Integrating the analysis of human biopsies using post‐translational modifications proteomics

Abstract

Proteome diversities and their biological functions are significantly amplified by post‐translational modifications (PTMs) of proteins. Shotgun proteomics, which does not typically survey PTMs, provides an incomplete picture of the complexity of human biopsies in health and disease. Recent advances in mass spectrometry‐based proteomic techniques that enrich and study PTMs are helping to uncover molecular detail from the cellular level to system‐wide functions, including how the microbiome impacts human diseases. Protein heterogeneity and disease complexity are challenging factors that make it difficult to characterize and treat disease. The search for clinical biomarkers to characterize disease mechanisms and complexity related to patient diagnoses and treatment has proven challenging. Knowledge of PTMs is fundamentally lacking. Characterization of complex human samples that clarify the role of PTMs and the microbiome in human diseases will result in new discoveries. This review highlights the key role of proteomic techniques used to characterize unknown biological functions of PTMs derived from complex human biopsies. Through the integration of diverse methods used to profile PTMs, this review explores the genetic regulation of proteoforms, cells of origin expressing specific proteins, and several bioactive PTMs and their subsequent analyses by liquid chromatography and tandem mass spectrometry.

1. INTRODUCTION

Proteins regulated by complex genomics and transcriptomics networks control most, if not all, cellular processes. The complexity of proteins increases as they are chemically and enzymatically decorated by post‐translational modifications (PTMs) and can be impacted by lipids or metabolites. These chemical modifications can alter protein localization and function, making PTM analysis vital in understanding proteomic signatures as biomarkers in human tissue (Klein et al., 2018; Mainoli et al., ²⁰²⁰). However, profiling PTMs is uniquely challenging given their transient nature and relatively low abundance (Castelo‐Soccio et al., 2023; Wang et al., ²⁰²¹). The emergence of sensitive liquid chromatography and tandem mass spectrometry (LC–MS/MS) techniques that enrich selective PTMs has generated crucial knowledge to better understand their roles in homeostasis and disease. However, barriers to capturing PTM complexity remain: How can we identify which cells produce a specific protein when we analyze a human biopsy with various cell types? Are all PTMs exerting essential biological functions or are there redundancies? How are PTMs affecting or being affected by metabolites and lipids? Can PTM analysis be extended to incorporate the proteomic signature of the human microbiota? This review will present a brief overview, discuss some of these complex questions, and provide insights on addressing these challenges.

2. INTEGRATION OF VARIOUS OMICS TECHNIQUES

2.1. Connecting the dots between single‐cell RNA sequencing, bulk RNA, and proteomics data

Genomics and transcriptomics techniques can reveal key information that can be integrated with proteomics investigations to identify which specific cells can produce a protein of interest (Figure 1). Recently, RNA sequencing (RNA‐seq) has become popular as an alternative to traditional microarrays for transcriptome analysis (Erfanian et al., 2022; Garber et al., ²⁰¹¹). Bulk RNAseq includes the sequencing of two types of libraries: mRNA‐only library and a whole transcriptome library containing all RNA species except rRNA. By sequencing type, the most common bulk RNAseq is a single‐end short sequencing that focuses on differentially expressed genes to comprehend molecular mechanisms in various diseases. For example, this technique was used to find genes implicated at different stages of tumorigenesis in cancer. This form of sequencing is simple, economical, and focuses primarily on the mRNA level (Li & Wang, 2021). Using bulk RNAseq helps identify potential biomarkers that are drivers of diseases (Karami et al., 2021; Nomiri et al., ²⁰²²). For example, multiple RNAseq‐based signatures have been developed and validated for cancer diagnosis, prognosis, and prediction across many tumor types (Li & Wang, 2021; Shukla et al., ²⁰¹⁷). Using bulk RNAseq data, the guanylate cyclase activator 2B (GUCA2B) gene was identified to be associated with colorectal cancer (CRC; Nomiri et al., ²⁰²²).

FIGURE 1.

Schematic representation of molecular diversity that can be analyzed using various omics techniques. NH₂ represents free amine from proteins or peptides. P inside the gray circle represents a phosphorylation group. Image created with Biorender.

However, bulk RNA‐seq measures the average expression of genes produced by all cells, making it difficult to distinguish between what cell type produces what gene products. This is as ambiguous as running a proteome‐wide investigation, limiting the detection of rare cell populations, as the gene or protein expression levels of these cells may be diminished by more abundant cells in the sample (Erfanian et al., 2022). Therefore, bulk RNA‐Seq is limited in capturing the spatial information of gene expression, which is crucial for comprehending tissue architecture and cell‐to‐cell interactions. A potential solution to this limitation is using single‐cell RNA‐seq (scRNA‐seq) or spatial transcriptomics to provide a more comprehensive picture of gene expression patterns for unique cell types (Erfanian et al., 2022; Li & Wang, ²⁰²¹). The heterogeneity and complexity of human diseases are two limiting factors but can better be studied using single‐cell analyses with a high dimension, such as single‐cell and single‐nucleus RNA sequencing (sc/snRNAseq; Healy et al., ²⁰²²). These data can later be integrated within a parallel workflow using shotgun proteomic analysis (Figure 1). Proteomics and mass spectrometry‐based proteomics are sensitive and accurate methods for large‐scale quantitative protein research, but low‐input protein samples can result in sensitivity and coverage issues, especially in complex human biopsies or samples where limited amounts are available (Aebersold & Mann, 2016; Buccitelli & Selbach, ²⁰²⁰).

As an example, during wound healing, multiple biological pathways are simultaneously activated and synchronized to respond following an injury. However, due to its complexity, characterizing the mechanisms and pathways involved remains challenging (Gurtner et al., 2008). By integrating omics data, we can more effectively investigate the complexities and heterogeneity of wound healing because of technological advancements and the introduction of potent analytical methods. These emerging capabilities enable a comprehensive integration of genomics, transcriptomics, proteomics, and metabolomics data, providing a multidimensional perspective. Using scRNA‐seq in parallel with quantitative proteomics and integrating data obtained from both techniques, we can better determine the molecular characteristics underlying a unique regenerative capacity, revealing a diverse skin fibroblast population producing different proinflammatory and fibrotic gene signatures in the back skin and velvet of reindeer, that was later translated to mouse models and human biopsies, and confirming a gene‐to‐protein axis production that correlated significantly (Sinha, Sparks, et al., 2022). These integrations were also used on COVID‐19 human patient samples in which COVID‐19‐enriched neutrophils were analyzed to better understand molecular mechanisms of dexamethasone action to develop targeted immunotherapies for COVID‐19 infections (Sinha, Rosin, et al., 2022). ScRNA‐seq is an effective method (Van de Sande et al., 2023; Sinha, Rosin, et al., ²⁰²²; Sinha, Sparks, et al., ²⁰²²) for measuring gene expression in single cells but mRNA expression levels do not always correspond precisely with protein expression levels from proteomics studies. However, other biological mechanisms impact gene‐to‐protein translation. For example, alternative splicing can produce multiple mRNA isoforms from a single gene, some of which may not be translated into protein (Buccitelli & Selbach, 2020). Additionally, regulatory elements such as microRNAs and RNA‐binding proteins can impact mRNA transcript stability and translation efficiency (Buccitelli & Selbach, 2020; Kooshkaki et al., ²⁰²⁰).

2.2. Single‐cell proteomics approaches

Due to the wide variety of protein copy counts in cells (approximately 10⁸ copies per cell for actin and with a median expression level of ~170,000 copies/cell in mammals) and the lack of PCR‐like amplification methods, single‐cell proteomics remains challenging but is rapidly evolving (Mansuri et al., 2023; Motone & Nivala, ²⁰²³). One of the most common approaches for quantifying protein abundance at the single‐cell level relies primarily on antibody‐based detection or tagging a fluorescent probe to the protein of interest, and after it is labeled, the proteins of interests can be detected using various methods such as flow cytometry or immunofluorescent microscopy (Labib & Kelley, 2020; Petrosius & Schoof, ²⁰²³). Furthermore, mass cytometry, also called cytometry by time of flight (CyTOF), can integrate mass spectrometry and flow cytometry, allowing for the investigation of single‐cell protein expression. This has allowed the simultaneous evaluation of DNA content and proteins, enabling the measurement of 40–100 parameters within individual cells (Tracey et al., 2021). Due to the limitations of using antibodies and tagged probes, there is still a need for a technique that can acquire more proteins. However, additional key clinical information about tumors can be gained using imaging mass cytometry. For example, over 1.1 million cells were analyzed at a single cell resolution of 139 high‐grade glioma and 46 brain metastatic tumor biopsies revealing pathogenic and protective cancer cellular neighborhoods (Karimi et al., 2023). Interestingly, using this approach, novel intra‐tumoral cellular neighborhoods associated with distinct survival were identified providing key differences in tumor and immune cell populations. Furthermore, a unique population of myeloperoxidase (MPO)‐positive macrophages was identified, and these newly identified cells were associated with long‐term survival (Karimi et al., 2023). A similar approach was used for lung cancer; using machine deep learning, this dataset can then be used to predict which patients will have a better survival outcome following surgery using only a single 1 mm² tumor (Sorin et al., 2023). Importantly, single cell profiling using CyTOF has revealed key novel biological mechanisms and unique cell populations previously uncharacterized.

Single‐cell analysis in living humans can be challenging, invasive, or impossible depending on the organ being studied. For example, eyes, lungs, hearts, or brains cannot be easily sampled in living humans because obtaining tissue biopsies would cause serious damage. Alternatives are thus needed to analyze proteomes in living humans. Using biofluids (or liquid biopsies), such as serum, plasma, cerebrospinal fluid (CSF), aqueous humor, or vitreous humor, offer new opportunities to analyze these proteomes. However, such biofluids can be challenging to analyze due to high dynamic ranges of proteins. To address this issue and better characterize protein levels abundance, a new data integration approach called TEMPO or Tracing Expression of Multiple Protein Origins (Wolf et al., 2023; Wolf et al., ²⁰²⁴). Using TEMPO, a high‐resolution proteomic profiling of aqueous and vitreous humor from 120 living patients was performed using a DNA aptamer‐based proteomics approach (SomaLogic) to quantify >5,900 proteins (Wolf et al., 2023). Integrating scRNA‐seq with proteomics data, multi‐omics approaches like TEMPO now make it possible to molecularly link individual proteins and their genes before tracing them back to a single cell origin (Wolf et al., 2023). However, only human samples can be analyzed this way at this time due to the protein detection platform, but LC–MS/MS approaches would offer new opportunities to use TEMPO in multiple species other than humans.

Using MS, typically, more than 100 nanogram of proteins are required (Petrosius & Schoof, 2023). However, it is estimated that a single mammalian cell contains 150 pg of protein on average (Wiśniewski et al., 2014). Therefore, a highly sensitive sample preparation is required in addition to an analytical pipeline that generates quantitative and reproducible data, thereby providing essential insights into pertinent biological processes (Rosenberger et al., 2023). The acquisition of a large number of peptides, however, by the mass spectrometer remains a significant challenge in the field (Petrosius & Schoof, 2023). For sample preparation, applying microfluidics platforms might be an effective way to overcome the challenges. The core principle of droplet‐based processing involves capturing individual cells within droplets, which serve a dual purpose: they act as carriers for the cells while also serving as confined reaction chambers. Moreover, these microfluidic platforms can seamlessly combine various droplets containing necessary reagents, streamlining subsequent sample processing steps (Gebreyesus et al., 2022). Although microfluidic devices have found routine use in the realm of transcriptomics, their application in the field of proteomics is recent but promising (Petrosius & Schoof, 2023). Single‐cell proteomics mass spectrometry (scp‐MS) is advancing rapidly and has the capacity to provide crucial insights into fundamental aspects of disease pathology and progression (Petrosius & Schoof, 2023). For example, several data independent acquisition (DIA) methods have been tested and optimized for scp‐MS (Petrosius et al., 2023). Therefore, it is now more amenable for long injection times in the mass spectrometer resulting in higher resolution but also being able to retain a low scan cycle time to achieve reliable quantification (Petrosius et al., 2023). Consequent challenges remain when combining scRNA‐seq and proteomics for all studies due to cost, time required for data analysis, and computational intensity. Measuring protein levels alone, however, is not fully indicative of protein function since a translated protein rarely goes unmodified. Therefore, a more in‐depth analysis must also include PTM analysis.

2.3. Post‐translational modifications

While traditional proteomic approaches that delve into upregulation or downregulation of given proteins in specific samples provide insight into potential mechanisms that may underlie a particular condition, they fail to account for the plethora of PTMs that play integral roles in proteins' structure, function, and activity. There are >300 PTMs in humans and most of which are either understudied or not ever analyzed in a traditional proteomics workflow (Figure 2) (Conibear, 2020; Prus et al., ²⁰¹⁹; Ramazi & Zahiri, ²⁰²¹). PTMs are typically low in abundance in cells and tissues despite having a potent biological impact on the fate of a specific cell or cell–cell interaction (Prus et al., 2019). Therefore, enrichment is needed to detect these chemical modifications. Mass spectrometry‐based proteomics offers the necessary sensitivity to profile thousands of PTMs within a given sample.

FIGURE 2.

Diagram of the chemical groups of 24 post‐translation modifications (PTMs) that can be studied in human biopsies. Image created with Biorender.

2.3.1. Proteolysis

All living organisms rely on regulated cell division and cell death to control biological functions and tissue homeostasis. Proteolysis is a key PTM as most, if not all, proteins are modulated by proteases to remove their signal peptide, activated via cleavage of a propeptide, processed to modulate their functions, or degraded to reduce their overall levels (Drag & Salvesen, 2010; Klein et al., ²⁰¹⁸; Wang et al., ²⁰²¹). There are 473 human proteases (including an additional 90 nonproteolytic homologues) and 151 human protease inhibitors (reviewed here [Kappelhoff et al., ²⁰¹⁷; Wang et al., ²⁰²¹]); yet, as demonstrated using the MEROPS protease repository (www.ebi.ac.uk/merops; Rawlings et al., ²⁰¹⁰), most proteases have not been extensively studied and little information is known about what substrates they cleave. Using N‐terminomics approaches such as TAILS (Anowai et al., 2022; Das et al., ²⁰²³; Gordon et al., ²⁰¹⁹; Kleifeld et al., ²⁰¹⁰), Subtiligase enrichment (Julien et al., 2016; Mahrus et al., ²⁰⁰⁸), COFRADIC/ChaFRADIC (Gevaert et al., 2003; Venne et al., ²⁰¹⁵), HYTANE (Chen et al., 2016), HUNTER (Weng et al., 2019), or CHOPS (Griswold et al., 2019), cells and tissue‐wide profiling of proteolysis, from targeted processing of substrates to degradation, can now be done routinely. Importantly, proteolysis is a key PTM used by bacteria (Eckhard et al., 2017; Lentz et al., ²⁰¹⁸; Marshall et al., ²⁰²⁰), viruses (Chen et al., 2023; Jagdeo et al., ^{2015, 2018}; Pablos et al., ²⁰²¹), fungi (Ball et al., 2019; Gutierrez‐Gongora et al., ²⁰²³; Gutierrez‐Gongora & Geddes‐McAlister, ²⁰²²; Pettersen, Dufour, & Arrieta, ²⁰²²), and parasites (Coffey et al., 2018; Li et al., ²⁰¹⁶), and N‐terminomic approaches can also be used to study these interactions. Using primary human cells, TAILS was used to identify substrates of mucosa‐associated lymphoid tissue lymphoma translocation protein 1 (MALT1) in B cells (Klein et al., 2015). It has also been used to profile proteolysis of ulcerative colitis patients' colonic biopsies (Gordon et al., 2019) and the synovial fluids of patients with osteoarthritis (Das et al., 2023). Overall, proteolysis is a crucial PTM but remains predominantly enigmatic in most cells and human tissues as most human proteases still have no known substrates or have been studied in primary cells or human biopsies. Paired with the severe lack of knowledge regarding bacterial, fungal, viral, and parasitic proteases, there is a large gap of knowledge that is gradually being filled in.

2.3.2. Phosphorylation

Phosphorylation of amino acids is a common PTM and can provide insight into signaling cascades, gene expression, and the activity state of kinases or other enzymes (Castelo‐Soccio et al., 2023). Phosphorylation generally occurs on residues that have a hydroxyl group present on their side chain such as serine, tyrosine, and threonine, but other residues can also be phosphorylated (Ubersax & Ferrell, 2007). Detection of phosphorylated residues often presents a challenge as phosphorylated proteins are generally low in abundance compared to the entire pool of peptides in a sample; therefore, enrichment is required for adequate detection (Urban, 2022). This challenge is further complicated by the complexity of signaling pathways and the non‐linearity of how kinases interact with their specific protein substrates (Jordan et al., 2000). Moreover, the stoichiometry of phosphorylation sites comparing the ratio of proteins with phosphorylated and non‐phosphorylated residues plays a critical role in driving biological processes (Mayya & Han, 2009). Essentially, due to the complexity of biological signaling processes, even a low phosphorylation level can have profound biological implications (Ferguson & Gray, 2018; Mayya et al., ²⁰⁰⁹; Zhang et al., ²⁰⁰⁹). While functional data comparing stoichiometries of such phosphorylation is becoming more readily available, its interpretability still presents a challenge, resulting in difficulty applying these results to a clinical setting. While what has been mentioned here may compel one to not pursue a phosphoproteomic approach, with the ever‐increasing availability of stoichiometric, residue site, and biological functionality data available, paired with the important role phosphorylation plays in nearly all biological processes, it can be an effective method to characterize causative effects based on kinase activity.

Using mass spectrometry‐based phosphoproteomics approaches, a high level of specific detection of phosphorylation events can be detected, while also being broadly applicable to many biopsy types and providing large amounts of data that can be further validated by complementary methods. The workflow for phosphoproteomics follows a similar approach to traditional, bottom‐up proteomics but has enrichment steps that permit the detection of phosphorylated amino acids (Urban, 2022). Mass spectrometry analysis of human biopsies using phosphoproteomics offers new opportunities to study the impacts of kinase signaling, stoichiometry of phosphorylated residues, and the cellular localization of such phosphorylation events and how these contribute to diseases. For example, a study examined phosphoproteomic profiles of CRC patients and how specific profiles predict metastatic versus non‐metastatic CRC (Li et al., 2020). Another study on acute myeloid leukemia used phosphoproteomic profiling to compare drugs that interact synergistically with Selinexor to reduce tolerance and sensitivity and found that the protein kinase B inhibitor MK‐2206 decreased cellular proliferation of cancerous cells and decreased intracellular trafficking of proteins contributing to this uncontrolled cellular division (Emdal et al., 2022). This comparison showed the impact of inhibiting a sole kinase and the resulting changes in entire phosphoproteomic networks, highlighting that even minor changes to stoichiometry levels can play a large role in biological processes. Phosphoproteomic analysis revealed that viral SARS‐CoV‐2 proteins were phosphorylated by host cellular machinery, such as protein kinase C, cyclin‐dependent kinase, and casein kinase II (Bouhaddou et al., 2020). The same study showed that SARS‐CoV‐2 infection activated mitogen‐activated protein kinase and casein kinase II, in which the negative charge associated with phosphate addition directly contributed to the functionality of viral proteins, namely the membrane, nucleocapsid, and non‐structural protein 9 (Bouhaddou et al., 2020). Overall, phosphoproteomics provides crucial information about potential diagnostic and therapeutic applications of protein phosphorylation.

2.3.3. Glycosylation

Glycosylation is prevalent PTM and defined by the addition of a glycan to an amino acid, namely asparagine (N‐linked) or serine and threonine (O‐linked; Cain et al., ²⁰²¹; Girgis et al., ²⁰²⁴; Morelle et al., ²⁰⁰⁶). Glycosylation impacts many functions of a protein: its structure, protein–protein interactions and physical dynamics (Xu et al., 2023). In complex biological samples, it is challenging to detect and categorize glycosylated residues due to their heterogeneity, stoichiometry constraints, and uncertainty of the specific residue that is glycosylated (Riley et al., 2021). Enrichment and labeling methods are thus needed. Lectin‐based glycoprotein enrichment is a common method but is constrained by the specificity of lectins to a single glycan and thus losing coverage of glycosylated proteins that do not contain the specific glycan (Xiao et al., 2018). While this may suffice in some cases, in complex biological samples, unless the query is a specific glycosylation event, broad scale glycoproteomic detection through mass spectrometry can be unfeasible. To circumvent this limitation, one can use an isotope‐targeted glycoproteomic (IsoTaG) approach, which was demonstrated to identify 32 N‐glycopeptides and over 500 O‐glycopeptides in Jurkat and PC‐3 cells with high sensitivity to low‐abundance glycopeptides (Woo et al., 2015).

Aberrant glycosylation is a hallmark of many diseases including cancer (Ohtsubo & Marth, 2006; Tuccillo et al., ²⁰¹⁴). For example, one study, using mass spectrometry techniques, identified aberrant CD43 glycosylation is detectable in several cancer tissues including breast and colon cancers and leukemia (de Laurentiis et al., ²⁰¹¹). Both the over‐ and under‐expression of glycan alterations have been implicated as a driver of tumor growth, invasion, and metastasis. Altered glycosylation patterns on cell surface receptors and adhesion molecules can promote tumor cell adhesion, migration, and immune suppression (Läubli & Borsig, 2019). Many changes in glycosylation of circulating proteins such as cancer antigens or glycoproteins shed by tumors can serve as biomarkers for cancer detection, prognosis, and monitoring treatment response (Kirwan et al., 2015). For example, elevated levels of serum glycoproteins such as CA 19–9 (Poruk et al., 2013) and CA 125 (Charkhchi et al., 2020) are currently used as biomarkers for pancreatic and ovarian cancer, respectively. Inflammatory and autoimmune diseases are also closely associated with dysregulated glycosylation (Radovani & Gudelj, 2022). Changes in glycosylation patterns of immunoglobulins, cytokines, and cell surface receptors can modulate immune cell function, contributing to disease pathogenesis. Detection of specific glycosylation signatures on immune molecules may aid in diagnosing autoimmune conditions and monitoring disease activity. Furthermore, glycoproteins and glycosylation alterations have been implicated in neurodegenerative disorders such as Alzheimer's and Parkinson's diseases (Abou‐Abbass et al., 2016). Abnormal glycosylation of proteins like amyloid beta and alpha‐synuclein can influence their aggregation and toxicity, contributing to disease progression (Videira & Castro‐Caldas, 2018). Glycosylation patterns of CSF proteins or circulating glycoproteins in blood or serum may serve as effective biomarkers for early detection and monitoring of neurodegenerative, inflammatory diseases, and cancer. Therefore, glycoproteomics is a crucical technique to better understand and characterize the pathogenic glycosylation processes in human diseases.

2.3.4. Ubiquitination

Ubiquitination is a PTM in which a ubiquitin protein is covalently attached to a target protein (Rape, 2018). It plays a role in several cellular processes, including protein degradation (Tai & Schuman, 2008), signal transduction (Haglund & Dikic, 2005), DNA repair (Ghosh & Saha, 2012), and cell cycle control (Zou & Lin, 2021). The mechanistic process of ubiquitination involves a series of enzymatic reactions mediated by three main types of enzymes: ubiquitin‐activating enzymes (E1), ubiquitin‐conjugating enzymes (E2), and ubiquitin ligases (E3; Rape, ²⁰¹⁸). First, an E1 enzyme activates ubiquitin by forming a bond with its C‐terminal glycine residue. Then, the activated ubiquitin is transferred to an E2 enzyme. Finally, an E3 ligase facilitates the transfer of ubiquitin from the E2 enzyme to a specific target protein, usually by forming a bond between ubiquitin and a lysine residue on the substrate target (Rape, 2018; Schulman & Harper, ²⁰⁰⁹). This can result in a single ubiquitin molecule (monoubiquitination) or multiple ubiquitin molecules forming a chain (polyubiquitination).

Ubiquitination is a key PTM due to its role in controlling several protein functions in homeostasis and is often dysregulated in human pathologies such as cancer and inflammatory diseases (Rape, 2018; Schulman & Harper, ²⁰⁰⁹; Tai & Schuman, ²⁰⁰⁸; Zou & Lin, ²⁰²¹). Identification and quantification of ubiquitinated proteins have been studied through mass spectrometry and allow the identification of ubiquitination sites in addition to the type of ubiquitin linkage present on the substrate (Udeshi et al., 2013). With the emergence of antibodies targeting the Lys‐ε‐Gly‐Gly (K‐ε‐GG) remnants following trypsin digestion at ubiquitination sites, we are able to label, quantify, and identify the ubiquitination sites of proteins on a large scale (Swatek et al., 2019; Udeshi et al., ²⁰¹³). This technique was first applied to characterize ubiquitin modifications in yeast samples and has since been expanded to several species and experimental models (Peng et al., 2003). Over the years, with advances in mass spectrometry, the mapping of individual ubiquitin modifications can be better described. Nevertheless, the extent and architecture of polyubiquitin signals remain challenging to characterize. Despite key advances in the field and that the “GlyGly” technique is an effective way of elucidating substrates of typical ubiquitination, because ubiquitylation can also occur on other residues such as cysteine, serine, and threonine, current methodologies may be insufficient (Ohtake, 2020). However, new negative and positive selection approaches have made this PTM easier to identify and quantify by mass spectrometry. For example, using a ubiquitin‐clipping approach (engineered viral protease called Lb^pro*), resulting in an enhancement of the signature C‐terminal GlyGly dipeptide on a modified residue, it is now easier to more effectively assess protein ubiquitination on substrates in addition to polyubiquitin (Swatek et al., 2019). Therefore, there is still a need to advance techniques such as antibody enrichment to study ubiquitination by mass spectrometry and continue providing new alternatives for large‐scale unbiased proteomics studies.

2.3.5. Palmitoylation

Palmitoylation is a PTM that modifies lipids, and it typically governs the localization, accumulation, stability, secretion, and function of proteins by modifying their affinity to cell membranes (Liu et al., 2022). There are three forms of palmitoylation: S‐palmitoylation, N‐palmitoylation, and O‐palmitoylation. Of these, S‐palmitoylation is the most common. N‐palmitoylation occurs when a fatty acid palmitate forms a stable amide bond with a cysteine residue at the N‐terminus of a protein. In contrast, O‐palmitoylation occurs when the monounsaturated form of palmitate forms an oxyester bond with the hydroxyl group of serine or threonine (Liu et al., 2022). S‐palmitoylation is regulated via the enzymatic attachment of a 16‐carbon saturated fatty acyl chain to the sulfhydryl group of cysteine residues in proteins, forming a temporary thioester bond. It is facilitated by palmitoyltransferases, primarily found in the zinc finger aspartate–histidine–histidine–cysteine (ZDHHC) family (Jeong et al., 2023; Ko & Dixon, ²⁰¹⁸). There are 23 ZDHHC proteins in mammals, which utilize palmitoyl‐CoA as the primary source of palmitoyl groups for acylating substrate proteins (Ko & Dixon, 2018). S‐palmitoylated proteins can undergo palmitoylation and de‐palmitoylation processes, transitioning between different forms, with time intervals ranging from seconds to hours (Jin et al., 2021; Ko & Dixon, ²⁰¹⁸). Based on the SwissPalm database (https://swisspalm.org), it is estimated that ~10% of the human proteome could undergo palmitoylation. This collection of palmitoylated proteins is referred to as the “palmitoylome,” and >600 substrates have been identified and characterized so far (Jin et al., 2021). Various techniques have been used to detect S‐palmitoylation including acyl‐biotin exchange (ABE), acyl‐resin‐assisted capture (Acyl‐Rac), and acyl‐PEG exchange (APE) followed by LC–MS/MS (Cheng et al., 2021; Jeong et al., ²⁰²³; Ji et al., ²⁰¹³). The dysregulation of protein S‐palmitoylation has been linked to the development of various human diseases, including viral infections, cancer, diabetes, as well as immunological and neurological disorders. Specifically, ZDHHC enzymes have been associated with cancer as potential oncoproteins, tumor suppressors, or prognostic indicators (Cheng et al., 2021; Jeong et al., ²⁰²³; Jin et al., ²⁰²¹). Overall, palmitoylation is a key PTM and additional studies are needed to better understand its functions in human diseases as it likely has potential diagnostic and therapeutic applications.

2.4. Complementing proteomics and PTMs analysis with metabolomics and lipidomics

2.4.1. Metabolomics

Metabolomics is used to study an organism's small molecules or metabolites. Recently, LC–MS/MS has emerged to support innovation and accuracy in metabolomics research. LC–MS/MS uses the separation capabilities of liquid chromatography with the sensitivity and specificity capabilities of mass spectrometry. During sample preparation, metabolites are extracted and purified to remove interfering compounds. Then, liquid chromatography separates the metabolite mixture, allowing the analytes to be eluted in a time‐dependent manner. Subsequently, tandem mass spectrometry employs multiple stages of mass analysis, including fragmentation, to obtain the precise identification and quantification of metabolites.

There are several advantages of LC–MS/MS in metabolomics:

1. Sensitivity and selectivity: LC–MS/MS enables the detection and quantification of metabolites across a wide concentration range, from low‐abundance signaling molecules to high‐concentration intermediates (Mehara & Roithová, 2020). The high selectivity of mass spectrometry ensures minimal interference from complex sample matrices, enhancing the accuracy of metabolite identification (Jia et al., 2023).

2. Comprehensive coverage: LC–MS/MS facilitates the simultaneous analysis of a vast array of metabolites due to its compatibility with a broad range of chromatographic and ionization techniques (Dettmer et al., 2007). This allows for comprehensive profiling of metabolomes, leading to a more holistic understanding of cellular processes.

3. Structural elucidation: The tandem mass spectrometry aspect of LC–MS/MS enables the elucidation of metabolite structures through fragmentation patterns (Heiles, 2021). By comparing these patterns to spectral libraries, metabolite identities can be confidently assigned, aiding in the discovery of novel biomarkers and pathways.

When analyzing complex human biopsies, metabolomics can aid in biomarker discovery as LC–MS/MS has played a pivotal role in identifying potential metabolite biomarkers associated with various diseases, such as cancer, diabetes, and neurodegenerative disorders (Chen et al., 2022; Nakayasu et al., ²⁰²¹). Additionally, LC–MS/MS can provide invaluable insights into drug metabolism and pharmacokinetics (Chen et al., 2007; Gwynne et al., ²⁰²²). With the analysis of metabolites generated during drug metabolism by LC–MS/MS, it is possible to understand the efficiency, toxicity, and interactions between drugs, contributing to the development of safer and more effective therapeutics (Reddy et al., 2021). Advancements in LC–MS/MS technology have led to improved sensitivity, resolution, and throughput. Integrating high‐resolution mass spectrometry (HRMS) and advanced data processing algorithms has enhanced metabolite annotation and identification (Naz et al., 2017). Additionally, coupling LC–MS/MS with other omics platforms, such as genomics and proteomics, enables a “multi‐omics” data integration to provide a more comprehensive understanding of biological systems. The LC–MS/MS approaches' sensitivity, selectivity, and structural elucidation capabilities make it indispensable for biomarker discovery, drug metabolism studies, and studying the biological function of metabolic proteins.

2.4.2. Lipidomics

Lipidomics characterizes lipids, their composition, and their interactions within biological systems. To dissect the complex lipidome of human biopsies, LC–MS/MS offers exceptional sensitivity, allowing for the detection and quantification of lipids present in minute quantities within biological samples. By coupling liquid chromatography with mass spectrometry, the technique enables the separation of lipid classes and species, followed by precise identification and quantification. This sensitivity facilitates the investigation of low‐abundance lipids, unveiling their significance in various cellular processes. LC–MS/MS provides invaluable structural information by generating fragmentation patterns of lipid ions (Pitt, 2009). The tandem mass spectrometry process involves isolating precursor ions and subjecting them to collision‐induced dissociation, resulting in the production of characteristic fragment ions (Johnson & Carlson, 2015). These fragments can be used to deduce the fatty acyl composition, double bond positions, and other structural features of lipids. Such structural insights are essential for deciphering lipid functions and exploring their roles in disease states.

Quantitative analysis is another critical aspect of lipidomics research, and LC–MS/MS excels in this regard. By incorporating stable isotope labeling, LC–MS/MS enables accurate quantification of lipid species (Triebl & Wenk, 2018). This approach accounts for variations in sample preparation, ionization efficiency, and instrument response, ensuring reliable and reproducible results. Moreover, LC–MS/MS allows for targeted lipid quantification by selectively monitoring specific lipid species of interest, aiding in the identification of biomarkers and the characterization of lipidomic alterations associated with the disease.

The high‐speed separations offered by modern liquid chromatography systems, combined with the rapid scan rates of mass spectrometers, facilitate high throughput lipidomics analysis. LC–MS/MS platforms can analyze many lipid samples within a short time frame, making it an ideal technique for large‐scale studies that require an efficient workflow. Furthermore, LC–MS/MS can simultaneously analyze several lipid classes, encompassing various molecular species, including glycerolipids, glycerophospholipids, sphingolipids, and sterols (Quehenberger et al., 2010). This comprehensive coverage of the lipidome allows for a more encompassing understanding of lipid metabolism and its implications for health and disease. As technology advances and challenges are addressed, LC–MS/MS will undoubtedly continue to pave the way for significant progress in the advancement of the field of lipidomics to help characterize human biopsies.

2.5. Added complexity to human biopsies: Exploring the microbial proteome

The evolution of specialized proteomics techniques that extensively profile proteolysis and PTMs in complex human tissue samples also extends our capacity to understand the contributions of the human microbial proteome, dubbed “metaproteogenomics.” The human microbiome, consisting of an individual's resident microbial community, which largely rivals the number of host somatic cells, provides significant insight into disease pathogenesis (Sender et al., 2016). While individual microbiome signatures are exceedingly personalized and heterogenous among individuals, advances in standardized microbiome sampling and targeted 16 s rRNA gene amplicon sequencing in case–control and cohort studies have uncovered distinct microbial biomarkers associated with chronic and acute disease, including obesity and early‐onset autoimmunity (Gilbert et al., 2018). With the advent of the second phase of the Human Microbiome Project (Integrative Human Microbiome Project, iHMP), studies of the microbiome have expanded toward longitudinal sampling and multi‐omics analyses of specific microbiome‐associated conditions such as preterm birth and inflammatory bowel disease (Proctor et al., 2019). Common metaproteogenomic methodology and relevant toolsets have been reviewed extensively (Lin et al., 2019; Schiebenhoefer et al., ²⁰¹⁹). Here, we will highlight recent key findings and challenges in integrating microbial and host proteomics datasets from gut biopsies and the potential use of specialized proteomics techniques, including N‐terminomics to profile proteolysis and discover activity‐based biomarkers for disease.

2.6. Metaproteogenomics of the microbiome

While distinct microbial communities have been detected on the skin and in mucosal body sites including the oral cavity and vagina, the most frequently studied niche is, by far, the human gut. Interactions between host cells and commensal microbiota broadly define mucosal pathophysiology and thus require rigorous profiling to appreciate the complexity of disease states.

Dysbiotic microbial communities have been implicated in numerous adverse health outcomes, including cardiovascular disease, neurodevelopmental disorders, and inflammatory bowel disease (Chopra et al., 2021; Pettersen, Antunes, et al., ²⁰²²; Pettersen, Dufour, & Arrieta, ²⁰²²; Shreiner et al., ²⁰¹⁵; Vijay & Valdes, ²⁰²²). Given the minimal invasiveness of obtaining fecal samples, microbial contributions to disease pathogenesis (diversity, metabolic activity, etc.) are evaluated on stool using multiple omics techniques. 16 s rRNA gene amplicon sequencing on fecal samples has thus revealed that the large intestine harbors a large and diverse microbial community ranging from 10¹⁰ to 10¹⁴ bacteria and dominated by the phyla Firmicutes and Bacteroidetes (Pettersen, Dufour, & Arrieta, 2022; Shreiner et al., ²⁰¹⁵; Turner, ²⁰¹⁸). Metaproteomic approaches have previously been used to taxonomically profile the gut microbiome in CRC (Long et al., 2020). Data‐independent acquisition label‐free quantitative LC–MS/MS identified specific bacterial genera previously implicated in CRC pathogenesis. Microbial proteins involved in iron acquisition and transport, oxidative stress, and DNA maintenance were differentially detected in patients with CRC compared to healthy controls. Integrating complementary genomics techniques (16s rRNA gene amplicon sequencing, shotgun metagenomic sequencing) into metaproteomic analysis of stool samples has the potential to provide a more complete picture of host–microbe interactions. Changes in the distribution of bacterial taxa in the gut have been explored as potential triggers for ulcerative colitis using 16s rRNA gene sequencing and label‐free LC–MS/MS (Thuy‐Boun et al., 2022).

Comparing these two methods revealed differences in the relative abundance of peptides from different kingdoms of life. Peptides from diverse kingdoms beyond bacteria were identified using LC–MS/MS compared to 16s rRNA sequencing, which only identified bacterial taxa. Additionally, the Bacteroidetes:Firmicutes ratio was distinct between methods, yielding a higher proportion of Bacteroidetes taxa by LC–MS/MS. However, most peptides detected by LC–MS/MS were human‐derived or derived from dietary plants. A large proportion of peptides were also unassigned using LC–MS/MS, likely contributing to the differences in taxonomic profiling using 16s rRNA sequencing, which is specific for bacterial identification. Both host and non‐host proteins were differentially detected in ulcerative colitis patients compared to healthy controls, particularly host serine‐endopeptidase activity (Denadai‐Souza et al., 2018; Gordon et al., ²⁰¹⁹; Motta et al., ²⁰¹⁹). Metaproteogenomic methods have evidently displayed potential to uncover targets for disease pathogenesis in the human gut but require improvements to begin to incorporate the impacts of microbial communities.

2.7. Challenges in integrating microbial data

Considering the individual and intra‐species complexity in the intestinal niche, proteomics techniques have been limited in their ability to capture microbial biological activity. These limitations have been described previously (Mayne et al., 2016; Muth et al., ²⁰¹⁵), emphasizing the weaknesses in current databases used to identify microbial proteins or peptides from MS spectra. However, new analysis tools have been developed to reduce the time and resources needed for such analysis. For example, MetaProClust‐MS1 (Simopoulos et al., 2022) can be used for large‐scale metaproteomics screening and allow for better prioritization of metaproteomic sample analysis. MetaProClust‐MS1 (Simopoulos et al., 2022) was developed to identify clusters of microbiome samples using MS1‐only profiles comparable to those identified in MS/MS spectra, therefore, providing a prioritization pipeline that can save time during analysis.

Additionally, custom proteogenomic databases are often generated to incorporate proteomes from specific taxa native to a particular niche. Still, variation at the strain level makes reproducibility between analyses complex and increases computational intensity. To address this, standardization of metaproteomic approaches using human and mouse gut microbial gene catalogs have been created to generate a universal workflow intended for the identification of microbial peptides in the intestine (MetaPro‐IQ). However, the limited pool of databases becomes more severe for specialized PTM analysis like N‐terminomics, which are curated for human proteases and processing or degradation of human substrates (Gordon et al., 2019; Luo et al., ²⁰¹⁹; Mainoli et al., ²⁰²⁰). Importantly, our current limited understanding of bacterial and fungal proteases cannot fully capture proteolytic activity at the host‐microbe interface. Current bioinformatics approaches are restricted to hypothesis generation for microbial protease activity in human biopsies.

Furthermore, specific signatures distinguishing between host and microbial cleavages have yet to be discerned, requiring extension into in vivo studies to validate microbial proteolysis of host substrates. Moreover, low abundance proteins derived from microbial taxa are frequently missed using data‐dependent acquisition (DDA), minimizing their perceived impacts on host biology. The bias toward high‐abundance proteins associated with DDA is amplified when characterizing the changes in low‐abundance post‐translationally modified microbial proteins. Coupling MS with techniques like immunoaffinity enrichment can help visualize protein modifications from cell lysates but requires more samples, which becomes incompatible when analyzing rare and often invasive human biopsies (Mainoli et al., 2020; Ouidir et al., ²⁰¹⁶).

Therefore, metaproteogenomics provides a significant opportunity to explore biomarkers for disease pathogenesis in complex human samples at the host‐microbe interface. However, successfully incorporating PTM analysis into the analysis of human biopsies, including vaginal swabs, saliva, and sweat, will require the development of specialized computational tools and databases that integrate complete microbial proteomic signatures.

2.8. Advancements in proteomics analysis of body fluids and biopsies for disease biomarker discovery

The search for disease‐related biomarkers has been significantly impacted by MS‐based proteomics of body fluids. Serum and/or plasma reflect the physiological state of an organism, and thus, it is constantly employed in proteomics studies. Given the relative ease and minimal invasiveness of collection, patient blood, serum, and plasma analysis offer major advantages when designing clinical studies (Figure 3a). However, the predominance of highly abundant proteins, such as albumin and immunoglobulins, hamper the detection of low abundant proteins therefore providing significant challenges for the analytical sensitivity, resolution, and reproducibility of MS analysis. Additionally, the presence of metabolites, lipids, and salts in serum and plasma introduce further confounds. This becomes more complicated when the low abundant proteins represent potential disease‐specific markers. However, with the advent of certain highly abundant protein depletion strategies and data‐independent acquisition (DIA) modes, the detection of low abundant proteins has improved significantly. In general, the proteomics analysis of target tissue makes more sense; however, for clinical diagnosis of disease at early stages or identification of non‐surgical cancers or metastases, the proteomics analysis of plasma or serum becomes indispensable. The collection of plasma or serum is relatively less invasive as compared to the extraction of disease‐affected target tissue (Pietrowska et al., 2019). For example, using a tandem mass tag (TMT)‐based shotgun proteomics approach, plasma of healthy pregnant females, non‐pregnant females, preeclampsia, and gestational hypertension patients was profiled to identify the changing protein signatures. FN1, ITIH2, and ITIH3 proteins were significantly enriched in gestational hypertension and preeclampsia but not in pregnant and non‐pregnant women (de Almeida et al., ²⁰²²). Another study used a top‐down proteomics approach coupling 2‐D gel electrophoresis and mass spectrometry to investigate PTMs and profile the proteome of first‐trimester maternal serum. The study found 30 significantly changing proteoforms in patients with spontaneous preterm delivery (D'Silva et al., ²⁰¹⁸).

FIGURE 3.

Proteomics analysis of human biopsies can profile distinct anatomical sites. (a) Depiction of biopsies previously used for proteomics analysis. (b) Benefits and challenges of using proteomics approaches to interrogate diverse biological samples based on invasiveness of collection method. Image created with Biorender.

The invasiveness of plasma and serum collection techniques has sparked the interest of researchers to use alternate biofluids. Amniotic and CSF (Figure 3a) are great for detecting changes in their neighboring tissues or organs but are hard to collect as they demand invasive procedures. Besides invasive liquid biopsies, alternate non‐invasive liquid biopsies also contain contributions from various organs and tissues and may indicate the health of the entire body (Figure 3b). They harbor the potential to detect systemic, organ‐specific, and tissue‐specific proteomics signatures. Body fluids such as urine, saliva, tears, and nasal secretion are non‐invasively acquired samples, which can provide adequate information about the ongoing pathological condition (Figure 3b). Urine is a non‐invasively collected fluid, which can be obtained in large quantities. Though the protein concentration in urine is low and varies according to the daily water intake, a single sample can still yield enough protein for proteomics analysis. A further benefit of using urine for proteomics studies is the ease with which samples can be obtained repeatedly over extended periods of time (Ponzini et al., 2022). A TMT‐based proteomics profiling of urine from COVID‐19 patients was performed, and the decrease in urinary ESCRT complex proteins was correlated with active replication of SARS‐CoV‐2 virus (Bi et al., 2022). Tears are another very simple‐to‐collect non‐invasive body fluid with low storage costs, harboring a high protein concentration and the ability to respond to ocular and systemic circumstances (Figure 3b). Based on the location, tears can predominantly indicate pathological disorders associated with the anterior region of the eye but can also reveal specifics about the status of the retina or vitrea in the eye. Saliva is another non‐invasive body fluid that is great for diagnostic purposes but is significantly impacted by microbiota in the oral cavity (Figure 3b). Human saliva proteomics has led to the discovery of protein biomarkers for the identification of various local and systemic disorders and has turned out to be an innovative method. However, the salivary proteome often poses problems due to its heavy contamination with oral cavity microbiome. A study used tear and saliva samples from Sjogren's syndrome and performed shotgun proteomics using isotopic dimethyl labeling. They identified 83 upregulated and 112 downregulated proteins in tears and 108 upregulated and 45 downregulated proteins in saliva (Das et al., 2021). Nasal mucus is another easy‐to‐collect body fluid and exhibits close proximity to the nervous system. The nasal mucosa is unique in that components of the olfactory neurons are embedded within it, providing a method to profile olfactory neuron secretions, which could be utilized as a source of biomarker discovery. The mass spectrometric analysis of nasal mucus from chronic rhinosinusitis patients with and without nasal polyps showed elevated cellular metabolism, altered immunologic pathways, and decreased cellular signaling (Kao et al., 2021). Body fluids are also a great source to detect sex‐specific changes in proteomes of diseased males and females. Diseases specific to females can be studied through secretions produced by females such as breast milk, cervicovaginal fluid, and amniotic fluid (Figure 3b). Cervicovaginal fluid is the discharge secreted from the vagina, cervix, or upper genital tract of females. Pathological situations can seriously disrupt the equilibrium of the healthy vaginal environment and cause changes in the quality and quantity of the proteins secreted in the cervicovaginal fluid. In a study, 29 enriched proteins were significantly elevated in the cervicovaginal fluid of pregnant women with a previous history of spontaneous preterm birth through shotgun proteomics analysis (Parry et al., 2020). Breast milk is another non‐invasively collected female secretion that contains epithelial cells, immune cells, and proteins (Figure 3b). Breast milk has been widely employed in studying breast cancer. Dysregulated proteins in breast milk such as galactosyltransferase, histone‐lysine methyltransferase, perilipin‐3 isoform 1, lactadherin isoform A, and recoverin have been identified and previously linked to the development of cancer (Aslebagh et al., 2018). Despite the diverse characteristics of body fluids, each one has distinct drawbacks. These limitations can be mitigated with appropriate strategies, making these body fluids desirable for proteomics analysis.

3. CONCLUSIONS AND OUTLOOK

There are an overwhelming number of omics approaches to study the functions of genes, proteins, and PTMs. However, changes in a disease state are not always reflected in a genome, transcriptome, or proteome alone. Single‐cell proteomics analysis is another emerging technology, which will complement scRNA‐seq analyses and help provide a cell‐to‐cell repertoire of information. Bioinformatics analysis also remains a challenge as new software and analytical approaches are constantly being created or ameliorated. Therefore, integrating various analytical platforms offers a more complete understanding of a disease. Still, integration is currently a challenge, and the cost of these techniques and data analysis currently limit their use. There are >300 PTMs in humans and we can still only analyze a subset of them. In clinical settings, PTMs analysis is practically never performed as it requires specialized tests that have been validated in thousands of patients. Current clinical tests typically use metabolomics, but these also present limitations. PTM analysis presents various benefits including knowledge about enzymatic activities of proteins that can become a new potential drug target. For example, protease and kinase inhibitors represent a high percentage of drugs used in the clinic. Despite the routine use of these drugs in the clinic, we frequently have only a vague understanding about their mechanisms of action in patients. Using PTM omics analyses will not only help in better prescribing the right drug at the right time for the patient but also generate new knowledge about disease mechanisms that can result in new preclinical drug programs about novel targets. These opportunities also present various challenges when studying proteomes, microbiomes, and PTMs in human biopsies. For example, the number of samples that can be collected at once (low‐throughput or high‐throughput collection), the complication of the collection procedure (invasive or non‐invasive), the timing of sample collection (e.g., before or after drug treatments, before or after flare ups), the heterogeneity between patients, and the opportunity for follow up sample collections (single sample vs. multiple collections over time) are all obstacles that need to be overcome. The choice of biopsy to study is another key factor, which can impact the outcome of a study. For example, serum/plasma is easier to collect and can be collected at various time points, but the analysis presents significant challenges such as the presence of highly abundant proteins. In contrast, tissue biopsies offer a large diversity of inflammatory or fibrotic protein signatures within a localized affected tissue but is typically highly invasive for the patient and can often be collected only once. Therefore, collections of multiple biopsies, when possible, from a single patient will offer an extended exploration of disease mechanisms and present a new perspective of dysregulated PTMs impacting immune pathways, cell signaling cascades, and tissue architectures. As new techniques emerge, additional data integration is required, but the resulting breakthroughs in knowledge offer new avenues of inhibiting a signaling pathway or an enzymatic cascade. Overcoming the limitations and challenges of omics approaches is critical to the exploration of complex protein PTMs that hold the secrets to mapping human diseases.

AUTHOR CONTRIBUTIONS

Antoine Dufour: Conceptualization; writing – original draft; funding acquisition; writing – review and editing; supervision. Sonali Bhardwaj: Writing – original draft. Mitchell Bulluss: Writing – original draft. Ana D'Aubeterre: Writing – original draft. Afshin Derakhshani: Writing – original draft. Regan Penner: Writing – original draft. MaryAnn Mahajan: Writing – review and editing. Vinit B. Mahajan: Writing – original draft; writing – review and editing; conceptualization; funding acquisition; supervision.

Bhardwaj S, Bulluss M, D'Aubeterre A, Derakhshani A, Penner R, Mahajan M, et al. Integrating the analysis of human biopsies using post‐translational modifications proteomics. Protein Science. 2024;33(4):e4979. 10.1002/pro.4979