1000 spider silkomes: Linking sequences to silk physical properties

Expanding the repertoire of silk genes

The transcriptomes of 1774 individual spiders were sequenced, which included 1098 species belonging to 441 genera and 76 families, globally sampled from four continents. Redundant sampling was performed for certain species to observe locality or sex differences in spidroin expression (22, 38) and sequence variations within species. After the curation of the assembled transcripts, a total of 11,155 putative spidroin genes were identified (Fig. 1 and data file S1). All of the data are openly accessible from the Spider Silkome Database (https://spider-silkome.org).

The present study greatly expands the number and diversity of known spidroin sequences; we report sequences from 58 spider families not previously represented in public database, including members of basal taxa (Mesothelae, Mygalomorphae, Synspermiata, and allied groups), Araneoidea (which comprises the ecribellate orb weavers), and previously poorly sampled but extremely diverse groups such as the RTA clade and other taxa. At the time of writing this manuscript, spidroin sequences in the National Center for Biotechnology Information (NCBI) Protein database come from only 52 species in 18 families, and 23% of these sequence is derived from the single genus Trichonephila, and majority (73%) of the registered sequences are of major/minor ampullate spidroins (MiSps). In Fig. 1, family names colored in red indicate those with species where spidroin sequence is previously unreported, and family names with orange circles in front indicate those without previously reported transcriptome data. As the number of species indicates to the right of the family names, the vast majority of species reported in this work is previously unreported for spidroins and transcriptome data.

Within the “haplogyne” spider groups (Synspermiata and allied groups), we obtained sequences from nine previously unexplored families, including the first aciniform spidroin (AcSp), pyriform spidroin (PySp), and cribellar spidroin (CrSp) from these taxa. These proteins are consistent with more specialized silk types tuned to distinct biological functions in contrast to the undifferentiated spidroins identified from more ancestral Mesothelae and Mygalomorphae.

The most extensive sampling was conducted within Araneoidea, including family Araneidae, where we identified previously-unidentified spidroin sequences from the major subdivisions within the family (39), and likewise from underrepresented web-building families such as Tetragnathidae and Linyphiidae. Our results showed that the greatest diversity of spidroin types existed within the araneoid taxa, and spidroins associated with the capture spiral and aggregate glue of orb webs [flagelliform spidroin (Flag) and aggregate spidroin (AgSp)] were conserved only within the superfamily. Enrichment of the diversity of paralogs of the MaSp dragline gene was also observed in the group, and clear distinctions were possible among the different ampullate sequences (MaSp and MiSp) in terms of both terminal domain and repetitive sequences [for instance, MaSp2 is characterized by the presence of glutamine (Q)–containing dipeptide motifs in diglutamine (QQ)/proline-glutamine (PQ)/serine-glutamine (SQ)] (40). This is in contrast to Synspermiata and RTA clade sequences, where it is often difficult to distinguish between MaSp and MiSp types. The existence of a third type of MaSp (MaSp3), including the nephilid variant MaSp3B, appears to be specific to Araneidae (see below) (41, 42).

The RTA clade accounts for approximately half of all spider biodiversity, yet silk sequences from these mostly non–web-building groups have thus far received little attention. Our sampling identified a wide range of spidroin types from the RTA clade. To illustrate, we have identified the first spidroin sequences originating from jumping spiders (Salticidae), which has the highest species diversity among all spider families, with multiple representatives of MaSp, MiSp, AcSp, and cyllindrical spidroin (CySp), as well as unclassified spidroins from 63 different genera. We also extensively sampled spider groups situated between the araneoid and RTA clades [the so-called Uloboridae, Deinopidae, Oecobiidae, and Hersiliidae (UDOH) grade] and obtained the first reported spidroin sequences for Nicodamidae, Oecobiidae, and Hersilidae.

Insights from sequence analysis: Some highlights

The sequencing and annotation of the huge number and high diversity of spidroin genes from diverse spider taxa enable a deeper look into the more poorly resolved spidroin classes than previously possible. Here, we provide some examples of analyses made possible by access to such an extensive spidroin sequence database.

Spider silkome: An integrated database of sequences and material properties

Along with the spidroin sequence data, dragline silk fibers were collected from selected spider species, which were then subjected to a comprehensive array of analyses to obtain the individual profiles across 12 index parameters, including mechanical performance (toughness, Young’s modulus, tensile strength, and strain at break), morphological and structural properties [fiber diameter, birefringence, and degree of crystallinity based on wide-angle x-ray scattering (WAXS) analysis], thermal degradation profiles (onset temperature for 1, 5, and 10% weight loss), and hydration properties (fiber water content and degree of maximum supercontraction), for the reeled dragline silk of 446 spider species (Fig. 3 and fig. S1). Spiders belonging to Araneoidea show particularly diverse uses of threads (43), and the majority of the dragline samples included in this project was obtained from this superfamily, because the relatively large body size and copious fiber production of these species facilitate extended fiber collection.

Together, these data represent the largest collection obtained to date linking genotype to phenotype for a particular type of protein biopolymer (Spider Silkome Database; Fig. 3C), a fully searchable platform with integrated Basic Local Alignment Search Tool (BLAST) search capability. All sequence data are also available from DNA Data Bank of Japan (data files S1 and S2).

Study on the sequence to property linkage of spider silk has been a challenge, since the source of variability is threefold: interspecific, intraspecific, and intraindividual (44). Varying protocols for silking and mechanical property measurement also complicate meta-analysis, for which the silking strain rate and humidity is known to have significant effects (45). Our data are entirely obtained under a single standardized protocol and realize comprehensive comparisons. We therefore first observed the distribution of mechanical properties by families and genera. The mechanical property data obtained in this project represent an almost continuous spectrum of toughness reaching up to 0.45 GJ/m³, a strain at break up to 60%, and a tensile strength up to 3 GPa (Fig. 3B and figs. S1 and S2); thus, this dataset seems promising for ascertaining relationships between the amino acid sequences of silk proteins and the physical properties of draglines across the spider phylogeny [see also Craig et al. (46)]. Toughness is highly correlated with the tensile strength and strain at break, as expected from its definition. Notably, the correlation between tensile strength and strain at break is low, indicating that the strength and elasticity of silk are independent factors (Fig. 3A and table S1). Birefringence reflects the degree of molecular orientation of silk protein chains and is a good predictor of tensile strength; crystallinity is a similar predictor for strain at break. Silk diameter is correlated with strain at break and supercontraction, but the latter probably represents a pseudo-correlation with Sparassidae and Araneidae silks, which tend to exhibit large diameters and high supercontraction. Overall, web-weaving spiders, or those belonging to the superfamily Araneoidea, tend to express superior mechanical, physical, structural, thermal, and water-based properties relative to basal spider groups (Fig. 3B and fig. S1). Diversity in the mechanical properties was also the largest in the family Araneidae, mirrored by the high variability in the repetitive region sequences of MaSp-type spidroins (fig. S3), whose diversity nearly covers the entire variability within the 1000 spiders encompassing 76 families.

We conducted variable selection to probe structure-function associations in dragline silks based on the mean differences in the physical properties of the silks according to taxonomic categories and spidroin types (figs. S4 to S6). Briefly, the different ampullate-like spidroin sequences found across the different spider taxa were classified according to conserved patterns within repetitive domains; this led to the categorization into 20 sequence groups, which comprised seven MiSp subtypes, seven MaSp1 subtypes, four MaSp2 subtypes, and two MaSp3 subtypes, including MaSpN. We then analyzed the contributions of the different groups to the different physical properties of the corresponding dragline fibers. For instance, the silks of spiders from the genus Argiope and family Araneidae showed significantly higher toughness (mean differences of +0.068 and +0.039 GJ/m³, respectively) and expressed unique spidroins, including MaSp3 (group 19), MaSp2 (group 11), and MaSp1 (group 17), resulting in mean differences in silk toughness of +0.041, +0.031, and +0.035 GJ/m³, respectively (Fig. 4A and fig. S7A). This suggests that the possession of MaSp3 (group 19) resulted in an increase in toughness of at least 0.041 GJ/m³, corresponding to an increase of approximately 32% relative to the overall average of 0.127 GJ/m³. However, this was most likely as combined effect of Araneidae-type MaSps, including MaSp2 (group 11) and MaSp1 (group 17), coinciding with the existence of MaSp3 (group 19). A similar significant superiority of Araneidae dragline fibers was observed in terms of strain at break, crystallinity, diameter, thermal degradation temperature, and supercontraction. Strain at break and supercontraction were the only properties for which the possession of the MaSp2 subtype was a greater determinant than belonging to family Araneidae, as tensile strength increased 3.7% in association with MaSp2 (group 13) and supercontraction increased 15.7, 15.8, 14.3, and 11.0% in association with MaSp2 (group 14), MaSp2 (group 13), MaSp2 (group 11), and MaSp1 (group 17), respectively (Fig. 4B and fig. S7B). The significant contribution of MaSp2 to spider dragline supercontraction and elasticity was in line with previous suggestions regarding the different roles of MaSp1 and MaSp2 (47, 48), but one spidroin subtype, MaSp2 (group 15), conversely influenced supercontraction (−8.3%; see fig. S7B). A close inspection of the repetitive motifs of MaSp2 (group 15) revealed longer polyalanine regions. Accordingly, the average β sheet region length (typically the polyalanine region but defined as stretches of multiple A, S, and V for more than five amino acid residues, as these amino acids tend to substitute for polyalanine) was negatively correlated with supercontraction (−0.508 for MaSp1 and −0.306 for MaSp2 β sheet regions). Furthermore, the correlation was higher when both the amorphous region and the polyalanine lengths were taken together in the ratio (figs. S8 and S9). The average amorphous to β sheet region length ratios for all repeats within the spidroins of interest were 0.526 for MaSp1 and 0.394 for MaSp2. Therefore, the proportion of amorphous regions within the spidroin is the key factor contributing to supercontraction. The contribution of the relaxation of orientation in the amorphous region of spidroins to supercontraction was suggested in previous works (49, 50) and was confirmed by the analysis of our comprehensive dataset. Considering the effects of the amorphous and crystalline regions on the measured physical properties, as described above, the repetitive sequences, rather than the terminal domains, can be considered to play the main roles in determining these physical and mechanical properties. Shrinkage of artificial spider silk threads and textiles due to supercontraction is often considered an undesirable property for industrial use, and these findings may contribute in designing primary sequences, avoiding supercontraction while preserving toughness of the material.

To further extract the sequence features contributing to the physical properties of spider silk, we screened the amino acid motifs correlated with the measured properties (data file S4), and the main findings are summarized in Table 1. Confirming the above analysis of categorical variable selection according to gene class and taxonomy, the degree of supercontraction was strongly negatively correlated with the frequency of the appearance of polyalanine sequences and was correlated with short (one– to four–amino acid) motifs corresponding to amorphous regions such as G, GG, and AGQG (Fig. 4C and data file S4). Likewise, strain at break was negatively correlated with polyalanine prefixed with Ser in MaSp2 and positively correlated with MaSp1/2 amorphous regions including Pro, which presumably adds to the elasticity of this region (51). Concerning tensile strength, the inclusion of Ala in the amorphous region of MaSp1 and Pro in that of MaSp2 had a negative effect, while the inclusion of Ser in the amorphous region of MaSp1 had a positive influence. The GYGQGG motif in MaSp1 was most strongly correlated with both tensile strength (r = 0.377) and strain at break (r = 0.416) and was consequently also correlated with toughness [YGQGG was ranked 1 (r = 0.547), and GYGQGG was ranked 2 (r = 0.531)] (Fig. 4C). The Tyr residues in the amorphous regions of MaSp1 may play a critical role in intermolecular chain packing in the spider dragline, similar to the intermolecular interactions suggested from the structural analysis of silkworm silk (52) . The inclusion of Pro in the MaSp2 amorphous region, along with the SY and SV motifs in MaSp1, was negatively correlated with toughness. The presence of GGS after the polyalanine region in MaSp1 was positively correlated with toughness. Confirmation of the contribution of these motifs to the physical properties using recombinant properties would be a future direction to fully understand the primary sequence designs, leading to the extraordinary mechanical properties of spider silk.

Together, our findings provide a thorough mechanistic evaluation of the pathways of spidroin evolution. First, the physical properties of spider dragline silk have significantly diversified and specialized with the deployment of orb webs related to Araneoidea species (43), and this is mirrored by the diversification of MaSp paralogs, as previously suggested through meta-analysis of silk mechanics and sequence motifs (46) . We propose that MaSp1 is specialized to increase fiber strength, while MaSp2 is specialized to increase fiber elasticity, and the combination of these paralogs results in the high toughness of dragline silk. Furthermore, species requiring extraordinary fiber toughness have evolved to produce a third paralog, MaSp3, whose presence was clearly shown to be one of the strongest determinants of high toughness in our analysis. The full complexity of the proteome composition of dragline silk is beginning to be elucidated. However, MaSp3 was shown to be the major component of Nephilinae and Araneus dragline silks, and the complexity of these silks extends beyond the composition of spidroins (42), involving other essential components referred to as spider silk-constituting elements (SpiCE), which has been shown to double the tensile strength of an artificial spider silk–based film in vitro (41) . Elasticity and supercontraction are related properties of dragline silk that are likely linked to the sequence features of MaSp2, in which the ratio of amorphous to β sheet regions plays critical roles. Similarly, the compositions of several amino acid motifs in the amorphous regions of MaSp1 were shown to be highly correlated with the toughness of dragline silk; these sequence-level design elements derived from the comprehensive analysis of 1000 spiders provide a foundation for the design and production of artificial spider silks. Many of these designs may also be applicable to other protein-based and polymeric materials.

In this study, we have provided a comprehensive dataset encompassing the genotypes and phenotypes (including the mechanotypes) of spider silks and identified the design elements responsible for the extraordinary mechanical and physical performances of these silks. Silk proteins have convergently evolved in various lineages (53), but the sequence motifs (54), amino acid composition (55), and the trade-off between tensile strength and elasticity as a function of ratio between amorphous and crystalline regions (56) have been shown to have a certain degree of shared characteristics, something supported by our spider silk data. Therefore, these data will serve as a framework for the future analysis of silk proteins and other structural proteins as biomaterials. Similar data-driven approaches encompassing protein materials excelling in properties other than toughness, such as elastomers and adhesive proteins, could also accelerate our understanding on the genetic design principles of the biomaterials. Methods including computational modeling and simulation that allow the prediction of the outcomes of molecular interactions between the multiple components of these biomaterials, such as multiple MaSp-type spidroins and SpiCE proteins, would be an important future direction. We focused on the silk mechanics in this work, but the 1000 spider transcriptome data should also facilitate arachnid and arthropod phylogenomics.

Spider sampling

Field work took place from 2014 to 2019 in Japan, Malaysia, United States, China, India, United Kingdom, Australia, Madagascar, and Italy (data file S1). In each field work session, the collected spiders were stored in a New PP Sample Tube (Maruemu Corporation, Osaka, Japan) and transported live back to the laboratory. Spider specimens were identified by the method described in the “Species identification” section. Immediately after arrival at the laboratory, photographs of the collected spiders were taken on 1-cm by 1-cm grids to measure their total body length, silk was sampled by the method described in the “Silk sampling” section, and specimens were preserved or RNA was extracted for transcriptome sequencing. In field work conducted in countries other than Japan, preserved specimens or RNA samples were transported back to the laboratory. The total body length of spiders was measured from the photographs by using ImageJ.

Silk sampling

Spider silks were forcibly silked from captured spiders immobilized using two pieces of sponge and locked with rubber bands. After immobilization, the silks spun from spinnerets were obtained with tweezers and attached to the end of the bobbin. Dedicated reeling devices were used for silking with a constant reeling speed (1.28 m/min). The duration of forcible silking was 1 hour at most. After silking, the bobbin with the reeled silk was placed in a plastic bag and stored in a cardboard preservation box at room temperature.

Species identification

All spiders were morphologically identified by A. Tanikawa and subsequently confirmed by Cytochrome c oxidase subunit I (COI) sequencing from the transcriptome assembly based on BLAST searches in the Barcode of Life database with a 90% identity threshold. If the COI sequence could not be recovered from the transcriptome assembly, additional Sanger sequencing was conducted by amplifying the cDNA with the primer sets COI1490 (5′-GGTCAACAAATCATAAAGATATTGG-3′)–COI2198 (5′-TAAACTTCAGGGTGACCAAAAAATCA-3′) and COI1718 (5′-GGAGGATTTGGAAATTGATTAGTTCC-3′)–COI2776 (5′-GGATAATCAGAATATCGTCGAGG-3′) (57, 58). Spiders that were difficult to identify on the basis of morphology or COI database searches were clustered into groups of operational taxonomic units (OTUs). The OTU clusters were defined by a 98% identity threshold in BLAST searches.

Transcriptome sequencing and assembly

Sample preservation, RNA extraction, sequencing, and assembly were conducted on the basis of methods previously described (59) with some modifications. Briefly, a single specimen of each of the spiders brought to the lab alive was flash frozen with liquid nitrogen and stored at −80°C until use. Samples of spiders that were difficult to transport alive were stored in RNAlater, in which they were initially held at 4°C for 24 hours and then stored below −20°C. RNA was extracted after homogenization on a multibead shocker with metal cones (Yasui Kikai) using TRIzol (Thermo Fisher Scientific), followed by purification with an RNeasy Plus Mini kit. Small specimens (body size, <5 mm) were extracted using a Direct-zol RNA Microprep kit (Zymo Research). RNA quality was checked using RNA ScreenTape on a TapeStation 2100 (Agilent) according to an RNA integrity number (RIN) of >6 and was quantified using Qubit v.3 (Life Technologies) and NanoDrop 2000 (Life Technologies) systems. The Illumina library was prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs); however, for samples with available amounts below the required input amount (<20 ng of total RNA), preparation was performed using the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Clontech), followed by fragmentation and cDNA library preparation with the KAPA HyperPlus Kit (KAPA Biosystems). The sequence library was then sequenced as paired-end reads on the NextSeq 500 platform (Illumina) via 300 cycles in high-output mode. The sequences were subjected to base calling and demultiplexing, and adaptor sequences were removed with bcls2fastq v.2 software (Illumina). Transcriptome assembly was performed using Bridger software with the default parameters using Illumina reads (60). To eliminate possible cross-contamination, transcripts with mapped read count per million values of less than 1 and “comp” numbers greater than 30,000 in the Bridger assembly were removed. See data file S1 for the list of Sequence Read Archive (SRA) and Transcriptome Shotgun Assembly (TSA) accession numbers.

Direct RNA sequencing

The direct RNA sequencing library was constructed with the SQK-RNA001 kit (Oxford Nanopore Technologies). More than 500 ng of mRNA was prepared from the extracted total RNA using the NucleoTrap mRNA Mini Kit (Clontech), and library generation was completed following the manufacturer’s protocol. Appropriate numbers of individual samples were prepared according to the amount of total RNA from each species (data file S2). Direct RNA sequencing was performed using a MinION device, and one v9.4 SpotON MinION flow cell (FLO-MIN106, Oxford Nanopore Technologies) per species was used. The produced reads were corrected by using proovread (v2.13.4) (61) .

Spidroin curation and nomenclature

Spidroin gene curation was performed using a previously reported spidroin motif collection algorithm (37, 42) . The BLAST search detected contigs, including the spidroin gene N/C termini (nonrepetitive region). The obtained spidroin terminus contigs were used as seeds to screen the short reads harboring exact matches of extremely large k-nucleotide oligomers (approximately 100) up to the 5′-end of the seed. The selected short reads were aligned on the 3′-side seed of the matching k-nucleotide oligomer to build a position weight matrix (PWM). Using very stringent thresholds, the seed sequence was extended on the basis of the PWM until there was a split in the graph (i.e., the neighboring repeats were not resolvable). By iterating this overlap-based extension process, we obtained the full subsets of the repeat units. The collected repeat units were mapped onto error-corrected long reads obtained by direct RNA sequencing. Last, the spidroin gene length or architecture data were manually curated on the basis of the mapped long reads. The obtained spidroins were categorized into the following groups based on sequence homology with known spidroin data: AcSp, AgSp1/AgSp2, CrSp, CySp, Flag, MaSp1 to MaSp3, MiSp, Pflag, PySp, ampullate spidroin (MaSp or MiSp), spidroin, and putative spidroin (no homology but a spidroin-like structure).

Spidroin grouping

Curated MaSp/MiSp/spidroins were categorized into groups based on repetitive motifs. The repetitive regions and terminal sequences of the curated spidroins were separated computationally using the frequency of 5-nucleotide oligomer amino acids. The tree containing all spidroins was created on the basis of the N-terminal region sequences. The phylogenetic trees were created with FastTree (v2.1.10, default option) (62) after alignment with MAFFT (v. 7.273, maxiterate option 1000) (63) and trimming with trimAl (v. 1.2rev59, gt option 0.2) (64). Because the definition of MaSp and MiSp proteins based on sequence information was ambiguous, very MiSp-like MaSp (or vice versa) proteins were scattered. Therefore, we redefined the groups by clade to discuss them separately. A clade consisting of only spidroins of the same type was defined as a group.

Measurements of morphological and structural properties

The surface morphology and cross sections of the dragline silk fibers were assessed via scanning electron microscopy (SEM) (JCM-6000, JEOL Ltd., Tokyo, Japan) according to a previous report (65) . The samples were mounted on an aluminum stub with conductive tape and sputter-coated with gold for 1 min with a Smart Coater (JEOL, Tokyo, Japan) before SEM visualization at 5 kV.

Birefringence measurements

The retardation provided by the silk fiber was measured with a WPA-100 birefringence measurement system (Photonic Lattice Inc., Miyagi, Japan) and was analyzed with WPA-VIEW (version 1.05) software in accordance with a previously reported method (65). The birefringence of the dragline silk fiber was calculated from the retardation value and silk fiber diameter, which was determined via SEM.

Measurements silk of mechanics

Tensility tests of the single dragline silk fibers were conducted with a mechanical testing apparatus (EZ-LX/TRAPEZIUM X, Shimadzu, Kyoto, Japan) at 25°C and a relative humidity of approximately 50% according to a previous report (45) . The initial length of the single dragline silk fiber was set to 5 mm. The extension speed was applied at 10 mm/min, and the force during testing was measured with a 1-N load cell. The tensile strength, Young’s modulus, elongation at break, and toughness were obtained from the resultant stress-strain curves. To assess the tensile strength, the cross-sectional areas of the fiber samples were calculated on the basis of the diameters determined by SEM observations.

Thermal property measurements

Simultaneous thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) were conducted in triplicate using spider silk samples with a total mass of 0.5 to 1.0 mg according to a previous report (65) . Samples were encapsulated in aluminum pans and heated under a nitrogen atmosphere at a rate of 20°C/min from 30° to 500°C using a TGA/DSC 2 instrument (Mettler Toledo, Greifensee, Switzerland). The device was calibrated with an empty cell baseline and with indium for heat flow and temperature. The degradation temperatures that yielded 1, 5, and 10% weight losses in the silk samples were defined as degradation temperatures of 1, 5, and 10% (T_d1, T_d5, and T_d10). The water content was calculated from the percent weight loss associated with the evaporation of bound water from the TGA data based on a previous silkworm silk study (65).

Synchrotron WAXS measurements

Spider silk fibers were aligned in bundles and subjected to synchrotron WAXS at 12.4 keV at the BL45XU beam line at SPring-8 (Harima, Japan), as described in previous literature (45, 66). The data collection parameters included a wavelength of 1.00 Å, a beam size of 250 μm by 150 μm (H × V), and an exposure time of 10 s at 25°C and 40% relative humidity. Diffraction patterns were recorded using Pilatus 2 M (Dectris Ltd., Switzerland) with a sample-to-detector distance of 179.6 mm. The module gaps of the detector according to offset measurement were complemented. The two-dimensional (2D) diffraction patterns were converted into 1D profiles using FIT2D (67), with corrections made for background scattering and detector geometry. The degree of crystallinity of the silks was calculated from the 1D profile. Each dataset was separated into crystalline and amorphous scattering components by curve fitting using Gaussian functions. The ratio of the total area of the separated crystalline scattering components to that of the crystalline and amorphous scattering components was used to determine the degree of crystallinity.

Maximum supercontraction

The supercontraction of spider silks was evaluated according to a previous method (68). Individual dragline fibers were prepared by cutting fragments of 5 to 10 cm (L₀), to which a small piece of vinyl tape was affixed on either end. The fibers were immersed in Milli-Q water for 1 min to allow supercontraction and then allowed to air dry overnight in an unrestrained state. The final length of the fiber (L_f) was measured, and the maximum supercontraction (%) was calculated as (L₀ – L_f)/L₀ × 100. At least six replicates were performed for each sample; all measurements were performed with a caliper.

Silkome database (https://spider-silkome.org)

Statistical analyses

For categorical variable selection, mechanical properties were tested to evaluate differences between their mean values for those belonging to the category and those not in the category, with the unpaired Student’s t test with a P value threshold of <0.01 using the G-language Genome Analysis Environment v.1.9.1 (69–71) . The categories used in this analysis were the family and genus of the spiders as well as the spidroin type. For example, the toughness value distribution of family Araneidae was compared with those of all other families. A minimum of five samples was required to belong to a category.

For motif extraction, repetitive regions of spidroin sequences were first extracted as the longest segments, spanning an amino acid motif composed of S, A, or V with a length greater than four. Subsequently, these regions were split into repeat units segmented by SAV motifs with a length greater than five. This SAV region was defined as the crystalline region, and the remaining amino acids within the repeat were considered the amorphous region. Amino acid motifs with lengths of one to eight were counted in the repetitive region and divided by the number of repeats. This occurrence value was averaged for all MaSp1, MaSp2, and MaSp3 paralogs, and the correlations of these values with the mechanical properties were calculated using Pearson correlation in the G-language Genome Analysis Environment v.1.9.1 (69–71) . Motifs appearing less than 100 times in total among all spidroin sequences and motifs appearing in less than 30 samples were discarded. Graphs were visualized using JMP v.15 software.

This PDF file includes:

Other Supplementary Material for this manuscript includes the following: