Introduction

As sessile organisms, plants do not have an immune system like animals. Therefore, when exposed to biotic or abiotic stresses, they have evolved some adaptive ways to deal with these environmental stresses (Jacobs et al. 1999), including hypersensitive response (HR), lignification of cell wall, and synthesizing a variety of proteins or compounds, such as, antioxidants (like reactive oxygen species, ROS), anti-microbial compounds (like phytoalexins), late embryogenesis-abundant (LEA) proteins, proline, sugars and pathogenesis related (PR) proteins. Based on amino acid composition, serological and biochemical properties, PR proteins have been classified into 17 different families, including β-1,3-glucanases (PR-2), chitinases (PR-3, 4, 8 and 11), thaumatin-like proteins (TLPs) or osmotin (PR-5), proteinase-inhibitor (PR-6), endoproteinase (PR-7), peroxidase (PR-9), defensins (PR-12), thionins (PR-13), lipid-transfer proteins (PR-14), etc. (van Loon et al. 2006).

Plant TLP belongs to the PR-5 family. Historically, it has been called TLP/PR5 or osmotin/osmotin-like protein (OLP). Here, the nomenclature TLP is used to represent this gene family. Most TLPs have 16-cysteine residues, which might form eight disulfide linkages. This structure can stabilize protein, and then let it to resist to pH, proteases and heat-induced denaturation (Ghosh and Chakrabarti 2008). Some smaller TLPs (only containing ten conserved cysteine residues) have also been identified in conifers and monocots (Fierens et al. 2009; Liu et al. 2010a; Petre et al. 2011).

TLPs are involved in plant defense system against various biotic and abiotic stresses (Petre et al. 2011). Over-expression of TLPs can induce stress resistance in different transgenic plants (Liu et al. 1994; Rajam et al. 2007; Datta et al. 1999; Wang et al. 2010; Munis et al. 2010; Subramanyam et al. 2012; Acharya et al. 2013). TLPs can inhibit hyphal growth or spore germination by a membrane permeabilizing mechanism (Abad et al. 1996) or by degradation of cell walls (Osmond et al. 2001; Zareie et al. 2002). In addition to antibiotic activities, TLPs have also been involved in other physiological and developmental roles, including antifreeze activities (Yu and Griffith 1999), abiotic stress tolerance (Subramanyam et al. 2012; Zhu et al. 1995; Zhang and Shih 2007), floral organ formation and fruit ripening (Neale et al. 1990; Salzman et al. 1998), seed germination (Seo et al. 2008), senescence (Sakamoto et al. 2006), and glucanase activity (Osmond et al. 2001; Grenier et al. 1999).

The recent availability of genome sequences of some models plant species provides an opportunity to study the evolution of TLP gene family. Considering the important roles associated with antibiotic activities and developmental and physiological functions, and the number of the TLP genes varied greatly among plant species, it’s of considerable interest to us to investigate how the TLP genes have evolved in Plantae. Here, our results suggest that the TLP gene family has an expansion process in number, and that tandem and segmental duplications play dominant roles in it. Our studies also reveal different expression profiles of the TLP genes in rice and functional network features of the TLP protein in Arabidopsis.

Materials and methods

TLP sequences retrieval and identification in six plant species

We used the Arabidopsis TLP sequences (Liu et al. 2010b) as queries to perform BLAST searches against EnsemblPlants database (http://plants.ensembl.org/index.html) to identify potential members of the TLP gene family in these species. Next, the Conserved Domain Database (CDD) (Marchler-Bauer et al. 2013) was used to confirm whether the returned sequences from such searches encode TLP domain. Predotar program (Small et al. 2004) was used to perform the protein subcellular location.

Phylogenetic analyses of the TLP gene family

We used MUSCLE 3.52 (Edgar 2004) to perform multiple sequence alignments of full-length protein sequences, and used MEGA v5 (Tamura et al. 2011) to carry out phylogenetic analyses of the TLP proteins based on amino acid sequences with the neighbor joining (NJ) method. NJ analyses were done using p-distance methods, pairwise deletion of gaps, and default assumptions. Support for each node was tested with 1000 bootstrap replicates.

Estimation of the maximum number of gained and lost TLPs

We first divided the phylogeny into different clades to determine the expansion degrees of the gene family in these species (Cao et al. 2015). Nodes were labeled as V: Viridiplantae; Em: Embryophyte; A: Angiosperm; M: Monocots; and E: Eudicots. Notung v2.6 (Chen et al. 2000) was used to infer gene duplication and loss events by reconciling the gene tree with the species tree.

Chromosomal location of the TLP genes and genomic duplication

Annotation information on TAIR (http://www.arabidopsis.org), Rice Genome Annotation Project (http://rice.plantbiology.msu.edu/index.shtml), Populus genome browser (http://www.phytozome.net/poplar) and MaizeSequence (http://www.maizesequence.org) was used to determine the chromosomal locations and intron–extron structures of the TLP genes. SyMAP v3.4 (Soderlund et al. 2011) was used to depict the paralogous regions of the putative ancestral constituents of the genomes. In this study, two patterns of gene expansion (tandem duplication and segmental duplication) were focused on. Tandem duplicated genes were defined as adjacent homologous genes on a single chromosome, separated by no more than one nonhomologous spacer gene (Hanada et al. 2008). Moreover, some tandem duplicated genes were further confirmed in the plant tandem duplicated genes database (PTGBase) (Yu et al. 2015). Segmental duplications of each TLP gene within the family in poplar, rice, maize, and Arabidopsis genomes were searched in the SyMAP v3.4 (Soderlund et al. 2011).

Microarray-based expression analysis

Plant Expression Database (PLEXdb, http://www.plexdb.org/index.php) (Dash et al. 2012) was used for the expression analyses of rice TLP genes. In this study, six experiments (OS8, OS10, OS25, OS85, OS65, and OS92) were selected. And the genesis (v 1.7.6) program (Sturn et al. 2002) was used to normalize the expression data.

Positive selection assessment

We first used the Selecton Server (http://selecton.tau.ac.il/) (Stern et al. 2007) to calculate site-specific selection. Four evolutionary models (M8, M8a, M7 and M5) were used in this study. Each of the models uses different biological assumptions to test different hypotheses. In addition, we also used FEL (fixed-effects likelihood), SLAC (single likelihood ancestor counting), and REL (random-effects likelihood) methods with default settings embedded in the Datamonkey web interface (Delport et al. 2010) to further identify selection in individual codons. Finally, PARRIS was also used to test for the signatures of selection.

Co-expressed network assembly

We used ANAP (Wang et al. 2012), a co-expressed network designed to convert and then integrate 11 Arabidopsis data sets, to analyze TLP co-expressed network in Arabidopsis. Arabidopsis TLP genes were mapped to their corresponding proteins in the network database. Seventeen TLPs were not present in the assembled network database. Resulting interactions were used to build the seven members interaction network.

Results and discussion

Identification of the TLP genes in Arabidopsis, rice, poplar, maize, Physcomitrella and Chlamydomonas genomes

We identified 24 TLP genes in Arabidopsis, 44 in rice, 49 in poplar, 49 in maize, 6 in Physcomitrella and only 1 in Chlamydomonas (Table 1). Compared with other four genomes as described above, Physcomitrella and Chlamydomonas species encoded a much smaller number of TLP genes. This suggested that the expansion of TLP family mainly occurred after the divergence of the Embryophyte leading to vascular plants. The number of TLP genes in rice, poplar and maize is very similar, which are about twice as many TLP genes than that in Arabidopsis.

Table 1 Number of TLP genes of Arabidopsis, rice, poplar, maize, Physcomitrella and Chlamydomonas in Groups I–VI
Full size table

Phylogenetic analyses and comparison of TLP proteins

A phylogenetic analysis of the predicted TLP protein sequences was performed based on the NJ method. As displayed in Fig. 1, tree branches are colored by species, TLP subclass, the number of introns, and predicted targeting to organelles. TLP genes can be classed into six groups based on their phylogenetic relationships (Fig. 1b). We found that most of the rice TLP proteins belong to the Group II and III. 14 of 15 TLP members come from poplar in Group IV, suggesting that species-specific expansion has happened in these groups (Fig. 1a, b).

Fig. 1
figure 1

NJ distant tree of all TLPs in Arabidopsis, rice, poplar, maize, Physcomitrella and Chlamydomonas. Terminal markers are colored to indicated: a species (Arabidopsis bright green; rice pink; poplar blue; maize black; Physcomitrella red; Chlamydomonas yellow); b subclass (Group I orange; Group II turquoise; Group III crimson; Group IV yellow; Group V deep yellow; Group VI bright green); c number of introns (genes with no intron gray; one intron yellow; two introns turquoise; three introns pink; four introns blue; more than four introns red); and d predicated organelle targeting (ER green; mitochondrial red; plastid blue; elsewhere gray), some no-Met proteins are discarding and are not shown in here

Full size image

Figure 1c displays the proportions of TLP genes with no intron, one intron, two introns, three introns, four introns and more than four introns in each species. The large majority of the TLP genes contain one or two introns in two eudicots (Arabidopsis and poplar). But, most of TLP genes in two monocots (rice and maize) are intronless. In addition, we also found that two lower plants (Physcomitrella and Chlamydomonas) have more introns. For example, EDP07492 and PP1S412_14V6.1 genes possess five introns, respectively. As we know, introns are important component of eukaryotic genes, and their loss or gain affect the complexity of genetic structure (Koonin 2006). Our results indicated that intron loss/gain events have occurred during the expansion and evolution of TLP paralogs.

Next, we also used the program Predotar (Small et al. 2004) to predict the organelles targeting of the family proteins. As a result, most of the TLP proteins were predicted to be targeted to the endoplasmic reticulum (ER) (Fig. 1d). Proteins targeted in the ER are usually experienced some protein processing, such as glycosylation, disulfide bond formation, folding and so on. Finally, these modified proteins are transported to their destinations when the signal peptides are removed (Trobetta 2003). Moreover, over 80 % of TLP proteins possess the signal peptide identified by the SignalP 4.0 server (Petersen et al. 2011).

Contrasting changes in the numbers of TLP genes

To better understand how TLP genes have evolved in these species, we estimated the number of TLP genes in the most recent common ancestor (MRCA) of Viridiplantae. There were about five ancestral TLP genes in the MRCA of Viridiplantae (V5) by reconciling the gene trees with the species phylogeny. Furthermore, we only identified one orthologous gene in the C. reinhardtii, implying that four of these five ancestral TLP genes have been lost when chlamydomonas is appearing (Fig. 2). The number of TLPs remained relatively stable before Angiosperm. Only after the emergence of Angiosperm species did TLPs once more expand significantly. It suggested that there were about 23 ancestral TLP genes in the MRCA of the green flowering plants. Interestingly, when compared with the MRCA of eudicots and monocots, it appeared that the expansion was uneven before their divergence. The MRCA of monocots has increased in size as much as two times (23/43), while the MRCA of eudicots has a similar number of TLP genes with that of Angiosperm (23/22). The expansion was also unbalanced between plant species since the divergence of eudicots and monocots (Cao and Li 2015). For example, poplar increased over two times (22/49) in size, while Arabidopsis only added two TLP genes (Fig. 2). When compared with the number of ancestral genes, it appeared that, except chlamydomonas, the TLP family had expanded in all the tested species. For instance, there are 24, 49, 44 and 49 genes in Arabidopsis, poplar, rice and maize, respectively; while the estimated number of genes in the MACA of eudicots and monocots is 23. Therefore, Arabidopsis, poplar, rice and maize have netted 1, 26, 21 and 26 genes, respectively, since their splits. Obviously, the numbers of genes gained in the poplar, rice and maize lineages are much greater than that in the Arabidopsis lineage.

Fig. 2
figure 2

Gene gain and loss of TLPs in the evolution of plants. The names of internal nodes are abbreviated (V Viridiplantae, Em Embryophyte, A Angiosperm, M Monocots, E Eudicots). The numbers of common ancestors at the five internal nodes (V, Em, A, M and E) are shown in the quadrates. Numbers after the plus signs the numbers of gene gain events, whereas numbers after the minus signs gene loss events

Full size image

Chromosomal location and duplication events of the TLP genes

Next, we also investigate the phylogenetic relationship and chromosomal location of each TLP gene. The results indicated that the TLP genes unevenly distributed among different chromosomes of these genomes (Fig. S1), and that the generation of 7 (29.2 % of 24) Arabidopsis, 25 (56.8 % of 44) rice, 28 (57.1 % of 49) poplar and 20 (40.8 % of 49) maize TLP genes could be explained by tandem duplication (Fig. S1). The largest TLP gene clusters are located on chromosome 1 of poplar genome and contain 13 tandem arrayed members, i.e. POPTR_0001s22810.1, POPTR_0001s22830.1, POPTR_0001s22850.1, POPTR_0001s22860.1, POPTR_0001s22870.1, POPTR_0001s22880.1, POPTR_0001s22890.1, POPTR_0001s22900.1, POPTR_0001s22910.1, POPTR_0001s22920.1, POPTR_0001s22930.1, POPTR_0001s22950.1 and POPTR_0001s22960.1 (Fig. S2). Moreover, these genes form a single clade, suggesting that they may come from the recent tandem duplications (Fig. S2). In addition to tandem duplication, segmental duplications also played an important role in the expansion of the TLP family gene. At least 3, 2, 7, and 4 pairs of paraloguos genes come from segmental duplication in Arabidopsis, rice, poplar and maize, respectively (Fig. S1). Within the identified duplication events, some pairs are retained as duplicates, whereas others lost them. It is likely that dynamic changes have occurred following segmental duplication. Therefore, tandem duplication and segmental duplication are the major factor that contributed to the expansion of this gene family.

Expression of the TLP gene family in rice

Expression profiling is a useful tool for understanding gene function (Durick et al. 1999). To assess the transcriptional characteristics of the TLP genes, we examined some publicly databases in rice. First, we analyzed the spatial- and temporal-specific expression profiles of rice TLP genes in embryo (6 days), endosperm (6 days), root, leaf and seedling. All of the 37 detected transcripts were divergent expressed in different tissues (Fig. S3). Some members of rice TLP gene (such as LOC_Os12g43490, and LOC_Os09g36580) were expressed at the highest levels in the root, implying that they may be involved in the root development. In addition, LOC_Os06g47600 and LOC_Os10g05600 genes were higher expressed in the embryo, suggesting that these TLPs might be associated with early embryonic development of rice. Similar results have also been observed in their homologs in Arabidopsis, which were highly expressed during seed germination (Seo et al. 2008).

Plant growth is affected by some abiotic cues (Han et al. 2015; Jayakannan et al. 2015; Liu et al. 2015). Here, we also examined the expression profiles of rice TLP genes under drought, salt, cold and heat shock stresses. Divergent expression patterns were present among TLP members when exposed to these stress conditions (Fig. S3). Four TLP genes (LOC_Os12g43450, LOC_Os03g46060, LOC_Os07g23470, and LOC_Os03g14050) displayed higher expression levels in these conditions. Interestingly, we also found that, compared with the control, over 70.2 % of rice TLP genes showed higher expression levels under heat shock stress, implying that most TLP genes might be involved in the heat shock response. Infections of some pathogenic bacteria and insect pest are key factors affecting crop quality and yield. Next, we examined some experiments infected by Xanthomonas oryzae pv. oryzae (Xoo) and Blumeria graminis (Bgh) and found that over 67.5 and 75.6 % rice TLP genes exhibited an increase in expression levels under X. oryzae and B. graminis infection, respectively (Fig. S3). In addition, we also examined the expression levels of TLP gene under an insect pest, striped stem borer (SSB) (Fig. S3). The results indicated that expression levels of over 59.4 % TLP genes were increased when attacked by this pest. An increasing number of evidence has suggested that TLPs may function in both biotic and abiotic stress tolerance. Previous studies reported that transgenic plants over-expressing TLP proteins showed enhanced resistance to Alternaria alternate (Velazhahan and Muthukrishnan 2003), Fusarium graminearum (Mackintosh et al. 2007), Verticillium dahliae (Munis et al. 2010), Phaeseoropsis personata (Singh et al. 2013), and so on. Moreover, over-expression of some TLP proteins could confer tolerance during salt, drought and other stresses (Rajam et al. 2007; Munis et al. 2010; Wang et al. 2011; Singh et al. 2013). In addition, several TLPs have been reported to be induced during insect attack (Johnson et al. 2011; Singh et al. 2013). Our study also indicated that most rice TLPs can be induced by these abiotic and biotic stresses, suggesting that they are likely to be required for enhancing resistance to stress.

Different selection regimes in different groups and amino acid sites

K a /K s ratio measures selection pressure on amino acid substitutions. A K a /K s ratio greater than 1 suggests positive selection and a ratio less than 1 suggests purifying selection. The amino acids in a protein sequence are expected to be under different selective pressures and to have distinct K a /K s ratios. To analyze positive or negative selection of specific amino acid sites within the full-length sequences of the TLP proteins in different groups, substitution rate ratios of nonsynonymous (K a ) versus synonymous (K s ) mutations were calculated with the Selecton Server (http://selecton.tau.ac.il) using a Bayesian inference approach (Stern et al. 2007). We performed the tests using four evolution models [M8 (ωs ≥ 1), M8a (ωs = 1), M7 (beta) and M5 (gamma)] implemented in this server. Selection models M8a and M7 do not indicate the presence of positively selected sites, whereas the M8 and M5 models do (Table S1). Moreover, statistical significance of positive selection has been testing for the identified positively selected sites. The results indicated that the K a /K s ratios of the sequences from different TLP groups were significantly different (Table S1). Higher K a /K s values existed in Group IV, indicating a higher evolutionary rate or selective relaxation within members of the Group IV. On the other hand, the K a /K s values in Group I are relatively small, implying a lower evolutionary rate or selective constraint within Group I members. However, despite the differences in K a /K s values, all the estimated K a /K s values are substantially lower than 1, suggesting that the TLP sequences within each group are under purifying selection pressure and that positive selection may have acted only on a few sites during the evolutionary process (Table S1). In addition, we also used SLAC, FEL and REL methods with default settings implemented in the Datamonkey web interface (Delport et al. 2010) to further identify selection in individual codon. The results were shown in Table 2. All the K a /K s ratios were less than 1, indicating that most codons in TLP sequences were under purifying selecting in these six groups. The FEL software detected the largest number of potential positively selected sites for each group. However, SLAC and REL analyses only detected a few. In this study, we used two programs (Selecton and Datamonkey) including seven methods to detect positively selected sites and got similar selection pressures for each group (Tables S1 and 2). Detecting positive selection will help to understand functional residues and functional shift of protein (Loughran et al. 2012; Chen et al. 2014). In this paper, we found that a few sites might undergo positive selection in evolution (Tables S1 and 2); implying that positive selection on these sites might have accelerated functional divergence and then result in the formation of gene subgroups.

Table 2 Predicted positive selection sites and evidence for positive selection within different TLP gene family
Full size table

Functional network analysis of the TLP genes in Arabidopsis

Genes involved in related biological pathways are usually expressed cooperatively (Eisen et al. 1998). To further investigate which genes are possibly regulated with the TLPs, we assembled a co-expression network (Fig. 3). 7 of 24 TLPs were present in the network, which exhibited 245 physical or functional interactions with 188 genes. Molecular function analysis of these 188 genes showed that genes with ATP or DNA binding, protein binding, kinase activity, hydrolase activity, and transporter activity were overly represented. Among the 188 interactors identified, 102 and 64 genes co-expressed with AT1G18250 and AT4G38660, respectively. Plant resistance is very important to the growth of plant (An et al. 2015). Our co-expressed analysis also revealed that TLP genes might function with some pathogen resistance proteins. C ysteine-rich repeat-like kinases (CRK) is an important group of enzymes involved in pathogen resistance (Chen et al. 2004). Here, three CRKs (AT4G23210, AT4G23150 and AT4G23160) were identified to be co-expressed with the TLP proteins, implying potential interactions between the TLP and CRK genes. In addition, some proteins with transporter activity are also involved in plant resistance. Some of these co-expressed genes included EDS5 (AT4G39030), DND2 (AT5G54250) and TIL1 (AT5G58070). EDS5 encodes an orphan multidrug and toxin extrusion transporter, which is a necessary component of salicylic acid-dependent signaling for disease resistance (Ng et al. 2011). DND2 is a second cyclic nucleotide-gated ion channel gene for hypersensitive response (Jurkowski et al. 2004). TIL1 encodes a temperature-induced lipocalin, which is involved in the thermotolerance (Chi et al. 2009). Plant chitinases are involved in defense responses against pathogen attacks and in tolerance of diverse environmental stresses (Takenaka et al. 2009). A TLP, AT4G11650, was found to be co-expressed with one chitinase, AT3G54420. In addition, another chitinase, CHI (AT2G43570), might be the potential interactors of the TLP, AT1G75040. Thus, whether chitinase could serve as a link to TLP molecular pathways need further experimental confirmation. Although the exact pathways mediated by these genes were still unclear, we speculated that these TLP genes might play critical roles in plant resistance. These observations have led us to hypothesize that TLP could regulate these plant responses through its involvement in different signal pathway in plants. This contributes to the selection of candidate genes for further functional genomics.

Fig. 3
figure 3

Functional network assembly of the TLP genes in Arabidopsis. Seven Arabdiopsis TLP genes are mapped to the co-expression database. This analysis reveals a total of 188 unique genes that exhibits 245 physical or functional interactions, and a network is then assembled based on these interactions

Full size image

Conclusions

This study provided a comparative genomic analysis addressing phylogeny, chromosomal location and duplication, selective pressures, expression profiling, and functional network analysis. Phylogenetic analyses revealed six well-supported groups in the TLP family. The TLP gene family had a birth process only after the emergence of Angiosperm species. Tandem and segmental duplication played a dominant role in the expansion of this gene family. In addition, TLPs were under purifying selection according to estimations of the substitution rates of these genes. Furthermore, comprehensive analysis of the expression profiles provided insights into possible functional divergence among members of the TLP gene family. Functional network analysis was also identified some resistance genes, which might work together with the TLPs. These data may provide valuable information for future functional investigations of this gene family.