Backbone Nitrogen Substitution Probes the Role of Glycine Residues at GxxxG Interfaces in Transmembrane Helices

Abstract

Association of transmembrane helices is central to the structure and function of membrane proteins. A common feature of these interfaces is the GxxxG motif, which allows for close contact between the protein backbones, resulting in the formation of a network of weak hydrogen bonds between C^α–H donors and carbonyl acceptors on opposing helices (i.e., C^α–H···O=C hydrogen bonds). Engineering these interfaces holds promise for the control of membrane protein activity, but few strategies exist for the substitution of the critical glycine residues, which lack mutable side chains. Herein, we replace glycine residues at the dimerization interface of single-pass transmembrane proteins (i.e., glycophorin A and BNIP3) with aza-glycine, which substitutes the alpha carbon with nitrogen. We hypothesized that aza-glycine incorporation would enhance intermolecular association by introducing a stronger hydrogen-bond donor and increasing solvophobic interactions within the lipid bilayer. In contrast, FRET-based measurements of helix association showed that aza-glycine substitution was insufficient to increase dimerization affinity, which we attribute to the decreased helical propensity of aza-glycine. Our results demonstrate a synthetic strategy to probe the determinants of membrane-protein folding and inform the design of transmembrane peptidomimetics.

Graphical Abstract

INTRODUCTION

Transmembrane (TM) proteins play central roles in diverse cellular functions, including signal transduction, molecular transport, and cell–cell communication.¹ The structural assembly and activity of these proteins are frequently mediated by the association of TM helices, which must occur with high specificity within complex lipid bilayers. TM helix association can occur either intramolecularly, wherein multiple helices within a single polypeptide chain pack together (e.g., GPCRs),² and/or intermolecularly, wherein multiple polypeptide chains form oligomeric complexes (e.g., glycophorin A).³ High-affinity interactions can result in relatively static assemblies like the TCR–CD3 complex,⁴ whereas other systems (e.g., receptor tyrosine kinases) require dynamic association/dissociation for function.^5,6 Understanding the determinants of TM helix assembly is therefore critical for controlling the activity of TM protein function, yet doing so poses challenges beyond those inherent to the engineering of soluble proteins. For example, whereas hydrophobic collapse plays a dominant role in the folding of soluble proteins, it is plays a lesser role in the folding of membrane proteins. Moreover, the ubiquity of TM helices poses a challenge for ensuring specific interactions between diverse molecules within crowded biological membranes.⁷ Some of these challenges can be overcome by focusing on the folding and interactions of unique ectodomains; however, understanding and controlling the assembly of membrane proteins through their less diverse TM helices offers a more broadly applicable strategy to modulate their activity.⁵ So far, this goal has been primarily advanced through de novo protein design of synthetic peptides that can bind effectively to TM segments, whose sequences are identified through computational and experimental screening.^8,9

One common sequence feature that imparts binding specificity to many TM helices is the GxxxG sequence motif at the dimerization interface.^10,11 As one example, this motif is found within the key seven residues that mediate the sequence-specific homodimerization of the glycophorin A (GpA) transmembrane domain (i.e., LIXXGVXXGVXXT).¹² Statistical analysis of TM sequences has shown that the GxxxG motif is highly over-represented in natural TM proteins.¹³ The mutation of glycine in this motif significantly disrupts the dimerization of TM proteins such as GpA and Bcl-2 interacting protein 3 (BNIP3),^14,15 underscoring its critical role. These glycine residues occur frequently at helix crossings and generally orient toward the helix–helix interface, thereby enabling interaction through multiple mechanisms.^3,16–18 The lack of a side chain allows for close backbone contact, which helps to maximize desolvation and apolar packing through van der Waals interactions, all while minimizing side-chain entropy loss upon association. The exposed backbone also helps drive association through the formation of a relatively polar interface that engages in solvophobic interactions within the nonpolar interior of the lipid bilayer. Together, these forces help TM helices to discriminate one another based on the chemistry of their exposed surfaces.

The polar interface within GxxxG motifs is stabilized in part by a network of relatively weak hydrogen bonds between C^α–H carbon donors and carbonyl acceptors on opposing helices (C^α–H···O=C hydrogen bonds, Figure 1A).¹⁹ These weak hydrogen bonding interactions are hypothesized to be major contributors to the free energy of association of GxxxG-mediated dimers.^11,20,21 Even though carbon is a weak hydrogen bond donor, the adjacent electron-withdrawing amide groups activate the C^α–H for donation. Ab initio calculations on model compounds predict that the strength of a C^α–H···O=C hydrogen bond is roughly half that of a N−H···O=C hydrogen bond,^22,23 suggesting a potentially significant contribution. Similarly, quantum calculations based on TM protein crystal structures estimated the average stabilization energy of a C^α−H···O=C interaction to be approx. 1.5 kcal/mol.²⁴ Hydrogen bonds from carbon donors might be particularly important within the lipid bilayer, which features reduced hydration and dielectric relative to bulk water, which would otherwise screen the effects of hydrogen bonding.^3,25 Therefore, if multiple C^α−H···O=C hydrogen bonds occur at the dimer interface of a TM domain, the cumulative effect of these weak interactions could impart stability and specificity to TM helix interactions.^3,25

Figure 1.

A) Molecular structure of the transmembrane domain of glycophorin A (1af0)³ with enlarged view of the dimerization interface. The glycine residue at position 83 is mutated to aza-glycine herein. B) Chemical structures of glycine and aza-glycine residues

Although computational and statistical studies suggest that C^α−H···O=C hydrogen bonds are critical to the stability and specificity in TM helix–helix interactions, relatively few studies have directly probed these interactions experimentally. For example, isotope-edited IR spectroscopy estimated the energy of one specific interaction (i.e., the Gly79 C^α–H···O=C Ile76 hydrogen bond in GpA) to be −0.88 kcal/mol,²⁶ which is in line with computational prediction. In contrast, mutation of Thr24 in bacteriorhodopsin (bR), whose side chain γ-O accepts an interhelical C^α−H···O^γ hydrogen bond from the C^α of Ala51, caused limited changes in overall stability in an SDS unfolding assay.²⁷ The discrepancy between these two studies was subsequently rationalized based on the different arrangement of the interacting groups, specifically the position of the acceptor with respect to the C^α−H, suggesting strong context dependence to the strength of these interactions,²⁸ which is consistent with the behavior of other hydrogen bonds.

Motivated by predicted importance of these interactions and the relative dearth of experimental methods to probe them, we sought complementary strategies to modulate these interactions as a handle for controlling membrane protein folding and function. Given the centrality of glycine at these interfaces, we were particularly interested in controlling the chemistry of the glycine hydrogen-bond donor; however, the lack of a side chain limits opportunities for optimizing interhelical interactions. To directly control the hydrogen bonding potential of the glycine α-position, we envisioned replacing the α-carbon with nitrogen to form aza-glycine (Figure 1B), which can strengthen hydrogen bonding through enhanced H^α acidity. This strategy has been successfully applied to the collagen triple helix, which similarly requires glycine residues to enable close backbone contact between its constituent helices;²⁹ substitution of glycine with aza-glycine led to a significant increase in its thermal stability as shown through circular dichroism (CD) spectroscopy.^30,31 Further computational, spectroscopic, and crystallographic analyses attributed this stabilization in part to the enhanced hydrogen bonding potential of aza-glycine relative to glycine and a unique conformational preorganization of the collagen backbone.^32,33 Crystal structures of collagen mimetic peptides containing aza-glycine residues revealed new cross-strand hydrogen bonds within the triple helix interior that are not present in native glycine-containing collagen; specifically, the N^α–H group of aza-glycine formed hydrogen bonds with up to two carbonyl groups on adjacent peptide chains.

Inspired by these studies, we hypothesized that the incorporation of the aza-glycine residue at the dimerization interface of TM proteins would increase the association of transmembrane helices through enhanced hydrogen bonding potential, though structural differences between glycine and aza-glycine (e.g., planarity of the N^α) might reorient the hydrogen-bond donor away from the ideal angle of approach. We also hypothesized that the enhanced hydrogen-bonding capacity of aza-glycine would increase the polarity of the dimer interface, which might further increase TM helix association through solvophobic interactions; the exposure of more polar groups to the hydrophobic interior of the lipid bilayer should create a thermodynamic driving force for dimerization by sequestering the polar surfaces from the hydrophobic local environment. Indeed, ClogP values, which estimate hydrophobicity, were lower for aza-glycine model compounds than for the corresponding glycine models (Table 1), supporting our prediction.

Table 1.

Transmembrane peptides incorporating aza-glycine at GxxxG dimer interfaces

Peptide	Sequence*	Kd (× 10−4 M)	MRE222 (× 103 deg cm2 dmol−1 res−1)	ClogPo/w
WT-GpA	SEPEITLIIFGVMAGVIGTILLISYGIRRLIKK	1.4 ± 0.7	−26.1 ± 1.5	−0.99
aG-GpA	SEPEITLIIFGVMAGVIGTILLISYGIRRLIKK	2 ± 1	−16.3 ± 2.9	−1.07
WT-BNIP3	SAEFLKVFLPSLLLSHLLAIGLGIYIGRRLTT	0.27 ± 0.05	−41.1 ± 5.1
aG-BNIP3	SAEFLKVFLPSLLLSHLLAIGLGIYIGRRLTT	5 ± 1	−28.2 ± 4.3

Peptide N termini are modified with either pyrene (Pyr), 7-dimethylaminocoumarin (Cou), or acetyl (Ac); peptides are C-terminal amidated. Aza-glycine residues are shown in bold. Mean residue ellipticity (MRE₂₂₂) values are calculated for 20 μM solutions at room temperature. Octanol/water partition coefficient (ClogP_o/w) values are calculated for N-acetylglycinamide and N-acetylazaglycinamide.

Despite the potential for aza-glycine to increase TM helix association, stereoelectronic effects imparted by backbone nitrogen substitution might undermine TM assembly by reducing helical propensity. Specifically, backbone nitrogen substitution creates lone pair–lone pair repulsion between the alpha and amide nitrogens, thereby biasing the φ dihedral angle (C′_i–1–N_i–C^α_i–C′_i) toward orthogonality, while also creating a urea-like moiety with the carboxyl group, thereby biasing the ψ dihedral angle (N_i–N^α–C′_i–N_i+1) toward planarity.^34–37 As a result, theoretical studies of aza-amino acid model compounds show conformational minima around (φ = ±90 ± 30°, ψ = 0 ± 30°) or (φ = ±90 ± 30°, ψ = 180 ± 30°),^34–42 which is significantly restricted relative to the conformational landscape of glycine. Importantly, these conformational propensities bias aza-glycine away from α-helical structure (φ = −60°, ψ = −45°),⁴³ which could undermine structural stabilization imparted by enhanced H^α acidity and high solvophobicity.

In this study, we evaluate the prospects for using aza-glycine substitution to probe and control interactions at GxxxG interfaces in TM helices, which we expect will provide further experimental characterization of their role in membrane protein folding and suggest strategies for the design of TM peptidomimetics.

MATERIALS AND METHODS

General experimental

All commercially available reagents were utilized without further purification. Protected amino acids were purchased from Chem-Impex International (Wood Dale, IL) and Ambeed, Inc (Arlington Heights, IL). Rink Amide ProTide Resin (LL) was obtained from CEM Corporation (Matthews, NC). Carbonyl ditriazole was obtained from Ambeed, Inc (Arlington Heights, IL). 1-pyreneacetic acid and 7-dimethylaminocoumarin-4-acetic acid were purchased from Sigma-Aldrich and Ambeed respectively. All other chemicals used were purchased from Fisher Scientific and Sigma-Aldrich. A rotary evaporator at water-aspirator pressure (<20 torr) and a water bath of 40 °C was used to remove solvents and other volatile materials.

Instrumentation

Peptide synthesis was performed on Liberty Blue 2.0 Microwave Peptide Synthesizer from CEM Corporation (Matthews, NC). The synthesized peptides were purified using an UltiMate 3000 Basic HPLC Systems from Thermo Fisher Scientific (Waltham, MA). Mass spectrometry was performed on an Agilent 6125 LC/MS from Agilent Technologies (Santa Clara, CA). FRET measurements were performed on BioTek Synergy H1 plate reader from Agilent Technologies (Cedar Creek, TX). Circular dichroism spectra were acquired on a J-1500 Circular Dichroism Spectrophotometer from JASCO North America (Easton, MD). The fluorophore labeled peptide concentrations were determined from absorbance measured using a NanoDrop OneC Microvolume UV-Vis Spectrophotometer from Thermo Fisher Scientific (Waltham, MA).

Peptide synthesis

All peptides were synthesized on a 0.1 mmol scale using standard Fmoc-based methods with DMF as the solvent on Rink Amide ProTide Resin (LL). Generally, Fmoc deprotection was achieved using 20% 4-methylpiperidine in DMF at 75 °C for 15 s then 90 °C for 50 s. 1 M N,Nʹ-diisopropylcarbodiimide (10 eq.) and 1 M ethyl cyanohydroxyiminoacetate (5 eq.) were used as coupling reagents. Coupling proceeded in 2 minutes at 90 °C. A fresh cleavage cocktail solution of either TFA:water:phenol:thioanisole:EDT (82.5:5:5:5:2.5, 10 mL) for GpA and TFA:phenol:TIPS:water (90:5:2.5:2.5, v/v/v/v, 10 mL) for BNIP3 were used for global deprotection and cleavage of the peptide. The cleavage reaction was stirred for 3 h, followed by the precipitation of peptides with cold diethyl ether. Isolated peptides were dissolved in 1,1,1,3,3,3-hexafluoroisopropanol (HFIP)/water (50:50, v/v) and were purified by reverse-phase HPLC using a preparative column (Vydac 214TP 10 μm C4 column from HiChrom [Wilmington, DE]) at a flow rate of 10 mL/min using a linear gradient of 5–95% v/v B over 30 min [A: 0.1% v/v TFA in H₂O and B: 0.1% v/v TFA in isopropanol/acetonitrile/water (60:30:9.9)]. The HPLC purified peptides were lyophilized to obtain white powders.

Incorporation of aza-glycine.⁴⁴

9-Fluorenylmethyl carbazate was synthesized according to Lubell and co-workers.⁴⁵ 9-Fluorenylmethyl carbazate (10 eq.) and carbonyl ditriazole (CDT, 10 eq.) were dissolved in 1–2 mL of DMF and stirred for 5–10 min. After activation, the solution was added to the resin-bound peptide and stirred for 24 hours at room temperature. The coupling solution was drained, and the resin was washed with DMF three times. The Fmoc group was removed using 5 mL of 20% 4-methylpiperidine (v/v) in DMF. The deprotection solution was stirred for 30 minutes, followed by washing with DMF 3 times.

Coupling to aza-glycine.

Immediately subsequent to aza-glycine incorporation, the next Fmoc-amino acid (5 eq.) was dissolved in 5 mL of DMF with HOAt (5 eq.), HATU (5 eq.), and DIPEA (10 eq.). This solution was stirred for 20 minutes and then transferred to the resin. The mixture was stirred for 1.5 hours at room temperature and drained. Another identical batch of coupling solution in DMF was added to the resin and stirred for another 1.5 hours. After coupling, the resin was washed with DMF.

Incorporation of fluorophores.

1-pyreneacetic acid (4 eq.) and 7-dimethylaminocoumarin-4-acetic acid (4 eq.) were used to incorporate pyrene as the donor fluorophore and 7-dimethylaminocoumarin as the acceptor fluorophore respectively. They were attached to the N terminus of the transmembrane peptides via a condensation reaction, with the formation of an amide functional group, using N,Nʹ-diisopropylcarbodiimide (4 eq.) and ethyl cyanohydroxyiminoacetate (4 eq.) as coupling reagents at room temperature overnight in a dark environment. The N terminus of unlabeled peptides was modified with the acetyl group using acetic anhydride (4 eq.) and DIPEA (8 eq.) at room temperature for 2 hours.

ClogP_o/w calculations

The calculated logarithm of n-octanol/water partition coefficient (ClogP_o/w) values were obtained for N-Acetylglycinamide and N-Acetylazaglycinamide using the web tool SwissADME (SIB Swiss Institute of Bioinformatics).⁴⁶

FRET experiments

Reconstitution of peptides into detergent.

The fluorophore labeled peptide was dissolved into trifluoroethanol and water was subsequently added to reach 50%, v/v. The dissolved peptide was then lyophilized and re-suspended in buffered SDS detergent. The buffer was 20 mM sodium phosphate with 200 mM sodium chloride at a pH of 7.0. Equivalent donor-only and acceptor-only samples were mixed, each with a protein concentration of 2.5 μM, to achieve a final total peptide concentration of 5 μM. The concentrations of the fluorophore labeled peptides were obtained spectrophotometrically in trifluoroethanol using ε₃₄₁ = 32000 mol⁻¹ cm⁻¹ for pyrene and ε₄₀₀ = 16500 mol⁻¹ cm⁻¹ for 7-dimethylaminocoumarin.⁴⁷ The concentration of buffered SDS detergent ranged from 5 mM to 200 mM to cover the mole fraction concentrations (moles of protein/moles of detergent) from 1×10⁻³ to 2.5×10⁻⁵. Stock solutions of the mixed donor–acceptor peptides at 5 mM and 200 mM SDS concentrations were prepared and mixed in different ratios to vary the concentration of buffered SDS detergent while keeping the total peptide concentration constant at 5 μM.

Reconstitution of peptides into lipid bilayers.

50 pmol each of donor-only and acceptor-only peptides dissolved in trifluoroethanol were mixed with 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) lipids dissolved in chloroform (25mg/mL) to achieve total peptide:lipid ratios from 1:100 to 1: 10,000. Samples were evaporated to dryness. The resulting lipid films were then redissolved in 10 mM sodium phosphate buffer with 500 mM sodium chloride at a pH of 7.0, vortexed, and subjected to repeated freeze–thaw cycles to hydrate the lipid mixtures. The hydrated liposomes were then extruded through a polycarbonate membrane filter with a size cutoff of 100 nm (Avanti Polar Lipids) to produce large unilamellar vesicles (LUVs).

Recording and fitting of FRET data.

Fluorescence emission spectra were obtained using a BioTek Synergy H1 plate reader (Agilent Technologies). Samples were excited at 345 nm, and emission spectra were recorded from 375 nm to 600 nm at 1-nm intervals. The dissociation constants for dimerization of each peptide were calculated as described previously.⁴⁸ Briefly, dissociation constants were obtained by globally fitting the FRET intensity at the acceptor emission maxima (454 nm) normalized to the respective isosbestic points (432 nm for GpA and 438 nm for BNIP3) (NF₄₅₄) as a function of mole fractions (moles of protein/moles of detergent) (χ_T) using the following equation:

$$N{F}_{454}=N{F}_D+\left(N{F}_M-N{F}_D\right)\frac{-K_D+\sqrt{{K_D}^2+8{\chi }_T{K}_D}}{4{\chi }_T}$$

where NF_D is the normalized fluorescence intensity of fully dimeric sample (i.e, at infinite concentration), NF_M is the normalized fluorescence intensity of fully monomeric sample (i.e, at infinite dilution) and K_D is the desired equilibrium dissociation constant. NF_D and NF_M were fit as global variables shared between WT and aG-modified peptides of respective proteins and K_D was fit individually for each peptide.

Circular dichroism (CD) spectroscopy

Unlabeled peptides were dissolved to a final concentration of 20 μM in buffered SDS detergent (5 mM SDS in 20 mM sodium phosphate buffer with 200 mM sodium chloride at a pH of 7.0). CD spectra of each peptide were collected from 200 to 260 nm (1-nm bandwidth, 2-s averaging) using a 1 mm path-length cuvette.

RESULTS AND DISCUSSION

To examine the effects of aza-glycine substitution on the association of TM helices, we first identified TM helices with canonical GxxxG motifs from two different single-pass TM proteins (i.e., glycophorin A [GpA] and Bcl-2 interacting protein 3 [BNIP3]). Based on available structural data, we identified Gly83 of GpA and Gly180 of BNIP3 as key hydrogen-bond donors at the dimerization interface of these proteins, and we designed peptides in which those glycine residues are replaced by aza-glycine (Table 1).^3,15 To measure the binding affinity of these TM complexes, we used a previously developed Förster resonance energy transfer (FRET)-based assay (Figure 2A).^47,48 Briefly, TM peptides were labeled with either pyrene (Pyr) or 7-dimethylaminocoumarin (Cou) as donor and acceptor fluorophores. Labeled peptides were mixed in equimolar ratios in varying concentrations of buffered detergent, and fluorescence spectroscopy was used to measure the monomer–dimer equilibrium. As detergent concentration increases, emission of the acceptor decreases due to dilution of the donor–acceptor pair by the detergent. Fitting of the spectral data then provides dissociation constants for dimerization of each TM peptide.

Figure 2.

A) Schematic of the FRET experiment. B) Emission spectra of WT and aG-GpA obtained upon excitation at 345 nm. C) Emission spectra of WT and aG-BNIP3 obtained upon excitation at 345 nm. The donor emission peak (392 nm) is denoted by a green triangle, isosbestic point (432 nm for GpA and 438 nm for BNIP3) by a blue triangle and acceptor emission peak (454 nm) by a red triangle. D) Fitting of the fluorescence data as a function of detergent concentration.

We began by collecting excitation (Figure S1) and emission spectra (Figure 2B/C) for each peptide FRET pair after reconstitution in detergent and saw qualitative agreement upon substitution with aza-glycine, indicating that aza-glycine did not disrupt the overall fold or assembly of the dimer complexes. We then recollected the excitation and emission spectra at a variety of detergent concentrations. As expected, acceptor emission decreased with increasing detergent, whereas donor emission increased with increasing detergent concentration for both GpA and BNIP3 (Figure 2B/C), which is consistent with reduced FRET efficiency of the peptide assembly upon dilution with detergent. Upon substituting the critical glycine residues with aza-glycine, both aG-GpA and aG-BNIP3 exhibited lower FRET efficiency than the corresponding glycine-containing peptides (Figure 2B/C), implying that aza-glycine causes a shift in equilibrium away from the dimer toward the dissociated monomers. Fitting of the fluorescence data to a two-state model revealed a modest ~1.4-fold increase in K_d upon aza-glycine substitution for GpA and a more dramatic ~18-fold increase in the K_d of BNIP3, which was consistent with the greater change in acceptor emission intensity upon aza-glycine substitution for BNIP3 compared to GpA. These data contradicted our expectations and instead suggest that aza-glycine destabilizes TM helix dimerization. Because these peptides do not sample the fully monomeric and dimeric states under the experimentally accessible conditions, wherein SDS remains soluble, our reported K_d values should be interpreted as only semiquantitative and used only to compare the relative association of variants with and without backbone modification.

Though SDS micelles provide a popular environment in which to study transmembrane peptide association,^12,15,47 they differ significantly from the structure of biological lipid bilayers. For example, the relatively short acyl chain and high particle curvature lead to increased hydration of the hydrophobic interior of the micelle relative to a typical lipid bilayer. These effects could disrupt the lipophobic and hydrogen-bonding interactions that are hypothetically enhanced by aza-glycine substitution, which could explain the inability of aza-glycine incorporation to stabilize dimerization. To address this possibility and overcome the limitations of SDS micelles as a membrane mimic, we repeated our FRET measurements in vesicles formed from bilayers of 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC). The fluorophore labeled peptides were mixed in equimolar ratios and reconstituted into large unilamellar vesicles (LUVs) made of POPC lipids at three different peptide:lipid ratios to sample the dimerization equilibrium. Emission spectra of the reconstituted peptides were then recorded as before (Figure 3). Consistent with the results in SDS micelles (Figure 2), the aza-glycine containing peptides exhibited reduced FRET efficiency relative to their glycine counterparts for both GpA and BNIP3 at all three peptide:lipid ratios studied. We therefore conclude that the effects of aza-glycine incorporation on transmembrane peptide association are qualitatively independent of model membrane.

Figure 3.

Emission spectra collected upon excitation at 345 nm of labeled WT and aG-GpA (left) and WT and aG-BNIP3 (right) in POPC LUVs at peptide:lipid ratios of A) 1:100, B) 1:1000, and C) 1:10,000.

As a further test of the effects of aza-glycine on TM helix association, we designed a competition experiment in which unlabeled TM peptides were titrated into a FRET pair to disrupt the complex (Figure 4A). In this paradigm, the unlabeled peptide acts as a competitive inhibitor of dimerization by the labeled peptides, providing a complementary measurement of the relative affinity of different constructs for the dimer interface. As expected, addition of an equal concentration of unlabeled WT-GpA or WT-BNIP3 to the WT-GpA and WT-BNIP3 FRET pairs, respectively, was sufficient to reduce acceptor fluorescence (Figure 4B and 4C), consistent with displacement of labeled peptides by the unlabeled inhibitor. The ability of the unlabeled peptides to displace labeled monomers from the dimer complex also suggests that dimerization does not occur through nonspecific aggregation, in which an increase in peptide concentration should increase the FRET efficiency, which we do not observe. Consistent with our previous results (Figure 2), unlabeled aG-GpA was less effective at reducing acceptor fluorescence than its glycine-containing counterpart WT-GpA (Figure 4B), again indicating reduced binding affinity at the aza-glycine interface. However, unlabeled aG-BNIP3 displaced the labeled WT-BNIP3 dimer to a similar extent as unlabeled WT-BNIP3 (Figure 4C), in contrast to their differences in K_d values (Table 1). This may reflect a greater tolerance of the aza-glycine substitution at the heterodimer interface (probed through the competition experiment) compared to the homodimer interface in BNIP3.

Figure 4.

A) Schematic showing the competition experiment in which unlabeled peptide displaces labeled monomers from the dimer interface. B) Fluorescence emission spectra of labeled WT-GpA dimer in the absence or presence of either unlabeled WT-GpA or unlabeled aG-GpA. C) Fluorescence emission spectra of labeled WT-BNIP3 dimer in the absence or presence of either unlabeled WT-BNIP3 or unlabeled aG-BNIP3.

Given the consistent reduction in TM helix binding affinity upon aza-glycine incorporation, we hypothesized that any gains in stabilizing enthalpy imparted by the increased hydrogen-bonding potential were offset by destabilization of helical secondary structure due to stereoelectronic effects within the aza-glycine residue. To test this hypothesis, we recorded circular dichroism (CD) spectra of unlabeled WT-GpA, aG-GpA, WT-BNIP3, and aG-BNIP3 peptides in SDS detergent micelles. As expected, all peptides adopted α-helical secondary structure when reconstituted in detergent micelles (Figure 5). Consistent with our hypothesis, incorporation of aza-glycine resulted in a ~37% and ~31% decrease in the mean residue ellipticity (MRE) value at 222 nm of aG-GpA and aG-BNIP3 respectively compared to their unmodified counterparts, indicating a reduction in the overall α-helical content (Figure 5A and 5B, Table 1).

Figure 5.

Far-UV CD spectra of A) WT-GpA and aG-GpA recorded in 5 mM SDS; B) WT-BNIP3 and aG-BNIP3 recorded in 5 mM SDS; C) WT-GpA and aG-GpA recorded in 5 mM (dotted line) and 40 mM (solid line) SDS. All spectra are collected at a peptide concentration of 20 μM.

To test if the reduction in helicity upon aza-glycine incorporation results primarily from loss of dimerization, rather than intrinsic destabilization of the α-helix, we recorded the CD spectra at different peptide:detergent ratios that support differing degrees of dimerization (Figure 2D). Specifically, we compared the CD spectra of WT-GpA and aG-GpA in either 5 mM or 40 mM SDS. CD spectra for both WT and aG-GpA were indistinguishable upon the change in detergent concentration (Figure 5C), indicating that the observed decrease in helicity for aG-GpA results not just from a reduction in the dimer population but also the intrinsic destabilization of the α-helix caused by the aza-glycine substitution. Thermodynamic destabilization of the requisite helical secondary structure therefore explains why increased hydrogen-bonding potential in aza-glycine-containing peptides was insufficient to increase the stability of the TM helix dimer.

CONCLUSIONS

Despite increased hydrogen-bonding potential from the more acidic H^α donor, substituting interfacial glycine residues with aza-glycine in TM GxxxG motifs is insufficient to enhance dimerization affinity. Any potential stabilization through enhanced hydrogen bonding appears to be offset by destabilization of helical secondary structure. Given that aza-glycine causes significant destabilization of incompatible secondary structures like β-sheets⁴⁹ and the i+1 position of type IIʹ β-turns,⁴⁴ the relative tolerance of dimeric TM helices toward aza-glycine incorporation suggests the presence of compensatory stabilizing forces, such as interhelical hydrogen bonds. Though aza-glycine substitution is insufficient to increase the overall stability of TM helix assemblies, the sensitivity of the assembly to such a subtle substitution highlights the importance of the GxxxG interface in TM protein structure.

Supplementary Material

The following files are available free of charge.

Supporting Information: LC-MS Data and fluorescence excitation spectra (Figure S1) (PDF)

ABBREVIATIONS

GpA

Glycophorin A

BNIP3

Bcl-2 interacting protein 3

FRET

Förster Resonance Energy Transfer

TM

transmembrane

WT

wild-type

aG

aza-glycine

Pyr

Pyrene

Cou

7-dimethylaminocoumarin

Ac

Acetyl

ClogP_o/w

calculated logarithm of 1-octanol/water partition coefficient

POPC

1-palmitoyl-2-oleoyl-glycero-3-phosphocholine

LUV

large unilamellar vesicle