Author + information
- Received December 6, 2016
- Revision received May 11, 2017
- Accepted May 16, 2017
- Published online July 10, 2017.
- Stephen D. Farris, MDa,
- Creighton Don, MD, PhDa,
- Deri Helterline, MSa,
- Christopher Costa, BSa,
- Tabitha Plummer, BSa,
- Susanne Steffes, DNP, RNc,
- Claudius Mahr, DOa,
- Nahush A. Mokadam, MDb and
- April Stempien-Otero, MDa,∗ ()
- aUniversity of Washington, Department of Medicine, Division of Cardiology, Seattle, Washington
- bUniversity of Washington, Department of Cardiothoracic Surgery, Seattle, Washington
- cUniversity of Washington, School of Nursing, Seattle, Washington
- ↵∗Address for correspondence:
Dr. April Stempien-Otero, Division of Cardiology, University of Washington, South Lake Union, 850 Republican Street, Box 358050, Seattle, Washington 98109.
Background Only limited data exist describing the histologic and noncardiomyocyte function of human myocardium in end-stage heart failure (HF).
Objectives The authors sought to determine changes in noncardiomyocyte cellular activity in patients with end-stage HF after left ventricular assist device (LVAD)-induced remodeling to identify mechanisms impeding recovery.
Methods Myocardium was obtained from subjects undergoing LVAD placement and/or heart transplantation. Detailed histological analyses were performed, and, when feasible, mononuclear cells were isolated from fresh, dissociated myocardium for quantitative reverse transcription polymerase chain reaction studies. Echocardiographic and catheterization data were obtained during routine care.
Results Sixty-six subjects were enrolled; 54 underwent 8.0 ± 1.2 months of LVAD unloading. Despite effective hemodynamic unloading and remodeling, there were no differences after LVAD use in capillary density (0.78 ± 0.1% vs. 0.9 ± 0.1% capillary area; n = 42 and 28, respectively; p = 0.40), cardiac fibrosis (25.7 ± 2.4% vs. 27.9 ± 2.4% fibrosis area; n = 44 and 31, respectively; p = 0.50), or macrophage density (80.7 ± 10.4 macrophages/mm2 vs. 108.6 ± 15 macrophages/mm2; n = 33 and 28, respectively; p = 0.1). Despite no change in fibrosis or myofibroblast density (p = 0.40), there was a 16.7-fold decrease (p < 0.01) in fibroblast-specific collagen expression. Furthermore, there was a shift away from pro-fibrotic/alternative pro-fibrotic macrophage signaling after LVAD use.
Conclusions Despite robust cardiac unloading, capillary density and fibrosis are unchanged compared with loaded hearts. Fibroblast-specific collagen expression was decreased and might be due to decreased stretch and/or altered macrophage polarization. Dysfunctional myocardium may persist, in part, from ongoing inflammation and poor extracellular matrix remodeling. Understanding these changes could lead to improved therapies for HF.
Heart failure (HF) affects almost 8 million Americans and carries >50% 5-year mortality from time of diagnosis (1). Although HF etiology and clinical trajectories are variable, myocyte–myofibril dysfunction, excess fibrosis, arrhythmias, and chronic volume overload uniformly contribute to disease progression. Detailed analysis of myocardium and in vivo biological mechanisms of dysfunction and remodeling were limited in human subjects before the advent of the left ventricular assist device (LVAD) as a bridge to transplant (BTT). Initial studies sought to evaluate histological changes after mechanical unloading to understand potential mechanisms of disease and recovery; however, subjects in these early studies primarily had now obsolete pulsatile LVADs, limited histological parameters were evaluated, and results were inconsistent (2–7). Newer-generation, continuous-flow LVADs have less pulsatility, greatly improved mechanical durability, and are an essential clinical tool in contemporary management of medical therapy refractory advanced systolic HF (8). Nevertheless, recovery after device implantation remains uncommon and histological studies continue to provide unclear answers (9).
Coincident with advances in LVAD technology, numerous human cell therapy trials have shown nonsignificant increases in systolic function with cell therapy; however, cellular and tissue characterization and potential mechanisms of action are absent (10–13). We sought to determine if recovery could be augmented by combining cell therapy and LVAD unloading. In a phase 1 study, with the purpose of defining safety and directly analyzing the effect of cell therapy on myocardial tissue in patients with end-stage HF, we injected purified, patient-derived bone marrow cell fractions into ischemic, viable myocardium (defined by single-photon emission computed tomography imaging) during LVAD placement (14). Upon cardiac explantation, we found that epicardial injection of either CD34+ or CD34-depleted cell fractions reduced activated fibroblast density compared with injected saline control without changes in fibrosis or microvessel density (14). These data provided evidence for a unique paracrine effect of cell injection independent of cell type in unloaded myocardium. Before initiating a larger clinical study to determine if larger scale inhibition of fibroblast activation would translate into significant reductions in fibrosis and augment recovery, we sought to better understand the effects of LVAD unloading on components of cardiac fibrosis.
Cardiac fibrosis is affected by diverse cell-specific pathways. Fibroblasts, the major collagen-expressing cell, can become activated by mechanical stretch to migratory, proliferative, and secretory phenotypes. Macrophages can adopt a pro-fibrotic (M2) phenotype that stimulates fibroblasts to express matrix or a pro-inflammatory (M1) phenotype to proteinases that degrade matrix. Hypoxemia from capillary rarefaction in advanced heart disease can also stimulate fibrosis. Because complete recovery on LVAD support remains a rare phenomenon, identifying histologic changes and which pathways are most active in response to unloading is critical to developing cell-based therapies to modulate fibrosis and potentiate reverse remodeling (15–20). Ultimately, such therapies may lead to improved myocardial recovery and survival.
To this end, we performed a prospective, observational study collecting clinical data and tissue from human subjects with end-stage HF undergoing LVAD implantation and/or cardiac transplantation. This study included histological analysis and gene expression studies of isolated cardiac fibroblasts and macrophages to test the hypothesis that unloading generates improvements in interstitial remodeling as manifest by: 1) increased capillary density; 2) decreased fibrosis; and 3) decreased expression of pro-fibrotic genes in isolated cardiac fibroblasts and macrophages.
Subjects gave informed consent and were prospectively enrolled at the University of Washington Medical Center before LVAD implantation or transplantation. Study groups include subjects with LVAD use (as either BTT or destination therapy [DT]), primary transplant, or total artificial heart implantation. Given the inherent, clinically driven crossover of DT and BTT indications, not all subjects enrolled underwent cardiac transplantation. All study protocols were approved by the University of Washington internal review board and adhered to the Helsinki Principles for Human Subjects Research.
Studies and analysis
Subjects underwent routine clinical care and established protocols at the University of Washington. Before LVAD implantation and/or heart transplantation, information was obtained via neurohormonal medication use, cardiac catheterization, echocardiography, and measurements of B-type natriuretic peptide (BNP). Left ventricular chamber measurements were made using 2-dimensional ultrasound techniques in the parasternal long axis.
Cell isolation was performed as previously described (21,22). In brief, ventricular myocardium was placed into cold phosphate-buffered solution without calcium or magnesium immediately in the operating room. Left ventricular myocardial tissue underwent immediate dissociation and cell isolation for RNA or formalin fixation and paraffin embedment for histology.
Fresh tissue was minced into 1-mm pieces and digested with thermolysin and deoxyribonuclease I. Sequentially, 100- and 70-μM filters were used. Ammonium-chloride-potassium lysing buffer was used to lyse red blood cells. Mononuclear cells were then incubated with anti-human CD14 antibody conjugated to magnetic beads and passed over a magnetic cell separation column to separate CD14+ macrophages and negative fibroblast fractions per protocol. RNA from eluted cells was isolated and quantified.
Tissue characteristics were analyzed for fibrillar collagen with a 50/50 mix of red and green dye and using immunohistochemistry for macrophages (mouse anti-human CD68), capillaries (mouse anti-human platelet endothelial cell adhesion molecule [PECAM-1] clone), and myofibroblasts (mouse anti-human anti-alpha smooth muscle actin [αSMA]). Density of myofibroblasts excluded αSMA+ blood vessels from analysis. Quantification was performed with automated image analysis by the University of Washington histology core or manual analysis in our lab.
Polymerase chain reaction primers were obtained (tested for linearity, Online Table 1). Gene expression analysis was performed using quantitative reverse transcription polymerase chain reaction and internally controlled with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) using the delta cycle threshold (ΔCT) method. Group mean expression levels in post-LVAD samples are compared with average pre-LVAD levels or as the negative (−ΔCT) per sample such that increased collagen 1α1 messenger RNA expression has a larger −ΔCT.
Demographic and clinical data are presented as median or mean, as indicated, ± SEM. Student t, Mann-Whitney U, Fisher exact, or paired Student t tests were used when applicable. Analysis of variance (ANOVA) tests were used for histological comparisons before and after LVAD support in ischemic and nonischemic subjects to identify hidden interactions. Unless stated otherwise, data are normally distributed. Analyses were done in a blinded fashion with respect to subject identity, HF etiology, and presence or absence of ventricular unloading.
Seventy-five tissue samples were collected from 64 subjects with end-stage HF resulting from an equal mixture of ischemic and nonischemic cardiomyopathies. Most subjects were male (80%) with a median age of 60 years (Table 1). Samples of failing cardiac tissue (loaded) were collected from 34 subjects at the time of LVAD implantation and 10 subjects who underwent primary transplantation or implantation of a total artificial heart. Tissue from LVAD cores included subjects undergoing implantation for both either DT or BTT. The DT group was older (median: 67 and 59 years, respectively; p = 0.01) but both groups were similar with respect to HF etiology, sex, LVAD type, preoperative medical therapy, and ejection fraction (Online Table 2).
Fifty-four subjects underwent LVAD support; of those, cardiac tissue (unloaded) was collected from 31 subjects who underwent cardiac transplantation after ventricular unloading for a median of 8.0 ± 1.2 months. Matched tissue samples from LVAD implantation (pre-LVAD) and cardiac transplantation (post-LVAD) were available from 17 subjects. Serial right heart catheterizations and BNP measurements confirmed significant ventricular unloading after LVAD use (Table 2). There was a significant decrease in mean pulmonary capillary wedge pressure (20.1 ± 1.3 mm Hg vs. 12.3 ± 0.9 mm Hg; p < 0.001) and BNP (991 ± 140 pg/ml vs. 40 ± 65 pg/ml; p < 0.001). Echocardiograms demonstrated ventricular remodeling with a mean decrease in end-diastolic dimension of 1.4 cm (p < 0.001) (Table 2). There was a significant increase in subjects with BTT tolerating beta blockade after LVAD use (97% post-LVAD from 55% pre-LVAD; p < 0.001), but there were no differences in angiotensin-aldosterone axis inhibitor use (Table 2).
Effects of LVAD unloading
Comparing all subjects, we found no differences in capillary density in pre-LVAD versus post-LVAD tissue (0.78 ± 0.12% vs. 0.88 ± 0.10% PECAM-1+ area; n = 42 and 28, respectively; p = 0.40) (Figure 1A). A direct comparison of patients (of mixed cardiomyopathy etiologies) who were assessed before and after unloading supported these data (n = 17; p = 0.10 by paired Student t test) (Figure 1B). By ANOVA, there were no hidden interactions (p > 0.1) for main effects of HF etiology and LVAD unloading (Figure 1C). In addition, there were also no differences in capillary density between all subjects with ischemic versus nonischemic HF etiology (0.74 ± 0.08% vs. 0.93% ± 0.11% PECAM-1 positive area; n = 43 and 27, respectively; p = 0.10).
Cardiac fibrillar collagen content was compared between subjects undergoing primary transplant or LVAD implantation and subjects undergoing transplant after LVAD unloading (25.7 ± 2.4% vs. 27.9 ± 2.4% picrosirius red area; n = 44 and 31, respectively; p = 0.50) (Figure 2A). To ensure that this finding was not purely from regional differences in fibrosis, we compared samples from apical cores from LVAD implantations to samples from the anterior wall from primary transplant and total artificial heart implantation. We found no differences in cardiac fibrosis (22.4% vs. 21.7% picrosirius red area, n = 39 and n = 7, respectively; p = 0.60 by Mann-Whitney U test) (Online Figure 1A). Furthermore, in a group of subjects with matched pre- and post-LVAD samples, we found no differences in cardiac fibrosis (n = 17; p = 0.50) (Figure 2B).
To exclude any hidden interaction with type of cardiomyopathy and unloading, we performed a 2-way ANOVA on ischemic and nonischemic tissue with and without LVAD unloading. There were no differences in the 4 groups, although there was a trend toward an interaction resulting from ischemic subtype (p = 0.1) (Figure 2C). This interaction was confirmed by comparing all myocardium (loaded, unloaded) from subjects with ischemic heart disease with those with nonischemic heart disease. Here we found a trend toward increased fibrosis (28.8 ± 2.4% vs. 23.1 ± 2.2% positive picrosirius red area; n = 43 and n = 34, respectively; p = 0.06) (Online Figure 1B). To determine if angiotensin blockade modified the fibrotic response, we compared changes in fibrosis in matched subjects on or off angiotensin-converting enzyme (ACE) inhibitor or angiotensin receptor blocker therapy after LVAD implantation. Neither group had a significant decrease in histological collagen by paired Student t test (Online Figures 2A and 2B). Comparison of all unloaded subjects by angiotensin blockade status confirmed this finding (Online Figure 2C).
Effect of LVAD unloading on fibrosis and inflammation
To determine the mechanisms behind persistent fibrosis after LVAD unloading, we examined gene expression in isolated cardiac fibroblasts. Although histological collagen content was unchanged, we found a 4-fold decrease in fibroblast-specific collagen expression after LVAD unloading when including all samples (p = 0.02) (Figure 3A) and 16.7-fold decrease when including paired samples only (p < 0.01) (Figure 3B). As expected, there was good correlation in pre-LVAD samples between Col1α1 expression in cardiac fibroblasts and histologic collagen accumulation (r = 0.70; n = 24; p < 0.002) (Figure 3C); however, this correlation was completely lost after unloading (r = 0.01; n = 12; p = 1.00) (Figure 3D).
To determine if the discrepancies between myocardial fibrosis and fibroblast collagen expression were due to differences in the number of collagen-expressing cells, we quantified myofibroblast (αSMA + interstitial cell) density because this subpopulation of fibroblasts may be most important in human cardiac fibrosis. Despite a marked decrease in collagen expression–cell after LVAD unloading, there were no differences in myofibroblast density after LVAD unloading (105.4 ± 36.5 αSMA+ cells/mm2 vs. 106.7 ± 17.3 αSMA+ cells/mm2; n = 8; p = 1.00) (Figure 3E). Additionally, there were no differences in αSMA+ area comparing all pre- and post-LVAD samples (1.72 ± 0.69 αSMA+ area vs. 1.4 ± 0.4 αSMA+ area; p = 0.70).
To test the hypothesis that LVAD unloading decreases inflammation, we quantified macrophage (CD68+ cells) density in our cohort. We found a trend toward increased macrophage density after LVAD unloading (80.7 ± 10.4 CD68+ cells/mm2 vs. 108.6 ± 15.2 CD68+ cells/mm2; n = 33 and 28, respectively; p = 0.10) (Figure 4A). Comparing matched samples (of mixed etiology), there were no differences in macrophage density (n = 16; p = 0.80) (Figure 4B). Differences were present in ischemic subjects with a significant interaction of ischemic tissue and an unloaded state (by 2-way ANOVA; p = 0.02) (Figure 4C).
To determine if unloading altered macrophage phenotype as a potential cause for decreased fibroblast collagen expression, we performed gene expression analysis of pro-inflammatory, pro-fibrotic, and proteinases from isolated cardiac macrophages. Pro-fibrotic M2 macrophage gene expression decreased after LVAD (Figure 5A). Expression of Kruppel-like factor-4, transglutaminase-2, mannose receptor-1, and interleukin-10 all decreased by at least 50% (p < 0.05; interleukin-10, p = 0.07). There was a reciprocal trend toward increased pro-inflammatory M1-polarization (2.5-fold and 5.8-fold increased expression for IL-1β and tumor necrosis factor-α; p < 0.1 for both comparisons) (Figure 5B). Furthermore, macrophage-derived matrix metalloproteinase (MMP)-2 expression is decreased 4-fold (p < 0.01) (Figure 5C).
Systolic HF remains a progressive disease, and noncardiomyocyte mechanisms of persistent myocardial dysfunction remain elusive. Unloading during LVAD support has the potential to exact multiple beneficial changes in the failing heart. Important physiological changes include improved cardiac output, reduced filling pressures, and “reverse remodeling” (23). These changes could have significant effects on the regulation of extracellular matrix and cellular composition. We measured histological parameters and cell-specific expression and sought to identify mechanisms potentially promoting or impeding recovery. We found no significant effect of LVAD unloading on microvascular density or cardiac fibrosis. We also found no significant changes in macrophage density; however, there were consistent, provocative alterations in pro-fibrotic gene expression in both cardiac fibroblasts and macrophages (Central Illustration).
Detailed analyses of capillary density in human disease are uncommon. In comparison to Drakos et al., who found a 35% increase in capillary density with unloading, we found no differences in capillary density between the apical core and anterior wall after ventricular unloading with pulsatile LVAD (7). Discrepancies in study findings may be related to device or methodological differences; in particular, the location of the tissue samples acquired. The post-LVAD specimens that we collected were much further from the apical core than those collected by Drakos et al. Proximity to the core might have resulted in post-surgical inflammatory changes in this area independent of the effects of unloading. Gupta et al. demonstrated there were no changes in serial radiotracer uptake during LVAD unloading, supporting that there are few, if any, notable changes in microvascular perfusion with LVAD unloading (24).
Prior studies on extracellular matrix changes were conflicting, limiting the understanding of myocardial remodeling in heart failure. Although Li et al. also found no changes in fibrosis after LVAD support (4), Lok et al. found an almost doubling (18% to 32%) of fibrosis, as assessed by Masson’s trichrome staining after LVAD unloading; their study was limited to nonischemic cardiomyopathy subjects (25). In other studies of pulsatile ventricular assist devices, the data were equally mixed (2,7). We found no differences in histological collagen after continuous-flow LVAD unloading. Because comparisons between pre- and post-LVAD samples were from different areas of the heart, we first determined that there were no baseline differences in the areas sampled (apex and anterior wall). We used picrosirius red positive area as a measure of fibrillar collagen because this stain provides more specificity, reproducibility, and detectability than Masson’s trichrome, which was used in earlier studies (26,27).
Extracellular matrix deposition by fibroblasts is regulated by mechanosignaling, neurohormones, and macrophage signals among other factors. We isolated fibroblasts and found markedly decreased collagen expression in response to unloading that did not translate to decreased histological fibrosis. Decreased collagen expression was likely due directly to the decreased wall stress with unloading as opposed to changes in angiotensin or aldosterone activity because there were no differences in ACE inhibitor/angiotensin receptor blocker or aldosterone inhibitor use after LVAD. Confirming that our fibroblast collagen expression was reasonable and consistent, there was a highly significant and strong correlation of fibrosis severity to collagen expression in fibrotic, failing hearts that was completely abolished after unloading. Milting et al. (28) and Felkin et al. (29) found small, nonsignificant decreases in collagen expression before transplantation; however, these and other studies measured extracellular matrix gene expression using whole tissue and/or were cardiomyocyte focused (25,28,30–32). Here we demonstrated changes in fibroblast-specific gene expression in different loading states ex vivo, introducing the possibility that modification of these factors in vivo could be harnessed to improve ventricular remodeling.
Macrophages have diverse contributions to fibrosis including the regulation of fibroblast number and function and degradation of matrix; thus, macrophages are likely important regulators of adverse remodeling in the face of chronic injury such as HF. Although we found no significant decrease in overall CD68+ macrophage density, we found decreased pro-fibrotic M2 macrophage gene expression with a trend toward increased M1-associated gene expression. This was consistent with an overall decrease in pro-fibrotic macrophage phenotype (Online Figure 3). Transglutaminase-2 and mannose receptor-1 are associated with alternatively activated macrophages and pathological fibrotic states. Kruppel-like factor-4, a transcription factor, has been implicated in M2 polarization and might be important in pathologic cardiac remodeling (33–35). Our understanding of Kruppel-like factor-4 in human disease is clearly in its infancy but represents a potentially important regulator in cardiac macrophage signaling.
Because immune cells are a major source of proteases, including MMPs, we evaluated macrophage-specific protease expression and found decreased macrophage MMP2 expression. The decrease in both macrophage-derived MMP2 and fibroblast-derived collagen expression could explain previously observed decreases in soluble to insoluble collagen after LVAD unloading (2,4). Further studies using a broader based screen to interrogate changes in macrophage function and proteinase expression (which are outside the scope of this study) would be useful to determine if altering cell phenotype or specific proteinase activity could modulate fibrosis in end-stage heart disease.
Our study was unique because we correlated clinical, hemodynamic, imaging, and laboratory findings with focused noncardiomyocyte histological changes and gene expression from isolated cells to gain greater insights into matrix remodeling and recovery mechanisms. It is, to our knowledge, the largest and most comprehensive analysis of these tissues in subjects with continuous-flow LVADs and the only one to examine cell-specific changes.
There are limitations to this study, however, including several uncontrollable variables, such as sex of the subjects, type of continuous-flow device, and duration of unloading. Tissue was not acquired from all subjects enrolled and, in some instance, was of insufficient quality or quantity for cell isolation. Nevertheless, our findings were robust because they were consistent across HF etiologies and duration of LVAD unloading. The genes tested were targeted by prespecified hypotheses; thus, they were not comprehensive. There were likely other important signaling pathways that were not tested. Furthermore, there were likely subpopulations of fibroblasts and macrophages not identified that are beyond the scope of this study.
Despite robust cardiac unloading with LVAD use, histological myocardial changes are minimal in patients with end-stage HF despite important cell-specific biologic changes, including changes in macrophage function and decreased collagen expression. Conventional therapies such as ACE inhibitors and beta-blockers appear to have no effects on this process. Dysfunctional myocardium might persist and impede recovery after LVAD implantation in part from persistence of inflammation, activated fibroblasts, and poor extracellular matrix turnover (Central Illustration). Understanding these mechanisms will be critical for developing effective pharmacological and cell-based therapies to support recovery.
COMPETENCY IN MEDICAL KNOWLEDGE: In patients with end-stage heart failure, both fibroblasts and macrophages contribute to fibrosis by different mechanisms that depend on the stage of disease and degree of ventricular unloading.
TRANSLATIONAL OUTLOOK: Future research should investigate the therapeutic potential of cell-specific modulators of fibrosis and inflammation to facilitate reversal of cardiac remodeling.
For supplemental figures and tables, please see the online version of this article.
This work was supported by grants R01 HL094384 and Tall Family Foundation (Dr. Stempien-Otero), UL1TR000423 ITHS (Dr. Don), Locke Family Foundation (Drs. Stempien-Otero and Farris), and an American College of Cardiology Foundation Fellowship (Dr. Farris). Dr. Modadam has consulted for Heartware and St. Jude. Dr. Mahr has consulted for HeartWare, St. Jude (Thoratec), and Abiomed. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- analysis of variance
- alpha smooth muscle actin
- B-type natriuretic peptide
- bridge to transplant
- destination therapy
- glyceraldehyde 3-phosphate dehydrogenase
- heart failure
- left ventricular assist device
- matrix metalloproteinase
- mouse anti-human platelet endothelial cell adhesion molecule
- Received December 6, 2016.
- Revision received May 11, 2017.
- Accepted May 16, 2017.
- 2017 American College of Cardiology Foundation
- Writing Group M.,
- Mozaffarian D.,
- Benjamin E.J.,
- et al.
- Klotz S.,
- Foronjy R.F.,
- Dickstein M.L.,
- et al.
- Thohan V.,
- Stetson S.J.,
- Nagueh S.F.,
- et al.
- Li Y.Y.,
- Feng Y.,
- McTiernan C.F.,
- et al.
- Drakos S.G.,
- Kfoury A.G.,
- Hammond E.H.,
- et al.
- Wever-Pinzon O.,
- Drakos S.G.,
- McKellar S.H.,
- et al.
- Chugh A.R.,
- Beache G.M.,
- Loughran J.H.,
- et al.
- Sanina C.,
- Hare J.M.
- Karantalis V.,
- Hare J.M.
- Stempien-Otero A.,
- Helterline D.,
- Plummer T.,
- et al.
- Drakos S.G.,
- Wever-Pinzon O.,
- Selzman C.H.,
- et al.
- Dandel M.,
- Weng Y.,
- Siniawski H.,
- et al.
- Mancini D.M.,
- Beniaminovitz A.,
- Levin H.,
- et al.
- Maybaum S.,
- Mancini D.,
- Xydas S.,
- et al.
- Topkara V.K.,
- Garan A.R.,
- Fine B.,
- et al.
- Carlson S.,
- Helterline D.,
- Asbe L.,
- et al.
- Birks E.J.
- Lok S.I.,
- Nous F.M.,
- van Kuik J.,
- et al.
- Felkin L.E.,
- Birks E.J.,
- George R.,
- et al.
- Saha B.,
- Bala S.,
- Hosseini N.,
- Kodys K.,
- Szabo G.
- Ke B.,
- Zhang A.,
- Wu X.,
- Fang X.