Native-like -structure in a Trifluoroethanol-induced Partially Folded State of the All- -sheet Protein Tendamistat

Transcription

1 J. Mol. Biol. (1996) 260, Native-like -structure in a Trifluoroethanol-induced Partially Folded State of the All- -sheet Protein Tendamistat Nancy Schönbrunner 1, Josef Wey 1, Joachim Engels 2, Holger Georg 3 and Thomas Kiefhaber 1 * 1 Biozentrum der Universität Basel, Abteilung Biophysikalische Chemie Klingelbergstr. 70, CH 4056 Basel, Switzerland 2 Institut für Organische Chemie der Universität Frankfurt, Marie Curie Str 11, D Frankfurt/Main Germany 3 Institut für Biophysik Universität Freiburg Albertstr. 23, D Freiburg, Germany *Corresponding author The effect of trifluoroethanol (TFE) on the structure of the all- -sheet protein tendamistat was investigated. At low concentrations TFE induces cooperative loss of the native tertiary structure leading to a partially folded state. The loss of specific side-chain interactions in the transition from the native state to the TFE-induced state is demonstrated by the disappearance of the CD bands in the aromatic region, a reduced chemical shift dispersion in the one-dimensional 1 H NMR spectrum and a broad, uncooperative thermal unfolding transition of the partially folded state. An increased line-width of the NMR bands in the TFE state compared with the unfolded state suggests the presence of multiple, rapidly interconverting conformations. Hydrogen-exchange studies of amide proteins in the TFE state reveal the existence of defined hydrogen bonds at the same locations as in the native state, but with largely reduced stability. This suggests the presence of most of the native -sheet structure. These results are supported by Fourier transformed IR measurements, which show nearly the same amount of -structure in the TFE state and in the native state. Far UV CD spectroscopy suggests the induction of some -helical structure upon addition of TFE, which appears to be located mainly in regions corresponding to loops or random structure in the native state and which seems to represent fluctuating conformations with preferred backbone angles rather than stable, hydrogen-bonded -helices. These results show that stable non-local interactions, as they occur in -sheets, can form in the absence of specific side-chain interactions. The presence of a subset of the native long-range interactions and the absence of stable non-native interactions suggests that the observed partially folded state might represent an early intermediate on a hierarchial folding pathway of tendamistat Academic Press Limited Keywords: protein folding; protein stability; tendamistat; hydrogen exchange; -sheets Introduction The large number of possible conformations of a polypeptide chain makes it impossible for a protein to find its native state by random search (Levinthal, 1969; Wetlaufer, 1973). This led to the postulate of distinct folding pathways, including partially Abbreviations used: CD, circular dichroism; FTIR, Fourier transformed infra-red; NMR, nuclear magnetic resonance; 1D, one-dimensional; 2D, two-dimensional; TFE, trifluoroethanol; Gdm, guanidinium; Gdn, guanidin. folded intermediates, which reduce the conformational freedom of the folding polypeptide chain. Experimental work on the mechanism of protein folding has consequently focused on the detection and on the characterization of partially folded states (Baldwin, 1975, 1993; Kim & Baldwin, 1982, 1990; Jaenicke, 1987; Dobson, 1992). In the case of several proteins, partially folded conformations can be induced at low ph or by addition of alcohols (Kuwajima, 1989; Dobson, 1992). The study of these structures gives valuable information on the conformational properties of proteins in the absence of specific side-chain interactions. For /96/ $18.00/ Academic Press Limited

2 TFE-induced State of Tendamistat 433 some proteins, it was shown that the conformations observed under these conditions share many properties with kinetic intermediates that accumulate transiently at early stages in the folding process (Hughson et al., 1990; Buck et al., 1993; Jennings & Wright, 1993; Shiraki et al., 1995). Thus, the investigation of partially folded states observed under equilibrium conditions is thought to provide insight into early events in the process of protein folding, before the formation of specific side-chain interactions occurs. Despite a wealth of information on the structure of partially folded states their actual role in folding is still enigmatic. Trifluoroethanol (TFE) is known to induce non-native partially folded states in proteins and to increase the helical content of short peptides. TFE-induced structures in proteins were shown to have a high -helical content and are usually characterized by the absence of specific tertiary interactions (Shiraki et al., 1995). Goto and coworkers found a close relationship between the -helix contents of proteins in the presence of TFE and their propensities to form -helix predicted from the amino acid sequence (Shiraki et al., 1995). This is in agreement with theoretical studies on the effect of TFE on protein structure, which suggested two different modes of action: (1) disruption of hydrophobic interactions; and (2) enhancement of local helix-stabilizing interactions (Thomas & Dill, 1993). The induced partially folded states have consequently lost the specific side-chain interactions and are chiefly stabilized by local backbone interactions. Little is known about the effect of TFE on -sheet structures, although there is evidence for -sheet to -helix conversion in some proteins and peptides (Sönnichsen et al., 1992; Fan et al., 1993; Yang et al., 1994; Liu et al., 1994; Kemmink & Creighton, 1995; Shiraki et al., 1995). We used the small -amylase inhibitor tendamistat (74 amino acid residues, 7952 Da) as a model system to investigate the effect of TFE on protein -sheets. The NMR and the X-ray structures of the protein have been solved (Kline & Wüthrich, 1985; Pflugrath et al., 1986; Kline et al., 1986; Braun et al., 1989; Figure 1). The secondary structure of tendamistat comprises only -strands (69%), which are located in two three-stranded anti-parallel -sheets, in addition to loops or random structure (31%). The unfolding and refolding reactions of tendamistat are completely reversible (unpublished results) making it a good model system to investigate the folding mechanism and the intrinsic properties of protein -sheets. The native structure of tendamistat is stabilized by two disulfide bonds connecting residues 11 and 27, and residues 45 and 73. Here, we show that TFE induces a partially folded state in tendamistat that has lost most of its specific side-chain interactions but that retains TFE state denotes the structure formed in wild-type tendamistat at concentrations of TFE beyond the cooperative unfolding transition (>8 M TFE). Figure 1. Schematic representation of the structure of tendamistat. The disulfide bonds connecting Cys11/ Cys27 and Cys45/Cys73 are shown as ball and stick models. The figure was drawn using the program Molscript (Kraulis, 1991). nearly all of the native -sheet structure with stable, well-defined hydrogen bonds. In the loop regions of the native protein, TFE seems to induce fluctuating helical structure. The presence of a subset of the native interactions in the TFE-state suggests that it might represent an important early intermediate in the kinetic process of tendamistat folding and it supports a hierarchical pathway for -sheet formation in proteins. Results Induction of a partially folded state by TFE Figure 2A shows CD spectra in the amide region of tendamistat in the presence of various concentrations of TFE. The far-uv CD spectrum of the native protein has a characteristic maximum near 230 nm and a minimum at 190 nm. Upon addition of TFE the spectrum changes and at 5 M TFE two distinct minima at 222 nm and at 208 nm, characteristic for -helical structure, become visible. Above 5 M TFE the shape of the spectrum does not change, but the intensity of the -helical bands increases slightly. Plotting the ellipticity at 222 nm versus the TFE concentration (Figure 3A) suggests a cooperative loss of the native structure in the transition from the native state to the -helical state between 4 M and 7 M TFE, and an uncooperative linear further increase in -helical content above 7 M TFE. The existence of an isodichroic point at 203 nm

3 434 TFE-induced State of Tendamistat Figure 2. A, Far-UV CD spectra of tendamistat (wild-type) at ph 2, 25 C in the presence of increasing concentrations of TFE. In order of increasing ellipticity at 190 nm: 0 to 2.8 M; 5.6 M; 6.3 M; 6.9 M; 8.3 M; 9.7 M; 11.1 M; 12.5 M. B, Near-UV CD spectra in 0 M and 12.5 M TFE at ph 2, 25 C. C, Thermal denaturation of tendamistat in 8.3 M TFE, ph 2, as monitored by ellipticity at 220 nm. provides evidence for a two-state transition. Fitting the unfolding transition at ph 2 to a two-state model gives G 0 (H 2 O) of 27.1 kj/mol and an m-value (d G 0 /d[tfe]) of 6.9 kj mol 1 M 1. The values for G 0 (H 2 O) and the m-value correspond well to the values measured in GdmCl-induced unfolding under the same conditions ( G 0 (H 2 O) = 28.4 kj mol 1 ; m = 7.0 kj mol 1 M 1 ). Near-UV CD spectra indicate that the defined tertiary structure of the protein is lost in the transition from the native state to the TFE state (Figure 2B). The absence of a specific tertiary Figure 3. A, TFE-induced cooperative structural transition at 25 C as monitored by ellipticity at 222 nm for tendamistat: ( ), wild-type, ph 2; ( ), C11A/C27T, ph 2; ( ), C45A/C73A, ph 3. The lines represent the least-squares fit assuming a two-state mechanism and linear dependence of the baselines for the native and TFE states according to the formalism of Santoro & Bolen (1988). The midpoints of the transitions lie at 5.7 M (wild-type, ph 2), 2.6 M (C11A/C27T, ph 2) and 3.6 M TFE (C45A/C73A, ph 3). B, Plot of the midpoints of the TFE-induced cooperative structural transition against the G NU values derived from the thermal denaturation of the protein at the respective conditions. The line represents a linear least-squares fit of the data with correlation coefficient structure and the loss of cooperativity is supported by a broad non-cooperative unfolding transition upon thermal denaturation monitored by the decrease in the negative ellipticity at 222 nm (Figure 2C). Effect of protein stability on the TFE state To test the influence of disulfide bonds and of protein stability on the TFE-induced transition, the effect of TFE on two variants of tendamistat, each with one of the two disulfide bonds replaced, was determined at different ph values (Figure 3A). For all variants and under all conditions investigated, induction of -helical structure was observed upon addition of TFE. As in the wild-type protein, TFE

4 TFE-induced State of Tendamistat 435 Figure 4. 1D 1 H NMR spectra of tendamistat: A, from 5.2 to 7.7 ppm (aromatic region) and B, from 0.5 to 3.5 ppm (aliphatic region) for the native state in 2 H 2O, ph 3, 25 C; the TFE state in 9.7 M TFE-d 3/ 2 H 2O, ph 3, 25 C and the denatured state, in 5.2 M Gdm 2 HCl, 2 H 2O, ph 2, 45 C. cooperatively unfolds the protein in a first step and, upon further raising the TFE concentration, the negative ellipticity at 222 nm increases linearly. The midpoint of the TFE-induced cooperative unfolding transition correlates linearly with the stability of the native protein towards thermal unfolding (Figure 3B), suggesting that the primary effect of TFE is the destruction of the native tertiary structure. The helical content of the TFE state is slightly influenced by the number of disulfide bonds. Both disulfide mutants show a minor increase in helical content in TFE compared with the wild-type protein, but the slope of the uncooperative induction of helical structure is steeper for the wild-type protein containing two disulfide bonds. Further characterization of the TFE-induced structure was performed at low ph in order to be able to use less TFE to induce the partially folded state. Intrinsic helix-forming propensity of tendamistat By comparing the effect of TFE on several proteins, Goto and co-workers found a good correlation between the intrinsic helix-forming propensity of a protein and the amount of helical structure induced by TFE (Shiraki et al., 1995). We used two different algorithms to determine the intrinsic helical propensity of tendamistat. Both methods predict very low levels of helicity. The joint prediction method (Nishikawa & Noguchi, 1991) suggests that the last six amino acid residues of tendamistat have some helical propensity. The remainder of the protein is predicted to have preference for random coil or -strand conformation. A stable -helix formed by the last six residues would correspond to 8% helical content referred to the complete protein. The helix prediction method of Serrano and co-workers (Munoz & Serrano, 1994, 1995a,b) suggests less than 1% helicity for both the wild-type protein and for the two disulfide mutants investigated. Tertiary interactions in the TFE state To detect specific side-chain interactions in the TFE state, 1D 1 H NMR spectra were recorded. Figure 4 compares spectra of unfolded and native

5 436 TFE-induced State of Tendamistat tendamistat with the spectrum of the TFE-induced state. The bands in the aliphatic region ( 0.5 to 3.5 ppm) in the spectrum of the TFE state show a largely reduced chemical shift dispersion comparable with the spectrum of the unfolded protein, indicating the loss of the rigid tertiary structure in the hydrophobic core of the protein (Figure 4B). The chemical shift dispersion in the aromatic region (6.5 to 8 ppm) and in the region of the downfield-shifted C protons (5 to 6 ppm) is also largely reduced in the TFE state compared with the native state (Figure 4A). However, there are significant differences in the chemical shifts in the aromatic region between the TFE state and the unfolded state, which might indicate some specific side-chain interactions in the TFE-induced state. Interestingly, all bands in the spectrum of the TFE state are much broader than in the unfolded state, indicating the presence of multiple conformations in the TFE state that interconvert rapidly on the NMR timescale. Reference measurements on short peptides performed under identical conditions show that the line-broadening is not due to viscosity effects caused by the presence of TFE. Stability of hydrogen bonds in the TFE-state Hydrogen/deuterium exchange of amide protons was performed to monitor the presence of stable amide hydrogen bonds in the TFE state. Amide protons that participate in hydrogen bonds are protected from exchange with solvent protons or deuterons. The exchange rate of each individual amide proton is proportional to the stability of the respective hydrogen bond, if exchange occurs under EX2 regime, i.e. if structural opening and closing of the hydrogen bond are both faster than the chemical exchange step (Hvidt, 1964). In order to detect stable hydrogen bonds in the TFE state of tendamistat, we performed hydrogen-exchange experiments at ph 3, where chemical exchange is extremely slow and structural opening and closing of the TFE state is fast. Figure 5A compares bulk hydrogen-exchange kinetics of all amide protons in the TFE state and in the completely unfolded protein at ph 3, 25 C. Exchange of unprotected amide protons, corresponding to the situation in the unfolded protein, is approximated by the sum of the individual exchange rates of each amide proton of tendamistat. These values are known from studies on model peptides (Bai et al., 1993; Connelly et al., 1993). The resulting curve for completely unprotected amide protons can be fit to the sum of two exponentials. The majority of the protons (39) exchange with comparable rates ( = 1/k HX = ten minutes). The remaining protons (31) exchange five times faster ( = two minutes) and correspond to sites next to charged residues. The given exchange rates represent average values of many individual exchange rates of similar magnitude. Amide proton exchange in the TFE state is significantly slower than predicted for Figure 5. Bulk hydrogen-exchange kinetics of all amide protons exchanging ( ) from the TFE state at ph 3, 25 C. The exchange kinetics are well described by the sum of three exponentials with 1 = 10.5 minutes (A 1 = 32.4), 2 = 41.7 minutes (A 2 = 24.5), 3 = minutes (A 3 = 8.0). ( ) Predicted exchange rates of the amide protons in unfolded tendamistat calculated from the intrinsic rates at ph 3, 25 C. The unprotected exchange kinetics are well described by the sum of two exponentials with 1 = 2.2 minutes (A 1 = 31), 2 = 10.0 minutes (A 2 = 39). completely unprotected amide protons in tendamistat and it can be described by a treble exponential curve. The rate of the fastest phase is identical with the rate of the slower-exchanging protons in the unfolded protein ( = ten minutes) and accounts for the exchange of about 32 protons. The intermediate phase has a time-constant of 42 minutes and an amplitude of 25 protons. The remaining eight amide protons exchange very slowly with a relaxation time of 350 minutes. As in the exchange kinetics of the unfolded protein, the measured rate-constants represent average values of several amide protons with similar rate-constants. Control experiments on model peptides showed that TFE does not significantly influence the intrinsic rates of amide hydrogen exchange (Buck et al., 1993; and unpublished results). Exchange rates in 9.7 M TFE were slightly faster than those reported for model peptides in water, probably due to changes in the effective ph value in the presence of TFE. These results reveal that amide proton exchange is retarded in the TFE state compared with the unfolded state, suggesting the presence of stable hydrogen bonds in the TFE state. Location of hydrogen bonds in the TFE state To determine the location of the protected amide protons in the TFE state, hydrogen exchange of protonated tendamistat in 9.7 M TFE-d 3 in 2 H 2 O, ph 3 was allowed for 70 minutes. At this point, exchange of the fast-exchanging protons is complete (see Figure 5A). In the next step tendamistat was diluted to final conditions of 0.1 M TFE, ph 4, where the native state forms very rapidly, whereas hydrogen exchange is still very slow. This procedure freezes the protonation pattern at the

6 TFE-induced State of Tendamistat 437 Table 1. The amide proton protection factors (k exp/k obs) for tendamistat in the native state ph 3, 50 C and in the TFE state at ph 3, 25 C Protection factor Protection factor Secondary Native Secondary Native Residue structure state TFE state Residue structure state TFE state Thr2 <200 ND Asp Ala Thr41 2(3) Val12 1(1) Glu42 2(3) Thr13 1(1) Gly43 2(3) 100 ND Leu14 1(1) 30 ND Leu44 2(3) Tyr15 1(1) Cys45 2(3) Gln16 1(1) <3 ND Tyr46 2(3) Ser17 1(1) 600 ND Ala47 2(3) 50 ND Tyr20 1(2) Val48 2(3) Ser21 1(2) Ala49 2(3) 100 ND Gln22 1(2) Gly ND Ala23 2(2) Gln52 1(3) Asp24 1(2) Ile53 1(3) 20 ND Asn25 1(2) 3000 ND Thr54 1(3) Gly ND Thr55 1(3) <3 ND Cys Val56 1(3) Thr30 2(2) 4 ND Gly57 1(3) 500 ND Val31 2(2) Asp58 2(3) <10 ND Thr32 2(2) 7 ND Ala67 2(1) 300 ND Val33 2(2) Arg68 2(1) 150 ND Lys34 2(2) Tyr69 2(1) Val35 2(2) Leu70 2(1) 200 ND Val36 2(2) 10, Ala71 2(1) Tyr37 2(2) 4000 ND Arg72 2(1) Glu ND Cys73 2(1) The extent of exchange in the TFE state after 70 minutes was determined after quenching and reconstitution to the native state, as described in Materials and Methods. The expected degree of exchange after 70 minutes was calculated according to reference data on peptide models (Bai et al., 1993; Connelly et al., 1993). The previously published hydrogen-exchange kinetics of tendamistat at ph 3, 50 C (Qiwen et al., 1987) were used to calculate the protection factors for the native state. The values for all residues with significant protection (P > 200) in the native state, as well as all residues located in -sheets are shown. The values for the TFE state represent lower limits, since exchange in model peptides in the presence of TFE was slightly faster than in water. ND, Not determined, no peak observed in the 1 H COSY spectrum after previous exchange in TFE-d3. stage of the 70 minute exchange pulse, since exchange in native tendamistat is extremely slow (Qiwen et al., 1987) and it allows the identification of the exchange-protected amide protons using 2D NMR spectroscopy. As expected from the bulk exchange experiments, different classes of protons can be distinguished in the 2D spectrum. Some of the amide sites had completely lost their protons during the 70 minute exchange pulse in 9.7 M TFE, while others had retained some degree of protonation. Table 1 shows protection factors of the individual amide sites in tendamistat calculated from the relative intensity of the amide proton peaks in the 2D spectrum. For comparison, the protection factors of the native state at ph 3, 50 C are shown. The protection factors in the TFE state are much smaller than the values in the native protein, but there is a good correlation between the protected sites in the native state and in the TFE state. The most highly protected amide protons of the native protein also show the strongest factors in the TFE state. One exception is the region from Asp40 to Glu42, which shows little protection in the native state but relatively high protection in the TFE state. As in the native state, the central strands of the two -sheets (Tyr20 to Asp24 and Val33 to Val36) are the most strongly protected from exchange in the TFE state and they show continuous protection in the complete strand, whereas the outer strands of the sheets show an alternating protection pattern. This kind of protection pattern is characteristic for hydrogen-bonding in -sheets. Figure 6 shows that the location of the hydrogen bonds in the native state is almost identical with that of the exchange-protected protons in the TFE state. This suggests that most of the native -structure is conserved in the TFE state, although with a much lower stability. Some amide protons located in -sheets of the native state are not highly protected in the native state (e.g. G43) and, consequently, these protons cannot be monitored with the applied hydrogen-exchange technique, since they readily exchange in the native state during sample preparation. The magnitude of protection factors observed for amide protons in the TFE state of tendamistat is comparable with values found in molten globule intermediates of other proteins (Hughson et al., 1990; Baldwin & Roder, 1991; Buck et al., 1993)

7 438 TFE-induced State of Tendamistat although most of these studies have been carried out at lower temperatures. The observed protection cannot be caused by small amounts of native molecules present under the experimental conditions or by EX1 exchange kinetics. Kinetic experiments show that the TFE state is formed rapidly from the unfolded state (k 21 < 1/3 s 1 ) and the intrinsic rate for hydrogen exchange is very low at ph 3 (k 23 1/500 s 1 ). Under these conditions the observed hydrogen-exchange kinetics are under EX2 regime, i.e. they depend directly on the fraction of non-hydrogenbonded amide protons and on the intrinsic exchange rate (see Materials and Methods). This is confirmed by a nearly tenfold ph dependence of the observed exchange rates in the TFE state (data not shown). We can also rule out residual amounts of native protein as the origin of observed protection from hydrogen exchange. The highest protection factors observed in the TFE state are larger than 100, which would correspond to more than 99% native protein if hydrogen exchange were to occur from the native state under EX2 conditions (see Materials and Methods). Figure 3B shows that the transition from the native state to the TFE state is complete at 5 M TFE at low ph, clearly ruling out the presence of a major fraction of native molecules in 9.7 M TFE. The experiments performed to identify the location of the hydrogen-bonded amide protons cannot rule out the presence of protected amide protons in the TFE state, which are located at sites that are not protected in the native state. These protons would be lost during the acquisition of the 2D NMR spectra. To detect such amide protons, the same 70 minute exchange pulse in 9.7 M GdmCl at ph 3 was performed as described above, but this time consecutive 1D NMR spectra were recorded immediately after the protein was transferred into native solvent conditions. Since the intrinsic exchange rates of the individual unprotected amide protons are low at ph 3 (see Figure 5A) and 1D NMR spectra can be recorded very rapidly, this procedure allows the detection of hydrogen exchange even of unprotected amide protons in the native state. No protons are lost in the first 15 hours after tendamistat was transferred from the TFE state to the native state (data not shown), indicating that the hydrogen bonds formed in the TFE state are a subset of the native hydrogen bonds and that additional hydrogen bonds are not formed in the TFE state. FTIR measurements of the structure of the TFE state The results from hydrogen-exchange experiments suggest that the stable amide hydrogen bonds present in the TFE state are a subset of the native hydrogen bonds that are exclusively located in the -sheet structure of native tendamistat. CD spectra, in contrast, indicate the induction of some -helical structure by TFE. In order to monitor the presence of -structure, FTIR spectra of tendamistat Figure 6. Schematic representation of the backbone structure of tendamistat with the hydrogen bonds formed in the native state indicated as arrows pointing from the NH donor to the carbonyl acceptor. Protected protons in the TFE state are indicated in black for the more strongly protected amide protons (P > 10) and in gray for the weakly protected amide protons (3 < P < 10).

8 TFE-induced State of Tendamistat 439 Figure 7. A, FTIR spectra of tendamistat in ( ) the native state, 20 mm potassium phosphate/h 2O, ph 3, 25 C; ( ) the TFE state, 9.7 M TFE in 20 mm potassium phosphate/h 2O, ph 3, 25 C. B, The difference FTIR spectrum of tendamistat (TFE native state). were recorded in the presence of 9.7 M TFE. In contrast to CD spectra, which are largely dominated by -helical structure, FTIR is very sensitive to the presence of -sheets. Figure 7 compares the FTIR spectrum of the TFE state with the spectrum of the native protein. The native state and the TFE-induced state show nearly identical spectra, with an absorbance maximum near 1640 cm 1, indicative for -sheet structure. The difference spectrum reveals a small decrease in the intensity near 1630 cm 1 and a comparable increase in the intensity at 1670 cm 1 in the TFE state. This could indicate the loss of a small amount of -sheet structure. Since the bands in the FTIR spectrum are very broad, and both random coil and -helices show bands near 1660 cm 1, we cannot judge whether the small positive band in the difference spectrum near 1670 cm 1 is due to formation of -helical structure. Discussion Effect of TFE on protein structure Co-solvents, such as alcohols, are commonly used to induce partially folded states in proteins. The basic physical mechanism by which alcohols influence protein structure is still unsolved. A consideration of the physico-chemical properties of TFE may help to interpret their mechanism of action and the relevance of the induced structural changes to protein folding. TFE has a low dielectric constant (about one-third that of water; Llinas & Klein, 1975), leading to a strengthening of charge interactions, including those between partial charges as they occur in hydrogen bonds. Simultaneously, TFE is a weaker base (pk a 8.2 versus 1.8 for water) but slightly more acidic than water (pk a = 12.4 versus 15.3; Llinas & Klein, 1975). It is thus a much weaker hydrogen-bond acceptor and slightly stronger donor than water (Nelson & Kallenbach, 1986). Additionally, the bulky -CF 3 group sterically hinders the interaction with the peptide backbone. These combined effects, of strengthening hydrogen-bond interactions in the protein while itself serving as a poor hydrogenbond partner, lead to an overall strengthening of intramolecular hydrogen-bonds. Furthermore, TFE acts to destroy hydrophobic interactions possibly by causing changes in water structure (Nelson & Kallenbach, 1986) and possibly by interacting with hydrophobic groups on the protein surface (Arakawa & Goddette, 1985). In summary, TFE decreases the strength of hydrophobic interactions while it increases intramolecular hydrogen bonds. This may mimic the environment of a protein during early folding, before fixed strong tertiary interactions have formed and when backbone conformational preferences override long-range interactions. TFE-induced structure of tendamistat We investigated the effect of TFE on the structure of the small -amylase inhibitor tendamistat. The secondary structure of tendamistat consists of two -sheets, each containing three strands. None of the residues is in an -helical conformation in the native state (Figure 1). Upon addition of TFE, the protein loses its defined tertiary interactions in a cooperative manner. The presence of an isodichroic point at 203 nm suggests that the loss of the native structure follows a two-state transition. The midpoint of the cooperative transition depends linearly on protein stability in the wild-type protein and in the two disulfide mutants investigated (Figure 3B), confirming the denaturant character of TFE as a co-solvent. This is in agreement with theoretical studies on the mechanism of alcohol denaturation by Dill and co-workers, who predict the major effect of TFE on protein structure to be disruption of the side-chain packing. The G 0 (H 2 O) values for the TFE-induced transition and for the GdmCl-induced unfolding of tendamistat agree very well (27.1 kj mol 1 versus 28.4 kj mol 1 ) and the m-values, which are a measure of the difference in denaturant accessibility of the native state and the unfolded state, are almost identical (6.9 kj mol 1 M 1 versus 7.0 kj mol 1 M 1 ). This could be the result of similar interactions of both co-solvents with proteins, and it confirms the denaturant character of TFE. The good agreement between the thermodynamic parameters obtained by GdmCl-induced denaturation and by TFE-induced unfolding suggests that the molar concentration of TFE, rather

9 440 TFE-induced State of Tendamistat than %, v/v, should be used for quantitative data analysis. In contrast to GdmCl-induced denaturation and thermal melting, unfolding by TFE does not produce a random coil conformation in tendamistat but a partially folded state with increased helical content, as judged by far-uv CD (Figure 2) and with most of the native -sheets still intact, as judged by hydrogen exchange (Figures 5 and 6) and by FTIR measurements (Figure 7). The ability to form the TFE-induced partially folded state is independent of the presence of the disulfide bonds and of the ph value of the solvent. Both single cystine variants of tendamistat show the same structural transition as the wild-type protein at low concentrations of TFE, but both have a slightly increased helical content in the TFE state. This suggests that the absence of each disulfide bond allows some additional helical structure to form. The effect is more pronounced in the variant where the disulfide bond has been removed. This disulfide bond is located at the end of a -hairpin in the native state and it encloses a shorter loop than the disulfide bond. The absence of the disulfide bond might allow some additional -helical conformation to be formed in this region in the TFE state. Near-UV CD and 1D NMR experiments show the loss of the majority of the specific side-chain interactions in the transition from the native state to the TFE state. Hydrogen-exchange experiments show the existence of protected amide protons in the TFE state that are identical with the most highly protected amide protons in the native state (Table 1 and Figure 6). Regions in the central strands of both native -sheets are involved in stable hydrogen bonds in the TFE state, and in these regions protection is observed for several consecutive amide protons. Amide protons located in the outer strands of the native sheets, in contrast, show an alternating protection pattern in the TFE state. The same protection pattern from hydrogen/deuterium exchange is observed in the native state and it is characteristic for -sheets, suggesting that the native -sheets are present in the TFE state. The high protection factor in the region from Asp40 to Gly43 in the TFE state might indicate that this strand is slightly extended in the TFE state compared with the native state, and it could reflect small changes in the location of the ends of the -sheet structure in the absence of specific side-chain interactions. The presence of native-like -sheet structure in the TFE state seems surprising, since CD measurements suggest induction of helical structure. From studies on model peptides the molar ellipticity for 100% helical structure is known to be around 42,000 deg cm 2 dmol 1 (C. A. Rohl & R. L. Baldwin, personal communication), which leads to a maximum value of about 25% induced helical structure in the TFE state of wild-type tendamistat, assuming that only -helical structure and random coil structure are present. If, in addition, the presence of -sheet structure is taken into account, which gives rise to weak negative CD bands around 222 nm, the estimate of induced helical structure drops to less than 25%. The secondary structure estimation program of Yang et al. (1986) gives values of 23% helical structure, 46% -sheets and 31% random coil structure for the TFE state of tendamistat. This shows that, although the shape of the CD spectrum is dominated by a small amount of helical structure, -sheets contribute significantly to the CD spectrum of the TFE state. The absolute values of the secondary structure estimation are, however, very prone to error. For the native state of tendamistat, which contains 70% -sheet structure, the secondary structure estimation algorithm suggests 18% -structure, 40% random coil and 42% turns. In native proteins, however, additional errors are introduced by the contributions of aromatic amino acid residues to the CD signal in the region of 225 nm (Adler et al., 1973). Hydrogen-exchange experiments show that no additional proton becomes involved in a hydrogen bond in the transition from the native state to the TFE-induced state, suggesting that the induced -helical conformations are of very low structural stability. The presence of most of the native -structure in the TFE state is supported by FTIR measurements, which show nearly the same amount of -structure in the TFE state and in the native state (Figure 7). The minor changes in the absorption bands around 1660 cm 1, which are assigned to -helices and random-coil structure, could indicate some conversion of -sheet to -helix. Studies on model peptides showed that -helical structure in aqueous solutions containing TFE exhibit a maximum at 1655 cm 1 (Haris et al., 1994). The 1D 1 H NMR spectrum of the TFE state does not show any chemical shift dispersion of C protons in the region of 5 to 6 ppm, which is characteristic for -structure. This could indicate very loosely packed side-chains of the -sheets. The high level of sensitivity of chemical shift dispersion in 1D 1 H NMR spectra towards changes in mobility of the protein structure was demonstrated for an equilibrium intermediate of equine lysozyme (Van Dael et al., 1993) and for a kinetic intermediate in ribonuclease A unfolding, where the chemical shifts dispersion of methyl protons was lost before major changes in CD or fluorescence properties could be detected (Kiefhaber et al., 1995). The presence of amide protons that are protected from hydrogen exchange indicates that stable hydrogen bonds are formed between the -strands and suggests a well-defined topology of the sheets. Interestingly, the region of the central strand of sheet 2 (Val31 to Glu38), which includes some of the most strongly protected amide protons in the TFE state, is predicted to have the highest -strand propensity of the tendamistat sequence. The increased chemical shift dispersion in the 1D 1 H NMR spectrum in the aromatic region in the TFE state compared with the unfolded state might indicate some specific

10 TFE-induced State of Tendamistat 441 side-chain interactions in the core region of the TFE-induced structure. These results show that far-uv CD spectroscopy is not suited to detect -structures in proteins. Due to the very strong bands of -helices, even a small amount of helical structure can completely dominate the shape of CD spectra in the far-uv region, even in the presence of major amounts of -sheet structure. Thus additional methods should be used to detect -sheet structures. Effect of TFE on -structure In peptide fragments corresponding to the sequence of -regions in native proteins, mainly unordered structure or -helices are observed in the presence of TFE (Dyson et al., 1992; Viguera et al., 1996). Most partially folded states of all- -sheet proteins show increased -helical content in TFE-induced structures compared with the native state (Shiraki et al., 1995; for a review, see Carlsson & Jonsson, 1995). In many proteins there is clear evidence for conversion of -sheet to -helix upon addition of TFE, as judged by CD measurements (Fan et al., 1993; Shiraki et al., 1995). In the case of -lactoglobulin, a predominantly -sheet protein, even complete -sheet to -helix conversion is postulated (Shiraki et al., 1995). In peptide fragments (Kemmink & Creighton, 1995) and in proteins (Shiraki et al., 1995), the helical content in the TFE state seems to correlate well with the intrinsic helix-forming propensity of the sequence. However, none of the studies on intact all- -sheet proteins used methods that are sensitive for -structure and the presence of -sheets might, in these cases, also be masked by the strong far-uv CD bands of the induced -helical structures. In tendamistat, which contains only -sheets and loop regions in the native state, TFE also seems to induce small amounts of -helical structure, as judged by the characteristic minima at 208 and 222 nm in the far-uv CD spectrum (Figure 2). Amide hydrogenexchange experiments, however, show that the induced -helical structure is not very stable. Compared with other proteins, where -sheet to -helix conversion was reported in TFE, tendamistat has a very low intrinsic helix-forming propensity and a very high -sheet propensity. This obviously leads to formation of -structure in the TFE-induced state and it shows that TFE does not exclusively allow local backbone interactions, as they occur in -helices. Non-local hydrogen bonds as they occur between -strands can form if the -sheet propensity of the protein is high. The observed induction of helical structure in tendamistat is probably restricted to the loop regions and seems to be very unspecific and ill-defined. A similar effect was observed in studies of peptide fragments of the -spectrin SH3 domain, where non-native -helical structure can be induced by addition of TFE in peptides corresponding to unordered regions of the native state. In contrast, no helical structure can be induced in peptides corresponding to -strand or -turn regions (Viguera et al., 1996). A possible origin of the observed CD spectra is a preference for certain low-energy backbone angles in the loop regions in the presence of TFE. A similar, although weaker, effect on the CD spectrum was observed by addition of perchlorate to unfolded RNase A. Here, a non-random CD spectrum was observed, but no protection of amide hydrogen bonds could be detected (Scholtz & Baldwin, 1993). The broad resonance lines in the 1D 1 H NMR spectrum in the TFE state suggest that it represents an ensemble of rapidly interconverting conformations. The presence of stable hydrogen bonds suggests a defined core region with the native -sheet structure being formed. The fluctuations are probably due to several possible side-chain interactions and/or conformational flexibility in the loop regions of the protein. A state very similar to the TFE state of tendamistat can be induced in the C11A/C27T variant of tendamistat at low ph. In contrast to the TFE state, the low-ph state is not monomeric but forms oligomeric structures of well-defined size (unpublished results). The monomeric nature of the TFE-state was directly measured by analytical ultracentrifugation experiments at different concentrations of tendamistat. In addition, the CD signal of the TFE state is not concentration-dependent up to concentrations of several mg/ml (data not shown). Implications for the mechanism of protein folding The study of partially folded conformations induced by extreme solvent conditions, such as low ph or addition of alcohols, is thought to provide a key to understanding the mechanism of protein folding. The structures of the observed partially folded states are believed to give information on early events in the folding process, at a stage when specific side-chain interactions are not yet formed. This view is supported by the finding that partially folded structures observed under equilibrium conditions often resemble structures that transiently accumulate in the process of protein folding. In the case of lysozyme from hen egg-white, a partially folded state induced by addition of TFE was shown to be nearly identical with an early folding intermediate (Radford et al., 1992; Buck et al., 1993, 1995). This state lacks specific side-chain interactions but contains some native secondary structure. The -domain of the protein forms native-like -helical structure in the TFE state, whereas the -sheet structure of the -domain is largely unordered, except for some additional helical structure in the region of one of the native -strands. The same partially folded state seems to populate during refolding of the majority of

11 442 TFE-induced State of Tendamistat lysozyme molecules. However, a faster pathway for the formation of the native state exists, where folding proceeds without any detectable intermediate state (Kiefhaber, 1995) and the role of the partially folded state for the mechanism of lysozyme folding is still unclear. It was shown for -lactoglobulin that the TFE-induced partially folded state, which contains non-native -helical structure, accumulates transiently during refolding in the absence of TFE (Kuwajima et al., 1987; Shiraki et al., 1995). This observation was interpreted in terms of non-hierarchical folding pathways for -sheet proteins, i.e. the initial formation of non-native -helical structure has to precede the formation of the native -sheets (Lim, 1978). Our results on the structure of the TFE state of tendamistat show that large parts of the native -sheet structures can form in the absence of specific side-chain interactions. In contrast to lysozyme and -lactoglobulin, where only -helices seem to be formed, the TFE state of tendamistat contains most of the native secondary structure, consisting of two three-stranded anti-parallel -sheets. Both -sheets of tendamistat are composed of two strands formed by a -hairpin and by a third strand from a distant part of the polypeptide chain (Figure 1). Thus the formation of the native -sheets requires non-local, long-range backbone interactions. The presence of native-like -sheet structure in the TFE state of tendamistat shows that long-range interactions between secondary structural elements should be able to form at a very early stage of the folding process in the absence of specific side-chain interactions. These results support a hierarchical model for -sheet formation. Stable native-like -structure can form rapidly and independently of the native tertiary structure. The observed non-native -helical parts seem to be very unstable and are not located in regions of the native -sheets. The presence of long-range interactions was observed also in a low-ph molten globule intermediate of -lactalbumin, which has no specific side-chain interactions but already contains the information to form some of the native disulfide bonds. This result can be explained only if part of the molecule contains a subset of native long-range interactions at the stage of the molten globule (Wu et al., 1995). A variant of bovine pancreatic trypsin inhibitor with only the disulfide bond intact forms a molten-globule-type structure in aqueous solution, for which it was shown that most of the central -hairpin is formed in the absence of specific side-chain interactions. As for the TFE state of tendamistat, it was shown that the observed state represents an ensemble of interconverting states with a stable -hairpin (Barbar et al., 1995; Ferrer et al., 1995) Our findings suggest that the TFE state of tendamistat might be important in the kinetic process of tendamistat folding. It contains a subset of the native interactions, including most of the native -structure. Its formation could present an early and crucial step in the folding process. The low stability of the -structure in the TFE-induced state shows that the correct secondary structure can form at a very early stage, but that the stability of the secondary structure is largely increased with the formation of the correct tertiary interactions. Kinetic experiments on tendamistat folding did not detect any populated intermediate states even under strongly native solvent conditions (unpublished results). This does not, however, rule out the importance of partially folded structures in tendamistat folding. There could be a nucleation/ condensation mechanism for folding, where the rate-limiting step is represented by the formation of a partially folded structure that contains a subset of the native interactions and thus rapidly converts to the native state. The observed partially folded state of tendamistat could represent an early intermediate on a nucleation/condensation pathway, or a sparsely populated intermediate on a hierarchical folding pathway. This view is supported by kinetic experiments, showing that refolding from the TFE-induced state occurs at the same rate as refolding from the GdmCl-unfolded protein (unpublished results). This is in contrast to results on lysozyme folding, where the observed partially folded state refolds more slowly than direct folding from the unfolded state to the native state occurs (Kiefhaber, 1995). The presence of most of the native -structure and the absence of stable non-native interactions in the TFE state indicates a very low tendency for tendamistat to form misfolded structures that are not on the direct folding pathway, even in the absence of the correct and specific tertiary interactions. In contrast, partially folded states containing stable, non-native secondary structural elements might represent kinetically trapped structures on slow folding pathways, as it was shown for the partially folded intermediate in lysozyme folding (Kiefhaber, 1995). For some proteins, these states seem to be readily accessible in the absence of the native tertiary interactions. Materials and Methods Materials Tendamistat was isolated from culture fluids of Streptomyces lividans and purified as described (Haas- Lauterbach et al., 1993). The disulfide-deficient mutants were constructed as described (Haas-Lauterbach et al., 1993). Protein concentration was determined by UV absorption measurements using an absorption coefficient of A1 0,1% cm = 1.61 (Vertesy et al., 1984). TFE-h3 was from Aldrich. TFE-d3 and 2 H 2O from Isotech. All other buffers and chemicals were from Merck (Darmstadt, Germany) and were reagent grade. CD spectroscopy CD measurements were carried out on a Jasco J-720 spectropolarimeter. A water-jacketed 1 mm cell and protein concentration of 12.5 M were used for measure-

12 TFE-induced State of Tendamistat 443 ments in the far-uv region and a water-jacketed 1 cm cell and protein concentration of 62.5 M for the near-uv region. TFE titrations were performed with samples containing a constant protein concentration and the indicated TFE concentrations in 20 mm potassium phosphate buffer at the indicated temperature and ph. The ph was measured with a Knick 761 Calimatic ph meter and the given values represent uncorrected glass electrode readings. Thermal denaturation Thermal denaturation was followed by the change in ellipticity at 220 nm on a Cary61 spectropolarimeter equipped with a thermostated quartz cell of 2 mm path-length and a heating rate of 1 deg.c/min using a water-bath (Laude RC3) equipped with a temperature programmer. CD and the temperature in the cell were recorded with an analog-digital converter (National Instruments, Austin, USA) and data collected by a computer (MacintoshII, Apple Computer, USA). The values of T m (temperature of transition midpoint) and H v.h. (van t Hoff enthalpy at T m) were obtained by van t Hoff analysis according to: d(ln K)/d(1/T) = H v.h./r (1) where K is the equilibrium constant for the native state and R is the gas constant. The free energy of folding at a given temperature (T) was then calculated using the following relationships: S 0 m = H 0 m/t m G 0 (T) = H 0 m T S 0 m + Cp[T T m T ln(t/t m)] (2) Sm 0 and Hm 0 represent the values of S and H at T m (Becktel & Schellman, 1987). The C p values for the wild-type protein and for the two disulfide mutants were taken from Vogl et al. (1995). Prediction of helix propensity Prediction of helix propensity was carried out using the joint secondary structure prediction method (Nishikawa & Noguchi, 1991) and the AGADIR1s program (Munoz & Serrano, 1994, 1995a,b). NMR spectroscopy All 1 H NMR spectra were recorded on a Bruker AMK NMR instrument operating at a proton frequency of MHz or on a Bruker AMX II 600 MHz spectrometer operating at MHz. 1D 1 H NMR spectra of the native state (in 2 H 2O, ph 3.0, 25 C) and of the unfolded states (in 5.2 M Gdn 2 HCl, 20 mm potassium phosphate/ 2 H 2O, ph 2, 45 C) were recorded 64 times with 16 K complex data points and a sweepwidth of 6024 Hz. The 1D 1 H NMR spectrum of the TFE state (in 70% TFE-d 3/30% 2 H 2O, ph 3.0, 25 C) was recorded with 4 K complex data points and a sweep width of 6666 Hz. Protein concentrations were typically 1 mm for 1D measurements and between 2 and 3 mm for 2D experiments. Amide hydrogen exchange Since the TFE state of tendamistat is formed slowly from the native state ( one minute), all hydrogenexchange studies were performed using tendamistat lyophilized directly from the TFE state (termed TFElyophilisate). Re-dissolving the TFE lyophilisate in 70% TFE, ph 3 results in formation of the TFE state within the dead-time of manual mixing of the CD-monitored kinetics (ca 20 seconds). For bulk exchange experiments the TFE lyophilisate was dissolved in 70% TFE-d3/30% 2 H 2O and the ph rapidly adjusted to 3.0 or D NMR spectra were then consecutively acquired at 25 C over a period of 16 hours to monitor hydrogen exchange. The amide intensities from 7.55 to 8.7 ppm were integrated and normalized against the non-overlapping intensities of the 31 aromatic hydrogen atoms from 6.5 to 7.55 ppm. To identify amide protons that are protected in the TFE state but do not show protection in the native state, exchange was allowed for 70 minutes as for the bulk exchange. The protein was then returned to the native state by dilution with 2 H 2O to final conditions of 20% TFE, ph 4.0, 25 C. Under these conditions the native state forms rapidly and hydrogen exchange is slow. The 1D NMR spectra were then acquired for a total of 15 hours and analyzed in terms of the relative amide intensities as above. For determining protection factors for individual amide protons in the TFE state, hydrogen exchange was performed in 70% TFE-d 3/30% 2 H 2O for 70 minutes before returning to native conditions as above. Consequently, residual concentrations of TFE were removed and the sample was brought to concentrations between 2 and 3 mm with a Centricon-3 concentrator (Amicon, Inc., Beverly, MA, USA). Absolute mode (magnitude) 1 H- 1 H COSY (two-dimensional-spin correlated) NMR spectra were collected at 20 C with time-domain data points, a sweep width of 6024 Hz and 32 scans. Amide peaks in the fingerprint region were assigned on the basis of the published COSY spectrum under the same conditions (Qiwen et al., 1987). The heights of the amide crosspeaks in the fingerprint region were scaled according to the heights of the C3, 5 H-C2, 6 H cross-peaks of the non-exchanging tyrosine residues. The relative peak heights were further normalized against the relative peak heights from the spectrum of a native reference sample prepared in the same manner except without prior exchange in TFE-d 3 and acquired using the same parameters. Comparing the proton occupancies after the exchange pulse in the TFE state with the maximum proton occupancies from the spectra of the native state allows the calculation of exchange rates for the individual amide protons. The data were analyzed according to the mechanism: A k 12 k 21 B k 23 C where A represents the folded, exchange-resistant form and B is the open, exchange-competent form. C represents molecules where exchange has taken place. Hydrogen exchange is under EX2 regime when the rate of structural closing (k 21) is larger than the rate of exchange (k 23), which applies to the conditions used in our experiments. Under these conditions the rate of exchange (k HX) is given by (Hvidt, 1964; Hvidt & Nielsen, 1966): k 12 K12 k HX = k k 12 + k 23 = k K 23 (3) 12 where K 12 is the equilibrium constant for structural opening (=k 12/k 21). Protection factors (k 23/k HX) were obtained by comparing the rate of exchange in the TFE state with the values for unprotected amide protons (k 23),