,
several lines of evidence have accumulated for a mechanism
termed GC-biased gene conversion whereby the frequency of
meiotic recombination affects the evolution of GC content (for a
review, see [6]). This mechanism relies on the fact that during
meiotic recombination, double-strand breaks (DSBs) are repaired
through a process involving the formation of DNA heteroduplexes
between the strands of the cut and the uncut chromosomes (see
[7,8] for reviews). As shown in Figure 1, these DNA heterodu-
plexes, which are systematically formed at the sites of DSBs, can
extend to variable distances away from it on both sides.
If the sequences of the strands forming an heteroduplex are not
perfectly complementary, mismatches occur that can be repaired by
several pathways, probably at multiple steps during the process of
DSB repair. Assuming that one strand is consistently used as a
template for the correction of the other strand, correcting the
sequence of the cut chromosome according to the sequence of the
uncut chromosome (this event is called gene conversion), leads to
three copies of the sequence from the uncut chromosome and only
one copy of the sequence from the cut chromosome in the
recombination products, instead of the original two copies for each
sequence. When comparing the sequences of the meiotic products,
this asymmetry appears as a so-called conversion tract (Figure 1). In
contrast, correcting the sequence of the uncut chromosome
according to the sequence of the cut chromosome (this event is
called gene restoration) preserves thesymmetry, and both sequences
remain present in two copies in the recombination products. One
could imagine complex patterns with mismatch repair alternating
between conversion and restoration events. However, the fact that
crossovers are frequently associated with simple, continuous
conversion tracts indicates that in the majority of the cases the
sequence of the cut chromosome is systematically converted [9].
For a given meiosis, gene conversion introduces an asymmetry
in the number of allelic sequences present in the meiotic products.
If DSBs at a given site occur with the same frequency in two pairs
(A and B) of homologous chromosomes, this asymmetry disappears
at the level of the population, and allelic frequencies are not
modified. In contrast, if DSBs occur more frequently at a given site
in one of the pairs of homologs (let’s say the A pair), the frequency
of the sequences of the A chromosomes close to this DSB site is
decreased by meiotic events, which lowers the probability of their
fixation in the population.
PLoS ONE | www.plosone.org1 January 2011 | Volume 6 | Issue 1 | e16109

Page 2

Regarding the evolution of the GC content, a higher probability
of AT-rich alleles to experience meiotic DSBs can thus lead to an
increase of the GC content in the sequences surrounding DSB
sites. This recombination-initiation bias is one of the models
proposed for GC-biased gene conversion. Other mechanisms are
possible, including biases in the repair enzymes [3], which would
favor GC in cases of AT/GC mismatches. Whatever the
molecular mechanism(s) ultimately responsible for it, the fact that
gene conversion linked to meiotic recombination increases GC
content was recently given a direct demonstration in Saccharomyces
cerevisiae by Mancera et al. who measured a significant 1.4 %
increase in the GC content of converted sequences by genotyping
* 52,000 markers in all the products of 51 meioses [9].
In most cases, the frequency of gene conversion on both sides of
a DSB site decreases exponentially with the distance from the DSB
site [10,11]. This can be explained by assuming that processes
leading to gene conversion extend from the DSB site with a fixed
probability p of stopping at each base pair. Let X be the random
variable corresponding to the distance over which a conversion
tract extends in one direction from a DSB site. Experimen-
tal observations thus indicate that Pr Xwx
majority of repair events leading to crossovers should involve
the extension of conversion tracts on both sides of a DSB (Figure 1).
Under the hypothesis that the distances over which conver-
sion tracts extend on both sides of a DSB correspond to indepen-
dent variables, the sum of the lengths of the diverging conversion
tracts (hereafter called the total length of conversion tracts) has
a mean value of
2{p
ð
experimental observations obtained by Mancera et al., who
found a median value of 2 kilobases (kb) for the total length of
conversion tracts associated with crossovers, results in an estimate
of p*0:001 [9].
fg~ 1{p
ðÞx. The
Þ=p [10]. Applying this model to the
Figure 1. Current model of the pathways involved in meiotic DSB repair. Two interacting, homologous chromatids (out of four) are
represented. Following DSB formation, 59 to 39 resection leads to 39 ssDNA tails. One of these tails invades the homologous DNA, forming a D-loop
which is then extended by DNA synthesis (dotted line). If the second end of the DSB is captured, either early D-loop cleavage (indicated by black
segments) leads to crossover products, or, in the DSBR (Double Strand Break Repair) pathway, a double Holliday junction is formed, whose resolution
or dissolution generate either crossovers or noncrossovers. Alternatively, in the SDSA (Synthesis-Dependent Strand-Annealing) pathway, the D-loop is
disassembled by displacement of the newly synthesized strand, which results in noncrossovers. Heteroduplex DNA structures are present at many
steps in all DSB repair pathways. Mismatch repair of heteroduplexes can probably occur at different steps, but is represented here as taking place at
the last step and as generating conversion tracts, which appear to be the most frequent outcome.
doi:10.1371/journal.pone.0016109.g001
Meiotic Conversion Tracts and GC Content
PLoS ONE | www.plosone.org2January 2011 | Volume 6 | Issue 1 | e16109

Page 3

Combining the exponential decrease of conversion tract
extension with the influence of these tracts on GC content, I
reasoned that DNA sequences devoid of meiotic DSB sites that
experience the extension of meiotic conversion tracts initiated
beyond their boundaries should present a specific profile of GC
content, with a higher GC content at both ends. Here I present a
simple, analytical model for the evolution of the GC content of
such sequences. Its predictions are consistent with the genomic
data of S. cerevisiae and of both sexual and presumably asexual
species of Candida and related yeasts. Finally, the model also seems
to provide a relevant description for the gene sequences of higher
eukaryotes, which suggests that a universal pattern of GC variation
could be induced in eukaryotic protein-coding sequences by
meiotic conversion tracts.
Results
A model for the evolution of the GC content in DNA
sequences devoid of meiotic DSB sites
Let’s consider a segment whose middle point is located at
position x (in bp) relative to the 59 end of a DNA sequence g
(Figure 2). The GC content of this segment evolves through the
appearance and the fixation of new alleles. We will suppose that
this process involves two kinds of mechanisms: (i) mechanisms
dependent on meiotic recombination, which modify the probabil-
ity of fixation of an allele through gene conversion and (ii)
mechanisms independent of meiotic recombination, which operate
uniformly on the segments of the sequence, independently of their
positions x.
These two kinds of mechanisms are characterized by the rates of
the substitutions they induce in the genome: let u1(x) and v1(x)
represent, respectively, the AT to GC and GC to AT substitution
rates linked to recombination-dependent processes, and u2and v2
represent the substitution rates linked to recombination-indepen-
dent processes, which we will consider as independent of x.
Let GS(x,t) and TS(x,t) be, respectively, the GC content
(proportion of GCs) of a segment located at position x, and the
frequency with which conversion tracts reach x in the sequence g
at time t. u1(x) and v1(x) are obviously dependent on TS(x,t). We
will suppose that u1(x) and v1(x) can be considered as
proportional to TS(x,t) and can be written as u1TS(x,t) and
v1TS(x,t), respectively, with u1 and v1 being two constants. We
thus have
LGS x,t
Lt
ðÞ
~ u1TS(x,t)zu2
½? 1{GS(x,t)
½?{ v1TS(x,t)zv2
½?GS(x,t) ð1Þ
Let’s calculate the frequency TS(x,t). We assume that the
sequences under study are devoid of DSB sites, which means that
the conversion tracts affecting them originate from DSB sites
located either in their 59 or in their 39 regions. Let’s first consider a
DSB site dilocated at a distance liupstream of the 59 end of the
sequence g, from which conversion tracts are initiated with a
frequency fi(t) (Figure 2). TS,i(x,t), the frequency with which
conversion tracts extend from DSBs at di to position x,
corresponds to the product of fi(t) and the probability that the
conversion tracts, once initiated, will extend up to x. Since
conversion tracts extend from DSB sites with a fixed probability p
to stop at each base pair, this probability is equal to (1{p)lizx. We
thus get TS,i(x,t)~fi(t)(1{p)lizx.
We consider now all the DSB sites located either upstream of
the sequence g, at distances lifrom its 59 end, or downstream of g
at distances l’jfrom its 39 end, i.e. at distance l’jzLgfrom the 59
end, Lgbeing the length of the sequence g (Figure 2). The position
of the DSB sites is unknown, so we simply consider all positions
upstream and downstream of the sequence g as potential DSB sites
at which conversion tracts are initiated at frequencies fi(t), with
fi(t) being negligible for most of them. We thus obtain
TS(x,t)~
X
n
i~1
fi(t)(1{p)lizxz
X
m
j~1
fj(t)(1{p)l’jzLg{x
ð2Þ
with n and m corresponding to the distances between the 59
(respectively, 39) end of the sequence g and the 59 (respectively, 39)
end of the chromosome on which g is located.
Let’s consider the situation in which the frequencies fi(t) in
Equation 2 can be replaced by the constants fi(in the Discussion,
we will see that GScan closely approach an equilibrium value only
if the frequencies fi(t) are either constant over time or undergo
only rapid changes around a constant, time-averaged value
fi(t)~fiat frequencies too high to be reflected by the GC content).
For a given sequence g, we can write Ig~Pn
i~1fi(1{p)liand
Jg~Pm
j~1fj(1{p)l’j, so that
TS(x,t)~TS(x)~Ig(1{p)xzJg(1{p)Lg{x
ð3Þ
G?
S(x), the equilibrium value of GS(x,t), is then obtained by
settingLGS x,t
Lt
ðÞ
~0 and can be written
G?
Sx
ð Þ~
u1TS x
ÞTS x
u1 Ig1{p
ð
u1zv1
Þ Ig1{p
ð Þzu2
ð Þzu2zv2
ÞxzJg1{p
ð
u1zv1
ð
~
ðÞLg{x
ÞLg{x
hi
zu2
i
ðÞxzJg1{p
ð
h
zu2zv2
ð4Þ
The GC content of the sequences of some organisms at least
appears to be close to equilibrium (see below), hence the relevance
of G?
S(x).
Figure 2. Schema illustrating some of the model parameters. A
sequence g of length Lg, devoid of meiotic DSB sites, is shown as a blue
rectangle. The orange rectangle corresponds to a segment whose
middle point is located at position x relative to the 59 end of g and at
position z~Lg{x relative to its 39 end. The red arrows represent the
extension towards g of different conversion tracts initiated either at the
DSB site di, located at the distance liupstream of g or at the DSB site dj,
located at the distance l’jdownstream of g.
doi:10.1371/journal.pone.0016109.g002
Meiotic Conversion Tracts and GC Content
PLoS ONE | www.plosone.org3 January 2011 | Volume 6 | Issue 1 | e16109

Page 4

Finally, Equation 4 can be simplified into
G?
Sx
ð Þ~
Ag 1{p
ð
D Ag1{p
ð
ÞxzBg 1{p
ÞxzBg1{p
ðÞLg{xzC
ÞLg{x
ð
hi
z1
ð5Þ
with Ag~u1Ig=(u2zv2) and Bg~u1Jg=(u2zv2) being specific to
the sequence g (depending on the position and on the time-
averagedactivity oftheneighboring
C~u2=(u2zv2) and D~1zv1=u1 being two constants a priori
identical for all sequences.
DSBsites) and
Analysis of Saccharomyces cerevisiae coding sequences
The model makes many simplifying assumptions (u1, v1, u2, v2
and p are considered to be constant over time and identical for all
sequences for example) and Equation 5 applies only to specific
cases, to sequences with low meiotic DSB density and whose GC
contents are close to equilibrium, in organisms exhibiting GC-
biased gene conversion. I first sought to evaluate its relevance by
analyzing the sequences of S. cerevisiae, an organism in which
meiotic recombination has been extensively studied and for which
the existence of GC-biased gene conversion has been experimen-
tally demonstrated [9].
Several studies have shown that the protein-coding sequences of
S. cerevisiae experience few meiotic DSBs. A thorough mapping of
meiotic DSBs on S. cerevisiae chromosome III at a resolution of
100–500 pb identified only 5 DSB sites in protein-coding
sequences out of 76 DSB regions [12]. DSBs were also found to
be practically excluded from intergenic regions containing two
terminators. These results were subsequently confirmed by a
genome-wide mapping of meiotic recombination hotspots [1].
Chromatin structure seems to be the most basic determinant
controlling the position of meiotic DSBs in yeast (reviewed in [13])
and the preferential localization of DSB sites in promoter regions
is often explained by the hypothesis that DSB-forming complexes
are more efficient on open chromatin regions, which are most
commonly established for promoting transcription.
The third codon position was selected for analysis because this
position is the least constrained by coding requirements. To test
whether the GC content of the third codon position (GC3) in S.
cerevisiae genes was close to equilibrium, the equilibrium GC3
contents G?
gof 3661 genes of S. cerevisiae were computed from the
inferred substitutions having occurred in S. cerevisiae lineage after
the divergence between S. cerevisiae and Saccharomyces paradoxus (see
Methods). A strong linear relationship was found between G?
the observed GC3 content Gob
g
Pv10{10): G?
g
with a~0:08+0:01 and b~0:83+
0:02. I therefore considered that the conditions were met for
Equation 5 to describe approximately the GC3 content of S.
cerevisiae protein-coding sequences.
The sequences of 5500 ORFs without intron, annotated as
verified or uncharacterized in the Saccharomyces Genome
Database were divided into non-overlapping segments of 66
codons (198 bp), starting either from the ATG or from the stop
codon. The orientation of the coding sequences was taken into
account because of the asymmetry in the distribution of the DSB
sites, which are preferentially located in promoter-containing
intergenes.
Equation 5 can be taken as describing the theoretical
equilibrium GC3 content G?
S(x) of a gene segment located at
position x relative to the ATG. Let z correspond to the distance
between a segment and the stop codon of the gene (z~Lg{x,
Figure 2). The theoretical equilibrium GC3 content of a gene
segment can also be written
gand
of S. cerevisiae genes (r~0:52,
g~azbGob
H?
Sz ð Þ~
Ag1{p
ð
D Ag1{p
ð
ÞLg{zzBg1{p
ÞLg{zzBg1{p
ðÞzzC
Þz
ð
hi
z1
ð6Þ
According to the model, the shape of the curves representing
Sand H?
between the middle and the end segments of a gene increases with
gene length, the middle segments of the longest genes rarely
experiencing the extension of conversion tracts. The genes were
therefore binned into classes according to the number of 198 bp-
segments they contain. The average observed GC3 contents
Gob
S(z) of each segment centered on position x from
the ATG or on position z from the stop codon, respectively, were
measured for the different classes of genes. As shown in Figure 3A,
Gob
Stends to decrease for segments located farther from the ends of
the genes. This trend could be observed for all classes of genes,
except for the shortest genes, and was most apparent for the
longest genes. A similar trend was also visible for Hob
Even if current recombination rates vary largely from one gene to
another [1,9,14,15], suggesting a similar heterogeneity for Agand
Bg, the potential gene-to-gene variations in Ag and Bg were
disregarded in a first analysis, and uniform values of Ag, Bg, C, D
and p that give the best fit to the observed GC3 contents Gob
Hob
similar results. The fact that the estimates of Ag(which reflects the
activity of DSB sites at the gene 59 ends) were higher than the
estimatesof Bg(which reflects the activity of DSB sites at the gene 39
ends) was consistent with the observation that in current S. cerevisiae
strains DSBs occur preferentially in promoter-containing regions
and are almost excluded from intergenes containing two termina-
tors. Similarly, the estimates of p, the probability of the conversion
tracts to stop at each base pair, were * 0.0009, in good agreement
with the value of p (* 0.001) determined experimentally from the
analysis of * 4200 crossovers in S. cerevisiae [9]. The Pearson’s
correlation coefficient between the observed GC3 contents Gob
the theoretical GC3 content G?
Equation 5 using the estimates of Ag, Bg, C, D and p given in
Table 1, was found equal to 0.25 (Pv10{10, n~36,167).
WehavepreviouslyseenthattheGC3contentofS.cerevisiaegenes
could be considered as close to equilibrium. In the frame of the
model,this observation meansthat the time-averaged frequencies of
conversion tract initiation fi(t) have been recently constant in S.
cerevisiae lineage, so that GShas adapted to them (see Discussion). In
that case, one possibility is that not only the time-averaged
frequencies fi(t), but also the frequencies fi(t) themselves could
have been recently constant in S. cerevisiae lineage, and therefore that
Ag and Bg, which are functions of fi(t), could be correctly
approximated by current estimates of these frequencies. The
relevance of this assumption can easily be assessed by comparing
the correlations between the data Gob
calculated with or without the approximations for Agand Bg.
Ag~u1Ig=(u2zv2) with Ig~Pn
initiated at DSB sites located at distance liupstream of the sequence
g extend up to its ATG. Buhler et al. recently measured the amounts
of ssDNA produced by the 59 to 39 resection of meiotic DSB ends at
* 41,000 positions in S. cerevisiae genome [15]. The amount of
ssDNA measured at a given position thus reflects the frequency with
which 59 to 39 resection tracts, originating from DSB sites located
either upstream or downstream, extend up to that position.
Conversion tracts and resection tracts probably do not coincide
exactly but they clearly overlap (Figure 1). I therefore hypothesized
that, for a given sequence, the time-averaged frequency of
G?
Sdepends on Lg as the difference in GC3 content
S(x) and Hob
S(Figure 3B).
S(x) or
S(z) gave
S(z) were determined (Table 1). Fitting Gob
S(x) and Hob
Sand
Sthat can be calculated from
S
and the theoretical G?
S
i~1fi(1{p)li. Ag
is thus
proportional to the frequency Ig with which conversion tracts
Meiotic Conversion Tracts and GC Content
PLoS ONE | www.plosone.org4January 2011 | Volume 6 | Issue 1 | e16109

Page 5

conversion tract extension Igcould be approximated by the current
frequency of conversion tract extension, which in turn could be
approximated by the current frequency of resection tract extension,
estimated from ssDNA measurements. The reasoning is the same
for Bg and leads to the hypothesis that Ag and Bg could be
considered as proportional to Aob
quantifications of ssDNA averaged over the 500 bp upstream of
the genes start codon or over the 500 bp downstream of the genes
stop codon, respectively (see Methods).
As alluded to above, these approximations of Agand Bghave
several limitations: (i) little is known about the extension of resection
tracts and how it correlates with the extension of conversion tracts,
(ii) Aob
gintegrate the amounts of ssDNA deriving from
DSBs located either upstream or downstream of the genes, whereas
Agand Bgcorrespond to conversion tracts originating exclusively
either from upstream (Ag) or from downstream (Bg) regions of the
genes and (iii) the measures of ssDNA appear noisy as the Pearson’s
g
and Bob
g, defined as the
gand Bob
correlation coefficient between the values determined by Buhler
et al. and by another study [14] analyzing the same strain with the
same microarrays and protocols is * 0.62 (n~40,478).
Replacing Agand Bgby aAob
gives
ÞxzaBob
D aAob
g1{p
ð
gand aBob
gin Equations 5 and 6
G?
Sx
ð Þ~
aAob
h
g1{p
ð
g1{p
ð
g1{p
ð
ÞLg{xzC
ÞLg{x
ÞxzaBob
i
z1
ð7Þ
H?
Sz ð Þ~
aAob
h
g1{p
ð
g1{p
ð
ÞLg{zzaBob
ÞLg{zzaBob
g1{p
ð
g1{p
ð
ÞzzC
Þz
D aAob
i
z1
ð8Þ
Estimates of a, C, D and p were determined by fitting the observed
GC3contentsGob
StoEquations7and8(Table1).Thevalues
SandHob
Table 1. Equation coefficients for S. cerevisiae genes.
Fit with x and Lg
Fit with x, Lg, Aob
gand Bob
g
Gob
S
Hob
S
Gob
S
Hob
S
a
NA NA0.030 + 0.002 0.026 + 0.001
Ag
0.12 + 0.03 0.12 + 0.04NA NA
Bg
0.09 + 0.02 0.08 + 0.02 NANA
C
0.321 + 0.004 0.324 + 0.0030.330 + 0.0010.331 + 0.001
D
1.0 + 0.41.0 + 0.4 1.40 + 0.041.32 + 0.04
p
0:92|10{3+7|10{5
0:94|10{3+7|10{5
1:04|10{3+3|10{5
0:98|10{3+3|10{5
Coefficients of Equations 5 and 6 (fit with x and Lg) and of Equations 7 and 8 (fit with x, Lg, Aob
genes. NA, not applicable.
doi:10.1371/journal.pone.0016109.t001
gand Bob
g) determined by fitting the data Gob
Sand Hob
Sof S. cerevisiae
Figure 3. Variations of GC3 content in S. cerevisiae protein-coding sequences. The mean GC3 contents Gob
segmentsareplotted as a functionof the positions x or z relative totheATGor tothestop codon,respectively, onwhich thesegmentsarecentered. For
a given position x or z, Gob
Sare averaged over classes of genes binned by their lengths. The genes of set 1 (n~1684, blue dotted line) contain 3
to5 66-codon segments, thegenes ofset2 (n~1316, blue dashedline) contain 6 to8 segments,thegenes ofset3 (n~717, bluedot-dashline)contain 9
to 11 segments, the genes of set 4 (n~379, blue long-dash line) contain 12 to 14 segments, and the genes of set 5 (n~477, blue solid line) contain at
least 15 segments. Only the average valuesforthesegments commontoallthegenes of a givensetareplotted (i.e. onlythesegmentscorresponding to
the shortest genes of the set). The thick, red, solid lines represent the mean Gob
segment (n~5374). The thin, red, solid lines represent the mean theoretical values G?
segmentandwhosevaluesofAob
7 and 8 (and the estimates of a, C, D and p given in Table 1), and averaged over segments with the same position x or z, respectively.
doi:10.1371/journal.pone.0016109.g003
S(A) or Hob
S(B) of 66-codon
Sand Hob
Sor Hob
Sor H?
SandH?
Saveraged over all genes containing at least one 66-codon
Saveraged over all genes containing at least one 66-codon
Swerecalculatedasfunctions ofx,Lg,Aob
gandBob
gcouldbedetermined(n~5162).G?
gandBob
gusingEquations
Meiotic Conversion Tracts and GC Content
PLoS ONE | www.plosone.org5 January 2011 | Volume 6 | Issue 1 | e16109