Advancements in Global Phosphoproteomics Profiling: Overcoming Challenges in Sensitivity and Quantification

ABSTRACT

Protein phosphorylation introduces post‐genomic diversity to proteins, which plays a crucial role in various cellular activities. Elucidation of system‐wide signaling cascades requires high‐performance tools for precise identification and quantification of dynamics of site‐specific phosphorylation events. Recent advances in phosphoproteomic technologies have enabled the comprehensive mapping of the dynamic phosphoproteomic landscape, which has opened new avenues for exploring cell type‐specific functional networks underlying cellular functions and clinical phenotypes. Here, we provide an overview of the basics and challenges of phosphoproteomics, as well as the technological evolution and current state‐of‐the‐art global and quantitative phosphoproteomics methodologies. With a specific focus on highly sensitive platforms, we summarize recent trends and innovations in miniaturized sample preparation strategies for micro‐to‐nanoscale and single‐cell profiling, data‐independent acquisition mass spectrometry (DIA‐MS) for enhanced coverage, and quantitative phosphoproteomic pipelines for deep mapping of cell and disease biology. Each aspect of phosphoproteomic analysis presents unique challenges and opportunities for improvement and innovation. We specifically highlight evolving phosphoproteomic technologies that enable deep profiling from low‐input samples. Finally, we discuss the persistent challenges in phosphoproteomic technologies, including the feasibility of nanoscale and single‐cell phosphoproteomics, as well as future outlooks for biomedical applications.

Abbreviations

DDA

data‐dependent acquisition

DDM

n‐Dodecyl β‐D‐maltoside

DIA‐MS

data‐independent acquisition mass spectrometry

EGFR‐TKI

epidermal growth factor receptor‐tyrosine kinase inhibitor

IMAC

immobilized metal ion affinity chromatography

IPA

isopropanol

MOAC

metal oxide affinity chromatography

NSCLC

non‐small cell lung cancer

PASEF

parallel accumulation‐serial fragmentation

PTM

post‐translational modifications

PTS

phase transfer surfactant

RTK

receptor tyrosine kinases

SCP

single‐cell proteomics

SCX

strong cation exchange

SDC

sodium deoxycholate

SILAC

stable isotope labeling with amino acids in cell culture

TMT

tandem mass tag

1. Introduction

Protein phosphorylation is the most widespread post‐translational modification (PTM), in which a phosphate group is added to a protein, typically on serine, threonine, or tyrosine residue. The incorporation of phosphate groups elicits alterations in protein's physicochemical properties, thus changing protein structure (conformation), localization or protein‐protein interaction (interaction interfaces), and protein activity (function) [1, 2, 3]. Phosphorylation in response to cellular cues is a central mechanism of signal transduction that drives various cellular processes such as intracellular signaling, protein synthesis, cytoskeletal rearrangements, and cell‐cycle regulation. The dynamics of phosphorylation are tightly regulated by networks of kinases and phosphatases, which coordinate cellular responses to internal or external signals under physiological or pathological conditions. Thus, aberrant activation or dysregulation of phosphorylation signaling, driven by altered kinase or phosphatase activity, has been implicated in many diseases, such as neurodegenerative diseases, cardiovascular diseases, diabetes, and cancer [1, 2, 3, 4].

It has been estimated that approximately 70% of all encoded proteins undergo phosphorylation on at least one residue [4, 5]. As of July 2024, 240,178 phosphorylation sites have been reported in the human PhosphoSitePlus database [6]. The human genome encodes 538 kinases, classified into 10 families, and 226 phosphatases, classified into six families, collectively accounting for nearly 2.5% of protein‐coding genes. These enzymes collectively orchestrate reversible phosphorylation to regulate phosphorylation‐mediated signal transduction events [7]. The majority of protein kinases phosphorylate multiple targets (substrates) and are themselves targets for upstream kinases resulting in a chain of signaling events to regulate different cellular processes. Aberrant protein phosphorylation is closely linked to the initiation or progression of numerous diseases and cancers [8, 9]. Therefore, these kinases are attractive candidates for drug targets. There are over 80 kinase inhibitors that have been approved by the Food and Drug Administration in 2024 and 180 kinase inhibitors are in clinical trials worldwide [10, 11].

Among kinases, receptor tyrosine kinases (RTKs), class of transmembrane cell surface receptors with 58 known types in humans and categorized into 20 subgroups, represent the most attractive druggable targets, such as epidermal growth factor receptor (EGFR or ERBB) family, vascular endothelial growth factor receptor (VEGFR), and hepatocyte growth factor receptor (MET) [12]. Among them, members of the ERBB family, including EGFR (ERBB1/HER1), ERBB2/HER2, ERBB3/HER3, and ERBB4/HER4, are the most frequently dysregulated oncogenic drivers across human cancers. The EGFR signaling pathway is a well‐established example of a cell signaling cascade that governs cell growth, migration, metastasis, and death. Due to the critical role of cell signaling in essential cellular activities, alterations or mutations in kinases induce dysregulation of signaling cascades leading to different types of cancer including lung cancer [13], breast cancer [14], colorectal cancer [15], endometrial cancer [16], and head and neck carcinoma [17]. A well‐studied example of such dysregulation of cell signaling cascade is the activation and amplification of EGFR (kinase) by mutations (L858R, del19) in lung cancer, leading to autophosphorylation of EGFR and its accompanying downstream phosphorylation cascade. Therefore, EGFR has become the most common target of various tyrosine kinase inhibitors (TKIs), particularly in the treatment of lung cancer [18, 19]. Thus, measuring dynamic phosphorylation events is essential to understanding the elements of the phosphorylation network, their role in relaying signals/information across complex networks, and their regulatory functions, which may offer new therapeutic opportunities.

Mass spectrometry (MS)‐based phosphoproteomics has emerged as the leading tool for measuring global protein phosphorylation events [5]. Mapping the dynamics of system‐wide signaling networks to understand complex cellular activities remains a challenging task due to the labile nature of phosphorylation, its rapid and dynamic alterations, low stoichiometry and predominant presence of free peptides (unmodified peptides) [20, 21]. Unlike proteomic analysis, phosphoproteomics analysis involves the identification of phosphoproteins, phosphorylation sites, and quantification of phosphorylation sites or phosphorylation stoichiometry [2, 4]. Recent improvements in miniaturization of sample preparation approaches along with improvements in MS have enabled sophisticated phosphoproteomic studies with reliable identification, localization and quantification of thousands of phosphosites [22]. Compared to other omics technologies, the sensitivity of phosphoproteomics still lags behind, limiting the scope of analysis in situations in which only small amounts of sample are available. In this context, advances in phosphoproteomic technologies have enabled the characterization of tens of thousands of phosphopeptides while reducing the initial sample requirement from approximately 1 mg to just 100 µg for large‐scale analysis and clinical cohorts [23]. However, phosphoproteomic technologies tailored for large‐scale phosphoproteomic profiling are prone to sample loss due to extended processing times, tube surface adsorption during multiple sample transfer steps, and large processing volumes. Learning from the evolution of development of proteomics approaches, multiple approaches including Phospho‐SISPROT, tandem tip, Hybrid‐DIA, µPhos, and SOP‐Phos, have significant improved sensitivity, making it possible to perform microscale to nanoscale phosphoproteomics today [24, 25, 26, 27]. These technologies are therefore applicable to challenging samples such as subpopulation of cells or single‐spheroids [26].

In this review, we aim to provide an overview of the fundamentals of protein phosphorylation, recent innovations, and the state‐of‐the‐art global and quantitative phosphoproteomic methodologies. Focusing on highly sensitive phosphoproteomic methodologies, this review examines recent developments in miniaturized sample processing strategies, data‐independent acquisition MS, quantitative methods, and phosphosite localization tools, along with their potential for in‐depth mapping of disease biology. Compared to previous literature, we particularly summarize advancements in evolving phosphoproteomic technologies, emphasizing their performance in achieving in‐depth coverage with low sample inputs.

2. Phosphoproteomic Sample Preparation and Challenges

Sample preparation for mapping global phosphorylation events is complex, lengthy, and technically challenging in comparison to proteomic analysis [20]. Although comprehensive proteome profiling can be achieved with a few thousand cells as the initial sample input, phosphoproteomic analysis requires significantly more input (typically 100–200‐fold more starting material) to provide in‐depth profiling. [20, 21, 28]. This disparity arises due to the inherent scarcity of phosphorylated peptides, which constitute less than 1% of all unmodified peptides [5, 29]. Therefore, phosphoproteomic workflows involve additional steps to reduce the complexity of peptides and enrich phosphopeptides from a predominantly non‐phosphorylated peptide background prior to MS analysis. Although conventional approaches have enabled deep phosphoproteome coverage, they often require further optimization based on several factors, such as sample type, sample amount, and the number of conditions to be measured [5]. The basics of sample preparation and phosphopeptide enrichment are summarized in Sections 2.1 and 2.2, respectively (Figure 1).

FIGURE 1.

General phosphoproteomic sample preparation workflow and the principle of phosphopeptide enrichment. (a) The typical sample preparation workflow for MS‐based phosphoproteomics involves cell lysis, protein extraction, protein digestion, peptide cleanup, phosphopeptide enrichment, and sample desalting prior to LC‐MS/MS analysis. (b) Phosphopeptides are enriched by three commonly employed strategies, including IMAC, MOAC, and IP with anti‐phosphotyrosine antibodies. IMAC and MOAC exploit the electrostatic interaction between positive metal ions and the negative phosphate groups on phosphopeptides. However, competitive binding from acidic amino acids may occur at high pH, leading to non‐specific enrichment. Thus, specific enrichment can be performed in pH/acid‐controlled conditions to achieve high specificity. Under appropriate pH, carboxylic groups on acidic residues are protonated while leaving the phosphate group deprotonated in the acidic environment. Furthermore, the addition of acids, such as acetic and lactic acids, can prevent the binding of acidic amino acids. S. T. Y.: Serine (S), Threonine (T), Tyrosine (Y). E.D.: Glutamic acid (E), Aspartic Acid (D).

2.1. Pre‐Enrichment Sample Preparation

In a typical phosphoproteomic sample preparation workflow, tissue or cell lysis is performed with powerful detergents, salts, or chaotropic agents including guanidine hydrochloride (GdmCl), or phase transfer surfactant (PTS) buffers to lyse cells, extract and solubilize proteins [22, 30, 31, 32, 33]. The efficient protein extraction is often achieved with strong detergents (sodium dodecyl sulfate [SDS], sodium deoxycholate [SDC], sodium laurate sarcosine [SLS]) or combination of detergents (PTS) to solubilize proteins [33]. In conventional workflows, reduction and alkylation are performed separately due to the reaction of iodoacetamide (IAA) with dithiothreitol (DTT) and nearly consume one and a half hours of reaction incubation time [34]. Recently, tris(2‐carboxyethyl)phosphine (TCEP) and 2‐chloroacetamide (CAA) have been reported to perform reduction and alkylation in one step by directly adding these buffers into lysis buffer, which eliminates the need to perform reduction and alkylation as separate steps [35]. In typical proteomic workflow, the enzymatic reaction (tryptic digestion) is incubated overnight to produce peptide lengths that are compatible with MS analysis [34]. Lys‐C is another commonly employed enzyme used in combination with trypsin so that trypsin can gain access to previously inaccessible proteome sites and enhance digestion efficiency, with the goal of reducing missed cleavages that often occur near phosphorylation sites. In comparison to proteomics datasets, phosphoproteomics datasets may show slightly more mis‐cleaved sites which are more prevalent when a phosphorylated residue is proximal to a cleavage site (lysine or arginine). Following enzymatic digestion, the digested peptides can be either desalted or fractionated (strong cation exchange [SCX], high‐pH reversed‐phase chromatography, hydrophilic interaction chromatography (HILIC), electrostatic repulsion–hydrophilic interaction), followed by enrichment of phosphopeptides and subsequent LC‐MS/MS analysis [23, 28].

2.2. Phosphopeptide Enrichment Strategies

Due to the prevalence of free peptides, which significantly suppress the ionization and detection of phosphopeptides, phosphopeptide enrichment is a key step in phosphoproteomic analysis. Despite recent improvements in the sensitivity of mass spectrometers, the direct identification of phosphorylated peptides from proteome digests remains elusive [36]. Therefore, typical phosphoproteomic workflow always involves phosphopeptide enrichment prior to mass spectrometer injection and likely introduces the most variation into a standard phosphoproteomic workflow [29, 37]. A wide range of enrichment approaches, each with unique strengths and limitations based on chemical characteristics, has emerged as the field has advanced and is reviewed in detail elsewhere [21]. Below, we provide a brief summary of popular enrichment strategies used for phosphoproteomic studies. The principle of each method is illustrated in Figure 1.

2.2.1. Immobilized Metal Ion Affinity Chromatography (IMAC)

The classic phosphopeptide enrichment approach that leverages the affinity (electrostatic interaction) of negatively charged phosphate moieties toward positively charged metal ions such as Fe³⁺, Ga³⁺, Al³⁺, Zr⁴⁺, Co²⁺, Ti⁴⁺, or Ni²⁺. These metals are immobilized on a supporting matrix through chelating groups such as nitrilotriacetic acid or iminodiacetic acid which form the IMAC material [38]. IMAC enrichment is performed in pH/acid‐controlled conditions to achieve high specificity in which carboxylic groups on acidic residues, such as aspartate (D) and glutamate (E), are protonated while leaving the phosphate group deprotonated in the acidic environment of pH 2.5–3. This is due to the pKa of aspartic acid and glutamic acid being around 4 while the pKa of phosphate group is about 2 [39]. Furthermore, acids such as acetic acid or citric acid are commonly employed to minimize non‐specific binding by competing with non‐phosphorylated peptides. Both binding and washing conditions are performed at similar pH so that carboxylic groups of acidic amino acids are protonated and washed off from the metal. The bound phosphopeptides are then eluted off the resin with basic buffers or buffers (e.g., NH₄H₂PO₄) that out‐compete phosphopeptides for binding to IMAC. Despite being a widely used approach for phosphopeptide enrichment, it has several caveats in terms of (1) low specificity due to non‐specific binding of acidic residues of peptides and (2) low tolerance toward salts or other reagents used in sample preparation. One approach to circumvent non‐specific binding is O‐methyl esterification in which carboxylic acid groups in acidic amino acid residues are derivatized to their corresponding methyl ester prior to IMAC workflow to limit non‐specific binding. Although this approach improves overall specificity, it produces reaction products, such as partial hydrolysis of peptides and deamidation of asparagine and glutamine residues, which increase sample complexity [40]. Over the years, the development of commercially available IMAC resins and improved buffer conditions have greatly enhanced the robustness and specificity of IMAC methods. These improvements enabled IMAC‐based enrichment strategy successful and effective for large‐scale studies without compromising specificity and phosphopeptide coverage. Recently, it has been used in micro‐to‐nanoscale phosphoproteomics to maximize the recovery of phosphopeptides from low‐input samples [25].

2.2.2. Metal Oxide Affinity Chromatography (MOAC)

MOAC has emerged as a highly effective and promising approach for phosphopeptide enrichment. Several metal oxides such as titanium dioxide, gallium oxide, aluminum hydroxide, and zirconium dioxide have been applied in phosphoproteomic studies [21, 41]. Among these, titanium dioxide (TiO₂) is a commonly employed metal oxide resin due to its high selectivity and chemical stability. TiO₂‐based enrichment exploits the affinity of the negatively charged phosphate group of the phosphopeptides toward the positively charged metal oxide surface in a highly acidic condition. The optimal enrichment is achieved by adjusting pH between 2.5–3.0, enabling the protonation of acidic residuals while maintaining the negative charge on phosphate groups [39]. The competitive binding of acidic residues on non‐phosphopeptides is further reduced by employing hydroxy acids such as lactic acid, or glycolic acid because their affinity for TiO₂ is greater than acidic residues and out‐compete non‐phosphopeptides for binding to metal ions [42]. These reagents out‐compete non‐phosphopeptides for binding to metal ions. The enriched phosphopeptides are eluted off titanium dioxide with increasing pH or out‐compete phosphopeptides for binding to titanium dioxide. Since different metal oxides have different affinities toward different phosphopeptide species, tandem MOAC could be employed using complementary metal oxides to comprehensively capture different phosphopeptides species. For example, gallium has higher affinity toward doubly phosphorylated peptides while titanium has higher affinity toward singly phosphorylated peptides. In comparison to IMAC strategies, MOAC is tolerant toward buffers and salts used in the upper stream of sample processing workflow, making it ideal for streamlining and integrating multiple sample preparation steps. TiO₂‐based MOAC enrichment has been broadly employed by various sample processing workflows aimed at miniaturization for micro‐scale to nano‐scale phosphoproteomic profiling [23, 43].

2.2.3. Immunoprecipitation (IP)

Phosphorylated tyrosine residues constitute only a very small fraction of the total cellular protein phosphorylation, with an estimated relative abundance of 1800:200:1 for serine, threonine, and tyrosine in mammalian cells.  [44]. Although tyrosine phosphorylation is highly dynamic with extremely low abundance, it plays a significant role as an upstream regulator in numerous cellular processes [45]. Fortunately, antibodies against phosphorylated tyrosine have been demonstrated to be specific and reliable on a variety of protein substrates. However, these reagents are costly and thus, other strategies for selective enrichment of phosphorylated species have been explored [44]. Due to the low abundance, a relatively large amount of protein (in the levels of mg) is often required to achieve an adequate coverage of the tyrosine phosphoproteome. Recently, Callahan et al. developed the broad‐spectrum optimization of selective triggering (BOOST) approach, which leverages pY‐loaded carrier channel to boost MS/MS triggering. This method enabled the quantification of over 2000 unique phosphotyrosine sites using 1 mg of T cell receptor‐stimulated CD8+ T lysate (approximately 20 million cells) [46]. Chang et al. developed rapid‐robotic phosphotyrosine proteomics (R2‐pY, the next version of R2‐P2) platform to enrich phosphotyrosine peptides using magnetic particle and pY antibodies on 96‐well plates. This method identified 4000 unique phosphotyrosine sites from 1 mg of HeLa and serum peptides treated with pervanadate [47].

2.2.4. Combined Enrichment Strategies

Different enrichment methods leverage distinct chemical functionalities to bind phosphate groups, leading to the enrichment of distinct phosphopeptide species [48]. The choice of metal ions (e.g., Fe, Ti, Ga, Zr) and enrichment materials (e.g., IMAC, MOAC) directly determine the types of phosphopeptides that are enriched [49, 50]. Consequently, no single approach can effectively profile the entire phosphoproteome. Instead, integrating different enrichment strategies offers complementary enrichment to enhance identification coverage. For example, Fe‐IMAC preferentially enriches multiply phosphorylated peptides, while Ti‐MOAC favors monophosphorylated peptides, making it particularly useful for studies focusing on low‐abundance monophosphorylated peptides, such as early‐stage signaling events [21]. Moreover, sequential enrichment strategies have been developed to increase both the depth and diversity of phosphopeptide identification. For example, Tsai et al. exploited the distinct binding affinities of Ga³⁺ and Fe³⁺ to enrich a wide range of phosphopeptides, including mono‐ and multiply phosphorylated peptides, as well as acidic and basophilic phosphopeptides [39]. Similarly, Thingholm et al. developed sequential elution from IMAC (SIMAC), which combines IMAC and MOAC to enrich both mono‐ and multiply phosphorylated peptides [51]. Additionally, MOAC was combined with immobilized anti‐pY antibodies (pY‐MIP‐TiO₂) to enhance the coverage of the tyrosine phosphoproteome [52]. Although these approaches offer comprehensive coverage and are particularly useful when analyzing complex biological samples, they are often labor‐intensive due to their lengthy, multi‐step processes, which may not be suitable for low‐input samples. To address these limitations, new strategies that either improve single‐shot coverage or leverage multiple metal ions for comprehensive phosphopeptide enrichment are required for advancing low‐input phosphoproteomics.

3. Enhancing Sensitivity Through Miniaturized Sample Processing Strategies

The rapid advancements in sample preparation approaches for low sample input have played a pivotal role in the rise of single‐cell proteomics (SCP), which has moved beyond being a mere technical development into a powerful tool for illuminating cellular phenotypes and cell type‐specific functional networks underlying biological functions [53, 54, 55]. Mapping dynamics of protein phosphorylation in trace samples may provide insights into heterogeneity in cell signaling events (in response to stimuli or microenvironment) across cell populations and cell types [2]. However, unlike proteomics, phosphoproteomics has largely lagged behind due to the low stoichiometry of site‐specific phosphorylation and low abundance of phosphopeptide species, and the complexity of enrichment protocols [20, 21, 28]. The sensitivity of existing phosphoproteomic approaches remains to be further enhanced to carry out phosphoproteomics at the nanoscale or even single‐cell level [21, 56, 57]. Current phosphoproteomic approaches are facing challenges in achieving a balance between sensitivity, coverage and reproducibility for elucidating signaling pathways from mass‐limited samples [25, 26, 28, 58]. The considerable efforts made to enhance the overall analytical sensitivity of phosphoproteomics sample preparation are summarized below (Figure 2a–f).

FIGURE 2.

Miniaturized sample preparation workflows for microscale and nanoscale phosphoproteomics. (a) Mini‐analytical system consists of PTS based workflow, miniaturized LC column (25 µm I.D.), and direct sample injection system. (b) Phospho‐SISPROT workflow consists of three integrated tips for protein digestion, enrichment, and peptide desalting. (c) Tandem tip workflow consists of three tips in tandem for desalting, enrichment, and second desalting. (d) EasyPhos workflow consists of a 96‐well plate for high‐throughput digestion and enrichment. (e) R2‐P2 workflow for single‐pot solid‐phase‐enhanced sample preparation (SP3) and 96‐well format on a magnetic particle processing robot. (f) SOP‐Phos workflow consists of DDM‐coated one‐pot sample preparation from lysis to enrichment. PTS indicates phase‐transfer surfactant; SDC, sodium deoxycholate; SCX, strong cation exchange.

3.1. Mini‐Analytical System

Masuda et al. introduced a miniaturized phosphoproteomic system that consisted of (1) PTS‐based workflow, (2) miniaturized LC column (25 µm I.D.), and (3) direct sample injection system (Figure 2a) [58]. In this approach, cells were lysed with low concentrations of PTS buffer (12 mM SDC and SLS) and a large volume (70% v/v) of acetonitrile (ACN) was employed to avoid acid‐induced precipitation to prevent precipitation‐associated sample losses. Phosphopeptides were enriched by TiO₂‐hydroxy acid‐modified metal oxide chromatography (TiO₂‐HAMMOC method), in which hydrophilic hydroxy acids such as high concentration of lactic acid was used to improve the enrichment efficiency. Following sample enrichment, the clean phosphopeptides are directly loaded to a miniaturized analytical column. In comparison to conventional workflow, miniaturized phosphoproteomic system improved sensitivity by enabling identification of 1000 phosphorylation sites from only 10,000 HeLa cells. This system dramatically improved the sample recovery rate by skipping the precipitation step via ACN. Despite these improvements, the throughput of the system is low due to manual operation of the sample injection system.

3.2. Phospho‐SISPROT

To facilitate highly streamlined proteomic workflow, Chen et al. introduced a simple and integrated spintip‐based proteomics technology (SISPROT) consisting of SCX beads and a C18 disk in one pipette tip to achieve proteomics sample preparation and high‐pH RP‐based peptide fractionation [59]. This method achieved coverage of > 1000 proteins from 2000 HEK 293 cells. They further developed a phosphoproteomic variant, termed Phospho‐SISPROT, which couples three tips in tandem for (1) SISPROT for protein digestion, (2) Ti⁴⁺‐IMAC StageTip for phosphopeptide enrichment, and (3) a C18 membrane StageTip for peptide desalting (Figure 2b) [24]. Phospho‐SISPROT greatly enhanced the overall sensitivity by minimizing sample losses and reducing sample preparation time to 6 h. This system enabled the identification of 600–5500 phosphopeptides from 1–20 µg cell lysate treated with pervanadate to reduce phosphatase‐induced dephosphorylation. The phospho‐SISPROT is an easy‐to‐use strategy for sensitive phosphoproteomic profiling of low‐micrograms of protein samples.

3.3. R2‐P2

Leutert et al. introduced the rapid‐robotic phosphoproteomics (R2‐P2) platform, a novel magnetic particle‐based automated workflow for phosphoproteomic profiling of low‐input samples [60]. R2‐P2 is an end‐to‐end workflow that utilizes the single‐pot solid‐phase‐enhanced sample preparation (SP3) method to process proteomic extracts followed by phosphopeptide enrichment via IMAC to deliver MS‐ready phosphopeptides. The entire workflow is executed in a 96‐well format on a magnetic particle processing robot (KingFisher Flex), allowing high‐throughput sample processing that starts from cell lysates. The extracted proteins (from cell lysates) are captured by carboxylated beads to form carboxylated beads–protein complexes, which are then digested on beads and simultaneously eluted. At this point, the peptides can be used for either total proteome analysis (R2‐P1) or subjected to on‐plate phosphopeptide enrichment using either IMAC (Fe³⁺, Ti⁴⁺, Zr⁴⁺) or metal oxides (TiO₂) magnetic beads (Figure 2e). The R2‐P2 platform demonstrated high sensitivity by detecting 4000 distinct phosphopeptides from 25 µg proteins (1.2 × 10⁵ yeast cells), and high throughput by analyzing 101 environmental and chemical perturbations that resulted in signaling network map of 25,000 regulated phosphosites in yeast. Its next version, R2‐pY was developed to enrich phosphotyrosine peptides using magnetic particles and pY antibodies on 96‐well plates. R2‐pY enabled identification of over 4000 tyrosine‐phosphorylation sites from 1 mg pervanadate treated HeLa and serum peptides [47].

3.4. EasyPhos, Updated EasyPhos, and µPhos

Humphrey et al. developed EasyPhos workflow, a streamlined approach to enhance phosphoproteomic profiling with small‐sized sample inputs [22]. EasyPhos workflow eliminates the peptide clean‐up step before enrichment by utilizing TFE (2‐2‐2‐trifluoroethanol)‐based digestion buffer that is compatible with phosphopeptide enrichment. This workflow enabled identification of >10,000 phosphopeptides using 1 mg protein starting amount. Its updated version, reported in 2018, further enhances the sensitivity by skipping protein precipitation step before enrichment, and improves throughput by performing digestion and enrichment in a 96‐well plate (Figure 2d) [23]. The updated EasyPhos simplified the transition from proteolysis to enrichment by utilizing isopropanol (IPA) to prevent acid‐induced precipitation of SDC buffer prior to phosphopeptide enrichment step, thereby reducing five‐fold sample input without compromising phosphoproteomics coverage. This revised workflow enhanced sensitivity by detecting 20,132 phosphopeptides from 200 µg proteins of EGF‐stimulated glioblastoma cells and > 4000 phosphopeptides with only 12.5 µg initial amount. [23]. In 2024, its newest version, microPhos (µPhos) was introduced as a scalable and sensitive phosphoproteomic platform for low‐input samples. In comparison to updated EasyPhos, µPhos enhanced sensitivity and reproducibility by (1) reducing volume from ∼1 mL to ∼100 µL, (2) processing 96 samples in parallel while minimizing transfer steps and hands‐on time [27]. The µPhos protocol combined with dia‐PASEF enabled quantification of > 30,000 and > 10,000 phosphosites from 20 µg and 1 µg HeLa lysate, respectively. Its application in mouse brain tissue revealed distinct phosphorylation pattern inferring spatially different kinase activities in subregions of brain. The scalability of µPhos was analyzed using time‐course analysis of leukemia cells treated with TKIs which uncovered drug‐specific responses [27].

3.5. Tandem Tip Phosphoproteomics

Tsai et al. developed a streamlined tandem tip‐based sample processing workflow for phosphoproteomic profiling of low‐input samples [25]. Tandem tip workflow streamlines the entire workflow by integrating three tips for (1) digestion/pre‐enrichment cleanup (C18 disk tip), (2) IMAC‐based tip for phosphopeptide enrichment, and (3) second cleanup tip for post‐enrichment desalting (Figure 2c). This approach combined with DIA identified 3000–9500 phosphopeptides from 1–10 µg proteins (5 × 10³–5 × 10⁴ cells). Combined with iBASIL strategy using 1 µg protein in the boosting channel, ∼600 phosphopeptides were identified from as low as 100 FACS sorted cells (1 µg protein for boosting signal from low sample input). The utility of this workflow was demonstrated by using small human spleen tissue which resolved distinct regions (white and red pulp regions) of the tissue. Each region was populated with different types of immune cells and exhibited distinct phosphorylated events on proteins, including LSP1, MAP2K2, and CD20, which are involved in immune responses [25].

3.6. SOP‐Phos‐DIA

To further reduce sample loss due to the multistep sample transfer, Muneer et al. introduced a simple and rapid one‐pot phosphoproteomics workflow (SOP‐Phos) for microscale phosphoproteomic analysis [43]. The SOP‐Phos workflow capitalizes one‐step cell lysis/protein extraction, reduction/alkylation, followed by direct proteolysis and phosphopeptide enrichment in a single‐tube format to minimize sample transfer steps (Figure 2f). The sample losses were further minimized by utilizing n‐Dodecyl β‐D‐maltoside (DDM) pre‐coated tubes throughout the workflow, downscaling processing volumes to < 50 µL, and shortening enzymatic digestion time particularly for low‐input samples. SOP‐Phos, combined with library‐based DIA enabled the identification of > 30,000 and > 6500 phosphopeptides from 50,000 and 2500 PC9 lung cancer cells, respectively. Furthermore, its feasibility for deep mapping of disease biology (high coverage of EGFR and downstream phosphosites along EGFR‐TKI resistance pathways) was demonstrated in the mechanistic study of EGFR‐TKI resistant/sensitive lung cancer cells [43]. The SOP‐Phos‐DIA workflow does not rely on specialized instruments/reagents, and thus, can be easily adapted for routine sample preparation for microscale phosphoproteomics.

3.7. Other Recent Methodological Developments to Boost Sensitivity

Aside from above mentioned methodologies, other exciting efforts have been reported to optimize and miniaturize sample preparation workflows, aiming to reduce sample losses and initial sample requirement. We have attempted to list some of these research workflows to acknowledge their advancement in enhancing phosphoproteomic profiling sensitivity. For example, early workflows, including TiO₂–SIMAC–HILIC (TiSH) [61], TiO₂ with Tandem Fractionation (TAFT) [62], miniPhos‐SL‐TMT [63], and other multiple automated enrichment methodologies [64], reduced sample processing steps by streamlining the workflow (combining enrichment, fractionation or tandem mass tag [TMT]) or automating enrichment steps. More recent efforts, including miniPhos [65], hybrid‐DIA [26], CoolTip workflow [66], and RUPE‐Phospho workflows [67], have been introduced to improve various dimensions of phosphoproteomic technology. For example, miniPhos workflow integrated (1) lossless proteomic sample preparation using SDC detergent and (2) a miniaturized enrichment workflow (SpinTip device) for low‐input samples [65]. This workflow enabled identification of 22,000 and 4500 phosphosites from 100 µg and 10 µg samples, respectively. Its application to small‐sized (approximately 10 µg peptides) mouse brain samples revealed distinct phosphorylation events (on ERBB signaling, drug metabolism, and opioid signaling) that are likely to contribute to brain function spatially [65]. The RUPE‐phospho workflow employed ultrasound‐assisted phosphoproteomic sample preparation to minimize sample processing time and enhance throughput [67]. This workflow identified over 5325 phosphosites from 5 µg of human embryonic kidney HEK293T cell lysate within 3–4 h of processing time. Lately, a systematic optimization of the phosphoproteomic workflow (peptide‐to‐beads ratio, enrichment buffer [glycolic acid], elution buffer [ammonium hydroxide], binding time, sample and enrichment buffer volumes) was carried out for low‐input samples to enhance sensitivity by further integrating with DIA method. The resulting optimized workflow identified 35,000 and 32,000 phosphopeptides by processing only 60 and 30 µg HeLa peptides, respectively. The phosphoproteomic profiling of low‐input samples limits the coverage of regulatory phosphorylation events. To address this challenge, the hybrid data‐independent acquisition (Hybrid‐DIA) approach was developed that exploits the benefits of conventional DIA and targeted acquisition methods [26]. Hybrid‐DIA approach enabled sensitive phosphoproteomic profiling of single spheroids and quantification of the predefined signaling events. These phosphoproteomic technologies have made significant leaps in terms of overall improvements in sensitivity and throughput for biomedical applications.

4. Coverage in Large‐Scale Phosphoproteomics

Deep quantitative phosphoproteomic profiling is commonly achieved either through extensive peptide fractionation or comprehensive phosphopeptide enrichment workflows, such as combination of TiO₂ and immunoaffinity or other strategies to increase tyrosine phosphorylation or low‐abundant phosphopeptides [5]. However, the increased sampling depth incurs substantial costs in data acquisition and is further compounded by additional overhead in sample preparation and data analysis. Advances in phosphoproteomic sample preparation approaches along with recent improvements in MS scanning speed and innovative data acquisition strategies have enabled the characterization of tens of thousands of phosphopeptides per experiment and improved identification of low‐abundance phosphosites from low amounts of starting samples [22, 60, 68, 69]. For comparison of the profiling sensitivity in terms of the number of identified phosphopeptides per ng of sample, Figure 3 shows a summary of the evolving methods reported after 2015 (see more details in Table 1). The essential role of the increased phosphoproteomics coverage in advancing our understanding of cell signaling can be exemplified by the continuing efforts of the model HeLa cells. Early phosphoproteomic investigations by Olsen et al. enabled the quantification of > 6000 phosphosites on > 2000 proteins which unveiled site‐specific phosphorylation dynamics in epidermal growth factor (EGF)‐treated HeLa cells [70]. This depth was further enhanced by using an extensive fractionation strategy along with a high‐resolution orbitrap mass spectrometer, which enabled the mapping of over 20,000 phosphosites and thus advanced our understanding of the complexity of the phosphoproteome in biological systems [70, 71]. Previous deep phosphoproteomic profiling in HeLa cells identified > 38,000 phosphosites from 51,000 phosphopeptides through 270 LC‐MS/MS experiments over 40 days of data acquisition time [5]. This in‐depth investigation of the phosphoproteome revealed the extent of protein phosphorylation, particularly tyrosine phosphorylation, in the proteome of HeLa cells. However, this level of phosphoproteomic depth was achieved using a large number of samples (270 experiments) and substantial input amount (8–12.5 mg), which limits its application for low‐input samples, such as small cell population and clinical specimens. To increase depth and throughput while reducing sample amount, Olsen et al. developed a 15‐min single‐shot DIA workflow, which enabled the identification of over  20,000 phosphopeptides with direct DIA and  >29,000 phosphopeptides with project‐specific library‐based DIA using 200 µg protein [69]. Using lung cancer as a model, we reported a DIA workflow using hybrid spectral libraries constructed from complementary data‐dependent acquisition (DDA) and DIA datasets using cell lines and tissues to enable deep quantitative phosphoproteomics profiling of low‐input samples [28]. This workflow allowed the quantification of over 21,000 phosphosites with direct DIA and over 38,000 phosphosites with the hybrid spectral library DIA from 200 µg of NSCLC (non‐small cell lung cancer) and CL68 cell lysate peptides. At lower levels of 50 µg and 100 µg of peptides, more than 20,000 and 30,000 phosphopeptides were identified, respectively, using library‐based DIA [28]. The feasibility of this strategy was demonstrated on drug‐resistant patient‐derived lung cancer tissues, which showed site‐specific phosphorylation events associated with tumor progression and EGFR‐TKI (epidermal growth factor receptor‐tyrosine kinase inhibitor) resistance. In summary, improvements in sample processing workflows, MS technology, and data acquisition strategies have allowedlarge‐scale phosphoproteomics analysis, while significantly reducing the initial sample requirement from around 2 mg to less than100 µg for cell lines and clinical specimens.

FIGURE 3.

Summary of recent phosphoproteomic approaches reported between 2015 and 2024, July. The figure shows the performance in the sensitivity, that is, number of phosphopeptides (shown with size of circle) identified per nanogram (ng) of samples under different sample input amounts.

TABLE 1.

Overview of representative miniaturized phosphoproteomics workflow aimed at reducing sample initial amount.

Method name	Cell type	Labelling approach	Data acquisition	MS instrument	Input amount (µg)	P‐peptide or site	References
Mini‐analytical	HeLa cells	Label‐free	DDA	Velos	2	1011	[58]
Auto phospho	HeLa cells	Label‐free	DDA	QE Plus	200	8269	[64]
					10	5726
					1	1903
					0.1	319
	Neuron cells				25	3615
					1	1963
Phospho‐SISPROT	HEK 293T cells	Label‐free	DDA	Fusion	20	5500	[24]
					10	3000
					5	1900
					1	600
Updated EasyPhos	U‐87 cells	Label‐free	DDA	QE HF‐X	100	11,000	[23]
			DDA‐MBR			14,500
			DDA		12.5	4000
			DDA‐MBR			8000
R2‐P2	Yeast	Label‐free	DDA	Lumos	25	4000	[60]
Single‐embryo phosphoproteomics	Frog embryo	Label‐free	DDA	QE HF	20	4583	[72]
Tandem tip	Leukemia	Label‐free	DDA‐MBR	Lumos	0.1	200	[25]
			DDA		10	9180
			Hybrid DIA		10	18,100
			DDA		1	7535
			Hybrid DIA		1	9180
	MCF10A cells	iBASIL	DDA		0.02	600
Hybrid‐DIA	A552 cells	Label‐free	Hybrid DIA	Exploris 480	2.5	2500	[26]
					5	3000
					10	3500
					20	4000
MiniPhos	HeLa cells	Label‐free	dia‐PASEF	timsTOF Pro 2	10	4500	[65]
					100	22,000
µPhos	HeLa cells	Label‐free	dia‐PASEF	timsTOF HT/Ultra	1	10,000	[27]
					20	30,000
STF	HeLa cells	TMT	DDA	Exploris	12.5	5485	[129]
					1	4577
SOP‐Phos	PC9 cells	Label‐free	lib‐DIA	Eclipse	10	30,433	[43]
					5	28,489
					1	23,693
					0.5	6548

Phosphoproteomic technology has made a significant leap in sensitivity, enabling its use in situations in which sample availability is limited. The optimal amount required for phosphoproteomic analysis has been reduced from 200 µg to tens of micrograms (micro‐scale), making it possible to analyze the phosphoproteome of small population of cells or single spheroids [26]. For example, the original EasyPhos workflow quantified over 10,000 phosphopeptides from approximately 1 mg of protein starting material [22]. Later, the updated EasyPhos workflow enabled the quantification of more than 20,000 and over 4000 distinct phosphopeptides from 200 µg and 12.5 µg of glioblastoma cell lysate proteins, respectively, unveiling major regulatory sites in EGFR signaling pathway [23]. Other workflows, such as automated enrichment platforms, MiniPhos, RUPE‐Phospho, R2‐P2, CoolTip, and Hybrid‐DIA, have pushed the boundaries of microscale phosphoproteomics, but at the cost of reduced identification coverage. For example, the Hybrid‐DIA workflow, empowered by the combined strengths of DIA and PRM, identified over 8000 phosphosites from single spheroids and allowed interrogation of drug effects on spheroids compared to monolayer cells [26]. Moreover, phosphoproteomic profiling of single‐embryo (∼20 µg proteins) detected over 4000 phosphosites, revealing dynamic phosphosites map during cell cycle transitions [72]. A tandem tip method coupled with library‐based DIA successfully identified over 9000 and 18,000 phosphopeptides from only 1 µg and 10 µg protein lysate, respectively [25]. Combined with the TMT‐boosting strategy, nano‐scale phosphoproteomics was attempted by profiling 100 EGF‐treated MCF10A cells (20 ng, boosted with 1 µg protein), resulting in the identification of 600 phosphopeptides that mapped the upregulated ERBB signaling cascade in EGFR‐treated MCF10A cells [25]. Unlike proteomics, phosphoproteomics studies have often used lysate or post‐digestion peptides to estimate starting amount. By counting cell numbers, the SOP‐Phos‐DIA workflow managed to identify over 30,000 and 6500 phosphopeptides from 50,000 (∼10 µg) and 2500 (∼0.5 µg) NSCLC PC9 lung cancer cells, respectively, covering key regulatory events in EGFR signaling pathway [43]. In summary, these advances successfully demonstrated highly sensitive phosphoproteomics down to microscale and nanoscale sample input, highlighting their potential for broader applications in biomedical research.

The integration of phosphoproteomics into multi‐omics frameworks provides a comprehensive understanding of regulatory mechanisms governing cellular processes and identifies druggable pathways for therapeutic development. Recent developments in multi‐omics, including proteogenomics, have been reviewed elsewhere and therefore will not be discussed here in depth [73, 74]. While multi‐omics approaches have provided important biological insights, they often require substantial sample amount to achieve sufficient phosphoproteomic coverage [75]. This limitation has often restricted multi‐omics studies to genomics, transcriptomics, and proteomics, with phosphoproteomics frequently being excluded due to perennial issues related to sample size and availability [76]. Conventional multi‐omics or proteogenomic workflows typically require tens to hundreds of milligrams of tumor tissue, which yield over 30,000 phosphosites per sample [77]. To improve sensitivity, Satpathy et al. introduced a microscale proteogenomic method that requires as little as 50–100 µg of protein for comprehensive proteogenomic analysis [75]. This method enabled the identification of over 25,000 phosphosites from core‐needle biopsy samples, using sample sizes at least five times smaller than the requirement from the Clinical Proteomic Tumor Analysis Consortium. Moreover, recent advances have facilitated phosphoproteomics in conjunction with other omics on the same samples. Notably, Rolfs et al. showcased the feasibility of performing phosphoproteomic analysis on RNA‐extracted samples [78]. Additionally, several efforts have been made to integrate phosphoproteomics with global proteomics, glycoproteomics, and even metabolomics to maximize biological insights [76, 79, 80]. Nevertheless, significant advancements in protocol optimization are still required to reduce material input for simultaneous DNA, RNA, protein, and phosphoproteomic analysis.

5. Data‐Independent Acquisition Enhances Identification Coverage

Advancements in data acquisition speeds and data analysis software tools have led to the rapid evolution of data‐independent acquisition as a powerful alternative for proteomic profiling with unprecedented sensitivity, coverage, and reproducibility [81]. Although, compared to proteomics, the complexity of DIA‐based global phosphoproteomics spectra poses substantial challenges in fragmentation spectra deconvolution and unambiguous site localization, making identification, quantification, and phosphosite localization often more difficult [28, 82]. Recently, DIA has demonstrated its superiority over DDA in terms of identification coverage, quantitative reproducibility (by reducing the missing values), particularly for global protein phosphorylation measurements [28, 69]. Early applications of DIA in phosphoproteomics focused on targeted analyses, such as insulin signaling networks [83], histone modifications [84, 85], and plasma proteins [86]. Later, DIA‐based targeted phosphopeptide quantification demonstrated high accuracy compared to the selective reaction monitoring method, as well as the capability to differentiate and quantify positionally isomeric phosphopeptides [87].

The accessibility of DIA‐based global phosphoproteomic measurements has significantly improved with the integration of localization algorithms into Spectronaut software (Biognosys) [88], DIA‐NN [89], and MaxDIA [90], addressing challenges arising from the complexities of DIA spectra. A pioneering DIA‐based study by Olsen et al. demonstrated deep phosphoproteomic coverage with DIA, and found that DIA‐based measurements significantly increase phosphoproteome coverage by approximately 3‐fold in comparison to DDA. while enabling reproducible quantification of over 10,000 phosphosites across hundreds of samples [69]. Alternatively, peptide‐centric analysis (also known as library‐based DIA), which involves mapping DIA‐MS² spectra to peptide fragment ion spectral library [91], has emerged as a promising strategy for deep phosphoproteomics. Using lung cancer as a proof‐of‐concept model, we showed that a deep hybrid (DDA and DIA dataset) library, containing 88,107 phosphosites derived from lung cancer cell lines and patient‐derived tissues, effectively increases the single‐shot LC‐MS/MS profiling coverage [28]. Using this hybrid spectral library, over 36,000 phosphosites were quantified in a single‐shot DIA, unveiling the feasibility of sensitive DIA for in‐depth coverage [28]. These studies highlight DIA as a promising alternative for achieving reproducible sampling and improved coverage in large‐scale phosphoproteomics.

Recently, DIA has been explored to enhance the phosphoproteomic depth from low‐input samples. For example, a streamlined tandem tip‐based workflowenabled quantification of > 18,000 phosphopeptides with DIA (9180 phosphopeptides identified with DDA) from 10 µg peptides along with improved quantitation in comparison to DDA method [25]. Martinez‐Val et al. introduced an intelligent data acquisition method, termed Hybrid‐DIA, which combines strengths of DIA for global profiling with multiplexed PRM for targeted quantitation to increase detection and quantification of signaling events [26]. The application of Hybrid‐DIA method demonstrated sensitive phosphoproteomics from single spheroids and targeted quantification of key signaling events (EGFR, AKT, FOXO, JUN, GS3K, mTOR) in EGFR signaling pathway and downstream MEK and PI3K kinase pathways. However, this workflow requires optimization and is compatible only with certain instruments. At low sample inputs, the size of the phosphopeptide spectral library has a more profound impact on the identification depth. We previously highlighted that peptide spectral libraries established with comparable sample input achieve optimal identification coverage [92]. Using one‐pot phosphoproteomics workflow, we evaluated the performance of classic direct DIA and sample‐size comparable library‐based DIA for microscale phosphoproteomics. We found direct DIA enhanced (2‐fold more unique phosphosites), lower missing values (< 1%), and improved reproducibility (8%–10% CV) compared to DDA [43]. The size‐comparable library DIA enabled the identification of > 30,000 and > 6500 phosphopeptides from 50,000 (∼10 µg) and 2500 (∼0.5 µg) PC9 cells, covering important regulatory sites and druggable targets.

With the launch of timsTOF platform, Meier et al. demonstrated parallel accumulation‐serial fragmentation (PASEF) in 2015, and further exploited its capabilities with DIA method, a novel scan mode termed dia‐PASEF [93, 94]. As the timsTOF mass spectrometer is relatively a recently released instrument, only a limited number of studies have utilized the dia‐PASEF technology for phosphoproteomic analysis. Skowronek et al. leveraged dia‐PASEF method with project‐specific libraries which enabled identification of over 46,000 phosphosites in only 21 min from 100 µg EGF‐stimulated HeLa cell digests [68]. Oliinyk et al. quantified over 9700 phosphopeptides from 20 µg HeLa digest using only a 7‐min gradient [95]. More recently, the µPhos workflow, integrated with dia‐PASEF, enabled quantification of > 30,000 and > 10,000 phosphosites from 20 µg and 1 µg HeLa lysates, respectively [27]. In summary, dia‐PASEF on the timsTOF platform enabled rapid phosphoproteome analyses without sacrificing depth or sensitivity.

6. Quantitative Phosphoproteomics for Small Sample Input

Generally, there are two types of quantitative measurements of phosphorylation events: (1) quantitation of relative changes in phosphorylation and (2) quantitation of phosphorylation stoichiometry [96]. The majority of phosphoproteomics applications focus on measuring relative changes in site‐specific phosphorylation under different events of cellular responses or phenotypic conditions. These dynamics in signal transduction networks often depend on activation of transient, protein‐ and site‐specific phosphorylation levels, which are not necessarily correlated with protein expression levels. The second category is quantitation of phosphorylation stoichiometry, which is defined as the ratio of the total amount of protein phosphorylated on a specific site to the total amount of protein [97]. The majority of quantitative phosphoproteomic investigations are based on the incorporation of stable isotopes by metabolic and chemical labeling approaches. The general principles of both metabolic and chemical labeling approaches have been reviewed elsewhere and therefore will not be discussed here in depth [98]. The simplest of these approaches is label‐free quantification, but the potential for sample preparation‐related inaccuracies is highest compared to label‐based methods [99]. In this section, we summarize different quantitative strategies and discuss their suitability and challenges in the context of quantitative phosphoproteomics (Figure 4a–c).

FIGURE 4.

Schematic overview of MS‐based quantitative phosphoproteomics approaches. (a) In label‐free approaches, every sample is individually processed and subjected to LC‐MS/MS analysis. Quantitation is performed by the peak area of extracted MS¹ or MS² ion chromatograms. (b) In metabolic labeling, amino acids with heavy isotopes are added to the cell culture medium. During cell growth, these amino acids are incorporated into proteins. Consequently, proteins from the heavy isotope condition can be distinguished from their light counterparts by a fixed mass shift. Quantitation is performed by the relative intensity or extracted ion chromatogram of the isotopic pairs. (c) In chemical labeling, isotope labeling‐based tags react with peptides (or protein) and samples are mixed and injected to LC‐MS/MS in one run. Quantitation is performed by comparing the relative intensities of the isotopic reporter ions. TMT indicates tandem mass tag.

6.1. Metabolic Labeling Strategies

In metabolic labeling strategy, stable isotope‐labeled amino acids are directly incorporated into the newly synthesized proteins in living cells. Stable isotope labeling with amino acids in cell culture (SILAC) is one of the pioneering and commonly used metabolic labeling strategies in which cells are grown in media containing either normal (light) or isotopically labeled (heavy) amino acids. Heavy (¹³C or ¹⁵N) and light (¹²C or ¹⁴N) lysine and arginine are the most commonly used amino acids in this approach [100]. Although SILAC labeling can be costly, particularly for in vivo applications, the protocol is relatively simple because labeling occurs during cell culture (or organism growth). During MS analysis, peptides are identified as pairs spaced by labeled and unlabeled forms, with their relative abundances quantified by comparing the peak area ratios of extracted ion chromatograms (Figure 4b) [101].

To date, a large number of quantitative phosphoproteomic studies are based on SILAC [102, 103, 104, 105]. This approach has become a widely adopted method for studying cancer signaling pathways due to its ability to provide accurate quantitative data on both protein expression and PTMs. For example, SILAC‐based protocol was used to investigate alterations in tyrosine phosphorylation following EGFR‐TKI treatment in lung cancer cells [106]. Cunningham et al. applied SILAC‐based global phosphoproteomics approach to map fibroblast growth factor receptor (FGFR) signaling networks in triple‐negative breast cancer [107]. Griffith et al. applied SILAC‐based phosphoproteomics to elucidate chimeric antigen receptor (CAR)‐mediated signal transduction in both CAR‐expressing and CAR‐targeted cells [108]. Recently, Ibáñez‐Molero et al. developed the HySic approach, hybrid quantification of SILAC‐labeled interacting cells, to measure signaling dynamics between tumor and T cells [109]. Zanivan et al. developed SILAC mouse model to investigate skin carcinogenesis at proteome and phosphoproteome level [104]. While SILAC is typically restricted to cell lines or whole organisms, the super‐SILAC approach, which uses SILAC‐labeled cells as spike‐in standards, has expanded its application to large‐scale clinical and tissue samples [110, 111]. The super‐SILAC strategy has been applied in vivo and in clinical applications, such as in mouse tissues [112], human lung tumor tissues [113], breast cancer tissues [110], and ovarian tumor tissues [114]. Moreover, the pulse‐chase SILAC method, which measures the rate of isotope incorporation by using two SILAC channels, has been utilized to quantify system‐wide protein turnover and explore how phosphorylation dynamically impacts this process [115, 116]. Wu et al. developed the DeltaSILAC approach to measure turnover rates for both modified and unmodified versions of the same peptide sequences from the same protein [105]. DeltaSILAC method demonstrated the regulatory role of site‐specific phosphorylation on protein turnover, revealing that many of these sites delay protein turnover in growing cancer cells [105]. With DeltaSILAC, the role of phosphorylation in regulating protein half‐lives have previously been identified, specifically in ubiquitin S65 phosphorylation and arginine phosphorylation, both of which serve as markers for protein degradation [117, 118]. Overall, SILAC remains a powerful and evolving tool in cell signaling research, yet the complexity, limited multiplexity, and cost, especially for in vivo experiments, can be limiting factors.

6.2. Chemical Labeling Strategies

Chemical labeling approaches have made significant advancements, using a variety of chemical tags or reagents to label peptides or proteins, which have been widely applied to different sample types [119]. Stable isotope dimethyl labeling, for instance, is a simple and cost‐effective labeling approach suitable for large‐scale comparative analysis [120, 121]. This approach is commonly used for the comparing two samples, particularly when relatively large amount of sample is available. For example, Senturk et al. used demethylation‐based quantitative phosphoproteomics to investigate aberrant phosphorylation signaling pathways in clear cell renal cell carcinoma [122]. Isobaric labels, including TMT and iTRAQ (isobaric tags for relative and absolute quantification), are the most commonly used chemical labeling approaches due to their advantages over chemical isotopic labeling (dimethyl labeling) and label‐free quantification strategies. These advantages include enhanced sensitivity, increased sample throughput, and superior quantitative accuracy and reproducibility (Figure 4c) [119]. In contrast to dimethyl labeling and SILAC, isobaric tags (e.g., TMT) offer substantially higher multiplexing capabilities, with up to 18 samples [123] and increased sample throughput.

Several TMT‐based quantitative phosphoproteomics protocols have been developed that reduce the quantity of labeling reagents and the initial sample input to enhance quantitative sensitivity. For example, Navarrete‐Perea et al. developed a streamlined TMT (SL‐TMT) workflow for quantitative proteomics and phosphoproteomics of low‐input samples [63]. Tsai et al. introduced a signal boosting strategy with isobaric labeling, termed as iBASIL (improved boosting to amplify signal with isobaric labeling), to quantify low‐abundance phosphopeptides. With this approach, more than 20,000 phosphorylation sites from human pancreatic islets (2 × 10⁵ cells) were quantified [124]. Thisapproach was further applied for deep quantitative measurement of low‐abundance tyrosine phosphorylation events [125, 126]. Ogata et al. developed nanoscale StageTip‐based solid‐phase reactor for TMT‐labeling of phosphopeptides after enrichment, which enabled the quantification of 10,000 phosphosites from only 50 µg peptide amount [127]. More recently, a streamlined tandem tip enabled quantification of 700 phosphopeptides from human spleen tissue voxels with a spatial resolution of 200 µm [25]. Additionaly, Koenig et al. reported a high‐throughput semi‐automated protocol for label‐free‐ or TMT‐based phosphoproteomics [128, 129]. By combining TMT labeling with high‐pH fractionation, they reported quantification of 13,250, 5485, and 4577 phosphopeptides from 12.5, 2.5, and 1 µg peptide inputs per TMT channel, respectively [129]. In summary, these workflows have significantly boosted the sensitivity and enabled multiplexed quantitative phosphoproteomics.

6.3. Label‐Free Quantification Strategy

Label‐free strategy, as the name suggests, do not require any chemical or metabolic tags on samples. Theoretically, an unlimited number of samples can be quantitatively compared; however, this comes at the cost of increased instrument time [99]. In this strategy, each sample is individually processed and measured by MS, making sample handling (especially during the phosphopeptide enrichment step) and chromatographic reproducibility critical for accurate quantification (Figure 4a) [20]. Therefore, most of the challenges with label‐free quantitation arise during post‐acquisition data analysis which can be circumvented by creating streamlined and reproducible pipelines. Rapid advancements in such pipelines and instrumentation have overcome these challenges and greatly expanded the popularity of label‐free strategies for phosphoproteomics [28, 45, 69]. The quantitation of a given phosphosite is typically achieved by comparing the peak areas of the chromatographic elution profiles of precursor or fragment ions across different biological conditions. Several large‐scale studies have successfully applied label‐free approaches to investigate different biological conditions. For example, Humphrey et al. employed precursor‐level label‐free quantification to investigate the responses of U‐87 human glioblastoma cells tost EGF treatment [23]. Similarly, Kitata et al. utilized a similar approach based on fragment‐level label‐free quantification to investigate differential expression of the quantified phosphosites between EGFR‐TKI‐sensitive and resistant lung cancer cells and tissues [28]. Label‐free quantification strategies offer several advantages for low‐input or sensitive phosphoproteomics, including a simple sample preparation workflow, flexibility to couple to different enrichment and fractionation strategies, broad dynamic range for quantitation, and versatility in sample types. Therefore, these strategies are simple to implement in high‐sensitive and high‐throughput sample preparation. For example, several reported miniaturized workflows (Session 3), such as EasyPhos [23], Phospho‐SISPROT [24], mini‐analytical system [58], R2‐P2 [60], tandem tip system, and SOP‐Phos [25, 43], are all based on label‐free approaches.

6.4. Phosphorylation Stoichiometry

Phosphorylation stoichiometry is the absolute percentage occupancy of a specific phosphosite that is calculated by comparing the abundance of phosphopeptide relative to the total abundance of corresponding protein. The stoichiometry of a phosphorylation site provides an understanding of the degree of phosphorylation which distinguishes novel functional sites from non‐functional phosphorylation events in regular signaling networks. Although, all the above‐mentioned isotopic labeling and label‐free methods can be used, determining phosphorylation stoichiometry is technically challenging due to low phosphorylation occupancy, dynamic and rapid alterations, and requirement for the presence of both phosphopeptide and corresponding non‐phosphorylated peptides [21, 29, 70]. Compared to the broad applications of relative quantitative phosphorylation, there is limited literature reporting methodologies for the measurement of phosphorylation stoichiometry. By measuring protein and phosphorylation levels, Wu et al. first raised awareness of both independent and concerted changes in protein expression and phosphorylation levels, emphasizing the need to normalizate data for the correct interpretation of dynamics of phosphorylation events [130]. Under basal conditions, low‐occupancy phosphorylated sites are predominant due to minimal activation of kinases and downstream signaling cascades [5]. Several conditions, including kinase‐phosphatase activities, activation or inhibition of signaling pathways, cell cycle states and other cellular conditions, can alter phosphorylation site stoichiometry. Several approaches have been developed to globally estimate the dynamic stoichiometries of phosphorylation sites [5]. A pioneering study by Olsen et al. reported that mitosis‐associated proteins exhibit over 75% site occupancy in HeLa cells during mitosis [70]. Wu et al. determined the distribution of site stoichiometries for 5033 phosphorylation sites in yeast and observed over 90% phosphorylation occupancy in 10% of the sites during exponentially growing yeast [131]. These investigations have uncovered that phosphorylation events are often low in abundance and occur at low stoichiometry; however, substantial fractions of sites exhibit very high phosphorylation occupancy under specific conditions, such as during mitosis. We reported a motif‐targeting quantitative approach to quantify phosphorylation stoichiometry of > 1000 phosphorylation sites in lung cancer cells and found dramatic differences in phosphorylation occupancy between drug‐resistant and drug‐sensitive lung cancer cells, identifying potential drug targets associated with acquired drug resistance [97]. Further analysis confirmed that S102 phosphorylation of HMGA1 is a critical node contributing to EGFR‐TKI resistance and its site mutagenesis reinforced the efficacy of gefitinib in resistant NSCLC cells through reactivation of the downstream EGFR signaling [132]. These investigations reveal the critical role of phosphorylation degree associated with aberrantly activated pathways in fundamental cellular processes and cancer biology.

7. Challenges in Phosphopeptide Identification and Site Localization

Current DIA‐based phosphoproteomic data analysis is constrained by the need to construct sample‐size comparable spectral libraries to boost identification coverage [43, 92, 133]. However, this approach often requires additional samples for library generation and efforts to scale libraries to the sample size, which may not always be feasible when sample amounts are limited. These custom libraries are also prone to excluding critical phosphorylation sites during the localization filtering step, limiting the detection of biologically important phosphorylation events [69]. Additionally, low‐abundance phosphopeptides are likely to get under‐scored for localization during library‐based DIA searches, which affects downstream analyses since only confidently localized class‐1 sites are typically considered [43, 69, 134]. To overcome these challenges, innovative data analysis approaches, particularly those leveraging machine learning and artificial intelligence models, are required to be developed.

The main advantage of MS‐based phosphoproteomics is its ability to provide site‐specific information on phosphorylation events at a system‐wide level [22, 60]. After identification of phosphopeptides, accurate localization of phosphorylated sites allows further analysis (differential analysis or pathway mapping) and functional characterization of the identified sites. Due to the low ionization efficiency of phosphopeptides, their fragmentation ions have lower intensities and less complete patterns compared to those from free peptides [135]. Furthermore, the labile nature of the phosphate moiety on the phosphopeptides often results in its neutral loss during fragmentation [135]. Phosphopeptide sequencing by MS/MS is particularly challenging challenging for low‐input samples due to two main issues: (1) loss of phosphosite during fragmentation and (2) difficulties in assigning phosphorylation probabilities to multiple potential sites (due to occurrence of phosphopeptide isomers with different phosphorylation site on the same peptide sequence) [136].

A common observation in phosphopeptide detection is the neutral loss of the phosphoryl group during the fragmentation step, potentially leading to unassigned phosphorylation sites and complicating its quantification [135]. Common fragmentation methods, such as collision‐induced dissociation (CID) and higher‐energy collisional dissociation (HCD) suffer from neutral loss of the phosphoryl group [29]. More advanced techniques, such as electron transfer dissociation (ETD), and a combination of ETD and HCD, generate rich MS/MS spectra with both b/y‐ and c/z‐type fragment ions that result in more complete peptide sequence coverage and more confident localization of phosphosites [137].

Low‐input phosphopeptide samples further pose challenges in confident phosphopeptide identification and site localization due to their very low intensity of precursor and fragment ions carrying phosphate groups. To address these challenges, several algorithms including Mascot delta score [138], AScore [139], PTM score [45], PhosphoScore [140], Phosphinator [141], PhosphoRS [142], and LuciPHOr [143, 144] have been developed to provide statistical estimation of phosphorylation site assignment. However, the majority of localization score algorithms provide arbitrary score cut‐offs (such as 0.75 or 0.99 to report confidently localized sites) but do not estimate false localization rate (FLR). As false discovery rate (FDR) is calculated in typical practice of protein/peptide identification, it has been suggested to employ decoy amino acids phosphorylation strategy to estimate FLR of phosphosites. One such strategy is the LuciPHOr that leverages peak intensities, mass accuracy, and the distribution of localization delta scores to estimate FLR [143, 144]. It employs a modified target‐decoy approach that adds a phosphate group to each amino acid residue in the target peptide sequence, thereby comparing candidate (S, T, and Y) and non‐candidate residues (other than S, T, and Y) against each other. Another elegant strategy is site localization in peptide (SLIP) approach, which generates decoys by adding phosphosites to glutamic acid and proline residues in the target peptide sequence to estimate FLR [145]. Recently, a deep learning‐based approach (deepFLR) has been proposed to estimate FLR for phosphoproteomics data [146]. The deepFLR approach generates decoys by randomly substituting phosphorylated amino acids (S, T, and Y) with alternative amino acid (other than S, T, and Y) in the target peptide sequence, and then deep learning model is employed to predict MS/MS spectra for both target and decoy peptide sequences to estimate the FLR. Currently, there is no universally accepted FLR approach for phosphoproteomic studies. Recently, Locard‐Paulet et al. have reviewed over 20 phosphoproteomic analysis pipelines, highlighting their strengths and weaknesses [147]. Nevertheless, user‐friendly toolboxes with integration of informatic modules for phosphopeptide identification,  accurate localization of site along with FLR estimation that are compatible with different types of phosphoproteomics datasets are urgently needed.

8. Lagging Progress of Single‐Cell Phosphoproteomics

Non‐MS‐based approaches are able to quantify a few phosphorylation sites in single‐cell signaling analyses, but such approaches rely on targeted detection, which requires prior information and specific antibodies. Technologies, such as immunocytochemistry, flow cytometry, or CyTOF enable the detection of up to tens of phospho‐proteins in individual single cells by using antibody‐bound reporter species [148, 149]. However, these approaches are constrained by the need for highly specific antibodies and limited multiplexing capacity. Recently, a non‐enrichment‐based phosphopeptide identification approach has been proposed to identify PTMs, including phosphorylation, from MS‐based large‐scale single‐cell proteomic datasets. This approach employs a “variable modification search strategy” to identify PTMs from single‐cell datasets due to lack of sensitive platforms that makes “phosphopeptide enrichment impractical” as reported in previous studies [36, 150].

MS‐based SCP has recently enabled the identification and quantification of thousands of proteins across hundreds of single cells. However, tools for measuring PTMs, such as protein phosphorylation at the single‐cell level or in small cell populations, are still inaccessible due to the lack of sensitive integrated platforms to detect PTMs at such low‐input sample sizes [57, 151]. This limitation makes certain biological systems inaccessible to existing phosphoproteomic technologies, particularly those with inherently limited sample amounts, such as rare cell populations or single spheroid. A few studies have successfully profiled the phosphoproteomes of single embryos and single spheroids [26, 72]. These measurements were facilitated by the fact that each of these samples contained relatively large cell sizes with micrograms of proteins, which can be processed using existing sample preparation technologies. At the current stage of technological progress, the lowest cell number for phosphoproteomic analysis is approximately 100 MCF10A cells, achieved by boosting strategy with an input of 5000 cells [25]. Notably, phosphoproteomic profiling at such small sample sizes (nanoscale levels) has only been achieved using TMT‐based boosting strategies, while label‐free phosphoproteomic profiling has not been reported for cell inputs below 5000–10,000 cells [25, 43]. We recently introduced an integrated phosphoproteomic chip (iPhosChip) coupled with DIA MS (Chip‐DIA) which enabled the first nanoscale (1000 to 10 cells) and single‐cell phosphoproteomic analyses [133]. The iPhosChip functions as an “all‐in‐one station” that accommodates the cell imaging/capturing/counting and the entire sample preparation workflow in a highly streamlined and multiplexed format. Using sample‐size‐comparable libraries, Chip‐DIA identified an average of 1076 ± 158 to 15,869 ± 1898 phosphopeptides from 10 ± 0 to 1013 ± 4 cells and revealed single‐cell phosphoproteomic landscape of over 200 phosphorylation sites, which included cancer‐associated signaling events and druggable sites. The single‐cell phosphoproteomics appears limited to detecting highly expressed phosphorylation sites, which may provide less information about phosphorylation‐mediated signaling networks. Therefore, continuing technological development is essential to advance this aspect.

9. Remaining Challenges and Future Prospectives

Over the past decade, quantitative phosphoproteomics has made significant advances, enhancing our understanding of regulatory mechanisms underlying various cellular functions and diseases. In spite of the advances, the field still faces a number of challenges, including (1) sample amount requirement, (2) inaccuracies in localization, (3) biological functions of identified sites, and (4) limited applications in biological or clinical research.

Compared to breakthroughs in proteomics, technology development in phosphoproteomics still needs to overcome analytical challenges in sensitivity and coverage to expand its full spectrum of application, especially at nanoscale to single‐cell levels. One of the remaining challenges in phosphoproteomics lies in its application to inherently scarce samples. Although the current demonstrated limit of sensitivity is as low as few thousand cells, further miniaturization of technologies along with improvements in data acquisition and analysis are necessary to expand their potential.

Future advances in microscale‐to‐nanoscale phosphoproteomics, reaching eventually to true mammalian single‐cell phosphoproteomics, will be an important path toward unraveling cell signaling networks to tailor targeted therapies against activated signaling pathways in cancer. Moreover, such advancements will provide insights into single‐cell signaling dynamics to drive innovative drug development for precision medicine. Another major challenge is the ambiguity in phosphorylation site localization, which affects quantitation accuracy and downstream functional interpretation. Despite the availability of various computational approaches, there is still room for improvement in methods for confidently localizing phosphorylation sites. For example, in addition to calculating localization cut‐offs, a universally accepted approach for estimating FLR needs to be developed and incorporated into search software suites. Furthermore, the incorporation of machine learning, particularly deep learning techniques, is essential to boost sensitivity, address missing values, and drive new biological discoveries. Notably, deep learning approaches for library‐based DIA analysis will be critical in advancing DIA methodologies. Such frameworks hold the potential to effectively interrogate large‐scale phosphoproteomics datasets, particularly in elucidating the biological function of phosphorylation sites and kinase‐substrate networks. With continuing advancements in instrumentation and database search tools, phosphoproteomics is expected to achieve significant improvements in sensitivity, throughput, reproducibility, and accuracy, enabling its broader application in both biomedical research and clinical settings.

After large‐scale exploration of phosphorylation sites, their functional and clinical relevance has to be actively pursued. Despite impressive coverage reported in many recent studies, only a small subset of phosphorylation sites has been associated with biological functions. An important goal of quantitative phosphoproteomics is to identify those phosphorylation events that have biological or clinical significance. Therefore, validation experiments following discovery studies, including kinase activation or inhibition, cell cycle states, or drug treatment effects over time, may help in understanding the functional roles of protein phosphorylation. In the context of clinical translation, proteomics, including phosphoproteomics has yet to be fully implemented in clinical assays. In this review, we have highlighted the successes of quantitative phosphoproteomics in biomedical research. A few representative examples of identification of phosphosites responding to drug treatments and associated with drug resistance include elevated phosphorylated tau at residues T181, T217, and T231 in cerebrospinal fluid and blood of Alzheimer's disease [152, 153], as well as phosphoprotein biomarkers for neuroendocrine cancer [154], human esophageal cancer [155], and lung cancer treatment [125]. To promote phosphoproteomics into clinical utility, significant efforts are required in key areas, including enhancing sensitivity, assay robustness, and high‐throughput capabilities; standardizing protocols, analytical pipelines, and data reporting; implementing stringent clinical validation and quality control systems; and ensuring compliance with regulatory guidelines. Additionally, raising awareness within the community and establishing infrastructure through collaboration between academia, industry, and healthcare sectors are essential steps to drive the implementation of phosphoproteomics in clinical settings. By addressing these challenges, phosphoproteomics can advance to become a valuable tool to contribute to precision medicine and improved patient care.

Considering recent developments highlighted in this review, it is conceivable that microscale and nanoscale phosphoproteomics will become more accessible in the next 5–10 years. Following the trajectory of SCP, advancements in microfluidic platforms and miniaturized tools are expected to streamline complex workflows to enable the profiling of PTMs in small‐cell populations and even at the single‐cell level. Concurrently, advancements in MS instrumentation, and data processing tools, particularly those integrating machine learning and deep learning techniques, are expected to improve sensitivity, site localization, and quantification precision.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Data sharing are not applicable to this article as no datasets were generated or analyzed during the current study.