Tuning orb spider glycoprotein glue performance to habitat humidity

Evolution in silk use has played a crucial role in the success of the diverse, over 47,000-species-strong arachnid order Araneae to which spiders belong (Vollrath, 2005; Vollrath and Selden, 2007; World Spider Catalog, 2017). The order Araneae is composed of two suborders: Mesothelae, which have segmented abdomens like scorpions and spinnerets that extend from the middle of their abdomen's ventral surface; and Opisthothelae, which have unsegmented abdomens and posterior spinnerets (Platnick and Gertsch, 1976). Opisthothelae contains two infraorders: Mygalomorphae, which includes tarantula and trapdoor spiders whose cheliceral fangs move parallel to the body's sagittal plane; and Araneomorphae, which contains over 95% of all living spider species whose fangs move more perpendicularly to the sagittal plane. Araneomorphae origin coincided with the appearance of a cribellum, a spinning plate formed of thousands of spigots that produces the nanofibers of a dry, fuzzy capture thread termed cribellate thread (see Glossary). Although some araneomorphs continue to spin cribellate threads (Opell, 2013), most no longer do so, constructing webs that are not sticky or, like jumping spiders and wolf spiders, abandoning web use in favor of other hunting tactics. The first orb webs contained cribellate threads but 110 million years ago members of the superfamily Araneoidea replaced these with moist viscous capture threads (see Glossary) (Peñalver et al., 2006). These viscous threads are considered a key innovation (Bond and Opell, 1998), contributing to the diversity of this clade, which contains 26% of all spider species and comprises 17 families of orb-weaving spiders and their descendants that spin webs with divergent architectures (Blackledge et al., 2009a,b; Dimitrov et al., 2016; Hormiga and Griswold, 2014).

Organisms employ adhesive secretions for a variety of other functions. For example, Polychaeta annelids construct protective tubes from cemented sand particles (Pavlovič et al., 2014), barnacles cement their cases to rocks and mussels attach themselves by byssal threads to the substrate to avoid being swept away by currents (Kamino, 2010; Waite, 2017). Like most commercial adhesives, bioadhesives typically have an initial low-viscosity phase, during which they establish surface contact, followed by a phase of increased stiffness, which allows them to resist the crack propagation that leads to failure (Gent, 1996). English ivy clings to tree trunks by secreting a low-viscosity adhesive solution that spreads before water evaporates, hardening it into a matrix (Huang et al., 2016). However, the challenge is much greater for aquatic animals (Stewart et al., 2011). Barnacles and mussels solve the problem by secreting adhesives that are subsequently enzymatically hardened (Dickinson et al., 2009; Naldrett, 1993; So et al., 2016; Waite, 2017). By contrast, the glycoprotein (see Glossary) glue of an orb-weaving spider's viscous threads remains hydrated and pliable in air because it is contained in tiny aquatic spheres (Fig. 1D,E) (Edmonds and Vollrath, 1992; Tillinghast et al., 1993; Townley et al., 1991). This ensures that their glycoprotein adhesive retains its viscoelasticity for effective adhesion (Sahni et al., 2010).

Orb-weaving spiders integrate silk produced from four distinct silk glands into a highly effective prey capture web (Fig. 1A). Attached by pyriform gland (see Glossary) secretions (Sahni et al., 2012a; Wolff et al., 2015), non-adhesive radial and frame threads produced by major ampullate glands (see Glossary) absorb and dissipate the kinetic energy of an insect's impact (Sensenig et al., 2012), while spirally arrayed, adhesive prey capture threads produced from flagelliform and aggregate glands (see Glossary) retain the insect (Sahni et al., 2013) until the spider can locate, run to and begin to subdue it. Material invested in non-adhesive threads influences the size and velocity of insects that a web can stop (Sensenig et al., 2012), and material invested in capture thread affects the time an insect is trapped (Opell et al., 2017). A retention time (see Glossary) difference of even a few seconds can be the difference between a prey being captured or lost (Eberhard, 1989).

Orb-weaving spiders are not unique in relying on extended phenotypes (see Glossary) for important functions (Dawkins, 1982), nor are they the only animals that use these products for prey capture. For example, parchment worms and caddisfly larvae employ nets to filter organic material from the water (Flood and Fiala-Médioni, 1982; Mackay and Wiggins, 1979). Like other extended phenotypes (see Glossary), orb spider threads and webs exhibit physical and architectural plasticity (Blamires, 2010; Blamires et al., 2014, 2016, 2017; Crews and Opell, 2006; Herberstein and Tso, 2011; Scharf et al., 2011; Townley et al., 2006; Tso et al., 2007; Wu et al., 2013). However, viscous threads are unusual in that, after being spun, they continue to exhibit plasticity as they respond to environmental conditions, most notably relative humidity (RH) (see Glossary) (Agnarsson et al., 2009; Opell et al., 2011a, 2013; Sahni et al., 2011; Stellwagen et al., 2015a, 2014).

Temperature and ultraviolet light influence viscous thread properties and performance (Stellwagen et al., 2015b, 2016, 2014), although humidity has the greatest and most universal effect. As RH decreases during daylight hours, temperature increases, mediating the decrease in absolute humidity and reducing glycoprotein viscosity. However, species experience the impact of humidity differently. Orb weavers that live in exposed, weedy vegetation experience greater daily oscillations in humidity than those whose webs are anchored in vegetation that provides shade and helps maintain humidity (cf. Fig. 2A,B). Although the first study of the effect of humidity on viscous thread adhesion was published over 30 years ago (Strohmenger and Nentwig, 1987), renewed interest in this topic is revealing details about the impact of humidity on this complex and highly integrated natural adhesive system.

Understanding the response of viscous threads to environmental humidity is key to understanding both the function and evolution of this unique adhesive system. A viscous thread is a compound, self-assembling adhesive produced by two aggregate gland spigots flanking a conical flagelliform gland (see Glossary) spigot (Fig. 1B) on each of a spider's paired posterior lateral spinnerets, a total of six spigots contributing to each thread (Coddington, 1989; Park and Moon, 2014; Peters, 1955). As an axial line (see Glossary) emerges from the flagelliform spigot's tip, it is coated with aggregate gland solution that contains glycoprotein and small hygroscopic molecules (Townley and Tillinghast, 2013). The coated axial fibers (see Glossary) from each of the two spinnerets merge to form a single cylindrical thread, after which Plateau–Rayleigh instability (see Glossary) causes the aggregate material to quickly form a series of evenly spaced droplets that exhibit a bead on a string (BOAS) morphology (Fig. 1C,D) (Edmonds and Vollrath, 1992; Mead-Hunter et al., 2012; Roe, 1975). Environmental humidity affects the size of the droplets that form through its impact on the viscosity of the aggregate material (Edmonds and Vollrath, 1992; Sahni et al., 2012b). Studies of viscous thread analogs and droplet formation in thin films show that the velocity of thread production and the size and shape of nozzle apertures affect droplet spacing (Sadeghpour et al., 2017; Sahni et al., 2012b), principles worth examining in viscous thread spinning. At the center of each droplet a glycoprotein core coalesces (Fig. 1E) (Vollrath and Edmonds, 1989). Although this is the only droplet region where protein can be visualized under light microscopy, proteins are also found in the remaining aqueous material, which covers both the thread's supporting axial fibers and its glycoprotein cores (Amarpuri et al., 2015a).

Four droplet regions have been identified: (1) a thin outer lipid coat, first identified by Hans Peters (Peters, 1995) and seen as a ‘skin’ in scanning electron microscope images of desiccated droplets (Opell and Hendricks, 2009) but poorly studied; (2) the aqueous layer (see Glossary) containing proteins and the small molecules that are described in the following section; (3) a distinct glycoprotein core; and (4) a granule in the core's center, which is thought to anchor the core to the thread's flagelliform fibers (Opell and Hendricks, 2010). Both the glycoprotein core and its granule are most clearly seen when a droplet has been flattened on a microscope slide or coverslip. Epi-illumination more clearly reveals the glycoprotein core, whereas the granule is more easily seen with transmitted light, where it appears as a cylinder or toroid within the core (Opell and Hendricks, 2010). Consequently, in some older literature the granule is assumed to be responsible for thread adhesion. It is not known if the granule is simply a region of the glycoprotein that has become associated with flagelliform fibers or a distinct protein or proteins. Although droplets resist being moved along the axial fibers, they are not permanently bonded and can slide (Opell et al., 2011a, 2013).

Despite the large percentage of water in a droplet, the adhesion of its glycoprotein is several orders of magnitude greater than the capillary adhesion of its aqueous layer (Sahni et al., 2010). Only one thread glycoprotein, aggregate spider glue 2 or ASG2 has been characterized (Choresh et al., 2009; Collin et al., 2016; Vasanthavada et al., 2012), with ASG1 subsequently being associated with mucin proteins that bind chitin to cells (Collin et al., 2016). Collin et al. (2016) showed that ASG2 is a member of the spidroin gene family and suggested that, consistent with spidroin nomenclature, it be named aggregate spidroin 1 (AgSp1). Spidroins are a class of scleroproteins that includes major ampullate and flagelliform fibers (Ayoub et al., 2007; Garb et al., 2010, 2007; Gatesy et al., 2001). However, the presence of AgSp1 proteins in glue droplets has not been confirmed and we do not know whether AgSp1 is the only glycoprotein gene or if this type of protein is the only adhesive in a droplet. The challenge of adhering to an insect's waxy epicuticle is great, and our understanding of AgSp1's mode of adhesion is poor relative to that of other bioadhesives, such as mussel glue (Forooshani and Lee, 2017). Although glycoproteins are known to be adhesives (e.g. von der Mark and Sorokin, 2002; Xu and Mosher, 2011), until information about possible post-translational modifications of AgSp1 proteins and their three-dimensional structure is available, it will be difficult to determine their mode of adhesion.

Viscous glue droplets contain abundant amounts of water-soluble organic and inorganic compounds that are hygroscopic in nature. These are often termed low molecular mass compounds (LMMCs) (see Glossary). Most are organic and only 10–20% are inorganic compounds (Townley and Tillinghast, 2013). Organic LMMCs are small polar aliphatic compounds (mostly amine and sulphate based), such as alanine, choline, betaine, proline, glycine, taurine, GABamide, putrescine, N-acetyltaurine, N-acetylputrescine and isethionic acid (Fig. 3) (Anderson and Tillinghast, 1980; Tillinghast et al., 1987; Townley et al., 1991, 2012, 2006; Townley and Tillinghast, 2013; Vollrath et al., 1990). Inorganic LMMCs include H₂PO₄⁻, K⁺, NO₃⁻, Na⁺, Cl⁻ and Ca²⁺ moieties (Anderson and Tillinghast, 1980; Townley and Tillinghast, 2013; Townley et al., 2006; Vollrath et al., 1990).

The LMMCs are hypothesized to have evolved in part from neurotransmitters (Edmonds and Vollrath, 1992), but are now distributed throughout the aqueous material where they function to take up water from the environment and interact with glycoproteins to render the glue functional in different humidity conditions (Amarpuri et al., 2015b; Opell et al., 2013; Sahni et al., 2011, 2014; Townley and Tillinghast, 2013). Individual LMMCs differ widely in hygroscopic response. Compounds such as choline and N-acetyltaurine are hygroscopic over a range of humidity conditions; GABamide, N-acetylputrescine and isethionic acid start adsorbing at approximately 55% RH, whereas glycine, potassium nitrate and potassium dihydrogen phosphate show less than 3% water uptake by mass even at high humidity conditions (Townley et al., 1991; Vollrath et al., 1990). LMMCs differ in their types and compositions across orb-weaving species living in habitats with different humidity levels (Fig. 3). However, it is important to note that even among individuals of the same species, LMMCs composition differs and is presumed to be affected by a spider's genetics and diet (Higgins et al., 2001).

The primary function of the LMMCs is to solvate and soften glycoproteins to enhance adhesion. The LMMCs interact with the glycoproteins to make viscid glue functionally responsive to humidity in the environment. Pristine thread droplets swell as RH increases, whereas removal of the hygroscopic compounds by washing threads with water leads to the collapse of the glue structure and renders it incapable of subsequently taking up more than 10–20% water even at high humidity. After this collapse, it becomes impossible to reintroduce LMMCs back into the washed glue to recover adhesion, and at 100% RH washed threads lose two orders of magnitude of adhesion compared with pristine threads (Fig. 4A,B). In all conditions (0%, 40%, 100% RH or wet), washed glue droplets fail to make intimate contact and do not adhere to the surface (Sahni et al., 2014). Various solid-state nuclear magnetic resonance (NMR) spectroscopy techniques have shown that the glycoproteins soften and become humidity responsive in the presence of LMMCs. Cross-polarization magic-angle spinning (CPMAS) NMR is sensitive to rigid molecules and demonstrates that the rigidity of glycoproteins in pristine glue decreases as humidity is increased from 0% RH to 100% RH (indicated by the decrease in intensity of the spectrum in Fig. 4C). This directly correlates with macro-level observations of glue getting softer as humidity rises, resulting in intimate contact with surfaces and enhanced adhesion. When LMMCs are washed off, the viscid glue is irresponsive to humidity (Fig. 4D) and the glycoproteins become rigid, corresponding to the collapse of the glue at a macro level (Sahni et al., 2014). Altering LMMCs composition provides a mechanism by which natural selection can optimize viscous thread performance to the humidity in a species' environment.

Viscous droplet volume responds dramatically to changes in humidity (Fig. 5A) (Opell et al., 2011a, 2013). However, as we will explain, the degree of droplet hygroscopicity differs among species and is related to the humidity of a species' habitat. Glycoprotein volume also responds to humidity (Fig. 5C), documenting that, after atmospheric water enters a droplet's aqueous layer, some of it is absorbed by the glycoprotein core. This results in an increase in droplet extensibility as humidity increases (Fig. 5B). Even after extension is adjusted for glycoprotein volume, this response differs among species (Fig. 5D,E). Compared with the lower hygroscopic droplets of species such as Neoscona crucifera and Verrucosa arenata that occupy humid environments, the more hygroscopic droplets of Argiope aurantia and Larinioides cornutus do not extend as far at higher humidities before releasing because their glycoprotein more easily becomes over lubricated, dropping in viscosity and more easily releases from a surface (Fig. 5D) (Opell et al., 2013; Sahni et al., 2011). Thus, the viscosity of A. aurantia glycoprotein at 55% RH is similar to that of N. crucifera at 90% RH (Fig. 5D,E). Although the greater hygroscopicity of A. aurantia threads might appear to be a deficiency, it is, in fact, an adaptation to remaining hydrated during the late morning and afternoon hours when humidity is low (Fig. 2A).

The level of humidity at which adhesion of viscid glues reaches a maximum in different spider species corresponds to their foraging habitats (Fig. 6A). Maximum adhesion occurs when the viscosity of the glue is such that the contribution of two factors is optimized: surface interactions (substrate–glue interaction energy and spreading area); and bulk dissipation (rate of peeling and viscosity) (Amarpuri et al., 2015b). As RH increases, spreading of the droplets improves as bulk dissipation decreases (Fig. 6B). At low humidity, droplets are stiff and do not spread efficiently. As humidity increases, droplets spread and resist peeling as the glycoprotein extends, leading to generation of high adhesive forces. At high humidity, droplets coalesce to form a sheet of glue that spreads completely but breaks easily. These changes in behavior represent a remarkable 1000-fold variation in glue viscosity, but adhesion is maximized in a relatively narrow range of viscosity that optimizes spreading and bulk contributions (Fig. 6C). Remarkably, this optimal viscosity is achieved at very different humidities in different species that closely matches where each forages (Fig. 6A). Thus, the diverse mixture of LMMCs (Fig. 3) adapts species to a range of habitat humidities (Amarpuri et al., 2015b; Opell et al., 2015, 2013). In the next section, we explain why maintaining glycoprotein extensibility plays an important role in thread adhesion.

In the milliseconds after an insect strikes a web, a viscous capture thread's glycoprotein cores must spread immediately to establish adhesion and then, as the insect struggles to escape, instantly resist shifting forces that threaten to pull threads from the insect's body and wings. If the axial lines and droplets were rigid, force applied to a thread would cause the terminal droplets to release and initiate serial droplet pull-off that would quickly lead to thread release. Compared with cribellate thread, the plesiomorphic, dry prey capture threads spun by araneoid ancestors (Garrison et al., 2016), viscous thread is more effective in this regard. Cribellate threads are formed of several thousand dry protein nanofibers arrayed around support lines and can adhere by van der Waals forces, capillary attachment, snagging on insect setae (Joel et al., 2015; Opell, 2013) and can even embed their nanofibrils in the waxy outer epicuticle of an insect's exoskeleton (Bott et al., 2017). Although versatile, the adhesion of this thread is limited by the stiffness of its internal supporting fibers. Its adhesion does not increase as increasing lengths of thread contact a surface, indicating that, after the adhesion of terminal thread regions fails, crack propagation ensues, preventing additional adhesion being recruited from more central thread regions (Opell and Schwend, 2008).

In contrast, viscous thread adhesion increases as the thread contact length increases (Opell and Hendricks, 2007, 2009). The pliable adhesive droplets of viscous threads combine with the thread's extensible flagelliform support lines (Blackledge and Hayashi, 2006) to create a dynamic adhesive system that assumes the configuration of a ‘suspension bridge’ as it sums the adhesive forces of multiple droplets (Fig. 7). Moreover, as force is applied to a thread, the extension of its droplets and flagelliform lines combines to dissipate the energy of a struggling prey (Piorkowski and Blackledge, 2017; Sahni et al., 2011). Thus, there are two ways to characterize viscous thread adhesion: the force required to pull a thread from a surface (e.g. Opell and Hendricks, 2007, 2009); and the work of adhesion required to bring a thread to the point of pull-off (e.g. Sahni et al., 2011).

The thread's hygroscopic aqueous layer also makes an essential contribution to the suspension bridge mechanism (see Glossary) by ensuring that flagelliform fibers remain hydrated and extensible. When threads were stretched experimentally to reduce axial fiber extensibility, but the number of contributing droplets was maintained by contacting longer thread lengths, the force required to pull a thread from a surface decreased (Opell et al., 2008). Flagelliform fiber extension is also crucial for a thread's ability to dissipate the energy of a struggling insect (Sahni et al., 2011), contributing more than twice the work of adhesion as combined droplet extensions (Piorkowski and Blackledge, 2017).

Because viscous threads rely on the extensibility of both flagelliform fibers and the glycoprotein cores of droplets the performance of these two components must have evolved in a complementary fashion. If glycoprotein is too stiff relative to a thread's flagelliform fibers, the outer droplets of a contacting strand will release before inner droplets have extended and contributed their adhesive forces. If, by contrast, glycoprotein extensibility is too great, there will be little resistance and the axial line will bow acutely, with little work being done and little adhesive force being summed. This is borne out by a comparison of the Young's modulus (see Glossary) of three species' flagelliform fibers and glycoproteins. Young's modulus (E) is a measure of a material's stiffness, with smaller values indicating a material that is more easily extended. When compared at 50% RH, flagelliform E ranged from 0.009 to 0.0300 GPa and glycoprotein E from 0.00003 to 0.0014 GPa, with flagelliform E being 21, 52 and 290 times greater than glycoprotein E for the three species (B.D.O., M. E. Clouse and S. F. Andrews, unpublished; Sensenig et al., 2010).

As the studies of Tillinghast, Townley, Vollrath and their colleagues have shown (Edmonds and Vollrath, 1992; Townley et al., 1991, 2012, 2006; Townley and Tillinghast, 2013; Vollrath et al., 1990; Vollrath and Tillinghast, 1991), environmental humidity plays a crucial role in the function of an orb web from the time that it is constructed until it is taken down and its silk ingested. High humidity during the later evening and early morning hours when most orb webs are constructed affects the self-assembly of the glue droplets of viscous capture threads. Changes in humidity over the course of a day (Fig. 2A–C) affect the web's ability to both withstand prey impact (Boutry and Blackledge, 2013) and retain intercepted prey (Opell et al., 2017). Finally, when ingested the fully hydrated glue droplets supply a spider with both water and recyclable nutrients (Edmonds and Vollrath, 1992; Townley and Tillinghast, 1988). In fact, some important LMMCs like choline are also necessary for spider physiology and are in short supply, being obtained only from insect prey and ingested threads (Higgins and Rankin, 1999; Townley and Tillinghast, 2013; Townley et al., 2006).

As we gain a greater understanding of viscous thread hygroscopicity and fine-scale, humidity-mediated changes in viscous droplets, it is important to determine how these features impact prey retention time because this is ultimately how natural selection must tune thread performance to the humidity of a species' environment. However, assessing prey retention, particularly in vertically oriented orb webs like most of those that have been studied, is challenging. Retention is affected by many factors, including the mass of an insect and its impact velocity, the number of capture threads that it strikes, the texture of the insect's body region that contacts a thread, the region of the web a prey strikes and whether, after struggling free from these threads, the insect tumbles into other capture threads (Blackledge and Zevenbergen, 2006; Opell and Schwend, 2007; Sensenig et al., 2013; Zschokke and Nakata, 2015).

To make humidity the focal variable, an anesthetized housefly was placed wings downward across three, equally spaced, horizontal capture thread strands from the large orb weaver Araneus marmoreus (Fig. 2F) and its escape captured in a video recording (Opell et al., 2017). The humidity maximizing retention time of the flies was predicted to be the humidity at which both the surface area and extensibility of the glycoprotein were greatest (Fig. 2D). This occurred at 72% RH; the same level at which the energy estimated to bring a 4 mm span of capture thread to the initiation of pull-off was greatest and thus, most difficult for a prey to achieve (Fig. 2E). This humidity is also similar to the afternoon humidity at the forest edge where A. marmoreus lives (Fig. 2B). At 72% RH, actively struggling flies were retained 11 s longer than at either 37% or 55% RH (Fig. 2F). This additional time is ecologically significant because it provides a spider more time to locate and reach an insect and to begin wrapping it with silk from numerous aciniform gland (see Glossary) spigots on the posterior median and posterior lateral spinnerets (Coddington, 1989; Tremblay et al., 2015) before the prey can escape the web.

Greater retention times also relate directly to the size of insects that a web can retain. For large orb weavers such as A. marmoreus, it is postulated that these large, rare prey are more profitable and comprise the greatest proportion of a spider's total food intake (Blackledge, 2011; Venner and Casas, 2005) but see Eberhard (Eberhard, 2013) for challenges to this hypothesis. Thus, there is solid evidence that longer prey retention time selects for changes in the composition of a viscous thread's hygroscopic compounds that tune thread performance to the humidity of a species' habitat. These findings are the first step in ascribing fitness values to the performance characteristics of viscous threads. As data for other species are added, it should be possible to rank the relative contributions of glycoprotein surface area, viscosity and extension to prey retention time.

Humidity poses serious problems to the stability of adhesive joints (Abdel Wahab, 2012; Brewis et al., 1990; Petrie, 2007; Tan et al., 2008; White et al., 2005). Most of the synthetic adhesives fail when a crucial RH is exceeded (Petrie, 2007; Tan et al., 2008). Therefore, it would be desirable to have synthetic adhesives that can either resist changes in RH and continue to strongly bind surfaces or respond with humidity similar to viscid silk. The unique natural designs of both cribellate and viscous prey capture threads have inspired researchers to develop similarly structured materials for a variety of applications, including adhesives, water collectors and solid–liquid hybrid materials (Bai et al., 2012; Chen and Zheng, 2014; Elettro et al., 2016; Sahni et al., 2012b; Song et al., 2014; Tian et al., 2011). In one of the first attempts, synthetic adhesive BOAS microthreads were fabricated by drawing a synthetic nylon thread through a pool of polydimethylsiloxane (PDMS) polymer (Sahni et al., 2012b). The process created a cylindrical coating that formed smaller droplets due to Plateau–Rayleigh instability and these threads were sticky when tested on a glass substrate (Fig. 8). The spacing and diameter of these synthetic thread droplets were varied by changing the capillary number (Ca=velocity×viscosity/surface tension), which depends on drawing velocity, PDMS viscosity and surface tension (Fig. 8A–C). A higher capillary number (higher velocity, higher viscosity and lower surface tension) produced larger and more widely spaced droplets (Fig. 8C), which exhibited greater adhesion (Fig. 8E). The study presented a simple and effective manner of creating BOAS adhesive mimics of viscous threads (Fig. 8D) and also helped in testing the fundamental principles behind the adhesion of viscid silk by using synthetic mimics (Sahni et al., 2012b). This successful strategy can also be used to generate humidity-responsive adhesives. For example, droplets can be laden with mixtures of LMMCs mimicking natural compositions (Fig. 3) incorporated within polymer matrices to generate viscous thread to synthesize humidity-sensitive adhesives. These synthetic adhesive structures can then be used in applications such as a bandages or adhesive tapes where adhesion is crucial in the presence of water.

Viscous thread adhesion relies heavily on water for both effective spreading of the adhesive glycoproteins and elasticity of the underlying axial thread. Water content also influences the Plateau–Rayleigh instability that determines the final size and spacing of glue droplets. These features act synergistically to generate substantial adhesion as viscous threads deform in a suspension bridge-like pattern while detaching from a variety of surfaces. Some of this water can be obtained directly from the atmosphere when threads are first spun, potentially resulting in a net gain of water by a spider when an orb web is taken down and its silk ingested. Most orb webs are spun under humid conditions, in the late evening or early morning, so that minimal hygroscopicity is likely to be necessary for droplet formation and adhesion (Blackledge et al., 2009a). However, we hypothesize that increased thread hygroscopicity was necessary to optimize thread adhesion as orb weavers diversified to occupy habitats where humidity drops during the course of a day. Thus, natural selection tuned the composition of LMMCs in a droplet's outer aqueous layer to meet this challenge (Townley and Tillinghast, 2013) and to maintain glycoprotein structure and enhance its surface interactions (Liao et al., 2015). However, this is largely based on investigation of a few temperate species of spiders and three key questions remain about viscid thread hygroscopicity. First, what about species in consistently arid or humid habitats such as deserts and rainforests? Do their glues perform similarly or show distinct LMMCs compositions? Second, can individual spiders control LMMCs composition physiologically to tailor thread structure and adhesion under different physiological conditions? Finally, did the hygroscopicity system arise to help spiders conserve water resources after viscid glue was already being produced (e.g. the ancestral condition was for orb spiders to exude wet sticky secretions from their aggregate glands) or as a mechanism to improve adhesion (Opell et al., 2011b; Piorkowski and Blackledge, 2017) with spiders adding LMMCs to dry adhesive secretions for some other functional benefit?

Our current model of the evolution of viscous thread environmental responsiveness relies entirely on describing variation in LMMCs composition. The amino acid sequence of only one glycoprotein has been characterized and details of this molecule's three-dimensional structure and adhesion are not well understood. Thus, the model we present here is clearly an oversimplified view. For instance, how much of the variation in the environmental responsiveness of different species' glue is explained by interactions between LMMCs and variation in glycoprotein sequence? Future investigation should also focus on understanding how LMMCs directly interact the glycoproteins to plasticize them and how this influences adhesion. Indeed, selection for optimal glycoprotein secondary structure may be as important as selection for optimal aqueous layer hygroscopicity.

The use of LMMCs to recruit water and control the self-organization of a hierarchically structured adhesive thread is simple in concept and therefore translatable to synthetic models. However, we still do not understand the specific functions of individual LMMCs and the mechanisms by which they plasticize the adhesive glycoproteins. In addition to optimizing the performance of synthetic adhesives, such research will also provide a powerful tool to test hypotheses about specific aspects of viscous thread function and spider web evolution.

We are grateful to two reviewers whose comments and suggestions allowed us to improve the clarity and completeness of this Review.