The anchorage of structures is a crucial element of construction, both for humans and animals. Spiders use adhesive plaques to attach silk threads to substrates. Both biological and artificial adhesive structures usually have an optimal loading angle, and are prone to varying loading situations. Silk anchorages, however, must cope with loading in highly variable directions. Here we show that the detachment forces of thread anchorages of orb-web spiders are highly robust against pulling in different directions. This is gained by a two-step back-and-forth spinning pattern during the rapid production of the adhesive plaque, which shifts the thread insertion point towards the plaque centre and forms a flexible tree root-like network of branching fibres around the loading point. Using a morphometric approach and a tape-and-thread model we show that neither area, nor width of the plaque, but the shift of the loading point towards the plaque centre has the highest effect on pull-off resistance. This is explained by a circular propagation of the delamination crack with a low peeling angle. We further show that silken attachment discs are highly directional and adjusted to provide maximal performance in the upstream dragline. These results show that the way the glue is applied, crucially enhances the toughness of the anchorage without the need of additional material intake. This work is a starting point to study the evolution of tough and universal thread anchorages among spiders, and to develop bioinspired ‘instant’ anchorages of thread- and cable-like structures to a broad bandwidth of substrates.
Thread- and cable-like objects are ubiquitous in our daily environment, and usually fastened to structures and buildings by cumbersome mechanical attachment devices like hooks, clamps, knots and screws. To find better attachment solutions for technological applications, animal models may reveal novel mechanisms of how to attach material to substrates. For example, orb-web spiders are both climbers and architects using silk cables with fascinating elegance and efficiency. A crucial part of both abseiling and web building is the attachment of the main silk thread (the dragline) to substrates, which is performed with remarkable speed and without any pre-preparation or penetration of the substrate. For substrate attachment spiders use numerous short glue-coated nanofibres drawn from the piriform glands [1,2]. Both the dragline-producing major ampullate gland and the piriform gland ducts lead to nozzle-like openings, the spigots, located on the paired anterior lateral spinnerets. In order to attach the dragline to a substrate the spinnerets are rapidly rubbed against each other and on the substrate [1,2], but a basic understanding of this process is pending. The result is a thin plaque, the attachment disc, consisting of a network of agglutinated piriform gland fibres, with the dragline embedded.
Spider attachment discs are intriguing and unique glue applications, because the spider can specifically and directly adjust the strength of the anchorage by the way it applies its glue . The anchorages of structural threads and draglines can yield a high pull-off resistance on highly unpredictable and problematic surfaces, at a minimum of material consumption [4,5]. Attachment discs even gain considerable adhesion on highly repellent non-polar surfaces like Teflon . The underlying mechanisms of this remarkable performance, however, are largely unknown. Furthermore, it is unclear how silk anchorages behave under different loading situations, naturally occurring in both web and dragline attachments.
First attempts to understand the structure and function of spider silk anchors relied on numerical models, using Kendall's theory on the peeling of elastic tapes from rigid surfaces , and the multiple peeling theory by Pugno . In tape-like structures the force that is necessary to break the bonding is ruled by the pull-off angle: at a steep angle (i.e. perpendicular to the substrate surface) the loading stress is highly concentrated on the peeling edge and a delamination crack is relatively easy induced, whereas low angles lead to a higher pull-off resistance, because the stress is distributed over a larger area . That means that the pulling direction of a single glued thread has an enormous effect on the strength of its attachment. However, if two contralateral tapes are combined and pulled off simultaneously, the peeling angle stabilizes at a lower value than the pulling angle, leading to much higher robustness and resistance [8,9]. Previous authors therefore highlighted the fact that in spiders the dragline is attached with accessory glue fibres and described the attachment disc as a sequence of parallel tape-like piriform threads, transversally overlying the dragline (‘staple-pin model’) [3,10,11]. In this configuration a vertical pulling of the dragline leads to a contralateral peeling, subsequently starting in each tape-like element. By that mechanism the pull-off force rises until the first tape is totally delaminated [8,9].
While the staple-pin model may explain the enhanced robustness of an anchorage based on accessory glue threads, its presumptions may differ from the empirically observed structure and mechanical behaviour of attachment discs. For instance, in most spiders, the dragline is not overlaid by the plaque, but enclosed in an envelope of conglutinated piriform gland threads (the connection piece) that is suspended in a network of branching piriform fibre bundles (bridge) connected to the basal plaque [4,6]. The piriform gland threads in the plaque rarely form separated, parallel glue trails, but frequently cross each other and form a fairly continuous film . Furthermore, the piriform gland threads are rarely orientated perpendicular to the dragline, but rather in variable directions [4,6]. These discrepancies between observations and numerical models may, in part, be explained by the lack of microscopical and experimental data. However, the way in which the dragline is joined to the plaque may have profound effects on the proposed fracture mechanics and the robustness of the anchorage. For instance, the separate joining of dragline and substrate with the attachment disc means that the initial tensile load may not act at the front edge of the plaque but rather more centrally and over a larger area, depending on how the bridge network is organized. Hence, the attachment discs of spiders may work like an axisymmetric membrane, where the peeling line is constantly growing and the pull-off force rises throughout the detachment process . In this configuration the delamination crack is initiated in the middle part of the plaque and grows radially around the loading point with an increasing circumference, which enhances resistance against detachment [12,13].
Depending on which model is used, we would predict a different resistance against variable pulling angles and different impact of geometrical parameters on the yielded peak pull-off force. (i) Single tape model: this model predicts a decrease in peak pull-off force with increasing pulling angle. The main determinant of pull-off force would be the tape (plaque) width. (ii) Staple-pin model: in the contralateral peeling of a bifurcating tape the mean of the peeling angles of both branches is hypothetically constant, regardless of the pulling angle [8,11]. Therefore, no difference in pull-off force between vertical and lateral pulling is predicted. Although, we would predict a decrease in peak pull-off force with increasing pulling angle along the longitudinal axis, since stress concentration at the front edge would increase. The main determinant of pull-off force would be both the plaque width and length. (iii) Axisymmetric membrane model: the maximal pull-off force is predicted to be similar in all loading situations, because all radial axes represent a double-peeling case. The main determinant of pull-off force would be the distance of the loading point from the front edge and both width and length of the plaque.
In order to quantify and describe the mechanical function of spider silk anchorages we studied the pull-off forces of attachment discs under four different loading situations. We further related measured forces to geometrical parameters of the attachment discs, which vary naturally. We describe the actual structure and formation of silk anchorages, especially the joint of the dragline with the attachment disc, using high-resolution three-dimensional electron microscopy and micro high-speed videography. Additionally, we used tape-and-thread models to test the influence of geometric parameters on the fracture mechanics in plaque-like attachment structures. Finally, we tested the directionality and tailoring of attachment discs, by perpendicular pull-off tests on upstream versus downstream draglines.
With our complementary approach we aimed to assess the robustness of silk anchorages for real-life situations where threads are loaded at various angles. Furthermore, we aimed to find structural principles that yield the strongest anchorages, to generate hypotheses about the evolution and plastic use of these structures, as well as the development of novel bioinspired glue applications and cable attachments.
2. Material and methods
2.1. Silk sampling and morphometry
Golden orb-web spiders (Nephila plumipes) are common along the East coast of Australia and build large, two-dimensional orb-webs with the addition of a barrier web . We collected individuals from different localities in the Northern Sydney metropolitan region (West Pymble Park, Brooklyn, and Hornsby). They were kept in plastic cups covered with gauze at 23–25°C and relative humidity of 45–55%. Spiders were fed every 7 days with a variable diet of house flies (Musca domestica), fruit flies (Drosophila spp.) and small crickets (Acheta domestica) (see  for spider husbandry).
Silk samples were collected on a stiff polypropylene film, which was mounted onto the walls of the containers housing the spiders. The spiders attached their threads to the film. Other samples were acquired by encouraging the spider to walk upside down over a microscopy glass slide, covered with the polypropylene film. Spiders attached their dragline onto the slide from time to time to secure themselves. This behaviour can be triggered by cutting the dragline of the spider, gently blowing at it or by a shake. Draglines were carefully cut at 2 cm above the attachment disc to prevent pre-stressing of the anchorage. For measurements single silk anchorages were isolated by carefully cutting the film into pieces. The pieces were then attached to a glass slide with double-sided tape. In total 105 attachment discs were harvested from 12 adult female spiders.
Each attachment disc was photographed prior testing with a Canon EOS 600D DSLR camera (Canon Inc., Tokyo, Japan) mounted onto a dissecting microscope (Motic Inc. Ltd, Hong Kong), using 6× to 10× magnification. To determine the pixel size of the micrographs, a micrometre scale was photographed at each magnification. Morphometric measurements were performed with ImageJ 1.5 . We took measurements of the following parameters: (i) projected plaque area (area), (ii) maximal width (width), (iii) maximal length (length), (iv) length of the connection piece (connection length) and (v) distance between the dragline insertion point and the front edge of the attachment disc (front shift). There was considerable variation in the morphometric parameters (see results), but there were no significant differences of means in samples used for the different pull-off tests (Kruskal–Wallis rank sum test in R; area: p = 0.24; width: p = 0.24; length: p = 0.29; connection length: p = 0.27; front shift: p = 0.42). Although hand-collected attachment discs tended to be smaller than such collected from containers, there was no significant difference in size and front shift.
Silk samples for 3view imaging were collected from a juvenile N. plumipes on Thermanox™ plastic coverslips by letting the spider walk over the slide (as described above). A juvenile spider was used, because the size of samples for 3view SBF-SEM is restricted, and because of a greater resolution of the microstructure due to the inclusion of fewer fibres than in the attachment discs of adult spiders.
2.2. Pull-off tests of silk anchorages and models
To measure peak detachment forces at different pull-off angles the glass slides were attached to a clamp with an adjustable arm that could be rotated and fixed in different positions. Tests were performed with an Instron 5542 tensile tester (Instron, Norwood, USA), using an ULC-0.5N load cell (Interface, Inc., Scottsdale, AZ, USA). A clamp with an attached piece of cardboard was mounted onto the sensor and the upstream dragline of the silk anchorage was attached to that cardboard with a small amount of cyanoacrylate glue applied at a length of 5 mm above the attachment disc insertion point. Samples were pulled off at a constant extension rate of 10 mm min−1. Load force–extension curves and peak pull-off forces were recorded with the Bluehill 3 software. Loading angles were 0° (i.e. parallel to the substrate, towards the spinning direction; 20 samples), 90° (i.e. perpendicular to the substrate; 25 samples), 180° (i.e. parallel to the substrate, against the spinning direction; 20 samples) and sideways (i.e. parallel to the substrate, perpendicular to the spinning direction; 20 samples). Additionally, perpendicular pull-off tests were performed on the downstream dragline (20 samples). Exemplary tests were video recorded with a Canon EOS600D SLR camera equipped with an extension tube. The environmental conditions in the lab varied between 20–25°C and 25–35% relative humidity during the experiments.
Data were statistically analysed with R v. 3.3  using a significance level of 0.05. Data of different loading situations were statistically compared using the Kruskal–Wallis test, followed by pairwise Wilcoxon rank sum tests with false discovery rate (FDR) p-value adjustment. Data of upstream versus downstream dragline pulling were compared using the Wilcoxon rank sum test. In order to identify which morphometric parameters significantly influenced pull-off forces, we performed a separate multiple regression (linear model) for each loading situation.
To test the effect of different morphometric parameters and spinning patterns we built simple tape-and-thread models. Twisted acrylic knitting thread was placed onto a glass slide and then attached with a piece of double-sided Sellotape®. The top side of the tape was covered with a thin polyethylene film to prevent fracture of the tape when the thread was loaded. Six different model types were produced, each with 10 specimens. In class 1 models, the thread was connected over its entire length with the tape. In class 2 models, there was a 6 mm notch in the front, simulating a front shift as observed in spider attachment discs. For each class, we produced three different types of models with type A using a 12 × 12 mm piece of tape (quadratic), type B a 20 × 12 mm (elongated) and C a 12 × 20 mm (broadened). The glass slides with the attached specimens were mounted onto the tensile tester using strong clamps and the thread was attached to the Instron ± 500 N Static Load Cell with a pneumatic clamp at a length of 25 mm above the tape insertion point. Samples were pulled off (initially) perpendicularly to the substrate at a constant extension rate of 1 mm s−1. Peak pull-off forces, peeling energy and yield resistance (force per area of tape) were statistically compared using ANOVAs followed by Tukey tests.
2.3. Three-dimensional electron microscopy
Silk samples on Thermanox™ plastic coverslips were prepared using the protocol by Deerinck et al.  and embedded in procure 812 epon resin (hard grade). Polymerized blocks were mounted on aluminium specimen pins, sputter coated with gold (15 nm thickness) and the edges were painted with conductive silver paint to reduce charging effects. Samples were imaged with a Zeiss Sigma VP field emission scanning electron microscope (Carl Zeiss AG, Oberkochen, Germany) equipped with a Gatan 3view serial block face system (Gatan, Inc., Pleasanton, CA, USA). Images were recorded at a field of view of 375 × 499 µm (178× magnification), and at an acceleration voltage of 4 kV in low vacuum (20 Pa). The section thickness was 100 nm, and approximately 800 slices were recorded at a pixel size of 50 nm to cover the complete silk anchorage at high resolution. For three-dimensional reconstruction and digital sectioning, the obtained datasets were converted, cropped and scaled down with Fiji  and processed with the Avizo 3D software (FEI, Hillsboro, OR, USA).
2.4. Spinneret movement tracking
To investigate the movement pattern of the spinnerets during attachment disc production, we used a micro high-speed video camera set-up. Spiders were shortly anaesthetized with CO2 and tethered to a wooden stick with a rolled piece of double sided adhesive tape that was carefully wrapped around the prosoma (hard front body part) between legs II and III. This fixation ensured attachment disc production in the relatively small field of view of the camera. It was important that movement was not impeded too much, since the spiders only produced silk anchorages when the opisthosoma (hind body where the spinnerets are located) and legs were freely moveable and could be moved into a natural position. Spiders were held in an either head-up or head-down vertical position, slightly above the surface of a vertically placed glass slide. Spiders often got hold of the edges of the glass slides with their walking legs and then anchored a thread onto the glass surface. This event was filmed through the glass slide with a Basler Ace 640 × 480pix USB 3.0 camera (Basler AG, Ahrensburg, Germany), equipped with a Navitar Precise Eye extension tube including a 1.33× magnification lens and a 0.25× accessory lens (Navitar, Inc., Rochester, NY, USA). The resulting field of view was 5.3 × 4.0 mm at a pixel size of 8.3 µm and a working distance of 310 mm. Videos were recorded at 500 frames per second, using the TroublePix software (NorPix, Inc., Montreal, QC, Canada) with continuous looping and post event trigger.
Videos were further processed with ImageJ 1.5 . The images were rotated such that the median symmetry axis between both anterior lateral spinnerets was orientated vertically. Images were cropped to the relevant field of interest (i.e. the area were all movements of the anterior lateral spinnerets took place) to reduce processing time for further adjustments and tracking. The contrast was enhanced to increase the visibility of the structures and silk. The movements of both anterior lateral spinnerets were manually tracked using the MTrackJ plugin , using the centre of the anterior lateral spinneret spinning field as a reference. Data points were exported into MS Excel (Microsoft, Redmond, WA, USA) and the x- and y-shift per time unit was calculated taking the final upstream dragline insertion point as point zero. We further calculated the movement direction (angle), velocity and change in total track length between each two subsequent data points.
3.1. Three-dimensional organization of the joint
The silk anchorage of golden orb-web spiders (N. plumipes) has the general structure as described by Wolff et al. : The main silk thread, the dragline, is combined with the attachment disc via a cylindrical envelope (connection) that is flexibly hinged with the adhesive plaque by a network of branching nanofibre bundles (bridge). The high-resolution three-dimensional reconstruction of silk anchorages via 3view SBF-SEM revealed the structure of the joint of the dragline with the attachment disc (figure 1). The dragline is a bundle of 2–8 major ampullate fibres (depending on the developmental state of the spider), that are separated in the connection piece and enclosed by closely packed, accumulated glue coated piriform silk bundles. In some sections the major ampullate fibres are highly agglutinated with the piriform silk envelope, whereas in other sections, it is freely suspended in tube-like cavities. The piriform silk envelope is relatively thin and rather close to the plaque towards the downstream side (figure 1f), whereas it is very thick and more elevated above the plaque towards the upstream dragline insertion (figure 1h). The connection piece is tied down to the extensive, very thin (submicron) plaque by bundles of piriform silk fibres, which emerge from the upper part of the connection piece and run laterally towards the plaque, thereby branching. The greatest amount and length of fibres in the bridge network is present around the upstream dragline insertion point, which represents the primary loading point in dragline anchorages. This point is rarely located at the front edge of the attachment disc, but is often shifted towards its centre (figure 1a). The plaque consists of crossed over piriform silk trails, forming sharp curves at the lateral fringes of the attachment disc.
The attachment discs of adult female N. plumipes have an average size of 2.34 ± 1.02 mm2 (mean ± s.d., N = 105), width of 1.60 ± 0.33 mm, length of 1.82 ± 0.57 mm, connection piece length of 0.59 ± 0.35 mm, and a front-to-centre shift of the dragline insertion point of 0.51 ± 0.18 mm.
3.2. Spinning pattern
By means of micro high-speed videography we observed that attachment discs are produced in a two-step process. Step 1 is the formation of the connection piece by multiple rubbing of the spinning fields of the anterior lateral spinnerets against each other. This rubbing rolls the dragline over the piriform spigot fields, thereby wrapping it in piriform silk. Step 1 is usually directly followed by step 2 (plaque production), but there may also be a considerable lapse of time (seconds to minutes) between steps 1 and 2, with spiders moving around with a ready-made connection piece between their spinnerets before they apply a plaque onto a substrate.
Step 2 is initiated by a spreading of the spinnerets, which pulls strands of the piriform silk out of the spigots. These are applied onto the substrate as long as the glue coat is still in a fluid state (see also Wolff et al.  for details on in vivo piriform silk extrusion in Nephila). The spinnerets are then moved in sickle-like patterns over the substrate, including a backwards and forwards lateral sweep. Such a movement unit is performed in about 0.2 s, covering a range of approximately 0.4 mm in both x- and y-directions and produces an approximately 3 mm long track (figure 2). Both spinnerets move simultaneously, but in opposite directions. In between both sweeps the spinnerets are rubbing along each other in a tilted position, which presumably strengthens the bridge between connection and plaque. A set of such movement units takes places along a line of 0.6–0.8 mm in the upstream direction (4–7 movement units), and then for 0.3–0.5 mm back in the downstream direction (2–3 movement units). This crucial step shifts the upstream dragline insertion point towards the attachment disc centre. In total the step 2 sequence consists of 6–10 movement units and is performed in 1.5–2.5 s. The movement speed of the anterior lateral spinneret tips fluctuates between 2 and 10 mm s−1, on average, throughout the sequence, with a mean speed of 5–7 mm s−1. The total track length is 7–16 mm. Given that approximately 250 piriform spigots are present on an anterior lateral spinneret of an adult Nephila spider , then in total 3.5–8 m of piriform silk nano-fibres are woven into the adhesive plaque of an attachment disc.
3.3. Pull-off resistance
We measured the peak pull-off forces of spider silk anchorages spun against a polypropylene sheet. We chose a substrate with low polarity in order to investigate the dynamics of plaque detachment, which would not be observable on polar substrates like glass, to which silk anchorages adhere so strongly that they break internally [4,6]. Total delamination occurred in all cases of perpendicular (90°) pulling, and in the majority of 180° pulling tests. In pre-stressed specimens (not used for measurements), a circular shape of the delamination crack around the dragline insertion point was observed (figure 3b). However, during tests, delamination usually occurred rather suddenly and crack induction and propagation was too quick to observe with the standard video frame and force data acquisition rate. At 0° loading, half of the samples and a quarter of samples at sideways loading showed a sliding dragline failure mode. That is, the adhesion between the major ampullate and the piriform silk fails, which leads to the dragline sliding through the otherwise intact piriform silk envelope of the attachment disc connection piece, thereby producing further friction forces. At all loading angles, except for 90°, two cases of dragline failure occurred. An internal fracture at the bridge was only observed in a single case of sideways loading. Details on the different failure modes of silk anchorages can be found in Grawe et al. .
Peak pull-off forces did not significantly differ between 0°, 180° and sideways loading, but were significantly reduced at perpendicular (90°) loading. Peak forces varied between 16.6 and 101.1 mN (mean ± s.d.: 45.1 ± 24.9 mN) at 0°, 2.5–51.4 mN (20.2 ± 10.6 mN) at 90°, 14.6–89.0 mN (43.9 ± 20.3 mN) at 180°, and 26.6–91.2 (50.2 ± 18.4 mN) at sideways loading (figure 3e). Extensibility of the anchorage increased with loading angle, indicating that the structure is much stiffer at low pull-off angles than when pulled vertically or against the spinning direction (figure 3d). The detachment forces measured when pulling vertically at the downstream dragline were only half as high as at a similar loading of the upstream dragline (10.9 ± 5.7 mN). This difference was significant (Wilcoxon rank sum test, p = 0.003).
At both horizontal and vertical loading, pull-off forces were highly correlated with the distance between the dragline insertion point and the front edge of the attachment disc (front shift) (multiple regression with five coefficients; front shift Pr = 0.009 at 0°, and Pr = 0.001 at 90° loading), while the effect of all other measured parameters, such as size, width and length, was negligible (figure 4). At 180° and sideways loading, forces did not correlate with any of the measured attachment disc parameters.
To gain more insight into the effect of structural parameters in attachment discs, we built different types of models consisting of a thread attached to a glass slide by adhesive tape. In class 1 models, the thread emerged from the front edge of the tape (figure 5a), whereas in class 2 models, we simulated an expressed front shift of the dragline insertion point by a short cut (notch) in the front edge (figure 5b). In each class there were three types of models differing in their tape geometry (A: quadratic; B: elongated; C: broadened plaque). We found a significant difference in the pull-off forces and yield resistance (force per area of tape) between different model types (figure 5e,f). The basic 1A (quadratic, non-notched) models showed the lowest pull-off forces (1.76 ± 0.30 N) and yield resistance (1.22 ± 0.22 N cm−2). Among class 1 models, tape broadening significantly enhanced pull-off resistance, whereas tape elongation did not. By contrast, tape elongation led to significantly increased pull-off forces in class 2 (notched) models, while tape broadening did not. Comparing models with similar tape size in both classes showed that the front notch nearly doubled the pull-off forces in type A and B models, while there was no significant difference in type C models. The highest pull-off forces (4.15 ± 0.48 N) were obtained by 2B models (elongated + notched). Relating the pull-off forces to the amount of tape used, 2A models showed the best performance (yield resistance of 2.10 ± 0.29 N cm−2). 2B models had a significantly lower efficiency than 2A models, but a significantly higher efficiency than all class 1 models. 2C models were not significantly more efficient than 1A and 1C models. Among class 1 models, elongation leads to a significant drop in efficiency, while broadening neither enhanced nor decreased it. For the peeling energy, model 2B and 2C types gained the highest values (0.077 ± 0.018 J and 0.073 ± 0.009 J respectively), and 1A models the lowest values (0.017 ± 0.003). Peeling energy was significantly higher in all type 2 models than in type 1 models. Among type 1 models, tape elongation did not yield a significant increase in the peeling energy, but tape broadening did. Among type 2 models, tape elongation and broadening both lead to a significant increase in peeling energy.
The difference in peak forces between models is reflected by the maximal length of the delamination crack (peeling line). The initial slope of force–extension curves is nearly similar between model types, but how much the force rises highly depends on the geometry of the tape and the placement of the thread insertion point (figure 5e). In our class 1 models the delamination crack had a nearly 90° angular shape (figure 5a,c). That means that in the quadratic 1A models the peeling line reaches the maximal length (that is when the crack reaches both lateral sides of the tape), when the thread is detached by the centre of the tape (approx. 6 mm). Then the peeling line is fairly constant, which leads to a force plateau, followed by a gradual decrease, when the delamination crack reaches the rear side of the tape. In the elongated 1B tape the maximal peeling line is rather similar, and accordingly peak pull-off forces are not significantly different (though the work of adhesion is higher due to a longer peeling event). The broadened 1C models gain higher peak forces, since the peeling line can further increase towards the lateral sides. By contrast, in class 2 models (with front notch) the peeling line is diamond-shaped, with a higher propagation speed along the thread axis than to the lateral sides (figure 5b,d), which explains that, in contrast to class 1, tape length has a higher effect on peak pull-off forces than width in class 2. Furthermore, the broadened 2C models also tended to fail at the thread–tape joint before total delamination.
4.1. Outstanding pull-off resistance of silk anchorages under various angles
Our results highlight that the localization of the dragline insertion point (loading point) within the attachment disc plays a crucial role for the pull-off resistance of the silk thread anchorages. The more the loading point is shifted towards the centre of the disc, the higher the pull-off resistance in vertical and horizontal directions. This is explained by a circular crack propagation around the loading point, which leads to an overall greater length of the peeling line and lower peeling angles than if the plaque was peeled off from the front edge. Comparable, so-called ‘mushroom-like’ microstructures, where a fibre- or pillar-like structure is medially attached to a circular plaque, can be found throughout animals, plants and bacteria, and have been used as a source of inspiration for glue-less adhesive tapes (called ‘mushroom-like’, because in mushrooms similarly the hat is medially attached to the stalk) [22,23]. A prominent example is the byssus threads of mussels [24,25]. Both the attachment discs of mussel byssus threads , and bioinspired, artificially produced mushroom-like micro-pillars [12,13] exhibit a high adhesive strength and a fracture mechanics with a circular propagation of the delamination crack, when pulled off the substrate at an ideal angle. However, they are prone to variation in the pull-off angle and rapidly fail when pulled at a flat angle [24,26]. This is because of a lack of flexibility of either the stalk or the plaque, leading to high stress concentrations at bending. By contrast, the pull-off force of spider silk anchors is greatest when pulled under a flat angle. Remarkably, it does not make any difference, if the dragline is pulled along, against or perpendicularly to the spinning direction. This significant robustness might be explained by the flexible and separated joint of dragline and substrate with the attachment disc. Because the bridge of piriform fibres is radially arranged around the loading point, stress might be distributed over a comparably large area in all pulling directions. When pulled at high angles the attachment disc seems softer and more extensible than under horizontal pulling, which can be explained by the difference in flexibility of the piriform silk bridge around upstream and downstream dragline joints. This might further contribute to a reduction in stress concentration when pulling in the counter-direction.
4.2. Rejection of the staple-pin model
Our findings on the attachment discs of Nephila do not support a staple-pin configuration as proposed by previous authors [3,10,11]. Spinning patterns and structural analysis show that the dragline is not attached by an overlay of transverse piriform threads, but rather embedded in between them. In the case of a staple pin-configuration, we would expect the delamination crack to start at the front edge and the peak pull-off force to decrease with increasing pulling angle. By contrast, we found no significant difference in pull-off resistance between 0° and 180° loading. We found that in Nephila the dragline does not emerge from the front edge of the disc, but a more median position, which enormously affects the pull-off resistance and fracture mechanics. The observed structure and crack propagation show characteristics of an axisymmetric membrane . However, at vertical pulling, the force is significantly reduced, which is not predicted by the axisymmetric membrane model, but rather by the single tape and the staple-pin models. This might be explained by a higher stress concentration in the vicinity of the loading point at vertical pulling, whereas when pulling under a flat angle stress may be more distributed along the dragline connection. These assumptions imply that the piriform silk must be very extensible, as predicted from previous theoretical studies [27,28]. However, to date, no data on the tensile properties of piriform silk are available.
One character of the proposed staple-pin configuration is a discontinuous peeling behaviour of the attachment disc, as a result of a discontinuous plaque (i.e. by spaces between parallel arranged piriform threads) [3,10,11]. After the so-called contact splitting theory this would lead to an accumulation of pull-off forces due to the repeated induction of delamination cracks . This is in contrast to the peeling theory of thin films, in which the peeling line, once induced, gradually proceeds . Indeed a discontinuous peel-off was experimentally observed, as a result of the glue–fibre composite structure in the plaque . However, this is only observable at very small timescales, using a high-speed video camera with a high frame rate. At the macroscopic timescale, neither the force curves of previous experiments (fig. 4a of ) nor that of our experiments reported here (figure 3d) show the characteristics of the staple-pin case (see fig. 12 of ).
Overall, our results indicate that spider silk anchorages show a mixture of characteristics of different models. This calls for the revision of current models of silk anchorages, which would enable the effect of different structural parameters (i.e. plaque geometry and position of the loading point) to be tested further on a more continuous scale.
4.3. Significance of the spider's ‘printing’ behaviour
Our high-speed video tracking analysis revealed that the structure of the attachment disc is determined by the spinneret movement programme. The first important step is the enclosure of the dragline in piriform silk prior to the application of the attachment disc onto the substrate. This ensures that the dragline is not embedded into the plaque, which would highly reduce the flexibility of the joint. Second, when spinning the plaque, a sequence of lateral sweeps is performed upstream along the longitudinal axis, with an eventual turn and a short continuation in the backwards (downstream) direction. This back-and-forth spinning pattern shifts the upstream dragline insertion point from the front edge towards the centre of the attachment disc and reinforces the radial bridge network around it. Both our morphometric and model approaches indicate that the way the spider applies its glue is crucial to gain tough thread anchorages. A shift in the loading point towards the centre of the plaque reinforces the anchorages even more than the intake of additional material (but see exception for wide discs in our model tests). In the future, it would be interesting to investigate if these features occur in the spinning sequence of other spider species, in order to retrace the evolution of these features.
4.4. Differential material distribution and directionality of attachment discs
We found evidence that the attachment discs of N. plumipes are optimized for loading of the upstream dragline. Pull-off forces are significantly higher when pulling at the upstream dragline than when pulling at the downstream dragline. This can be explained by a more extensive bridge around the upstream dragline insertion point. At the downstream dragline insertion point, the bridge includes less and shorter suspended piriform threads. Therefore, the dragline is located much closer to the plaque in this section and the tensile load is presumably less distributed. A better performance in the upstream dragline seems reasonable, because during locomotion the load is always acting on this part of the disc. By reinforcing the downstream dragline insertion point the spider would waste material.
It has been noted previously that dragline attachment discs can be directional (dissimilar between upstream and downstream side): in wolf spiders (Lycosidae) this character is used as a mechanical cue to find conspecifics when following draglines . It would be interesting if web attachments in which both upstream and downstream threads are simultaneously loaded (personal observation both in webs found in the field and spun in captivity) are less directional due to an altered spinneret movement programme.
5. Conclusion and outlook
We have shown that the silk anchorages of N. plumipes are more resistant to loading at various angles than usual tape-like or mushroom-shaped contact elements. This is due to the specific way in which spiders ‘print’ silk into a three-dimensional attachment microstructure. We revealed that a shift in the dragline insertion point towards the plaque centre is the most efficient way to enhance pull-off resistance of the anchorage and flexibility in the thread–substrate joint. This implicates that former contact models to describe the function of silk anchorages are not sufficient and should be revised. In the future, further experimental, comparative and theoretical approaches will help to enhance our understanding of the relationships between structure and function in silk thread anchorages. Important aspects to be addressed in future studies are the effect of the glue–fibre composite structure of the plaque on its peeling strength, the mechanical properties of piriform silk, and the evolution of silk anchor performance in spiders. This is of high significance for the understanding of the evolution of the high material performance of spider silks and webs, and the bio-inspiration of novel non-mechanical cable attachments.
J.O.W. and M.E.H. conceived and designed the study. J.O.W. performed the experiments and analysed the data. J.O.W. and M.E.H. wrote the paper.
The authors declare no competing interests.
This work was funded by a Macquarie Research Fellowship to J.O.W.
We thank Lydia Wolff for help with animal care, video recordings and building of models. We are grateful to Minh Huynh, Patrick Trimby and Matthew Foley (ACMM, University of Sydney) for their great, dedicated help with the preparation and processing of 3view samples and three-dimensional analysis. We thank Joshua Madin (Macquarie University) and Aaron Harmer (Massey University, NZ) for their help and discussion on the tensile testing of silk. We further thank Walther Adendorff and Vivek Hegde (METS, Macquarie University) for building a tripod mounting plate for our micro high-speed video camera. The authors acknowledge the facilities and the scientific and technical assistance of the Australian Microscopy & Microanalysis Research Facility at the Australian Centre for Microscopy & Microanalysis at the University of Sydney.
Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9.figshare.c.3691873.
- Received September 26, 2016.
- Accepted January 31, 2017.
- © 2017 The Author(s)
Published by the Royal Society. All rights reserved.