Uniting Native Capillary Electrophoresis and Multistage Ultraviolet Photodissociation Mass Spectrometry for Online Separation and Characterization of Escherichia coli Ribosomal Proteins and Protein Complexes

Abstract

With an overarching goal of characterizing the structure of every protein within a cell, identifying its interacting partners, and quantifying the dynamics of the states in which it exists, key developments are still necessary to achieve comprehensive native proteomics by mass spectrometry (MS). In practice, much work remains to optimize reliable online separation methods that are compatible with native MS and improve tandem MS (MS/MS) approaches with respect to when and how energy is deposited into proteins of interest. Herein, we utilize native capillary zone electrophoresis coupled with MS to characterize the proteoforms in the Escherichia coli 70S ribosome. The capabilities of 193 nm ultraviolet photodissociation (UVPD) to yield informative backbone sequence ions are compared to those of higher-energy collisional dissociation (HCD). To further improve sequence coverage values, a multistage MS/MS approach is implemented involving front-end collisional activation to disassemble protein complexes into constituent subunits that are subsequently individually isolated and activated by HCD or UVPD. In total, 48 of the 55 known E. coli ribosomal proteins are identified as 84 unique proteoforms, including 22 protein-metal complexes and 10 protein-protein complexes. Additionally, mapping metal-bound holo fragment ions resulting from UVPD of protein-metal complexes offers insight into the metal-binding sites.

Graphical Abstract

INTRODUCTION

Owing to the prevalence of noncovalent interactions between proteins and metals, ligands, or other proteins, the definition of such networks is necessary to fully understand cellular processes or elucidate disease mechanisms.¹ Preservation of such interactions and characterization of the protein-protein and protein-ligand partners poses a significant problem, especially in the quest to analyze more elaborate mixtures of proteins that are better representatives of a biological system. Numerous strategies have been developed to evaluate protein interactions, ranging from spectroscopic to microscopic to mass spectrometric to other molecular biology methods.^2,3 Mass spectrometry (MS) in particular offers the potential both to identify individual proteins via various proteomic techniques and to characterize interactions via application of native-MS methods that allow preservation of non-covalent interactions of the protein-protein and protein-ligand partners.^4–6 However, despite substantial inroads in the performance of mass spectrometric methods, characterization of mixtures of protein complexes remains challenging. Bottom-up MS-based proteomics provides a robust method for characterization of the primary sequences of proteins but is less suitable for mapping protein interactions due to the use of denaturing conditions and proteolytic digestion of proteins into peptides.⁷ Alternative top-down methods enable direct analysis of intact proteins and now successfully demonstrated even for complex mixtures.⁸ As such, top-down proteomic uniquely offers the potential to derive a complete picture of all of the combinations of molecular forms in which a protein resulting from a single gene exists [i.e., caused by genetic variation, alternative splicing, and post-translational modification (PTM)], referred to as proteoforms.⁹ However, the wealth of information related to noncovalent interactions is typically not retained owing to the use of denaturing solvents that facilitate conventional reversed phase liquid chromatography necessary for efficient protein separations but disrupt all levels of higher order structures. An alternative approach, native MS, involves the use of volatile salts during electrospray ionization (ESI) to efficiently transfer proteins into the gas phase while maintaining noncovalent interactions. This method allows analysis of multi-protein complexes that retain both metals and ligands with higher order structures reminiscent of those adopted in solution.^4–6 While this technique has become widely used in studies aimed at probing secondary, tertiary, and quaternary protein structures, it is generally limited to highly purified single proteins or simple protein mixtures.^10,11 Achieving more comprehensive characterization of the proteoforms present in multimeric macromolecules in complex biological mixtures requires technical advances in two primary areas: (1) robust, high-resolution separation methods compatible with the volatile salts necessary for native MS and (2) MS instrumentation capable of detecting high m/z species as well as enabling proficient tandem MS (MS/MS) to allow identification of the constituent proteins.

Several separation techniques have been adapted to be compatible with native MS conditions, including size exclusion chromatography (SEC),^12–14 ion exchange chromatography (IEX),^15,16 hydrophobic interaction chromatography (HIC),¹⁷ and electrophoresis methods.^18–24 While SEC is routinely used to separate higher order structure variants or aggregates of antibodies under native MS conditions,^12–14 the simplicity of this method often yields relatively low-resolution separations insufficient to distinguish the subtle differences between various proteoforms of a single protein. IEX and HIC have demonstrated superior resolving powers in the analysis of computationally designed oligomers,¹⁵ biopharmaceutical protein products,¹⁶ and antibody-drug conjugates.¹⁷ However, both of these methods operate on principles requiring relatively high salt concentrations which can cause disruption of weak noncovalent interactions. Additionally, the amenability of these separation methods (SEC, IEX, and HIC) to nanoscale flow rates has yet to be demonstrated. In this regard, they require significantly larger sample quantities, may cause sample dilution, and ultimately result in lower sensitivity compared to conventional reversed-phase LC approaches used for bottom-up and top-down proteomics. In contrast to such techniques that rely on interactions of analytes with a solid phase, electrophoresis methods utilizing a high electric field to separate molecules based on charge and size have become attractive options for performing native separations.^18–24 Isoforms and subcomplexes of relatively stable protein–protein complexes have been separated by capillary isoelectric focusing.^18,19 However, this method requires denaturing sheath flow buffers to maintain stable ESI, potentially causing protein unfolding. Additionally, native gel-eluted liquid fraction entrapment electrophoresis has been used for off-line separation of endogenous protein complexes but has not yet been coupled online to a mass spectrometer owing to the use of non-MS compatible detergents and salts.^20,21 Stemming from its ability to achieve high-resolution separations using negligible quantities of samples and buffers under both denaturing and native conditions, capillary zone electrophoresis (CE) has emerged as a top contender for facilitating analysis of protein complexes in conjunction with native MS.^22–27 Although non-MS compatible buffers are typically used as the background electrolyte (BGE) in CE, volatile salts such as ammonium acetate or ammonium formate are easily substituted. Additionally, commercially available MS sources (both sheath flow and sheathless) circumvent initial issues with completing the electrical circuit for separation while simultaneously maintaining appropriate voltages necessary for ESI.^28,29

Two key features of native MS, the deposition of fewer charges during the ESI process and the retention of noncovalent interaction partners, result in the production of high m/z ions that demand mass analyzers with extended mass ranges. This is why initial native MS studies were generally limited to time-of-flight (TOF) or Fourier-transform ion cyclotron resonance instruments.^30–33 More recently, the development of Orbitrap instruments with ultra-high mass range (UHMR) capabilities has provided a new high performance platform for native MS.³⁴ With respect to MS/MS capabilities, collisional activation, including collisional-induced dissociation and higher-energy collisional dissociation (HCD), has proven proficient for disassembly of protein complexes into subunits or even monomeric proteins, but the low degree of backbone cleavages results in limited sequence coverage of proteins, impeding both identification and characterization.³⁵ Several ion activation methods based on alternative mechanisms and/or offering higher energy deposition have been developed for the characterization of intact protein complexes.³⁶ Electron-based methods, including electron-transfer dissociation, electron-capture dissociation, and electron ionization dissociation, offer improved sequence coverage of the proteins within protein complexes.^37–40 Conversely, surface-induced dissociation mainly causes disruption of multimeric complexes into constituent protein subcomplexes and individual subunits, thus affording a remarkable new strategy for probing architectures of protein assemblies.⁴¹ Another high-energy activation method, ultraviolet photodissociation (UVPD), produces both intact subunits and extensive series of sequence ions resulting from backbone cleavages.^42–44 In addition to returning unsurpassed sequence coverages and retention of labile PTMs even for high-throughput workflows,^45–48 UVPD maintains non-covalent interactions, thus resulting in metal- or ligand-bound holo fragment ions that offer insights into binding sites.⁴⁹

While these alternative MS/MS methods provide high sequence coverages and impressive characterization of increasingly large protein complexes,⁵⁰ signal averaging of several minutes may be required to achieve sufficient resolution and signal-to-noise necessary for identifying backbone cleavage sites. Thus, the amenability to high-throughput workflows is limited. Several recent studies posit a multistage MS approach in which an initial stage of collisional activation of intact protein complexes [referred to as in-source trapping (IST)] is used to disassemble oligomeric proteins into constituent monomeric subunits.^51,52 The resulting monomeric species are then individually isolated and activated using HCD or UVPD. These successful multistage MS/MS approaches (IST-HCD or IST-UVPD) have afforded higher sequence coverages compared to single-step activation methods and improved characterization of model protein complexes,⁵¹ enzymes implicated in either metabolism⁵³ or chemotherapeutic resistance of glioblastoma tumors,⁵⁴ and protein complexes in a human cell lysate pre-fractionated off-line.²¹ More recently, endogenous lipids bound to membrane proteins were identified using a strategy that integrated native MS of intact protein-lipid complexes, collisional activation to separate the lipids from the protein, and subsequent HCD or 213 nm UVPD of the lipids.⁵⁵

Here, we unite CE with a multistage IST-UVPD approach to characterize the proteins comprising the Escherichia coli 70S ribosome. The tolerance for high salt concentrations characteristic of CE allows up to 500 μM magnesium acetate to be included in the BGE solution and results in the observation of the intact 30S and 50S subunits. By reducing the magnesium concentration in the BGE and removing the ribosomal RNA (rRNA), subcomplexes and/or single proteins are generated for subsequent MS/MS characterization. HCD and UVPD as well as multistage IST-HCD and IST-UVPD methods are used to characterize the ribosomal proteins and protein complexes. PTMs, including methylation, acetylation, and phosphorylation, are identified in addition to noncovalently bound metal cofactors (Mg²⁺ and Zn²⁺). UVPD consistently yields higher sequence coverages of the ribosomal proteins, protein-metal, and protein-protein complexes compared to HCD. Incorporating front-end collisional activation (IST-UVPD) further improves the protein characterization capabilities of UVPD, yielding significantly higher sequence coverage and offering insights into metal-binding sites based on mapping holo fragment ions. The demonstration of CE in conjunction with a multistage MS/MS approach represents a significant advance in establishing a robust pipeline for native proteomics studies.

EXPERIMENTAL SECTION

Ribosomal Sample Preparation.

E. coli 70S ribosomes were purchased from New England BioLabs (Ipswich, MA) and exchanged into 25 mM ammonium acetate (pH 6.8) containing 1 mM magnesium acetate using 10 kDa molecular weight cutoff filters (MilliporeSigma, Burlington, MA) for CE-MS analysis using a BGE containing a high magnesium concentration. Alternatively, ribosomal nucleic acids were precipitated as previously described.⁵⁶ Briefly, 1:4 (v/v) 100 mM magnesium acetate/ribosome and 1:1 (v/v) glacial acetic acid/ribosome were added to the ribosome suspension. The sample was incubated at 4 °C for 1 h before centrifugation at 10,000 rpm for 5 min. The resultant supernatant was removed and exchanged into 25 mM ammonium acetate (pH 6.8) containing 0 or 100 μM magnesium acetate for CE-MS analysis.

Native Capillary Electrophoresis.

Ribosomal samples prepared at ~30 μg/μL were hydrodynamically injected at 5 psi for 30–45 s using a CMP Scientific (Brooklyn, NY) ECE-001 capillary electrophoresis autosampler into a 100 cm capillary (50 μm inner diameter) coated with linear polyacrylamide as previously described.²⁴ In short, a bare fused silica was flushed successively with 1 M hydrochloric acid, water, 1 M sodium hydroxide, water, and methanol prior to exposure to 3-(trimethoxysilyl)propyl methacrylate to introduce carbon-carbon double bonds along the capillary wall. An aqueous acrylamide solution containing ammonium persulfate was used to fill the treated capillary before incubation at 50 °C for 2 h. Extensive flushing with water removed unreacted reagents, and introduction of the BGE solution (25 mM ammonium acetate, pH 6.8, containing no, 100, or 500 μM magnesium acetate) conditioned the capillary. CE separations were achieved using a +30 kV separation voltage with 0.5 psi of pressure applied throughout the separation. Applying a higher pressure generally shortens the total separation time at the expense of the resolution. An electrokinetically pumped sheath flow interface (CMP Scientific, Brooklyn, NY) allowed coupling of the CE capillary with the mass spectrometer. ESI emitters (~20 μm tip opening, 4 cm length) were fabricated using a Sutter Instrument (Novato, CA) P-1000 micropipette puller from borosilicate capillaries (1.0 mm outer diameter, 0.75 mm inner diameter) for the CE-MS interface. The sheath liquid was 10 mM ammonium acetate throughout the study. Following CE separation, online ionization of proteins was carried out using ESI spray voltages of 2.4–2.6 kV.

Mass Spectrometry.

A Thermo Scientific Q Exactive UHMR mass spectrometer (Bremen, Germany) previously modified⁵⁴ to perform UVPD in the HCD cell by a Coherent ExciStar XS 500 (Santa Clara, CA) pulsed excimer laser operating at 193 nm (ArF gas) was used for all CE-MS experiments. The source temperature was set at 200 °C, and all ion optics were optimized for the transmission of species of interest. Specifically, for CE conditions with high magnesium acetate concentrations (100 and 500 μM), values were tuned to transfer higher m/z species, whereas for CE-MS without magnesium acetate present in the BGE solution, parameters were adjusted for lower m/z species. During ribosomal analysis, ESI mass spectra, without and with IST, were collected at a resolving power of 6250. MS/MS spectra were acquired at a resolving power of 140k for the top three most abundant precursors in a data-dependent manner. For all spectra, the ion population was controlled by the ion time (IT) as automated gain control was turned off during CE-MS analysis. MS spectra represent two averages using an IT of 20 ms, and MS/MS spectra are based on 10 averages with an IT of 300 ms. HCD spectra were collected using collision energies of 200–250 eV/q, and a single laser pulse at 3 mJ was used for UVPD spectra. For multistage analyses (IST-HCD and IST-UVPD), the desolvation voltage was increased from −25 V to −225 V i.e.(, resulting in disassembly of protein complexes instead of simply desolvation). The nitrogen bath gas pressure of the HCD cell was lowered for CE-MS analysis using none or 100 μM in the BGE solution. Specifically, the value was adjusted from a pressure corresponding to 1 × 10⁻⁹ to 4 × 10⁻¹⁰ mbar or 1 × 10⁻¹⁰ mbar in the ultra-high vacuum region for HCD and UVPD, respectively.

Data Analysis.

Triplicate CE-MS runs were collected each using HCD, UVPD, IST-HCD, and IST-UVPD for solutions containing no and 100 μM magnesium acetate in the BGE. ProSight Native was used to de-charge low resolution ESI and IST mass spectra and deconvolute corresponding high resolution MS/MS spectra. Lists of the average masses observed in the ESI mass spectra (acquired with and without IST) were assigned as 70S ribosomal proteins or protein-protein complexes within ±3 Da. Searches included methylation (+14.0 Da), acetylation (+42.0 Da), and phosphorylation (+80.0 Da) as possible PTMs. The presence of metal cofactors was identified by a mass difference corresponding to Mg²⁺ (+22.3 Da) or Zn²⁺ (+63.4 Da). Fragment ion matches were made using a tolerance of ±15 ppm. For HCD mass spectra, only b- and y-type ions were considered, whereas for UVPD mass spectra, nine ion types were searched (a, a + 1, b, c, x, x + 1, y, y - 1, z). Table S1 summarizes the observed average mass, sequence coverage, and P-score (Poisson-based score) for each identified ribosomal protein and protein complex based on CE-MS analysis in the absence of magnesium acetate [(-) Mg²⁺] or containing 100 μM magnesium acetate in the BGE. Holo fragment ions (bound to Mg²⁺ or Zn²⁺) produced upon UVPD were identified by inclusion of the corresponding mass shifts at the N- and C-terminus. All reported sequence coverages derived from UVPD include both holo (metal-bound) and apo (metal-free) fragment ions. Only holo fragment ions identified in all three replicates at a given magnesium acetate concentration in the BGE were considered confidently identified and reported.

RESULTS AND DISCUSSION

Separation of Ribosomal Proteins by Native CE Using Various Mg²⁺ Concentrations.

Comprised of 55 proteins held together by three rRNA strands, the E. coli 70S ribosome requires a minimum concentration (10 mM) of Mg²⁺ to maintain its intact 2.3 MDa structure.^57,58 At lower concentrations (<1 mM), this macromolecular complex dissociates into 50S (1.4 MDa, composed of 33 proteins) and 30S (850 kDa, consisting of 22 proteins) subunits.⁵⁹ The dependence of subunit association and activity on the presence and concentration of Mg²⁺ suggests that this metal cofactor intricately governs the dynamics of the ribosome by controlling the balance between flexibility and stability.^60,61 The Mg²⁺-dependent reorganization of the ribosome has been studied by hydrogen/deuterium exchange and MALDI TOF-MS to localize regions impacted by varying magnesium concen-trations.^62,63 Additionally, intact E. coli ribosomes have been previously detected by native MS⁶⁴ and the heterogeneity of such macromolecules studied using various concentrations of magnesium acetate in the ESI spray solution.⁶⁵ Similarly, native MS in conjunction with top-down and bottom-up MS methods have been used to dissect ribosomal protein complexes and define the ribosomal proteoforms present across the kingdoms of life, including bacteria, plants, and humans.⁶⁶ These prior studies relied on denaturation of the ribosomal proteins to delineate the proteoforms present based on application of bottom-up or top-down MS techniques. In the present study, the use of CE for front-end separation facilitates the characterization of the ribosomal proteins and protein complexes present at various magnesium concentrations under native conditions.

Base peak electropherograms for the CE-MS analysis of ribosomal proteins with various concentrations of magnesium acetate (no, 100, and 500 μM) in the BGE solution are shown in Figure 1. Although CE is typically tolerant of salts, inclusion of greater than 500 μM magnesium acetate in the BGE resulted in poor spray stability and a low MS signal (and thus likely accounts for the lack of detection of the intact 70S ribosome which typically requires a minimum magnesium concentration of 1 mM for stabilization). Nevertheless, two dominant multimeric species were observed in the presence of 500 μM magnesium acetate (Figure 1A). Intact masses in the ESI mass spectra shown in Figure 1B allow these species to be identified as the 30S (migration time 53.72 min, 850.1 ± 0.2 kDa) and 50S (migration time 65.41 min, 1450.4 ± 0.9 kDa) subunits. A previous study⁶⁵ utilizing direct infusion ESI of an E. coli ribosomal sample similarly observed these two subunits using 500 μM magnesium acetate in the spray solution. Decreasing the magnesium concentration of the BGE alone failed to result in production and detection of significantly more subcomplexes or individual proteins. However, removal of the rRNA by precipitation, as described in Experimental section, yielded the electropherograms in Figure 1C,D. After removal of the nucleic acids scaffolding the complex, a variety of proteins and protein complexes were present in solution for characterization by MS/MS methods.

Figure 1.

(A) Base peak electropherogram of E. coli ribosomal proteins containing 500 μM magnesium acetate in the BGE solution. (B) Experimental masses of the dominant species in the ESI mass spectra collected at migration times of (1) 53.72 and (2) 65.41 min with a high Mg²⁺ concentration (500 μM) correspond to the theoretical masses of intact 30S and 50S subunits. The broad peaks of low abundance around m/z 21,000 in [B(1)] and m/z 16,000 in [B(2)] may correspond to the 50S and 20S subunits, respectively. Extracted ion chromatograms for observed m/z values of the 30S (turquoise) and 50S (orange) subunits are overlaid with the base peak electropherogram in (A). Removal of rRNA allowed separation of smaller subcomplexes and individual proteins shown as the base peak electropherograms in (C) no Mg²⁺ and (D) 100 μM Mg²⁺ inthe BGE. Table S1 summarizes all species identified under these latter two conditions.

MS/MS Methods for the Analysis of Ribosomal Proteins.

The absorption of 193 nm photons by the protein backbone results in direct dissociation from excited states and allows access to more fragmentation pathways during UVPD compared to collisional activation methods.⁴⁴ Owing to its higher energy deposition, UVPD has consistently yielded greater levels of sequence coverage for intact proteins under denaturing^45–48 and native^42,43,50,54 conditions alike. For proteomics applications, higher sequence coverage generally affords better characterization of the various proteoforms present.⁴⁷ HCD and UVPD mass spectra were collected in a data-dependent manner after CE separation of ribosomal proteins with no or 100 μM magnesium acetate present in the BGE solution. Additionally, increasing the IST voltage enabled implementation of a multistage MS/MS approach (referred to as IST-HCD and IST-UVPD) for protein characterization.

Table S1 summarizes the sequence coverage values and P-scores for all identified proteoforms using each of the four activation methods (HCD, UVPD, IST-HCD, and IST-UVPD) obtained for both CE separation conditions (e.g., no and 100 μM magnesium acetate). Representative MS and MS/MS spectra are shown in Figure S1 for three proteoforms containing covalent modifications (RL7 (Me, Ph)·Mg²⁺, RS16 (Ac)·Mg²⁺, and RS11 (Me)) confidently identified and characterized using IST-UVPD. Identified proteoforms are categorized into three groups: individual proteins, protein-metal complexes, and protein-protein complexes (Figure 2). In the absence of magnesium acetate during CE separation, HCD and IST-HCD resulted in the identification of the same 54 proteoforms, whereas UVPD alone confirmed the presence of 61 proteoforms and IST-UVPD allowed assignment of 62 proteoforms. Almost half of the species identified by both HCD and UVPD corresponded to protein–metal complexes with only 3 protein–protein complexes confirmed by each of the four MS/MS approaches. As expected, inclusion of 100 μM magnesium acetate in the BGE for the CE separation resulted in the identification of significantly more protein–protein complexes (10 each for HCD and UVPD). The identification of an overall lower number of proteoforms for CE–MS with 100 μM magnesium acetate in the BGE solution compared to no magnesium acetate is attributed to the signal suppression and MS peak broadening caused by the presence of a high concentration of Mg²⁺, a trend commonly noted in native MS studies.^65,67

Figure 2.

Graph displaying the total number of E. coli ribosomal proteoforms identified using HCD, IST-HCD, UVPD, or IST-UVPD to activate proteins after CE separation with no (left) or 100 μM (right) magnesium acetate present in the BGE solution. Identified proteoforms are divided into three categories: protein (turquoise), protein-metal complex (orange), or protein-protein complex (tan). A majority of the protein-protein complexes also contain a metal bound to at least one of the subunits. A complete list of identified proteoforms is shown in Table S1.

While there is significant agreement between the overall number of proteoforms identified by HCD and UVPD (Figure S2), examination of sequence coverage values confirms that UVPD generally offers better characterization of a given proteoform. Figure 3A displays the sequence coverages obtained by HCD and UVPD for all identified proteoforms for the CE-MS analysis of the ribosome containing no magnesium acetate. Corresponding bar graphs for IST-HCD vs IST-UVPD in the absence of magnesium acetate in the BGE solution and all activation methods (HCD vs UVPD and IST-HCD vs IST-UVPD) for CE-MS using 100 μM magnesium acetate are shown in Figure S3. With the exception of proteoforms of lower molecular weight (<10 kDa), UVPD resulted in higher sequence coverage values across the board. Plots of the differences in sequence coverages for HCD vs UVPD and IST-HCD vs IST-UVPD for both CE conditions (no and 100 μM magnesium acetate) highlight this trend (Figure 3B). A list detailing the ranked proteoform pairs is shown in Table S2. The orange dots and triangles indicate those proteoforms in common for which greater sequence coverage (>2%) was afforded by UVPD or IST-UVPD compared to HCD or IST-HCD. The turquoise dots and triangles denote the opposite outcome with higher sequence coverages for HCD or IST-HCD compared to UVPD or IST-UVPD. Under both CE conditions (no and 100 μM magnesium acetate in the BGE), over 75% of the dots/triangles are orange which suggests sequence coverages produced by UVPD and IST-UVPD are consistently higher than those resulting from HCD and IST-HCD for native protein complexes analyzed in a high-throughput manner by CE-MS. Greater sequence coverages of E. coli ribosomal proteins (unmodified or carbamylated) by UVPD compared to HCD have also been previously using denaturing LC-MS/MS for analysis.^46,68 In general, UVPD sequence coverages reported here using native CE-MS are somewhat lower (by ~10–15%) than six similar ribosomal proteins reported in a previous study (RS19, RL15, RL18, RL24, RL34, and RL36),⁶⁸ an outcome attributed to the fact the MS parameters in the present study were optimized for identification of multimeric protein complexes (i.e., higher m/z species).

Figure 3.

(A) Bar graph depicting sequence coverages afforded by HCD (turquoise) and UVPD (orange) for each E. coli ribosome proteoform identified after CE separation with no magnesium acetate in the BGE solution. The floating labels correspond to the proteoforms of the subunits comprising the identified protein-protein complexes. Corresponding graphs for IST-HCD and IST-UVPD as well as for the separation containing 100 μM magnesium acetate using each activation method are shown in Figure S3. The differences in sequence coverage of proteoforms identified in common are shown for (B) HCD vs UVPD (dots) and IST-HCD vs IST-UVPD (triangles) and (C) UVPD vs IST-UVPD. Proteoforms were ranked by increasing difference in coverage. Lists of ranked proteoform pairs are included in Table S2. In (B), orange indicates proteoforms for which UVPD (dots) or IST-UVPD (triangles) generated greater sequence coverage than HCD (dots) or IST-HCD (triangles) by more than 2%, turquoise highlights proteoforms for which HCD (dots) or IST-HCD (triangles) yielded higher sequence coverage than UVPD (dots) or IST-UVPD (triangles) by more than 2%, and tan indicates proteoforms for which the sequence coverage differed by less than 2%. Similarly, in (C), orange denotes proteoforms for which IST-UVPD afforded greater sequence coverage than UVPD (>2%), turquoise indicates proteoforms for which UVPD resulted in higher sequence coverage than IST-UVPD (>2%), and tan identifies proteoforms for which the sequence coverage differed by less than 2%. Note that the y-axis for (1) (-)Mg²⁺ and (2) 100 μM Mg²⁺ in (C) is scaled differently than (B) (maximum value of 35% instead of 55%) since the differences in sequence coverages between UVPD and IST-UVPD for corresponding proteoforms were smaller. Sequence coverages for all identified proteoforms are listed in Table S1.

A comparison of the sequence coverage values associated with UVPD and IST-UVPD is also shown for both CE conditions (no and 100 μM magnesium acetate in BGE) in Figure 3C. Again, orange dots denote proteoforms in common whose sequence coverages using IST-UVPD were higher compared to UVPD, while turquoise dots and triangles indicate higher sequence coverage values resulting from UVPD compared to IST-UVPD. In the absence of magnesium acetate, there is minimal improvement in sequence coverage when using the multistage approach (IST-UVPD) relative to UVPD. However, at a magnesium acetate concentration of 100 μM (favoring the survival of more multimeric protein-protein complexes), IST-UVPD resulted in higher sequence coverage than UVPD for over two-thirds of the proteoforms. The proteins for which UVPD alone outperformed IST-UVPD tended to be smaller proteins, but a more extensive study including a larger number of proteins would be needed to confirm this finding.

Multistage MS/MS Approach for the Improved Characterization of Ribosomal Protein Complexes.

As demonstrated in Figure 3C, a multistage MS/MS approach is particularly beneficial for improved characterization of protein-protein complexes compared to HCD or UVPD alone. Closer examination of the largest protein-protein complex (66.4 kDa) identified at no or 100 μM magnesium concentrations further demonstrates this trend. The hetero-pentameric RL8 complex [consisting of RL10(RL7/RL12)₄] comprises the stalk of the ribosomal complex.⁶⁹ Owing to its role in promoting translation factors to the ribosome, it remains highly mobile and is absent from most crystal structures of the ribosome.^57,58,69 As such, the development of alternative structural biology methods in which this protein complex can be reliably detected as part of the ribosome is important. While subcomplexes of this assembly were detected in the absence of magnesium acetate in the BGE during CE separation [i.e., (RL7/RL12)₂, (RL7/RL12)₄, and RL10(RL7/RL12)₂], the intact complex was only identified at the higher magnesium acetate concentration (100 μM). The ESI mass spectra collected at a migration time of 61.81 min during the CE separation performed with 100 μM magnesium acetate in the BGE (Figure 1D) are shown in Figure 4A. The 14+ to 16+ charge states of the pentameric RL10(RL7/RL12)₄ complex bound to four equivalents of Mg²⁺ (observed average mass: 66,325 ± 9 Da; theoretical average mass: 66,409.7 Da) were detected. Increasing the desolvation voltage (-225 V) for a multistage MS/MS approach yields the IST mass spectrum displayed in Figure 4B in which the monomeric subunits corresponding to each individual protein are observed [6+, 7+ charge states of RL7/RL12·Mg²⁺; 5+, 6+ charge states of RL7/RL12 (Ac)·Mg²⁺; and 6+ to 8+ charge states of RL10]. Asymmetric partitioning of the charge from the heteropenta-meric complex, common for collisional activation methods (such as IST), yields monomeric subunits with a disproportionate amount of charge.

Figure 4.

(A) ESI mass spectrum collected at a migration time of 61.81 min during online CE separation of E. coli ribosomal proteins using a BGE solution containing 100 μM magnesium acetate (base peak electropherogram shown in Figure 1D). The 14+ to 16+ charge states of the pentameric RL10(RL7/RL12)₄·4Mg²⁺ complex are labeled. (B) IST spectrum, collected at a migration time of 61.27 min during CE separation with 100 μM magnesium acetate in the BGE, resulting from disassembly of the protein-protein complex observed in (A) into its constituent monomeric subunits: RL7/RL12·Mg²⁺ (6+, 7+ charge states labeled in turquoise), RL7/RL12 (Ac)·Mg²⁺ (5+, 6+ charge states labeled in orange), and RL10 (6+ to 8+ charge states labeled in tan). (C) Maps and sequence coverage values for each protein comprising the complex afforded by (1) UVPD of the 15+ charge state of the pentamer and (2) IST-UVPD or (3) IST-HCD of each individual subunit [7+ charge state of RL7/RL12·Mg²⁺, 6+ charge state of RL7/RL12 (Ac)·Mg²⁺, 7+ charge state of RL10]. Observed fragment ion types are labeled as a/x (green), b/y (blue), and c/z (red). MS/MS spectra for the UVPD maps shown in (C) are included in Figure S4.

The UVPD mass spectrum of the 15+ charge state of the pentameric RL10(RL7/RL12)₄ complex and IST-UVPD mass spectra of each individual subunit (7+ charge state of RL7/RL12·Mg²⁺, 6+ charge state of RL7/RL12 (Ac)·Mg²⁺, 7+ charge state of RL10) and the corresponding deconvoluted MS/MS spectra are shown in Figure S4. The sequence coverage maps and sequence coverages corresponding to each of the three proteins contained in the complex demonstrate several remarkable features (Figure 4C). First, UVPD (without IST) of the pentameric complex yields a modest number of diagnostic fragment ions of each of the constituent proteins, yielding 11 to 16% sequence coverage. More notably, the multistage IST-UVPD approach results in significantly better characterization of the proteoforms constituting this complex (Figure 4C-2). In particular, sequence coverage values using IST-UVPD increased to 43% for RL7/RL12·Mg²⁺, 43% for RL7/RL12 (Ac)·Mg²⁺, and 35% for RL10. These improvements are attributed to the simplification of the UVPD spectra owing to disassembly of the pentamer and activation of the subunits individually instead of activation of the pentameric complex. Fewer averages are necessary to achieve quality MS/MS spectra for the subunits than the multi-protein complex making the multistage approach more amenable for the high-throughput analysis of protein-protein complexes.

Mapping UVPD Holo Fragment Ions to Examine Metal Cofactor Binding to Ribosomal Proteins.

In addition to higher sequence coverages, the accessibility to higher energy fragmentation pathways allows the preservation of non-covalent interactions between the protein and metals or ligands during cleavage of covalent backbone bonds of the protein, a phenomenon reported for UV photoactivation of protein-metal and protein-ligand complexes.^{49,54,70–72} This counter-intuitive outcome is rationalized by the fast cleavage of the polypeptide backbone occurring from ions in excited states prior to vibrational energy redistribution which favors disruption of lower energy covalent or noncovalent bonds.⁴⁴ In practice, mapping these ligand- or metal-retaining holo fragment ions generated during UVPD has offered insight into the residues involved in ligand binding for a wide variety of protein-ligand complexes such as an NADPH cofactor and inhibitor methotrexate bound to dihydrofolate reductase,⁷⁰ GDP or a GTP-analogue within the oncogenic rat sarcoma protein K-Ras,⁷¹ adenosine phosphate ligands interacting with the phosphotransferase enzyme adenylate kinase,⁷² and a pyridoxal 5'-phosphate cofactor pulled down with branched-chain amino acid transferase 2 (BCAT2).⁵⁴

Two specific examples of UVPD holo fragment ion mapping are shown in Figure 5 for the protein-metal complexes RL29· Mg²⁺ and RL31·Zn²⁺ identified by IST-UVPD in the CE separation of ribosomal proteins with no magnesium acetate present in the BGE. The backbone cleavage sites that lead to holo fragment ions are represented either as lines above the protein sequence or colored along the corresponding crystal structure of the proteins (PDB ID: 4V4Q).⁷³ Backbone cleavages that result in N-terminal-containing holo fragment ions (a, b, c) are colored turquoise, C-terminal-containing holo fragment ions (x, y, z) are shaded orange, and backbone cleavages that produce complementary N-terminal- and Cterminal-containing ions (“bi-directional” holo fragment ions) are shown in dark red. Corresponding 1ST–MS spectra and IST-UVPD MS/MS spectra for both protein-metal complexes are shown in Figure S5. For RL29, the Mg²⁺ ion is expected to interact with acidic residues (D or E) which correlates with the interaction site identified by mapping holo fragment ions (L22–Q25). Similarly, the putative binding site of Zn²⁺ in RL31 is C16. This aligns with the binding region elucidated from tracking the holo fragment ions (S15–N20). These types of holo ions are not generally created upon HCD as collisional activation tends to disrupt electrostatic interactions preferentially relative to cleavage of backbone bonds of the protein.

Figure 5.

Backbone cleavage sites of holo (Mg²⁺- or Zn²⁺-containing) fragment ions observed upon IST-UVPD of (A) RL29·Mg²⁺ and (B) RL31·Zn²⁺ following online CE separation of E. coli ribosomal proteins using no magnesium acetate in the BGE solution represented as lines above the corresponding sequences or colored along a crystal structure of the protein (PDB ID: 4V4Q). Cleavage that results in N-terminal holo fragment ions (a, b, c) are colored turquoise, C-terminal holo fragment ions (x, y, z) are shaded orange, and complementary N-terminal and C-terminal ions (bi-directional holo fragment ions) are shown in dark red. Corresponding ESI-MS spectra, UVPD MS/MS spectra, and sequence coverage maps including apo (metal-free) fragment ions are shown in Figure S5.

CONCLUSIONS

Characterization of E. coli 70S ribosomes at various Mg²⁺ concentrations is successfully accomplished by coupling native CE with a UHMR mass spectrometer equipped with UVPD. The capabilities of this instrument for detecting high m/z species mitigates bias evident in previous studies toward the identification of lower mass complexes (<30 kDa) and enables identification of protein-protein complexes (up to 66.4 kDa) larger than that previously reported using a native high throughput workflow.^23,24 In agreement with these previous studies,^23,24 front-end separation of such protein complexes by native CE is fairly reproducible and robust based on observation of minimal variations in triplicate runs in the present study. Additionally, the ability to access different structural states of a macromolecular complex through adjustment of the salt concentration of the BGE solution posits CE as a versatile separation partner for MS. While not necessary for analysis of ribosomal proteins, more complex samples might require the use of preconcentration or sample stacking methods to overcome the lower sample loading capacity of native CE compared to liquid chromatography approaches.²⁸ Four different MS/MS approaches are compared, with UVPD consistently outperforming HCD in the characterization of ribosomal proteoforms across individual protein, protein-metal complexes, and protein-protein complexes. Incorporation of a multistage approach (IST-HCD and IST-UVPD) further increases observed sequence coverage values, particularly for protein-protein complexes owing to the simplification of the precursor species. Although previous studies using native MS in conjunction with denaturing top-down and bottom-up LC-MS/MS⁶⁶ or CE-MS/MS under denaturing and native conditions²³ to analyze E. coli 70S ribosomes both resulted in the identification of more proteoforms, the current study was designed specifically to improve the characterization of intact protein-protein complexes (higher m/z species). Additionally, mapping holo fragment ions generated by UVPD offers further confirmation of putative binding sites for protein-metal complexes. Such fragment ions provide another layer of information relevant for deciphering the organization of protein complexes containing metals or small molecule ligands. As this multistage native CE/UVPD-MS approach is further scaled up, the high-throughput analysis of more complex samples such as the endogenous proteins in cell lysates represents the next frontier and will require development of automated algorithms that matches primary and quaternary structural information in an integrated manner.