Molecular evolution of psbA gene in ferns: unraveling selective pressure and co-evolutionary pattern
- First Online:
- Cite this article as:
- Sen, L., Fares, M.A., Su, YJ. et al. BMC Evol Biol (2012) 12: 145. doi:10.1186/1471-2148-12-145
The photosynthetic oxygen-evolving photo system II (PS II) produces almost the entire oxygen in the atmosphere. This unique biochemical system comprises a functional core complex that is encoded by psbA and other genes. Unraveling the evolutionary dynamics of this gene is of particular interest owing to its direct role in oxygen production. psbA underwent gene duplication in leptosporangiates, in which both copies have been preserved since. Because gene duplication is often followed by the non-fictionalization of one of the copies and its subsequent erosion, preservation of both psbA copies pinpoint functional or regulatory specialization events. The aim of this study was to investigate the molecular evolution of psbA among fern lineages.
We sequenced psbA , which encodes D1 protein in the core complex of PSII, in 20 species representing 8 orders of extant ferns; then we searched for selection and convolution signatures in psbA across the 11 fern orders. Collectively, our results indicate that: (1) selective constraints among D1 protein relaxed after the duplication in 4 leptosporangiate orders; (2) a handful positively selected codons were detected within species of single copy psbA, but none in duplicated ones; (3) a few sites among D1 protein were involved in co-evolution process which may intimate significant functional/structural communications between them.
The strong competition between ferns and angiosperms for light may have been the main cause for a continuous fixation of adaptive amino acid changes in psbA , in particular after its duplication. Alternatively, a single psbA copy may have undergone bursts of adaptive changes at the molecular level to overcome angiosperms competition. The strong signature of positive Darwinian selection in a major part of D1 protein is testament to this. At the same time, species own two psbA copies hardly have positive selection signals among the D1 protein coding sequences. In this study, eleven co-evolving sites have been detected via different molecules, which may be more important than others.
Fixed effects likelihood
Inverted repeat region(s)
Large single copy region
Likelihood ratio test
Million (1 × 106) years ago
Multiple sequence alignment(s)
- PS II
Photo system II
Random effects likelihood
Row and column weighing of mutual information
Single likelihood ancestor counting
The photosynthetic oxygen-evolving photo system II is a unique biochemical system that is capable of oxidizing water molecules  and is responsible for producing the almost totality of oxygen on earth [2, 3]. psbA gene, along with three other chloroplast (cp) genes, namely psbB , psbC and psbD , encodes the core proteins complex in the chlo-roplasts of ferns [4, 5, 6, 7]. Precisely, psbA gene encodes D1 protein (also known as PsbA protein) in ferns. Physical mapping and pastime sequencing unra-veled a set of genome rearrangements around psbA gene in “higher” fern lineages [8, 9, 10]. Based on completely and partially sequenced fern pastimes [11, 12, 13], the psbA gene often locates either to inverted repeat regions (IRs) or to the large single copy region (LSC). Moreover, gene order around it is only found in one state in IRs (rps7psbAtrnH), but in two possible arrangements in LSC (trnKpsbAtrnH or matKpsbAtrnH). Impor-tantly, psbA gene duplicated when its location shifted from LSC to IRs . This shift mainly affected four fern orders: Schizaeales, Salviniales, Cyatheales and Polypodiales, accounting for over 90% of extant fern diversity. Because both gene copies were preserved since the duplication, we sought to investigate the selective pressures that may drive the evolution of this interesting gene.
Evolution after gene duplication has been a fundamental issue in evolutionary biology chiefly because of its direct link to the generation of novel functions and adaptations [14, 15, 16]. The opportunity for generating novel functions is, however, often balanced with the effects such duplications have on gene dosage [17, 18, 19, 20]. The photosynthesis environment changed mainly as a result of the emergence of flowering plants , and as the open sunny ground transformed into the closed shadowy canopy. For this reason, the biodiversity of other vascular plants largely decreased, and some lineages underwent extinction [22, 23, 24]. Remarkably, rather than the result of being adept at holding on in the face of angiosperm domination, the leptosporangiate ferns may have the ability to capitalize upon it . The underlying molecular functioning of this ability remains a mystery for evolutionary biologists, and the present study sheds some light on it. The colonization by these species of canopies angiosperm-dominated light environments sparks the idea that such extraordinary diversification may have been fuelled by functional innovation following psbA duplication. To disclose this obscurity, we performed a comprehensive analyze of the action of natural selection following the duplication in psbA .
Coevolving sites are non-independent amino acid sites during the course of protein evolution [26, 27]. Among the different sites with physical or functional relationships in one protein, one mutation is likely to trigger the corresponding mutations at related sites . Natural selection theory and nearly neutral theory have different explanations for the mechanism of these dynamic changes [29, 30, 31]. Despite the debates on the underlying mechanism, the compensatory mutations among the interdependent amino acid sites provide an important approach to understand protein structure and function . The identification of coevolving amino acid sites will not only play a prominent role in the annotations of the function of D1 protein, but will also reflect the evolutionary pattern of the particular protein.
To achieve a better understanding on the evolutionary biology of psbA genes in ferns, we focused on three aims: 1) identifying gene order states around psbA genes in 11 fern orders; 2) detecting the selective constraints in consequence of psbA gene duplication; and 3) unraveling the co-evolution pattern of D1 protein among fern lineages.
Sampling and sequencing
Plant materials and sampling locations in this study
P. bifurcatum (Cav.) C. Chr.
M. fortunei (T. Moore) Ching
A. australasicum (J.Sm.) Hook.
D. mariesii Wall.
N. exaltata (L.) Schott
C. lepifera (Hook.) R.M. Tryon
C. barometz (L.) J. Sm.
P. japonica Nakai
P. distinctissima Ching
S. molesta Mitchell.
L. microphyllum Cav. R.Br.
D. linearis (Burm.f.) Underw.
D. chinensis (Rosenst.) DeVol
D. chinensis Christ
V. radicans (Sw.) Copel.
O. vachellii Hook.
O. angustifolia Ching
H. zeylanica Hook.
O. vulgatum L.
B. strictum Underw.
Since the analysis of molecular adaptive evolution strictly rejects the termination codon (i.e. TAA, TGA, TAG) in a sequence , coding sequences in dataset 1, 2 and 3 showed in Additional file 1 Table S3 were applied in the investigation of selective pressures.
Selective pressures amongpsbAgene sequences examined via SLAC and FEL models
Numbers and positions of positively/negatively selected site*
p = 0.01
p = 0.05
p = 0.1
p = 0.15
p = 0.2
p = 0.25
3(4, 155, 352)/93
4(4, 155, 350, 352)/120
5(4, 53, 155, 350, 352)/139
In accordance with the phylogenetic structure in pre-vious documents and present analyses (Figure 2), we found that the combined dataset 6 might not be an accurate indicator for the reconstruction of fern phylogenetic structure under geological timescale. Even though the currently gained tree partially coincided with other relevant results from Smith’s system  and Lehtonen’s conclusion , we could found several disagreements.
The estimated parameters among phylogenetic tree via BEAST software
Estimated Diverge Time**
Parameter estimates and log-likelihood values under models of variableωratios among codons
Estimated value of parameters
Positively selected codons**
M0: One ratio
ω = 0.0184
M1a: Near neutral
p0 = 97.14%, ω0 = 0.00618; p1 = 2.86%, ω1 = 1
p0 = 97.14%, ω0 = 0.00618; p1 = 2.86%, ω1 = 1; p2 = 0%, ω2 = 79.26
91 L (97.2%)
p0 = 96.31%, ω0 = 0.00488; p1 = 3.39%, ω1 = 0.5646; p2 = 0.3%, ω2 = 3.9847
91 L (99.8%)
p = 0.0267, q = 0.614
M8: β &ω > 1
p0 =97.84%, p = 0.01575, q = 0.46508; p1 = 2.15%, ω = 1.08788
91 L (99.7%)
M0: One ratio
ω = 0.02403
M1a: Near neutral
p0 = 96.57%, ω0 = 0.00127; p1 = 3.43%, ω1 = 1
p0 = 96.57%, ω0 = 0.00127; p1 = 3.43%, ω1 = 1; p2 = 0, ω2 = 34.2681
p0 = 46.14%, ω0 = 0; p1 = 49.52%, ω1 = 0; p2 = 4.31%, ω2 = 0.61912
p = 0.00981, q = 0.19983
M8: β &ω > 1
p0 =96.61%, p = 0.01744, q = 2.58197; p1 = 3.38%, ω = 1
347E (81.6%), 349P (61.5%), 353 G (54.4%)
M0: One ratio
ω = 0.01934
M1a: Near neutral
p0 = 96.39%, ω0 = 0.00538; p1 = 3.6%, ω1 = 1
p0 = 96.47%, ω0 = 0.0054; p1 = 3.55%, ω1 = 1; p2 = 0, ω2 = 4.46
71 L (72.4%), 91 L (78.0%)
p0 = 90.54%, ω0 = 0; p1 = 8.51%, ω1 = 0.5308; p2 = 0.943%, ω2 = 1.71
71 L (96.1%), 91 L (97.1%), 351 V (81.3%)
p = 0.03078, q = 0.63009
M8: β &ω > 1
p0 =98.21%, p = 0.01555, q = 0.36594; p1 = 1.78%, ω = 1.19067
71 L (96.4%), 91 L(97.2%), 351 V (84.9%)
Gene order around psbA gene in ferns
Results of the PCR amplification indicate that gene order around psbA gene in fern species could be mainly classified into three types: trnK-psbA-trnH , matK-psbA-trnH and rps7-psbA-trnH . When universal primers (Additional file 1 Table S1) from conserved sequence of trnK and psbA were applied in PCR systems, five species (Botrychium strictum, Ophioglossum vulgatum, Helminthostachys zeylanica, Osmunda angustifolia and Osmunda vachellii) have positive reaction with the right sizes of the fragments. Four species (Vandenboschia radicans, Dipteris chinensis, Diplopterygium chinensis and Dicranopteris linearis) showed positive results when the primers were from conserved sequence of matK and psbA (Additional file 1 Table S2). Other eleven species (Lygodium microphyllum, Salvinia molesta, Plagiogyria distinctissima, Plagiogyria japonica, Cibotium barometz, Cyathea lepifera, Nephrolepis exaltata, Davallia mariesii, Asplenium australasicum, Microsorum fortune and Platycerium bifurcatum) showed positive results when the primers were from conserved sequence of rps7 and psbA (Additional file 1 Table S2).
Selective pressure in different psbA genes
As showed in Figures 1 and 2, we tested the action of natural selection in D1 protein using three datasets (Dataset 1, 2 and 3 in Additional file 1 Table S3) comprising both presently determined sequences (20 species in Table 1) and sequences retrieved from GenBank (7 species in Additional file 1 Table S4).
For the detection of positive selection we used nested maximum likelihood models allowing for variation in the ratio of non-synonymous to synonymous substitutions rates (dN/dS) across codons, as implemented in PAML and Selecton [41, 42], and three models in Datamonkey as well [43, 44]. For each dataset, one Likelihood Ratio Test (LRT) was performed for dN/dS heterogeneity across codons (M0/M3 test). The Discrete model (M3) fitted the data significantly better than the one-ratio model (M0), which suggested the existence of significant variation in selective constraints among codons (Table 4, Additional file 1 Table S5). Further, four LRTs were performed for the positively selected codons: M1a/M2a, M7/M8, M8a/M8 and M8a/Mechanistic empirical model (MEC). In order to take the differences between amino-acid replacement rates into account, MEC model employed a cpREV matrix to expand a 20 by 20 amino acid replacement rate matrix into a 61 by 61 sense-codon rate matrix . By this means, a position with radical replacements will obtain a higher Ka value than a position with more moderate replacements (Figure 3). Log-likelihood values and estimates of parameters under various models were given in Table 4, and the likelihood ratio tests in Additional file 1 Table S5. In dataset two, alternative models (M2a and M8), permitting ω > 1, failed to detect positively selected codons and showed non-significance against their null tests (M1a and M7). In M2a, though the estimated value (ω2 = 34.2681) is greater than one, no codon (p2 = 0) belongs to this kind of ω; and the estimated value for ω was not greater than 1 in M8. Meanwhile, M8 significantly outperformed M7 in both datasets; M2a outperformed M1a as well (Additional file 1 Table S5). Collectively, we can conclude three main points under nested maximum likelihood models (Table 4): 1) subtle positive selection signal was detected in D1 protein from fern species; 2) codons within D1 protein encoded by duplicated psbA genes were mainly under negative selection; 3) several codons encoded by single psbA gene might have undergone adaptive evolution.
The random-site models, SLAC, FEL, REL and MEC methods were employed to examine the adaptive evolution of the D1 protein in eleven fern orders [41, 42, 43, 44]. For dataset one (Table 2 and Additional file 1 Table S6), one positively selected codon (91 L) has been detected via maximum likelihood methods (Table 4), none via SLAC model, one (4 T) via FEL model (p = 0.05), two (4 T and 155 T) via REL model (PP > 95%). Our results indicate that 96.99% codons are highly conserved, 2.69% evolve neutrally, and a few (0.31%) are positively selected with ω > 1 (Table 4). In the evolution of singleton psbA gene, positive selection at a handful of codons has played an important role in the evolutionary dynamics. On the contrary, the majority of codons are under negative selection in dataset two, while a small part is under neutral selection. No positively selected codon was found in the four fern lineages (Order Schizaeales, Salviniales, Cyatheales and Polypodiales). Our results showed that most of the sites are under purifying selection, while a small part is under neutral selection. No sites in the duplicate psbA gene were found under positive selection, while 4 have statistical significance in the singleton.
The selective relaxation in duplicated psbA genes could be observed in the adaptive selection analyses based on single likelihood ancestor counting (SLAC) and fixed effects likelihood (FEL) methods. Results in Table 2 showed a decrease of selective pressure in duplicated psbA genes (Table 2). For instance, 22 sites in dataset 1 were under selection while only 4 in dataset 2 via SLAC (p = 0.01). In accordance with FEL, 69 sites in dataset 1 were under selection while only 22 in dataset 2 (p = 0.01). Less codons from dataset 2 against dataset 1 were under either positive or negative selection (p = 0.01). On the other hand, no codon was identified as positive selection in dataset 2 under REL (Additional file 1 Table S6), while two amino acid sites (4 T and 155 T) were identified in dataset 1.
Co-evolutionary pattern among D1 protein
Co-evolution pairs with the D1 protein
Software or Website*
It is a great challenge to develop DNA barcodes for land plants [40, 48]. Kress and Erickson (2007) recommended rbcL gene and psbA-trnH spacer region as universal markers . Our results indicated that the psbA and psbA-trnH sequences have high resolution at order and family level, but low resolution at several genera level (Figures 1 and 2, Table 5). Further investigation with extra locus will have a deeper reflection on this task.
Genome rearrangement and psbA duplication
The PCR amplification results coincide with the previous conclusion based on the completely and partially sequenced plastomes [10, 11, 12, 13]: three kinds of gene order (trnK-psbA-trnH , matK-psbA-trnH and rps7-psbA-trnH) were found in the 20 species from 8 fern orders (Figure 1). The data of the extant complete fern cp genomes showed that the trnK-psbA-trnH and matK-psbA-trnH fragments are located in the LSC and the rps7-psbA-trnH fragment is in the IRs [10, 13, 50]. Since the large-scale inversion events involving the duplication of psbA gene are rare during the evolution history of fern species [8, 9], one copy of psbA exists in those species of trnK-psbA-trnH and matK-psbA-trnH types and two in rps7-psbA-trnH. Their full-length psbA encoding sequences have not been determined in previous studies.
Evolutionary trajectory and survival strategy
The strong competition between ferns and angiosperms for light may have been the main cause for the evolutionary trajectory of psbA gene. Nevertheless, the occurrence of psbA duplication provided an alternative survival strategy versus the molecular adaptive evolution at the codons of D1 protein. Several positively selected sites were found in single psbA gene copy and none in duplicated psbA (Table 2). These results indicated that single psbA gene might have functional adaptation via the replacement at certain positions among the D1 protein while the duplicated one might have not. Conversely to the case of single gene copies, the leptosporangiate ferns with duplicate psbA gene may have acquired further fitness gain through the existence of a new transcription locus for the synthesis of D1 protein, which might directly increase the efficiency of photosynthesis by protein dosage effects [20, 51, 52].
The structure and function of the D1 protein are conserved among cyanobacteria, red algae and plants [53, 54]. The purifying selection of psbA in Lejeuneaceae has been noticed in a recent study . Regretfully, not like the elegant investigation in rbcL gene , the detail function of the codons in D1 protein remains obscure. Mutation experiments of D1 protein will unveil the functional importance of the positively selected positions. Moreover, the further research on the relationships between positively selected mutations among psbA gene from fern species and the biodiversity will have great impacts on the understanding of fern evolutionary biology.
Complicated intra-molecular evolution under selective pressure
Eleven co-evolving sites (Table 5) have been detected via different molecules, and they (site 4, 19, 53, 71, 72, 92, 281, 350, 351, 352 and 353) may be more important than others in D1 protein. Point mutation experiments may have distinctive outcomes, some could cause severe functional consequences and others could result in completely undetectable change. This fact indicates that a protein is a network of interacting residues, and the core nodes in this network determine the function of the protein . Future point mutation experiments aiming at the eleven sites may have butterfly effect on the protein structure and function.
psbA gene, along with three other chloroplast genes, encodes the core proteins complex in the chloroplasts of ferns. However, as we have concluded before the functional adaptation of this complex might be caused by the inter-coevolution among different proteins . The current intra-molecular evolution analysis may have shortcomings on predicting the complicated coevolutionary networks among the entire functional core complex. Further conclusion could not be declared without overall analysis based on the four related genes: psbA, psbB, psbC and psbD.
In the current research we present evidence that point to a complex adaptive process mediating the functional innovation of the D1 protein. This process involves a multiply checking of the structural and functional consequences of the fixation of functionally novel mutations and the amelioration of the effects by such mutations may have through compensatory replacement events. A serial amino acid sites are identified as co-evolution positions while significant positive selection signals are detected in the single copy psbA gene from the fern species. One hypothetical scenario is put forward: i) single copy psbA gene fern species may adapt to the newly formed living circumstance by the modification of amino acids in D1 protein; ii) by the meantime, the dosage effects of D1 protein are the possible strategy against the rising of angiosperm in psbA duplicated ones. The selective relaxation in duplicated psbA genes could be observed throughout different models. However, no evidence stands for the functional divergence in the duplicated psbA genes. Future investigation will shed new lights on this question. Although this research covers only a little of the diversity of species in fern, our sampling included all 11 orders. Our research however opens exciting new avenues that will hopefully lead to a more complete understanding of the functional novelties and dosage effects in the D1 protein among ferns.
Sampling of plant materials
Plant materials of 20 fern species for the present investigation were collected from Wuhan Botanical Garden, Fairy Lake Botanical Garden and South China Botanical Garden, respectively (Table 1).
Isolation of total genomic DNA
For each species, three pieces of fresh leaves were collected to isolate genomic DNA with the modified CTAB protocols . Each sample was dissolved in 50 μl TE buffer. Roughly, the quality was determined by 1% agarose/TAE gel electrophoresis and the quantity was estimated via DNA Ladder (Takara). The bright sample under UV-light with right size was used as the template in PCR reactions. The absorption at 260 and 280 nm of qualified template DNA was measured using a 752 spectrophotometer. The purity and concentration was resolved and calculated by the A260/A280 ratio and A260 absorption value.
PCR amplification and DNA sequencing
PCR amplification was carried out in 100 μl volumes containing 50 mM KCl, 10 mM Tris–HCl (pH 8.0), 0.1% Triton X-100, 1.5 mM MgCl2, 0.2 mM each deoxynucleoside triphosphate, 2 U Taq DNA polymerase, 0.3 μM primer, 30 ng genomic DNA and DNA-free water. The 3-step and 2-step PCR protocols employed species-specific and universal primers, respectively. Individually, the annealing temperature in 3-step PCR reaction (Additional file 1 Table S1) of species-specific primers was calculated via Primer 3 (http://frodo.wi.mit.edu/primer3/). The thermo-cycling program was set as: 5 min at 95°C (1 cycle); 45 s at 94°C, 60s at Ta°C, 90s at 72°C (34 cycles); 10 min at 72°C (1 cycle). However, the annealing temperature was ignored in 2-step PCR reaction (Additional file 1 Table S2). The thermo-cycling program was set as: 5 min at 95°C (1 cycle); 60s at 94°C, 150 s at 60°C (35 cycles); 20 min at 72°C (1 cycle). Except as normal reactions, the genomic DNA was excluded from the reaction mix for negative control. Then the molecular weight of PCR products was verified in 1% agarose/TAE gels.
Each qualified DNA fragment amplified by the above steps was recovered and purified with a quick PCR Purification Kit (Promega), and then cloned into PMD19-T vectors (Takara). The plasmids, composed of the vectors and the DNA fragments, were transformed to Escherichia coli strain DH5α. Plasmids within positive clones were extracted and sequenced with an ABI PRISM 3730 DNA analyzer. Three clones were sequenced for each amplicon to control Taq polymerase errors. The overlapping sequences from various amplification steps were assembled as a single contig. To ascertain the contigs’ locations among cp genomes, the sequencing results were submitted to DOGMA website .
Multiple sequence alignments and best-fit nucleotide substitution model
Six multiple sequence alignments (hereafter MSAs) were established for the present investigation (Additional file 1 Table S3). Nucleotide sequences obtained experimentally (Table 1) plus those retrieved from the public databases (Additional file 1 Table S4) were aligned using the MUSCLE software . Partially fern psbA sequences from GenBank were excluded from the present study.
The best-fit nucleotide substitution model for each MSA was selected via jModeltest package . And it was also estimated via the automatic model selection tool at the Datamonkey website (http://www.datamonkey.org) for the coding regions [43, 44].
Reconstruction of phylogeny along with time-scale
Isoetes flaccida (GenBank Accession GU191333) was selected as outgroup in the phylogenetic analyses. Two nodes were chosen to constrain for a rate consistent with the known fossils: 1) Since Osmunda fossils have been described from the Upper Triassic , the Osmundaceae clade was constrained to 199.6 million years ago (Mya, Node 4); 2) According to the fossil Gleichenia, node 14 were dated to have been originated 99.6 Mya . Moreover, following previous estimates, another node 21 and 22 were separately dated to 110.5 Mya and 42.5 Mya at the beginning of the running: 3) Pteridaceae clade  and 4) Polypodiaceae clade .
According to authors’ suggestion , to avoid the misspecification of dating and taxon sampling, the empty alignment was run before the real MSAs. Then BEAST 1.7 was allowed to infer topology, branch lengths, and dates for combined datasets . A uniform distribution is applied over the estimating of the absolute ages via the MCMC process. For each MSA, BEAST runs 6 × 107 generations, saving data every 1,000 generations, producing 60,000 estimates of dates under a Yule speciation prior under the uncorrelated lognormal distributed relaxed clock model. Convergence statistics was analyzed in Tracer v 1.5, resulting in 54,000 post-burn-in trees. Before the consensus tree was graphically illustrated by Figtree v.1.3.1, TreeAnnotator v.1.6.1 was utilized to produce maximum clade credibility trees from the post-burn-in trees and to determine the 95% probability density of ages for all nodes in the phylogenetic tree.
Detection of positively and negatively selected codons
Since identification of positive/negative (non-neutral) evolution is fundamental to the understanding of the process of diversifying/purifying selection, this subject has been the focus of several decades of mathematical and computational efforts. Different scientists have developed numerous analysis models and methods, and each has its own advantages [42, 43, 44, 65]. However, the general consensus of them is that non-synonymous nucleotide substitutions (dN), whose alternatives leading to a change in the codon and its corresponding amino acid, can be time-scaled by the number of synonymous replacements (dS), which are nucleotide changes that only change the codon but not the amino acid and are consequently neutrally fixed and proportional to the divergence time between the sequences.
The random-site models (M0 vs. M3, M1a vs. M2a, M5 vs. null test, M7 vs. M8, and MEC vs. null test), contained in PAML package version 4.1 and Selecton version 2.2, allow the ω ratio (ω = dN/dS) to shift among codons within the MSA and this parameter is estimated by maximum-likelihood value via Bayesian inference approach [41, 42]. Additionally, the results from Selecton version 2.2 are visualized with seven-color scale for representing the different types of selection. To identify the statistical significant levels of the results, the LRT was conducted to compare the nested models .
Besides, another three models for detecting codons under selection are implemented on Datamonkey website : SLAC, FEL and REL. SLAC originated from the Suzuki–Gojobori counting approach  and is quite efficient on detecting non-neutral evolution in large MSAs. Less conservative than SLAC, FEL fits a site-specific dN and dS in the context of codon substitution models and tests whether dN = dS, outperforming other two models in MSAs of intermediate size. As the most powerful model of the three, REL is improved based on the Nielsen–Yang approach . Before the analysis of Natural selection, the best-fit nucleotide substitution models in the MSAs were calculated via the model-selection molecule online. Different from the posterior LRTs in the nested models, the parameter for statistical significant level (p value or Bayesian factor) was pre-set prior to the estimating processes .
Analysis of inter-dependent evolutionary sites
To understand the broad implications of the amino acid replacements in D1 proteins from fern species, the analysis of the evolutionary dependencies among sites to identify functional/structural dependencies among residues were conducted via five.
Based on a tree-ignorant strategy, CAPS outperforms the tree-aware strategy methods as reported by previous published work , which compares the correlated variance of the evolutionary rates at 2 sites corrected by the time since the divergence of the 2 sequences . The significance of the results was evaluated by randomization of pairs in the alignment, calculation of their correlation values, and comparison of the real values with the distribution of 10,000 randomly sampled values. The step-down permutation procedure was employed to correct for multiple tests and non-independence of data . An alpha value of 0.001 was applied to minimize Type I errors. The correlated variability between amino acid sites was weighted by the level of substitutions per synonymous site in order to normalize parameters by the time of sequence divergence .
The mathematical models for detecting the co-evolving residues in protein from InterMap3D and Datamonkey websites are based on a tree-aware strategy [47, 73]. Currently, three different models, namely Row and Column Weighing of Mutual Information (RCW-MI), Dependency and Mutual Information/Entropy (MI/E), were implemented at the former website , and one (Spidermonkey) at the later [43, 73]. RCW-MI and Dependency extract both the entropy dependency and the phylogenetic signal [74, 75], while the MI/E extracts the entropy dependency from the signal by dividing mutual information by the joint site’s entropy [76, 77]. Spidermonkey molecule reconstructs the substitution history of the MSAs by ML-based methods, and analyses the joint distribution of substitution events by Bayesian graphical models .
We thank Dr. Yefu Wang at State Key Laboratory of Virology, College of Life Sciences, Wuhan University, China, for the guidance for wording, and we also thank Bo Wang and Dr. Lei Gao at the Wuhan Botanical Garden, Chinese Academy of Sciences, China, for the experimental assistance. We thank Dr. Jianqiang Li at the Wuhan Botanical Garden, Chinese Academy of Sciences, China, for the advices on species sampling. We appreciate two anonymous reviewers and other editors for their helpful suggestions. The present work was financially supported by the National Nature Science Foundation of China (No. 30970290, 31070594), and by the Knowledge Innovation Program of the Chinese Academy of Sciences (No. KSCX2-EW-J-20, KSCX2-YW-Z-0940). Dr. Mario A. Fares was supported by Spanish Ministerio de Ciencia e Inovación (No. BFU2009-12022) and Research Frontiers Program (No. 10/RFP/GEN2685) from Science Foundation Ireland.
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.