Author + information
- Received November 14, 2005
- Revision received April 27, 2006
- Accepted May 30, 2006
- Published online October 3, 2006.
- Xifu Liu, PhD⁎,
- Xinhua Gu, MD⁎,
- Zhaoming Li, PhD⁎,
- Xinyan Li, MD, PhD⁎,
- Hui Li, MD⁎,
- Jianjie Chang, MD⁎,
- Ping Chen, MD⁎,
- Jing Jin, BSc⁎,
- Bing Xi, MSc⁎,
- Denghong Chen, BSc⁎,
- Donna Lai, PhD†,
- Robert M. Graham, FAA, MD† and
- Mingdong Zhou, PhD⁎,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Mingdong Zhou, Zensun Science and Technology Ltd., 328 Bibo Road, Zhangjiang Science Park, Pudong, Shanghai 201203, P.R. China.
Objectives We evaluated the therapeutic potential of a recombinant 61-residue neuregulin-1 (beta2aisoform) receptor-active peptide (rhNRG-1) in multiple animal models of heart disease.
Background Activation of the erbB family of receptor tyrosine kinases by rhNRG-1 could provide a treatment option for heart failure, because neuregulin-stimulated erbB2/erbB4 heterodimerization is not only critical for myocardium formation in early heart development but prevents severe dysfunction of the adult heart and premature death. Disabled erbB-signaling is also implicated in the transition from compensatory hypertrophy to failure, whereas erbB receptor-activation promotes myocardial cell growth and survival and protects against anthracycline-induced cardiomyopathy.
Methods rhNRG-1 was administered IV to animal models of ischemic, dilated, and viral cardiomyopathy, and cardiac function and survival were evaluated.
Results Short-term intravenous administration of rhNRG-1 to normal dogs and rats did not alter hemodynamics or cardiac contractility. In contrast, rhNRG-1 improved cardiac performance, attenuated pathological changes, and prolonged survival in rodent models of ischemic, dilated, and viral cardiomyopathy, with the survival benefits in the ischemic model being additive to those of angiotensin-converting enzyme inhibitor therapy. In addition, despite continued pacing, rhNRG-1 produced global improvements in cardiac function in a canine model of pacing-induced heart failure.
Conclusions These beneficial effects make rhNRG-1 promising as a broad-spectrum therapeutic for the treatment of heart failure due to a variety of common cardiac diseases.
The erbB family of receptor tyrosine kinases and their ligands, neuregulins (NRGs), are critical involved not only in cardiac development but also in the maintenance of structural and functional integrity of the adult heart (1–3). ErbB2 (Her2/neu) functions as a non–ligand-binding, pre-activated co-receptor (4) and in the myocardium heterodimerizes with ErbB4 (Her4), upon activation by NRG-1; the latter is expressed by the endocardium and endothelium of the cardiac microvasculature (5). Initial evidence for a functional role of the NRG/erbB signaling system in adult heart came from trials of the monoclonal antibody, trastuzumab (Herceptin), in patients with erbB2-overexpressing breast cancers (6). Downregulation of erbB2 by Herceptin resulted in a low incidence of cardiac dysfunction when used as monotherapy but, importantly, the development of cardiomyopathy in almost 30% of patients when combined with an anthracycline. In addition, whereas ErbB2-deficient conditional mutant mice are viable at birth, adult animals develop severe dilated cardiomyopathy with wall thinning and decreased contractility, leading to premature death from heart failure (7).
ErbB2/4 receptors downregulate with the onset of heart failure in animals with pressure-overload hypertrophy, indicating involvement of disabled erbB signaling in the transition from compensatory hypertrophy to failure (8). In contrast and consistent with the latter findings in animals, expression of erbB receptors increases in heart failure patients subjected to mechanical unloading, particularly those with ischemic cardiomyopathy (9). In adult cardiomyocytes, NRG-1 enhances sarcomere organization, protects against myofibrillar disarray (10), and unlike other hypertrophic stimuli such as angiotensin II or interleukin-1beta/IFNgamma, promotes cell-survival rather than death (1,5,11).
Despite these pleiotropic “cardiotonic” effects of NRG-1/erbB-activation, its therapeutic potential has not been evaluated. We show here that, in each of 4 distinct models of severe left ventricular (LV) failure, short-term administration of a recombinant receptor-active NRG-1 peptide-fragment significantly improves or protects against deterioration in myocardial performance and, importantly, in a rodent model of ischemic cardiomyopathy, prolongation of survival is additive to that observed with converting enzyme inhibitor-treatment.
Production and activity of rhNRG-1
Deoxyribonucleic acid encoding the epidermal growth factor-like domain of human NRG-1 (beta2aisoform; residues Ser177-Glu237) was polymerase chain reaction (PCR)-amplified (Fig. 1A),and the resulting E. coli-expressed 61-residue peptide (rhNRG-1) was purified by sequential ion-exchange and hydrophobic chromatography (final purity >95%, as determined by sodium dodecyl sulfate polyacrylamide gel electrophoresis [SDS-PAGE]) (Fig. 1B). Primary cultures of rat neonatal cardiomyocytes were used to evaluate the activity (erbB4 phosphorylation) of the purified protein (Fig. 1C), as previously described (12). Activity of rhNRG-1 was also evaluated in vivo by IV bolus administration to mice, rats, and dogs, as detailed in Figure 1, and this response was used in dose-finding studies to determine the doses of rhNRG-1 required in vivo for the subsequent hemodynamic studies.
Sprague Dawley rats (200 to 250 g, Animal Center of Fudan University, Shanghai, China) were anesthetized (ketamine hydrochloric acid [HCl], 100 mg/kg, IP), and the LV was infarcted by ligation of the left anterior descending coronary artery (LAD), as described (13). Sham-operated animals were similarly treated except that the suture around the LAD was left untied; rhNRG-1 (10 μg/kg/day, IV, for 5 or 10 days) or vehicle (0.2% albumin/phosphate buffered saline, pH 6.0) was administered as indicated in Figure 2.Sham-operated animals received no treatment. Angiotensin I and II, aldosterone levels, and plasma renin activity were determined by radioimmunoassay (Institute of Beijing North Biotechnology), in plasma isolated from blood samples obtained at the time of the final hemodynamic study in sham-operated and LAD-ligated animals treated with rhNRG-1 or vehicle for 5 days. All animal studies were approved by the Animal Care and Use Committee, Shanghai City, China.
Drug-induced cardiomyopathy model
Doxorubicin was administered to 2 groups (n = 20/group) of Sprague Dawley rats (300 to 350 g), as detailed in the text. Simultaneously, 1 group received rhNRG-1 (20 μg/kg/day, IV, for 5 days), and another, vehicle alone. A 3rd group (n = 20) served as untreated control subjects. Hemodynamics and cardiac pathology were evaluated after 4 weeks. Myocardial cell damage was assessed from serum cardiac troponin I (cTnI) levels (Optus Plus Immunoassay System, Boehringer Diagnostics, Germany).
In vitro, trypsin-digested Sprague Dawley neonatal rat cardiomyocytes were cultured in 24-well plates at a density of 5 × 105cells/well with 20% fetal bovine serum in Eagle’s Minimum Essential Medium at 37°C, 5% carbon dioxide. After 18 h, cells were infected with Coxsackie virus B3 (CV-B3; Nancy strain) (Institute of Cardiovascular Disease, Zhongshan Hospital, Fudan University) for 1 h. Viral titers were expressed as the 50% tissue culture infectious dose (TCID50) (14). Cells incubated with medium alone were used as control subjects. The CV-B3–infected cells were then washed with virus-free medium and divided into 2 groups: 1 treated with rhNRG-1 (100 ng/ml), and 1 with vehicle. Cells were then cultured for 3 to 5 days and evaluated for cTnI release into the culture medium. Each group consisted of 5 replicate wells.
In vivo, 4-week-old male Balb/c mice (Animal Center, Fudan University) were inoculated with CV-B3 (1 × 104TCID50in 0.2 ml virus solution, IP) or were untreated (timed control subjects). The CV-B3–infected mice were treated with rhNRG-1 (30 μg/kg/day, IV, for 5 days) or with vehicle. Seven days later, echocardiography was performed, and blood was sampled for cTnI levels. The hearts were then harvested for histological evaluation and determination of viral titers as described (14), except using Hep2 cells.
Chronic rapid pacing model
Beagle dogs (8 to 10 kg) were anesthetized with sodium pentobarbital (25 mg/kg, IV). A unipolar pacing lead (Model 4023-58 cm, Medtronic, Minneapolis, Minnesota) was inserted into the right ventricular apex through the jugular vein under X-ray guidance and connected to a programmable pacemaker (Prevail 8086, Medtronic) that was implanted into a subcutaneous pocket. The dogs were allowed to recover for 7 to 10 days before experimentation. A control, sham group had the pacing lead implanted, but the pacemaker was not activated. Two experimental groups were paced at 230 beats/min for 3 weeks, at which time 1 received rhNRG-1(3 μg/kg/day, IV, for 5 days) and the other received only vehicle. Pacing was continued during these 5 days of treatment. Six animals were studied per group. Hemodynamics and cardiac contractility were evaluated as detailed in the following section.
Hemodynamics, echocardiography, and contractility studies
Arterial pressure and heart rate were determined by micromanometry (15) in rats and dogs anesthetized with sodium pentobarbital (30 mg/kg, IP, or 25 mg/kg, IV, respectively). Echocardiography was performed on rats and dogs anesthetized with ketamine HCl (100 mg/kg, IP) or sodium pentobarbital (25 mg/kg), respectively, and on immobilized conscious mice, as described previously (15). The LV was imaged in the short-axis view at the mid-papillary muscle level, and M-mode measurements of left ventricular end-diastolic diameter (LVEDD) and left ventricular end-systolic diameter (LVESD) were recorded and transferred on-line to a computer for analysis. Ejection fraction (EF) and fraction of shortening (FS) were measured off-line with the Teichholtz method (16). Cardiac output in dogs was obtained by Doppler echocardiography. This allowed cross-sectional area of the aortic root, LV ejection time, and mean aortic flow velocity to be determined, which were then used to calculate stroke volume and thus cardiac output, given that cardiac output = stroke volume × heart rate.
Ten serial transverse sections of formalin-fixed and paraffin-embedded rat heart were cut from each of 3 regions: apex, mid-ventricle, and just below the atrioventricular junction. Sections were hematoxylin and eosin (H & E)-stained and examined under light microscopy (× 10) by an experienced investigator blinded to the treatments, for myocardial lesions induced by doxorubicin (cytoplasmic vacuolization and/or myofibrillar loss) or viral infection (necrosis and inflammation). Findings were graded with a semiquantitative score from 0 to 4 (0 = no pathological changes; 1 = lesions involving <25% of the field; 2 = lesions involving 25% to 50%; 3 = lesions involving 50% to 75%; and 4 = lesions involving 75% to 100%). Representative views were photographed. Capillary density (vessels/mm2) in fibrotic areas was also determined by counting 3 slides and 3 views per slide.
Data are presented as mean ± 1 SEM. Statistical comparisons (p < 0.05 was considered significant) were by 1-way ANOVA followed by Tukey’s post hoc analysis. Cumulative survival data was evaluated by Kaplan-Meier nonparametric regression analysis and the Mantel-Cox log rank test. Statistical analyses were performed with the SPSS software (SPSS Inc., Chicago, Illinois). All data from hemodynamic and echocardiographic recordings and subsequent analyses were made without knowledge of treatments.
NRG-1 peptide and activity
We synthesized a 61-residue peptide (rhNRG-1) corresponding to the epidermal growth factor-like domain of human NRG-1 and containing 3 disulfide bonds and used it to activate erbB signaling (Fig. 1). This peptide retains the binding and erbB-activating properties of the 637 amino acid parent protein and can be synthesized recombinantly in high yield. Correct folding and disulfide bond-connectivity (17) were confirmed by peptide digestion and mass spectrometry (data not shown) and are evident from the ability of rhNRG-1 to stimulate erbB2/4 phosphorylation in vitro and in vivo (Fig. 1).
rhNRG-1 signaling and effect on sarcomere organization
We also demonstrated, consistent with the enhanced myofibrillogenesis observed by Saywer et al. (10) and Baliga et al. (11), with NRG-1 treatment of cultured cardiomyocytes, that sarcomere organization was enhanced with rhNRG-1 (data not shown). This cardiomyocyte response could be blocked by the mitogen-activated protein/extracellular signal-regulated kinase (ERK) (MEK)-1 inhibitor, PD98059, indicating involvement of the ERK1/2 signaling pathway but not by the PI-3 kinase inhibitor, Wortmannin (data not shown). Consistent with these in vitro findings, intravenous administration of rhNRG-1 in mice, rats, and dogs activated cardiac ERK1/2 (p42/44 mitogen-activated protein kinases) (Figs. 1E–1H).
Rat myocardial ischemia model
Myocardial infarction, induced here by LAD ligation, is widely used to produce a model of LV dysfunction. Impaired cardiac performance is due to loss of myocardium and progressive remodeling characterized by LV dilatation (increased LV dimensions), contractile dysfunction (reduced EF percent and percent FS), and increased filling pressure (LVEDP) (18), as evident here in sham-operated versus vehicle-treated animals subjected to LAD ligation (Figs. 2 and 3,protocol I). rhNRG-1, given 1 week after LAD ligation, significantly attenuated the dysfunction in all parameters evaluated (Figs. 2A and 2B, protocol I). Although this attenuation was less marked if rhNRG-1-treatment was delayed until 2 months after LAD ligation, improvements in cardiac performance (LVESD, EF, and FS) (Figs. 2E and 2F, protocol II) were still evident.
In additional groups of animals subjected to LAD ligation followed, 1 week later, by rhNRG-1 (10 μg/kg/day for 5 days) or vehicle, histological analysis revealed no difference in infarct size, which averaged 30% to 35%; but consistent with an angiogenic effect of erbB receptor-activation (19), capillaries were more abundant in the fibrotic peri-infarct regions in the rhNRG-1-treated group (vessels/mm2: 6.5 ± 2.2, vehicle-treated; 10.1 ± 3.0, rhNRG-1-treated, p < 0.01). Moreover, cardiac failure, evident by activation of the renin-angiotensin-aldosterone axis, was attenuated by rhNRG-1 treatment (Table 1).
To evaluate the effects of erbB receptor-activation on survival, we compared LAD-ligated animals given rhNRG-1 (10 μg/kg/day for 5 days) with those given either vehicle alone, the angiotensin-converting enzyme inhibitor, captopril (2.5 mg/kg twice daily for the duration of the study), or combined captopril plus rhNRG-1 treatment. We commenced treatment 1 week after LAD ligation. In vehicle-treated animals surviving 172 days from the commencement of treatment, cardiac dysfunction was evident from marked impairment of LV pressure development at both end-systole (LVESP, −27%) and end-diastole (LVEDP, −43%), reduced rates of contractility (+dP/dt) (−46%) and relaxation (−dP/dt) (−44%), and lower arterial pressure (−21%) (Fig. 3). In contrast, changes in all these parameters were significantly less marked in animals treated with rhNRG-1 or combined therapy but not in those receiving captopril alone (Fig. 3). In a related study in which the initiation of treatment was delayed until 2 months after LAD ligation and the animals evaluated 84 days later, we observed similar hemodynamic responses (Table 2).In addition to these favorable cardiovascular responses, survival was markedly improved with rhNRG-1 or captopril (Figs. 4Aand 4B), and these survival-effects were additive in animals receiving combined therapy (Fig. 4A).
Compared with these effects in rats with myocardial infarction, administration of rhNRG-1 to normal rats, at a dose of 10 μg/kg/day, IV, for 5 days, had no effect on echocardiographically determined LVESD, LVEDD, EF, or percent FS (Table 3).
Doxorubicin cardiomyopathy model
Anthracyclines, such as doxorubicin, are well-characterized cardiotoxic chemotherapeutic agents. Myocardial injury is associated with myofibrillar disarray and contractile abnormalities (10,11). Administration of doxorubicin (3.3 mg/kg/week, IV, for 4 weeks) to rats induced marked cardiac dysfunction (impaired dP/dt and LV pressure generation) (Table 4)as well as myocardial cell damage, evident both histologically (Fig. 5and Table 4) and biochemically (increased plasma cTnI levels) (Table 4). These changes were prevented (Fig. 5, Table 4) and survival markedly improved (Fig. 4) in animals treated concomitantly with rhNRG-1 (20 μg/kg/day, IV, for 7 days).
Infections with enteroviruses, such as CV-B3, are a common cause of acute myocarditis that can lead to heart failure and dilated cardiomyopathy (20). Both direct viral cytopathic effects and host cellular immune responses play a role in disease pathogenesis (21). We observed cell damage in neonatal rat cardiomyocytes infected with CV-B3. This was evident morphologically (data not shown) and from increased cTnI release (n = 5, 49 ± 4 ng/ml), as compared with vehicle-treated control-cultures (n = 5, 25 ± 4 ng/ml, p < 0.01). These effects were attenuated in cells treated with both virus and rhNRG-1 (100 ng/ml for 5 days) (n = 5, cTnI: 33 ± 3 ng/ml, p < 0.01 vs. virus-infected). Next we investigated whether this protective effect is also operative in vivo. Consistent with the induction of myocarditis, infection of mice with CV-B3 produced marked lymphocytic infiltration of the myocardium (Fig. 5) as well as histological (Table 4) and biochemical (increased plasma cTnI levels) evidence of myocardial damage, LV dilatation (increased LVESD and LVEDD), and impaired cardiac function (decreased EF and percent FS) (Table 4). rhNRG-1 (30 μg/kg/day for 5 days) not only protected against these histological (Fig. 5, Table 4), biochemical (decreased cTnI levels), and echocardiographically determined changes in EF percent, percent FS, LVEDD, and LVESD (Table 4) but significantly improved survival (Fig. 4). Moreover, the CV-B3 titer was reduced in the rhNRG-1 mice (Table 4), consistent with reduced inflammation with NRG treatment (22).
Chronic pacing model
Compared with unpaced control (sham) animals, those paced at 230 beats/min for 3 weeks developed heart failure, as evidenced by reductions in EF and cardiac output of approximately 50% and 30%, respectively (Fig. 6).In addition, in the animals treated only with vehicle, pacing resulted in significant increases in LVEDP (averaging >25 mm Hg), falls in mean arterial pressure of approximately 25 mm Hg, and reductions in both +dP/dt and −dP/dt (Fig. 6). Despite continued pacing, treatment with rhNRG-1 (3 μg/kg/day, IV, for 5 days) improved LVEDP by more than 25% (from 27.7 to 20.5 mm Hg) and LVESP by more than 35% (from 98.7 to 133.2 mm Hg) compared with the vehicle-treated group (Fig. 6). Cardiac contractility and relaxation were also significantly improved, by 47% (+dP/dt: from 1,740 to 2,558 mm Hg/s) and 55.7% (−dP/dt: from −1,107 to −1,723 mm Hg/s), and echocardiographically determined percent EF increased by 78% (from 24.6% in the vehicle-treated to 43.8% in the rhNRG-1-treated group) (Fig. 6). By contrast to these effects in dogs with pacing-induced heart failure, acute bolus administration of rhNRG (3 μg/kg/day, IV, for 5 days) in normal dogs had no effect on heart rate, maximum or minimum LV pressure (LVPmax, LVPmin), or rates of contraction or relaxation (+dP/dt or −dP/dt), recorded continuously over 60 min after administration (Table 5).
We and others have shown that in vitro erbB receptor-activation promotes myofibrillar organization in the face of serum-starvation and, as we show here, limits the cytopathic effects of Coxsackie infection. In addition, in each of 4 different models of LV failure, erbB receptor-activation improves both cardiac performance and survival. In the rat ischemia model, favorable cardiovascular responses do not occur acutely (within minutes), as would be expected with agents that activate protein kinase A or other inotropic pathways. Nevertheless, the early divergence of the cumulative survival curves indicates that improvements in cardiac function begin to occur within a few days of commencing treatment. Moreover, despite a short circulating half-life (approximately 30 min; unpublished data), rhNRG-1-induced improvements in cardiac function and survival were sustained for at least 60 to 80 days after treatment was discontinued. After this, decrements in survival began to parallel those seen with vehicle-treatment.
In contrast to captopril, arterial pressure was well maintained with rhNRG-1 (Fig. 3). This suggests that these 2 agents have distinct modes of action, and in support of this notion is the finding that combined treatment was superior to either agent alone, both in terms of improvements in cardiac performance and survival. Potential therapeutic effects of erbB receptor-activation to improve cardiac performance by promoting organization of sarcomere structure, cell integrity, and cell-cell adhesion (10,11) as well as suppress apoptotic pathways (5) and enhance angiogenesis (19) (evident here by the increase in infarct-region capillary density, which might contribute to enhanced cardiac function by improving LV compliance) are also likely distinct from the mechanisms underlying the favorable effects of other agents presently used for heart failure. Calcium channel antagonists, for example, predominantly act to reduce afterload, whereas beta-blockers reduce myocardial metabolic demand.
Although increased mortality was reported in mice given NRG-1 (23), the dose administered was almost 100 times higher than the rhNRG-1 used here, the isoform (beta1) differed, and NRG was given by continuous infusion over a prolonged period (14 days). More importantly, with highly purified cyclic guanosine monophosphate (current Good Manufacturing Practice)-prepared rhNRG-1, no deaths or other major side effects were observed in formal toxicology studies in mice, rats, and Rhesus monkeys or in a phase 1 clinical trial that we have completed involving 80 normal humans (Liu et al., data to be reported in a separate publication). Also, malignant transformation was not observed in any female rats given injections of rhNRG-1 (3 mg/kg/day) into breast tissue for 1 month (data not shown).
With the cardiomyopathy and myocarditis models, rhNRG-1 seems to limit myocardial-cell damage, as was evident by reduced troponin I release and histological changes, and in both models divergence in the survival curves from those of the vehicle-treated groups occurred early and was sustained. Protection against anthracycline cardiotoxicity involves NRG-1-mediated activation of the PI3-kinase pathway, which limits apoptosis and caspase-3-activation (11,24). Activation of the Stat transcription-pathway augments host-cell antiviral systems and protects against Coxsackie B3 myocarditis (19). This likely explains the cardioprotection observed here with rhNRG-1, because although NRG-1/erbB-mediated Stat-activation has not yet been demonstrated in myocardial cells, it has in other cells; a response involving recruitment of Src to NRG-1-promoted erbB/Stat complexes rather than erbB-mediated Jak-activation (25).
The finding that rhNRG-1 improved function and survival in the rat infarct model, even when the start of therapy was delayed for 2 months, has important implications for the clinical management of heart failure, where myocardial-cell damage might already be advanced. Moreover, given that rhNRG-1 largely prevented rather than merely attenuated the deleterious cardiac effects of doxorubicin and Coxsackie B3, it might also be more efficacious in the setting of ischemic heart disease, if treatment is not delayed after the onset of a myocardial infarct.
These considerations, coupled with the finding that rhNRG-1 improves cardiac function despite continued pacing in dogs with pacing-induced heart failure, suggest that its key action in all these disorders is improved integrity and performance of surviving cardiomyocytes. Given that survival is a “hard” end point for any therapeutic and given that the cardiac performance and survival benefits of rhNRG-1 were additive to those of angiotensin-converting enzyme inhibitor therapy, the findings of the present study auger well for the development of erbB-activation as a novel and potentially powerful approach to the treatment of a variety of common forms of heart failure.
The authors thank A. F. Finch for help with figures and H. Nobbs for manuscript preparation.
This work was supported in part by grants from the National Foundation of China (#863) and the Shanghai Science & Technology Committee, an Australian Postgraduate Award (Dr. Lai), and a grant from the National Health and Medical Research Council, Australia (Dr. Graham). Drs. Liu, Xinyan Li, and Zhou hold shares in Zensun Sci. & Tech. Ltd. All authors except Drs. Lai and Graham are employees. Drs. Liu and Gu contributed equally to this work.
- Abbreviations and Acronyms
- cardiac troponin I
- Coxsackie virus B3
- ejection fraction
- extracellular signal-regulated kinase
- fraction of shortening
- left anterior descending coronary artery
- left ventricle/ventricular
- left ventricular end-diastolic diameter
- left ventricular pressure development at end-diastole
- left ventricular end-systolic diameter
- left ventricular pressure development at end-systole
- mitogen-activated protein /ERK
- recombinant human neuregulin-1
- 50% tissue culture infectious dose
- Received November 14, 2005.
- Revision received April 27, 2006.
- Accepted May 30, 2006.
- American College of Cardiology Foundation
- Liu X.,
- Hwang H.,
- Cao L.,
- et al.
- Zhao Y.Y.,
- Sawyer D.R.,
- Baliga R.R.,
- et al.
- Speyer James
- Rohrbach S.,
- Yan X.,
- Weinberg E.O.,
- et al.
- Baliga R.R.,
- Pimental D.R.,
- Zhao Y.Y.,
- et al.
- Sawyer D.B.,
- Suppinger C.,
- Miller T.A.,
- et al.
- Liu X.,
- Hwang H.,
- Cao L.,
- et al.
- Lin F.,
- Owens W.A.,
- Chen S.,
- et al.
- Sam F.,
- Sawyer D.B.,
- Chang D.L.,
- et al.
- Russell K.S.,
- Stern D.F.,
- Polverini P.J.,
- et al.
- Olayioye M.A.,
- Beuvink I.,
- Horsch K.,
- et al.