Small-Scale Serial Size Exclusion Chromatography (s³SEC) for High Sensitivity Top-Down Proteomics of Large Proteoforms

Abstract

Top-down-mass spectrometry (MS) based proteomics has emerged as a premier technology to examine proteins at the proteoform level, enabling characterization of genetic mutations, alternative splicing, and post-translational modifications. However, significant challenges that remain in top-down proteomics include the analysis of large proteoforms and the sensitivity required to examine proteoforms from minimal amounts of sample. To address these challenges, we have developed a new method termed “small-scale serial Size Exclusion Chromatography” (s³SEC), which incorporates a small-scale protein extraction (1 mg of tissue), serial SEC without postfractionation sample handling, coupled with online high sensitivity capillary reversed-phase liquid chromatography tandem MS (RPLC-MS/MS) for analysis of large proteoforms. The s³SEC-RPLC-MS/MS method significantly enhanced the sensitivity and reduced the proteome complexity across the fractions, enabling the detection of high MW proteoforms previously undetected in one-dimensional (1D)-RPLC analysis. Importantly, we observed a drastic improvement in the signal intensity of high MW proteoforms in early fractions when using the s³SEC-RPLC method. Moreover, we demonstrate that this s³SEC-RPLC-MS/MS method also allows the analysis of lower MW proteoforms in subsequent fractions without significant alteration in proteoform abundance and equivalent or improved fragmentation efficiency to that of the 1D-RPLC approach. Although this study focuses on the use of cardiac tissue, the s³SEC-RPLC-MS/MS method could be broadly applicable to other systems with limited sample inputs.

Graphical Abstract

Top-down mass spectrometry (MS)-based proteomics has become a premier technology enabling the characterization of proteoforms that arise from genetic mutations, alternative splicing, and post-translational modifications (PTMs).^1–4 However, two major challenges facing the field are the analysis of high molecular weight (MW) proteoforms and the sensitivity required to examine proteoforms from minimal amounts of sample.⁵ The difficulty in effectively studying large proteoforms stems from significant analytical limitations, with the predominant obstacle being the exponential decrease in the signal-to-noise ratio (S/N) as a function of increasing MW, which can be attributed to the dispersion of the signal across several charge states and isotopic distributions.⁶ In complex mixtures, the overall reduction of the signal is exacerbated by the coelution of smaller proteins which further suppresses the signal of the larger proteins.⁵

To overcome these limitations, size-based separations have typically been employed to isolate the high MW species from the low MW species.^7–16 Although gel-based techniques^7,8 provide good separation based on MW, the use of sodium dodecyl sulfate (SDS) requires significant sample cleanup prior to MS analysis, resulting in reduced throughput and protein recovery.^17,18 We have previously shown the advantages of using liquid chromatography, specifically size exclusion chromatography (SEC), as a suitable alternative for size-based intact protein separation.^9–12 Having the flexibility to use a wide range of MS-compatible mobile phases without the need for surfactant removal is highly advantageous when coupling offline fractionation to direct infusion or a second dimension of separation.^10,11 However, SEC using wider column diameters (i.e. 9.4 mm i.d.) has limitations, including large sample requirements and postfractionation sample handling (sample pooling and/or concentration), preventing its application to sample-limited systems.

To address these challenges, we set out to develop a method, termed small-scale serial size exclusion chromatography (s³SEC), that enables size-based separation of proteoforms extracted from minimal amounts of sample, without additional postfractionation sample handling. s³SEC implements a small-scale protein extraction and offline size-based fractionation before coupling to high sensitivity capillary reversed-phase liquid chromatography tandem MS (RPLC-MS/MS) analysis (Figure 1). To convey the scale of this approach, 1 mg of tissue was compared to a nickel for a visual representation (Figure S1). In this method, proteins were extracted from approximately 1 mg of human cardiac tissue using a previously established two-stage extraction that was modified for the reduced sample consumption.^19,20 Cytosolic proteins were first depleted using a HEPES buffer (pH = 7.4) after which the tissue pellet was homogenized in TFA buffer (pH = 2.0) to enrich for sarcomeric proteins. Following protein normalization, 20 μg of total protein – a 15-fold reduction from our previous work¹⁰ – was injected onto two serially connected 2.1 mm i.d. PolyHYDROXYETHYL A (PolyHEA) columns (1000 Å/300 Å) from PolyLC Inc. for size-based separation. Following UV detection, one-minute fractions were collected into LC-MS vials and approximately 150 ng of each fraction was then analyzed using high sensitivity capillary RPLC-MS/MS on a Bruker maXis II modified with a Newomics microflow multiemitter nanoelectrospray (MnESI) source.^21,22 DataAnalysis (ver. 4.3; Bruker Daltonics) software was used to process and analyze the LC-MS raw files. Tandem mass spectra were exported from DataAnalysis software and analyzed using MASH Native (ver. 1.1).²³ Detailed sample preparation procedures and instrument parameters can be found in the Supporting Information.

Figure 1.

Schematic of the s³SEC-RPLC-MS/MS method from a small amount of tissue for top-down proteomic analysis.

To accommodate loading amounts amenable to protein extractions from minimal amounts of tissue, we chose to perform the s³SEC fractionation using 2.1 mm i.d. columns. Although previous work has demonstrated the effective use of smaller i.d. SEC columns for protein characterization of simple mixtures,^24,25 2.1 mm i.d. columns provide a strong balance between separation efficiency and sample consumption,²⁶ which will be critical for complex mixture analysis. Independently, the 1000 Å column enabled more effective separation of the high MW species (Figure S2B) compared to the 300 Å column (Figure S2A). Upon the serial connection of the 1000 Å and 300 Å columns, the total column length containing two different pore sizes provided additional partitioning of large proteins in fractions 1–4 while also enabling more separation in the lower MW region in fractions 5–12 (Figure S2C). It is important to note that by decreasing the inner column diameter to 2.1 mm, the separation efficiency of the s³SEC separation was reduced compared to the previous studies that utilized 9.4 mm i.d. columns.¹⁰ However, this result was expected due to the reduced column volume of the 2.1 mm i.d. columns and increased impact of extra-column dispersion.²⁶ With column volumes being scaled by the ratio of their squared inner diameters, there is roughly 5% of the packing material in the 2.1 mm i.d. columns compared to the 9.4 mm i.d. columns. Despite significantly less pore volume for separation to occur, the serial setup sufficiently and reproducibly separated high MW proteins from low MW proteins across multiple biological replicates (Figure 2A, Figure S3).

Figure 2. s³SEC-RPLC enables analysis of high molecular weight proteoforms not detected in the traditional 1D-RPLC approach.

(A) Representative UV chromatogram and SDS-PAGE gel of sarcomeric extract following offline s³SEC fractionation. Equal volumes (10 μL) of unnormalized fractions were loaded onto the gel. L, ladder; F#, fraction number. (B) Representative TICs of 1D-RPLC and s³SEC-RPLC-MS analysis coupling offline s³SEC fractionation with RPLC-MS. (C) TIC overlay from the 1D-RPLC separation (black) and fraction 1 (F1, red) of the s³SEC-RPLC separation. (D) MS spectra from averaged TIC regions of 1D (black) and F1 (red). 1D spectrum shows cardiac and skeletal actin (CAA/SKA) proteoforms while F1 shows MHC (223.1 kDa). (E) TIC overlay from 1D-RPLC separation (black) and fraction 2 (F2, blue) of the s³SEC-RPLC separation. (F) MS spectra from averaged TIC region of 1D (black) and F2 (blue). 1D spectrum shows myosin light chain (MLC-2v) proteoforms, while F2 shows a 154.8 kDa proteoform.

Operating at a lower flow rate (29 μL/min) that is compatible with the smaller column diameters proved beneficial in reducing postfractionation sample handling. By collecting 1 min fractions directly in LC-MS vials, there was no additional sample transfer step needed prior to MS analysis. Importantly, the concentrations of each fraction (Figure S3) were sufficient for our high-sensitivity RPLC-MS/MS setup using a capillary column and the Newomics MnESI source.²⁷ Representative total ion chromatograms (TICs) corresponding to each fraction (F#) and the traditional one-dimensional (1D)-RPLC analysis displayed unique profiles, suggesting the presence of different proteins in each fraction and further validating the success of offline fractionation (Figure 2B). Additionally, the chromatographic profiles of each fraction were consistent across different biological replicates, demonstrating the reproducibility of both the offline fractionation and the online RPLC-MS analysis (Figure S4).

With the goal of applying s³SEC-RPLC to detect and analyze large proteoforms, we first focused on the early fractions (F1–F4) where we detected the presence of distinct bands in the high MW range with minimal presence of lower MW species (Figure 2A). In comparing the 1D-RPLC TIC (black) with the F1 TIC (red), we observed a clear profile change in the chromatogram from 32.2–32.8 min (Figure 2C). After averaging across that specific elution window in the 1D-RPLC analysis, the m/z and deconvoluted spectra corresponded to cardiac and skeletal alpha-actin, highly abundant 41.8 kDa sarcomere proteins (Figure 2D, black). However, when averaging over that same elution window in the F1 sample, a charge state distribution characteristic of a large protein was present instead (Figure 2D, red). Deconvolution of the spectrum and the use of the Sum Peak algorithm determined the most abundant mass to be approximately 223.1 kDa. The observed mass corresponds to the theoretical MW of myosin heavy chain (MHC), a prominent thick filament protein in the sarcomere.^28,29 Although precursor ions were selected for fragmentation, the high charge and resulting low S/N prevented the detection of resolved tandem MS fragments given the small amount of sample used here. Thus, we validated the raw m/z spectrum to ensure the charge state distribution is consistent with the MW of MHC, similar to what was done in a previous study.²⁷ Furthermore, the detection of MHC in F1 is consistently observed in similar elution windows across six different biological replicates (Figure S5). In F2 (blue), we noticed an additional chromatographic profile alteration during the 28.1–28.7 min elution window (Figure 2E). Rather than detecting an abundant 18.7 kDa isoform of ventricular myosin light chain (Figure 2F, black), we were able to detect a charge state envelope corresponding to a previously undetected 154.8 kDa proteoform (Figure 2F, blue). Together, these results demonstrate that with the separation provided by the offline fractionation, our s³SEC-RPLC-MS/MS method enabled the detection of high MW proteins that were not detected in the traditional 1D-RPLC approach.

In addition to the detection of previously unobserved high MW proteins, the s³SEC-RPLC method is also advantageous for improving the S/N of large proteoforms that were, in some capacity, detected using the 1D-RPLC approach. In F3, there are two proteins in particular where we observe significant improvement in their detection as a result of reduced coelution with smaller proteins (Figure 3). When averaging across the 34.7–35.0 min elution window (Figure 3A), both the 1D and F3 analyses were able to detect a 68.4 kDa proteoform. However, the mass spectrum in the 1D-RPLC analysis showed significantly more ion suppression of the larger protein due to the increased presence of troponin C (Figure 3B). Although troponin C is still present during the same elution window for F3, there was sufficient size-based separation to result in over an 11-fold increase in the deconvoluted intensity of the 68.4 kDa proteoform (3.1E5) compared to the intensity observed with 1D-RPLC (2.6E4) (Figure S6A). A similar trend is observed in the 35.0–35.5 min elution window (Figure 3A) where coelution of troponin C, an ATP synthase subunit and multiple other species significantly suppresses the signal of a 71.4 kDa proteoform (Figure 3C). In this case the deconvoluted intensity increased over 8-fold in the F3 spectrum (4.5E4) compared to the 1D-RPLC spectrum (5.2E3) (Figure S6B). The increased signal and minimized coelution attributable to the reduced proteome complexity highlight another benefit in using s³SEC-RPLC-MS. Moreover, the 71.4 kDa species discussed above was the largest MW proteoform observed in the 1D-RPLC approach. The s³SEC-RPLC-MS approach, which enabled the detection of a 223.1 kDa protein, effectively increased the observable MW range over 3-fold.

Figure 3. Increased detection of high molecular weight proteoforms.

(A) TIC overlay from 1D-RPLC separation (black) and fraction 3 (F3, blue) of the s³SEC-RPLC separation. (B) MS spectra from the averaged TIC region of 1D (black) and F3 (blue). F3 shows significantly less coelution of troponin C (TnC), enabling better detection of a 68.4 kDa proteoform. (C) MS spectra from averaged TIC region of 1D (black) and F3 (blue). F3 shows significantly less coelution of multiple species including TnC and an ATP synthase subunit (AT5F1), enabling better detection of a 71.4 kDa proteoform.

Beyond examining the early fractions where the high MW proteoforms eluted, we also sought to demonstrate the capability of the s³SEC-RPLC method in capturing the proteoform landscapes of the latter fractions, where smaller proteins typically reside. Examining s³SEC F7 specifically, the TIC showed the elution of several proteins across the gradient (Figure S7A). High-resolution deconvoluted mass spectra of these elution windows enabled the identification of multiple sarcomeric, Z-disk, and nuclear proteins with a wide variety of PTMs including acetylation, phosphorylation and trimethylation (Figure S7B). To ensure there was no specificity in the way proteoforms fractionate, we compared the total phosphorylation levels (Ptotal) from ventricular myosin light chain-2 (Figure S8A), cardiac troponin I (Figure S8B), and alpha-tropomyosin (Figure S8C) in F6, F7, F8, F9, and 1D-RPLC analyses. Importantly, s³SEC fractionation showed no significant alteration in the relative abundance of these proteoforms across the four fractions or 1D-RPLC analysis. Thus, although proteins may elute across multiple fractions, it is valid to report the relative abundance of proteoforms using any particular fraction. For proteoforms where PTMs may result in a large increase in MW compared to unmodified forms, s³SEC fractionation windows can be altered and further optimized to better target a specific protein and its respective modifications.

For a top-down proteomics experiment, fragmentation efficiency is necessary for protein identification and to gain information on a protein’s sequence. Thus, for each fraction, untargeted data-dependent AutoMS/MS was performed. Across the s³SEC-RPLC analysis of different fractions, we observed fragmentation for a variety of proteins of different MWs and abundances. In F5 (Figure S9A), we observed fragmentation of desmin, a 53.4 kDa intermediate filament protein essential to cellular integrity and organization as it links sarcomere and Z-disk proteins to the mitochondria, sarcoplasmic reticulum, and cell membrane.³⁰ Likewise, in F7 and F9, we observed quality fragmentation of tropomyosin (Figure S9B) and cardiac troponin I (Figure S9C), two sarcomere proteins critical to cardiac contraction.²⁸ Importantly, we wanted to ensure that there was no reduction in fragmentation efficiency when comparing the 1D analysis and the s³SEC-RPLC method. To show this, we specifically investigated whether there were any differences in the fragmentation of cardiac troponin I or desmin in both approaches. For cardiac troponin I, while we observed minor variations in the specific ions produced for the 1D-RPLC (Figure S10A) and F9 (Figure S10B) fragmentation, there was no difference in the overall sequence coverage of the protein between methods. In desmin, however, we observed an increase in the overall sequence coverage in the s³SEC-RPLC method (Figure S10D) compared to 1D-RPLC (Figure S10C). Therefore, in addition to improved detection of high MW proteins provided by the earlier fractions, the s³SEC-RPLC-MS/MS method enables an equivalent or improved fragmentation efficiency in the later fractions, allowing MS coverage across a wider MW range. It is important to note that because this fragmentation was untargeted, a higher overall sequence coverage could be achieved by conducting a more targeted tandem-MS approach with isolation of specific precursor ions and tuned collision energy. Despite the improved detection of high MW proteoforms, effective fragmentation of proteins greater than 60 kDa proved difficult. The combination of low S/N for selected precursor ions and elution occurring on a LC time-scale, resulted in minimal fragmentation of high MW proteoforms. This result was not surprising as fragmentation of large proteoforms is widely recognized as a limitation of current instrumentation and top-down proteomic analysis.^6,31

In summary, we have developed an s³SEC-RPLC-MS/MS method that enables the detection of high MW proteoforms in a complex mixture by providing additional size-based separation not observed in the traditional 1D-RPLC approach. Notably, the method uses a high-sensitivity setup that allows for significantly less sample usage than previous approaches, while preventing the need for any additional sample handling postfractionation. This method also enables the detection of low MW proteins in later fractions, retaining the ability to do relative quantification of different proteoforms. We envision that the s³SEC-RPLC method will be generally applicable to other experiments requiring high MW proteoform detection and mobile-phase flexibility from sample-limited sources, including but not limited to human biopsy tissue, minimal amounts of cells, and extracellular vesicles.

Supplementary Material

Supplementary Information

Materials and Methods.

Supplementary Figure S1. Visual representation of sample requirements for s³SEC-RPLC method.

Supplementary Figure S2. Evaluation of individual and serially connected 2.1 mm SEC columns.

Supplementary Figure S3. Reproducibility of offline s³SEC fraction.

Supplementary Figure S4. Reproducibility of online RPLC-MS analysis of individual s³SEC fractions.

Supplementary Figure S5. Detection of myosin heavy chain across six biological replicates.

Supplementary Figure S6. Reduced coelution in s³SEC fraction enables increased detection of high MW proteins.

Supplementary Figure S7. Diverse proteoform landscape detected in s³SEC fraction 7.

Supplementary Figure S8. Comparison of 1D-RPLC-MS and s³SEC-RPLC-MS Ptotal analysis.

Supplementary Figure S9. Top-down proteomic analysis of proteins across different s³SEC fractions.

Supplementary Figure S10. Comparison of 1D-RPLC and s³SEC-RPLC fragmentation efficiency.

Data Availability Statement:

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD#047680 and the MassIVE repository with identifier MSV#000093625.