OUP user menu

Aprotinin impairs coronary endothelial function and down-regulates endothelial NOS in rat coronary microvascular endothelial cells

Sibel Ülker, Pascal P McKeown, Ulvi Bayraktutan
DOI: http://dx.doi.org/10.1016/S0008-6363(02)00489-3 830-837 First published online: 1 September 2002

Abstract

Objective: The non-specific serine protease inhibitor aprotinin is currently used to reduce blood loss and the need for blood transfusion after cardiopulmonary bypass. We have recently reported that aprotinin impairs endothelium-dependent but not endothelium-independent relaxations in rat thoracic aortic rings due to its inhibitory effect on endothelial nitric oxide (NO) production. In light of these findings, the current study was designed to investigate the effects of aprotinin on coronary endothelial function in isolated rat hearts and on the expression of endothelial NO synthase (eNOS) in cultured rat coronary microvascular endothelial cells (CMEC). Methods: Hearts obtained from Sprague–Dawley rats were perfused on a constant flow Langendorff isolated heart system and coronary perfusion pressure and cardiac parameters were recorded. The coronary relaxant responses to bolus infusions of bradykinin (BK) and sodium nitroprusside (SNP) were recorded in the absence and presence of aprotinin. Total RNA and protein samples were extracted from CMEC incubated with aprotinin or the vehicle, 0.9% NaCl. RT-PCR-Southern blotting and Western analyses were carried out to assess eNOS mRNA and protein levels, respectively. Results: Aprotinin (125 and 250 kIU/ml) increased coronary perfusion pressure without changing the heart rate and cardiac contractility. Aprotinin inhibited BK-induced coronary vasodilatation at 250 kIU/ml, but not at 125 kIU/ml concentrations. The relaxant response to SNP did not change in response to either concentration of the drug. Incubation of CMEC with aprotinin down-regulated eNOS mRNA and protein at 250 kIU/ml, but not at 125 kIU/ml concentration. Conclusion: These data suggest that aprotinin selectively inhibits NO synthesis at higher doses (≥250 kIU/ml) and therefore impairs endothelium-dependent coronary vascular tone. This effect of the drug may contribute to its ‘blood-sparing’ action, but may also account for the increase in the incidence of postoperative graft thrombosis observed in clinical practice during aprotinin therapy.

Keywords
  • Coronary circulation
  • Endothelial function
  • Gene expression
  • Nitric oxide
  • Vasoconstriction/dilation

Time for primary review 24 days.

1 Introduction

Excess bleeding possibly due to abnormal platelet function is a common source of morbidity and mortality after cardiopulmonary bypass operations [1,2]. Aprotinin, a non-specific serine protease inhibitor, is currently used to inhibit the need for blood transfusions after cardiac operations [3,4]. It has been shown that aprotinin infusion preserves platelet glycoprotein Ib receptor function and therefore improves platelet adhesion [5]. Furthermore, aprotinin also inhibits plasmin [4] and some other proteases in the coagulation cascade and fibrinolysis process, i.e. kallikrein, activated protein C and tissue plasminogen activator [6]. Since endothelial cells play pivotal roles in the regulation of coagulation and fibrinolysis as well as platelet adhesion and activation via regulation of the synthesis and/or release of several vasoactive mediators, it is possible that aprotinin could directly affect endothelial cell function. Indeed, it has been reported that aprotinin increases the release of von Willebrand factor and thromboxane B2 but decreases the release of 6-keto-prostaglandin F in cultured human umbilical vein endothelial cells thereby leading to enhanced platelet aggregation [7,8].

Nitric oxide (NO) is known to be the most important endothelium-derived substance that regulates platelet function, vasomotor tone as well as blood flow and inhibits neutrophil adhesion to the endothelium. NO is generated by a class of enzymes known as NO synthases (NOSs). Three main isoforms of NOS have so far been described; the constitutive isoforms neuronal NOS (nNOS) and endothelial NOS (eNOS) and the inducible isoform (iNOS) (for a recent review, see Ref. [9]). It has been shown that the production of NO is significantly increased during and after cardiopulmonary bypass [10]. The systemic endotoxaemia that occurs with the institution of cardiopulmonary bypass is a potent stimulus for the release of proinflammatory cytokines, which induce iNOS expression and subsequent NO release [11]. Several studies have reported that aprotinin reduces NO production in vivo in the airways of patients undergoing cardiopulmonary bypass and decreases cytokine-induced iNOS expression in a dose-dependent manner in vitro in cultures of a murine epithelial cell line [12,13]. Apart from an inhibitory effect on iNOS expression in inflammatory cell lines, aprotinin has recently been shown to inhibit both nNOS and iNOS in whole rat brain homogenates; hence it is accepted as the ‘first competitive protein inhibitor of NOS activity’ [14].

Although extensive studies have shown the beneficial ‘blood-sparing’ effect of aprotinin in patients undergoing cardiopulmonary bypass, there have been some clinical reports questioning the overall safety of aprotinin therapy [15,16]. The concern about the drug is that aprotinin therapy may lead to a subclinical hypercoagulable state resulting in an increase in the incidence of postoperative graft thrombosis and thromboembolic phenomena that may adversely influence patient outcome. The inhibitory effect of aprotinin on the excessive release of NO by iNOS is suggested to have beneficial effects in reversing the inflammatory response to cardiopulmonary bypass. However, similar inhibitory effects on eNOS-derived NO may lead to enhanced platelet activation, thrombosis and endothelial dysfunction, i.e. impaired endothelium-derived vascular relaxation. Indeed, incubation of rat thoracic aortic rings and cultured rat CMEC with aprotinin resulted in impaired endothelium-dependent relaxation and NO release, respectively [17].

In light of these findings, the aim of the current study was to determine whether aprotinin alters coronary vascular tone in isolated rat hearts and affects eNOS mRNA and protein expressions in rat cultured CMEC.

2 Methods

The investigation conforms with the ‘Guide for the Care and Use of Laboratory Animals’ published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996).

2.1 Isolated rat heart perfusion

Hearts were excised from anticoagulated, anaesthetised male Sprague–Dawley rats (350–450 g; 100 U i.v. heparin and 100 mg/kg i.p. pentobarbitone sodium) and immersed in ice-cold Krebs–Henseleit solution of the following composition (mM): NaCl 118.0, KCl 4.5, KH2PO4 1.4, MgSO4 1.2, NaHCO3 25, CaCl2 1.4 and glucose 11; pH 7.4. Hearts were initially perfused retrogradely through the aorta by means of a Langendorff heart set-up at a constant flow with Krebs–Henseleit solution maintained at 37 °C and gassed with 95% O2 and 5% CO2. The flow rate was determined according to animal weight using the formula: flow (ml/min)=x0.56×7.43 (x is the heart weight), heart weight (mg)=0.0027y+0.6 (y is the body weight) [18]. Coronary perfusion pressure was monitored continuously as an index of coronary microvascular tone with a pressure transducer (SensoNor Sp 844, Norway) connected to a side arm of the aortic cannula. Since total flow through the coronary vascular system changes depending on the cardiac contractile activity, left ventricular pressure, as an index of cardiac contractility, was also measured with a fluid-filled latex balloon inserted into the left ventricle and connected to a second pressure transducer. Balloon volume was adjusted to obtain an end-diastolic pressure of 10 mmHg. All variables were recorded continuously on a computer through a data-acquisition system (Chart v4.1, Powerlab/4SP, ADInstruments, UK). Heart rate, left ventricular developed pressure (LVDP, the difference between systolic and diastolic left ventricular pressures) and the positive and negative differentiated pressures (+dP/dt and −dP/dt) were monitored simultaneously as computed inputs derived from left ventricular pressure. The maximum value recorded for contraction rate (+dP/dtmax) was used as an inotropism index and the minimum value (−dP/dtmax) served as a relaxation index. After completion of the instrumentation, temperature was maintained by placing the hearts in a temperature-regulated glass jacket throughout the experiment.

The heart was allowed to equilibrate for at least 30 min in order to obtain stable cardiac parameters. Then the perfusate was changed to Krebs solution containing 3.2 mM potassium, which was prepared by equimolar replacement of KCl by NaCl. This small reduction in extracellular potassium concentration resulted in an increase in coronary perfusion pressure thus allowing vasodilator response to be observed. After an equilibration period of 20 min, vasodilator agents, bradykinin (endothelium-dependent, 0.0001–100 nM) or sodium nitroprusside (endothelium-independent, 0.1–1000 nM) were administered to the coronary vascular bed as bolus infusions. Only one vasodilator agent was used on each heart. After obtaining the control response, the perfusate was switched to 3.2 mM potassium Krebs containing aprotinin (125 or 250 kIU/ml) and concentration-dependent vasodilator response to the same agent was recorded in the presence of aprotinin. Aprotinin used in this study was the commercially available form of the drug which contained 1.4 mg/ml (10 000 kIU/ml) of the active compound in sterile 0.9% NaCl. Therefore, in order to exclude NaCl overload, NaCl concentration of the Krebs solution containing aprotinin was adjusted prior to use. Nω-Nitro-l-arginine methyl ester (l-NAME, 0.1 mM) was also used, as a reference drug known to inhibit eNOS activity, to record bradykinin-induced vasodilator response.

2.2 Isolation and culture of rat CMEC

CMEC were isolated as previously described [19]. Briefly, the hearts were mounted and perfused retrogradely on a constant-flow Langendorff system with 0.04% collagenase. The ventricles were chopped and collagenase digestion was quenched by the addition of bovine serum albumin to the perfusate. CMEC were obtained by sedimentation of myocytes and incubated in 0.01% trypsin at 37 °C for the prevention of nonendothelial cell attachment. Cells were then activated by washing in calcium and suspended in Medium 199 supplemented with l-glutamine, foetal calf serum and antibiotic/antifungal agents. Cell suspensions were plated and incubated at 37 °C under 5% CO2. After 1-h incubation, unattached cells were washed off with saline and remaining cells were cultured to confluence.

Cultured CMECs were characterised by their typical ‘cobblestone’ morphology, uptake of fluorescently labeled acetylated LDL by >99% of cells and the rapid formation of capillary-like tubes on the basement membrane preparation Matrigel. For further culture, cells were trypsinized and subcultured. Confluent cells between passages 3 and 5 were incubated with aprotinin (125 and 250 kIU/ml) overnight. The cells incubated with the vehicle, 0.9% NaCl, served as controls.

2.3 Analysis of the eNOS mRNA expression by RT-PCR and Southern blotting

After incubation with aprotinin, cells were washed twice in ice-cold phosphate buffered saline (PBS) and total RNA was extracted using the guanidinium thiocyanate–phenol–chloroform extraction method [20]. Reverse transcription (RT)-polymerase chain reaction (PCR) reactions were carried out using the Superscript One-Step RT-PCR kit according to manufacturer's instructions (Invitrogen). The gene-specific primers used to amplify a 324-bp eNOS PCR product were based on previously published sequences [21]; 5′-GGGCCAAGGGTGATGAGCTCTG-3′ and 5′-CCCTCCTGGCTTC CAGTGTCC-3′ (sense and antisense, respectively). GAPDH was used as a reference gene and the primer sequences were as follows [22]; 5′-TATGACAACTCCCTCAAGAT-3′ and 5′-AGATCCACAACGGATACATT-3′ (sense and antisense, respectively).

The amplified PCR products were size fractionated using agarose gel electrophoresis and visualised under UV using ethidium bromide staining. The products were transferred to nylon Hybond-N membrane (Amersham Pharmacia) by capillary transfer [23] and hybridised with fluorescein-labelled eNOS or GAPDH cDNA probes. The immunocomplex was developed according to manufacturer's instructions (Amersham Pharmacia) and exposed to X-ray films. The intensities of eNOS and GAPDH bands were quantified by scanning densitometry. The values obtained for eNOS were normalised with respect to GAPDH [24].

2.4 Analysis of eNOS protein expression by Western blotting

After incubation with aprotinin, cells were washed twice in ice-cold PBS and lysed in boiling lysis solution containing 1% SDS and 10 mM Tris pH 7.4. The insoluble material was removed by centrifugation and protein concentration in the supernatant was measured by the Lowry method (Bio-Rad). Equal amounts of protein were run on 8% SDS–polyacrylamide gels and electroblotted onto nitrocellulose membrane (Hybond N+). Equal rate of transfer among lanes was confirmed by reversible staining with Ponceau S (Sigma). The membrane was then incubated with 1/1500 dilution of a mouse monoclonal antibody raised against a polypeptide (amino acid residues 1030–1209) of the human eNOS protein (BD Transduction Labs). A horseradish peroxidase-linked antimouse antibody was used as a secondary antibody (Amersham Pharmacia). The immunocomplex was developed using an ECL Plus detection kit (Amersham Pharmacia) and exposed to X-ray film. The autoradiographs were analysed by scanning densitometry with subtraction of the background counts measured outside loaded lanes. The values for aprotinin samples were normalised by the mean value obtained with control PBS samples.

2.5 Statistical analysis

Data are expressed as mean±S.E.M. Changes in coronary perfusion pressure were expressed as the percentage of the baseline value. Statistical analysis of the isolated heart data was carried out by Student's t-test for paired observations or one-way analysis of variance followed by a post-hoc Bonferroni's test, as appropriate. Student's t-test for unpaired observations was used for the comparison of molecular biology data. P-values less than 0.05 were considered to be statistically significant.

3 Results

3.1 Effects of aprotinin on cardiac parameters

The hearts were initially perfused in 5.9 mM K+ containing solution for 30 min in which time the small increases in coronary microvascular pressure were stabilised. Reduction of potassium concentration to 3.2 mM in the perfusate significantly increased coronary perfusion pressure from 46.3±2.2 to 81.7±3.4 mmHg (P<0.001) but decreased heart rate (P<0.01; 238±7 vs. 210±7 beats/min), +dP/dtmax (P<0.01; 3647±240 vs. 3314±237 mmHg/s) and the −dP/dtmax (P<0.001; 1895±113 vs. 1546±92 mmHg/s). LVDP, however, was not altered (99.8±6.6 vs. 99.0±7.7 mmHg) by the reduction in K+ levels.

In contrast, constant perfusion of the hearts with aprotinin (125 and 250 kIU/ml) significantly increased the coronary perfusion pressure (P<0.01, Table 1) without changing the heart rate, LVDP, +dP/dTmax and −dP/dTmax values.

View this table:
Table 1

Effects of aprotinin (125 and 250 kIU/ml) on cardiac parameters

Aprotinin 125 kIU/mlAprotinin 250 kIU/ml
BeforeAfterBeforeAfter
Perf. pressure (mmHg)89.1±8.1119.0±9.1*79.4±8.3100.9±8.6*
Heart rate (beats/min)185±13181±12195±10190±14
LVDP (mmHg)94.8±8.796.7±8.785.2±8.783.3±8.0
+dP/dtmax (mmHg/s)2825±2242807±2242893±3862862±405
−dP/dtmax (mmHg/s)1439±1141454±1071346±1331342±144
  • Perf. pressure indicates coronary perfusion pressure; LVDP, left ventricular developed pressure; n = 8. *P<0.01, after versus before aprotinin.

Bolus infusions of bradykinin (BK, 0.0001–100 nM) resulted in a dose-dependent reduction in coronary perfusion pressure (Fig. 1). While perfusion of the hearts with 125 kIU/ml concentration of aprotinin did not alter the BK-induced reduction in coronary perfusion pressure (Fig. 1A), 250 kIU/ml of aprotinin significantly inhibited relaxant responses to BK (P<0.01, Fig. 1B).

Fig. 1

Effect of aprotinin at 125 (A) and 250 kIU/ml (B) on the bradykinin-induced coronary vasodilator response. Results are expressed as the percentage of reduction in coronary perfusion pressure from basal values. Data are shown as the mean±S.E.M. *P<0.01 after versus before; n = 8.

The contribution of endogenous NO release to BK-induced coronary vasodilatation was investigated by use of l-NAME (0.1 mM), a competitive inhibitor of eNOS. Perfusion of the hearts with l-NAME resulted in approximately 36% (P<0.01) increase in coronary perfusion pressure compared to the values obtained with 3.2 mM K+. However it did not affect the heart rate, LVDP, +dP/dtmax and −dP/dtmax (data not shown). l-NAME also significantly inhibited BK-induced reduction in perfusion pressure (P<0.001, Fig. 2), an effect that was similar to that of 250 kIU/ml concentration of aprotinin.

Fig. 2

Effect of l-NAME (0.1 mM) on the bradykinin-induced coronary vasodilator response. Results are expressed as the percentage of reduction in coronary perfusion pressure from basal values. Data are shown as the mean±S.E.M. *P<0.01 or **P<0.001 after versus before; n = 7.

Endothelium-independent relaxing agent sodium nitroprusside (SNP, 0.1–1000 nM) caused a dose-dependent reduction in coronary perfusion pressure (Fig. 3). The coronary relaxant responses to SNP were independent of the concentrations of aprotinin.

Fig. 3

Effect of aprotinin at 125 (A) and 250 kIU/ml (B) on the sodium nitroprusside (SNP)-induced coronary vasodilator response. Results are expressed as the percentage of reduction in coronary perfusion pressure from basal values. Data are shown as the mean±S.E.M.; n = 6. P>0.05.

3.2 Effects of aprotinin on the expression of eNOS mRNA

Treatments of CMEC with 125 kIU/ml concentration of aprotinin did not significantly alter the eNOS mRNA expression. However, the use of 250 kIU/ml of aprotinin resulted in a significant decrease in eNOS mRNA levels by approximately 60% (P<0.01, Fig. 4a,c). Neither concentration of aprotinin altered GAPDH mRNA levels, hence the alterations in eNOS mRNA levels were normalised to GAPDH (Fig. 4b).

Fig. 4

Representative Southern blots of eNOS transcripts (A) and GAPDH transcripts (B) amplified from control CMEC (lane 1) and 125 or 250 kIU/ml aprotinin-treated CMEC (lanes 2 and 3, respectively) and histogram (C) showing changes in eNOS mRNA in rat CMEC before and after aprotinin treatments following normalisation with respect to the levels of GAPDH levels (n = 4). **P<0.01.

3.3 Effects of aprotinin on the expression of eNOS protein

Similar to its effect on eNOS mRNA, incubation with 250 kIU/ml but not 125 kIU/ml concentration of aprotinin significantly decreased eNOS protein levels (by approximately 60%) in CMECs (P<0.01, Fig. 5a,b).

Fig. 5

Western blot showing expression of eNOS protein in cultured rat CMEC (lane 1) and 125 (lane 2) and 250 kIU/ml (lane 3) aprotinin-treated CMEC. Similar results were obtained in four different experiments. **P<0.01.

4 Discussion

NO plays pivotal roles in the regulation of vascular tone and inhibition of platelet aggregation and adhesion [25,26] when generated by eNOS in a large variety of cells including cardiac myocytes and the endothelial cells of the coronary vasculature [27]. In contrast, a large production of NO by iNOS is involved in pathological phenomena [28], hence the regulation of NO is critical in many biological systems.

The data presented in this study reveal that perfusion of isolated rat hearts with a serine protease inhibitor, aprotinin, increases the basal tone of coronary vessels and inhibits BK-induced endothelium-dependent relaxation of coronary vessels. The increase in the coronary resistance following perfusion with aprotinin may be due to two possibilities. Firstly, a direct inhibition of synthesis and/or release of basal NO from the coronary endothelium and secondly, an inhibitory effect of the drug on kallikrein, the enzyme necessary for the synthesis of endogenous BK within the coronary vessel wall. Clinical studies with aprotinin have revealed that low-dose (140 mg i.v. loading dose, 140 mg pump prime and 35 mg/h i.v. constant infusion) and high-dose (twofold the doses given in low-dose regime) protocols result in plasma concentrations of around 125 and 250 kIU/ml, respectively [29], therefore, concentrations used in the present study were adjusted to remain within this range throughout the study. The observation that both concentrations of aprotinin increased coronary resistance, while only the higher concentration inhibited BK-induced relaxation, suggests that the drug at higher concentrations may, in addition to inhibition of basal NO availability in endothelial cells, indirectly inhibit local BK formation through the kallikrein pathway. Indeed, approximately 200 kIU/ml concentration of aprotinin has been reported to be sufficient to inhibit plasma kallikrein [6]. It is noteworthy in this context, tissue kallikrein, a smaller molecule than plasma kallikrein, may be inhibited by aprotinin at similar concentrations.

In the current study, aprotinin at 250 kIU/ml inhibited the coronary relaxation induced by exogenous administration of BK and did not alter SNP-induced coronary relaxant responses. This may suggest that aprotinin selectively impairs agonist-induced release of endothelium-derived NO from coronaries and also the possibility that aprotinin might inhibit cGMP effector cascade is unlikely. This hypothesis was confirmed by the inhibition of BK-induced coronary relaxation by l-NAME, a nonselective inhibitor of NOS. The impairment of endothelium-dependent coronary dilatation following aprotinin perfusion might further enhance vasoconstrictor tone in coronary vasculature and predispose to vasospasm. This in vitro observation is consistent with the clinical reports indicating an increase in the incidence of graft thrombosis and myocardial infarction in cardiopulmonary bypass (CPB) patients receiving aprotinin [15].

Although, aprotinin is used in heart surgery mainly for its haemostatic effects, it has also been suggested to provide an anti-inflammatory benefit due to its combined protective effect on neutrophils. Indeed, aprotinin initially prevents neutrophil activation within the circulation, as assessed by diminished expression of Mac-1 (CD11b/CD18) [30], and later prevents secretion of histotoxic mediators such as reactive oxygen species within the tissues [31]. In addition to these, a more recent study has shown that aprotinin inhibits leucocyte extravasation through endothelial cells, thereby preventing rapid accumulation of large numbers of neutrophils and other inflammatory cells within major organs [32]. It has been reported that there are two main types of aprotinin target in the inflammatory system, i.e. soluble proteases such as kallikrein and plasmin and cell-associated proteases, e.g. protease-activating receptors (PARs), a subset of a larger family of seven transmembrane G protein-coupled cell surface receptors [33]. Soluble proteases may be involved in the inhibitory effect of aprotinin on leukocyte extravasation [34]. PARs are expressed on leucocytes, platelets, throughout the vasculature and endothelium and may be involved in the activity of aprotinin in reduced endothelial cell activation and reduced vascular permeability. It has been reported that PARs 1, 2 and 3 are expressed on endothelial cells (see Ref. [33] for a review). However, the effect of aprotinin on each of these receptors with regards to regulation of thrombosis and endothelial cell responsiveness to inflammatory signals remains to be established.

This study also demonstrates for the first time that aprotinin decreases eNOS mRNA and protein expression in rat CMEC in a dose-dependent manner. The mechanism by which aprotinin down-regulates eNOS in CMEC is unclear; however, previous reports have revealed that some protease inhibitors of the chloromethylketone group prevent expression of iNOS in vascular smooth muscle cells by blocking the activation of nuclear factor κB (NFκB) [35]. Consistent with this concept, the promoter region of the eNOS gene also contains a nuclear factor-1 (NF-1)-like binding site [36]. Therefore, it is likely that inhibition of eNOS expression by aprotinin, a protease inhibitor with pharmacologic properties similar to chloromethylketones, may result from a similar mechanism.

Despite its inhibitory effects on eNOS mRNA and protein expression in CMEC, aprotinin (125, 250 and 500 kIU/ml doses) altered neither the basal release of NO, as measured by Griess reaction, from these cells nor basal tonus of the rat thoracic aortic rings [17]. However, aprotinin inhibited calcium ionophore A23187-stimulated release of NO metabolites from CMEC and selectively enhanced phenylephrine-induced contractions in aortic rings thereby suggesting an effect on Ca2+-dependent activation of eNOS [17]. Although, these results may, in part, be extrapolated to the isolated heart system, it is noteworthy that in Langendorff systems, the coronary arteries, compared to vascular rings, due to constant perfusion are subjected to shear stress which affects the expression of eNOS and perhaps the generation of NO metabolites [37].

In summary, this study suggests that at higher doses (≥250 kIU/ml) aprotinin selectively inhibits NO synthesis and therefore impairs endothelium-dependent coronary vascular tone, i.e. it induces coronary endothelial dysfunction as a result of its inhibitory effect on eNOS gene and protein expression. This effect of the drug may suggest a new perspective in explaining its ‘blood-sparing’ action, but also may account for the increase in perioperative thrombosis risk observed in cardiopulmonary bypass patients receiving aprotinin.

Acknowledgements

This work was supported by funding from the Heart Trust Fund (Royal Victoria Hospital), Belfast, UK.

References

  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
  7. [7]
  8. [8]
  9. [9]
  10. [10]
  11. [11]
  12. [12]
  13. [13]
  14. [14]
  15. [15]
  16. [16]
  17. [17]
  18. [18]
  19. [19]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
  30. [30]
  31. [31]
  32. [32]
  33. [33]
  34. [34]
  35. [35]
  36. [36]
  37. [37]
View Abstract