Evolution of the RALF Gene Family in Plants: Gene Duplication and Selection Patterns

Sequence identification and conserved motif analysis of RALF genes

To identify potential members of the RALF gene family in Arabidopsis, poplar, rice and maize, we performed multiple database searches. Published Arabidopsis RALF gene sequences ³² were retrieved and used as queries in BLAST searches against NCBI (http://www.ncbi.nlm.nih.gov) and Phytozome (http://www.phytozome.net). The program MEME ³³ (http://meme.sdsc.edu) was used to identify motifs in the candidate RALF protein sequences. MEME was run locally with the following parameters: number of repetitions = any, maximum number of motifs = 6, and with optimum motif widths constrained to between 6 and 200 residues.

Alignment and phylogenetic analysis

To generate the alignment of the 91 RALF proteins from the Arabidopsis, rice, poplar and maize, COBALT ³⁴ program was used. Phylogenetic analyses of the RALF proteins based on amino acid sequences were carried out using Neighbor-Joining (NJ) methods in MEGA 5. ³⁵ NJ analyses were done using p-distance methods, pairwise deletion of gaps, and the default assumptions that the substitution patterns among lineages and substitution rates among sites were homogeneous. Support for each node was tested with 1,000 bootstrap replicates.

Estimation of the maximum number of gained and lost RALFs

To determine the degrees of gene family expansion in the analyzed plant lineages, we divided the phylogeny into ancestral clades (those containing at least one representative of monocots and eudicots), recent clades (monocot specific or eudicot specific) and species-specific clades. Nodes basal to the split among lineages denoted the most recent common ancestor (MRCA) and were labeled as N0 to N4. Gene duplication and loss events were inferred by reconciling the gene tree for each cluster/subcluster with the species tree using Notung v2.5. ³⁶

Divergence levels analysis

To analyze positive or negative selection of the RALF sequences, substitution rate ratios of nonsynonymous (K _a ) versus synonymous (K _s ) mutations were calculated. We first identified the closest orthologs for each gene in the genome of the close relative A. lyrata (Fig. S6) and included only those A. thaliana genes that had a single ortholog in A. lyrata. Moreover, gene pairs were considered orthologs when they clearly formed a single subclade. Pairwise alignment of nucleotide sequences of the RALF orthologs was performed using MEGA 5. ³⁵ Alignments were performed using Clustal W (Codons). K _a and K _s values of the orthologous genes were estimated using K-Estimator 6.0. ³⁷ To calculate the K _a /K _s ratios in different Groups, the Selecton server^38,39 was also used. It implements several evolutionary models that describe, in probabilistic terms, how characters evolve. The models are expressive enough to describe the biological reality. In this study, five models (M8, M8a, M7, M5 and MEC) were used. Each of the models uses different biological assumptions so that different hypotheses can be tested.

Inference of duplication time

Pairwise alignment of nucleotide sequences of the RALF paralogs was performed using MEGA 5. ³⁵ Alignments were performed using Clustal W (Codons). The K _a and K _s values of the paralogous genes were estimated using K-Estimator 6.0. ³⁷ To better explain the patterns of macroevolution, estimates of the evolutionary rates were considered extremely useful. Assuming a molecular clock, the synonymous substitution rates (K _s ) of the paralogous genes will be expected to be similar over time. Thus, K _s can be used as the proxy for time to estimate the dates of the segmental duplication events. The K _s value was calculated for each of the gene pairs and then used to calculate the approximate date of the duplication event (T = Ks/2λ), assuming clock-like rates (λ) of synonymous substitution of 1.5 × 10⁻⁸ substitutions/synonymous site/year for Arabidopsis, ⁴⁰ 6.5 × 10⁻⁹ for rice and maize ⁴¹ and 9.1 × 10⁻⁹ for poplar ⁴² .

Codon bias analysis

Codon bias can reflect the degree of selective constraint in a gene. To measure the extent of codon bias, effective number of codons (ENC) and codon bias index (CBI) were estimated using DnaSP v.5.10.01. ⁴³ The ENC values range from 20 to 61, meaning from the maximum codon bias (only one codon is used for each amino acid) to no codon bias (all synonymous codons for each amino acid are equally used). ⁴⁴ The CBI values range from 0 to 1, meaning from uniform use of synonymous codons to maximum codon bias. ⁴⁵ We also estimated some parameters related to codon bias, such as GC1,2 (the GC content at the first and second codon positions), GC3 (the GC content at the third codon positions) using DnaSP v.5.10.01. ⁴³

Correlation analysis of expression data and protein subcellular localization

Expression profiling can provide useful clues to gene function. To examine the expression patterns of the RALF genes, a comprehensive expression analysis was performed using the publicly available microarray data from Genevestigator.^46,47 For genes with more than one set of probes, the median of expression values was used. Finally, the expression data were gene-wise normalized and hierarchically clustered based on Pearson coefficients with average linkage in the Genesis (version 1.7.6) program. ⁴⁸ Protein subcellular localization was predicted using WoFL PSORT software (www.wolfpsort.org). ⁴⁹

Identification, motif organization and phylogenetic analyses of the RALF genes

We identified 33, 23, 16, and 19 putative RALF genes from Arabidopsis, poplar, rice and maize, respectively. Arabidopsis has about doubled the collection of RALF genes than rice, whereas poplar and maize have fewer (30.3% and 42.4%, respectively) genes than Arabidopsis. By searching the PlantGDB (http://www.plantgdb.org), ⁵⁰ we found that the predicted genomes of poplar, rice and maize contain 45,778, 30,192 and 32,540 genes, respectively, which are 67.2%, 10.3% and 18.8% larger than that of Arabidopsis (27,379), respectively. This suggested that the numbers of the RALF genes are not proportional to the sizes of the predicted genomes. All the RALFs in the four species possess only one RALF domain through the CDD^51,52 and Pfam (http://pfam.sanger.ac.uk) analysis. While these tools are suitable for defining the presence or absence of recognizable domains, they are unable to recognize smaller individual motifs and more divergent patterns. Thus, we further used the MEME program ³³ to study the diversification of RALF genes in Arabidopsis, poplar, rice and maize. Six distinct motifs were identified in these genes (Table 1 and Fig. S1). Details of the six motifs were presented in Fig. S2.

OPEN IN VIEWER

Table 1

Number and motif structure of RALF proteins from Arabidopsis (At), poplar (Pt), rice (Os) and maize (Zm).

Group	At	Pt	Os	Zm	Structure
I	15	1	0	0	Motif 4-1 / Motif 5-1
II	7	0	0	0	Motif 5-6-1
III	0	0	3	2	Motif 5-1
IV	0	4	0	1	Motif 4-1 / Motif 5-1
V	1	5	0	0	Motif 5-3-1 / Motif 3-1 / Motif 5-3-2-1
VI	2	0	0	0	Motif 5-3-1
VII	0	0	10	13	Motif 5-2-1 / Motif 5-3-2-1 / Motif 5-6-2-1 / Motif 3-2
VIII	1	3	0	0	Motif 5-3-2-1 / Motif 5-2-1 / Motif 2-1
IX	1	2	3	3	Motif 5-2-1 / Motif 5-3-2-1 / Motif 2-1
X	6	8	0	0	Motif 5-3-2-1 / Motif 3-2-1 / Motif 2-1 / Motif 3

Phylogenetic analyses can allow us to identify evolutionarily conservative and divergent of gene family. To achieve this goal, phylogenetic analyses of the 91 RALF members were performed. Based on phylogenetic relationships, we divided the RALF members into 10 groups (Fig. S1). Most of these genes encode proteins with the same or similar motif organization, while others are scattered in the families formed by proteins with other motifs, suggesting their complex evolutionary history. For convenience, we categorized the ortholog clades into 3 classes: (i) superstable: clades with orthologs containing at least one representative of monocots and eudicots, (ii) stable: clades including orthologs with monocot specific or eudicot specific, and (iii) unstable: lineage-specific clades. From Figure S1, it was clear that the superstable clade (Group IX) contained similar numbers of genes from each species, suggesting that major expansion/contraction in gene number had not occurred since the divergence between eudicots (Arabidopsis and poplar) and monocots (rice and maize). This result was also consistent with the number of RALF genes in Selaginella moellendorffii, in which only two RALF genes were found (Table S1). Figure S1 also showed that some genes formed lineage-specific clusters. The largest of such cluster had seven Arabidopsis genes. Moreover, of 16 RALF genes in Group I, 15 genes came from Arabidopsis. All of these suggested that many subsets of the RALF gene family had experienced extensive gene duplications.

Contrasting changes in the numbers of RALF genes

To better understand how RALF genes have evolved in these species, we estimated the number of RALF genes in the MRCA of eudicots and monocots. Reconciliation of the gene trees with the species phylogeny suggested that there were about two ancestral RALF genes in the MRCA of eudicots and monocots (N1). Furthermore, we identified 5 orthologous genes in the eudicots MRCA (N2) and 11 in the MRCA of monocots (N3) (Fig. 1). We also found that the number of RALFs remained relatively stable through evolutionary history from the land plants (N0, Physcomitrella patens) to the vascular plants (N0, Selaginella moellendorffii) and the angiosperms (N1). Only after the separation of the eudicot and monocot species about 145 million years ago ⁵³ did RALFs once more expand significantly. When compared the number of ancestral genes with those in the extant species, it appeared that the RALF family had expanded in all the analyzed species. For example, the number of RALFs increased approximately 6.6-fold since the divergence of the various eudicot species from their respective MRCA in Arabidopsis. However, the expansion was uneven between these plant species. For example, there are 33, 23, 16 and 19 genes in Arabidopsis, poplar, rice and maize, respectively, while the estimated number of genes in the MACA of eudicots and monocots are two. Therefore, Arabidopsis, poplar, rice and maize have gained 31, 22, 14 and 17 genes, respectively, since their splits. Only one lost gene is found in poplar. Clearly, the numbers of genes gained in the Arabidopsis lineage are much greater than that in other three lineages. OPEN IN VIEWER

Evolutionary patterns of RALF gene family

It has been suggested that the Arabidopsis genome experienced three duplication events within the past 250 million years, ⁵⁴ while the rice genome is believed to have experienced a genome-wide duplication approximately 70 million years ago.^55,56 To investigate the relationship between the RALF genes and potential genomic duplications within the genome, the location of the genes in previously identified Arabidopsis and rice chromosomal duplications^57,58 was noted. The distributions of the RALF genes relative to the corresponding duplicated genomic blocks were also illustrated in Arabidopsis (Fig. S3) and rice (Fig. S4). This result suggested that the generation of 17 (50.0% of 34) Arabidopsis and 7 (43.7% of 16) rice RALF genes could be due to tandem duplication. In Arabidopsis, the largest RALF gene cluster was located on chromosome 2 and contained four tan-demly arrayed members: ie, At2g19020, At2g19030, At2g19040 and At2g19045 (Fig. 2). Phylogenetically, these four genes formed a single sub-clade in Group II, suggesting that they may result from recent tandem duplications. Because Group II also contains genes from other locations (At3g25165 and At3g25170 are located on chromosome 3, whereas At4g13075 is on chromosome 4), these genes may be the result of more ancient duplication events. OPEN IN VIEWER

While segmental duplications were not the major factors that led to the expansion of the RALF gene family, it might be that dynamic changes occurred following segmental duplication, leading to loss of many of the genes. In contrast to Arabidopsis and rice, where 50.0% and 43.7%, respectively, of the RALFs were arranged in tandem repeats as described above, considerably fewer RALFs were arranged in tandem repeats in poplar (22.7%) and maize (10.5%), indicating that, in these species, RALFs mainly emerged by mechanisms other than tandem duplication.

Next, we also investigated the distributions of the unstable and stable RALFs in Arabidopsis. This result indicated that unstable genes are strongly clustered (about 66.7%), while stable and superstable genes are evenly scattered (or only 38.1% genes clustered) over the chromosomes (Fig. 3). It is clear that the majority of unstable genes in Arabidopsis emerged after the most recent whole genome duplication event.^57,59 We also found that, with the exception of Arabidopsis, three other species did not contain unstable genes, indicating divergent expansion of the RALF genes in different higher plants. In summary, our results suggested that after stable evolution of the RALF gene family in Angiosperms that followed the divergence from Tracheophyta (such as, only two RALF genes are identified in Selaginella moellendorffii), dramatic expansion had been largely occurred. OPEN IN VIEWER

In addition, when distantly related species compared, the newly added genes tended to form species-specific clusters or sub-clusters in the eudicots. For example, seven Arabidopsis RALF genes formed the most basal cluster within Group II. In Group IV, four poplar RALF genes also clustered. This suggested that, as F-box genes, ⁵³ the RALF genes in different species might have been derived from a series of gene duplication events that occurred after the split of the different lineages. A similar situation was found in the well supported clade of the monocot genes, in which most of the maize and rice genes also formed species-specific clades (such as Groups III and VII, see Fig. S1).

The phylogenetic tree topology revealed several pairs of RALF members with a high degree of homol-ogy in the terminal nodes of each group, suggesting that they were putative paralogous pairs (Fig. S1). Totally, 13, 7, 6 and 3 pairs of putative paralogous RALF proteins were identified, accounting for more than 78.8%, 60.9%, 63.2% and 37.5% of the entire family in Arabidopsis, poplar, maize and rice, respectively, with sequence identities ranging from 30% to 100% (Table S2). These pairs of RALF members are evolutionarily very closely related, and each pair of genes has very similar structure (Fig. S1), indicating that they originated from duplications. About 38.4% of the paralogous RALF pairs in Arabidopsis have very consistent K _s values (from 0.66038 to 0.74663), suggesting that the duplication events in this species occurred within the last 22.01 to 24.89 million years. This period was consistent with the time when a recent large-scale genome duplication event was thought to have occurred in Arabidopsis.^42,60 We also found that duplication of three of six RALF pairs originated from the recent large-scale duplication events (about 15.4 million years ago) in maize. ⁴⁰ This suggested that, as plant Sm and OPT genes,^61,62 the recent genome wide duplication events contributed partially to expansion of the RALFs. In addition, in evolutionary terms, some of these RALF gene duplications appeared to have occurred relatively recently, such as Poptrdraft673738-Poptrdraft672089 (about 1.53 million years ago) and Poptrdraft578381-Poptrdraft578382 (about 0.8 million years ago). It might be associated with novel functional divergence and adaptation.

Since codon bias can provides some examples of weak selection at the molecular level. Moreover, several researches have verified that selection on synonymous sites is correlated with stability of mRNA secondary structure, translation efficiency and accuracy, ribosome traffic and protein folding.^63–65 We also verified the codon usage bias of RALF genes. Some information is list in Table S3. In which, CBI and ENC were calculated to measure the degree of codon bias. We can see that CBI showed a marked negative correlation with CBI, so, in this study, ENC was used to measure the degree of codon bias. To determine the relative effects of mutation pressure versus natural selection on codon composition, the relationship between GC3 content and GC1,2 content was examined. The result showed a tendency of positive correlation between GC3 and GC1,2, suggesting that the GC content is most likely the result of mutation pressure since natural selection acts differently on different codon position. In addition, we also confirmed that K _s was positively correlated with K _a (R ² = 0.655, P < 0.001), and very weakly negatively correlated with ENC and GC3 (but this was not significant) (Fig. S5), implying that codon bias might be a factor in K _s variation among RALF genes and might be under natural selection.

Different signatures of selection in RALFs

To examine whether RALFs confer adaptational properties, we determined K _a /K _s ratios for superstable, stable and unstable genes of A. thaliana with A. lyrata (Fig. S6). K _a /K _s ratios of 0.0269 for superstable RALFs (Fig. 4) strongly indicated purifying selective pressures. In contrast to that, unstable and stable genes seemed to be closer to neutral selection, as inferred by significantly higher K _a /K _s ratios (0.5257 and 0.4237) for stable genes and unstable genes, respectively (Fig. 4). We also analyzed the selection properties of the RALFs in different Groups. The results showed that the K _a /K _s ratios of the sequences from the different Groups were significantly different (Table S4). However, despite the differences in K _a /K _s values, all the estimated K _a /K _s values were substantially lower than 1, suggesting that the RALF sequences within each of the Groups were under strong purifying selection pressure and that positive selection might have acted on only a few sites during the evolutionary process. OPEN IN VIEWER

Different expression profiles of the RALFs in Arabidopsis

We also examined the expression patterns of the Arabidopsis RALF genes. The results indicated that the divergent expression profiles were present in stable and unstable RALFs across the eight tissues/developmental stages assessed. Furthermore, the stable genes in different evolutionary branches also displayed different expression patterns (Fig. 5). Whether do duplicated genes have similar expression patterns? To answer this question, we investigated their expression profiles and found that none of the pairs of genes shared similar expression patterns (Fig. 5), indicating that substantial neofunctionalization might have occurred during subsequent evolution of the RALF duplicated genes. It seems that the expression patterns of the paralogs have diverged during long-term evolution, suggesting functional diversification of the duplicated genes. ⁶⁶ Such a process ensures the duplicated genes to increase adaptability to environmental changes, thus conferring a possible evolutionary advantage. ⁶⁷ We also found that over 82% of the assessed genes were likely to be localized in the extracellular space. At2g32885, At2g19030, At2g19040, At1g61563, At1g61566 and At2g19045 have 100% probability of being localized to the extracellular space. For all the other RALFs, although the extracellular space was predicted as the most likely location, it was also possible that they were localized to the membranes of organelles such as the cytosol, vacuolar membrane or chloroplast. Taken together, while the selected RALFs showed similar subcellular localizations, they differed considerably in their expression profiles, indicating that possible functional diversification may be achieved by selection. OPEN IN VIEWER