Introduction

The groups of genes showing similarity with each other are referred to as a gene family, reflecting the assumption that all arise from a common ancestor. Gene families arise essentially by gene duplication, either by wholesale duplication of entire genes or by duplication and shuffling of exons from different genes (Conant and Wolfe 2008). Functional differences between duplicate genes can originate in several different ways, including mutations that directly impart new functions, subdivision of ancestral functions, selection for changes in gene dosage and gene regulation (Louis 2007). It has been established that the positive selection plays an important role in retention and diversification of duplicate copies in plant defense genes (Moore and Purugganan 2005).

The Soybean Kunitz super-family inhibitors are one of the important arsenals in plant defense mechanism. These proteins inhibit the enzymatic activity of digestive proteases of pests and pathogens important for their survival (Ryan 1990). Kunitz family inhibitors have been established as a rapidly evolving multigene family, and they are characterized by the presence of conserved N-terminal signature sequence (Talyzina and Ingvarsson 2006). Over the long period of time, these inhibitors have evolved from a common ancestral gene by gene duplication and gene conversion events and adapted themselves for defense role in different plant species as per biotic and abiotic stresses. In this process, sequence and thereby the structural and functional evolution in their specificities have occurred as per the local environmental conditions. In general, plant gene families are largely conserved. Much of the diversity is thought to have been caused by gene duplication and adaptive specialization of pre-existing genes. Neofunctionalization and subfunctionalization are considered to be important processes responsible for retention of duplicate genes (Flagel and Wendel 2009).

The miraculin-like proteins (MLPs) are a group of proteins which exhibit significant sequence identity (~39–55%) to miraculin protein. The native miraculin, a 24.6-kDa plant protein purified from red berries of Richadella dulcifica, possesses unique taste-modifying property (Theerasilp and Kurihara 1988; Hirai et al. 2010). Till date, there are no reports of any MLP having the taste-modifying property. Many of the characterized MLPs have been shown to have an important role in plant defense and possess trypsin inhibitory activity (Tsukuda et al. 2006). Recently, remarkable up-regulation of MLPs was observed at different stages of citrus huanglongbing (HLB) disease development (Fan et al. 2011). HLB is the most destructive citrus pathosystem threatening citrus production worldwide (Gottwald 2010; Callaway 2008).

Both miraculin and MLPs belong to Kunitz super-family and have shown amino acid sequence similarity (~30%) to soybean Kunitz family trypsin inhibitors. The plant Kunitz-type soybean trypsin inhibitors (STIs) are the proteins with a molecular mass of approximately 20 kDa with four cysteine residues arranged into two disulfide bridges (Richardson 1991). They play an important role in plant defense against pathogens and predators and are known to be involved in many biological functions, such as blood coagulation, platelet aggregation, and anti-carcinogenesis (Kennedy 1998; Oliva et al. 2000; Ryan 1978). The notable variations, however, have been reported among Kunitz inhibitors. Crystal structures of Kunitz-type inhibitors having single or no disulfide bridge have been reported (Cavalcanti et al. 2002; Hansen et al. 2007). The variations in sequence and structure with functional diversifications indicate that members of Kunitz family have undergone rapid evolutionarily changes because of varying selective pressures. Plant proteinase inhibitors are encoded by multigene family and have rapidly evolved in response to different biotic and abiotic stresses (Talyzina and Ingvarsson 2006).

The comparison of known MLPs with classical Kunitz family members has shown significant variations in sequence, structure, and function. The present study elucidates the molecular evolution of MLPs in soybean Kunitz super-family. The analysis was performed on 34 gene sequences which included newly cloned Rutaceae MLP type 1 and 2 genes, known MLPs, native miraculin, typical Kunitz family inhibitors which showed considerable similarity with MLPs, and classical STI. Our results demonstrate the sequence and thereby the structural and functional divergences of MLPs which have evolved into a distinct group within Kunitz super-family.

Materials and Methods

Genomic DNA was isolated from leaves of different members of Rutaceae family using CTAB plant genomic DNA isolation kit (Bangalore Genei, Bangalore, India). Two sets of primers were used to amplify orthologous and paralogous genes. The first set of primers designated MLP type-I were designed on basis of Murraya koenigii miraculin-like protein (MKMLP) gene sequence [MLP-1 Forward: 5′-AAT ACC ATG GGA TCC TTT GCT TGA TAT CAA TG-3′; MLP-1 Reverse: 5′-AAT ACT CGA GTC AAG ACA CGC ATG AG-3′]. PCR conditions were 94°C (4 min) 94°C (1 min)/62°C (1 min)/72°C (1 min) for 30 cycles, followed by 72°C (10 min). The second set designated as MLP type-II was designed targeting the conserved sites of MLPs that are available in the database [MLP-2 Forward 5′-CCT TCT TTC CTT ATC CTT (AG)CC TTG (AG)CC (AT)CA-3′; MLP-2 Reverse 5′-AAC CA(GC) AA(GC) CAG ACG T(AC)G AAC GCC ATC-3′]. PCR conditions were 94°C (4 min) 94°C (1 min)/66°C (1 min)/72°C (1 min) for 30 cycles, followed by 72°C (10 min). PCR products obtained were purified from 1% agarose gel using a gel elution kit from Zymo research and sequenced three times from Ocimum Biosciences, Hyderabad, India. The obtained sequences were submitted to NCBI Gene databank.

Multiple sequence alignments were made using ClustalW (http://www.ebi.ac.uk/Tools/msa/clustalw2/) taking default parameters. Coding sequences lacking signal sequences and few residues of C-terminus were taken for analysis. Sequences showing significant similarities to MLPs were taken from NCBI database (http://www.ncbi.nlm.nih.gov/) (Table 1). Phylogenetic trees were constructed by means of the MEGA version 5 program from amino acid alignments using the Maximum Likelihood method based on the JTT matrix-based model (Tamura et al. 2011). The reliability of the branching was tested by bootstrap statistical analysis (1,000 replications). Synonymous and non-synonymous substitution rates were calculated from a set of codon-aligned nucleotide sequences through SNAP server (Korber 2000). Selecton version 2.4 server was used for the identification of site-specific positive selection and purifying selection in a protein (Stern et al. 2007).

Table 1 Sequences retrieved from NCBI database for phylogenetic analysis with accession numbers
Full size table

Structural modeling of a representative member of type 2 MLP protein of Citrus jambhiri (Cj2) was done using Modeller9v8 program (Sali and Blundell 1993), taking MKMLP structure (PDB code-3IIR chain A) (Gahloth et al. 2010) as template. The best model was selected using Procheck after evaluating the stereo-chemical quality. Protein–protein docking of MKMLP and porcine trypsin (PDB code-1AVW chain A) was done through ZDOCK server (Chen et al. 2003), In Trypsin, active site residues (His57, Asp102, and Ser195) and in MKMLP putative reactive loop and N-terminal region are chosen to be present in interaction region. Various types of protein interactions between MKMLP and trypsin complex were predicted by PIC server (Tina et al. 2007). Tools from Expasy server were used to identify signal peptides, conserved domains, and various types of sequence motifs (Gasteiger et al. 2003).

Results

The genomic DNA was used for amplification of new MLP genes as they do not contain introns (Gahloth et al. 2010). Fourteen new MLP genes were amplified from members of Rutaceae family plants. Among them, six genes were obtained from the first set, and eight from the second set of primers. Around 516 and 519–522 base pair fragments were obtained after sequencing MLP type 1 and 2 genes, respectively. All the 14 gene sequences have been submitted to the NCBI Gene databank (Table 2).

Table 2 MLP sequences from different Rutaceae members submitted to NCBI database with accession numbers
Full size table

The dataset used for sequence and phylogenetic analysis included new Rutaceae MLP type 1 and 2 genes, known MLPs, native miraculin, typical Kunitz family inhibitors which showed considerable similarity with MLPs and classical STI. In total, 34 sequences were used in the analysis. Maximum Likelihood tree constructed from the complete dataset divided the above sequences into four major groups where group I and II cluster together suggesting that they belong to a common gene family (Fig. 1). Group I included type 1 MLPs, group II included type 2 MLPs, and group III consisted of native miraculin, and stress-induced MLPs. Talisin shared the common ancestor of Group I and II but formed separate group. Classical soybean Kunitz inhibitor did not feature in these three groups and formed a separate group. MLPs in group I and II represent orthologous gene copies separately in the respective groups.

Fig. 1
figure 1

Phylogenetic tree of MLP sequences was constructed by the maximum likelihood method dividing them into three groups. The numbers above and below the branch points indicate the confidence levels for the relationship of the paired sequences as determined by bootstrap statistical analysis. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site

Full size image

Amino acid sequence analysis of all the new and old sequences presented interesting results. Multiple sequence alignments of new Rutaceae type 1 and 2 MLP sequences separately showed 98 and 75–95% identities, respectively among themselves, while 49–58% identity was observed between the two groups (Fig. 2). The sequence comparison of I and II groups with known MLPs and native miraculin protein showed ~40–95 and ~35% identities, respectively. With classical Kunitz inhibitor like STI, ~30% identity was observed with both group I and II MLP sequences. Group III MLPs among themselves showed 42–95% identity. Comparison with group I and II showed ~39–45 and ~30–43% identities, respectively (Fig. 3). Except for native miraculin, no other group III MLPs showed the presence of two histidines (His30 and His60 miraculin numbering), which is considered to be responsible for taste modifying property.

Fig. 2
figure 2

Multiple sequence alignments of new Rutaceae type 1 and 2 MLP sequences. Multiple sequence alignments of sequences (submitted to NCBI GenBank) were done by ClustalW, deleting the signal sequences and few C-terminal residues. The predicted Glycosylation site (bold), STI (Kunitz) protease inhibitors family signature (gray shade), and phosphorylation site (underlined) are indicated. The putative P1 residue is shown by the arrow, and conserved cysteines are shown in boxes. Abbreviations used: ClI, C. limonia MLP-I; ClII, C. limonia MLP-II; CaI, C. aurantifolia MLP-I; CaII, C. aurantifolia MLP-II; CmI, C. maxima MLP-I; CmII, C. maxima MLP-II; AmII, Aegle marmelos MLP-II; CrI, C. reticulata MLP-I; CrII, C. reticulata MLP-II; CsI, C. sinensis MLP-I; MpI, M. paniculata MLP-I; MpII, M. paniculata MLP-II; MkII, M. koenigii MLP-II; and CamII, C. aurantium MLP-II

Full size image
Fig. 3
figure 3

Multiple sequence alignments of MLP sequences from NCBI database and Rutaceae MLPs were done by ClustalW, deleting the signal sequences and few C-terminal residues. The predicted glycosylation site (bold), STI (Kunitz) protease inhibitors family signature (gray shade) and phosphorylation site (underlined) are indicated. Abbreviations used: MkI, M. koenigii MLP-I; MkII, M. koenigii MLP-II; CrI, C. reticulata MLP-I; CrII, C. reticulata MLP-II; Ccvs, Citrus cv shiranuhi MLP-2; Cxp2, Citrus × paradisi MLP-2; Cxp3, Citrus × paradisi MLP-3; Cxpm, Citrus × paradisi mRNA; Cu2, C. unshiu putative MLP-2; Cum, C. unshiu mRNA; Cj1, C. jambhiri Rlem MLP-1; Cj2, C. jambhiri Rlem MLP-2; Pt, P. trichocarpa Kunitz trypsin inhibitor; All, A. lyrata trypsin inhibitor; Sp, S. palustre MLP; Nt-T, N. tabacum tumor related protein; Nt, N. tabacum Kunitz trypsin inhibitor; Tb, T. bicolor trypsin inhibitor; and STI, G. max soybean trypsin inhibitor

Full size image

The primary amino acid sequences of the most of the MLPs possess a conserved Kunitz signature motif [L/I/V/M]-x-D-x-[E/D/N/T/Y]-[D/G]-[R/K/H/D/E/N/Q]-x-[L/I/V/M]-(x)5-Y-x-[L/I/V/M] at the N-terminal of the sequence (Laskowski and Kato 1980) (Fig. 3). The presence of three disulfide bonds is a typical feature of MLPs, and their positions in the most of the sequences of type 1 and 2 are conserved. In classical Kunitz inhibitors, only two conserved disulfides are present. Multiple sequence alignment comparative analysis of sequences at reactive site loop revealed major changes at both the primary and secondary specificity sites in groups I and II MLPs as compared with classical Kunitz inhibitors. Compared to native miraculin and classical Kunitz inhibitors like STI, the conventional Arginine/Lysine at P1 position has been replaced by the conserved Asn65 residue in both groups I and II MLPs (Fig. 2). The differences in secondary specificity sites at all the positions were observed in MLPs when compared with native miraculin and soybean Kunitz trypsin inhibitor. In MLP type 1, Tyr63 (P3), Asn64 (P2), Thr66 (P1′), and Ser67 (P2′); and in MLP type 2 Tyr63 (P3), Asp64 (P2), Ser66 (P1′), and Thr67 (P2′) are present in place of Pro61 (P3), Tyr62 (P2), Ile64 (P1′), and Arg65 (P2′) in STI. Likewise, native miraculin showed differences at secondary specificity sites where Asn66 (P3), Pro67 (P2), Glu69 (P1′), and Asp70 (P2′) are present in native miraculin. Interestingly, Asn13 which plays an important role in stabilizing the reactive loop conformation in STI is replaced by Ala13 and Ser13 in group I and II MLPs, respectively, whereas in group III MLPs, this position shows variability consisting of any one of Thr/Val/Pro/Ala (Fig. 3).

The sequences were compared for the presence of functional motifs. A significant feature is the presence of a phosphorylation motif, created by insertion of four amino acids at position 85 in type 2 MLP sequences, and its absence in type 1 MLPs. Also, differences in glycosylation sites were observed when group I and II MLPs were compared with native miraculin and STI. Glycosylation motif is located in the putative reactive loop of both type 1 and 2 MLPs (Fig. 2).

The three groups were analyzed separately as well as among themselves to assess the rates of synonymous and non-synonymous substitutions. The average dN and dS values compared within groups as well as for all pairwise comparisons across groups were compared, and Z-test results from MEGA 5 and SNAP server suggested that dS level was higher than dN implying that sequences are under purifying selection (Table 3). To determine site-specific positive Darwinian selection and purifying selection, a web-server Selecton version 2.4 was used. The ratio between non-synonymous (K a) and synonymous (K s) substitutions at each site of the protein is graphically displayed on each site using a color-coding scheme indicating either positive or purifying selection (online resource 1). For type 1, based on K a/K s ratios, a total of 11 residues (Asp16, Leu33, Leu40, Leu52, Glu82, Trp103, Thr130, Gln132, Gly133, Thr134, and Phe149) were identified to be undergoing positive selection (K a/K s ratio >1) in MEC model. The residues Glu82, Gly133, and Thr134 had high K a/K s ratios of 1.8, 1.5, and 1.8, respectively. For type 2 MLPs, a total of 15 residues (Gln16, Leu33, Tyr40, Thr52, Leu82, Gly85, Arg86, Asp87, Tyr88, Trp107, Asn134, Pro136, Gly137, Thr138, and Lys153) were identified to be undergoing positive selection (K a/K s ratio >1) in MEC model in Selecton server. The residues Leu82, Gly85, Arg86, Asp87, Pro136, Gly137, and Thr138 had high K a/K s ratios of 1.8, 1.5, 1.5, 2, 1.9, 1.5, and 1.8, respectively. For group III MLPs, a total of nine residues (Thr36, Val42, Val54, Phe85, Tyr108, Gly135, Ser136, Phe138, and Val153) were identified to be undergoing positive selection (K a/K s ratio >1) in MEC model in Selecton server. The residues Phe85, Gly135, and Phe138 had high K a/K s ratios of 1.8, 1.6, and 1.8, respectively. Significance test for positive selection was obtained by likelihood ratio test (LRT) and Akaike Information Criterion (AIC_c) (online resource 2). All the residues, except residues Leu33 in both type 1 and 2 MLPs, undergoing positive selection were located in the loop regions in crystal structures of MKMLP and Cj2 model, representatives of type 1 and 2 MLPs, respectively. Residue Leu33 was located in β-strand. It is interesting to note that there are four additional residues in type 2 as compared to type 1, which are undergoing positive selection. These residues (85–88) are particularly noteworthy as they constitute phosphorylation motif created by insertion of four amino acids and are absent in type 1 MLPs.

Table 3 Observed evolutionary properties of miraculin-like and its related proteins
Full size table

Secondary structure analyses of MLPs and STI were done by ESPRIPT server (http://espript.ibcp.fr/ESPript/ESPript/) taking one member from each group. It showed that overall structures are conserved with the presence of three disulfide linkages in MLPs but only two disulfide linkages are present in STI (Fig. 4). The presence of 12 β-sheets and two short helices are predicted in all the three group members.

Fig. 4
figure 4

Secondary structure analysis of MLPs was made by ESPRIPT server taking MKMLP structure (PDB id: 3IIR) as template. MkI (M. koenigii MLP), Cj2 (C. jambhiri Rlem MLP-2), Mir (miraculin) each representing group I, II, and III, respectively, and STI. β-Sheets, α helices, and disulfide linkages are shown as arrows, spirals, and numbers (1, 2, and 3), respectively. PS phosphorylation motif

Full size image

Overall model of Cj2 protein is similar to that of MKMLP consisting of β-trefoil fold made up of 12 antiparallel β-sheets connected by coils and two short stretches of α-helices (Fig. 5a). Procheck analysis shows only three residues (Ala120, His117, and Ala186) in disallowed regions in the model. The Cj2 protein possesses six cysteine residues leading to three disulfide bridges. The central core superimposes well with the MKMLP structure with RMSD of 0.178, but subtle deviations are seen in conformations of connecting loops (Fig. 5b). A loop having four residues (85–88) insertion possessing phosphorylation motif is observed.

Fig. 5
figure 5

a The three-dimensional model of Cj2 was created by comparative modeling using the template model of MKMLP [PDBcode: 3IIR] from M. koenigii by MODELLER9v8. Central β-sheets, two short helices, three cysteines, and putative inhibitory loop are shown. b Superimposition of C-α atoms of MKMLP 3D-structure and Cj2 model

Full size image

The MKMLP structure (PDB code: 3IIR Chain B) was docked on to the porcine trypsin (PDB code: 1AVW) by ZDOCK server to analyze the intermolecular contacts between MKMLP reactive loop and trypsin (Fig. 6a). The reactive loop interacting residues and types of interactions are shown (online resource 3). The mode of interactions observed in the MKMLP–trypsin complex at both primary and secondary specificity sites are significantly different from that of interactions in STI–trypsin complex. Asn65 (P1) in MKMLP is not able to interact extensively like Arg63 (P1) in STI with the residues forming S1 subsite of trypsin. It mainly interacts with Cys191, Gln192, Ser195, and Cys220 of trypsin (Fig. 6b) whereas Arginine in STI interacts additionally with Asp189, Ser190, Gly193, Asp194, Ser214, Trp215, Gly216, Gly219, and Gly226 of trypsin. It occupies the same position as Arginine or Lysine at P1 position in S1 subsite of trypsin, but it is too short to interact with Asp189, a key residue in binding pocket of trypsin. The geometry of the carbonyl group at the P1 position, important for the interaction between inhibitor and proteinase during catalysis, showed that carbonyl carbon atom is within van der Waals distance from Ser195 of trypsin. Also, carbonyl group forms a hydrogen bond with NH of Ser195 of trypsin. Asn64 (P2) does not interact with any of the residues of trypsin, whereas Tyr62 (P2) in STI interacts with His57, Phe94, Gly96, Leu99, Asp102, Gln192, Ser214, and Trp215 of trypsin. Tyr63 (P3) in MKMLP makes hydrogen bond with Gln192 of trypsin and interacts with Leu99, Trp215, and Tyr217 of trypsin, whereas Pro61 (P3) in STI interacts only with Trp215 and Gly216 of trypsin. Thr66 (P1′) in MKMLP interacts only with Ser195 of trypsin, whereas Ile64 (P1′) in STI also makes contacts with Phe41, Cys42, His57, Gln192, and Gly193 of trypsin. In P2′ position, Ser67 of MKMLP makes one hydrogen bond with His40 and interacts with Phe41, Cys42, and Cys58, whereas Arg65 in STI interacts with His40, Phe41, Gly193, and Tyr151 of trypsin. A total of 12 hydrogen bonds present between MKMLP and trypsin involve residues from reactive site loop.

Fig. 6
figure 6

a MKMLP–trypsin complex model: The MKMLP backbone is shown as cartoon and trypsin backbone is shown as C-α ribbon. Residues present at different positions (P1, P2, P3, P1′, P2′, and P3′) of the putative reactive loop of MKMLP are depicted as ball and stick representation. b Reactive loop of STI–trypsin complex superimposed on putative reactive loop of MKMLP–trypsin complex model. Arg63 of STI interacts with Asp189 of trypsin (lines), Asn65 of MKMLP cannot interact with Asp189′ of trypsin. Asn13 of STI interacts with Tyr62 and leads to stabilization of reactive loop. Interacting residues with P1 residues of STI and MKMLP are shown as lines

Full size image

Discussion

This study focuses on sequence and thereby structural and functional evolution of MLPs within Kunitz super-family. MLPs constitute an important group of plant defense proteins within Kunitz super-family. It is to be noted that most MLPs have been characterized from Rutaceae family plants (Tsukuda et al. 2006; Gahloth et al. 2010), whereas most typical Kunitz inhibitors including STI have been characterized from Leguminosae family plants (Oliva et al. 2010). This shows that proteins of Kunitz super-family have evolved with refined specificities in plants of each family to serve as defense agents. MLPs show ~30% identity with classical soybean Kunitz inhibitor (STI) and ~35% identity to native miraculin, also a member of Kunitz family. Most MLPs possess conserved Kunitz signature motif (Fig. 2) and many characterized MLPs (Tsukuda et al. 2006; Shee and Sharma 2008) have shown trypsin inhibitory property and therefore, MLPs can be appropriately categorized under Kunitz family.

Phylogenetic analysis of MLPs along with STI demonstrated that MLPs and native miraculin clustered separately from STI implying that they have diverged long back from common ancestor. MLPs and native miraculin are broadly divided into three groups in phylogenetic analysis (Fig. 1). All the new MLPs cluster together in two groups, Groups I and II, in which group II can be further subdivided into two minor groups, whereas native miraculin and related MLPs cluster in a distinct group. There are orthologous and paralogous copies of type 1 and type 2 MLPs present in different members of Rutaceae family. Orthologous sequences across species are highly conserved implying functional maintenance. Multiple sequence alignments showed 98 and 75–95% identities among type 1 and 2 MLPs, respectively, while 49–58% identity was observed between the two groups. Group III MLPs showed 42–95% identity among themselves. Comparison with groups I and II showed ~39–45 and ~30–43% identities, respectively.

The data demonstrate that MLPs constitute a multigene family and have evolved from common ancestor by gene duplication, and later, functional diversification occurred because of speciation and specific local environments. Apart from trypsin inhibitory activity, new functions have been attributed to these proteins. They include taste-modifying property demonstrated only in native miraculin and antifungal property demonstrated in type 2 MLPs (Hirai et al. 2010; Tsukuda et al. 2006; Shee and Sharma 2007). The taste-modifying property has not been reported in MLPs. As demonstrated by mutagenesis studies, two histidine residues (His30 and 60 miraculin numbering), located in exposed regions, are considered responsible for miraculin taste-modifying activity (Ito et al. 2007). In type 1, only one of the Histidines is present (corresponding to His60) in exposed region, and the other is absent. Also, there are differences in gene expression patterns among these proteins. MKMLP, a type 1 MLP, is expressed constitutively in seeds and has been shown to be a seed storage protein (Shee and Sharma 2008), while type 2 MLPs characterized from C. jambhiri Lush, and tomato and coffee protein (LeMir and CoMir, respectively), both clustering with native miraculin in phylogenetic tree, are induced because of biotic and abiotic stress conditions like fungal, nematode, insect infestations ,and chemical treatments (Tsukuda et al. 2006; Brenner et al. 1998; Mondego et al. 2010). These observations suggest that regulatory differentiation might also take place allowing differential, spatial, and temporal gene regulations. These novel properties may have resulted from neofunctionalization.

In order to understand the selective pressure operating on the members of MLPs due to adaptive protein evolution, the average rates of non-synonymous and synonymous substitutions were calculated within and between phylogenetic groups. Surprisingly, the average dN/dS values for all pairwise comparisons were found to be less than 1, and other tests showed that sequences are under purifying selection (Table 3). It has been suggested that, when average rates are assessed, the positive selection gets masked, as only few residues in protein sequences are positively selected. Therefore, the site-specific positive Darwinian selection and purifying selection were estimated using a web server Selecton version 2.4. This server enables detecting the selective forces at a single amino-acid site by calculating the ratio of non-synonymous to synonymous substitutions (K a/K s ratio). Analysis of selection operating on type 1, type 2, and group III MLPs revealed a total of 11, 15, and 9 residues having K a/K s ratios >1, respectively, under MEC model and undergoing positive selection. All the residues were located in loop regions in all the three group members, except Leu33 (in type 1 and 2 members). Previous crystallographic analyses of the Kunitz inhibitors, interleukins-1β and 1α, and the acidic and basic fibroblast growth factors have shown that they contain β-trefoil fold (Murzin et al. 1992). Although these different proteins have very similar structures, many of their sequences have no significant similarities overall, therefore differences in function. Recent studies suggest that these proteins have similar key structural residues that are distributed symmetrically in their structures (Feng et al. 2010). MLPs are also characterized by β-trefoil fold with pseudo-threefold symmetry and consist of a six stranded-barrel capped by a triangular hairpin triplet. The loops connecting the β-strands vary in length and structure. The loops give the fold its varied binding capability, and the binding sites lie in different parts of the fold. In our results, we found that the residues undergoing positive Darwinian selection were in loop regions, suggesting evolution of protein for novel interactions. We could recognize one such site at residue 40 in type 1 protein sequences undergoing positive selection where a change to amino acid proline at this site might render chymotrypsin inhibitory activity to the protein, as rest of the loop residues match exactly with double-headed, winged bean chymotrypsin inhibitors second reactive site loop (Dattagupta et al. 1999). In many cases, reactive loop residues have been observed to be likely targets for positive selection (Chakrabarty et al. 2006), but in some cases, reactive loop residues are not involved (Talyzina and Ingvarsson 2006). Amino acid divergences in the reactive loop are known to result in differences in binding affinity of inhibitors to various proteases. Although most residues undergoing positive selection in MLPs were located in loop regions, none of them was present in reactive loop. When type 1 and 2 MLPs are compared, the position of P1 residue asparagine is well conserved, but there are changes around the P1 residue in paralogous MLPs. We found that a few putative active site loop residues in MLPs are undergoing purifying selection perhaps to maintain reactive loop conformation. This suggests convergent evolution among paralogous MLPs and advocates that MLPs putative active site loop residues might have optimized for specific interactions, and purifying selection is acting to maintain it. Positive selection was also detected in predicted phosphorylation sites, and its implications are difficult to predict.

The comparison of amino acid sequences of MLPs among themselves, with miraculin and related proteins, and STI demonstrates notable features which points toward adaptive evolution and acquiring of new functions in this group of proteins. Many of these MLPs from both groups have been characterized at biochemical and structural levels. It has been shown that MLPs, like STI, possess trypsin inhibitory activity (Shee and Sharma 2008). However, drastic alterations in primary and secondary specificity sites have been observed when MLPs were compared with STI and native miraculin (Fig. 6). Compared to classical Kunitz inhibitors like STI and native miraculin, the conventional arginine/lysine at P1 position has been replaced by asparagine residue in type 1 and 2 MLPs, which is too short to interact with Asp189, a key residue in binding pocket of trypsin. This suggests that MLPs may not act as substrate-like inhibitors, as they lack arginine/lysine as active site residue, essential for trypsin specificity. The alterations in secondary specificity residues of putative reactive loop of MLPs as compared with STI will certainly result in differences in interactions with proteases important for activity. Also, subtle differences have been observed at secondary specificity sites among different MLPs, which may have resulted in response to specific needs. This clearly indicates alterations in specificities of Kunitz super-family inhibitors to counter digestive proteases of local pests and predators. Crystal structure analysis and biochemical characterization of one of the type 1 MLPs, MKMLP, helped in understanding some of the properties of this group of proteins (Gahloth et al. 2010; Shee and Sharma 2007). MKMLP in both native and heat-treated forms is remarkably stable against proteolysis, but its functional stability reduces with increase in temperature (Shee et al. 2007) as compared with soybean Kunitz trypsin inhibitors which are functionally stable even at high temperature. The remarkable structural stability of native and heat-treated proteins against proteolysis could be attributed to the presence of three disulfide bridges in the structure. The functional stability of STI has been attributed to the stabilization of reactive loop by a conserved Asn13 which forms a network of hydrogen bonds with reactive loop (Iwanaga et al. 2005). In type 1 and type 2 MLPs, the corresponding asparagine residue in STI is replaced by alanine and serine, respectively. Some of these type 2 MLP proteins have been shown to possess protease inhibitory as well as anti-fungal activities. The anti-fungal activity of type 2 MLPs has been attributed to the presence of thaumatin motif (Tsukuda et al. 2006). It has been suggested that the mechanism of anti-fungal activity may have nothing to do with trypsin inhibitory activity. Thaumatin and thaumatin-like proteins have pI > 8 and are proposed to bind to fungal cell wall components through an acidic cleft between their protein structures (Ghosh and Chakrabarti 2008) (online resource 4). Type 2 MLP sequences, however, have pI < 5.5, and the acidic residues distributed on surface are slightly clustered, but not present in cleft, which suggests a different mechanism of action. This clearly shows that the anti-fungal properties in MLP type 2 proteins have been acquired during the course of evolution to develop new specificities to counter local environmental stresses. A notable modification found only in type 2 MLPs is the presence of short phosphorylation motif created due to insertion of four residues after position 84, and all these four residues are undergoing strong positive selection. These features are absent in type 1 MLPs, native miraculin, and STI. Certainly, this will add to new unknown function in type 2 MLPs, which will be absent in other Kunitz super-family members.

From our data, we conclude that MLPs belongs to Kunitz super-family and represent a rapidly evolving gene family. Driven by gene duplication, neofunctionalization, and later by positive Darwinian selection, MLPs are being optimized for novel functions to counter local biotic and abiotic stress conditions.