To understand the evolution of the three-finger toxins in S. catenatus edwardsii venom, we determined their gene structures (Figure 2A). We obtained the full length gene of 3FTx 3 by gDNA PCR. However, we obtained only partial genes (from exon II to exon III) of the other four toxins by gDNA PCR even after several attempts, perhaps due to suboptimal annealing of the primers or the thermal cycling profile. We performed GenomeWalking to obtain the remaining gene segments from exon I to exon II (Figure 2A). At least 16 clones were sequenced from each PCR product to obtain the gene sequences of all five three-finger toxins (2.246 kb for 3FTx 1, 2.145 kb for 3FTx 2, 2.186 kb for 3FTx 4, 2.167 kb for 3FTx 5 and 2.986 kb for 3FTx 3). The cDNA sequences were used to determine the exon-intron boundaries (Figure 2B), which follow the GT-AG rule of splice-donor and -acceptor sites 25. Genes of all five three-finger toxins have similar architecture, with two introns and three exons, similar to those of elapid three-finger toxin genes. Sequence data was deposited in GenBank under accession numbers EU 293789, EU 293790, EU 293791, EU 293792 and EU 293793, respectively. The gene sequences were used to determine the phylogenetic relationship with three-finger toxin genes from the snake families Elapidae and Colubridae. The phylogenetic tree was constructed using DNAMAN, and viperid three-finger toxin genes form a separate cluster in the tree (Figure 3).

Exons also appeared to have segments (Figure 1). As expected because of overall size, the segments in exons are much smaller as compared to those in introns (4–7 bp in segment vii in exon II to 53 bp in segment i in exon III, compared to 21–22 bp in segment X to 680 bp in segment Vd in intron 1). Analysis of the gene sequences reveals that the gain/loss of segments occurs in both introns and exons (Figure 6). The segments in introns, when present, show high similarities (>85% identity). In contrast, while some exon segments show high sequence identity (60–100% identity; shown in same colors), other segments show low identity (12.5–50% identity; shown in different colors). Such similarity/dissimilarity of segments is more prominent in exon II and exon III. In exon II, there are four different kinds of segment ii; in 3FTx 4 and 5 they are identical, whereas in the other genes they are totally different (Figure 6). In the same way, segment v is similar in 3FTx 1, 4 and 5, whereas in 3FTx 2 and 3 it is different. Segment vi is absent in 3FTx 1 but present in the other genes; this segment is similar in 3FTx 4 and 5 but different in 3FTx 2 and 3. Thus exon II appears to have evolved through "switching" of segments, although the origins of the "new" segments are not known. Similarly, exon III also appears to have evolved through "switching" of segments. However, this phenomenon is not observed in introns of these genes, although there are additions/deletions of a few short segments in introns (see Additional file 1). The switching of segments in exons has far reaching effects on the biological effects of the toxins. The changed segments will affect the overall surface properties of the expressed protein (such as charge density and hydrophobicity) and hence the function. In contrast, the small additions/deletions in introns will not affect the nature of the protein product, but may affect its expression. It is important to note that in spite of this segment switching, positions of cysteine residues are conserved, thereby maintaining similar disulfide pairing, folding architecture and the overall three-finger fold of the mature protein.

There are several possibilities that could explain the observed switching of segments.

1. Splicing variations: The difference in isoforms of some proteins due to change of segments can be easily explained based on splicing variation 30. However, unlike these proteins, the segment switching in viperid 3FTx occurs within exon II and exon III. Among three-finger toxins, long-chain neurotoxins arose from short-chain neurotoxins through an error in the splicing site 31. Such an error leads to insertion of a short segment with the fifth disulfide in the second loop. In viperid toxin genes, however, the insertions of segments do not occur at the intron-exon boundaries, so the mechanism of insertions/deletions or switching of segments does not occur due to errors in splicing.

2. Recombination: Distinct genes encoding isoforms of proteins are also generated through recombination of two related/unrelated genes 3233. In general, the segments involved in such recombination events are fairly large (700 to 2500 bp), and the segments that are exchanged in exons of 3FTxs are probably much too small. Therefore, canonical recombination may not be involved in these exchange events.

3. Accumulation of point mutations: Accelerated point mutations in three-finger toxins are common and they lead to the evolution of several isoforms 101114. Although these point mutations occur in exon segments, they may not explain such a distinct change in the sequences of segments. This possibility requires the repeated occurrence and accumulation of multiple point mutations within each segment. Further, all the intermediates have to be selected through evolution. The absence of intermediate isoforms, however, contradicts this possibility; and

4. Independent recruitment events: Venom protein genes are thought to be recruited to the venom gland genome by gene duplication of a normal physiologically important gene and recruitment of the duplicated gene for expression in the venom gland 34. It is possible, but not probable, that each of these isoforms has an independent origin and their ancestral three-finger toxin genes were recruited at different times. High similarity across numerous 3FTx genes in exon I, and introns I and II, supports instead a single recruitment and lineage of these genes and not multiple recruitment events. In the unlikely events of independent recruitments, introns will have to undergo convergent evolution to explain the high similarity while the exons will be undergoing divergent evolution. Therefore, segment switching results in divergence of functional regions of exons (see below) while maintaining the basic fold, rather than convergence upon a single scaffold motif during independent recruitment.

Although the mechanism of the exchange of segments in exons is unknown, it is apparent that these events play important roles in the evolution of these toxins, in addition to the role that accelerated point mutations in the exons plays in toxin evolution 101114. These point mutations appear to alter the interaction surfaces of toxins 35. However, they affect one residue at a time and a smaller molecular surface, and they may help in fine-tuning of functional sites of the molecule for interaction with specific molecular targets. They may (a) enhance the affinity to a specific receptor or ion channel; (b) change the specificity to another closely related receptor or ion channel; and (c) change the species specificity of the toxin to a particular receptor. In contrast, accelerated exchange of larger segments drastically changes parts of the interacting loops or toxin surface. Therefore, these exchanges of segments may help in switching the molecular targets of toxins and hence affecting their pharmacological properties. We propose that

a

s

s

e

t

To understand the impact of switching of segments on the molecular surfaces of three-finger toxins in S. catenatus edwardsii, we modeled all four distinct three-finger toxins. As shown in Figure 7, the three β-sheeted loops in these toxins are distinctly different from one another, as most of them are replaced by segment switching (Figure 7, top row). Further, the electrostatic potentials of these toxins indicate that the charge distributions on their molecular surfaces are also different. 3FTx 1 and 4 have more acidic residues on the surface as compared to 3FTx 2 and 3 (Figure 7, middle and bottom rows). This drastic difference in the charge residues on the surface is due to the exchange of segments (see Additional file 1), but retention of the similar molecular fold. Such a change in the charged residues might play an important role in switching the molecular targets. Since most of the functional sites are located on these β-sheeted loops (for a review see, 36), it is logical to propose that all of these novel viperid toxins have distinct pharmacological properties. Therefore, ASSET phenomena affect the molecular surfaces of three-finger toxins significantly and alter their molecular targets, playing a crucial role in the evolution of the three-finger toxins.

Extraction of venom glands and liver from Sistrurus catenatus edwardsii (Desert Massasauga) were reported previously (Pahari et al. 2007). RNeasy mini kit, DNeasy DNA extraction kit, Qiaquick gel extraction and PCR purification kit, and a QiaPrep mini prep kit were purchased from Qiagen (Valencia, CA, USA). GenomeWalker™ kit was purchased from Clontech Laboratories Inc (Palo Alto, CA, USA), Long PCR Enzyme Mix was procured from Fermentas (Burlington, ON, Canada), TOPO^® XL PCR Cloning kit was obtained from Invitrogen (Carlsbad, CA, USA), and ABI PRISM^® BigDye^® terminator cycle sequencing ready reaction kit was purchased from Perkin Elmer (Foster City, CA, USA). Oligonucleotides were custom synthesized by 1^st BASE (Singapore). All other chemicals and reagents used were of the purest grade available.

Genomic DNA (gDNA) was extracted from the S. catenatus edwardsii liver tissue (30 mg, previously stored in RNAlater) using the DNeasy Tissue kit (Qiagen, USA) according to the manufacturer's instructions. RNaseH was used during the extraction to remove contaminating RNA. The integrity of the isolated gDNA was confirmed by 0.8% agarose gel electrophoresis and DNA was quantified spectrophotometrically.

gDNA PCR was performed to obtained the gene sequence of 3FTx 3 using a gene-specific forward primer 5'-ATGAAAACTCTGCTGTTGATCCTGGGGGT-3' and gene-specific reverse primer 5'-GCCAATAGTCACTTTTAGAACTATTTGTTGCAGTTGTCTG-3'. The PCR reaction contained 1.0 unit of Long PCR Enzyme Mix, 1 μg of gDNA. 0.2 mM dNTP, 0.2 μM primers and 1 × Long enzyme mix polymerase buffer in a total of 50 μL. The thermal cycling reaction involved 35 cycles of one step each at 95°C for 15 s, 60°C for 15 s, 68°C for 3 min followed by a final extension step at 68°C for 10 min. The amplified PCR products were extracted and cloned as mentioned below.

The GenomeWalking libraries were constructed using the Universal GenomeWalker™ kit (Clontech Laboratories Inc, USA) according to the manufacturer's instructions. Briefly, libraries were constructed with 3 μg of gDNA restriction digested with DraI, EcoRV, PvuII and StuI. The 'genome walk' involved two sets of primers: adaptor primer 1 (AP1-sense) 5'-GTAATACGACTCACTATAGGGC-3' and nested PCR adaptor primer 2 (AP2-sense) 5'-ACTATAGGGCACGCGTGGT-3', both provided in the kit, and 25-mer and 27-mer gene-specific primers designed from the signal peptide regions of cDNAs of all the three-finger toxins. Primary and nested PCRs were performed as recommended by the manufacturer (BD GenomeWalker™) using Advantage Polymerase 2 Mix obtained from Clontech Laboratories Inc (Palo Alto, CA, USA). The 50.0 μl reaction mixture consisted of 1 μl of DNA template (0.1 μg) (either from each library or from primary PCR products), 1× PCR buffer (provided in the kit), 0.2 mM dNTPs, 0.2 μM appropriate adaptor primers, 0.2 μM of appropriate gene-specific primers, 1× Advantage™ 2 polymerase mix. The thermal cycling profile used was as follows: 7 cycles of 94°C for 2 s, 72°C for 3 min; 32 cycles of 94°C for 2 s, and 67°C for 4 min followed by a final extension at 67°C for 7 min. The PCR products were purified, cloned and sequenced.

PCR products were subjected to 1% agarose gel electrophoresis, visualized by ethidium bromide staining and purified using a gel extraction kit or PCR purification kit. Purified PCR products were ligated either to pDrive vector (Qiagen, Hilden, Germany) or pCR^®-XL-TOPO^® vector (Invitrogen, USA). Ligated vectors were transformed to DH5a competent cells by heat shock. Kanamycin (100 mg mL^-1) was used for antibiotic resistance selection. Blue/white colony screening was done on LB agar plates using 80 mg mL^-1 X-gal and 0.5 mL L^-1 of 100 mM isopropyl-β-D-thiogalacto-pyranoside (IPTG) to select the positive colonies.

DNA sequencing reactions were carried out using the ABI PRISM^® BigDye^® terminator cycle sequencing ready reaction kit (BDV3.1) according to manufacturer's instructions (Applied Biosystem, Foster City, CA, USA). DNA sequencing was carried out using an ABI PRISM^® 3100 automated DNA sequencer.

Sequence analysis was carried out using the BLASTX program at National Center for Biotechnology Information website. Multiple sequence alignment was done using DNAMAN and online DIALIGN Multiple Sequence Alignments tool at BiBiServ 37. A neighbor-joining tree was constructed using DNAMAN version 4.1.5.1 (Lynnon BioSoft).

Three dimensional structures of three-finger toxins from Desert Massasauga (Sistrurus catenatus edwardsii) venom gland transcriptome were modeled using the online I-TASSER server for protein 3D structure prediction 38. The server predicts the folds and secondary structure by profile-profile alignment (PPA) threading techniques. For each protein, 3–4 models were obtained. The model with the lowest free energy was used for further analysis. Ribbon structure diagrams and surface charge models were created using the DS ViewerPro software to compare potential differences in electrostatic charges of these viperid 3FTx.