Diversity, Structures, and Collagen-Degrading Mechanisms of Bacterial Collagenolytic Proteases

Abstract

Bacterial collagenolytic proteases are important because of their essential role in global collagen degradation and because of their virulence in some human bacterial infections. Bacterial collagenolytic proteases include some metalloproteases of the M9 family from Clostridium or Vibrio strains, some serine proteases distributed in the S1, S8, and S53 families, and members of the U32 family. In recent years, there has been remarkable progress in discovering new bacterial collagenolytic proteases and in investigating the collagen-degrading mechanisms of bacterial collagenolytic proteases. This review provides comprehensive insight into bacterial collagenolytic proteases, especially focusing on the structures and collagen-degrading mechanisms of representative bacterial collagenolytic proteases in each family. The roles of bacterial collagenolytic proteases in human diseases and global nitrogen cycling, together with the biotechnological and medical applications for these proteases, are also briefly discussed.

INTRODUCTION

Collagen is the most abundant and ubiquitous material that occurs in the extracellular matrices of animals. Nearly 30% of the total proteins in mammalian bodies are various types of collagens, which are essential structural components of all connective tissues (1). Collagen serves as an important nitrogen source in the global nitrogen cycle, primarily due to its tremendous distribution in the animal kingdom all over the world. Collagens are a large family consisting of 28 different members, which share a helical structure made up of three polypeptide chains (1). The polypeptide chains are mostly composed of repeating Gly-X-Y triplets, where X and Y are frequently proline and hydroxyproline, respectively. Each polypeptide chain is composed of triple-helical domains, which are flanked by nonhelical regions (2). Nonhelical regions are present in all procollagens (1). Type I collagen has one major helical domain, but there are 21 to 26 interruptions in the helix structure of type IV collagen (3). For type I and III collagens, a large part of each nonhelical region in procollagens, which is called propeptide, is proteolytically removed when procollagens are processed into mature molecules, and the remaining parts at the N and C termini of each mature collagen molecule are called telopeptides, which are capable of forming intramolecular and intermolecular cross-links to enhance the tensile strength of collagen fibrils (4). In type VI collagen, the C-terminal nonhelical domain is kept during posttranslational processing (5).

Collagens can be classified into several subgroups when they are organized into supramolecular assemblies. Fibril-forming collagen, such as type I or III collagen, forms the structural basis of connective tissues in all mammals. The molecules of this type of collagen axially polymerize to form a microfibril, and then further aggregate structures, fibrils and fascicles, are successively formed (6). The individual molecules in fibrillar collagens are staggered from each other by approximately 67 nm, allowing covalent bonds with adjacent molecules that extend beyond their individual ends (7). Type IV collagen, a major constituent of mammalian basement membranes, represents another type of supramolecular organization which forms three-dimensional networks (3). Other major subgroups include fibril-associated collagens with interrupted triple helices, beaded-filament-forming collagens, anchoring fibril collagens, and transmembrane collagens (8).

Because of its special structure, collagen is resistant to most common proteases and can be degraded only by a few types of proteases from mammals or bacteria. These proteases that can degrade one or more types of collagens are regarded as collagenolytic proteases, and these include mammalian matrix metalloproteinases (MMPs), mammalian cysteine proteases, and some bacterial proteases. MMPs have been widely studied for their importance in mammal development and for their involvement in human diseases, and they have been introduced in other reviews (9–11). Bacterial collagenolytic proteases include some metalloproteases and serine proteases. The bacterial collagenolytic metalloproteases reported so far are some members of the M9 family in the MEROPS database (http://merops.sanger.ac.uk) (12), which are Clostridium collagenases and Vibrio collagenases. Bacterial collagenolytic serine proteases are distributed in the S1, S8, and S53 families. In addition, some members of the U32 family, mainly from pathogenic bacteria, are also reported to be collagenolytic proteases.

In recent years, there has been remarkable progress in identifying new bacterial collagenolytic proteases and their degradation mechanisms on collagen. In this paper, a comprehensive introduction of various bacterial collagenolytic proteases is presented, with a focus on the structures and collagen-degrading mechanisms of representative bacterial collagenolytic proteases. In addition, the roles of bacterial collagenolytic proteases in human diseases and global nitrogen cycling, as well as their biotechnological and medical applications, are also briefly discussed.

Before further discussing bacterial collagenolytic proteases, it is necessary to discuss the relationship between bacterial collagenolytic proteases and bacterial collagenases to avoid confusion. Bacterial collagenases are usually considered to be enzymes that cleave helical regions of fibrillar collagen molecules under physiological conditions (13). However, this concept of bacterial collagenase is not always used in such a strict standard in publications. Bacterial proteases which can only hydrolyze denatured collagen or type IV collagen are sometimes called collagenases. Here, we suggest that the term “bacterial collagenolytic proteases” be used to refer to all proteases that can degrade at least one type of collagen. Bacterial collagenolytic proteases include true collagenases and other proteases with collagenolytic activity.

CLOSTRIDIUM COLLAGENASES FROM THE M9 FAMILY

Diversity.

The collagenase secreted by Clostridium histolyticum reported first by MacLennan et al. in 1953 was the first bacterial collagenase to be discovered and studied (14). Since then, the collagenases from Clostridium histolyticum have been studied in detail and extensively used in basic science laboratories and for medical treatment. These Clostridium collagenases are divided into two classes, I and II, with class I being encoded by the gene colG and class II by colH. The C. histolyticum genome contains both genes. colG encodes a 1,118-amino-acid precursor, and colH encodes a 1,021-amino-acid precursor. ColG is secreted as a 1,000-amino-acid protein (Y119 to K1118), and ColH is secreted as a 981-amino-acid protein (V41 to R1021) (15–17). In addition to collagenases ColG and ColH, at least five gelatinases were identified from C. histolyticum (16, 17). The N-terminal amino acid sequences of these gelatinases coincide with that of either ColG or ColH, indicating that these gelatinases are produced from a full-length collagenase by the proteolytic removal of C-terminal fragments (17). Collagenases from other Clostridium species, such as ColA from Clostridium perfringens (18) and ColT from Clostridium tetani (19), have also been reported. All Clostridium collagenases are members of the M9B subfamily in the MEROPS database (12).

Structure.

A Clostridium collagenase molecule is composed of two parts: the N-terminal collagenase unit (or module) and the C-terminal recruitment domains (20, 21). The collagenase unit contains an N-terminal activator domain and a C-terminal peptidase domain that contains a conserved zinc-binding motif and functions as a catalytic domain. The collagenase unit is able to independently show collagenolytic activity (20). The recruitment domains usually contain one or two collagen-binding domains (CBDs) and one or two polycystic kidney disease (PKD)-like domains, but ColT, which has no PKD-like domain, is an exception (18–21). The recruitment domains are not needed for degradation of triple-helical collagen but are most likely needed for larger collagen entities such as fibrils (20, 22). Clostridium collagenases usually show differences in their C-terminal recruitment domains and/or their zinc-binding motifs (Fig. 1).

FIG 1.

Schematic diagrams of the domain organization of typical collagenolytic proteases from bacteria. The domain structures of mature clostridial collagenases ColG (BAA77453), ColH (BAA34542), and ColT (AAO37456) were drawn with reference to a paper by Eckhard et al. (21), and the others were deduced from the sequences of mature ColA (BAA02941) secreted by Clostridium perfringens, VMC (AAC23708) from Vibrio mimicus, VAC (CAA44501) from Vibrio alginolyticus, VHC (BAK39964) from Grimontia (Vibrio) hollisae, SOT (BAI44325) from Streptomyces omiyaensis, MO-1 (BAF30978) from Geobacillus collagenovorans MO-1, MCP-01 (ABD14413) from Pseudoalteromonas sp. SM9913, myroicolsin (AEC33275) from Myroides profundi D25, kumamolisin-As (BAC41257) from Alicyclobacillus sendaiensis, and PrtC (AAA25650) from Porphyromonas gingivalis. Both Clostridium and Vibrio collagenases are gluzincins with Glu as the third proteinaceous zinc ligand.

Although the overall structure of Clostridium collagenases is still not completely known, the crystal structures of the collagenase unit of ColG (20), the peptidase domains of ColH and ColT (21), the PKD-like domains of ColG and ColH (22, 23), and the CBDs of ColG and ColH (24, 25) have been solved. The N-terminal activator domain and the C-terminal peptidase domain in the collagenase unit of ColG form a saddle-shaped architecture. The activator domain, the left saddle flap, comprises an array of 12 parallel α-helices, and the peptidase domain, the right saddle flap, adopts a thermolysin-like peptidase fold (Fig. 2) (20). In addition to an active zinc ion, a calcium ion is also seen in the active site of Clostridium collagenases. Both ions are required for the full enzymatic activity (21). The three PKD-like domains from ColG and ColH are structurally similar, adopting a V-set Ig-like fold. Each PKD-like domain contains a calcium ion, which is likely involved in interdomain alignment and aids the domain stability. However, some structural differences between the ColG and ColH PKD-like domains have also been found. There are surface aromatic residues in the ColH PKD-like domains but not in the ColG PKD-like domain (23). The CBDs from ColG and ColH adopt a β-sheet sandwich fold, in which a collagen-binding cleft is seen. Two calcium ions are coordinated in a CBD molecule, and these are critical for maintaining the function and stability of the CBDs (24, 25). It has not yet been determined whether the CBD can function as a helicase to unwind the collagen triple helix.

FIG 2.

Structures of ColG (M9), ColH (M9), MCP-01 (S8), and kumamolisin-As (S53). For ColG, the activator domain and the peptidase domain of the collagenase unit (PDB accession number 2Y50) (20) are colored in purple and gray, respectively, and the PKD (PDB accession number 4AQO) (21) and the CBD (PDB accession number 1NQD) (24) are colored in yellow and orange, respectively. For ColH, the peptidase domain (PDB accession number 4AR1) (21), the PKD (PDB accession number 4U7K) (23), and the CBD (PDB accession number 3JQW) (25) are colored in gray, yellow, and orange, respectively. The catalytic domains of both MCP-01 (PDB accession number 3VV3) (65) and kumamolisin-As (PDB accession number 1SN7) (72) are shown in gray. The catalytic triads of serine collagenolytic proteases MCP-01 (Asp49, His104, and Ser269) and kumamolisin-As (Ser278, Glu78, and Asp82) are shown in stick representation. The Zn²⁺ in ColG and ColH is shown in orange, and Ca²⁺ is colored in green for all enzymes.

Collagenolytic mechanism.

The class I and II collagenases are true endopeptidases. ColG is suggested to have two states for collagen hydrolysis. The activator domain and the catalytic domain in the collagenase module remain mostly closed during collagen cleavage but relax to the open ground state once the collagen is degraded (20). Unlike MMPs, which cleave native collagens into characteristic three-quarter and one-quarter fragments, the collagenases from Clostridium histolyticum hydrolyze native collagens into a mixture of small peptides (26, 27). The class I and II collagenases initially attack distinct hyperreactive sites at Y-Gly bonds in the repeating Gly-X-Y collagen sequence (27). The amino acid preference at P1 to P3 and P1′ to P3′ of both classes of collagenases has been determined with synthetic peptides (28, 29), and this has been comprehensively described by Van Wart (30). In recent years, new materials and techniques, such as immobilized positional peptide libraries and proteomic identification of protease cleavage sites, have been used to study the substrate specificity of Clostridium collagenases, and new profiles of the prime and nonprime cleavage site specificities and their molecular bases have been illustrated (31, 32). The two classes of collagenases bind to different portions of the collagen and have different specificities for cutting native collagens, showing synergy in collagen degradation (33, 34).

VIBRIO COLLAGENASES FROM THE M9 FAMILY

Diversity and structure.

The extracellular proteases from Vibrio spp. have been widely studied because of their important roles in virulence. Vibrio extracellular proteases are divided into three classes based on their structure and function (35, 36). The class I proteases, which are members of the M4 family, are not discussed in this article, because they have no collagenolytic activity. The proteases in class II and class III all have collagenolytic activity and are all members of subfamily M9A in the MEROPS database (12, 36). However, the enzymes in these two classes have significant differences in structure and function (Fig. 1 and Table 1). Notably, the class II proteases have a zinc-binding motif (HEYTH), contain no C-terminal domain, and cannot hydrolyze casein (37, 38), while the class III proteases have a zinc-binding motif (HEYVH), contain a PKD-like domain and a prepetidase C-terminal (PPC) domain at their carboxyl terminus, and can hydrolyze casein (36, 39, 40). However, the collagenase from Grimontia hollisae 1706B, which was first named Vibrio hollisae 1706B in 1982 (41) and reclassified in 2003 (42), is an exception among the class III proteases (43). Although it has the same zinc-binding motif as the other class III members, this enzyme does not have a C-terminal PKD-like domain and cannot hydrolyze casein. It has been suggested that it represents a new group of Vibrio collagenases (43). Thus far, no crystal structure of a Vibrio collagenase or its domain has been solved.

TABLE 1.

Characteristics of representative collagenolytic proteases

Family	Class	Name	Origin	Mol mass (kDa)	Motif/catalytic triad	GenBank accession no.	Substrates
M9B	I	ColG	Clostridium histolyticum	116	HEYTH	BAA77453	Type I, II, and III collagens (15, 28)
	I	ColA	Clostridium perfringens type C NCIB 10662	116	HEFTH	BAA02941	Type I collagen, Pz peptide, azocoll (18)
	II	ColH	Clostridium histolyticum	98	HEYTH	BAA34542	Type I, II, and III collagens (16)
M9A	II	VMC	Vibrio mimicus ATCC 33653	62	HEYTH	AAC23708	Type I, II, and III collagens, gelatin, Cbz-GPLGP, Cbz-GPGGPA (37)
	II	PrtVp	Vibrio parahaemolyticus 93	62	HEYTH	CAA86734	Type I, II, III, and IV collagens, FALGPA (38)
	III	VAC	Vibrio alginolyticus	82	HEYVH	CAA44501	Gelatin, casein, collagen, synthetic substrate (39)
	III	VPPC	Vibrio parahaemolyticus 04	90	HEYVH	AAG59883	Type I collagen, gelatin, casein, Cbz-GPGGPA (36)
	III	VPM	Vibrio parahaemolyticus FYZ8621.4	90	HEYVH	ABF19104	Type I, II, III, and IV collagens, gelatin, casein, Cbz-GPGGPA (40)
S1		SOT	Streptomyces omiyaensis	23	His75, Asp120, Ser210	BAI44325	Type I and IV collagens (50)
		SGT	Streptomyces griseus ATCC 10137	23	His73, Asp118, Ser208	AAA26820	Type I collagen (50)
S8			Geobacillus collagenovorans MO-1	105	Asp196, His260, Ser590	BAF30978	Type I and IV collagens (52)
		MCP-01	Pseudoalteromonas sp. SM9913	66	Asp49, His104, Ser269	ABD14413	Type I, II, and IV collagens, gelatin, casein, Pz peptide (63)
		Myroicolsin	Myroides profundi D25	56	Asp129, His174, Ser378	AEC33275	Type I collagen, gelatin (66)
S53		Kumamolisin-As	Alicyclobacillus sendaiensis NTAP-1	37	Glu78, Asp82, Ser278	BAC41257	Gelatin, relaxed collagen (69)
U32		PrtC	Porphyromonas gingivalis ATCC 53977	70 (dimer)		AAA25650	Reconstituted type I collagen, heat-denatured type I collagen, azocoll (78)

Collagenolytic mechanism.

Vibrio collagenases cleave triple-helical collagen in a manner similar to MMPs in the first step of degradation. They attack at a point three-quarters of the way from the N terminus by hydrolyzing the preferential peptide bond, Xaa—Gly. Vibrio collagenase cleaves native collagen at a much higher rate than vertebrate MMP1 (44, 45). Vibrio collagenase hydrolyzes the Pz (4-phenylazobenzyloxycarbonyl) peptide (Pz-Pro-Leu-Gly-Pro-d-Arg) by cleaving the same peptide bond (Leu—Gly) as Clostridium collagenase (44, 45). Because the crystal structure of Vibrio collagenase is not solved, it is still unclear how Vibrio collagenase recognizes and degrades collagen. Although the class II proteases do not contain a C-terminal extension, it is reported that the FAXWXXT motif in the carboxyl terminus of Vibrio mimicus collagenase is involved in binding to collagen (46). In contrast, neither the PKD-like domain nor the PPC domain of the class III proteases is yet demonstrated to function as a collagen-binding domain.

BACTERIAL SERINE COLLAGENOLYTIC PROTEASES FROM THE S1 FAMILY

Of the reported collagenolytic serine proteases in the S1 family, most are from invertebrate animals. The protease isolated from a fiddler crab (Uca pugilator) was the first serine protease found to be capable of cleaving native type I triple-helical collagen. The crystal structure of this protease was solved in 1997, and structural analysis showed that it has no collagen-binding domain and that the substrate-binding pocket is enlarged relative to that of trypsin, owing to the insertion of several amino acids (47). Its cleavage pattern on collagen is similar to that of MMPs (47, 48). Fewer reports on the S1 collagenolytic serine proteases from bacteria exist. Uesugi et al. (50) purified an S1 collagenolytic protease, SOT (Streptomyces omiyaensis trypsin), from Streptomyces omiyaensis, and it has 77% identity with the trypsin SGT (Streptomyces griseus trypsin) from S. griseus in the primary structure (49, 50). SOT and SGT show similar pH and temperature optima, thermostabilities, and substrate preferences. Both SOT and SGT greatly hydrolyze type I collagen, showing much higher activities than commercial C. histolyticum collagenase type I (Sigma). SOT also has a high activity towards type IV collagen, but the ability of SGT to hydrolyze type IV collagen is poor. A comparison of the substrate specificities of the constructed chimeras of SOT and SGT indicates that the N-terminal domains of these enzymes are associated with their specificity toward structural protein substrates (50). Because SOT and SGT can degrade type I collagen at 37°C, these S1 collagenolytic proteases can be regarded as bacterial collagenases.

BACTERIAL SERINE COLLAGENOLYTIC PROTEASES FROM THE S8 FAMILY

Diversity.

The S8 family, also known as the subtilisin or subtilase family, is the second-largest family of serine proteases. Proteases in this family are all characterized by an Asp/His/Ser catalytic triad and an alpha/beta-fold catalytic center containing a seven-stranded parallel β-sheet (51). While most proteases in the S8 family have no collagenolytic activity, some S8 proteases have been reported in recent years to be collagenolytic proteases. The thermostable protease from Geobacillus collagenovorans MO-1 was the first reported S8 collagenolytic protease. This enzyme has a C-terminal collagen-binding domain, and its cleavage sites on collagen are varied but specific (52, 53). Several more S8 proteases with collagenolytic activity have been reported from environmental bacteria and archaea (54–59) and from human pathogens (60–62). Recently, two S8 proteases from deep-sea bacteria, deseasin MCP-01 from Pseudoalteromonas sp. strain SM9913 and myroicolsin from Myroides profundi D25, were characterized as collagenolytic proteases, and their actions toward collagen degradation were also revealed (63–66).

Structure and collagenolytic mechanism.

Deseasin MCP-01 contains an S8 catalytic domain, a linker, a proprotein convertase domain (P domain), and a PKD-like domain (Fig. 1) (67). Its C-terminal PKD-like domain is responsible for collagen binding, and Try36 in the PKD-like domain plays a key role in collagen binding (63). The PKD-like domain of MCP-01 can swell collagen but cannot unwind the collagen triple helix (64). The structure of the catalytic domain of MCP-01 has been solved (65), and this showed that it adopts a fold similar to subtilisin Carlsberg, the protype of the S8 family, which has no collagenolytic activity. The catalytic triad of MCP-01 is composed of Asp49, His104, and Ser269 (Fig. 2 and Table 1). Structural and mutational analyses indicated that compared to subtilisin Carlsberg, MCP-1 has an enlarged substrate-binding pocket, which is necessary for collagen recognition. The substrate-binding pocket is composed of three loops, and the acidic and aromatic residues on these loops form a negatively charged, hydrophobic environment for collagen binding. The catalytic domain of MCP-01 alone can degrade type I collagen, but with a lower efficiency than the intact MCP-01, suggesting that the PKD domain in MCP-01 is helpful for collagen degradation (65).

MCP-01 cleaves type I, II, and IV collagens, gelatin, and fish collagen (63). Atomic force microscopy observation and biochemical analysis of the action of MCP-01 on bovine type I collagen fibers confirmed that both MCP-01 and its catalytic domain progressively release single fibrils from collagen fibers and release collagen monomers from fibrils mainly by hydrolyzing proteoglycans and telopeptides in the collagen fibers (65). The catalytic domain can further degrade type I triple-helical collagen and cleave collagen chains with multiple sites (63, 65). MCP-01 displayed a nonstrict preference for peptide bonds with Pro or basic residues at the P1 site and/or Gly at the P1′ site in collagen. His211 in the S1 pocket is the key residue responsible for the preference of MCP-01 for the P1 basic residues (65).

The S8 protease myroicolsin from Myroides profundi D25 also has a C-terminal domain, a β-jelly roll domain (Fig. 1) (66), which, however, has no collagen-binding ability. Myroicolsin had broad specificity toward several types of collagens, including types I, II, and IV collagens, gelatin, and fish collagen. Scanning electron microscope and atomic force microscope observations combined with biochemical analyses confirmed that myroicolsin exerts a similar action on collagen fiber disassembly as MCP-01. Myroicolsin showed different cleavage patterns on native collagen and denatured collagen. On native collagen, the P1 position is often occupied by Gly, Arg, Pro, or Phe, and the P1′ position is occupied by Gly. On denatured collagen, the P1 position is always a basic residue (Lys and Arg), and the P1′ position is Gly. Ran et al. proposed a collagen degradation model of myroicolsin based on their results (Fig. 3) (66). Myroicolsin could firstly break the interfibrillar proteoglycan bridges, leading to disassembly of the tight structure of collagen fiber and exposure of collagen fibrils. Then, telopeptides at the molecular ends are hydrolyzed by myroicolsin, which leads to the breaking of cross-links between collagen monomers and the release of monomers. Thus, myroicolsin gains access to the attack sites on collagen monomers. Finally, the fibrillar structure of collagen is completely destroyed, and collagen monomers are degraded into peptides and free amino acids (66).

FIG 3.

Schematic model for stepwise collagen degradation by the S8 protease myroicolsin from Myroides profundi D25. Black arrows indicate the collagenolytic process of myroicolsin from collagen fiber to peptides and amino acids. Green arrows indicate the cross-links or bonds destroyed by myroicolsin in each step. Red arrows indicate the cleavage sites of myroicolsin on synthetic peptides. T1, T2, and T3 are the three polypeptide chains in a collagen monomer, and dotted lines indicate the hydrogen bonds between the polypeptide chains in a collagen monomer. (Adapted from reference 66 with permission of the publisher [copyright 2014, the American Society for Biochemistry and Molecular Biology.]).

Bacterial collagenolytic proteases in the S8 family, including MCP-01, myroicolsin, and the thermostable protease from Geobacillus collagenovorans MO-1, can degrade native type I collagen at 37°C and below, though they have high temperature optima (50 to 60°C). Therefore, these S8 collagenolytic proteases can be regarded as bacterial collagenases.

BACTERIAL SERINE COLLAGENOLYTIC PROTEASES FROM THE S53 FAMILY

Diversity and structure.

The proteases of the S53 family, or the sedolisin family, have been identified recently and found in a wide variety of organisms (12, 68). These enzymes usually exhibit maximum activity at low pH and high temperature. Crystal structures of some members of this family are available, and these adopt a fold that resembles subtilisin but is significantly larger. However, the enzymes of the S53 family have a unique catalytic triad, Ser-Glu-Asp, distinct from that of subtilisin-like proteases (Asp-His-Ser) (68).

Kumamolisin-As, originally named ScpA, is the only known member with collagenolytic activity in the sedolisin family (69). Kumamolisin-As was originally identified in the culture filtrate of a thermoacidophilic soil bacterium, Alicyclobacillus sendaiensis strain NTAP-1 (70, 71). Mature kumamolisin-As contains 361 amino acid residues. Crystal structures of the uninhibited enzyme and its complex with a covalently bound inhibitor, N-acetyl-isoleucyl-prolyl-phenylalaninal, have been solved (72). The catalytic triad is composed of Ser278, Glu78, and Asp82 (Fig. 2). In recent years, the catalytic mechanism of kumamolisin-As has been elucidated by quantum mechanical/molecular mechanical molecular dynamics and free energy simulations and site-directed mutations (73–77).

Collagenolytic mechanism.

Specificity analyses showed that kumamolisin-As is highly specific for collagen under thermal and acidic conditions and has maximum collagenolytic activity at pH 4.0 and 60°C (69, 71). Collagen is denatured under these conditions, and thus, kumamolisin-As actually acts on denatured collagen. Because strain NTAP-1 is a thermoacidophilic bacterium (70), collagen should be relaxed under the bacterial growth conditions before proteolysis can occur (71). These analyses suggest that kumamolisin-As may not be a true bacterial collagenase. Structural and docking studies have also shown that a groove that encompasses the active site provides a sufficient space for binding a relaxed collagen molecule and that the negative charges clustered on the surface of this groove facilitate the binding of the positively charged collagen. The low-pH activity of kumamolisin-As may be due to the presence of Asp¹⁶⁴ in the oxyanion hole. Kumamolisin-As shows a strong preference for arginine at the P1 position on collagen, due to the presence of a negatively charged residue (Asp179) in the S1 pocket of the enzyme (72).

BACTERIAL COLLAGENOLYTIC PROTEASES FROM THE U32 FAMILY

Diversity and structure.

The U32 family contains several collagenolytic proteases of unknown catalytic type (12). Most of these proteases are the virulence factors of human-pathogenic bacteria. Protease PrtC from Porphyromonas gingivalis ATCC 53977 (which is no longer available from the ATCC) was the first reported and is also the most studied U32 protease (78). The other members of this family are from Helicobacter pylori, Salmonella enterica serovar Typhimurium, Clostridium beijerinckii, Escherichia coli, and Bacillus subtilis (79–82). The prtC gene from Porphyromonas gingivalis ATCC 53977 contains 1,002 bp and encodes a 333-residue PrtC protein (78). The structure of the 10-kDa C-terminal domain of a putative U32 protease from Geobacillus thermoleovorans has been solved. This C-terminal domain shows a compact distorted open β-barrel made up of eight β-strands and may function in protein binding based on structure similarity analysis (83). If this is true, it can be concluded that the N-terminal domain of U32 proteases should function as a catalytic domain (Fig. 1).

Collagenolytic mechanism.

PrtC degraded soluble and reconstituted fibrillar type I collagen, heat-denatured type I collagen, and azocoll at temperatures below 37°C but not gelatin or the Pz peptide (78). This characteristic indicates that PrtC is likely a true bacterial collagenase. The activity of PrtC was enhanced by Ca²⁺ and inhibited by EDTA, sulfhydryl-blocking agents, and the salivary peptide histatin (78). The substrate specificity of other U32 proteases has not been characterized in detail. Due to the lack of structural information, the protein fold and active site residues of the catalytic domain for any member of this family are unclear. However, because most U32 proteases are the virulence factors of human-pathogenic bacteria, it is important to understand the structures and collagen degradation mechanisms of these virulent proteases, as such information is helpful for the development of therapies for the diseases caused by relevant pathogenic bacteria.

ROLES IN HUMAN DISEASES AND ENVIRONMENTAL NITROGEN CYCLING AND BIOTECHNOLOGICAL AND MEDICAL APPLICATIONS

Roles in human diseases.

Bacterial collagenolytic proteases from pathogens have been of concern mainly because they are potential virulence factors. The collagenases from several pathogenic Clostridium species are suggested to be involved in the degradation of specific cell membrane or extracellular matrix components, which have been documented in a 2009 review by Popoff and Bouvet (84). The collagenase of Vibrio vulnificus, a pathogen which can cause cellulitis or septicemia, was reported to have a potential contribution to the invasiveness of the bacteria into human tissue through open wounds (85, 86). Many collagenases from Clostridium and Vibrio pathogens are supposed to have similar roles. Some S8 proteases with high collagenolytic activity that are secreted by the pathogens Stenotrophomonas maltophilia (opportunistic agent of hemorrhagic pneumonia) and Acanthamoeba spp. (causative agent of amebic keratitis and granulomatous amebic encephalitis) probably function as important pathogenic factors and may serve as targets for the development of therapeutic agents (60–62).

The U32 family collagenolytic proteases from several human pathogens, such as Porphyromonas gingivalis, Aeromonas veronii, and Helicobacter pylori, are also suggested to be virulence factors (78, 81, 87). Although a lot of pathogens are found to secrete collagenolytic proteases, the roles and mechanisms of the proteases involved in their pathogenic processes are still largely unknown. The exploration of the mechanisms of human pathogen collagenolytic proteases on collagen degradation is essential for establishing treatments for related human diseases. More detailed information on bacterial collagenases related to human diseases can be found in a recent review (88).

Roles in environmental nitrogen cycling.

Besides pathogens, a lot of collagenolytic-protease-secreting bacteria have been isolated from terrestrial soil and marine sediments. Thus, it is reasonable to believe that collagen degradation by extracellular collagenolytic proteases from various environmental bacteria is an important biological process for the release of fixed nitrogen into the global nitrogen cycle. It is well known that Clostridium spp. are involved in carcass putrefaction, and their collagenases play obvious roles in this process. Collagenolytic proteases from other environmental bacteria can be supposed to play similar roles in degrading various collagens from dead or even living animals. Revealing the mechanisms of environmental bacteria on collagen degradation is helpful for the study of the global nitrogen cycle. Moreover, bacterial collagenases have shown much biotechnological potential, which is documented in detail in the following section.

Biotechnological and medical applications.

Clostridium collagenase has been used in laboratories to dissociate tissues and isolate cells (89) and used as a therapeutic drug for the removal of necrotic wound tissues for several decades (90, 91). It is successfully applied in the treatment of third-degree burns (92, 93), soft tissue and pressure ulcers (94–97), diabetic ulcers (98), ischemic arterial ulcers (99–101), and Dupuytren's disease (102). Recently, clostridial collagenases (Xiaflex/Xiapex) were approved by both the FDA (December 2013) and the European Union (February 2015) to be used as nonsurgical treatment for Peyronie's disease. It has been shown that both Clostridium collagenase itself and its collagen degradation products can promote cell migration (103). In addition, the serine collagenolytic protease from Pseudoalteromonas sp. SM9913 has shown good potential in meat taste improvement (104) and meat tenderization (105). These examples indicate that bacterial collagenolytic proteases may have wide applications in research, medicine, and food processing in the future.

SUMMARY AND PROSPECTS

Summary.

Due to their essential role in global nitrogen cycling and/or their vital virulent role in some diseases, bacterial collagenolytic proteases have drawn increasing attention. During the last decade, the intact or partial structures of some bacterial collagenolytic proteases, including ColG, ColH, and ColT from M9, deseasin MCP-01 from S8, and kumamolisin-As from S53, have been solved, and their mechanisms for collagen degradation have been revealed by structural and biochemical studies. Several kinds of noncatalytic domains, including PKD-like domains, PPC domains, and CBDs, are found at the C termini of some bacterial collagenolytic proteases. Unlike PKD-like domains and PPC domains, which are identified by sequence homology, CBD is a functional name of the domains in bacterial collagenolytic proteases that have collagen-binding ability. The CBDs from different bacterial collagenolytic proteases may have little sequence homology. Except for the M9 Vibrio collagenase and the S1 protease SOT, whose cleavage patterns on collagen are similar to those of MMPs, the other bacterial collagenolytic proteases (the M9 Clostridium collagenase, the S8 protease, and the S53 protease) have multiple cleavage sites on collagen and can degrade collagen into small peptides and amino acids. Therefore, these bacterial proteases are very effective at environmental collagen degradation.

With more and more serine collagenolytic proteases being studied, common characteristics of these proteases from different families can be observed. The catalytic domains of serine collagenolytic proteases have activity toward collagen, independently of whether they have a collagen-binding domain. Serine collagenolytic proteases have an enlarged and negatively charged substrate-binding pocket for collagen recognition compared to their homologs that have no collagenolytic activity. Serine collagenolytic proteases usually have a preference for basic residues at the P1 position on collagen. The basic residue Arg is abundant in human collagens. The numbers of Arg residues in the sequences of chains α1 and α2 of human type I collagen are 71/1,464 and 72/1,366, respectively, and the numbers of Arg residues in the sequences of subunits α1 to α6 of collagen IV are 45/1,669, 79/1,721, 65/1,672, 70/1,690, 37/1,685, and 47/1,691, respectively. The peptide bonds with a C-terminal Arg residue in human collagens are all potential cleavage sites for serine collagenolytic proteases with a strong specificity for Arg at P1.

Prospects.

It is most likely that there are still many undiscovered bacterial collagenolytic proteases hidden in various environmental or pathogenic bacteria. Some of these proteases may have novel structures and collagenolytic mechanisms. The exploration of these proteases could provide further understanding of bacterial degradation of collagen globally and could uncover promising proteases with biotechnological potential. In addition, understanding the structures and collagen degradation mechanisms of the collagenolytic proteases from human pathogens, such as Vibrio proteases and the U32 proteases, will be helpful for the development of therapies for the diseases caused by relevant pathogenic bacteria.