Mechanisms of Assembly and Remodeling of the Extracellular Matrix

Abstract

The extracellular matrix (ECM) is the complex meshwork of proteins and glycans that forms the scaffold surrounding and supporting cells. It exerts key roles in all aspects of metazoan physiology, from conferring physical and mechanical properties to tissues and organs to modulating cellular processes such as proliferation, differentiation, and migration. Understanding the mechanisms orchestrating the assembly of the ECM scaffold is thus critical to understand ECM functions in health and diseases. This Review will first discuss novel insights into the compositional diversity of matrisome components and the mechanisms leading to tissue-specific assemblies and architectures tailored to support specific functions. The Review will then highlight recently discovered mechanisms modulating ECM secretion, assembly, and remodeling in homeostasis and human diseases, including a discussion of post-translational modifications and metabolic pathways such as amino acid availability and the circadian clock. Last, the Review will explore the potential of “matritherapies,” i.e., strategies to normalize ECM composition and architecture to achieve a therapeutic benefit.

Introduction

The extracellular matrix (ECM) is the complex meshwork of proteins and glycosaminoglycans that forms the scaffold surrounding and supporting cells^1,2. Phylogenetically, the build-up of the ECM has coincided with the emergence of multicellularity. The diversification of the repertoire of ECM components and the complexification of ECM protein structures have accompanied critical steps in metazoan evolution, like the appearance of chordates, vertebrates, and mammals^3–13.

The ECM exists in two structurally distinct forms, basement membranes, that are thin, sheet-like assemblies, and meshwork-like interstitial ECMs, that differ in molecular composition, architecture, localization, and functions. The core of the ECM scaffold is composed of structural proteins including collagens^14–16, fibronectin¹⁷, fibrillins, and elastin^18–23. The ECM scaffold contributes to the organization of cells into tissues, guides cell polarization, and confers mechanical properties to tissues and organs. In addition to its structural roles, the ECM both provides and relays mechanical and biochemical signals^24–28. These signals are transduced by cell surface receptors, such as the integrins^29–32 or the discoidin domain receptors (DDRs)³³, and modulate a plethora of cellular processes including adhesion^34,35, migration^36–38, proliferation³⁹, survival^40,41, and differentiation⁴². These processes are all critical for the proper development and homeostasis of multicellular organisms^43–45. Consequently, disruption of the ECM structure or its signaling properties — resulting from ECM gene variants or altered ECM composition — lead to a broad spectrum of clinical presentations, from impaired development and tissue repair to cardiovascular and musculoskeletal diseases to fibrosis and cancer^42,46–50.

The mechanisms involved in the assembly and maintenance of the ECM scaffold have been primarily studied in vitro and on a small subset of exemplary proteins like collagens^14–16,51, fibronectin^17,52, fibrillins^53,54, and elastin^18,21,55. The study of genetic disorders of the ECM, accelerated by the sequencing of the human genome, has greatly contributed to our understanding of the mechanisms of ECM assembly of a few dozen ECM components^56–60. However, we now know that the human “matrisome” comprises nearly 300 genes encoding structural components of the ECM and ~700 genes encoding proteins capable of modulating ECM protein structures and/or functions⁶¹. In addition, advances in proteomic technologies have revealed that the composition of the ECM is tissue-specific and that any given tissue is composed of well over 200 distinct matrisome components^62,63. Yet the consequences of this diversity on the architecture and functions of the ECM remain to be fully elucidated.

The aim of this Review is to present recent insights into the tightly regulated and highly dynamic multi-step process leading to the assembly of the ECM scaffold. With examples primarily taken from research on the mammalian ECM, this Review will discuss the compositional diversity of ECM building blocks and the mechanisms leading to tissue-specific assemblies resulting in different architectures tailored to support different functions. The Review will then highlight recently discovered mechanisms modulating ECM protein secretion, assembly and remodeling, including post-translational modifications, and the impact of metabolic pathways such as amino acid availability, endocytic recycling, and the circadian clock. Last, the Review will explore the potential of “matritherapies”, i.e., therapeutic approaches targeting the mechanisms of ECM assembly or cell–ECM interactions, to either maintain or restore ECM homeostasis and achieve therapeutic benefits.

Compositional Diversity Of The Ecm

The sequencing of the genomes of different organisms and advances in our understanding of structural features of proteins have provided opportunities to perform de novo sequence analysis to identify genes encoding proteins sharing common features. This approach has been used to predict many protein repertoires, including the ensemble of genes encoding protein kinases termed the kinome⁶⁴, the ensemble of G-protein-coupled receptors⁶⁵, or the adhesome, i.e., the complement of genes involved in cell adhesion⁶⁶.

Using characteristic features of known ECM proteins, namely, the presence of a signal peptide marking ECM proteins for secretion and the presence of protein domains commonly found in ECM proteins (e.g., type III fibronectin domain, triple helical collagen domain, EGF-like domain; Figure 1)⁶⁷, we predicted that about 1,000 genes in the human genome encode potential ECM components. We termed this ensemble the “matrisome”^1,61,68. This section will provide an overview of the different categories of matrisome components.

Figure 1 |. Diversity of the core matrisome components.

a. The human core matrisome comprises 277 genes, including 44 genes encoding collagens, 36 genes encoding proteoglycans, and 197 genes encoding ECM glycoproteins.

The domain-based organization predicted by the simple modular architecture research tool (SMART) and canonical members of each of these categories are depicted in the following panels. The length of the canonical form of each protein is indicated in amino acids (aa). All core matrisome components are characterized by the presence of an amino-terminal signal peptide, which is required for secretion into the extracellular space. Readers are invited to visit the SMART website for more details about each protein domain depicted in the figure.

b. The domain-based organization of exemplary collagen proteins (encoded by the COL1A1, COL6A1, and COL18A1 genes, respectively) is shown. Blue boxes indicate regions rich in triple helical Gly–X–Y repeats (where X and Y are commonly proline and hydroxyproline residues) characteristic of collagen proteins. Other protein domains commonly found in collagens include the fibrillar collagen c-terminal domain (COLFI), the von Willebrand factor (vWF) type A domain (VWA), the frizzled domain (FRI), or the laminin G domain (LamG). Cleavage products stemming from the carboxy-terminal non-collagenous region of certain collagens exert biological activities, such as in the case of endostatin, an anti-angiogenic protein derived from the proteolysis of the α1 chain of collagen XVIII.

c. Domain-based organization of extracellular proteoglycans representative of each of the three classes proposed by Iozzo and Schafer⁷⁹: the hyaluronan- and lectin-binding proteoglycans (hyalectans) and lecticans (represented by aggrecan, which is encoded by ACAN), the small leucine-rich proteoglycans (SLRPS; represented by decorin, which is encoded by DCN), and the SPARC/osteonectin, CWCV and Kazal-like domain (SPOCK) proteoglycans (represented by testican-1, encoded by SPOCK1). Protein domains commonly found in proteoglycans include the immunoglobulin domain (IG), Link (Hyaluronan-binding) domains (LINK), Epidermal growth factor-like domain (EGF), the C-type lectin or carbohydrate-recognition domain (CLECT), the complement control protein (CCP) domain (also known as short consensus repeat SCR or SUSHI repeat), different leucine-rich repeats (LRR-NT, LRR, and LRR-TYPE), the KAZAL domain, or the Thyroglobulin (TY) type I repeat. For more information on these domains, readers can refer to the SMART database mentioned above.

d. Domain-based organization of exemplary non-collagenous ECM glycoproteins such as fibronectin, the latent TGFβ binding protein 1 (LTBP1), thrombospondin 1 (TSP1), tenascin-C, and laminin α5. Certain matrisome genes, such as fibronectin or tenascin C, can be alternatively spliced, resulting in different isoforms containing different domains. Protein domains commonly found in ECM glycoproteins include fibronectin type I (FN1) domains that form two anti-parallel β-sheets, type II (FN2) domains that contain cysteines involved in disulphide bonds and protein-protein interactions, and type III (FN3) domains that form β-sandwich folds. Another set of domains commonly found in ECM glycoproteins are epidermal growth factor (EGF) domains, including EGF-like domains, calcium-binding EGF (EGF CA) domains, and laminin-type EGF-like (EGF Lam) domains. Other domains characteristic of ECM glycoproteins include von Willebrand factor type C (VWC) domains first identified in von Willebrand factor (vWF), thrombospondin N-terminal-like (TSPN) domains and thrombospondin type 1 (TSP1) domains were first identified in thrombospondins but also found in other ECM glycoproteins, and domains first identified in laminins but broadly present in other glycoproteins such as laminin N-terminal (LamNT) domains, globular laminin G (LamG) domains, and laminin B (LamB) domains. All these domains are involved in protein-protein interactions.

The core matrisome

The core matrisome comprises approximately 300 genes that encode proteins contributing to the structure of the ECM scaffold⁶¹. Based on biochemical features, these genes can be grouped into three categories: those encoding collagens, proteoglycans, and other ECM glycoproteins [Figure 1a].

Collagens

In the human genome, 44 genes encode distinct collagen chains that selectively assemble to form 28 different trimeric collagen proteins^14,51,69. Fibrillar collagens are one of the most, if not the most, abundant proteins in mammals. Their main structural function is to confer resistance to tensile force or stretching. Collagens are characterized by the presence of triple-helical motifs composed of glycine (Gly)-X-Y repeats where the amino acids X and Y are commonly proline (Pro) and hydroxyproline (Hyp) residues [Figure 1b, left panel]¹⁴. As a result, the amino acid composition of collagens is biased towards proline and glycine residues. For example, interrogation of the protein database UniProt reveals that the α1 chain of collagen I (encoded by the gene COL1A1) is composed of 1,464 amino acids, of which 391 (26.7%) are glycine residues and 278 (19%) are proline residues. In comparison, glycine residues and proline residues represent 7% and 5% of the amino acid composition of non-matrisome proteins, respectively.

Collagen chains, encoded by distinct collagen genes, form homo- or heterotrimers that further assemble into different higher-order structures allowing their classification into six groups: fibril-forming collagens (I, II, III, V, XI, XXIV and XXVII), fibril-associated collagens with interrupted triple helices (FACITs; IX, XII, XIV, XVI, XIX, XX, XXI and XXII), network-forming collagens (IV, VIII and X), transmembrane collagens (XIII, XVII, XXIII and XXV), multiplexins (XV and XVIII), and a group of 4 collagens characterized by the presence of multiple von Willebrand factor A domains (collagen VI forming microfibrils, collagen VII forming anchoring fibrils, and collagens XXVI and XXVIII)¹⁴ [representative members of these groups are presented Figure 1b and details of their supramolecular assemblies are discussed below]. Whereas certain collagens are only encoded by a single gene, resulting in homotrimeric proteins (e.g., the COL18A1 gene encodes the α1 chain of collagen XVIII, and the functional collagen XVIII is a homotrimer composed of three α1 chains, others result from the assembly of different chains encoded by distinct genes. For example, six collagen IV genes, COL4A1 and COL4A2 localized on chromosome 13, COL4A3 and COL4A4 localized on chromosome 2, and COL4A5 and COL4A6 localized on chromosome X, have emerged from gene duplication. These genes are expressed in a tissue-specific manner and encode six distinct αchains (1–6) that selectively assemble to form three functional heterotrimeric type IV collagens: a form composed of two α1 and one α2 chains [α1(IV)₂α2(IV)], a form composed of one α3, one α4, and one α5 chains [α3(IV)α4(IV)α5(IV)], and a form composed of two α5 and one α6 chains [α5(IV)₂α6] (ref. ⁷⁰). This exemplifies different mechanisms that contribute to the compositional diversity of the ECM.

Of note, although, arguably, the presence of a triple-helical motif is the defining feature of a collagen, not all proteins containing triple-helical motifs are classified as collagens. Interrogation of the human reference proteome in the UniProt database (UP000005640)⁷¹ revealed that 78 genes encode proteins containing a triple helical motif (InterPro IPR008160)⁷². Of the 34 non-collagen genes, some, such as emilin-1 or the EMI domain-containing protein 1 (EMID1), are bona fide ECM components, whereas others, such as ficolins, or collectins known as soluble defense collagens⁷³, have been classified as ECM-affiliated (see below). Yet other proteins containing triple helical motifs like the macrophage receptor MARCO or some members of the scavenger receptor class A are cell-surface receptors^1,14. Interestingly, despite several decades of active research, criteria to definitively classify a protein as a collagen remain to be defined⁷⁴.

Glycosaminoglycans and proteoglycans

Glycosaminoglycans (GAGs) are polysaccharides composed of anionic disaccharide repeats comprising one amino sugar and one uronic sugar molecule (or a galactose molecule in keratan sulfate). Hyaluronan is a GAG polymer exclusively composed of non-sulfated monosaccharides and forms a hydrated gel filling most of the extracellular space⁷⁵. In contrast, proteoglycans are composed of a core protein decorated by at least one covalently-bound sulfated GAG chain^76,77. The nature and position of the disaccharides, linkage, and sulfation levels define four classes of GAGs: chondroitin sulfate, dermatan sulfate, keratan sulfate, and heparin or heparan sulfate. The primary structural function of the hydrogel formed by GAGs and proteoglycans is to resist compressive forces⁷⁸.

Proteoglycans can be classified based on the size of their core proteins into small proteoglycans of tens of kilodaltons (e.g., biglycan, decorin) and large proteoglycans of several hundreds of kilodaltons (e.g., aggrecan, perlecan, versican) [Figure 1c]. A nomenclature of proteoglycans integrating information on cellular localization and structure has been proposed and categorizes proteoglycans into four major classes: a single intracellular proteoglycan, serglycin, located in secretory granules of mast cells, 13 cell surface proteoglycans, four pericellular proteoglycans, and 25 extracellular proteoglycans⁷⁹. Extracellular proteoglycans are further divided into three categories: hyaluronan- and lectin-binding proteoglycans (hyalectans) and lecticans, small leucine-rich proteoglycans, and the SPARC/osteonectin, CWCV and Kazal-like domains proteoglycans (SPOCK1–3) family of proteins encoding testicans (1–3)⁷⁹. All extracellular and pericellular proteoglycans are part of the matrisome nomenclature⁶¹ but some differences between classifications are of note. Collagens XV (COL15A1) and XVIII (COL18A1) are classified as pericellular proteoglycans but are listed as collagens in the matrisome nomenclature. In addition, agrin (AGRN), listed as a pericellular proteoglycan, and tsukushin (TSKU), listed as a non-canonical small leucin-rich proteoglycan, are classified as ECM glycoproteins under the matrisome nomenclature. The matrisome nomenclature lists an additional 11 genes encoding putative proteoglycans or proteins forming complexes with proteoglycans (e.g., members of the hyaluronan and proteoglycan link protein family)⁶¹. This highlights the difficulty of categorizing structurally and functionally multifaceted proteins and exemplifies how using different properties and parameters will result in somewhat different classifications.

Non-collagenous ECM glycoproteins

In addition to collagens and proteoglycans, nearly 200 other glycoproteins can be found in the ECM^1,61 [Figure 1d]. This number was initially predicted from sequence analysis⁶¹ and has now found experimental support from large-scale ECM-focused proteomic studies⁶³ (see below). In contrast to proteoglycans that include GAG chains, the glycan moiety covalently linked to a glycoprotein can be a mono-, di-, oligo-, or polysaccharide. Many of these glycoproteins, including fibronectin¹⁷, fibrillins²², latent TGFβ-binding proteins (LTBP1–4)⁵⁸, form fibrils that contribute to the architecture of the ECM scaffold. Of the non-fibrillar components of the ECM, elastin is an amorphous protein that forms, in combination with microfibril-associated proteins (MFAPs), elastic fibers^20,23. Laminins are characteristic of a specialized type of ECM called the basement membrane⁸⁰ (see below). 11 genes in the human genomes encode laminin chains (five αchains, fourβchains, and three γchains) that assemble selectively to form, as of today’s knowledge, 16 distinct functional α–β–γ trimers. These assemblies are tissue- and cell-specific⁸¹. Last, matricellular proteins are non-fibrillar glycoproteins that incorporate into the ECM scaffold and exert modulatory roles within the ECM^82–84. Examples of matricellular proteins include CCN proteins (CYR61, CTGF, NOV, and the Wnt-induced secreted proteins WISP1–3)⁸⁵, tenascins (TNC, TNR, TNN also referred to as TNW, TNX)⁸⁶, thrombospondins (1–5)⁸⁷ and fibulins (1–7)⁸⁸ in addition to individual proteins like periostin⁸⁹, or secreted protein acidic and rich in cysteine (SPARC)⁹⁰.

ECM-associated proteins

In addition to structural ECM proteins, a plethora of proteins associate with core matrisome components and modulate the structure and/or the mechanical and signaling properties of the ECM. This subset of ECM-associated components, also defined by de novo sequence analysis and the presence of canonical protein domains, is further subdivided into 171 genes encoding ECM-affiliated proteins, 238 genes encoding ECM regulators, and 344 genes encoding secreted factors [Figure 2a]⁶¹. Categorizing these components as part of the matrisome allowed the definition of a broad, yet comprehensive, framework for system-level studies of the ECM (see below).

Figure 2 |. Diversity of the matrisome-associated components.

a. 753 genes encode matrisome-associated proteins, including 171 genes encoding ECM-affiliated proteins, 238 genes encoding ECM regulators, and 344 genes encoding secreted factors.

The domain-based organization predicted by the simple modular architecture research tool (SMART) of canonical members of each of these categories is depicted in the following panels. The length of the canonical form of each protein is indicated in amino acids (aa). As with core matrisome components shown in Fig. 1, matrisome-associated proteins carry a signal peptide required for their extracellular secretion.

b. Domain-based organization of exemplary ECM-affiliated proteins, including galectins (encoded by LGALS genes), collectins (encoded by COLEC genes), and ficolins (encoded by FCN genes). Blue boxes indicate regions rich in triple helical Gly–X–Y (where X and Y are commonly proline and hydroxyproline) repeats characteristic of collagen proteins (see text).

c. Domain-based organization of exemplary secreted factors known to interact with core matrisome components (e.g., transforming growth factors β (TGFβ) or vascular endothelial growth factors (VEGFs)). Like core matrisome components, matrisome-associated components, such as VEGFA, can be alternatively spliced and generate isoforms of different lengths. In the case of VEGFA, these different isoforms resulting from alternative splicing (introns are indicated by vertical bars) generate proteins that bind with different affinities to the ECM.

d. ECM regulators can be broadly grouped into three groups: enzymes responsible for additive post-translational modifications (PTMs) (e.g., prolyl hydroxylases, lysyl hydroxylases, and kinases), cross-linking enzymes (e.g., lysyl oxidases (LOXs), transglutaminases (TGMs), peroxidasin) and their regulators, and proteinases (e.g., matrix metalloproteinases (MMPs), A Disintegrin and metalloproteinases (ADAMs), cathepsins) and their regulators. The schematic represents amino acid residues that are commonly post-translationally modified in ECM proteins and the enzymes that mediate these PTMs and whether they take place intra- or extracellularly.

ECM-affiliated proteins

As illustrated for collagens and proteoglycans, defining proteins strictly based on sequence features is challenging. For this reason, we proposed to group proteins that share structural or functional features with bona fide core ECM components but are not yet broadly accepted as ECM proteins under the umbrella term of “ECM-affiliated” proteins⁶¹ [Figure 2b]. Examples of ECM-affiliated proteins include triple-helical-motif-containing proteins such as ficolins or collectins, galectins, mucins, and surfactant proteins. Several families of transmembrane proteins were also included such as the proteoglycans syndecans and glypicans, semaphorins and plexins. This was motivated by experimental evidence having shown that the extracellular moiety of some proteins of these families could be released and found in association with the ECM⁶¹.

ECM-bound secreted factors

To quote Pr. Richard Hynes, one of the founders of the field of ECM research, the ECM is “not just pretty fibrils”. Indeed, many ECM proteins bind secreted factors such as growth factors, morphogens, or cytokines, regulating their bioavailability and modulating their signaling functions⁹¹. A prime example of the interplay between ECM fibers and growth factors is that of TGFβ, which is secreted in an inactive form as part of a complex with fibril-forming LTBP proteins and whose activation depends on proteolytic cleavage and mechanical stretching⁹². The interplay between ECM and growth factors can also be exemplified by the different isoforms of the vascular endothelial growth factor A (VEGFA) that are generated via alternative splicing and have different affinities for different GAGs (heparin or heparan sulfate proteoglycan) and as a result exert different signaling functions⁹³ [Figure 2c].

ECM regulators

Genes grouped under the category of ECM regulators encode enzymes whose main function is to modify the physical and chemical properties of ECM components [Figure 2d]. Examples of ECM regulators include several families of proteases that cleave ECM components such as metalloproteinases of the matrix metalloproteinases (MMPs), A Disintegrin and Metalloproteinases (ADAMs) and ADAMs with Thombospondin-1 motifs (ADAMTS)⁹⁴, pappalysins (PAPPAs), or meprins families, cathepsins (CTSs)⁹⁵, or elastases⁹⁶. This category also includes proteins whose function is to modulate the enzymatic activity of ECM proteases, such as Tissue Inhibitors of Metalloproteinases (TIMPs)^97,98, cystatins (CSTs), or serpin peptidase inhibitors (SERPINs). For further information on these proteolytic enzymes and regulators, readers are invited to explore the MEROPS peptidase database⁹⁹. The ECM regulators also comprise enzymes inducing other types of post-translational modifications (PTMs), such as the cross-linking enzymes of the lysyl oxidase (LOX) family¹⁰⁰, transglutaminases, and peroxidasin. The roles of these ECM regulators in contributing to ECM remodeling is discussed below.

Originally strictly limited to enzymes exerting an activity in the extracellular space, the list of ECM regulators was revised to include some intracellular enzymes, such as the procollagen-lysine 1, 2-oxoglutarate 5-dioxygenases (PLODs), prolyl hydroxylase domain-containing proteins (EGLNs or PHDs), prolyl 4-hydroxylases (P4Hs), or sulfatases, based on experimental evidence that those enzymes remained tightly bound to their substrates and were detected in ECM-enriched protein samples by proteomics (see below)¹⁰¹. It is reasonable to speculate that a better understanding of the mechanisms leading to the post-translational modifications of ECM proteins will result in the expansion of the ECM regulator compendium to include more systematically all enzymes, including those exerting intracellular functions, affecting any aspects of ECM protein metabolism and perhaps even those involved in GAG metabolism.

Proteoforms increase the compositional diversity of the matrisome

Although sequence-based prediction revealed that ~1,000 genes encode the human matrisome, the number of proteoforms resulting from genomic (e.g., single nucleotide polymorphisms), transcriptional (e.g., use of alternative promoters), post-transcriptional (e.g., alternative splicing), and/or post-translation modifications is at least an order of magnitude or two higher^102,103, contributing to increasing the diversity of matrisome components. The following sections will describe some examples of this added layer of diversity.

Alternative splicing and the example of fibronectin

ECM protein isoforms result from the alternative splicing of several matrisome genes^104,105. Perhaps the most striking example to illustrate how alternative splicing drives molecular diversity is that of fibronectin. Interrogation of the UniProt database reveals that fibronectin (entry P02751; canonical isoform P02751–15; 2,477 amino acids) exists in 17 isoforms produced by alternative splicing. Fibronectin isoforms present specific patterns of expression during development and diseases and can bind to different partners, and thus exert different functions^17,106. For example, the isoforms EIIIA and EIIIB, also termed EDA and EDB, contain an additional type III fibronectin domain (13^th or 8^th, respectively) [Figure 1d] and are predominantly expressed during development and in the pathological tumor or atherosclerotic vasculature¹⁰⁷. Other isoforms differ in the size of the variable “V” domain located between the 15^th and 16^th type III fibronectin domains [Figure 1d]. Recent studies have begun to uncover the complexity of the mechanisms leading to the alternative splicing of fibronectin in the context of atherosclerosis and aneurysm: A study using mice lacking both EIIIA and EIIIB showed that these two domains exerted protective effects on the endothelium from hemorrhage and aneurysm¹⁰⁸. Mechanistically, the expression of spliced fibronectin variants by endothelial cells was driven by a cross-talk between endothelial cells, macrophages and platelets^108,109. A follow-up study using ECM proteomics further showed that the inclusion of splice variants of fibronectin regulated the composition of the arterial ECM¹⁰⁹. In particular, the absence of EIIIA- and EIIIB-containing fibronectin led to a decrease in the immobilization of circulating fibrillin-1 in the arterial ECM under low flow, which was proposed as a mechanism to explain why mice deficient in fibronectin splicing isoforms are at a higher risk of intimal rupture. Isoforms of fibronectin containing the EIIIA and EIIIB domains have also been shown to be expressed specifically in the tumor-associated vasculature but not in the normal vasculature¹⁰⁷. It has been proposed to exploit this specificity by developing bi-functional antibodies directed against these isoforms to target the delivery of drugs or imaging agents to tumors^110–114. Isoforms arising from alternative splicing are not unique to fibronectin and similar examples can be provided for tenascin-C (six isoforms resulting from alternative splicing)^105,115,116, elastin (13 isoforms resulting from alternative splicing), but also many collagens and proteoglycans. In addition, ECM protein isoforms can arise from the usage of alternative promoters (e.g., COL9A1, COL13A1, COL18A1, LTBPs)¹⁰⁴.

Post-translational modifications

Post-translational modifications (PTMs) encompass a broad range of reversible or irreversible changes affecting the properties of a protein by the addition of chemical groups or proteolytic cleavage resulting, per the definition of the Proteomics Standards Initiative (PSI), “in an alteration of the measured molecular mass of a protein”^117–119. Hence, PTMs constitute another major source of diversity of matrisome proteoforms. Major PTMs of ECM components are described here [Figure 2d] and readers interested in additional details are invited to refer to a recent review for a comprehensive overview of the mechanisms and functions of ECM PTMs¹²⁰.

Apart from glycosylation, which is extensively reviewed elsewhere¹²¹, hydroxylation (or oxidation) of proline and lysine residues constitute perhaps the most extensively studied PTMs as they are obligatory modifications of collagens¹²². The major 4-hydroxyprolines and minor 3-hydroxyprolines contribute, in large part, to the stability of the collagen triple helix at physiological temperatures^122,123. Hydroxylysine residues, upon extracellular oxidative deamination, produce reactive aldehydes that can participate in the formation of intra- and inter-molecular cross-links, a first step towards supramolecular assembly of collagen fibers^122,124,125 (see below).

Tyrosine bromidation is a PTM that occurs upon oxidation of bromide into hypobromous acid (HOBr) by the extracellular enzyme peroxidasin (PXDN)¹²⁶. Interestingly, this enzyme, through its oxidizing potential, is also involved in the oxidation of methionine residues, which initiates the formation of covalent sulfilimine cross-links between methionine and hydroxylysine residues¹²⁷, a critical step to stabilize basement membrane ECMs^128–130.

It has been long appreciated that ECM proteins could undergo phosphorylation¹³¹. However, the kinases responsible for the phosphorylation of secreted proteins, including ECM components, have only been recently discovered. The FAM20C kinase, discovered a decade ago, is responsible for the phosphorylation of serine residues of a large panel of secreted proteins, including matrisome components^132–134. The vertebrate lonesome kinase (VLK) was discovered around the same time, and is responsible for the phosphorylation of tyrosine residues of ECM proteins and exerts key roles in modulating cell adhesion, migration and organogenesis^135–138.

In addition to hydroxylation, brominations and phosphorylation, many of the post-translational modifications observed on intracellular proteins are also found on matrisome proteins, including ubiquitination to direct proteins to proteasome-mediated degradation¹³⁹, acetylation¹⁴⁰, and citrullination (deamination of arginine residues)^141–143. Of clinical relevance, citrullinated ECM components including tenascin-C and collagen II are the targets of auto-antibodies in patients with rheumatoid arthritis^144–147, hence a better understanding of the mechanisms leading to citrullination could lead to future therapeutic opportunities.

In contrast to these additive PTMs, ECM proteins can also be enzymatically cleaved by a plethora of ECM regulators described above. As I will describe below, proteolysis has key roles in ECM biology, including in the maturation of some collagens as pro-forms, the release of biologically active products referred to as matricryptins or matrikines [Figure 1b] (ref. ¹⁴⁸), and in the remodeling of the ECM after it has assembled.

Defining the matrisome of organs and tissues using mass spectrometry

Transcriptomic analyses have revealed that one of the key drivers contributing to the building of tissue-specific matrisomes is the tissue-specific regulation of matrisome gene expression^149,150. However, gene- and transcript-level regulations only account for certain aspects of matrisome diversity. Over the past decade, mass-spectrometry-based proteomics has emerged as a robust method to decipher the compositional complexity of the ECM^62,151 and has revealed that the matrisome of any given tissue or organ is composed of well over 200 distinct proteins⁶³. The aggregation of datasets from multiple proteomic studies on the ECM of human and murine tissues into the ECM proteomics database, MatrisomeDB, has provided experimental evidence for nearly 98% of the predicted matrisome components⁶³. Comparison of the matrisome of different organs has further shown that, in addition to a subset of ubiquitous ECM proteins (such as collagens I, III, and IV, fibronectin, fibrillins, LTBPs and perlecan) expressed in overall similar relative abundance, tissue-or disease-specific ECM proteins or proteoforms can be identified. As described elsewhere, this new knowledge can be leveraged to identify novel diagnostic or prognostic biomarkers and to develop novel therapeutic strategies^{116,152–157}.

Mass spectrometry has also been instrumental in the identification of ECM PTMs^{118,158–160}, including of glycosylation sites by glycoproteomics^161–163, the nature of the glycans coupled to ECM proteins using glycomics^164–166, and even of novel proteoglycans^167–169. Mass spectrometry has also helped map hydroxylation of proline and lysine residues¹⁷⁰. For example, the substrate specificity of the prolyl 4-hydroxylases P4HA1 and P4HA2 was revealed by mass spectrometry, where P4HA1 was shown to preferentially hydroxylate proline residues preceded by a positively charged or a polar uncharged side chain amino acid, whereas P4HA2 preferentially hydroxylates proline residues preceded by a negatively charged amino acid¹⁷¹. Tissue-specific hydroxylation patterns were also revealed by mass spectrometric analysis. For example, in the tendon, P4HA2 is responsible of the hydroxylation primarily of proline residues of Gly-Glu-Pro or Gly-Asp-Pro triple helical collagen motifs¹⁷². Mass spectrometry has also allowed the identification of brominated proteins, substrates of peroxidasin, in the context of the renal basement membrane¹⁷³ and the healthy and fibrotic lung ECM¹⁷⁴. Lastly, technologies like degradomics and N-terminomics have contributed to the identification of ECM cleavage rules and products^175,176.

However, it is worth noting that certain regions of ECM proteins, sometimes large, still remain undetected by mass spectrometry, either because they are not accessible to proteolytic or chemical cleavage or because of chemical considerations that make them more difficult to resolve by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS)⁶². As a result, low sequence coverage significantly hinders the comprehensive mapping of ECM PTMs and the identification of isoforms and single amino acid variants¹⁷⁷ in the context of diseases. Hence, as a community we should shift our focus from aiming to increase protein identification to devising methods to increase sequence coverage. This effort will be further facilitated by creating custom protein databases, for example through in-silico translation of RNA sequencing datasets that would include these proteoforms^178,179.

In summary, the ECM environment is incredibly large and diverse, and the mere number of 1,000 matrisome genes only begins to capture the compositional complexity of the ECM. Importantly, different ECM compositions assemble into different architectures resulting in different tissue organizations that support broadly diverse physiological functions.

Diversity Of Ecm Structures

As mentioned above, the ECM exists in two structurally distinct forms: basement membranes and interstitial ECMs that differ in molecular composition, architecture, localization, the nature of cell–ECM interactions and functions [Figure 3].

Figure 3 |. Structural diversity of extracellular matrices.

a. The basement membrane is a thin sheet-like assembly of ECM proteins primarily composed of a self-assembled lattice of laminin trimers, a network of collagen IV, and the proteoglycans agrin and perlecan. Additional proteins contribute to the linkage between these components (e.g., nidogen). Cells express at their surface ECM receptors such as integrins or the laminin receptor dystroglycan. These receptors create a bridge between the ECM and the cellular cytoskeleton.

b. Examples of anatomical locations of basement membranes. Basement membranes are found in various locations, such as the basal pole of epithelial cells, surrounding endothelial cells, for example, at the blood–brain barrier interface, or around Schwann cells.

c. The interstitial ECM is a 3-dimensional meshwork found in connective tissues and primarily made of fibrillar ECM proteins (collagens I, III, V, IX; fibronectin) and glycosaminoglycans and proteoglycans. Cells found in connective tissues (e.g., fibroblasts) are surrounded by the interstitial ECM and interact with proteins of interstitial ECMs through receptors such as integrins, the tyrosin kinase discoidin domain receptors (DDRs) or the hyaluronic acid receptor CD44.

Basement membranes

Basement membranes are thin (10s-100s of nanometers) sheet-like assemblies. The core components of basement membranes include collagen IV, the glycoproteins laminins and nidogens, and the proteoglycans agrin (AGRN) and perlecan (HSPG2)^180,181 [Figure 3a]. The now archived Matrixome Project website hosts the mouse basement membrane bodymap that provides information on the distribution of these core basement membrane proteins in the mouse embryo. Beyond this “basement membrane toolkit”, proteomics experiments have revealed that dozens of other matrisome components can associate with the basement membrane including collagens VI, VII, XV, XVIII and ECM glycoproteins like fibulins, nephronectin and papillin. These are now described in the database basement membraneBase¹⁸².

Basement membranes are found in discrete locations including lining the basal pole of epithelial cells and creating the interface between epithelia and connective tissues (e.g., epidermis–dermis interface), lining the vascular endothelium, and surrounding certain cell types like Schwann cells , adipocytes, and myofibers [Figure 3b]. Basement membranes are also key components of stem cell niches (e.g., satellite cells of skeletal muscle^183,184, epidermal stem cells located in the hair follicle bulge or the interfollicular epidermis^185,186), although the specific ECM proteins capable of sustaining stemness remain to be identified.

Cell–basement membrane interactions are mediated by cell surface receptors including integrins and the dystroglycan receptor, and establish a linkage with the actin and intermediate filament cytoskeletons through focal adhesions and hemidesmosomes, respectively¹⁸⁰. Functionally, basement membranes contribute to supporting and stabilizing the structures they underline, enable cell polarization, and form a mechanical and physical barrier that contributes to key processes including glomerular filtration or the formation of the blood-brain-barrier^181,187. Breaching of the basement membrane is a critical step in the early dissemination of tumor cells on their metastatic journey^188,189.

Interstitial extracellular matrices

In contrast to basement membranes, interstitial ECMs are porous meshworks constituting the bulk of connective tissues (dermis, cartilage, bone) and fill the parenchyma of all organs. Fibroblasts are the primary cell type producing the fibrillar proteins of interstitial ECMs including collagens I, II, III, V, IX, fibronectins, and the components of elastic fibers (e.g., elastins, fibrillins, microfibril-associated proteins) [Figure 3c]⁵⁰. Functionally, interstitial ECMs define the mechanical properties (rigidity, elasticity) of tissues and organs and support the functions of the mesenchymal cells they surround. Fibrotic disorders are characterized by an excessive accumulation of interstitial ECMs^42,50.

Assembly Of the Extracellular Matrix

The assembly of an ECM is a dynamic multi-step process. This section highlights novel insights into the mechanisms leading to ECM protein synthesis, secretion and assembly in the extracellular space. This section will also discuss newly identified regulators of these processes, including amino acid availability, post-translational modifications, and the circadian clock.

Metabolic requirements of ECM protein synthesis

ECM proteins are first synthesized in the lumen of the rough endoplasmic reticulum (ER). ECM protein synthesis is energitically demanding due to their large size and presents unique challenges due to their unique amino acid composition, in particular for collagens with high proline and glycine content^190–192. An emerging focus of ECM biology is to study the metabolic pathways contributing to ECM protein synthesis. The neutral amino acid transporter SLC38A2 was recently shown to be responsible for proline uptake by differentiating osteoblasts and to be required for the synthesis of proline-rich proteins such as the α1 chain of collagen I (COL1A1) and osteocalcin (also known as bone gamma-carboxyglutamic acid-containing protein, BGLAP) to sustain bone matrix production¹⁹³. Proline and glycine are non-essential amino acids, meaning that they can also be synthesized by cells directly. A recent study showed that cancer-associated fibroblasts synthesize proline from glutamine through the conversion of 1 pyrroline-5-carboxylate by the pyrroline-5-carboxylate reductase (PYCR1) to increase collagen I synthesis¹⁹⁴. Interestingly, mutation of PYCR1 in humans results in cutis laxa, a connective tissue disorder characterized by excessively loose skin¹⁹⁵. Along the same line, cancer-associated fibroblasts have been shown to use pyruvate carboxylase to fuel the TCA cycle to sustain levels of α-ketoglutarate, a precursor of glutamate, that can be used to synthesize proline and sustain collagen I production¹⁹⁶. Metabolomics has further demonstrated the existence of a feedback mechanisms between ECM stiffness and arginine and proline metabolisms in murine dermal fibroblasts in vitro and showed that it was mediated by the mechano-sensitive channel Piezo1 (ref.¹⁹⁷), offering potential novel insights into the mechanisms of skin fibrosis. The concept of “metabo-reciprocity” has recently emerged from this body of work to describe the bi-directional interplay between ECM mechanics and cellular metabolism¹⁹⁸.

ECM protein transport and secretion

In addition to being the site of protein synthesis, the ER is also the site of initial polymer assembly for collagen trimers or fibronectin dimers. The mechanisms leading to the vesicular trafficking of ECM proteins, in particular collagens, from the ER through the Golgi apparatus and their extracellular secretion have also been the focus of active research over the past decade [Figure 4]. We now understand that specific transmembrane cargo receptors (e.g., the transport and Golgi organization 1 or TANGO1 protein, the cutaneous T-cell lymphoma-associated antigen 5 cTAGE5, the KDEL receptor that recognizes the recognises the Lys-Asp-Glu-Leu amino acid sequence^199,200), chaperones that link cargo to receptors such as heat shock protein 47 (HSP47) (refs. ^201,202), and components of the coat protein complex II (COPII) (e.g., members of the SEC23 and SEC24 families) are required to load large-sized ECM cargos (up to several hundreds of nanometers for fibrillar collagens) into COPII-coated vesicles that transport proteins from the ER to the Golgi [Figure 4]^203,204. It has also emerged that different ECM cargos engage different partners. For example, it was recently shown that procollagen I export preferentially involves SEC24A/B, whereas fibronectin preferentially involves SEC24C/D²⁰⁵. The ER and the Golgi apparatus are also the sites of post-translational modifications. Importantly, only adequately post-translationally modified ECM cargos, i.e., those that adopt proper folding, will be trafficked on and secreted. For example, a recent study demonstrated that the secretion of collagen IV, but not of I or III, was decreased in cells lacking lysyl hydroxylase 3 (ref. ²⁰⁶).

Figure 4 |. Synthesis, secretion, and supramolecular assembly of collagens.

a. ECM proteins are synthesized in the endoplasmic reticulum and transported in coat protein complex II (COPII)-coated vesicles to the Golgi network, where additional modification steps take place. From the Golgi, ECM proteins are transported to the extracellular space in secretory vesicles. Inset (1) depicts collagen chain synthesis and trimerization in the endoplasmic reticulum and the formation of triple helical procollagen. Inset (2) illustrates the mechanisms of ECM cargo loading into COPII vesicles. Inset (3) shows the cleavage of collagen pro-peptides by N- and C-proteinases upon secretion into the extracellular space to form a mature collagen helix. Inset (4) depicts the diversity of supramolecular structures resulting from collagen assembly. Collagens I, II, III, V, XI, XXIV, and XXVII assemble to form fibrils and higher-order fibers. Collagens IX, XII, XIV, XVI, XIX, XX, XXI, and XXII are termed fibril-associated collagens with interrupted triple helices (FACITs) since they interact with fibrillar collagens. Collagens IV, VIII, and X assemble to form geometrical networks (“chicken-wire-like” for collagen IV and hexagonal networks for collagens VIII and X), where the triple helical collagens are connected by their respective non-collagenous domains (shown in grey). Collagen VI forms beaded and collagen VII forms anchoring fibrils bridging the basement membrane and the interstitial ECM.

Mechanisms of ECM protein assembly and ECM deposition

The building of the ECM is a multi-step and multi-scale process that is mediated by protein interactions, some being initiated intracellularly and some occurring in the extracellular space. The following sections lay out the main pathways and regulatory mechanisms leading to ECM assembly.

Self-assembly of ECM proteins

Many ECM components can spontaneously assemble. For example, collagens first assemble into triple helical monomers in the ER [Figure 4, inset 1] and can further assemble in supramolecular structures that allow their classification into six groups: fibril-forming collagens (I, II, III, V, XI, XXIV and XXVII), fibril-associated collagens with interrupted triple helices (FACITs; IX, XII, XIV, XVI, XIX, XX, XXI and XXII), network-forming collagens (IV, VIII and X), beaded-filament-forming collagen VI, and anchoring-fibril-forming collagen VII [Figure 4, inset 4]¹⁴. The self-assembly of collagens has been extensively studied for certain collagens in vitro²⁰⁷. For example, we now know that, upon secretion and cleavage of their carboxy-terminal pro-peptides that maintain collagens soluble, fibrillar collagens can form fibrils and higher-order fibers [Figure 1b, Figure 4, insets 3 and 4]. In contrast, triple helical monomers of collagen VI first assemble intracellularly into antiparalellel dimers via disulfide bonds and then into tetramers. Once secreted, collagen VI tetramers further assemble into beaded filaments via interactions between their amino-terminal VWA domains [Figure 4, inset 4], a termed given based on the appearance of these structures by electron microscopy²⁰⁸.

Laminin trimers self-assemble into a polymeric sheet-like lattice in the basement membrane [Figure 3a]¹⁸⁰. Recently, a combination of biochemical and imaging approaches has led to the mapping of the precise domains and residues involved in the trimerization of recombinant laminin chains mixed together in vitro^209,210. Although developed based on purified proteins in vitro, the structure resulting from these recent studies has the potential to help uncover the basis of lamininopathies and the development of therapeutic strategies for these disorders^210,211.

ECM assembly relies on heterotypic protein–protein and protein–glycan interactions

Beyond the self-assembly of some ECM components, the formation of the ECM scaffold depends on heterotypic interactions involving proteins and glycans.

The example of fibronectin will be used to illustrate this [Figure 5]: Fibronectin is secreted into the extracellular space as a dimer and its assembly into an insoluble fibrillar scaffold is a step-wise process involving, in brief, 1) interaction with the cell surface receptors integrins, 2) force-dependent conformational changes of fibronectin upon integrin binding and cell contractility, 3) end-to-end intermolecular interactions between fibronectin dimers, and 4) the lateral extension of fibronectin fibrils^52,212,213. Although we now have a good understanding of the molecular requirements governing fibronectin–integrin binding, the details of the molecular mechanisms leading to the formation of insoluble fibronectin fibrils continue to be unraveled. For example, the early step of fibronectin assembly was recently shown to depend on the interaction of fibronectin through its 13^th type III fibronectin domain with heparan sulfate^214,215 [Figure 5b]. Moreover, the cross-linking enzyme transglutaminase 2 was shown to bind fibronectin and facilitate its interaction with integrins in the early step of fibronectin fibril assembly²¹⁶.

Figure 5 |. Mechanism of fibronectin fibrillogenesis.

a. Fibronectin fibrillogenesis is a multi-step process involving 1) the binding of folded fibronectin dimers to heterodimeric integrin receptors expressed at the cell surface, 2) conformational changes leading to the stretching of fibronectin dimers, 3) the lateral extension of the fibronectin fibril array, and 4) the assembly of collagen fibers using fibronectin fibers as a template.

b. Map representing the domain-based organization of fibronectin and its known interacting partners and their binding sites. Notably, fibronectin interacts with cell surface receptors such as integrins primarily via its 9^th and 10^th FN3 domains and syndecans. It also interacts with glycosaminoglycans like heparin (via amino-terminal FN1 domains or carboxy-terminal FN3 domains), with core matrisome proteins (e.g., collagens, fibrillins, and tenascin C), and with growth factors (e.g., VEGF, TGFβ, or FGF). Fibronectin can also be cleaved by proteolytic enzymes cathepsin D. A circulating form of fibronectin is found in the plasma, where it plays a role in clotting by interacting with fibrin and thrombin and in hemostasis and fibrinolysis by being degraded by the proteolytic enzymes urokinase and plasmin.

Once assembled, the fibronectin fiber scaffold serves as a substrate for the assembly of other meshworks including the collagen^217–219 and fibrillin^220,221 meshworks [Figure 5b]. The interdependence of different meshworks in the assembly of the ECM scaffold extends beyond the fibronectin meshwork. For example, the small proteoglycans decorin and biglycan can modulate the assembly and arrangement of the collagen I meshwork in the mouse tendon²²². Similarly, the incorporation of fibrillins 1 and 2 in the ECM is dependent on its interaction with LTBP1 (ref. ²²³). Interestingly, the deletion of a segment of fibrillin 1 causing Weill–Marchesani syndrome impacts not only the ultrastructure of fibrillin microfibrils but also the interaction of fibrillin 1 with LTBP1, demonstrating how mutation in one ECM gene can have consequences on the entire ECM scaffold²²⁴.

The establishment of interaction networks within the ECM and at the cell–ECM interface is of paramount importance to build the ECM scaffold and sustain ECM functions. Readers interested in exploring the interaction networks formed by ECM proteins can refer to MatrixDB , a database that repertories experimentally validated interactions between intact ECM proteins or ECM protein fragments and other ECM proteins, ECM receptors, GAGs, lipids, and cations²²⁵.

Circadian regulation

In recent years, we have gained an appreciation for the highly dynamic nature of ECM assembly processes. In particular, new research has shown that the circadian clock is a key regulator of ECM biosynthesis and homeostasis. Because of its extremely well-characterized architecture and somewhat simpler composition (predominantly fibrillar collagen I), the ECM of the tendon has been the focus of investigations on what is now termed the “chrono-matrix”²²⁶. In this system, the expression of several genes encoding proteins involved in collagen secretion (e.g., SEC61, TANGO1) was shown to be rhythmic and to result in procollagen synthesis during the nighttime and collagen fibril assembly during the daytime²²⁶. Interestingly, in this system, the expression of Col1a1 and Col1a2 was not found to be rhythmic. The induction of collagen hydroxylation by P4HA1 in the cartilage was also recently shown to be rhythmic and to be under the control of the transcription factor and clock protein Bmal1²²⁷. Indeed, exposing mice to artificial light at night led to a downregulation of Bmal1 and a decrease in proline hydroxylation and collagen trafficking in chondrocytes²²⁷. Beyond these two examples, we now have evidence for the circadian rhythm to modulate ECM metabolism in the skin²²⁸, cartilage²²⁹, intervertebral discs²³⁰ and kidney²³¹.

Mechanisms Of Physiological and Pathological Ecm Remodeling

Once assembled, the ECM undergoes remodeling that results in new topologies (e.g., bundling, alignment, pore size) and mechanical (e.g., visco-elasticity, stiffness) and signaling properties. This remodeling is essential to maintain ECM homeostasis and to support cell, tissue and organ functions.

The mechanisms of physiological ECM remodeling are diverse and involve enzymatic and non-enzymatic processes, as well as cell–ECM interactions. Importantly, impairment of these mechanisms can lead to altered ECM architecture and functions.

ECM protein cross-linking

Cross-links between ECM components can be mediated by different enzymes including lysyl oxidases (LOXs)¹⁰⁰, transglutaminases such as transglutaminase 2 (TGM2) or Factor XIII (F13A1), and peroxidasin (PXDN, see above). Cross-linking can also occur through non-enzymatic reactions involving advanced glycation endproducts (AGEs), which are proteins that become glycated as a result of exposure to glucose, fructose, or derivatives²³². Whereas basal cross-linking contributes to the stability and mechanical properties of the ECM meshwork and modulates physiological processes^233–236, excessive cross-linking is a hallmark of many pathological processes, including scarring²³⁷, cancer²³⁸, and fibrosis⁴⁹. Interrogation of MatrisomeDB reveals that cross-linking enzymes are more abundant in the ECM of tumor microenvironments than in healthy tissues⁶³. Functionally, LOX-mediated increased ECM stiffness has been shown to accompany all steps of cancer progression, from initiation to sustaining tumor cell proliferation to metastatic dissemination^239–242. Increased collagen stiffness was shown to up-regulate the expression of pro-fibrotic genes and matrisome genes in adipocytes²⁴³. Similarly, AGEs contribute to the initial modeling of the fibronectin and collagen meshworks^244,245. However, excessive AGE-mediated cross-links have been shown to accompany tissue aging^246,247, liver fibrosis²⁴⁸, breast cancer metastasis²⁴⁹, and clinical presentations of diabetic patients^{246,247,250,251}.

ECM protein degradation and endocytosis-mediated recycling

As extensively reviewed elsewhere, proteolysis is another major mechanism leading to the remodeling of the composition and structure of the ECM^252,253, thereby generating ECM proteoform diversity in structure and function. Proteolytic cleavage can also result in the release of bioactive molecules that exert functions different from native ECM proteins. For example, proteolytic cleavage of certain collagens leads to the release of anti-angiogenic peptides, including arresten derived from the non-collagenous domain of the α1 chain of collagen IV, canstatin derived from the non-collagenous domain of the α2 chain of collagen IV or endostatin derived from the non-collagenous domain of the α1 chain of collagen XVIII [Figure 1b] (ref. ²⁵⁴). Research has begun to decipher the mechanism of action of these naturally-occurring bioactive fragments and has shown that they can engage ECM receptors as well as growth factors and cytokine receptors, opening the possibility of leveraging these molecules for therapeutic purposes^148,255.

Beyond the generation of biologically active fragment, we do not yet fully understand the fate of ECM cleavage products. It has now emerged that these fragments can be further catabolized or recycled. For example, proteolytic fragments of collagen can be endocytosed upon binding to the collagen receptor urokinase plasminogen activator receptor-associated protein (uPARAP or Endo180) and directed to lysosomes where they will be further digested, for examples by intracellular cathepsins^256,257. Interestingly, the remodeling of the collagen ECM has been shown to be under the control of the circadian clock as the expression of the collagen-degrading enzymes, cathepsin K and MMP14, was shown to be rhythmic in the tendon^226,258. Of note, collagen degradation represents an important source of proline and glycine available for collagen neo-synthesis. Similarly, the matricellular proteins thrombospondins 1 and 2 can undergo uPARAP-mediated endocytosis²⁵⁹ as a way to regulate their extracellular availability. Another matricellular protein, tenascin-R, characteristic of the brain ECM, contributes to ECM remodeling through cycles of endocytosis and resurfacing²⁶⁰. This constitutes a novel regulatory mechanism that can rapidly fine-tune ECM-mediated signaling.

Similarly to cross-linking enzymes, ECM-degrading enzymes, in particular MMPs and elastases are highly abundant in the ECM of tumor microenvironments as compared to healthy tissues⁶³. These enzymes contribute to different steps of the metastatic cascade, from local invasion and intravasation to extravasation and seeding of distant organs²⁶¹. Excessive ECM degradation also correlates with other diseases like osteoarthritis²⁶², rheumatoid arthritis²⁶², or chronic obstructive pulmonary disease^263,264. Although ECM protein degradation primarily impacts the architectural integrity of the ECM meshwork, it can also lead to the release of elevated levels of bioactive fragments. Consequently, the potential of using ECM degradation fragments either as biomarkers or therapeutic target is being actively explored^265–268.

Collectively, these examples illustrate how physiological ECM remodeling results from a finely tuned balance between ECM build-up and break-down and show how any disruption will lead to ECM disorganization and the initiation of a pathological cascade.

ECM remodeling mediated by cell–ECM interactions

As previously discussed, ECM protein interactions with their respective receptors are a critical step in the initiation of ECM assembly. However, cell–ECM interactions, through dynamic reciprocity, are also key to remodel the ECM meshwork and create new ECM topologies in normal and pathological contexts [Figure 6]^30,269,270. In brief, through outside-in signaling, cell–ECM interactions initiate the remodeling of the intracellular acto-myosin cytoskeleton that results in changes in integrin conformation modulating, in turn, via inside-out signaling, the affinity of integrins for their ECM ligands. Readers are invited to refer to the following references for in-depth descriptions of the mechanisms of integrin activation^28,32.

Figure 6 |. Dynamic reciprocity and mechanisms of ECM remodeling.

Dynamic reciprocity describes the loop established by outside-in (blue arrows) and inside-out (red arrows) signalling between cells and the ECM. Dynamic reciprocity initiated by ECM ligand–ECM receptor interactions (e.g., fibronectin–integrins) contributes to ECM remodelling in several ways: (1) ECM ligand–integrin interaction results in a change of integrin conformation that increases the integrin’s affinity for its ligand. (2) This change in affinity results in changes in mechanical forces exerted by integrins and the cytoskeleton on the ECM ligand, but also, by proxy, on other ECM proteins binding to fibronectin (e.g., fibrillins, collagens, LTBPs). (3) Activation of integrin signalling — from the cell surface or from the endocytic compartment — can lead to the activation of ECM gene transcription, leading to de novo ECM protein synthesis and consequently (4) changes in overall ECM composition and architecture.

Importantly, cell–ECM interactions trigger biochemical and mechanical signals that result in structurally and compositionally different ECMs. Indeed, the forces exerted by integrins on their ligands modulate physical parameters like ECM fiber alignment [Figures 5 and 6]. In addition, ECM-induced signaling pathways regulate the expression of matrisome genes through mechanisms that remain to be elucidated. An example of the importance of cell–ECM interactions in ECM remodeling recently emerged from a study that showed that interactions of cells with the basement membrane components collagen IV and laminin via theα2β1 and α3β1 integrins respectively, serve as sites of initiation of fibronectin assembly. Upon initiation of fibronectin fibrillogenesis, cells can form a new type of “sliding” α5β1-integrin-rich focal adhesions that, through interactions with the acto-myosin cytoskeleton, lead to the further extension of the fibronectin array²⁷¹. Beyond signaling events that directly affect the ECM, ECM-induced signaling pathways transduced by integrins, such as the mitogen-activated protein kinase (MAPK) or the PI3K pathways primarily control many cellular phenotypes including proliferation and survival²⁷². Of note, recent studies have demonstrated that ECM-bound integrins could also signal from intracellular vesicles upon endocytosis^273,274.

Remodeling of the ECM during aging

Dysregulation of ECM homeostasis in aging exemplifies the importance of the tight regulation of the mechanisms leading to the building of the ECM. First, transcriptomic and proteomic studies have provided evidence that the expression of matrisome genes and matrisome protein biosynthesis and incorporation in the ECM vary with aging^{150,275–279}. Structural, proteomic and biophysical measurements have also highlighted that ECM proteins are altered in aged tissues through the accumulation of damaging PTMs (e.g., glycations, cross-links) that alter protein structure and consequently functions²⁸⁰. Altogether, this results in the disruption of the integrity (destabilization or over-stabilization) of the ECM scaffold, with macroscopic consequences on tissues like the skin or the cartilage. The remodeling of the ECM during aging also creates a microenvironment favorable (in composition and/or structure) to the etiology or progression of diseases primarily occuring in aging patients, such as age-related macular degeneration²⁸¹, prostate cancer²⁸², or melanoma metastasis²⁸³.

Development of Matritherapies to Achieve Ecm Normalization

With a better understanding of the molecular makeup of the ECM, of the mechanisms leading to physiological ECM assembly and remodeling, and of the mechanisms causing altered ECM homeostasis in disease, we can now envision the development of what we propose to coin “matritherapies”, i.e., strategies to normalize ECM composition and/or architecture to restore ECM homeostasis and achieve a therapeutic benefit [Figure 7].

Figure 7 |. Mode of actions of matritherapies.

Several levels of action are envisioned to maintain or restore ECM architecture and functions therapeutically. (1) Modulation of ECM gene expression can be achieved by interfering with upstream signaling pathways. One such example is the angiotensin II receptor blocker, Losartan, which reduces the levels of pro-fibrotic TGFβ and, consequently, levels of collagens. (2) Direct modulation of ECM gene expression and/or transcript levels can be achieved by gene editing or exon skipping, for example, by using antisense oligonucleotides. (3) Modulation of the translation of ECM proteins can help restore function in disease models where mutation-induced stop codons lead to pathological protein variants. For example, drugs such as aminoglycoside and non-aminoglycoside premature termination codon (PTC) readthrough drugs have been used to restore the synthesis of nonsense variants of chains of collagen IV. (4) Modulation of the activity of ECM remodeling enzymes such as cross-linking enzymes (e.g., lysyl oxidase (LOX) inhibitors) or proteases (e.g., matrix metalloproteinase (MMP) inhibitors). (5) Blocking of the interaction between ECM ligand and their receptor (e.g., integrin blocking antibodies). (6) Inhibition of the enzymatic activity of effectors activated downstream of ECM ligand–ECM receptor interactions (e.g., inhibition of the focal adhesion kinase, FAK).

Modulation of ECM gene expression, post-transcriptional and translational modifications

Targeting the signaling pathways leading to the up-regulation of ECM genes (e.g., TGFβ signaling pathway) has been an active route of investigation and has achieved significant success. A successful example is the discovery that the angiotensin II receptor type 1 inhibitor, Losartan, could dampen TGFβ signaling in an animal model of Marfan syndrome caused by mutations in fibrillin 1 (ref. ²⁸⁴) and improve the cardiovascular presentations (e.g., reduction of aortic dilatation) of Marfan syndrome patients²⁸⁵. Losartan is now being explored in the treatment of a broad spectrum of fibrotic disorders^286,287.

The possibility of modulating matrisome gene expression by intervening post-transcriptionally is also being explored. For example, gene editing and exon-skipping strategies using anti-sense oligonucleotides are showing promise in the context of epidermolysis bullosa caused by COL7A1 variants²⁸⁸. A recent study reported the development of an anti-sense oligonucleotides targeting exon 73 of COL7A1 in a topical gel formulation. After demonstrating that collagen VII lacking exon 73 had properties similar to that of wild-type collagen VII, the authors showed that delivery of the anti-sense nucleotide restored close-to-basal collagen VII abundance in a human skin model in vitro²⁸⁹.

Last, the possibility of modulating matrisome gene expression by intervening at the translational (i.e., protein synthesis) level is being researched in the context of genetic variants of matrisome genes. Non-sense variants, leading to the introduction of premature termination codons (PTCs) and the synthesis of truncated ECM proteins have been found in the COL4A5 gene of Alport syndrome patients²⁹⁰. An in vitro luciferase-based assay has demonstrated the potential of compounds permitting stop-codon readthrough (e.g., aminoglycoside and non-aminoglycoside premature termination codon readthrough drugs) to restore the expression of certain but not all variants of the α5 chains of collagen IV and further showed that the restored proteins, now including a complete carboxy-terminal domain, can trimerize with an α3 and an α4 chains of collagen IV²⁹¹. However, PTC readthrough-based approaches come with limitations and the risk of side effects²⁹², and it will be important to overcome these to allow successful clinical translation. Nonetheless, these examples demonstrate the promise of modulating ECM gene expression or ECM protein translation to achieve therapeutic benefits.

Modulation of cell–ECM interactions and ECM signaling

Modulating cell–ECM interactions has been another active area of investigation, with a prominent focus on attempting to inhibit ECM–integrin interactions or downstream signaling effectors such as the focal adhesion kinase (FAK) to disrupt ECM-triggered signaling pathways^293–295. Successful examples include the development of biologics or small molecules blocking leukocyte-specific α4β7 and α4β1 integrins for the treatment of inflammatory disorders (e.g., inflammatory bowel disease, ulcerative colitis, Crohn’s diseases, rheumatoid arthritis) or the platelet-specific αIIbβ3 integrins for the management of thrombotic events [Figure 7]²⁹⁴. However, in these examples, the therapeutic benefit is exerted through the inhibition of integrin-dependent signaling pathways, not by achieving ECM remodeling. Nonetheless, these precedents can serve as models to devise future approaches to modulate cell-dependent remodeling of ECM architecture.

A recent example epitomizes the promise of modulating ECM signaling to achieve ECM normalization or restoration. Osteoarthritis (OA) is a debilitating degenerative disease of the joints, characterized by ECM degradation. An appealing approach to counter-balancing OA is to devise strategies to prevent ECM breakdown and/or promote ECM buildup. In a recent study, a short fragment of the secreted matrisome factor angiopoietin-like 3 (ANGPTL3) encompassing the carboxy-terminal fibrinogen-like (FBN-like) domain, termed LNA043, was shown to bind to α5β1 integrin and to elicit signals leading to the synthesis of cartilage ECM proteins in vitro and in vivo in a randomized phase I clinical trial. Although the mechanisms of action of LNA043 are not fully elucidated yet and its long-term benefits in vivo remain to be demonstrated, this is a promising example of matritherapy.

Modulation of the activity of ECM remodeling enzymes

Lastly, strategies aiming to modulate the action of ECM remodeling enzymes (e.g., hyaluronidase, LOX, LOXL2, MMPs) to normalize ECM architecture are also envisioned. And while early attempts ended prematurely due to lack of therapeutic benefits or, in certain cases, severe adverse effects²⁹⁶, new developments, permitted by significant advances in our understanding of the fundamental mechanisms of action of these ECM remodeling enzymes, hold tremendous potential^261,297. For example, the LOX inhibitor β-aminopropionitrile (BAPN) has been shown to revert ECM stiffening in different tumor models, resulting in improved efficacy of immune checkpoint inhibitors²⁹⁸. However, the clinical use of BAPN has been impeded due to its toxicity. A newly developed pan-LOX inhibitor shows promise in decreasing tumor stiffness and in improving tumor perfusion and the efficacy of chemotherapy in a mouse model of pancreatic ductal carcinoma, a tumor type characterized by its stroma content and excessive ECM accumulation²⁹⁹.

Conclusions and perspectives

The past few decades of ECM research have demonstrated that the ECM is not an inert scaffold. Rather, ECMs are complex and diverse in their compositions and architectures, and undergo constant, yet tightly regulated, remodeling. Many questions pertaining to ECM assembly and functions remain unanswered. For example, we have yet to build a spatially-resolved atlas of the ECM at the single-cell resolution to determine the positioning of ECM components with regard to each other. The ECM of a given tissue or organ is the result of a multi-scale assembly and we still do not know precisely which cell populations secrete ECM proteins, which populations secreted ECM modulators, and finally which cell populations express receptors for these ECM proteins. Consortium-led mapping efforts, such as the NIH Human Biomolecular Atlas Program, should help address these open questions³⁰⁰. In addition, until now, many aspects of ECM biology have been studied in isolation and in vitro, but all aspects of ECM biology are intertwined: the biochemical composition of the ECM directly affects its physical and mechanical properties and signaling functions. There is thus a critical need to devise integrated approaches to perform system-level analyses of native ECMs together with the cells they support. Last, sustained efforts aimed at deciphering the fundamental aspects of ECM metabolism, from synthesis to secretion to assembly and post-assembly remodeling are of paramount importance if we want to be able to devise successful matritherapies to sustain healthy aging and benefit patients suffering from pathologies originating from or aggravated by ECM disorders.

Glossary

Intimal Rupture

Rupture of the innermost — intinal — layer of a blood vessel wall. The tunica intima is composed of the endothelium and a subendothelial ECM including a basement membrane and elastic layer in arteries.

Schwann Cells

Glial cells that form the myelin sheath of axons of the peripheral nervous system.

Proteoforms

Different molecular forms of a proteinproduced from a single gene. Can arise from single amino acid variant, alternative splicing, or post-translational modification.

Liquid Chromatography Coupled To Tandem Mass Spectrometry (LC-MS/MS:)

Analytical proteomic methods that combines peptide separation by liquid chromatography and the subsequent analysis of these peptides by tandem mass spectrometry.

Epidermolysis Bullosa

Group of rare genetic skin disorders characterized clinically by skin and mucosal blistering and skin fragility.

Premature Termination Codon Readthrough Drugs

Drugs that allows the translation machinery to read through a premature STOP codon and promotes the synthesis of a full-length protein.

Immune Checkpoint Inhibitors

Drugs that block proteins involved in immune checkpoints, the pathways that regulate the activation of the immune system. In cancer imunotherapy, these drugs (e.g., anti-CTLA4, anti-PD-1) are used to activate anti-tumor immunity.