Protein–metal coordination interactions were recently found to function as crucial mechanical cross-links in certain biological materials. Mussels, for example, use Fe ions from the local environment coordinated to DOPA-rich proteins to stiffen the protective cuticle of their anchoring byssal attachment threads. Bioavailability of metal ions in ocean habitats varies significantly owing to natural and anthropogenic inputs on both short and geological spatio-temporal scales leading to large variations in byssal thread metal composition; however, it is not clear how or if this affects thread performance. Here, we demonstrate that in natural environments mussels can opportunistically replace Fe ions in the DOPA coordination complex with V and Al. In vitro removal of the native DOPA–metal complexes with ethylenediaminetetraacetic acid and replacement with either Fe or V does not lead to statistically significant changes in cuticle performance, indicating that each metal ion is equally sufficient as a DOPA cross-linking agent, able to account for nearly 85% of the stiffness and hardness of the material. Notably, replacement with Al ions also leads to full recovery of stiffness, but only 82% recovery of hardness. These findings have important implications for the adaptability of this biological material in a dynamically changing and unpredictable habitat.
Countless life processes have evolved a dependence on the ability of specific metal ions to interact with proteins as essential cofactors with functions, including enzyme catalysis, hormone signalling and gas transport . Negative physiological effects are often associated with a deficiency of essential metal ions, as well as with the introduction of anomalous metal ions, that may compete for protein-binding sites [1,2]. In addition to these physiological roles, protein–metal coordination complexes contribute to mechanical properties of biological materials, including the jaws of certain arthropods and annelids, as well as mussel byssus attachment threads . Currently, however, very little is understood about how metal substitution affects the mechanical performance of metal-dependent materials. The primary goal of this study is to investigate this question using the mussel byssus as a model system.
Mussels employ stiff, tough and extensible proteinaceous attachment fibres known as byssal threads in order to effectively dissipate energy from crashing waves in rocky seashore habitats (figure 1a) . Loss of mechanical integrity in byssal threads owing to environmental factors (e.g. temperature and pH) can lead to increased dislodgement in mussel beds . The extensible fibrous core of the thread is coated by a thin protective cuticle (thickness 5–10 µm), which possesses hardness and stiffness values similar to epoxies, while at the same time exhibiting extensibility up to 100% before cracking (figure 1b) [6–8]. This unusual combination of high hardness/stiffness and extensibility is critical to the proposed abrasion-resistant function of the cuticle and was attributed largely to the presence of protein–metal cross-links organized in a hierarchical particle-reinforced composite-like structure (figure 1b,c) [6,8,9]. Specifically, the cuticle is comprised largely of a single protein, mussel foot protein-1 (mfp-1), which is enriched in the post-translationally modified amino acid 3,4-dihydroxyphenyl-l-alanine (DOPA). DOPA is renowned for its capacity to coordinate specific metal ions (e.g. Fe, V, Al) in highly stable tris complexes (three DOPA: one metal) with bidentate chelation via oxygen ligands (figure 1c). Metal coordination cross-links between DOPA in mfp-1 and Fe were demonstrated previously to be a critical component of cuticle mechanical behaviour [6,7,9]; thus, biological acquisition of metal ions is crucial to proper material function.
Natural and anthropogenic influences cause variations in ocean metal concentration on both large and small spatio-temporal scales . Mussels and other molluscs possess elaborate pathways and mechanisms for acquiring and trafficking metal ions within their tissues . Studies using radiolabelled Fe have demonstrated that seawater metals accumulate in the soft tissue of the mussel and are subsequently actively integrated into the byssus . The broad range of metal ions found in threads (e.g. Zn, Cu, Fe, V, Al, Si, Ni, Pb, U, etc.) and the concentration of each varies immensely between different environments and even seasonally in the same environment [13,14]. For example, V can become concentrated in the byssus, especially following oil spills, by up to 10 000-fold [15–17]. However, it is not clear how or if this variation affects metal-dependent mechanical properties of the byssus. To address this question, we investigate here the influence of ‘anomalous' DOPA-binding metal ions (e.g. V and Al) on the byssal thread cuticle properties using biochemical, spectroscopic and nanomechanical characterization methods. Results demonstrate that mussels are opportunistic in their use of reinforcing metal ions, possibly as an adaptive response to life in wave-swept marine habitats in which metal bioavailability is highly unpredictable and in which the consequences of altered mechanical performance are catastrophic [5,18,19].
2.1. Sample preparation
Mytilus californianus mussels, collected from Goleta Pier (Goleta, CA 93117, USA), were grown in a tank with flowing seawater from the Santa Barbara Channel. Fresh threads, grown for up to one week, were removed from the mussel and stored in distilled water for up to four weeks prior to measurement or embedding. Native threads were not further treated prior to experimental testing. Threads for EDTA treatment and metal replacement studies were stirred for 1 day in 100 mM citrate buffer (pH 3.86), which was followed by treatment with 200 mM EDTA, 10 mM Tris-buffer (pH 4.34) for 3 days exchanging EDTA solution every day. Threads were then washed in 0.1 mM EDTA solution twice for 1 h. EDTA-treated threads were stored in 0.1 mM EDTA. For metal-recovery experiments (Fe, V and Al), EDTA-treated threads were stirred for 1 h in solutions of 1 mM FeCl3 (pH 2.33), 1 mM VCl3 (pH 2.53) or 1 mM AlCl3 (pH 4.23), respectively. Afterwards, threads were washed in Millipore H2O twice for 1 h. EDTA-treated and metal-recovery threads were stored for between 1 and 3 days until measurement or were embedded immediately for mechanical measurements.
2.2. Inductively coupled plasma–optical emission spectrometry
Following the general sample preparation described above, native and treated byssal thread samples for inductively coupled plasma–optical emission spectrometry (ICP-OES) were freeze dried, weighed and subsequently dissolved in 500 µl aqua regia (167 µl HNO3 + 333 µl HCl) at 40°C for 24 h. The resulting solutions were diluted at a 1 : 10 ratio with Millipore H2O, and emission spectra were recorded with an ICP-OES analyser (Optima 8000, Perkin-Elmer). A concentric glass nebulizer (type A, Meinhard, Golden, CO, USA) equipped with a glass cyclonic spray chamber (Meinhard) was used, together with an argon plasma torch, to atomize the sample solutions. Sample spectra were recorded with a dual échelle polychromator in combination with a CCD detector.
2.3. Amino acid analysis
Thread samples for amino acid analysis were freeze dried, weighed and hydrolysed for 24 h in vacuo at 110°C in 6 M HCl with 10% phenol. Hydrolysed samples were flash evaporated, resuspended in sample dilution buffer and run on an amino acid analyser S433 (Sykam, Eresing, Germany) to determine the amino acid composition by ion-exchange chromatography coupled with post-column derivatization of amino acids with ninhydrin.
Elemental composition maps of cross sections of native and treated byssal threads were obtained by energy-dispersive X-ray spectroscopy (EDS) with a Jeol JEM 7500F electron microscope equipped with two Oxford X-Max 150 Silicon drift detectors. An accelerating voltage of 20 kV was used. Measurements were performed on 150 µm thick cryo-sectioned thread cross sections. To avoid charging, samples were carbon-coated prior to the investigation.
2.5. Raman spectroscopy
Spectra from treated and untreated byssal thread cross sections were acquired with a confocal Raman microscope (CRM200, WITec, Germany) equipped with a piezo-scanner (P-500, Physik Instrumente, Karlsruhe, Germany). The diode-pumped near infrared laser (λ = 785 nm, Toptica Photonics AG, Gräefelfing, Germany) was focused on the sample through a 60× water-immersed objective (Nikon, NA = 1.0). The laser power on sample was set to 15 mW. The spectra were acquired using an air-cooled CCD (DU401A-DR-DD, Andor, Belfast, North Ireland) behind a 300 g mm−1 grating spectrograph (Acton, Princeton Instruments Inc., Trenton, NJ, USA) with a spectral resolution of 6 cm−1. The ScanCtrlSpectroscopyPlus software (v. 1.38, WITec, Ulm, Germany) was used for measurement set-up. At least five single measurements of 10 accumulations with an acquisition time of 2 s each were acquired from different regions within the cuticle area of thread cross-sectional surfaces and averaged. Two-dimensional imaging of thread surfaces was performed with a 0.5 s acquisition time and a lateral resolution of 0.5 µm. OPUS software package v. 7.0 (Bruker Optik GmbH) was used to subtract a polynomial background and smooth spectra with a Savitzky–Golay function.
2.6. Electron paramagnetic resonance
For each thread treatment type, distal portions of five threads were dried, cut into approximately 5-mm long pieces and pooled for electron paramagnetic resonance (EPR) measurements. EPR spectra were recorded at approximately 9.4 GHz (X-band) with a Bruker CW Elexsys E 500 spectrometer. Simulation of spectra was performed with the program X-Sophe v. 188.8.131.52  using the experimental parameters.
Native and treated byssal thread samples for nanoindentation were embedded in Epofix and cured overnight. An ultra-microtome (PowerTome XL, RMC Products) was used to obtain smooth transverse cross-sectional surfaces. All indentation experiments were carried out within 1–3 days after embedding according to a previously established protocol . During measurements, the samples were fully submerged in distilled water or in the case of EDTA-treated threads, a solution of 0.1 mM EDTA. Testing was performed on a TriboScan UBI-1 (Hysitron, Minneapolis, MN, USA) equipped with a Berkovich tip. The location of each indent was chosen after obtaining an image of the sample surface via scanning probe microscopy. Between 60 and 90 indents of the cuticle were performed per thread sample, and two different threads per sample treatment were tested. All indentations were performed in the open loop mode with a loading and unloading rate of 100 µN s−1. The peak load of about approximately 1000 µN was held for 30 s to eliminate creep effects. A PMMA standard was used to obtain the tip area function. Hardness and reduced elastic modulus values were obtained with the method described by Oliver & Pharr , using the TriboScan measurement software (Hysitron). Values were statistically analysed with the R software for statistical computing (R Core Team 2014), using the Kruskal–Wallis rank-sum test for analysis of variance of treatment groups with post hoc multiple comparisons according to Siegel & Castellan  and Dunn–Šidák correction for analysis of specific pairs of treatment groups for significant differences.
Table 1 summarizes ICP-OES results for various metal ions present in Mytilus californianus byssal threads. Fe is present in threads at a concentration consistent with previous reports for this species , and V was found at concentrations nearly 34% that of Fe on average. Notably, V was also detected in threads of Mytilus galloprovincialis mussels grown in the same seawater and in an aquarium with artificial seawater (electronic supplementary material, table S1), suggesting that its incorporation in the byssus is neither species-specific nor an anomalous behaviour of mussels in the oil-seep-rich Santa Barbara Basin. Additionally, Ca, Al and Si, which were previously shown to be present in the cuticles of M. galloprovincialis threads [7,23], were also detected in threads of M. californianus (table 1). Other DOPA-binding metals, such as Mn and Ti were not detected in significant amounts.
Following extensive treatment with the metal chelator EDTA, more than 70% of the Fe and 90% of the V ions were removed from threads on average, whereas Al was only reduced by approximately 45% and still appeared in the cuticle in EDX images (electronic supplementary material, figure S1 and table S2). Soaking EDTA-treated threads in 0.1 M solutions of either FeCl3 (Fe-recovery threads), VCl3 (V-recovery threads) or AlCl3 (Al-recovery threads) for 1 h, followed by extensive washing in distilled water, led to the recovery of the respective ion in ICP-OES measurements, but in amounts exceeding values measured in the native thread by threefold, 18-fold and 1.4-fold for Fe-, V- and Al-recovery threads, respectively (electronic supplementary material, table S2). However, EDX measurements of metal distribution in thread cross sections of V- and Al-recovery threads suggest that some of the excess V and Al is present in the thread core (electronic supplementary material, figure S1). Consistent with previous reports of other Mytilus species [23,24], amino acid analysis showed that threads contained 0.82 ± 0.16 mol% DOPA (78 ± 18 µmol g−1 dry thread). Compared with the total concentration of all metals measured in the native state (table 1), DOPA is always in significant excess of metal ions with respect to forming tris–DOPA–metal complexes (three DOPA : one metal ion).
Confocal Raman microspectroscopy imaging was used to investigate DOPA–metal coordination within the cuticle (figure 2). Resonance Raman spectra obtained from native thread cuticles are consistent with those previously reported, exhibiting a triad of peaks between 500 and 700 cm−1 corresponding to the oxygen–metal interaction and four peaks between 1150 and 1450 cm−1 corresponding to vibrations in the catechol ring [9,25] (figure 2a). Raman confocal imaging of the thread cross section indicates that the DOPA–metal interaction is isolated to the cuticle (figure 2b). As previously reported, extensive EDTA treatment of threads significantly reduced the resonance signal as indicated by the loss of signal in the cuticle relative to the core (figure 2b) . Fe-recovery threads showed the reappearance of lower intensity resonance peaks, some of which are shifted by up to 10–20 cm−1 compared with the native thread (figure 2 and electronic supplementary material, table S3), whereas V-recovery threads exhibited a very strong Raman signal nearly identical to the native signal. The observed peak positions of Fe- and V-recovery threads correspond well to those observed in DOPA-containing polymers mixed with Fe and V ions, respectively (electronic supplementary material, table S3) , confirming that V is coordinated by DOPA in a bidentate fashion within native byssal thread cuticles .
The intensity of the core in the two-dimensional Raman images provides an internal reference for the relative intensity of the cuticle signal, as it does not appear to be affected by the various treatments (figure 2b). In this light, the large intensity differences between the native, Fe- and V-rescue threads become very apparent. However, it is important to emphasize here that no quantitative conclusions can be drawn about the relative amount of DOPA–V and DOPA–Fe interactions based on resonance Raman spectroscopy, because the intensity of the resonance signal is highly dependent on the absorption maxima of the complexes and the laser wavelength used . This is due to the fact that when the incident laser wavelength is close to the absorption maximum of a complex, the resonance effect is typically more enhanced, which has been observed for metalloproteins bound to various metal ions . Interestingly, the absorption maximum for tris–DOPA–Fe3+ is approximately 490 nm, whereas the two absorption maxima for catechol–V complexes assigned to tris–DOPA–V4+ were observed at approximately 400 and 650 nm . DOPA–V coordination clearly exhibits a dominant Raman resonance signal relative to DOPA–Fe complexes in this study (figure 2), which is almost certainly a by-product of the laser wavelength used (λ = 785 nm) and not an indication of increased amount of V coordination in the thread cuticle. Thus, EPR measurements were performed to examine DOPA–Fe coordination in native threads.
EPR measurements of native threads, Fe-recovery threads and byssal thread adhesive plaques (figure 3 and electronic supplementary material, figure S3) yielded nearly identical spectra with two dominant features centred at g values of approximately 4.25 and approximately 2.00. Previous measurements on adhesive plaques from Mytilus edulis byssal threads exhibited nearly identical peaks, which were assigned to tris-coordinated DOPA–Fe3 + (g ∼ 4.24) and organic radicals (g ∼ 2.00) arising from oxidized DOPA residues . Notably, present in the native threads, but not in the Fe-recovery threads are peaks (g ∼ 1.99) assigned to V with a nuclear spin of I = 7/2 (figure 3, insets). This ensemble of peaks resembles the EPR spectrum of V-recovery threads (electronic supplementary material, figure S4); however, the peaks in the native threads are less resolved, which could be due to several effects including low concentration or slight variations in the coordination sphere. While a similar spectral pattern previously observed in catechol–V hydrogels was interpreted as an organic radical , here, we assign these features instead to coordinated V because they are absent in Fe-recovery threads, which have had nearly all V removed (table 1 and figure 3). This is further supported by Raman spectra of native and V-recovery threads, which clearly show DOPA–V coordination (figure 2).
Taken together, EPR and Raman data confirm the presence of both DOPA–Fe and –V coordination in native byssal thread cuticles, and show that the cross-linking metal can be easily exchanged following metal removal with EDTA. In order to investigate the role of Fe and V in influencing the mechanical properties of the cuticle, nanoindentation was performed as previously described  (figure 4 and electronic supplementary material, table S4). The cuticles of hydrated native threads exhibited reduced elastic modulus (Er) and hardness (H) values of 1.2 ± 0.3 GPa and 38 ± 13 MPa, respectively. Metal removal via EDTA-treatment resulted in a reduction of Er and H values to 0.2 ± 0.1 GPa and 6 ± 2 MPa, respectively, which far exceeds the approximately 50% loss of hardness in EDTA-treated threads reported previously for M. galloprovincialis threads . Fe- and V-recovery threads both exhibited a full mechanical recovery of Er and H values that were not statistically different from native threads (p < 0.01), whereas all three groups were significantly different from the EDTA-treated threads (p < 0.01; figure 4). Notably, Al-recovery threads exhibited full recovery of Er that was not statistically different from native threads (p < 0.01), but showed significantly different (approx. 18% lower) H values (p < 0.01). It is important to emphasize that the cuticle is a viscoelastic material and thus, the reported values are dependent on the indentation loading and unloading rates, indenter tip geometry and other factors. Notably, previously reported values for reduced modulus of M. californianus cuticle are similar to currently reported values (Er = ∼1.6 GPa), whereas hardness values were higher (H = ∼90 MPa) . Thus, while care should be taken when comparing the current values to other materials or studies, comparisons between the various treatments reported within this study are highly informative and provide a window into the mechanical role of metal coordination.
This study demonstrates convincingly that, in addition to Fe, mussels naturally use V, and likely Al, to mediate DOPA coordination cross-linking in the cuticle. Loss of nearly 85% of cuticle Er and H following EDTA treatment of threads confirmed that metal ions are necessary cross-linking agents and moreover, provided a platform for exploring the mechanical contribution of specific metal ions. Metal recovery studies demonstrated that Fe and V were both equally sufficient to recover native mechanical behaviour, whereas Al fully recovered Er and largely recovered H (82%). As previously demonstrated, altered mechanical performance of the byssus owing to environmental changes can be detrimental for the ability of mussels to create a secure attachment and colonize high-impact sites in the intertidal habitat [5,19]. The current findings suggest that opportunistic use of a range of DOPA-binding metal ions in the cuticle may be an adaptive trait that minimizes such consequences by preservation of constant performance (i.e. ‘mechanical homeostasis') in ocean environments where iron availability is highly variable, and often limiting for many organisms [30,31]. While some marine organisms can substitute metal cofactors in proteins in metal-limited waters (e.g. Cd and Co substitution for Zn in diatoms) , to the best of our knowledge, the current findings are the first example of a biological material that substitutes metals in coordination cross-links without an observed change in performance. This contrasts previous research on the Zn-reinforced jaws of the sediment-dwelling Nereis marine worms, which did not fully recover mechanical properties during metal recovery experiments following EDTA treatment and showed large differences between recovery experiments with different metal ions . Considering the differences in the coordination chemistry of Fe, V and Al [34,35] and previously observed mechanical differences between Fe-, V- and Al-reinforced DOPA-enriched hydrogels [26,36], it is striking that cuticle performance was not greatly affected and raises the question of the mechanism underlying this behaviour.
The Er of a non-rigid polymeric material arises as a consequence of the number of cross-links present in the material . Thus, it can be inferred that regardless of the specific chemical properties of the interactions (e.g. geometry, oxidation state, number of ligands), the number of DOPA–metal cross-links in native, Fe-, V- and Al-recovery threads is essentially the same and accounts for at least 83% of the total number of cross-links. Although compositional studies showed that DOPA is in significant excess of the total metal ions (table 1), increasing the metal ion content in rescue threads did not result in an increase in stiffness or hardness, suggesting that only a specific amount of the total free DOPA present is able to participate in metal-mediated cross-links. Thus, a key determinant of the ‘mechanical homeostasis' exhibited by the cuticle may be the ability to fix the number of DOPA binding sites because it limits the number of cross-links to an acceptable range and prevents over-stiffening under high metal concentration conditions. At the same time, employing DOPA as a ligand likely facilitates the formation of very stable complexes with a broad range of metal ions , increasing the likelihood of filling all cross-linking sites under low metal concentration conditions. Variations in hardness, as exhibited in Al-recovery threads, are more difficult to interpret—especially because this is a viscoelastic material. Nonetheless, assuming that the protein network comprising the cuticle remains unaltered during treatment and only the metal ions are exchanged, these results suggest a lower breaking force for DOPA–Al cross-links under the testing conditions, which is consistent with previous rheological studies on metal-reinforced DOPA-enriched hydrogels [26,36].
Under basic seawater conditions, DOPA coordinates with metal ions in a bidentate fashion via its two oxygen atoms and typically exists as bis (two DOPA : one metal) or tris (three DOPA : one metal) complexes [26,39]. EPR and Raman evidence confirm that Fe in the native thread is coordinated in the tris mode [9,26,28]. At the pH of the V recovery solution, the vanadyl form [V = O]2+ should dominate . Vanadyl is able to complex DOPA in the bis mode when V is in excess; however, the oxygen can be displaced by another catechol moiety when DOPA is in excess, leading to tris complexation [34,41]. In this case, the large molar excess of DOPA in the cuticle compared with V as indicated in table 1 would appear to favour formation of tris–DOPA–V complexes. Based on Badger's rule, which describes the empirical relationship between Raman vibrational positions and metal–oxygen bond lengths, the nearly identical spectra of native and V-recovery threads indicates similar coordination in both cases, at least with regards to bond length (electronic supplementary material, table S3) . This is remarkable considering that native threads are formed under biological conditions, whereas metal ions in recovery threads are integrated into the structure via passive diffusion. While spectroscopic investigations applied in this study are not suitable for the characterization of Al coordination owing to intrinsic electronic properties of Al ions, we postulate that Al also binds DOPA in a tris complex based on the tendency for tris–DOPA–Al3+ formation under neutral to basic conditions [35,36]. Further work is required to elucidate the exact coordination and oxidation states of various metals in the cuticle (e.g. with X-ray absorption spectroscopy); however, the present mechanical analysis suggests that Er and H of the cuticle are independent of these factors.
Based on these considerations, the cuticle can be considered a hard and flexible organic scaffold that augments mechanical performance by fivefold by filling fixed DOPA cross-link sites with metal ions such as Fe, V, Al and possibly also others. Several studies have proposed that DOPA in the byssus not only participates in the formation of metal coordination cross-links, but also in the formation of diDOPA covalent cross-links that may help fix the cuticle scaffold structure during the thread formation process . DiDOPA formation has been speculated to proceed either via enzymatic means with catechol oxidase [5,44] or via a redox-driven reaction coupling the reduction of coordinated metal ions (e.g. Fe3+ → Fe2+) with the conversion of DOPA to DOPA-quinones, which further leads to aryl coupling of two DOPA rings [45,46]. Regardless of the formation mechanism, the current findings indicate that covalent cross-linking provides at maximum 17% of the cross-links in the cuticle, in spite of the presence of redox active metal ions (e.g. Fe, V) capable of oxidizing DOPA and an apparent high content of organic radicals as indicated by EPR (figure 2) . Mechanical recovery observed with Al, a non-redox active metal, offers further support in this respect. Interestingly, recent in vitro studies on the cuticle protein mfp-1 indicated that redox-driven formation of diDOPA cross-links can be initiated at low pH, but is prevented at basic pH . Consistent with our current findings, the authors proposed that covalent bond formation occurs at early stages of thread formation in which protein contents are secreted in an acidic medium (pH < 5), but that further oxidative cross-linking is prevented at the higher pH of seawater. However, this must be further examined.
In summary, this study demonstrates that mussels naturally integrate V ions into their byssus where they were shown to contribute to mechanical reinforcement of the cuticle without an observed change in mechanical stiffness or hardness. This is particularly relevant, for example, in V-rich waters near oil effluents or following an oil spill . Al-contaminated waters, on the other hand, might have an adverse effect on the longevity of byssal threads based on the slightly decreased hardness of Al-recovery threads. Relevant to these findings, the emerging field of marine ecomechanics focuses on understanding how organisms interact with the physical and chemical constraints of demanding ocean habitats and the material adaptations employed in this endeavour [5,18]. The byssus provides a lifeline for mussels, and its mechanical performance is strongly linked to organismal survival—as evidenced by the catastrophic effects of altered material properties under changing oceanic conditions . From this perspective, the ability of the mussel cuticle to employ V, and possibly Al, in place of Fe while maintaining reliable mechanical performance provides an adaptive advantage in a tumultuous environment in which metal ion bioavailability varies unpredictably over a large range of temporal and spatial scales. This is especially important in the case of the mussel owing to its sessile lifestyle and cosmopolitan distribution. From a broader perspective, the current findings are relevant for the growing field of mussel-inspired materials based on DOPA chemistry [39,47] and suggest that a fixed scaffold of cross-linking sites may be a critical design consideration for synthesizing hard and extensible materials for coating applications.
M.J.H. and C.N.Z.S. designed research; M.J.H., C.N.Z.S., A.W. and L.B. performed research; M.J.H., C.N.Z.S., L.B. and P.S. analysed data; M.J.H. and C.N.Z.S. wrote the paper. All authors discussed results and commented on manuscript.
We declare we have no competing interests.
The authors acknowledge the German Research Foundation (DFG, Priority Programme 1568 HA6369/1-1) and the Max Planck Society for financial support.
We thank J. H. Waite and E. Danner for providing M. californianus threads, as well as P. Fratzl and I. Zlotnikov for helpful discussion.
- Received May 22, 2015.
- Accepted July 31, 2015.
- © 2015 The Author(s)
Published by the Royal Society. All rights reserved.