Humidity-mediated changes in an orb spider's glycoprotein adhesive impact prey retention time

An orb spider's web is a behavioral and material extension of the spider's phenotype on which the spider relies for prey capture. The web integrates non-adhesive radial and frame threads that absorb and dissipate the kinetic energy of an insect's impact (Sensenig et al., 2012) with spirally arrayed, adhesive prey capture threads (Sahni et al., 2013) that retain the insect until the spider can locate, run to, and begin to subdue it (Chacón and Eberhard, 1980). Thus, material invested in non-adhesive threads increases the size and velocity of insects that a web can successfully stop and material invested in capture thread increases the time a spider has to prevent an insect from escaping. A retention time difference of even 5 s often means the difference between a prey captured or a prey lost (Blackledge and Zevenbergen, 2006; Eberhard, 1989).

Both the non-adhesive and adhesive components of orb webs exhibit phenotypic plasticity (Blamires, 2010; Boutry and Blamires, 2013; Dawkins, 1982). Differences in diet and prey type cause striking differences in both a web's architecture and the properties of its threads (Blamires et al., 2016; Herberstein and Tso, 2011; Scharf et al., 2011; Townley et al., 2006; Tso et al., 2007), as can environmental factors such as wind (Wu et al., 2013). After a web has been spun, its frame and axial threads and its capture threads respond to daily changes in environmental conditions, with humidity having the greatest impact (Agnarsson et al., 2009; Sahni et al., 2011; Opell et al., 2011, 2013; Stellwagen et al., 2014, 2015; Amarpuri et al., 2015).

As explained more fully in the following paragraphs, studies of the effect of humidity on the adhesion of capture threads have concluded that they evolved to perform optimally in the humidity regime typical of a species' habitat (Opell et al., 2013; Amarpuri et al., 2015). In characterizing this performance, these studies have used flat glass and steel contact surfaces and constant velocity extensions to ensure uniform and precise comparisons. However, in nature, capture threads must resist forces whose direction and magnitude change randomly and rapidly and must adhere to three-dimensional, non-uniform insect surfaces, across which adhesion differs greatly (Opell and Schwend, 2007). Because successful prey capture appears to be the only mechanism by which natural selection tunes the humidity-specific performance of an orb web, it is important to verify the link between a thread's fine-scale material properties and its ability to retain prey. The objective of the current study was to document this association.

Modern (Araneoidea) orb spiders employ viscous capture threads, which are formed of regularly spaced droplets (Fig. 1A), each composed of an adhesive glycoprotein core surrounded by a hygroscopic aqueous layer (Fig. 1B; Edmonds and Vollrath, 1992; Townley et al., 1991; Sahni et al., 2013; Townley and Tillinghast, 2013; Opell et al., 2015). These threads are spun as a continuous cylinder of aggregate gland dope surrounding a pair of supporting flagelliform fibers. Low molecular weight organic compounds and salts in this aggregate material confer hygroscopicity (Townley and Tillinghast, 2013), causing it to attract atmospheric moisture, enlarge and then form droplets (Edmonds and Vollrath, 1992). Within each droplet, a glycoprotein core condenses (Fig. 1B). Over the course of a day, the volumes of these hygroscopic droplets change as they respond to changes in environmental humidity (Townley et al., 1991; Opell et al., 2013). Although the allocation of absorbed atmospheric water has not been fully studied, some of it solvates the glycoprotein, ensuring its interaction with a surface, some of it hydrates the glycoprotein core, reducing its viscosity, and some of it probably increases flagelliform fiber extensibility (Sahni et al., 2011, 2014; Opell et al., 2011, 2013; Amarpuri et al., 2015).

Optimal glycoprotein viscosity is crucial for droplet adhesion (Amarpuri et al., 2015). To resist adhesive failure, glycoprotein must be pliable enough to establish contact with an insect surface (Fig. 1B). But the glycoprotein must also be stiff enough to resist cohesive failure when it extends as an insect struggles to escape from a web (Fig. 1C; Amarpuri et al., 2015). When glycoprotein absorbs too much water, it becomes over lubricated and its extensibility is reduced (Sahni et al., 2011; Opell et al., 2013). Moreover, droplet extensibility, in conjunction with flagelliform fiber extensibility, is also responsible for recruiting and summing the adhesion of multiple droplets (Fig. 1D; Opell and Hendricks, 2009), and for absorbing the energy of prey struggle (Sahni et al., 2011). Droplet hygroscopicity and thus glycoprotein viscosity appears to have been tuned to the humidity regime typical of a species' habitat by selection on the salts and low molecular weight organic materials in the droplet's aqueous layer (Townley and Tillinghast, 2013; Amarpuri et al., 2015). When the droplets of species that occupy different habitats were examined, not only did they perform optimally at the humidity typical of each species' habitat but also at these optimal humidities their glycoprotein viscosities were remarkably similar (Amarpuri et al., 2015).

This study tested the hypothesis that humidity-mediated differences in viscous droplet performance translate into differences in insect retention times. We studied threads of the orb weaver Araneus marmoreus Clerck 1757, a diurnal species that constructs vertically oriented orb webs in vegetation on forest edges, where humidity remains elevated during the day, and monitors its web using a signal line that extends from the web to the protective confines of the spider's folded leaf retreat (Bradley, 2013; Fasola and Mogavero, 1995; Jennings and Graham, 2007). We inferred the optimal humidity for the adhesive performance of this species' threads from: (1) the conditions of habitats where this species is found, (2) measurements of individual viscous droplet glycoprotein surface area and extensibility, and (3) a model that used these droplet values to estimate the work required to initiate the pull-off of a 4 mm-long capture thread span from a surface. We then tested this prediction by comparing the active struggle time required of house flies to escape from capture threads under different humidities.

Assessing prey retention, particularly in vertical orb webs, is challenging, as it is affected by many factors. These include: (1) the mass of an insect and its impact velocity, (2) the number of capture threads that it strikes, (3) the region of the web a prey strikes, and (4) whether, after struggling free from these threads, the insect tumbles into other capture threads (Blackledge and Zevenbergen, 2006; Zschokke and Nakata, 2015). Moreover, insects often exhibit bouts of escape behavior with intervals of inactivity. To control these variables, we assayed escape behavior by placing an anesthetized house fly wings-downward on three, equally spaced, horizontally oriented capture thread strands (Fig. 2) and capturing a video of the fly's escape behavior, permitting us to determine the total active struggle time required to overcome thread adhesion.

We placed a Hobo U23 temperature/relative humidity (RH) data logger (Onset Computer Corporation, Bourne, MA, USA) in each of two habitats where we have observed A. marmoreus over the past 12 years. The probe of one logger extended from vegetation of a southeast-facing forest edge at a height of 1.5 m and the probe of the other logger extended from vegetation at the edge of a north-facing forest edge at a height of 1 m. Each logger recorded humidity and temperature at 12 min intervals from 15 August to 15 October, 2016. This interval extended from the time during which A. marmoreus are in the last one or two juvenile stadia to the time that most adult females were killed by frosts or died of other causes.

To ensure that spiders were well nourished and their webs constructed under natural conditions, we collected sectors from orb webs constructed by A. marmoreus adult females from their natural habitat on vegetation at the edge of forests near Blacksburg, Montgomery County, VA, USA, between 06:00 h and 08:30 h during September and October. Flagging tape marked the position of each sampled web to prevent resampling. The 13 samples used to characterize droplets were collected during 2013 and 2014 and all still and video images used to describe their properties were completed by 16:00 h on the day of collection. The 14 samples used to evaluate prey retention were collected in 2015 and all videos of insect escape were made by 11:00 h on the day of collection.

Web sectors were collected on 18 cm diameter aluminium rings with a bar across the center. Double-sided 3M tape (no. 9086K29550360, 3M, Maplewood, MN, USA) applied to the 0.6 cm width of the ring and center bar secured the threads without altering the original structure and tension of the web's threads. Immediately after collecting a web sample, this ring was placed in a closed plastic container to prevent contamination by dust and debris. After transporting the webs to the laboratory, we placed 4.0 mm brass bars covered in double-sided carbon tape (cat no. 77816, Electron Microscope Sciences, Hatfield, PA, USA) along the sample's radial threads and across the ring's rim and center bars, enabling us to collect single capture thread strands without compromising their native tensions.

Capture thread samples were removed from rings on the carbon tape-covered tips of forceps that were blocked open to accommodate the supports on microscope slide thread samplers and burned free of the web sample using a hot wire probe. These thread samplers consisted of five U-shaped brass struts epoxied at 4.8 mm intervals to microscope slides with their free ends extending upward and covered with double-sided carbon tape (fig. 3 in Opell et al., 2011). Two sets of thread samples were prepared from each web, one to measure glycoprotein contact surface area and one to measure droplet extension. Each set of measurements was made within a temperature- and humidity-controlled chamber that rested on the mechanical stage of a Mitutoyo FS60 inspection microscope (Mitutoyo America Corp., Aurora, IL, USA), where a temperature of 24°C and RHs of 20%, 37%, 55%, 72% and 90% were established (Opell et al., 2013). These values correspond to absolute humidity values of 4.35, 8.05, 11.97, 15.67 and 19.58 g m⁻³, respectively.

In preparing suspended droplets for extension measurements, we ensured that the probe used to contact and extend a droplet contacted only a single droplet. We did so by using a minuten insect pin moistened with distilled water to move away droplets that were adjacent to the test droplet located at the center of the thread strand. This process retained the aqueous coating of the strand's axial fibers, as demonstrated by the formation of small droplets similar to those often present between the large primary droplets of many viscous threads (Fig. 1A). Prior to each extension, we photographed the isolated suspended droplet and used this image to infer its glycoprotein volume as described above. A steel probe, cleaned with 100% ethanol on a Kimwipe^® before each use, was then inserted through a port in the side of the test chamber and its 413 µm­-wide polished tip aligned and brought into contact with the focal droplet. To ensure full droplet adhesion, the probe was pressed against the droplet until it was deflected by a distance of 500 µm. A 60 frames s⁻¹ video then recorded the droplet's extension as the thread was withdrawn from the probe at a velocity of 69.6 µm s⁻¹ by a stepping motor, which moved the mechanical stage on which the observation chamber rested. From these videos, we measured the length of the extending droplet filament just before its release from the probe.

To determine the change in water volume within a viscous droplet, we subtracted the mean droplet volume at 20% RH from that at each of the other RHs. We determined the change in water volume within a droplet's glycoprotein core by subtracting the mean glycoprotein volume at 20% RH from that at each of the other RHs. We then determined the change in water content of the aqueous layer at each of the four higher RHs by subtracting the change in water content of the droplet's glycoprotein core from the total change in droplet water content.

When force is applied to a viscose thread that has adhered to a surface, the droplets extend (Fig. 1D) until the outermost one pulls from a surface and inner droplets successively extend and release. The thread's configuration just prior to the initiation of pull-off, as marked by the release of the outermost droplet, represents the point at which maximum thread adhesion is achieved (Opell and Hendricks, 2009). The work required to extend the droplet to this point is directly related to the volume of droplet's glycoprotein (V), the stiffness of its glycoprotein (S) and the distance over which the glycoprotein (droplet) is extended (D). Volume and extension increase, up to a point, as humidity increases, whereas stiffness decreases.

Araneus marmoreus viscous threads have an average of 3.7 droplets per mm (Opell and Hendricks, 2009), or 15 droplets per 4 mm span. Because the viscous thread's axial lines are very extensible, the two outermost droplets have extended much further than the inner droplets by the time the outer droplets release (Fig. 1D). Our simple model assumes that the thread's paired support lines take the shape of a parabola, as suggested by our current studies (B.D.O., unpublished observations). Thus, as humidity increases, the total extension of the droplets in a span will increase because the stiffness of their glycoprotein decreases (Fig. 3A,B, 20–72% RH). This continues until droplet extension diminishes as the glycoprotein within a droplet begins to experience adhesive failure at its contact footprint and cohesive failure in its extending filament (Fig. 3A, 90% RH).

We used tweezers with tips that were blocked open to a width of 2 cm and wrapped in carbon tape to collect a single span of capture thread from the web sample. Three 18 mm thread strands were placed 2.3 mm apart on parallel bars covered in 3M tape (no. 9086K29550360, 3M) (Fig. 2). The bars were glued to a 50 mm ring from which two mounting rods extended, allowing us to position this sample holder horizontally during observations of insect retention.

We assayed insect retention using adult Musca domestica house flies (item 100002365, Evergreen Growers Supply, Oregon City, OR, USA) reared from pupae. A sample of six flies had a mean mass of 12.01±0.88 mg, a head width of 222±4 µm, a notum width of 211±7 µm and an individual wing width of 247±11 µm. Each fly was anesthetized with CO₂ gas for 3–4 s and was then placed wings-downward on the three capture threads with its head touching the first thread and the wings touching the second and third threads (Fig. 2).

Target RHs of 37%, 55% and 72% were established in three 30.5×49.5×31.8 cm Plexiglas desiccator chambers. The 37% RH chamber was equipped with a small Ivation Peltier dehumidifier (model 236072) and a tray of silica gel desiccant. Two ice packs were placed in the chamber to offset the dehumidifier's thermal input. As laboratory humidity was slightly below 55%, we regulated the humidity of the 55% RH chamber by gently exhaling into the chamber. A small Foxwill Volcano Ultrasonic Humidifier 160 ml USB mini humidifier and circulating fan established 72% RH. We monitored humidity with a Fisher digital hygrometer (model 11-661-7A) and temperature with a Fisher digital thermometer (model 11-661-8), recording the values under which each test was conducted.

We placed a thread sample holder into a humidity chamber before anesthetizing a fly, allowing capture threads to acclimate to the test humidity for 1–2 min and then placed a fly, wings downward, on the threads. After the anesthetized fly was placed on threads, we filmed its behavior at 60 frames s⁻¹ with a Canon Vixia HF 610, HD camcorder (Movie 1). Using iMovie, we analyzed each video to determine the length of time a fly was actively struggling to leave the web. We characterized active struggle as rapid wing flapping, which was sometimes also associated with rapid leg movements. We recorded both the number of these active bouts and their duration, summing them to determine a fly's total struggle time.

We used JMP (SAS Institute, Cary, NC, USA) to analyze data and considered comparisons with P≤0.05 as significant. All tests are two-tailed. The normal distribution of values was assessed with a Shapiro–Wilk W-test, with values of P>0.05 being considered normal.

At the two representative A. marmoreus sites, RH decreased to 67–76% and temperature increased to 23–24°C by mid-afternoon, resulting in absolute humidity remaining between 14.5 and 15.8 g m⁻³ (Fig. 4), values that most closely match the 15.73 g m⁻³ mean value of our 72% RH treatment (Table 3).

Table 2 summarizes the effect of humidity on droplet features and performance. Mean droplet dimension (average of droplet length and width) increased slightly in the two highest RHs, although the ratio of droplet length (direction parallel to droplet axial lines) to width changed little as humidity increased (Fig. 5). Despite this, there was a precipitous drop in free absorbed water present in a droplet's aqueous material at 55% RH followed by increases at 72% and 90% RH. The volume of water absorbed by a droplet's glycoprotein core increased at 55% and 72% RH and decreased at 90% RH.

To reduce the effect of inter-individual droplet size, we plotted both glycoprotein contact area and droplet extension values relative to the glycoprotein volume of an individual's droplet. The greatest relative glycoprotein contact area occurred at 55% and 72% RH, with relative droplet extension peaking at 72% (Fig. 6), causing us to hypothesize that capture threads should retain house flies for the greatest length of time at 72% RH.

The model of relative changes in glycoprotein features (Table 1, Fig. 3B) showed that relative glycoprotein volume increased slightly as humidity increased to 72% and then decreased, whereas relative glycoprotein stiffness decreased greatly as humidity increased. Relative droplet extension increased continually to 72% and then decreased. When these values were used to approximate the relative work of adhesion necessary to extend a 15 droplet, 4 mm span of capture thread to the point of pull-off, they resulted in steadily increasing work up to 72% RH and a marked decrease in work at 90% RH (Fig. 3C). Thus, these results support those of single droplet performance described in the previous paragraph in predicting that A. marmoreus capture threads should perform optimally at 72% RH.

Table 3 summarizes insect retention results and Table S1 provides the core results. We controlled RH within narrow limits, but mean temperature differed among treatments, being 1.2°C higher at 37% RH and 1.4°C lower at 72% RH than at 55% RH. However, regressions showed no relationship between temperature and total struggle time for 37%, 55% and 72% RH treatments (P=0.2779, 0.7174, 0.1322, R²=0.097, 0.011, 0.179, respectively). There were no inter-humidity differences among the slopes of these regression lines (P=0.1638). Neither total in-web time nor the number of active struggle bouts differed among RHs (Table 3). Although a Wilcoxon test showed no overall difference in fly struggle time among RHs (Table 3), the struggle time at 72% RH was twice that at 37% and 55% RH and a Wilcoxon each-pair test identified a difference between 37% and 72% RH values (P=0.0386).

Mean individual struggle time ranged from 3.33 to 34.67 s and averaged 14.26±2.98 s (mean±s.e.m.). We attributed some of this variance to inter-spider differences in capture thread adhesion. We addressed this by first averaging the active struggle times required for flies to escape from an individual spider's threads at 37%, 55% and 72% RH. We then subtracted this mean value from the active struggle time required to escape from this individual's threads at each experimental humidity to compute an index we term fly struggle time difference (FSTD).

A linear regression showed no experiment-wide effect of temperature on FSTD (P=0.3259, R²=0.024), with quadratic and exponential 3P models showing little improvement (R²=0.031 and 0.030, respectively). A linear regression of FSTD on measured RH and temperature was marginally insignificant (P=0.0503), with RH but not temperature contributing to this model (P=0.0259 and 0.5068, respectively). Measured RH was significantly and positively related to FSTD (P=0.0179). We also determined the difference between each individual's mean experimental temperature (which ranged from 23.17 to 27.07°C) and the temperature at which each of its humidity trials were conducted, an index we term temperature difference (TD). Regressions showed no relationship between TD and FSTD across the data set or for 37%, 55% and 72% RH treatments (P=0.1806, 0.6131, 0.1093 and 0.4843, respectively). A regression of FSTD on measured RH and TD was significant (P=0.0317), with RH but not TD contributing to this model (P=0.0244 and 0.2454, respectively). Therefore, we assume that temperature can be eliminated as a contributing variable in subsequent analyses of the effect of humidity on FSTD.

Shapiro–Wilk tests showed that FSTD was normally distributed at 55% and 72% RH, but not at 37% RH (P=0.7732, 0.0814 and 0.0170, respectively), requiring the use of non-parametric statistics. A Wilcoxon test showed that RH affected FSTD (P=0.0187), with Wilcoxon each-pair tests identifying differences between values at 37% and 72% RH (P=0.0082) and between those at 55% and 72% RH (P=0.0326), but not between those at 37% and 55% RH (P=0.8362) (Fig. 7).

Environmental humidity plays a critical role in the function of an orb web from the time that it is constructed until the time that it is taken down. High humidity during web construction facilitates the self-assembly of viscous capture threads. Changes in humidity over the course of a day affect the web's ability to both withstand prey impact and retain intercepted prey. Webs of Argiope trifasciata and Nephila clavipes were better able to absorb the kinetic energy of simulated prey strikes at high humidity than at low humidity (Boutry and Blackledge, 2013). Our study demonstrates that fine-scale, humidity-mediated changes in capture thread droplet adhesion affect prey retention times. Finally, when a web is taken down and its silk ingested during the high humidity of the evening or early morning, the fully hydrated viscous droplets supply a spider with both recyclable protein and water (Townley and Tillinghast, 1988; Edmonds and Vollrath, 1992).

Our observations support the hypothesis that humidity-mediated changes in A. marmoreus viscous thread droplets and threads, as assessed by laboratory tests, translate into measurable differences in the retention time of actively struggling house flies. House flies were retained an average of 11 s longer at 72% RH than at 37% and 55% RH. This difference in prey retention is ecologically significant, as the short time after an insect strikes a web and before a spider commences wrapping the insect is the insect's only opportunity to escape from the web. A study of the retention of hanging flies, and small and large grasshoppers by orb webs from which spiders were removed showed that of the insects that escaped from webs within a 180 s trial period, 18% did so within 1 s of striking the web (Blackledge and Zevenbergen, 2006). Our study used prey retention time to evaluate the performance of capture threads. However, retention time should be directly related to the size of prey that a web can retain. For large orb weavers like A. marmoreus, these large prey are crucial because they comprise the greatest proportion of a spider's total food intake (Venner and Casas, 2005), although aspects of this large profitable prey hypothesis have been called into question (Eberhard, 2013). Thus, the hygroscopicity of viscous threads has probably been tuned to the humidity of a species' habitat by selection favoring an increase in both the number and size of prey that a web captures in ways that facilitate not only spider survival but also the total number of eggs produced.

The performance of A. marmoreus threads is consistent with this species' occupation of forest edge habitats (Bradley, 2013; Fasola and Mogavero, 1995; Jennings and Graham, 2007), where humidity remains elevated during the late morning and early afternoon hours (Fig. 4). While recordings from only two sites do not fully characterize this species' habitat, they combine with analyses of droplet and thread performance to suggest that A. marmoreus viscous threads have been selected to function in moderately high humidities. Moreover, in the two sampled habitats, absolute humidity changed little from mid-morning to late afternoon, indicating that these threads encounter more stable conditions than suggested by RH alone.

Humidity affects viscous droplets and threads in many ways. As humidity increases, more water is absorbed by a droplet's glycoprotein, reducing its viscosity and causing it to have a greater surface area of contact and extend further before releasing until, at the highest humidity, the glycoprotein becomes over-lubricated (Sahni et al., 2011; Opell et al., 2013) and extension decreases (Fig. 6). However, the effect of humidity on water allocation within a droplet appears more complicated, as reflected by a large drop in the free water content of the droplet's aqueous material at 55% RH (Fig. 5). A similar deflection in droplet volume at 55% RH was observed in Argiope aurantia (see fig. 9A of Opell et al., 2013). The formula used to compute droplet volume is unaffected by changes in droplet volume at different humidities (Liao et al., 2015) and the drop in free water within the aqueous layer at 55% RH seems too great to be explained by small inaccuracies in our method of assessing the change in glycoprotein water content, particularly as glycoprotein volume is only 20–30% of droplet volume (Table 1, GV/DV ratio). This dip in water content is consistent with a study of Araneus diadematus orb webs, which showed that at about 48% RH, whole-web samples, water-soluble and -insoluble, and ethanol-soluble and -insoluble­ fractions of webs exhibited a dip in weight gain as humidity increased (see fig. 2 of Vollrath et al., 1990). That study also showed that the N-acetyltaurine, GABamide, and sodium isethionate found in these viscous threads did not begin to absorb atmospheric moisture until RH increased to 42–46%. Thus, it appears that two overlapping phenomena may underlie the apparent dip in the aqueous layer's water volume at 55% RH: (1) more water is taken up by the flagelliform fibers that pass through the droplet and by the low molecular weight compounds within the aqueous layer when humidity exceeds 40% and (2) less-hygroscopic compounds, like those mentioned above, begin to add water to the droplet when humidity exceeds 55%. It may be possible to distinguish these two events by measuring droplet and glycoprotein volumes at smaller intervals within the 40–60% RH range.

Our simple model of the work required to extend a thread span to the initiation of pull-off shows how humidity-mediated changes in glycoprotein volume and viscosity impact the thread span's ability to resist an insect's struggle (Fig. 3). Our model showed that changes in glycoprotein volume have the least effect and changes in glycoprotein extensibility the greatest effect on thread performance. As humidity increases and the glycoprotein becomes less viscous, less energy is required for its extension. Thus, in the context of a thread span, greater droplet extensibility allows more droplets to extend a greater distance, resulting in more energy being absorbed by the thread span. However, actual thread performance begins to deviate from this model at very high humidity, because some droplets slide and merge or their extending filaments combine to form a sheet, leading to configurations that are more challenging to model (Amarpuri et al., 2015; B.D.O., unpublished observations).

Sheree Andrews provided technical support and helpful comments on the manuscript. Sarah Stellwagen captured some of the images and videos used in droplet analyses. Sarah Helweg measured the lengths and widths of droplets. Robin Andrews discussed and provided perspective on the statistical analyses.