Proteomic insights into the extracellular matrix: a focus on proteoforms and their implications in health and disease

Abstract

Introduction:

The extracellular matrix (ECM) is a highly organized and dynamic network of proteins and glycosaminoglycans that provides critical structural, mechanical, and biochemical support to cells. The functions of the ECM are directly influenced by the conformation of the proteins that compose it. ECM proteoforms, which can result from genetic, transcriptional, and/or post-translational modifications, adopt different conformations and, consequently, confer different structural properties and functionalities to the ECM in both physiological and pathological contexts.

Areas covered:

In this review, we discuss how bottom-up proteomics has been applied to identify, map, and quantify post-translational modifications (e.g., additions of chemical groups, proteolytic cleavage, or cross-links) and ECM proteoforms arising from alternative splicing or genetic variants. We further illustrate how proteoform-level information can be leveraged to gain novel insights into ECM protein structure and ECM functions in health and disease.

Expert opinion:

In the Expert opinion section, we discuss remaining challenges and opportunities with an emphasis on the importance of devising experimental and computational methods tailored to account for the unique biochemical properties of ECM proteins with the goal of increasing sequence coverage and, hence, accurate ECM proteoform identification.

1. Introduction

The extracellular matrix (ECM) is a highly organized and dynamic network composed of proteins and glycosaminoglycans, which provides essential biophysical and mechanical support to cells [1,2]. The ensemble of ~1,000 genes in mammals encoding the constituting components of the ECM has been termed the matrisome [3]. It includes genes encoding core matrisome proteins like collagens, which provide tensile strength; elastin, which imparts elasticity; fibronectin and fibrillin, which contribute to the structural integrity of the ECM and the organization of other ECM components [1,2]; and proteoglycans, with their glycosaminoglycan side chains, that play crucial roles in hydration of the ECM gel, growth factor binding and the regulation of ECM interactions [2]. In addition to these structural components, matrisome-associated proteins, including ECM-affiliated proteins, ECM-modulating enzymes, and secreted factors, exert modulatory roles on the composition, architecture, and functions of the ECM [1,4]. These functions include governing cell adhesion, guiding cell migration, and modulating proliferation, survival, and differentiation [4,5]. Dysregulation in ECM composition or assembly can result in impaired cell and tissue function, leading to clinical conditions ranging from structural defects and cardiovascular and skeletal diseases to fibrosis and cancer [6–9].

Importantly, ECM proteins undergo extensive post-translational modifications (PTMs) both intracellularly during their synthesis and secretion, as well as extracellularly [6,10–12]. These PTMs, alongside alternative splice variants and single nucleotide variants (SNVs) including those leading to single amino-acid variants (SAAVs), generate various molecular forms of a single ECM protein termed proteoforms [13–16]. Different ECM proteoforms, characterized by different conformations and the ability to form different supra-molecular assemblies, thus exert different mechanical and biochemical signals and hence impart the ECM with different functionalities.

To introduce our readers to the importance of PTMs and proteoforms on ECM protein structure and functions, we will take the example of collagen V and its implication in Ehlers-Danlos syndromes (EDS) (Figure 1). Collagen V is a minor fibrillar collagen of the interstitial ECM found in most connective tissues and plays a central role in the fibrillation, or supramolecular assembly, of major fibrillar collagens, collagens I and III [17,18]. Three genes in the human genome, COL5A1, COL5A2, COL5A3, encode three distinct α chains that assemble in the endoplasmic reticulum to form triple helical collagen V monomers, a collagen V trimer for example can be composed of two α1 chains and one α2 chain. Through the biosynthetic and secretory pathway, and after being secreted in the extracellular space, collagen V undergoes extensive PTMs such as hydroxylation, glycosylation, proteolytic cleavage and cross-linking. These modifications are critical for the proper assembly of collagen V fibers and its interactions with other ECM proteins in the ECM, and hence for the integrity of ECM architecture (Figure 1, left panel; [17]). Variants of collagen V or of its modifying enzymes are linked to multiple clinical manifestations of EDS (Figure 1, right panel) characterized by connective tissue disorders [19]. Specifically, SNVs and their resulting SAAVs, along with exon skipping in the COL5A2 gene, produce alternative collagen V isoforms. These isoforms can lead to structural defects in collagen assembly, characteristic of classical EDS. In addition, aberrant PTMs due to disease-causing variants in enzymes involved in the collagen V biosynthetic, secretory, and assembly pathways all lead to different forms of EDS (Figure 1, right panel; Table 1). Of note, most core matrisome components undergo similar processing. There is thus a critical need to develop methods to study the structure/function relationship of ECM PTMs and proteoforms to understand their contribution to diseases.

Figure 1. Schematic representation of the diversity of collagen V proteoforms arising through the biosynthetic pathways in a healthy tissue or in Ehlers-Danlos syndrome (EDS).

Left panel: Biosynthetic pathway leading to collagen V incorporation into the ECM. The biosynthetic pathway consists of four steps: (a) The transcription and processing of the COL5A1 and COL5A2 mRNA, (b) the translation of COL5A1 and COL5A2 into the α1 and α2 chains of type V collagen, respectively, in the endoplasmic reticulum (ER), subsequent post-translational modifications by prolyl-3- and prolyl-4-hydroxylases, lysyl hydroxylases, and galactosyltransferases, and trimerization to form triple-helical procollagen V, (c) the trafficking of procollagen V through the secretory pathway and secretion into the extracellular space, (d) the proteolytic cleavage of N- and C-terminal peptides to mature collagen V fibrils that are further cross-linked by lysyl oxidase (LOX) to create a complex ECM network.

Right panel: Structurally abnormal collagen V in EDS can result from disease-causing COL5A2 gene variants, alternatively spliced variants (red), or disease-causing variants of enzymes involved in the collagen V biosynthetic pathway (red boxes). For example, disease-causing variants of the hydroxylation enzymes prolyl-3-hydroxylase (P3H3) and lysyl hydroxylase 1 (PLOD1) disrupt procollagen trimerization, contributing to kyphoscoliotic EDS; variants of ADAMTS2 impair its ability to process procollagen N-propeptide, resulting in dermatosparactic EDS; abnormal lysyl hydroxylation disrupts cross-linking, contributing to EDS type V. Altogether, these alterations result in improper collagen V structure, leading to disrupted ECM assembly and, hence, function.

Created using BioRender.

Table 1.

Clinical conditions associated with dysregulation or disease-causing variants of enzymes involved in post-translational modifications of ECM proteins

PTM	Enzyme	Gene	Phenotype/Disorder	OMIM ID
Hydroxylation	Prolyl 4-Hydroxylase	P4HA1	Joint hypermobility, contractures, axial and appendicular hypotonia with congenital weakness, Mild skeletal dysplasia without bone fragility, and high myopia. Congenital disorder of connective tissue	176710
		P4HA2	Myopia 25, autosomal dominant	600608
		P4HA3	Implicated in atherosclerotic carotid artery lesions	608987
	Prolyl 3-Hydroxylase	P3H1	Osteogenesis imperfecta, type VIII with bone fragility, low bone mass, and susceptibility to fractures	610339
		P3H2	Myopia, high, with cataract and vitreoretinal degeneration	610341
		P3H3	Phenotype similar to Ehlers-Danlos syndrome, kyphoscoliotic type 1	610342
	Lysyl hydroxylases	PLOD1	Ehlers-Danlos syndrome, kyphoscoliotic type 1	153454
		PLOD2	Bruck syndrome 2	601865
		PLOD3	Bone fragility, arterial rupture, deafness, and connective tissue disorder	603066
Cross-linking	Lysyl oxidases	LOX	Aortic aneurysm, familial thoracic 10	153455
		LOXL1	Exfoliation syndrome	153456
		LOXL2	Werner syndrome
		LOXL3	Myopia 28, Autosomal recessive	607163
	Transglutaminases	TGM1	Autosomal recessive congenital ichthyosis	190195
		TGM2	Proliferative vitreoretinopathy, Huntington disease	190196
		TGM3	Uncombable hair syndrome 2	600238
		TGM5	Peeling skin syndrome 2	603805
		TGM6	Spinocerebellar ataxia 35	613900
	Peroxidasin	PXDN	Anterior segment dysgenesis 7, with sclerocornea, i.e., corneal opacification, congenital cataract, and microcornea	605158
Phosphorylation	Extracellular Serine/Threonine Protein Kinase	FAM20A	Amelogenesis imperfecta	611062
		FAM20B	Phenotype resembling Desbuquois dysplasia	611063
		FAM20C	Raine syndrome	611063
	Vertebrate Lonesome Kinase	VLK	Rhizomatic limb shortening with dysmorphic features, Skeletal dysplasia	614150
Citrullination	Peptidyl arginine deiminases	PADI1	Susceptible to rheumatoid arthritis	607934
		PADI2	Susceptible to rheumatoid arthritis	607935
		PADI3	Uncombable hair syndrome; Susceptible to rheumatoid arthritis	606755
		PADI4	Susceptible to rheumatoid arthritis	605347
		PADI6	Oocyte/zygote/embryo maturation arrest basal cell carcinoma	610363
Proteolytic fragments	A disintegrin and metalloproteinase with thrombospondin motifs	ADAMTS2	Ehlers-Danlos syndrome, dermatosparactic type	604539
		ADAMTS3	Hennekam lymphangiectasia-lymphedema syndrome-3	605011
		ADAMTS5	Osteoarthritis	605007
		ADAMTS10	Weill-Marchesani syndrome	608990
		ADAMTS13	Thrombotic thrombocytopenic purpura	604134
		ADAMTS14	Usher syndrome-1D	607506
		ADAMTS15	Arthrogryposis, distal, type 12	607509
		ADAMTS16	Associated with human resting systolic blood pressure and diastolic blood pressure.	607510
		ADAMTS17	Weill-Marchesani 4 syndrome, recessive	607511
		ADAMTS18	Microcornea, myopic chorioretinal atrophy, and telecanthus	607512
		ADAMTS19	Cardiac valvular dysplasia 2	607513
	Matrix  metalloproteinases	MMP1	Chronic obstructive pulmonary disease Epidermolysis bullosa dystrophica Preterm Premature Rupture of Membranes	120353
		MMP2	Multicentric osteolysis, nodulosis, and arthropathy	120360
		MMP3	Susceptibility to coronary heart disease	185250
		MMP7	Preterm Premature Rupture of Membranes	120355
		MMP9	Metaphyseal anadysplasia 2	120361
		MMP12	Associated with arthritis, and atherosclerosis, development of chronic obstructive lung disease	601046
		MMP13	Spondyloepimetaphyseal dysplasia, Missouri type Metaphyseal anadysplasia 1 Metaphyseal dysplasia, Spahr type	600108
		MMP14	Winchester syndrome	600754
		MMP19	Cavitary optic disc anomalies	601807
		MMP20	Amelogenesis imperfecta, type IIA2	604629
		MMP21	Heterotaxy, visceral, 7, autosomal	608416
	Cathepsins	CTSA	Galactosialidosis	613111
		CTSB	Keratolytic winter erythema	16810
		CTSC	Haim-Munk syndrome Papillon-Lefevre syndrome Periodontitis 1, juvenile	602365
		CTSD	Ceroid lipofuscinosis, neuronal, 10	116840
		CTSF	Ceroid lipofuscinosis, neuronal, 13 (Kufs type)	603539
		CTSH	Associated with myopia, facial dysmorphism, and hearing loss	116820
		CTSK	Pycnodysostosis	601105
		CTSV	Keratoconus	603308

The large size, extensive PTMs, and insolubility of ECM proteins have posed significant challenges to our ability to analyze these proteins using mass spectrometry. Additionally, the wide dynamic range of ECM protein abundance has, so far, limited the depth of analyses [20]. In recent years, advancements in sample preparation techniques, such as de-glycosylation and fractionation, combined with improved instrumentation, have enabled mass spectrometry-based proteomic analysis of the ECM. These developments have expanded our understanding of ECM biology by allowing the identification and quantification of over 150 distinct ECM proteins in any given human tissue, as well as the profiling of ECM PTMs both in terms of patterns and site-specific modifications under various physiological and pathological conditions [1,21–24]. Despite these successes, challenges remain in the detection and characterization of proteoforms and PTMs.

The aim of this review is to highlight the power of proteomics in identifying ECM proteoforms that arise from PTMs, genetic variations, and transcriptional changes. Using examples from the recent literature, we will begin by discussing how bottom-up proteomic analysis can be applied to detect various ECM PTMs (e.g., addition of chemical groups, proteolytic cleavage, cross-linking) and resulting proteoforms. We then examine how these proteoforms affect the micro- and macro-molecular structure of the ECM. Additionally, we provide a comprehensive overview of computational tools and resources to aid the identification of these modifications. In the “Expert opinion” section of this review, we discuss remaining challenges and opportunities for improving ECM proteoform identification and advancing our understanding of ECM biology.

2. Proteomic approaches to detect additive post-translational modifications of matrisome proteins

The addition or removal of functional groups from amino acid residues markedly alters their mass, charge, hydrophobicity, and both inter- and intra-molecular interactions, thereby affecting their structure and function. In this section, we introduce the roles of major PTMs known to date, to mark ECM proteins and discuss how bottom-up proteomics has been applied to detect these modifications. Of note, we will not discuss glycomics and glycoproteomics applied to the study of the ECM protein glycosylation and glycosaminoglycans decorating proteoglycans, but invite readers interested to refer to recent reviews [25–30].

2.1. Hydroxylation of lysines and prolines of collagens

2.1.1. Mechanisms, biological roles, and implications in disease

Hydroxylation is a crucial post-translational modification in collagens that occurs intracellularly and involves the addition of hydroxyl groups to proline or lysine residues within the collagen triple helix region (rich in Gly-Xaa-Yaa repeats) and their telopeptides. This process is mediated by either prolyl hydroxylases or lysyl hydroxylases in the presence of oxygen, 2-oxoglutarate, Fe²⁺, and ascorbate [31]. Proline residues within collagens undergo 4-hydroxylation (4-OH) by prolyl-4-hydroxylases (P4HA1-3) or 3-hydroxylation (3-OH) by prolyl-3-hydroxylases (P3H1-3). proline 3-OH preferentially occurs when the proline is at the X position of the Gly-Xaa-Yaa triplet, whereas 4-OH occurs on prolines at the Y position [6,31].

4-OH of proline stabilizes the collagen triple helix through interchain hydrogen bonding and conformational changes in the imino ring [6,31,32]. Knocking out of P4ha2 in mice results in no obvious phenotypic abnormalities but reduced 4-prolyl hydroxylation, decreased thermal stability of collagens and consequently, secretion and deposition in the ECM [33,34]. Variants of the P4HA1-3 genes lead to congenital connective disorders (Table 1) [35,36]. 3-OH is less understood but has been implicated in collagen supramolecular assembly [37], P3h1 knockout in mice results in altered collagen fiber assembly in collagen I rich tissues such as bone, tendon and skin [31]. Notably, variants of the P3H1-3 genes are associated with osteogenesis imperfecta, basement membrane defects, and EDS (Table 1) [19,35,36].

5-hydroxylation of lysine residues is mediated by lysyl hydroxylases LH1-3 encoded by procollagen-lysine,2-oxoglutarate 5-dioxygenase 1-3 (PLOD1-3), preferentially when lysine is at the Y position of the Gly-Xaa-Yaa triplet [6,31,38]. This hydroxylation step is essential for subsequent glycosylation and cross-linking of collagen chains, influencing their secretion, assembly, and the regulation of the biophysical properties of the ECM (Figure 1) [31,39]. Plod1 inactivation in the mice resulted in altered cross-linking and consequently increased the risk of aortic ruptures [31]. Variants of PLOD1-3 are associated with EDS, Bruck syndrome 2, and other disorders leading to bone fragility and other basement membrane defects (Table 1) [19,35,36]. Moreover, aberrant hydroxylation of lysine residues is observed in fibrosis [40]. Despite the association of altered hydroxylation with various disorders, pharmacological modulation of hydroxylation patterns remains challenging due to limited knowledge of enzyme-specific roles across tissues [3,12].

2.1.2. Proteomics unveils hydroxylation patterns of collagens and substrate specificity of hydroxylases

Hydroxylation results in a mass shift of 15.9949 Da, which can be observed in both proline and lysine residues (Table 2). Since hydroxylation is a mandatory PTM of functional collagens, the inclusion of this PTM as a variable modification during database search to analyze mass spectrometric output has been shown to significantly improve collagen peptide and protein identification [41,42]. For example, allowing for this PTM to analyze the composition of the ECM of omental metastases from high-grade serous ovarian tumors resulted in a four-fold increase in spectral counts and the detection of an additional 17 distinct collagen chains [41]. However, expanding the list of variable modifications increases the database search time and may also increase false-positive identifications.

Table 2.

Mass shifts and masses resulting from post-translational modifications commonly found in ECM proteins

Post-translational modification	Modification type	Amino acids involved	Mass shift (Da)
Additive PTMs * indicates the mass shift due to neutral loss of the functional group due to fragmentation.	Hydroxylation	Lys, Pro	+15.9949
	Phosphorylation	Ser, The, Tyr	+79.9663, −17.9989*
	Bromination (79Br)	Tyr	+77.9105
	Bromination (80Br)	Tyr	+79.9085
	Citrullination	Arg	+0.9840, −43.0058 *
	Acetylation	Lys	+42.0105
Cross-linking “ ” indicates the masses of the amino acid residues after the addition of the cross-linked functional groups	Allysine	Lys	−1.0316, “145.1564”
	Hydroxyallysine	Lys	−13.9792, “162.1004”
	Lysinonorleucine (LNL)	Lys	−17.0304, “276.3525”
	Hydroxylysinonorleucine (HLNL)	Lys	−1.0311, “292.3519”
	Dihydroxylysinonorleucine (DHLNL)	Lys	+14.9633, “308.3513”
	Histidinohydroxylysinonorleucine (HHL)	Lys	−5.0626, “443.4747”
	Pyridinoline (Pyr)	Lys	−10.1257, “429.4449”
	Desmosine (Des)	Lys	−58.1471, “527.6110”
	Histidinohydroxymerodesmosine (HHMD)	Lys	−22.0984, “572.6318”
	γ-glutamyl-ε-lysyl	Lys, Gln	−17.0027, “276.3095”
	Pentosidine (PEN)	Lys, Arg	+58.0576, “379.4341”
	Glucosepane	Lys, Arg	+108.0211, “429.4912”
Proteolytic fragment	Trideuteroacetate	N-termini	+45.0340

Over the past decade, proteomics has significantly advanced the understanding of hydroxylation patterns and site specificity, providing novel mechanistic insights into their roles in health and disease. The degree of hydroxylation, or the hydroxylation pattern, varies between tissues and changes during aging, with specific residues exhibiting tissue-specific hydroxylation [34,43,44]. For instance, lysine at position 1030 (Lys¹⁰³⁰) was found to be partially hydroxylated (65%) in the tendon, whereas the same residue was almost entirely hydroxylated (90%) in ligaments [44]. Proteomic analysis has also elucidated substrate specificity, revealing that P4HA1 and P4HA2 preferentially hydroxylate the proline in the Y position of Gly-Xaa-Yaa triplet when the X position is occupied by either an acidic or a positively charged amino acids, respectively [34,45]. Characterization of various sites of 3-prolyl hydroxylation using enriched collagens from different tissues and species uncovered a tissue-specific, conserved pattern of 3-OH among mammalian species. Notably, two such conserved sites within collagen V and XI were spaced by 231 residues, suggesting a structural role of 3-OH in interchain hydrogen bonding [37].

Building on the preferential site-specific hydroxylation within the Gly-Xaa-Yaa triplet, Basak and colleagues established an in-silico motif-based pipeline to globally detect hydroxylation and glycosylation of collagen chains [46–48]. This approach enabled the identification of several evolutionarily conserved hydroxylation sites within lysyl- and 3-prolyl-hydroxylated residues across humans, mice, and zebrafish [46,48]. The in-silico pipeline also facilitated the re-analysis of publicly available proteomic datasets, revealing dynamic changes in the site-specific hydroxylation of proline and lysine residues during zebrafish heart regeneration [48]. Furthermore, a similar re-analysis strategy uncovered elevated collagen hydroxylation and site-specific modifications in both proline and lysine residues in the ECM of stent-induced neointima [42]. This data suggests the role of site-specific hydroxylation patterns in tissue development and disease.

To further enhance the understanding of hydroxylation in collagen chains, a knowledgebase of hydroxylation and glycosylation sites in collagen chains, ColPTMScape (https://colptmscape.iitmandi.ac.in) and a comprehensive database of hydroxylated proteins beyond collagens, HypDB (https://www.HypDB.site), have been developed (Table 3) [49,50]. We envision that these databases will enable future mechanistic studies on the role of hydroxylation in homeostasis and disease.

Table 3.

Major databases containing PTM information

Type	Database	Links
General	UniProt	https://www.uniprot.org
	Protein Data Bank (PDB)	https://www.rcsb.org
	AlphaFold	https://alphafold.ebi.ac.uk
Post-translational modifications	PSP (PhosphoSitePlus)	https://www.phosphosite.org
	EPSD (Eukaryotic Phosphorylation Site Database)	https://epsd.biocuckoo.cn
	dbPTM (Database Post-translational modification)	https://awi.cuhk.edu.cn/dbPTM
	BioGRID (The Biological General Repository for Interaction Datasets)	https://thebiogrid.org
	qPTM	http://qptm.omicsbio.info
	PTMCode v2	http://ptmcode.embl.de
	HPRD	http://www.hprd.org
	PHOSIDA	http://www.phosida.com
	PTM-SD	http://www.dsimb.inserm.fr/dsimb_tools/PTM-SD
	Unimod	https://www.unimod.org
	PTMD 2.0	https://ptmd.biocuckoo.cn
Hydroxylation	ColPTMScape	https://colptmscape.iitmandi.ac.in
	HypDB	http://www.hypdb.site
Lysine modification	CPLM (Compendium of Protein Lysine Modifications)	http://cplm.biocuckoo.org
Phosphorylation	EPSD	http://epsd.biocuckoo.cn
	PhosphoNET	http://www.phosphonet.ca
	RegPhos	http://140.138.144.141/~RegPhos
	Phospho.ELM	http://phospho.elm.eu.org
	dbPSP	http://dbpsp.biocuckoo.cn/indExp.php
Protein termini and protease processing	TopFIND	https://topfind.clip.msl.ubc.ca
Design of targeted proteomic experiments	Assay portal of the Clinical Proteomic Tumor Analysis Consortium (CPTAC)	https://proteomics.cancer.gov/assay-portal
Curated collection of ECM proteomic studies	MatrisomeDB	https://matrisomdb.org

2.2. Phosphorylation of ECM proteins

2.2.1. Mechanisms, biological roles, and implications in disease

The phosphorylation of matrisome proteins on serine, threonine, and tyrosine residues has been recognized for a long time [6,51]. In bone, phosphorylation of osteopontin (OPN) enhances fracture toughness by facilitating cation cross-linking of OPN polymers and their binding to hydroxyapatite through negatively charged phosphorylated serine residues [52]. It is hypothesized that the phosphorylation of intrinsically disordered proteins (IDPs) within the small integrin-binding ligand N-glycosylated (SIBLING) proteins, including OPN, bone sialoprotein (BSP), and dentin sialophosphoprotein (DSPP), modulate structural and functional interactions, thereby influencing tissue mechanics [52]. Yet, the kinases responsible for these PTMs, the Family with Sequence Similarity 20 Member C (FAM20C) kinase and the Vertebrate Lonesome Kinase (VLK), have only recently been discovered [53,54].

FAM20C phosphorylates the Ser-Xaa-Glu/pSer motifs of secretory proteins, including casein in milk and SIBLING proteins in the bone [53]. Variants of FAM20C have been linked to Raine syndrome with osteosclerotic bone dysplasia (Table 1) [55]. Conditional knockout of Fam20c in mice resulted in hypophosphatemia rickets, a phenotype rescued by expression of the Fam20c transgene [56]. Moreover, tissue-specific Fam20C knockout in mice revealed altered tissue architecture and function in the heart and salivary glands [57,58].

VLK, or Protein Kinase Domain Containing Cytoplasmic (PKDCC), is a tyrosine kinase. VLK is found at high levels in dense granules in platelets and is released during degranulation alongside ATP, facilitating extracellular tyrosine phosphorylation [54]. ECM proteins such as MMPs, laminins, OPN, and collagen I are substrates of VLK [54]. Homozygous knockout of Pkdcc in mice resulted in neonatal lethality and skeletal deformities, including cleft palate, shortened limbs, and reduced bone mineralization [21]. Notably, variants of PKDCC have also been associated with rhizomelic deformities and skeletal dysplasia (Table 1) [59]. These findings highlight the importance of both FAM20C- and VLK-mediated phosphorylation of ECM proteins in homeostasis and disease.

2.2.2. Identifying ECM protein phosphorylation using proteomics

Phosphorylation of either serine, threonine, or tyrosine residues causes a net mass addition of approximately 79.9663 Da. However, serine and threonine phosphoester bonds are labile and can easily fragment within the collision cells, losing phosphoric acid (−98 Da) and causing a deduction of 17.9989 Da (Table 2) [60]. Proteomic analysis of conditioned medium from wild-type HEK293T and FAM20C knockdown cells identified nearly 100 secreted proteins as FAM20C substrates, including the matrisome proteins fibronectin, OPN, bone morphogenetic protein 4 (BMP4), and tenascin C [61]. Moreover, annotation of phosphorylated sites in proteins, based on their extracellular location, including the extracellular domain of transmembrane proteins, identified 770 phosphorylated sites in 66 proteins across different species, including the ECM proteins OPN, BSP, laminins, tenascins, and fibronectin [62]. LC-MS/MS analysis of the matrix metalloproteinase, MMP1 co-expressed with VLK not only validated the phosphorylation Tyr³⁶⁰ but also identified additional sites of phosphorylation on Ser⁵⁷ and Thr²⁷⁴ demonstrating the ability of VLK to phosphorylate serine and threonine alongside tyrosine [54]. Proteomics also validated five of the 29 predicted phosphorylated sites on MMP2 (Ser³², Ser1⁶⁰, Ser³⁶⁵, Tyr²⁷¹, Thr²⁵⁰) and these sites were found to be conserved across mammals. Dephosphorylation of MMP2 using alkaline phosphatase increased its catalytic activity, while phosphorylation of MMP2 led to decreased catalytic activity, which is thought to be due to its change in its confirmation following phosphorylation [63,64].

While much focus has been on ECM protein phosphorylation in bone homeostasis and diseases, phosphorylation has also been implicated in other tissues [57,58,65]. However, studies focused solely on ECM protein phosphorylation remain limited. Global phosphoproteome databases (Table 3), focused in vivo and in vitro studies, can illuminate phosphorylation patterns in ECM proteins, offering insights into their roles as potential diagnostic and prognostic markers

2.3. Peroxidasin-dependent bromination of ECM proteins

2.3.1. Mechanisms, biological roles, and implications in disease

The bromination of tyrosine residues occurs as a by-product of peroxidasin (PXDN) catalysis. Following PXDN-mediated sulfilimine cross-link formation, Br⁻ ions are released, forming hypobromous acid (HOBr) in presence of hydrogen peroxide. HOBr subsequently interacts with the aromatic ring of tyrosine, resulting in the substitution of bromine (Br) at the 3- and 5- positions of the tyrosine residue, yielding 3-bromotyrosine (3-Br-Tyr) or 3,5-dibromotyrosine [11,66,67]. The importance of PXDN in ECM structure is highlighted by its involvement in cross-link formation between non-collagenous domains (NC) of collagen IV [66]. It is also speculated that tyrosine bromination may influence the phosphorylation of tyrosine, which could further alter the ECM structure, function, and signal transduction.

Catalytic-null mutations in Pxdn in mice and Pxn in Drosophila exhibited reduced viability and tissue stiffness [68]. Mutations in Pxdn in mice lead to developmental defects in the anterior eye segment, resembling the defects observed in patients with PXDN variants characterized by anterior segment dysgenesis and corneal opacification (Table 1) [69,70]. Moreover, pharmacological inhibition of PXDN using phloroglucinol in adult mice and flies exhibited a similar phenotype of reduced tissue and basement membrane stiffness [66,68]. Elevated levels of tyrosine bromination in the urine and higher Br⁻ enrichment in the thickened glomerular basement membrane are observed in patients with diabetic nephropathy, suggesting a correlation between ECM turnover, stability, and tyrosine bromination [11,71].

2.3.2. Proteomics reveals extensive bromination of ECM proteins

The isotopes of Br, ⁷⁹Br, and ⁸¹Br yielding 3-Br-Tyr induce mass shifts of 77.9105 and 79.9085 Da, respectively (Table 2) [11,72,73]. Tyrosine bromination has been detected by mass spectrometry on Tyr¹⁴⁸⁵ and Tyr¹⁴⁹⁰ residues in the C-terminal non-collagenous (NC1) domain of collagen IV in murine and human glomerular basement membranes, respectively, and reduced bromination of Tyr¹⁴⁸⁵ was observed in the Pxdn^−/− mice [11]. Recent proteomic analyses have extended the detection of tyrosine bromination to other basement membrane proteins, including laminins, tubulointerstitial nephritis antigen-like 1 (TINGAL1), and nidogen-2 [72,74]. The detection of the tyrosine bromination in collagen IV and its interacting proteins suggests a potential bystander effect of PXDN-mediated bromination via HOBr following cross-linking.

Kinetic reaction analyses have suggested the potential for bromination of other amino acids, such as tryptophan, cysteine, histidine, and methionine [75]. Recent proteomic studies have indeed detected tryptophan bromination in a subset of human and mice lung tissues, as well as cell-derived basement membrane, underscoring the need for a comprehensive exploration of the scope and significance of this PTM in both ECM structure and function [74]

2.4. Citrullination of ECM proteins

2.4.1. Mechanisms, biological roles, and implications in disease

Citrullination, or deimination, is a reaction that converts arginine into citrulline. It is catalyzed by peptidyl arginine deiminases (PAD1-4 and PAD6 encoded by the PADI1-4 and PADI6) [76]. This process replaces a nitrogen atom in the positively charged guanidium group of arginine with an oxygen atom, resulting in a neutral peptidyl citrulline that can alter hydrogen bonding [76,77]. The presence of citrullinated ECM proteins and their role in inflammation is underscored by the detection of anti-citrullinated protein antibodies (ACPAs) directed against various citrullinated ECM proteins, including collagen II, fibrinogen, fibronectin, and tenascin C, in rheumatoid arthritis (RA) patients [22,23,78,79]. Even though PAD enzymes lack a signal peptide, it is postulated that the extracellular citrullination is mediated by the transport of these enzymes via extracellular vesicles or via its release after apoptotic or necrotic cell death [77,80]. PADs influence different physiological processes; knockout of Padi1 in mice leads to embryonic lethality, while Padi2 knockout mice display no apparent physical abnormalities but develop emphysema and higher lung compliance [81,82]. Elevated levels of PAD2 and citrullinated ECM proteins are detected in colorectal carcinomas with liver metastasis [80]. Furthermore, PAD2-mediated citrullination of ECM proteins like fibulin 5 (FBLN5) and latent transforming growth factor beta binding protein 4 (LTBP4) plays a vital role in the formation of elastic fibers within the ECM. Decreased citrullination of FBLN5 was observed in aging lungs and impaired elastogenesis was observed in both neonatal lung fibroblasts isolated from Padi2 knockout mice or following pharmacological inhibition of citrullination using BB-Cl-amidine. Interestingly, impaired elastogenesis could be rescued by exogenous addition of citrullinated FBLN5 [82]. Several MMPs, including MMP1, MMP3, MMP9, and MMP13, were found to undergo citrullination following PAD2 treatment.

Notably, hypercitrullination of MMP9 enhanced its activation by MMP3. The physiological occurrence of citrullinated MMP9 was validated in sputum samples of cystic fibrosis patients [76]. Citrullination also alters growth factor function and protein resistance to proteolysis [77,83]. Importantly, trypsin hydrolysis of arginine residues is reduced after citrullination, possibly due to the changes in its charge and confirmation, underscoring the role of citrullination in modulating ECM protein structure and function [84] but also presenting a potential limitation to mass spectrometric detection in tryptic digests.

2.4.2. Proteomic analyses of citrullinated ECM proteins in rheumatoid arthritis and cancer

Deimination of arginine causes a mass shift of 0.9840 Da, which is equivalent to the mass shift caused by spontaneous deamidation of asparagine and glutamine (Table 2). However, citrullination of peptides increases their hydrophobicity, resulting in an increased retention time, which can be used to differentiate citrullination from the deamidation of asparagine and glutamine [85]. Moreover, fragmentation within the collision cell could lead to the removal of isocyanic acid (HNCO), causing a neutral loss of 43.0058 Da, detectable by mass spectrometry (Table 2; [86]). Proteomic analysis has enabled the identification of novel citrullinated residues in ECM proteins like collagen II and tenascin C, as well as their associated antibodies in the sera of RA patients [22,23]. Proteomic analysis of the PAD2-treated collagen II revealed that PAD2 preferentially citrullinates arginine residues in the Arg-Gly-Xaa motif when X was either occupied by a hydrophobic or an acidic amino acid [22]. Analysis of PAD2-treated transforming growth factor-beta (TGF-β) latency-associated peptide identified citrullination of the arginine in the RGD peptide, which inhibited integrin binding and subsequent activation [77]. Proteomic analysis of PAD2 and PAD4 treated fibronectin identified 24 sites of citrullination with one such site, Arg¹⁴⁴⁰, in the synergy site, i.e., near the RGD integrin-binding motif [5]. Re-analysis of ECM proteomic datasets of different cancers detected citrullinated ECM proteins [85]. However, the functional consequences of citrullination on ECM protein structure and functions remain to be determined. In addition, this PTM is not systematically included as variable modification, suggesting that we may be underestimating the prevalence of this PTM. Utilizing in vivo model systems and proteomic approaches can provide new insights into the mechanistic roles of PADs and citrullinated ECM proteins in cancer, potentially improving therapeutic outcomes.

2.5. Implications of ECM protein acetylation in fibrosis and cancer

2.5.1. Acetylation modulates ECM protein assembly

Lysine Acetylation is the reversible addition of an acetyl group from acetyl coenzyme A, which can occur both enzymatically by lysine acetyltransferases as well as non-enzymatically [87,88]. The addition of an acetyl group to the positively charged lysine residues neutralizes its positive charge, altering hydrogen bonding and resulting in changes in protein conformation and protein-protein interactions [88]. Dysregulated expression of lysine acetyltransferases and their deacetylating enzymes, lysine deacetylases or histone deacetylases (HDACs), have been implicated in numerous disorders, including fibrosis and cancer. Consequently, HDAC inhibitors are being explored in both pre-clinical and clinical settings [89,90]. However, the investigation of ECM protein acetylation is currently in its early stages. N-acetylation of collagens using sulfosuccinimidyl acetate disrupts their self-assembly by neutralizing the positive charge of lysine residues [91]. Acetylation of Lys⁷⁹⁴ in the cytoplasmic domain of integrin β1 has been shown to regulate fibronectin assembly, likely due to the neutralization of the positively charged lysine residue, facilitating hydrogen bonding with the hydrophobic pocket of kindlin-2 and initiating an inside-out signaling [12]. These findings highlight the potential role of acetylation in ECM protein assembly, representing an exciting frontier in ECM research.

2.5.2. Proteomics reveals acetylation of ECM proteins

Protein acetylation of lysine residues results in a mass shift of 42.0105 Da. Additionally, neutralization of positively charged lysine residue during acetylation hinders trypsin digestion, serving as an indicator during peptide detection (Table 2) [92]. Given the low stoichiometry of lysine acetylation, peptide enrichment and fractionation before LC-MS/MS analysis aids in the effective detection of this PTM [92,93]. Analysis of global lysine acetylation patterns in three cell lines MV4-11 (acute myeloid leukemia), Jurkat (acute lymphocyte leukemia), and A549 (adenocarcinoma alveolar basal epithelial) cells identified lysine acetylation in 18 matrisome proteins including microfibrillar-associated proteins, galectin-1, and annexins [93]. Proteomic analysis of cervical cancer tissue detected tumor-specific acetylated matrisome proteins including fibronectin type III domain containing 1 (FNDC1), serpin B3, S100-A8, S100-A9, and S100-A16, while peptide corresponding to fibrinogen was specifically acetylated in healthy cervical tissues [94]. Global lysine acetylome analysis of hepatocellular carcinomas and adjacent normal tissues detected 296 acetylated peptides corresponding to 72 matrisome proteins, including collagens I, VI, XIV, fibrinogen, fibronectin, vitronectin, tenascin-X, fibrillin-1, and transforming growth factor-beta-induced protein (TGFBI) [95]. However, the specific roles of lysine acetylation in these ECM proteins remain to be elucidated.

An interplay between acetylation and phosphorylation is known to regulate protein function and may similarly influence ECM protein structure and function, thereby impacting tissue architecture and homeostasis [96]. Integrating proteomic methodologies with in vitro and in vivo model systems to profile acetylation and deacetylation of ECM proteins will thus help delineate the contributions of acetylation to ECM assembly and function and, hence, may offer new insights into disease mechanisms.

The examples provided above demonstrate how mass spectrometry has enhanced our understanding of the nature of ECM PTMs and the mechanisms governing their catalysis. It is worth noting that other reactions such as carbamylation [97–99] or, the more recently discovered glutathionylation [100], can modify ECM proteins (collagen I and fibronectin, respectively). However, the substrates of these PTMs, and their structural and functional roles remain largely understudied. We postulate that the development of advanced proteomic methodologies will contribute to map comprehensively these PTMs (and perhaps help identify new ones), as well as help decipher the mechanisms responsible for their catalysis.

3. Proteomic approaches to detect ECM protein cross-links

3.1. Mechanisms of enzymatic and non-enzymatic cross-linking and consequences on ECM structure

Cross-linking of ECM proteins represents a critical step in ECM assembly and remodeling. Initial divalent cross-links provide a structural foundation for other ECM protein assemblies, while mature trivalent cross-links regulate the biophysical properties of the ECM and thus tune the mechanical properties of tissues [8,101]. Enzymatic and non-enzymatic mechanisms generate different types of cross-links, such as pyridinoline, pyrrole, pentosidine, and glucosepane, each with distinct chemistries and functions discussed elsewhere and summarized here [8,102].

3.1.1. Enzymatic cross-linking

Various enzymes, including lysyl oxidases (LOX), transglutaminases (TG), Factor XIII (F13A1), and PXDN, can induce cross-links between ECM proteins. LOX and LOX-like (LOXL) proteins catalyze the oxidative deamination of the ε-amino groups of lysine and hydroxylysine residues, producing highly reactive aldehydes (allysine and hydroxyallysine). These aldehydes interact with other lysine residues to form divalent cross-links such as lysinonorleucine (LNL), hydroxylysinonorleucine (HLNL), and dihydroxylysinonorleucine (DHLNL), which are essential for stabilizing newly formed collagen and elastin molecules [8,103]. These divalent cross-links can further undergo spontaneous reactions to form mature trivalent and tetravalent cross-links, including histidinohydroxylysinonorleucine (HHL), pyridinoline (Pyr), desmosine (Des), and histidinohydroxymerodesmosine (HHMD), which are critical for the maintenance of tissue stability and protecting the ECM from proteolysis [8,39,101]. For example, LOXL2-mediated cross-linking of tropoelastin led to reduced susceptibility of this protein to trypsin. Interestingly, variants of LOXL2 are implicated in mid-dermal elastolysis (Table 1) [103]. TGs and F13A1 catalyze the formation of γ-glutamyl-ε-lysyl which mediates the γ-γ cross-links between the lysine and glutamine residues [104,105]. While TGs are involved in cross-linking of the ECM proteins and maintaining their stability, F13A1-mediated cross-linking is vital in fibrin clot formation and, thereby, hemostasis and wound healing [102,105].

3.1.2. Non-enzymatic cross-linking

Cross-linking can also occur non-enzymatically by formation of advanced glycation end products (AGE). AGE cross-links are formed when sugars react with an ε-amino-lysine, forming a Schiff’s base like glycosyl lysine. Glycated lysines then undergo spontaneous Amadori rearrangement and interact with the carbonyl group of lysine to form cross-links such as pentosidine (PEN), glucosepane, or non-cross-linked lysine abducts such as N^ε-(1-carboxymethyl)-L-Lysine (CML), and N^ε-(1-carboxyethyl)-L-Lysine (CEL) [8]. AGE cross-links are implicated in the initial assembly of collagen fibers, onto which other ECM proteins can be deposited and assembled [1,96]. Long-lived proteins like collagens and elastin are particularly susceptible to AGE-mediated cross-linking, which further increases with aging [8,108]. Although AGE cross-linking increases ECM stiffness, it also decreases viscoelasticity and mechanical load-bearing properties, contributing to increased fragility of bone and is also known to play a role in other connective tissue disorders [24,107]

In addition to maintaining tissue stability and regulating ECM mechanical properties, aberrant cross-links are implicated in disorders such as osteoporosis, chronic kidney disease, atherosclerosis, glaucoma, various cancers, and fibrosis [8,102]. In cancers, increased LOX-mediated cross-linking and ECM stiffness correlate with enhanced tumor invasiveness and metastatic potential [109]. Despite promising results from various LOX and LOXL inhibitors in experimental settings, no such inhibitors have been clinically approved yet. This reflects the limited understanding of the roles of different enzymatic and non-enzymatic cross-links in homeostasis and disease [109].

3.2. Detection of ECM protein cross-links using proteomics

Detection of cross-links within the ECM is achieved by identifying characteristic fragment ions generated by different cross-linking groups following either complete hydrolysis or by digestion of cross-linked proteins to peptides (Table 2) [101,104,110]. Utilizing known standards of cross-linking groups for calibrating the mass spectrometer to estimate their mass shifts and retention times aids in the quantification of respective cross-links [24,110]. These mass shifts can be manually incorporated into proteomic data analysis software for cross-link detection (Table 4) [111]. Additionally, enriching cross-linked peptides by removing non-cross-linked amino acids or fractionating samples can further enhance the detection of cross-links [101,110].

Table 4.

Commonly used tools and software packages to analyze mass spectrometry data

Type	Tools	Links
Bottoms-up proteomics	Skyline	https://skyline.ms
	FragPipe	https://fragpipe.nesvilab.org
	Protein Prospector	https://prospector.ucsf.edu/prospector/mshome.htm
	MaxQuant	https://www.maxquant.org
	Trans Proteomic Pipeline	http://www.tppms.org/
	Mascot	https://www.matrixscience.com
	Open MS	https://openms.de
	MS-PyCloud	https://github.com/huizhanglab-jhu/ms-pycloud
	Comet	https://uwpr.github.io/Comet
	X!Tandem	https://www.thegpm.org
Post-translational modifications	InsPect	https://bio.tools/inspect_ms
	MetaMorpheus	https://smith-chem-wisc.github.io/MetaMorpheus
	PTM Select	https://sites.google.com/site/fredsoftwares/products/ptm-select
	PTM-Shepherd	https://ptmshepherd.nesvilab.org
	DeepRescore2	https://github.com/bzhanglab/DeepRescore2
Degradomics	CLIPPER 2.0	https://github.com/UadKLab/CLIPPER-2.0
	ClipsMS	https://github.com/loolab2020/ClipsMS
Proteoform	Proteoform Suite	http://smith-chem-wisc.github.io/ProteoformSuite
	SEPepQuant	https://github.com/bzhanglab/SEPepQuant
Data independent analysis (DIA)	DIA-NN	https://github.com/vdemichev/DiaNN
	Spectronaut	https://biognosys.com/software/spectronaut
Top-down proteomics	MASH Native	https://labs.wisc.edu/gelab/MASH_Explorer/MASHNativeSoftware.php
	TopPIC	https://www.toppic.org
	PTM-TBA	https://github.com/wenronchen/PTM-TBA
Multi-dimensional protein identification	Manchester Peptide Location Fingerprinting (MPLF)	https://github.com/maxozo/PLF
	MudPIT (Multi-Dimensional Protein Identification Technology)	https://medicine.yale.edu/keck/proteomics/technologies/proteinidentification/mudpit
Data storage and organization	Panorama	https://panoramaweb.org
Database generation	CustomProDB	https://bioconductor.org/packages/release/bioc/html/customProDB.html

Proteomic analysis of cross-linked amino acids from collagen I following hydrolysis of bovine fetal and adult ligament and tendon tissues revealed high levels of immature divalent cross-links in fetal tissues. Moreover, adult ligaments and tendons exhibited varied compositions of mature cross-links, with HHMD and pyrrole cross-links predominating in tendons and pyridinoline cross-links being more abundant in ligaments, suggesting the distinct roles of different cross-links in tissue function [44]. TG-2 mediated incorporation of biotinyl-TVQQEL-OH and 5-(biotinamido) pentylamine at glutamine and lysine residues of the N and C terminals of OPN identified various reactive glutamine and lysine residues. Moreover, the most reactive glutamine residues were located in the N-terminal region, while most reactive lysine residues were found at the C-terminus of OPN. This distribution potentially plays a role in the polymerization of OPN [104].

Adapting chemical cross-linking and ECM proteomic methodologies, 61 intermolecular and 193 intramolecular cross-linked peptides, accounting for 125 cross-links, were identified in FXIIIA-mediated fibrin clots. Most cross-links were observed within the αC region of fibrinogen (Gln²⁴⁰, Gln²⁵⁶, Gln³⁴⁷, Gln⁵⁸², Lys⁴³⁷, Lys⁵⁵⁸, Lys⁵⁷⁵, and Lys⁶²⁰). A subset of these cross-links was also observed in clots prepared from human plasma and in the thrombi from mediastinal wounds following cardiac surgery [105]. Cross-link analysis of peptides generated from the decellularized mouse liver with cirrhosis revealed increased AGE-related cross-links (glucosepane and pentosidine) when compared to LOX- (pyridinoline) and TG- (γ-glutamyl-ε-lysyl) mediated cross-links. Additionally, treatment with rosmarinic acid reduced AGE cross-links and alleviated fibrosis. Notably, the level of AGE-cross-links correlated with the fiber thickness and the elastic modulus in different stages of fibrotic liver, indicating the role of AGE-mediated cross-linking in liver fibrosis [24]. These findings underscore the relevance of proteomic analyses in aiding the detection of various cross-links and their patterns in homeostatic and diseased tissues.

However, the analysis of naturally occurring cross-links remains challenging due to their structural complexities, low abundance, and limited understanding of their fragmentation patterns. Moreover, mature stable cross-links exhibit resistance to enzymatic digestion, complicating their identification. We envision that improved methodologies and further in-depth proteomic analyses of ECM protein cross-links will offer insights into the mechanistic roles of these PTMs and potential avenues for therapeutic interventions.

4. Proteomic approaches to detect protease substrates and their proteolytic fragments

4.1. Role of proteolysis in ECM remodeling and matrisome proteoforms diversity

Proteolysis, the cleavage of peptide bonds, is an irreversible PTM that significantly contributes to proteoform diversity [10,112]. A plethora of proteases mediate this process, including MMPs, a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS), and cathepsins. The proteolysis of ECM proteins can lead to their activation, inactivation, alter their conformation, or generate biologically active fragments called matricryptines or matrikines [113,114]. These modifications are critical for ECM remodeling during homeostasis and tissue repair. For instance, the proteolytic processing of the C-terminal region of collagen is essential for its folding and deposition within the ECM [113,115]. Aberrant expression of these proteases and excessive ECM degradation have been implicated in cancer, fibrosis, and inflammation (Table 1) [10,113]. The role of various proteases and proteolytic neo-peptides within the ECM has been extensively reviewed elsewhere [10,112,113].

Beyond the generation of neo-peptides, proteolysis of ECM proteins also plays a vital role in ECM remodeling. ADAMTS10 and ADAMTS6, for example, modulate fibrillin-1 assembly through cleavage of fibrillin-2, which is essential for normal skeletal development [116]. Adamts6 knockout mice exhibit high levels of fibrillin-2 and skeletal abnormalities, which are mitigated by reducing fibrillin-1, demonstrating a role for ADAMTS6 in modulating fibrillin structure during development [116]. Variants of ADAMTS10 or fibrillin-1 result in Weill-Marchesani syndrome (Table 1) [116]. Variants of procollagen I, V, and ADAMTS2 lead to distinct forms of EDS, arthrocalastic, classical, and dermatosparactic types, respectively (Figure 1) [115]. Despite all affecting the amino-terminal procollagen cleavage, differences in cleavage sites and resulting proteoforms may account for the diversity of clinical manifestations [115]. Proteases are appealing drug targets [117]. However, substrates for many proteases remain unknown, necessitating further studies to improve our current understanding of their role in homeostasis and disease to aid biomarker discovery and the development of specific therapeutics.

4.2. ECM degradomics

Degradomics is the study of proteolytic events and includes the identification of protease/substrate pairs and their resulting proteolytic fragments [118]. Various experimental strategies with or without enrichment have been applied to the study of the ECM degradome [10,112,119]. Non-enrichment strategies, such as fast profiling of protease specificity (FPPS), involve labeling neo-N termini generated by proteases with trideuteroacetate, detectable by mass spectrometry as a +45.0340 Da mass shift (Table 2) [120]. Protease cleavage generates semi-tryptic N-terminal and C-terminal peptides. These N-terminal peptides generated by the proteases are labeled and enriched prior to mass spectrometric analysis. Various enrichment strategies like terminal amine isotopic labeling of substrates (TAILS), combined fractional diagonal chromatography (COFRADIC), and subtiligase-based enrichment have been used [112,119]. TAILS is one of the most extensively used enrichment strategies for degradome analysis of matrisome proteins [112,117]. It involves negative enrichment of N-terminal fragments by blocking the N-termini and lysine side chains through isotopic or demethylation labeling, followed by trypsin digestion. The N-terminal peptides generated by trypsin are excluded through reductive amination-mediated binding to a polyaldehyde polymer, which can be separated by ultrafiltration based on their molecular mass [112,115,117]. Blocking lysine side chains ensures that trypsin preferentially cleaves at the C-terminal of arginine, aiding in the validation of protease-generated fragments. The mass shifts due to the labeled neo-N-termini alongside the signature cleavage pattern of C-arginine can be used for the detection of the protease-generated peptides [112,119]. Further bioinformatic tools like CLIPPER (https://github.com/UadKLab/CLIPPER-2.0) or Trans Proteomic Pipeline (http://www.tppms.org/) can be used to analyze and annotate neo-N-terminal peptides (Table 4) [121,122]. Incorporating an inactivated test protease as a control further validates neo-N-terminal identification [117].

Isobaric tag for relative and absolute quantification (iTRAQ)-TAILS analysis of the secretome from MMP2-deficient and inactivated MMP9 mouse embryonic fibroblasts digested in vitro with MMP2 and MMP9 respectively identified 3152 cleavage sites in 1054 proteins, including matrisome proteins like galectin-1 and thrombospondin-2. Moreover, novel cleavage sites within both galectin-1 and thrombospondin-2 were detected. However, the role of the generated proteolytic fragment is yet to be investigated [117]. TAILS analysis of proteolytic fragments of the C-terminal half of fibrillin-2 generated by ADAMTS6 and a catalytically inactive ADAMTS6 revealed a cleavage site at Gly²¹⁵⁸-His²¹⁵⁹, which is hypothesized to hinder its multimerization and consequently its, microfibrillar assembly [123]. TAILS analysis of cell-derived ECMs treated with ADAMTS16 identified fibronectin as its substrate. Furthermore, a 30-kDa fibronectin fragment containing a heparin-binding domain was detected and was shown to further increase the expression of MMP3, initiating a feed-forward loop. This loop resulted in decreased fibronectin fibrillogenesis and ECM protein assembly [123]. TAILS identified previously unknown cleavage sites in collagen I and collagen V NC1 domains mediated by ADAMTS2 and ADAMTS14 in mouse skin. Novel cleavage sites in the triple helical domains of NC2 of COL5A1 and COL5A2 were also detected, which were similar to those in collagen I, collagen II, and collagen III [115]. These findings suggest that the clinical manifestations observed in different types of EDS (Refer 3.1), which share the pre-processing pathway of collagen I, may arise from altered procollagen V cleavage. These examples underscore the potential of proteomics and degradomics to identify proteolytic events and decipher their consequences on ECM structure and functions. However, most degradome studies have focused on secreted proteins and cell lysates, with limited studies on native insoluble ECMs.

5. ECM proteoforms arising from alternative splicing and genetic variants

Beyond post-translational modifications, genomic and transcriptional events further expand the diversity of the proteoform repertoire. These proteoforms include protein isoforms arising from alternative splicing or usage of alternative promoters and often disease-causing missense SAAVs resulting from SNVs, which differ from their canonical counterparts in both structure and function. (Figure 1) [124]. This section will illustrate the importance of these proteoforms in ECM homeostasis and will describe how proteomics can aid in the detection of these proteoforms and advance our understanding of their structural and functional consequences.

5.1. Roles of ECM isoforms and amino-acid variants in health and disease – the example of fibronectin

A prime example highlighting the importance of isoforms in ECM biology is that of fibronectin. Fibronectin exists in multiple isoforms, but two isoforms arising from alternative splicing have been the focus of investigations. Indeed, EIIIA (or ED-A), characterized by the insertion of an additional type III fibronectin domain between the 11^th and 12^th type III fibronectin domains, and EIIIB (or ED-B), characterized by the insertion of an additional type III fibronectin domain between the 7^th and 8^th type III fibronectin domains, are highly expressed during development but are only present at low levels in adult tissues [125]. Of clinical relevance, the expression of these two isoforms is upregulated during wound healing and in the tumor vasculature. Notably, cells from ED-B null mice showed defective fibronectin assembly [125]. Moreover, these isoforms are also involved in regulating the composition of the arterial wall, where a decreased immobilization of fibrillin-1 in the arterial ECM under low flow was observed in mice deficient in the ED-A and ED-B containing fibronectin. Interestingly, these mice also present a high risk of intimal rupture, indicating the role of these isoforms in regulating ECM architecture [126].

Variants of the fibronectin gene are associated with spondylometaphyseal dysplasia with corner fractures. These variants predominantly occur in regions critical for the proper 3D conformation, folding, and assembly of fibronectin, with cysteine residues in the N-terminal assembly domain being most frequently mutated. Modeling of the c.718T>G variant, which results in a p.Tyr24Asp substitution, predicts destabilization of the amino-terminal fibronectin type I domains 1 to 5 and improper protein folding. Consequently, variants of fibronectin accumulate intracellularly in an in vitro model system, potentially disrupting the secretion and assembly of other extracellular matrix proteins. [127]. The biological and clinical implications of the diverse array of matrisome isoforms arising from alternative splicing have been extensively reviewed elsewhere [124]. The modulation of the structural integrity and assembly of the ECM by fibronectin isoforms highlights their importance; however, there is still limited understanding of the structural and signaling consequences of these isoforms in homeostasis and disease. To address this, efforts should focus on the comprehensive identification of these isoforms at the protein.

Of note, what is illustrated here with fibronectin or with collagen V in Figure 1 extends to all components of the matrisome and to a broad panel of diseases [6,35,36]. For example, the analysis of the mutational landscape of 14 different cancers from The Cancer Genome Atlas (TCGA) program revealed a higher mutation rate in matrisome genes across all matrisome categories as compared to the rest of the genome. These mutations frequently occur in protein domains or at sites of splicing events, likely affecting protein structure and function. Interestingly, mutations in some of the matrisome genes were also predictive of overall survival [7].

5.2. Current landscape and challenges in detecting ECM protein isoforms using proteomics

Traditional protein identification techniques, such as western blotting, often fall short in distinguishing isoforms that differ by subtle mass increments or in that they would require isoform-specific antibodies. In global proteomic studies, ECM proteoform or isoform identification is dependent on the detection of isoform-specific peptides with high confidence [20]. If different proteoforms are present in different abundance, global proteomics will favor the identification of the most abundant form. In addition, overall low sequence coverage significantly hinders our ability to systematically identify the proteoform landscape of tissues (Figure 2).

Figure 2. Detection of ECM proteoforms by bottom-up mass spectrometry.

Top panel: Extracellular matrix (ECM) proteoforms arise from genetic (e.g., single amino acid polymorphisms; red star), post-transcriptional (e.g., alternative splicing events; see isoform-specific peptides in turquoise), and/or post-translational events (e.g., additive PTMs, cross-links, proteolytic cleavage). These molecular events contribute to a diverse array of proteoforms.

Bottom panel: In bottom-up proteomics, proteoforms are first denatured and then digested into peptides (dark grey boxes). These peptides are then separated by liquid chromatography and identified using tandem mass spectrometry (LC-MS/MS). Mass spectra are used for database search, peptide identification, and sequence matching. Green boxes denote peptides detected by LC-MS/MS analysis, while dotted lines indicate unseen protein regions. Some tryptic peptides will remain undetected due to several factors, including low abundance, hydrophobicity, filtering out peptides of suboptimal length, omitting PTMs during database search, or insufficient confidence in peptide assignment. Incomplete sequence coverage results in the underestimation of ECM proteoform diversity.

Created using BioRender.

Selective reaction monitoring and parallel reaction monitoring involving selecting a target (e.g., isoform-specific) precursor ion based on the retention time and mass-to-charge ratio of a peptide of interest for fragmentation, enables the sensitive detection of product ions and can thus become instrumental in aiding protein isoform detection [128–130]. A recent study utilizing parallel reaction monitoring to characterize different isoforms of periostin, revealing differential expression patterns of periostin isoforms 9 and 10 in atopic dermatitis lesions as compared to healthy samples, thus offering an opportunity for biomarker discovery [131]. Targeted proteomics holds tremendous promise to become a method of choice to enhance isoform and SAAV identification.

Proteogenomics, namely, the integration of genomics, transcriptomics, and proteomics, represents another promising approach to enhance proteoform detection. This approach involves predicting theoretical protein spectra from genomics or transcriptomics data to create custom databases, using tools such as customProDB [132] (Table 3) and subsequently comparing these with experimental spectra generated by proteomic analysis of the same samples [133,134]. This approach has been broadly applied to efforts led by the Clinical Proteomic Tumor Analysis Consortium (https://proteomics.cancer.gov/programs/cptac) but has not yet been specifically applied to enhance matrisome proteoform discovery.

6. Conclusion

One of the primary challenges in proteomics is to achieve sufficient protein coverage to identify proteins and, more importantly, proteoforms with confidence. For ECM proteins, this challenge is further pronounced due to their insolubility and extensive PTMs, some of which were, until recently, unknown. Enhanced bottom-up proteomic workflow, including ECM enrichment and peptide fractionation, together with increased knowledge of the modifications affecting ECM proteins and their inclusion as variable modifications during database search, has resulted in improved sequence coverage and has permitted the identification of ECM proteoforms, mapping of PTMs, and protease/substrate pairs and cleavage sites. Yet, interrogation of MatrisomeDB (https://matrisomedb.org), a database of curated ECM proteomic studies reprocessed through a common analytical workflow [135], revealed that, while the most abundant ECM proteins collagen I and fibronectin are relatively well-covered (~70% sequence coverage in any given ECM samples), other ECM proteins such as laminins and thrombospondin exhibit limited sequence coverage (10-35% on average across datasets included in MatrisomeDB). This limited coverage, combined with the dynamic range of protein abundances within the ECM, complicates the reliable identification and mapping of PTMs residues, isoforms, and SAAVs. Overcoming these challenges will further necessitate the development of experimental and computational methods tailored to account for the unique biochemical properties of ECM proteins. If successful, these methods will result in reliable proteoform detection, which, in turn, will significantly advance our understanding of the pathways regulating ECM protein conformation, assemblies, and functions.

7. Expert opinion: Opportunities in ECM proteoform identification

7.1. Enhancing peptide recovery to enhance proteoform identification

ECM proteomics is a field marked with significant advancements and enduring challenges. Bottom-up proteomics, a widely used approach due to is high sensitivity, established data deconvolution and analysis pipelines, remains the method of choice to study highly insoluble and often very large ECM proteins [20]. Yet, this methodology often relies on only a few peptides for protein inference, limiting the comprehensive and in-depth profiling of the matrisome of tissues and underscoring the need for complementary approaches to capture the full spectrum of PTMs and proteoforms in the ECM [14]. One avenue worthy of exploration is to adapt methodologies used to characterize the cellular proteome at the proteoform resolution, including the experimental enrichment of post-translationally modified peptides, like in phosphoproteomics [136]. However, the insolubility of ECM proteins and the lack of validated reagents to enrich ECM-specific PTMs, such as hydroxylation, constitute significant challenges.

Trypsin is the gold standard protease for peptide digestion. PTMs, including hydroxylation, glycosylation, citrullination, acetylation, cross-linking, but also protein-protein interactions, can hinder efficient peptide cleavage by standard proteases like trypsin. For instance, trypsin cannot efficiently cleave hydroxylated, glycosylated and crosslinked lysine or citrullinated arginine residues, resulting in large peptides that may be filtered out during proteomic data analysis. LysC is now broadly used together with trypsin in ECM proteomics [3,137]. In addition, digesting the α1 chain of collagen IV with trypsin and GluC significantly improved sequence coverage to 84.7% [46]. The combination of multiple proteases to improve global ECM protein sequence coverage is thus a promising approach. However, compatibility of two or more proteases that have different pH or temperature requirements in a single experiment may pose certain limitations [138].

7.2. A step toward structural matrisomics

Structural and conformational changes resulting from proteoform arising during aging or pathological conditions can alter protease susceptibility, adding another layer of complexity to proteomic analysis [139,140]. Peptide location fingerprinting, a peptide mapping strategy that estimates peptide yields, has been employed to identify structural changes in proteoforms under homeostasis versus pathological conditions [139,140]. This method has determined structural changes and protease susceptibility of different domains within fibrillin-1 and the α3 chain of collagen IV induced by photoaging through exposure to ultraviolet broadband and solar-simulated radiation [140]. Using a similar strategy and the Manchester Peptide Location Fingerprinting (MPLF) tool enabled the re-analysis of ECM datasets, identifying conserved structural differences in aging and atherosclerotic tissues [139].

Recently, we have adapted limited proteolysis [141–143] to improve sequence coverage of ECM proteins. Data generated from partial digestion of ECM proteins with trypsin at four different time points resulted in improved coverage of various ECM glycoproteins, collagens, and matrisome-associated proteins [144]. By overlaying the peptides identified onto experimental or predicted protein structures from the Protein DataBank (PDB, https://www.rcsb.org; [145]) or AlphaFold (https://alphafold.ebi.ac.uk/; [146]), we could also gain structural insights into protease accessibility, and thus, folding of ECM proteins [144].

7.3. Applying top-down proteomics to the study of ECM proteoforms?

Top-down proteomics (TDP) provides an alternative approach by analyzing intact proteins rather than peptides. This approach facilitates the detection of proteoforms arising from alternative splice variants, SNVs, SAAVs, and other PTMs. This approach has been effectively applied to detect combinatorial PTMs and proteoforms in various diseases, including cancer, diabetes, neurodegenerative, and cardiovascular disorders [147]. Despite its advantages, traditional TDP methods face challenges with proteins larger than 30 kDa, and the average size of a canonical core matrisome protein is 1045 amino acids (>100kDa). Moreover, issues such as solubilization, co-elution, separation, ionization, and data deconvolution further complicate the analysis of larger proteins, such as ECM proteins [148]. Recent advancements in TDP methodologies offer solutions to some of these challenges. Capillary zone electrophoresis (CZE) and polyacrylamide-gel-based prefractionation have been developed to enhance the separation and analysis of larger proteins [148]. Additionally, the Individual Ion Mass Spectrometry (I2MS) technique, which uses direct charge measurements of individual ions, enables the detection of proteins larger than 30 kDa [149,150]. The recent integration of size exclusion chromatography (SEC) with CZE-MS/MS has successfully detected 276 integral membrane proteoforms in mouse brain tissue [151]. The most critical limitation hindering the analysis of the ECM is its insoluble nature due to the high degree of post-translational modifications and cross-linking. The use of photocleavable surfactants like Azo has improved the extraction of membrane proteins, which are often difficult to solubilize in MS-compatible reagents [148]. Similar reagents and strategies can be employed to improve the solubility of the ECM proteins for its subsequent TDP analysis. Integrating detailed PTM information from TDP with the peptide ion coverage and localization data from bottom-up methods can aid in improving the confidence in detection and validation to achieve a more complete picture of ECM protein modifications. Bioinformatics tools like PTM-TBA (https://github.com/wenronchen/PTM-TBA) further enhance this integration by improving the sensitivity and coverage of PTM detection [147].

ECM proteoforms represent an untapped reservoir of potential disease biomarkers and therapeutic targets. The integration of bottom-up and top-down proteomics approaches and the development of more powerful experimental and analytical pipelines tailored to enhance ECM proteoform identification will undoubtedly result in significant advances in our understanding of the mechanisms guiding ECM assembly and functions in health and disease. These combined efforts will also accelerate biomarker discovery and the development of effective matritherapies [1], that is, interventions targeting the ECM to achieve a therapeutic benefit.

Article Highlights.

• ECM proteoforms generated by pre- and post-translational modifications adopt different conformations and thus modulate the macromolecular architecture of the ECM meshwork and its functions.

• Proteomic analysis of post-translational modifications can shed light on their consequences on ECM protein structure and signaling properties.

• Sample preparation methods devised to account for the unique biochemical properties of ECM proteins result in increased sequence coverage and ECM proteoform detection.

• Comprehensive databases containing information on PTMs, as well as tailored computational workflows, are essential resources for the accurate identification and characterization of ECM proteoforms.