Online Mixed-Bed Ion Exchange Chromatography for Native Top-Down Proteomics of Complex Mixtures

Abstract

Native top-down mass spectrometry (nTDMS) allows characterization of protein structure and noncovalent interactions with simultaneous sequence mapping and proteoform characterization. The majority of nTDMS studies utilize purified recombinant proteins, with significant challenges hindering application to endogenous systems. To perform native top-down proteomics (nTDP), where endogenous proteins from complex biological systems are analyzed by nTDMS, it is essential to separate proteins under nondenaturing conditions. However, it remains difficult to achieve high resolution with MS-compatible online chromatography while preserving protein tertiary structure and noncovalent interactions. Herein, we report the use of online mixed-bed ion exchange chromatography (IEC) to enable separation of endogenous proteins from complex mixtures under nondenaturing conditions, preserving noncovalent interactions for nTDP analysis. We have successfully detected large proteins (>146 kDa) and identified endogenous metal-binding and oligomeric protein complexes in human heart tissue lysate. The use of a mixed-bed stationary phase allowed retention and elution of proteins over a wide range of isoelectric points without altering the sample or mobile phase pH. Overall, our method provides a simple online IEC-MS platform that can effectively separate proteins from complex mixtures under nondenaturing conditions and preserve higher-order structure for nTDP applications.

Graphical Abstract

Introduction

Native top-down mass spectrometry (nTDMS) has emerged as a powerful tool to study the structure and dynamics of proteins and protein assemblies, providing complementary insights to conventional structural biology techniques such as X-ray crystallography, nuclear magnetic resonance spectroscopy (NMR), and cryo-electron microscopy (cryo-EM).^1–9 nTDMS enables the localization of binding sites for ligands and metal cofactors,^6,10–12 analysis of subunit organization,^12–14 and investigation of higher-order structure.^2,15–17 Nonetheless, nTDMS is generally limited to the analysis of overexpressed recombinant proteins and purified standard proteins.^1–6 Significant challenges remain in the application of nTDMS to endogenous proteins in biological systems due to the high complexity and large dynamic range of the proteome.^18,19

To perform nTDMS on a proteome scale, known as “native top-down proteomics” (nTDP), it is essential to separate proteins under nondenaturing conditions prior to nTDMS analysis.^7,20–22 Commonly available MS-compatible liquid chromatography (LC) modes, such as reversed-phase liquid chromatography (RPLC), separate proteins in their denatured state, resulting in a loss of noncovalent interactions and higher-order structural information.^22,23 Maintaining native-like conditions places strict requirements on online chromatography, forbidding not only all nonvolatile additives, but also organic solvents, non-physiological pH, and high temperatures.^24,25 As a result, nondenaturing separations often have lower resolution and are poorly suited for LC-MS compared to their denaturing counterparts.^5,26,27 Few online, nondenaturing chromatographic methods have been developed, with applications primarily to relatively simple systems including antibody-drug conjugates, biotherapeutic charge variants, and recombinant proteins utilizing hydrophobic interaction chromatography (HIC), ion-exchange chromatography (IEC), or size-exclusion chromatography (SEC).^28–35 Additionally, capillary zone electrophoresis (CZE) has been used as an online, nondenaturing separation and was successfully applied to an E. coli lysate.^20,36–38 There remains an urgent need to develop high-resolution, nondenaturing chromatographic separations of endogenous proteins from complex mixtures for nTDP.

Under nondenaturing conditions, IEC separates proteins based on surface charge characteristics which can generally be predicted from isoelectric point (pI).^39,40 The distribution of isoelectric points in the proteome is wide and bimodal,⁴¹ meaning that a complex biological sample will have a large quantity of both acidic and basic proteins, complicating IEC method development. One approach is to alter the pH of the mobile phase to bring the majority of proteins to the desired charge,^39,42,43 however, the structure of many protein complexes are sensitive to changes in pH.^44–46 Alternatively, a mixed-bed ion exchange column (containing both cation-exchange and anion-exchange material) is capable of retaining both acidic and basic proteins at neutral pH.^47,48

Herein, we report the use of online mixed-bed IEC for nTDP analysis of complex, endogenous mixtures. We demonstrate the utility of mixed-bed IEC separations for nTDP using standard protein mixtures and protein extract from human heart tissue. We observed that noncovalent interactions such as metal binding and oligomerization are preserved following online mixed-bed IEC, and successfully detected proteins up to 146 kDa. Furthermore, we found that retention behavior in online mixed-bed IEC provides an additional source of structural information when combined with nTDP data. Overall, this work demonstrates the utility of online mixed-bed IEC for nTDP analyses of complex, endogenous mixtures.

Experimental Section

Reagents and Consumables

All reagents were purchased from MilliporeSigma (Burlington, MA, USA) unless otherwise noted. Solvents were prepared with water from an in-house Milli-Q system (Millipore, Corp., Billerica, MA, USA). Ion exchange columns were generously provided by PolyLC Inc. (Columbia, MD, USA).

Human Cardiac Tissue Collection

Healthy donor hearts with no history of heart disease were obtained from the University of Wisconsin–Madison Organ and Tissue Donation-Surgical Recovery and Preservation Services. Tissue was kept in cardioplegic solution until dissection, after which they were flash-frozen in liquid nitrogen and stored at −80 °C. The procedures for the collection of human donor heart tissues were approved by the UW–Madison Institutional Review Board.

Protein Extraction from Human Heart Tissue

Human heart tissue (20 mg) was homogenized in 150 μL of HEPES extraction buffer (25 mM HEPES pH=7.4), 60 mM NaF, 1 mM L-methionine, 1 mM PMSF, 1x HALT protease/phosphatase inhibitor cocktail using a Teflon pestle at 4 °C. The homogenate was centrifuged for 30 min at 21,000 g at 4 °C and the supernatant was transferred to Eppendorf Protein Lo-Bind tubes, snap frozen in liquid nitrogen, and stored at −80 °C.

IEC-nTDP Analysis of Standards and Complex Extracts

A mixture of 5 standard proteins that spanned a range of isoelectric points, noncovalent interactions, and PTMs were selected for method development. The proteins chosen were carbonic anhydrase (CA; 29 kDa) from bovine erythrocytes (UniProt ID: P00921), ovalbumin (Ova; 44 kDa) from chicken egg white (UniProt ID: P01012), enolase (Eno; 93 kDa) from brewer’s yeast (UniProt ID: P00924), cytochrome C (CytC; 12 kDa) from equine heart (UniProt ID: P00004), and lysozyme (Lys; 14 kDa) from chicken egg white (UniProt ID: P00698). The standard protein mixture was desalted into 10 mM ammonium acetate using a 10 kDa molecular weight cut-off (MWCO) filter to reach an approximate final concentration of 1.0 mg/mL, 2.0 mg/mL, 2.5 mg/mL, and 0.5 mg/mL for CA, Ova, Eno, and Cyt/Lys, respectively. Human heart protein extracts were desalted into 10 mM or 50 mM ammonium acetate using a 30 kDa MWCO filter. A Bradford protein assay was used to determine the final concentration.

Online liquid chromatography–tandem mass spectrometry (LC–MS/MS) was performed using a Waters ACQUITY UPLC M-Class and Bruker maXis II quadrupole time-of-flight mass spectrometer. The mobile phases were 10 mM ammonium acetate (A) and 800 mM ammonium acetate (B) with a flow rate of 5 μL/min. PolyCATWAX and PolyWAX capillary columns (0.3×100 mm, 3μm, 1500Å) (PolyLC Inc., Columbia, MD, USA) were used for online LC–MS/MS. For the standard protein mixture, 500 nL (3.5 μg total) was injected and subjected to a salt gradient (Table S1). For the human heart HEPES and LiCl protein extracts, 4.0 μL (12.6 μg and 5.5 μg, respectively) was injected and subjected to a salt gradient (Table S1). The column was kept at room temperature, and the autosampler temperature was 4 °C. For MS analysis, the source conditions were a capillary voltage of 4500 V, end plate offset of 500 V, nebulizer pressure of 0.8 bar, dry gas flow rate of 4 L/min, and dry gas temperature of 180 °C. A scan rate of 1 Hz was used with a mass range from 500-6000 m/z, quadrupole low mass cutoff set at 300 m/z, and a quadrupole energy of 4 eV. The ion transfer optics were set at a funnel 1 RF of 400 Vpp, multipole RF of 800 Vpp, and isCID energy of 150 eV. The collision cell energy was 10 eV with a collision cell RF of 4000 Vpp, a transfer time of 135 μs, and a pre-pulse storage time of 35 μs. For MS/MS analysis, autoMS/MS was enabled (Table S2), selecting the four most abundant precursors for fragmentation, with active exclusion of precursors after 5 scans and an exclusion time of 2 minutes.

Data Analysis

Bruker Compass v4.3 was used to analyze MS spectra, performing charge deconvolution by MaxEnt. Proteoforms above 50 kDa were deconvoluted with MaxEnt using a resolving power of 10,000 followed by the SumPeak algorithm to determine average mass. To determine monoisotopic masses, spectra were deconvoluted by eTHRASH in MASH Native v1.1.⁴⁹ Theoretical isotopic fit profiles were generated in Bruker Compass IsotopePattern and MASH Native v1.1. Extracted ion chromatograms (EICs) shown are smoothed with 1 point Gaussian smoothing. For MS/MS analysis, data files were converted from Bruker *.d format to *.mzML using the MS ProteoWizard⁵⁰ v3.0.22068 MSConvert GUI with peak picking. The *.mzML files were then deconvoluted using the TopFD GUI in TopPIC suite v1.6.2.⁵¹ Deconvoluted results were searched against the UniProtKB/Swiss-Prot reviewed human proteome in FASTA format using TopPIC. TopPIC parameters included an E-value cutoff of 0.01, a 15 ppm error tolerance, ±500 Da unexpected mass shift, and five variable modifications: oxidation, methylation, dimethylation, acetylation, and phosphorylation. The maximum number of variable modifications was set to 2. TopPIC output was manually validated in MASH Native v1.1.

Results and Discussion

Online IEC-nTDP Analysis of Standard Proteins

The initial online mixed-bed IEC method was developed using a mixture of standard proteins chosen to cover a range of isoelectric points (~5.9 to ~10.7; Table S3) and noncovalent interactions including metal binding and oligomerization. The standard mixture was buffer exchanged into ammonium acetate to match the starting mobile phase condition and prevent nonspecific adducts. We found that the mixed-bed stationary phase was able to retain both the acidic and basic standard proteins and separate them with a simple salt gradient (Figure 1a). The resulting mass spectra were characteristic of native MS, indicated by low charge states and narrow charge envelopes²⁵ (Figure 1b). The dimerization of Eno and iron binding of CytC was preserved and confirmed with high mass accuracy. To show the applicability of this method to nTDP, CA was subjected to collisionally activated dissociation (CAD) which resulted in substantial C-terminal fragmentation (Figure S1). A closer look at the online PolyCATWAX separation of Ova revealed multiple peaks deriving from different charge variant phosphorylation states (Figure 1c). Ova is known to have two primary phosphorylation sites at S68 and S344, with the bisphosphorylated form being most abundant.⁵² While the largest peak in the chromatogram corresponded to bisphosphorylated Ova, a well-resolved peak and a shoulder corresponded to two monophosphorylated Ova proteoforms. One of these peaks was a monophosphorylated charge variant with a mass difference of approximately 1 Dalton (Table S4).

Figure 1. Online mixed-bed IEC of standard proteins.

a) Representative base peak chromatogram (BPC) and extracted ion chromatograms (EIC) for a separation of a mixture of standards. b) Native mass spectra and theoretical isotope fitting of standard proteins. Reported deconvoluted masses are monoisotopic except for Eno where average mass is reported. Eno crystal structure: PDB 1ebh. c) Online IEC separation of Ova including a native mass spectrum (left), labeled chromatogram displaying separation of phosphorylated proteoforms (middle), and zoom-in of a single Ova charge state for each species (right). d) Online IEC separations of CA including a native mass spectrum (left), a comparison of separations by PolyWAX and PolyCATWAX (middle), and two views of the surface electrostatic potential of the protein calculated by the Adaptive Poisson-Boltzmann Solver (APBS). Blue corresponds to positive regions and red corresponds to negative regions. CA crystal structure: PDB 1v9e.

To better understand retention patterns in mixed-bed ion exchange, further experiments were conducted with standard proteins. First, all standards were subjected to analytical-scale separations using the weak anion-exchange material (PolyWAX) and weak cation-exchange material (PolyCAT) in isolation (Figure S2). Surprisingly, the acidic proteins CA and Eno were primarily retained by the PolyCAT phase. Based on their isoelectric points (pI<7), one would expect CA and Eno to be retained by the PolyWAX phase at the near-neutral pH of unadjusted ammonium acetate. The impact of the stationary phase material on pH during the separation^39,53,54 was tested by collecting mobile phase fractions across a salt gradient (Supplementary Experimental Section) and was found to have a negligible effect. Across the gradient, the maximum pH difference between any tested eluent fractions was approximately 0.3 pH units (Figure S3). A plausible reason for the retention of CA by PolyCAT was revealed by mapping the electrostatic surface potential of the protein using the Adaptive Poisson-Boltzmann Solver (APBS)⁵⁵ in PyMOL. When folded, many of the negative residues are buried in the zinc binding pocket of the protein and thereby less accessible to the stationary phase (Figure 1d). Moreover, the exposed surfaces appear to have sufficient positive charge to explain retention. To investigate the limits of this behavior, CA was then separated by both PolyCAT and PolyWAX columns with the pH increased to 7.9 (Figure S4). With elevated pH, retention of CA by PolyCAT was lost, indicating a loss of positive surface charge. Retention of CA by PolyWAX remained low at pH 7.9, suggesting that the inaccessibility of the negative pocket continued to prevent retention. Analysis of the PolyWAX pH 7.9 condition by online IEC-nTDMS revealed that CA charge variants were being separated (Figure S5). Single-Dalton mass shifts likely corresponding to deamidation (resulting in a reduction in pI)⁵⁶ slightly increased the retention of CA by PolyWAX but decreased overall retention by PolyCATWAX. This supports that the PolyCAT phase is the main determiner of CA retention near neutral pH, and that surface charge distribution makes important contributions to the retention of folded proteins in IEC. In summary, these observations from targeted experiments indicate that retention patterns can provide additional structural information when correlated with native MS data.

Online IEC-nTDP Analysis of Human Heart Protein Extracts

To demonstrate the applicability of this method to complex biological samples, we performed a PolyCATWAX separation of a human heart native protein extract. Soluble proteins were extracted from ~20 mg of heart tissue with a HEPES extraction buffer and exchanged into ammonium acetate to match the gradient starting condition. The overall shape of the total ion chromatogram (TIC) displayed a wide spread of elution times throughout the gradient, notably lacking a large “flow-through” peak which would indicate the presence of many unretained proteins (Figure 2a). Average peak width was calculated to be approximately 1.1 minutes (Table S5), comparable to what has been achieved in higher-resolution, online nondenaturing separations such as native CZE.^20,38 This corresponds to a peak capacity⁵⁷ of approximately 28 over the 30-minute gradient, a number which could be significantly increased by utilizing longer gradients and multidimensional separations. The charge envelopes observed in the mass spectra exhibited low charge states and narrow distributions as expected in native MS (Figure 2b), indicating that the extraction protocol and separation method were nondenaturing. We detected charge envelopes corresponding to the presence of multiple high molecular weight proteins, including a 126 kDa and a 146 kDa species. Notably, we detected an 85.9 kDa protein that we suspect to be a homodimer due to the concurrent elution of a species of exactly half its molecular weight in the mass spectrum (Figure 2b-1).

Figure 2. Detection of high molecular weight species from online mixed-bed IEC of proteins from a human heart extract.

a) Representative total ion chromatogram (TIC) and extracted ion chromatograms (EICs) from an online mixed-bed IEC separation of a HEPES protein extract from human heart tissue. EICs are scaled to the same height to show peak shape. b) Native mass spectra with labeled charge states and deconvoluted average masses corresponding to each labeled EIC.

Auto MS/MS could be performed only on ions below 3000 m/z and resulted in confident identification through top-down sequencing of 10 proteins across two injections (Figure 3, Supplementary Spreadsheet S1). Highlighting our nondenaturing workflow, we detected the monomeric and homodimer forms of malate dehydrogenase in the same elution window with high mass accuracy (Figure 3b-5). We also identified the Cu-Zn superoxide dismutase monomer with both metal cofactors bound (Figure 3b-3), confirming that this method can retain bound metal ions throughout the extraction and separation steps. Additionally, we identified S100-A1 and S100-A6, key proteins in intracellular Ca²⁺ signaling⁵⁸ and ion channel modulation⁵⁹ (Figure 3b-6,8).The primary proteoform of both S100-A1 and S100-A6 were N-terminal methionine-removed and acetylated, features not currently annotated on UniProt. Moreover, both S100-A1 and S100-A6 were detected with bound calcium (Ca²⁺) and magnesium (Mg²⁺) ions (Figure S7), while S100-A1 was also detected as a homodimer (Figure 3b-8). Competitive binding of Mg²⁺ is common in Ca²⁺-binding proteins and in some cases relevant to disease.^60–62 Myoglobin was also identified (Figure 3b-2), with some portion bound to heme and iron (III). The higher oxidation state of iron and higher abundance of apo-myoglobin suggested that further optimization of the native extraction and preparation steps is needed to maintain protein structural integrity. The loss of the proteins’ native environment during the extraction and buffer exchange step may result in alteration to concentration-dependent interactions, especially weaker interactions.^9,24,25 It is also possible that the separation step had some effect on protein structural integrity. Although our experiments with standard proteins did not show any loss of noncovalent binding, the wider range of proteins and their noncovalent interactions present in a tissue lysate may respond differently.⁹

Figure 3. Native top-down identification of proteins from human heart extract.

a) Representative total ion chromatogram (TIC) and extracted ion chromatograms (EICs) from an online mixed-bed IEC separation of a HEPES extract from human heart tissue. EICs are scaled to the same height to show peak shape. b) Identified proteins from labeled EICs including fitted theoretical isotope distributions, monoisotopic mass error, and example fragment ions. Due to the lack of isotopic resolution, the deconvoluted average mass profile is shown for the malate dehydrogenase dimer. The PDBs associated with the representative crystal structures are 1AZV (superoxide dismutase), 1BD9 (PE-binding protein), 1K8U (S100-A6), and 2C95 (adenylate kinase). The full charge state distributions for each species are shown in Figure S6.

Since the salt concentration of a solution can impact the folding of a protein and the strength of its noncovalent interactions,^63–65 we suspected that a starting condition of 10 mM ammonium acetate was too low and may be affecting protein complex stability. Therefore, we conducted an additional experiment where the protein extract was exchanged into a starting condition of 50 mM ammonium acetate and the gradient was shifted to run from 50-500 mM ammonium acetate. We observed limited differences in the relative abundance of the detected dimers and metal-bound species between these conditions and found that the higher initial salt concentration had a negative impact on the separation (Figure S8). However, the use of a higher initial salt concentration was found to be beneficial when analyzing native extracts of proteins with lower solubility. To demonstrate this, we performed a two-step protein extraction from human heart tissue, first extracting soluble proteins in HEPES followed by further extraction with a buffer containing high salt in the form of LiCl, a method previously reported by our group^8,9 (Supplementary Experimental Section, Figure S9). The online mixed-bed IEC-nTDP separation of the LiCl extract exchanged into 50 mM ammonium acetate identified more species including the calcium handling protein calmodulin and detected various unknown species up to 76 kDa (Figure S10). Further experiments are necessary to find a balance between the gradient conditions and initial sample salt concentration for these lower solubility extracts. Although this method can be applied to samples more complex than those traditionally analyzed by nTDMS, multidimensional chromatography will likely be necessary to achieve deep coverage of nondenatured proteins for nTDP.

Conclusions

We report the nTDP analysis of complex protein mixtures using a novel online mixed-bed IEC method. We found that the use of a mixed-bed stationary phase allows the protein surface charge specific retention over a wide range of isoelectric points without altering the pH of the mobile phase. Experiments with standard proteins demonstrated the ability of online IEC to inform and complement nTDP results. Application to endogenous protein complexes from human heart tissue identified metal-binding and oligomeric proteins, with proteins detected up to 146 kDa. We expect this method to be an important tool enabling direct structural characterization of endogenous proteoforms from complex biological mixtures by nTDP.

Data Availability

The mass spectrometry proteomics data generated in this study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository under the accession code PXD052354 and MassIVE repository under the accession code MSV000094787.