INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

To prevent the formation of large ice crystals in blood and tissue at low temperatures, many species, including polar fish and insects, manufacture high concentrations of proteins that inhibit the growth of ice crystals, lowering the freezing point, with only a small effect on the melting temperature. This non-colligative property is distinct from equilibrium freezing point depression, and can be observed as hysteresis in ice formation in cooling/warming cycles. This is possibly due to binding on one or more ice facet, preventing further growth (for a review, see Yeh and Feeney, 1996).

Several antifreeze proteins have been isolated and described, and they fall into four distinct structural classes (Davies and Sykes, 1997). All of these proteins form ordered, compact structures similar to other globular proteins. Although there are numerous theories of their action, there is little consensus as to the mechanism of ice binding of these diverse proteins. In Antarctic fish species, the blood and tissues contains high concentrations of a number of related antifreeze glycoproteins. These proteins are based on a repeated sequence of AAT*, where T* is threonine with the disaccharide Gal-1, 3GalNAc bonded to the O of the threonine residue. The various fractions that have been described differ mainly in the number of repeats, though fraction 8 in particular is not only short (14-16 residues), but also contains some of the alanine residues substituted by proline. These short glycopeptides are the most abundant, but on a weight basis are less potent in preventing ice crystallization.

Recently we described the conformational preferences and local dynamics of AFGP8 (Lane et al., 1998). According to the NMR spectra, there was no significant long-range order, and substantial segmental flexibility. However, for the Thr-Pro-Ala tripeptides, evidence was obtained of some local order forming a small segment of secondary structure reminiscent of the polyproline helix (Lane et al., 1998). The high degree of flexibility of these peptides suggests that the mechanism of action may be fundamentally different from the non-glycosylated AFPs (Yeh and Feeney, 1996; Davies and Sykes, 1997; Harding et al., 1999; Haymet et al., 1999), as there is no fixed long-range structure present. To provide further information on the conformational preferences of these glycoproteins, we have investigated the solution properties of AFGP1-5, which contain no proline residues, and vary in length from 60 to 120 residues. We have assigned the ¹H- and ¹³C-NMR spectra of AFGP1-5, and determined its dynamic properties by ¹³C-NMR relaxation measurements at two temperatures. We have also used FTIR spectroscopy to provide further information about the local backbone conformation of both AFGP1-5 and AFGP8. The amide I region of the spectrum (~1600-1700 cm¹) reports on the peptide carbonyl stretching frequency, which is sensitive to the local conformation (Byler and Susi, 1986; Dong et al., 1990; Pestrelski et al., 1991; Surewicz et al., 1993). To reconcile the NMR and FTIR results, we invoke a dynamic model that incorporates the different time scales probed by these methods. In addition, results from MD calculations on short peptide sequences related to those present in AFGPs show that local backbone fluctuations of the appropriate magnitude do occur on the correct general time scale.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Materials

AFGP1-5 was isolated from the Antarctic fish Trematomus borchgrevinki and purified as previously described (DeVries et al., 1970). This preparation contains five homologs of the repeating tripeptide AAT* that differ in the number of repeats n. AFGP3-5 account for >80% of the total, and vary from 11 to 22 kDa. (DeVries et al., 1970). Integration of signals in both ¹H and ¹³C 1D-NMR spectra acquired with long relaxation delays showed that the relative concentrations of each sugar and amino acid residue were equal to within better than 10%.

For NMR analysis, the AFGP1-5 sample was dissolved in d₃-acetate buffer pD* = 5 at 17 mg/ml. For FTIR measurements, AFGP1-5 and AFGP8 were dissolved in 10 mM phosphate buffer pH 7.4 at 50 mg/ml.

PolyPro (1-10 kDa) was purchased from Sigma Chemical Co. (St. Louis, MO) and used without further purification. Aprotinin, hen egg white lysozyme, and bovine carbonic anhydrase were purchased from Sigma Chemical Co. (Poole, Dorset, UK). For NMR they were dissolved in a deuterated buffer containing 20 mM sodium phosphate and 100 mM KCl, pH 7.

The peptides NAcAATAA and NAcAAPAA were synthesized using fmoc chemistry on an Applied Biosystems 430A peptide synthesizer and purified by reverse-phase HPLC on a preparative C18 column using a 1% acetonitrile-water gradient. The composition and purity were verified by NMR spectroscopy.

Methods

NMR spectroscopy

(1)

(2)

(3)

1

6

(4)

(5)

Infrared spectroscopy

1

1

1

1

MD calculations

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

NMR assignments of AFGP1-5

Protons were assigned to spin type using DQF-COSY in D₂O (Fig. 1) and TOCSY (Fig. 2 A) and NOESY in H₂O (Fig. 2 B), all at 10°C. Only three kinds of amino acid resonances were observed, corresponding to Thr, and two Ala residues. This is consistent with a repeated unit of (AAT)_n. All of the protons could be assigned to residue type. In addition, all of the sugar protons could be assigned to -Gal or GalNAc (Table 1). Only one set of resonances for the disaccharides was observed. This is again consistent with a simple repeated unit (AAT*)_n. ¹³C assignments were obtained for all protiated carbons (Table 1) using the ¹³C-¹H HSQC experiment (Fig. 3). The improved resolution of the spectra allowed unambiguous and complete assignments of both sugar residues, unlike previously for AFGP8 where the four GalNAc residues were non-equivalent (Lane et al., 1998; Dill et al., 1992). The shifts in Table 2 revise some previous assignments of comparable fragments (Homans et al., 1985). The 1D ¹³C and HSQC spectra (Fig. 3) showed only a single set of resonances for the (AAT*) repeating unit. All of the ¹³C resonances were narrow compared with linewidths predicted for a 20-30-kDa globular protein, indicating the presence of significant internal mobility in the molecule. Furthermore, there is substantial variation in the ¹³C linewidths, especially comparing the carbons atoms of the amino acid backbone and the GalNAc with the carbon atoms of -Gal, which suggests differences in the mobility of these residues.

View larger version (16K): [in this window] [in a new window]   FIGURE 1   DQF-COSY spectrum of AFGP1-5 in D2O. The spectrum was recorded at 11.75 T, 303 K with acquisition times of 0.4 s in t2 and 0.098 s in t1. The data table was zero-filled once in t2 and twice in t1, and multiplied by a shifted Gaussian function in each time dimension before Fourier transformation. The final digital resolution was 1.2 Hz per point in F2 and 2.4 Hz per point in F1. The lines show spin systems of Thr and the two Gal residues.

View larger version (17K): [in this window] [in a new window]   FIGURE 2   NOESY and TOCSY spectra of AFGP1-5 in H2O. Spectra were recorded at 11.75 T and 278 K. The acquisition times were 0.41 s in t2 and 0.102 s in t1. Data tables were zero-filled once in t2 and twice in t1, and multiplied by a Gaussian function centered at t = 0 before Fourier transformation. The relaxation delay was 2 s. (A) TOCSY with a mixing time = 47 ms, B1 = 8.2 kHz. (B) NOESY with a mixing time of 200 ms.

View larger version (26K): [in this window] [in a new window]   FIGURE 3   13C-NMR spectra of AFGP1-5. (A) 1D 13C spectrum recorded at 303 K and 9.4 T with Waltz decoupling. (B) HSQC spectrum. The spectrum was recorded at 14.1 T and 30°C with acquisition times of 0.1 s in t2 and 0.057 s in t1. Data tables were zero-filled twice in t2 and twice in t1, and multiplied by a Gaussian function centered at t = 0 before Fourier transformation, giving final digital resolutions of 2.5 Hz/point in F2 and 4.39 Hz/point in F1.

¹H- and ¹³C-NMR assignments were also obtained for the pentapeptides, which showed ¹H and ¹³C chemical shifts very close to those observed for random coil peptides (Wishart et al., 1995). No significant non-sequential NOEs were observed, indicating that there is no stable secondary structure present in these peptides.

NMR dynamics

To characterize the internal dynamics of the glycoprotein in greater detail, we have measured ¹³C-NMR relaxation rate constants (R₁, R₂, and NOE) at two temperatures (Table 2). The large {¹H}-¹³C-NOEs and relatively small R₂ values imply substantial motion on the sub-nanosecond time scale. The substantial variation in R₂ values is also evident in the linewidths in 1D ¹³C spectra (Fig. 3); the linewidths of the -Gal carbon resonances are substantially narrower than those in the NAcGal residues or the C of the amino acids.

The relaxation data were quantitatively analyzed with the Lipari-Szabo (1982) method to give values for ₀, S², and _e for each C-H vector (Table 2) as described for AFGP8 (Lane et al., 1998). The R₂ values of some residues were relatively poorly determined at 10°C as the decay was so fast, leading to a relatively large residual error. However, the fast decay is consistent with the independently observed linewidths measured from 1D ¹³C-NMR spectra (cf. Fig. 3). The three-parameter fit gave very small residuals (<1%), implying that the simple model adequately accounts for the relaxation data. However, for a few sites, especially associated with the NAc-galactosamine residue, the fit was considerably poorer, and the correlation time estimated from R₁ and the NOE alone was smaller than that implied by the large value of R₂. This could be modeled as an exchange contribution (Table 2). Such an exchange contribution to relaxation has often been observed in ¹⁵N relaxation of proteins, and is usually attributed to a slow conformational equilibrium (Clore et al., 1990; Szyperski et al., 1993; Marsden et al., 1998). It is notable that the exchange contribution is larger at 10°C than at 30°C, which is in accord with the linewidths at the two temperatures. This suggests that the exchange contribution is fast compared with the chemical shift difference, and that at lower temperatures, the system approaches the intermediate exchange regime as the rate constant decreases.

The mean value of ₀ was 4.8 ± 0.9 ns (range = 3-6.8 ns) at 10°C and 3.5 ± 0.8 ns (range = 2.3-4.7 ns) at 30°C (Table 2). For comparison, AFGP8 gave ₀ = 3.6 ± 1.8 ns at 5°C when analyzed in the same way using the previously reported relaxation data (Lane et al., 1998). Thus, the effective correlation time of the large AFGP1-5 protein is about the same at 30°C as that of AFGP8 at 5°C. The ratio of the site correlation time at 283 K to that at 303 K is 1.6 ± 0.3, which is comparable to the that expected from the temperature dependence of the viscosity of D₂O (Wilbur et al., 1976). As expected, the order parameters for the methyl carbons are much smaller than for the other carbons owing to free rotation of the methyl group (S²(methyl) = 0.11). The order parameters for the C6 positions of the two sugars are intermediate in size, indicating substantial freedom of motion of the CH₂OH group. The order parameters for C-H groups are slightly larger at 10°C than at 30°C. They also can be ranked in terms of peptide backbone (Ala and Thr) and the two sugars, such that S²(amino acid residues) > S²(NAc-Gal) > S²(Gal). This reflects the relative distance of the -Gal residues from the backbone of the molecule, and shows that these sugars are highly mobile with respect to the remainder of the molecule. In addition, the C generally have higher order parameters than the side chains.

The range of ₀ observed at either temperature is not consistent with a spherical rotor. However, the assumption of a rod-like structure (or other symmetric top rotor model) leads to physically implausible axial ratios; a rigid anisotropic rotor is an inappropriate model. In fact, the relatively low order parameters at the backbone sites (CH) indicate the presence of substantial internal mobility.

The ¹³C relaxation data show that AFGP1-5 is a rather flexible molecule, such that local segmental correlation times are comparable to those observed in the much shorter AFGP8 (Lane et al., 1998). The short effective correlation time is consistent with a persistence length of a few amino acid residues (Schwalbe et al., 1997). However, the presence of a disaccharide every third residue should make the persistence length comparable to the length of the repeating unit, namely AAT*. In addition, the sugar residues (especially -Gal) must be fully exposed to solvent, and essentially unconstrained. The simple picture of the dynamics of the glycoprotein is that of essentially a random coil showing large-scale segmental flexibility on the sub and nanosecond time scale (Bush and Feeney, 1986). The presence of only a few sequential, short-range NOEs for the amino acids is also consistent with the absence of substantial long-range order.

The value of ³J measured or the Thr residue was 4-5 Hz, which is larger than expected for either g⁺ or g rotamers (³J < 4 Hz) and much smaller than for the trans rotamer (³J > 12 Hz). This indicates that the population of the trans rotamer is relatively small. This coupling constant is similar to that observed in the peptide NAcAATAA.

Translational diffusion

The translational diffusion coefficient of AFGP1-5 was measured in D₂O at 10°C and 30°C, under the same conditions as the ¹³C-NMR relaxation measurements. Fig. 4 shows the decay data at the two temperatures, fitted to a single Gaussian. A slight improvement in the fit was obtained with a sum of two Gaussians, but the amplitude of the second Gaussian was relatively small, and the diffusion coefficient was not well determined. It is possible that this second component corresponds to some of the minor species present in the AFGP1-5 mixture. The translational diffusion coefficients were determined as 5.1 · 10⁷ cm² s¹ at 30°C and 2.71 · 10⁷ cm² s¹ at 10°C. Normalization to water at 20°C gives D_20,w = 4.8 and 4.6 10⁷ cm² s¹, respectively. Thus, the diffusion coefficient follows the expected dependence on solvent temperature (and viscosity). This normalized value is similar to that previously reported for fraction 4 (Ahmed et al., 1975) and for fractions 1-5 determined by quasi-elastic light scattering (Wilson and deVries, 1994). The normalized value is ~60% of that determined for the globular protein carbonic anhydrase (M_r = 30 kDa). As diffusion coefficients scale as M_r^1/3, this implies a particle that generates much more friction than a globular protein, and is consistent with an extended molecule that is fully solvated. Thus, the rotational and translational diffusion data indicate an unstructured, segmentally flexible molecule.

View larger version (21K): [in this window] [in a new window]   FIGURE 4   Diffusion of AFGP1-5. 1D spectra were recorded at 303 K and 283 K in D2O at 14.1 T as a function of the z-gradient strength. Intensity versus gradient strength for the regions 4.51-3.15 ppm (, ) and 1.56-0.95 (, ) ppm at 30°C and 10°C.

Interactions with water

At 5°C, two resonances at 5.75 and 5.9 ppm were observed in H₂O (Fig. 5 A) that do not appear in spectra recorded in D₂O. Thus, these protons must be readily exchangeable with solvent hydrogens, such as NH or OH. As all of the NH have been accounted for at much lower field, these resonances are likely to arise from carbohydrate hydroxyls (Poppe and van Halbeek, 1994; Leroy et al., 1985). One of these resonances appears as a doublet. This shows crosspeaks in both TOCSY (Fig. 2 A) and ROESY (Fig. 5 B) with the H2/3 of -Gal, whereas the broader peak shows an interaction with the C6 of -Gal and/or -GalNAc. These resonances belong to sugar hydroxyls, and plausibly are those on the C2/C3 of -Gal and the C6, respectively.

View larger version (28K): [in this window] [in a new window]   FIGURE 5   Hydroxyl protons in AFGP1-5. Spectra were recorded at 11.75 T and 278 K. (A) The 1D spectrum was recorded with an acquisition time of 1.5 s and a recycle time of 4 s. (B) The ROESY spectrum was recorded with a continuous wave spin lock (B1 = 4.3 kHz) centered at the water frequency with a mixing time of 40 ms. The acquisition times were 0.41 s in t2 and 0.102 s in t1. The single contour is positive (diagonal and exchange peaks), the multiple contours are negative (ROEs).

The ROESY spectrum (and NOESY) also shows crosspeaks between the water resonance, and protons in the two sugar moieties. Most of the sugar carbons (not the anomeric) have a hydroxyl group. As only two hydroxyl protons have been accounted for by the downfield-shifted resonances, it is probable that the remaining hydroxyl protons are coincident with the water resonance. These additional crosspeaks therefore represent ROEs between hydroxyl protons and the proton attached to the same carbon atom. These ROEs are likely also to be mediated by exchange with water protons (Gyi et al., 1998).

FTIR spectroscopy of AFGP1-5 and AFGP8

FTIR provides information about conformation of the peptide backbone via the frequency of the amide I band. The IR spectra of AFGP8 and AFGP1-5 are quite different (Fig. 6); the distribution of resolvable bands is more even across the frequency range for AFGP8 than for AFGP1-5. Independent of the precise assignment of the observed IR bands, AFGP1-5 is significantly different from AFGP8. The frequencies and relative peak areas are given in Table 3. The different C==O stretch frequencies have been associated with different types of secondary structure (Dong et al., 1990, 1992, 1994; Pestrelski et al. 1991), which may mean that the amide I frequency reflects the local ø,  angles, either directly through hybridization, by affecting hydrogen bonding to the solvent, or indirect local electrostatic influences of the orientation of the amino acid with respect to the plane of the peptide bond. The precise assignment of bands to secondary structure types, however, remains controversial (Surewicz et al., 1993).

View larger version (32K): [in this window] [in a new window]   FIGURE 6   FTIR spectra of AFGP1-5 and AFGP8 and peptides. Spectra were recorded as described in the Experimental section. They are displayed as second derivatives, with Gaussian band fits. (A) AFGP8; (B) AFGP1-5; (C) NAcAATAA; (D) NAcAAPAA.

In AFGP8 there are only 13 peptide bonds, including two from proline. It is rather unlikely that both helical and extended strand conformations could be simultaneously present in such a short peptide. The alternative would be that at any one moment a substantial fraction of molecules would be essentially all helix, and a similar fraction of molecules would be in an extended conformation. Similarly, for AFGP1-5, the IR data indicate substantial amounts of -helix and -sheet, yet the NMR data show that such structures must be present in low amounts.

CD and NMR experiments have indicated the presence of polyproline type II helix in AFGP8 (Bush et al., 1984; Lane et al., 1998). With FTIR analysis, a pure polyproline sample (50 mg/ml) has over 80% of the amide region centered on 1618 ± 2 cm¹ (Hays, 1998). Most FTIR users consider this wavenumber to be outside of the amide I region, and it is usually defined as the vibrational frequency of side chains (Chirgadze et al., 1975). However, Ala, Thr, and Pro do not absorb in this region. This region may also include intermolecular -sheets due to protein aggregation (for review see Dong et al., 1995). However, the IR spectrum of polyproline lacks the 1685 cm¹ peak that is also present in protein aggregates. The NMR spectra also show no evidence of significant aggregation. In the AFGP8 spectra, a substantial peak was observed at 1620 cm¹, whereas for AFGP1-5, there was a much lower intensity at this wavenumber (Table 3). However, it is not obvious that the IR frequency of the polyPro II conformation will be the same for non-Pro residues. The band at 1645-1650 cm¹ has been assigned to an unordered conformation (Table 3). However, it has been suggested, based on ultraviolet circular dichroism and vibrational circular dichroism, that this frequency corresponds at least in part to the polyPro II conformation (Woody, 1992). This suggests that the two classes of antifreeze glycoproteins differ mainly in the distribution of possible conformers, which we expect to reflect the influence of the Pro residues on the local conformation in AFGP8 (Lane et al., 1998). Because of the possible distortions of band areas introduced by taking the second derivative, and because the absorption coefficients for the different conformers are not known to be the same, the actual populations cannot be accurately calculated from the IR data.

NMR spectra of the two pentapeptides NAcAATAA and NAcAAPAA show dynamically unstructured peptides (data not shown). The FTIR spectra of the same peptides show numerous bands ranging from 1610 to 1680 cm¹, analogously to those observed in the AFGPs, and the pattern of observed bands is significantly different for the Pro- and Thr-containing peptides (Fig. 6). This clearly shows that these short peptides populate the allowed regions of Ramachandran space, and interconvert slowly compared with the difference between their respective vibrational frequencies. The similarity to the FTIR spectra of the AFGPs indicates that the carbohydrate is not responsible for the slow interconversion rates between the various conformations.

MD calculations

From molecular dynamics simulations in water it is possible to assess the mobility of protein backbone atoms on the sub-nanosecond times scale and compare them with experimental NMR data. Typically, for folded proteins, the order parameters of NH and CH vectors are in the range 0.8-0.9 in defined secondary structure elements, and the local correlation times are usually all very similar (Clore et al., 1990; Kördel et al., 1992; Lüginbuhl et al., 1997; Nicholson et al., 1996). Unstructured portions of the protein, especially at chain termini, generally show a decreased local correlation time and lower order parameters (van Heijenoort et al., 1998; Alexandrovich et al., 1999). These general results are reproduced by MD simulations (J. O. Trent, S. J. Smerdon, and A. N. Lane, unpublished data). Preliminary calculations in both water (400 ps) and in vacuo (1-5 ns) show that although the backbone torsions ø and  oscillate around mean values with an rmsd of ~30°, transitions between different regions of Ramachandran space are comparatively rare (~1-3 times every nanosecond), and not obviously correlated between neighboring dipeptide units. This is slow on the infrared time scale, such that separate bands for extended, turn, and helix would give rise to separate bands. It is, however, fast on the time scale of NMR coupling constants and chemical shifts, which would therefore not give rise to separate lines, and coupling constants should reflect averaged values. The high-frequency fluctuations of C and C are fast on the NMR relaxation time scale, and would be expected to give NMR order parameters substantially less than unity.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

According to the NMR data, AFGP8 has no long-range order, but displays significant local order (Lane et al., 1998). In contrast, AFGP1-5 is a dynamically disordered molecule that shows neither significant long-range nor short-range order on a time scale from 100 ps to milliseconds. At face value the IR data indicate that although the AFGP fractions are different at the backbone level, they both contain substantial amounts of secondary structure elements. The IR results can be rationalized with the NMR data by considering the time scales probed by the two methods. The spread of wavenumbers for different regions of Ramachandran space is ~60 cm¹ centered at 1650 cm¹ (Table 1), which is equivalent to a characteristic time scale of >10¹² s¹. Interconversion between conformational states on a slower time scale leads to individual peaks corresponding to all conformations that are present in an ensemble. It has recently been shown by Raman spectroscopy that unordered polypeptides show several vibration frequencies, which was interpreted in terms of an ensemble of conformation states (Sane et al., 1999). This is supported by molecular dynamics simulations of Ala-8 peptides in water, which showed substantial high-frequency fluctuations (a few picoseconds) about ø and , and rarer transitions between the _R, _L, and polyPro II regions of Ramachandran space (Sreerama and Woody, 1999). In contrast, the same interconversion rates would be fast on the NMR chemical shift time scale (~10³ s¹), and therefore the NMR resonances appear at a single averaged frequency. This is similar to the observed behavior of short hormonal peptides in solution (Feeney, 1977). In addition to the relatively slow rotational motions of groups of residues (~5-8) on the 3-7-ns time scale, there are higher-frequency motions (tens to hundreds of picoseconds) of substantial amplitudes, as shown by the NMR order parameters. Values of S² ~0.5 are roughly equivalent to angular fluctuations of ~35-40° on the sub-nanosecond time scale. Hence, there are large amplitude motions of the C-H vectors that require substantial changes in the ø,  angles, presumably within the allowed regions of Ramachandran space. There is an ensemble of conformations interconverting rapidly on the NMR time scale, and slowly on the IR time scale. Hence, in IR even a random coil will show different bands corresponding to an ensemble of molecules that occupy different regions of Ramachandran space, whereas in NMR there will be a single resonance for each proton or carbon atom. The simplest interpretation of the IR and NMR data together is that the molecules are flexible, such that each dipeptide unit occupies several regions of Ramachandran space for various amounts of time. Because ø in proline is essentially fixed by the covalent structure, proline-containing dipeptides have fewer degrees of freedom than other dipeptides. This can be expected to restrict the conformational space accessible to the proline-containing AFGP8 compared with the proline-free AFGP1-5. For AFGP1-5, a wide variety of backbone torsions are significantly populated, giving rise to numerous bands in the IR spectrum, whereas in AFGP8 the degree of Ramachandran space that is accessed is lower than in the proline-free AFGP1-5, giving rise to a narrower distribution of IR bands having significant intensity (>15-20%). This is supported by the NMR and FTIR spectra of the pentapeptides. These findings have implications for the analysis of secondary structures of proteins by IR, as the so-called unordered frequency does not account for all of the non-secondary structure peptides. The helical, turn, and extended segments can be overestimated, which is likely to be more of a problem in partially folded proteins. Even in denatured or unstructured proteins such as casein, the observation of a very broad IR band (Byler and Susi, 1986) does not necessarily imply a single conformation, but rather an unresolved envelope of conformers.

In conclusion, the combination of NMR, IR, and MD is a powerful means of obtaining detailed information about the conformational properties and dynamics of macromolecular systems. It is becoming increasingly clear that biological activity is not restricted to compact semi-rigid proteins. However, many of the methods of analysis and the philosophical underpinnings of protein action are couched in terms of single conformations or simple conformational changes. The activity of highly flexible molecules that may or may not fold into a unique conformation on binding to a "receptor" poses a challenge for conformational analysis. Not least is the need to describe ensemble properties and the dynamic of interconversion within the ensemble. It is usually not possible to obtain sufficient experimental data to determine the conformations present in such an ensemble. However, the combination of experimental methods and molecular dynamics may be able to provide a description of the general conformational properties and the extent to which conformational space is populated, albeit in a non-unique manner.