Abstract
Collagen VI is a ubiquitous extracellular matrix component that forms extensive microfibrillar networks in most connective tissues. In this study, we describe for the first time, to our knowledge, that the collagen VI von Willebrand factor type A–like domains exhibit a broad-spectrum antimicrobial activity against Gram-positive and Gram-negative bacteria in human skin infections in vivo. In silico sequence and structural analysis of VWA domains revealed that they contain cationic and amphipathic peptide sequence motifs, which might explain the antimicrobial nature of collagen VI. In vitro and in vivo studies show that these peptides exhibited significant antibacterial activity against Staphylococcus aureus, Escherichia coli, and Pseudomonas aeruginosa through membrane disruption. Our findings shed new light on the role of collagen VI–derived peptides in innate host defense and provide templates for development of peptide-based antibacterial therapies.
Introduction
Rapid host defense mechanisms are essential to overcome the harmful actions of pathogenic bacteria. Antimicrobial peptides are powerful molecules of the innate immune defense system providing a rapid and nonspecific response against invading pathogens. Antimicrobial peptides exhibit broad-spectrum activity against Gram-positive bacteria, Gram-negative bacteria, fungi, and viruses (1). Despite their diverse origins, antimicrobial peptides share several physicochemical properties, which are crucial for their direct antimicrobial nature. These peptides are relatively small molecules (10–50 amino acids) and, with few exceptions, carry a net positive charge (2–4). Another reoccurring feature is that they have enough hydrophobic residues to adopt distinct amphiphilic characteristics upon contact with a lipid interface (e.g., the bacterial membrane) (4). Presently, more than 2000 antimicrobial peptides have been identified from a wide range of species including plants, insects, and mammals (Antimicrobial Peptide Database, http://aps.unmc.edu/AP/main.php). The predominant mechanism of action among antimicrobial peptides includes the adsorption to and disruption of cytoplasmic membrane of the target cells (5). However, several models describing the precise mechanics of this type of activity have been proposed (6), and in most cases, the detailed mechanism, such as the nature of membrane permeabilization, is still unknown (7). Apart from their direct antimicrobial activity, a mounting body of evidence shows that these peptides also have immunomodulatory functions such as being chemotactic, neutralizing endotoxin (e.g., LPS), wound healing, angiogenesis, and regulating production of proinflammatory cytokines (8, 9). Interestingly, their broad-spectrum antimicrobial activity together with a seemingly slow and incomplete bacterial resistance development process makes these antimicrobial peptides attractive therapeutic agents against infections (10, 11).
Extracellular matrix (ECM) proteins such as collagens, fibronectin, laminins, and vitronectin are attractive targets for pathogenic bacteria to adhere, invade, and colonize the connective tissue of the host (12). Among these, collagen VI is a ubiquitous ECM component which is present in all connective tissues and often associated with basement membranes. The predominant form is composed of three distinct polypeptide chains, α1(VI), α2(VI), and α3(VI), which form triple-helical monomers. Inside the cell, the monomers assemble into dimers and tetramers that are secreted into the extracellular space. There, the tetramers aggregate end-on-end to form microfibrils that become part of extended supramolecular matrix assemblies. More recently, three additional chains (α4, α5, and α6) were discovered, which may substitute for the α3-chain in some tissues (13, 14). In terms of structure, each α-chain is characterized by a short extended triple-helical region flanked by two large N- and C-terminal globular regions, which share homology with von Willebrand factor type A (VWA) domains (15–17). The α1(VI) and α2(VI) chains contain one N-terminal (N1) and two C-terminal (C1 and C2) VWA domains, whereas the α3(VI) is much larger and is composed of some ten N-terminal (N10-N1) VWA domains and two C-terminal VWA domains. Additionally, the α3(VI)-chain has three C-terminal domains (C3–C5) that share homology with salivary gland proteins, fibronectin type III repeats, and the Kunitz family of serine protease inhibitors (18).
Peptides derived from von Willebrand factor–containing consensus heparin-binding sequences (Cardin and Weintraub motifs) (19) are known to exhibit antimicrobial activity against Gram-positive and Gram-negative bacteria (20, 21). We have previously demonstrated that collagen VI binds to Streptococcus pneumonia and group A, C, and D streptococci (22), leading to bacterial killing (23). Recently, we observed similar adhesive and bactericidal effects of collagen VI microfibrils against Moraxella catarrhalis and other Gram-negative and Gram-positive human pulmonary pathogens (24). Although these data implicate collagen VI in antibacterial defense, no structural or mechanistic explanations were presented. The links between VWA domains and antimicrobial peptides prompted us to explore whether the globular domains of collagen VI may play an important role in host tissue defense during the early phase of infection. In this study, employing various experimental approaches, we describe for the first time, to our knowledge, that VWA domains of the collagen VI α3-chain harbor cationic sequence motifs which display antimicrobial activity in vivo.
Materials and Methods
Bacterial strains and culture conditions
Streptococcus pyogenes strain AP1 (40/58) of serotype M1 was from the World Health Organization Collaborating Centre for Reference and Research on Streptococci (Prague, Czech Republic). Staphylococcus aureus strain 111 and Escherichia coli strain B1351 were collected at the Department of Clinical Microbiology, Lund University Hospital (Lund, Sweden). The Pseudomonas aeruginosa strain used in this study was PAO1 (American Type Culture Collection, Teddington, U.K.), originally isolated from a wound. All bacteria were routinely grown in Todd Hewitt broth (THB; Difco, Detroit, MI) and incubated at 37°C in a humid atmosphere with 5% CO2.
Recombinant expression and purification of N- and C-terminal VWA domains of collagen VI α- chains
Collagen VI microfibrils were extracted from bovine cornea by collagenase digestion as described by Abdillahi et al. (23). The cDNA constructs coding for the noncollagenous domains of collagen VI were generated by RT-PCR on total RNA from mouse brain and cloned with 5′-terminal NheI or XhoI and 3′-terminal BamHI or XhoI restriction sites using the primers listed in Table I (25). Each of the amplified PCR products were inserted into a modified pCEP-Pu vector containing an N-terminal BM-40 signal peptide and a C-terminal tandem strepII-tag downstream of the restriction sites (26). HEK293-EBNA cells (Invitrogen, Carlsbad, CA) were transfected with the recombinant plasmids using FuGENE 6 reagent (Roche, Mannheim, Germany) according to the manufacturer’s protocol. The cells were selected with puromycin (1 μg/ml) (Sigma-Aldrich, St. Louis, MO), and the recombinant proteins were purified directly from DMEM (Invitrogen) supplemented with FCS (Biochrom, Berlin, Germany). After filtration and centrifugation (1 h, 10,000 × g), the cell culture supernatants were applied to a Strep-Tactin column (1.5 ml; IBA, Göttingen, Germany) and eluted with 2.5 mM desthiobiotin (Sigma-Aldrich), 10 mM Tris–HCl (pH 8).
Sequence and structural analysis
The amino acid sequence of human collagen VI α-chains can be accessed through UniProtKB database: α1(VI) (UniProt no. P12109), α2(VI) (UniProt no. P12110), and α3(VI) (UniProt no. P12111). Swiss PDB Viewer DeepView version 4.1 was used for sequence alignment and to analyze three-dimensional structures. Only a crystal structure of mouse α3N5 was available, with Protein Data Bank code 4IGI (27). No crystal structures were available for human VWA domains of collagen VI, and predicted models from ModBase (http://modbase.compbio.ucsf.edu) were therefore used to generate the figures. There were no predicted models for N1 and C2 domains of α3(VI).
Peptide synthesis
GVR28 (GVRPDGFAHIRDFVSRIVRRLNIGPSKV), FYL25 (FYLKTYRSQAPVLDAIRRLRLRGGS), FFL25 (FFLKDFSTKRQIIDAINKVVYKGGR), VTT30 (VTTEIRFADSKRKSVLLDKIKNLQVALTSK), SFV33 (SFVARNTFKRVRNGFLMRKVAVFFSNTPTRASP), and DVN32 (DVNVFAIGVEDADEGALKEIASEPLNMHMFNL) were synthesized by Biopeptide (San Diego, CA). The purity (>95%) and molecular mass of these peptides were confirmed by MALDI-TOF mass spectrometry analysis. All peptides used were water-soluble except DVN32, which was dissolved in <0.01% DMSO.
Heparin-binding assay
LL-37 (5 μg) or recombinant fragments from collagen VI (10 μg) were applied to nitrocellulose membranes (Hybond-C; GE Healthcare, Uppsala, Sweden). Membranes were blocked with 2% BSA in PBS (w/v) for 2 h at room temperature (RT), followed by washing steps with PBS with Tween 20, and incubated with 60 μg of heparin–biotin (Sigma-Aldrich) overnight at 4°C. In some experiments, unlabeled heparin (6 mg/ml) was added for competition of binding. After washing, the membranes were incubated with HRP streptavidin (Sigma-Aldrich) for 30 min at RT and washed, and the bands were visualized by the Supersignal West Pico Chemiluminescent Substrate developing system (Thermo Fisher Scientific, Roskilde, Denmark).
Patients and isolates
Human skin biopsies were obtained from fasciotomy patients who were judged to have an infection with group A streptococci by the responsible physician. Fibrin slough from patients with a chronic venous ulcer was fixed and processed as previously described (28, 29). The research project was approved by the ethics committee of Lund University Hospital. Informed consent was obtained from the patients.
Viable count assay
Bacteria were grown to midlogarithmic phase (OD620 ≈ 0.4) in THB medium at 37°C with 5% CO2. The bacterial solution was subsequently washed and adjusted to 2 × 109 CFU/ml in Tris buffer (10 mM, pH 7.4) containing 5 mM glucose. S. pyogenes, S. aureus, E. coli, or P. aeruginosa (cell viability 95–98% for all strains) were then incubated with 2 μM (200 μg/ml) of purified collagen VI at 37°C for 2 h. In another experiment, S. pyogenes was incubated with recombinant collagen VI fragments at various concentrations (0.125, 0.25, 0.5, 1.0, and 2.0 μM) for 2 h at 37°C. Bacteria incubated with Tris buffer (containing 5 mM glucose) or 3 μM (13 μg/ml) LL-37 (Innovagen, Lund, Sweden) were used as negative and positive controls, respectively. To quantify the bactericidal activity, serial dilutions of the incubation mixtures were plated on blood agar plates, followed by incubation at 37°C overnight, and the number of CFU was determined. One hundred percent survival was defined as total survival of bacteria in the same buffer and under the same condition in the absence of collagen VI or recombinant proteins.
Radial diffusion assay
Radial diffusion assay (RDA) was performed as described earlier (30) with some minor modifications. Bacteria were grown to midlogarithmic phase (OD620 ≈ 0.4) in 10 ml of full-strength (3% w/v) trypticase soy broth (TSB) (Becton Dickinson, Franklin Lakes, NJ). Bacteria were then washed once with Tris buffer (containing 5 mM glucose). Subsequently, 4 × 106 CFU/ml bacteria was added to 5 ml of the underlay agarose gel consisting of 0.03% (w/v) TSB, 1% (w/v) low electroendosmosis-type agarose, and 0.02% (v/v) Tween 20 (both from Sigma-Aldrich). The underlay was poured into a Ø 90-mm petri dish. After agarose solidification, Ø 4-mm wells were punched out, and 6 μl of 10 mM Tris–HCl buffer alone or containing peptide (100 μM) was added to each well. Plates were incubated at 37°C for 3 h to allow diffusion of the peptides. The underlay gel was then covered with 5 ml of the overlay (6% TSB and 1% low electroendosmosis-type agarose in dH2O). To test peptide activity in physiological salt concentrations, 150 mM NaCl in the presence or absence of 2.5 mM of calcium or/and 4 mM of magnesium were added to the underlay agarose gel. Antimicrobial activity was seen as a clearing zone around each well after incubating 18–24 h at 37°C.
calScreener isothermal microcalorimetry assay
The growth rate and inhibitory potential of the different selected peptides were determined using the calScreener isothermal microcalorimetry assay (Symcel Sverige, Spånga, Sweden). The use of calorimetry to determine bacterial growth and antimicrobial compound potency has been described (31–34). For all experiments, the calScreener 48-channel isothermal microcalorimeter with its corresponding presterilized plastic insert 48-well plate (calPlate) was used. The calPlate was placed in a thermostatic chamber set at 37°C and rested upon a heat-flux detecting sensor, the thermopile. The sensor is attached to a heat-sink with a large mass compared with the cell culture cups. All heat produced is transferred to the heat-sink, giving rise to a signal in the thermopile sensor proportional to the heat flow. For the assays, bacterial cultures were grown overnight and diluted into fresh Tris buffer (containing 5 mM glucose) to an inoculum of 102 CFU/ml. Subsequently, 200 μl of culture was added with collagen VI–derived peptides (1, 5, 10, 50, and 100 μM) to the calorimetric vial prior to calorimetric measurement. The measurements were performed in closed vessels containing 200 μl of growth media and 400 μl of air. The assay was run with continuous data collection for a minimum of 24 h. Data were analyzed using the CALview software (version 1.0.33.0, 2016; Symcel Sverige) for visualization of growth kinetics and total energy release. General handling and device manipulation were done according to the manufacturer’s recommendations.
Liposome leakage assay
The liposome production and leakage assay were performed similar to what has been previously described (35). In short, dry lipid films of E. coli polar lipid extract (Avanti Polar Lipids, Alabaster, AL) were formed on round-bottom flask walls by dissolving lipids in chloroform, followed by evaporation under N2 flow, and subsequently placed in vacuum overnight. Lipid films were resuspended at 55°C in an aqueous solution of 100 mM of 5(6)-carboxyfluorescein in Tris buffer. The suspensions were subjected to repeated extrusion through a 100-nm polycarbonate membrane to reduce multilamellar structures and polydispersity. Untrapped carboxyfluorescein was removed by gel filtration. Membrane permeability was measured by monitoring carboxyfluorescein efflux from the liposomes to the external low concentration environment, resulting in loss of self-quenching and an increased fluorescence signal. The 96-well plates were prepared with a 2-fold serial dilution of the peptides in Tris buffer as well as controls without peptides (background) and 0.16% Triton X-100 (maximum leakage). The plates were preheated to incubation temperature (37°C), and liposome solution was administered to a final lipid concentration of 10 μM in 200 μl. The effects of each peptide concentration on the liposome systems were monitored for 45 min, at which point the initial leakage had largely subsided. Results shown represent the average from triplicate experiments with SD and are expressed as percentage of total leakage generated with Triton X-100 and subtraction of the baseline value. The EC50 values are calculated from sigmoidal dose-response curves of the leakage percentage as a function of the peptide concentration (log10).
Light microscopy and immunohistochemistry
Skin biopsies were fixed in PBS containing 4% formaldehyde (18 h, 4°C) and processed for paraffin sectioning. Deparaffinized and rehydrated 5-μm sections were then incubated with primary Abs (1:1000 dilution in Da Vinci Green) and peroxidase-conjugated AffiniPure rabbit anti-chicken (secondary Ab, 1:500 dilution in Da Vinci Green) and detected with the MACH 1 Universal HRP-Polymer Detection Kit (Biocare Medical, Concord, CA) according to the manufacturer’s instructions. Ab 1014+ against collagen VI was a kind gift from Dr. R. Timpl (Max Planck Institute, Martinsried, Germany). Polyclonal Abs (IgY) against collagen VI–derived peptides (GVR28, FYL25, FFL25, VTT30, and SFV33) were purchased from Capra Science Antibodies (Ängelholm, Sweden). For testing unspecific binding, primary chicken IgG Ab (2.1 mg/ml; Sigma-Aldrich) diluted 1:1000 was used. Specimens were examined in an Olympus BX43 equipped with an Olympus XC10 camera.
Fluorescence microscopy
Bacteria were grown to midlogarithmic phase in THB medium, washed, and resuspended in 10 mM of Tris–HCl containing 5 mM of glucose to obtain a suspension of 2 × 107 CFU/ml. One hundred microliters of the bacterial suspension was incubated with 2 μM (200 μg/ml) of purified collagen VI or 3 μM (13 μg/ml) of LL-37 at 37°C for 30 min. Untreated bacteria were used as control. Two hundred microliters of FITC (6 μg/ml; Sigma-Aldrich) was added to all samples and incubated for 30 min at 37°C. Bacteria were washed and immobilized onto poly-l-lysine (Sigma-Aldrich)–coated glass slides by incubating for 45 min at 37°C. The slides were washed with Tris–HCl/glucose and were fixed with 4% paraformaldehyde by incubating at 4°C for 15 min followed by 45 min incubation at RT. The glass slides were subsequently mounted on coverslips using Prolong Gold antifade reagent mounting medium (Invitrogen). The bacteria were visualized in an Olympus BX43 fluorescence microscope equipped with an Olympus XC10 camera, a Plan Apochromat (100× objective), and a high numerical aperture oil condenser.
Scanning electron microscopy
S. pyogenes, S. aureus, E. coli, and P. aeruginosa (2 × 109 CFU/ml) were incubated with purified collagen VI at a concentration of 2 μM (200 μg/ml) for 0, 30, 60, and 120 min at 37°C with 5% CO2. In some experiments S. pyogenes was incubated with 2 μM of recombinant collagen VI fragments for 2 h at 37°C. Three micromolars (13 μg/ml) of LL-37 was used as a positive control, and bacteria in Tris–HCl (pH 7.4) was used as negative control. For bacterial adhesion, bacteria were applied onto full-length collagen VI–coated (1 μM; 150 μg/ml) titanium discs and incubated at 37°C with 5% CO2 for 30 min. For heat inactivation, bacteria were incubated for 20 min at 65°C. For chemical fixation, bacteria were incubated for 1 h in 2.5% glutaraldehyde in 0.15 M sodium cacodylate. All samples were then fixed with 2.5% glutaraldehyde in 0.15 M of sodium cacodylate (pH 7.4) (cacodylate buffer), washed with cacodylate buffer, and dehydrated with an ascending ethanol series as previously described (36). Paraffin sections of patient biopsies were treated in a similar way (24). The specimens were then subjected to critical-point drying with carbon dioxide, and absolute ethanol was used as an intermediate solvent. The tissue samples were mounted on aluminum holders, sputtered with 20 nm palladium/gold, and examined in a Philips/FEI XL-30 FEG scanning electron microscope operated at 5 kV accelerating voltage.
Transmission electron microscopy
The binding of recombinant collagen VI fragments to the bacterial surface was visualized by negative staining and transmission electron microscopy as described previously (22). Briefly, bacteria were incubated with recombinant collagen VI fragments in presence or absence of heparin (10 μg/ml) for 1 h at 37°C. For visualization in the electron microscope, the different recombinant fragments were conjugated with 5 nm colloidal gold (37). In some experiments, human skin and fibrin slough specimens were fixed in 0.15 M sodium cacodylate and 2.5% glutaraldehyde (pH 7.4), embedded in Epon, and ultrasectioned. Samples were subjected to Ag retrieval with sodium metaperiodate and subsequently incubated with primary Abs against collagen VI–derived peptides (dilution 1:50), followed by detection with species-specific secondary Ab–gold conjugates. Bacteria were detected by specific Abs directed against S. pyogenes, S. aureus, P. aeruginosa, and E. coli. A standard procedure for controls, omitting the primary Ab, was also performed. Specimens were examined in a Philips/FEICM 100 TWIN transmission electron microscope operated at 60 kV accelerating voltage. Images were recorded with a side-mounted Olympus Veleta camera with a resolution of 2048 × 2048 pixels (2k × 2K) and the ITEM acquisitions software.
Statistical analysis
Statistical analysis was performed using GraphPad Prism, version 7.0. The p value was determined by using ANOVA. All experiments were performed at least three times if not otherwise mentioned. Values were expressed as mean ± SE.
Results
Collagen VI kills Gram-positive and Gram-negative human pathogens by membrane permeabilization
Prompted by our recent findings that collagen VI kills several human pathogens (23, 24), we wished to investigate these properties in further detail. For this purpose, an integrated approach was established to combine microbiological and biochemical assays with high-resolution scanning electron microscopy. Viable count assays were performed by incubating the Gram-positive bacteria S. pyogenes and S. aureus as well as the Gram-negative bacteria E. coli and P. aeruginosa with purified collagen VI for 2 h at 37°C. The results showed that collagen VI indeed displayed antibacterial activity against S. aureus, E. coli, and P. aeruginosa in a similar way as observed for S. pyogenes, our model organism (Fig. 1A). The human benchmark antimicrobial peptide LL-37 was used as a positive control and showed almost 100% killing of all the bacteria strains. To examine whether collagen VI disrupts the bacterial membrane, high-resolution scanning electron microscopy was used to visualize bacterial architecture during killing. Bacteria were either incubated with buffer alone or with collagen VI (Fig. 1B). The results showed extensive disruption of the bacterial membrane structure and extravasations of cytoplasmic components in the presence of collagen VI, indicating damage to the bacterial membranes (Fig. 1B, right panel, Fig. 1C). These findings were similar to those seen after treatment with LL-37 (data not shown). In contrast, bacterial cell wall architecture in the control samples remained unaffected (Fig. 1B, left panel). In addition, heat-killed or chemical-fixated bacteria were also treated with collagen VI, and there was no disruption of bacterial membrane or precipitation of collagen VI (Supplemental Fig. 1). None of the bacterial strains underwent aggregation, which was also confirmed by electron microscopy (data not shown). These observations were further substantiated by the use of the impermeant dye FITC. Fluorescence microscopy analysis showed that the uptake of FITC was only visible in samples treated with collagen VI or LL-37 (Fig. 1D), thus demonstrating permeabilization of the bacterial membrane. Similar observations were made for a variety of other Gram-positive and Gram-negative human pathogens (S.L. Nordin et al., manuscript in preparation). Taken together, these data demonstrate that collagen VI exhibits a broad-spectrum antimicrobial activity against Gram-positive and Gram-negative bacteria.
Antibacterial effect of collagen VI against different strains of Gram-positive and Gram-negative bacteria. (A) S. pyogenes, S. aureus, E. coli, and P. aeruginosa (2 × 106 CFU/ml) were incubated with collagen VI (200 μg/ml; 2 μM) for 2 h at 37°C with 5% CO2. Bacteria incubated with Tris–HCl/glucose (pH 7.4) buffer or with LL-37 served as negative or positive controls, respectively. Six separate experiments. (B) For visualization of antimicrobial activities, bacteria (2 × 109 CFU/ml) were treated with collagen VI (2 μM) for 2 h at 37°C and subsequently subjected to scanning electron microscopy. Extensive membrane damage, blebbing, and ejection of cytoplasmic components were observed in the presence of collagen VI (right panel) compared with untreated bacteria (left panel). Scale bar, 2 μm. (C) Kinetics studies of bacterial membrane disruption induced by collagen VI. S. pyogenes and P. aeruginosa (green pseudocolor) were treated with collagen VI (200 μg/ml; 2 μM) for 0, 30, 60, and 120 min at 37°C and visualized with scanning electron microscopy. Arrowheads indicate membrane blebbing. Cytoplasmic exudates are highlighted in purple pseudocolor. Scale bars, 1 μm (S. pyogenes) and 2 μm (P. aeruginosa), respectively. One representative experiment out of three is shown. (D) In the fluorescence microscopy analysis, bacteria were treated with collagen VI as described above, and permeabilization was assessed by using the impermeant probe FITC (lower panels). The same positions were visualized with light microscopy (upper panels). LL-37 (3 μM) was used as a positive control for membrane damage, and bacteria with buffer only was used as negative control. Green color indicates bacterial lysis. Images were taken at original magnification ×1000. One representative experiment out of five is shown.
Antibacterial effect of collagen VI against different strains of Gram-positive and Gram-negative bacteria. (A) S. pyogenes, S. aureus, E. coli, and P. aeruginosa (2 × 106 CFU/ml) were incubated with collagen VI (200 μg/ml; 2 μM) for 2 h at 37°C with 5% CO2. Bacteria incubated with Tris–HCl/glucose (pH 7.4) buffer or with LL-37 served as negative or positive controls, respectively. Six separate experiments. (B) For visualization of antimicrobial activities, bacteria (2 × 109 CFU/ml) were treated with collagen VI (2 μM) for 2 h at 37°C and subsequently subjected to scanning electron microscopy. Extensive membrane damage, blebbing, and ejection of cytoplasmic components were observed in the presence of collagen VI (right panel) compared with untreated bacteria (left panel). Scale bar, 2 μm. (C) Kinetics studies of bacterial membrane disruption induced by collagen VI. S. pyogenes and P. aeruginosa (green pseudocolor) were treated with collagen VI (200 μg/ml; 2 μM) for 0, 30, 60, and 120 min at 37°C and visualized with scanning electron microscopy. Arrowheads indicate membrane blebbing. Cytoplasmic exudates are highlighted in purple pseudocolor. Scale bars, 1 μm (S. pyogenes) and 2 μm (P. aeruginosa), respectively. One representative experiment out of three is shown. (D) In the fluorescence microscopy analysis, bacteria were treated with collagen VI as described above, and permeabilization was assessed by using the impermeant probe FITC (lower panels). The same positions were visualized with light microscopy (upper panels). LL-37 (3 μM) was used as a positive control for membrane damage, and bacteria with buffer only was used as negative control. Green color indicates bacterial lysis. Images were taken at original magnification ×1000. One representative experiment out of five is shown.
The recombinant VWA domain–containing globular regions of collagen VI bind to the bacterial surface in a heparin-dependent manner
The affinity for the negatively charged surfaces on bacterial membranes is perhaps the most important factor for antibacterial peptides, regardless of their specific mode of action. Given this, most antibacterial peptides and proteins show affinity for heparin (20, 38). Therefore, we wished to determine whether collagen VI exhibited similar properties. To this end, biotin-labeled heparin was tested in a slot blot assay for binding to immobilized recombinant fragments (for construction of these fragments see Table I) of collagen VI (as denoted in Fig. 2A). Heparin bound to the different N- and C-terminal regions with varying intensity (Fig. 2B, left panel). Interestingly, the affinity of heparin to α3C was comparable to LL-37, the positive control. The binding to all fragments was blocked by nonlabeled heparin (Fig. 2B, right panel). For visualization of these interactions, collagen VI fragments were directly conjugated to colloidal gold and incubated with S. pyogenes bacteria. Negative staining and transmission electron microscopy revealed that all fragments bound to the bacterial surfaces in the absence of heparin (Fig. 2C, upper panel). No binding was observed in the presence of heparin, and instead, gold conjugates were distributed randomly in the background (Fig. 2C, lower panel). Similar results were obtained for S. aureus, E. coli, and P. aeruginosa in the absence or presence of heparin (Fig. 2D). This was also the case for the full-length protein (data not shown).
Primer . | Sequence . | Restriction Enzyme . |
---|---|---|
αlN(fw) | 5′-AGAGCTAGCATGCCCTGTGGATCTATTC-3′ | NheI |
αlN(rev) | 5′-GCACTCGAGAATCATGTCCACAATGGTGT-3′ | XhoI |
αlC(fw) | 5′-GCAGCTAGCTGCACATGTGGACCCATTGA-3′ | NheI |
αlC(rev) | 5′-AACCTCGAGGCCCAGTGCCACCTTCCT-3′ | XhoI |
α2N(fw) | 5′-AGAGCTAGCAAGGCCGACTGCCCAGTC-3′ | NheI |
α2N(rev) | 5′-GCACTCGAGGACCTTGATGATGCGGTT-3′ | XhoI |
α2C(fw) | 5′-GAAGCTAGCTGTGAGAAGCGCTGTGGT-3′ | NheI |
α2C(rev) | 5′-GCAGGATCCACAGATCCAGCGGATG-3′ | BamHI |
α3N(fw) | 5′-TATCTCGAGCTGATGGATCTGCTGTGAGGTTA-3′ | XhoI |
α3N(rev) | 5′-AGGAACCAGGGATCCCAGGGGCCTGTCATACATGAAGCC-3′ | BamHI |
α3C(fw) | 5′-AAAGCTAGCCTGGAGTGCCCTGTATTCCCAAC-3′ | NheI |
α3C(rev) | 5′-TTTGGATCCTCAAACTGTTAACTCAGGACTAC-3′ | BamHI |
Primer . | Sequence . | Restriction Enzyme . |
---|---|---|
αlN(fw) | 5′-AGAGCTAGCATGCCCTGTGGATCTATTC-3′ | NheI |
αlN(rev) | 5′-GCACTCGAGAATCATGTCCACAATGGTGT-3′ | XhoI |
αlC(fw) | 5′-GCAGCTAGCTGCACATGTGGACCCATTGA-3′ | NheI |
αlC(rev) | 5′-AACCTCGAGGCCCAGTGCCACCTTCCT-3′ | XhoI |
α2N(fw) | 5′-AGAGCTAGCAAGGCCGACTGCCCAGTC-3′ | NheI |
α2N(rev) | 5′-GCACTCGAGGACCTTGATGATGCGGTT-3′ | XhoI |
α2C(fw) | 5′-GAAGCTAGCTGTGAGAAGCGCTGTGGT-3′ | NheI |
α2C(rev) | 5′-GCAGGATCCACAGATCCAGCGGATG-3′ | BamHI |
α3N(fw) | 5′-TATCTCGAGCTGATGGATCTGCTGTGAGGTTA-3′ | XhoI |
α3N(rev) | 5′-AGGAACCAGGGATCCCAGGGGCCTGTCATACATGAAGCC-3′ | BamHI |
α3C(fw) | 5′-AAAGCTAGCCTGGAGTGCCCTGTATTCCCAAC-3′ | NheI |
α3C(rev) | 5′-TTTGGATCCTCAAACTGTTAACTCAGGACTAC-3′ | BamHI |
fw, forward primer, rev, reverse primer.
(A) Schematic diagram of collagen VI domain structures. Collagen VI consists of three α-chains, namely α1(VI), α2(VI), and α3(VI). The N- and C-terminal globular domains of collagen VI are numbered as described previously (46). The brackets indicate the region in which the recombinant fragments were expressed. (B) Heparin-binding activity of recombinant globular domains of collagen VI was determined by slot blot using biotinylated heparin. Recombinantly expressed fragments of α1(VI)-, α2(VI)-, and α3(VI)-chain (10 μg) showed binding to heparin. LL-37 (5 μg) was used as a positive control (right panel). Unlabeled heparin (6 mg/ml) inhibited the binding of biotinylated heparin to the recombinant fragments and LL-37 (left panel). One representative experiment out of three is shown. (C) Binding of recombinant fragments to S. pyogenes was visualized by negative staining and transmission electron microscopy using colloidal gold labeling. Recombinant fragments at final concentration of 2 μM displayed binding to bacterial membrane (upper panel). Upon preincubation with unlabeled heparin, recombinant fragments did not bind to bacterial membrane (lower panel). Scale bar, 100 nm. (D) The amounts of recombinant fragments bound to the bacterial surface in the absence (−) or presence (+) of heparin were calculated as gold label per square millimeter of bacterial surface area. Four separate experiments. ****p < 0.0001.
(A) Schematic diagram of collagen VI domain structures. Collagen VI consists of three α-chains, namely α1(VI), α2(VI), and α3(VI). The N- and C-terminal globular domains of collagen VI are numbered as described previously (46). The brackets indicate the region in which the recombinant fragments were expressed. (B) Heparin-binding activity of recombinant globular domains of collagen VI was determined by slot blot using biotinylated heparin. Recombinantly expressed fragments of α1(VI)-, α2(VI)-, and α3(VI)-chain (10 μg) showed binding to heparin. LL-37 (5 μg) was used as a positive control (right panel). Unlabeled heparin (6 mg/ml) inhibited the binding of biotinylated heparin to the recombinant fragments and LL-37 (left panel). One representative experiment out of three is shown. (C) Binding of recombinant fragments to S. pyogenes was visualized by negative staining and transmission electron microscopy using colloidal gold labeling. Recombinant fragments at final concentration of 2 μM displayed binding to bacterial membrane (upper panel). Upon preincubation with unlabeled heparin, recombinant fragments did not bind to bacterial membrane (lower panel). Scale bar, 100 nm. (D) The amounts of recombinant fragments bound to the bacterial surface in the absence (−) or presence (+) of heparin were calculated as gold label per square millimeter of bacterial surface area. Four separate experiments. ****p < 0.0001.
The recombinant VWA domain–containing globular regions of the three collagen VI α-chains induce bacterial killing by membrane disruption
To correlate the antibacterial activity of collagen VI to individual N- and C-terminal globular regions, we investigated the bactericidal effect of the recombinant proteins on streptococci. Bacteria from the S. pyogenes strain AP1 were incubated with increasing concentrations of protein and analyzed by viable count assays. The bacteria showed dose-dependent killing in the presence of the different fragments (Fig. 3A). Interestingly, N- and C-terminal regions from the α3(VI) were more potent than the respective domains from the α1(VI) and α2(VI). To analyze these findings at the ultrastructural level, specimens of bacteria incubated with VWA domains were examined by high-resolution scanning electron microscopy. As depicted in Fig. 3B, bacteria indeed showed significant structural alterations such as membrane perturbations, blebbing, and exudation of cytoplasmic constituents. Notably, streptococci treated with α3N and α3C displayed significantly more cell structure decay. In contrast, control bacteria treated under similar conditions with only buffer were not affected (Fig. 3B, top). These results implicate that the individual VWA domain–containing globular regions of collagen VI exhibit a mode of bacterial killing that is similar to the holoprotein.
Dose-dependent killing of S. pyogenes by recombinant globular domains of collagen VI. (A) Bacteria (2 × 106 CFU/ml) were incubated with recombinant fragments at the concentrations indicated for 2 h at 37°C with 5% CO2. Six separate experiments. ***p < 0.0003, ****p < 0.0001. (B) Representative scanning electron microscopy imaging of membrane disruption by the recombinant globular domains of collagen VI. Extensive membrane disruption and leakage of intracellular contents are observed in the presence of these proteins (2 μM) and are indicated with arrowheads. Scale bar, 5 μm. One representative experiment out of three is shown.
Dose-dependent killing of S. pyogenes by recombinant globular domains of collagen VI. (A) Bacteria (2 × 106 CFU/ml) were incubated with recombinant fragments at the concentrations indicated for 2 h at 37°C with 5% CO2. Six separate experiments. ***p < 0.0003, ****p < 0.0001. (B) Representative scanning electron microscopy imaging of membrane disruption by the recombinant globular domains of collagen VI. Extensive membrane disruption and leakage of intracellular contents are observed in the presence of these proteins (2 μM) and are indicated with arrowheads. Scale bar, 5 μm. One representative experiment out of three is shown.
The VWA domains of the collagen VI α3-chain contain amphipathic amino acid motifs with putative antibacterial activity
Cationic and amphiphilic properties are essential for membrane permeabilizing peptides (39). The combination of these properties generally governs the extent to which these peptides exert their bioactivity (2, 40). Therefore, as a next approach, we decided to perform an in silico sequence analysis of the VWA domains of the α3(VI)-chain to identify such amino acid motifs with putative antimicrobial activity. The α3(VI) was chosen because it turned out to be most efficient in bacterial killing as described above. First, we defined the possible secondary structure by aligning the sequences of the N- and C-terminal VWA domains (N10-N2 and C1, see Fig. 2A) using Swiss PDB Viewer program. This analysis revealed that these domains are predicted to assume α-helices (Fig. 4, rectangular boxes) as well as β-strands (Fig. 4, black arrows). Furthermore, three-dimensional models generated with ModBase proposed that these domains consist of a central six-stranded hydrophobic β-sheet flanked on either side by three amphipathic α-helices (data not shown). These findings are in general accordance with structural data obtained by x-ray crystallography of the mouse α3N5 domain (27). The VWA domains N1 and C2 (see Fig. 2A) were not included in this study because there were no molecular models available in any database. Next, we determined amino acids that were likely to be exposed on the surface (Fig. 4, bold letters) to predict possible interaction site(s) between these domains and the bacterial membrane. By combining these results together with positively charged areas (Fig. 4, highlighted in yellow) in the sequence, we were able to predict putative antimicrobial regions as indicated in blue boxes.
Structural alignment of VWA domains in human collagen VI α3-chain generated by structural superimposition of the VWA domain models. Underneath the sequence, α-helices and β-strands are indicated with rectangular boxes and black arrows, respectively. The exposed amino acids are denoted in bold letters, and cationic stretches are highlighted in yellow. The blue boxes in the sequence indicate the location of the cationic peptides as well as the control peptide (DVN32).
Structural alignment of VWA domains in human collagen VI α3-chain generated by structural superimposition of the VWA domain models. Underneath the sequence, α-helices and β-strands are indicated with rectangular boxes and black arrows, respectively. The exposed amino acids are denoted in bold letters, and cationic stretches are highlighted in yellow. The blue boxes in the sequence indicate the location of the cationic peptides as well as the control peptide (DVN32).
Peptides derived from the VWA domains of the collagen VI α3-chain exert bactericidal activity
The identification of putative antimicrobial regions prompted us to select peptide sequences thereof with a high total net charge and hydrophobicity, previously demonstrated to be important prerequisites for antimicrobial peptides (41, 42). In total, five peptides were chosen from the N3, N2, and C1 domains (Fig. 5A, 5B). Surface representation models were also generated to get an overview of the net charge of these domains, and indeed, the C1 and N3 domains displayed a large number of cationic regions on their surface (Fig. 5A). Similar patterns, although to a somewhat lesser extent, were found for N2. The N10 domain showed more anionic residues on its surface (Fig. 5A), and a peptide synthesized from that domain was used as a negative control (DVN32).
(A) Surface representation of VWA domains of α3(VI)-chain show the electrostatic properties (blue = positive charge; red = negative charge). (B) The ribbon diagrams show the location of the cationic peptides and the negative control peptide (DVN32). The biophysical properties of the peptides. aPeptides are identified by their first three N-terminal residues using the single-letter code, followed by the total number of residues constituting the peptide. bSequences of peptides are given in single-letter code. cTheoretical isoelectric point (pI) calculated by using the Protparam tool available at http://us.expasy.org/tools/protparam.html.
(A) Surface representation of VWA domains of α3(VI)-chain show the electrostatic properties (blue = positive charge; red = negative charge). (B) The ribbon diagrams show the location of the cationic peptides and the negative control peptide (DVN32). The biophysical properties of the peptides. aPeptides are identified by their first three N-terminal residues using the single-letter code, followed by the total number of residues constituting the peptide. bSequences of peptides are given in single-letter code. cTheoretical isoelectric point (pI) calculated by using the Protparam tool available at http://us.expasy.org/tools/protparam.html.
To verify the antibacterial activity of the selected VWA-derived peptides, all peptides were screened in RDA for bactericidal activity against E. coli, S. aureus, and P. aeruginosa. All the peptides exhibited significant bactericidal activity against all tested strains to varying extent (Fig. 6A). Interestingly, in most cases, the observed bacterial killing potential was considerably higher than our positive control, the benchmark host defense peptide LL-37. In the presence of physiological NaCl, some of the peptides retained their activity against Gram-negative bacteria but not against Gram-positive bacteria (Fig. 6B). Addition of calcium and magnesium, respectively, showed that the antimicrobial activity of the peptides was totally blocked (data not shown). In another experimental set-up, the effect of bacterial killing potency was studied by determination of the dose dependence by microcalorimetric growth assays. The calorimetric assay determines the metabolic rate of the bacteria by continuous measurement of the heat flow (joules per second) Fig. 7. The data readout gives rise to a specific kinetic profile determined by the combination of the bacterial species and the growth conditions, determined by the media used and the addition of antibacterial compounds. The heat output data are integrated over time to show the accumulation of released energy (Supplemental Figs. 2 and 3). The heat (joule) accumulated over time is proportional to the formation of biomass and thus the bacterial growth rate. The slope of the curve is a representation of growth rate. By presenting the total energy released for a given time interval against compound concentration, as shown in Supplemental Figs. 2 and 3, it is possible to determine the dose response based on total bacterial metabolism for the experiment. A shift in the lag time to maximal growth rate is indicative of a bactericidal effect of the peptide tested, and a shift in maximal growth rate is indicative of a bacteriostatic effect. It should be noted that the total released energy can be the same for multiple concentrations if there is an initial delay in bacterial growth followed by a later onset of growth (e.g., Supplemental Fig. 2, LL-37). Three different aspects of growth and compound inhibitory effects can be followed using this system: shift in lag phase duration, maximum growth rate, and maximum peak output energy. Because calorimetric measurement is not dependent on the samples, properties such as turbidity of the media or geometric shape of the sample bacterial growth and metabolism can be followed in both liquid and solid samples without any intervention or additions for a prolonged period of time.
Antibacterial activity of peptides derived from α3(VI)-chain. (A) For determination of antibacterial activities, the indicated bacterial isolates (4 × 106 CFU) were inoculated in agarose gel and loaded with peptides (at 100 μM). LL-37 or buffer only were used as a positive and negative control, respectively. The clearance zones correspond to the inhibitory effect of each peptide after incubation at 37°C for 18–24 h. (B) The antimicrobial activity of the peptides in physiologic salinity (150 mM NaCl). Four separate experiments.
Antibacterial activity of peptides derived from α3(VI)-chain. (A) For determination of antibacterial activities, the indicated bacterial isolates (4 × 106 CFU) were inoculated in agarose gel and loaded with peptides (at 100 μM). LL-37 or buffer only were used as a positive and negative control, respectively. The clearance zones correspond to the inhibitory effect of each peptide after incubation at 37°C for 18–24 h. (B) The antimicrobial activity of the peptides in physiologic salinity (150 mM NaCl). Four separate experiments.
Collagen VI host defense peptides exert a long-term bactericidal effect in vitro. The curves shown are representative data for the power output of bacterial growth, heat flow (microwatt) over time, for test of concentration dependent inhibition of the growth of P. aeruginosa. All experiments were performed in triplicates.
Collagen VI host defense peptides exert a long-term bactericidal effect in vitro. The curves shown are representative data for the power output of bacterial growth, heat flow (microwatt) over time, for test of concentration dependent inhibition of the growth of P. aeruginosa. All experiments were performed in triplicates.
Comparison of the different peptides tested shows significant variations in mode of action as well as potency of inhibition. The benchmark peptide LL-37 displays a clear shift in the lag time but no shift in growth rate even at the highest concentrations, and the total energy is consistent for all concentrations, all indicative of a bactericidal activity. The growth of the bacteria is, however, not inhibited when incubated for a prolonged time-period, and a completed growth curve is acquired for all peptide concentrations. The VTT30 peptide displays both a shift in lag phase duration and some shift in maximal growth rate combined with a distinct decrease in maximal peak–power output. The combined power output also shows that growth is sustained even at the highest concentrations. The remaining peptides FYL25, FFL25, GVR28, and SFV33 all display much more potent antimicrobial properties as shown by a combined shift in lag phase time, decrease in growth rate for the higher concentrations, and decrease in peak output power (Fig. 7, Supplemental Fig. 2). The total energy release shown in Supplemental Fig. 3 also shows higher dose-response dependence for these peptides, with an almost complete inhibition of growth for the highest concentrations. Interestingly, there are signs of residual metabolic activity for all peptides except LL-37 and SFV33. As seen in Fig. 7, there is remaining metabolic activity at the end part of each growth curve except for SFV33, in which a very steep decline in metabolic output is monitored. Prolonged metabolic activity at the end of a growth curve is consistent with the formation of bacterial biofilm and/or the presence of quiescent cells (not tested in this study). There is a very sharp concentration step in which total inhibition of P. aeruginosa growth is inhibited by the SFV33 peptide, between 5 and 20 μM under the conditions used in our study.
The bacterial assays exhibited decay of the structural integrity similar to LL-37, which exerts its direct bactericidal activity by disruption of the phospholipid bilayer (43). To establish that the activity observed for the collagen-derived peptides shared the same general mechanism of action, a liposome leakage assay was performed. The unilamellar liposomes used were produced from an E. coli polar lipid extract. This extract, in terms of lipid headgroup composition, is virtually identical to the typical P. aeruginosa membrane and can therefore be seen as a model for both organisms. All of the cationic collagen–derived peptides exhibited substantial membrane permeabilizing activity (Fig. 8). The two peptides with highest net positive charge, FYL25 and SFV33, were the most potent, even surpassing that of LL-37. Although still very active on the liposomes, VTT30 was the least membrane disrupting of the cationic collagen peptides, which correlated well with its slightly lower activity on, for example, P. aeruginosa bacteria in the RDA assay. The anionic collagen peptide, DVN32, did not exhibit any signs of leakage generation (up to 20 μM) on these model membranes. The liposome leakage results showed that all the cationic collagen–derived peptides were exceptionally potent at inducing membrane permeabilization of bacterial phospholipid membranes at physiological pH. This indicates that the antibacterial effects observed are similar in nature to that of LL-37 and include perforations of the cytoplasmic membrane by for example pore formation.
Membrane permeabilization on E. coli liposomes generated by the collagen VI–derived peptides. The graph shows the efflux levels of the self-quenching fluorophore and carboxyfluorescein, after 45 min of incubation, as a function of peptide concentration. The 100-nm sized unilamellar liposomes were manufactured from an E. coli polar lipid extract. Each marker represents the mean leakage at 37°C in 10 mM Tris buffer (pH 7.4) with SD from triplicate experiments done at individual peptide concentrations (i.e., no cumulative additions). The EC50 values are calculated from the plotted sigmoidal dose-response curves. The well-characterized membrane disruptive peptide LL-37 is used for activity comparison. Three independent experiments were performed.
Membrane permeabilization on E. coli liposomes generated by the collagen VI–derived peptides. The graph shows the efflux levels of the self-quenching fluorophore and carboxyfluorescein, after 45 min of incubation, as a function of peptide concentration. The 100-nm sized unilamellar liposomes were manufactured from an E. coli polar lipid extract. Each marker represents the mean leakage at 37°C in 10 mM Tris buffer (pH 7.4) with SD from triplicate experiments done at individual peptide concentrations (i.e., no cumulative additions). The EC50 values are calculated from the plotted sigmoidal dose-response curves. The well-characterized membrane disruptive peptide LL-37 is used for activity comparison. Three independent experiments were performed.
Collagen VI and its derivative peptides show bactericidal activity in vivo
The bacterial killing observations were also confirmed by immunohistochemistry of patient biopsies obtained from individuals with skin infection and diabetic foot ulcer. High levels of the collagen VI holoprotein and its derived peptides were found in skin samples with group A Streptococcus infection. In contrast, in healthy skin controls, only full-length collagen VI was observed, showing a normal dermal distribution with highest levels just underneath the epidermis (Fig. 9). High-resolution scanning electron microscopy and immunoelectron microscopy of ultrasections from the same specimens showed large amounts of Gram-negative and Gram-positive bacteria in different patient biopsies (Fig. 10). Frequently, the bacteria were embedded in biofilm (Fig. 10A–C). The collagen VI–derived peptides bound to all the different bacteria (Fig. 10D–G) and, interestingly, were often identified in accumulations of membrane vesicles of killed bacteria (Fig. 10H–K). Fig. 10 shows these properties for GVR28, but the same observations were made for all peptides (data not shown). Taken together, these findings demonstrate that the VWA domains of the α3(VI)-chain contain several antimicrobial motifs that may account for the observed bactericidal properties of the collagen VI molecule and highlight the role of collagen VI as a bactericidal barrier in vivo.
Collagen VI and its derivative peptides are expressed in human skin infections in vivo. Skin biopsy specimen from healthy individuals (skin control) or patients infected with group A streptococci (skin infection) were processed for immunohistochemistry staining for collagen VI and collagen VI–derived peptides (GVR28, FYL25, FFL25, VTT30, and SFV33) as described in 2Materials and Methods. In healthy skin, collagen VI expression was depicted just beneath epidermis (i) and in dermis (ii) region, whereas no detection was observed for collagen VI peptides. In infected skin, both collagen VI and collagen VI–derived peptide expressions were elevated in epidermis (iii) and dermis regions (iv). Upper panels (i and iii), Overviews at low magnification (scale bars, 1 μm). Lower panels (ii and iv), Higher magnifications of the dermis, respectively (scale bars, 200 nm). One representative experiment out of five is shown.
Collagen VI and its derivative peptides are expressed in human skin infections in vivo. Skin biopsy specimen from healthy individuals (skin control) or patients infected with group A streptococci (skin infection) were processed for immunohistochemistry staining for collagen VI and collagen VI–derived peptides (GVR28, FYL25, FFL25, VTT30, and SFV33) as described in 2Materials and Methods. In healthy skin, collagen VI expression was depicted just beneath epidermis (i) and in dermis (ii) region, whereas no detection was observed for collagen VI peptides. In infected skin, both collagen VI and collagen VI–derived peptide expressions were elevated in epidermis (iii) and dermis regions (iv). Upper panels (i and iii), Overviews at low magnification (scale bars, 1 μm). Lower panels (ii and iv), Higher magnifications of the dermis, respectively (scale bars, 200 nm). One representative experiment out of five is shown.
Antimicrobial activity of collagen VI–derived peptides in vivo. (A–C) Scanning electron micrographs of patient biopsies infected with Gram-negative (arrows) and Gram-positive (arrowheads) micro-organisms. (A) Skin fasciotomy specimen from a patient infected by group A Streptococcus. Note the granular appearance of internal surfaces due to massive colonization by streptococci (arrows). Scale bar, 20 μm. (B) Higher magnification of the same area. Globular streptococci embedded in biofilm are visible all over the biopsy (arrows). Scale bar, 5 μm. (C) Fibrin slough sample from a patient with diabetic leg ulcer, colonized by Gram-negative (arrowheads) and Gram-positive (arrows) micro-organisms. Scale bar, 5 μm. (D–K) Transmission immunoelectron microscopy of colocalization of collagen VI peptide GVR28 in patient biopsies. (D and H) S. pyogenes, (E and I) S. aureus, (F and J) P. aeruginosa, and (G and K) E. coli bacteria are identified in skin biopsies and fibrin slough by specific Abs (15-nm gold conjugates). (D–G) All bacterial species contain GVR28 particles (12-nm gold conjugates) bound to the bacterial membrane and internalized. Scale bar, 500 nm. (H–K) Killed bacteria showing extensive membrane rupture are frequently observed in all specimens. Scale bar, 200 nm. One representative experiment out of three is shown.
Antimicrobial activity of collagen VI–derived peptides in vivo. (A–C) Scanning electron micrographs of patient biopsies infected with Gram-negative (arrows) and Gram-positive (arrowheads) micro-organisms. (A) Skin fasciotomy specimen from a patient infected by group A Streptococcus. Note the granular appearance of internal surfaces due to massive colonization by streptococci (arrows). Scale bar, 20 μm. (B) Higher magnification of the same area. Globular streptococci embedded in biofilm are visible all over the biopsy (arrows). Scale bar, 5 μm. (C) Fibrin slough sample from a patient with diabetic leg ulcer, colonized by Gram-negative (arrowheads) and Gram-positive (arrows) micro-organisms. Scale bar, 5 μm. (D–K) Transmission immunoelectron microscopy of colocalization of collagen VI peptide GVR28 in patient biopsies. (D and H) S. pyogenes, (E and I) S. aureus, (F and J) P. aeruginosa, and (G and K) E. coli bacteria are identified in skin biopsies and fibrin slough by specific Abs (15-nm gold conjugates). (D–G) All bacterial species contain GVR28 particles (12-nm gold conjugates) bound to the bacterial membrane and internalized. Scale bar, 500 nm. (H–K) Killed bacteria showing extensive membrane rupture are frequently observed in all specimens. Scale bar, 200 nm. One representative experiment out of three is shown.
Discussion
Our major finding in this study is the identification of amphipathic cationic motifs with bactericidal activity in VWA domains of noncollagenous globular regions of collagen VI. Our in vitro and in vivo data suggest that bacterial killing occurs by direct membrane targeting and disruption of both Gram-negative and Gram-positive bacteria. We also show for the first time, to our knowledge, that all six different recombinant N- and C- terminal VWA domain–containing regions from all three α-chains display this effect. The bactericidal activity is, at least in part, mediated through the antimicrobial peptide regions described in this article that are localized in these domains. The data further substantiate our concept of collagen VI as an innate host defense molecule in connective tissues with broad-spectrum bactericidal properties.
Collagen VI is broadly expressed in a variety of connective tissues. It forms characteristic extended microfibrillar networks that anchor interstitial structures and larger collagen fibrils. Collagen VI is thought to be involved in maintenance of tissue structure and function by contributing to ECM homeostasis and remodeling. It belongs to the superfamily of proteins containing VWA domains that mediate protein–protein interactions and cell adhesion (27). In addition to these properties, we have recently reported that collagen VI displays adhesive and bactericidal effects against respiratory tract pathogens at physiological conditions (23, 24). The observations in the current study extend our previous findings and demonstrate that collagen VI also recruits (Supplemental Fig. 4) and kills pathogens such as S. pyogenes, S. aureus, E. coli, and P. aeruginosa by membrane destabilization and cytoplasmic exudation, thereby supporting the hypothesis that collagen VI is a broad-spectrum antimicrobial innate host defense molecule against a variety of Gram-positive and Gram-negative pathogens. The combined adhesive and bactericidal properties of collagen VI networks contribute to the innate connective tissue host defense and may protect against pathogen intruders by containment and rapid clearance.
During the past years, the strong correlation between heparin-binding and bactericidal properties of antimicrobial peptides has become commonly accepted knowledge (19, 20). Over recent years, several thousand different host defense molecules of the innate immune system have been reported. They have been found within all taxonomic kingdoms of cellular life. They can be isolated from many different sources, such as neutrophils, macrophages, and epithelial cells. In contrast, only a limited number of heparin-binding peptides derived from ECM components, such as laminin, vitronectin, and fibronectin, have been described as host defense molecules (20). However, to our knowledge, collagen VI is the only ECM molecule reported so far with direct killing properties, per se, as a holoprotein. Interestingly, this property could be mimicked by and assigned to particular domains in the molecule. Thus, all recombinantly expressed VWA domains showed binding to heparin. The strongest effect was observed for α3N, α3C, and α2N, whereas the other domains showed somewhat less binding. Similarly, all VWA domains bound to the bacterial surface in a heparin-dependent way as shown by negative staining electron microscopy. These findings are in general agreement with the heparin-binding properties of the recombinant N-terminal globule of the collagen VI α3-chain as was described earlier (15). Correspondingly, all domains were able to kill bacteria. Interestingly, the domains with the strongest affinity showed the most pronounced killing effect at micromolar concentrations similar to the benchmark peptide LL-37. This observation may reflect the fact that the number of VWA domains within a given collagen VI α-chain plays a significant role in bacterial clearance efficiency because the α3-chain contains 10 N-terminal VWA domains and two C-terminal VWA domains compared with α1- and α2-chains, which only have one N-terminal VWA domain and two C-terminal domains, respectively. Taken together, the ability of the globular VWA domains of collagen VI to interact with negatively charged bacterial surfaces and subsequently induce bacterial cell death is a feature compatible with the function of the majority of other antimicrobial peptides.
Based on these results, we wished to gain more detailed insight in the bactericidal nature of collagen VI. We searched the amino acid sequence for putative antimicrobial peptide stretches in an in silico approach. Our initial focus was on the VWA domains of the α3-chain, as this chain was most effective in bacterial killing. Indeed, both N-and C-terminal globular regions of the α3-chain contained several amphipathic stretches (i.e., cationic, hydrophobic, and helical). These properties are prerequisite for most antimicrobial peptides described so far for membrane targeting, insertion, and destabilization. In combination with a structural analysis and molecular modeling, we could pinpoint particular VWA domains with an overall positive charge on their surface. Interestingly, the N3 and C1 domains had a significantly higher number of positive residues and thus a more pronounced positive net charge than other domains such as N10, which turned out to be more anionic. By combining these data, we selected five peptide candidates from the N3, N2, and C1 domains for synthesis. Indeed, the corresponding peptides exerted killing activity against both Gram-positive and Gram-negative bacteria in vitro and in vivo by membrane disruption according to leakage assays and immunoelectron microscopy. The liposome leakage results showed that all the cationic collagen VI–derived peptides were exceptionally potent at inducing membrane permeabilization of bacterial phospholipid membranes at physiological pH. Striking similarity was observed between the cationic collagen VI–derived peptides and LL-37, both in the liposome leakage kinetics (fast onset that reaches a plateau within a few minutes) and the profile of its peptide concentration dependence in Fig. 8. This suggests that the antibacterial effects observed are similar in nature to that of LL-37 and could likely include perforations of the cytoplasmic membrane by, for example, pore formations. Noteworthy in this respect was the evidence of membrane destabilization, blebbing, and disruption by collagen VI on the bacterial surface. The observed morphology of bacteria after treatment suggests that the breakdown is comparable with what others have associated with the so-called carpet model mechanism (44). This is essentially positive curvature strain, and similar structural phenomena have been monitored on liposomes with other linear cationic peptides, such as melittin (45). Interestingly, in RDA, collagen VI–derived peptides had a considerably stronger inhibitory effect on bacterial growth as compared with LL-37. Some of these peptides even retained their activity under physiological salt concentrations. Similar observations were made with several other synthetic peptides derived from the α3-chain (data not shown). The calorimetric growth assays clearly confirm the antimicrobial properties of the tested peptides. The metabolic rates assay further shows the different impact on bacterial growth on inhibition properties as well as the potency of bacterial killing in a planktonic growth system using P. aeruginosa as model organism. Calorimetric assessment of growth inhibition is a valuable tool to determine the properties of lead drug candidates and rank the desired properties. The antibacterial properties are mostly consistent with bactericidal action at lower concentrations with total inhibition of growth for the highest concentrations tested and a very complete shutdown of bacterial growth using the SFV33 peptide.
In vivo, they interacted with both structurally intact and killed Gram-negative and Gram-positive bacteria and were observed both bound to the membrane and internalized. These findings suggest that the described antimicrobial peptides account, at least in part, for the bactericidal properties of the collagen VI VWA domains. In addition, they kill bacteria at least as efficiently as the well-known cathelicidin peptide LL-37. Nevertheless, it remains an exciting future challenge to investigate how these peptides are generated in vivo. Such active collagen VI peptides may be released by proteolysis through granulocyte proteases during specific phases in inflammation.
Taken together, these findings suggest that collagen VI is part of the constitutive host defense present in connective tissues during normal and inflammatory conditions. The collagen VI–derived peptides from VWA regions could be further characterized and used in the development of novel antibacterial agents for the treatment of infectious disease.
Acknowledgements
We thank Ann-Charlotte Strömdahl for technical support. We thank the staff from the Core Facility for Integrated Microscopy, Panum Institute, University of Copenhagen, for providing a cutting edge environment for electron microscopy. We thank IQ Biotechnology Platform, Infection Medicine, Lund University, for preforming light microscopy. We also thank Magnus Jansson and Göran Conradson from Symcel Sverige AB for providing calScreener isothermal microcalorimetry assay.
Footnotes
This work was supported by the Swedish Foundation for Strategic Research (Grant SB12-0019), Lennanders Stiftelse, the Swedish Research Council (Project 7480), the Crafoord Foundation, the Johan and Greta Kock Foundation, the Alfred Österlund Foundation, the King Gustav V Memorial Fund Foundation, the Royal Physiographic Society of Lund, the Medical Faculty at Lund University, and Deutsche Forschungsgemeinschaft (SFB829).
The online version of this article contains supplemental material.
References
Disclosures
S.M.A. and M.M. have submitted a patent application regarding the collagen VI–derived peptides. The other authors have no financial conflicts of interest.