Phenylalanine Side Chain Behavior of the Intestinal Fatty Acid-binding Protein

The equilibrium unfolding behavior of the intestinal fatty acid-binding protein has been investigated by 19F-NMR after incorporation of 4-fluorophenylalanine and by pulsed field gradient diffusion 1H-NMR. At low urea concentrations (0-3 m) but prior to the global unfolding that begins at 4 m urea, the protein exhibits dynamic motion in the backbone and an expanded hydrodynamic radius with no major change in the side chain orientation. As monitored by two-dimensional 19F-19F nuclear Overhauser effect, the distance between two phenylalanine residues (Phe68 and Phe93) located in the two different β-sheets that enclose the internal cavity did not change up to 4 m urea. Additionally, the chemical shifts of these two residues changed almost identically as a function of denaturant. At all urea concentrations, as well as in the native protein, multiple conformations exist. These conformers interconvert at different rates under different conditions, ranging from slow exchange by showing separate peaks in the native state to intermediate exchange at intermediate urea concentrations. Residual structure persisted around Phe62 even at very high concentrations of denaturant, suggesting that region as a nucleation site during folding. The results were compared with previous studies examining the backbone behavior (Hodsdon, M. E., and Frieden, C. (2001) Biochemistry 40, 732-742) and suggest that the side chains show more stability than the backbone prior to global unfolding of the protein.

The equilibrium unfolding behavior of the intestinal fatty acidbinding protein has been investigated by 19 F-NMR after incorporation of 4-fluorophenylalanine and by pulsed field gradient diffusion 1 H-NMR. At low urea concentrations (0 -3 M) but prior to the global unfolding that begins at 4 M urea, the protein exhibits dynamic motion in the backbone and an expanded hydrodynamic radius with no major change in the side chain orientation. As monitored by two-dimensional 19 F-19 F nuclear Overhauser effect, the distance between two phenylalanine residues (Phe 68 and Phe 93 ) located in the two different ␤-sheets that enclose the internal cavity did not change up to 4 M urea. Additionally, the chemical shifts of these two residues changed almost identically as Incorporating fluorine-labeled amino acids into proteins has allowed investigation of the effects of ligand addition, conformational changes, and protein folding on the behavior of side chains (1,2). Such data should be complementary to data examining the response of the backbone to these effects by methods such as, for example, circular dichroism or hydrogen/deuterium (H/D) 2 exchange. Previous studies from our laboratory using 19 F-labeled amino acids have focused on the stabilization of side chains during the folding process and have found that, in general, such stabilization appears to be associated with the last step in the formation of the native structure (3,4). Examination of side chain behavior of fluorine-labeled residues, however, can also be used to monitor the behavior and relative distances between the side chains of labeled residues under various conditions.
The intestinal fatty acid-binding protein (IFABP) is one of a class of small (15 kDa) proteins that bind ligands into a large cavity surrounded by 10 antiparallel ␤-strands. IFABP contains eight phenylalanine residues, and all of them are included in the 29 residues that line this binding cavity (5) (Fig. 1). We have previously assigned the 19 F-NMR resonances for each residue (6). Monitoring the behavior of each phenylalanine under a variety of conditions should allow the ability to compare and contrast the role of the different phenylalanine residues.
In the present work we have examined the role of phenylalanines in protein stability by measuring their 19 F-NMR properties as a function of urea concentration. Combined with the global property from PFG diffusion experiments, the results are compared with 1 H-15 N heteronuclear single quantum coherence (HSQC) and H/D exchange data obtained earlier as a function of urea concentration (7). Here we show that side chain behavior may differ from backbone behavior with respect to unfolding by urea under equilibrium conditions.

MATERIALS AND METHODS
Chemicals-Ultrapure urea was purchased from United States Biochemical. The concentration of urea was determined by index of refraction at 25°C (8). 4-19 F-Phe was obtained from Acros Organics (Morris Plains, NJ). All other chemicals were of reagent grade.
Protein Production and Purification-The unlabeled and 4-19 F-labeled IFABP was produced and purified as described elsewhere (6). Protein in the NMR experiments was used within ϳ2 weeks of its preparation.
Sample Preparation for 19 19 F probe was cooled and kept at 20 K with the Varian Cryo-Q open cycle cryogenic system. All one-dimensional spectra were recorded at 20°C with 64 scans and processed by NMRPipe (9) with 12-Hz exponential line broadening unless otherwise indicated. For quantitative measurements (relative to the intensity of 6-19 F-tryptophan) a delay of 5 s was used. Under these conditions, no saturation of 19 F signals was observed. For the two-dimensional NOESY experiments at different urea concentrations, the mixing time was selected by acquiring a series of NOESY spectra with different mixing times from 50 to 500 ms. An optimal value of 250 ms was used to probe the distance change between Phe 68 and Phe 93 as a function of urea concentration. The two-dimensional NMR data were analyzed by the PIPP program (10). The spectra were referenced to an internal standard of 6-19 F-tryptophan (ϳ0.2 mM; the actual chemical shift is Ϫ46.293 ppm relative to trifluoroacetic acid), and all samples contained 8% (v/v) D 2 O. There was no correction to the pH value due to D 2 O.
PFG Diffusion 1 H-NMR Spectroscopy-PFG-NMR diffusion measurements were made with a bipolar PFG longitudinal eddy current delay pulse sequence (11) on a Varian Inova 500 spectrometer at 20°C. Twenty-two spectra were acquired with the strength of the diffusion gradient varying between 2.5 and 100% of its maximum value. The length of the diffusion gradient was optimized for each sample to give a total decay of ϳ90% of the initial intensity. For protein samples at 0, 1.5, 2.5, and 4 M urea the length of the diffusion gradient was set to 3 ms, whereas in 5.5 and 7 M urea it was set to 5 ms. The echo time for diffusion was set to 150 ms, and all spectra were acquired with 10,000 complex points and a sweep width of 7000 Hz. Because many contaminants in protein solution give rise to resonances in the aliphatic region of the spectrum, only the integration in the aromatic region (6.5-8.2 ppm) was used as a function of gradient strength, following standard methods (12). The reported hydrodynamic radius (R h ) of dioxane, 2.12 Å (13), was used to calculate the R h of IFABP under different urea concentrations.
Fluorescence Spectra-Steady state fluorescence experiments as a function of urea were carried out at 20°C using a PTI Alphascan fluorometer (Photon Technology International, South Brunswick, NJ). The excitation wavelength used was 290 nm, and the emission spectra were recorded between 300 and 400 nm. Equilibrium data were fit to a twostate model as described by Ropson et al. (14) with an equation from Santoro and Bolen (15). Fig. 2 shows the change in intrinsic fluorescence of fully labeled IFABP with urea concentration. The data were obtained under similar conditions as used in NMR experiments except at much lower protein concentrations. Fully labeled 19 F-Phe-IFABP has a denaturation midpoint (4.9 M) that is slightly higher than that obtained for wild-type protein (ϳ4.8 M) (16). We have shown that circular dichroism data are superimposable on the fluorescence data (not shown), indicating that the fluorescence data monitor global unfolding. For the discussion below, it is important to note that the global unfolding of the protein only begins at Ͼ4 M urea.

Denaturation of Labeled IFABP-
One-dimensional 19 F-NMR Spectra of IFABP as a Function of Urea Concentration-The equilibrium unfolding of fully labeled IFABP was studied by 19 F-NMR as a function of urea concentration. Fig. 3A shows the spectra acquired at different urea concentrations for fully labeled IFABP. The assignments for each of the phenylalanine residues has been made previously (6). At urea concentrations Ͼ3 M the decrease in native peaks is accompanied by an increase in unfolded peaks around Ϫ40.5 ppm, characteristic of the denatured resonances and indicating that the exchange between native form and denatured forms is slow on the NMR time scale. Fig. 3B shows the chemical shift changes as a function of urea concentration. It is obvious that major chemical shift changes occur at Ͻ3 M urea, well before the protein begins to undergo global unfolding. All phenylalanines show a similar denaturation midpoint of ϳ4 M urea (data not shown). The midpoint is well above the urea concentrations influencing the chemical shifts and well below that for the global unfolding.
Other changes occur at higher urea concentrations. Careful examination of Fig. 3A shows, for example, that a new small peak for Phe 62 , in addition to the major peak at Ϫ47.15 ppm, appeared at Ϫ46.72 ppm in 0.5 M urea, indicating the presence of a new conformation. The intensity of these two peaks (at Ϫ46.72 and Ϫ47.15 ppm) initially decreases with urea but increases at urea concentrations of Ͼ5 M, concentrations at  which all other native peaks are gone (supplemental Fig. 1, available in the on-line version of this article). Although not obvious from Fig. 3A (because the intensities of the spectra from 5 M to 8.5 M urea have been reduced in the figure for plotting purposes), there is a loss of the total intensity between 3.5 and 6 M urea, which is discussed below.
In Fig. 3A, Phe 68 and Phe 93 behave differently by exhibiting broader line widths than those for the other residues. Our previous study has shown that Phe 68 has two conformations at native state, as it could be deconvoluted into two peaks. The same is true for Phe 93 at temperatures of Ͼ34°C (6). To further characterize the urea denaturation behavior, we used proteins singly labeled either at Phe 68 or Phe 93 (Fig. 4), as well as the protein labeled at both Phe 68 and Phe 93 (supplemental Fig. 2, available in the on-line version of this article). These proteins show a slightly lower denaturation midpoint than does the fully labeled protein. For protein singly labeled at Phe 68 , the separation of the two resonances, corresponding to the two conformations, is slightly more extensive than that for the fully labeled protein. Also, the major upfield conformation is less stable because it decreases more rapidly with increasing urea (Fig.  4A). Multiple resonances were also observed for Phe 93 and Phe 68 at higher urea concentrations around the denatured region, Ϫ40.5 ppm (Fig. 4, B and C, respectively), with one peak continuing to increase whereas the others decrease until merging with noise. Those decreasing peaks represent some unfolded like intermediates in slow exchange with the unfolded conformers.
Homonuclear 19 (18), where I is the intensity of the cross peak and r is the distance between two nuclei that gives rise to the nuclear Overhauser effect cross-peak. Using two-dimensional NOESY measurements, it is possible to monitor the distance between 4-19 F-Phe 68 and 4-19 F-Phe 93 at 0, 1.5, 2.5, and 4 M urea (Fig. 5). A cross-  peak between Phe 68 and Phe 93 appears at all these urea concentrations. The peak intensities at 1.5, 2.5, and 4 M urea were normalized relative to that at 0 M urea by dividing the intensities by the percentage of native structure. The latter was normalized relative to that at 0 M urea by integrating five different regions of the spectra, namely, the combined integration of Phe 2 , Phe 17 , and Phe 93 , the integration of Phe 47 , Phe 62 , and Phe 68 separately, and the integration between Ϫ40.3 and Ϫ41.3 ppm (Fig. 3A), and all gave similar results. The average was used as the percentage of native structure. From 0 to 4 M urea, the distance between 4-19 F-Phe 68 and 4-19 F-Phe 93 barely changes (supplemental TABLE ONE, available in the on-line version of this article). Thus, any native (or native-like) structures that remain show little change in the distance between the two residues.
Under native conditions, minor conformations, with ϳ0.1-0.5% of the population for Phe 47 and Phe 62 and ϳ8% of the population for Phe 2 were observed by the appearance of exchange cross-peaks in phasesensitive 19 F-NOESY (6) (Fig. 5). With increasing urea the exchange cross-peaks for these two conformations get weaker for two reasons. First, the overall concentration of native-like structure is decreased with the increasing urea and, second, the population of the minor conformation is very low and the resonance of minor conformation broadens relative to the major form (19).

PFG Diffusion 1 H-NMR as a Function of Urea-
The apparent hydrodynamic radius of IFABP increases substantially at high concentrations of urea (TABLE ONE), indicating protein unfolding. The hydrodynamic radii at 1.5 and 2.5 M urea increased ϳ 10%, which is an ϳ33% increase in hydrodynamic volume even though major chemical shift changes occur for most resonances (Fig. 3B) without global unfolding (Fig. 2).
As intermediates undetectable by 19 F-NMR are expected to be beyond detection by diffusion NMR, the intensity loss observed by PFG-NMR comes from two factors, namely increase of gradient and exchange broadening. Therefore, we expect the midpoint determined by PFG-NMR to be similar to that determined by 19

DISCUSSION
Structural Changes Prior to Global Unfolding-It is instructive to compare the data dealing with side chain behavior obtained in this study with data from previous NMR studies of IFABP dealing with backbone behavior as a function of urea concentration. As an example, data with respect to Phe 68 are shown in Fig. 6. Data for the other phenylalanine residues (not shown) are essentially similar to those shown in the Fig. 6.
Fundamentally, what this comparison shows is that the backbone, at least for phenylalanine residues, behaves differently from the side chains.
Hodsdon and Frieden (7) monitored amide H/D exchange rate and collected a series of two-dimensional 1 H-15 N-HSQC spectra to examine the properties of the backbone as a function of urea concentration (7). That study made several points. First, all amide H/D exchange rates increased significantly at low urea. Several of the backbone amide protons have exchange times on the order of days, but it was found that by slightly over ϳ2 M urea all amides exchanged too rapidly to be measured within experimental dead time (about 10 m). This result is shown by the leftmost curve (-x-) in Fig. 6. The data suggest that, at low urea concentrations, amide protons become more susceptible to exchange as a consequence of the protein undergoing a conformational change from a closed form (e.g. H-bonded) to an open form (20). Second, except for regions of residual structure, all of the native backbone resonances show midpoints between 2 and 3 M, disappearing by ϳ4 M urea. These data indicate that amide resonances are broadened by conformational fluctuations on a millisecond to microsecond time scale. The HSQC data for Phe 68 are shown in Fig. 6 and can be interpreted to reflect a transition from a well formed secondary structure to one or more intermediate forms with a more dynamic backbone. It is important to note that these changes occur without any appreciable fluorescence (or circular dichroism) change.
At higher urea concentrations, but prior to global unfolding, the resonances for the 19 F-labeled phenylalanines start to disappear. These data suggest that the stability of the hydrophobic phenylalanine side chains persists even after the backbone becomes more dynamic. Finally, as measured by fluorescence, the protein undergoes global unfolding.
It is intriguing to examine the behavior of the protein between the urea concentration where the H/D exchange is rapid and the concentration where global unfolding starts. This is the region where major chemical shift changes of phenylalanines occur (2.5-3.0 M urea), H/D exchange is fast, most backbone amide resonances are broadened by at least half, and the hydrodynamic radius is increased by ϳ10%. These phenomena suggest that the native structure (or native-like) has become more expanded and dynamic. These expanded conformations could be very similar to the mechanically softened conformations under low denaturant concentrations observed by atomic force microscopy (21). The chemical shift changes also indicate some structural perturbation of the expanded conformations. Although no good theory exists to correlate fluorine chemical shift with environment, a systematic chemical shift change with the change of conditions may give some clues provided that the structure is known. As a function of urea the 19 F chemical shifts of Phe 68 and Phe 93 , whose distance remains at ϳ 2 Å up to 4 M urea, change downfield in an almost identical way. This observation may indicate that they experience the same or very similar environmental changes. Because these two strands represent two different ␤-sheets, the overall structure remains up to the point of global unfolding. The behavior of Phe 47 and Phe 62 is somewhat different from that of Phe 68 and Phe 93 , showing much smaller chemical shift changes with urea and suggesting that these regions may experience fewer environmental changes at low urea concentrations. As with Phe 68 and Phe 93 , Phe 62 and Phe 47 are spatially close.
All of the phenylalanine residues, except Phe 55 , are located in structural elements other than turns. In particular, Phe 47 , Phe 62 , Phe 68 and Phe 93 are stacked around one of the critical turns between the D and E strands. The observed changes relate to the mechanism of protein unfolding showing that there are no major changes in side chain orientation prior to global unfolding but rather movement of structural regions of the protein.
Evidence of Intermediates during Urea Unfolding-The denaturation midpoint observed by 19 F-NMR is lower than that observed by fluorescence. We should emphasize, however, that the discrepancy does not necessarily mean that phenylalanines globally destabilize before tryptophans do. The difference could be a consequence of the fact that any folding intermediate that is exchange-broadened would be undetectable on the NMR time scale and therefore would be recorded as an unfolded state, yielding an apparent lower midpoint compared with global unfolding. In the case of IFABP, there is no evidence of residual structure in the unfolded form around tryptophan measured by fluorescence (22), but there is strong evidence that intermediates exist during urea unfolding by 19 F-NMR. In Fig. 3A there is significant intensity loss (ϳ20 to ϳ30%) from ϳ3.5 to 6 M urea (compared with the reference peak of 6-19 F-Trp at Ϫ46.293 ppm). However, no substantial change in the line width was observed for the native peaks or denatured peaks during the urea titration course, indicating the intensity loss was not caused by exchange between unfolded states and folded states. The most reasonable explanation for the intensity loss is the presence of intermediates exchanging on a time scale undetectable by NMR, with those intermediates being more like the unfolded state. In Fig. 3A (and supplemental Fig. 2), the sharpening and the increasing intensity of the resonances of the unfolded states also points to the existence of NMR-undetectable intermediates.
Residual Structures at High Denaturant Concentrations-The central issue in the mechanism of protein folding is the nature of the nuclei around which the protein folds. The common perception is that such nuclei are more stable and may show persistent structure under unfolding conditions. In previous work, we have identified several regions for nuclei formation based on single site mutations (23,24) as well as NMR studies (7,22). In general the data have been consistent in proposing turns between ␤-strands as nucleation spots.
As shown in Fig. 3A (see also supplemental Fig. 1), two native-like peaks (at Ϫ46.72 ppm and Ϫ47.15 ppm) persist above urea concentration of 5 M urea, concentrations at which all other native peaks are essentially gone. These data suggest that a core region remains collapsed in the most hydrophobic region even when the majority of the protein is completely unfolded. The data also indicate that Phe 62 and other hydrophobic residues are involved in hydrophobic collapse during the very early stage of folding. This result is consistent with other studies indicating residual structure near this region (7,22).
The chemical shift of Phe 62 is the most upfield, possibly due to its being surrounded by other aromatic side chains. At higher urea concentrations (e.g. 8.5 M) residual Phe 62 still resonates upfield the most (Ϫ46.72 ppm), only slightly differently than in its native state (Ϫ47.15 ppm). This large shielding effect may also indicate that the side chain topology around Phe 62 is largely maintained at higher urea concentration.
Conclusions-Altogether, the current data in combination with previous results suggest the following structural changes as a function of urea concentrations. At low urea, chemical shift changes and H/D exchange experiments indicate that the structure is loosened to an extent to allow solvent penetration. As the urea concentration increases, the backbone structure may convert from rigid, well formed structures to intermediate states that are poorly formed or more dynamic. Little or no fluorescence change occurs when the backbone undergoes these transitions. The phenylalanine side chains remain organized until just prior to global unfolding, suggesting that clusters of hydrophobic residues remain stable until global unfolding occurs. Substantial intermediates are present between 3.5 M and 6 M urea. Even after the protein unfolds, some transient structural regions exist at high denaturant concentrations.
Acknowledgments-We thank Dr. Andre D'Avignon for discussions about pulse field gradient experiments and Robert Horton for excellent technical assistance.  The hydrogen/deuterium exchange data reflect the half-times of exchange. By slightly over 2 M urea the rate was too fast to measure by the techniques used (7). These data and the HSQC data are taken from Hodsdon and Frieden (7).