Foraging is one of the main evolutionary driving forces shaping the phenotype of organisms. In predators, a significant, though understudied, cost of foraging is the risk of being injured by struggling prey. Hunting spiders that feed on dangerous prey like ants or other spiders are an extreme example of dangerous feeding, risking their own life over a meal. Here, we describe an intriguing example of the use of attachment silk (piriform silk) for prey immobilization that comes with the costs of reduced silk anchorage function, increased piriform silk production and additional modifications of the extrusion structures (spigots) to prevent their clogging. We show that the piriform silk of gnaphosids is very stretchy and tough, which is an outstanding feat for a functional glue. This is gained by the combination of an elastic central fibre and a bi-layered glue coat consisting of aligned nanofibrils. This represents the first tensile test data on the ubiquitous piriform gland silk, adding an important puzzle piece to the mechanical catalogue of silken products in spiders.
Severe injury by prey is a high cost of predatory foraging, and the danger imposed by prey may affect a predator's choices considerably (Mukherjee and Heithaus, 2013). Counter-intuitively, some predators do not avoid but specialize on risky prey, and although different hypotheses have been raised to explain the evolution of such specializations, it remains an enigma (Pekár and Toft, 2015). Ground spiders (Gnaphosidae) are free-hunting spiders, some of which are very abundant in harsh arid environments. With currently 2200 described species in 125 genera and a worldwide distribution from sub-arctic to tropical regions, they are one of the major spider families (World Spider Catalog, 2016). Many gnaphosids have been reported to prey on ants or spiders (Bristowe, 1958; Grimm, 1985; Jäger, 2002; Jarman and Jackson, 1986; Pekár et al., 2012; Michálek et al., 2017), but the degree of specialization is unknown. Gnaphosids exhibit some distinct characters, which are presumably adaptations towards a specialization on hazardous prey. The most conspicuous characteristic is a strong modification of the spinning apparatus. Among araneomorph spiders, the first pair of spinnerets (ALS, anterior lateral spinnerets) usually bears the openings of single large major ampullate (MA) glands and numerous tiny piriform (PI) glands (Eberhard, 2010). The MA silk produces the main structural thread, the dragline, and from the PI glands short glue-coated micro-fibres emerge that fasten the dragline to substrates (Apstein, 1889; Wolff et al., 2015). This configuration is extremely conserved among araneomorph spiders, presumably because it is the basis for a versatile application of silk (Coddington, 1989; Coddington and Levi, 1991; Eberhard, 2010; Murphy and Roberts, 2015). However, in the Gnaphosidae, this configuration is strongly derived.
The gnaphosid PI glands and their nozzle-like openings, the spigots, are enormously enlarged and retractable, whereas the MA gland is comparably small (Kovoor, 1987; Murphy, 2007; Platnick, 1990). This was related to the following special technique to subdue hazardous prey (i.e. other spiders or ants). The spider quickly runs past the prey, thereby leaving a band of sticky silk behind, which immobilizes the prey's legs (Bristowe, 1958; Grimm, 1985). This mechanism, however, has never been studied in detail, because of the lack of suitable techniques. Furthermore, it is unclear how the morphological derivations of the ALS to suit a prey capture function affect the original function, namely the spinning of silk anchorages (attachment discs). Is prey capture a functional substitution or extension, and are there trade-offs between the two functions? Generally, silk anchorages with attachment discs should be much more robust against pull-offs than a thread's own glue coat, because the attachment disc structure generates a much higher contact area and effectively controls the peel-off angle (Pugno et al., 2013; Sahni et al., 2012; Wolff, 2017; Wolff and Herberstein, 2017). Nonetheless, an enormous mechanical impact is expected in the case of a struggling prey attached to a substrate by glue-coated silk. To understand the efficacy and significance of this predatory technique, it is therefore of high importance to know the tensile properties of the swathing silk. Furthermore, the mechanism and functional role of the infolding mechanism of the PI spigots is unclear.
We approached these questions by a multi-methodological approach, including: (1) behavioural observations using high-speed videography to reveal the use of silk during predatory attacks; (2) morphological investigations of the ALS (in the active and inactive state) and their glands using light microscopy, cryo-scanning electron microscopy (cryo-SEM) and micro-computed tomography (µCT), to reveal the modification of glands and the biomechanics of the spigots; and (3) micro-tensile tests and fracture analysis of isolated PI silk fibres to reveal the mechanical properties of the sticky swathing silk.
MATERIALS AND METHODS
Collection of spiders and silk samples
Gnaphosid spiders were collected by turning stones, peeling bark and sifting litter. If spiders were found resting inside a silken webbing, its structure was photo documented. A list of studied material is provided in Table 1. Spiders were kept in plastic containers with paper tissue that was slightly moistened once a week. Alpine species were kept at 15°C; all other species were kept at room temperature. Glass slides were laid into the containers and removed after a few days, in order to collect silk samples. Silk samples were studied with dissecting microscopes.
Gland preparation and light microscopy
The spinning glands of freshly killed spiders were dissected in embryo dishes using physiological solution (0.9% aqueous solution of sodium chloride) and viewed under an Olympus SZX12 stereomicroscope. They were subsequently transferred in a drop of physiological solution onto a microscope glass slide with a small prefabricated circular dimple and photographed under a Nikon Eclipse 80i light microscope.
Air-dried silk samples were sputter coated with 5 nm Au-Pd and viewed in an S4800 scanning electron microscope (Hitachi Ltd, Tokyo, Japan) at an acceleration voltage of 3.0 kV.
A juvenile Drassodex cf. heeri was attached to a sample holder using Tissue-Tek® compound, and then shock frozen in liquid nitrogen. The frozen specimen was directly sputtered with 10 nm Au-Pd using the Gatan ALTO-2500 cryo-system (Gatan Inc., Abingdon, UK) and viewed in an S4800 scanning electron microscope equipped with a stage cooled to −120°C.
A female individual of Scotophaeus scutulatus was fixed in 70% ethanol, dehydrated in a series of increasing ethanol concentrations and then critical-point dried. Dried samples were glued onto plastic pipette tips with cyanacrylate glue and scanned with a SkyScan 1172 HR micro-computer tomograph (Bruker microCT, Kontich, Belgium) with an acceleration voltage of 40 kV and a voxel size of 0.5 µm. 3D images were reconstructed using NRecon 1.6.6 software and processed with AMIRA 6.0.0.
Prior to prey capture trials, spiders were starved for 1–2 weeks. To investigate the use of silk during prey capture, we placed a gnaphosid and a prey animal into a cylindrical plastic container (diameter 4 cm, height 12 cm). The bottom of the container was removed and replaced by a clear glass slide, which was set in a custom-built 3D-printed frame containing lateral tunnels, in which the lenses of a gooseneck lamp were inserted. This induced frustrated reflection in all strands of silks contacting the glass surface (see Kleinteich and Gorb, 2015, for details). Additional lighting was applied from below the glass slide. As prey items, we used other spiders (Eratigena atrica, Amaurobius fenestralis, Zygiella x-notata), collected on the campus of Kiel University, and crickets (Acheta domesticus), obtained from a pet shop. Videos were recorded from below, using a Fastcam SA 1.1 (Photron Inc., San Diego, CA, USA) at 250 or 500 frames s−1.
We obtained high-speed video recordings of: D. cf. heeri capturing E. atrica (N=3) and A. fenestralis spiders (N=5); S. scutulatus capturing E. atrica (N=2), Z. x-notata (N=1) and a cricket (Acheta domesticus, N=1); and Zelotes sp. capturing E. atrica (N=1). Additionally, we observed S. scutulatus capturing an A. fenestralis (N=1); and Gnaphosa sp. capturing a cricket (N=1). All prey spiders and crickets were of equal or larger size than the gnaphosid, except for Z. x-notata, which was approximately 1/3 of the gnaphosid body size. One additional video was captured on a reflection interference contrast microscope (RICM) to investigate the extrusion and application of the PI silk and the change in its optical properties shortly after extrusion. As the RICM was an inverted microscope, the setup was basically similar, and the image was directed onto the camera chip via a beam splitter. After prey capture trials, glass slides with silk on them were stored and further studied by means of light microscopy.
To obtain single PI fibres for tensile testing, starved gnaphosids were placed into a Petri dish with a polymer film (ACLAR®-foil, Plano GmbH, Wetzlar, Germany) as a ground substrate and a prey spider (E. atrica). After the attack, often trails of parallel PI fibres were found on the plastic substrate. The adhesion of the PI glue to the polymer film was so low that fibres could be carefully peeled off without damaging stress. For this purpose, 0.3–2.0 mm long pieces of PI silk trails were cut at both ends with a razor blade. One end of the PI silk fibre was glued to the tip of a minute insect pin with a tiny amount of cyanoacrylate glue and then carefully detached from the substrate. The pin with the attached silk fibre was then attached to a three-axis micromanipulator (F-131.3SS, Physik Instrumente GmbH & Co. KG, Karlsruhe, Germany), and the free end of the fibre was glued onto another pin that was attached to a force transducer (FORT-10, 10 g capacity; World Precision Instruments, Inc., Sarasota, FL, USA). The thread was positioned such that it was pulled perpendicular from the force transducer at a constant rate of 50 µm s−1, which represents a quasistatic measurement. Forces were recorded with a Biopac data acquisition system (MP-100, Biopac Systems Ltd, Goleta, CA, USA). Tensile tests were recorded with the Photron Fastcam video camera using a frame rate of 50 frames s−1 and a shutter speed of 0.002 s, and equipped with ×5 to ×10 macro lenses.
Stress–strain curves were calculated from the force–time curves after Blackledge and Hayashi (2006). First, the curves were smoothed (averaging of each 25 data points, which corresponds to the frequency of the static noise of the transducer), to reduce the inherent noise of the data signal. True stress was calculated from the tensile forces divided by the cross-sectional area of the fibre. The initial fibre diameter was determined from SEM images of untested pieces of the same fibres. The cross-sectional area of the fibre was modelled assuming constant volume throughout the test and simplifying the geometry of the fibre as a half cylinder. This cross-sectional shape was found in the SEM observation of fracture faces of failed piriform threads. We thereby neglected the thin lateral extensions of the glue coat, as these do not significantly contribute to the thread volume. True strain was calculated as the natural logarithm of the actual length divided by the initial length, whereby the actual length was the initial length plus the test time multiplied by the strain rate. Because the deformation of the thread could potentially be non-linear, we additionally determined the strain from the video recordings at chosen time points and compared this with the calculated values. No clear difference was found.
From stress–strain curves, certain mechanical parameters were determined, namely extensibility (true strain at breakage), tensile strength (stress at breakage), yield strength (stress at the transition between elastic and plastic deformation, seen as a clear change in slope), toughness (integral of the stress–strain curve until breakage) and Young's modulus (initial slope).
In total, 17 PI silk fibres of Drassodex and two PI silk fibres of Scotophaeus were tested. Some test data had to be omitted because they contained not single but paired fibres. The fibres of Scotophaeus were not included in the statistics because of the small sample size.
Silk utilization and prey capture behaviour
In more than half of the observed cases, silk was extruded during attacks and applied both to the substrate and to the prey's legs and mouth parts (Fig. 1F–L). Only PI silk was involved in such attacks. Prior to silk extrusion, a protrusion and spreading of the spinnerets and inflation of the ALS apex was observed, leading to a wide spreading of PI spigots. However, there were cases in which no silk was extruded, and the prey was directly grabbed with the front legs and then overwhelmed with a leg basket (Fig. 1O–Q). This occurred three times in D. cf. heeri capturing A. fenestralis of equal size, in S. scutulatus capturing a smaller Z. x-notata and a cricket of equal size, and in Zelotes sp. capturing E. atrica of equal size. From our anecdotal observations, it is unclear which cues trigger either the use of silk or a direct attack. However, most spiders first tried a direct attack but quickly extruded silk if the prey turned out to be too large after the first physical contact (e.g. Fig. 1A–E). In the case of large prey items, the gnaphosid started several swathing attacks and rested still in between. The prey was then often already entangled so that it was significantly hampered in its movements.
The prey spiders often tried to defend themselves by biting. Whereas E. atrica was never successful in defence, A. fenestralis frequently succeeded in biting the predator (all observed cases), which led to a fatality in at least one case.
Silk left behind after the attacks included irregular puddles of solidified silk material (presumably resulting from an uncontrolled flow in the beginning of silk emergence) (Fig. 2S), parallel PI silk trails on the substrate with a tape-like morphology (Fig. 2K,O,P), and suspended PI threads with a cylindrical glue coat. We frequently observed plumose setae attached to the glue (Fig. 2U), which presumably originated from the spider prey. The PI silk exhibits a distinct core-coat structure that is easily discernible in a dissecting microscope (Fig. 2Q,R). In D. cf. heeri, the central fibre has a diameter of 2.2±0.3 µm (N=14), and the glue strip may spread to a width of 8–15 µm, depending on the wettability of the substrate. In the rapidly extruded PI threads, the width of the glue strip may vary, and the central fibre may sometimes appear bloated, but in the parallel trails, the structure is usually very regular. With means of reflection interference microscopy, we recorded the emergence and application of the PI silk onto a glass slide during a predatory attack (Fig. 2L). This revealed that the core-coat structure is already present at emergence, and that the glue coat completely cures within less than 1 s, as indicated by a change in refractive index (Fig. 2M,N).
Gnaphosids produce silken shelters, in which they hide during periods of inactivity (usually daytime) and when guarding an egg sac (Fig. 2A–C). The shelters consist of a meshwork of different thread types, including very fine fibres of sub-micrometre diameter as well as thicker ones (Fig. 2D,G). In contrast to comparable shelters in Clubionidae and Salticidae, the threads are not anchored to substrates by attachment discs. However, short irregularly curved trails of PI silk are occasionally applied to the loose meshwork to hold it in place on the substrate (Fig. 2D–H). Whereas in Scotophaeus and Drassodex, only a few such glue points were found, their use was more frequent in the webbings of Hemicloea. Overall, the silk shelters could be easily removed from smooth surfaces without causing major damage. Considerable adhesion of the gnaphosid webbing was observed on rough substrate surfaces, such as rocks and tree bark, indicating that the fibres are attached by mechanical interlocking. Silk is also extensively applied in egg sacs, which we did not analyse here in detail.
Of all studied species, occasional draglines and abseiling behaviour were only observed in Arboricaria sociabilis. Arboricaria sociabilis are very small gnaphosids, which do not exhibit such a high degree of PI spigot enlargement, and hence no widened PI silk trails. Nevertheless, the attachment discs used to fasten the draglines to the substrate (glass slide) exhibit an irregular shape as in the silk anchorages of silk shelters in other gnaphosid species (see above) (Fig. 2I).
Functional morphology of spinnerets
In the cryo-SEM study of a juvenile D. cf. heeri, we observed both the activated and the deactivated state of the ALS. In the deactivated state, the PI spigots are folded inwards and not visible from outside (Fig. 3A). However, the single MA spigot is situated on a separate part of the ALS apex that is not retracted, and thus is durably erect (Fig. 3E). With the help of µCT, the position of the PI spigots in the resting position in the ALS of S. scutulatus was visualized. The basal ALS apex is invaginated in this state, such that the conical PI spigots are clustered and situated in a horizontal position (Fig. 3D–F).
In the activated state, the ALS apex is inflated, causing the PI spigots to become erect and widely spread (Fig. 3B,C). The PI spigots are approximately 40 µm wide at the base and 8–10 µm wide at the tip, which is very large for a silk spigot in spiders. The cuticle appears rather thin and flexible at the tip, which leads to the opening being expanded under high pressure (when silk emerges) and collapsing under low pressure (in the resting state). This passively opens and closes the spigot opening, which presumably prevents uncontrolled silk loss and glue curing in the duct.
Two types of spinning glands open on the ALS of studied gnaphosids, the PI and MA glands (Fig. 3G–J).
We found a set of usually fewer than 10 large PI glands and the same number of reduced ones that were active in the previous instar (corresponding number of tartipores is visible next to well-developed spigots, towards the middle of the spinneret). The PI glands are large and whitish (Fig. 3I); they possess a long cylindrical ampulla with a narrow proximal part, composed of the secretory zone B and a tiny tail composed of less translucent secretory zone A. The proximal and distal secretory zones are not clearly separated, but two products can be distinguished inside the lumen of the gland based on their different colour and/or transparency. The PI glands occur in one cluster, including the long, relatively wide ducts that constitute an opened loop.
The MA glands (Fig. 3J) produce liquid crystalline material (the content of the gland lumen keeps its shape even when it is taken out – the material behaves like paste). Three pairs of MA glands can be seen: the large functional primary one (the only functional one, sensu Townley et al., 1993), the smaller secondary open one (which was functional during the last moulting) and the dwarf secondary blocked one (which was functional during the moult before last) (e.g. Fig. 3G).
The secretory part is tubuliform; its distal part is curled. The secretory zones are not apparent (perhaps there is only one); both are transparent to white and of the same width, but seem to slightly differ in their translucence. The duct is relatively short. It possesses an opened loop in its distal third.
The ALS silk gland system of the representatives of the clade Gnaphosinae (Gnaphosa and Zelotes) differs from that of the phylogenetically more basal Drassodes by reduction of the MA glands on the one hand and further enlargement of PI glands on the other hand (Fig. 3H).
Tensile properties of giant piriform silk
Of 11 useful tensile test replicates of PI fibres of Drassodex, we obtained the following mechanical properties (means±s.d.): extensibility 0.51±0.26 mm mm−1; true strength 511.0±123.6 MPa; yield strength 250–350 MPa; toughness 140.7±74.3 MPa; Young's modulus 5.59±1.75 GPa. The loading curve exhibits a shape that is characteristic for silks, with an initially high slope during elastic deformation, followed by a drop of the force increase and an extensive section of plastic deformation (Fig. 4G).
During tensile tests, we observed the occurrence of cracks in the glue coat after exceeding the yield point (at 5–10% extension; Fig. 4B,C). A study of the fractured PI silk by SEM revealed that in these cracks only the outermost skin layer of the glue coat was ruptured, which is a thin homogeneous film (Fig. 4E,H). Underneath, aligned nanofibrils were apparent, which form the bulk of the glue material (Fig. 4H). At cracks in the surface layer, especially at the underside of the thread, we often observed crystals (Fig. 4E,H,J,K), which may indicate salt-like substances that are embedded in the glue in the native state. Apparently, cracks in the glue coat were only present in the lateral extensions of the glue strip, but were evenly scattered above and next to the embedded central PI fibre (Fig. 4E,F). Fractured threads always curled towards the upper side (Fig. 4I), indicating an elastic behaviour of the embedded fibre. Transverse fracture faces exhibited smooth breaking edges of the glue coat and irregular, fibre-like fractures of the central thread (Fig. 4K,L). However, in longitudinal fractures, the core fibre showed a comparably smooth fracture face (Fig. 4J,K). This may indicate that the central core fibre is a highly anisotropic material. The central fibre was often pulled out of the glue coat at the breaking edge.
Consequences of the functional shift
We have shown here that gnaphosids are bold predators, able to subdue prey that are extraordinarily large and hazardous. All tested species attacked other spiders and preyed on them. Accessibility to a wide prey spectrum may explain why gnaphosids are especially successful in barren habitats with low arthropod abundance. Araneophagy is also known from related families, such as Lamponidae (Platnick, 2000; Michálek et al., 2017) and Cithaeronidae (Edwards and Stiles, 2011), although these do not exhibit a modified spinning apparatus. This may indicate that araneophagy evolved earlier than the ALS modification. Of all families within the Gnaphosoidea and allies, the Gnaphosidae are the most diverse and widespread, and their ecological success may be linked to their novel use of piriform silk.
The use of sticky silk is an efficient strategy to immobilize the prey before handling it, in order to reduce the risk of injury. Thus, the modification of the spinning apparatus might be an adaptation to handle hazardous prey. However, this special adaptation comes with the cost that gnaphosids (except for Arboricaria) cannot spin functional draglines any more, and the function of attachment discs is extremely reduced. The ability to attach threads to substrates via attachment discs (Fig. 2J) is regarded as one of the key innovations of araneomorph spiders that presumably highly enhanced the versatility of silk use and made the building of webs in a 3D space possible (Coddington and Levi, 1991). Draglines play a role in protecting the spider against unpredicted falls (Ortlepp and Gosline, 2008), controlling jumps (Chen et al., 2013) and on-water locomotion (Gorb and Barth, 1994), assisting navigation in webs (Barth et al., 1998), as elemental structures for webs (Denny, 1976), egg sac suspension (Gheysens et al., 2005), shelter building and intraspecific and interspecific communication (Leonard and Morse, 2006; Tietjen, 1977). Hence, a deviation from the usual ALS configuration (single large MA and an array of multiple small PI spigots) is extremely rare among araneomorph spiders (Coddington, 1989; Coddington and Levi, 1991; Eberhard, 2010; Murphy and Roberts, 2015). Although many (but not all) gnaphosids live on the ground and do not build webs, the reduced functionality of attachment discs may represent a significant draw-back, especially for the security of locomotion through a structured terrain, and the stability of shelters and egg sacs, all of which may potentially increase vulnerability to predation. For instance, we observed that the silk shelters of gnaphosids are easily removable from smooth surfaces and relatively easily torn open. In contrast, silken shelters of Clubionidae and Salticidae adhere strongly to smooth glass surfaces and are destroyed when attempting to pull them apart under the application of high forces.
Furthermore, silk use for prey capture is costly. Accordingly, in our prey capture trials, the spiders did not always make use of their silk. We presume that the spiders start swathing after an assessment of the prey's strength and dangerousness, because it was never started without previous physical contact with the prey. For direct attacks, the dense hairy adhesive pads (scopulae) in the front legs help the spider to hold onto the prey's body and subdue it (Eggs et al., 2015; Grimm, 1985; Wolff and Gorb, 2012; Wolff et al., 2013) (Fig. 1R). The fact that spiders precisely budget secretions that are potentially metabolically costly is also known for venom, the amount of which is adjusted to the type of prey (Boevé, 1994).
The use of sticky silk for prey immobilization is well known from various web-building spiders, such as orb web spiders (Araneidae) and cobweb spiders (Theridiidae) (Foelix, 1982). However, these have evolved an additional set of glands, the aggregate glands, which produce viscid glue, and the ALS are not modified (Coddington, 1989; Peters, 1987; Sahni et al., 2010, 2011). However, in daddy longlegs spiders (Pholcidae), the piriform glands have diversified, including a highly enlarged gland and spigot (Huber, 2000; Kovoor, 1987). This modification may be related to special wrapping attacks, too (Huber and Fleckenstein, 2008; Jackson and Brassington, 1987). Despite the modifications of the ALS spigots, pholcids retain the ability to spin attachment discs, but with a modified shape. Pholcids rarely spin draglines during locomotion (J.O.W., personal observation), but it is unclear whether this is due to an inefficiency in dragline attachment or because the glue material must fulfil both dragline attachment and prey immobilization functions, and should therefore not be used excessively.
Modifications in the ALS spinning apparatus related to the functional shift
Both gland types that open on the anterior lateral spinnerets differ morphologically from the situation in related spider groups. The MA glands are usually the largest spinning glands of spiders. However, in Gnaphosidae (especially in the representatives of the clade Gnaphosinae: Gnaphosa and Zelotes) we found them to be reduced, and smaller than the PI glands. The secretory part is not remarkably elongated and is not widened proximally (the structure for storing silk precursor before its usage, called the ampulla, is missing). The spinning duct is relatively short. In contrast, the PI glands and their spigots are highly enlarged, compared with those of other spiders, especially their proximal zone, which produces the glue coat (Kovoor, 1987; Kovoor and Zylberberg, 1980). Accordingly, the PI silk threads are 10–15 times wider than usual (compare with images in Wolff et al., 2015). Despite being significantly enlarged, the secretion product of these PI glands has a general appearance similar to the PI silk of other spiders, with a clear core-coat structure and a glue composed of aligned nanofibrils and a thin isotropic skin (Wolff et al., 2015).
The enlargement probably has two functions. First, it permits the quick expelling of large amounts of glue, in order to ensure spreading on rather complex surfaces (like the setose cuticle of arthropod prey). Silk trails left behind after a prey capture event often contain plumose setae of the prey. This indicates that gluing of body parts bearing such setae is hampered because of setal discharge. This effect is responsible for the escape of insect prey from sticky webs (Nentwig, 1982), and is also known from springtails, which are densely covered in scale-like setae, reducing the efficiency of glue as a means to capture them (Wolff et al., 2016, 2014). A thicker glue coat may ensure that the glue spreads not only on the loose superficial structures but also on the underlying stable surface (Voigt and Gorb, 2010). Second, a correlated increase of the diameter of the PI silk core may enhance its breaking force to meet the increased demand of mechanical resistance in single PI fibres due to high forces elicited by the struggling prey.
The widened ducts and spigots come with three potential problems. First, the pressure in the duct and nozzle is reduced. In most silks, shear forces in the duct and nozzle play an important role in aligning and elongating the proteins and formation of the fibre structure (Knight and Vollrath, 1999). The extent to which this is relevant in PI silk is unknown. As both the glue and the core are composed of aligned nanofibrils, it is conceivable that shear forces and/or self-assembly driven by weak intermolecular forces play a role in the formation of such anisotropy. Second, there might be an increased risk of desiccation of the aqueous silk dope in the spigot, which could lead to clogged nozzles. The flexibility of the spigot openings and their self-closing mechanism at reduced pressure might effectively prevent this. Interestingly, such flexible and self-closing spigots seem also to be present in the glue glands of other spiders, such as the aggregate glands of orb web and cobweb spiders (Coddington, 1989), and the enlarged PI spigot of daddy longlegs spiders (Huber, 2000). Third, in such voluminous nozzles, it might be difficult to control pressure and silk flow. Silk material might emerge from the openings and contaminate the exterior of the spigots and surrounding structures. We interpret the unique spigot (de-)activation mechanism as an evolutionary consequence to control silk flow.
Properties of piriform silk
Our tensile test data represent the first mechanical assessment of piriform silk. Blackledge and Hayashi (2006) have previously assembled mechanical data from tensile tests of most of the silken products of the orb web spider Argiope argentata, including major ampullate silk (dragline silk), minor ampullate silk (auxiliary spiral and bridging silk), tubuliform silk (egg sac silk), flagelliform silk (capture spiral silk) and aciniform silk (wrapping silk). The dragline silk is the strongest silk, with a strength of 0.6–1.2 GPa in different orb web spiders, whereas flagelliform silk is the stretchiest, with an extensibility of 1.2–1.8 (Blackledge and Hayashi, 2006; Denny, 1976; Köhler and Vollrath, 1995). The toughest silk is aciniform silk, with a toughness of approximately 240 MPa (Blackledge and Hayashi, 2006). The piriform silk of Drassodex, with a strength of 0.5 GPa, is not as strong as dragline silk and the wrapping silk of orb weavers (Fig. 4G, inset). However, it is more extensible than any other silk, except for the capture spiral threads of orb webs, which are three times stretchier, but less strong. In consequence, the PI fibres are as tough as the dragline silk of orb web spiders. The PI silk of Drassodex is less stiff than most silks, with the exception of flagelliform silk. These results are in line with previous theoretical estimations, which predicted that piriform silk should be rather stretchy, although extensibility and softness were extremely overestimated (Pugno et al., 2011). These properties can be related to random coil structures, caused by regularly spaced proline domains in the piriform spidroin, which were found to enhance flexibility (Chaw et al., 2017; Geurts et al., 2010). The fractographic analysis revealed that after the yield was exceeded, cracks occurred in the thin isotropic surface layer of the glue coat. Whereas wide cracks occurred in the lateral extensions of the glue strip, they are regularly distributed over the central fibre. This indicates that the fibre exhibits elastomeric properties, which evenly distribute the stress in the superficial glue coat. The glue seems far less elastic, as fractured threads always curled towards the upper side of the fibre, where less glue material was deposited. Furthermore, fracture faces always showed an even breaking edge of the glue coat, indicating a highly ordered, crystalline-like structure, whereas the central fibre exhibited irregular breaking edges. This was likewise found in the piriform silk of orb web spiders (Wolff et al., 2015), showing that the overall structure of PI silk has been kept constant in the course of the evolutionary transition. The PI glue is composed of aligned nanofibrils that are apparently pulled along each other during plastic deformation, which might delay crack propagation and makes the glue highly ductile and tough, as shown for similar materials (Brown et al., 2012).
For a glue, these are outstanding features. In attachment discs of both web-building and wandering spiders, PI silk breaks before detachment for surfaces with moderate to high polarity (Grawe et al., 2014; Wolff et al., 2015). This means that the adhesion of the glue coat exceeds its strength in these cases. Assuming that the pull-off stress works only in a small zone near the detachment (peeling edge), as proposed by previous authors (Pugno et al., 2011; Sahni et al., 2012), it would mean that the glue can withstand a shear strength of 400–700 MPa. For comparison, conventional artificial glues reach shear strengths of 0.9–1.7 MPa (Graham et al., 2016). It is unsurprising that the PI silk of Drassodex is considerably stretchy, which is likewise the case in the glue-coated threads of the capture spiral in orb webs (Blackledge and Hayashi, 2006). In orb webs, the capture spiral must absorb the mechanical impact of an insect flying in at high speed (Denny, 1976; Köhler and Vollrath, 1995). The swathing silk of gnaphosids must likewise absorb high mechanical stresses exerted by the struggling prey, because it is preferably applied onto the mobile and strong appendages of the prey. It is anticipated that in thread anchorages, the demands are rather similar, because sudden and heavy load may occur when a spider drops and stops its fall with an attached dragline, or in frame threads of webs that must contribute to the shock absorption of the prey impact. Thus, the changes in the spinning apparatus of gnaphosids are probably related to quantitative rather than qualitative adjustments.
The functional shift of PI silk use in Gnaphosidae is an intriguing example of a trade-off. We think that it can only have evolved because of the pre-existing frequent predation on hazardous prey, like spiders or ants. The immobilization of the prey with sticky silk presumably strongly reduces the risk of death during foraging, which may have contributed to the ecological success of this spider family. This must have been so beneficial that it outweighed the cost of a reduced ability to anchor silk threads. Future behavioural experiments may shed light on the degree of specialization and the balanced use of PI silk in different species of gnaphosids. Further, our mechanical data on PI silk add an important puzzle piece to the catalogue of properties of silken products in spiders. Whether the mechanical and chemical properties of PI silk remained constant throughout the functional shift from thread attachment to prey capture remains to be studied.
We thank Arno Grabolle for the identification of Drassodex cf. heeri and worthy communications on his extensive observations on European Gnaphosidae. Additionally, we thank Axel Schönhofer for the organization of the collection trip to the Southern Alps. Thanks to Mariella Herberstein for some discussions on araneophagic spiders. Thomas Kleinteich is acknowledged for the development of a 3D-printed glass cover slide frame for observation of frustrated reflection. Fabienne Frost assisted during prey capture experiments.
J.O.W. was supported by a doctoral scholarship of the German Merit Foundation (Studienstiftung des Deutschen Volkes) and a Macquarie Research Fellowship of Macquarie University. M.Ř. was supported by the Czech Ministry of Agriculture (project MZe RO0415). T.K. was supported by Internal Grant Agency of the Faculty of Environmental Sciences, CULS Prague (project 4211013123183). The µCT was founded by the German Science Foundation (Deutsche Forschungsgemeinschaft, DFG) (µCT Großgeräteantrag) to S.N.G.
Conceptualization: J.O.W., M.Ř.; Methodology: J.O.W., M.Ř.; Validation: J.O.W., M.Ř.; Formal analysis: J.O.W., M.Ř.; Investigation: J.O.W., M.Ř., T.K.; Resources: J.O.W., M.Ř.; Data curation: J.O.W., M.Ř.; Writing - original draft: J.O.W.; Writing - review & editing: J.O.W., M.Ř., S.N.G.; Visualization: J.O.W., M.Ř., T.K.; Supervision: S.N.G.; Project administration: S.N.G.; Funding acquisition: J.O.W., S.N.G.
The authors declare no competing or financial interests.