Author + information
- Received February 2, 2015
- Revision received June 24, 2015
- Accepted July 14, 2015
- Published online September 22, 2015.
- Merry L. Lindsey, PhD∗,†,‡,
- Rugmani Padmanabhan Iyer, PhD∗,†,
- Rogelio Zamilpa, PhD†,
- Andriy Yabluchanskiy, MD, PhD∗,†,
- Kristine Y. DeLeon-Pennell, PhD∗,†,
- Michael E. Hall, MD§,
- Abdullah Kaplan, MD∗,†,
- Fouad A. Zouein, PhD∗,†,
- Dustin Bratton, BSc∗,†,
- Elizabeth R. Flynn, BSc∗,†,
- Presley L. Cannon, BSc∗,†,
- Yuan Tian, PhD∗,†,
- Yu-Fang Jin, PhD†,‖,
- Richard A. Lange, MD†,¶,
- Dorota Tokmina-Roszyk, MSc†,#,
- Gregg B. Fields, PhD†,# and
- Lisandra E. de Castro Brás, PhD∗,†∗∗∗ ()
- ∗Mississippi Center for Heart Research, Department of Physiology and Biophysics, University of Mississippi Medical Center, Jackson, Mississippi
- †San Antonio Cardiovascular Proteomics Center, University of Texas Health Science Center, San Antonio, Texas
- ‡Research Service, G.V. (Sonny) Montgomery Veterans Affairs Medical Center, Jackson, Mississippi
- §Division of Cardiology and Department of Physiology and Biophysics, University of Mississippi Medical Center, Jackson, Mississippi
- ‖Department of Electrical and Computer Engineering, University of Texas at San Antonio, San Antonio, Texas
- ¶Paul L. Foster School of Medicine, Texas Tech University Health Science Center, El Paso, Texas
- #Florida Atlantic University, Department of Chemistry and Biochemistry, Jupiter, Florida
- ∗∗Department of Physiology, Brody School of Medicine, East Carolina University, Greenville, North Carolina
- ↵∗Reprint requests and correspondence:
Dr. Lisandra E. de Castro Brás, Department of Physiology, The Brody School of Medicine, East Carolina University, 600 Moye Boulevard, Greenville, North Carolina 27834.
Background Proteolytically released extracellular matrix (ECM) fragments, matricryptins, are biologically active and play important roles in wound healing. Following myocardial infarction (MI), collagen I, a major component of cardiac ECM, is cleaved by matrix metalloproteinases (MMPs).
Objectives This study identified novel collagen-derived matricryptins generated post-MI that mediate remodeling of the left ventricle (LV).
Methods Recombinant collagen Ia1 was used in MMPs cleavage assays, the products were analyzed by mass spectrometry for identification of cleavage sites. C57BL6/J mice were given MI and animals were treated either with vehicle control or p1158/59 matricryptin. Seven days post-MI, LV function and parameters of LV remodeling were measured. Levels of p1158/59 were also measured in plasma of MI patients and healthy controls.
Results In situ, MMP-2 and -9 generate a collagen Iα1 C-1158/59 fragment, and MMP-9 can further degrade it. The C-1158/59 fragment was identified post-MI, both in human plasma and mouse LV, at levels that inversely correlated to MMP-9 levels. We synthesized a peptide beginning at the cleavage site (p1158/59, amino acids 1159 to 1173) to investigate its biological functions. In vitro, p1158/59 stimulated fibroblast wound healing and robustly promoted angiogenesis. In vivo, early post-MI treatment with p1158/59 reduced LV dilation at day 7 post-MI by preserving LV structure (p < 0.05 vs. control). The p1158/59 stimulated both in vitro and in vivo wound healing by enhancing basement membrane proteins, granulation tissue components, and angiogenic factors.
Conclusions Collagen Iα1 matricryptin p1158/59 facilitates LV remodeling post-MI by regulating scar formation through targeted ECM generation and stimulation of angiogenesis.
Despite current treatment strategies, progression to heart failure can occur in up to one-third of myocardial infarction (MI) patients as a result of adverse remodeling of the collagenous scar that replaces the necrotic myocardium (1). Although excessive collagen deposition may result in a stiff and noncompliant left ventricle (LV), insufficient collagen deposition may produce LV thinning and dilation. Therefore, collagen levels are a critical determinant of LV remodeling.
Cardiac extracellular matrix (ECM) dynamically interacts with cells to regulate cell–cell and cell-ECM interactions (2). ECM proteins modulate cell proliferation, migration, adhesion, differentiation, and survival (3). Although ECM protein synthesis is controlled by growth factors, degradation is regulated by proteases, particularly matrix metalloproteinases (MMPs) (2). Collagen Iα1 is a major cardiac ECM component of the post-MI scar and a substrate for several MMPs, including MMP-1, -2, -8, and -9 (4). Davis et al. (5) introduced the term matricryptin to describe proteolytically released biologically active ECM fragments. Elastin-derived matricryptins induce leukocyte activation, smooth muscle cell proliferation, and MMP expression (6–8). Hyaluronan-derived matricryptins interact with Toll-like receptor-2 and -4 to modulate the inflammatory and fibrotic responses (9). The collagen Iα1 matricryptin DGGRYY, amino acids 1200 to 1205, is a potent neutrophil activator, and the tripeptide GHK (derived from both collagen Iα2 and secreted protein acidic and rich in cysteine [SPARC]) is chemotactic for monocytes, macrophages, and mast cells and has proangiogenic properties (10–12). Endostatin, an MMP-9–released C-terminal fragment from collagen XVIII, has antiangiogenic properties, whereas matricryptins from collagen IV are both pro- and antiangiogenic (13,14). MMP-9 cleaves collagen IVα3 to produce tumstatin, which inhibits angiogenesis (15). These reports suggest that ECM-generated matricryptins may have important biological activity during the healing cascade that follows MI.
Plasma MMP-9 levels directly correlate with LV dysfunction post-MI, in both human and animal models. In mouse models, MMP-9 deletion improves LV remodeling and cardiac function post-MI (16). Analysis of the wild-type LV infarct proteome showed increased levels of a low-molecular-weight collagen Iα1 fragment, compared with baseline day 0 levels (17). Collagen and collagen fragments are known MMP-9 substrates, and MMP-9–null mice have better LV function post-MI (16). However, whether this is a result of accumulation of collagen matricryptins has not been investigated. In this report, we evaluated the mechanisms of generating a novel collagen fragment and its roles in post-MI remodeling.
Animal studies were performed according to protocols approved by the institutional animal care and use committee at the University of Mississippi Medical Center. Written informed consent was collected from all patients under protocols approved by the University of Mississippi Medical Center institutional review board committee. Experimental details are provided in the Online Appendix.
Results and Discussion
Previously, we identified an ∼30-kDa C-terminal collagen Iα1 fragment in the infarcted LV (LVI) at 7 days post-MI (Figure 1A) (17). Because MMP-9 is prominent post-infarction, we performed cleavage assays using a recombinant collagen Iα1 that contains the N- and C-terminal regions and 66 Gly-X-Y triplet repeats (residues Gln23-Lys277 + Gly1094-Leu1464). Mass spectrometry analysis of the cleavage products revealed a new MMP-9 cleavage site between amino acids 1158/59 in the C-terminal region, 37 amino acids upstream of the collagen C-telopeptide (Figure 1B). The generated C-collagen Iα1 fragment (C-1158/59) is downstream of the canonical collagenase cleavage site at Gly775-Ile776 (18).
Using an antibody specific for the 1158/59 cleavage site, we evaluated the temporal generation of C-1158/59 fragment in the LVI and noninfarcted remote control regions in a mouse MI model. C-1158/59 LVI levels robustly increased at day 7 post-MI; it was >4-fold higher versus all earlier time-points (all p < 0.05) and 3.5-fold higher than the day 7 LV remote noninfarcted control (p < 0.05) (Figure 2A, Online Figure S1A). Post-MI, loss of myocytes and contractility in the infarct region increases wall stress in the remote LV and induces compensatory hypertrophy. Our data suggest a small degree of collagen turnover in the remote noninfarcted region in response to LV remodeling post-MI. In the LVI, MMP-9 protein concentrations increased significantly in the first week post-MI, returning to baseline (day 0) levels by day 7 (Figure 2B). This increase was consistent with our past reports and with the timing of the generation of the C-1158/59 fragment (19).
Because C-1158/59 fragment levels were higher at day 7 post-MI, when MMP-9 levels were returning to baseline, and were generated in the remote region where MMP-9 expression is low, we expanded our search to other MMPs that could cleave collagen at this site. We tested the gelatinase MMP-2 that shares highest homology and several substrates with MMP-9 and the collagenase MMP-8. MMP-11 was used as a negative control, because this stromelysin does not cleave collagen. Analysis of cleavage assays for MMP-2, -8, -9, and -11 showed that both MMP-2 and -9, but not MMP-8 or -11, can generate the C-1158/59 fragment (Online Figures S1B and S1C). Our results show, for the first time, that MMP-2 and -9 cleaved collagen Iα1 in the C-terminus at the amino acids 1158/1159.
Of note, increasing incubation times or MMP-9 doses (but not MMP-2) resulted in degradation of the C-1158/59 fragment (Figure 2C). To confirm this in vivo, we quantified C-1158/59 levels from LVI protein extracts of MMP-9-null mice at day 7 post-MI, incubated in the presence or absence of MMP-9. We established that C-1158/59 is generated in the MMP-9–null mice and incubation with MMP-9 degraded the fragment (Figure 2D). This confirmed that C-1158/59 can be generated by other MMPs and suggested that decreased C-1158/59 levels at days 3 and 5 post-MI in the wild-type mice might be caused by MMP-9 degradation.
In vitro treatment with the P1158/59 peptide
To study the biological activity of the C-1158/59 fragment (Central Illustration), we synthesized a short peptide containing the 15 amino acids downstream of the cleavage site, beginning at amino acid 1159 (p1158/59; molecular weight 1,360.4 g/mol). The use of small synthetic peptides to mimic the biological activities of naturally occurring molecules has great potential in overcoming difficulties associated with synthesis of larger peptides. A short peptide is easier to synthesize, purify, and concentrate. Although technological advances have improved peptide synthesis strategies, the purity and yield of synthesized peptides can be limited by the sequence length. Additionally, a small size peptide limits response variability, which can occur with a longer peptide that may possess multiple activities; by using a small peptide, we ensured that cellular effects occurred as a direct response to the cleavage site sequence and not peptide structural properties. Cardiac fibroblasts were stimulated with increasing doses of p1158/59 or the spanning peptide negative control. The peptides did not induce cell apoptosis or affect cell proliferation (n = 6 to 10/dose/assay) (Online Figures S2A and S2B).
We performed automated wound healing assays and observed that wound closure rates were enhanced with 1 μmol/l p1158/59 treatment and peaked at 22 h compared with untreated controls (Figure 3A). Because proliferation rates were not different, the increase in wound healing rates was due to increased cell migration. Gene array analysis on the treated cells revealed that Lama1 (laminin α1), Mmp-9, and Sparc expression were all increased in cells treated with p1158/59 (Figure 3B) (all p < 0.05). Laminin α1 is a glycoprotein involved in basement membrane assembly and architecture, and plays an important role in cell migration through α3β1 and α6β1 integrins (20). MMP-9 is also known to facilitate migration by cleaving adhesion molecules, and SPARC has been suggested to have antiadhesive properties, therefore promoting cell migration by reducing cell–substrate adhesion (21,22). Cell migration can be enhanced by a directional cue, such as a chemoattractant, disruption of cell–cell contact (mechanical cue), internal signaling cascades, and extracellular cues such as ECM proteins, growth factors, and ligands (23). Our data suggest that the p1158/59 matricryptin acts as an extracellular cue and stimulates cell migration by inducing overexpression of migration-mediator factors. Similarly, others have reported collagen, fibronectin, and elastin-derived matricryptins that stimulate directed cell migration of leukocytes and fibroblasts (24,25). Because p1158/59 induced fibroblast MMP-9 expression, we looked at collagen levels in the conditioned media to evaluate if MMP-9 increased collagen degradation. The p1158/59 peptide did not affect levels of collagen secreted in vitro (Online Figure S2C).
Human umbilical vein endothelial cells were used to assess p1158/59 angiogenic properties. After 6 h, cells cultured in basal media and with the negative control spanning peptide showed initial stages of alignment and elongation, whereas cells treated with p1158/59 showed robust induction of a network of 3-dimensional capillary-like tubular structures (p < 0.05) (Figure 3C). Our data indicated that p1158/59 had strong proangiogenic in vitro properties.
Systemic treatment with the P1158/59 peptide
The in vitro data suggested that increased levels of p1158/59 post-MI could lead to accelerated fibroblast migration into the infarcted area, which in turn would result in earlier deposition of granulation tissue that could provide a sounder support for scar formation and maturation. Thus, we treated mice with p1158/59 or saline vehicle control 3 h post-MI for a total of 7 days. Our aim was to determine whether an early intervention strategy would stimulate scar formation and angiogenesis to attenuate LV remodeling. There was no difference in infarct size between groups (p = 0.15), indicating equal initial injury. This was expected, because treatment was started 3 h post-MI, after myocardial necrosis was fulminant.
In vivo, p1158/59 treatment resulted in attenuated increased LV end-diastolic diameter and end-systolic diameter post-MI compared with the saline group (p < 0.05) (Figure 4A), indicating that p1158/59 treatment preserved LV structural integrity and reduced LV dilation in the infarct region. Lower end-diastolic dimensions indicate decreased stretching of the LV muscle fibers during preload, which in turn would result in preserved LV structure. A study evaluating chronic effects of the peptide would be beneficial to fully understand p1158/59 effects on cardiac function post-MI. After MI, a person immediately encounters decreased cardiac output and increased strain on the heart. Prolonged inflammation and/or delayed scar formation result in progressive LV wall thinning and LV dilation, both associated with increased incidence of heart failure and patient mortality. Preserved chamber geometry and reduced dilation, compared with the saline group, indicated better LV contractility. Overall, our data demonstrated that treatment with p1158/59 reduced LV dilation 7 days post-MI, which opens therapeutic avenues for this and other matricryptins. The literature describes several studies in which matricryptins or other peptides are used in MI models with different rates of success. In a rat model, treatment with B-type natriuretic peptide for 7 days improved LV function at 8 weeks post-MI (26). Ladage et al. (27) delivered recombinant periostin peptide into the pericardial space post-MI in a pig model and observed a decrease in infarct size and improved LV function compared with controls.
We further investigated which changes were occurring at the molecular level during scar formation that could explain preserved LV structure after peptide delivery. We determined that p1158/59-treated mice showed less fibrosis in the LVI, evidenced by lower levels of collagen deposition in the infarcted area at day 7 post-MI (p < 0.05 vs. saline) (Figure 4B). We used picrosirius red staining for histological visualization of collagen I and III fibers (27). Sirius red dye molecules align parallel to the long axis of collagen molecules to enhance their natural birefringence (28). Under polarized light, interstitial collagens show different colors and intensities of birefringence. Collagen I displays yellow or red thick, strongly birefringent fibers that are highly cross-linked, whereas collagen III appears as thin, more compliant green fibers with lower birefringence (28). Of note, although the saline group showed thick, strongly birefringent fibers consistent with collagen I deposition, p1158/59 treatment resulted in thin green fibers with lower birefringence, suggestive of increased collagen III deposition. By gene array, we confirmed that p1158/59 stimulated Col3a1 expression by 3.3-fold (Table 1). Collagen III is the first fibrillar collagen produced by fibroblasts during the proliferative phase after injury, later being replaced by collagen I, which is stiffer and less compliant than collagen III (29). These data suggest that p1158/59 stimulated collagen III deposition, resulting in a decreased ratio of collagen I/collagen III, which would generate a less stiff scar and may explain the better LV geometry (Figure 4C). Fetal tissues show increased expression of collagen III compared with adult tissue, which has led some to hypothesize that collagen III is a main contributor to the scarless phenotype observed in the early stages of gestation (30). We further looked at the expression of MMPs and tissue inhibitors of metalloproteinases (TIMP), because MMP activity and MMP/TIMP ratios can determine ECM turnover and remodeling (31). At 7 days post-MI, only MMP-8, -11, and -14 expression were increased in the peptide-treated group compared with saline (Online Tables S1 and S2). Our data demonstrated that p1158/59 treatment improved scar formation through reduced stiffness and fibrosis, and preserved LV structure post-MI.
Treatment with p1158/59 induced 35 ECM- and adhesion molecule-associated genes in the infarct region compared with saline-treated MI controls, out of 85 genes examined (Table 1) (Online Tables S1 and S2 shows the 50 unaltered genes). Expression of Adamts2 was increased 2-fold in the treated group compared with the saline control (p < 0.05). ADAMTS2 cleaves the N-terminal propeptides of collagen types I and III during collagen maturation, therefore playing a key role in collagen turnover. Connective tissue growth factor (CTGF) expression was elevated in the treated group, CTGF promotes fibroblast proliferation, migration, adhesion, and ECM and vessel formation (32). CTGF directly binds to integrin receptors and proteoglycans, and thus participates in signal transduction events. Most biological activities of matricryptins are mediated by integrins, proteoglycans, and growth factor receptors (14). Endostatin (MMP-2 generated) and tumstatin (MMP-9 generated) exert their angiogenic effects by interacting through αvβ3 integrins, whereas angiogenic laminin peptides bind to αvβ3 and α5β1 integrins (33,34). Table 1 shows overexpression of Itga5 (integrin alpha-5), Itgb1 (integrin beta-1), and Itgb3 by p1158/59 treatment, suggesting that this matricryptin may be mediated by these integrins.
Our data suggest p1158/59 reduced the stiffness of the provisional matrix post-MI by enhancing expression of collagen III and CTGF, which would reduce fibrosis during early wound healing. The accelerated formation of a stable matrix may further support cell migration, adhesion, and vessel formation, which would contribute to preserved LV structure and reduced LV dilation.
Animals that received p1158/59 for 7 days post-MI showed enhanced expression of Col4a1, Col4a2, and Lama1 compared with the saline MI group (Table 1) (p < 0.05 for all). These genes are components of the basement membrane and stimulate angiogenesis. Collagen IV provides a stable substrate for vessel formation and promotes and regulates the formation, extension, and stabilization of microvessels during angiogenesis. Similarly, Lama1 stimulates angiogenesis and is critical for vessel integrity. Therefore, increased collagen IV and Lama1 expression by p1158/59 may promote angiogenesis and quicker formation of adaptive capillary growth during LV remodeling. Extracellular matrix protein-1 (Ecm1), a secretory glycoprotein that promotes angiogenesis by inducing endothelial cell proliferation and neovessel formation, was overexpressed in the p1158/59 group (Table 1). Interestingly, this protein has been shown through high-affinity protein–protein interactions to inhibit MMP-9 activity (35). Mmp-11, also significantly elevated in the p1158/59-treated animals (Table 1), is highly associated with cell migration and invasion (36).
To validate what we observed at the gene level, we quantified by immunoblotting and immunohistochemistry Griffonia (Bandeiraea) simplicifolia lectin I (GSL-1), which specifically binds to endothelial cells (37). Mice treated with p1158/59 displayed higher binding of GSL-1 and increased vessel numbers in both the remote and infarct regions compared with saline (Figures 5A and 5B) (p < 0.05), indicative of increased vascularization. The remote region maintains perfusion after MI; however, remodeling occurs due to changes in wall strain patterns that lead to rapid regression of cardiac mass through cardiomyocyte atrophy. The cardiomyocytes respond to the increased load by increasing protein synthesis, and the resulting hypertrophic response may lead to relative insufficient vascular growth. To assess whether the increased vessel numbers observed with p1158/59 treatment were due to myocyte atrophy, we quantified the expression of the Forkhead Box O3 (Foxo3), a transcription factor that, when activated, promotes myocyte atrophy and autophagy (38,39). No difference was observed in Foxo3 activation between groups (Online Table S2), giving evidence that p1158/59 treatment stimulated angiogenesis did not occur through Foxo3 signaling. Increased angiogenesis in both remote and infarct regions may offer the necessary cardioprotection to mitigate LV dysfunction post-MI. Taken together, our results established that p1158/59 matricryptin had robust proangiogenic properties in vivo. Endorepellin, a perlecan-derived matricryptin, displayed proangiogenic and neuroprotective properties in a rodent model of ischemic stroke and has shown promising results for the treatment of stroke (40). It would be interesting to determine whether the use of proangiogenic matricryptins such as p1158/59 in the settings of ischemia and compromised healing, such as diabetic wounds, could delay or even reverse the tissue necrosis caused by poor vascularization and chronic inflammation. Further studies are necessary for a deeper understanding of the modes of action. For example, doses to be administered in vivo require careful optimization because many angiogenic factors display biphasic response curves. Another critical issue is the dual activities of some matricryptins, which depend on environment, concentration, and/or cell type. Moreover, time and duration of intervention is also important. In our study, we delivered the matricryptin for 7 days, beginning very early post-MI, but long-term effects will need to be studied.
Finally, we quantified the C-1158/59 fragment in the plasma of healthy subjects and patients at 24 to 48 h post-MI. Patient characteristics are in Online Table S3. Of interest, the post-MI patients showed reduced levels of C-1158/59 (Figure 6A) (p < 0.05 vs. healthy controls), consistent with high MMP-9 levels at this time. Plasma MMP-9 levels peak at day 1 in post-MI patients, which may increase collagen fragment degradation (41). This decrease may also reflect a difference between sampling from the tissue and plasma. In post-MI patients, the plasma levels of the C-1158/59 fragment negatively correlated with E/e′ ratios (r = 0.332; p = 0.025) (Figure 6B). An E/e′ ratio <8 is indicative of normal filling pressures, whereas an E/e′ >15 is indicative of markedly elevated filling pressures (42). Thus, our data suggest that lower levels of C-1158/59 collagen fragment correlated with reduced cardiac function.
One limitation was the time difference in sampling between the mouse and human MI studies. Another was that we collected peripheral venous plasma samples, but did not collect coronary sinus levels, which would have allowed us to evaluate the myocardium as a direct source of the peptide. Further studies will be necessary to investigate the natural occurrence of this collagen fragment at later time points and its effects on long-term cardiac remodeling.
We have identified a novel collagen Iα1 cleavage site and the generation of a matricryptin downstream of this site, p1158/59, stimulated in vitro and in vivo wound healing. Our data show that p1158/59 is a biologically active molecule with potent proangiogenic properties and has direct effects on the quality of scar tissue post-MI by promoting early ECM deposition. Together, these effects preserved LV structure and geometry, resulting in reduced LV dilation post-MI. These data strongly promote this matricryptin as a molecule of interest to treat adverse LV remodeling.
COMPETENCY IN MEDICAL KNOWLEDGE: Following MI, collagen I, a major component of the cardiac extracellular matrix, is proteolytically cleaved by matrix metalloproteinases into fragments known as matricryptins that mediate left ventricular remodeling in a mouse model by enhancing fibroblast activity, basement membrane and granulation tissue integrity and angiogenesis.
TRANSLATIONAL OUTLOOK: Clinical studies are needed to assess the safety and efficacy of matricryptins and other fragments that modulate cardiac remodeling at the molecular level in patients recovering from MI.
This work was supported by the National Institutes of Health, Heart, Lung, and Blood Institute, Bethesda, Maryland: HHSN268201000036C (N01-HV-00244), HL075360, HL051971, GM104357; Biomedical Laboratory Research and Development Service of the Veterans Affairs Office of Research and Development: 5I01BX000505 (to Dr. Lindsey); American Heart Association, Dallas, Texas: 13POST14350034 (to Dr. DeLeon-Pennell), 14POST18770012 (to Dr. Iyer), 14SDG18860050 (to Dr. de Castro Brás); and Bernard and Audre Rapoport Foundation, Waco, Texas (to Dr. Lange). The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- extracellular matrix
- left ventricle
- infarcted left ventricle
- myocardial infarction
- matrix metalloproteinase
- tissue inhibitor of metalloproteinase
- Received February 2, 2015.
- Revision received June 24, 2015.
- Accepted July 14, 2015.
- American College of Cardiology Foundation
- Go A.S.,
- Mozaffarian D.,
- Roger V.L.,
- et al.
- Jarvelainen H.,
- Sainio A.,
- Koulu M.,
- Wight T.N.,
- Penttinen R.
- Daley W.P.,
- Peters S.B.,
- Larsen M.
- Mann D.L.,
- Spinale F.G.
- Brassart B.,
- Fuchs P.,
- Huet E.,
- et al.
- Garnotel R.,
- Monboisse J.C.,
- Randoux A.,
- Haye B.,
- Borel J.P.
- Sage E.H.,
- Reed M.,
- Funk S.E.,
- et al.
- Clark R.A.,
- Wikner N.E.,
- Doherty D.E.,
- Norris D.A.
- Webb C.S.,
- Bonnema D.D.,
- Ahmed S.H.,
- et al.
- Faye C.,
- Moreau C.,
- Chautard E.,
- et al.
- Cao D.J.,
- Jiang N.,
- Blagg A.,
- et al.
- Schips T.G.,
- Wietelmann A.,
- Hohn K.,
- et al.
- Geske J.B.,
- Sorajja P.,
- Nishimura R.A.,
- Ommen S.R.