Molecular Architecture and Evolution of a Modular Spider Silk Protein Gene

See allHide authors and affiliations

Science  25 Feb 2000:
Vol. 287, Issue 5457, pp. 1477-1479
DOI: 10.1126/science.287.5457.1477


Spider flagelliform silk is one of the most elastic natural materials known. Extensive sequencing of spider silk genes has shown that the exons and introns of the flagelliform gene underwent intragenic concerted evolution. The intron sequences are more homogenized within a species than are the exons. This pattern can be explained by extreme mutation and recombination pressures on the internally repetitive exons. The iterated sequences within exons encode protein structures that are critical to the function of silks. Therefore, attributes that make silks exceptional biomaterials may also hinder the fixation of optimally adapted protein sequences.

Araneoid spiders are capable of spinning up to seven unique silks. Some of these silks are renowned high-performance fibers. For example, dragline silk has exceptional tensile strength whereas flagelliform silk, the elastic filament that forms the capture spiral of an orb-web, may have >200% extensibility (1). In Nephila clavipes (Araneae: Tetragnathidae), the gene that encodes flagelliform silk (Flag) is transcribed into an mRNA of about 15.5 kb (2). Most of Flag is composed of numerous iterations of three different amino acid motifs: GPGG(X)n, GGX (G, Gly; P, Pro; X, other), and a 28-residue “spacer” (Fig. 1A). These motifs are organized into complex ensembles of about 440 amino acids that have similar tandem arrays of the glycine-rich motifs combined with a single spacer motif (Fig. 1B).

Figure 1

The Flag gene contains hierarchical sets of components. (A) The repetitive coding region is composed of codons for three different amino acid sequence motifs. (B) Iterations of the three motifs are organized into complex ensemble repeats of about 440 amino acids. (C) Each ensemble repeat is encoded by a single exon. These repeated exons are separated by repeated introns. (D) The Flag gene spans about 30 kb. Exons and introns are numbered, and regions of nonrepetitive sequence are shaded gray.

Here we report genomic sequences of Flaggenes that encompass both 5′ and 3′ ends and substantial portions of the intervening repetitive region. We cloned partial Flaggenes from genomic DNAs of N. clavipes (N.c.) andNephila madagascariensis (N.m.) (3). The resulting 36 kb of sequence (GenBank accession nos.AF218621AF218624) represents the most extensive DNAs known for any spider silk. In contrast to other spider silk genes, Flag is not encoded by a single enormous exon (4, 5). Instead, the Flag gene is evenly divided into exonic and intronic regions (6). In all, the Flaglocus is estimated to span 30 kb and to contain 13 exons (Fig. 1D).

The first two exons encode the nonrepetitive amino-terminal region. The final exon contains both repetitive sequence and the nonrepetitive carboxy terminus (Fig. 1D). Exons 3 to 13 each encodes an individual ensemble repeat (Fig. 1, B to D). These repeated exons are of similar length [about 1320 base pairs (bp)] and identical organization. When these exons are aligned, the spacer motifs emerge as the most conserved sequences within and between species (Fig. 2A). The main differences among the repeated exons are the variable numbers of tandem GPGG(X)n and GGX motifs. In comparisons of species, only exons 3 and 13 of N.c. are more similar to the corresponding exons of N.m. than to the other repeated exons of N.c. (Fig. 2B). Even given these exceptions, on average the repeated exons are more alike within (73%) than between (68%) species at the DNA level. Given that both species have identical motifs and ensemble repeat structures, differential selection cannot easily account for this divergence pattern. Instead, such homogenization of sequence repeats within species is indicative of concerted evolution (7).

Figure 2

Each ensemble repeat is encoded by an individual repeated exon. (A) The translated alignment of nucleotides (14) from N.c. and N.m. is depicted with the motif symbols from Fig. 1A and the numbering of exons in Fig. 1D. (B) Parsimony analysis (15) was done to show the similarity among exons by the possession of shared characters. The tree shown is one of the two shortest trees (2609 steps) that was chosen after one round of successive approximations (16). Bootstrap values (17) are shown above the internodes; Bremer support scores (18) are below. Branch lengths are proportional to the number of character changes. An additional 605 steps are required to make all the exons pair by corresponding number (e.g., N.c.5 with N.m.5).

The introns separating the repeated exons also share high similarity (Fig. 1, C and D). Introns 3 to 12 are each about 1420 bp long and, in general, are easily aligned to one another (Fig. 3A) (8). Within species, the introns are on average 87% similar. The degree of homogenization inN.c. is extreme, with introns 5, 6, and 7 sharing 99.9% identity. Between species, the introns are much less similar (75%). Thus, the repeated introns are less divergent within a species than between species (Fig. 3B).

Figure 3

Iterated introns are homogenized within species. (A) A portion of the alignment (14) is shown, and variable sites that are fixed within a species are highlighted. (B) Parsimony analysis (15) of the aligned intron sequences resulted in three trees (1158 steps), which differed only in the resolution of N.c.5, N.c.6, and N.c.7, that share 99.9% identity. Support indices and branch lengths are as in Fig. 2B. To force the pairing of introns by position in the gene (e.g., N.c.3 with N.m.3) required an additional 801 steps.

The corresponding exons and introns from N.c. andN.m. were compared to further examine between-species divergence of repetitive and nonrepetitive regions. The nonrepetitive exons 1 (591 bp) and 2 (405 bp) and the nonrepetitive 3′ portion of exon 13 (372 bp) are the most highly conserved coding sequences (Fig. 4). These regions differ by less than 5% between the two species. In contrast, the repeated exon sequences are five times more divergent between species. Thus, the repeated exons are not only homogenized within species but are also more divergent between species than the nonrepetitive coding sequences (Fig. 4).

Figure 4

Corresponding gene regions betweenN.c. and N.m. were aligned pairwise (19) and percent divergence is plotted. Colors of exons and introns are as in Fig. 1. Arrows point to the low divergence in exons 1 and 2 and to the carboxy terminal portion of exon 13.

Even with the high divergence among repeated exons, there is evidence of purifying selection. The most obvious indication is that the organization of motifs within the ensemble repeats is strictly maintained in both species. Each ensemble repeat has just one centrally located spacer that is flanked by GPGG(X)nand GGX repeats (Fig. 2A). More detailed comparison of the exons shows that the observed types of synonymous substitution are biased toward adenine and thymine, which is consistent with the substantial codon usage preferences present in spider silk genes (2, 4, 9). Furthermore, there are additional, apparent constraints on the observed types of nonsynonymous substitution. Most amino acid differences occur in the X positions of the GPGG(X)n and GGX motifs. Although these positions are the most variable among the repeats, they tend to be replaced from a very small subset of amino acids—namely, alanine, serine, tyrosine, and valine. The final apparent constraint is that insertions and deletions are often entire GPGG(X)n or GGX motifs. In fact, most of the divergence between the aligned exons is due to short insertions and deletions of these motifs and not point mutations (Fig. 2A).

Because Nephila rely on their orb-webs for prey capture, there are strong selective forces that operate on Flag to produce a functional capture spiral. Thus, the high divergence of the repeated exons might be due to purely molecular mechanisms. The internally repetitive nature of the exons is likely to promote errors from slippage during replication. Furthermore, because codons for glycine, proline, and alanine compose almost 75% of the coding sequence, there is an abundance of guanine and cytosine bases. Strings of these base pairs are thought to create recombination hot spots (5, 10). The resulting unequal crossover events could, in part, account for the length differences observed in the numbers of tandem GPGG(X)n and GGX motifs (Fig. 2A).

The large variation in length that has been observed in alleles of spider silk genes is further evidence for the prevalence of replication slippage and unequal crossing-over events (5). Many of the differences between alleles are insertions and deletions of ensemble repeats or shorter motifs that can be explained by replication slippage. Similarly, allelic length differences are common in the lepidopteran silkworm fibroin gene, which is also rich in cytosine and guanine and highly internally repetitive (10, 11).

The repeated introns are more similar (87%) (8) within a species than are the repeated exons (73%) (Figs. 2B and 3B). The intron sequences could have been homogenized by recombination events, selection, or some combination of the two processes. A possible function of the intron sequences may be stabilization of the Flag precursor mRNA. However, the extreme 99.9% similarity ofN.c. introns 5 to 7 is strong evidence for homogenization by recombination. To argue for selection alone would require the convergent evolution of hundreds of sequence changes among the introns of N.c. and N.m. (Fig. 3A). Thus, as with the exons, recombination and/or conversion events are likely to have limited the within-gene divergence of the repeated introns. However, in contrast to the exons, the introns are not internally repetitive and lack a sequence rich in cytosine and guanine. Without these factors that promote replication slippage and unequal crossovers, the introns exhibit less divergence than the repeated exons and tend to require fewer alignment gaps (Figs. 2B and 3).

The greater divergence of the Flag repetitive exons relative to the introns could be due to different directional selection regimes operating on the Flag fibers of N.c. versusN.m. If so, then a unique sequence element that is advantageous to one species is expected to be present in all of its repeated exons and absent from the repeated exons of another species. Given that there are very few fixed differences among the ensemble repeats within N.c. or N.m., directional selection does not seem to account for rapid sequence divergence relative to the repeated introns. Also, purifying selection on the silk protein structure is reflected in the strict maintenance of the ensemble organization (Fig. 2A). Instead of speeding up evolution, functional constraints should decrease sequence divergence of the exons. Thus, the greater exonic divergence still can be best explained by the molecular architecture of the Flag gene.

Flag was known to be a modular protein with three basic motifs composing a large ensemble repeat. The genomic organization of theFlag gene suggests a new hierarchical level of modularity. Not only are the ensemble repeats encoded by repeated exons, but the intervening introns are also iterated copies. This molecular architecture results in efficient within-gene concerted evolution. Probably through some combination of gene conversion and unequal crossing-over at repetitive exons, Flag remains fairly homogenized over its entire 15,500-bp coding sequence. However, this same highly repetitive architecture apparently prevents the coding sequences of Flag from completely homogenizing. The evolution of the Flag gene represents a case in which homogenization of repeats through purifying selection and recombination is offset by mutational mechanisms inherent in the basic structure of the DNA sequences (12). Thus, the repetitive genetic architecture of spider silk encourages sequence homogenization as well as rapid sequence divergence. This conflict has implications for the interpretation of high-performance silks as optimally adapted supermolecules (1, 13).

  • * To whom correspondence should be addressed. E-mail: hayashi{at}


View Abstract

Navigate This Article