Intersubunit Hydrophobic Interactions in Pf1 Filamentous Phage*

Magic angle spinning solid-state NMR has been used to study the structural changes in the Pf1 filamentous bacteriophage, which occur near 10 °C. Comparisons of NMR spectra recorded above and below 10 °C reveal reversible perturbations in many NMR chemical shifts, most of which are assigned to atoms of hydrophobic side chains of the 46-residue subunit. The changes mainly involve groups located in patches on the interfaces between neighboring capsid subunits. The observations show that the transition adjusts the hydrophobic interfaces between fairly rigid subunits. The low temperature form has been generally more amenable to structure determination; spin diffusion experiments on this form revealed unambiguous contacts between side chains of neighboring subunits. These contacts are important constraints for structure modeling.

The filamentous bacteriophage Pf1 consists of a closed circular, single-stranded DNA of 7349 nucleotides, encased by ϳ7300 copies of a major capsid protein in a shell capped at the ends by a few copies of key proteins involved in virus assembly and host specificity (for reviews see Refs. [1][2][3]. It is 1 of over 60 different filamentous phages that have been reported, about a dozen of which have been structurally characterized to various extents. Many of these phages integrate their genomes into the genomes of their bacterial hosts and some contribute to the virulence of pathogenic bacteria. The host of Pf1 is Pseudomonas aeruginosa, strain K (4). Pf1 has virtually the same structural genes as the much larger phage Pf4 (12,437 nucleotides), which occurs as a latent prophage integrated in the chromosome of P. aeruginosa PA01 that has been implicated as a virulence factor of this human pathogen (5,6). Among the characterized filamentous phages, Pf1 is unique in having a unit chemical ratio of one major capsid subunit for each nucleotide in the packaged genome (1). Related to this stoichiometry, Pf1 does not show any evidence of long length scale gentle coiling characteristic of several other phages (7). Through many structural studies by a variety of techniques, Pf1 capsid structure has become a paradigm for the capsids of a number of filamentous phages (2,3), even though key aspects of that structure are still puzzling.
An approach to the full determination and understanding of the Pf1 structure is through the use of a well known reversible structural phase transition near 10°C between two ordered states that were first observed by x-ray fiber diffraction (8,9). This transition temperature is well below the physiological temperature, and the structural response of Pf1 is associated with a change in the helical pitch and rise. A modification in salt concentration is also associated with a small change in subunit rise in both forms of the phage (10). Interestingly, a similar response to temperature has not been observed in any of Ff, Pf3, Xf, or PH75. One difference between Pf1 and the others that might account for the presence of a transition in Pf1 and not the others is its 1:1 stoichiometry. Diffraction studies of oriented phage in solution showed that the low and high temperature forms differ slightly in the orientation of the subunit with respect to the filament axis (10,11). The two states have slightly different numbers of subunits per turn in the helical arrangement of their capsids, which produce very different diffraction patterns even though the structures are closely similar. The low temperature form (Pf1 L ) 2 has an apparent structure repeat of about 21.9 nm containing 71 subunits in 13 turns, or 5.46 subunits/turn (2). To account for the unit stoichiometry, one can consider repeats of 71 dimers in 26 turns, where subunits in the dimer differ in structure slightly because of up-strand versus down-strand interactions (12). For Pf1 H , the apparent repeat is about 7.8 nm, and the symmetry assigned is nominally 27 subunits per 5 turns (or 5.40 subunits/turn), but this symmetry does not account fully for features of some diffraction patterns and NMR data (13)(14)(15). The transition between the two symmetries is estimated to occur between 280 and 286 K depending on pH and salt concentration. The transition is clearly a property of a single virion particle, as shown by studies done at varying concentrations of 100 mg/ml by Raman optical activity (16), 50 mg/ml by NMR (17), 20 mg/ml by diffraction from oriented gels (10), and 0.5 mg/ml and lower by calorimetry and CD (18). However, the transition is apparently blocked at the very high concentrations of 400 mg/ml or more in fibers; dry fibers made at the high temperature exhibit the same diffraction pattern when the temperature is lowered to 4°C and vice versa (9,19). The electronic and vibrational spectroscopies revealed little if any change in protein conformation, although changes involving DNA have been observed by CD and Raman optical activity (16,20).
Static solid-state NMR experiments on oriented samples at ϳ50 mg/ml by Thiriot et al. (17) showed large changes in 15 N chemical shift values in 15 N-Ile enriched virus, revealing structural details for the two-state transition. A decrease of the tilt angle of the C terminus starting at residue alanine 29 was clearly observed in the dipolar wave analysis (see Fig. 5b in Ref. 17). This change results in a smaller overall diameter of the virion at low temperature, in agreement with measurements of the birefringence of aligned concentrated solutions of Pf1 in varying magnetic fields (21). Others had suggested that the N terminus is tilted 4°toward the filamentous axis in the low temperature form (10).
This study of Pf1 and its phase transition is by means of magic angle spinning solid-state NMR on fully hydrated, infectious samples. Based on our prior assignments of the coat protein subunit (15), we have used 13 C homonuclear correlation spectroscopy to map transition-induced perturbations in the assigned NMR chemical shifts. We report on a significant number of such perturbations. We also report the first intermolecular constraints on the virus; spin diffusion experiments on Pf1 have revealed internuclear contacts between neighboring subunits.

EXPERIMENTAL PROCEDURES
Sample Preparation-Pf1 phage was isolated and purified from infected cultures of P. aeruginosa, strain K, grown in minimal M9 media with 15 NH 4 Cl and [U-13 C]glucose as the sole nitrogen and carbon sources as described earlier (15). Following density gradient ultracentrifugation, the sample in high CsCl was precipitated with 4% w/v polyethylene glycol (PEG) 8000 (Fluka Biochemicals), pelleted, and redissolved to a concentration of ϳ1 mg/ml in 10 mM Tris, pH 8.4. The cryoprotectant ethylene glycol was then added (22) to a final concentration of 30% v/v. A second precipitation was done by mixing MgCl to 5 mM and PEG8000 to 10% w/v. After several minutes of gentle mixing, the suspension was centrifuged for 10 min at 3700 ϫ g. The loose pellet was transferred to a pipette tip sealed at the bottom and further pelleted at ϳ20,000 ϫ g for 2 h in an Eppendorf microcentrifuge to reduce excess supernatant. The concentration of Pf1 in the pellet was estimated to be ϳ200 mg/ml based on the amount of Pf1 before precipitation (6 mg) and the approximate total volume of the resulting pellet (30 l). The bottom seal of the pipette tip was then sliced off, and the pellet was sedimented from the tip into a 4-mm ZrO 2 magic angle spinning NMR rotor. For Pf1 H , all steps were carried out at room temperature. For Pf1 L , the second precipitation was done in a cold room at ϳ4°C, and the pelleting into the pipette tips and the NMR rotor was done at 0°C. The small NMR rotor was kept cold at all times and transferred to a pre-cooled NMR probe for the magic angle spinning experiments. Concentrations were based on UV spectra with an extinction coefficient ⑀ ϭ 2.1 mg Ϫ1 cm 2 at 271 nm (7).
NMR Experiments-Two-dimensional homonuclear experiments were based on dipolar-assisted rotational resonance (DARR) (23) also known as RF-assisted spin diffusion (RAD mixing) (24), with heteronuclear scalar J CN decoupling; they were performed on a Varian Infinity Plus spectrometer operating at a 600-MHz proton frequency. Cross-polarization (25) was achieved by tangential ramp (26) on the 13 C channel with the magic angle spinning Hartman-Hahn condition set to n ϭ 1 ( 1H -1C ϭ r , 1X being the RF power on channel X, and r is the sample spinning rate) and a contact time of 1.25 ms. Twopulse phase modulation (27) was used for proton decoupling with 1H ϳ75 kHz, phase 15°, and pulse segments of 6.5 s. Relaxation delay was 3 s, and acquisition lengths were 20 ms (t 2 ) and 11.5 ms (t 1 ). 16 -32 transients were collected for each FID using 32 dummy pulses to equilibrate the RF-produced sample heating. Experiments were performed with mixing times of 10, 50, and 200 ms. Temperature control was achieved using an FTS cooling unit operating at Ϫ20°C for the low temperature samples and 15°C for the high temperature samples. The low temperature of Ϫ20°C ensures that friction and RF-induced heating do not elevate the temperature above the phage transition temperature.
Data Processing and Analyses-The data were processed using NMRPipe (28) and involved various apodization schemes, including Gauss-to-Lorentz transformation, cosine bell apodization, and exponential multiplication. All were used to verify the position and integrity of the cross-peaks. The resulting spectra were analyzed using the SPARKY program (29). The assignments of Pf1 L were based on BMRB (30) accession number 15138 for the high temperature form of Pf1 and aided by cross-peaks from the long mixing time experiments. We have used the following strategy to assign the low temperature form. (i) For peaks with a similar location at Pf1 L and Pf1 H , we assumed an identical assignment. (ii) Side-chain resonances, which make up most of the perturbations, were tracked from the backbone peaks using a "side-chain walk" strategy. (iii) The existence of a peak shift between Pf1 H and Pf1 L , when other peaks of the same amino acid type overlap, allowed unambiguous determination of its identity. (iv) In cases of ambiguities, cross-peaks from long mixing time experiments were used to verify the identities of the peaks. In some cases, such peaks did not exist for all amino acids of the same type; however, by observing cross-peaks of other specific amino acids, we assigned others by elimination. For example, out of four leucine residues in the sequence, Leu-33 and Leu-38 have a similar C␣, but the identification of Leu-33 C␤-Ala-34 C␣ cross-peak in the DARR50 experiment (DARR with 50-ms mixing time) allowed us to assign unambiguously Leu-38 C␣ and Leu-38 C␤ to 58.4 and 41.8 ppm, respectively. In our final shift perturbation table, the entire side chain of Leu-38 shows significant shift deviations between the two forms (0.4 -1 ppm), whereas the shifts of Leu-33 are similar in both forms.

RESULTS
We performed a set of two-dimensional 13 C-13 C chemical shift correlation experiments (23, 24) on both Pf1 H and Pf1 L . An overlay of the aliphatic carbon regions from spectra of the two forms is shown in Fig. 1, a and b. The observed resonances in the DARR/RAD spectra are those of the coat protein, which contributes ϳ94% of the total mass of the virion. Cross-peaks resulting from a 10-ms magnetization transfer period mainly correspond to correlations between carbons that are one or two bonds apart. Initial inspection of the C␣-C␤ region ( Fig. 1b) suggests that the two spectra are similar, and therefore it is highly likely, as suggested by other studies, that significant backbone structural changes do not occur. However, several chemical shift perturbations induced by the temperature difference were observed, and they can be attributed to individual resonances. The most apparent ones are indicated by labeled arrows in Fig. 1, a and b, based on the prior assignment of Pf1 H (15). The ␣-helical nature of Pf1 causes high congestion in some regions of the NMR spectra, and therefore verifications of many of the chemical shifts for Pf1 L were sought through analyses of spectra obtained with longer mixing times (50 and 200 ms, discussed below). These spectra include cross-peaks of carbons that are further apart, up to 5-6 Å, providing more spectral dispersion for detection and assignment of the perturbations in the chemical shifts.
In Fig. 1, c and d, we compare a portion of the 50-ms DARR spectra of both forms of Pf1 (Fig. 1d) and the 200-ms spectrum of Pf1 L (Fig. 1c). The region shown in the spectra clearly indi-FIGURE 1. a and b, overlay of spectra for the low and high temperature forms (blue, Pf1 L ; red, Pf1 H ). Both samples were separate aliquots of the same Pf1 preparation. Pf1 L was precipitated at a temperature below the transition point and kept at that temperature, and Pf1 H was prepared at high temperature throughout (see "Experimental Procedures"). The DARR spectra were measured on a 600-MHz Varian InfinityPlus spectrometer with a mixing time of 10 ms and identical experimental parameters other than the temperature. The signal-to-noise of the lowest contour is 15, and 12 contours are drawn at multiples of 1.3. a, side-chain cross-peaks region; b, mostly C␣-C␤ region; c and d, medium range correlations observed by long mixing time DARR experiments. d, typical region from the 50-ms DARR spectra of Pf1 L (blue) and Pf1 H (red) is shown. The black arrows indicate positions of intrasubunit contacts in Pf1 L (red arrows point to similar cross-peaks in Pf1 H ). Especially important is the shift in Ala-29 C␤ between the two forms (⌬␦ ϭ 1.1 ppm), which is verified by connectivities to Leu-30 C␣ and Gly-28 C␣. The other peaks in this region indicate residues that are largely unchanged. c, a similar region is extracted from a 200-ms spectrum measured on Pf1 L . Despite the decrease in intensity of many peaks, other peaks are well resolved and reinforce our assignments of Pf1 L . The dashed and dotted rectangles link residues associated with the C␤ atoms of Ala-29 and Ala-11, respectively. Also indicated are the intersubunit contacts of the C␥2 atom of Ile-32 with the C␣ atom Met-19 belonging to a different subunit. In the inset, the correlation to Met-19 C␤ is shown, which resolves the ambiguity of Met-19 C␣ with Lys-20 C␣.  NOVEMBER 19, 2010 • VOLUME 285 • NUMBER 47 cates significant perturbation for residue Ala-29 by virtue of its correlations to its adjacent amino acids Gly-28 and Leu-30. Other typical peaks show mostly unchanged residues as indicated by the arrows in Fig. 1d. In general, the low temperature spectra of Pf1 L show more intense peaks and more signals than Pf1 H spectra, probably because of reduced dynamic effects. The same set of peaks is also detected in the 200-ms spectrum (Fig.  1c). In this spectrum, many of the peaks already lose intensity because of relaxation effects during the long longitudinal mixing period. The advantage is that some peaks, for example the alanine residues in the solvent-exposed region, become more resolved. The correlations of Ala-29 and Ala-11 are indicated in Fig. 1, c and d, by the dashed and dotted rectangles, respectively.

Intersubunit Hydrophobic Interactions in Pf1 Filamentous Phage
Significantly, the 200-ms DARR spectrum for Pf1 L reveals three intersubunit contacts. One is between the well resolved C␥2 atom of Ile-32 in one subunit (29.3 ppm) and the C␣ (60.6) and C␤ (35.5) atoms of Met-19 in a neighboring subunit. This correlation is undoubtedly an intersubunit contact because the internuclear distances between atoms of residues 19 and 32 within a given ␣-helical subunit are too long to produce the observed data. By the same logic, the well resolved Ala-29 C␤ (20.7 ppm) with either of the overlapping Leu-38 C␦ (25.1) and Val-35 C␥ (25.2) is an intersubunit cross-peak, as is the well resolved Ile-12 C␤ (36.6 ppm) with either of the same two overlapping atoms, Leu-38 C␦ and Val-35 C␥.
Based on our previous analysis of Pf1 H spectra, we were able to assign many of the carbon atoms in Pf1 L and hence establish the sites of the perturbations. A summary of the assigned chemical shifts for Pf1 L appear in supplemental Table S1. The absolute values of the perturbations in chemical shifts observed in all our spectra appear in Fig. 2 and are tabulated in supplemental Table S2. Overall, 155/205 carbon atoms have been identified in both Pf1 L and Pf1 H (76% in 41 amino acids, missing are 4 glycines and Lys-45; 169 or 82% identified in Pf1 L spectra). Of these, 23 atoms in 14 amino acids, all of which are hydrophobic, show shifts Ն0.4 ppm. In Fig. 2, backbone and side-chain carbons are plotted for each residue independently and are marked according to their positions in the chain. For a threshold of 0.4 ppm, 3 it is immediately apparent that the most significant changes appear in the hydrophobic region of the capsid protein (residues 21-43) and are primarily observed for the side-chain atoms. The atoms that could not be resolved in the various two-dimensional spectra are indicated at the bottom of Fig. 2 (marked as NR).
After sufficient data had been collected for Pf1 L , the sample temperature was increased to 15°C to test whether the structural transition is reversible at the high concentration of ϳ200 mg/ml used. Fig. 3 shows an overlay of the low (Pf1 L ) and low to high (Pf1 L3 H ) forms, compared with an overlay of Pf1 L3 H and Pf1 H forms. Arrows in Fig. 3 indicate chemical shift changes stemming from the transition, which have been fully reversed upon the temperature increase.

DISCUSSION
Characterization of the variations between the high and low temperature forms of Pf1 is of significant importance in obtaining a more accurate structural insight of Pf1 H and in our understanding of the transition itself. In our data we observe chemical shift perturbations and some long range intersubunit contacts. The perturbations we observe are relatively small and are restricted mostly to side-chain atoms. Such changes are similar to those observed in "structure-activity relationship" studies by NMR of drug and ligand binding to proteins (31). In this way, our data are more compatible with changes in the intersubunit contact surfaces that occur upon the transition, rather than significant changes in the backbone structure of the coat protein. Fig. 2, the largest chemical shift changes occur in well defined hydrophobic patches of side chains near the C terminus. Fig. 4, based on the capsid model 4IFM, illustrates the positions of these patches. Some minor perturbations also exist near the N terminus. A surface plot of the significant shifts (⌬␦ Ն 0.4 ppm; ⌬␦ ϭ ͉␦(Pf1 H ) -␦(Pf1 L )͉) shows clearly the strong perturbations in the C-terminal patches. This indicates that reorganization of the packing of intermolecular hydrophobic contacts occurs in this structural transition, as was suggested earlier by Hinz et al. (18).

Locations of Atoms with Perturbed Chemical Shifts on Hydrophobic Surfaces-According to the perturbation summary in
A model of the interfaces within the capsid (Fig. 4b) locates the perturbations at the interaction surface between the subunits, based on Protein Data Bank code 4IFM. The assumption  (Fig. 2). For ⌬␦ Ն0.4 ppm, van der Waals surfaces are shown; otherwise stick models without surfaces are used in c1-c3. c1, axial view is from the N-terminal ends of the subunits on the outside toward their C-terminal ends near the center. c2, axial view is from the opposite direction as in c1. In the side view of c3, surfaces are only for ⌬␦ values Ն0.4, and in c4, surfaces are for all residues regardless of the shift perturbation. The gray colors indicate unassigned atoms including all non-carbon atoms but not hydrogens. NOVEMBER 19, 2010 • VOLUME 285 • NUMBER 47 that the side chains are packed in the manner depicted agrees well with previous observations on residues that are not amenable to chemical modifications (32,33). Fig. 4, b1, shows the three close-packed subunits 0, ϩ6, and ϩ11 (34) as viewed from the DNA axis toward the outside. Fig. 4, b2, is an enlargement of the principal contact region involving residue Ile-12 of subunit 0 (marked Ile 0 -12) and residue Ile 6 -26 located nearer the axis, as well as a bottom region involving residues Gln 0 -16, Ala 6 -29, Ile 6 -32, Val 11 -35, Leu 11 -38, and Ile 11 -39. A view from the outside in Fig. 4, b3 (a 180°rotation of Fig. 4, b2), shows that the largest chemical shift perturbations are not located in the external surface of the virion, but rather in the packing surfaces between the subunits.

Intersubunit Hydrophobic Interactions in Pf1 Filamentous Phage
To illustrate the location of the residues exhibiting significant shifts in the context of the entire virion, different views of an assembly of 60 subunits (Ͻ1% of the total virion length) are shown in Fig. 4c. Views in opposite directions along the axis of the virion are shown in Fig. 4, c1 and c2). The surfaces shown here emphasize where the regions of the largest chemical shift changes are concentrated. The side view of the virion (Fig. 4, c3) emphasizes the location of the main shift perturbations with respect to the exterior of the phage, and these perturbations become almost invisible in Fig. 4, c4, where the surface of the entire phage is drawn using a similar color scheme. It is apparent that the surface of the virion does not undergo any change, and the shifts we detect are not due to a different solvent exposure at the low temperature. Clearly these changes are buried between subunits.
It is important to note that interactions with the DNA could not be probed in these studies. Unfortunately, many of the atoms belonging to residues involved in interactions with the DNA (e.g. Met-42, Arg-44, and Lys-45) were assigned although broad in the high temperature form, and are missing or could not be unambiguously assigned in the present data for the low temperature form. Therefore, it is not clear if these residues are or are not shifted during the transition.
Correlation of Observed Perturbations with Capsid Models-To further explore correlations between the observed perturbations and intersubunit hydrophobic contacts at the atomic level, we first analyzed four models in the data base in terms of closest neighbors of adjacent subunits; Protein Data Bank codes 4IFM and 1PFI for the low temperature form and codes 2IFN and 1QLI for the high temperature form. We marked the amino acids belonging to an arbitrary subunit k ϭ 0, for which at least one carbon or nitrogen atom is within 3.5 Å of a neighboring subunit, and binned those atoms to the contacted neighbor. The distance of 3.5 Å is approximately a van der Waals interaction surface of two carbon atoms. We then considered the models in low/high pairs to predict surface modifications that may induce chemical shift changes. One difficulty in the comparisons was that model 1PFI has DNA, although the others do not, and 1PFI was not fully refined. Also, there is no high temperature model that directly corresponds to 1PFI. Nevertheless, results are in agreement with the pair 1PFI/2IFN on 17 occasions in which a chemical shift change corresponds to a contact difference and with 4IFM/1QL1 (15 occasions). For both pairs, 12 unique shifts are explained by changes in the interface (out of 15 total), but then for the other two pairs we found 8 shifts (4IFM/2IFN, 8 unique) and 11 shifts (1PFI/ 1QL1, 9 unique).
Because the low temperature structure is inherently better defined, we focused on it and chose the empty capsid model The colored areas enclose amino acids for which at least one carbon or nitrogen is within 3.5 Å of a neighboring subunit in the model, and the identity of this subunit is indicated by a colored number. a, contacts with subunits ϩ11 (blue), Ϫ11 (green), and Ϫ17 (red) are shown (for ϩ17 and only contacts with Ile-3 exist). b, contacts with subunits ϩ6 (blue), Ϫ6 (green), and the two residues of subunit Ϯ5 (orange and red) are shown. Amino acids circled in black show ⌬␦ Ն0.4 ppm in our data. The dashed circles in a correspond to shifts in van der Waals contact areas of the subunits emphasized in b and vice versa. See text for further details. c, a lattice diagram corresponding to the low temperature symmetry of the Pf1 filamentous bacteriophage. A rotation of 65.9°and rise of 3.05 Å were used to generate the dots. A representative subunit k is sketched as a broad bent line, and the neighboring subunits Ϯ6 are indicated by narrower lines. Subunits in contact with subunit k are numbered according to a right-handed sense of the Pf1 capsid helix.
4IFM. In Fig. 5, a helical wheel representation of residues 10 -45 is shown for the capsid. The surface lattice of the Pf1 capsid allows, for any subunit k, close contacts with neighboring subunits k Ϯ5, k Ϯ6, k Ϯ11, and k Ϯ17. Residues having atoms in contact with the particular subunit k are indicated by the colored areas in Fig. 5. The lattice diagram of the capsid arrangement for this symmetry is provided in Fig. 5c. In the helical wheels (Fig. 5, a and b), residues having atoms with ⌬␦ Ն0.4 ppm are indicated by the black circles. They correlate well with the 3.5 Å interface. For the region shown, all shifts Ն0.4 ppm (with the exception of Tyr-25) can be attributed to one of the van der Waals interaction surfaces of the low temperature model, and many of the other residues involved in contacts are either unassigned in our data or show shifts just below 0.4 ppm. A similar analysis of both the high temperature models 2IFN and 1QL1 shows that Tyr-25 is in proximity (i.e. Ͻ3.5 Å) to subunit k Ϫ6, and therefore every shift perturbation we see in this region can be associated with a surface interaction in either Pf1 L or Pf1 H .
It is important to note that approximately one-half of the residues involved in the intersubunit surfaces exhibit chemical shift perturbations, whereas only a couple of residues not involved show substantial shift perturbations (Ala-7 and Thr-5). Among the residues belonging to subunit k ϭ 0 in the 4IFM model of Pf1 L that are within 3.5 Å of a neighboring subunit we found the following: for k ϩ6, three fully assigned residues (Gln-16, Leu-30, and Ile-12) are perturbed, and C⑀ of Met-19 is unassigned; for subunit k Ϫ6, Ile-32 and Ala-29 (Leu-33 has two atoms shifted by ϳ0.3 ppm and C␦ 2 was detected only in Pf1 L ); for subunit k ϩ11, Ile-26 and Ile-22 (Gly-15 and Gly-23 are unassigned); and for subunit k Ϫ11, Val-35 and Leu-38 and probably Met-42 (Val-31 is unassigned and Lys-20 contacts are at the very end of the chain, atoms C⑀ and N. It was suggested that Lys-20 is in contact with the N terminus, potentially bal-ancing the charge of the aspartic acid at position 4.). Overall, of the 26 residues involved in surface contacts, we could resolve resonances of 21, of which 11 or 13 show shift perturbations (for 0.4 or 0.3 ppm thresholds). Ignoring the first five N-terminal residues that may have sample-and condition-dependent structures and positions, we estimate a probability of 60 -70% of observing at least one resonance shifted for a residue involved in the interfaces between subunits.
The perturbations also potentially provide a useful predictive tool to recognize interaction surfaces or changes in interaction surfaces. Among the 14 non-terminal residues in which chemical shift perturbations were detected (based on a 0.4 ppm cutoff), 11 or 78% are at an interaction surface of the low temperature model (8 or 57% for the high temperature model). Two additional residues (Tyr-25 and Val-8) are found on the interaction surface of the high temperature model 1QL1, increasing the number to 13, or 93%.
Intersubunit Contacts and Their Correlation to Capsid Models-Further characterization of the contact interfaces can be done via the identification of intersubunit cross-peaks. The helical nature of the coat protein causes significant ambiguity in the spectra of the fully enriched phage, and thus only very few contacts in our fully enriched sample can be expected to yield unambiguous correlations. In the spin diffusion experiments described above, we have identified three such intersubunit contacts. The cross-peak between Ala-29 C␤ and Leu-38 C␦ and/or Val-35 C␥, despite the ambiguity in its assignment, is clearly not an intramolecular peak and therefore provides evidence regarding intersubunit contacts. A similar conclusion holds for Ile-12 C␤ and Leu-38 C␦/Val-35 C␥, and the Ile-32 C␥-Met-19 C␣/C␤ contact (spectrum and inset shown in Fig.  1c). The latter was detected in the data for both Pf1 H and Pf1 L . Such quaternary contacts are useful for constraining the overall assembly of the coat protein. The contacts are generally compatible with existing models for the virion assembly, as shown in Fig. 6. For example, according to model 4IFM, the ␦-carbons of Leu 11 -38 (but not Val 11 -35 C␥) are 3.7-4.5 Å from the C␤ of Ala 6 -29 (located 5 positions down the virion axis). This proposed interpretation is depicted in the bottom portion of the binding pocket in Fig. 4b. For Ile 0 -12 C␤, the ␥-carbons of Val 11 -35 are within 4.5-5.5 Å but not Leu 11 -38. The contact between Ile-32 and Met-19 also appears at the interface linking subunits k ϭ 0 (for Met-19) and k ϭ 6 (Ile-32), and therefore reports on the pitch of the helix of subunit 0. (It is also expected that atoms from Tyr 6 -25 and Ile 6 -26 would also be in contact with Val 11 -35, and we identified a potential cross-peak between Val-35 C␥ and Tyr-25C␤ at 37.5-25.2 ppm; however, this contact is also ambiguous in that it is identified in relatively con- The dashed lines connect atoms in the model that correspond to the assigned cross-peaks. In magenta are carbon atoms within 5Å of others in the model that show perturbations. a, contact between Ala 6 -29 C␤ (Ala residue in red) and Leu 11 -38 C␦2. Although Gln 0 -16 C␥ is within 5 Å in the model, it does not show any cross-peak correlation for Pf1 L , but its chemical shift is significantly perturbed during the transition (Fig. 2). Also, Val 6 -35 and Val 11 -35 (in blue), initially part of Leu-38/Val-35 ambiguity are far from Ala 6 -29-C␤ and hence are not involved in a contact. b, left, contact between Ile 6 -32 C␥2 (Ile residue in red) and Met 0 -19 C␣/␤. Despite their proximity to Ile 6 -32 in the 4IFM model, we could not detect any unique Lys 0 -20 atoms in correlation with Ile 6 -32 atoms. On the right, a zoom-out view of the same contact region showing the location of other Met residues (blue), demonstrating the clear assignment of this peak to an intersubunit contact.
gested chemical shift regions of the spectra.) All three of these intersubunit interface contacts are important constraints for structure models of Pf1. Their presence in these spectra is encouraging regarding the feasibility of structural analysis.
Comparison with Other Studies Characterizing the Transition-On the basis of radial electron density distributions derived from x-ray diffraction by the virus both in fibers and in solution, Specthrie et al. (10) proposed that more changes are associated with the outer layer of the capsid than with the inner layer. Based on the calorimetry data, Hinz et al. (18) hypothesized that one or more hydrophobic side chains change positions to lower their surface contact with water. Welsh et al. (11), to account for their proposed conversion of a monomeric asymmetric unit in Pf1 L to a trimeric asymmetric unit in Pf1 H , suggested more extensive changes in subunit contacts and structure on the basis of proposed similarities to molecular crystals.
More in line with our observations, NMR measurements of oriented phage particles by static solid-state NMR by Thiriot et al. (17) led to the suggestion that the C-terminal region of the subunit ␣-helix has a slightly lower tilt angle at lower temperatures, producing a reduced radius of the virion and an overall slight extension. Dipolar wave analysis of these data show that the changes in the coat protein tilt angle occur at residues Gln-16 and Ala-29. In our data, we observe a significant C␤ chemical shift perturbation for Ala-29 (⌬␦ ϳ1.1 ppm), that may be associated with a backbone shift. Torsion angle analysis of our data using either TALOS or PREDITOR of the fragment Gly-28 -Ala-29 -Leu-30 is not sufficiently accurate to verify that this perturbation is consistent with the proposed backbone tilt. We detect a small change for the C␤ of Gln-16 (0.3 ppm, below the threshold); however, a perturbation for C␥ of Gln-16 is large (1.0 ppm), in contrast with most other residues belonging to the N-terminal half of the protein. This change may be associated with an intersubunit interaction pocket discussed above. It was also observed in the same study of oriented phage solutions (17) that the changes of the 15 N chemical shifts of all isoleucine residues occur simultaneously, suggesting that the whole protein is affected by the transition. Our data also indicate that chemical shift perturbations occur for all isoleucine residues (in addition to other residues) and that these perturbations occur mostly at the side-chain atoms. Nevertheless, the perturbations of the Ile chemical shifts in this case are not necessarily an indication of the whole protein, and our data show that most changes are located at the hydrophobic C-terminal part of the protein coat.
It is noteworthy that we observed a reversible transition in solutions of virus with concentrations as high as 200 mg/ml, suggesting that interactions between different virus particles even at this very high concentration are not strong enough to prevent the transition. In fact, for an approximate particle mass of ϳ36 MDa, and assuming a rod-shaped particle of size 2000 ϫ 7 nm, this concentration of 200 mg/ml results in an occupation of about one-third of the total volume, so that full hydration is maintained. This will leave enough solvent space between individual particles for the transition to occur.
Summary-We have used magic angle spinning solid-state NMR to study the intersubunit interface of the Pf1 filamentous phage, to better understand its structure via the structural phase transition that occurs at ϳ10°C, thus providing important insight toward understanding the assembly and morphogenesis of this bacteriophage. Our approach included re-assignment of the NMR resonances of the low temperature form of Pf1, identification of unambiguous intersubunit contacts, and detection of the chemical shift differences between its high and low temperature forms. These data provide evidence for the identities of side chains that interact at the intermolecular surfaces in the assembly and are most amenable to structural reorganization during the transition. We identify specific patches on the coat protein that show these interactions most clearly.