Enhanced Characterization of Membrane Protein Complexes by Ultraviolet Photodissociation Mass Spectrometry

Abstract

Development of chemical chaperones to solubilize membrane protein complexes in aqueous solutions has allowed for gas-phase analysis of their native-like assemblies, including rapid evaluation of stability and interacting partners. Characterization of protein primary sequence, however, has thus far been limited. Ultraviolet photodissociation (UVPD) generates a multitude of sequence ions for the E. coli ammonia channel (AmtB), provides improved localization of a possible post-translational modification of aquaporin Z (AqpZ), and surpasses previous reports of sequence coverage for mechanosensitive channel of large conductance (MscL). Variations in UVPD sequence ion abundance have been shown to correspond to structural changes induced upon some perturbation. Preliminary results are reported here for elucidating increased rigidity or flexibility of MscL when bound to various phospholipids.

Graphical Abstract

Membrane proteins represent a significant portion of current drug targets, but remain one of the most challenging classes of proteins to characterize.^1,2 Aggregation and precipitation are prevalent problems that have impeded analysis, and significant effort has been dedicated to exploring new methods for solubilization of membrane proteins as well as new tools to characterize them. Owing to the diminutive sample quantity requirements and speed of analysis, mass spectrometry is one such tool that has been essential in the field of proteomics and more recently in structural biology. Advances in hydrogen–deuterium exchange have been used to probe conformational changes of an ATP-binding cassette (ABC) transporter³ and to study proteins directly from lipid nanodiscs to characterize membrane association and their interactions with lipids and other ligands.^4,5 The use of mass spectrometer-friendly detergent micelles or other means (nanodiscs, lipid bilayers, amphipols, etc.) has facilitated the solubilization of membrane proteins, thus allowing them to be studied from aqueous solutions.^6–11 Combining the use of these types of chemical chaperones with gentle ionization methods such as nanoelectrospray ionization (nESI) offers the opportunity to transfer intact membrane proteins in native-like states to the gas phase, as reported in an increasing number of studies over the past decade.^7–15

Subunit stoichiometry can be confirmed and relative stability of membrane protein complexes can be evaluated directly from ESI mass spectra. The addition of various detergents and lipids can influence both aspects, prompting further studies into the role that the membrane bilayer plays in protein function.^9–12,16 Evaluation of membrane protein stability using mass spectrometry has also directed X-ray structure experiments by screening reagents that may lead to successful crystal-lization.^10,17 Our group has previously explored the thermodynamics and allostery of lipid binding to membrane proteins, demonstrating that different phospholipids can have positive, neutral, or negative allosteric effects on subsequent protein:lipid and protein:protein interactions on the ammonia channel of E. coli.^17–19 Protein stability can also be monitored using ion mobility, which paints a picture of the structural landscape of the membrane protein assemblies and how they may shift to more compact or extended conformations in the gas phase using different solution conditions, namely detergent or lipid compositions.^10–12,20

Until recently, predominantly just the intact masses of membrane protein complexes have been measured by mass spectrometry, thus leaving many unresolved questions about the primary sequences, post-translational modifications (PTMs), and local structural features that may be addressed by employing various ion activation methods to disassemble protein complexes or fragment individual proteins. Several groups have explored the use of tandem mass spectrometry, primarily collision-induced dissociation (CID), to achieve modest levels of sequence coverage by cleavage of the amide bonds in the protein backbone.^21–25 Enhanced cleavage in the transmembrane domains (TMDs) of non-native proteins was observed with CID; however, the intact, noncovalent assembly is lost when using organic solvents, prohibiting structural analysis from MS/MS spectra.^22–25 In one of the first MS/MS studies of native-like membrane protein complexes, Sobott and co-workers demonstrated that sequence coverage ranging from 11% to 73% of multimers from tetramers to hexamers is obtained using CID to characterize the intact precursor complexes.²¹ While CID is the most popular and widely available ion activation method, its low energy deposition and stepwise energization process cause few backbone cleavages and favor disruption of noncovalent interactions and labile PTMs. These preferential pathways upon CID lead to loss of lipids and disassembly of multimeric structures, thus impeding characterization of intact complexes.

Complementary to CID, electron-transfer and electron-capture dissociation (ETD and ECD, respectively) have been used to extend sequence coverage into the soluble regions of non-native membrane proteins,²³ although the utility of electron-based activation methods for analysis of native-like membrane complexes has yet to be explored. Another new option for characterization of protein assemblies is surface-induced dissociation (SID).^26–29 In fact, the architecture of quaternary structures can be inferred from MS/MS spectra generated by SID of protein complexes.^26–29 Unfolding and rearrangement is minimal, allowing interpretation of subunit connectivity and topology. Combined with ion mobility, surface labeling, cross-linking, and course-grained modeling methods, SID has been successfully used to determine structures of soluble protein complexes and, in one case, an integral membrane complex.^26,28

Ultraviolet photodissociation (UVPD) has recently emerged as a versatile new method for analysis of intact proteins, providing unsurpassed sequence coverage for elucidation of primary sequences and localization of PTMs.^30–32 The fast, high energy deposition of UVPD promotes backbone cleavages while simultaneously allowing retention of bound ligands, thus enabling the detection of holo-fragment ions that can be used to localize binding sites and/or monitor changes in native structure introduced by mutations or ligand binding.^32–35 However, UVPD has not been employed toward the analysis of native-like membrane proteins.

Here, we employ an Orbitrap mass spectrometer with an extended mass range and customized by addition of a 193 nm ArF excimer laser to allow characterization of both apo- and lipid-bound membrane protein complexes by UVPD. We demonstrate that UVPD provides a greater number of sequence ions, affording higher sequence coverage compared to conventional collisional activation methods. Importantly, UVPD also proves successful for membrane proteins adducted to salts, detergents, or lipid ligands without significant decreases in fragment ion abundances or sequence coverage. This latter attribute is particularly beneficial as membrane proteins are notoriously difficult to purify in the absence of adducted molecules, such as copurified lipids, often leading to broadened ion peaks that contribute to spectral complexity.

EXPERIMENTAL SECTION

Protein Expression and Purification.

Aquaporin Z (AqpZ), mechanosensitive channel of large conductance (MscL), and the ammonia channel from E. coli (AmtB) were expressed and purified as previously described.¹⁰ Purified membrane proteins were loaded onto a Superdex 200 Increase 10/300 GL column (GE Healthcare) equilibrated in GF buffer (130 mM sodium chloride, 10% glycerol and 50 mM TRIS, pH 7.4 at room temperature) supplemented with 0.5% tetra-ethylene glycol monooctyl ether (C₈E₄). Peak fractions containing membrane proteins were pooled, concentrated, flash frozen in liquid nitrogen, and stored at −80 °C. Lipids including 1-palmitoyl-2-oleoyl phosphotidylcholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphoinositol (POPI), and 1,1′,2,2′-tetraoleoyl-cardiolipid (TOCDL) were obtained from Avanti Polar Lipids (Alabaster, AL) and diluted in 200 mM ammonium acetate with 0.5% C₈E₄ detergent.

Native Mass Spectrometry.

Native mass spectrometry (MS) was performed on a Q Exactive Plus Ultra High Mass Range Orbitrap mass spectrometer (Thermo Fisher, Bremen, Germany). UVPD was implemented in the HCD cell through incorporation of a Coherent Excistar 193 nm ArF excimer laser (Santa Cruz, CA) as described previously.³⁶ Nanoelectrospray ionization was performed using gold/palladium-coated glass capillaries prepared in-house. Each purified protein was exchanged into 200 mM ammonium acetate solution containing 0.5% C₈E₄ detergent for native MS experiments for a final protein concentration of 2 μM or less. Instrument parameters were tuned to minimize activation of complexes and include an applied ESI capillary voltage between 1.3 and 1.5 kV, resolution setting of 100 000 at m/z 400, and in-source trapping desolvation potentials between −75 and −150 V. Ions were selected with a mass window of 10–20 m/z for MS/MS experiments. For MS/MS experiments, collision energies up to the maximum setting of 300 V or laser pulse energies up to 4 mJ were used for HCD and UVPD, respectively. For all UVPD analyses, a single laser pulse was used. For analysis of MscL bound to lipids, either POPC, POPI, or TOCDL was added in excess to the MscL protein up to 40 μM. At least three replicates were collected for each MS/MS acquisition.

Data Analysis.

Recorded spectra were deconvoluted using Xtract (Thermo Fisher Scientific) with a noise threshold of 3 and fragment ions were identified using a 10 ppm tolerance with UV-POSIT³⁷ and ProSight Lite.³⁸ Fragment ion abundances were mapped onto structures sourced from the Protein Data Bank using custom python scripting alongside PyMOL.³⁹ The subunit interfacial regions were identified from the crystal structures using the PISA tool from the European protein data bank (PDBePISA).⁴⁰

RESULTS AND DISCUSSION

Characterization of Membrane Proteins on a High-Resolution, Ultrahigh Mass Range Mass Spectrometer.

To evaluate the utility of ultraviolet photodissociation for characterization of intact membrane protein complexes, three integral membrane protein complexes were analyzed, including aquaporin Z (AqpZ, a 98.9 kDa tetramer), mechanosensitive channel of large conductance (MscL, an 85.6 kDa pentamer), and the ammonia channel (AmtB, a 127 kDa trimer). Each protein was solubilized in 0.5% C₈E₄ detergent and 200 mM ammonium acetate to maintain the native-like assemblies. The amplitude of the in-source trapping voltage proved critical to liberate the protein complex from the detergent micelle. For example, in-source trapping voltage of −50 V was insufficient to completely release the intact AqpZ tetramer from the detergent micelle, as evidenced by the numerous C₈E₄ adducts observed in the MS1 spectrum in Figure S1a. Raising the in-source trapping voltage to −75 V liberated the tetramer from the detergent micelle, yielding a mass spectrum with predominant signal for the apo tetrameric complex (Figures 1a and S1b). This desolvation voltage setting was used for all further analysis of AqpZ. The observed mass of the complex was 98,848 Da in good agreement with the calculated mass of 98,851 Da (Table S1).

Figure 1.

(a) ESI mass spectrum of AqpZ tetramer in 200 mM ammonium acetate and 0.5% C₈E₄ detergent. Inset is expanded to show the AqpZ tetramer (14+). Asterisks indicate residual C₈E₄ detergent molecules that are retained from nonspecific binding during the ESI process. MS/MS spectra of the AqpZ tetramer (14+) activated by (b) HCD, 4.2 keV E_Lab collision energy and (c) UVPD, 4 mJ, 1 pulse. (d) Locations of backbone cleavages resulting in sequence ions from UVPD are shown on the crystal structure (PDB 2ABM) of the AqpZ tetramer, colorized for a single subunit to reflect fragment abundances.

The AqpZ tetramer (14+) was isolated and subjected to HCD and UVPD, resulting in the spectra displayed in Figure 1b,c, respectively. Using the maximum HCD collision voltage of 300 V (4.2 keV E_lab collision energy) resulted predominantly in asymmetric charge partitioning, producing monomers (6+ to 8+) and trimers (6+ to 8+), similar to dissociation patterns observed for soluble protein complexes. Note that the abundances of the monomer ions are likely somewhat exaggerated relative to the complementary trimer ions owing to the faster decay of larger ions in the Orbitrap analyzer.^41,42 An array of sequence ions that originate from cleavage of the protein backbone were also observed using 4.2 keV E_lab collision energy, the relative abundances of which are mapped along the primary sequence of one AqpZ subunit in Figure 2a. The sequence ions generated from HCD yield an average of 30% coverage and 87 fragments, comprised of roughly an equal number of N-terminal (a- and b-ions) and C-terminal (y-ions) fragments (Figure 3a). The most abundant ions, indicated by red lines in Figure 2a, arose from preferential cleavages C-terminal to aspartic acid residues (b₁₁₀ and y₁₃₀) or N-terminal to proline residues (y₂₁₁ and y₈₃).⁴³ These dominant preferential cleavages have been previously observed for proteins in low charge states and are consistent with the underpinnings of the mobile proton model.^43,44

Figure 2.

Locations of backbone cleavages and corresponding abundances of the resulting N-terminal (a,b,c) and C-terminal (x,y,z) fragment ions from activation of the AqpZ tetramer (14+) using (a) HCD, 4.2 keV lab frame collision energy or (b) UVPD, 4 mJ, 1 pulse. (c) Structural features from the crystal structure (PDB 2ABM) differentiate variations in (i) protein solubility from transmembrane (teal) to soluble regions (black), (ii) secondary structure including helices (green) and coils (purple), and (iii) subunit interfaces (navy) and lipid or solvent exposed interfaces (maroon).

Figure 3.

Fraction of identified sequence fragments generated from HCD or UVPD of (a) the AqpZ tetramer (14+), (b) the MscL pentamer (13+), and (c) the AmtB trimer (16+).

UVPD of the AqpZ tetramer (14+) using a single 4 mJ laser pulse (Figure 1c) generates a larger array of sequence ions and nearly no ejected monomers. A minor amount of trimer ions might be produced, but they are obscured by the density of low abundance ions in the higher m/z range above the precursor ion. The congested region from m/z 6500 to m/z 13 000 presumably corresponds to semi-intact complexes containing truncated subunits owing to release of sequence ions. The low abundance of released monomers contrasts the findings described previously for UVPD of soluble protein complexes.^45–47 Soluble proteins of different oligomeric proteins were reported to disassemble into various subcomplexes depending on the laser power used for UVPD.^45–47 No such trend was observed for AqpZ in the present study using laser pulse energies from 1 to 4 mJ-pulse⁻¹, suggesting that the different interactions at subunit interfaces, such as hydrophobic interactions, salt bridges, or hydrogen bonds, may contribute to different photodissociation behaviors. One may assume that membrane protein complexes have weaker subunit interfaces in vacuum owing to the loss of hydrophobic interactions. Disassembly of the complexes would, therefore, be expected to occur more readily. Indeed, subunit ejection is a favorite pathway during HCD, demonstrated by the greater abundance of highly charged monomers (Figure 1b). It may be possible that absorption of a UV photon is sufficient to both cleave the protein backbone and disrupt sufficient noncovalent interactions to release sequence ions; however, disruption of noncovalent interactions should also contribute to monomer ejection. Many factors influence complex unfolding and dissociation, and further studies are required to fully understand how they affect UV photodissociation.

Though detection of monomers and other subcomplexes was sparse, an average of 193 sequence fragments were matched to ten ion types (a, a +1, b, c, x, x + 1, y, y − 1, Y, z; Figure 3a), corresponding to 45% coverage of AqpZ. The sequence ions predominantly correspond to ones originating from backbone cleavages close to the termini of AqpZ (Figure 2b). The most abundant product ion in the UVPD spectrum is consistent with a preferential cleavage N-terminal to a proline residue (a₂₉ + 1; Figures 1c and 2b), yet preferential backbone cleavages are observed to a lesser extent for UVPD compared to HCD.

A 57 Da deviation from the theoretical mass of the AqpZ monomer revealed the presence of a modification or mutation in the N-terminal region of the protein. As shown in Figure S2a, localization of the modification by HCD is limited to the first 25 amino acids. The enhanced coverage provided by UVPD improves localization slightly and can more precisely pinpoint the modification between residues 14 and 23 (Figure S2b). In addition to the 57 Da modification, variable formyl-Met was also observed as described previously.⁴⁸

In addition to the many N- and C-terminal fragments identified and displayed in Figure 2a, it is likely that sequence fragments that do not contain either terminus (i.e., internal fragments) are generated during activation. Internal fragments as well as fragments with side-chain losses (d/v/w-ions) are left unassigned in most fragment-matching software because of the massive increase in search space and increased likelihood of false positive matches.^49,50 Further discussion of the false discovery rate for HCD and UVPD of AqpZ is provided in the Supporting Information. Because only terminal ions are searched, many of the ions deconvoluted using Xtract remain unassigned using ProSight Lite (e.g., 91% unassigned from HCD and 89% unassigned from UVPD for AqpZ).³⁸ A manual search for side-chain loss fragments revealed that at least one such fragment was generated during UVPD (Figure S3) and likely many others. The production of internal fragments is amplified using higher collision energies or laser powers as the increased energy deposition can lead to multiple backbone cleavage events within a single ion. Indeed, the activation of the AqpZ tetramer using lower laser powers (2 and 3 mJ) resulted in more fragments extending deeper into the middle of the primary sequence (Figure S4). However, the signal of sequence ion fragments is lower when using lower laser powers, as is displayed in Figure S4a, and fewer ions were identified overall. Therefore, a laser power of 4 mJ-pulse⁻¹ was utilized to maintain higher sequence coverage and increased signal-to-noise of fragments. Though the fragmentation patterns vary slightly for different pulse energies, the distribution of sequence ion-types remained fairly consistent as a function of laser energy (Figure 3a). Most notably, a + 1 ions become more predominant in the UVPD spectrum using a single, 4 mJ pulse when weighted for abundance (Figure S5a). Similar fractions of explained/assigned ions from HCD and UVPD (9% and 11%, respectively) suggest there is no disadvantage in the increased number and array of fragmentation pathways from UVPD relative to HCD in this case. Despite the large portion of unassigned or unexplained fragments and the potential for extracting high information content from these ions, the reliance on manual inspection to identify side-chain loss and internal ions makes it impractical to explore these types of products in detail at this time.

To evaluate whether the fragmentation is more or less likely to occur in transmembrane domains (TMDs) or soluble regions, as has been suggested in a previous study,²³ the protein solubility across the primary sequence is displayed in Figure 2c. Structural information in Figure 2c was obtained from the crystal structure (PDB 2ABM) with TMDs colored in teal. Comparison of the TMD regions and backbone cleavage maps reveals no obvious preference for fragmentation of TMDs versus soluble regions for HCD or UVPD of the native-like assembly. Figure 1d shows the crystal structure of AqpZ imprinted with the sites of backbone cleavages from UVPD of the native-like AqpZ tetramer (14+), and fragmentation occurs in both the soluble loop regions as well as the transmembrane helices for the native protein. The lack of region-specific fragmentation of the AqpZ tetramer by HCD and UVPD is further demonstrated in Figure S6a, where the sequence ions are binned based on the residue’s presence in transmembrane or soluble regions. The distribution of backbone cleavage sites correlates with the total number of residues located in the TMDs and soluble regions, also shown in Figure S6a (gray bars), depicting no significant preference for cleavage based on hydrophobicity. This trend holds true for both HCD and UVPD. Additional structural features including secondary structure (α-helix, random coil, etc.) and location of subunit interfaces are also displayed in Figure 2c for qualitative assessment of preferential cleavage based on native structure. No trends in dissociation as a function of structure are apparent for either HCD or UVPD.

To maximize sequence coverage, a pseudo-MS³ strategy was implemented in which the AqpZ tetramers were activated using in-source trapping (−300 eV desolvation energy) to release highly charged monomers (Figure S7a). Highly charged monomers ejected using similar collisional activation on other platforms have shown to possess larger collisional cross sections owing to partial protein unfolding. Here, the reduction of noncovalent interactions allowed more efficient release of sequence ions during HCD (Figure S7b), resulting in an increase in the number of identified fragments (208), especially C-terminal ions (Figure S8c,d), and sequence coverage up to 66% using an E_Lab collision energy of 1050 eV (Figure S8a). The improvement in UVPD results was less dramatic but still significant with the number of identified fragments increasing to 329 and sequence coverage up to 69% using a single pulse of 3 mJ (Figure S8b). A similar shift to more abundant C-terminal ions, x and x + 1 ions in particular, was also observed for UVPD (Figure S8c,d). Higher collision or laser pulse energies produced poorer results presumed to be caused by overdissociation (e.g., production of small, uninformative fragment ions and inability to assign fragment ions in heavily congested regions of the MS/MS spectra).

A second membrane protein complex, MscL, was analyzed in a similar manner to AqpZ. The ESI mass spectrum, shown in Figure 4a, displays the intact pentamer with a typical native charge state distribution and an observed mass of 85 600 Da (theoretical 85 603 Da, Table S1). The pentamer (13+) was isolated and subjected to HCD (Figure 4b) or UVPD (Figure 4c). Interestingly, no monomers or other subcomplexes were released using either activation method across the range of collision or laser energies applied. Instead, exclusively sequence-type ions originating from backbone cleavages are the dominant products upon HCD and UVPD. This unique behavior may be caused by structural features in the C-terminal region of the protein, as discussed below.

Figure 4.

(a) ESI mass spectrum of MscL pentamer in 200 mM ammonium acetate and 0.5% C₈E₄ detergent. Asterisks indicate residual lipopolysaccharide molecules leftover from the expression and purification process.⁹ MS/MS spectra of the MscL pentamer (13+) activated by (b) HCD, 3.9 keV E_Lab collision energy and (c) UVPD, 4 mJ, 1 pulse. (d) Locations of backbone cleavages resulting in sequence ions from UVPD are shown on the crystal structure (PDB 2OAR) of the MscL pentamer, colorized for a single subunit to reflect fragment abundances.

The location of the backbone cleavages and the abundances of the sequence ions generated from UVPD are imprinted on one protein subunit of the crystal structure of pentameric MscL in Figure 4d, demonstrating exceptional coverage of the transmembrane helices. With few exceptions, sequence ions produced from HCD and UVPD are limited to ones containing the N-terminus (a,b,c), as illustrated in the fragment maps in Figure 5, panels a (HCD) and b (UVPD), and in the distributions of fragment ion types in Figures 3b and S5b. The large array of N-terminal fragments from both activation methods provides excellent characterization of the two TMDs of MscL that are located in the N-terminal half of the primary sequence as shown in Figure 5c (teal). Cleavage occurring C-terminal to aspartic acid residues was preferentially enhanced for HCD of the MscL pentamer (Figure 5a), similar to HCD of the AqpZ tetramer, and accounts for the most abundant N-terminal (b₁₆, b₃₆, and b₅₃) and C-terminal fragments (y₁₉). Also similar to the results for AqpZ was the absence of any clear specificity for soluble regions or TMDs of MscL using HCD, as shown in Figure S6b. Although the fragment ions observed in the HCD mass spectra were ones driven by mobile-proton pathways, an average of 61 total fragments were observed using 3.9 keV E_lab collision energy, amounting to 34% coverage, which triples the previously reported coverage of 11% using CID.²¹ UVPD extends this record even further with an average of 53% sequence coverage from 153 fragments using a single pulse of 4 mJ. A greater proportion of backbone cleavages occurred adjacent to transmembrane residues during UVPD (Figure S6b), potentially suggesting a proclivity for UVPD to occur in TMDs. However, the scarcity of C-terminal ions leaves just half of a picture for the MscL pentamer. As was observed for HCD of the MscL pentamer, the most abundant C-terminal sequence ions from UVPD are preferential cleavages adjacent to an aspartic acid residue (y₁₉) and a proline residue (y₁₈), with few others detected. Fragmentation patterns for the N-terminus of MscL do not appear to be dictated by protein solubility, secondary structure, or interface exposure (Figure 5c). Such claims cannot yet be made for the C-terminal fragmentation patterns (i.e., the lack thereof) owing to the insufficient electron density that left the 36 C-terminal residues of the protein (126–161) uncharacterized in the crystal structure (PDB 2OAR).⁵¹

Figure 5.

Locations of backbone cleavages and corresponding abundances of the resulting fragment ions from activation of the MscL pentamer (13+) using (a) HCD, 3.9 keV lab frame collision energy or (b) UVPD, 4 mJ, 1 pulse. (c) Structural features from the crystal structure (PDB 2OAR) differentiate variations in (i) protein solubility from transmembrane (teal) to soluble regions (black); (ii) secondary structure including helices (green), coils (purple), and strands (red); and (iii) subunit interfaces (navy) and lipid or solvent exposed interfaces (maroon).

One possible explanation for the lack of fragment ions containing the C-terminus (x-, y-, and z-ions) may stem from their inefficient release from the intact MscL complex. Lack of subunit ejection from the pentamer using both HCD and UVPD supports a hypothesis of strong subunit interactions. Cleavage of the protein backbone will not result in a detected sequence ion if noncovalent interactions cause the fragment(s) to be retained on an adjacent subunit. The C-termini of MscL subunits are predicted to be soluble, located in the cell cytoplasm, and disordered (assumed due to the poor electron density). Subunit and sequence ion detection may be suppressed if the C-termini of the subunits are involved in some interfacial region that impedes release of subunits or sequence fragments. Varying the laser pulse energy did not result in vast differences in backbone cleavage locations (Figure S9) and provided no additional C-terminal fragments of MscL. Whatever is precluding characterization of the C-terminus is independent of activation method, as neither HCD nor UVPD generated a substantial array of C-terminal fragments nor subunit ejection. The low charge density of native-like membrane proteins, which results in few charges available per subunit (average of 2.6 positive charges per monomer from the 13+ pentamer), may also rationalize the scarce detection of C-terminal fragments, as the positive charges necessary for ion detection are less likely to be localized along the acidic C-terminus. Because release of monomers was not observed, a pseudo-MS³ approach was not possible to increase coverage of the C-terminus. Further experiments are required to confirm the influence of protein structure on fragmentation propensity. Such experiments may include structural elucidation using X-ray crystallography or NMR spectroscopy or MS/MS analysis of the non-native protein using an unfolding technique prior to activation.

As one of the most well-studied membrane protein complexes using mass spectrometry, AmtB was also subjected to HCD and UVPD analysis. The ESI mass spectrum of the trimer is displayed in Figure S10a with a standard, native-like charge state distribution with an average mass roughly in agreement with theoretical values (Table S1). MS/MS spectra of the most abundant charge state (16+) are shown in Figure S10b,c using a lab frame collision energy of 4.8 keV and a single laser pulse of 4 mJ, respectively. Owing to the larger size of each subunit (42.2 kDa) and the subsequent increased number of noncovalent interactions, sequence coverage of AmtB was limited to 15% and 12% from HCD and UVPD, respectively. The enhanced coverage by HCD relative to UVPD is surprising and possibly explained by an unfolding of the protein allowing more efficient release of sequence ions. Alternatively, UVPD occurs faster than unfolding, which may prevent release of fragments caused by retained noncovalent interactions. The diversity of fragments from UVPD is also greater than that achieved from HCD, which leads to increased spectral complexity, fewer resolved and identified fragments, and ultimately lower sequence coverage. More UVPD fragments were, nonetheless, identified (80) than HCD fragments (68). Similar to MscL, sequence ions are localized primarily toward the N-terminus, as shown in Figures S11 and 3c, with few identified C-terminal ions, apparent by the limited x-, y-, and z-ions. The scarcity of C-terminal fragments is amplified when weighted for abundance in Figure S5c. It is not clear if the limited C-terminal coverage is caused by insufficient release of such fragments or ineffective detection. If the former, gentle collisional activation prior to HCD or UVPD (to unfold the protein) or after (to release sequence ions that are retained by noncovalent interactions on the complex) may be used to increase coverage. However, in-source trapping did not provide sufficient energy to release suitable amounts of monomer for a pseudo MS³ event, shown in Figure S12, and instead began to induce covalent backbone cleavages. The release of monomers and dimers from the trimer suggest that the lack of C-terminal coverage is not solely due to noncovalent interactions that might prevent release of the fragments from the complexes. Indeed, the C-terminal region of the protein does not appear to be involved in the subunit interface (PDB 1U7G, not shown). Post-HCD or UVPD activation is not possible using the current instrument design. It is possible that C-terminal fragments are being generated but are not detected owing to spectral congestion and low signal, potentially caused by fewer charges at the C-terminus relative to the N-terminus (also suggested for MscL). In this case, supercharging or non-native solvents may be useful for increasing sequence coverage of protein subunits.

Evaluation of UVPD for Structural Characterization of Membrane Proteins.

The ability to localize ligand binding sites and reveal conformational rearrangements within proteins using UVPD^32–35 motivated our interest in evaluating the capabilities of UVPD to probe phospholipid binding to MscL. The ESI mass spectra of MscL bound to phosphoinositol (POPI), phosphocholine (POPC), and cardiolipin (TOCDL) are displayed in Figure S13. Zero to five lipids are bound for each charge state (11+ to 17+) of the pentamer. Each MscL·lipid_n complex was isolated and subjected to UVPD. The sequence coverages, number of matching fragments, and distribution of fragment ion types in the resulting MS/MS spectra were remarkably similar for each complex (Figure S14). These results are particularly advantageous for the study of membrane proteins that often retain lipids and detergents that are coisolated in the purification process and difficult to remove.

As demonstrated in previous reports, the pattern and distribution of backbone cleavage sites that result in ligand-containing (holo) sequence ions upon UVPD can be used to convey the binding sites of various ligands.^32,34,35 To date, the protein–ligand interactions that have been retained during UVPD have correlated largely with salt bridges and hydrogen bonding. Because transfer to the gas phase results in disruption of the hydrophobic interactions between lipid tails and the transmembrane regions of these protein complexes, retention of the lipid may be driven by interactions at the lipid headgroup. However, a search for MscL fragments bound to phospholipids revealed no abundant holo-ions, prohibiting localization. The absence of assignable holo-product ions in this case may be caused by (1) heterogeneous binding of lipids to multiple locations on the protein or (2) photodissociation of the lipid during activation. Because many fragments are generated during UVPD, the signal is distributed across many more types of ions than observed upon collisional activation. Heterogeneous lipid-binding amplifies this issue and reduces the signal-to-noise ratio for holo-fragments, prohibiting confident identification. UVPD has proven to be a versatile activation method that also promotes fragmentation of phospholipids.^52,53 It is possible, therefore, that sequence ions may contain fragments of phospholipids, ones which go undiscovered if the mass of the intact lipid is incorporated into the fragment ion search. Truncated portions of the phospholipids were not considered in the search for holo-ions in order to minimize false positive identifications.

Although holo-fragments retaining the lipids are not observed here, an alternative strategy for assessing the impact of lipid binding entails monitoring the changes in abundances of the apo fragment ions produced upon UVPD of the apo-MscL pentamer versus the MscL·lipid_n complexes. Figure 6a displays the significant differences in apo-fragment abundances generated from the apo-MscL (13+) and MscL·lipid₅ (13+) assemblies, shown as heat maps extending from the N-terminus to the C-terminus across the protein. The heat maps are displayed for MscL complexes containing POPI, POPC, and TOCDL lipids. These data do not necessarily localize the lipid-binding location but may suggest regions that have undergone some structural reorganization upon lipid-binding.

Figure 6.

Differences in apo-fragment ion abundances generated from UVPD (1 pulse, 4 mJ) of apo-MscL and holo-MscL precursor (13+) bound to five phospholipids plotted on (a) the primary sequence and (b–d) on a single subunit of the pentamer crystal structure (PDB 2OAR). The predicted transmembrane domains are contained within the dashed blue lines between the periplasmic and cytoplasmic domains. Only differences calculated using t test with a significance level of 95% are included. Results are shown for three lipids: 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphoinositol (POPI), 1-palmitoyl-2-oleoyl phosphotidylcholine (POPC), and 1,1′,2,2′-tetraoleoyl-cardiolipid (TOCDL).

The information from each lipid-specific heat map is imprinted on thecrystal structure in Figure 6b–d, shaded to mirror the heat map gradient. The locations of the significant changes in fragment ion abundances are similar for the MscL pentamers containing different lipids, with the most significant variations for the apo versus holo proteins occurring in the first transmembrane helix (residues 15–44) and spanning into the periplasmic domain (residues 45 to 68), which are pointed out in Figure 6c. The variations that occur in the transmembrane helix are consistent with previous molecular dynamics simulations and collisional cross section measurements.¹⁰ The significant changes in backbone cleavages that occur in regions of the protein that are expected to reside in the periplasm may be rationalized by a global shift in protein conformation MscL upon binding lipids or by association of the polar head groups.

Interestingly, when MscL is bound to POPI or TOCDL, backbone cleavages on average are suppressed. This outcome may suggest an increase in rigidity of the protein and enhancement of noncovalent interactions that would subsequently reduce the backbone fragmentation in this region or decrease the separation and release of fragment ions upon UVPD, thus reducing the abundances of the detected fragment ions diagnostic for this region. Surprisingly, when MscL is bound to POPC, backbone cleavages of this same region of the protein appear to be enhanced upon UVPD, leading to higher abundances of the diagnostic fragment ions relative to that observed upon UVPD of apo-MscL. This result may suggest that POPC promotes conformational dynamics rather than stabilizing the protein like POPI and TOCDL. Further studies will be required to fully understand the structural effects of MscL binding various phospholipids.

CONCLUSION

Membrane proteins were characterized using UVPD for the first time. A diverse array of sequence ions was observed for three membrane protein complexes, AqpZ, MscL, and AmtB, contributing to the enhanced sequence characterization commonly afforded by high energy UV photoactivation. Sequence coverages of 45% and 53% were obtained from UVPD of AqpZ and MscL, respectively, surpassing values of 30% and 34% obtained from HCD. Pseudo-MS³ events, in which HCD and UVPD were used to dissociate ejected monomers from the AqpZ tetramer, increased sequence coverage to 66% and 69%, made possible by the reduced number of noncovalent interactions. Collisional activation of AmtB resulted in 15% coverage compared to 12% from UVPD. The diminished performance of UVPD in this instance is likely due to the generation of ions with overlapping m/z leaving many fragments unresolved and subsequently unidentified, which is more likely with increasing subunit mass.

Interestingly, subunit ejection was largely unobserved using UVPD, regardless of pulse energy, which contrasts previous reports for soluble complexes. Typical subunit ejection patterns were generated from collisional activation of the AqpZ tetramer, providing complementary information to UVPD. However, cleavage of the protein backbone occurred preferentially during HCD of AmtB and MscL, particularly within the N-terminal half of the sequence. Further studies will be necessary to understand what is precluding release and detection of subunits and C-terminal sequence ions.

UV photoactivation of MscL bound to various phospholipids resulted in comparable levels of characterization across several different lipidated states, showing promise for future membrane-protein analysis using UVPD. Changes in UVPD fragment ion abundance for the apo-MscL and MscL:lipid₅ assemblies may suggest global structural changes of the protein when bound to lipids. Suppression of fragment abundance from MscL bound to POPI and TOCDL hints toward a shift to a more rigid MscL complex, whereas an enhancement in fragment ion abundance when bound to POPC may indicate a shift to a more flexible or dynamic conformation. Localization of the lipid binding region to specific residues was impeded by the lack of observed holo-fragment ions that retain intact lipid molecules. Future studies to assess retention of lipids bound to membrane proteins should begin with the analysis of well-characterized systems with discrete binding motifs.