Biomacromolecules 2004, 5, 661-665
661
A DECODER NMR Study of Backbone Orientation in Nephila clavipes Dragline Silk under Varying Strain and Draw Rate Philip T. Eles and Carl A. Michal* Department of Physics and Astronomy, University of British Columbia, 6224 Agricultural Road, Vancouver, British Columbia V6T 1Z1, Canada Received July 31, 2003; Revised Manuscript Received January 29, 2004
Using DECODER (direction exchange with correlation for orientation distribution evaluation and reconstruction) NMR, we probe the orientations of carbonyl carbons in [1-13C]glycine-labeled dragline silk under conditions of varying strain and fiber draw rate. A model-specific reconstruction of the molecular orientation distribution incorporating β sheets and polyglycine II helices indicates that the structures’ alignment along the fiber can be described by a pair of Gaussian distributions with full width at half-maxima of 20 and 68° and ∼45 and ∼55% relative contributions to the signal intensity. The alignment along the fiber was found to change appreciably when the drawing tension on the fiber was relaxed in a sample drawn at 4 cm/s while little change was observed in a sample drawn at 2 cm/s. The degree of alignment along the fiber was found to increase with fiber draw rate. Introduction The dragline silk of the golden orb weaver spider, Nephila claVipes, is widely regarded as a high performance fiber with a combination of strength and extensibility that has been difficult to achieve in man-made materials1 especially under benign processing conditions. These unique physical properties are attributed to the secondary and tertiary structure of the silk protein2 in the fibrous state. An understanding of this structure-function relationship will complement ongoing efforts of biomimeticists to spin fibers3,4 based on recombinant silk proteins5 in commercial quantities, having mechanical properties matching those of fibers produced in nature. The size of the silk protein, its insolubility in the fibrous state, and its inherent inhomogeneity precludes it from conventional solution nuclear magnetic resonance (NMR) or X-ray crystallographic protein structure determination, and other techniques must be employed. Because of its sensitivity to structure and dynamics, solid-state NMR is a powerful tool for the study of polymer fibers,6 and several groups have applied various solid-state NMR techniques to silks.7-9 On the basis of solid-state NMR,7,9-12 fiber X-ray diffraction,13,14 and amino acid sequences from cDNA sequencing,2,15 it is known that dragline silk is a semicrystalline polymer comprised of rigid alanine-rich β sheet crystallites connected by a network of soft glycine-rich chains, with the latter believed responsible for silk’s extensibility. Computer simulations suggest that such a model accounts well for silk’s mechanical properties.16 X-ray diffraction14 and solid-state NMR7 studies have shown that the crystalline regions exhibit a high degree of orientational order and that these regions * To whom correspondence should be addressed. E-mail: michal@ physics.ubc.ca. Fax: (604) 822-5324.
reorient under tensile deformation.14 The structure of the glycine-rich regions has been described as amorphous,14 random coil,1,17 and most recently a highly extended polyglycine II (or 31) helix,9,12,18 and although the orientation distribution (OD) of these structures is expected to be affected by macroscopic fiber deformation, such changes have not been observed. In this work, we use a new variation of DECODER (direction exchange with correlation for orientation distribution evaluation and reconstruction) NMR19 to reconstruct the OD of glycine’s carbonyl groups within the silk fiber under varying strain and fiber draw rate. Our implementation of DECODER NMR is novel due to the sample geometry, which complicates the reconstruction but provides a number of practical benefits. DECODER NMR. As a result of the anisotropy of the local electronic environment, the NMR chemical shift of a nucleus depends on the molecular orientation in an external magnetic field20 according to
(
[(23 cos Θ - 21) + η sin Θ cos 2Φ])
ω ) ω0 δiso + δ
2
2
where δiso, δ, and η describe the chemical shift anisotropy (CSA) tensor, which is attached to the molecular frame, and (Θ, Φ) specify the orientation of the magnetic field in this frame. Thus, the NMR spectrum contains information about the OD of molecular fragments within the sample. However, the problem of reconstructing an OD from a spectrum is illposed because each NMR frequency does not correspond to a unique orientation. This degeneracy can be largely lifted by a two-dimensional (2D) NMR experiment known as DECODER, which correlates the NMR frequency at two sample orientations using a macroscopic rotation of the sample during the mixing time. Each point in the 2D
10.1021/bm0342685 CCC: $27.50 © 2004 American Chemical Society Published on Web 02/25/2004
662
Biomacromolecules, Vol. 5, No. 3, 2004
Communications
sample for the large fiber bundles required to obtain adequate signal to noise from multidimensional NMR experiments. Experimental Methods
Figure 1. Sample holder designed to vary fiber strain with the removal of an insert.
Figure 2. Simulated DECODER spectra of wrapped fiber samples for discrete orientations of the carbonyl carbon in the fiber for a sample rotation of 90°.
spectrum corresponds to one of eight orientations, which are not further differentiable because of the symmetry of electronic shielding. Here, we introduce a new implementation of DECODER that incorporates a novel geometry with uniaxial fibers wound around a cylindrical support rod (Figure 1). Such a geometry complicates the reconstruction of the underlying OD by breaking the direct correspondence between points in the 2D spectrum and orientation in the fiber. Instead, each molecular orientation in the fiber corresponds to a distinct spectral signature as shown in Figure 2, and we employ simulation and fitting techniques to reconstruct the underlying OD. This sample geometry offers several practical advantages over an oriented fiber geometry by allowing for (1) a large amount of sample to fit inside the measurement volume, thus increasing the signal-to-noise ratio and opening the door for measurement of low-abundance amino acids and unlabeled fiber samples, (2) uniform variation of the sample strain throughout the sample with changes in the diameter of the poly(tetrafluoroethylene) (PTFE) rod, (3) smooth and constant silking rates during sample collection, and (4) the possibility of performing measurement of hydrated samples with the PTFE rod fitted into a water-filled NMR tube. Such practical benefits are difficult to achieve in an oriented
Silk Samples. Selectively labeled silk from 8-10 adult female N. claVipes spiders collected in central Florida was harvested by forced silking21 during five sessions over the course of two weeks to yield ∼100-mg samples. The spiders were fed Dulbecco’s modified eagle’s medium (SigmaAldrich, Steinheim, Germany) enriched with 5% (w/v) [1-13C]glycine (Cambridge Isotope Laboratories, Andover, MA) and supplemented with 2.5% (w/v) unlabeled alanine to approximately reflect the amino acid composition of major ampullate silk. Silk was drawn at 2 and 4 cm/s, within the range typically found in nature, and these rates have recently been shown to yield silk fibers with appreciably different crystallite ordering.22 Silk was drawn onto specially designed PTFE rods, which could collapse to ∼90% of their original diameter when a poly(chlorotrifluoroethylene) insert was removed, releasing some of the inherent fiber drawing strain (Figure 1). We measured a drawing strain of ∼10% for the 2 cm/s drawing rate. Thus, the decrease in the sample rod circumference represents an appreciable change in fiber strain. NMR Spectroscopy. DECODER spectra were collected on a home-built NMR spectrometer23 based upon an 8.4 T magnet (Oxford Instruments, Oxford, U.K.) providing a 13C resonance frequency of 91.480 MHz. Hartmann-Hahn crosspolarization from protons was performed with 50-kHz radio frequency (rf) fields and 1.5-ms contact times. Two-pulse phase modulation (TPPM)24 proton decoupling at 50 kHz was used during both evolution periods. A 90° sample flip about the rf coil axis was performed by a spectrometercontrolled stepper motor during the 0.1-ms mixing time. Fifty complex points were obtained in the indirect dimension with an 8-µs dwell time. Free induction decays were zero-filled to achieve the same spectral resolution in both dimensions and were apodized before Fourier transformation with an exponential corresponding to a 50-Hz full width at halfmaximum (fwhm). The sample’s starting orientation was cycled in 1.8° increments while signal averaging to ensure cylindrical symmetry about the rotation axis. Spectral Reconstruction. DECODER spectra of wound uniaxial fibers were simulated for discrete orientations of the chemical shift tensor in the fiber with a resolution of 5° in θ and φ (see Figure 2). These angles describe the orientation of the CSA in the fiber and are defined here to be successive rotations of the CSA tensor about the fiberfixed z and y axes with the CSA initially oriented with the largest component along x and the smallest component along z (the fiber axis). The glycine carbonyl CSA was assumed to lie with the most shielded component perpendicular to the peptide plane and the intermediate component tilted by 5° away from the carbonyl bond.25 CSA tensor values of (242, 178, 96) ppm9 were used in the simulations. A linear superposition of the simulated subspectra was used to reconstruct the experimental spectrum with the resulting weight of each subspectrum, indicating the relative number
Biomacromolecules, Vol. 5, No. 3, 2004 663
Communications
Figure 3. Physical model of silk, including secondary structures.
of residues that have the corresponding orientation. As a result of the intrinsic broadening of the experimental spectra, the nonorthogonality of the simulated subspectra, and our discrete sampling of orientation space in the simulations, our reconstruction technique is limited in the degree to which the underlying OD can be resolved. For this reason, we choose a model-dependent approach for reconstructing the experimental spectrum where the molecular OD in the silk is described by a small number of free parameters in a physically relevant model, from which weights for individual subspectra can be calculated. Because glycines reside in both the crystalline and soft interconnecting domains, we model our data with a two-component distribution, assuming that the two environments have varying degrees of alignment along the fiber. We introduce a pair of intermediate structures, shown in Figure 3, and assume that the OD of each structure is described by a Gaussian,
(
P(R) ) A exp -
)
sin2(R) 2∆R2
parametrized by a width ∆R, in radians. The model assumes that all possible orientations of the intermediate structures rotated about the structure axis (by an angle γ, as in Figure 3) are equally likely. By defining the set of rotations that connect the molecular and intermediate frame, we write R in terms of θ and φ and, thus, the Gaussian probability, P(R), as P(θ, φ), yielding the appropriate weighting of each subspectrum. On the basis of recent results,9,12,18 we chose to model the two-component nature of our data in terms of extended antiparallel β sheets (θβ ) 75°, φβ ) 5°) and polyglycine II helices (θII ) 55°, φII ) 10°), where the angles given here describe the orientation of the CSA tensor in the fiber frame when the secondary structures are aligned with their long axes along the fiber (Figure 3). We henceforth refer to the two respective components in the model as “crystalline” and “helical” but do not, as will be discussed, mean to imply a rigorous assignment of secondary structure. We implemented other more complicated two-component models that also reproduced our data well, including one model containing random coils and another model with the angles that define the intermediate structures left as fitting parameters. However, the complexity of these models (i.e.,
Figure 4. (a) Carbonyl peak from the 13C DECODER spectrum of isotopically labeled silk with an aliphatic carbon peak due to naturalabundance 13C shown in the inset, (b) best fit spectrum, (c) residuals, and (d) OD, P(R), from the fit with an 8.5 and a 27° Gaussian width for the β sheet and polyglycine II helix, respectively.
the number of fitting parameters) resulted in many sets of parameters yielding fits of similar quality (i.e., a shallow minimum in the multidimensional χ2 space), making an interpretation of the changes as a function of fiber strain and draw rate difficult. The simplicity of the model described here, containing very few free parameters, facilitates interpretation of our results under varying conditions of strain and draw rate. Fitting. Simulations were fit to experimental spectra with a nonlinear least-squares fitting algorithm from the PORT Mathematical Subroutine Library (AT&T Bell Labs, Murray Hill, NJ).26 The relative contribution of the signal from each intermediate structure and the width of each Gaussian distribution were left as fitting parameters, as were spectral broadening parameters and frequency offsets. Results The carbonyl region of the 13C DECODER spectrum of silk drawn at 2 cm/s under residual draw strain is shown in Figure 4a. No background subtraction was deemed necessary because, on the basis of electrospray mass spectroscopy of a portion of the sample following hydrolysis in 12 M HCl for 24 h at 100 °C, we estimate that 84% of the carbonyl signal intensity arises from glycine residues. Spectra similar to Figure 4a were collected for silk drawn at 4 cm/s, both under drawing tension and with tension relaxed. Spectra obtained under varying conditions showed qualitative differences and were appreciably different from spectra obtained on unoriented samples.27 Parts b and c of Figure 4 depict the best-fit spectrum and the residuals, respectively. The OD resulting from the [1-13C]glycine-labeled
664
Biomacromolecules, Vol. 5, No. 3, 2004
Communications
Table 1. Best-Fit Parameters from the Model and Calculated Order Parameters β sheet 2 cm/s 4 cm/s
as drawn relaxed as drawn relaxed
helix
Table 2. Distance of Glycine Residues from the Crystalline Domain total
glycine position
% of glycinesa
within crystalline domain 3 residues 4-6 residues >6 residues
31 24 22 23
%
∆R
f
%
∆R
f
f
47 44 47 32
8.5 8.6 8.5 9.2
0.93 0.93 0.93 0.92
53 56 53 68
27 24 23 29
0.33 0.42 0.45 0.29
0.62 0.64 0.68 0.49
fit to the data, shown in Figure 4d, reveals that 47% of glycines are well-aligned along the fiber (∆R ) 8.5°) in the crystalline components while 53% are poorly aligned (∆R ) 27°) in the helical component. Table 1 summarizes the fitting results for the various glycine-labeled samples. Although we have not undertaken a detailed error analysis, we find the goodness of fit to the data changes on a scale of about (2% for the percent contribution and (1° for the distribution widths. These uncertainties do not reflect significant correlations among the parameters, whereby changes in the relative populations of the components may be offset by a compensating change in a distribution width. The order parameter f ) (3〈cos2 R〉 - 1)/2 is calculated for comparison with previously published data. Discussion Molecular Orientation. DECODER experiments were recently reported9 for [1-13C]glycine- and alanine-labeled dragline silk from N. edulis spiders in an oriented fiber geometry. Although a quantitative analysis of orientation was not performed in that study, those data appeared to be consistent with two components of roughly equal intensity, one well-aligned, the other poorly aligned, in qualitative agreement with our results. The degree of alignment of our model’s crystalline component corresponds to a fwhm of 20° [where fwhm ) 2 arcsin(1.18∆R)], while that of the helical component is about 68°. These results compare to a 16° fwhm crystalline component (12%), a 30° fwhm oriented amorphous component (34%), and an isotropic component (54%) reported by Grubb and Jelinski14 from X-ray diffraction measurements. We also compare to the results of Simmons et al.,7 who report a 40% well-aligned crystalline component with a 5° fwhm and a 60% poorly aligned crystalline component with a 75° fwhm, based on 2H NMR measurements of alanines. The sequence of amino acids generally accepted as forming the core of the crystalline domains (see Table 2) is the alanine-rich sequence bounded by glutamine residues. Because glycines are found near the ends of these domains, where deviations from perfect crystallinity, or “fraying”, may be expected, it is not surprising that the glycines show a greater distribution width than did the alanines. It has also been suggested18 that a staggering of polyalanine sequences in the β sheets would create an interphase domain between the crystalline and amorphous domains that would contain glycine and glutamine residues that, because of the differing size and shape from alanines, would break the crystal symmetry and result in a broader OD than that of the bulk crystallites.
a Fractions are based on the peptide sequence of Xu and Lewis15 containing 656 residues, a representative sample of which we reproduce here with the crystalline region in bold: ...AGRGGLGGQGAGGAAAAAAGGAGQGGGYGGLGSQG...
Thus, we make a connection between our crystalline component and the combination of Grubb’s crystalline and oriented amorphous components and between our model’s helical component and Grubb’s isotropic component. To account for the 47% of glycines within our crystalline component, we must include (Table 2) all of the glycines within the crystalline sequence as well as the majority of glycines within three residues of the terminating glutamines. Such an assignment appears to be at odds with the Simmons et al. results, because on the basis of the Simmons et al. model, in which linker regions are unoriented, we would expect to find much less orientation in the glycines than we do. One potential explanation for this discrepancy is a different spectroscopic weighting of the components due to a difference in cross-polarization efficiency or a difference in 13C or 1H relaxation. We have ruled out this possibility by observing single-exponential behavior in carbonyl carbon (T1 ) 50 s and T2 ) 0.8 ms) and proton relaxation times (T1 ) 0.7 s) and observed overall cross-polarization efficiencies similar to those in glycine powder, leading to the conclusion that more glycines are well-aligned along the fiber than expected from the Simmons et al. model. The results from the relaxed fibers drawn at 4 cm/s show a reduced (32%) contribution from the aligned component, suggesting that fiber tension and drawing conditions affect the relative contribution of these poorly and highly aligned components to the makeup of silk. Tension and Fiber Drawing Rate. Our data suggest that the dependence of OD on tension is a function of fiber drawing rate. The sample drawn at 2 cm/s actually showed a slight decrease in alignment of the helical component when the draw tension was relaxed. The overall order parameter, however, remained nearly constant, as the increase in alignment of the poorly aligned helical component was offset by an increase in the population of that component. As discussed below, our data, especially for the poorly aligned component, are insensitive to the details of the secondary structure and so we do not interpret this change in population to indicate a shift from sheets to helices. Rather, we claim that the fitting results for this sample reflect subtle changes in the OD. In contrast, the sample drawn at 4 cm/s exhibited a significant broadening of the distribution (i.e., a decrease in the total order parameter) when tension was released. From the fits, it is clear that the decreased degree of alignment along the fiber is due to an increase in width of the distribution for the helices, as well as an increased contribution from the helical component to the total signal intensity.
Communications
Again, we do not mean to suggest that our data provide evidence for sheet structures becoming helical when the tension is relaxed. The fits are not sensitive enough to local structure to warrant such an interpretation. We attribute the changes in OD observed in the faster drawn fibers to the presence of metastable conformations that exist under tension and arise from expedited fiber processing in the duct. The release of tension allows for molecular rearrangement into a lower energy conformation. The slower drawing rate may allow time for this stable conformation to be reached in the duct during processing. Alternatively, the fact that we observed greater changes in OD in the fibers drawn more quickly can be attributed to the dependence of inherent drawing strain on draw rate. Fibers drawn faster may be closer to their breaking strain, whereas fibers drawn slower would be nearer a fully relaxed state. Our results would imply that more chain reordering occurs at high values of strain. Comparison of the OD in the two samples as drawn seems to indicate that the drawing rate affects the distribution of the poorly aligned component much more than the wellaligned component. An increased drawing rate yields fibers with helical components that are better aligned along the fiber, as would be expected from alignment due to elongational and sheer flow.28,29 These results are consistent with recent X-ray diffraction measurements of orientation in the ordered regions of as-drawn silk,22 which found an increase in order with increasing draw rate, with appreciable changes in the order parameter occurring between 1 and 40 mm/s draw rates. Our results suggest that these data should be interpreted as an increased alignment of the amorphous region, as opposed to more perfect alignment of the already well-aligned crystalline regions. Secondary Structures. We have chosen to simulate our spectra with a two-component model consisting of β sheets and polyglycine II helices. A more complete model might incorporate more components or other secondary structures. We find, however, that our results are fairly insensitive to the exact nature of the intermediate structures and cannot justify the inclusion of additional fitting parameters to describe a third component. This insensitivity is due to the similarity of (θβ, φβ) and (θII, φII), reflecting the fact that the protein backbones in both structures are well-aligned along the structure axis. What sensitivity there would be to the secondary structure is further obscured by a broad distribution in R, especially for the component we have labeled “helices”. We believe it unlikely that long, regular 31 helices exist in silk but have found these structures to be useful for building a model to simulate our data. Thus, our DECODER data appears to be consistent with β sheets and polyglycine II helices but cannot be taken as evidence for their existence. It is clear, however, that our data are not well-described by a single component but require at least two overlapping components to be adequately simulated. Conclusions and Future Work We have shown that DECODER NMR in a wound fiber geometry is sensitive to molecular orientation within the fiber and that this method presents some practical benefits over
Biomacromolecules, Vol. 5, No. 3, 2004 665
oriented-sample DECODER by allowing for uniform changes in strain, constant drawing rates, sample hydration, and increased signal to noise, rendering low-abundance amino acids and unlabeled samples accessible to such techniques. We plan to further exploit these advantages with studies of lower-abundance amino acids in dragline silk and effects of fiber hydration.27 We have been able, for the first time, to directly visualize the changes in orientation of the soft parts of the silk as functions of strain and draw rate. Our results indicate that the orientation of the helical component increases with increased draw rate and that the OD appears to change appreciably when the drawing tension on the fiber is released in a sample drawn at 4 cm/s but not in one drawn at 2 cm/s. Acknowledgment. This work was supported by a grant from the Natural Sciences and Engineering Research Council (NSERC) of Canada. P.T.E. thanks NSERC for a postgraduate scholarship. References and Notes (1) Gosline, J. H.; DeMont, M. E.; Denny, M. W. EndeaVour 1986, 10, 36. (2) Guerette, P. A.; Ginzinger, D. G.; Weber, B. H.; Gosline, J. M. Science 1996, 272, 112. (3) Vollrath, F.; Knight, D. P. Nature 2001, 410, 541. (4) Liivak, O.; Blye, A.; Shah, N.; Jelinski, L. Macromolecules 1998, 31, 2947. (5) Lazaris, A.; Arcidiacono, S.; Huang, Y.; Zhou, J.-F.; Duguay, F.; Chretien, N.; Welsh, E. A.; Soares, J. W.; Karatzas, C. N. Science 2002, 295, 472. (6) Schmidt-Rohr, K.; Spiess, H. W. Multidimensional Solid-State NMR and Polymers; Academic Press: New York, 1994. (7) Simmons, A. H.; Michal, C. A.; Jelinski, L. W. Science 1996, 271, 84. (8) Zhao, C.; Asakura, T. Prog. Nucl. Magn. Reson. Spectrosc. 2001, 39, 301. (9) van Beek, J.; Hess, S.; Vollrath, F.; Meier, B. Proc. Nat. Acad. Sci. U.S.A. 2002, 99, 10266. (10) Simmons, A.; Ray, E.; Jelinski, L. W. Macromolecules 1994, 27, 5235. (11) Yang, Z.; Liivak, O.; Seidel, A.; LaVerde, G.; Zax, D. B.; Jelinski, L. W. J. Am. Chem. Soc. 2000, 122, 9019. (12) Kummerlen, J.; van Beek, J.; Vollrath, F.; Meier, B. Macromol. 1996, 29, 2920. (13) Warwicker, O. J. Mol. Biol. 1960, 2, 350. (14) Grubb, D. T.; Jelinski, L. W. Macromolecules 1997, 30, 2860. (15) Xu, M.; Lewis, R. V. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 7120. (16) Termonia, Y. Macromolecules 1994, 27, 7378. (17) Gosline, J.; Denny, M.; DeMont, M. Nature 1984, 309, 551. (18) Valluzzi, R.; Szela, S.; Avtges, P.; Kirschner, D.; Kaplan, D. J. Phys. Chem. B 1999, 103, 11382. (19) Schmidt-Rohr, K.; Hehn, M.; Schaefer, D.; Speiss, H. J. Chem. Phys 1992, 97, 247. (20) Abragam, A. The principles of nuclear magnetism; Clarendon Press: Oxford, 1961. (21) Work, R.; Emerson, P. J. Arachnol. 1982, 10, 1. (22) Riekel, C.; Madsen, B.; Knight, D.; Vollrath, F. Biomacromolecules 2000, 1, 622. (23) Michal, C.; Broughton, K.; Hansen, E. ReV. Sci. Instrum. 2002, 73, 453. (24) Bennett, A.; Rienstra, C.; Auger, M.; Lakshmi, K.; Griffin, R. J. Chem. Phys. 1995, 103, 6951. (25) Teng, Q.; Iqbal, M.; Cross, T. J. Am. Chem. Soc. 1992, 114, 5312. (26) Dennis, J.; Gay, D.; Welsch, R. ACM Trans. Math. Software 1981, 7, 348. (27) Eles, P. T.; Michal, C. A. Macromolecules 2004, 37, 1342. (28) Doi, M.; Edwards, S. The theory of polymer dynamics; Clarendon Press: Oxford, 1986. (29) Forest, M. G.; Ueda, T. J. Non-Newtonian Fluid Mech. 1999, 84, 109.
BM0342685
