Cobweb-weaving spiders produce different attachment discs for locomotion and prey capture

Abstract

Spiders' cobwebs ensnare both walking and flying prey. While the scaffolding silk can entangle flying insects, gumfoot silk threads pull walking prey off the ground and into the web. Therefore, scaffolding silk needs to withstand the impact of the prey, whereas gumfoot silk needs to easily detach from the substrate when contacted by prey. Here we show that spiders accomplish these divergent demands by creating attachment discs of two distinct architectures using the same pyriform silk. A 'staple-pin' architecture firmly attaches the scaffolding silk to the substrate and a previously unknown 'dendritic' architecture weakly attaches the gumfoot silk to the substrate. Gumfoot discs adhere weakly, triggering a spring-loaded trap, while the strong adhesion of scaffolding discs compels the scaffolding threads to break instead of detaching. We describe the differences in adhesion for these two architectures using tape-peeling models and design synthetic attachments that reveal important design principles for controlled adhesion.

ARTICLE

Received 20 Apr 2012 | Accepted 30 Aug 2012 | Published 2 Oct 2012 DOI: 10.1038/ncomms2099

Spiders cobwebs ensnare both walking and ying prey. While the scaffolding silk can entangle ying insects, gumfoot silk threads pull walking prey off the ground and into the web. Therefore, scaffolding silk needs to withstand the impact of the prey, whereas gumfoot silk needs to easily detach from the substrate when contacted by prey. Here we show that spiders accomplish these divergent demands by creating attachment discs of two distinct architectures using the same pyriform silk. A staple-pin architecture rmly attaches the scaffolding silk to the substrate and a previously unknown dendritic architecture weakly attaches the gumfoot silk to the substrate. Gumfoot discs adhere weakly, triggering a spring-loaded trap, while the strong adhesion of scaffolding discs compels the scaffolding threads to break instead of detaching. We describe the differences in adhesion for these two architectures using tape-peeling models and design synthetic attachments that reveal important design principles for controlled adhesion.

Cobweb-weaving spiders produce different attachment discs for locomotion and prey capture

Vasav Sahni1, Jared Harris1, Todd A. Blackledge2 & Ali Dhinojwala1

1 Department of Polymer Science, Integrated Bioscience Program, The University of Akron, Akron, Ohio 44325-3909, USA. 2 Department of Biology, Integrated Bioscience Program, The University of Akron, Akron, Ohio 44325-3908, USA. Correspondence and requests for materials should be addressed to A.D. (email: [email protected]).

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms2099

Spiders attach their major ampullate (MA) silk dragline bres to surfaces using attachment discs spun from pyriform silk1. These attachments allow spiders to move safely from place to

place while secured to a dragline and to attach their webs to a variety of surfaces. Cobweb-weaving spiders spin three-dimensional webs that target both walking and ying prey2. While the scaolding silk acts as a catching net for ying prey, the gumfoot silk threads pull walking prey o the ground and into the web2,3. This requires that, upon agitation by the momentum of the incoming prey, the scaffolding silk remains rmly attached to the substrate while the gum-foot silk threads easily detach from the surface.

Here, we show that cobweb-weaving spiders accomplish these divergent functions by creating attachment discs of two distinct architectures using the same pyriform silk. A previously known staple-pin architecture is used to rmly attach the scaolding silk to the ceiling and a newly described dendritic architecture is used to weakly attach the gumfoot silk to the substrate. Interestingly, orb-weavers (evolutionary ascendants of cobweb weavers) also use the staple-pin architecture to rmly attach their webs to dierent substrates. Peeling tests conducted on discs conrm that gumfoot dendritic discs adhere much more weakly than the scaolding discs. We explain the dierence in adhesion strengths of the two architectures by simulating the two attachment discs and using Kendalls classical tape-peeling model4. Small peeling angles combined with signicant elastic deformation of the pyriform bres in the scaolding disc explain its much higher adhesion. Gumfoot discs, on the other hand, peel at much higher angles, explaining their relatively lower adhesion. These results provide unique ideas and principles for controlling structure in the design of novel tunable adhesives for various biomedical and material science applications.

ResultsDierent disc architectures. The cobweb-weaving spider Achaearanea tepidariorum spins the same type of attachment disc used to secure scaolding silk when it is made to walk on a glass slide and secure its dragline. Multiple pyriform gland spigots surround the MA gland spigot5 such that the dragline silk produced by spiders is attached to surfaces using multiple nanobres, composed of a core and a shell6. To sample the gumfoot disc, the spiders cage is layered with glass slides and samples are collected aer the cobwebs are spun. Figure 1 shows SEM micrographs of the scaolding disc and the gumfoot disc. The dierence in the architectures of the discs is clear. The scaolding disc resembles a staple-pin architecture, while the gumfoot disc shows a dendritic architecture. Each pyriform bre is coated with a uid that likely facilitates adhesion to both smooth and rough surfaces (Supplementary Fig. S1). The number of pyriform bres in staple-pin discs is much higher than in dendritic discs (1,550 100 versus 188 20), which translates into a greater number of load-bearing attachment points in staple-pin discs. (There are observable intra-species dierences in the size of the joints and the density of pyriform bres within each joint. To avoid these dierences, the numbers were calculated using gumfoot and scaolding joints spun by the same spider). Moreover, in gumfoot discs, pyriform bres are suspended in the air as they converge with gumfoot threads and are bound to the surface only at the periphery of the discs (Supplementary Fig. S2). This results in only the thinnest pyriform bres contacting the surface, as pyriform bres bifurcate successively, becoming thinner towards the periphery of the discs (Supplementary Fig. S2). In contrast, in scaolding discs, most of the length of the pyriform bres, which are thicker than in gumfoot discs, is in contact with the surface.

Attachment discs are formed without the involvement of legs7, so dierences result solely from movements of the abdomen and spinnerets. The anterior spinnerets are brought into direct contact with the substrate and are rubbed against it while a secretion is exuded through the pyriform gland spigots7. The dierent architectures

of the attachment discs therefore imply dierent movement patterns of the anterior spinnerets against the substrate. Based on high-resolution images of the discs, and the hairpin bends as shown in Supplementary Fig. S1, we postulate that the spinnerets sweep back and forth across an MA bre several times to create a scaolding disc, whereas the successive splitting of the pyriform bres in the gumfoot disc (Supplementary Fig. S2) indicates that the anterior spinneret rubs against the substrate once, going inwards from the periphery of the gumfoot disc. This also implies that the pyriform bres are not completely solidied when secreted and that they coalesce to form thicker bres, going inwards from the periphery of a gumfoot disc. Incidentally, cribellate spiders also use nanobres for adhesion8. However, unlike in pyriform bres, these nanobres are drawn from the cribellum of the spiders by a setal comb on their legs8. Moreover, the cribellar bres solidify upon drawing, as indicated by their uniform diameter, unlike pyriform bres.

Eect of architecture on adhesion. To determine the eect of architecture on attachment strength, the discs were peeled from a clean glass surface at a controlled rate, by pulling on the dragline silk/gumfoot silk thread perpendicular to the plane of the respective discs. Interestingly, scaolding discs (for both cob- and orb-webs) could not be induced to peel from glass by pulling the underlying dragline silk vertically. Instead, the dragline silk thread itself always failed before peeling of the scaolding discs was completed. This demonstrates that the adhesion force of the scaolding discs is greater than the breaking force of MA silk (dragline silk), which is the strongest (highest breaking force) silk used in the web. Spiders were hence made to spin scaolding discs on glass slides laid across a much thicker nylon thread (30 m diameter) (Fig. 2a), to test the discs maximum adhesive strength. Using the nylon thread, a scaffolding disc was peeled as shown in the Supplementary Movie 1. The pyriform threads deform and peel sequentially, starting from one end of the disc, at low peeling angles leading to eventual fracture. Figure 2b shows the post-peeling SEM picture of a scaolding disc spun on a nylon thread. An SEM micrograph of broken pyriform bre from the scaolding disc is shown in Fig. 2c.

To quantify the adhesion of gumfoot discs, the discs were peeled by pulling on the lower region of the gumfoot thread, which is composed of four MA bres2. To sample the gumfoot discs for adhesion measurements, the spider cages were layered with glass slides

Figure 1 | Attachment disc morphology matches different functions. Figure shows SEM images of scaffolding discs (a) and gumfoot discs (b) spun by the cobweb-weaving spider A. tepidariorum. Insets show optical microscope images of the respective discs. The black arrows point towards dragline silk (MA silk) in (a) and its inset, while the white arrows point towards pyriform bres, arranged in a staple-pin architecture, attaching the dragline silk to the surface. Black arrows point towards gumfoot thread (MA silk covered with aggregate glue) in (b) and its inset, while the white arrows point towards pyriform bres, arranged in a dendritic architecture, attaching the gumfoot thread to the surface. Scale bars in both the gures are 100 m.

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms2099

ARTICLE

Figure 2 | Scaffolding disc adhesion testing. (a) Shows the cobweb-weaving spider A. tepidariorum spinning a scaffolding disc on a nylon thread. The inset shows an optical microscope image of a scaffolding disc spun on a nylon thread (30 m diameter). The black arrows point towards the nylon thread in both (a) and its inset. (b) Shows a scaffolding disc post peeling. The black arrows in (b) and its inset point towards the peeling zone of the scaffolding disc spun on a nylon thread. Pyriform bres peel a certain distance followed by breaking, as is shown in the inset. The scale bar in (b) and its inset is100 m and 25 m, respectively. (c) Shows a broken pyriform bre from a scaffolding disc that has been peeled. The scale bar is 1 m.

Figure 3 | Gumfoot disc adhesion testing. (a) Shows a schematic of a typical cobweb. The black arrow points towards the ag used to collect individual gumfoot threads with the attached pyriform attachment disc (light green). (b) Shows a gumfoot disc post peeling. The picture shows a bre-dense two-lobed structure, Scale bar is 50 m. (c) and (d) show

SEM images of a gumfoot disc post fracture. Black arrows indicate broken pyriform bres. The scale bars are 100 m and 1 m, respectively.

such that aer the webs were spun, a ag, like that shown in Fig. 3a, was used to collect the gumfoot silk thread with the attachment disc undisturbed. Figure 3b,c show an optical image and an SEM micro-graph of a gumfoot disc, post peeling. The fracture image shows two bre-dense lobes in the gumfoot discs. Figure 3d shows an SEM micrograph of broken pyriform bres from the gumfoot disc. The diameters of the broken end of the pyriform bres in both scaolding and gumfoot discs are similar. Supplementary Movie 2 shows the behaviour of pyriform bres during peeling of a gumfoot disc. Depending on the angle at which the gumfoot thread is pulled, pyri-form bres in the two lobes behave dierently. The peeling angles of gumfoot bres are much higher than the peeling angles in scaffolding discs. Also, unlike in scaolding discs where pyriform bres deform and peel sequentially from one side of the disc, in gumfoot discs, all the pyriform bres respond simultaneously.

Figure 4a shows the force-extension behaviour obtained from peeling scaolding and gumfoot discs spun on clean glass surfaces. The maximum force required to peel the discs is considered the adhesion force or the pull-o force. Figure 4b compares the

pull-o forces of the discs on glass. Scaolding discs bind by an order of magnitude stronger than in gumfoot discs. The total work done in peeling the discs can be calculated by integrating the area from the force-displacement results. The total work done involves work incorporated in stretching the thread (nylon in the case of scaolding discs) and the energy expended in peeling the discs. The strain energy stored in nylon threads and gumfoot silk threads was independently determined from tensile stress-strain measurements and was subtracted from the total work done (calculated by integrating the area under the force-extension curve obtained from the disc peeling measurements) to calculate the work done owing to peeling (as shown in Fig. 4c) for scaolding discs and gumfoot discs, respectively. Scaolding discs adhere much more strongly than gumfoot discs, as anticipated. The work done owing to peeling comprises of stretching, separating and breaking of pyriform bres, as shown by high-speed videos taken during peeling, and should thus, in theory, be much higher than the thermodynamic work of adhesion of pyriform bres.

The dierences in pull-o forces can be attributed to dierences in the number, diameter, chemistry (intrinsic adhesion strength) and the peeling angles of the pyriform bres between these two discs. To test the eect of number of bres, we controlled the length of the scaolding disc as shown in Fig. 5ag and measured the pull-o forces as a function of number of bres (Fig. 5h). The pull-o force initially increases with the number of pyriform bres and then plateaus for higher numbers of bres. The plateau in pull-o forces is expected because in the staple-pin architecture only a few bres resist the pull-o forces at a time since the peeling starts from one end of the disc and the deformation zone is smaller than the scaolding disc. Interestingly, for the same number of bres in the scaolding and gumfoot discs, the pull-o force of the scaffolding disc is almost six times that of the gumfoot disc (Fig. 5h). Moreover, scaolding discs with fewer than half the number of bres in a gumfoot disc still generate a pull-o force three times as big. Therefore, the dierences in the pull-o forces between the scaolding and gumfoot discs cannot be solely caused by the differences in the number of pyriform threads. The SEM images of the broken bres in both the discs (Figs 2c and 3d) suggest that the eect also cannot be explained by dierences in diameters of the bres. Instead, scaolding discs generate high pull-o forces owing to their architectures.

To test the possibility that the pyriform bres in the scaolding and gumfoot discs could be chemically dierent, we cite the work of Moon et al.9 that shows that female A. tepidariorum spiders possess 90100 pairs (total 180200) of pyriform glands that feed into discrete pyriform spigots on the anterior spinnerets. The total number of pyriform spigots is consistent with the total number of bres measured by us in the gumfoot discs (188 20). Also, the

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms2099

10

5

4 3 2 1 0

10

Force (mN)

Force (mN)

Energy (J)

0.30

5

0.0

0.6

0

0.0 0.5 1.0

Gumfoot

disc

Scaffold

disc

Dragline

silk

Gumfoot

disc

Scaffold

disc

Extension (mm)

Figure 4 | Adhesion measurements. (a) Shows the force-extension curves obtained by peeling the scaffolding discs (blue) and gumfoot discs (green) spun on a clean glass surface. Data are plotted as mean s.d. from three measurements each. Inset emphasizes the initial region of the force-extension curve of the gumfoot disc. (b) Shows the force required to separate a gumfoot disc and a scaffolding disc from a clean glass surface. Gumfoot disc adhesion is an order weaker than scaffolding disc adhesion. The force required to break a dragline silk (MA silk) thread is also plotted. A dragline silk thread breaks when it is pulled to peel a scaffolding disc (hence thicker nylon threads were used to peel scaffolding discs). Gumfoot discs were peeled using the gumfoot silk threads (four intertwined MA threads). Data are plotted as mean s.d. from ve measurements each. In theory, however, gumfoot discs could be peeled using just one MA thread as the breaking force of MA thread is almost twice the pull-off force of gumfoot discs with glass.

This indicates that scaffolding disc adhesion is much stronger than gumfoot disc adhesion. (c) Shows the energy of adhesion of a gumfoot disc and a scaffolding disc. Gumfoot discs adhere an order weaker than scaffolding discs adhere. The energy of adhesion is determined by subtracting the strain energy contribution of nylon threads (while peeling scaffolding discs) and gumfoot threads (while peeling gumfoot discs) from the total work done in peeling the discs. Data are plotted as mean s.d, from ve measurements each.

300 m 500 m 750 m 1,000 m

2,500 m

1,050

20

10

Force (mN)

100

210

320

500

690

870

1,500 m

2,000 m

a b c d e f g

Scaffolding joint no.

Figure 5 | Effect of number of bres on scaffolding disc adhesion. (ag) Show scaffolding discs spun on nylon thread containing 100, 210, 320, 500, 690, 870 and 1,050 pyriform bres, respectively. (h) Shows the adhesion forces of these discs. The green line shows the breaking force of an MA thread, which, as shown in Fig. 4b, is higher than the adhesion force of a gumfoot disc.

number of pyriform bres in scaolding discs is 1,550 100 (which, we hypothesize, result from several strokes of spiders abdomen across an MA thread). This suggests that all the pyriform glands are employed for creating gumfoot discs, and hence the scaolding and gumfoot discs are spun by the same spinnerets and thus have the same chemistry. This also implies that the dierence in the pull-o forces of the gumfoot and scaolding discs is not due to difference in chemistry.

The remaining hypothesis we put forth to explain the dierences in the pull-o forces is the dierences in the peeling angles (as shown in Supplementary Movies 1 and 2). The eects of peeling adhesive tapes as a function of peeling angles was studied by Kendall4,10 and he showed that peeling force at low peeling angles depend on the modulus of the bres and physical factors such as width and thickness of the tape. At higher peeling angles, the pull-o forces were considerably lower and they only depended on the adhesion energy (referred to here as fracture energy, G) and the width of the tape. Therefore, we anticipate that the peeling angles could explain the dierences in pull-o forces between the two dierent architectures of attachment discs used by spiders.

Synthetic joints inspired by pyriform attachment discs. We design two models using adhesive tapes to illustrate the dierences in the architecture used in these two attachment discs. These macroscopic models also allow us to understand the functional implications of materials used by spiders and test conditions that we could not control in these natural materials. A scaold disc is modelled by attaching a bundle of nylon bres on a clean glass surface using six strips of a stretchable tape (3M, MMM8884) perpendicular to the long axis of the strand (Fig. 6a). Peeling of the tape strips is induced by pulling on the peeling end of the nylon strand perpendicular to the plane of the tapes, at a xed rate (inset, Fig. 6a). The rate of pulling is kept xed throughout the study because the fracture energy G, and hence the peeling force F, depend on the rate.

The gumfoot discs are simulated by placing six strips of the same tape in a radial pattern, such that all the strips are securely attached to a strand of nylon threads at the centre of the circular disc (Fig. 6b). Here, peeling of the tapes is induced by pulling the nylon strand perpendicular to the plane of the tapes, at a xed rate. Figure 6c,d shows the adhesion forces of both geometries. For the same surface area of the tape, the staple-pin geometry (mimicking

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms2099

ARTICLE

Force

Nylon

Force

Tape strips

Peeling scenario

Front view

Nylon

Isp

Top view

Peeling end Fixed end

Tape strips

Irp

Wrp

Wsp

100

100

Load (N)

50

0

0 150

50 100 150

Extension (mm) Extension (mm)

10

5

0

Load (N)

Extension (mm)

50

Load (N)

0

0 50 100 150

0

Figure 6 | Modelling attachment discs. (a) Shows a schematic of the simulated scaffolding disc. The length lsp is 100 mm while the width wsp is such that 6*wsp equals 36 mm. The inset shows how the peeling measurements are conducted. The nylon strand is pulled on perpendicular to the plane of the tapes while the load-extension behaviour is registered. (b) Shows a schematic of the simulated gumfoot disc. The length lrp and the width wrp of the tape strips are equal to lsp and wsp, respectively. All six strips are rmly secured to the strand of nylon threads at the centre of the circle. The insets on the left and right show the top and front views, respectively, of the modelled tapes. (c) Shows the load-extension results obtained from the adhesion measurements of the simulated scaffolding discs. Circles denote the case with length = lsp and width wsp, squares with length = lsp/2 and width = wsp, and triangles with length = 6*wsp and width = lsp. The inset shows a plot comparing the adhesion forces of simulated scaffolding silk with the simulated gumfoot disc, both using the same dimensions of the strip: length = lsp, width = wsp. (d) Shows the load-extension results obtained from the adhesion measurements of the simulated gumfoot discs. Circles denote the case with length lrp and width wrp. Squares denote length ~lrp/2 and width wrp, inverted triangles denote length lrp*3/4 and width wrp/2, and upright triangles denote width wrp and length ~3 mm. For (c) and (d), data are plotted as mean s.d. from ve measurements each.

scaolding discs) requires a much larger force to detach than the dendritic geometry (mimicking gumfoot discs) (blue circles versus green circles and blue squares versus green squares, Fig. 6c,d). The energy expended (area under the load-extension curve) is also higher for the staple-pin geometry.

Peeling the modelled scaolding disc involves deformation of the tape strips and their concomitant detachment from the glass surface. The detachment of the tape from the glass surface follows a near-zero degree peeling behaviour, with a small gradual increase in peeling angle during further peeling. Consistent with Kendalls observations with tapes, the thickness of the tape strips, d, aects the fracture energy of the simulated scaolding disc (Supplementary Fig. S3a, solid blue squares).

Detachment of simulated gumfoot discs, however, initiates as a near-ninety degree tape peeling, the peel angle reducing only slightly during further peeling. The testing method for both the modelled discs involved slight changes in the peeling angle during the course of peeling. This was done to simulate the real-life scenario and also the method used to test the natural attachment discs). In the radial peeling of six-tape strips of width wrp, each was observed to be equivalent to peeling of a single tape strip of width 6*wrp. Each curve shows a fracture-initiation region, followed by a fracture-propagation region. The propagation region scales with the width of the tape: a tape of width < wrp shows a propagation region of smaller magnitude (inverted triangles,

Fig. 6d), while a shorter tape of the same width has a propagation

region of the same magnitude (squares, Fig. 6d). Successively reducing the length of the tape, keeping width constant, results in the extreme case where just the initiation region is observed: fracture propagation is absent (green upright triangles, Fig. 6d). Also, varying the thickness of the tape has no eect on radial peeling, unlike in the staple-pin case, consistent with Kendalls observations (green solid squares, Supplementary Fig. S3b). The slight increase in the propagation regions of all the curves is explained by the small reduction in peeling angle during peeling.

To accomplish peeling of the modelled discs, a nylon strand of breaking force 200 N was used. In reality, however, MA silk thread breaks before peeling of the scaolding discs. This was simulated by using a thinner strand of nylon threads (breaking strength ~60 N). Using this strand, the simulated gumfoot disc still peels completely, but an attempt to peel a simulated scaolding disc instead causes fracture of the nylon strand before the disc detaches. These simulation results clearly demonstrate the design principles employed by spiders to make strong versus weak discs using the same material.

Discussion

Why do spiders create two dierent architectures for attachment discs? The dierence in pull-o forces for these two divergent architectures has major implications on locomotion and prey capture by spiders. When an insect encounters an orb-web (which is also attached to dierent surfaces using pyriform discs spun in the staple-pin architecture), the hysteretic behaviour of the MA bres,

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms2099

as they stretch and relax, accounts for majority of the energy absorbed by the web11. Also, when an insect becomes completely entangled, the weight of the insect is supported by the MA bres, transferring the load to the attachment points of the web with the surface. Stiness, strength, toughness and viscoelasticity of the MA silk are key to the functioning of both the orb and cobwebs. However, the excellent properties of MA silk can only be utilized for capturing prey if the threads are rmly xed to surfaces. Spiders accomplish this by creating attachment discs out of pyriform silk that have staple-pin architectures. The observation that the pull-o forces of these discs are higher than the breaking strength of MA silk has major implications for optimized web engineering: the mass and velocity of the incoming prey are limited solely by the breaking strength of the strong MA silk and not by its attachment to surfaces.

In contrast, the gumfoot discs are attached very weakly to the substrate, which enables the gumfoot threads to detach easily. Energy stored in the scaolding silk suspends the insect helplessly in the air. A gumfoot thread is composed of four MA silk threads at its base2. During adhesion measurements, gumfoot discs could be easily peeled from a surface by vertically pulling the gumfoot thread attached to the discs. The pull-o force of a gumfoot disc is ~50% the breaking force of one MA silk thread (~1/8th the breaking strength of a gumfoot thread, as there are four MA threads in the lower region of a gumfoot thread).

A macroscopic tape model (Fig. 6) illustrates not only the importance of peeling angles in controlling pull-o forces but also the requirement for the gumfoot discs to act as an efficient trap: discs should release with ease only aer being agitated by an insect. Gum-foot threads are spring-loaded traps always held in tension. These traps are triggered when agitated by a walking insect. An ideal trap acts spontaneously upon triggering by prey, but not upon minor perturbations of the web owing to wind or even the spider moving through its own web. We postulate that the initial over-shoot region acts as a trigger, going past the region in which the propagation region slopes downwards (Supplementary Fig. S3b, green diamonds). Once the agitation force caused by the insect surpasses the initial over shoot, the gumfoot discs rapidly release with ease, suspending the insect helplessly in the air for the spider to subdue. The force prole during peeling of a staple-pin disc instead slopes upwards continuously, which is ideal for developing a strong disc that resists catastrophic failure but it does not full the trigger requirement of an ideal spring-loaded trap. The force-extension proles demonstrated by the tape models mimic those measured in actual scaolding and gumfoot discs (Fig. 4a).

The Kendall model for elastic adhesive tapes of width w, the peeling force for large peeling angle () is equal to w*G/(1 cos()), where G is the fracture energy. Aer the initial zone, the peeling force should be independent of the length of the tape (or length of pyriform bres in dendritic discs). In addition, the force should be proportional to the width (or the diameter of the thread). Therefore, in the tape model, varying the amount of material in a gumfoot disc does not change the maximum force required to peel the disc (Fig. 6d, green circles and green squares). In the actual gumfoot discs, bres in contact with the surface are mostly at the periphery of the disc and become successively thinner going towards the periphery (Supplementary Fig. S2), which is most accurately modelled by using tapered strips of tape represented by green diamonds in Supplementary Fig. S3b. As the peeling force is a function of width, the progressively thinner diameter insures that there will be a quick release of the thread.

Another interesting design parameter used in spinning the staple-pin architecture is the width of the attachment disc. The lengths of the pyriform bres are 250 times the diameter of the secured MA silk thread (~1,000 m versus 4 m), which seems to suggest that a lot of expensive material used to spin pyriform bres is wasted,

and that pyriform secretions could either have been saved or used to cover a longer length of MA silk thread with shorter pyriform bres. We tested these conditions using the tape model to illustrate the importance of the width of the attachment disc (Fig. 6c). Keeping the width of each strip constant (wsp), if the length of each strip is reduced to lsp/2 (so that the volume of tape is essentially halved), the adhesion force is reduced by almost 50% (blue squares versus blue circles, Fig. 6c). In contrast, using the same volume of the material to produce wider strips of width lsp/6, where length is reduced to 6*wsp (blue triangles, Fig. 6c) reveals an interesting behaviour: even though the length of the nylon strand covered by the tape to immobilize it on the glass surface is now much longer, that is, lsp

compared with the previous value of 6*wsp, the force and energy required to detach is smaller by 50%. Interestingly, this value is even lower than when 50% of the material (length of each strip = lsp/2)

was used (blue triangles versus blue squares, Fig. 6c). Thus, for a given volume of material more adhesion is generated by covering a shorter length of dragline with long and thin pyriform bers.

During staple-pin peeling, the tape strips have to be deformed for the fracture to initiate and propagate. In the rst case (width wsp and length lsp), peeling requires higher force to deform and detach the longer strips of length lsp. However, in the third case (width lsp/6 and length 6*wsp), much smaller force is required because the length to be deformed and detached is only 6*wsp. This might explain why spiders use long bres (higher disc width) to make the scaolding disc.

The pull-o forces in these studies were measured on a clean glass surface, which is a simplied perspective considering that the spiders attach to rough surfaces, including wood, leaves and even concrete. Based on our measurements (Fig. 4) and our eld observations, it appears that scaolding disc always adheres more strongly than the force required to break the MA silk thread. However, the gumfoot disc adhesion depends on the surface, for example, they fracture cohesively on a clean glass surface (green curve, Figs 4a and 3a,b), but fracture adhesively on a hydrophobic surface. The successive bifurcation of glue threads (Supplementary Fig. S2) would either facilitate breaking or precipitate detachment through peeling depending on the surface.

In summary, the cobweb-weaving spiders exploited the dierence in architecture of the attachment discs, rather than the chemistry of the silk, to achieve either a very strong attachment of the discs to hold the dragline permanently to substrates or controlled attachment points that release rapidly when contacted by prey. This is further demonstrated by studying the attachment discs of the evolutionary ascendants of cobweb-weaving spiders, the orb-weaving spiders. The orb-weavers also use staple-pin architecture to rmly attach their orb-webs to dierent surfaces, similar to how cobweb weavers attach their scaolding sheet to a variety of surfaces (Supplementary Fig. S4) (Supplementary Fig. S4 shows a disc spun by the furrowed orb-weaver Larinioides cornutus. Their distant relatives, Nephila clavipes, also spin discs with the same structure, indicating the prevalence of this structure among orb-weavers. Similarly, black widow spiders, Latrodectus hesperus, also spin both staple-pin and dendritic discs, indicating the prevalence of the architecture philosophy among cobweb weavers.). The chemical composition of the proteins in the pyriform silk for orb-weavers diers from the cobweb weavers12. However, just like the scaolding discs of cobweb-weavers, the adhesion of orb-weavers staple pin discs is higher than the breaking force of their MA thread, emphasizing the importance of the architecture rather than the dierences in the chemical composition of the pyriform silk. Our models show that this variation in pull-o forces and behaviour does not require specialization of the chemistry in the glues, but is clearly controlled by architecture. It is now recognized that the spinning process is very important in achieving high tensile strength, super contractibility and other excellent properties of silk1315.

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms2099

ARTICLE

Here, we emphasize how spiders also use the architecture of attachments to control pull-o forces and detachment behaviour. These results provide design principles of using synthetic adhesives to fabricate tunable structural adhesives by using the known chemistry available today.

Methods

Disc sample preparation. Scaolding disc samples were produced by making the spider (A. tepidariorum, L. cornutus, L. hesperus or N. clavipes) walk on a clean glass slide layered with a single nylon thread (diameter 30 m) such that the spider spun the scaolding disc symmetrically to the nylon bre (Fig. 2b). For making discs of a desired length, a blade was used to remove the extra pyriform bres from that disc. Gumfoot disc samples were prepared by layering a cobweb-weavers cage with glass slides and using a ag (Fig. 3a) to collect the gumfoot samples with the gumfoot disc. To acquire an SEM image (JEOL) of the discs, spiders were made to walk and spin a disc on an SEM stub, and for the gumfoot disc case, the cages were layered with SEM stubs.

Adhesion testing on discs. Nylon threads used for the scaolding disc were xed on the top clamp of a NanoBionix (Agilent Tech.) while the glass slide with the scaolding disc was xed on the bottom clamp. The nylon thread was pulled perpendicular to the plane of the disc at 2 mm s 1 while the load-extension behaviour was recorded. To test the gumfoot disc, gumfoot threads were xed on the top clamp, and the glass slide with the gumfoot disc was xed on the bottom clamp. The gumfoot thread was pulled perpendicular to the plane of the disc at 2 mm s 1 while the load-extension behaviour was recorded.

Adhesion testing on simulated discs. Strips of a stretchable tape (3 M, MMM8884) were used to attach a strand of nylon threads to a clean glass plate in the manner described in the text and shown in Fig. 6a,b. For the adhesion testing, the glass plate was xed on the bottom clamp of an Instron, and the nylon strand was xed on the top clamp. The nylon strand was pulled perpendicular to the plane of the tape strips at a rate of 50 mm min 1 while the load-extension behaviour was recorded.

References

1. Andersen, S. O. Amino acid composition of spider silks. Comp. Biochem. Physiol. 35, 705711 (1970).

2. Blackledge, T. A., Summers, A. P. & Hayashi, C. Y. Gumfooted lines in black widow cobwebs and the mechanical properties of spider capture silk. Zoology 108, 4146 (2005).

3. Argintean, S., Chen, J., Kim, M. & Moore, A. M. F. Resilient silk captures prey in black widow cobwebs. Appl. Phys. A 82, 235241 (2006).

4. Kendall, K. Thin-lm peeling-the elastic term. J. Phys. D 8, 14491452 (1975).5. Coddington, J. A. Spinneret silk spigot morphology: evidence for the monophyly of orbweaving spiders, Cyrtophorinae (Araneidae), and the group Theridiidae plus Nesticidae. J. Arachnol. 17, 7195 (1989).

6. Kovoor, J. & Zylberberg, L. Fine structural aspects of silk secretion in a spider (Araneus diadematus). I. Elaboration in the pyriform glands. Tissue Cell 12, 547556 (1980).

7. Hajer, J. & ehkov, D. Spinning activity of the spider Trogloneta granulum (Araneae, Mysmenidae): web, cocoon, cocoon handling behaviour, draglines and attachment discs. Zoology 106, 223231 (2003).

8. Hawthorn, A. C. & Opell, B. D. Van der Waals and hygroscopic forces of adhesion generated by spider capture threads. J. Exp. Biol. 206, 39053911 (2003).

9. Moon, M. J. & An, J. S. Microstructure of the silk apparatus of the comb-footed spider, Achaearanea tepidariorum (Araneae: Theridiidae). Entomol. Res. 36, 8 (2006).

10. Gent, A. N. & Hamed, G. R. Peel mechanics of adhesive joints. Polym. Eng. Sci. 17, 462466 (1977).

11. Lin, L. H., Edmonds, D. T. & Vollrath, F. Structural engineering of an orb-spiders web. Nature 373, 146148 (1995).

12. Geurts, P. et al. Synthetic spider silk bers spun from pyriform spidroin 2, a glue silk protein discovered in orb-weaving spider attachment discs. Biomacromolecules 11, 34953503 (2010).

13. Lewis, R. V. Spider silk: ancient ideas for new biomaterials. ChemInform 37, 37623774 (2006).

14. Vollrath, F. & Knight, D. P. Liquid crystalline spinning of spider silk. Nature 410, 541548 (2001).

15. Hayashi, C. Y. & Lewis, R. V. Molecular architecture and evolution of a modular spider silk protein gene. Science 287, 14771479 (2000).

Acknowledgements

We express sincere gratitude to Professor Gary Hamed and Professor Alan Gent for stimulating discussions. This work is supported by National Science Foundation.

Author contributions

V.S. and J.H. performed the experiments. V.S. and A.D. analysed the data. V.S., T.A.B. and A.D. wrote the paper. V.S., A.D. and T.A.B. discussed the results and commented on the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/ naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions/

How to cite this article: Sahni, V. et al. Cobweb-weaving spiders produce dierent attachment discs for locomotion and prey capture. Nat. Commun. 3:1106 doi: 10.1038/ncomms2099 (2012).

NATURE COMMUNICATIONS | 3:1106 | DOI: 10.1038/ncomms2099 | www.nature.com/naturecommunications 2012 Macmillan Publishers Limited. All rights reserved.