Environmental response and adaptation of glycoprotein glue within the droplets of viscous prey capture threads from araneoid spider orb-webs

Many biological materials respond in adaptive ways to environmental humidity. For example, keratin in gecko toe pads softens as humidity rises to increase toe adhesion (Puthoff et al., 2010) and wheat awns shrink and swell cyclically with humidity changes, burying seeds in the soil (Elbaum et al., 2007). High humidity during the late evening or early morning hours when most orb-weaving spiders construct their webs is crucial for configuring a web's prey capture threads (Fig. 1) (Peters, 1986; Vollrath, 1992). These sticky threads form the web's capture spiral, which retains insects that strike the web, giving a spider more time to locate and subdue these prey before they escape from the web (Blackledge and Zevenbergen, 2006; Chacón and Eberhard, 1980). The material that forms these viscous threads issues from three spinning spigots on each of the spider's paired posterior spinnerets: a flagelliform gland that spins a supporting protein axial fiber and two aggregate glands that coat this fiber with an aqueous solution (Coddington, 1989). When these coated fibers first merge to form a thread, their aqueous covering forms a continuous cylinder. Salts and as many as 12 low molecular weight compounds, such as GABamide, glycine and choline (Edmonds and Vollrath, 1992; Fisher and Brander, 1960; Higgins et al., 2001; Townley et al., 1991; Tillinghast and Christenson, 1984; Townley et al., 2006; Vollrath et al., 1990) make this coating hygroscopic, causing it to rapidly absorb atmospheric moisture and be reconfigured by Rayleigh instability into droplets (Edmonds and Vollrath, 1992). Within each droplet, ASG1 and ASG2 glycoproteins condense into a viscoelastic core that is both adhesive and extensible (Choresh et al., 2009; Sahni et al., 2010; Tillinghast et al., 1993; Townley et al., 2006), properties that combine to generate an effective adhesive delivery system (Opell and Hendricks, 2007; Opell and Hendricks, 2009). The ASG1 protein contains a non-repetitive region rich in charged amino acids and is considered to be hydrophilic (Choresh et al., 2009).

This hygroscopic legacy continues to impact a viscous thread over the course of a day. Both the volume and extensibility of Larinoides cornutus thread droplets are directly related to relative humidity (RH) (Opell et al., 2011a). The extension bridge mechanism by which viscous threads sum the adhesion of multiple droplets relies on droplet extensibility (Opell and Hendricks, 2007; Opell and Hendricks, 2009). Therefore, these humidity changes have the potential to alter viscous thread adhesion. Thread adhesion can be gauged in two ways: the force required to pull a thread from a surface (e.g. Agnarsson and Blackledge, 2009; Opell and Hendricks, 2009) and the energy required to do so (Sahni et al., 2011). The latter value gauges a thread's ability to dissipate the force of an insect struggling to free itself from a web. At rapid extension velocities, an increase in RH from 15% to 40% resulted in a 360% increase in the energy required to pull L. cornutus viscous threads from a surface. However, this energy decreased by 67% at 90% RH, indicating that water uptake at intermediate humidity optimized glycoprotein performance, whereas at higher humidity excess water over-lubricated the glycoprotein, causing it to more easily pull from a surface (Sahni et al., 2011).

As a viscous thread's response to humidity has the potential to significantly affect thread performance, we hypothesized that selection has optimized thread hygroscopicity to the humidity regime of a species' habitat. To test this hypothesis and to more fully characterize the humidity responses of viscous thread droplets, we examined the performance of threads spun by two species: Argiope aurantia Lucas 1833, a species commonly found in exposed, weedy vegetation (Carrel, 2008; Enders, 1977; Young and Edwards, 1990), and Neoscona crucifera (Lucas 1839), a species that builds its webs in forests and along forest edges (Adams, 2000; Edwards, 1984). Neoscona crucifera can both be nocturnal and continue to use its webs during daylight hours. We hypothesized that the viscous droplets of A. aurantia are more hygroscopic than those of N. crucifera, permitting them to remain optimally hydrated during the drier times of the day. We tested this hypothesis by using an expanded number of experimental humidities (20%, 37%, 55%, 72% and 90% RH) that permitted us to more fully characterize droplet response (Fig. 2) and a new technique that permitted us to visualize (Fig. 3) and quantify a droplet's glycoprotein core and to compute glycoprotein volume-specific indices of droplet extension (Fig. 4).

We collected viscous threads from orb-webs constructed by 14 adult females of each species between 06:00 h and 08:00 h from sites near Blacksburg, Montgomery County, VA, USA, from 29 August to 30 September 2011. Argiope aurantia threads were collected from webs constructed in weedy vegetation along the margins of a meadow (Heritage Park) and N. crucifera threads from trees and shrubs along the edge of a forest adjacent to a Virginia Tech agricultural field. A Hobo data logger at each site recorded humidity at 10 min intervals during the 33 days that web samples were collected. To prevent theft, the data logger at the A. aurantia site was situated in an area where the surrounding vegetation probably maintained humidity at higher levels throughout the day than in more exposed places where many A. aurantia were found. Therefore, the humidity reported for this habitat is probably higher than some or many A. aurantia experience.

Samples of two individuals' webs were collected on each study day and the positions of these webs marked with flagging tape to prevent resampling. Images and videos from which measurements of droplet extension were later made were completed by 12:00 h and photographs used to determine glycoprotein area were completed by 16:00 h of the same day. To minimize temporal effects, we alternated the species studied, characterizing the threads of 2–4 individuals of one species and then switching to the other.

We collected an orb-web sector on a 17 cm diameter aluminum ring with a bar across its center. Scotch double-coated tissue tape (Tape 4101T; 3M, St Paul, MN, USA) applied to the 5 mm wide rim of the ring and bar secured threads at their native tensions. After transporting web samples to the laboratory in a closed container, we placed 4 mm wide brass bars covered with double-sided carbon tape (product 77816, Electron Microscopy Sciences, Hatfield, PA, USA) across the ring's center bar and rim along the sample's radial threads. This allowed us to collect short spans of capture thread without damaging other regions of the web sample or altering the native tension of adjacent viscous threads. We then collected individual threads from the outer third of an orb-web's capture region using a pair of tweezers whose tips were covered with double-sided carbon tape and blocked open to accommodate the support spacing of thread samplers. We accomplished this by pressing the tweezer tips against a thread and cutting it free using iris scissors. This thread sector was then transferred to the carbon tape-covered rails of adjacent brass, U-shaped supports glued at 4.8 mm intervals to a microscope slide to form a thread sampler [see fig. 3 in Opell et al. (Opell et al., 2011b)].

As adjacent droplets usually merged upon contacting the tip of the probe used to extend a droplet, we isolated an individual thread droplet near the center of each 4.8 mm long thread span. Exhaling gently on a thread span temporarily more fully hydrated its droplets, allowing unwanted droplets to be pushed away from the focal droplet using the moistened tip of a probe made from an insect minuten pin. This procedure did not remove the thin aqueous layer that covers axial fibers in inter-droplet regions and maintains their supercontraction. After droplets were moved, small secondary droplets, like those often found between the large primary droplets of native threads (Fig. 1B), reformed in the regions where larger droplets had been.

To control RH, we placed thread samplers in a humidity chamber [see fig. 4 in Opell et al. (Opell et al., 2011a)], which incorporated a holder for the thread sampler and ports for air inlet and outlet tubes, the droplet extension probe and the probe of a Fisher Scientific Instant Digital Hygrometer (Waltham, MA, USA), whose readings were updated in less than 10 s. A sheet of anti-Newton glass resting on a Sorbothane gasket sealed the top of this aluminum chamber. We controlled humidity by placing a small Petri dish filled with either silica gel beads or a Kimwipe moistened with distilled water into the chamber. Humidity was then adjusted by either exhaling into a tube connected to a chamber port to introduced humid air or withdrawing air from the tube to introduce drier room air. Temperature within the chamber was maintained at 23–24°C. Samples were measured first at 20% RH and then at successively higher values (37%, 55%, 72% and 90%). As described below, three sets of measurements were required to assemble this study's data. Our ability to establish RH values within tight limits resulted in the five experimental RH values being nearly identical for each set of the three measurement sets (Table 1).

After photographing a suspended droplet, we extended it using procedures that are described in detail a previous paper (Opell et al., 2011a) while a video was recorded at 60 frames s⁻¹. Before each trial we used 95% ethanol on a Kimwipe to clean the flat tip of the single steel probe, which was used to contact and extend all droplets. This probe was inserted through a port in the side of the chamber, and its tip was aligned with and then brought into contact with the droplet, advanced 500 μm to firmly attach the glycoprotein core to the probe's flat tip, and then withdrawn from the droplet at a velocity of 69.5 μm s⁻¹, while a video was recorded. The last video frame in which the extending droplet was attached to the probe's tip was captured (Fig. 4) and the length of the droplet filament measured with ImageJ using the probe's 413 μm wide tip as a scale.

For a representative A. aurantia out-of-frame droplet extension at 90% RH, the combination of measurement inaccuracies that produced the most extreme DE values resulted in a maximum deviation of ±161 μm or 3.7% of the median DE computed for this individual.

The glycoprotein core of droplets becomes visible when a glass coverslip is placed on a thread that is suspended between the supports of a thread sampler (Fig. 3). Unlike similar preparations viewed with a standard compound microscope (Fig. 5A) (Opell and Hendricks, 2010), the epi-illumination of the Mitutoyo microscope used in this study resolved the glycoprotein's outline, but did not reveal the smaller granule at its center (Fig. 3). To flatten droplets without opening the chamber after a desired RH was established and suspended droplets were photographed, a 22 mm diameter glass coverslip was suspended over the thread in a device attached by magnets to the underside of the chamber's glass cover. After three suspended droplets were photographed through the coverslip and their positions recorded, a magnet triggered the coverslip's release onto the suspended thread. All coverslips used to flatten droplets came from the same box and each was cleaned with a Kimwipe moistened with 95% ethanol immediately before use.

To ensure that the coverslip was firmly and uniformly pressed against the droplets, a probe was inserted through the port in the side of the chamber, positioned over the coverslip and centered between the supports of the microscope slide sampler, on which the coverslip rested. The shaft of the probe was then pressed downward until its movement was stopped by the lower edge of the slit in the chamber's side that permitted the probe to be moved along the length of the chamber for alignment with the droplets. Thus, we believe that the force applied in flattening droplets was standardized by three factors: (1) coverslips of the same mass were dropped on to threads from the same height, (2) the deflection of threads upon impact of the coverslips was controlled by the uniform length of the suspended thread spans, as determined by the equal spacing of thread sampler supports, and (3) the final force applied to the coverslip was set by the stiffness of the coverslip, the uniform width of the space into which a coverslip deformed when pressed by the probe and the uniform limit to downward movement of the probe, which was less than that required to break a coverslip.

We used ImageJ to measure the flattened surface area of each of the three droplets and of their glycoprotein cores. As an additional control for droplet flattening, we divided droplet volume by flattened droplet surface area to determine average flattened droplet thickness and used only the two droplets with the most similar thicknesses in subsequent computations. Flattened droplet measurements were not made at 72% RH for the first three A. aurantia studied, but were taken for all subsequent samples. Neoscona crucifera droplets were more viscous than those of A. aurantia. Consequently, we were unable to measure N. crucifera flattened areas at 20% RH and obtained measurements of only six individuals' threads at 37%. Glycoprotein could not be clearly resolved in the droplets of one A. aurantia individual at 90% RH or in the droplets of one N. crucifera individual at 55% and 72% RH.

For each RH we determined glycoprotein volume by multiplying the flattened area of a droplet's glycoprotein core by the droplet's average thickness. Subtracting glycoprotein volume from droplet volume yielded the volume of the droplet's outer, aqueous material. Because we could not both determine the volume of a droplet's glycoprotein core and extend this droplet, we inferred the volume of glycoprotein within an extended droplet by multiplying the droplet's volume by the mean glycoprotein volume-to-droplet volume ratio of the two flattened droplets of that individual's thread with the most similar droplet thicknesses. Indices of droplet extension per glycoprotein volume and relative stress on extended droplet filaments, as described below, were computed from a droplet's inferred glycoprotein volume, as this value was most closely associated with the droplet's performance.

We believe that this method provides a useful measure of glycoprotein volume. Nomarski images of flattened droplets show that the glycoprotein core extends above the level of the aqueous material (Fig. 5A). A side view of a 10 μl droplet of distilled water on glass shows the edges of the droplet to be tapered (Fig. 5B), a configuration that is consistent with that seen in a flattened droplet (Fig. 5A). Thus, a side view of a flattened droplet without its two axial fibers would be similar to the reconstruction in Fig. 5C. Actual glycoprotein volume is probably slightly greater than that determined from its surface area, although, after factoring out the volume of axial fibers passing through the glycoprotein, these two volumes should be similar. Increases in humidity appear to affect the flattened surface areas of glycoprotein and aqueous material in a similar manner. At higher humidity, glycoprotein plasticity increases and droplet cores probably flatten to a greater extent. At higher humidity, aqueous material becomes more dilute and less viscous and forms a thinner layer on the glass coverslip. Therefore, we believe that our method of determining glycoprotein volume was not biased by humidity level.

We computed two indices of droplet performance at each humidity: (1) extension per glycoprotein volume at pull-off (EGV) and (2) relative glycoprotein stress at pull-off (RGS). We computed EGV by dividing maximum droplet extension by glycoprotein volume. Values of this index increase as absorbed water makes glycoprotein more plastic, and decrease as absorbed water over-lubricates a droplet's contact footprint, allowing it to be pulled free with less force. Our instrument was not equipped with a load cell to allow us to directly measure the force on an extending droplet. Therefore, as the stiffness of these two species' axial fibers is similar (Young's modulus for A. aurantia E=0.009 GPa, N. crucifera E=0.010 GPa) (Sensenig et al., 2010), we inferred the force on a glycoprotein filament from the deflection of its supporting axial line, which we measured from the screen capture of the extended droplet immediately before pull-off using ImageJ (Fig. 4 and Fig. 6, angle A). Subtracting this axial line angle (Fig. 6, angle A) from 180 deg and dividing the result by 2 produces an angle (angle B, Fig. 6), whose sine is proportional to the force on the glycoprotein filament.

As noted previously, some droplets, predominantly droplets of A. aurantia extended at 72% and 90% RH, did not release from the probe until the axial lines had passed from the camera's field of view. In these cases, we measured axial line deflection in the last frame where axial line angle could be determined. At this point, the axial line angle of A. aurantia was nearly 180 deg, averaging 175 deg at 72% and 178 deg at 90%.

The RGS on an extended filament just prior to droplet pull-off was then determined by dividing either the relative force (or the humidity-corrected relative force, described below) by the cross-sectional area of the glycoprotein filament, determined by dividing glycoprotein volume by the length of the extended droplet at pull-off. Although RGS is not an engineering grade index, it does allow a general assessment and comparison of the two species' glycoproteins.

An issue that may impact the estimation of relative force at droplet pull-off is an apparent increase in axial fiber supercontraction as relative humidity rises. This phenomenon of supercontraction is well documented for dry dragline threads (Agnarsson et al., 2009; Blackledge et al., 2009; Boutry and Blackledge, 2010; Work, 1981), but has not been studied in viscous threads. Although it might appear that the already water-covered axial lines of viscous threads would not further supercontract, when immersed in water they do (Gosline et al., 1999) and become more taut. Three lines of evidence suggest that even humidity alters viscous thread supercontraction. (1) Videos of droplets extending at low humidities, particularly those of N. crucifera droplets, often showed the thread's axial line pair to be slack and sometimes convoluted as it flew back upon droplet release, whereas at higher humidities the supporting line remained taut. (2) As building humidity dropped, viscous threads in a few web samples showed some slackness, but became taut when gently exhaled upon, suggesting that water uptake increased supercontraction. (3) We placed four to six single droplet isolates on microscope slides so that we did not have to open the humidity chamber during a trial series. At low humidities these strands nearly always remained intact, but as humidity increased an increasing number of them spontaneously pulled from the adhesive holding them to the sampler's supports. We attribute this to two factors: increased supercontraction, which increases axial line tension, and increased water content of the thread's aqueous layer, which lubricates axial lines allowing them to be more easily pulled from the adhesive.

Interspecific differences in such supercontraction probably have only minor impacts on our estimates of the stress on glycoprotein filaments at pull-off and this phenomenon is tangential to the focus of our study. However, because supercontraction increases the tension on viscous threads, increased force is required to achieve the same angle of axial line deflection as humidity increases. In an attempt to account for this, we multiplied the sine of angle B by the humidity-specific, supercontraction multiplier values shown in Table 2. These correction factors describe an exponential increase in super contraction from zero at the lowest humidity (20% for A. aurantia and 37% for N. crucifera) to a factor of 2 at 90% RH. The values used for the two species differ because at 37% RH, N. crucifera droplets have just become pliable enough to respond in a manner similar to those of A. aurantia at 20% RH. These supercontraction-corrected relative stress values are reported in Table 2.

We assembled, summarized and analyzed data using JMP (SAS Institute, Cary, NC, USA), considering P≤0.05 as significant. As a previous study determined that ambient humidity affects the volume and extensibility of whole viscous thread droplets (Opell et al., 2011a), this analysis capitalizes on the increased number of test humidities to compare the response curves of aqueous material, glycoprotein and glycoprotein-specific indices. If the mean value of a feature at each RH was normally distributed (as shown by Shapiro–Wilk W-tests with P≥0.05), we used ANOVA to test the differences in mean values; these results are presented in Figs 9, 10, 11, 12 and 13. If the mean at one or more humidites was not normally distributed, we log-transformed data. In most cases the means then became normally distributed. In four cases (glycoprotein volume at 20% RH for A. aurantia, droplet extension at 20% RH for A. aurantia and 37% RH for N. crucifera, and all relative stress values of A. aurantia droplet filaments) mean values remained non-normal. For the first three indices we performed both ANOVA and Wilcoxon/Kruskall–Wallis tests of means. For the relative stress of A. aurantia, we used only a Wilcoxon test.

The curves that describe changes in droplet features over the five experimental humidities (e.g. Fig. 9A,B) are accompanied by the percentage change in mean values from those expressed at 20% RH for A. aurantia and 37% RH for N. crucifera, whose computation required a droplet to flatten (e.g. Fig. 9B). When possible, the mean values at the five humidites were fitted to exponential or logarithmic curves, whose formulas are provided. When these curves did not adequately describe the values, polynomial curves were fitted. Below each feature of these figures we provide a histogram of the percentage maximum value of this feature, based on the greatest value expressed by both species (e.g. Fig. 9C,D).

From 11:00 h to 18:00 h, humidity was lower in the A. aurantia habitat than in the N. crucifera habitat, being significantly different from 12:00 h to 16:00 h (Fig. 7). Table 1 summarizes the volumes and extensibilities of droplets and Table 2 presents the values used to compute relative droplet stress. For N. crucifera, features computed from flattened droplet measurements were available only for 37% RH and greater, although values inferred for 20% RH from regression plots are given in brackets in Table 1 and included in the percentage histograms of Figs 9, 10, 11, 12 and 13.

Across the five RH values the relationship between flattened droplet area (FDA) and flattened glycoprotein area (FGA) was similar for the two species (Fig. 8), showing that the species' droplets flattened in a similar manner and therefore would not bias our computation of glycoprotein volume in favor of one of the species. An insignificant effect of the interaction between FDA and species on FGA (P=0.58) showed that the slopes of the species' regressions did not differ. However, a significant contribution of species (P=0.0049) to the model ‘FGA=FDA×species’ showed that the Y-intercept of A. aurantia was greater than that of N. crucifera, as would be expected from the larger size of the former species' droplets (Fig. 1, Table 1).

With the exception of N. crucifera droplets measured at 37% RH, droplet volume increased as humidity increased, with A. aurantia droplets being more hygroscopic than N. crucifera droplets (Table 1). At 90% RH, A. aurantia droplets were 258.5±26.5% (mean ± s.e.m.) larger than at 20% RH, whereas N. crucifera droplets were only 160.6±8.5% larger at 90% than at 20% RH, a difference that was significant (t-test, P<0.0014). The aqueous component of both species' droplets increased as humidity increased (Table 1, Fig. 9). With the exception of A. aurantia droplets measured at 55% RH, glycoprotein volume increased as humidity increased (Table 1). At 37% and 55% RH, glycoprotein volume appeared to decrease or stabilize in both species (Fig. 10), although in N. crucifera the high variance and narrow range of values made this difficult to assess.

When viewing Figs 9, 10, 11, 12 and 13, which present the responses of viscous droplets to increasing humidity, it is useful to remember that viscous threads first experience high humidity during the early morning hours and humidity decreases as the day progresses. In each species, droplet extension increased with humidity, with the two species reaching similar pull-off lengths at 90% RH, despite the smaller volume of glycoprotein in N. crucifera droplets (Table 1, Fig. 11). However, the exponential increase in N.crucifera droplet extension resulted in this species' droplet filaments being much shorter than those of A. aurantia at all other humidities.

Pronounced differences were observed in the two species' droplet EGV (Table 1, Fig. 12). In A. aurantia, EGV increased in a linear fashion from 20% to 55%, presumably as absorbed water increased glycoprotein plasticity, and then declined as droplets became over-lubricated and more easily released from the probe (Tukey HDS test rank, from low to high index values: 20%, 37%=90%, 55%=72%). In contrast, N. crucifera EGV values increased exponentially as humidity increased, reaching a maximum value at 90% RH that was 3.58 times the maximum value reached by A. aurantia at 55% RH.

In A. aurantia, uncorrected RGS just prior to pull-off decreased in a linear fashion as humidity increased, whereas RGS increased logarithmically in N. crucifera (Table 1, Fig. 13A,B), reaching a maximum value at 90% RH, 13.9 times the maximum value of A. aurantia, which was observed at 20% RH. When corrected for supercontraction, A. aurantia RGS did not differ among humidities (Wilcoxon χ², P=0.1582), whereas N. crucifera corrected RGS did (ANOVA, P<0.0001), increasing exponentially (Y=1.596⁻⁷×X^3.39) and reaching a maximum value at 90% RH, 26.7 times the maximum value of A. aurantia, which was observed at 37% RH (Table 2).

Our results support the hypothesis that A. aurantia droplets are more hygroscopic than N. crucifera droplets, permitting them to remain hydrated and perform optimally, and adapting them to drier environments and times of the day. However, like droplets of L. cornutus (Sahni et al., 2011), those of A. aurantia become progressively over-lubricated at RH greater than 55%, resulting in decreased EGV (Fig. 12A,C, Fig. 14). In contrast, the EGV of N.crucifera increased exponentially as humidity increased, and showed no evidence of over-lubrication (Fig. 12B,D, Fig. 14). The lower hygroscopicity of N. crucifera droplets also caused their glycoprotein to remain more viscoelastic at higher humidity, resulting in an exponential increase in the relative stress on the glycoprotein filaments at pull-off (RGS) as humidity increased (Fig. 13B,D). This contrasts with the low and decreasing RGS of A. aurantia droplets as RH increased. As RGS is an indicator of the energy required to extend a droplet to pull-off, this suggests that increased humidity enhances the ability of N. crucifera threads to absorb the energy of struggling prey through extension, whereas it has a negative effect on the ability of A. aurantia threads to absorb energy in this manner.

These interspecific differences in EGV and RGS are consistent with the hypothesized environmental adaptiveness of droplet hygroscopicity. At 14:00 h–15:00 h, when the humidity of both species' habitats was at its lowest (Fig. 7), RH in the A. aurantia habitat dropped to a mean of 66% and in the N. crucifera habitat to 76%. For A. aurantia, this low humidity mark corresponds with the humidity range at which droplets have reached their maximum EGV and just begin to show signs of over-lubrication (Fig. 12A,C, Fig. 14). In contrast, at times of lowest daily RH, N. crucifera droplets exhibited only half of their maximum EGV (Fig. 12B,D) and over half of their maximum RGS (Fig. 13B,D). Consequently, when differences in EGV and supercontraction corrected RGS are examined from mid morning to late afternoon (10:00 h–18:00 h), A. aurantia droplets exhibited their maximum values during mid afternoon and N. crucifera during the morning (Fig. 15). Thus, hygroscopic differences in the two species' viscous threads appear to adapt them to function optimally not only in different humidity regimes (Fig. 14) but also at different times of the day (Fig. 15). However, it is important to note that the minimum midday EGV of N. crucifera was 1.9 times the maximum value for A. aurantia and that the minimum midday RGS of N. crucifera was 23 times the maximum value for A. aurantia.

As a biomimetic system, N. crucifera droplets appear to offer more promise than A. aurantia droplets because N. crucifera EGV and RGS increase exponentially and logarithmically, respectively, as RH increases (Fig. 11B, Fig. 12B, Fig. 13B). Moreover, N. crucifera droplets attain a maximum EGV that is 3.6 times that of A. aurantia and an RGS at 90% RH that is 64 times that of A. aurantia at 90% RH. However, A. aurantia droplets illustrate how an adhesive system might be engineered to fail at a desired humidity. The simplest explanation for these interspecific differences is that the aqueous material of A. aurantia is more hygroscopic than that of N. crucifera as a result of having more and/or different low molecular weight compounds. Argiope aurantia glycoproteins may also be more hygroscopic, causing them to take up a higher percentage of the atmospheric water provided by the aqueous material and, thereby, become more extensible at a given RH than the glycoprotein within N. crucifera droplets.

Although it is convenient to attribute increased droplet extensibility at higher humidities to the dilution of glycoprotein by absorbed atmospheric water, this may be an oversimplification given the complex interactions that establish the three-dimensional configuration of glycoproteins. Intramolecular hydrogen bonding establishes most of the directional interactions that underpin protein folding. However, the burial of hydrophobic (apolar) side-chains within the molecule plays a key role in organizing and stabilizing the folded molecule (Rose, 1993; Bolen and Rose, 2008). Residues within these buried hydrophobic regions are themselves configured by hydrogen bonding. Poor osmolytes, like water, disrupt intramolecular hydrogen bonding and suppress secondary structure because the strength of these bonds is weakened by competition with water molecules that supply alternative bonding sites, thereby reducing hydrogen bonding as water content increases. These structural changes may increase glycoprotein unfolding as humidity increases. Moreover, increased water content may also cause additional apolar groups to be forced to the molecule interior, a change that is associated with increased entropy and volume (Dill, 1990). The decrease in glycoprotein flattened area and volume observed at 55% RH in A. aurantia and the probable stabilization of N. crucifera glycoprotein flattened area and volume at intermediate humidities (Table 1, Fig. 10A,B) may reflect such changes in molecular configuration.

This study shows that a single model cannot explain the performance of viscous threads produced by all orb-weavers and suggests that the environmental responsiveness of these threads has been shaped by natural selection. A full understanding of viscous thread evolution and operation will require investigations that characterize this diversity and address the impact of web environment on a viscous thread's molecular structure, material properties and prey capture performance.

Mary Lee Hendricks and Sarah Stellwagen provided useful comments on the manuscript.