Extensive environment-dependent rearrangement of the helix-turn-helix DNA recognition region and adjacent L-tryptophan binding pocket is reported in the crystal structure of dimeric E. coli trp aporepressor with point mutation Leu75Phe. In one of two subunits, the eight residues immediately C-terminal to the mutation are shifted forward in helical register by three positions, and the five following residues form an extrahelical loop accommodating the register shift. In contrast, the second subunit has wildtype-like conformation, as do both subunits in an isomorphous wildtype control structure. Treated together as an ensemble pair, the distorted and wildtype-like conformations of the mutant apoprotein agree more fully than either conformation alone with previously reported NOE measurements, and account more completely for its diverse biochemical and biophysical properties. The register-shifted segment Ile79-Ala80-Thr81-Ile82-Thr83 is helical in both conformations despite low helical propensity, suggesting an important structural role for the steric constraints imposed by β-branched residues in helical conformation.
The tryptophan repressor of Escherichia coli (TrpR) has served as an important early example of the complexity of relationships among protein flexibility, stability, and function. The TrpR dimer has extraordinary thermal stability, unfolding cooperatively to monomers in vitro with a midpoint temperature of ∼92°C.1 Yet many complementary lines of experimental evidence indicate that the helix-turn-helix region (HtH) that contacts DNA is conformationally mobile even in the presence of its ligands L-tryptophan and DNA. The earliest evidence came from crystallographic studies showing that HtH helices D and E (Fig. 1) can adopt different positions relative to the dimeric protein core (helices A, B, C, and F of both subunits), independent of L-tryptophan occupancy.2 NMR studies by Jardetzky and coworkers3–7 demonstrated that segments D and E are helical in solution on a nanosecond, but not a millisecond, timescale even in the presence of L-tryptophan. Thermodynamic analysis of TrpR binding to L-tryptophan and to DNA8 revealed very large heat capacity changes, the signature of protein surface burial.9 Both crystallography and NMR showed that DNA binding is accompanied by conformational adjustments in HtH helices D and E.10, 11 Further indications of HtH mobility derive from proteolytic sensitivity in helix D,12–14 molecular dynamics simulations,15–18 and dramatic conformational rearrangement of the HtH region in a domain-swapped crystalline TrpR gel.19
A genetic screen designed to identify mutations conferring temperature-sensitivity in TrpR function led to isolation of the point mutant L75F.20 In wildtype (wt) TrpR, the partially solvent-exposed side-chain of Leu75 at the C-terminus of helix D contacts residues of the HtH and the dimeric core. Compared to wt, the L75F mutant protein with Phe75 has ∼10-fold weaker affinity for L-tryptophan, 2- to 5-fold weaker affinity for operator DNA, and slight increases in helix content and stability toward urea denaturation.20 Although NMR spectra indicate that the average solution structure of L75F apoTrpR resembles wt apoTrpR, long-range effects on dynamics are evident from significant changes in NOE patterns and amide-proton exchange rates.20 Very large NMR chemical-shift differences are detected for residues ∼25 Å distant from Phe75, much too far away to be ascribed to ring-current effects, and seemingly inconsistent with the wt-like average solution structure.21 Measurements of apoTrpR backbone amide-bond dynamics by 15N relaxation indicate that helix D is more ordered in L75F than in wt, and that helix E is less ordered.22
We initiated crystallographic study of the L75F apoTrpR mutant to investigate the unusual biophysical properties conferred by the L75F point mutation. We report here the structure elucidation and analysis of isomorphous crystals of L75F and wt apoTrpR, which enable resolution of effects of the L75F mutation from effects of crystal packing and permit comparison with the NMR data.
Results and Discussion
Isomorphous wt and L75F apoTrpR crystal structures
Structures of wt and L75F apoTrpR were obtained from monoclinic crystals grown under identical conditions and analyzed in parallel (see Methods). Each structure consists of one complete homodimer with two crystallographically independent subunits. Summary crystal, data collection, and refinement statistics are provided in Table I.
Table I. wt and L75F apoTrpR Crystal Structure Statistics
In each structure the two subunit chains differ strongly in relative order as judged by main-chain atomic displacement parameters, particularly in N-terminal arm and HtH regions [Fig. 2(A,B)]. The nonequivalent HtH regions are designated here as HtH1 and HtH2. The asymmetry is a consequence of differences in local crystal packing environments (Supporting Information Fig. S1). HtH1 has extensive contacts with the N-terminal arm of a neighboring dimer, whereas HtH2 is largely free of crystal contacts. The N-terminal arm-HtH1 contact buries ∼900 Å2 of total accessible surface area and resembles the arm-HtH interaction between TrpR dimers bound to DNA as well as interdimer contacts seen in other TrpR crystal structures (Fig. 3). The relative orientations of the interacting dimers and the extent of arm-to-HtH contact differ considerably among these structures, but in each case the N-terminal arm Tyr7 phenol ring fills a shallow, largely hydrophobic depression lined by residues within or near the HtH (Leu61, Leu62, Ser86, Leu89, Lys90, and Arg97). The presence of arm-HtH interactions in multiple TrpR crystal structures lends strong support to the suggestion that TrpR N-terminal arms mediate higher-order association of dimers in solution.25–27
The two subunits of the wt dimer have nearly identical conformations [Fig. 2(A)], but the two L75F subunits differ substantially in the conformations of their HtH regions [Fig. 2(B)]. The better-ordered HtH1 of L75F [Fig. 2(B)] participates in the arm-HtH tandem-like contact [Fig. 3(C)], (Supporting Information Fig. S1) and has a wt-like fold. The more poorly ordered HtH2 of L75F has a substantially deformed helix E and D-E turn. Because the HtH2 region is relatively free of crystal contacts in both structures, the altered conformation in the L75F structure is likely a consequence of the mutation.
Comparison of wt and L75F HtH1 regions
Within the better-ordered HtH1 region the wt and mutant structures are only subtly different from each other. The Phe75 phenyl ring occupies the same position as the Leu75 isobutyl group in the wt structure [Fig. 2(C)]. Accommodation of the slightly longer, bulkier Phe75 is achieved not locally but through a ∼1 Å rigid-body shift of the entire HtH1 [Fig. 2(C)]. The numerous contacts made by the side-chain of residue 75 with residues of helix C are conserved.
The L-tryptophan binding site adjacent to HtH1 has a slightly larger pocket volume in L75F (316 Å3; see Methods) when compared with wt (302 Å3), but these are both much larger than in previously determined crystal structures for apoTrpR (178 Å3) and L-tryptophan-bound TrpR (246 Å3).2, 28 The variation in L-tryptophan pocket volume results directly from flexibility of the HtH helices, which adopt different positions in different crystal structures.2 Both wt and L75F pockets contain partially ordered solvent. The L75F binding site holds in addition a tris cation ((HOCH2)3CNH3+) bound with partial occupancy; the tris amino nitrogen and bulky central carbon sit roughly at L-tryptophan α-amino nitrogen and α-carbon positions, respectively [compare Fig. 1 and Fig. 2(B)]. The presence of tris in only the L75F structure suggests that the mutant may be more permissive than wt to entry of molecular entities bulkier than the native ligand. Although the binding of tris adjacent to HtH1 is unique to the L75F structure, it is unlikely that its presence influences HtH2 secondary structure, since there is no cooperativity between the two L-tryptophan binding sites of the TrpR dimer.8, 20, 29
Comparison of wt and L75F HtH2 regions
Within the HtH2 region the two structures differ more substantially. Helix D of the mutant is shifted essentially as a rigid body by ∼3 Å relative to wt [Fig. 2(D)]. In addition, the residues that belong to the interhelical turn and helix E in the wt structure undergo a remarkable conformational change in the mutant (Fig. 4). Residues 76–83 form an alternative helix (E′) with the first turn in 310 conformation. Mutant helix E′ occupies approximately the same position as wt helix E residues 79–86; in other words, the mutant helix is shifted out-of-register by three residues relative to wt (Fig. 4). In the register-shifted conformer, Ile82 takes the position occupied in wt by Gly85. In the wt conformer the absence of a side-chain at Gly85 accommodates the L-tryptophan indole18, 30; therefore, L-tryptophan binding would be incompatible with the distorted mutant conformation.
In the mutant, the five residues following helix E′ form a large loop, whereas in wt the same residues form the central part of helix E [Fig. 2(D)]. The entire loop is heavily solvated and makes no crystal contacts. The largest Cα-atom position shift between wt and L75F structures is ∼10 Å and is found in this segment at Ser86, 11 residues distant in primary structure from the mutation site. The sequence of loop residues 85–88, GlySerAsnSer, is turn-like with low propensity for both helix and strand (Supporting Information Fig. S3).
Residues 89–92 are in roughly the same position and orientation in both mutant and wt structures, forming the last turn of helix E in wt and a single-turn helix (E″) in L75F. Mutant helices E′ and E″ are tip-to-tip and nearly coaxial (offset ∼30°). Atom geometries and distances imply that the two segments are linked by a single helical-like backbone hydrogen bond between Ile82 and Leu89, and by a backbone hydrogen bond between Thr83 and Lys90 bridged by a water oxygen. Although the main-chain conformation of Ser88 is extended beta rather than helical, its carbonyl group makes the only i, i + 4 backbone hydrogen bond within helix E″, to the Ala92 amide. The Ser88 side-chain hydroxyl oxygen can also form a hydrogen bond with the Ala91 amide, providing an N-terminal cap31 for the short E″ helix. There are no side-chain interactions between helices E′ and E″ instead, residues of both helical segments participate in hydrophobic contacts with residues of surrounding helices as discussed further below.
Resolving effects of crystal contacts and mutation
The dramatic difference in conformations of the two chains of the L75F mutant apoTrpR dimer raises the question of whether the distortions in HtH2 arise from mutation or from crystal packing. When the two HtH regions are superimposed on each other within the context of the crystal lattice, the relatively open environment of the deformed HtH2 appears to offer no steric hindrance to adoption of the wt-like conformation. A refinement test designed to investigate the possibility that wt-like conformation might be adopted by at least a small percentage of L75F apoTrpR dimers at the HtH2 position within the crystal gave a clear negative result (see Methods). In contrast, the relatively restrictive HtH1 lattice environment, which includes the tandem-like arm-HtH contact, cannot accommodate the deformed conformation owing to severe clashes. These observations indicate that the HtH conformation of the L75F mutant depends on the environmental context, i.e., whether or not there is contact with a neighboring dimer in the crystal lattice. Analogously, the HtH2 lattice environment of wt apoTrpR offers no steric hindrance to adoption of a distorted conformation. Thus, the conformational difference between wt and mutant HtH2 conformations can be attributed directly to the L75F mutation.
To better understand how the L75F mutation enables the conformational change, side-chain packing interactions in the apoTrpR structures were analyzed. The two alternative packing arrangements can be described as differing in the register of the helical Ile79-Ile82 segment relative to invariantly positioned residues of the L-tryptophan binding pocket: Ile57 and Val58, which are located in the C-terminal half of helix C, and Leu89, located near the C-terminus of helix E or within helix E″ (Fig. 5). In wt and wt-like mutant conformations, Leu/Phe75 and helical residues Thr81 and Ile82 pack against Ile57 [Fig. 5(A,B)]. In the distorted mutant conformation, the helix-register shift and 310 conformation result in Ile79 occupying the wt position of Ile82, and Ile82 occupying the wt position of Gly85 [Fig. 5(C)]. In the distorted conformation only, Ile79 joins the cluster of packed residues. Thus, the alternative packing of two segments with relatively invariant local structures, 79–82 and 57–58, links the burial of Ile79 to the looping out of residues 84–88 from helix E.
As expected, the solvent-accessible surface areas of the two packing arrangements differ significantly. When the whole chain or only the HtH regions are considered, the distorted conformer of apo L75F TrpR buries slightly less hydrophobic as well as total surface area than the wt-like conformer, and both conformers bury slightly more surface area than wt apoTrpR (Supporting Information Table S1). However, when only the alternatively packed β-branched segments are considered, the distorted L75F conformer has hydrophobic and total surface area burial that is approximately 2-fold more efficient than wt or wt-like L75F. Although the crystal structures indicate that both conformations are accessible to L75F apoTrpR, the enhanced local surface burial of L75F may explain why the distorted conformation is favored for the mutant in the absence of crystal contacts. Differences in solvent-accessible surface areas cannot be used directly to assess the relative stabilities of the two conformers because crystallization is an inherently hysteretic process and thus crystal structures need not reflect equilibrium distributions of conformers. On the other hand, because the differences in structure between the two conformers of L75F are strictly local, and appear to be controlled by the alternative packing of these β-branched segments, the relative stabilities of the two conformers may also reflect the local packing efficiency of these segments rather than the global packing efficiency.
Comparison with NMR data
NMR data for L75F TrpR clearly indicate the long-range nature of the effects of the L75F mutation. Although early NMR spectra were fully consistent with a wt-like average structure, the number of NOESY cross-peaks was much larger than in wt spectra acquired under the same conditions.20 ROESY spectra ruled out spin diffusion as a trivial cause of the NOESY results.20 The rate of deuterium exchange of unresolved amide protons was ∼50% slower than in wt, as were exchange rates for the resolved indole protons of both Trp19 and Trp99.20 All of the proton exchange data were adequately fit by single-exponential decay functions, indicating generally similar rate-limiting steps for exchange in the wt and mutant proteins. Two well-resolved, slowly-exchanging amide protons were detected in 1D spectra of the mutant only, as well as a prominent NOESY crosspeak between them20; these were later assigned to Met42 and Leu43 in the B-C turn.21 In addition, a prominent HSQC crosspeak unique to the mutant was detected20 that was later assigned to the Thr44 amide.21
Full assignment of L75F apoTrpR21 revealed extremely large chemical shift changes for residues up to ∼26 Å from Phe75, far too distant to be explained by effects of the ring current of the new aromatic residue. Furthermore, many distant residues were very strongly shifted while others closer to the mutation site were not. Among amide nitrogens assigned in both mutant and wt, the largest 15N chemical shift differences were found for Thr81, Ile82, and Thr83, which shift upfield by 2.03, 3.90, and 4.42 ppm, respectively; the largest shifts for assigned amide nitrogens closer to the mutation site are 1.32 ppm upfield for Asn73 and 1.75 ppm downfield for Gly76. For comparison, the distances from Phe75 to each of these residues on the wt-like side of the L75F crystal structure reported here are Thr81, 10.4 Å; Ile82, 10.2 Å; Thr83, 14.0 Å; Asn73, 5.9 Å; and Gly76, 3.8 Å. Examples of residues that are similarly distant from Phe75 as residues 81–83 but experience little or no change in chemical shift include Gly78, 8.7 Å, −0.17 ppm; and Glu70, 8.3 Å, −0.08 ppm. The chemical-shift results in the L75F HtH region were previously rationalized as reflecting a global response to the mutation with effects at nearby residues complicated by ring-current shifts.21 The ring current was assumed to extend as much as 10 Å from Phe75 to include potential anisotropic effects. Nevertheless, very significant 15N chemical-shift changes (∼1–2 ppm upfield for backbone amides) as distant as ∼28 Å away in the same subunit and ∼26 Å away in the other subunit could not be explained.
The average NMR structure of L75F apoTrpR derived from NOESY restraints is very similar to that of wt TrpR, as are all the individual members of the family of accepted structures.21 Like all NMR structures of TrpR solved to date, the dimeric core of L75F apoTrpR (helices A, B, C, and F) is well-constrained by the data. In contrast, considerable variation in the HtH regions of L75F was attributed to instability of these domains, following the interpretation of more extensive data available for wt apoTrpR.3, 5 When compared with the dimeric core, the L75F and wt HtH regions share a general paucity of NOEs, limited protection of amides from exchange, and smaller chemical-shift deviations from random coil, indicating that they are much less well-defined. However, other NMR measures indicate apparently more helical structure for the HtH of L75F than for wt. L75F residues 67–83 have several i, i + 3 NOEs and two i, i + 4 NOEs between α-carbon and backbone amide protons that are absent in wt, as well as larger helical chemical shift deviations for α-carbons and α-protons.21 These data provide the restraints that result in models for L75F apoTrpR that are more helical in the HtH region than corresponding models for wt.
Although the environment-dependent distortion in the L75F crystal structure suggests that both conformations might also occur in solution, none of the HtH region conformations in the family of accepted NMR solution structures for L75F apoTrpR (PDB id 2xdi)21 closely resembles either conformation in the crystal structure. This finding prompted evaluation of both wt-like and distorted conformations against NOE-based distance restraints. Of 728 measured NOEs, only ten conflict with either one or both crystal structure conformations, with inferred interproton distances that are too long to generate an NOE of the measured strength (Table II). Significantly, the discrepancies are not found throughout the molecule but involve exclusively residues involved in the alternative packing depicted in Figure 5. Many of these discrepancies do involve residues relatively close to the Phe75 aromatic ring, but ring-currents that can affect chemical shift positions do not affect relaxation processes such as NOEs.
Table II. Comparison of NOEs and Inferred Interproton Distances for L75F apoTrpR
Interproton distances or distance ranges were measured between predicted hydrogen atom positions in the L75F apoTrpR crystal structure (PDB id 3SSX). Bold ranges indicate values consistent with the observed NOE.
Phe 75 Hβ
Ile 79 Hγ2
Phe 75 Hδ1
Ile 79 Hγ2
Gly 76 H
Ile 79 Hβ
Gly 76 Hα
Ile 79 H
Ala 77 Hα
Ile 79 H
Ile 57 Hγ2
Thr 81 Hγ2
Phe 75 H
Ala 77 Hβ
Phe 75 Hβ
Ala 80 Hβ
Gly 76 H
Ile 79 Hγ1
Asn 87 Hα
Leu 89 Hδ1
Of the 10 identified NOEs, five are consistent with the distorted conformation but inconsistent with wt-like conformation. In particular, two strong NOEs with upper limit distances of 4.5 Å from Ile79 to Gly76 and to Ala77 are much better accounted for by the distorted structure. Conversely, 2 of the 10 NOEs are fully consistent with wt-like conformation but inconsistent with distorted conformation. Thus, neither structure alone can account for all the data. The final 3 of the 10 NOEs involve residue pairs Phe75-Ala80, Gly76-Ile79, and Asn87-Leu89; neither wt-like nor distorted conformations have interproton distances sufficiently short to satisfy the corresponding restraints. However, an intermediate conformation representing a transition between the two crystallographic conformations might reasonably be expected to bring each of these residue pairs into sufficiently close proximity to yield an NOE.
NOE strength, though proportional to distance, is not necessarily proportional to population or duration; thus, NOEs can over-represent transitory close encounters. Indeed, van Gunsteren32 has developed an approach that uses NOE-derived restraints as time-averaged bounds. In this method the ensemble as a whole satisfies all the observed NOEs, rather than the conventional treatment that requires every structure in the ensemble to satisfy all restraints. If the HtH region of L75F apoTrpR alternates between two conformations in solution, each conformation would account for only some of the NOEs, and the structure calculated by the conventional approach would not reflect these dynamics, but present only an apparently disordered region. This situation appears to best account for the NMR results on L75F apoTrpR.
Reanalysis of NMR evidence for L75F apoTrpR thus suggests that more than one conformation must be populated in solution to account for all the data. The highly nonuniform patterns of NMR changes include those that are predicted if the distorted crystal conformer of L75F were present in solution. A recent analysis of 15N backbone relaxation data for L75F apoTrpR22 indicates that the spin–spin relaxation of Ile82 is strongly affected by an exchange process, supporting the existence of a conformational transition involving the HtH region. The relaxation data also indicate enhanced flexibility for residues 84–88 in L75F relative to wt. The picture that emerges is that in solution, L75F apoTrpR samples conformations that likely include both conformations observed in the crystal structure. The transition between the two crystallographic conformations could be accomplished by relatively modest chain excursions compared with conversion to the domain-swapped gel form that both wt and L75F TrpR undergo.19
In addition to accounting for NOEs in the immediate vicinity of the mutation, the presence of the distorted structure in solution would provide a highly satisfactory explanation for the effects detected by NMR involving the distant amide groups of Met42, Leu43, and Thr44. In wt apoTrpR and the wt-like side of L75F apoTrpR, these residues are closely approached by the guanidinium group of Arg84, whereas in the distorted L75F conformation the looping-out of segment 84–88 between helices E′ and E″ moves Arg84 away and brings several uncharged residues nearby (Thr81, Ile82, Ser86, Asn87). This change represents a substantial perturbation of the electronic environment that could readily affect nearby amide protons. Furthermore, since Leu43 methyl protons display NOEs to Trp19 indole ring protons, the change might be responsible for the large chemical shift changes experienced by all the hydrophobic residues on the Trp19 face of helix A in the L75F mutant, as well as the blue-shifting of the fluorescence emission spectrum attributed to Trp19 (see discussion below).20 All of these puzzling perturbations detected at so many residues far removed from the site of mutation have not otherwise been explained by any NMR structures or relaxation analyses conducted to date.
Comparison with biophysical data
A large body of data that has accumulated on L75F apoTrpR can be partly rationalized by the structures and analysis reported here.
The urea denaturation midpoint monitored by circular dichroism (CD) is slightly higher than for wt (∼7.0 M vs. ∼6.5 M), but the thermal midpoint assessed by differential scanning calorimetry (∼91°C) is unaltered.20 The heat capacity change upon denaturation is similar to that of wt (∼90 kcal/mol), and the ratio of van't Hoff to calorimetric enthalpy is near one for both proteins, indicating a two-state unfolding process.20 These data are consistent with the general picture of TrpR as a highly stable but reversibly folding dimer.33, 34 The interdependent tertiary and quaternary organization of the intertwined central core is expected to be relatively insensitive to structural changes in the more peripheral HtH DNA-binding domains, consistent with local control of HtH stability based on the packing data. The independence of the HtH domains from the central dimer fold is dramatically illustrated by the finding that chymotrypsin cleavage of wt apoTrpR dimers within the HtH yields four fragments that reassociate spontaneously to regenerate the native NMR spectrum.14
Although apoTrpR stability is largely unaltered by the L75F mutation, the CD spectrum shows an increase in helix content compared to wt from ∼77 to ∼85% of residues.20 However, the structures show that WT apoTrpR has two more residues in helical conformation than the distorted conformer, in apparent opposition to the CD results, although the effects of helix length and distortion can complicate interpretation of the fraction of helical residues.35 On the other hand, the increased helical CD intensity and slightly higher resistance to urea denaturation are consistent with the modest improvement in relative order seen in the crystal (Fig. 2). Both proteins reach ∼87% helix content upon titration with trifluoroethanol,20 implying that even in the absence of cosolvent L75F TrpR has nearly the maximal helix content permitted by its primary structure. Only the domain-swapped crystalline gel formed by wt or L75F TrpR in isopropanol has higher helix content (∼89% as calculated from the structure), due to conversion of the C-D and D-E turns to helical conformation.19
The intensity-weighted fluorescence emission maximum is blue-shifted by ∼2.3 nm in L75F apoTrpR relative to wt, mainly due to changes at the red edge of the spectrum, with an overall increase in intensity of ∼14%.20 With excitation at 280 nm, emission is confined almost entirely to the two native Trp residues at positions 19 and 99. Based on analysis of the NMR results,21 the spectral changes could be assigned to an altered electronic environment at Trp19 despite its being ∼25 Å away from Phe 75 in the same subunit and ∼15 Å away in the other subunit in wt conformation. However, the distorted conformation of L75F better explains the fluorescence results as reflecting the proximity of Trp19 to Leu43 in the L-trp binding pocket, where the electronic environment is strongly affected by the alternating proximity of Arg84. The altered L-trp binding affinity of L75F20 is also consistent with occlusion of the binding pocket by the distortion.
Significance of β-branched residues in helical segments
In the course of comparing the 20 accepted NMR structures of L75F apoTrpR with the crystal structure reported here, the isolated HtH regions of the forty subunits of the NMR models and the two crystal subunits were superimposed on each other independently of their positions with respect to the dimer core. Residues 80–82 were found to adopt a highly similar conformation in all structures. This result was unexpected considering the limited set of NMR constraints available in the HtH region, and prompted a more detailed evaluation based on the variation in main-chain phi, psi angles in these structures (Supporting Information Fig. S4). The variation in main-chain conformation is high in turns and low in helical regions, with the notable exception of the HtH region. Helix E residues 83–89 present almost as much variation as do turn regions, but residues 80–82 are nearly as invariant as helical residues of the dimeric core, with phi, psi angle values near the helical minimum on the Ramachandran plot. Thus, even though the wide distribution of HtH conformations in the family of NMR structures implies apparent instability of this region, residues 80–82 appear to have intrinsic helical order. This finding is consistent with the alternative packing patterns in the crystal structure in which this helical segment appears to move as a unit.
The seemingly high degree of helical order for residues 80-82 is also unexpected given their local sequence environment, as they are embedded within a segment rich in β-branched and Gly residues: 76-GlyAlaGlyIleAlaThrIleThr-83. This sequence has relatively low helical propensity (Supporting Information Fig. S3). The occurrence of β-branched residues in helical segments is entropically disfavored because their side-chains can access only one rotamer conformation due to steric hindrance between the bulky Cβ branch point and the helical backbone.36 In principle, the steric hindrance could be relieved by unraveling of the segment into a more extended structure, but the resulting excursion of the chain might compromise tertiary interactions. A similar argument can be made regarding residues 76-GlyAlaGlyIle-79, which adopt 310-helix conformation in the distorted L75F apoTrpR subunit. Their side-chains offer minimal steric hindrance that would promote extension of the helix rise from α- to 310 conformation, and this sequence also has overall low helical propensity (Supporting Information Fig. S3). The fact that it adopts 310 conformation rather than α-helix or random coil is consistent with a role for tertiary constraints. Tertiary constraint of secondary structure was invoked to explain why additional residues inserted within helical segments of lysozyme are accommodated not by helix lengthening but by displacement of adjacent helical residues into flanking turn regions.37 The coupling of secondary and tertiary structures is thought to be the molecular origin of protein folding cooperativity.38
In TrpR, the tertiary interactions made by the β-branched residues of helix E or E′ involve Ile57 and Val58, which are within a segment of helix C that is also rich in β-branched residues, 53-ThrArgValArgIleVal-58. Thus, the distortion and register shift of the E′ helix act together to accommodate the en bloc movement of one β-branched segment (79–83) while maintaining efficient packing with another β-branched segment (53–58) that preserves its conformation. The participation of β-branched residues in tertiary constraint calls to mind the transmembrane helix of glycophorin A,39 in which such residues play a key role at the quaternary level as well. β-branched residues are common in helical segments of membrane proteins, and have been thought to reflect altered helical propensity in the membrane environment.40 However, the bilayer acts as a constraint on helix length and on helix–helix packing through its influence on the exposure of polar and nonpolar residues to the surrounding lipid, analogous to the tertiary constraint in soluble proteins. Helical β-branched residues present side-chains that are preorganized for interaction with intra- or intermolecular partners, making them perhaps ideally suited to enforce super-secondary constraints whether these arise from the tertiary structure in the case of soluble proteins or the bilayer in the case of membrane proteins.
Sigler and coworkers had earlier coined the term “hydrophobic brace” to describe the interactions between residues in these β-branched-rich segments because they appeared to fix the intrinsic conformation of the HtH motif and its orientation relative to the rest of the molecule in TrpR as well as in several other HtH repressors.28 In Figure 4, brace residue positions are indicated by dots. Mutation of Leu75 to Phe crowds the brace, resulting in rearrangement as reflected in the alternative packing arrangements shown in Figure 5. Rearrangement of the brace to accommodate the mutation implies that the brace is not a rigid enforcing structure as its name suggests, but rather is adaptable. The majority of brace residues in Sigler's group of repressors are Ile, Leu, and Val, all of which in addition to being hydrophobic have severe restrictions on their side-chain conformations. The present analysis thus suggests that a more complete definition of the brace would highlight steric constraints on conformational mobility.
Clusters of the branched aliphatic side-chains Ile, Leu, and Val (ILV) have recently been implicated as early organizing features of both native and off-pathway folding intermediates of TIM barrel and Rossmann fold proteins.41 In native proteins ILV clusters appear to correlate with regions in which backbone amides are highly protected from hydrogen exchange, an effect that has been ascribed to the solvent-excluding property of the branched side-chains that shields the underlying backbone.42, 43 Residues Val55, Ile57, Val58, and Ile82 from the β-branched segments of helices C and E, as well as Leu75 from helix D, belong to the ILV cluster of wt TrpR (S. Kathuria and C. R. Matthews, personal communication). However, the backbone amides of both Leu75 and Ile82 are among the most rapidly exchanging in wt apoTrpR.4 Thus, cluster residue Ile82 is unlikely to contribute to the apparently intrinsic helical character of the 80–82 segment of helix E by a solvent-exclusion mechanism, supporting a role for steric hindrance. It is worth noting that even Leu, though not β-branched, has significant steric restriction that may contribute to the role of ILV clusters. In the rotamer library of Dunbrack and Karplus,36 only 2% of the Leu chi1 conformers lie in the range 60° ± 60°; the remainder are divided about 2:1 between the other two conformers, regardless of backbone conformation. The relative importance of steric hindrance and solvent exclusion in ILV clusters may be resolvable with a larger database of protein hydrogen-exchange rates.
The crystal structures of isomorphous wt and L75F apoTrpR that are reported here reveal considerable perturbation in the L-tryptophan-binding and DNA-binding HtH region caused by the Leu to Phe substitution. In crystals of the L75F mutant the extent and nature of structural distortion depend on local environmental context: an extensive crystal lattice interaction at one HtH supports a wt-like conformation, whereas the other HtH lacking crystal contacts presents a major rearrangement of secondary and tertiary structure. Given that (i) the lattice interaction supporting the wt-like conformation of L75F involves an N-terminal arm-HtH contact that is also likely to be relevant in solution, and (ii) the distorted HtH conformation largely lacks crystal contacts, it is likely that both conformations observed in the crystal are sampled by L75F apoTrpR in solution. Inclusion of the these conformers within the family of accepted NMR structures of L75F apoTrpR provides a more fully satisfactory explanation of all existing biophysical and NMR data. The results indicate that distinct alternative conformations can be overlooked despite full NMR structure determination as well as detailed relaxation measurements. Without structural models of specific alternative conformers to guide analysis, NMR data may appear to indicate poor ordering of a segment.
Crystallization and data collection
wt and L75F apoTrpR were purified as described.20, 44 Protein stocks were stored frozen at ∼12 mg/mL in 250 mM NaCl, 0.1 mM PMSF, 0.1 mM sodium phosphate, pH 7.0, at −20°C prior to crystallization trials. Crystals were obtained by standard hanging-drop vapor diffusion at 294 K with Hampton Research Crystal Screen condition #6 (30% (w/v) PEG 4000, 200 mM MgCl2, 100 mM Tris HCl pH 8.5). Single crystals grew overnight with typical dimensions 0.2 mm × 0.2 mm × 0.05 mm. One wt apoTrpR crystal and one L75F apoTrpR crystal were soaked for 30 s in Hampton Research crystal screen cryo condition #6 (24% (w/v) PEG 4000, 180 mM MgCl2, 80 mM Tris HCl, pH 8.5, and 20% (v/v) glycerol), mounted on nylon loops, and flash-cooled to 100 K in a N2 gas stream. X-ray diffraction data were collected at the Brookhaven National Laboratory National Synchrotron Light Source beamline X26C with an ADSC Quantum 4 CCD detector. The incident radiation wavelength was 1.1 Å. Reflections were integrated, scaled, and merged using HKL2000.45 Data collection statistics are provided in Table I.
Initial phases were obtained for both wt and L75F mutant structures by the molecular replacement method using the program AMORE.46 In each case the search model was orthorhombic TrpR dimer2 (PDBid 2oz9), with L-tryptophan ligands and solvent excluded. Rotation and translation solutions were unambiguous. Each structure was subjected to multiple cycles of refinement and manual model-building using Phenix47 and COOT.48 Atomic displacements were modeled with individual isotropic B-factors and a single set of translation-libration-screw (TLS) parameters for all atoms.
Analysis and display
Conformation occupancy test: a composite L75F apoTrpR model was constructed with residues 74–95 from the HtH2 chain in both wt-like and deformed conformations, each initially at half occupancy. The model was subjected to refinement against the X-ray diffraction data for occupancy only, with no solvent model imposed. The occupancy of the deformed conformation increased to 0.95, while the occupancy of the wt-like conformation decreased to 0.07, indicating that wt-like conformation is not present in the HtH2 region of the L75F structure.
The Molprobity server49 was used to evaluate main-chain conformation, as reported in Table I; the Molprobity server was also used to add hydrogen atoms to the crystal structures for comparison of the L75F apoTrpR crystal structure with NOE restraint data. Each proton–proton distance was measured and compared to the cutoff distance assigned on the basis of the strength of each NOE, taking into account distance ambiguities for chemically equivalent protons.
L-tryptophan ligand pocket volumes were measured using the Castp server.50 Because the position of the Arg84 side-chain varies significantly between the models compared, the calculations were performed on models in which Arg84 side-chain atoms beyond Cβ were removed.
Secondary structure propensities (Supporting Information Fig. S3) were determined using four different propensity scales as follows. Stapley-Doig α: ΔG values for residue in helix relative to residue in coil.51 Stapley-Doig β: ΔG values for residue in β-strand relative to Ala in β-strand.51 Minor-Kim β: ΔΔG values for residue in β-strand relative to Ala in β-strand.52 Pace-Scholtz α: ΔΔG values for the residue in α-helix relative to Ala in α-helix.53 For each propensity scale, values for individual residues were summed and the sum divided by the number of residues in the segment.
Variance of main-chain torsion angles phi and psi (Supporting Information Fig. S4) in the 20 L75F apoTrpR NMR model chains and two X-ray crystal structure model chains were calculated using Procheck-NMR.54
Calculations of accessible surface area (Supporting Information Table S1) used the program NACCESS.55 The solvent probe was a water molecule of 1.4 Å radius, and the slice size (Z-step) was 0.05 Å.
Graphical images were created using Pymol56 (Figs. 1, 2, 3, 1–3, 5, and Supporting Information Fig. S1) and UCSF Chimera57 (Supporting Information Fig. S2).
We thank M. Gryk, G. Rose, and C. Yanofsky for valuable discussions, S. Kathuria and C.R. Matthews for bringing ILV clusters to our attention and for applying their analysis to TrpR, G. Montelione for advice on comparing NOEs to calculated distances, V. Copié for providing the ensemble of models of the L75F apo-trpR mutant minimized against NMR restraints prior to their deposition and the NMR relaxation results prior to publication, and A. Heroux for assistance with data collection at National Synchrotron Light Source beamline X26C, which is supported by the Department of Energy and the National Institutes of Health (NIH).