Spider silk colour covaries with thermal properties but not protein structure

Abstract

Understanding how and why animal secretions vary in property has important biomimetic implications as desirable properties might covary. Spider major ampullate (MA) silk, for instance, is a secretion earmarked for biomimetic applications, but many of its properties vary among and between species across environments. Here, we tested the hypothesis that MA silk colour, protein structure and thermal properties covary when protein uptake is manipulated in the spider Trichonephila plumipes. We collected silk from adult female spiders maintained on a protein-fed or protein-deprived diet. Based on spectrophotometric quantifications, we classified half the silks as ‘bee visible’ and the other half ‘bee invisible’. Wide angle X-ray diffraction and differential scanning calorimetry were then used to assess the silk's protein structure and thermal properties, respectively. We found that although protein structures and thermal properties varied across our treatments only the thermal properties covaried with colour. This ultimately suggests that protein structure alone is not responsible for MA silk thermal properties, nor does it affect silk colours. We speculate that similar ecological factors act on silk colour and thermal properties, which should be uncovered to inform biomimetic programmes.

1. Introduction

Biomimetics is an emergent scientific endeavour wherein researchers explore the properties of natural phenomena as inspiration for new synthetic designs and developments [1]. For instance, the mechanisms enabling animals to vary the colour of secretions such as hair, feathers, mucous and silk over time or across environments intrigues biomimetics research as the underlying principles inspire the development of innovative new coloured materials [2–6]. Likewise, the thermal properties of animal secretions may inspire the creation of new heat and light sensing materials [7].

Spider major ampullate (MA) silk is considered one of nature's toughest materials [8]. It thus comes as no surprise that many researchers have attempted to understand the intricacies of its structure–function relationship with a vision to create silk biomimetic products for medical and engineering applications (see reviews by Blamires et al. [8], Eisoldt et al. [9], Hsia et al. [10] and Ebrahimi et al. [11]). Consequently, we know that MA silk is hierarchically organized with a lipid, protein and glycoprotein-rich skin around a fibrous outer and inner core [8,12,13]. The silk core is traditionally thought to be composed of two proteins called spidroins: MA spidroin 1, or MaSp1, and MA spidroin 2, or MaSp2 (but see Babb et al. [14] and Correa-Garhwal et al. [15] for the potential inclusion of additional spidroins). MaSp1 consists of repeating polyalanine, (GA)_n, (GGX)_n and (A)_n amino acid motifs (G = glycine, A = alanine and X = other amino acids) that are expected to combine to promote the formation of β-sheet secondary structures in the assembled fibres [16,17]. The MaSp2 protein on the other hand is thought to consist of additional multiple (GPGXX)_n motifs (where P = proline) and has been predicted to form the so-called ‘amorphous region’, consisting of disordered type II β-turns and other secondary structures [18,19]. These secondary structures form a range of crystalline and amorphous protein tertiary structures, which have been considered the basis for spider silk's remarkable mechanical performance [8,13,20].

In addition to its impressive mechanical properties, spider MA silk has inimitable optical and thermal properties [21–23]. Understanding silk optical properties may inspire biomimetic applications such as the development of new highly efficient lenses, sensors or optic fibres [24–26]. It thus provides for a fruitful line of research. We know that MA silk has a high refractive index despite being largely translucent [21,22,27], which explains why it is visible to the naked human eye even though it is exceptionally thin (diameter: approx. 1–5 µm). We also know that the visibility, coloration and brightness of MA silk varies across environments, both among spider species and between individuals [28–31].

The skin, which primarily functions to prevent fibre biodegradation [32], is consistently smooth and homogeneous across species and individuals [33,34] so surface features are unlikely to induce inter- or intraspecific variation in coloration. Why MA silk should vary in colour among and between species and individuals is thus unknown. Nevertheless, there are some potential explanations. First, the shape and thickness of the skin, the appearance of cracks or microfibrils, twisting and accumulation of particulate debris may affect some of the silk's optical properties [26,35,36]. Second, pigments and other compounds, such as phenols, porphyrins, quinones and carotenoids, may reside within the skin layer and are thought to be responsible for some of the species-specific silk coloration variation [6,30,37,38]. Third, the β-sheet secondary structures and crystalline tertiary structures of MA silk are anisotropic, thus highly birefringent [39,40]. Given that the size, density and alignment of β-sheets can vary across spider species [41], this may contribute to variation in MA silk optical properties across species. Fourth and finally, the thermal properties of MA silk, particularly the silk's enthalpy (a measure of the total internal energy in the silk thread), glass transition temperature (T_g) and melt temperature, are hypothesized to covary with the size and alignment of the silk's crystalline structures and, thus, correlate with its birefringent properties [42,43]. The interactions between thermal properties and protein structure are observably manifested upon exposure to water as a downward shift in T_g due to changes in structural alignment [44].

While it might be deduced that there should be evolutionary and ecological consequences for environmentally induced variation in silk coloration across species and/or individuals, it is not readily apparent what these may be. It has been deduced that more colourful or visible silks might alert prey to the presence of the web or predators to the presence of the spider [45]. Accordingly, MA silk should always be inconspicuous against its background for the spider to effectively capture prey and avoid predators. In some instances, nonetheless, MA silk seems to reflect light at wavelengths selectively attractive to certain prey insects [28,46,47]. The consequent fitness benefit of such selective visibility is greater prey capture success, which may provide an evolutionary explanation for the conspicuous yellow (to the human eye) coloration of the silks of Nephila or Trichonephila spp. [29].

Here, we investigated how a spider's diet relates to silk colour and the underlying structural mechanisms and covarying properties. We know that the protein composition of the spider's diet can, to a certain extent, influence spider silk coloration [48]. Whether this is a product of a change in pigment deposition as different nutrients are made available, or a consequence of changes in birefringence accompanying the changes in protein structural properties associated with such dietary shifts [41,49] are unknown. If variation in silk colour is a consequence of changes to protein structures, then both the structures responsible and thermal properties should covary with coloration across diets. We accordingly tested our prediction by quantifying MA silk coloration (as spectral reflectance functions), crystalline secondary structures and thermal properties in the spider Trichonephila plumipes (formerly Nephila plumipes, see Kuntner et al. [50] for nomenclature).

2. Material and methods

2.1. Spiders

We used adult female T. plumipes collected from suburban Sydney, Australia for the following experiments. We measured their body length and width to ±0.1 mm using digital Vernier calipers (Caliper Technologies Corp., Mountain View, CA, USA) and mass to ±0.001 g using an electronic balance (Ohaus Corp., Pine Brook, NY, USA). Using these data, we took 30 individuals of similar mass that appeared not to be gravid to the laboratory at the University of New South Wales, Sydney, for the following procedures.

2.2. Experimental feeding regimes

Spiders were initially placed in 115 mm (wide) × 45 mm (high) plastic circular containers with perforated wire mesh lids, as described by Blamires et al. [41,48,49], and fed a 30% weight per volume concentration (w/v) glucose solution daily over 5 days. This feeding regime ensured that all influences of recent diet on protein structure or silk coloration were minimized [48]. We reweighed all of the spiders after the fifth day of feeding and removed one individual that lost excessive mass after constructing an eggsac. The remaining spiders were randomly allocated into two treatments: protein-fed (P) or protein-deprived (N).

The solution used for the protein-fed treatment was identical to that described by Blamires et al. [41,49]: a mixture of 10 g of a 10% albumin solution with 6 g of sucrose in 60 ml of water. The solution for the protein-deprived treatment was 8 g of sucrose in 30 ml of water. We fed the spiders the solutions by placing a 20 µl droplet onto their chelicerae using a micropipette. The spiders were fed the respective treatments over 10 days. After completing the experiment, we re-weighed all of the spiders and excluded any that lost excessive mass and/or built an eggsac (i.e. one from each treatment).

The above experiment was performed under controlled temperature (approx. 25°C) and humidity (approx. 50% RH) in still air.

2.3. Silk collection

Upon completing the feeding experiments, we collected MA silk from all of the 26 remaining spiders (13 per treatment) for spectrophotometry, wide-angle X-ray diffraction (WAXS) and differential scanning calorimetry (DSC) analyses as follows. We firstly anaesthetized all of the spiders using carbon dioxide and placed them ventral side up on a foam platform and immobilized them using non-adhesive tape and pins. We then carefully collected a single MA silk thread from the spinnerets. This procedure was performed under a dissecting microscope to ensure that a single thread from one spinneret was collected. An electronic spool rotating at 1 m min⁻¹ was used to reel silk threads from each spider by two different collection methods.

First, to collect the silk for both spectrophotometry and WAXS a single MA silk thread was pulled across 3 mm × 1 mm titanium frames containing 0.5 mm × 0.5 mm windows. We then spooled the silk for 1 h, which resulted in approximately 1000 rounds being collected across the frame windows (see Benamú et al. [51] for details). Second, the silk was wrapped around a glass tube connected to the rotating spool for another 1 h. This resulted in approximately 15 mg of thread being wound onto the glass tube. The thread was scraped off the tube using a scalpel and bundled and placed into a 2 ml Eppendorf tube.

2.4. Spectrophotometry

We measured the reflectance spectra over the 300–700 nm waveband for the frame mounted silks using a Jaz spectrophotometer (Ocean Optics Inc., Largo FL, USA). We used the frame mounted silks for this process as each of the rounds of thread were in parallel, thus thread orientation could not interfere with our spectrophotometric measurements. The spectrophotometer was connected to a QR200-7UV-VIS reflection probe, held perpendicular to the sample, and a laptop running the program OceanView 1.6.3 (Ocean Optics, Inc.) via a series of SMA 905 optical fibres (to 0.22 numerical aperture), a reading fibre and a halogen light source (DT-1000, Ocean Optics, Inc.). Labsphere certified white (SRS-99-010) and black (SRS-02-010) reflectance standards [52] were used to calibrate the spectrometer to 100% and 0% reflectance, respectively. The area captured in all instances was 2 mm², and the integration time was 150 ms. Three repeated measurements for each silk bundle and the standards were taken in a dark room against a black cardboard background (corroborated against the Labsphere black standard). We subsequently plotted reflectance curves representing an average of the three repeated measurements per individual spider.

To interpret the reflectance spectra attained for the silks in a biologically relevant context (i.e. how they were viewed by prey), we performed visual models to ascertain whether any of the silks produced were viewed differently by honeybees, a common prey for T. plumipes in Sydney. We accordingly calculated the relative quantum flux absorbed by honeybee photoreceptors (P) across the 300–700 nm waveband using the following equation:

$$P=R{\int }_{300}^{700}{I}_S(\lambda )S(\lambda )D(\lambda )\mathrm{d}\lambda ,$$ 2.1

where I_S(λ) is the spectral reflectance function of the silk, S(λ) is the combined spectral sensitivity function of the six photoreceptors within a honeybee ommatidium [53], and D(λ) is the daylight spectrum (CIE Standard Illuminant D65 [54]). The sensitivity factor (R) is determined by the following equation:

$$\begin{array}{rl}R=\frac{1}{{\int }_{300}^{700}{I}_B(\lambda )S(\lambda )D(\lambda )\hspace{0.17em}\mathrm{d}\lambda },& \\ & \end{array}$$ 2.2

where I_B(λ) is the spectral reflection function of a typical vegetation background ascertained according to Blamires et al. [55]. We assumed that the honeybee photoreceptors were adapted to background coloration and that the photoreceptors displayed half their maximum response at all times.

The excitation (E) index of honeybee UV, blue and green photoreceptors while viewing each silk against a green background were subsequently estimated from their quantum flux values (P), using the equation below, derived from the colour hexagon model of Chittka [56]:

$$E=\frac{P}{P+1}.$$ 2.3

We also estimated the discriminability of the silks by a honeybee viewer using the receptor-noise limited model, assuming a Weber fraction of 0.13, log-transformed quantum catches and a relative UV: blue: green receptor density of 1: 0.47: 4.41 [57,58]. Our results were qualitatively unchanged from those detailed below for the hexagon model, with ‘white’ silks estimated to be indiscriminable (t₁₀ = −1.25, p = 0.881) and ‘yellow’ silks discriminable (t₁₃ = 2.34, p = <0.001) considering a threshold of one just noticeable distance. Estimated colour-distances were also highly correlated between hexagon and receptor-noise models (r = 0.98). We thus refer to the results of the hexagon model henceforth, since it represents the most rigorously validated model for honeybee viewers [56,59]. We performed the visual models using the R package ‘pavo’ v. 2.2 [60].

2.5. Wide-angle X-ray diffraction analysis

WAXS provides detailed information on silk protein structures across secondary and tertiary levels. We accordingly measured the size, density and orientation of the crystalline and amorphous structures of the frame-mounted silks using WAXS at the SAXS/WAXS beamline at Australian Synchrotron, Melbourne, Australia. We taped each of the silk-containing frames to a sample plate supplied by the Centre. The plates were mounted onto a holdfast a distance of 330 mm away from the incident X-ray beam. The beam size was confined by a collimator 0.5 mm in diameter. A digital camera was set up enabling us to move the specimens into the beam line from outside the experimental hutch. We exposed each silk sample to the beam for 10–60 s depending on its density. The radiation scattered from each sample was detected by a Mar 165 imaging plate over a Q range of approximately 1.45 Å. Two-dimensional WAXS images were developed using the program Scatterbrain (Australian Synchrotron, Melbourne, Australia), from which the: (i) scattering parameter (q), (ii) diffraction angles (2θ), (iii) azimuthal angles, (iv) intensity peaks (I_x) and (v) full width and half width maximum intensities of the 2θ and azimuthal angles were calculated. These parameters allowed us to thereupon calculate: (i) d-spacing (or relative distances between pleated β-sheet) among the crystalline structures, (ii) relative crystalline intensity ratios (I₀₂₀/I_amorphous and I₂₁₀/I_amorphous) with I₀₂₀, I₂₁₀ and I_amorphous, representing the sum of the scattering intensity at the (0 2 0) and (2 1 0) Bragg reflections and the amorphous halo, respectively, (iii) crystallinity index (X_c), representing an estimate of size of β-sheet crystallites and (iv) Herman's orientation function (f_c). Details pertaining to the application, analytical procedures and equations used in the WAXS analyses are found elsewhere [41,49,51].

2.6. Differential scanning calorimetry

DSC detects the heat flow into or out of biological polymeric materials, such as silks undergoing phase transitions [61,62]. It has thus been used to measure the enthalpy, degradation, crystallization, melt temperatures and T_g of silks (see [62–64] and the electronic supplementary material). While not a direct measure of protein structure, it has been used by some researchers to infer protein secondary and tertiary structures in silk [63–65]. We nevertheless used DSC herein to determine the parameters: (i) melt temperature, (ii) T_g and (iii) total internal enthalpy of the silks collected on glass tubes. The full procedures are outlined in the electronic supplementary material.

Each silk sample was placed individually within Tzero aluminium DSC pans with hermetic lids sealed with a Tzero press (TA Instruments, Austin, MN, USA) at the Centre for Advanced Macromolecular Design (CAMD), Faculty of Engineering, University of New South Wales. Conventional mode DSC was performed using a Q20 differential scanning calorimeter (TA Instruments, Austin, MN, USA). The rate of heating was set to 3°C min⁻¹ over the −50 to +200°C temperature range and modulation amplitude was set at 1°C with a period of 60 s. A blank pan (i.e. one without a sample added) was prepared as a reference and simultaneously added to the calorimeter along with the sample pans. We thereupon estimated T_g, melt temperature and enthalpy, as described in the electronic supplementary material.

2.7. Statistical analyses

We used one sample t-tests to compare the colour contrast values, calculated in hexagon units, as the Euclidean distances between the summed excitation indices for the calculated honeybee UV, blue and green photoreceptors over colour space, with a discrimination threshold value of 0.10, i.e. we assumed differentially conditioned bees to always view the silks [59,66]. We also calculated the contrasts between the green photoreceptor excitation values and the vegetation background to estimate the achromatic contrast values and compared them with the 0.10 discrimination threshold [67,68]. Where an individual silk exceeded the chromatic or achromatic discrimination threshold we considered them to be ‘bee visible’, whereas those that did not exceed any of the discrimination thresholds were considered ‘bee invisible’.

We used discriminant function and least significant difference post hoc analyses to determine whether the: (i) d-spacing, crystalline intensity ratios, crystallinity indices and/or Herman's orientation functions and (ii) degradation temperature, T_g, and enthalpy varied between the protein-fed and protein-deprived treatments. We used a categorical Pearson's χ²-test to ascertain whether the proportion of silks that we classified as bee visible and bee invisible differed across treatments. Prior to performing these analyses we checked the data for normality, linearity and homoscedasticity using quantile–quantile and probability plots.

To determine whether changes to silk protein structures and thermal properties covaried with silk coloration across diets we plotted redundancy analysis (RDA) ordinations [69] of protein structures and thermal properties against silk visibility (classified as bee visible or bee invisible as explained above). The individual WAXS and DSC parameters were collapsed to first redundancy variables for the RDA. Where ordinations showed across parameter differentiation between the bee visible and bee invisible silks we interpreted this as indicating that silk coloration was associated with its protein structures and/or thermal properties.

3. Results

The spectral reflectance functions for the MA silks of all individual spiders were highly variable across both of the treatments (figure 1). The mean chromatic and achromatic contrast of silks from each of the feeding treatments fell below the detection threshold for (${\chi }_{23}^2=0.110$; p = 0.597; see mean values in figure 2a,b). Nevertheless upon closer examining the data points in figure 2, it is evident that approximately half of all silks (i.e. 14 of 26) could be classifiable as bee visible and the others (12 of 26) could be classifiable as bee invisible based on their absolute colour contrast values, independent of the treatment (see data points in figure 2a,b). Representative reflectance spectra for each of these silks are shown in figure 3a. The silks that were classified as bee visible appeared yellow or golden to the naked human eye, while those classified as bee invisible appeared white to silver to the naked eye (figure 3b).

Figure 1.

Reflectance spectra of the silks from the P-treatment spiders (blue curves) and N-treatment (red curves) spiders. (Online version in colour.)

Figure 2.

Scatter and means-error plots of chromatic (a) and achromatic (b) contrast values of silks from both N and P treatments. Yellow data points of the scatter plots represent measurements for each silk classified as bee visible and the silver data points represent those for silks classified as bee invisible. Dots in the means-error plots represent mean values and the whiskers represent ±1 s.e. The dotted line at 0.10 in each plot shows the threshold value for classification as bee visible. The coloured dots in the figure refer to the colours of each silk as seen by the human eye. (Online version in colour.)

Figure 3.

(a) Reflectance spectra of the bee visible silks (yellow line) and bee invisible silks (silver line). (b) Examples of bee visible silk (the yellow/golden silk on the left) and bee invisible silk (the white/silver silk on the left) as they appeared to the human eye when wound around the titanium frames in preparation for spectrophotometry and WAXS. (Online version in colour.)

Our discriminant function analyses revealed that the WAXS derived d-spacing, I₂₁₀/I_amorphous and Herman's orientation functions differed between the silks of the protein-deprived and protein-fed spiders (tables 1 and 2), as did the DSC derived degradation temperature, T_g, and enthalpy (tables 3 and 4; see also electronic supplementary material, figures S1 and S2, respectively, for WAXS images and heat flow versus temperature thermographs across treatments).

Table 1.

Results of a discriminant function and least significant difference post hoc analyses determining whether silk crystal sizes, crystalline intensity ratios, crystallinity indices and Herman's orientation functions varied between the four species. Comparisons of means (±1 s.e.) across treatments. Significance of p-values is indicated by asterisk.

	d-spacing (nm)	I020/Iamorphous	I210/Iamorphous	crystallinity index	Herman's orientation
P-treatment	0.710 ± 0.161	1.223 ± 0.005	1.011 ± 0.030	1.342 ± 0.048	0.235 ± 0.205
N-treatment	0.592 ± 0.092	1.222 ± 0.004	1.059 ± 0.018	1.375 ± 0.072	0.132 ± 0.102
Wilk's lambda	0.101	0.088	0.141	0.091	0.138
partial lambda	0.738	0.897	0.561	0.867	0.572
F-remove (d.f. = 1,24)	3.598	1.481	10.577	1.983	9.692
p-value	0.021*	0.234	<0.001*	0.132	<0.001*
tolerance	0.792	0.269	0.238	0.672	0.584
1-tolerance	0.208	0.731	0.762	0.328	0.416

*p < 0.05.

Table 2.

Results of a discriminant function and least significant difference post hoc analyses determining whether silk crystal sizes, crystalline intensity ratios, crystallinity indices and Herman's orientation functions varied between the four species. Raw coefficients of the canonical variables.

roots removed	eigenvalue	canonical R	Wilk's lambda	χ2	d.f.	p-value
0	3.567	0.883	0.079	105.093	15	<0.001
1	1.719	0.795	0.363	42.052	8	<0.001
2	0.012	0.113	0.987	0.534	3	0.911

Table 3.

Results of a discriminant function and least significant difference post hoc analyses to determine whether melt temperature, glass transition temperature and enthalpy varied between the four species. Comparisons of means (±1 s.e.) across the four species. Significance of p-values is indicated by asterisk.

	melt temperature (°C)	glass transition temperature (°C)	enthalpy (J g−1)
P-treatment	193.56 ± 4.36	147.93 ± 1.64	229.54 ± 94.80
N-treatment	196.28 ± 3.55	152.34 ± 1.11	258.10 ± 86.81
Wilk's lambda	0.006	0.001	0.007
partial lambda	0.252	0.143	0.229
F-remove (d.f. = 1,24)	43.913	79.433	45.921
p-value	<0.001*	<0.001*	<0.001*
tolerance	0.821	0.917	0.884
1-tolerance	0.179	0.083	0.116

*p < 0.05.

Table 4.

Results of a discriminant function and least significant difference post hoc analyses to determine whether melt temperature, glass transition temperature and enthalpy varied between the four species. Raw coefficients of the canonical variables.

roots removed	eigenvalue	canonical R	Wilk's lambda	χ2	d.f.	p-value
0	373.879	0.998	<0.001	369.192	9	<0.001
1	12.941	0.963	0.063	117.311	4	<0.001
2	0.1336	0.343	0.882	5.329	1	0.021

Our RDA ordinations nevertheless showed that while silk colour explained approximately 53% of the variance in protein structures, it was not associated with any of the structural parameters we measured using WAXS (figure 4a). On the other hand, silk colour explained approximately 95% of the thermal property variance, and our ordinations (figure 4b) and the DSC thermographs (electronic supplementary material, figure S3) showed that the degradation temperature and T_g of the bee visible silks were distinctly different from those of the bee invisible silks.

Figure 4.

RDA ordinations of the association between the variance in silk colour and variance in (a) structural parameters measured using WAXS and (b) thermal property parameters measured using DSC. (Online version in colour.)

4. Discussion

We deprived the spider Trichonephila plumipes of protein, and performed spectrophotometry, WAXS and DSC analyses on its MA silks to investigate whether the changes in silk protein structures and/or thermal properties were associated with any changes in silk coloration. We found that while the protein structures and thermal properties varied between treatments, only the thermal properties were associated with silk colour, and that this association occurred irrespective of protein intake. We thus suggest that protein structures alone are not responsible for MA silk thermal properties and they do not drive silk colour.

About half of all of the silks we examined across both treatments could be classified as bee visible and the other were classifiable as bee invisible. Moreover, the bee visible silks always appeared yellow/golden to the human eye, while the bee invisible silks always appeared white/silver to the human eye. We would like to note that in the field we almost always observe spiders in webs comprised of yellow silk. Nevertheless, the fact that we had substantial colour variation in the silks collected in the laboratory (yellow to white) allowed us to compare the characteristics of the different coloured silks more closely and make two broad conclusions about the colour of T. plumipes silk. Firstly, the yellow and white colours are not related to the spider's protein intake, at least under the regimes tested herein. Secondly, silk thermal properties but not protein structure correlated with silk colour. The ecological implications for silk coloration variation across species/individuals are not known, partly because the colour variation remains, until now, unquantified. Hence our findings provide valuable insights into the mechanisms and consequences of silk coloration variation.

Our results suggest that changes in silk protein structures do not affect silk colour via variation in the silk's refractivity or birefringence. We thus expect the most likely cause of the yellow or white coloration observed in T. plumipes silk to be pigment deposition, as has been speculated for Trichonephila clavipes [6,37]. We do not currently know which pigments are present in T. plumipes MA silks. Putthanarat et al. [37] speculated that depositions of hydroxylated benzoquinone, naphthaquinone and other quinones within the silk skin are responsible for the yellow coloration in T. clavipes silks. Hsiung et al. [6] on the other hand found evidence that β-carotene deposition was responsible for yellow coloration in Nephila pilipes MA silk.

Various carotenoids induce orange and yellow coloration in moth cocoon silks [70–73]. Since spiders cannot synthesize β-carotene de novo, the pigments need to be taken up via the spider's diet [6,73]. Our finding that half of the silks we retrieved were yellow irrespective of protein intake suggested that the yellow coloration of T. plumipes silk is unlikely to be due to carotinoid intake and deposition within the silk skin. We nonetheless occasionally noticed that the yellow-coloured silk became white toward the end of silk collecting, suggesting that the spiders might have been depleted of a pigment, or other colour inducing compound after approximately 2 h of being forcibly silked. Ongoing analyses using nuclear magnetic resonance, Fourier transform infrared spectroscopy and pigment transcript analyses should, in time, confirm which pigments, if any, induce the yellow/golden colour of T. plumipes MA silks.

An examination of figure 1 shows a more marked modulation at approximately 470–500 nm, i.e. within the blue to green region, in the reflectance spectra of the silks of spiders from the protein-fed treatment (blue curves) compared to those of spiders from the protein-deprived treatment (red curves). The modulations were however not manifested, as both of our visual models found that the protein-fed treatment spider silks were not any more discriminable by honeybees than the protein-deprived treatment spider silks. This was likely a consequence of the green background against which the silks were contrasted also having a modulation in the green region of the spectrum. We cannot say from examining reflectance spectra alone whether these modulations indicate the presence of pigments, such as porphyrins, variation within the surface or structural features, or are an anomaly. We nevertheless expect that the presence of a pigment is unlikely because porphyrin, or another other green pigment, would need to be consumed in order to be expressed in silk [74].

The protein structural properties, particularly the orientation and alignment of crystalline and amorphous secondary structures, are thought to influence the thermal properties of MA silk [36,63,75]. This likely explains why both protein structures and thermal properties varied between our feeding treatments. Nevertheless, a lack of any significant association between T. plumipes silk's protein structures, thermal properties and coloration suggests that factors other than protein structures are affecting the thermal properties and coloration. We do not know why the thermal properties and coloration of T. plumipes MA silk covaried independent of protein structures, but we suspect a combination of changes within the structural integrity and composition of the skin, including any pigment deposition, to be responsible.

Water infiltrates hydrogen bonds between amorphous region proteins in MA silk, causing the crystalline and amorphous region proteins to become misaligned along the fibre's axis without changes in density [76,77]. A change in crystalline alignment alone may be enough to significantly shift T_g so the silk becomes rubbery at room temperature (20–30°C) [36,38,76,77]. Water is therefore an agent that may simultaneously affect MA silk thermal properties and protein structure. However, our experiments were performed under controlled temperature (approx. 25°C) and humidity (approx. 50% RH) conditions so water infiltration probably had a minor effect on the protein structures and/or thermal properties measured herein. It might nonetheless explain why we calculated slightly lower T_g values than that deduced for other spider silks.

We found differences in the thermal properties of T. plumipes silks between treatments and among the different coloured silks as a consequence of significant variation in heat flow across treatments and the different coloured silks. An examination of our DSC thermographs revealed that there were no major shifts in topography between treatments or across the different coloured silks (i.e. they always showed an endothermic response across the temperature range), and T_g was consistent at approximately 140–150°C, with melt temperature peaks always observed at greater than 160°C, and thermal degradation consistently commenced at around 180–200°C (see electronic supplementary material, figures S2 and S3). These values approximate the ranges reported for other MA silks [63,75], with the exception of T_g which was marginally lower. We also found that silk crystallinity did not vary greatly between feeding treatments, or among the different coloured silks. It seems therefore evident that there was no major phase transition or crystalline region breakdown across treatments or among the different coloured silks, ruling out further water infiltration as a cause of thermal property differences across the bee visible and bee invisible silks.

An additional possibility is that the two distinct silk colours that we identified (bee visible and bee invisible) are unique to the assumed honeybee viewer which, while appropriate given our knowledge of prey composition for our focal spiders, does not capture the full diversity of potential viewers nor viewing conditions. Furthermore, we measured the spectra of silks herein by wrapping them tightly around metal cards, meaning the silks appeared different from what would be expected in a web, where they are distributed more sparsely and orientated in a multitude of different directions. With that said, the models used and sample preparations also might affect the colour classifications of the silk based on its visibility to different insects. More testing is clearly required to ascertain the visibilities of the silks to different prey or predators. Notwithstanding, there are difficulties ascribing an ecological function to any estimated colour classifications because the light weight, translucency and exceptional thinness of silks render them extremely difficult to discern by insects under field conditions.

A further possibility that our findings suggest is that the different silk colours might be more associated with the regulation of silk thermal properties than any perceptual ecological advantages, such as attracting prey [28,46,47]. Perhaps subtle changes to skin size or composition (e.g. by the deposition of pigments), or appearance (e.g. by twisting), ensures that the silk proteins do not undergo unintended denaturation as temperature and/or humidity changes over time. Conversely, silk coloration may be selected for as a mechanism for attracting prey under certain conditions and any subsequent changes in thermal properties are an unavoidable artefact. Regardless of the evolutionary or ecological consequences and mechanisms, there are likely to be trade-offs between prey/predator attraction and avoidance as consequences of enhanced/reduced silk visibility and thermal properties as spider diets vary over time and/or space. It is the balance of this trade-off that likely determines whether T. plumipes silks appear visible or invisible to bees in any environment. We have seen other spiders using similar yellow-coloured silks when they build their webs in similar open habitats to those exploited by T. plumipes (see the electronic supplementary material, figure S4 for an example), suggesting that other spiders may vary their silk coloration in a similar way across diets or other ecological circumstance. Certainly, more comparative analyses of silk coloration, spider diets, web building, habitat preferences and the fitness benefits associated with the use of different coloured silks are warranted to test these hypotheses.

A broader implication of our findings is that they suggest that it is possible to develop biomimetic silk-like fibres with variable protein structures, hence variable mechanical properties, while holding the optical properties constant and vice versa. Nevertheless, thermal properties might inadvertently covary with optical properties. We thus recommend further research focus on the mechanism of this covariation so we can ascertain how to go about disentangling the covarying parameters during fibre construction.

Data accessibility

Data are accessible from Dryad.

Authors' contributions

S.J.B. designed the study and drafted the manuscript. G.C., T.E.W. and M.E.H performed the experiments, and carried out data analysis. S.J.B., M.E.H. and M.M.K. supplied materials. All authors gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

S.J.B. was supported by Australian Research Council and Hermon Slade Foundation grants.