OUP user menu

Selenate reduction by Enterobacter cloacae SLD1a-1 is catalysed by a molybdenum-dependent membrane-bound enzyme that is distinct from the membrane-bound nitrate reductase

Carys A. Watts , Helen Ridley , Kathryn L. Condie , James T. Leaver , David J. Richardson , Clive S. Butler
DOI: http://dx.doi.org/10.1016/S0378-1097(03)00782-1 273-279 First published online: 1 November 2003

Abstract

Enterobacter cloacae SLD1a-1 is capable of reducing selenium oxyanions to elemental selenium under both aerobic and anaerobic conditions. In this study the enzyme that catalyses the initial reduction of selenate (SeO42−) to selenite (SeO32−) has been localised to isolated cytoplasmic membrane fractions. Experiments with intact cells have shown that the putative selenate reductase can accept electrons more readily from membrane-impermeable methyl viologen than membrane-permeable benzyl viologen, suggesting that the location of the catalytic site is towards the periplasmic side of the cytoplasmic membrane. Enzyme activity was enhanced by growing cells in the presence of 1 mM sodium molybdate and significantly reduced in cells grown in the presence of 1 mM sodium tungstate. Non-denaturing polyacrylamide gel electrophoresis (PAGE) gels stained for selenate and nitrate reductase activity have revealed that two distinct membrane-bound enzymes catalyse the reduction of selenate and nitrate. The role of this membrane-bound molybdenum-dependent reductase in relation to selenate detoxification and energy conservation is discussed.

Keywords
  • Nitrate reductase
  • Selenate reductase
  • Molybdenum
  • Tungsten
  • Activity
  • Membrane-bound enzyme

1 Introduction

Selenium (Se) is a naturally occurring trace element that is found in fossil fuels, shale, alkaline soils and as a constituent in over 40 minerals [1]. It is essential for both prokaryotes and eukaryotes, but is toxic at high concentrations [2,3]. Microbes that can reduce selenium oxyanions, selenate (SeO42−) and selenite (SeO32−), are not restricted to any particular group/subgroup of prokaryotes and examples are found throughout the bacterial and archaeal domains. It has been suggested that selenate reduction may be catalysed in many cases by bacterial nitrate reductases and the selenate reductase activity of both the membrane-bound nitrate reductase (NAR) and periplasmic nitrate reductase (NAP) has been reported [4,5]. However, it is evident that nitrate reductases are poor reducers of selenate [4] and may not contribute significantly to global selenate reduction, particularly in areas enriched with both selenate and nitrate. Consequently, novel enzyme systems that catalyse the reduction of selenate selectively have been sought and to date detailed biochemical studies have been limited mainly to two species: Thauera selenatis [6,7] and Sulfurospirillum barnesii [8]. Under anaerobic conditions, both these species can respire with selenate as the sole terminal electron acceptor. Selenate is reduced to elemental selenium that accumulates exogenously, outside the cell envelope in the surrounding growth medium rather than as internal precipitates. The dissimilatory selenate reductase (SER) of T. selenatis is a periplasmic trimeric enzyme [7]. The complex has an apparent molecular mass of 180 kDa with an α subunit of 96 kDa (SerA), a β subunit of 40 kDa (SerB) and a γ subunit of 23 kDa (SerC) [7]. The gene encoding an additional protein has been sequenced, suggesting a specific chaperone assembly component (SerD) [9]. The SerC subunit is a b-type cytochrome subunit and the presence of conserved cysteine rich motifs in SerB suggests the coordination of iron–sulphur clusters. The enzyme contains molybdenum, iron and acid-labile sulphur and has an apparent Km for selenate of 16 µM. The enzyme is very substrate specific, reducing only selenate to selenite. Alternative substrates nitrate, chlorate or sulphate were not reduced by SerABC when benzyl viologen was used as the artificial electron donor [7].

The selenate reductase from S. barnesii has been studied to a lesser extent and its biochemical characterisation has been reported only in a general review article [8]. Selenate reductase activity in S. barnesii has been localised in the membrane fraction [8,10] suggesting a membrane-bound enzyme analogous to NAR [11]. Again, a b-type cytochrome has been detected in the membrane fraction from selenate grown S. barnesii, but its reduced minus oxidised difference spectrum is different to that detected in T. selenatis. The membrane fractions isolated from selenate grown S. barnesii, although exhibiting greatest activity towards selenate, also reduced substrates nitrate, thiosulphate and fumarate, even though components of these pathways were not readily detectable by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE). These results suggest that this membrane-bound selenate reductase has much broader substrate specificity than the periplasmic enzyme from T. selenatis.

Enterobacter cloacae SLD1a-1 (ATCC 700258), a bacterium isolated from Se-contaminated drainage water in the San Joaquin Valley, California, USA, can also reduce Se oxyanions to elemental selenium [12]. However, it is the ability of this organism to catalyse the reduction of selenate to selenium under aerobic conditions [12] that has interested those developing bioremediation strategies. The observation that elemental selenium accumulates near the cytoplasmic membrane prior to expulsion has led to the suggestion that the reduction of selenate to selenite by E. cloacae may occur via a membrane-bound reductase expressed under aerobic conditions [12]. In the present paper we demonstrate that E. cloacae SLD1a-1 expresses two distinct membrane-bound reductases for the reduction of nitrate and selenate. The evidence presented shows that the selenate reductase activity is associated with the cytoplasmic membrane, the enzyme involved is orientated such that its active site faces the periplasmic compartment, its activity is significantly enhanced by growth in the presence of molybdenum and it is inhibited by growth in tungsten-containing medium.

2 Materials and methods

2.1 Growth of E. cloacae SLD1a-1 and the preparation of cell extracts

E. cloacae SLD1a-1 was purchased from the American Type Culture Collection (ATCC 700258) and grown under both aerobic and anaerobic conditions in the presence of potassium nitrate, sodium selenate or sodium selenite. When grown aerobically, E. cloacae SLD1a-1 was cultured in sterile Luria–Bertani (LB) broth medium (500 ml) in 2-l flasks at 37°C (unless otherwise stated). Flasks were agitated at 180 rpm to maintain high aeration. Where appropriate the following were added to a final concentration of 10 mM: sodium selenate (Na2SeO42−), sodium selenite (Na2SeO32−), sodium molybdate (Na2MO4) and sodium tungstate (Na2WO4). For optimum enzyme activity cultures were routinely grown for 9 h. Harvested cells were broken by sonication at 14 µ amplitude (20 s on, 20 s off for 8 min) followed by low speed centrifugation (6466×g) to remove any unbroken cells debris. Soluble protein was separated from the membrane fraction by ultracentrifugation (100 000×g) for 1 h. Membrane protein was solubilised by treating with 2.5% (w/v) octyl β-d-glucopyranoside in 50 mM Tris buffer pH 8.6 for 30 min followed by ultracentrifugation at 100 000×g for 1 h. Outer membranes were prepared as described previously [13]. For anaerobic growth, E. cloacae SLD1a-1 was cultured in 500-ml medical flat bottles on minimal medium BSM supplemented with trace element solution (1 ml l−1) [14], containing either glucose, glycerol (plus formate as electron donor), lactate or succinate as the sole source of carbon and electrons in the presence of selenate or nitrate as the sole electron acceptor. To ensure that the medium was oxygen free bottles were sparged with filtered oxygen-free nitrogen for 15 min. Cultures were incubated at 37°C and growth was monitored (OD600nm) for up to 14 days. Protein concentration was determined using the Bio-Rad protein assay following the manufacturer's instructions.

2.2 PAGE and activity staining

Solubilised membrane proteins were separated using non-denaturing PAGE (6% Tris–glycine gels) and gels were stained with dithionite-reduced methyl viologen (5 mM) under anaerobic conditions using a small in-house built anaerobic chamber sparged with N2. Protein bands with selenate/nitrate reductase activity were identified by the addition of either selenate or nitrate (100 mM) for 15 min and observed as clear bands due to the substrate-dependent re-oxidation of reduced methyl viologen, changing from dark blue to colourless.

2.3 Enzyme activity assays

Reductase activity in cell fractions was measured at 20°C by following the oxidation of either reduced methyl viologen or benzyl viologen spectrophotometrically at 600 nm [11], coupled to the reduction of substrates or potential substrates; sodium selenate, potassium nitrate, potassium chlorate, potassium arsenate, potassium sulphate, potassium nitrite, dimethyl sulphoxide (DMSO) and trimethylamine N-oxide (TMAO). Values for Km and Vmax cited in the text are the mean values±S.D. (n=3) calculated using non-linear regression analysis in Grafit v3.0 (Erithacus software).

3 Results

3.1 Growth of E. cloacae SLD1a-1 on selenate- or nitrate-containing medium and the effect of molybdenum and tungsten on reductase activity

It has been established that E. cloacae SLD1a-1 can grow aerobically on selenate-containing medium and reduce selenate to elemental selenium [12] and further to dimethyl selenide [15]. It has also been reported that anaerobic growth can be supported using selenate as the sole electron acceptor when grown on glucose as the sole carbon source [12]. Since glucose fermentation can support anaerobic growth in the absence of a terminal electron acceptor we have attempted to grow E. cloacae SLD1a-1 on the non-fermentable carbon substrates: lactate and glycerol plus formate (at final concentration 15 mM) using either selenate or nitrate as the sole electron acceptor. Although lactate could support aerobic growth, no growth was observed under anaerobic conditions with selenate or nitrate. Nitrate could be used as a terminal electron acceptor to support anaerobic growth on glycerol/formate whereas selenate could not. Nitrate reductase activity was upregulated 10-fold in cells cultured under anaerobic conditions in the presence of nitrate. Aerobic growth on LB medium in the presence of selenate (1–10 mM) resulted in the formation of the red elemental selenium precipitate (Fig. 1A). The observed selenate reductase activity was enhanced 10-fold by the addition of molybdate to the culture medium suggesting that a molybdo-enzyme may be involved in the selenate reduction pathway. To test this hypothesis, cultures were grown in the presence of tungstate (10 mM). Tungstate has been shown to inhibit a number of molybdo-enzymes, including nitrate reductase [16], by substituting W for Mo at the active site. E. cloacae SLD1a-1 was grown on LB medium containing sodium tungstate in the presence of either sodium selenate or sodium selenite (Fig. 1). Cells cultured with either selenate (10 mM) or selenite (10 mM) alone grew to optical densities of OD600selenate∼6 units and OD600selenite∼1 unit respectively, and both reduced the selenium oxyanions to the red elemental selenium precipitate. Cells grown in the presence of tungstate and selenite also reached a maximum OD600 of 1 unit and maintained the ability to reduce selenite to selenium as evident by the formation of the red precipitate (Fig. 1D). In contrast, cells cultured with tungstate and selenate grew as normal reaching a maximum OD600∼5 units but failed to reduce selenate to elemental selenium (Fig. 1B), demonstrating that the enzyme responsible for selenate reduction was inhibited by tungstate, indicating that it is likely to be a molybdo-enzyme. In control experiments, cell growth was unaffected by the presence or absence of 10 mM tungstate in the growth medium (Fig. 1E,F). Selenate reductase activity does not appear to be enhanced by growth in the presence of selenate, confirming the previous observation that this is not an inducible process [12]. However, selenate reductase activity, as measured in whole cells by monitoring the substrate-dependent re-oxidation of reduced methyl viologen, was maximum in cells grown to early exponential growth phase (data not shown), after which selenate reductase activity significantly decreased as cells enter mid-late exponential phase. The effect of increasing the molybdate concentration from 1 to 20 mM and varying the temperature from 37 to 30°C had no effect on selenate reductase activity.

1

Aerobic growth of E. cloacae SLD1a-1 on medium supplemented with either selenate or selenite in the presence or absence of tungstate. A: LB medium supplemented with selenate (10 mM). B: LB medium supplemented with selenate (10 mM) and tungstate (10 mM). C: LB medium supplemented with selenite (10 mM). D: LB medium supplemented with selenite (10 mM) and tungstate (10 mM). E: LB medium only. F: LB medium supplemented with tungstate (10 mM) only.

3.2 The cellular localisation of nitrate and selenate reductase activity in E. cloacae SLD1a-1

To determine the cellular location of the selenate/nitrate reductase activity, cell fractions (soluble protein and membrane protein) from cells grown both aerobically and anaerobically were prepared and assayed for both selenate and nitrate reductase activity using reduced methyl viologen as the electron donor. Selenate and nitrate reductase activity was readily detected in the membrane fraction (specific activities: 0.29 µmol SeO42− min−1 mg−1 and 0.35 µmol NO3 min−1 mg−1). No selenate reductase activity could be detected in the soluble fraction. In contrast, a low level of nitrate reductase activity (>0.03 µmol NO3 min−1 mg−1) was observed in the soluble fraction from aerobically grown cells and probably arises from a NAP. Membrane fractions from tungstate grown cells showed a 65- and 45-fold reduction in selenate and nitrate reductase activity, respectively, providing further evidence for the selenate and nitrate reductase activity arising from a molybdenum-dependent enzyme(s). With the selenate reductase activity located in the total membrane fraction it was important to establish whether the enzyme was associated with the cytoplasmic membrane or the outer membrane. Outer membranes were prepared as described [13] and assayed as above. No selenate-dependent re-oxidation was observed using outer membranes. Selenate reductase activity was however retained by spheroplasts, confirming that the selenate reductase is associated with the cytoplasmic membrane. Using whole cells of E. cloacae SLD1a-1 grown both aerobically and anaerobically in the presence of molybdate (1 mM), selenate and nitrate reductase activity was measured using both the membrane-impermeable methyl viologen and the membrane-permeable benzyl viologen as electron donors (Fig. 2). This experimental approach has been used successfully to determine the orientation of NAR from Thiosphaera pantotropha [17]. Results show that selenate reductase activity is enhanced under aerobic conditions and that the selenate reductase can react with both methyl and benzyl viologen, with the activity being ∼2-fold higher when using methyl viologen. Given that no selenate reductase activity was detected in the soluble fraction from a possible periplasmic selenate reductase, it would seem highly likely that the membrane-associated putative selenate reductase is orientated such that its active site faces the periplasmic compartment. In contrast, nitrate reductase activity was higher in cells grown anaerobically on nitrate and was again able to accept electrons from both methyl and benzyl viologen but the nitrate reductase activity was ∼2-fold higher with benzyl viologen, consistent with the anaerobic expression of NAR that has its active site facing the cytoplasmic compartment [17]. Following cell lysis, levels of selenate/nitrate reductase activity measured in total cell extracts were equal when using either methyl or benzyl viologen as the electron donor.

2

Selenate and nitrate reductase activity measured in whole cells of E. cloacae SLD1a-1. A: Cells grown both aerobically and anaerobically and assayed for selenate reductase activity. B: Cells grown both aerobically and anaerobically and assayed for nitrate reductase activity. Enzyme activity was assayed using either methyl viologen (MV) or benzyl viologen (BV) with substrates sodium selenate and potassium nitrate at 20 and 1 mM, respectively. The data presented are mean values±S.D., n=3.

3.3 Two distinct membrane-bound enzymes catalyse the reduction of selenate and nitrate

Having established that both nitrate and selenate reductase activity was associated with the cytoplasmic membrane, a further experiment was conducted to address whether the selenate and nitrate reductase activities arose from a single membrane-associated enzyme. Solubilised membrane proteins were resolved using native PAGE, stained with dithionite-reduced methyl viologen and developed by the addition of either selenate or nitrate. Fig. 3 shows a composite photograph of a native gel cut into three sections. Lanes 1, 2 were stained to show total protein. Lanes 3, 4 and 5, 6 were stained for selenate and nitrate reductase activity, respectively. It is evident that the clear bands that arise from the substrate-dependent re-oxidation of methyl viologen are in different locations for the two substrates. The selenate reductase activity is associated with a protein that has migrated further than the protein that has detectable nitrate reductase activity. These data provide direct evidence that the selenate and nitrate reductase activities detected in the membrane fraction arise from two distinct enzymes.

3

Identification of selenate and nitrate reductase activity in native PAGE resolved membrane fractions. Composite native PAGE gel. Lanes 1 and 2, resolved membrane fractions (192 µg of protein in duplicate) stained with Simply Blue™ SafeStain (Invitrogen). Lanes 3 and 4, resolved membrane fractions (192 µg of protein in duplicate) stained for selenate reductase activity using reduced methyl viologen (5 mM) and developed by the addition of selenate (100 mM). Lanes 5 and 6, resolved membrane fractions (192 µg of protein in duplicate) stained for nitrate reductase activity using reduced methyl viologen (5 mM) and developed by the addition of nitrate (100 mM). Protein bands that display nitrate and selenate reductase activity are indicated.

3.4 Kinetics of nitrate and selenate reduction

Both nitrate and selenate reductase activity of solubilised membranes from aerobic and anaerobic cultures was measured spectrophotometrically (OD600nm) in an anaerobic cuvette using benzyl viologen as an artificial electron donor, at a range of substrate concentrations. The observed nitrate reductase activity in the solubilised membranes from nitrate grown anaerobic cultures, was typically Vmax=6.0±0.5 µmol NO3 min−1 mg−1, and the observed Km for nitrate was 0.32±0.1 mM. The nitrate reductase activity was inhibited ∼70% by sodium azide (1 mM), a known potent inhibitor of NAR [11]. Solubilised membranes from aerobically grown E. cloacae SLD1a-1 were used to study selenate reduction kinetics. From the dependence of the observed rate of benzyl viologen oxidation on selenate concentration, an observed Km of 6.25±0.5 mM was determined. The maximum specific activity was Vmax=0.20±0.02 µmol SeO42− min−1 mg−1. Selenate reductase activity was not inhibited by sodium azide (1 mM). Nitrate reductase activity in the solubilised aerobic membrane fractions was 0.09 µmol NO3 min−1 mg−1 and again was inhibited by azide. No activity (<0.01 µmol min−1 mg−1) was detected using other potential substrates including sodium arsenate, potassium nitrite, potassium sulphate, DMSO, and TMAO. Chlorate was reduced at a rate approximately three times that of nitrate (∼0.3 µmol ClO4 min−1 mg−1), but again was sensitive to azide inhibition (70% decrease in activity; 0.09 µmol ClO4 min−1 mg−1), suggesting that chlorate was being reduced by the nitrate reductase, as reported previously [11].

4 Discussion

In the present study a selenate reductase has been identified in E. cloacae SLD1a-1. The enzyme distinguishes between selenate and nitrate, is expressed under both aerobic and anaerobic conditions, is associated with the cytoplasmic membrane and is orientated such that its active site faces the periplasmic compartment. Selenate reductase activity was enhanced by growing cells on molybdate-enriched medium and inhibited by growth in the presence of tungstate. These data strongly suggest that the membrane-bound selenate reductase activity arises from a molybdo-enzyme like the well-characterised periplasmic selenate reductase from T. selenatis [7]. The solubilised membrane fractions displayed a high Km for selenate (∼6 mM) when assayed using benzyl viologen as the electron donor, compared to the low Km of 16 µM reported for the purified periplasmic selenate reductase from T. selenatis [7]. The Km for selenate of the membrane-bound enzyme is significantly higher than the range of affinities reported for in situ selenate reduction activities in a variety of sediments [18]. Furthermore, cell suspensions of E. cloacae SLD1a-1 have been demonstrated to convert 94.5% of 127 µM selenate to elemental selenium ([12], Watts, C.A. and Butler, C.S., unpublished data), suggesting a physiological Km for selenate in the micromolar range. The observed Km (selenate) may be misleadingly high owing to the use of the non-physiological electron donor benzyl viologen. A similar observation has been reported for NAR from Paracoccus denitrificans [11]. The Km (nitrate) measured using duroquinol was reproducibly ∼20-fold lower than the Km value determined using methyl viologen as the electron donor. Morpeth and Boxer [19] have also reported Km values for NAR from Escherichia coli to be ∼200-fold greater when assayed using viologen dyes compared to the values obtained with duroquinol. These authors have suggested that the apparent differences arise due to the electrons from duroquinol entering the active site via a b-type cytochrome, whereas methyl viologen can act as a direct reductant of the active site molybdenum. A similar mechanism might be in operation with the membrane-bound selenate reductase. It is likely that viologen will be donating electrons directly to the active site Mo. Currently, no information is available as to the subunit composition of the putative membrane-bound selenate reductase from E. cloacae SLD1a-1 and whilst we speculate that ubiquinol is the physiological electron donor, this remains to be established.

The novel membrane-bound enzyme detected in the present study appears unable to reduce nitrate to nitrite and this selectivity may reflect the structural distinction between nitrate and selenate. Nitrate (NO3) is trigonal planar and selenate (SeO42−) is tetrahedral. Similar substrate specificity is observed for the T. selenatis periplasmic selenate reductase [7], but is different to the broad substrate selectivity reported for the S. barnesii enzyme [8]. The combined evidence suggests that E. cloacae SLD1a-1 expresses a membrane-bound molybdo-enzyme that is biochemically distinct from the periplasmic selenate reductase and the membrane-bound nitrate reductase that is capable of reducing selenate. Currently, the full breadth of substrate selectivity of this enzyme is unknown, but this study indicates that as well as nitrate, other oxyanions including arsenate, sulphate, chlorate, nitrite, and DMSO and TMAO are not substrates.

Recently, a putative selenate reductase has also been identified in E. coli using genetic analysis based upon transposon mutagenesis [20]. These authors have shown that the mutation of a novel gene (ygfK), encoding a putative oxidoreductase, resulted in a strain unable to reduce selenate in vivo. The ygfK gene is located upstream of ygfN and ygfM in the ygfKLMN operon. Genes ygfN and ygfM encode a molybdo-protein and a FAD-containing protein, respectively. Attempts to provide direct biochemical evidence that YgfKMN is a selenate reductase complex were unsuccessful. Proteins YgfK, YgfM and YgfN are predicted to be soluble proteins without leader peptides at their N-terminus [20], suggesting that the E. coli selenate reductase is located in the cytoplasm. The location of the enzyme in the cytoplasm is interesting and suggests that the putative E. coli selenate reductase may have a physiological role in selenium assimilation rather than in respiration or detoxification.

Finally, what is the physiological function and source of electrons of this putative selenate reductase in E. cloacae SLD1a-1? Unlike the selenate reductases from T. selenatis and S. barnesii, the enzyme in E. cloacae SLD1a-1 cannot support anaerobic growth on the non-fermentable carbon substrates, such as lactate or glycerol/formate. Nevertheless, the location of the active site at the periplasmic face of the membrane makes the respiratory chain the most likely source of electrons. Assuming that the quinol pool provides the source of electrons for the reductase, the overall reaction Embedded Image will consume the two scalar protons from the periplasmic compartment released by quinol oxidation. Thus the free energy in the QH2/SeO42− couple is not conserved as proton motive force (pmf). Selenate reduction could though still be associated with generation of pmf if electrons are transferred into the Q pool through pmf-generating dehydrogenases (e.g. a proton pumping reduced nicotinamide adenine dinucleotide (NADH) dehydrogenase). Even if this is the case, there is threshold dependence between pmf and growth rate [21] and thus if the steady-state pmf generated during selenate reduction is below this, no significant growth will be detected. Thus the main role of this molybdo-enzyme appears not to be in supporting anaerobic growth. The location of the active site is, however, consistent with a role in detoxification and enables the organism to protect itself against elevated levels of both selenate and the reduction product selenite. The enzyme will be well-suited to this, because, since QH2-selenate electron transfer is not electrogenic, the process will not be subject to negative thermodynamic feedback on the system via pmf generated through aerobic respiration. Whilst the enzyme that catalyses the reduction of selenite to selenium is currently unknown, it has been suggested that a periplasmic nitrite reductase may be responsible [22,12]. Enteric bacteria, as exemplified by E. coli, express both periplasmic (Nrf) and cytoplasmic (Nir) nitrite reductases during anaerobic growth on nitrite (reviewed by Potter et al. [23]). NrfA is a 50-kDa cytochrome c552 that catalyses the six-electron reduction of nitrite to ammonia. Given the location of the putative selenate reductase active site, we speculate that Nrf may also catalyse the four-electron reduction of selenite to elemental selenium. We are currently seeking evidence for the involvement of NrfA in selenite reduction in E. cloacae SLD1a-1, but as selenite reduction is readily observed under aerobic conditions we cannot exclude the possibility of a specific selenite reductase.

Acknowledgements

This work was supported by a BBSRC research grant (13/P17219) to C.S.B. and D.J.R. We would like to thank Dr John Stolz (Duquesne University) for providing advice on culturing cells on selenate and we acknowledge Dr James Allen, Prof. Stuart Ferguson (University of Oxford) and Prof. Jeff Cole (University of Birmingham) for useful discussions.

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].
View Abstract