Exercise builds the scaffold of life: muscle extracellular matrix biomarker responses to physical activity, inactivity, and aging

Skeletal muscle extracellular matrix (ECM) is critical for muscle force production and the regulation of important physiological processes during growth, regeneration, and remodelling. ECM remodelling is a tightly orchestrated process, sensitive to multi‐directional tensile and compressive stresses and damaging stimuli, and its assessment can convey important information on rehabilitation effectiveness, injury, and disease. Despite its profound importance, ECM biomarkers are underused in studies examining the effects of exercise, disuse, or aging on muscle function, growth, and structure. This review examines patterns of short‐ and long‐term changes in the synthesis and concentrations of ECM markers in biofluids and tissues, which may be useful for describing the time course of ECM remodelling following physical activity and disuse. Forces imposed on the ECM during physical activity critically affect cell signalling while disuse causes non‐optimal adaptations, including connective tissue proliferation. The goal of this review is to inform researchers, and rehabilitation, medical, and exercise practitioners better about the role of ECM biomarkers in research and clinical environments to accelerate the development of targeted physical activity treatments, improve ECM status assessment, and enhance function in aging, injury, and disease.


I. INTRODUCTION
Muscle cells (myofibres) adapt rapidly to acute and chronic changes in their environment (Frontera & Ochala, 2015). All animal cells are connected through a network-like scaffold known as the extracellular matrix (ECM); generated by the cells to support their biological functions including cellular integrity, signalling, differentiation, migration, proliferation, fusion, and survival (Hynes, 2009), and this complex environment also shows significant adaptive potential (Grounds, 2008). The skeletal muscle ECM provides both structural support and biochemical cues and is composed of many interacting molecules, with some key proteins indicated in Fig. 1. Collagens are major ECM components (Table 1), especially collagens I and III in the interstitial connective tissues (where force is transferred to the skeleton), while collagens IV and VI are closely associated with the specialised ECM called the basement membrane (rich in laminins) connected to the myofibre surface (sarcolemma) that transfers the force generated by myofibre contraction out to the interstitial collagens (Grounds, Sorokin & White, 2005;Grounds, 2008). Many basement membrane molecules are in intimate contact with satellite cells (muscle precursor stem cells or quiescent myoblasts) and contribute to the regulation of myogenesis and new muscle formation. Moreover, proteoglycans such as decorin and perlecan, and glycosaminoglycans have multiple interactions and as a consequence have diverse functions in skeletal muscle, including sequestering of key growth factors with roles in myogenesis and fibrosis (Grounds, 2008;Schaefer & Schaefer, 2010;Brandan & Gutierrez, 2013;Chaturvedi et al., 2015). Despite its importance, surprisingly little is known about the molecular markers available for assessing changes in ECM remodelling and status. By volume, the ECM makes up a large proportion of the total connective tissue network throughout the body. In muscles, the connective tissues form a mesh-like structure that encloses the myofibres (endomysium, with basement membrane), muscle fascicles (perimysium), and entire muscles (epimysium) (Purslow, 2020). The endomysium encompasses single myofibres and includes their ECM. It constitutes 0.47-1.2% of the dry mass of muscles (Purslow, 2010), and contains the network-forming collagen IV, and to a lesser extent the fibril-forming collagens I, III, and V, as well as collagens VI, and XII (Listrat, Picard & Geay, 1999;Listrat et al., 2000;Passerieux et al., 2006;Jakobsen et al., 2017). The perimysium ensheathes muscle fascicles and constitutes from 0.43% to 4.6% of the muscle dry mass, with the amount varying much more between muscle groups than for the endomysium (Purslow, 2010). It contains much higher quantities of collagens I and III compared with the endomysium (Kurose et al., 2006). Collagens VI, V, VI, and XII are also present in smaller quantities (Listrat et al., 1999(Listrat et al., , 2000Petibois et al., 2006). The epimysium surrounds each muscle and creates a layer that is continuous with the tendons, and mechanically connected to the perimysium (Purslow, 2010(Purslow, , 2020. Whilst the exact amount of ECM is not known in humans, these structures account for 1-7% of the total adult bovine skeletal muscle dry mass (Light et al., 1985), and collectively are known as muscle connective tissue or deep muscle fascia. Although the skeletal muscle ECM is traditionally referred to as 'intramuscular connective tissue' (Jozsa et al., 1988;Lapier et al., 1995;Järvinen et al., 2002;Kovanen, 2002;Purslow, 2020), this term implies that it exists only within (i.e. 'intra-') muscles, and therefore excludes the epimysium surrounding the musclesthat is, the muscle fascia. Thus, in the present review, the term 'muscle connective tissue' will refer collectively to muscle endomysium, perimysium, and epimysium.
Markers of ECM breakdown or synthesis have been commonly used for research on bone, tendon, and cartilage as well as in conditions such as rheumatoid arthritis or hepatic fibrosis, and these can include ECM proteins, ECMmodifying enzymes, and molecules that alter ECM-cell interactions. However, despite the increasing interest in examining biochemical changes to muscle ECM components Biological Reviews 98 (2023)  to assess ECM status, there is little information on, and recommendations for, the use of ECM remodelling biomarkers in tissues and biofluids. The present review summarises the responses of key ECM-related molecules to exercise, disuse, and aging, and provides recommendations for the use of these molecules as biomarkers to assess underlying ECM remodelling processes.
Strengths and weaknesses exist in the numerous methods currently available for assessing ECM structural changes in terms of their ease of use, the number of analyses required, invasiveness, tissue specificity, and other factors. Therefore, an understanding of the limitations of each method is important when making appropriate methodological choices to ensure correct result interpretation. For example, the staining of muscle biopsy samples provides a visual representation of collagen and other specific proteins, which often allows for accurate protein expression assessment. However, biopsy samples only provide site-specific information and previous studies have suggested that changes in the physiological state in response to aging, loading (unloading), etc., can vary between ECM components associated with different slow and fast-twitch myofibre types (e.g. predominantly slow twitch type I, or fast-twitch type II myofibres) (Takala et al., 1983;Kovanen et al., 1988;Koskinen et al., 2001a). Moreover, human muscle biopsies are invasive and only a small portion of tissue is sampled, so human ECM changes are often assessed through lessinvasive methods involving the quantification of relevant blood-and urine-based biomarkers. Thus, studies relating to ECM molecule alterations in biofluids (e.g. blood serum, plasma, urine, sweat, etc.) will be reviewed to summarise the current understanding of underlying ECM processes when direct muscle sample assessment is unavailable.
Further complications can arise from the incorrect interpretation of biofluid marker data. For example, an increase in the estimated quantity of a biomarker in the blood does not always reflect the same physiological phenomenon as increases in muscle; this is illustrated by increased blood hydroxyproline (Hyp) predominately indicating collagen breakdown (Murguia et al., 1988) while increased muscle Hyp indicates increased collagen content (Zimmerman et al., 1993). To clarify tissue-specific changes, biomarker changes in muscle and biofluids will be reported separately.  Table 1. Types, groups, and functions of collagen. The different types of collagens can be divided into groups according to their composition, structure, and location. Fibrillar collagens contain one major triple-helical domain (Ricard-Blum, 2011). Collagens belonging to multiplexins have multiple triple helical domains separated by unique non-triple helical regions that may exist to give flexibility to the molecule, and as they carry glycosaminoglycan chains they can also be considered proteoglycans (Muragaki et al., 1994). Transmembrane collagens have dual functions as cell surface receptors and matrix molecules. They are responsible for interactions between the epithelium and the mesenchyme during morphogenesis, cell adhesion in epithelium and nerves, and protection against microbial agents (Franzke et al., 2003). Fibril-associated collagens (FACIT) do not form fibrils but associate with the surface of collagen fibrils and aid in their stabilisation and integrity. Network-forming collagens create tetrameric or hexagonal networks that are the most important structural components of basement membranes and enable the integration of molecules. Beaded collagens interweave with other collagen fibrils and contain a characteristic beaded filament structure. ECM, extracellular matrix; FACIT, fibril-associated collagen; RGD, arginine-glycine-aspartic acid.

Type
Group Function I Fibril-forming Accounts for 90% of all the collagen in mammals; major constituent of the skin, tendon, bone, ligaments, and muscular connective tissues (Wu et al., 2010;Ricard-Blum, 2011). II Fibril-forming Composes cartilage (Wu et al., 1992(Wu et al., , 2009Wu & Eyre, 1995), creating an intricate network along with collagens IX and XI. III Fibril-forming Usually acts together with collagen I as co-polymers, forming reticular fibres (Keene et al., 1987). Creates a looser meshwork of fibres compared to collagen I which gives compliance to the tissue (Kovanen, 2002). IV Networkforming Located in the basement membrane, and unlike collagens I, II, and III, forms networks instead of fibres (Bailey et al., 1984), creating a stable scaffold that integrates different ECM molecules (Kühn, 1995). V Fibril-forming Particularly abundant in the cornea (Bruckner, 2010), but is also widely distributed in different tissues (Martin et al., 1985). Its functions are still poorly understood, however, it regulates collagen fibril assembly initiation (Wenstrup et al., 2004). COL5A1-knockout mice synthesise normal amounts of collagen I but fail to create fibrils, and the animals die at the onset of organogenesis (Leeming & Karsdal, 2016). VI Beaded filament Found in a wide range of tissues with many different roles, including cytoprotection, regulation of autophagy, cell differentiation and apoptosis, protection against oxidative damage, and tumour growth (Cescon et al., 2015). Additionally, it is responsible for myofibre integrity, and deficiency of this collagen results in Bethlem myopathy (Bonaldo et al., 1998). VII Anchoring Involved in ECM function, binding collagen I and III, and providing stability to ECM structures of the basement and interstitial membranes (Sakai et al., 1986). Major component of the anchoring fibrils connecting the dermis and the epidermis of human skin (Villone et al., 2008). VIII Networkforming Synthesised by endothelial cells and is the major component of Descemet's membrane, forming hexagonal networks (Illidge et al., 1998). Found in the brain, lung, liver, heart, and muscles, and around chondrocytes in the cartilage (Kittelberger et al., 1990). Involved in endothelial organisation and differentiation (Iruela-Arispe et al., 1991), smooth muscle cell migration (Adiguzel et al., 2013), cellular proliferation, and angiogenesis (Paulus et al., 1992). IX FACIT Contributes to the stability of the network found in the cartilage, where it is cross-linked to the surface of collagen II fibrils (Eyre et al., 1987). X Networkforming Forms networks and is usually expressed in hypertrophic chondrocytes in cartilage, where it typically constitutes about 1% of total collagen (Eyre, 1991). Involved in calcification and distribution of proteoglycans and matrix vesicles within the growth plate (Kwan et al., 1997). XI Fibril-forming Crucial for fibrillogenesis in cartilage controlling the diameter and spacing of collagen II fibrils (Kadler et al., 2008). XII FACIT Found in the skeletal muscle. Interacts with ECM proteins such as decorin, tenascin, and fibromodulin, and also thought to stabilise the structure of collagen fibrils during development and rapid remodelling (Tzortzaki et al., 2006;Chiquet et al., 2014). XIII Transmembrane A non-fibrillar transmembrane collagen involved in various biological maturation and differentiation processes. Has multiple functions in the myotendinous junctions of the skeletal muscle Sund et al., 2001;Franzke et al., 2003;Määttä et al., 2006). Assists in muscle fibre and basement membrane integrity, as lack of it causes abnormal myofibres, disorganised myofilaments, Z-line streaming, and increased susceptibility to exercise-induced muscle damage (Kvist et al., 2001). Its mutations could explain myopathies with unidentified causes (Kovanen, 2002). XIV FACIT Found in tendon, cornea, skin, and articular cartilage; involved in fibrillogenesis (Ansorge et al., 2009). Found in areas where high mechanical stress is experienced, indicating the involvement of this collagen in tissue biomechanical function (Niyibizi et al., 1995). XV Multiplexins Located predominantly in vascular, mesenchymal, neuronal, and epithelial basement membrane zones (Myers et al., 1996). To date, there is no known human disease caused by mutations in the COL15A1 gene, but mice deficient in it display impaired cardiac function, skeletal myopathy, and a poorly organised fibrillar collagen matrix (Rasi et al., 2010).
(Continues on next page)   (Lai & Chu, 1996). As a FACIT collagen, it associates with other types of collagen and other non-collagen molecules such as fibrillin and integrin. Involved in both structural integrity and in integrin-mediated signalling (Ratzinger et al., 2010). XVII Transmembrane Binds to different molecules both intracellularly (such as integrin-β4) and extracellularly (such as laminin-332), indicating a role as an adhesion-stabilising molecule adhering epithelial cells to the ECM ( Van den Bergh et al., 2011). XVIII Multiplexins Present in three isoforms: short, middle, and long. Being part of the multiplexin family, it contains noncollagenous domains. The short isoform is found in epithelial and vascular basement membranes, the long isoform is highly expressed in liver, while the intermediate isoform is not as widely expressed (Saarela et al., 1998;Suzuki et al., 2002). All three isoforms are involved in basement membranes, where they interact with perlecan, laminin, nidogen, and fibulin (Kalluri, 2003). The carboxyl-terminal fragment of this type of collagen is called endostatin and it inhibits angiogenesis and tumour growth in murine hemangioendothelioma cells (O'Reilly et al., 1997). This suggests that this collagen has both structural properties when intact and signalling potential when degraded. XIX FACIT Found in vascular, mesenchymal, neuronal, and epithelial basement membrane zones of human skeletal muscle, prostate, spleen, liver, placenta, kidney, skin, and colon (Myers et al., 1997). Involved in muscle development and differentiation (Sumiyoshi et al., 2004). XX FACIT Has similarities to collagens XII and XIV and is therefore expected to interact with other fibrillar collagens (Ricard-Blum, 2011). Has been detected in the corneal epithelium and in lesser amounts in sternal cartilage, embryonic skin, and tendon (Koch et al., 2001). XXI FACIT Located in the fetal and adult heart, stomach, placenta, jejunum, kidney, skeletal muscle, pancreas, lung, and lymph nodes (Fitzgerald & Bateman, 2001). As those tissues are rich in collagen I, collagen XXI may colocalise with collagen I, similar to collagens XII and XIV which also co-localise with collagen I (Karsdal, 2016). XXII FACIT Localised in the myotendinous junctions of the heart and skeletal muscle, skin, and cartilage. In myotendinous junctions, it interacts with integrins to assist in stabilising the integrity of the ECM (Koch et al., 2004). Also involved in force transmission and resistance to rupture of the skeletal muscle (Charvet et al., 2013). XXIII Transmembrane Expressed in normal human heart and retina as well as in metastatic prostate cancer cells (Banyard et al., 2003). Contains arginine-glycine-aspartic acid (RGD) motifs that allow binding to different integrin types. Interactions with the RGD motifs could indicate a role for this collagen as a signalling molecule that assists in integrin-mediated signalling, or interactions in a soluble form following cleavage. XXIV Fibril-forming Expressed during development and the formation of the skeleton in mouse embryos, as well as in the periosteum and trabecular bone. Involved in collagen I fibrillogenesis. Gene expression of COL24Α1 is detectable in the early stages of development and increases concurrently with osteoblast differentiation (Wang et al., 2012). XXV Transmembrane Membrane-bound collagen that includes both intracellular and extracellular domains. Required for the fusion of myoblasts into myofibres (Gonçalves et al., 2019). Found in brain and spinal cord neurons as well as in muscles (Tanaka et al., 2014). COL25A1-knockout mice die immediately after birth due to respiratory failure (Tanaka et al., 2014), however, overexpression of collagen XXV promotes amyloid plaque formation which contributes to the development of Alzheimer's disease (Tong et al., 2010). XXVI Beaded filament Expressed in the testes and ovaries of adult tissues and in low levels in the kidney. During early development, testicular and ovarian collagen XXVI is highly expressed in myoid and pre-theca cells, respectively. Speculated to have a functional role in the adult reproductive organs as well as in their early development (Sato et al., 2002). XXVII Fibril-forming Its triple-helix domain is shorter than other fibrillar collagens. Additionally, it lacks a minor triple-helical domain and an N-terminal telopeptide-like region, both of which are fibrillar collagen characteristics (Plumb et al., 2007;Hjorten et al., 2007). Involved in developmental phases, assisting in cartilage calcification during skeletogenesis and in the developing lungs, ears, and colon. XVIII Beaded filament Structurally resembles collagen VI. Mainly located in peripheral nerves and surrounds all non-myelinating glial cells (except collagen II terminal Schwann cells in the hairy skin) and dorsal root ganglia. Also found in the nodes of Ranvier and is a component of the peripheral nervous system nodal gap. Predominantly synthesised during development, however, it persists in the ECM for an extended period after synthesis (Veit et al., 2006). XXIX Beaded filament Belongs to the class of collagens containing von Willebrand factor type A domains. The highest expression of its gene (COL29A1) is in the skin, followed by the lung, colon, small intestine, and testis (Söderhäll et al., 2007). Involved in epidermal integrity and function, as its lack of expression is associated with atopic dermatitis. Not classified as a new type, as the associated COL29A1 gene was shown to be identical to the COL6A5 gene (Gara et al., 2008;Ricard-Blum, 2011 The overall purpose of this review is to provide information relating to acute and chronic changes in tissue and biofluid ECM molecules following altered physical loading states and aging. Additionally, the sometimes-unrecognised limitations of different assessment methods are described, and recommendations for future research are provided.

II. COLLAGENS
Collagen is the main ECM structural protein, the main fibrilcreating molecule, and the most abundant protein in vertebrates, constituting 30% of the total protein mass (Ricard-Blum, 2011). Collagen is a glycoprotein trimer composed of three left-handed polypeptide chains of 1000 amino acids twisted into a triple, right-handed helical structure. This process is made possible by specific repeats in the amino acid collagen sequence, with glycine-X-hydroxyproline and glycine-proline-X being the most common, where X represents any amino acid other than glycine, proline, or Hyp (Ricard-Blum, 2011). Approximately 33% of the residues are glycine, 12% proline, and 13% Hyp (Hall & Reed, 1957); as glycine is the smallest amino acid, a sharp twisting of the helix is possible (Kovanen, 1989). So far, 28 types of collagen have been discovered (Ricard-Blum, 2011); different collagen types serve different functions while a wide variety exists within each tissue (Table 1), and even within individual collagen fibres. Additionally, there are differences in co-and post-translational modifications between collagen types (Kivirikko & Myllylä, 1979). Understanding how different collagen types respond to physical activity and inactivity is of paramount importance as collagens provide essential connective tissue integrity and forcetransmission properties. Mutations in collagens and other ECM proteins result in a wide range of diseases, with several collagens (e.g. collagen VI) being of special interest for skeletal muscles (Voermans et al., 2008;Ricard-Blum, 2011;Lamandé & Bateman, 2020).
Collagen molecules circulate in the blood, and exercise affects their levels. For example, increased serum collagen I concentration has been found at 3 and 4 days after 70 maximal eccentric forearm flexions and extensions in men and after 4 days in women, while equivalent concentric exercise produced no response (Saxton & Donnelly, 1994). Similarly, no differences in serum collagen I concentrations were detected after 50 maximal concentric knee extensions but increases (49%) developed 1, 7, and 9 days after maximal eccentric exercise (Brown, Day & Donnelly, 1999). The outcomes of these studies are consistent with our understanding of ECM remodelling following muscle injury (Saxton & Donnelly, 1994;Brown et al., 1999). Although elevated collagen I concentrations were attributed to muscle tissue disruption caused by eccentric exercise (Saxton & Donnelly, 1994), the muscle cannot be confirmed as the source of the collagen as it is omnipresent throughout the body and may thus have, at least partly, originated from the tendons or bone. Nonetheless, after low-or non-impact exercise that does not substantially stimulate bone turnover, serum collagen I concentration can quantify exercise-induced stimulus to muscle ECM.
Other biofluid collagen types may be more specific to the skeletal muscle ECM following exercise. For example, increased plasma collagen IV concentration manifested in both femoral veins and arteries during 27 min of bicycling at 50-65% of maximum oxygen consumption rate (VO 2max ) in men (Rullman et al., 2013). Levels were significantly different in venous (1.11-fold) versus arterial (1.14-fold) samples, possibly because of the molecules being released into venous blood, although by 120 min post-exercise they were equally elevated. Collagen IV is largely contained within the basal lamina (i.e. basement membrane) and it may be assumed with more certainty that increases during and up to 120 min after bicycling result from intramuscular rather than skeletal collagen pools, especially if the samples are drawn from blood vessels adjacent to the exercising muscles (Rullman et al., 2013). Bicycling is a low-impact exercise so both muscle and vascular tissues (and not the bone) would be subjected to the greatest stress. Based on these findings as well as the known role of collagen IV, collagen IV blood concentrations might quantify exercise-induced muscle collagen turnover more accurately than collagen I. Nonetheless, further studies are required explicitly to determine the relationship between biofluid collagen IV and other indirect markers of muscle and connective tissue damage after exercise of different modalities and intensities.
Direct muscle-sample staining can be used to estimate the content of ECM molecules such as collagen (Fig. 2). Muscledamaging exercise can trigger increases in collagen concentration after several days (i.e. during recovery). For example, in humans, collagen I and III expressions increased 30 days after electrical stimulation exercise in gastrocnemius in men ( Fig. 3; earlier time points were not analysed) (Mackey et al., 2011) while vastus lateralis (VL) collagen IV expression increased (26%) 22 days after 100 maximal eccentric knee extensions (Mackey et al., 2004) (no increase was detected at 4 days, and no other time points were assessed). In rats, quadriceps femoris collagen IV content increased 7 days after 130 min of downhill treadmill running (−13.5 , 17 m/min) in type II myofibres without changes at 6 h or 4 days (Koskinen et al., 2001a). In the same study, no difference existed in collagen IV content in quadriceps femoris type I myofibres or soleus although a decrease was detected 24-48 h later. Speculatively, the proteolytic action of matrix metallopeptidases (MMPs), which can degrade collagens and other ECM components, could underpin this initial reduction, clearing away damaged structures before subsequent collagen production (see Section V.2.d). This hypothesis is supported by decreased muscle prolyl 4-hydroxylase (PH) activity, a collagen-synthesising molecule, shortly after exercise (Koskinen et al., 2002) (see Section V.2.a). Based on the findings of animal studies (Tables 2 and 3), the time course of collagen production following damaging exercise suggests that robust increases are detectable after at least  (Koskinen et al., 2001a), consistent with the time course of changes in collagen messenger RNA (mRNA) levels (described later in this section). However, the lack of data quantifying changes in human muscle collagen concentrations shortly after exercise prevents clear conclusions from being drawn.
Muscle collagen concentrations increase after long-term endurance training in humans and animals, which is consistent with observed increases in the collagen triple helixstabilising amino acid, Hyp (Kovanen et al., 1988;Kovanen, 1989;Mackey, Donnelly & Roper, 2005;Makhnovskii et al., 2020). Human muscle collagen I concentrations were unchanged after low-impact endurance or short-term (e.g. 4 weeks) resistance training, while collagen XIV increased after 4 weeks of resistance training (Mackey et al., 2005;Jakobsen et al., 2017). This indicates that distinct collagen types respond differently to exercise, most likely because of differences in their functional roles. For example, endomysial staining of collagen XIV is more intense closer to the myotendinous junctions than mid-muscle ( Jakobsen et al., 2017), which might suggest an important role for collagen XIV in the tissue's biomechanical function and force transmission. In addition to increased concentrations, collagen fibrils undergo hypertrophy and macromolecular reorganisation after several weeks of loading (Michna, 1984;Coutinho et al., 2006). The ECM reorganisation in particular may involve force vector-specific collagen fibril alignment and denser molecular packing (Oakes, Parker & Norman, 1982;Michna, 1984;Vilarta & Vidal, 1989;Coutinho et al., 2006), or a change in its quantity or molecular cross-linking profile (type of cross-links) (Avery & Bailey, 2005) (see Section II.2). This was supported by rat muscles demonstrating higher tensile strength under passive stretch after 4 weeks of treadmill training, despite the rats exhibiting no changes in soleus or rectus femoris (RF) collagen concentrations (measured as Hyp concentration) versus controls (Kovanen, Suominen & Heikkinen, 1984b). Therefore, future research investigating the long-term effects of exercise training on collagen expression should focus not only on muscle concentration or content but also on qualitative and architectural changes. Immobilisation (i.e. disuse) strongly influences muscle collagen expression and connective tissue area. Increased muscle connective tissue area was observed in human soleus after 60 days of bed rest (Thot et al., 2021), and in animals after immobilisation at longer (Jozsa et al., 1988;Lapier et al., 1995) and shorter lengths (J ozsa et al., 1990). Moreover, increased animal muscle collagen (Kovanen et al., 1988;Kovanen, 1989) and Hyp (Kovanen, Suominen & Peltonen, 1987;Kovanen, 1989;Gosselin et al., 1998;Ducomps et al., 2003) concentrations occur during aging, however without accompanying increased PH and galactosylhydroxylysyl glucosyltransferase (GGT) activities (Kovanen, 1989;Kovanen & Suominen, 1989). This contradiction could be caused by a decreased collagen turnover rate during aging (Kovanen & Suominen, 1989), which is supported by decreases in collagen mRNA levels (Han et al., 1999b;Ahtikoski et al., 2001Ahtikoski et al., , 2003. During 7 days of animal muscle immobilisation, increased levels of collagen I and III (Salonen et al., 1985) and decreased collagen IV Table 2. Acute muscle biomarkers: time course of changes following eccentric contractions in animals. Animal models of acute responses to eccentric contractions included: 240 forced tetanic eccentric contractions (Koskinen et al., 2002), 130 min of downhill (−13.5 ) treadmill running at 17 m/min (Han et al., 1999a), 130 min of downhill (−13.5 ) treadmill running at 17 m/min (Koskinen et al., 2001a), 5 min of downhill (−16%) treadmill running at 18 m/min (Graham et al., 2015), 30 min of downhill (−20 ) treadmill running at 15 m/min (Boppart et al., 2006), 40 min of downhill (−10 ) treadmill running at 17 m/min (Mahmassani et al., 2017), 40 min of downhill (−16 ) treadmill running at 20 m/min (Nourshahi et al., 2013). ": significant increase, #: significant decrease, $: not different to control values (significance at least P < 0.05).

Markers
2 h 6 h 12 h 24 h 2 days 4 days 7 days Enzymes PH #TA "TA, "RFI, "SOL "TA, "RFI "TA, $RFI, "SOL MMPs proMMP-2 "SOL, "TA "RFI, "SOL, "TA "RFI, "RFII, "SOL, "TA proMMP-9 "TA "TA MMP-2 "RFI "RFI, "TA "RFI, "TA MMP-9 "TA TIMPs TIMP-1 "RFI $RFI TIMP-2 #RFI "TA "RFI, "TA "RFI, "TA  (Ahtikoski et al., 2003) were observed, without changes in collagen V levels (Salonen et al., 1985). In humans, increased collagen I levels were observed in the peri-and endomysial areas of soleus after 60 days of bed rest (Schoenrock et al., 2018), and in another study, human vastus lateralis collagen IV levels increased after 48 h but collagen III levels decreased (Urso et al., 2006). As collagen IV functions as a cell membrane scaffold, the immobilisation-induced collagen IV decrease might be expected to reduce cell membrane structural integrity. However, surprisingly, there was a decreased susceptibility of immobilised muscle to damage (measured as a smaller reduction in extensor digitorum longus (EDL) tetanic force after electrically stimulated foot flexions) after 3 weeks of rat EDL immobilisation, albeit the muscle forcegenerating capacity, when normalised to cross-sectional area, was unaffected (Lapier et al., 1995). Moreover, the slower collagen turnover compared to other myofibre proteins causes the connective tissue area to appear enlarged (Salonen et al., 1985;Jozsa et al., 1988;J ozsa et al., 1990;Lapier et al., 1995) (see discussion of Hyp responses to immobilisation in Section II.3). Nevertheless, daily stretching can reorganise collagen molecules and partially reverse the decrease in myofibre cross-sectional area despite immobilisation causing collagen fibre disorganisation in rats (Järvinen et al., 2002;Coutinho et al., 2006). These effects are dependent upon stretching frequency (and presumably the tension generated), as daily stretching was more effective than stretching every other day, at least in rat muscles (Coutinho et al., 2006). This effect could be mediated by transmembrane mechanosensors (see Section III.5), which regulate muscle cell gene expression depending on stretching intensity and frequency. Intracellular protein synthesis (including that of enzymes such as PH) can be explored by the quantification of specific mRNA levels. This indicates the gene transcription rate prior to protein synthesis and is typically reported as the foldchange from the pre-intervention level. Exercise consistently increases collagen mRNA levels in rat muscle and tendon, but human responses are variable. In animals, quadriceps femoris collagen I mRNA levels tend to remain unchanged for the first 6-12 h following eccentric exercise (Han et al., 1999a;Koskinen et al., 2001a), after which they are elevated for up to 7 (Han et al., 1999a;Koskinen et al., 2001a) or even 14 days (10-14-fold) (Han et al., 1999a). This coincides with temporal changes in muscle PH activity, as described previously. Acute electrically stimulated eccentric (5.5-fold) and isometric (3.7-fold) exercise can increase muscle collagen mRNA levels more than concentric exercise (2-fold) in animals , but this is unexplored in Table 3. Acute synthesis of muscle biomarkers: time course of changes following eccentric contractions in animals. Animal models of acute responses to eccentric contractions included: 13 sets of 10 electrically induced eccentric gastrocnemius actions spread over 4 days , 130 min of downhill (−13.5 ) treadmill running at 17 m/min (Koskinen et al., 2001a), 240 forced tetanic eccentric contractions (Koskinen et al., 2002), 130 min of downhill (−13.5 ) treadmill running at 17 m/min (Han et al., 1999a), 30 min of downhill (−20 ) treadmill running at 17 m/min (Boppart et al., 2008). ": significant increase, #: significant decrease, $: not different to control values (significance at least P < 0.05).
Isotope-labelled amino acids can be used to measure protein synthesis by venous infusion and subsequent tissue tracing. The exercise-induced responses detected with this technique align well with mRNA expression responses. In fact, collagen mRNA expression was unchanged during the first 6-12 h after eccentric exercise in rats (Han et al., 1999a;Koskinen et al., 2001a), while no change or decreases were observed up to 24 h after eccentric exercise in human VL muscle (Hyldahl et al., 2015). However, assessment of collagen fractional synthesis rate (FSR; the rate that a radio-labelled amino acid precursor is incorporated into proteins) in the VL muscle of men increased (28%) as early as 4.5-6 h after resistance exercise (Moore et al., 2005;Miller et al., 2005), peaking at 24 h (68%) in VL muscle and patellar tendons (Miller et al., 2005). Although exercise types differ between studies (i.e. conventional concentric/eccentric versus eccentric-only resistance exercise), these results suggest a difference between the two techniques. Unfortunately, like mRNA, limited data exist documenting the effects of endurance exercise on muscle collagen FSR. While mRNA levels can indicate longer-term trends in protein production, the presence of mRNAs themselves does not necessarily determine the final collagen amount produced since this is dependent upon translation efficiency and other factors such as genetic differences (Cenik et al., 2015). Indeed, the variation in mRNA levels accounts for only 40% of the final variation in protein concentration (Vogel & Marcotte, 2012). Therefore, collagen FSR is likely more valid in quantifying muscle and tendon collagen synthesis after exercise than collagen mRNA quantification. However, the labelled isotope method is invasive, as it requires muscle biopsy samples. Nonetheless, collagen FSR may prove to be a useful tool for 'calibrating' other less-invasive methods and help to determine similarities in collagen synthesis between animals and humans.
Interestingly, collagen FSR was stable across menstrual cycle phases after acute single-leg exercise as well as at rest (Miller et al., 2006). However, greater increases and longer synthesis periods developed in men versus women (Miller et al., 2007). The authors concluded that higher incidences of musculoskeletal injuries during physical activity in women Biological Reviews 98 (2023)  than in men could be attributable to lower rates of tissue repair following exercise, indicated by a lower tendon collagen FSR (Miller et al., 2006). Therefore, researchers should be mindful of biological sex when examining exercise-or immobilisation-induced collagen synthesis changes.
Of potential clinical importance is that immobilisation decreases human muscle and tendon collagen FSR (de Boer et al., 2007), a finding consistent with changes in mRNA levels (Han et al., 1999b;Ahtikoski et al., 2001). Aging also decreases collagen FSR, as it was found to be lower (10-fold) in older (24 months) rat gastrocnemius versus young (1 month) (Mays et al., 1991), but this remains unexamined in humans. These findings also support the hypothesis of increased muscle collagen and Hyp concentrations during aging and immobilisation resulting from lower collagen and Hyp turnovers versus other proteins, rather than increased collagen or Hyp synthesis per se (see discussion of Hyp responses to immobilisation).
In conclusion, collagen can be insightful when studied in response to exercise, disuse, and aging status in biofluids and muscle, however, care must be given to methods and collagen types assessed. Biofluid collagen IV levels might be more specific to exercising muscles than collagen I, especially when assessed close to the muscle of interest. Regarding longterm exercise training or disuse, it appears that it takes at least 7 days to detect changes reliably in animals, but information about human muscle responses is limited. Valuable information can be obtained from qualitative and architectural changes in collagen in addition to quantitative collagen (content) changes. Assessing synthesis requires attention to the methods used, as collagen mRNA levels appear to respond differently than FSR.
(1) Collagen peptides Procollagen, the precursor molecule of collagen, contains N-and C-terminal extensions removed by specific peptidases during the conversion of procollagen to tropocollagen. These extensions are referred to as the N-and C-terminal propeptides of procollagen (PINP and PICP, respectively, for collagen I). Immunoassays for the cross-linked carboxyterminal telopeptide region of type I collagen (ICTP) and PICP allow for assessment of collagen I degradation and synthesis, respectively. Serum collagen propeptide and telopeptide concentration analyses have been widely used to quantify bone turnover and synthesis rates and are also used to examine tendon collagen turnover.
Prolonged low-impact (intensity) activities, such as running, increase collagen synthesis in humans (indicated by serum PICP increases), as detected 30 min after exercise and which peaked 3 days after (1.2-fold) (Salvesen, Piehl-Aulin & Ljunghall, 1994;Thorsen et al., 1997;Langberg et al., 2000). However, it is not possible to specify the tissue of origin from serum samples, and bone turnover may significantly contribute to the release of PICP. Microdialysis is an alternative method with greater tissue specificity for assessing collagen peptide concentration, as it enables the analysis of biochemical substances in the extracellular space of various tissues via the insertion of a small, capillary-sized catheter, allowing for continuous monitoring up to several days after needle insertion (Kristoffersson et al., 1995;Langberg et al., 1999Langberg et al., , 2007Langberg, Rosendal & Kjaer, 2001;Heinemeier et al., 2003;Crameri et al., 2004;Miller et al., 2005). Using microdialysis of muscle and tendon, increases in human PICP of 5-and PINP of 4-fold, with a peak at 8 days post-exercise, were detected after eccentric exercise (one-leg drop-downs from 45 cm platforms) in VL muscle (Crameri et al., 2004) as well as chronic increases in Achilles tendon collagen synthesis . However, differences may exist when assessing collagen peptides in serum versus dialysate (the sample from microdialysis). For example, three studies assessing PICP and ICTP using both radioimmunoassay in venous samples and microdialysis in the Achilles peritendinous region found different acute and long-term exercise responses at each time point (Langberg et al., 1999(Langberg et al., , 2001Heinemeier et al., 2003), possibly resulting from the earlier collagen peptide detection in the interstitial and peritendinous regions than in venous blood. This suggests that the assessment of biomarkers in blood may not accurately reflect the temporal changes in the tissues of origin. Microdialysis can be useful for detecting changes in PICP and PINP concentrations, and conclusions from these assessments are tissue specific to muscles and tendons. The specificity of microdialysis can be further improved if participants perform single-limb (or single muscle group) exercise, while the contralateral limb is assessed as a non-exercise control since the trauma caused by microdialysis catheters can significantly affect PICP values . Future studies should assess the time course of changes of the three molecules in muscle dialysate, particularly changes immediately after different types of exercise performed with different programming parameters, to provide a better understanding of the effects of exercise on ECM remodelling.
Immobilisation stimulates ECM remodelling, as indicated by collagen and procollagen propeptide responses. Increased PINP and ICTP concentrations were detected in both blood and dialysate during 7 weeks of Achilles tendon immobilisation in 12 humans with fractures (Christensen et al., 2008). Although there was possible interference from the bone fractures, PINP (12-fold) and ICTP (5-fold) concentrations increased in the immobilised versus the control leg after 7 weeks of immobilisation. During remobilisation, levels of both peptides decreased in the immobilised leg without differences in the control leg. This indicates that the continuous ECM remodelling process during immobilisation consists of both increased collagen degradation and synthesis, rather than exclusively one or the other. However, this hypothesis requires explicit testing as the influence of bone fractures in these patients cannot be determined. Nevertheless, PINP and ICTP concentrations are promising blood-based biomarkers for the assessment of long-term collagen synthesis. Future studies should quantify both molecules and explore the potential usefulness of their ratio for quantifying ECM remodelling (i.e. PINP/ ICTP concentrations). Extracellular matrix biomarkers related to physical activity and aging Ν-terminal propeptide of procollagen type III collagen (P3NP) is a peptide cleaved from procollagen type III in the synthesis of tropocollagen type III and then released in the blood, allowing for the quantification of collagen III synthesis. Unfortunately, P3NP concentration changes poorly indicate exercise-induced ECM damage and repair compared to other collagen peptides. Most studies have not found changes in human serum concentrations even after 5 days of monitoring following 50 maximal concentric contractions or 24 h after relay skiing (Takala et al., 1989;Virtanen et al., 1993). No associations between serum P3NP concentration and muscle damage markers (serum creatine kinase and lactate dehydrogenase) existed after a competitive 24-h run despite increased P3NP concentrations (36%, 2 days post-exercise) . This led Takala et al. (1986) to attribute increased GGT activity and P3NP concentration to exerciseinduced injury of collagen-synthesising cells in the connective tissue rather than to exercise-induced myofibre damage. Interestingly, 6 weeks of biweekly resistance exercise in elderly men and women increased serum P3NP concentration (8%), which did not reach statistical significance, although the changes were significantly associated with increases in muscle mass (Fragala et al., 2014). Collagen III is co-located with collagen I in connective tissues but in much smaller quantities. Therefore, quantifying P3NP concentration may be more difficult than collagen I peptides and is thus not recommended.
Some investigators have quantified muscle P3NP expression following exercise. Increased PINP and P3NP expressions in both VL muscle endomysial and perimysial spaces developed 2-8 days following 50 single-leg drop landings from a 45-cm box and 160 maximal isokinetic eccentric knee extensions in men (Crameri et al., 2004). P3NP was more immunoreactive from days 2 to 8 post-exercise, however, the authors did not report the data. These findings are supported by increased P3NP concentration (2.4-fold) in the VL muscle of men 4-7 days after 300 maximal voluntary eccentric knee extensions (Raastad et al., 2010). P3NP concentration appears more reliable when measured in muscle instead of biofluids, however, there are insufficient data to conclude about its usefulness in quantifying collagen synthesis.
(2) Collagen cross-links Collagen molecules are subjected to great mechanical stress, especially where they bind to (interdigitate with) other tissues, such as at the myotendinous junctions. The collagen triple helix by itself is unable to withstand such stretching forces, and to strengthen these structures, spontaneous molecular cross-links are formed between the aldehydes created by lysyl oxidase (LO), producing immature, divalent cross-links, which are possible to quantify in their reduced forms (hydroxylysinorleucine and dihydroxylysinorleucine) (Knott & Bailey, 1998). These cross-links react further to form mature, trivalent cross-links such as lysylpyridinoline (LPyr), hydroxylysylpyridinoline (HPyr, also known as pyridinoline), deoxypyridinoline (DPyr) and pyrroles (McNerny, Gardinier & Kohn, 2015). They bind different tropocollagen molecules and elastin together, increasing fibre tensile strength while preventing molecular sliding under load. Cross-links are released into the blood during collagen remodelling, and some cross-links have been traditionally used for monitoring bone turnover (Welsh et al., 1997;Woitge et al., 1998).
With regard to exercise, trends towards statistically significant increases (P = 0.058) in urinary HPyr levels were detected after 50 maximal unilateral eccentric knee extensions, peaking 2 days after exercise (54.5 μmol per mol of creatine, versus a baseline of 33.6 μmol/mol) (Brown et al., 1997). However, low-intensity exercise such as brisk treadmill walking increased urinary HPyr and DPyr concentrations (38.7% and 42.3%, respectively, versus baseline and standardised to creatinine) after 1 day (Welsh et al., 1997). Therefore, blood or urine HPyr could potentially be used to quantify collagen degradation, although more research is required to understand its responses to exercise. Existing assays can quantify HPyr in sweat (Sarno et al., 2001), however, as with collagen I or Hyp, these cross-links may derive from several tissue types and have so far been used primarily for assessing bone turnover. These markers must be used with caution when measured in biofluids as their connective tissue origin cannot be accurately determined.
Regarding long-term exercise training, serum levels of HPyr and DPyr were reduced (10-15%) in humans after 8 weeks (3 per week) of aerobic exercise (40-60 min of running) but increased (22%) in the anaerobic exercise group (five sprint runs of 80-300 m) (Woitge et al., 1998). Although HPyr and DPyr were used as bone resorption markers, the repetition of high-intensity contractions together with the impact from sprinting may have caused increased collagen turnover in both muscle (including tendon) and bone. Therefore, to examine muscle-specific collagen remodelling, muscle biopsy samples may be required.
Collagens are long-lasting molecular structures modified by glycation. Glycation increases with age as it is caused by random reactions of glucose with collagen lysines. This process forms many advanced glycation endproducts (AGEs) which act as cross-links, increasing insolubility and inducing non-optimal stiffness, resulting in collagen fibril brittleness in senescent tissues. Glycation occurs at high rates in diabetes mellitus, which accelerates aging (Avery & Bailey, 2005). Therefore, studying collagen cross-link concentration and connective tissue profiling (i.e. cross-link types and AGE concentration) in different diseases is clinically important and could result in new and more effective treatments for improving the elasticity and tensile strength of muscles and tendons.
Higher cross-linking rates may therefore help to counteract slower collagen synthesis rates in aging by creating stiffer Biological Reviews 98 (2023)  and more durable collagen molecules (Reddy, 2004), whereas increased collagen stiffness reduces tissue elasticity and could create more brittle bones and injury-prone soft tissues (Saito, Fujii & Marumo, 2006). AGE accumulation also decreases collagen turnover (DeGroot et al., 2001), thereby creating a feedback loop accelerating non-optimal cross-linking. However, most studies have investigated skin and bone collagen (i.e. non-muscular) cross-link profiles during aging. The surgical removal of the excess connective tissue in conditions such as arthrofibrosis (fibrosis of joints) (Usher et al., 2019), may allow direct analyses of the cross-link profiles in humans, at least in fibrotic conditions. Exercise increases collagen turnover and therefore assists in the renewal of old collagen, cross-linked by glycation or byproducts of lipid oxidation. For example, treadmill training in young chickens triggered large decreases in tendon and meniscal HPyr content (Pedrini-Mille et al., 1988;Curwin, Vailas & Wood, 1988) versus controls, suggesting that exercise could improve the cross-link profile in these tissues, possibly through the increased breakdown of old collagen. In addition to collagen renewal, exercise appears essential for maintaining an optimal collagen cross-link profile in senescence, even when started in middle age. For example, a positive correlation manifested between muscle stiffness and HPyr concentration, with old sedentary rats having higher stiffness than their exercise-trained (10 weeks of treadmill running) age-matched counterparts (Gosselin et al., 1998). However, there was no difference between old-trained, young-trained, and young-sedentary groups, suggesting that exercise-induced reductions in collagen cross-linking attenuate the age-induced increased muscle stiffness, and this is further supported by other studies (Zimmerman et al., 1993). Moreover, reduced rat HPyr and LPyr concentrations occurred in gastrocnemius and soleus following 8 weeks of inclined treadmill running (Carroll et al., 2015). In addition, rats with lathyrism, a condition inhibiting LO activity (and therefore the collagen cross-linking process), possess weaker collagen and experience much greater strain versus healthy controls with the same force applied to their myofibres (Kovanen et al., 1984b). However, treadmill exercise counteracts the effects of this condition by promoting mature cross-link (HPyr and pyrrole) formation and improving muscle cross-link profiles; in one study this effect was so large that no differences in mechanical properties (e.g. structural stiffness and strength) existed between oldexercised rats and healthy sedentary controls (McNerny et al., 2015). Eccentric exercise stimulates collagen breakdown, indicated by increased blood Hyp and muscle MMP concentrations. Future studies should examine the effects of different exercise forms on contraction-specific changes in muscle cross-linking profiles. This may improve our understanding of optimal exercise interventions, especially in elderly and sedentary populations. Potential associations between exercise-induced collagen turnover and crosslinking profiles remain uninvestigated; to our knowledge, no study has assessed the effects of exercise on human muscle cross-linking profiles, although cross-sectional associations between improved muscle cross-linking profiles and regular exercise have been reported (Haus et al., 2007;Couppé et al., 2014).
(3) Hydroxyproline and hydroxylysine (Hyp and Hyl) Hydroxyproline (Hyp) and hydroxylysine (Hyl) are nonproteinogenic amino acids and major collagen components that stabilise the collagen triple helix (Kotch, Guzei & Raines, 2008). Muscle Hyp or Hyl concentrations are traditionally used to estimate total collagen concentration (Zimmerman et al., 1993), however, increases within biofluids most prominently result from collagen turnover (Murguia et al., 1988) while elevated plasma levels have been associated with increased injury risk (Murguia et al., 1988). Thus, Hyp or Hyl biofluid concentration assessments may serve as easy and minimally invasive methods for collagen turnover assessment and connective tissue injury-risk status.
Assessment of Hyp and Hyl in blood and urine can reliably quantify post-exercise collagen turnover in humans. A previous investigation examined changes in urine Hyp levels in eight participants after 50 maximal voluntary unilateral eccentric knee extensions which increased the Hyp-tocreatine ratio on day 2 after exercise with a 96% increase above baseline (Brown et al., 1997). In the same study, the Hyl-to-creatine ratio increased 69% above baseline on day 2 and was elevated again on days 5 and 9. Similarly, plasma Hyp concentration increased (40-53%) as soon as 24 h after a single volume-matched low-or high-intensity (no difference between groups) eccentric-only cycling and remained elevated above baseline at 72 h (last assessment) in men (Mavropalias et al., 2021). Other muscle damage-inducing activities such as plyometric exercise can also affect Hyp urine concentration. For example, serum Hyp increased 24-72 h after intense plyometric jumping exercises (200 jumps), with a peak at 48 h (80% increase from baseline) (Tofas et al., 2008). In the same study, Hyl concentration increased at 48 h post-exercise. Post-exercise increased Hyp and Hyl concentrations result from connective tissue damage by forceful muscle lengthening and subsequent breakdown through MMPs. However, some data contradict these findings, with no changes detected even after eccentric exercise (Wheat et al., 1989;Brown et al., 1999). For instance, after 60 min of intermittent running at 80% of peak oxygen uptake (VO 2peak ) on a − 10% slope, no changes in urine Hyl concentration were detected for up to 96 h post-exercise in 10 men (Wheat et al., 1989). Also, biofluid Hyp and Hyl concentrations are unaffected by concentric exercise (Virtanen et al., 1993;Brown et al., 1999;Nogueira et al., 2011). For example, no plasma Hyp changes were detected for up to 9 days after 50 maximal voluntary concentric unilateral knee extensions (Brown et al., 1999), or up to 4 days after 50 maximal concentric bilateral knee extensions (Virtanen et al., 1993). These findings led Brown et al. (1997) to suggest that eccentric but not concentric contractions induce muscle damage and collagen breakdown. This hypothesis was supported by another study (Brown et al., 1999)  conclusion that indices of collagen breakdown such as Hyp and Hyl can assess exercise-induced connective tissue damage. Likewise, Tofas et al. (2008) supported the use of these biomarkers, stating that serum Hyp and Hyl levels can be promising for detecting exercise-induced collagen degradation.
Regarding changes after long-term exercise training, lifelong endurance exercise can increase rat muscle Hyp concentration (Kovanen et al., 1987;Kovanen, 1989). However, no increases in muscle Hyp concentration existed after shorter durations (up to 10 weeks) of treadmill running (Kovanen et al., 1984b;Kovanen, Suominen & Heikkinen, 1984a;Zimmerman et al., 1993;Gosselin et al., 1998), although increased PH activity has also been reported (by extension suggesting increased Hyp synthesis) (Kovanen, Suominen & Heikkinen, 1980;Takala et al., 1983). Others reported increased rat Hyp content in gastrocnemius (72%) and soleus (63%) following 8 weeks of endurance running exercise (Carroll et al., 2015). Moreover, high-volume endurance exercise in both men and mice Hellsten et al., 1996), as well as 60 electrically stimulated tetanic eccentric gastrocnemius contractions in rats, increased muscle Hyp concentration and content (Takagi et al., 2016). Interestingly, the increase in Hyp concentration in rat gastrocnemius after one eccentric exercise session (+60%, 28 days later) (Takagi et al., 2016) was greater than the increase (+40%) in rat quadriceps after endurance training (daily running for 20 days) ), but it is not possible to assign this discrepancy to the difference in exercise type, muscle force produced (or work done), muscle morphology, measurement unit used (i.e. concentration versus content), or human versus animal difference due to the limited information provided. Nevertheless, at least 10 days of exercise appear necessary for detectable increases to be observed in animal muscles  as no increase for up to 7 days after 240 eccentric contractions or downhill running existed in some studies in rats (Han et al., 1999a;Koskinen et al., 2002). However, rather than the number of sessions per se, the time allowed for the concentration to increase (i.e. number of days) might be the key factor as 60% increased Hyp content was observed 28 days after eccentric gastrocnemius contractions in rats (Takagi et al., 2016). Therefore, at least in animals, muscle-damaging exercise can adequately increase muscle Hyp content but may require multiple days to manifest.
Interestingly, basal decreases in collagen turnover may occur after long-term eccentric but not concentric exercise training. After 8 weeks of either concentric or eccentric upper-body exercise training, decreased resting urine Hyp concentrations (−50%) were detected only in the eccentric group compared to the non-exercising control (Nogueira et al., 2011). Elevated resting plasma Hyp levels may therefore act as connective tissue injury risk indicators during muscle overuse (Murguia et al., 1988). Based on the above, it is possible that long-term eccentric exercise training may reduce connective tissue injury risk, as reflected by the lower Hyp concentrations. Assessment of resting biofluid Hyp concentrations in predicting connective tissue injury risk warrants further research.
Apart from exercise, environmental factors can also affect muscle Hyp synthesis. Hypoxia (with or without exercise) can increase rat muscle Hyp content (Perhonen et al., 1996). In fast-twitch rat EDL, Hyp increased (32%) in non-exercised rats after 56 days in a hypobaric chamber versus a 46% increase in rats that did treadmill running (5 per week; 2 per day). By contrast, decreased Hyp content (21%) was observed in the normobaric-exercise group, indicating that exercise itself provokes Hyp release. However, hypoxia seems to act independently of exercise, given that the response of the hypobaric-exercise was 68% greater than the normobaric-exercise condition. These results were somewhat mirrored in the slow-twitch soleus, where total Hyp content increased 78% more in the hypobaric-sedentary than normobaric-sedentary group after 56 days. Thus, there are clear and mutually exclusive effects of exercise and hypoxia on rat muscle Hyp concentrations. It is currently unclear whether this occurs in humans, although examination of this hypothesis might lead to novel treatments in injury rehabilitation or disease management.
Two more factors affecting muscle Hyp synthesis are aging and immobilisation. Aging increases muscle Hyp concentration in animals (Kovanen et al., 1987;Kovanen, 1989;Gosselin et al., 1998;Ducomps et al., 2003). Specifically, Hyp concentration in rat RF muscle increased between 10 and 24 months, and between 1 and 2 months (i.e. during maturation to adulthood) as well as 10 and 24 months in soleus (Kovanen et al., 1987). However, as described previously (in Section II.1), this increase contrasts with unchanged PH and GGT activities (Kovanen, 1989;Kovanen & Suominen, 1989) and might instead result from decreased collagen turnover during aging (Kovanen & Suominen, 1989). For similar reasons, immobilisation increases connective tissue area in human and animal muscle, which coincide with increased Hyp concentration (Williams & Goldspink, 1984;Savolainen et al., 1988a;Koskinen et al., 2000) in muscles immobilised at longer and shorter lengths (Williams & Goldspink, 1984;Savolainen et al., 1988a;Ahtikoski et al., 2001), although others have not detected changes up to 42 days of immobilisation (Karpakka et al., 1991;Han et al., 1999b;Ahtikoski et al., 2003). Hyp concentration would be expected to decrease during immobilisation as previous studies have clearly shown decreases in PH activity (Savolainen et al., 1988a,b;Ahtikoski et al., 2001). This discrepancy between PH and Hyp could be caused by slower collagen turnover rates compared to non-collagenous proteins, resulting in increased Hyp concentration (Ahtikoski et al., 2001). This hypothesis is supported by two studies (Savolainen et al., 1988a;Karpakka et al., 1990a) in which Hyp concentration (mg/g of soluble protein) of immobilised rat soleus increased but Hyp content (mg/whole muscle) was unchanged (Savolainen et al., 1988a;Karpakka et al., 1990a) or decreased (Savolainen et al., 1987), relative to baseline or controls. Moreover, it has been shown that the Biological Reviews 98 (2023)  endomysium-to-myofibre number remained unchanged despite the endomysium-to-myofibre area ratio increasing after 55 days of bed rest in human soleus muscle, further indicating that the absolute amount of connective tissue remains unchanged during immobilisation and it is the myofibre atrophy that causes the greater proportion of connective tissue (Thot et al., 2021). Therefore, although Hyp concentration may appear to increase during short-term immobilisation, this is likely only due to a slower turnover rate than other proteins (Williams & Goldspink, 1984;Savolainen et al., 1988a;Ahtikoski et al., 2001). Additional support for this is provided by reduced collagen synthesis during immobilisation (see Section II.1). These results highlight the need to select and report analysis methods carefully.
Regarding recommendations for Hyp or Hyl concentration tests, biofluid Hyp concentration may be more sensitive than Hyl for collagen degradation assessment as Hyl constitutes only about 0.4-0.7% (Segrest & Cunningham, 1970;Wheat et al., 1989), as opposed to the 12-14% Hyp content, in collagen molecules (Hall & Reed, 1957). Additionally, only about 25% of Hyp from degraded collagen is present in rat urine (Weiss & Klein, 1969), so blood is potentially a more sensitive biofluid in which to detect collagen degradation. Nevertheless, thought should be given to the selection of appropriate sample types as serum protein concentrations are lower than plasma concentrations (Lum & Gambino, 1974). Moreover, the diet of the participants requires consideration as dietary collagen and gelatin consumption can affect urinary Hyl concentrations (Askenasi, 1975). Unfortunately, Hyp and Hyl are constituents of all collagen types, so it is impossible to identify the tissue origin or collagen type. Bone turnover increases after exercise (Maïmoun et al., 2006;Kish et al., 2015), thus blood-or urine-based Hyp and Hyl assessments do not exclusively reflect collagen breakdown originating from exercised muscles. Moreover, blood Hyp concentration may be better for assessing collagen turnover than Hyl, however, researchers should be aware of its limitations.
In conclusion, despite inconsistencies, Hyp (in preference to Hyl) blood concentration can be useful for assessing collagen turnover, while muscle Hyp content (in preference to concentration) can assess total collagen content. In either blood or muscle, Hyp responds to exercise and disuse, but care should be given to standardise diet for biofluid assessment, while adequate time should be given for long-term content changes in tissue. Eccentric exercise appears to be a potent trigger of Hyp turnover (Table 4), which can lead to greater collagen synthesis, and thus is important for the renewal of old cross-linked collagen (see Section II.2).

III. OTHER PROTEOGLYCANS AND GLYCOPROTEINS
Proteoglycans are highly glycosylated proteins that contain a main protein body with different covalently linked glycosaminoglycan types. Proteoglycans can be found in almost all connective tissue types and have extensive molecular diversity (Schaefer & Schaefer, 2010). Many proteoglycans have multiple physiological functions, and many genetic diseases are linked to mutations in proteoglycan genes. Glycoproteins are protein molecules with covalently bound carbohydrate molecules (glycosylation). Many secreted ECM proteins are glycosylated, and this carbohydrate component serves numerous functions. Glycoproteins provide structure and can be found in the form of hormones, enzymes, and receptors in addition to possessing regulatory and transport functions.
Physical activity affects proteoglycan and glycoprotein concentrations. For example, more than half of the proteoglycans expressed in muscle significantly increased (more than 1.5-fold) following 12 weeks of bicycling (2 per week, 9-11 high-intensity intervals) and whole-body resistance exercise (2 per week), in the VL muscle of 40-65-year-old men. These were also upregulated after a single 45-min bicycling exercise session (Hjorth et al., 2015). In animal studies, increased (20%) total glycosaminoglycan concentration developed in chicken tendons after 4 weeks of treadmill running (Hae Yoon et al., 2003). In addition, decreased glycosaminoglycan content (48%) occurred in arthritic rabbit knees (versus healthy knees) after 24 h of immobilisation, and further reduction (37%) was observed at 48 h, yet continuous passive motion mitigated this decrease, resulting in only 12% and 20% reductions at the same time points (Ferretti et al., 2005). Therefore, even passive movement contributes to ECM preservation during immobilisation. Nonetheless, others have not detected acute exerciseinduced changes. For example, hyaluronic acid staining in the VL muscle of young and elderly humans did not change for up to 72 h after 300 maximal eccentric knee extensions (Sorensen et al., 2018). Future research is required to clarify human responses, for example by focusing on the effects of passive motion on ECM status in people with limited muscle activation capacity (e.g. after spinal cord injury) or in post-surgery patients. Aging can also negatively impact proteoglycan content; for example, as shown in Fig. 4, reduced perlecan concentration (a proteoglycan with a wide variety of biological roles) occurred in mouse muscle during aging. Although limited research exists documenting muscle glycosaminoglycan concentrations following exercise, disuse, and aging, moleculespecific or total concentrations of glycosaminoglycans could provide information on ECM status.

(1) Decorin
Decorin is a fibronectin-interacting proteoglycan regulating collagen fibrillogenesis and influencing transforming growth factor beta 1 (TGF-β1) activity (see Section V.1) (Grounds, 2008). As collagen synthesis is important in ECM tensile integrity, decorin responses to physical activity could elucidate the effectiveness of different exercise types on ECM remodelling. Unfortunately, few studies exist detailing changes in decorin mRNA levels following exercises of varying modes and intensities, although acute increases tend to occur after resistance exercise. For example, decorin Biological Reviews 98 (2023)  mRNA levels increased in VL muscle 6 h after 1 h of singleleg knee extensions (67% of maximum workload) in men although they remained unchanged in the patellar tendon (Heinemeier et al., 2013). However, no changes in decorin mRNA levels, or other proteoglycans such as biglycan, fibromodulin, and versican, were detected 4 h following 30 unilateral knee extensions (70% of concentric maximum) in the patellar tendon of either men or women (Sullivan et al., 2009). In another study, a single full-body resistance exercise session significantly increased plasma decorin concentrations (1.3-fold) in 10 well-trained healthy males compared to their pre-exercise values (Kanzleiter et al., 2014). In long-term training studies, a significant increase in VL decorin concentration (20%) was observed in physically inactive men (40-65 years) who completed a 12-week combined strength and endurance training programme (4× per week), which interestingly correlated with increases in leg press strength (Kanzleiter et al., 2014). Following 12 weeks of bicycling (2 per week, 9-11 high-intensity intervals) and whole-body resistance exercise (2 per week), the VL muscle of 40-65-year-old men showed no changes in decorin mRNA levels (Hjorth et al., 2015). Although little information exists regarding the effects of exercise type, a single resistance exercise session is sufficient to evoke prolonged increases in muscle decorin mRNA levels (Heinemeier et al., 2013), but those increases may attenuate after repeated exposure to the same exercises (Hjorth et al., 2015). Decorin expression has been shown to correlate positively with fat-free mass and strength improvements following chronic resistance training (Kanzleiter et al., 2014), and as such, additional research is needed to understand responses to different exercise types, aging, and immobilisation.
(2) Tenascin C (TNC) Tenascin C (TNC) is an ECM glycoprotein inhibiting fibronectin adhesion and critical for ECM remodelling initiation following altered physical loading (Sarasa-Renedo & Chiquet, 2005;Grounds, 2008). Although normally present at low levels in uninjured muscles, expression increases rapidly following mechanical loading or damage associated with inflammation and mechanical stress (Sarasa-Renedo & Chiquet, 2005;Grounds, 2008). Muscle TNC expression displays a consistent pattern of increase 24-48 h after exerciseinduced muscle damage in both animals and humans (Crameri et al., 2004(Crameri et al., , 2007Raastad et al., 2010;Mackey et al., 2011;Hyldahl et al., 2015), coinciding with reductions in rat muscle collagen concentration (Koskinen et al., 2001a). This fibronectin-inhibiting action might be an important driver of remodelling, initiating ECM-tomyofibre detachment which facilitates degradation of damaged structures and subsequent synthesis. Indeed, TNC expression can reliably quantify connective tissue damage, and further supports the suggested associations between delayed-onset muscle soreness and muscle extracellular matrix disruption (Raastad et al., 2010;Hyldahl et al., 2015;Sorensen et al., 2018). For example, muscle and connective tissue damage were examined in humans after performing maximal voluntary eccentric knee extensions with one leg and electrically stimulated contractions with the other (Crameri et al., 2007). VL muscle Z-line and intracellular disruption were more pronounced in the electrically induced (40%) than the voluntary contractions (10%). However, TNC expression increased equally (1.8%) in both training modes 24 h after exercise and remained elevated to the last measurement (192 h), with similar delayed-soreness responses in both conditions. The results suggest greater ECM involvement in the delayed-soreness response than myofibre damage. This is further supported by greater increases in TNC staining in the VL muscle of young versus elderly participants up to 72 h after 300 eccentric knee extensions (Fig. 5) (Sorensen et al., 2018). The older group experienced less functional decline while exercising and faster peak torque and average power recovery following exercise, as Table 4. Acute biofluid marker time course of changes following eccentric contractions in humans. Human models of acute responses to eccentric contractions included: 70 maximum voluntary eccentric contractions of the forearm flexors and extensors (Saxton & Donnelly, 1994), 100 maximal voluntary eccentric knee extensions (Mackey et al., 2004), 45 min −10 downhill running at 60% of each participant's maximal velocity (Koskinen et al., 2001b), 60 min −10 downhill running at 15 km/h (van de Vyver et al., 2016), 30 min of −10 downhill running at 70% of the age-predicted maximum heart rate (Welsh et al., 2014), 70 eccentric knee extensions at 110% of 10 repetition maximum (Nascimento et al., 2016), 50 maximal voluntary unilateral eccentric knee extensions (Brown et al., 1997(Brown et al., , 1999. ": significant increase, #: significant decrease (significance at least P < 0.05). Lack of changes versus baseline ($) is not shown, due to the large number of measurement time points.
Biological Reviews 98 (2023)  well as a non-statistical trend for reduced post-exercise soreness, compared to the young group. The reduced muscle strength during, and immediately after such a large number of eccentric contractions could be partially attributed to muscle connective tissue disruption, which can reduce lateral force transmission and change architectural gearing (Grounds et al., 2005;Eng, Azizi & Roberts, 2018). TNC expression might therefore be promising for quantification of ECM de-adhesion following different stimuli and provide information relating to injury status. It would be valuable to investigate potential associations between ECM damage markers such as TNC expression and force reduction following exercise-induced damage.
In muscle, TNC mRNA levels increase following muscle reloading after immobilisation. While no effect was observed on TNC expression, increases in TNC mRNA levels (up to 40-fold) and endomysial protein expression occurred in soleus 1-5 days after reloading of the muscle following 14 days of hindlimb suspension (Flück et al., 2003). As animal muscle collagen concentrations increase at 7 days after injurious exercise (Han et al., 1999a;Koskinen et al., 2001a), it is likely that the return to baseline of TNC mRNA levels were cryosectioned and mounted on polylysine-coated glass slides. Sections were fixed to the glass slide with 4% paraformaldehyde in phosphate-buffered saline (PBS), washed with PBS, and blocked with PBS containing 10% goat serum and 1% bovine serum albumin. Sections were incubated with anti-perlecan antibody (clone CCN-1) diluted 1/500 in blocking buffer for 1 h at room temperature (RT), washed three times in PBS, and then treated with anti-rabbit Alexa Fluor 488 diluted 1/500 in blocking buffer for 1 h at RT in the dark. Slides were washed and mounted with Vectashield. Slides were imaged using a Zeiss fluorescent microscope and SPOT Advanced software. Identical settings were used to image all samples and image analysis was performed using Image J. For each sample, the mean fluorescent intensity per fibre (N = 200 fibres) was calculated by obtaining the mean fluorescence for a field of view, divided by the number of fibres present. Statistical analysis was performed using the non-parametric Kruskal-Wallis test. Mean fluorescence was significantly reduced in both 18-and 22-month-old mice compared to 2-month-old mice (*P < 0.01). Scale bar = 100 μm. The perlecan antibody was kindly provided by John Whitelock and Megan Lord. Samples were collected, and staining and analysis performed, in the laboratory of Danielle E. Dye and Deirdre R. Coombe, in collaboration with Miranda D. Grounds (unpublished data 4-5 days after remodelling-provoking stimuli allows for subsequent ECM-to-myofibre reattachment. However, changes in muscle collagen concentrations or mRNA levels were not analysed, and it is therefore not possible to conclude the time course of remodelling (Flück et al., 2003). Nevertheless, these findings suggest that both immobilisation and remobilisation may trigger ECM remodelling. Muscle TNC expression appears useful in quantifying exercise-induced muscle damage, and is strongly associated with delayed onset of muscle soreness, however, its use in human research is hampered by the necessity for muscle biopsy samples.

(3) Fibronectin
Fibronectin is a glycoprotein essential for cell adhesion and migration, binding to integrin, collagen, and other ECM molecules (Grounds, 2008). A tripeptide sequence of arginine, glycine, and aspartate (RGD domain) contained within fibronectin is recognised by cell adhesion molecules, allowing for ligand binding. However, fibronectin cell adhesion is inhibited by TNC, which in turn inhibits cell migration and ECM remodelling (Grounds, 2008).
Exercise affects muscle fibronectin expression, as it increased above baseline in VL muscle 72 h after 300 eccentric knee extensions in both young (22.7 years) and old (70.9 years) adults (2.3-versus 1.3-fold), however, the overall expression was higher in the older group (Sorensen et al., 2018). Interestingly, smaller TNC changes occurred in the elderly (Fig. 5), who also experienced less functional decline than the younger group during exercise and faster peak torque and average power recovery. This further supports the inverse relationship between changes in muscle TNC and fibronectin expressions, possibly explained by the fibronectin-inhibiting action of TNC. Peak torque had not recovered in the young even up to 72 h after exercise (Sorensen et al., 2018), which may result from greater exercise-induced muscle damage than in the older group as the process of disassembling damaged structures would potentially require more time. It is possible that increased fibronectin expression would develop at later stages (after 72 h), coinciding with peak torque recovery. Fibronectin is a force mediator, transferring stress from the myofibre to the ECM and then laterally to adjacent myofibres (Tidball, 1991). Therefore, the reduced muscle function lasting multiple days after eccentric exercise could reflect a TNC-induced inhibitory period of fibronectin-to-integrin binding, subsequently reducing lateral force transmission. Future studies are required to investigate the inverse relationship between muscle TNC and fibronectin expressions following exercise-induced muscle damage, in addition to the hypothesised positive relationship between increased fibronectin expression and peak torque recovery.
Fibronectin expression increases during immobilisation (Salonen et al., 1985), coinciding with increased integrin-β1 content (Li et al., 2013). Fibronectin mediates integrin-to-ECM attachment and may assist in integrin signalling function, initiating ECM remodelling after exercise or immobilisation (see Section III.5). In denervated rat gastrocnemius, fibronectin expression increased in endomysia and perimysia 7 days after denervation versus the contralateral functioning limb, and levels remained elevated for at least 28 days after denervation (Salonen et al., 1985). This finding suggests that immobilisation triggers active ECM remodelling in addition to muscle atrophy while muscle fibronectin (and integrin-β1; see Section III.5) expression may be important when investigating exercise-or disuseinduced ECM remodelling.
Fibronectin is important for post-exercise muscle regeneration due to its ability to pair with fibrin to create linkages providing fibroblast anchorage (Grinnell, Billingham & Burgess, 1981). Increased fibronectin expression precedes collagen formation in new granulation tissue (new connective tissue forming in healing wounds) (Grinnell et al., 1981). In addition, fibronectin could assist in the rapid removal of damaged collagen molecules. For example, fibronectin mRNA levels in the VL muscle of men were unchanged at 2 h but increased (1.2-fold) 6 h after 1 h of single-leg knee extensions at 67% of maximal workload, before returning to baseline at 26 h (Heinemeier et al., 2013). When fibronectin binds to collagen fragments following collagenase digestion, it promotes phagocytosis and subsequent fragment clearing by fibroblasts and macrophages (Grinnell et al., 1981). The increases in fibronectin mRNA levels 6 h after exercise, together with increased TNC mRNA levels 3 h after exercise (Heinemeier et al., 2013), support the hypothesis that the initial fibronectin increase following injurious exercise may be more important for clearing damaged structures than for ECM-to-myofibre adhesion as was traditionally thought.

(4) Laminin
Laminins are ECM glycoproteins and major components of the basal lamina, where they bind with integrins and other molecules and influence cell differentiation, migration, and adhesion: about 16 isoforms are described, with laminin α2 forms being of major importance in skeletal muscle (Grounds, 2008). However, laminin concentration appears unresponsive to exercise (see Figs. 2 and 3). Both endurance-trained (for 2 years, 5 per week) and sedentary rats demonstrated similarly increased laminin concentrations in RF muscle between 4 and 10 months of age (Kovanen, 1989). Although both exercise and aging increased total collagen and collagen IV (a basement-bound collagen) concentrations in rat muscles, surprisingly, this did not influence laminin concentration. This was further confirmed in a follow-up study with the same design, where laminin concentration remained unchanged despite a trend toward increased training-induced collagen IV concentration (Kovanen et al., 1988). Therefore, laminin concentration might not be particularly responsive to exercise training. Moreover, following acute or long-term exercise, no changes were detected in laminin-α2 concentration in rat gastrocnemius after 1 or 18 electrical stimulation sessions (Ogasawara et al., 2014). Yet other studies have demonstrated that injection of a muscle laminin isoform (laminin-111) increased satellite cell content and enhanced new myofibre formation after downhill running-induced muscle damage in mice (Zou et al., 2014). Further, it attenuated disease symptoms in mouse Duchenne muscular dystrophy by increasing integrin-α7 synthesis and sarcolemma stabilisation (Rooney, Gurpur & Burkin, 2009). As these are important benefits, future research should aim to determine which exercise modalities can increase muscle laminin concentration in humans.
With regard to aging, decreases (−50%) in laminin concentration in rat soleus occurred between 1 and 10 months of age (Kovanen et al., 1988). However, laminin concentration in rat RF muscle increased between 2 and 10 months of age (210%) and was statistically different from the changes in soleus. The soleus is predominantly composed of type I myofibres, which have a 40-50% greater collagen concentration than muscles with predominantly type II myofibres such as RF (Kovanen et al., 1980(Kovanen et al., , 1984a(Kovanen et al., ,b, 1988. Greater muscle collagen concentrations are speculated to require greater laminin-dependent basal lamina anchoring, but the findings from the above study do not support this hypothesis (Kovanen et al., 1988). It is intriguing as to why laminin concentrations differed between muscles with predominately different fibre types during aging, at least in rodents.
Regarding laminin synthesis, laminin mRNA levels (LAMC1 and LAMC2) were up-regulated (2-fold) in VL muscle 3 h after 100 maximal eccentric knee extensions in adult men (Hyldahl et al., 2015). However, only a small increase (4-fold; non-significant) in LAMC2 mRNA levels occurred in VL muscle 2 days after 300 eccentric knee extensions (no change was detected at 27 days) and no change (2-fold increase; non-significant) occurred after a second session. Nonetheless, increased mRNA levels for laminin-β1 (5-fold) and laminin-β2 (1.5-fold) were observed 30 days after electrically stimulated isometric gastrocnemius contractions, without changes seen immediately after or following repeated exercise 30 days later (Mackey et al., 2011). Thus, only small changes in mRNA levels occur, even after muscle-damaging exercise, but from the limited information, it is not possible to conclude how voluntary exercise affects laminin mRNA levels. Although the small increases in laminin mRNA levels following acute exercise may not necessarily translate to long-term increased protein concentration, current evidence suggests that laminin concentration may increase after long-term high-tension muscle contractions, accompanying increased integrin expression (see Section-III.5) (Li et al., 2013;De Lisio et al., 2015). For example, following 12 weeks of bicycling (2 per week, 9-11 highintensity intervals) and whole-body resistance exercise (2 per week), the VL muscle of 40-65-year-old men showed significant increases in mRNA levels for laminin-α4 (1.7-fold), laminin-β1 (1.7-fold), laminin-β3 (1.5-fold), and laminin-γ3 (1.7-fold) (Hjorth et al., 2015). Other studies have also found similar increases in mRNA levels after long-term exercise training (Timmons et  Extracellular matrix biomarkers related to physical activity and aging concentration remains to be investigated in both humans and animals.

(5) Integrin
Integrins are transmembrane receptors enabling cell-to-cell and cell-to-ECM interactions in animals (Grounds, 2008;Hynes, 2009). They form costamere complexes, which connect the inside of the cell to the ECM, binding to various ECM components such as fibronectin, collagen, laminin, and tenascin. Due to this mechanical continuity, integrins can signal bi-directionally and regulate important intracellular and extracellular processes (Grounds, 2008;Hynes, 2009). Integrins are heterodimers, composed of 18 αand eight β-subunits that assemble in 24 different combinations with cell-specific expression patterns and overlapping substrate specificity (Grounds, 2008;Hynes, 2009). In addition, those subunits are further subdivided into isoforms. For example, the predominant integrin in adult skeletal and cardiac muscle is the β1D isoform (Belkin et al., 1996). Integrins are important for muscle force transmission and signalling, and thus studying their responses to physical activity could promote an understanding of exercise-induced strength or hypertrophic adaptations and inform exercise strategies in aging and pathological conditions. Conflicting data exist for animal studies quantifying integrin-α7 and -β1 concentrations following a single bout of exercise. For example, integrin-β1 protein concentration in soleus (−22%) and EDL (−17%) muscles decreased in rats 48 h after 40 min of downhill running (−16 , 20 m/min) (Nourshahi et al., 2013), but remained unchanged in gastrocnemius after resistance exercise (electrical stimulation) (Ogasawara et al., 2014) or in soleus at 2 h or even 48 h after 90 min of downhill running (18 m/min, −16%) (Graham et al., 2015). Discrepancies have also been observed for integrin-α7 concentration after exercise, including no changes in rat gastrocnemius after resistance exercise (electrical stimulation) (Ogasawara et al., 2014) or in either rat or mouse gastrocnemius-soleus complex 3 h after 40 min downhill running (−20 , 17 m/min) (Mahmassani et al., 2017). However, decreases in integrin-α7 concentration were observed in rat soleus 2 h but no change at 48 h after 90 min of downhill running (18 m/min, −16%) (Graham et al., 2015), while increases (70%) in mouse gastrocnemiussoleus complex 24 h after 30 min downhill running (−20 , 15 m/min) were observed, which persisted for up to 1 week (Boppart, Burkin & Kaufman, 2006). These discrepancies might be explained by differences in myofibre composition between the muscles (or between species). However, due to the limited number of studies, the role of each of the two subunits in regulating anabolic signalling, and their exercise responses, remains unclear.
The various integrin subunits have specific functions and respond differently to similar stimuli. For example, 3 h after 30-min downhill running (−20 , 17 m/min), mRNA levels in the mouse gastrocnemius-soleus complex increased for integrin-α7 (5.4-fold) but decreased for integrin-α4 (57%; persisting for 1 week), with no changes for integrin-α5 mRNA, and for integrin-α6 there was no change at 3 and 24 h post-exercise but 84% decrease at 1 week (Boppart et al., 2008). The acute decrease in integrin-α4 mRNA level may result from this subunit adhering to fibronectin (Rüegg et al., 1992). Fibronectin-to-integrin adhesion is blocked by TNC (Sarasa-Renedo & Chiquet, 2005;Grounds, 2008), and manifests following maximal eccentric contractions (Crameri et al., 2007;Hyldahl et al., 2015). Therefore, reduced integrin synthesis might represent another strategy promoting ECM-to-myofibre detachment, which together with increased TNC mRNA levels and protein expression, could lead to lower fibronectin adhesion. However, the lack of change in integrin-α5 mRNA (an adhesion-promoting subunit; Boppart et al., 2008) following downhill running remains uninvestigated.
Following acute exercise in humans, integrin-α7, -β1A, and -β1B (different isoforms of -β1) concentrations were unchanged at 3 h following 150 maximal eccentric knee extensions. However, these concentrations increased at 24 h following exercise by 3.8-fold, 3.6-fold, and 3.9-fold, highlighting the importance of carefully selecting the sampling window (De Lisio et al., 2015). Notably, in another study, increased integrin-α7 mRNA levels (4.6-fold), were detected 2 days after 300 maximal eccentric knee extensions (Hyldahl et al., 2015), which, together with the aforementioned protein concentration changes, suggests that it is preferable to test at least 24 h after exercise to quantify exerciseinduced integrin responses, as these appear to require more than 3 h to manifest (De Lisio et al., 2015) but persist for at least 2 days thereafter (Hyldahl et al., 2015).
These integrin subunits are found in dimers that are important for muscle health and function (Burkin et al., 2005;Lueders et al., 2011;Zou et al., 2011;Mahmassani et al., 2017). Specifically, integrin-α7β1 dimers promote hypertrophic signalling, muscle growth, and muscle force production , protect against eccentric exercise-induced muscle damage (Mahmassani et al., 2017), and compensate for the lack of dystrophin in dystrophic mice. For example, transgenic mice with increased integrin-α7β1 levels show greater exercise-induced muscle hypertrophy (Burkin et al., 2005;Zou et al., 2011) and larger increases in upstream anabolic signalling pathways (Lueders et al., 2011) than their normal counterparts. From the studies above, it can be concluded that even a single exercise session changes human muscle integrin concentration, possibly resulting in stronger signal transduction and satellite cell proliferation as well as insulin growth factor 1 and TGF-β secretion (see Section V.1), both of which mediate muscle hypertrophy (Burkin et al., 2005). However, little information exists describing the role and responses of α7 and β1 subunits in human muscle despite potentially important clinical applications, for example in muscle dystrophy. Therefore, the effects of different exercise interventions on the sustained responses in integrin-α7β1 levels require further exploration.
Contrary to acute exercise, integrins may respond more cohesively to long-term exercise training in both animal Biological Reviews 98 (2023)  and human models. Integrin-α7 concentrations were unchanged after 18 resistance exercise sessions (electrical stimulation) in rat gastrocnemius (Ogasawara et al., 2014) or in mouse gastrocnemius-soleus complex after 4 weeks of downhill running (−20 ; 3 per week, 30 min, 17 m/min) . Similar results were observed in humans, including unchanged integrin-α7 concentrations in VL muscle after 24 eccentric cycling sessions in men (Mavropalias et al., 2022). Thus, integrin-α7 appears to respond only acutely and only in response to muscle damage-inducing exercise.
Apart from exercise, muscle disuse also affects integrin concentrations, however, the responses vary among studies. During 5 days of bed rest or neuropathy-induced physical inactivity, α7 and β1 integrin signalling (transcriptome) (Mahmassani et al., 2019), mRNA levels (Anastasi et al., 2008), and concentrations (Anastasi et al., 2008;Mahmassani et al., 2019) were downregulated (to a greater extent in old human muscle versus young). However, in another study, human integrin-β1 concentration increased (22%) in VL muscle after both 10 and 34 days of bed rest (although this was not statistically significant) (Li et al., 2013). Interestingly, in people with sensorimotor polyneuropathy, integrin-α7B expression in gastrocnemius was lower than in healthy patients, whereas integrin-α7A mRNA was higher despite decreased integrin-β1 mRNA levels and expression (Anastasi et al., 2008). These results highlight the importance of carefully investigating the different integrin isoform functions and responses to both exercise and disuse. The findings also suggest that different isoforms (α7A versus α7B) might respond differently to the same stimulus (Anastasi et al., 2008).
Taken together, integrin quantification provides unique insights into muscle structural integrity and anabolic signalling. Integrins seem to be more responsive to both acute and long-term exercise compared to laminins. While discrepancies between responses to both acute and long-term exercise exist (Tables 2 and 3), investigating responses of different integrin subunits (and isoforms of each subunit) should improve our understanding of muscular myopathies and tissue adaptations following different exercise regimes.
For example, specialised exercise programmes targeted at increasing integrin-α7 and -β1 concentrations could speculatively benefit people with muscular dystrophy to attenuate disease effects.

IV. FIBROBLASTS
While many cells (including myoblasts and myotubes) synthesise diverse ECM molecules, fibroblasts are the main cells that synthesise collagen and many other ECM components and are present in both tendon and muscle connective tissues. Fibroblasts detect mechanical stress via integrin attachments and regulate their own protein synthesis following mechanical stimuli (Sarasa-Renedo & Chiquet, 2005). In one study, fibroblast numbers in healthy young men were greater at 30 days (4-fold) than 7 days (2.5-fold) after 200 electrical stimulation-induced eccentric contractions . Fibroblasts are required for muscle ECM synthesis, and they may facilitate faster ECM remodelling following injury. Interestingly, fibroblast proliferation persists for longer than 7 days after exercise-induced injury , which, speculatively, may strengthen the muscle against future exercise-induced injuries that would require restructuring. A greater number of collagensynthesising fibroblasts could explain the greater increases in Hyp (i.e. muscle collagen) expression 24 h after a second session (28 days later) of eccentric exercise compared to 24 h after the first (Takagi et al., 2016), and account for the faster muscle function recovery and reduced soreness (Hyldahl et al., 2015). Thus, fibroblast numbers can quantify eccentric exercise-induced muscle adaptation and potentially ECM synthesis, however, muscle biopsy samples are required to assess their numbers, limiting their utility in human experiments.
It is also important to note that excessive activity of fibroblasts and myofibroblasts can result in pronounced collagen deposition and fibrosis with adverse consequences in many tissues (Serrano & Muñoz-C anoves, 2010). Muscle fibrosis creates stiff contracted tissue with reduced range of motion and strength and is associated with several conditions. Fibrosis of skeletal muscle results from chronic inflammation, repeated injury, and hypoxia, which activate TGF-β (see Section V.1) and stimulate fibroblasts to transform into myofibroblasts (Valle-Tenney et al., 2020). Myofibroblasts are the effector cells of fibrosis, and continue to produce ECM (especially collagen I, fibronectin, and TNC) and inflammatory cytokines after the initial insult has resolved (Usher et al., 2019). Although some myofibroblasts are present in the healthy fascia, numbers increase in response to injury, and repeated micro-injuries can lead to pathology, contraction, and increased pain.
Although mechanical strain activates myofibroblasts, cyclic mechanical loading reduces the expression of TGF-β in lung cells in vitro and reduces myofibroblast transformation when conditions mimic the in vivo environment Biological Reviews 98 (2023)  Extracellular matrix biomarkers related to physical activity and aging (Blaauboer et al., 2011). Continuous passive movement is cyclic stretching that mimics natural movements, and continuous passive movement within the pain-free zone has been shown to decrease inflammation (Ferretti et al., 2005). Supporting this, acute inflammation in healthy rats was attenuated by stretch, reducing neutrophil migration and increasing the production of pro-resolving mediators (Berrueta et al., 2016). These results were replicated with connective tissue in vitro, suggesting that the anti-inflammatory effects of stretch are mechanically induced by the deformation of the ECM and are independent of muscle activation, blood, and lymphatic flow (Berrueta et al., 2016). The authors of the latter study suggested that connective tissue is an important mediator and regulator of the immune system, with non-injurious deformation of ECM having powerful effects on immune-mediated processes that are anti-inflammatory. However, exercise increases the release of TGF-β and connective tissue growth factor (CTGF; also referred to as CCN2), so it would seem wise to approach exercise in the presence of muscle fibrosis with caution since muscle damage (even micro-injuries) resulting from exercise can lead to further fibrosis and atrophy. Pain-free stretching (as opposed to forced stretch that creates micro-tears), including gentle continuous passive movement, is likely to be beneficial.

V. GROWTH FACTORS, ENZYMES, AND MODULATION OF ECM COMPONENTS
While a multitude of interacting proteins provide the dynamic structure of the skeletal muscle ECM, their properties are modulated by a very wide range of growth factors and enzymes (e.g. for synthesis, degradation, and sulphation) and some of these key growth factors and enzymes are worthy of discussion.
(1) Growth factors In addition to binding to growth factors/cytokines, ECM components can trigger the release and activation of large amounts of growth factors stored in the ECM, thereby directly and rapidly influencing cell receptor signalling and cellular activity and function. TGF-β and CTGF are key regulatory cytokines for ECM synthesis and degradation (Schild & Trueb, 2002;Grounds, 2008;Robertson & Rifkin, 2016;Weiskirchen, Weiskirchen & Tacke, 2019). They regulate cell adhesion, survival, migration, and proliferation in diverse cell types. Both growth factors are pivotal for the turnover and synthesis of collagen I and other matrix components (Schild & Trueb, 2002;Grounds, 2008;Robertson & Rifkin, 2016;Weiskirchen et al., 2019). Although CTGF is produced downstream of TGF-β during fibrogenesis, the relationship of these growth factors with ECM components is bidirectional, as CTGF can bind to TGF-β to enhance receptor activation (Schild & Trueb, 2002;Grounds, 2008;Robertson & Rifkin, 2016;Weiskirchen et al., 2019). TGF-β and CTGF gene expression respond rapidly to structural damage, and therefore quantification in muscle and potentially biofluids can provide information about ECM synthesis.
TGF-β is a 'master switch' driving collagen gene expression, consisting of a family of three subtypes, with TGF-β1 and -β2 upregulating fibronectin, collagen I, and collagen III mRNA levels (Weiskirchen et al., 2019). Serum TGF-β1 levels respond acutely to exercise. For example, plasma TGF-β1 concentrations increased (30%) in men after 1 h of uphill treadmill running (12 km/h, 3% incline) and remained elevated 6 h afterwards albeit unchanged in the Achilles tendon peritendinous region (Heinemeier et al., 2003). Collagen I synthesis directly depends on TGF-β1 activity, thus serum TGF-β1 concentration may be useful for quantifying collagen synthesis. With regard to long-term exercise training, participants after a 4-week heavy-strength training programme (5 per week) had increases in plasma TGF-β1 concentration at 14 and 21 days during training, returning to baseline by the 28th training day (Hering et al., 2002). Large amounts of latent TGF-β1 are stored in ECM (Robertson & Rifkin, 2016), and are activated and released by damage and mechanical stress, suggesting that muscles and connective tissues are likely sources of plasma TGF-β1 after exercise. However, while increased circulating TGF-β1 after exercise may originate from mechanically loaded skeletal and cardiac muscle, it could also originate from other collagen Icontaining structures such as bone (as also observed for Hyp). Regardless of origin, current evidence suggests that plasma TGF-β1 levels increase rapidly following exercise and likely mediate collagen synthesis and ECM remodelling.
CTGF and TGF-β mRNA levels are acutely affected by exercise. Human muscle CTGF and TGF-β2 mRNA levels increase as early as 2 h (Heinemeier et al., 2013) or 3 h (Hyldahl et al., 2015) after exercise, but it is unclear when they return to baseline. For example, mRNA levels of these molecules remained elevated in human VL muscle at 6 h but returned to baseline 26 h after 1 h of knee extensions (Heinemeier et al., 2013), yet were elevated at 30 days after one session of electrically stimulated contractions in human gastrocnemius versus the control limb (Mackey et al., 2011). Eccentric contractions elicit large increases in CTGF mRNA levels in rats and humans Hyldahl et al., 2015). CTGF and TGF-β1 mRNA levels in rat gastrocnemius were higher after eccentric (TGF-β1: 10-fold, CTGF: 2.8-fold) and isometric (TGF-β1: 4-fold, CTGF: 2.3-fold) than concentric contractions (TGF-β1: 2-fold, CTGF: 1.2-fold), with this pattern also observed for LO mRNA levels. However, it is unclear whether these changes were triggered by the higher forces inherent to eccentric (and isometric) contractions, the contraction mode, or possibly the greater muscle and ECM damage observed after electrically stimulated versus voluntary contractions Fouré & Gondin, 2021). Interestingly, increases in CTGF and TGF-β mRNA levels in VL muscle were almost identical when measured 3 h after 100 eccentric knee extensions in men, with CTGF increasing 7.6-fold and Biological Reviews 98 (2023)  TGF-β2 transcripts increasing 7.8-fold (Hyldahl et al., 2015). Similar exercise-induced increases in TGF-β2 (1.8-fold) and CTGF (2.1-fold) mRNA levels in VL muscle occurred in men 2 h after an hour of single-leg knee extensions (67% of maximum workload), but the increased TGF-β1 (1.7-fold) was not statistically significant versus the control limb (Heinemeier et al., 2013). Possibly, changes in CTGF mRNA levels are more closely associated with increased TGF-β2 than TGF-β1, although information is limited. Nevertheless, these findings further highlight the potential importance of eccentric exercise for stimulating collagen synthesis.
Aging and immobilisation influence growth factors. For example, increases in TGF-β and CTGF mRNA levels were more pronounced in old than young rats after 12 weeks of loaded ladder-climbing exercises (Guzzoni et al., 2018). Of clinical interest, immobilisation appears not to impact CTGF mRNA levels although reloading after immobilisation does. For example, rat tendon CTGF mRNA levels were unchanged after hindlimb suspension for 7 or 14 days followed by 2, 4, 8, or 16 days of reloading, however, levels were moderately increased on days 4 and 8 of reloading in muscle (Heinemeier et al., 2009). Unfortunately, due to limited evidence, conclusions cannot be drawn concerning CTGF and TGF-β mRNA responses to immobilisation. Future research could examine changes following immobilisation and aging in human muscle, together with direct quantification of muscle collagen synthesis such as collagen FSR.
TGF-β is important for ECM hypertrophy as well as its remodelling. This was shown in a study in which TGF-β inhibition following eccentric exercise-induced muscle damage resulted in rapid recovery of muscle strength but also incomplete structural regeneration and reduced long-term muscle strength (Gumucio et al., 2013). Conversely, chronically elevated TGF-β reduces myotube and myofibre diameter and causes muscle atrophy, possibly reducing muscle strength (Abrigo et al., 2016). In fact, in many fibrotic diseases, excessive connective tissue proliferation, myofibre atrophy, and chronically elevated TGF-β are all observed (Usher et al., 2019). Collectively, rapid TGF-β and CTGF increases following exercise may influence collagen synthesis, ECM structure, and transmission of muscle force, possibly by inducing fibroblast proliferation (Schild & Trueb, 2002;Grounds, 2008;Robertson & Rifkin, 2016;Weiskirchen et al., 2019). The large increases in serum TGF-β concentrations and mRNA after both acute and long-term exercise highlight potential roles as indicators of ECM hypertrophy and remodelling in healthy subjects. However, as this cytokine is pleiotropic and health and inflammatory status can alter its levels and effects, the health status of test participants must be considered. (

2) Enzymes
Enzyme activity orchestrates all aspects of ECM remodelling during growth, regeneration, and adaptation to mechanical loading or unloading. This enzymatic activity includes a wide range of proteins involved in the biosynthesis of the diversity of ECM proteins such as collagens, MMPs in combination with tissue inhibitors of MMPs (TIMPs); and enzymes involved in the synthesis and sulphation of glycosaminoglycans and their many complex protein interactions (Gottschalk & Elling, 2021). Typically, enzymes in biofluids (blood or urine) and muscles are assessed through their total activity (referred to below simply as 'activity', i.e. the amount of substrate converted to product over time) or their specific activity (activity per milligram of total protein).

(a) Prolyl 4-hydroxylase (PH) as a biomarker of collagen synthesis
A key biomarker of collagen biosynthesis is the enzyme prolyl 4-hydroxylase (PH), which is essential for collagen posttranslational modifications by catalysing Hyp formation, and its activity can thus quantify collagen synthesis (Gorres & Raines, 2010). This modification step, which is critical for collagen stiffness development, also requires ascorbic acid as a co-factor. In diseases such as scurvy, long-term vitamin C deficiency causes a lack of lysine and proline hydroxylation and a loose collagen fibre triple helix, resulting in tensile weakness (Peterkofsky, 1991). Therefore, alterations in PH activity or its co-factor can profoundly affect collagen mechanical stiffness.
Rat muscle PH activity changes significantly following acute exercise Han et al., 1999a;Koskinen et al., 2002). For example, reduced PH activity (−30% versus controls) 6 h after 240 electrically evoked eccentric contractions was subsequently followed by increases at 2, 4, and 7 days (peak 330% at 7 days) in rat tibialis anterior (TA) (Koskinen et al., 2002). This initial decrease in PH activity could reflect the removal of damaged collagen after exercise prior to the subsequent onset of collagen synthesis. Large increases in PH activity also developed in rat muscles from 2 to 10 days after treadmill running (up to 250% versus baseline)  and from 12 h to 14 days after downhill running (100%) (Han et al., 1999a), indicating that, like high-force eccentric contractions, prolonged running exercise might evoke acute changes in PH activity. However, changes during the first hours after exercise were not reported, so it is unknown whether these exercise models also initially decrease PH activity. Collectively, eccentric or running exercises induce large increases in rat muscle PH activity, and both exercise modalities provide a model with which to study acute exercise-induced collagen synthesis changes (Tables 2 and 3). Nonetheless, these limited data do not permit conclusions to be drawn regarding acute exerciseinduced changes in human muscle PH activity, while no data exist to determine acute PH responses to other exercise types, including heavy resistance, sprint, non-running endurance, or concentric-only exercise.
Increased muscle PH activity has been consistently observed following longer-term exercise training (Suominen & Heikkinen, 1975;Suominen, Heikkinen & Parkatti, 1977;Kovanen et al., 1980;Takala et al., 1983;Kovanen, 1989;Kovanen & Suominen, 1989; Biological Reviews 98 (2023)  Extracellular matrix biomarkers related to physical activity and aging Perhonen et al., 1996). In rats, running for as little as 4 weeks (5 per week) increased quadriceps femoris PH activity (67% versus controls) (Takala et al., 1983), and this increase persisted (60% versus non-training controls) for the entire lifespan of the animal (2 years) (Kovanen, 1989). Thus, running is a potent stimulus for continuous collagen turnover in rats, which is an important process for muscular health (discussed in Section II.2). Similar human responses are documented, as 8 weeks of endurance training (1 h; 5 per week) increased PH activity in the VL muscle of elderly women (69 years, 160% versus baseline) but not men (Suominen et al., 1977). The same research group found a 50% greater PH activity in the VL muscle of habitually endurance-trained men (running or cross-country skiing) than in untrained controls (Suominen & Heikkinen, 1975). These findings suggest that long-term endurance training is a potent PH activity stimulus and, subsequently, a promoter of collagen synthesis.
Maturation to adulthood in rats is associated with decreased muscle PH activity (Kovanen, 1989;Kovanen & Suominen, 1989). For example, PH activities in rat soleus and RF muscles decreased sharply (about −80%) between 1 and 2 months of age but did not change further over the following 22 months of life (Kovanen, 1989). Therefore, at least in rats, PH activity increases during the early growth period, decreases during maturation to adulthood, and then remains unchanged into old age. Thus, muscle growth and regeneration might be quantified through muscle PH activity, but this needs to be investigated in humans.
By contrast, studies show that immobilisation reduces PH activity (Savolainen et al., 1987;Karpakka et al., 1990aKarpakka et al., , b 1991Han et al., 1999b), although the effect on muscle and tendon is attenuated when muscles are immobilised at longer lengths (Savolainen et al., 1988a,b;Ahtikoski et al., 2001). Three weeks of rat hindlimb immobilisation at a short muscle length reduced PH activities in gastrocnemius (−63%) and soleus (−77%) (Savolainen et al., 1988b), however, no difference to controls existed when muscles were immobilised at long lengths. Of interest, PH specific activity was greater in TA after immobilisation at both short and long muscle lengths versus controls, which could suggest that TA is less dependent upon stretch as a stimulus for collagen synthesis than gastrocnemius and soleus muscles, and this has been observed in other studies (Karpakka et al., 1990b). Reloading of muscles provides a strong remodelling stimulus, with increased PH activity (versus baseline) in rat soleus and TA observed as soon as 3 days (32% above control values) after remobilisation (Karpakka et al., 1990b(Karpakka et al., , 1991. Taken together, these findings suggest that immobilisation could impair collagen synthesis by decreasing muscle PH activity, however, placing the muscle at a longer length might at least partly attenuate this decrease in some muscles. Whether this response is similar in humans has yet to be investigated. Few studies have examined exercise-or immobilisationinduced changes in PH mRNA levels. Downhill treadmill running increased PH α-subunit mRNA levels in type I myofibres in rat RF muscle from 12 h to at least 2 days (1.9-fold peak increase), while β-subunit mRNA levels were increased from 12 h to 7 days after exercise (6-fold increase); much smaller increases in β-subunit mRNA were detected after 2 days in soleus (Han et al., 1999a). By contrast, decreases in rat soleus α-subunit mRNA levels after 7 days of short-but not long-length immobilisation occurred (Ahtikoski et al., 2001), although no changes manifested in EDL, TA, or gastrocnemius. These changes in rat muscle mRNA levels are consistent with changes following physical activity and immobilisation at longer versus shorter muscle lengths as well as in the responsiveness of different muscle groups to immobilisation (Savolainen et al., 1988a,b;Ahtikoski et al., 2001). Unfortunately, no inferences can be made for humans as relevant information is lacking.
In summary, PH activity is readily measured in muscles, but no study has examined its responses in biofluids. Immobilisation-induced changes in muscle PH activity and mRNA levels respond consistently in animals, highlighting the need for further study of these biomarkers following altered mechanical loading in humans (Savolainen et al., 1988b;Ahtikoski et al., 2001). The differential responses between muscle groups suggest that a specific focus on the properties (and possibly the major functional role) of different muscle groups is needed to gain a better understanding of the mechanisms that affect collagen synthesis. Moreover, additional animal and human studies examining responses to long-term endurance and resistance exercise training are required to explore further the usefulness of PH activity and mRNA levels as reliable biomarkers. In addition, research into the effects of exercise intensity, duration, and other aspects of exercise programming on PH activity and mRNA levels is needed to determine the most effective exercise prescription for collagen synthesis. It is unclear whether this prescription varies between people of different ages, exercise-training histories, or disease status.

(b) Galactosylhydroxylysyl glucosyltransferase (GGT) and collagen synthesis
Galactosylhydroxylysyl glucosyltransferase (GGT; also known as procollagen glucosyltransferase) is an enzyme required for collagen post-translational modifications during synthesis (Kivirikko & Myllylä, 1979). Like PH, GGT activity during collagen post-translational phases is critical to collagen quality and stability. Measurement of its concentration has been traditionally used to monitor the progression of human diseases associated with increased collagen synthesis such as pulmonary fibrosis . In animals, muscle GGT activity closely follows collagen synthesis. For example, type I mouse myofibres have 25% higher GGT activity than type II (Takala et al., 1983), which may be explained by the greater collagen amounts in muscles predominantly composed of type I myofibres (+40-50% than those predominantly composed of type II) (Kovanen et al., 1980(Kovanen et al., , 1984a(Kovanen et al., ,b, 1988. GGT could therefore be useful in quantifying collagen synthesis following altered mechanical loading. Serum GGT concentration in men increased immediately (17%) and 3 days (24%) after 50 maximal concentric (i.e. muscle shortening) knee extensions (Virtanen et al., 1993). Interestingly, Hyp, a serum marker of collagen breakdown, did not detectably increase, suggesting that concentric contractions do not induce connective tissue 'damage', even at maximal exercise intensities. Instead, the authors attributed the increased GGT concentration following exercise to leakage from collagen-synthesising cells (such as osteoblasts and/or fibroblasts), rather than collagen synthesis increases, and this possibility warrants further research. Similarly, gradually increased serum GGT concentration manifested during a competitive 24-h run in men, which peaked (68%) 24 h after the run, before returning to baseline 48 h after the race . The same group found a 40% increase in serum GGT concentration immediately after a 24-h cross-country skiing event in men, which returned to baseline after 24 h (Takala et al., 1989). However, increases in serum GGT concentration as early as 24 h after exercise are incompatible with the time course of increased muscle collagen expression (Mackey et al., 2004) (see Section-II) and it is more likely that tissue GGT concentration increases at 96 h onwards to support collagen posttranslational modifications. Macrophage GGT leakage may also explain some of the early increases in biofluid GGT concentrations since GGT activity is relatively high in macrophages (Myllylä & Seppä, 1979), and macrophage numbers increase with exercise-induced inflammation. Therefore, biofluid GGT activity assessment may be unreliable for quantifying acute collagen synthesis but could potentially quantify exercise-induced inflammation. Whether human biofluid GGT activity can reliably quantify collagen synthesis following long-term exercise training remains to be investigated.
With regard to aging, human serum GGT concentration is 30% higher in newborn infants than in adults (Anttinen, 1977). Speculatively, this might result from a greater collagen synthesis within bones and connective tissues during infant growth. Nevertheless, future studies should consider the developmental stage of participants when using collagen synthesis-specific enzymes such as serum PH or GGT concentration, as this may influence test results. In support, maturation towards adulthood is associated with decreased rat muscle GGT activity (Kovanen, 1989;Kovanen & Suominen, 1989). Like PH, GGT activities in soleus and RF muscles decreased sharply (about −80%) between 1 and 2 months of age (maturation to adulthood) but remained constant over the following 22 months (Kovanen & Suominen, 1989). These findings further support the hypothesis (see Section V.2.a) that GGT activity, together with PH, may be reliable biomarkers of long-term growth or muscle ECM regeneration. However, these inferences are derived from animal studies, and it is unclear whether human muscle GGT activities respond similarly.
Long-term exercise training increases muscle GGT activity in animals (Takala et al., 1983;Kovanen & Suominen, 1989). In a study of short-term daily running exercise, GGT specific activity in rat RF muscle was lower than PH specific activity and increased (30%) only after 10 days of exercise before returning to baseline at 20 days (Takala et al., 1983). Long-term effects occurred after 2 years of treadmill running (5 per week), with increased GGT activity (35-50%) in the soleus of trained rats versus untrained controls (Kovanen & Suominen, 1989). Moreover, exercise training using concentric-only contractions, such as 3 weeks of swimming (5 per week), strongly stimulates rat TA GGT activity (42% versus controls) (Karpakka et al., 1990a). Evidently, unlike PH, increased GGT activity manifests only after a relatively large number of exercise sessions, at least in animals (Takala et al., 1983). This could mean that muscle GGT activity is more associated with long-term collagen modifications and might be more useful as a long-term connective tissue adaptive marker. Moreover, activity decreases after a few sessions, which may indicate that collagen modification has slowed due to unchanged mechanical stimulus (Takala et al., 1983). This could have implications for monitoring the point at which sufficient collagen synthesis occurs following muscle connective tissue injury; however, to our knowledge, no study has quantified muscle or biofluid GGT activities following long-term exercise training or assessed human myoskeletal injury status.
Immobilisation reduces rat muscle GGT activity (Savolainen et al., 1987;Karpakka et al., 1990bKarpakka et al., ,a, 1991, however, this is attenuated when muscles are immobilised at longer lengths (Savolainen et al., 1988a,b). Decreased GGT activity in rat gastrocnemius (−37%) and soleus (−53%) occurred after 3 weeks of immobilisation at short muscle lengths, whereas no differences relative to healthy controls were detected after immobilisation at long lengths (Savolainen et al., 1988b). The same group reported a 25% decrease in GGT activity in distal rat tendon following 3 weeks of immobilisation in shortened but not lengthened muscle, however, there were no differences in GGT activity detected between rat gastrocnemius, soleus, or TA immobilised at short or long lengths, albeit the shortened position resulted in greater muscle atrophy (Savolainen et al., 1988a). This discrepancy could have been caused by the first study inducing immobilisation through fixation (Savolainen et al., 1988b), whereas the latter used denervation (Savolainen et al., 1988a). Noticeably, in the denervation study where both units were measured, GGT specific activity increased whereas total activity decreased (Savolainen et al., 1988a), highlighting the need to carefully consider the reporting units before conclusions are drawn. Therefore, muscle lengths during immobilisation affect muscle and tendon collagen synthesis. Similar to PH, remobilisation after 1 week of immobilisation triggered considerable increases in total GGT activity in rat soleus in as little as 3 days (42%), with a peak at 7 days (135%), and had returned to control values by 14 days of movement (Karpakka et al., 1991). Despite the considerable information from animal studies, there is currently no information relating to the effects of immobilisation or remobilisation on human muscle GGT activity.

Extracellular matrix biomarkers related to physical activity and aging
Taken together, the current evidence suggests that changes in GGT muscle activity following immobilisation and long-term exercise training are similar to changes in PH activity and are also affected by the developmental stage. However, GGT activity is unaltered following a small number of exercise sessions. In addition, acute changes in biofluid GGT concentrations could be derived from sources other than the muscle connective tissues. Until this is clarified, biofluid GGT concentrations should not be considered a viable biomarker of exercise-induced connective tissue damage but may serve to monitor post-injury connective tissue healing (Risteli & Kivirikko, 1974;Risteli, Tuderman & Kivirikko, 1976;Anttinen, 1977;Anttinen et al., 1985). Unfortunately, the usefulness of biofluid GGT concentration in quantifying collagen synthesis following long-term exercise training in humans remains largely unexamined. Of practical importance, changes in rat GGT concentrations after treadmill running and swimming appear reliable to indicate enhanced collagen synthesis although these exercise types are not thought to impose high muscle tensile stress. Future studies are needed to describe changes in human muscle and biofluid GGT levels following loading and unloading and specifically to examine responses to conventional resistance or eccentric-only exercise training.

(c) Lysyl oxidase (LO) and collagen synthesis
Lysyl oxidase (LO) is a copper-dependent extracellular enzyme that forms crosslinks with Hyp and Hyl, through the LO-dependent oxidisation of lysine and Hyl residues to aldehydes in tropocollagen (Kagan & Li, 2003). The resulting aldehydes are highly reactive and cross-link between multiple tropocollagen molecules and elastin to create a flexible but force-resistant network, ultimately resulting in collagen fibril formation. In Menkes syndrome, the reduced ability to distribute copper reduces LO effectiveness, resulting in defective collagen (Kagan & Li, 2003). Lathyrism is another disease that manifests due to LO inhibition, which causes the creation of mechanically weak collagen (Kovanen et al., 1984b). Thus, LO is a critical enzyme for collagen fibril production, while LO inhibition has negative consequences for collagen integrity.
LO mRNA levels increase after exercise in animals (Han et al., 1999a;Heinemeier et al., 2007). For example, LO mRNA levels in rat gastrocnemius increased 10-and 6.5-fold after 4 days of eccentric and isometric exercise respectively but did not change statistically after concentric exercise . Of importance is that force production during the eccentric contractions was greater than during isometric contractions, which was greater than during concentric contractions. Moreover, electrical stimulation causes greater muscle damage than voluntary contractions, which would potentially cause further ECM disruption (Fouré & Gondin, 2021). Thus, it is not immediately clear whether the between-condition differences were unique to contraction mode, reflected muscle tensile force, or were caused by greater muscle and ECM damage. Additionally, LO mRNA levels in rat soleus and RF muscles increased 10-and 46-fold, respectively, 2 days after downhill treadmill running, returning to baseline by 14 days (Han et al., 1999a). These findings suggest that eccentric and isometric contractions, or at least the large muscle tension generated by them, are potent stimuli for this preeminent step in collagen fibril synthesis (Table 3). Interestingly, the large differences between the soleus and RF may be associated with different activation between the two muscles, or with the increased susceptibility of type II myofibres (predominantly in RF) to eccentric exercise-induced muscle damage than type I myofibres (predominantly in soleus) (Fridén & Lieber, 1992), which would subsequently require a greater degree of collagen synthesis for repair. In conclusion, LO mRNA levels increase 48 h post-exercise (Han et al., 1999a;Heinemeier et al., 2007), coinciding with the onset of increased collagen mRNA levels . Based on the available evidence, LO mRNA levels appear sensitive to the magnitude of muscle tension , although exercise volume could be another influential parameter (Han et al., 1999a).
There is no information regarding muscle and biofluid LO responses to exercise, immobilisation, or aging in animals and humans, and more research is required to determine whether LO mRNA levels show similar exercise-induced changes as LO activity. Nonetheless, LO activity may be valuable in quantifying collagen cross-link remodelling (see Section II.2) following altered mechanical loading states.

(d) Matrix metallopeptidases and inhibitors (MMPs and TIMPs)
Matrix metallopeptidases, also widely called matrix metalloproteinases (MMPs), are a zinc-dependent protease family capable of degrading collagens and other ECM components. So far, 23 MMP members have been identified in humans and, along with TIMPs that act as MMP inhibitors, they are vital for regulating collagen molecule degradation (Chen & Li, 2009). A delicate balance of MMP and TIMP activity is required during ECM remodelling, which is essential for embryonic morphogenesis, development, reproduction, tissue remodelling, and resorption (Chen & Li, 2009). Collagenases (MMPs 1, 8, 13, and 18) cleave collagens I, II, and III while gelatinases (MMPs 2 and 9) degrade denatured collagens IV, VII, and X. ProMMPs are MMP precursors, and active MMPs can activate proMMPs, creating positive feedback loops (Chen & Li, 2009). MMPs can degrade a wide range of ECM molecules and existing studies suggest that ECM protein fragments, when degraded and released, can act as soluble ligands (Hynes, 2009). The ECM has been described as a 'reservoir of growth factors', which are bound to the glycosaminoglycan components of proteoglycans (see introduction of Section III) and released during degradation. MMPs can thus serve a dual role, removing damaged structures and simultaneously stimulating the initiation of exercise-induced collagen synthesis. Moreover, MMP and TIMP release into the interstitial space and bloodstream has potential signalling and remodelling effects elsewhere in Biological Reviews 98 (2023)  the body through the action of matricryptins (Ricard-Blum & Ballut, 2011), or by facilitating glycosylated cross-linked collagen degradation. Therefore, muscle or biofluid MMP or TIMP assessments could provide important information relating to ECM breakdown or synthesis, in addition to other important health-related processes.
TIMP-1 demonstrates variable post-exercise blood activity. For example, serum TIMP-1 concentration increased (26%) immediately after 45-min −10 downhill running at 60% of each participant's maximal velocity (Koskinen et al., 2001b) but in other studies, plasma concentrations remained unchanged after 30 min of −10 downhill running at 70% of maximum heart rate (Welsh, Allen & Byrnes, 2014), or after a maximal exercise treadmill test (Tayebjee et al., 2005). Discrepancies also exist following maximal eccentric exercise, as serum TIMP-1 concentration increased (28%) and persisted for several days following 100 maximal eccentric knee extensions (Mackey et al., 2004), but in another study plasma TIMP-1 concentrations were unchanged after 60 eccentric elbow flexions (Madden et al., 2011). Current TIMP-1 assessment methods might not be sufficiently accurate or reliable. Due to the potentially unpredictable responses, assessment of TIMP-1 activity to infer ECM responses to exercise cannot currently be recommended.
Muscle MMP-9 activity has been linked to damage and early inflammatory stages and, as such, post-exercise increases in blood in otherwise healthy animals and humans might be expected to reflect exercise-induced muscle damage and inflammation (Kherif et al., 1999). However, exercise that typically does not cause significant muscle damage, such as endurance exercise, can increase blood MMP-9 levels (Saenz et al., 2006;Suhr et al., 2007Suhr et al., , 2010Reihmane, Jurka & Tretjakovs, 2012;Reihmane et al., 2013;Rullman et al., 2013;Schild et al., 2016). Such increases have been found in human serum and plasma following 50-90 min of bicycling (Suhr et al., 2007;Reihmane et al., 2012;Rullman et al., 2013;Schild et al., 2016), a maximal treadmill incremental test (Suhr et al., 2010), and after a half or a full marathon (Saenz et al., 2006;Reihmane et al., 2013). Nonetheless, no changes existed after a maximal treadmill test (Tayebjee et al., 2005) or 40-60 min of submaximal bicycling (Nourshahi, Hedayati & Ranjbar, 2012;Rocha et al., 2015). Further discrepancies exist, with serum MMP-9 concentration increasing immediately after 30-60 min of downhill running (Koskinen et al., 2001b;Welsh et al., 2014;van de Vyver et al., 2016) and at 8 days after 100 maximal eccentric knee extensions (Mackey et al., 2004), but decreasing immediately after 70 eccentric leg extensions (at 110% of 10 repetition maximum) and persisting for up to 48 h afterward (Nascimento et al., 2016). Furthermore, no changes existed after 60 eccentric elbow flexions at 120% of the concentric maximum (Madden et al., 2011). Overall, plasma or serum MMP-9 concentrations are responsive to exercise, but the patterns of changes are dissimilar between studies using similar exercise protocols. These discrepancies could result from different methods. This is evident in studies using maximal incremental tests and enzyme-linked immunosorbent assay (ELISA) analyses in which increases in serum (Suhr et al., 2010) but not plasma (Tayebjee et al., 2005) MMP-9 concentrations were found. However, even when the same assessment method (ELISA) and medium (serum) were used, MMP-9 levels increased after 60 min of submaximal bicycling in one study (Nourshahi et al., 2012) whilst remained unchanged after 90 min of high-load bicycling in another (Suhr et al., 2007). Therefore, with the currently available methods and practices, plasma or serum MMP-9 concentration might not be reliable to assess acute post-exercise ECM changes.
Unlike MMP-9, increased muscle MMP-2 activity is associated with late-stage myofibre regeneration (Kherif et al., 1999). However, similar to MMP-9, acute increases in serum and plasma concentrations (Suhr et al., 2007(Suhr et al., , 2010Rocha et al., 2015;Schild et al., 2016) and decreases (Nourshahi et al., 2012) immediately after bicycling were reported, a modality which does not typically induce extensive myofibre damage and regeneration. Furthermore, response discrepancies in blood MMP-2 concentrations were observed after eccentric exercise. In one study, serum MMP-2 concentration was unchanged even after 14 days following 100 maximal eccentric knee extensions (Mackey et al., 2004) while in another study 70 eccentric leg extensions decreased plasma MMP-2 concentration up to 24 h postexercise before returning to baseline at 48 h (Nascimento et al., 2016). Others report that serum MMP-2 concentrations remained unaffected for up to 7 days following 45 min of downhill running in either cold (5 C) or room (22 C) temperature (Koskinen et al., 2001b). MMP-2, like MMP-9 and TIMP-1, does not respond uniformly to similar stimuli, and other factors might affect the blood kinetics.
Factors other than exercise type or contraction mode could underpin the response discrepancies of biofluid MMP and TIMP concentrations. Acute increases in biofluid MMP concentration observed immediately post-exercise might be linked to leucocyte mobilisation and caused by factors other than ECM remodelling (Koskinen et al., 2001b;Mackey et al., 2004). Nonetheless, data from some of the aforementioned studies suggest that biofluid MMP-2 and -9 concentrations are more responsive to large numbers of either concentric or eccentric contractions, typically performed during bicycling or downhill running, respectively, rather than by mechanical strain or muscle damage per se (Mackey et al., 2004). Additionally, it appears that the activation of larger muscle groups, such as those in the lower limbs, induces greater increases than in smaller muscles of the arms since no changes in MMP-9 or TIMP-1 existed after 60 eccentric elbow flexions (Madden et al., 2011). Nevertheless, as mentioned previously, biofluid MMP and TIMP assessments display variable responses regardless of whether the same assessment method or different methods [e.g. ELISA versus zymography (Nourshahi et al., 2012;Rocha et al., 2015)] are used. By contrast, the sampling site (vein versus artery) does not appear to affect the results (Rullman et al., 2013). Based on these findings, blood MMP and TIMP concentration assessments are not currently recommended as reliable ECM remodelling biomarkers. Further studies in humans and animals are required to clarify the circumstances under which these indicators are affected by exercise mode and by the muscles involved. Urine has not been assessed for these biomarkers, but speculatively, as a blood-originating biofluid, it is most likely unsuitable.
Acute exercise can affect MMP levels in muscle, as 65-min of bicycling increased MMP-9 levels in VL at 120 min postexercise (125%) in men (Rullman et al., 2007). Moreover, acute muscle damage provokes increases in muscle MMP and TIMP activities in animals. For example, MMP-9 activities in EDL and TA muscles were increased during the early inflammatory phase following experimental cardiotoxininduced muscle necrosis (Kherif et al., 1999). MMP-9 activity increased at 24 h and persisted for up to 3 days in that study (Kherif et al., 1999), whereas in another study an increase in rat TA MMP-9 activity was detected only after 4 days of eccentric exercise (Koskinen et al., 2002). This different response probably results from extensive acute muscle necrosis in the former study compared with mild tissue damage from downhill running. However, although studies of rat muscle show that MMP, proMMP, and TIMP activities increase consistently for at least 7 days following eccentric exercise (Koskinen et al., 2001a(Koskinen et al., , 2002, no changes in serum MMP-2 or −9 activities existed after 100 maximal eccentric knee extensions in humans (Mackey et al., 2004). Repeated resistance exercise sessions (containing eccentric contractions) evoke a smaller remodelling response (indicated by no changes in muscle MMP-2 and MMP-9 activities) in rats (Ogasawara et al., 2014), potentially affected by the eccentric-induced muscle damage and function-protective phenomenon known as the repeated bout effect (Lapier et al., 1995;Hyldahl et al., 2015). These data suggest that muscle MMP activity is sensitive to eccentric loading, at least in animals. In animals and humans, increases in muscle MMP-2 and -9 activities are attenuated after repeated eccentric exercise, possibly due to reduced muscle damage and the need for ECM remodelling.
Altered muscle MMP activities suggest strong ECM remodelling responses after long-term exercise in humans and animals (Carmeli et al., 2005;Rullman et al., 2007;Scheede-Bergdahl et al., 2014). Rats that underwent 2 weeks of daily treadmill running training at low (50% of VO 2max ) and high exercise intensities (70-75% of VO 2max ) displayed no changes in MMP-9, while MMP-2 levels increased only in gastrocnemius (high intensity 156%, low intensity 62%) and RF muscles (high intensity only, 160%) but remained unchanged in soleus and vastus intermedius (Carmeli et al., 2005). In another study examining the effects of exercise intensity (8 weeks of loaded ladder climbing; body weight versus 75% overload of rat's weight increased to exhaustion), only the high-intensity condition increased MMP-2 activity (2-fold) in the TA (Deus et al., 2012). Similar changes were observed in rat gastrocnemius after 8 weeks (5 per week) of resistance exercise (loaded climbs) where, despite no change in MMP-9 activity, MMP-2 increased (1.7-fold) only in the group that performed double the volume (8 climbs of 1.1 m, with 50, 75, 90 and 100% of the animal's maximal carrying capacity) of the other resistance-training group (4 climbs, same height and weight) or non-exercising controls (de Sousa Neto et al., 2017). In humans, long-term exercise training using 45-min knee extensions (4 per week; 5 weeks) increased MMP-2 activity (5-fold) after 10 days in the VL muscle of both the freely exercising leg and the leg subjected to blood-flow restriction, but both conditions were only 3-fold greater than baseline at 5 weeks (Rullman et al., 2009). In the same study, MMP-9 activity in VL muscle increased 7.5-fold immediately after one exercise session when blood flow was restricted, but only 3.5-fold in the freely exercising leg, however, that difference was not statistically significant, and no differences existed after 10 days or 5 weeks of exercise. Based on these findings, changes in muscle MMP-9 and -2 activities are more aligned with the hypothesis that muscle MMP-9 levels increase acutely whilst MMP-2 levels increase only after long-term exercise training (Kherif et al., 1999;Rullman et al., 2009), in contrast to biofluid responses. Interestingly, increased MMP-2 levels in rats developed only in muscles that are predominantly composed of type II myofibres, such as gastrocnemius and RF, rather than those with larger proportions of type I myofibres such as soleus and vastus intermedius (Ariano, Armstrong & Edgerton, 1973). Higher running intensities and greater resistance exercise volumes in animals induce greater increases in muscle MMP-2 but numerous underlying factors could affect this increase (e.g. muscle tension, oxidative stress, vascularisation, etc.). The lack of differences between the blood flow-restricted and freely exercising limbs along with the greater increases after resistance exercise at higher intensities suggest that increased MMP activity, and by extension ECM remodelling, may be more associated with mechanical strain rather than metabolic stress. Immobilisation strongly affects animal and human muscle MMP activities. Increases in proMMP-2 (1.4-fold) and MMP-2 (2.5-fold) activities in rat soleus developed after 3 days of immobilisation, but only when the muscles were held in shortened positions, while longer lengths did not cause changes versus controls (Ahtikoski et al., 2003). Short immobilisation periods in humans, however, do not elicit significant ECM remodelling, as 48 h of knee immobilisation did not affect MMP-28 or TIMP-1 levels in VL muscle (Urso et al., 2006). Previous research has established that longer immobilisation periods increase MMP activity, as proMMP-2 (2-fold) and TIMP-2 (2-fold), but not proMMP-9 or TIMP-1 activities, were higher in the VL muscle of people with spinal cord injury versus non-injured controls (Koskinen et al., 2000), although one study found significant decreases in MMP-2 activity in rat spinal muscles after 21 days of immobilisation compared to controls (Huang et al., 2018). Passive movement of otherwise-immobile limbs attenuates increased MMP levels, as indicated by the lower MMP-1-positive cell percentage in arthritic rabbit knees exposed to continuous passive motion for 24 h (26% versus 79%) or 48 h (24% versus 84%) versus fully immobilised knees  (Ferretti et al., 2005). This suggests that immobilisation increases muscle MMP activity (after 2 days), while long muscle lengths during immobilisation inhibit this increase. In addition, both active and passive movement can protect against non-optimal ECM remodelling during immobilisation.
Apart from direct enzyme activity assessment, muscle MMP and TIMP synthesis can be assessed indirectly through mRNA level analysis. Changes in muscle MMP mRNA levels appear as reliable for ECM remodelling assessment as activity levels. For example, rat medial gastrocnemius and its tendon showed an up to a 3-fold increase in MMP-2 mRNA levels after 4 days of electrically evoked isometric and eccentric but not concentric contractions . In the same study, increased TIMP-1 and TIMP-2 mRNA levels manifested after exercise of all contraction modes but eccentric contractions increased muscle mRNA levels (60-and 4.5-fold, respectively) significantly more than concentric contractions (18-and 2.5-fold, respectively). However, force production during eccentric contractions was greater than in isometric contractions, which was greater than in concentric contractions. It is therefore not immediately clear whether between-condition differences were contraction mode dependent due to different muscle tension or due to the greater muscle and ECM damage after electrically stimulated versus voluntary contractions (Fouré & Gondin, 2021). Nevertheless, other studies have consistently shown increased TIMP mRNA levels in animal muscles after different forms of eccentric exercise that last up to 7 days (Koskinen et al., 2001a(Koskinen et al., , 2002. Although rat MMP-2 mRNA levels were unchanged after eccentric exercise in the first 24 h, TIMP-1 and -2 mRNA increased as soon as 6 h, while MMP-2, TIMP-1, and TIMP-2 mRNA levels peaked at around 2 days after and remained elevated for up to 7 days after (Koskinen et al., 2001a(Koskinen et al., , 2002. Higher exercise intensities (3.1-fold), even during activities such as treadmill running, induce greater increases in rat quadriceps MMP-2 mRNA levels than lower intensities (2.4-fold) (Carmeli et al., 2005). In animals and humans, increased muscle MMP-9 mRNA levels coincide with an early post-exercise inflammatory phase (Rullman et al., 2007(Rullman et al., , 2009Hoier et al., 2012;Guzzoni et al., 2018) while increased MMP2 mRNA levels are related to prolonged remodelling (Koskinen et al., 2001a(Koskinen et al., , 2002Carmeli et al., 2005;Scheede-Bergdahl et al., 2014;Guzzoni et al., 2018). These trends are in general agreement with the muscle MMP activity findings.
Following 3 days of immobilisation, MMP-2 mRNA levels increased (2.3-fold) in rat soleus; however, in contrast to findings in MMP muscle activity, and although immobilisation at longer lengths still induced increases (1.5-fold), it was significantly lower than in the shortened position (Ahtikoski et al., 2003). Additionally, while 48 h of knee immobilisation did not change MMP-28 or TIMP-1 muscle activities in human VL muscle, mRNA levels decreased (1.9-and 1.8 fold, respectively) (Urso et al., 2006). The discrepancies between muscle MMP activities and mRNA levels cannot be explained by the available data, but it is well known that increased mRNA levels do not always translate to increased protein concentrations or activities (Czerwinski, Martin & Bechtel, 1994;Vogel & Marcotte, 2012).
Based on the above findings, blood MMP and TIMP concentrations can produce different results in muscle activities. MMP, proMMP, and TIMP activities and mRNA levels are sensitive for quantifying muscle ECM remodelling after different exercises. High-tension exercises, such as resistance and eccentric exercise, can be equally effective in increasing biofluid (Table 4) and muscle (Tables 2 and 4) MMP activities. Immobilisation increases muscle MMP activity and synthesis (mRNA levels), however long muscle lengths during immobilisation may inhibit the increase. Thus, muscle MMP activity appears reliably to quantify both acute and chronic ECM remodelling but unfortunately requires invasive procedures.

VI. FUTURE PERSPECTIVES
The results from the studies included in this review suggest that completion of relatively few exercise sessions, and perhaps even passive movement (Ferretti et al., 2005), can evoke substantive adaptations in muscle ECM composition and health. Conclusions can also be drawn regarding how different exercise interventions and programming parameters (e.g. intensity, volume, frequency) influence the human muscle ECM. However, definite conclusions cannot be drawn due to the limited number of human studies and small number of biomarkers tested concurrently. Well-controlled, randomised trials are therefore needed that assess multiple markers and compare across several exercise modalities, examining acute (during and in the days after exercise) and long-term (over multiple sessions) effects.
Due to the difficulties associated with obtaining human tissue samples, animals, and especially rodents, are widely used to study collagen synthesis and ECM remodelling. Although non-human mammalian physiology is similar to that of humans, important differences must be considered before strong conclusions are drawn. Rodents have different metabolic rates to humans, and the findings of animal studies may not always represent human post-exercise responses. For example, it has been shown that the number of HPyr cross-links in cartilage and tendon (and hydroxypyridinium cross-links in the lung) decrease with aging in humans but increase in rats (Moriguchi & Fujimoto, 1978). This may result from the shorter lifespan or the different protein synthesis rates of short-lived animals. Other factors that have recently been highlighted are the different temperature requirements of rodents compared to humans, nocturnal lifestyle, need to eat constantly, greater reliance on hepatic glycogen to fuel exercise, and the use of stressful stimuli to force rodents to exercise (Fuller & Thyfault, 2021). The fasting blood glucose and free fatty acid profiles of rats are closer to human profiles than are those of mice, which have twice Biological Reviews 98 (2023)  the circulating levels (Fuller & Thyfault, 2021), suggesting that rats may better replicate human physiology. Nonetheless, inferences can be made from animal studies where the mechanisms are understood by data from human studies.
Assessment of muscle and tendon ECM components through the analysis of biopsy samples (in humans) or dissected tissues (in animals) is a more direct and informative method than biofluid analysis of studying changes within a region of interest. However, differences in biopsy depths, muscles or muscle regions yield different results Mackey et al., 2004) and complicate data interpretation. Human muscle biopsies are invasive and access to such biopsies can be very limited in some countries. Additionally, only a limited range of muscles are routinely biopsied with the VL muscle (in the leg) being the most widely used. This contrasts with a huge range of muscles available for analyses in mice of both sexes at all ages. In studies where human muscle biopsies can be obtained, reliable indicators of ECM status such as collagen staining and FSR could be used as 'gold standards' against which biofluid markers could be validated (and calibrated), enabling testing of reliability and tissue-source effects (e.g. bone, muscle, tendon, ECM). Biomarker validation against collagen staining and FSR could be extensively performed using animal models before confirmation in humans, which would provide the additional benefit of validating the use of animal models. Until these comparative 'gold standard' validation studies are conducted, care must be taken when making inferences as the same mechanical stimulus might produce different results in different tissue samples (e.g. blood versus muscle) (de Sousa Neto et al., 2017).
Validation and calibration using 'gold standard' biomarkers for muscle ECM breakdown and synthesis would boost confidence in the analyses and interpretation of results as well as determine the contribution of fragments from nonmuscle tissues. The results would permit well-powered human studies to determine how muscle ECM adapts to exercise, aging, and disuse. Indeed, finding reliable biofluid markers for muscle ECM synthesis and breakdown will be an important step in understanding human responses to exercise and disease; this should be prioritised in research since dependence on biopsy sample analysis restricts the number of large human trials. In general, the interpretation of results from muscle mRNA analyses is complicated and may produce inconsistent results and lead to incorrect conclusions, especially when extrapolating to long-term effects, because of the impact of post-transcription and post-translation processes as well as individual variation in translation efficiency, which obscures the mRNA-protein relationship (Czerwinski et al., 1994;Vogel & Fig. 6. Schematic summary of the course of skeletal muscle extracellular matrix (ECM) remodelling following loading, mechanical stress, or disruption. CGTF, connective tissue growth factor; ECM, extracellular matrix; GGT, galactosylhydroxylysyl glucosyltransferase; TGF, transforming growth factor. Marcotte, 2012). Nonetheless, mRNA analyses may be useful trend indicators for short-term investigations.
Researchers should be aware that apart from quantitative assessment, there is also a qualitative aspect to ECM status and health that reflects real-world impacts on quality of life. Alterations in a muscle's collagen cross-link profile (i.e. cross-linking through mature trivalent cross-links versus non-enzymatic glycation-induced AGEs; see Section II.2) can substantially reduce muscle elasticity, tensile integrity, and force transmission, as well as increasing chronic inflammation (Haus et al., 2007;Olson et al., 2021). Moreover, changes in collagen fibre organisation, alignment, and packing conditions following changes in muscle loading (Oakes et al., 1982;Michna, 1984;Vilarta & Vidal, 1989;Coutinho et al., 2006), can influence the muscle's physiological response to different stimuli. For example, both muscle ECM immune cell content (i.e. inflammation) and collagen organisation are significantly associated with muscle growth in response to progressive resistance training in older individuals (Long et al., 2022). Therefore, due to considerable effects on both muscle properties and physiological responses, qualitative assessments of collagen fibre organisation, alignment, packing conditions, and cross-link profiles in muscle ECM in response to physical activity, disuse, and aging are needed, apart from the quantitative assessment of protein concentration or enzyme activity changes. Fig. 7. Candidate extracellular matrix biomarkers for assessment of alterations in response to acute and long-term exercise, disuse, and aging in human studies. Biomarkers above the 'worth-it line' have the strongest potential or reliability. Green background indicates muscle damage and collagen breakdown markers; orange indicates structural markers; blue indicates collagen-synthesis markers. The muscle symbol refers to the assessment of tissue levels; the tube symbol refers to the assessment of biofluid levels; the mRNA symbol refers to the assessment of mRNA in tissues; the mouse symbol indicates notable biomarkers not yet investigated in humans, but reliable when assessed in rodent tissue, and recommended for future research. COL, collagen; CTGF, connective tissue growth factor; DEC, decorin; FBN, fibronectin; FIBRO, fibroblasts; FSR, fractional synthesis rate (collagen); GGT, galactosylhydroxylysyl glucosyltransferase; HPyr, hydroxylysylpyridinoline; HYP, hydroxyproline; ICTP, carboxyterminal telopeptide region of type I collagen; INT, integrin; LO, lysyl oxidase; MMP, matrix metallopeptidase; P3NP, Ν-terminal propeptide of procollagen type III collagen; PH, prolyl 4-hydroxylase; PICP, C-terminal propeptide of procollagen; PINP, Nterminal propeptide of procollagen; TGF-β, transforming growth factor β; TIMP, tissue inhibitor of metallopeptidases; TNC, tenascin-C.
Biological Reviews 98 (2023)  Another focus for future research should be on determining the most effective exercise interventions for stimulating the breakdown of non-optimal cross-linked collagen caused by age, disease, and disuse. In addition to guiding exercise-therapy programmes, this will help to clarify exercise-induced health benefits. Responses to different exercise modalities could be assessed by analyzing HPyr and AGE in biofluids. However, as discussed herein, those cross-links can originate from bone or other connective tissues. Therefore, future studies should explore ways to specify their tissue of origin. Moreover, exercise-induced increased expressions of different muscle integrin isoforms, which are increasingly being recognised as important for muscle signalling and protection, provide clinically relevant information regarding the adaptive potential of muscles for rehabilitation (e.g. tissue structural integrity, and increased muscle hypertrophy) and disease (e.g. muscular dystrophy and fibrosis).
Care should be taken when interpreting experimental results. In addition to issues relating to the effects of blood sampling sites (i.e. artery or vein; Rullman et al., 2013) and the usefulness of animal models, significant differences exist between enzyme-specific or total activity (or simply, activity) outcomes (Savolainen et al., 1987(Savolainen et al., , 1988a as well as between protein concentration or total content (Savolainen et al., 1987(Savolainen et al., , 1988aKarpakka et al., 1990a), which could lead to different conclusions being reached. To mitigate this, researchers should report both specific and total activities as well as both concentration and total content, whenever possible. Finally, measurements should be taken at baseline before interventions are started to control for the possibility of genetic variations and undiagnosed diseases influencing concentrations in biofluids.

VII. CONCLUSIONS
(1) A continuous, delicate, and finely regulated balance of enzymatic activity and protein expression is necessary for optimal muscle health and function in resting conditions.
(2) Even a single exercise session can evoke a powerful and dynamically orchestrated time course of changes resulting in substantive improvements in muscle ECM biomechanical and physiological properties when repeated over time (Fig. 6).
(3) Both physical activity and disuse significantly influence the skeletal muscle ECM. Physical activity, particularly endurance (aerobic) and eccentric exercises (Tables 2-4), are especially potent in inducing favourable remodelling.
(4) Disuse causes non-optimal changes including muscle connective tissue proliferation. However, maintaining the muscle at a longer length during immobility can attenuate some of the negative effects as seen through most of the biomarkers explored in this review, possibly through greater activation of mechanosensing components, and passive movement also appears to be beneficial.
(5) Promising biofluid markers for ECM remodelling, ECM status, and responses to exercise, disuse, or aging, are shown in Fig. 7. Collagen FSR appears to be the most consistent biomarker for assessing acute exercise-induced collagen synthesis in humans, whereas collagen IV and Hyp appear reliable in assessing acute exercise-induced collagen breakdown. (6) When assessing tissue samples, integrin-α7, TNC, and collagen IV can reliably assess acute exercise-induced changes in the basal membrane, ECM-to-cell adhesion, and the collagen scaffold itself, respectively. Hyp and PH are tissue markers that appear reliable for the assessment of long-term exercise-induced collagen synthesis, whereas integrin-β1 increases consistently with long-term exercise, and indicates changes in cell-to-ECM anchorage. (7) No reliable human biofluid markers currently exist for the assessment of long-term ECM adaptations, however, research should focus on promising biofluid markers such as P3NP and TGF-β for assessing collagen synthesis, and HPyr for structural changes in collagen profile. ECM markers for disuse and aging require more research in humans, but PH and GGT seem very promising indicators of collagen synthesis changes in animal muscles. Nevertheless, muscle ICTP, PINP, and MMP2 for disuse, and GGT for aging hold potential as human biomarkers of collagen synthesis to be explored further in research.

VIII. ACKNOWLEDGEMENTS
Dr Jacob Sorensen, Dr Abigail Mackey, adnd Dr Robert Hyldahl are greatly appreciated for providing high-resolution versions of their unpublished figures to be included in this review. The authors declare that they have no conflicts of interest. No funding was received for this work. Open access publishing facilitated by Edith Cowan University, as part of the Wiley -Edith Cowan University agreement via the Council of Australian University Librarians.