Toward Precise Interpretation of DEER-Based Distance Distributions

Article pubs.acs.org/biochemistry

Toward Precise Interpretation of DEER-Based Distance Distributions: Insights from Structural Characterization of V1 Spin-Labeled Side Chains Aidin R. Balo,† Hannes Feyrer,† and Oliver P. Ernst*,†,‡ †

Department of Biochemistry and ‡Department of Molecular Genetics, University of Toronto, Toronto, Ontario M5S 1A8, Canada S Supporting Information *

ABSTRACT: Pulsed electron paramagnetic resonance experiments can measure individual distances between two spin-labeled side chains in proteins in the range of ∼1.5−8 nm. However, the flexibility of traditional spin-labeled side chains leads to diffuse spin density loci and thus distance distributions with relatively broad peaks, thereby complicating the interpretation of protein conformational states. Here we analyzed the spin-labeled V1 side chain, which is internally anchored and hence less flexible. Crystal structures of V1-labeled T4 lysozyme constructs carrying the V1 side chain on α-helical segments suggest that V1 side chains adopt only a few discrete rotamers. In most cases, only one rotamer is observed at a given site, explaining the frequently observed narrow distance distribution for doubly V1-labeled proteins. We used the present data to derive guidelines that may allow distance interpretation of other V1-labeled proteins for higher-precision structural modeling.

X

minimize this effect, computational approaches have been developed to simulate the rotamer space of spin-labeled side chains, especially R1, at particular sites on protein structural models.9−13 Unlike R1, the similar cysteine-derived V1 side chain (Figure 1A), recently introduced by Hubbell and coworkers, possesses no dihedral angle χ5, and its χ4 is restricted by an internal interaction between the 3-nitrogen in the imidazoline ring (3N) and Sγ in the side chain (Figure 1B).14,15 Like all cysteine-derived side chains, V1 contains at least three rotatable bonds with dihedral angles χ1, χ2, and χ3. It has been shown by crystallography that in the case of R1, χ1, χ2, and in some cases χ3 exist in somewhat discrete states as a result of interactions with the proximate peptide backbone.16 The present work aimed to identify, using X-ray crystallography of V1-labeled T4 lysozyme (T4L) constructs, the dihedral angle combinations {χ1, χ2, χ3, χ4} of the V1 rotamers that occur on α-helices. The resultant experimental rotamer library potentially allows for modeling of these rotamers onto α-helical surface sites on other proteins to greatly simplify the EPR structural model building process by limiting the likely rotamer space. As previously determined by crystallography, the {χ1, χ2} set in the R1 side chain on α-helices exists predominantly in three states, approximately {−60°, −60°}, {180°, −60°}, and {180°, 60°} (or {m, m}, {t, m}, and {t, p}, respectively), presumably as a result of interactions between the disulfide

-ray crystallography can yield highly detailed structural models of single proteins and protein complexes. However, proteins in crystals are restricted in low-energy conformations that result in the best packing, and in some cases, a wealth of information about the ensemble of states in solution is lost. To retrieve this information, electron paramagnetic resonance (EPR) spectroscopy using site-directed spin-labeled proteins is sometimes applied as a complementary approach to probe different protein conformations when a model, such as a crystal structure, is available.1−4 These EPR approaches are attractive partly because of their low sample requirements, their lack of molecular weight restrictions, and their compatibility with complex environments such as biological membranes.5 Double electron−electron resonance (DEER) spectroscopy is a pulsed EPR method that is widely applied to study protein structural ensembles by measuring individual distances of ∼1.5−8 nm between two or more paramagnetic side chains. Typically, these side chains are nitroxides generated by the reaction between a thiol-reactive spin-labeling reagent and a solvent-accessible cysteine residue introduced via site-directed mutagenesis.1,6 Structural models can be built by trilateration of the distances measured in several doubly spin-labeled proteins.7 Most used and studied of these modified side chains is the cysteine-derived R1 modification (Figure 1A). However, the flexible nature of the R1 side chain, especially about the dihedral angles χ4 and χ5, often interferes greatly with the translation from EPR-based distances between spin densities to protein backbone structural ensembles.8 In an effort to © 2016 American Chemical Society

Received: June 14, 2016 Revised: August 8, 2016 Published: August 17, 2016 5256

DOI: 10.1021/acs.biochem.6b00608 Biochemistry 2016, 55, 5256−5263

Article

Biochemistry

R1 side chain or the spin-labeling reagent bis(1-oxyl-2,2,5,5tetramethyl-3-imidazolin-4-yl) disulfide (IDSL) (Figure 1A) to generate the V1 side chain. After the 1 h incubation, the excess spin-labeling reagent was removed by buffer exchange. MTSSL was purchased from Toronto Research Chemicals (www.trccanada.com). IDSL was purchased from Enzo Life Sciences (www.enzolifesciences.com). DEER Spectroscopy Measurements and Interspin Distance Interpretation of R1- and V1-Labeled T4L Constructs. The DEER spectroscopy measurements were conducted using 10 μL of ∼100 μM spin-labeled T4L with 20% glycerol as a cryoprotectant on a Bruker Elexsys 580 spectrometer with a Super Q-FTu bridge at 80 K. A 32 ns π pump pulse was applied to the low-field peak of the nitroxide field-swept spectrum, and the π/2 (16 ns) and π (32 ns) observer pulses were positioned 50 MHz (17.8 G) upfield, which corresponds to the nitroxide center line. The DEER traces reached near-maximal signal-to-noise ratios within 1 h of averaging scans. The distance distributions were obtained from the raw DEER traces using the MATLAB program DEER Analysis 2013 (developed by the Jeschke group, Eidgenössische Technische Hochschule Zürich), which can be downloaded from http://www.epr.ethz.ch/software.html. The LabVIEW (National Instruments) program LongDistances (developed by Christian Altenbach, University of California, Los Angeles) gave virtually identical results. It can be downloaded from https://sites.google.com/site/altenbach/labview-programs/ epr-programs. With either program, the final distance distributions were obtained from the background-corrected DEER time traces by Tikhonov regularization. In each case, the regularization parameter at the elbow of the L curve was chosen as the best fit of the DEER data. Crystallography of V1-Labeled T4L Constructs. All of the crystals were grown using the hanging-drop vapor diffusion method over a 500 μL well solution containing 2.0−2.4 M NaH2PO4/K2HPO4 (pH 6.6−7.4), 150 mM NaCl, 100 mM 1,6-hexanediol, and 3% 2-propanol. Crystals grew in 1−4 days in a 5 μL drop composed of 2.5 μL of a solution containing ∼5 mg/mL T4L, 50 mM MOPS (pH 6.8), and 25 mM NaCl and 2.5 μL of the well solution suspended over the well. Diffraction data for the ∼100−200 μm crystals of the T4L constructs I9V1, A73V1, V131V1, T151V1 (PDB access codes 5JGN, 5JGV, 5JGX, and 5JGZ, respectively) were collected at the Structural Genomics Consortium in Toronto, ON, using a Rigaku FR-E SuperBright X-ray generation system. Diffraction data for the ∼25 μm T4L K43V1 and R119V1 crystals as well as the monthold I9V1/V131V1 crystals (PDB access codes 5JGR, 5JGU, and 5KGR, respectively) were collected at the Advanced Photon Source at beamline 23-ID. Data processing was done with applications provided by SBGrid.20 Diffraction data were processed using imosflm (CCP4).21 The indexed and integrated reflections were scaled and merged using Aimless (CCP4).22 Molecular replacement was conducted with Phaser (Phenix)23 using a T4L model (PDB access code 1C6T). Structure and density maps were refined and validated using Phenix.refine (Phenix).24 See Table S1 in the Supporting Information for data collection and refinement statistics. Mass Spectrometry of T4L T109V1/V131V1. In order to identify the degradation products of V1-labeled proteins, a mass spectrum of 10 μL of 100 μM T4L T109V1/V131V1 in 50 mM MOPS (pH 6.8) and 25 mM NaCl incubated at room temperature for 4 days was obtained. The electrospray ionization mass spectrum was obtained from the AIMS Mass

Figure 1. (A) Generation of the R1- (black) and V1-labeled (green) side chains from a cysteine residue. The dihedral angles of bonds with free rotation and the paramagnetic nitroxide group are both shown in red. (B) (left) Three superimposed cysteine disulfide rotamers {χ1, χ2} = {t, m}, {t, p}, and {m, m}. The interactions between Sδ and peptide backbone atoms are believed to restrict {χ1, χ2} to these states.16 (right) The 3N−Sγ interaction that restricts χ4 in the V1 side chain.14 (C) Six {χ1, χ2, χ3, χ4} V1 side chain rotamers based on the internal side chain interactions (see B). The number of checkmarks denotes the number of V1-labeled T4L crystal structures solved with a particular rotamer.

moiety in the side chain and the peptide backbone, especially Hα in the case of {m, m} (∼2.8 Å) and {t, p} (∼3.2 Å) and the carbonyl O in the case of {t, m} (∼3.5 Å) (Figure 1B).17−19 The dihedral angle of the disulfide bond, χ3, usually exists as approximately −90° or 90° (or m or p, respectively). On the basis of the eight V1-labeled sites visualized by crystallography, the trends of {χ1, χ2, χ3} of the R1 side chain appear to be the same for the V1 side chain. The electron density of the imidazoline ring in all eight V1 side chains is resolved at least in part with close 3N−Sγ proximities (∼3.2 Å), such that χ4 is approximately 0° (or c). This affords six potential {χ1, χ2, χ3, χ4} combinations of V1 on α-helices (Figure 1C).

■

EXPERIMENTAL PROCEDURES Expression, Purification, and R1 and V1 Labeling of T4L Constructs. T4L constructs were expressed, purified, and spin-labeled as previously described.14 Briefly, the purified cysteine mutants were subjected for 1 h at room temperature to a 10-fold molar excess of either the spin-labeling reagent (1oxyl-2,2,5,5-tetramethyl-Δ3-pyrroline-3-methyl)methanethiosulfonate (MTSSL) (Figure 1A) to generate the 5257

DOI: 10.1021/acs.biochem.6b00608 Biochemistry 2016, 55, 5256−5263

Article

Biochemistry

ones where the side chains made minimal contacts with symmetry mates in the standard T4L packing system, and the solution-phase DEER spectroscopy distance measurements of the V1-labeled constructs were found to be within 1 Å of the observed distances between the midpoints of the N−O bonds in the V1 side chains in the corresponding crystal structures. Rotamer Model of the V1 Side Chain on a Generic αHelix Based on Crystal Structures. The crystal structures of the singly V1-labeled T4L constructs I9V1, K43V1, A73V1, R119V1, V131V1, and T151V1 (PDB access codes 5JGN, 5JGR, 5JGV, 5JGU, 5JGX, and 5JGZ, respectively) were solved (Figure 3) in addition to the previously solved structure of the

Spectrometry Laboratory at the University of Toronto using an Agilent 6538 Q-TOF mass spectrometer.

■

RESULTS Effect of Internal Interactions of the V1 Side Chain on DEER Spectroscopy Measurements. To further illustrate the enhanced rigidity and narrower rotamer space of the V1 side chain over the R1 side chain previously shown by Hubbell and co-workers,14 DEER spectroscopy was used to measure distances between either R1 or V1 side chains on the T4L constructs I9C/V131C and A73C/T151C, generated from the cysteine-free (C54T/C97A) background of T4L, with the engineered cysteine residues on various solvent-exposed αhelical sites (Figure 2). In each case, the R1 side chain

Figure 2. (left) Background-corrected DEER time traces and their fits and (right) derived distance distributions of T4L mutants (top) I9C/ V131C and (bottom) A73C/T151C with either R1 (black) or V1 (green) modifications. To the right of plots are superimposed crystal structures of the T4L constructs I9V1 and V131V1 or A73V1 and T151V1 (PDB access codes 5JGN, 5JGX, 5JGV, and 5JGZ, respectively). V1-labeled side chains (modeled as green sticks onto the structure shown in cartoon representation) produce narrower distance distributions (∼3 Å fwhm for both pairs) than R1-labeled side chains at the same sites (∼9 and ∼4 Å fwhm for the I9R1/V131R1 and A73R1/T151R1 pairs, respectively). The DEER distance distribution (P(r)) maxima of the V1-labeled pairs (shown with dashed red lines) have values very similar to the obtained crystallographic distances between nitroxides (shown in red). See Table 1 for the dihedral angles χ1−χ4 and the distances of the internal interactions 3N−Sγ, Hα−Sδ, and O−Sδ.

Figure 3. 2Fo − Fc electron density maps contoured at 1.0σ of V1modified sites on T4L as determined by crystallography. Cysteine and thiocysteine side chains were produced by the degradation of the V1 side chain in the crystal over time and were present in some structures in significant quantities (see Figure 6). See Table 1 for the dihedral angles χ1−χ4 and the distances of the internal interactions 3N−Sγ, Hα−Sδ, and O−Sδ. See Table S1 for data collection and refinement statistics.

produced multiple distance distribution maxima, which generally could be the result of multiple protein conformations, multiple side chain rotamers, or both. In contrast, the V1 side chains produced predominantly single, sharp distance distribution maxima, which is consistent with the well-characterized intradomain rigidity of T4L between the chosen cysteine pairs.25,26 Because of crystal contacts, the orientations of the side chains on the surface of a protein observed in a crystal do not always reflect the orientations of the same side chains in solution. Nevertheless, the sites chosen for crystallization were

T4L construct K65V1/R76V1 (PDB access code 3K2R14). By analysis of these structures, ones with replicate V1 side chain rotamers were found to have very similar dihedral angles, resulting in small deviations in the locus of the spin density (≤1 Å). The dihedral angles χ1−χ4 and the distances of the internal interactions 3N−Sγ, Hα−Sδ, and O−Sδ are displayed in Table 1. χ1−χ4 were found to be stable within ±2° and 3N−Sγ, Hα−Sδ, 5258

DOI: 10.1021/acs.biochem.6b00608 Biochemistry 2016, 55, 5256−5263

Article

Biochemistry Table 1. V1 Side Chain Dihedral Angles and Internal Distances Observed by Crystallography

a

site

PDB ID

rotamer

χ1

χ2

χ3

χ4

73 119 151 43 65 131 76 9

5JGVa 5JGUa 5JGZa 5JGRa 3K2Rb 5JGXa 3K2Rb 5JGNb

{m,m,m,c} {m,m,m,c} {m,m,m,c} {t,p,p,c} {t,p,p,c} {t,p,p,c} {t,m,m,c} {m,m,p,c}

−73° −70° −74° 191° 180° 182° 180° −74°

−51° −52° −62° 100° 86° 80° −79° −81°

−96° −87° −92° 76° 77° 75° −79° 108°

5° ∼−7° ∼17° −2° −22° 3° −7° 2°

3N−Sγ

Hα−Sδ

O−Sδ

3.17 3.19 3.31 3.23 3.19 3.21 3.23 3.24

2.69 2.66 2.92 3.63 3.36 3.17 4.30 3.21

4.78 4.79 4.80 5.13 4.72 4.48 3.35 4.92

Å Å Å Å Å Å Å Å

Å Å Å Å Å Å Å Å

Å Å Å Å Å Å Å Å

See Figure 3 for the electron density of side chains and Table S1 for crystallographic statistics. bStructure from ref 14.

and O−Sδ within ±0.02 Å between each crystallographic refinement cycle. For visualization, we modeled the V1 side chains from the crystal structures of the V1-labeled T4L constructs onto a generic, α-helical polyalanine peptide (Figure 4A) to illustrate where on a standard α-helical site the spin density of the side chain is most likely confined relative to the side chain’s Cα.

when modeling side chains directly onto sites on a V1-labeled protein of interest, it is more likely that the side chain will exist as either the {m, m, m, c} or {t, p, p, c} rotamer, more specifically, whichever rotamer is least hindered by its environment. Our experience with the interpretation of DEER distance distributions measured from other doubly V1labeled proteins is that these two rotamers function well as an initial guess and can be corrected to another rotamer through a trilateration process as more distance constraints are acquired. Effect of Tertiary Contacts on V1 Side Chain Rotamers. Side chains with the rotamers {m, m, p, c} and {t, m, m, c} were each observed once by crystallography. The {m, m, p, c} rotamer arose only when the two previously mentioned rotamers, {m, m, m, c} and {t, p, p, c}, were restricted by tertiary contacts (Figure 5A). Similarly, a possible explanation for the occurrence of the {t, m, m, c} rotamer at site 76 in the previously published T4L crystal structure14 is that the rotamers {m, m, m, c} and {t, p, p, c} of the relatively hydrophobic V1 side chain would cause the side chain to reside in a highly polar cavity in the crystal lattice (