Optimization of a High-Throughput Shotgun Immunoproteomics Pipeline for Antigen Identification

Abstract

Using a systematic, hypothesis-driven approach, we identified potential improvements for three individual steps of a previously published approach for antigen-identification. Optimization of each step created a methodology which resolved many of the persistent issues associated with previous antigen identification approaches. The optimized high-throughput shotgun immunoproteomics approach described herein identifies more than five times as many unique antigens as the previously published method, greatly reduces protocol cost and mass spectrometry time per experiment, minimizes both inter- and intra-experimental variability, and ensures each experiment is fully quantitative. Ultimately, this optimized antigen identification approach has the potential to facilitate novel antigen identification studies, allowing evaluation of the adaptive immune response in a longitudinal manner and encourage innovations in a wide array of fields.

Graphical Abstract

1. Introduction

Identification of the antigenic determinants responsible for adaptive immune activation has broad application to enhance the understanding of auto- (e.g., autoimmune disease, cancer biology), allo- (e.g., solid organ transplantation) and xeno- (e.g., animal-derived biomaterials and organs) antigenic immune responses in health and disease. [1–6] To date, studies investigating the drivers of adaptive immune system activation have generally focused on cell-mediated approaches (T-cell epitope mapping) versus humoral approaches (B-cell antibody production). [1–7] T-cell epitope mapping methodologies rely on the cellular processing of antigenic proteins into peptides, which are then presented in the binding site of the major histocompatibility complex (MHC). [5, 7–9] This requirement for protein processing and presentation of resultant peptides from the original protein on MHC presents a major barrier to determining the identity of the original parent protein in T-cell epitope mapping studies. [7–9] The major advantage of humoral approaches to antigen identification is that antibodies recognize intact proteins, thus allowing for direct identification of the antigenic protein of interest. [1, 2, 6] Furthermore, restriction of such searches to IgG interactions ensures that T-helper cell help has occurred during the process of B-cell activation and isotype switching, allowing for the inference that both cell-mediated and humoral responses are involved in the observed antigen-specific response. [1–4] Consequently, approaches employing the humoral immune response have become the mainstay for identification of antigenic proteins in a wide range of preclinical and clinical applications. [3, 4, 6]

Early attempts at antigen identification typically employed large format 2-dimensional electrophoresis western blots (2-DE-WB), followed by MALDI-TOF identification of selected protein spots. [1, 10–13] While this workflow yielded results, numerous issues associated with this method were identified, including, but not limited to, a high degree of variability between replicate gels, low number of antigens identified per experiment, labor intensive workflows, and high costs associated with each experiment. [1, 10–13] Furthermore, the challenging solubility profile of highly lipophilic proteins, particularly during isoelectric focusing, further complicated identification of certain antigens (e.g., integral membrane proteins). [1,14]

Alternative strategies for antigen identification have been developed, such as serological screening of peptide microarrays, serological analysis of recombinant cDNA expression libraries (SEREX), and immunoprecipitation techniques. [15–17] However, as reviewed elsewhere, in addition to the specific limitations of each method (e.g., lack of tissue specificity and underrepresentation of membrane proteins with microarrays, incomplete reproduction of protein tertiary structure with SEREX), such methods are highly time consuming and costly. [6] Due to the numerous challenges of such approaches, especially for high-throughput discovery applications, our group identified a critical unmet need in the field and set out to develop an alternative methodology for antigen identification.

To overcome the limitations of previous antigen identification strategies, we developed an affinity chromatography approach in which poly-polyclonal (i.e., polyclonal antibodies toward multiple antigens) IgG molecules are crosslinked to sepharose beads and used to capture antigenic proteins from solubilized protein mixtures. [2] Although LC-MS/MS analysis of antigens captured in such columns greatly increased the number of antigens identified in each experiment and allowed for the identification of integral membrane antigens, other limitations of previously reported antigen identification approaches remained. [2] Specifically, the affinity chromatography antigen identification approach relied upon semi-quantitative MS methods (e.g., spectral counts) and required separate MS runs for each sample, increasing the variability, MS time, and cost of antigen identification experiments. [2] Consequently, there exists a critical need to develop a high-throughput, highly reproducible, quantitative, and cost-effective antigen identification pipeline to facilitate discovery and longitudinal monitoring of antigen-specific responses toward tumor antigens, implanted allografts, and xenogeneic biomaterials, among other applications.

In this study, we sought to optimize our previously published affinity-chromatography antigen identification approach, to maximize antigen capture and subsequent elution from the column, reduce inter- and intra-sample variability, and thereby maximize the number of antigens detected in a rabbit model of xenogeneic tissue implantation. We hypothesized that protein sample solubilization conditions, antigen elution conditions, and proteomic sample preparation approach would all affect resultant antigen identification efficiency (Fig. 1). Furthermore, we hypothesized that despite the small volume (i.e., 200 μl) and relatively low levels of antigens contained in column eluates (i.e., ~1–5 μg), tandem mass tag (TMT) labeling would allow longitudinal quantification of antigen-specific immune responses toward a clinically relevant biomaterial (i.e., bovine pericardium (BP)) following unadjuvanted implantation in a small-scale proof-of-concept rabbit model.

2. Materials and Methods

2.1. Rabbit Serum Generation

Polyclonal anti-native BP serum was generated as previously described. [18] Briefly, New Zealand white rabbits (n=1) were immunized by subcutaneous injection with 1 ml of a 1:1 mixture of homogenized native BP and Freund’s complete adjuvant at Day 0, followed by 1:1 mixture of homogenized native BP and incomplete Freund’s adjuvant on Days 14, 28, and 42. Blood was obtained prior to inoculation (D0) and at Day 84. Following collection, blood samples were centrifuged at 3000 g and the anti-native BP serum was stored at −80°C for further use.

For longitudinal antigen response experiments, 1 × 1 cm pieces (~200 mg) of native BP were implanted in the subpannicular region of New Zealand white rabbits (n=2). Blood samples were obtained prior to implantation (D0) and at days 7, 14, 28, 42, and 56 post-implantation.

2.2. Native Bovine Pericardium Protein Extraction

2.2.1. Tissue Harvest

BP from young adult cattle was obtained (Spear Products, Coopersburg, PA) and processed to remove any residual epicardial fat and/or connective tissue. Cleaned BP was cut into 1×16 cm circumferential strips and stored at −80°C in Dulbecco’s Modified Eagles Medium (DMEM) with 15% (v/v) dimethyl sulfoxide (DMSO), as previously described. [18]

2.2.2. Tissue Disruption

Previously frozen native BP strips were thawed and subsequently cut into pieces approximately 1×1 cm in size, with a mass ~0.2 g each. Prior to protein extraction, cut BP pieces were either manually minced or pulverized using a Cellcrusher (Schull Co, Cork, Ireland). For the manual mincing process, BP pieces were laid flat and snap-frozen in a liquid nitrogen bath. Once frozen, each piece of BP was retrieved from the liquid nitrogen and minced into ~1 × 1 mm chunks with Metzenbaum scissors. Alternatively, BP pieces pulverized using the Cellcrusher device were homogenized according to the manufacturer’s directions. Briefly, the entire Cellcrusher was first submerged in a liquid nitrogen bath and allowed to equilibrate to temperature. Once the device was chilled, the cut BP piece was added to the mortar portion of the Cellcrusher and flash-frozen. Following the flash freezing step, the pestle portion was positioned, and the device was struck with a mallet to ensure complete homogenization of the tissue. Both the manually minced chunks and pulverized BP powder from the Cellcrusher were immediately transferred to a microcentrifuge tube for protein extraction.

2.2.3. Protein Extraction

Following tissue disruption, proteins were extracted from homogenized BP via either a one-step (IP Buffer) or two-step (Hydrophile & Lipophile Buffers) protocol. The IP buffer utilized for the one-step extraction contained 0.25% (w/v) sodium deoxycholate (Sigma-Aldrich, St Louis, MO), 1% (w/v) octyl-β-D-glucopyranoside (Sigma-Aldrich, St Louis, MO), 1 mM ethylenediaminetetraacetic acid (EDTA) (pH 8) (ThermoFisher Scientific, San Francisco, CA), 1 mM phenylmethylsulfonylfluoride (PMSF) (Sigma-Aldrich, St Louis, MO), & 1x Halt Protease Inhibitor Cocktail (ThermoFisher Scientific, San Francisco, CA). [19] Alternatively, the hydrophile and lipophile extraction buffers employed during the two-step extraction protocol contained either 134 mM NDSB-256 (Hydrophile buffer) or both 134 mM NDSB-256 (Sigma-Aldrich, St Louis, MO) and 1% (w/v) n-dodecyl-β-D-maltoside (Sigma-Aldrich, St Louis, MO) (Lipophile buffer) in a standard extraction buffer (10 mM tris-HCl (pH 8.0), 1 mM dithiothreitol (DTT), 2 mM MgCl₂-6H₂O, 10 mM KCl, 0.5 mM Pefabloc (Sigma-Aldrich, St Louis, MO), respectively, as previously described. [14] To begin the extraction process, homogenized BP was incubated at 4°C in 1 mL of extraction buffer (either IP or hydrophile) for 1 hr with constant mixing at 1,400 rpm on a thermomixer (Eppendorf, Hamburg, Germany). Following mixing, samples were centrifuged at 17,000 g for 25 min at 4°C to pellet any insoluble material from the solution. For both protocols, the supernatant was subsequently collected and stored at −80°C, with the extracts collected using the two-step protocol being designated as the hydrophile protein extract and extracts collected using the IP protocol designated as the IP extract. During the two-step protocol, upon collection of the first supernatant, the insoluble pellet of samples was resuspended and washed with hydrophile extraction buffer twice, and then incubated in 0.5 mL of lipophile extraction buffer for 1 hr with constant mixing at 1,400 rpm, 4°C. Samples were again centrifuged at 17,000 g for 25 min at 4°C to pellet any insoluble material from the solution, with the resulting supernatant subsequently being collected, designated as the lipophile protein extract, and stored at −80°C.

2.3. One-dimensional Electrophoresis & Gel Analysis

All protein gel analyses were performed using one-dimensional electrophoresis with the Invitrogen NuPAGE system. For all experiments, an equal volume of protein solution was loaded into each well of a 4–12% Bis-TRIS NuPage gel as a standardized loading control. Following loading, samples were electrophoresed at 200 V for 55 min. All gels & western blots were imaged using a FluorChem M imaging system (BioTechne, Minneapolis, MN) and the Optical Density (OD) of each lane was subsequently quantified using ImageJ (National Institutes of Health, Bethesda, MD) with all lanes corrected for background intensity.

2.3.1. Coomassie Gel Staining & Quantification

Following electrophoresis, gels designated for total protein analysis were stained using BioSafe Coomassie G-250 stain (Bio-Rad, Hercules, CA) according to the manufacturer’s instructions. For any individual experiment, all Coomassie-stained gels were imaged at the same exposure time.

2.3.2. Western Blotting & Quantification

Following electrophoresis, proteins from gels designated for western blotting were transferred to nitrocellulose membranes using a wet transfer system and probed, as previously described. [18] Briefly, gel sandwiches were assembled, submerged in 1x NuPAGE Transfer buffer (NuPAGE Transfer Buffer (Invitrogen, Waltham, MA), 0.1% v/v NuPAGE antioxidant (Invitrogen, Waltham, MA), 20% v/v methanol), and transfers were subsequently performed at 30 V for 3 hrs. Upon the completion of the protein transfer, membranes were initially blocked with Pierce Protein-free Blocking Buffer (ThermoFisher Scientific, San Francisco, CA) + 0.5% v/v 10% Tween-20 (Biorad, Hercules, CA). Following blocking, membranes were stained with a 1:100 dilution of rabbit Day 84 polyclonal anti-native BP serum as the primary antibody. IgG-specific binding was visualized via HRP-conjugated goat anti-rabbit IgG Fc fragment secondary antibody (1:5,000) (Jackson ImmunoResearch, West Grove, PA).

2.4. Affinity Chromatography

2.4.1. Column Preparation

Rabbit poly-polyclonal IgG affinity chromatography columns were created using Protein A HP SpinTrap columns (GE Healthcare, Chicago, IL) according to the manufacturer’s specified protocol, as previously described. [2] In brief, per column, serum from rabbits immunized with native BP (D84) or a pooled baseline from unvaccinated rabbits (n=6) (D0) was first diluted 1:10 in binding buffer (50 mM Tris, 150 mM NaCl (pH 7.5)). For rabbit longitudinal non-immunized experiments, separate columns were formed using serum (D0, D7, D14, D28, D42, and D56) from rabbits implanted with native BP. Following column equilibration, 200 μL of the diluted rabbit serum was added to the column and the system was mixed with end-over-end rotation for 30 min. Columns were subsequently washed with 400 μL binding buffer and 400 μL of triethanolamine (200 mM) (pH 8.9). Antibodies were cross-linked to Protein A by incubating 400 μL of dimethyl pimelimidate dihydrochloride (DMP) (50 mM) in triethanolamine (200 mM) in the column for 1 hr with end-over-end rotation. Following crosslinking and an additional wash, columns were blocked with 400 μL of ethanolamine (100 mM (pH 8.9)) for 15 min, after which unbound antibodies were eluted from the column with 200 μL of pH 2.9 elution buffer (0.1 M glycine and 2 M urea (pH 2.9)). Prior to protein extract application/antigen capture, columns were washed twice with binding buffer to re-equilibrate the column to neutral pH. Depending on the experiment being conducted, columns were next exposed to 200 μL of protein extract prepared from native BP (IP, hydrophile, or lipophile protein extract), as described above for 1 hr with end-over-end rotation. Columns were washed a total of five times with wash buffer (50 mM Tris, 150 mM NaCl, 2 M urea (pH 7.5)) prior to elution of captured antigens.

2.4.2. Elution of Captured Antigens

After the antigen capture and column wash steps, bound antigenic proteins were then eluted off the columns via a series of sequential step-down elution steps. All columns were initially washed with 200 μL of glycine-urea elution buffer (pH 5) (0.1 M glycine & 2 M urea), followed by 200 μL of glycine-urea elution buffer (pH 4) (0.1 M glycine & 2 M urea) to remove non-specifically bound proteins. According to the individual experiment, antigenic proteins were eluted from columns with either 200 μL of glycine-urea elution buffer (pH 2.9) (0.1 M glycine & 2 M urea) or 1% trifluoracetic acid (TFA). All elution samples were collected in tubes containing 15 μL of neutralizing buffer (1 M Tris-HCl (pH 9)) and stored at −80°C for downstream analysis.

2.5. LC-MS/MS & Proteomic Analysis

2.5.1. Sample Preparation/Digestion

Proteins were reduced by incubating samples in 5 mM TCEP for 1 hour at room temperature and further alkylated using 20 mM iodoacetamide for 30 min in dark. Eluates were then diluted 2-fold with 100 mM TEABC buffer and incubated with trypsin enzyme overnight in thermomixer at 37 °C. Trypsin digestion was stopped by acidifying the digest and peptides were cleaned using C₁₈ stage tips.

2.5.2. TMT Labeling

Peptides were then labeled with TMT isobaric reagents (Thermo Fisher Scientific, Waltham, MA) as per the manufacturer’s instructions. Labeling was quenched using 5% hydroxylamine. A small aliquot of the labelled peptides was pooled from all the samples and analyzed by LC-MS/MS to determine the normalized factors. For normalized experiments, peptides were pooled as per the normalization factors to account for the protein loading differences between samples. Alternatively, in the volume-normalized experiments, peptides from differing TMT-labelled samples were pooled using equal proportions of the eluted sample volume from each affinity chromatography column. Pooled peptides were further cleaned using C₁₈ stage tips, vacuum dried, and subjected to LC-MS/MS analysis.

2.5.3. LC-MS/MS Analysis

Peptides were resuspended in 0.1% formic acid and subjected to LC-MS/MS analysis on Orbitrap Fusion Lumos connected to Ultimate 3000 RSLC nano system or Orbitrap Eclipse Tribrid mass spectrometer connected to Vanquish Neo UHPLC system (Thermo Fisher Scientific, Waltham, MA). To enable peptide separation, two different columns were used: initial trapping column PepMap C₁₈, 2 cm× 100 μm, 100 Å, (Thermo Fisher Scientific, Waltham, MA) and an analytical column (PepMap RSLC C₁₈ 2 μm, 75 μm × 50 cm, 100 Å, (Thermo Fisher Scientific, Waltham, MA)). Peptides were first loaded on a trap column using solvent A (0.1% formic acid) and then separated on analytical column using a gradient from 3% to 40% solvent B (100% acetonitrile, 0.1% formic acid). Both columns were equilibrated with solvent A for 5 min before loading the next sample. Precursor ions were analyzed in a data-dependent manner with cycles of survey MS scan followed by MS/MS. In MS scan, precursor ions with mass range of 350–1500 m/z were analyzed by Orbitrap at 120,000 resolution, normalized AGC target of 250%, and 50 s injection time. After the MS scan, top intense ions were sequentially filtered by quadrupole with 1.6 m/z isolation width and fragmented using 34% normalized high-energy collision-induced dissociation (HCD) energy.

Fragmented ions were analyzed in Orbitrap analyzer as MS/MS spectra using 30,000 resolution, normalized AGC target of 250%, and injection time of 100 ms. Precursor ions selected for MS/MS were filtered for a minimum intensity threshold of 10,000, charge state of 2–6, and excluded for repeated MS/MS analysis for 40 s. A cycle time of 2 sec was used for MS followed by MS/MS scans.

2.5.4. Database Searching

Raw files were analyzed using Proteome Discoverer software version 2.5 (Thermo Fisher Scientific, Waltham, MA). MS/MS spectra were extracted from the raw files and searched against Bovine and Rabbit UniProt protein database using Sequest search engine with full tryptic cleavage specificity, 2 missed cleavages, precursor and fragment ion tolerance of 10 ppm and 0.05 Da respectively. Contaminating rabbit proteins were subsequently excluded from downstream statistical analyses. Oxidation of methionine as dynamic modifications and TMTpro at peptide N-terminus, lysine, and carbamidomethylation of cysteine were specified as fixed modifications. False discovery rate was maintained at 1% at peptide and protein levels using Percolator. TMT channel intensities were calculated using an integration tolerance of 20 ppm. Label-free quantitation was performed using Minora feature detector node and iBAQ values were calculated from the protein intensities.

2.5.5. Antigen Identifications and statistical analysis

Antigens were defined as those proteins for which LC-MS/MS abundance increased by ≥ 2-fold in day 84 eluates compared to day 0 eluates, with a p-value ≤ 0.05. Volcano plots of antigenicity were constructed by plotting log2 fold change (Day84/Day0) against −log10 of the p-value for each protein. Comparisons between groups were made using one-way or two-way ANOVA depending on the experimental design and Tukey post-hoc testing employed where relevant, with significance set at p = 0.05.

3. Results

3.1. Overview

Fig. 1. Schematic overview of the high-throughput shotgun immunoproteomic antigen identification pipeline.

Red boxes indicate the three steps in the protocol (protein extraction, antigen elution, and LC-MS/MS analysis) which were targeted for optimization.

3.2. A one-step protein extraction method increases the yield and diversity of native bovine pericardium protein extracts

In order to optimize the protein solubilization step of the affinity chromatography approach, various protein isolation methods (extraction buffers and tissue homogenization methods) were assessed in a full factorial experiment (n = 7/group; minced vs. pulverized, 1-step IP vs. 2-step HES&LES buffers) for their ability to increase the number and diversity of proteins and antigens solubilized from native BP. Quantification of protein yield for each extraction showed that 1-step IP resulted in a significantly greater percent yield compared to the LES step for minced samples (p ≤ 0.0001). Similarly, using 1-step IP buffer significantly increased protein yield for crushed samples compared to either of the 2-step HES & LES buffers (p ≤ 0.05, p ≤ 0.0001 respectively). However, tissue disruption method did not significantly alter the total protein yield regardless of extraction buffer (p > 0.05) (Fig. 2A). Visualization of protein profiles of SDS-PAGE gels loaded with equal volumes of each extract qualitatively indicated a similar trend, with 1-step IP buffer samples containing more total protein than samples prepared with either the HES or the LES buffers (Fig 2B), which was confirmed by quantitative densitometry analysis of total band intensity (Fig. 2C) (p ≤ 0.001, p ≤ 0.0001 respectively). Additionally, 1-step IP protein extracts showed a greater number of higher molecular weight proteins present on SDS-PAGE gels than did either HES or LES extracts. Furthermore, LC-MS/MS analysis of protein extracts confirmed that the total number of proteins present in 1-step IP extracts was greater than that found in the combined 2-step HES & LES extraction approach (Fig. 2D). Assessment of antigen content, using 1-DE western blotting visualized with Day 84 immunized serum as the primary antibody, showed that 1-step IP extracts had greater diversity (e.g., more high molecular weight antigens present) and overall intensity of antigenic bands than did either HES or LES lanes from the 2-step protein extraction approach (Fig. 2E). Quantification of antigen content using densitometry of western blot images confirmed that 1-step IP resulted in significantly more antigens present in the resultant protein extracts than either of the 2-step fractions (HES p ≤ 0.0001 for minced, p ≤ 0.005 for crushed, LES p ≤ 0.0001 for minced, p ≤ 0.001 for crushed), with no significant differences found between the tissue disruption method (Fig. 2F) (p > 0.05).

Fig. 2. Optimization of the protein extraction increases total protein yield, antigen number, and diversity in BP extracts.

A.) Percent yield of protein extracted from pieces of native BP as determined via Detergent Compatible (DC) protein assay. B.) Representative Coomassie-stained gel loaded with equal volumes of total protein extracts prepared via various experimental conditions in each lane. C.) Densitometry quantification of total protein gels as measured via ImageJ analysis. D.) Venn diagram of all proteins identified via LC-MS/MS in protein extractions prepared using the 2-step extraction protocol and the optimized 1-step method. E.) Representative Western blot loaded with equal volumes of total protein extracts prepared via various experimental conditions. F.) Densitometry quantification of all antigens identified on western blots as measured via Image J analysis. * p ≤ 0.05, *** p ≤ 0.001, **** p ≤ 0.0001, ns=no statistically significant difference found

3.3. Optimization of antigen elution conditions from affinity chromatography columns reduces variability and improves the total number of antigens identified

Effect of differing elution conditions (i.e., glycine-urea (pH 2.9) versus 1% TFA elution (n = 7/group)) on the quantity and quality (i.e., variability) of antigenic proteins eluted from affinity chromatography columns was analyzed using 1-step IP protein extracts. Qualitatively, images of SDS-page total protein gels stained with Coomassie Blue demonstrated stronger band intensity in lanes loaded with 1% TFA eluates, than those from glycine-urea (pH 2.9) eluates (Fig. 3A). Quantification of total band intensity confirmed this observation, with 1% TFA samples having a significantly increased total lane intensity, as compared to the glycine-urea (pH 2.9) buffer group (Fig. 3B) (p ≤ 0.01). LC-MS/MS analysis of Day 84 post-immunization column eluates (n = 2/group) using either 1% TFA or glycine-urea (pH 2.9) elution buffers, demonstrated significant overlap among the proteins identified via either elution method (752 proteins), however 1% TFA eluates had a greater number of unique proteins identified (277) as compared to the glycine-urea (pH 2.9) eluates (77) (Fig. 3C). Similar trends were observed when we assessed antigenic protein content in the eluate samples via western blots. Specifically, representative images of western blots probed with Day 84 serum from immunized rabbits qualitatively showed a stronger banding pattern and more antigens in the high molecular weight region of the blot for eluates prepared using 1% TFA, as compared to equal volume of eluates prepared with glycine-urea (pH 2.9) elution buffer (Fig. 3D). Densitometry analysis of total lane intensity in the western blot images quantitatively confirmed that lanes loaded with 1% TFA eluates had significantly increased intensity, versus lanes loaded with an equal volume of glycine-urea (pH 2.9) buffer eluate samples (Fig. 3E) (p ≤ 0.0001). We further confirmed these findings on the effect of elution conditions on antigen identification using TMT-based quantitative proteomic analysis of day 0 and day 84 rabbit serum (i.e., >2-fold change). Although significant overlap in antigen identification was found regardless of elution method (222 unique antigens), the 1% TFA group was found to have a greater quantity and diversity (i.e., uniqueness) of identified antigens (344), compared to the glycine-urea (pH 2.9) elution buffer samples (141 unique antigens) (Fig. 3F). Further, the variability in the summed intensity of all identified proteins was significantly reduced in the Day 84 samples eluted with 1% TFA, compared to paired samples eluted with the glycine-urea (pH 2.9) buffer (Fig. 3G) (p ≤ 0.01). Additionally, when the coefficient of variability for each individual identified protein in each group was calculated, variability in 1% TFA eluates from both Day 0 and Day 84 columns was significantly lower compared to glycine-urea (pH 2.9) eluates at the corresponding time points (p ≤ 0.0001 for each time point).

Fig. 3. Elution of antigens from affinity chromatography columns using 1% TFA increases the quantity and diversity, while simultaneously reducing the variability, of antigens identified.

A.) Representative Coomassie-stained gel loaded with equal volumes of affinity chromatography eluate samples. B.) Densitometry quantification of total protein in affinity chromatography eluate gels as measured via ImageJ analysis. C.) Venn diagram of all identified proteins by LC-MS/MS analysis in 1% TFA and glycine eluates from day 84 affinity chromatography columns. D.) Representative Western Blot loaded with equal volumes of affinity chromatography eluate samples. E.) Densitometry quantification of antigens in western blots of affinity chromatography eluate samples as measured via Image J analysis. F.) Venn diagram of antigens identified in day 84 columns compared to day 0 by TMT-based quantitative proteomic analysis of affinity chromatography eluate samples prepared with various elution buffers. G.) Bar graph of the coefficient of variability for the total abundance of identified proteins (via LC-MS/MS analysis) across various elution buffer groups. H.) Violin plot of the coefficients of variability for all proteins identified in affinity chromatography elution samples prepared using various elution buffers. * p ≤ 0.05, ** p ≤ 0.005, **** p ≤ 0.0001, ns=no statistically significant difference found.

3.4. Sample loading conditions for TMT-labeling and multiplexed LC-MS/MS are critical to optimizing quantitative antigen identification

The effect of protein loading conditions on TMT-labelled LC-MS/MS proteomic quantification of antigens captured in affinity chromatography columns was assessed. TMT-labeled samples multiplexed based on equal protein concentration per sample in the -plex led to the identification of 38 antigens (Fig. 4A). When the dilution factor for each eluate sample was mathematically corrected, the number of identified antigens increased to 583 antigens (Fig. 4B). However, multiplexing TMT-labeled eluates based on an equal volume of sample eluted from each column (i.e., Day 0 and Day 84) resulted in identification of a total of 791 antigens (Fig. 4C). Assessing the commonality of the antigens identified in each of these three proteomic/data analysis approaches revealed that multiplexing samples based on equal protein concentration did not identify any antigens exclusively in that group. 32 of the 38 antigens found in the equal concentration group were also identified in both the equal volume and dilution factor corrected methods. While both the dilution factor corrected and equal volume groups contained a large number of uniquely identified antigens (137 unique antigens in protein concentration corrected group, 351 unique antigens in the volume loaded group), multiplexing based on equal volume led to a greater number of exclusive antigens being found in those eluates (Fig. 4D). Finally, using STRING analysis to determine the subcellular distribution for antigens detected in the equal volume group, of the 708 mapped proteins, antigens from all subcellular locations, including 219 membrane protein antigens, were identified (Fig. 4E).

Fig. 4. Multiplexing equal volumes, not equal protein concentrations, of TMT-Labeled samples results in a greater number of antigens identified via LC-MS/MS analysis.

Volcano plots of all identified antigens in LC-MS/MS analysis of TMT-labeled affinity chromatography eluate samples multiplexed by either A.) equal protein concentration, B.) equal protein concentration (corrected for dilution factor), or C.) equal volume. D.) Venn diagram of antigens identified in each of the three LC-MS/MS loading strategies (i.e., equal protein content, equal content corrected for dilution factor, and equal volume loading).

3.5. Longitudinal response toward individual antigens in rabbits implanted with native non-adjuvanted bovine pericardium

The ability of the optimized affinity chromatography approach to monitor responses toward multiple individual antigens was assessed using a non-adjuvanted rabbit subpannicular implantation model. Response toward the majority of antigens in native BP was negligible until after day 28 post-implantation (Fig 5A). All antigens showed a >2-fold increase in response from day 42 onward, which persisted until endpoint (day 56). The magnitude of response varied dramatically between antigens, with fold increase in captured protein abundance ranging from 2-fold to >40-fold in both rabbits. (Supplemental Tables 1 & 2) The overlap in antigens recognized between individual rabbits was 91.7% (Fig. 5B).

Fig. 5. Optimized affinity chromatography antigen identification approach is capable of quantifying longitudinal response toward over 700 individual antigens in a non-adjuvanted rabbit subpannicular implantation model.

A.) Longitudinal response curves for every individual antigen in a representative rabbit implanted with native BP. B.) Venn diagram of overlap between individual rabbits in their response toward identified antigens (proteins with a Fold Change ≥2) at endpoint (Day 56) following implantation with native BP.

4. Discussion

The ability to identify longitudinal immune response toward antigenic molecules responsible for inciting (i.e., primary antigens) and perpetuating (i.e., secondary antigens) adaptive immune activation has the potential to provide tremendous insight into the basic mechanisms of disease and inform novel therapeutic strategies for a wide range of clinical scenarios. [1–6] However, despite the enormous value of determining patient longitudinal response toward individual antigens, to date, numerous issues associated with antigen identification have limited its widespread utilization. The results in this manuscript demonstrate that the affinity chromatography antigen identification approach optimized herein is capable of high-throughput, quantitative, cost-efficient, and highly reproducible antigen identifications using longitudinal serum samples. Importantly, data from the unadjuvanted rabbit xenogeneic biomaterial implantation study (Fig. 5) demonstrate that this optimized antigen identification approach is applicable for multiplexed monitoring of post-implantation longitudinal graft-specific immune responses toward clinically relevant biomaterials (i.e., BP).

The current study sought to optimize three independent steps of our previously published affinity chromatography antigen identification pipeline to overcome the limitations of previous antigen identification methodologies. [1, 6, 10–14] Since solubilization of the source proteins is critical for the proteomic success, we sought to optimize extraction (i.e., physical tissue homogenization and chemical solubilization steps) of protein from a candidate xenogeneic biomaterial (i.e., BP) which is extensively utilized in heart valve replacements and other surgical applications. [20–22] Since antigens captured by their respective antibodies in the affinity chromatography column are only available for MS analysis following elution from the column, we optimized column elution conditions to maximize captured antigen elution. Finally, to provide quantitative results between samples from different time points in longitudinal experiments, we implemented micro-preparative TMT labeling, followed by sample multiplexing of the antigen eluates. Collectively, these improvements culminated in an optimized antigen identification approach, which directly addresses the previously reported limitations of antigen identification approaches.

Two major factors are involved in the protein extraction process, namely, the tissue disruption method and extraction buffer. In the previous antigen identification pipeline, manual mincing with Metzenbaum scissors was employed for tissue disruption. [1–4] While tissue homogenization by manual mincing was a reasonably consistent method for disrupting tissue, the resulting pieces were still relatively large (~1 mm× 1 mm), decreasing the tissue surface area to extraction buffer ratio for efficient protein extraction. Additionally, anecdotally, we observed a steep learning curve and relatively high variability between protein extraction runs with the manual mincing method. Despite the highly collagenous nature of BP (i.e., ~70% collagen DW) the Cell Crusher device was able to produce a powder of flash-frozen tissue pieces. [23] Consequently, the Cell Crusher method greatly decreased the size, and by extension, increased the surface area to extraction buffer ratio, of tissue pieces being prepared for protein extraction. Yet, despite these theoretical advantages, no significant (i.e., quantitative) difference in either protein or antigen solubilization was identified using the Cell Crusher method compared to manual mincing. However, the Cell Crusher method was faster to perform (homogenization in seconds, versus minutes for manual mincing) and eliminated the learning curve previously identified for the mincing method.

In the poly-polyclonal IgG affinity chromatography approach, a further consideration is that the solubilized proteins must remain capable of binding to their respective antibodies in the column. Indeed, this concern was previously demonstrated for traditional sodium dodecyl sulfate (SDS) solubilization methods, whereby 1% SDS protein isolation was reported to be incompatible with later antibody-antigen binding in the affinity chromatography column. [2] This concern may be due to the denaturing action of SDS and/or high binding affinity of the molecule to solubilized proteins. [24] The previously published antigen identification protocol, therefore, employed a two-step protein solubilization, non-denaturing, and zwitterionic buffer system (HES & LES buffers) to ensure solubilization/extraction while retaining protein conformations for downstream capture. [1, 2, 14] However, the two-step protein solubilization buffer system required the subsequent analysis of two extraction fractions, both greatly increasing the cost (i.e., both labor and financial) of each experiment and creating a source of potential error (as the total abundance of any individual protein was not directly measured, but rather, had to be calculated by summing the HES and LES abundance). Simply combining the fractions in this two-step protein solubilization approach would risk altering the delicate solubilization conditions required for proteins in each extraction solution, resulting in the potential for loss of proteins via precipitation. Therefore, to overcome these limitations, we performed protein extractions using a 1-step IP buffer protocol which significantly increased the percent protein yield compared to the 2-step HES & LES method. Importantly, unlike previously reported SDS solubilization methods, this increase in protein yield was associated with both an increase in the total quantity and overall diversity of antigens captured within the IgG affinity chromatography columns, as determined by WB and LC-MS/MS analysis. The ability of IP buffer to simultaneously increase protein yield while avoiding interference with antigen binding in the column is presumably related to the ability of sodium deoxycholate (a non-denaturing ionic detergent) and octyl-β-D-glucopyranoside (non-ionic surfactant) to solubilize antigenic proteins without disrupting their antigenic epitopes. [19] We therefore concluded that the 1-step IP buffer-crushed protocol should be utilized, as it addressed several qualitative limitations of previous antigen identification protocols while simultaneously increasing the pool of potentially identifiable antigens.

Following capture by their respective antigen-specific IgG molecules in the affinity chromatography column, antigens must then be eluted from the column for downstream LC-MS/MS analysis. The manufacturer’s recommendation for protein elution from IgG protein A or protein G affinity columns employ a glycine-urea (pH 2.9) elution buffer. However, since TFA is commonly utilized as the mobile phase for rapid ion-pairing and subsequent protein elution from a variety of HPLC column applications, we theorized that this approach may be applicable for antigen elution from IgG affinity chromatography columns. [25, 26] Compared to glycine-urea (pH 2.9)) elution buffer, we observed that affinity chromatography columns eluted with 1% TFA had significantly more total protein (~20%) in the resulting antigen fraction. This increase in total protein content translated to an increase in diversity of identifiable proteins by LC-MS/MS. Further, the coefficient of variability of protein abundances in 1% TFA eluates was lower compared to glycine-urea (pH 2.9). While the increase in total protein was encouraging, it was crucial to ensure that the 1% TFA solution was not increasing the total protein content of the eluates by causing the elution of non-specific proteins from the column. Western blot analysis confirmed that 1% TFA was, specifically increasing the antigen content of elution samples. We further confirmed this using LC-MS/MS analysis by performing a small-scale antigen identification experiment in which we assessed the adjuvanted rabbit longitudinal response (Day 0 vs. Day 84) towards adjuvanted BP. This longitudinal data confirmed that the 1% TFA elution increased the number of identified antigens by 56% in this model. We therefore concluded that the use of 1% TFA to elute captured antigens from affinity chromatography columns increases both the quantity and diversity, while simultaneously significantly reducing the variability, of antigens identified in downstream LC-MS/MS analyses.

Finally, we employed a micro-preparative TMT-labelling approach to quantitatively measure the antigens eluted from the affinity chromatography column. Since their introduction, tandem mass tags (TMT) have been routinely employed for quantitative proteomic analysis. [9, 10] TMT-labeling allows for the multiplexing and analysis of multiple individual samples in one mass spectrometry experiment, thus reducing the variability of protein quantitation across the samples and simultaneously reducing the overall cost per experiment. To our knowledge, this quantitative methodology has not been utilized in the context of longitudinal immunoproteomic antigen identification experiments, despite its numerous advantages, most notably that it allows for the direct measurement of the magnitude of the longitudinal humoral response towards any individual antigen. However, the way samples should be multiplexed for longitudinal antigen identification experiments was a lingering question that needed to be resolved. Traditional proteomics logic would dictate that samples should be “normalized” and batched based on equal protein amounts. However, the nature of antigen identification experimental designs results in “endpoint” samples (e.g., Day 84 in the adjuvanted rabbits) being expected to contain a greater amount of antigens (i.e., protein), as compared to “baseline” (i.e., Day 0) samples. Consequently, we assessed the results of multiplexing based on traditional normalization of protein loading between samples (with or without post-hoc correction for dilution factor) compared to multiplexing based on equal volumes of sample eluted from affinity chromatography columns. The data presented here demonstrate that multiplexing samples based on equal volumes resulted in identification of ~2,000% more antigens than when samples were multiplexed based on equal concentration. Further, when the dilution factor in the Day 84 samples was mathematically corrected during the data analysis process, multiplexing samples based on equal volume still identified 36% more antigens than those identified using equal protein loading. Additionally, the increased number of antigens identified in the equal volume group subsequently led to an increase in the diversity of antigens identified.

Previous publications have documented the challenge of solubilizing and subsequently identifying highly lipophilic antigens by 2-DE western blot-based immunoproteomic methods. [1, 14] Recent publications have also demonstrated the importance of highly lipophilic (e.g., integral membrane) antigens in initiating immune responses in the clinical setting (e.g., heart transplantation). [4] Consequently, it is important to ensure that any antigen identification method is capable of identifying antigens from a variety of subcellular compartments, and particularly those associated with the cell membrane. Based on STRING analysis, over 30% of antigens identified using the equal volumes multiplexing approach were membrane-associated, confirming the capacity of the final optimized protocol to identify this important antigen subgroup, while also identifying antigens from a range of intracellular compartments. We therefore concluded that multiplexing of TMT-labeled samples should be performed based on equal volumes of sample, rather than equal amounts, and, doing so, the number and diversity of identified antigens (i.e., both hydrophilic and lipophilic antigens) is significantly increased.

Finally, to provide “proof-of-concept” for the utility of our optimized, high-throughput, shotgun immunoproteomics pipeline for antigen identification, we utilized a model of xenogeneic tissue implantation. As discussed previously, the majority of current generation prostheses employed during heart valve replacement surgery are manufactured from bovine pericardial tissue. Although these valve prosthetics are chemically treated to minimize immunologic recognition, a low level of graft-specific rejection response persists in clinical patients. [32] As such, in order to engineer the next generation of heart valve prosthetics, it is imperative to identify any and all xenoantigens present in the tissue which are responsible for eliciting such a response. Therefore, in the current proof-of-concept study, we collected serum samples from unadjuvanted rabbits implanted with a 1 cm× 1 cm piece of native BP across a longitudinal time course. The ability to combine and analyze multiple samples into one LC-MS/MS run utilizing the micro-preparative TMT-labeling approach, coupled with the optimized affinity chromatography pipeline, allowed response curves for ~700 individual antigens across six time points to be constructed. As expected, the adaptive immune response towards a majority of these antigens was relatively minimal until the Day 28 timepoint, as evidenced by the relatively flat response curves. However, at the Day 42 timepoint, the response towards these antigens increased and persisted until endpoint. We determined that the magnitude of response towards any individual antigen was highly variable compared to other antigens, supporting the conclusion that individual antigens are being bound in the column rather than protein complexes being captured. Finally, when we compared the identity of each identified antigen between animals, a minimal number of unique antigens (~8%) was identified between individuals. The high degree of overlap between individuals further demonstrates high biological and technical reproducibility (i.e., 92% of antigens shared between rabbits) using our optimized high-throughput shotgun immunoproteomic pipeline for antigen identification in this proof-of-concept antigen identification study for this clinically important biomaterial (i.e., BP).

5. Limitations & Future Directions

Although the work presented herein alleviated many of the persistent deficiencies associated with previous methodologies for antigen identification, several technical challenges associated with this approach remain.

One limitation of the high-throughput affinity chromatography approach is that it only identifies the protein which elicited the immune response, not the specific epitope against which the antibody binds. In theory, it is possible some of the antigens identified may have post-translational modifications which form the epitope rather than the protein itself being the epitopic residue. One example of this occurs in the field of xenotransplantation, in which investigators must consider the presence/absence of carbohydrate post-translational modifications, such as galactose-1–3-α-galactose, in instigating a graft-specific response. [27–31] However, since the tissues of both species employed in this model of xenotransplant express similar patterns of glycosylation, the likelihood of such epitopes being antigenic is dramatically reduced. However, in applications in which post-translational modifications might be suspected drivers of the immune response (e.g., discordant xenotransplantation from animals to humans), we propose that additional steps should be taken during the protein extract preparation phase, such as enzymatically deglycosylating the prepared protein extracts, to minimize the risk of capturing these modifications in the affinity chromatography column. Indeed, such approaches have been successfully employed in previous studies to identify graft-specific protein antigen responses in patients implanted with glutaraldehyde fixed bovine pericardium heart valves. [32] Ultimately, epitope mapping studies may be required for each identified antigen to confirm epitope location and assist in follow up mechanistic studies (e.g., antibody purification).

Another technical limitation of the affinity chromatography method for antigen identification is the potential that it does not have the resolution to determine whether identified antigens were directly bound by antibodies in the affinity chromatography column, or if these proteins were subunits in captured protein complexes/associated with other proteins via protein-protein interactions. While both the protein solubilization conditions and extensive washing of the columns during the affinity chromatography step of the pipeline would likely disrupt these complexes, this concern may persist for tightly bound complexes. However, the differing magnitudes and temporal responses detected for individual identified antigens in the current work support the conclusion that a large proportion of the detected antigens are captured as individual proteins in the affinity column. This finding has also been confirmed in previous publications of affinity chromatography antigen identification in human heart transplant recipients, where STRING analysis showed no evidence of antigens being identified as part of known protein complexes. [4] Yet, the possibility that a small number of proteins identified using this approach are part of protein-protein complexes persists and requires future investigation.

6. Conclusion

The results presented herein outline an optimized, fully quantitative, high-throughput shotgun immunoproteomics pipeline for antigen identification which greatly increases the number of antigens identified, minimizes both inter- and intra-experimental variability, and reduces the overall cost-per-experiment. Additionally, in a small-scale “proof-of-concept” study, we demonstrate the utility of this optimized antigen identification protocol and its ability to monitor the longitudinal immune response towards individual antigens in a clinically relevant model of xenotransplantation, while avoiding many of the issues which have plagued previous pipelines for antigen identification. To this end, this optimized methodology for antigen identification has the capacity for widespread adoption and utilization for numerous applications and in many diverse fields of biomedical science. Ultimately, this has the potential to spur novel discoveries and provide tremendous insight into both basic mechanisms of disease, as well as support development of novel therapeutic strategies.

Highlights.

• Three steps of a previously published high-throughput shotgun immunoproteomic method for antigen identification were individually optimized in a systematic manner

• Using a model of xenogeneic biomaterial implantation, five times more individual antigens were identified using the optimized method than those found using the previously reported protocol

• Following optimization, both the cost and mass spectrometry time per experimental run was greatly decreased

• Unlike previous antigen identification protocols, the optimized methodology minimizes interexperimental variability and is fully quantitative

Abbreviations

BP

Bovine Pericardium

IP

Immunoprecipitation

HES

Hydrophilic extraction solution

LES

Lipophilic extraction solution

TMT

Tandem Mass Tag

TFA

Trifluoroacetic acid

2-DE-WB

2-Dimensional Electrophoresis-Western Blot

TCEP

tris(2-carboxyethyl)phosphine

TEABC

Triethylammonium bicarbonate

HCD

Hexachlorodisilane