Spiders are renowned for ensnaring their prey with a complex and finely-tuned structure-the orb-web. Approximately 10 % of all spider species hunt using orb-webs, which are constructed entirely of different types of silk, both solid and liquid in form. Orb-webs need to absorb the energy of rapidly flying insect prey without collapsing or catapulting their potential prey out of the web. These functional requirements necessitate that webs possess strength, extensibility and adhesion which is accomplished through a combination of design innovation [1][2][3][4] and silk specialization [5][6][7][8]. Orb-web spiders produce a variety of silk types, each manufactured by different glands, with highly divergent functional capabilities [9][10][11][12][13]. Perhaps no web component exhibits more evolutionary and functional diversity across orb-web weaving spiders than the silks responsible for prey adhesion. In cribellate orb-weavers (Uloboridae and Deinopidae), prey capture is achieved with a dry, adhesive silk that likely represents the ancestral condition for capture thread production in the Entelegynae [14][15][16]. In these spiders, billowy mats of cribellate silk are laid on support threads of pseudoflagelliform silk and bind to insect cuticle through a combination of hygroscopic and van der Waals' forces, and physical interactions with prey cuticle [17][18][19]. Ecribellate orb-weaving spiders, on the other hand, utilize a wet adhesive that coats the highly elastic capture threads spanning the radial threads of an orb-web. The transition from a dry to a wet adhesion system improved prey capturing performance with substantially less energetic demands [20][21][22] and is hypothesized to represent a key evolutionary innovation that drove the extensive diversification of the araneoid superfamily of spiders [23].
The wet adhesive capture thread of orb-web spiders (Araneidae) is composed of silk from two gland types: aggregate (two pairs) and flagelliform (one pair). Each silk gland has a duct that leads to its own spigot on the spinnerets. The pair of spigots connected to the aggregate glands are tightly arranged on either side of the flagelliform gland spigot, forming a triad, and coat the flagelliform silk fiber in an aqueous glue as the fiber is extruded during orb-web construction [24][25][26]. The liquid aggregate silk quickly coalesces, though Plateau-Rayleigh instability, to form minute droplets evenly spaced along the flagelliform fiber [27,28]. The aggregate glue droplets are comprised primarily of glycoproteins and hygroscopic low-molecular mass compounds (LMMCs) that appear to form multiple layers [24,27,29]. A dense granule of proteins anchored to the flagelliform fiber lies at the inner-most layer [30] and is encapsulated first by a less dense glycoprotein core and then an aqueous layer containing additional proteins, salts and LMMCs. Aqueous glues are also used in the prey capture threads of cobweb spiders (Theridiidae), a group that is likely derived from an orb-web building ancestor. In cobwebs, however, the aggregate glue is attached to major ampullate silk fibers ('gumfoot' threads) and exhibits different mechanical properties and droplet composition relative to the capture threads of orb-weavers [31][32][33][34].
The functional requirements for aggregate glue performance are complex and involve a diverse set of interactions and behaviors including binding to insect cuticle and maintaining cohesion within the droplet and connection with the flagelliform fiber as prey struggle. Orbweb glue has been characterized as a viscoelastic solid that exhibits liquid-like properties during initial prey contact and becomes more fiber-like as the prey is retained [35]. Optimal performance of viscid capture threads involves a trade-off between surface contact and cohesion [36,37]. At high viscosity, the aqueous glue will not spread sufficiently to form a large enough contact area with the prey's cuticle, while at low viscosity the cohesion strength of the glue is too weak to retain the struggling insect. Glue droplet viscosity is controlled by the absorption of water from the environment and is fine-tuned, via the composition of the hygroscopic LMMCs, to optimize prey retention. Glue performance appears to be optimized to the relative humidity of a species' habitat [37][38][39][40]. Cobweb glue, on the other hand, exhibits minimal responsiveness to changes in humidity and has been described as a viscoelastic liquid with much lower extension-to-break values than orb-web glue [31].
While glue hydration is controlled largely by LMMC composition, adhesion and cohesion are controlled primarily by the aqueous glue proteins [31,35]. The glue droplet contains an assortment of proteins [32,34] but is dominated by two proteins encoded by members of the spidroin gene family-aggregate spidroin 1 (AgSp1) and aggregate spidroin 2 (AgSp2). Spidroin genes encode large repetitive proteins and have undergone dramatic evolutionary diversification within spiders, producing an array of functionally distinct genes [5][6][7]10]. Each specialized silk gland primarily expresses one type of spidroin gene that has specific performance properties [9,12,41]. For instance, major ampullate spidroin (MaSp) genes expressed in the major ampullate glands underlie the high strength and toughness of draglines and orbweb radial threads [42][43][44][45], while flagelliform spidroin (Flag) genes expressed in the flagelliform glands produce the highly extensible fibers used in capture threads [46,47].
Several studies have examined the differences in mechanical performance of aqueous glue among spider species [31,37,39,[48][49][50][51], but we still know little about the molecular variation associated with aggregate glue glycoprotein function. Coding sequence for aggregate glycoprotein was first described for Trichonephila clavipes [52] and Collin et al. [53] subsequently identified it as AgSp1, while also presenting the first sequence data for AgSp2. Stellwegen and co-authors [54,55] provided the first detailed analysis of AgSp gene structure. They found that AgSp genes encode the largest spidroins sequenced to date, possessing several different regions of repetitive sequence. The core repeat unit [56] is enriched for glycine and proline, like MaSp genes [45,57,58], but has a more complex amino acid composition than the typical MaSp repeats. Proteomic analysis has revealed extensive glycosylation and phosphorylation across both AgSp1 and AgSp2, which likely improves adhesion [32,34].
Here, we present the first detailed comparative analysis of AgSp gene structure among orb-web weaving spiders. We classify a previously described spidroin of unknown function as an aggregate spidroin, AgSp3, and identify several species-specific gene duplications. We find that AgSp1 in orb-weaving spiders has two primary repeat types that are both present in different regions of most gene copies but the relative proportions of these repeat types can vary substantially among paralogs. AgSp2 is similar to AgSp1 in repeat structure but is distinguished by the sporadic occurrence of unusually large, simple repeats with an abundance of motifs common to silk fiber proteins. By contrast, AgSp3 has diverged greatly at the molecular level from the other AgSp paralogs and is distinguished by high levels of charged residues, poly-threonine blocks and extensive interspecific variation. Similar to other spidroins, the repetitive regions of most AgSp genes are highly homogenized but with a more complex structure relative to other spidroins.
Individual spiders were collected at several localities in the USA (Table S1). A phylogeny of the species examined in this study based on previous analyses [59][60][61][62] is presented in Fig. S1. Mature spiders were housed at the American Museum of Natural History (AMNH) in ambient conditions until dissection. Spiders used for molecular work were euthanized with carbon dioxide and then dissected to extract individual silk glands. Dissected tissue was flash frozen in liquid nitrogen and stored at -80 • C until processed for DNA or RNA extraction.
Genomic sequencing for the majority of species in this study was conducted using Oxford Nanopore (ON) PromethION technology. Latrodectus hesperus was sequenced with the PacBio HiFi platform. All genomic preps for ON sequencing were conducted on dissected silk gland tissue from a single individual using the Gentra Puregene tissue kit (Qiagen cat 158667). These samples were then treated with the Circulomics short read eliminator kit (PacBio cat SS-100-101-01). Approximately 5 μg of DNA for each species was sent to the DNA Technologies Core at UC Davis for long-read sequencing. Samples were sequenced on one or two PromethION flowcells, producing between 62 GB and 124 GB of sequence (Table S1). Genomic DNA used for ON sequencing was also submitted to Novogene (Sacramento CA, USA) for library preparation and short read sequencing. Approximately 60 GB of 150 bp paired-end reads were generated from each sample for use in error correction. Dissected silk gland tissue for L. hesperus was sent for processing at the BYU DNA sequencing center which conducted library prep using the SMRTbell Express Template Prep Kit 2.0 and sequenced across two 30 h 8M SMRT cells on the PacBio Sequell II. HiFi reads were generated from subreads using PacBio SMRTlink.
For genome assembly, ON reads were first corrected using the long read self-correction step in Canu v2.1.1 [63] with default parameters except correctedErrorRate = 0.105. Corrected reads were assembled using Flye v2.8 [64] with default parameters. Short reads for each species were then used to error-correct assembled genomes with the polishing tool POLCA that is part of MaSuRCA v4.0.9 [65,66]. The L. hesperus HiFi reads were assembled with Hifiasm v0.13-r307 [67,68]. All genomes were masked for repetitive regions using RepeatMasker v4.1.5 [69]. A set of protein coding genes was generated for each genome with the Braker3 pipeline [70][71][72][73][74][75][76][77] using information from both short read RNA-seq evidence (see below) and a set of homologous proteins derived from eight recently sequenced spider genomes (Argiope bruennichi, Caerostris extrusa, Nephila pilipes, Oedothorax gibbosus, Parasteatoda tepidariorum, Stegodyphus dumicola, Trichonephila clavata and Uloborus diversus) and the arthropod OrthoDB v11 protein set [78].
To identify all potential AgSp gene copies, uncorrected genomic contigs for each assembly were blasted against a custom database of published spidroin amino (N)-and carboxy (C)-terminal regions (excluding repetitive sequence). Contigs that had a spidroin hit were extracted from the genome and subjected to three rounds of short read error correction using minimap2 v2.17 [79] to align the short reads and Pilon v1.23 [80] to correct the sequences. To prevent artificial homogenization of the repeat nucleotides due to reads from dominant repeat sections aligning to and correcting variant repeat sections, Pilon only corrected frameshifts (with '-fix indels' command) not base pairs. Intron/exon boundaries were identified by aligning RNA-seq reads generated from aggregate tissue (see below) to the spidroin contigs with minimap2 and viewing shifts in read mapping distribution in IGV [81]. Canonical repeat units for AgSp1 and AgSp2 were identified by searching for the starting motif ('PGTTPG') identified in a previous study [56]. To measure the degree of homogenization among the repeat units within a given section of an AgSp gene we combined all the units within that section into a single assembly using MAFFT [82] and calculated the average pairwise distance among the units using Geneious (v10.2.4 https://www.geneious.com). In most cases, the repeat sequences consisted of translated amino acids and indels were included in the calculation.
For most species, RNA samples (generally three replicates per species) for aggregate gland tissue and non-aggregate gland tissue (a combination of silk glands other than the aggregate) were prepped using the PureLink RNA Kit (Invitrogen) with bead mill homogenization and Trizol isolation. For three species-Argiope aurantia, Argiope trifasciata and L. hesperus-a set of individual glands (aciniform, flagelliform, major ampullate, minor ampullate and tubuliform) were prepped separately rather than as a non-aggregate composite. All RNA samples were submitted to Novogene for library preparation and Illumina sequencing, with approximately 20 million 150 bp paired-end reads generated for each sample. RNA-seq libraries were mapped to the genomes of each species using STAR [83,84] and the resulting bam files were used in the Braker annotation. Gene expression values in the form of Transcripts per Million (TPM) were generated by mapping RNAseq libraries to each species' Braker gene set using RSEM [85]. In many cases, Braker failed to correctly predict all full length spidroins in a species. Therefore, for the RSEM analysis, all the predicted Braker genes that had a blast hit to a spidroin were replaced with a hand-curated set of N-and C-termini for all spidroins.
Several phylogenetic trees were produced for termini and repeat region sequences. Termini sequences consisted of the conserved spidroin N-and C-terminal regions that are hallmarks of the spidroin family. Both nucleotide alignments (translated alignments for protein coding regions) and protein alignments were generated using MAFFT [82]. Maximum likelihood trees were calculated with PhyML [86] using the models GTR + I + G (nucleotides) and LG + I + G (amino acids), generally with 1000 bootstrap replicates.
Assembly quality varied substantially across species (Table S1) but full-length aggregate spidroins were reconstructed in most cases. Between two to six genes with significant homology to AgSp1 or AgSp2 in our spidroin database were identified for each species, with the cobweb species, L. hesperus containing the most paralogs (Table S2). To understand the pattern of evolutionary change among these paralogs we conducted a phylogenetic analysis based on the translated N-and Cterminal sequence for each spidroin and included a set of non-AgSp spidroin sequences from a few species with well characterized genomes. Sp-8175, a putative spidroin whose gland affinity had not clearly been identified but exhibits high expression in the aggregate gland in the golden orb-weaver, Trichonephila clavipes [87], was also extracted from each study species and included in the phylogenetic analysis. Contrary to previous analyses [88], the prey capture spidroins from cribellate (cribellate and pseudoflagelliform) and ecribellate (aggregate and flagelliform) spiders group together. The basal split within this large clade is between cribellate spidroins and aggregate spidroins plus flagelliform spidroins (including pseudoflagelliform and Sp-5803) (Fig. 1). Sp-8175 is clearly embedded within the aggregate spidroin clade and likely originated from a duplication event involving AgSp2 (Fig. 1). The two genes are adjacent to each other in most genomes, including the theridiid cobweb spider L. hesperus. Overall, the AgSp diversification into three distinct genes occurred prior to the divergence of cobweb and orbweb spiders as both L. hesperus and the tetragnathid species Tetragnatha kauaiensis have representative genes in each of the three primary clades. Therefore, given its evolutionary affinity to the other AgSp genes and its broad taxonomic distribution, Sp-8175 is now designated as AgSp3.
Compared to the major ampullate spidroins [58,87,89,90], gene family composition for each AgSp gene is relatively conserved but there are several instances of species-specific duplications. For AgSp1, L. hesperus and the araneids Verrucosa arenata, Micrathena sagittata, Araneus marmoreus, and Larinioides cornutus each have two or more paralogs while AgSp2 has duplicated in the araneid Metepeira labyrinthea (Fig. 1). Three of the genes in L. hesperus (AgSp1.1, AgSp1.2 and AgSp1.3) and the duplicate pairs of V. arenata, L. cornutus and M. labyrinthea are tandemly arranged next to each other in the genome. The contigs containing the M. sagittata and A. marmoreus genes are relatively small so it is not possible to determine the relative proximity of different AgSp copies. Previous studies [54,55] have shown that AgSp genes are among the longest spidroins and this is confirmed for our sampling. Each of the three AgSp gene types has an average coding sequence length greater than 13 kb across species (Table S2). AgSp1 coding sequences (41,099 bp on average) are approximately 3 times longer than AgSp2 (14,403 bp) and 1.5 times longer than AgSp3 (27,289 bp). Particularly noteworthy is the immense size of AgSp1.1 in L. hesperus, which has a coding region over 133 kb making it the single longest spidroin, and a candidate for one of the largest genes of any type, ever found.
The gene organization for AgSp1 and AgSp2 in orb-web weavers has been detailed in previous studies [54][55][56] and is relatively conserved across the species in this study. In most species, both genes are characterized by the presence of two exons separated by an intron near the N-terminal region with the majority of repetitive sequence located in the larger, second exon (Fig. 2, Fig. S2). The AgSp1 genes in the cobweb spider L. hesperus (Fig. 2), as well as those in T. clavipes and Caerostris darwini, which occupy a basal position among the araneids (Fig. S1), do not have an intron suggesting that the intron arose at the common ancestor of the remaining araneids. In addition, for some species with an AgSp gene duplication, one copy (in V. arenata and L. cornutus) or both copies (M. sagittata) lack introns. A previously unrecognized feature of AgSp2 gene structure in most orb-web species is the presence of an open reading frame region roughly in the middle of the intron that contains a full N-terminal domain and some additional coding sequence (Fig. 2, S2). This secondary N-terminal domain (termed N2) may have resulted from an exon duplication after the diversification of the AgSp genes, as the primary and secondary N-terminal domains in AgSp2 are more closely related to each other than either is to the AgSp1 or AgSp3 Nterminal domains (Fig. S3). Expression values for the alternative N-terminus region in AgSp2 are low (Table S3) suggesting this region is either not part of a functional transcript or is expressed at a different life stage than adult females.
AgSp3 gene organization is similar to the other AgSp genes in orbweavers, having a smaller N-terminal exon and a larger C-terminal exon containing most of the repetitive sequence. The N-terminal region of AgSp3, however, includes an insertion of several hundred nucleotides that is not homologous to spidroin N-terminal sequence such that the standard N-terminal sequence is split into two sections (Table 1, Fig. S4). This inserted sequence is enriched for the hydrophilic residues asparagine and glutamine which comprise 33.2 % of the sequence across all species compared to only 10.5 % in the standard N-terminal regions of all three AgSp genes. Conversely, the hydrophobic residues, alanine, isoleucine, leucine and valine, are substantially reduced in the inserted sequence, representing only 4.7 % of the sequence versus 34.3 % of the standard N-terminus regions. Therefore, the insertion in the AgSp3 Nterminal region represents the most hydrophilic block of sequence located within any of the AgSp genes (Table 1). AgSp3 in orb-weavers also contains, in the middle of the gene, a stretch of small, virtually Fig. 2. Gene organization of AgSp paralogs in an orb-web (A. aurantia) and cobweb spider (L. hesperus). Rectangle boxes represent exons and black lines are introns. The red boxes are terminal regions, the dark gray boxes are core repeat regions and the light gray regions are non-canonical repeat regions. The white box in A. aurantia AgSp2 is an alternative N-terminal region. The middle CDS region of AgSp3 in A. aurantia contains several small, nearly identical, exons separated by small introns. Exon/intron and repeat structure in the AgSp genes for the other orb-weaving spiders sampled in this study are similar to those presented for A. aurantia but with variation in exon and intron lengths.
Abundance of amino acid features and motifs for the different repeat regions of the three AgSp genes in the orb-weavers in this study and the cobweb spider, L. hesperus. Len column indicates the total number of amino acids comprising each repeat section across all the paralog genes in that group. All other columns (except Hydrophob.) provide the relative percentage of each motif. Hydrophob indicates the average hydrophobicity score for the region based on the Kyte and Doolittle measure (lower values are more hydrophilic). Y[GP] motif measures the percentage of the total sequence that is comprised of blocks of at least four amino acids that contain at least one tyrosine along with glycine and proline residues. Poly-amino acid columns provide the percentage of total sequence comprised of motifs of two or more of that amino acid. NTR -N-terminus transition region. N-insert -inserted sequence within the N-terminus.
Len (aa) GP-block Y[GP] Serine Poly-I Poly-Q Poly-T Charged Hydrophob. AgSp1-Orb EY region 10,985 2.9 0.018 12.4 0.492 0.182 1.6 20.3 -92.6 NTR 16,956 20.0 0.029 5.1 3.320 0.189 4.6 16.1 -25.1 Core GP-rich 115,512 29.5 0.866 2.3 0.149 0.014 9.5 13.2 -55.8 Core S-rich 52,849 20.2 0.013 5.8 0.299 0.011 6.3 15.5 -39.9 AgSp1-Lhesp EY region 1,431 4.5 0.000 3.0 0.140 0.000 2.1 23.1 -77.7 Core 67,766 25.0 0.015 9.5 4.129 0.000 3.9 12.2 -20.3 AgSp2-Orb NTR 5,516 67.6 4.450 3.6 0.000 1.180 0.2 9.5 -140.1 Core 39,262 28.3 0.005 4.1 1.987 2.733 4.8 10.1 -58.9 Long 14,030 31.4 0.014 4.6 0.527 7.391 4.5 5.7 -85.3 AgSp2-Lhesp Core 397 8.3 0.504 6.0 1.763 2.519 0.0 21.1 -43.5 AgSp3-Orb N-insert 2,424 8.1 0.000 12.5 0.000 7.600 0.0 19.3 -203.8 Core RM1 16,730 11.7 0.012 6.5 0.741 0.000 19.6 20.4 -88.9 Core RM2 66,040 14.5 0.003 6.6 0.132 0.000 10.9 9.8 -55.3 AgSp3-Lhesp Core 3,092 0.0 0.000 1.5 0.000 0.259 29.1 10.5 -18.1
identical, exons separated by small introns (Fig. 2, Fig. S2). The number of exons, when present, ranges from 8 (M. labyrinthea) to 27 (L. cornutus) with a median of 10. Mastophora hutchinsoni does not possess an AgSp3 gene.
All three AgSp genes are highly expressed within the aggregate glands, with AgSp1 having the highest expression level in most species, followed by AgSp2 (Fig. 3, Table S3). The AgSp genes comprise, on average, 5.7 % of the total expression within the aggregate gland but there is substantial variation across species with M. labyrinthea AgSp genes exhibiting the highest proportional expression (14.3 %) and A. marmoreus the lowest (1.1 %). AgSp3 expression is lower than the other two AgSp genes in all species, but it is still relatively high compared to other genes (TPM greater than 1000 in seven of the species measured for AgSp3 expression). In addition, expression of AgSp3 is substantially higher in the aggregate gland than the non-aggregate glands in all species (Table S3), further supporting the classification of this gene as an AgSp gene.
Typical of spidroins, the AgSp coding regions are overwhelmingly made of repeating units. The AgSp1 repeat sequence is divided into a number of distinct regions [55] and its general organization is conserved across the orb-weaving spiders sampled in this study (Fig. 4, Fig. S5). Following the N-terminus is a block (termed the EY-region) of simple, non-homogenized repeats comprised primarily (>75 %) of five amino acids (glutamic acid, glycine, serine, threonine and tyrosine) and ranges in size from 401 to 1127 aa. Next downstream is the N-terminal transitional repeat region (NTR) that has non-canonical repeats that are similar to, but distinct from, the core repeat regions that make up the majority of the gene (Fig. 4, Fig. S5). The NTR is highly conserved among the orb-weavers with only a single indel and an average percent identity of 85.21 % in the alignment of the protein sequence from all AgSp1 paralogs. The NTR is also characterized by a high proportion of poly-isoleucine blocks (i.e. any consecutive stretch of two or more isoleucine residues), with over 15 times the proportion of this motif than occurs in the core repeat region (Table 1). There is another, smaller and less conserved region of transitional sequence preceding the C-terminus (Fig. S5).
Previous analysis of the AgSp1 core repeat structure [54][55][56] identified different regions (termed repeat modules, RM) in which the primary repeat unit (termed PGTTPG-unit, [56]) is homogenized within the module but variable between modules. In some species, such as M. hutchinsoni [54,56], T. clavipes and M. labyrinthea (Fig. 4), there is hierarchical structure to the repeat organization in which two slightly variable PGTTPG-units are part of a larger unit that repeats (the larger units are called ensemble repeats, ER). To examine the pattern of evolutionary change among PGTTPG and RM units, we constructed a single amino acid sequence for each RM in each of the orb-weaver AgSp1 paralogs by taking the consensus sequence based on an alignment of all the PGTTPG-units within that RM. A phylogenetic analysis of all of these repeat unit consensus sequences reveals two distinct types of PGTTPGunits (Fig. 4a). The most prevalent type, termed the GP-rich repeat, has an abundance of glycine and proline and a variable region that ranges in size (0-32 bp) across paralogs (this region was called 'the tail' in [55]). The variable region is also characterized by the occurrence of tyrosine, which comprises 8 % of this region but only 0.3 % of the other core repeat sequence. This residue often occurs in combination with glycine and proline (Table 1), a motif that is common in both major ampullate and flagelliform spidroins [58,91,92]. The alternative type of repeat, termed the S-rich repeat, has over twice the proportion of serine Fig. 3. Relative gene expression of the AgSp genes. For each species, the total expression of all paralogs for each of the three AgSp gene types is depicted on a log2 scale but the relative proportion of each type within the total is calculated for non-log values. as the GP-rich repeat and exhibits minimal length variation (98 bp in all but one paralog). As noted in previous analyses [54,55], both repeat types have a high proportion of poly-threonine motifs but with the GPrich unit containing about 50 % more than the S-rich unit (Table 1).
Another area of divergence among the repeat types is the prevalence of variation involving charged amino acids. These residues comprise only 14 % of the GP-rich and S-rich consensus sequences, but half of the variable sites between the two consensus sequences encode a change in the charge of the residue (χ = 13.58, p < 0.001, df = 1; Fig. 4b). In addition, the repeat sequences in V. arenata, which exhibits the most divergent mechanical glue properties among orb-weavers with a greater than 10-fold increase in glue stiffness [39,93], has a substantially higher proportion of charged amino acids (Fig. S6), particularly the positively charged arginine residues which have arisen in thirteen separate locations among the three RM regions in the two AgSp1 copies (Fig. S7). Most significantly, the distribution of these repeat types varies considerably among the different AgSp1 gene copies, with some genes comprised entirely of a single repeat type and others a mixture (Fig. 4c). Even when one repeat dominates, the core repeat region is sometimes divided into different modules. For instance, in several species (Argiope appensa, A. aurantia, A. trifasciata, Araneus marmoreus AgSp1.1 and C. darwini) over 90 % of the core repeat sequence is GP-rich, but there are multiple RM regions of this type within each gene (Fig. 4c). In some AgSp1 genes, introns provide a divider between different RM blocks but, in other cases, one RM can transition directly into another. It is noteworthy that for three of the species with species-specific duplications of AgSp1 genes (V. arenata, M. sagittata and L. cornutus), the two copies differ substantially in the repeat type abundance highlighting that a given repeat can rapidly sweep through most of the gene following gene duplication. In addition, neither T. clavipes nor C. darwini, have any Srich repeat sequence suggesting the GP-rich repeat was ancestral for orbweaving spiders. In this scenario, the S-rich repeat arose later within the family, probably as a small unit at the base of the gene, but when favored, such as for some gene duplicates and in M. hutchinsoni, the Srich repeat was able to spread across most of the gene sequence.
AgSp1 gene structure in the cobweb spider, L. hesperus, is similar to that in the orb-web weavers but with several notable differences. All four of the L. hesperus gene copies except AgSp1.1 contain an EY-region, but this sequence has a reduced proportion of serine (3.0 %) compared to the orb-weaver species (12.4 %). None of the AgSp copies in L. hesperus contain an NTR region directly homologous to the conserved region in orb-weavers. The core repeat region of the L. hesperus AgSp1 genes are comprised of PGTTPG-units but they are divergent from the GP-rich and S-rich repeats in orb-web species and exhibit more complex hierarchical structure and more divergence among paralogs than in orbweavers (Fig. S8). The most noteworthy amino acid feature of the repeat sequence is the elevated abundance of isoleucine (cob:11.8 %, orb:5.0 %). Similar to the NTR region in orb-web spiders, these residues are generally clustered in poly-isoleucine or isoleucine-rich motifs (Table 1, Fig. S8) suggesting the orb-web NTR and the cobweb core regions may share some homology. The prevalence of poly-isoleucine motifs in these sequences also causes an increase in their hydrophobicity. The AgSp1 core sequence in L. hesperus is 2-3 times more hydrophobic than either the GP-rich or S-rich core repeats in the orb-weavers AgSp1 gene (Table 1). Both AgSp1.1 and AgSp1.2 in L. hesperus have ensemble repeat units that vary in size throughout the gene. AgSp1.1 has a 2-unit and 3-unit ER, with the 3-unit repeat occupying approximately the first 85 % of the core repeat region and the 2-unit region located in a small section near the C-terminal domain. AgSp1.2 primary ER comprises 5 PGTTPG-units, but there are also ERs made up of 3 and 7 units and the overall placement of each ER type within the gene is not regularly spaced (Fig. S9).
The repeat unit sequence and organization in the AgSp2 genes of orbweb spiders is similar to the AgSp1 genes but with less distinct repeat modules and more idiosyncratic repeat units. The transitional region (NTR) from the N-terminus to the core repeat region of AgSp2 is shorter (average size = 394 aa) and less conserved than the transitional regions in AgSp1 and contains a high proportion of GP and YGP motifs (Table 1, Fig. S10). A motif that has been identified as a core unit in the fiberforming dragline and flagelliform silk genes, GPGGX [41,94], is common in this NTR region with the specific motif GPGGQ comprising 12.4 % of the sequence. The core repeat sequence, however, in the AgSp2 genes is comprised of PGTTPG-units homologous to those in AgSp1 but with some noteworthy deviations. First, most AgSp2 genes contain a few PGTTPG-units that are much longer than the standard units (452 aa vs 114 aa on average). These long units occur irregularly, approximately 2-3 times within each gene (A. appensa AgSp2 has the most units, 8, and M. hutchinsoni contains no long units). These long units are homologous to the standard units at their ends but have evolved an expanded middle region rich in glycine, proline and poly-glutamine motifs (Table 1, Fig. 5, Fig. S10). Similar to the NTR region, GPGGX is common in the long units and, along with another common motif, PQQPG, comprise 22.5 % of the long unit sequence across all species. QQ motifs are a distinguishing feature of MaSp2 genes and are often found in conjunction with glycine and proline residues [45,58]. Given the differences in the occurrence and placement of the long repeat units in the AgSp2 genes between species, there is a 12-fold difference in the overall proportion of QQ motifs among species with A. appensa (6.48 %) having the highest proportion and M. hutchinsoni (0.54 %) the lowest.
A couple of species possess repeat structure that deviates from the standard pattern. M. sagittata has a short stretch of standard PGTTPG units but the majority of the protein consists of simplified PGTTPG-units that are smaller than typical repeats and contains primarily glycine, proline and poly-glutamine motifs (Fig. 5). M. hutchinsoni repetitive sequence is structured as an ensemble repeat that contains four PGTTPG units that have a high proportion of charge residues (Fig. 5, Fig. S6). As has been noted previously [55], L. hesperus AgSp2 is small (636 aa), with no repeat structure and is enriched for negatively charged residues (15.6 % of sequence).
AgSp3 in orb-weavers have two primary repeat regions: RM1 which contains several small exons and RM2 which is encoded by a single large exon (Fig. 2, Fig. S2, Fig. S11). AgSp3 RM1 and RM2 are highly divergent both across species and compared to AgSp1 and AgSp2. Neither region contains the stereotypical AgSp PGTTPG-units, but the overall amino acid composition in these regions is similar to that in AgSp1 and AgSp2, with glycine, proline and tyrosine comprising over half of the sequence (RM1:52.8 %; RM2:62.9 %). The RM1 repeat units within a given gene are nearly identical in sequence and size, with an average size across Fig. 4. AgSp1 repeat unit differentiation. A) Phylogenetic relationships among the consensus sequences of the core repeat modules within the orb-web AgSp1 genes. The tree indicates there exist two primary repeat types (bootstrap value presented for node differentiating the groups). B) Alignment of the consensus sequences of each primary repeat type derived from all orb-web genes that contain that type. Differences between the sequences that involve a charged residue are highlighted with a red asterisk. The underlined sequence for the GP-rich repeat indicates a length-variable region that differs across species (see Fig. S7 for full alignment). C) AgSp1 gene organization. Coding sequence depicted for each gene copy. Red blocks = terminal regions, gray blocks = EY region, green blocks = N-terminal random repeat region (NTR), tan block = GP-rich core repeat region, blue block = S-rich core repeat region, black line = transitional or non-canonical repeat regions. Core repeat regions are divided into distinct repeat modules (RM) that contain similar repeats (see Fig. 4a for phylogeny of the consensus sequence for each RM). Hatched blocks indicate that the repeat regions contain alternating units organized as ensemble repeats. Asterisks show the location of the introns. Taxa abbreviations as in Fig. 1.
Variation in repeat structure and sequence across AgSp2 genes. A) Gene organization provided for four orb-web weaving spiders with divergent structure (other species are generally similar in structure and sequence to A. aurantia). The height of each bar represents the size of that unit and the total length of the units on the Y-axis is proportional to the size of the coding region for each species (scale bar provide above M. hutchinsoni). Red bars = terminal regions, gray bars = transitional or noncanonical repeat regions, blue bars = core repeat regions, green bars = long repeat regions, yellow bars = simple repeat sequence found only in M. sagittata. Consensus or representative sequences are provided for each of the repeat regions. The L. cornutus sequence highlights an atypical long unit low in glutamine and rich in valine. M. hutchinsoni is presented as an alignment because four distinct, but homologous, repeat units are organized into an ensemble repeat (ER). B) Phylogeny of consensus sequences for core repeat regions (blue bars) in each orb-web species. Taxa abbreviations as in Fig. 1. species of 204 aa (range: 100-295 aa). While there is substantial sequence divergence and length variation in the AgSp3 RM1 sequence across the orb-weavers, all eight species that contain this region (M. sagittata and V. arenata do not have a small exon region) have an abundance of poly-threonine motifs (Table 1) and possess a conserved, positively-charged block with the consensus sequence RVHLHKHKRRIRNPHYR (Fig. S11, Fig. S12). The RM2 repetitive sequence is less structured than RM1, with only the Argiope species (unit size: 41-45 aa) and V. arenata (unit size: 198 aa) having clearly defined repeat units. There is also more dramatic variation in amino acid composition across species than in any other RM from all three AgSp genes. For instance, glycine constitutes an average of a quarter of the sequence across all the orb-weaver AgSp3 RM2 but ranges fourfold from 12.3 % (M. labyrinthea) to 49.4 % (M. sagittata). Similarly, charged amino acids range nearly sevenfold between 2.4 % (A. aurantia) and 16.7 % (M. labyrinthea). Perhaps the most striking difference involves the poly-threonine motifs which range in proportion from 38.4 % (Neoscona crucifera) to 0.3 % (M. sagittata). This divergence, however, is not driven by the overall proportion of threonine in these species as M. sagittata contains 27.1 % of this amino acid (exactly the average across all species), but rather by the distribution of threonine as either individual residues or motif blocks. Of the 1925 threonine residues that occur within the N. crucifera RM2 region, 93.1 % are organized into poly-threonine blocks, whereas only 10 of the 3729 threonine residues in the M. sagittata sequence are located next to another threonine. Unlike AgSp2, the AgSp3 sequence in L. hesperus contains repeat sequence with a very high abundance of poly-threonine motifs (Table 1, Fig. S12). There are numerous occurrences of TTTTT, a motif that is rare in any of the other AgSp genes.
The variation in repeat organization among AgSp genes provides an opportunity to examine aspects of homogenization dynamics. Overall, nearly all AgSp genes are highly homogenized (Table S4) with most RM exhibiting pairwise amino acid identity across all repeat units within that RM greater than 95 %. The primary source of variation among repeats units is large indels rather than amino acid replacements. For instance, as mentioned earlier, the ER units of L. hesperus AgSp1.2 consist of 3, 5 or 7 PGTTPG-units (that range in size from 933 to 2085 bp) so there are sizable indels when all the ER units are aligned with each other. However, independent of this indel variation, the amino acid sequence is virtually identical among all ER units with an average pairwise identity of 99.2 % (Fig. S9), indicating the homogenization process can operate across units of substantially different sizes. In addition, the homogenization is maintained by recombination processes rather than extreme stabilizing selection because the repeat similarity is maintained at the nucleotide level, even at synonymous sites. The average nucleotide pairwise identity (excluding indel variation) among the large RM units of the four Argiope AgSp1 genes is 99.2 % (range: 97.8-99.9 %; Fig. S13). Even more notable, this homogenization process can operate over vast distances. The immense gene, AgSp1.1 in L. hesperus, has a complex ER unit comprised of 2 or 3 PGTTPG-units that span the entire repeat region. Across a total of 145 ER units, each nearly 1 kb in length (total span 126.5 kb), the average pairwise nucleotide identity (excluding indels) is greater than 99 %.
The presence of distinct RM regions indicates there are constraints on the process of homogenization such that a single repeat unit rarely takes over the entire gene. In some cases, introns provide a barrier between different RMs (Fig. 4), but divergent repeat regions of the same type (e.g. A. aurantia, C. darwini, and M. sagittata AgSp1.1, and V. arenata AgSp1.2) can be maintained in a gene without the presence of intervening introns (Fig. 4 and Fig. S5). The transitions between GP-rich and S-rich units within AgSp1 genes also are not governed by the presence of an intron but there are often one or two units in the transition zone between RM that are divergent from both the GP-rich and S-rich units. These idiosyncratic units appear to be homologous across species and are characterized by an unusually high proportion of asparagine (Fig. S14). These transitional repeats are comprised of 5.5 % of this amino acid compared to 1.5 % in the GP-rich repeats and 1 % in the S-rich repeats. AgSp1 gene in M. labyrinthea is unusual in that it has several transitions between GP-rich and S-rich regions. Despite the physical separation between different regions of a given type, the repeat units appear to be homogenized. The average nucleotide pairwise identity (excluding indels) among all of the GP-rich units from the two separate regions (RM2.1 and RM2.2) is 99.0 % and there is no clear differentiation between repeat units from the two regions (Fig. 6). The three S-rich regions exhibit some level of differentiation within each region but this specific pattern of divergence is shared across all three regions (Fig. 6). The last ER unit of each RM is substantially diverged from the other ER units in that RM but similar across the three RM as they share 63 unique SNPs relative to all the other S-rich ER units (Fig. 6).
The functional diversity of spider silks is primarily the result of the abundant gene duplication and sequence diversification in a single gene family-spidroins [5][6][7]9]. Orb-weaving species generally possess 30-40 spidroin members that belong to seven primary types relating to their gland affinity. Cribellate spiders contain additional spidroin types (cribellate, paracribellate and pseudoflagelliform) related to their unique prey capture system. The evolutionary transition from cribellate silk to capture threads with aqueous glue required both the formation of specialized glands (flagelliform and aggregate glands) and the creation of new spidroins that are expressed in these glands. Previous studies [53][54][55][56] have distinguished two core paralogs of aggregate spidroins, AgSp1 and AgSp2, but the evolutionary conservation of this pattern was not well established. In this study, we show that the presence of these two genes is relatively stable throughout the Araneoidea and that an additional spidroin, termed Sp-8175 in previous studies [87,95], is clearly an aggregate spidroin, which we now call AgSp3, that arose from a duplication event involving AgSp2 at the base of the Araneoidea (Fig. 1).
Despite the phylogenetic affinity between AgSp2 and AgSp3, there is more structural similarity between AgSp1 and AgSp2 than either has to AgSp3. Unlike AgSp1 and AgSp2, AgSp3 does not possess PGTTPG repeat units and has a different gene organization. Expression of the three AgSp types is also relatively conserved across orb-weaving spiders. All AgSp genes are highly expressed in the aggregate glands with AgSp1 comprising the most abundant protein, a pattern that is also found in proteomic analysis [34]. This pattern indicates that a core structure of three AgSp silk glue genes likely arose in conjunction with the origination of the aggregate glands. Therefore, much of evolutionary change in glue performance and functionality within the Araneoidea likely involved modification to these three core spidroin components and how they interact with each other and other components of the glue droplet.
While some spidroins, such as those expressed in the major ampullate gland (MaSps), are characterized by extensive duplication at different taxonomic levels [5,58,87,89], the proliferation of gene copies within the three main AgSp types is moderate and limited to some species-specific duplications. Five species possess additional copies of AgSp1 while AgSp2 has duplicated within M. labyrinthea. In the orbweavers, the AgSp1 duplicates are generally distinguished by the relative proportion of the two repeat types (GP-rich and S-rich) with one copy comprised primarily of one type and the other copy possessing largely the second type. The dramatic shift in repeat composition among these species-specific paralogs represents the most rapidly evolving feature of AgSp genes. In addition, the differences in the repeat organization between genes suggest that an increased level of gene expression is not the primary selection pressure driving the duplications. Instead, it is likely either that the two copies interact to improve the functional performance of the glue droplet, similar to how MaSp1 and MaSp2 are complementary components of dragline silk, or perhaps the two AgSp paralogs provide alternative functionality and can be selectively expressed at different times, allowing the spider to fine-tune performance based on ecological or environmental conditions. For instance, serine is a prominent residue in the adhesive proteins of aquatic organisms such as sandcastle worms [96], mussels [97], sea cucumber [98] and caddisfly larvae [99] and often improves adhesion through phosphorylation [98,100]. Protein analysis of aqueous glues in both orb-web and cobweb weavers indicates that there are numerous phosphorylated sites in AgSp1 and AgSp2 [32,34]. Therefore, the increased serine content in the alternative repeat module of AgSp1 may facilitate adhesion in more humid environments. In all of the species with two copies of an AgSp1 gene except A. marmoreus, both paralogs 3) are distinct from the other units but similar to each other (grouping supported by 100 % bootstrap). C) Alignment of each of the last three RM1 units relative to the consensus sequence of all other RM1 units. Only three nucleotide differences exist between the last units of RM1.1 and RM1.2. are expressed at high levels (Table S3) suggesting they are likely to be found together in a glue droplet and may interact. The large number of differences at homologous positions between the GP-rich and S-rich repeats involving charged residues provides an opportunity for electrostatic interactions between the proteins.
Another noteworthy difference among the AgSp types is gene length. Like many genes coding for fibrous proteins, silk genes in both insects and spiders [55,101] produce extremely long proteins that facilitate extensibility, strength and toughness. The AgSp genes, however, stand out as including the longest silk genes in any organism. The average length of all orb-weaver AgSp genes excluding introns (29,347 bp) is longer than any other spidroin in the genomes of orb-weavers spiders [58,89,[102][103][104] or any insect H-fibroin [101]. Furthermore, one of the AgSp1 paralogs in the black widow, L. hesperus (AgSp1.1, 133.4 kb), is likely among the largest genes ever characterized [105,106]. Given that the extreme length of the AgSp genes is clearly derived since they are the most recently evolved spidroins [8,23], it is likely there have been strong selection pressures favoring this gene feature. It is noteworthy that AgSp2 in L. hesperus has lost its repetitive structure and also undergone an extreme reduction in length, highlighting the potential functional connection between these attributes. Recent analysis of allelic variation in silk genes found extensive length variation among haplotypes [107], and there are considerable interspecific length differences among the orb-weavers sampled in this study. Despite this variation, there are consistent differences among the three AgSp genes, with AgSp1 having the largest coding region followed by AgSp3 and then AgSp2. This pattern of substantial variation within a AgSp gene type but minimal overlap across AgSp gene types suggests that there is a functional range in which variation is selectively tolerated.
The glue droplet of an orb web constitutes a highly complex structure involving the interactions among numerous componentsmultiple proteins, salts, LMMCs, and waterthat result in emergent functional properties [30]. Across orb-weaving spiders there is large interspecific variation in droplet morphology (e.g. volume, shape and spacing), glue behavior (e.g. hygroscopicity) and mechanical performance (e.g. stickiness and elasticity) [30,31,[37][38][39][49][50][51]. Ultimately one of the primary goals of comparative genomic analysis of aggregate spidroin is to understand how each of the glycoproteins and their respective components (modules, repeats, motifs) contribute to the overall functional performance of aqueous glue. While results from this study cannot fully elucidate the structure-function relationships among AgSp sequence and aqueous glue properties, they do provide insights regarding critical areas of molecular variation. Overall, the complexity of the functional requirements of glue performance are mirrored at the genetic level. The typical gene organization for spidroins involves N-and C-terminal regions encasing a repetitive region of tandem repeats all similar in structure [57,58,[108][109][110][111]. Occasionally there are small linker regions interspersed with the repetitive sequence [108] but, in most cases, spidroins consist of three distinct regions. The AgSp genes, however, contain substantially more organizational diversity both within and among paralogs than other spidroins. Rather than having three distinct regions, most AgSp1 genes have 6-10 distinct regions (Fig. 4, Figs. S5, S10, S11) and it is likely that these different regions each contribute to different aspects of the aqueous glue's mechanical performance. For instance, the AgSp1 EY region near the N-terminus is highly differentiated from the rest of the repeat sequence but is present in all orb-weavers and some of the cob weaver L. hesperus AgSp genes but with reduced serine and elevated alanine. This conservation across species and families suggests that this region has a specific functional role, possibly impacting the hygroscopicity of the glue droplets through the abundance of hydrophilic residues.
Across the diversity of the AgSp genes examined in this study, certain molecular features, such as fiber-like motifs and charged amino acids, stand out as likely critical to mechanical performance. Orb-weaver aqueous glue behaves as a viscoelastic solid with substantial extensibility [31,35]. When an insect hits an orb-web, numerous glue droplets on a capture thread come into contact with the insect and work jointly to secure the prey. As the prey struggles, the capture thread stretches and the droplets form a 'suspension bridge' configuration in which the outermost droplets undergo substantial extension, assuming a fiber-like morphology [39,40,50], and do the most work [4]. The ability of these droplets to extend substantially without pull-off allows the inner droplets to be recruited to the adhesion process and the force to be transferred to the capture threads [4,51,112]. Despite the liquid state of glue droplets in nature, AgSp proteins have the capacity for fiber formation. Experiments using recombinant constructs comprised of AgSp1 repeat sequence and wet-spinning techniques were able to produce synthetic fibers with considerable strength and extensibility [113]. Several sequence characteristics that contribute to the extensibility of silk fibers, such as GP blocks [10][11][12]44,114], QQ motifs [45,58] and tyrosine residues [91,92], have been identified in MaSp and Flag spidroins and these are also present to varying degree in the AgSps. Specifically, the variable region of the GP-rich repeat of AgSp1 and the NTR and long repeats of AgSp2 are the primary repositories of these motifs. Therefore, species, such as the Argiope taxa, whose AgSp1 is primarily composed of the GP-rich repeat with a large variable region and contain numerous long repeat units in AgSp2 have a high proportion of fiber-like motifs that may increase the extensibility of the glue droplet during prey capture.
It is striking that the two species in our study that do not utilize a suspension bridge mechanism, the cobweb spider L. hesperus and the bolas spider M. hutchinsoni, have the lowest proportion of fiber-like motifs. Cobweb spiders utilize gumfoot glue droplets on vertical support threads that are under tension and that release from the ground with prey contact, dangling the prey in the air. Highly extensible capture gumfoot lines or glue might allow the prey to come into contact again with the ground and escape. In addition, because only the upper end of the gumfoot line is anchored, droplet extension contributes little to prey retention and, therefore, strength, not extensibility, is more important for retaining prey [32,115]. As a result, cobweb glue has a lower extension-to-break than orb-weaving glue [31,32]. Consistent with the requirements for greater stiffness and less extensibility, L. hesperus has AgSp1 with less GP-blocks and tyrosine residues than AgSp1 in most orbweavers and a truncated AgSp2 with no QQ-rich long repeats.
The bolas spider, on the other hand, produces a single large glue droplet that spreads rapidly across a moth's cuticle after contact with the insect and then quickly hardens. Thus, bolas spider AgSp is not expected to be rich in sequence elements associated with fiber formation or extensibility. M. hutchinsoni is the only araneid species in this study with an AgSp1 comprised almost entirely of the S-rich repeats that are low in GP-blocks and tyrosine and a AgSp2 without any long repeats. In addition, M. hutchinsoni AgSp2 is expressed at a much lower level (Fig. 3) and contains a more complex repeat organization (Fig. 5) than other orbweaving AgSp2 genes, suggesting it may play a different role in the adhesion process of this species.
While the fiber-like motifs increase silk extensibility, charged amino acids generally impact strength and stiffness both internally (cohesion) and externally (adhesion). Research on adhesives in other systems has shown that charged regions enact their effect through cross-linking [116][117][118], binding to charged surfaces [96,119], polymerization and self-assembly [120] and, in extreme cases, promoting liquid-liquid phase separation prior to adhesive bonding [121,122]. There are several areas of noteworthy molecular variation among the AgSps involving charged residues, all of which may influence the binding potential of the glycoproteins. These include 1) the AgSp genes of V. arenata and M. hutchinsoni, 2) difference between the GP-rich and S-rich AgSp1 repeats and 3) repeat module type 1 (RM1) of AgSp3.
The most striking potential relationship between mechanical properties and molecular structure involves V. arenata. This species' glue is dramatically stiffer than other orb-weavers, with a more than 10-fold greater elastic modulus relative to most species [39,93]. At the molecular level, AgSp1 V. arenata exhibit elevated levels of charged amino acids (Figs. S6-S7). This is largely due to numerous arginine replacements in both the GP-rich and S-rich repeats of both paralogs. The unique level of glue stiffness and charged molecular feature in this species suggests a strong functional relationship between these attributes that is likely driven by unusually strong cross-linking within and among the AgSp proteins. In addition, M. sagittata, which is closely related to V. arenata, has low glue stiffness and no elevated frequency of charged sequences, but another Micrathena species, M. gracilis, has glue that is also much stiffer than most orb-weavers [37,50]. M. gracilis wasn't sampled in this study but, if the charged molecular regions are responsible for the high glue stiffness, then we expect that the AgSp sequences in this species will contain charged sequence features similar to V. arenata.
The other species with elevated levels of charged residues is the bolas spider, M. hutchinsoni. For the standard AgSp2 repeat regions, the percentage of charged residues in M. hutchinsoni is over double that found in the other orb-weaving spiders. That this species would be grouped with V. arenata based on this feature is counterintuitive since their glue droplets exhibit dramatically different behavior. While V. arenata glue is stiff and viscous, the bolas spider's droplet is highly hygroscopic and spreads quickly across a moth's cuticle after contact [56,123]. However, despite the rapid spreading, the M. hutchinsoni glue must also maintain some elasticity. When the spider flicks the droplet towards a moth, it extends over five times its original diameter and, if it doesn't contact the prey, recoils back to its original shape [123]. In addition, it is possible that the bonding provided by charged residues in the two species is initiated at different time points in the adhesion process. In this scenario, elevated crosslinking is present in the V. arenata glue droplet when the aggregate glue is laid down and prior to any prey interception while in M. hutchinsoni there is minimal bonding when the droplet is made but strong ionic interactions are activated within the glue as it contacts the moth and spreads. There are examples of adhesives in other organisms that utilize a multi-part system in which components are kept separate initially but rapidly form complex bonds when mixed at the point of adhesion. This occurs either through the sequential addition of different proteins [124] or the compartmentalization of distinct protein types in heterogeneous sub-granules that then form strong bonds when they come into contact with each other [121,122]. Whether the bolas spider's droplet exhibits any type of phase separation is not known at this point but the rapid hardening of glue required to retain a thrashing moth is consistent with extensive crosslinking providing strong cohesion and adhesion.
While the phylogenetic analysis of the terminal domains (Fig. 1) indicates that the gene previously referred to as Sp-8175 is a core AgSp gene that originated with AgSp1 and AgSp2, little is known about the molecular structure or potential function of this spidroin. Proteomic analysis, however, has verified that AgSp3 produces a functional protein that is extruded as part of the glue droplet, albeit in relatively low abundance [34]. In two Argiope species, AgSp1 and AgSp2 combined represent approximately 40 % of all the glue protein whereas AgSp3 comprises roughly 0.5 % of the protein within the droplet [34]. Therefore, it is unlikely that AgSp3 is a major structural protein in aqueous glues, but it may facilitate interactions among other proteins or LMMC components to contribute to the mechanical properties of the glue.
Our comparative analysis of AgSp3 molecular variation highlights several notable molecular features that distinguish it from AgSp1 and AgSp2 and may be relevant to its function. First, there is abundant sequence and structural variation both across the gene and among species. RM1 and RM2 exhibit little sequence homology to one another and in species such as L. cornutus the gene is almost entirely composed of small exons while the repetitive sequence is in a single exon in V. arenata (Fig. S2). There is also substantial variation in amino acid composition across species. Future studies will need to assess whether this interspecific variation is driven by species-specific selective forces. Second, the N-terminal region of AgSp3 is interrupted by a highly hydrophilic block that is divergent from standard AgSp N-terminus sequence. We are not aware of another spidroin that possesses this type of modified N-terminus sequence. N-terminal regions influence storage and dimerization in fiber forming spidroins [125][126][127] but their role in liquid glue spidroins is less well understood. How the N-terminal regions of AgSp3 affects its molecular behavior will require additional experimentation. However, it is noteworthy that the presence of an insertion within the Nterminus and its hydrophilic composition is conserved between orb-web and cobweb weaving spiders (Fig. S4) suggesting that it has some functional significance. Third, most of the AgSp3 genes possess a highly positively charged block in part of its repetitive region (Fig. S12). Proteomic analysis [34] indicates that AgSp3 has the highest isoelectric point of any protein in the glue droplet so its functional significance is likely to involve electrostatic interactions. Fourth, AgSp3 genes possess a high proportion of poly-threonine motifs that may improve adhesion through glycosylation [34,35,112] but the extent of protein modification for this gene in not yet determined.
Overall, the AgSp molecular features associated with substantial variation in glue properties across species suggest specific roles for the different proteins within a droplet. For instance, AgSp2 appears central to cohesion and extensibility. The high proportion of fiber-like motifs in this gene in most orb-weaving spiders, the low extension-to-break value and the lack of an anchoring granule in L. hesperus which lacks a long repetitive AgSp2 support this hypothesis. Alternatively, the high proportion of residues and motifs (such as poly-threonine regions and serine) that can be glycosylated or phosphorylated [32,34] in AgSp1 and AgSp3 suggests these proteins play a larger role in adhesion. Given its high level of expression and relative conservation, AgSp1 is likely the primary adhesive protein while the high interspecific variation in AgSp3 may indicate the protein provides species-specific adaptive response to certain prey types or environmental variations. Orb-web glue droplets have a structured organization characterized by the presence of a granule that binds to the flagelliform capture thread and a core that contains a higher concentration of proteins [30]. Given the potential roles of the different AgSp proteins, we expect AgSp2 is concentrated in the interior, possibly forming the bulk of the capture thread anchoring granule, with AgSp1 distributed throughout the droplet and comprising most of the central region and AgSp3 occurring largely at the outer edges of the droplet. It is important to note, however, that many aspects of functional performance are not isolated to individual glycoproteins but likely emerge from interactions among the different spidroins, various non-spidroin proteins [34], and other components of the droplet such as LMMCs [27].
Most types of spidroins are highly homogenized, often exhibiting near-identical levels of sequence similarity across repeat units [5,57,58,108,109]. Homogenization of repetitive proteins, however, is not common across eukaryotes [128,129], suggesting that selection may be driving the pattern in spidroins. In addition, silk genes in other arthropods, such as the h-fibroin gene in the lepidopteran insect Bombyx mori, also exhibit complex, homogenized repeat structure [130,131], indicating that this organization is beneficial for many aspects of silk function. Not surprisingly, the aggregate spidroins examined in this study are homogenized in most repetitive regions but differ from most other spidroins in possessing numerous different regions of homogenized sequence within a single gene. This pattern of structured homogenization again suggests that selection favors a complex gene organization with identical repeats in most, but not all, regions. It is particularly noteworthy that the NTR region of the orb-weavers' AgSp1 is comprised of several PGTTPG-units that are strongly conserved across species but are not homogenized within the gene. This indicates that homogenization is not an automatic outcome for repeats with spidroinlike amino acid structure. The divide between the non-homogenized NTR and homogenized RM regions that are adjacent to each other highlights the potential variation in functional requirements for different sections of these genes. A combination of selection and molecular mechanisms such as mutation, recombination, and gene conversion likely shape this pattern. In addition, the observation that several species-specific gene duplicates contain different repeat types that are homogenized across the majority of the gene relative to the other AgSp gene copy points to the speed and efficiency of the homogenization process to transform a gene.
The primary organizational divide within the orb-weavers' AgSp1 involves the transition between the GP-rich and S-rich regions as nearly all of the gene copies in different species contain both types of repeat sequences. How each type contributes to performance and whether selection favors a specific proportion of each type in each species is not known at this point but the maintenance of each type across the araneids sampled here suggests there is functional significance to the combination and that selection favors mechanisms that preserve the repeat purity of each region. The PGTTPG-units that lie in the transition between GP-rich and S-rich regions are distinct from either of the two primary types and are relatively conserved across species and possess an unusually high proportion of asparagine (Fig. S14). The occurrence of this distinct repeat unit may serve to constrain recombination across subregions of the gene with different repeat types. By minimizing crossing over events between GP-rich and S-rich regions, the integrity of repeats within a given subregion are in effect isolated from each other. This is similar to the role that repetitive, homogenized introns potentially play in the maintenance of complex ensemble repeat structures in MaSp and Flag spidroin genes [58,132].
The potential complexity of the homogenization process in aggregate spidroins is best exemplified by AgSp1 in M. labyrinthea which has alternating regions of S-rich and GP-rich repeats that are nearly identical to each other. This is the only spidroin we are aware of in which sequences are homogenized across different regions that are separated by long stretches of alternative repeats. One possible explanation is that the gene was originally one type of repeat (e.g. S-rich) that was very recently invaded in two locations by GP-rich sequence, perhaps from another allele that contained a more standard S-rich/GP-rich split. However, the last repeat units in each of the three S-rich regions are both divergent from other S-rich repeats in the repeat modules and nearly identical to each other across the three separated units, suggesting that these three units have been distinct for some period of time but are still interacting with each other in a precise manner across the gene (Fig. 6). It will be critical to examine the AgSp1 gene structure from other M. labyrinthea individuals, as well as from other species within the genus, to reconstruct the pattern of RM evolution and homogenization dynamics that produce this unusual alternating repeat module structure.
Overall, selection likely impacts the pattern of repeat similarity within spidroins in two ways. First, selection favors repeat units that are virtually identical and often organized into hierarchical units of smaller repeats nested within larger repeats. Presumably, this improves silk performance. Second, selection favors specific gene features such as introns or divider units that constrain the homogenization process to produce the optimal repeat structure. Numerous molecular features, such as sequence motifs and histone modifications, have been associated with diversity in recombination patterns [133,134] and it will be important for future studies to investigate the molecular details impacting recombination and homogenization in the AgSp genes.