OUP user menu

Identification of low-dye-binding (ldb) mutants of Saccharomyces cerevisiae

Isaac Corbacho, Isabel Olivero, Luis M Hernández
DOI: http://dx.doi.org/10.1016/S1567-1356(03)00162-4 437-444 First published online: 1 January 2004


We have completed the identification of Saccharomyces cerevisiae genes that are defective in previously isolated ldb (low-dye-binding) mutants. This was done by complementation of the mutant's phenotype with DNA fragments from a genomic library and by running standard tests of allelism with single-gene deletion mutants of similar phenotype. The results were as follows: LDB2 is allelic to ERD1; LDB4 to SPC72; LDB5 to RLR1; LDB6 to GON7/YJL184W; LDB7 to YBL006C; LDB9 to ELM1; LDB10 to CWH36; LDB11 to COG1; LDB12 to OCH1; LDB13 to VAN1; LDB14 to BUD32; and LDB15 to PHO85. Since the precise function of some of the genes is not known, these data may contribute to the functional characterization of the S. cerevisiae genome.

  • Saccharomyces cerevisiae
  • Glycosylation
  • Deletion mutant
  • ldb mutant
  • Oligosaccharide phosphorylation
  • ldb, low-dye-binding
  • mnn, mannan-defective
  • MP, mannosylphosphate

1 Introduction

The N-linked oligosaccharides of Saccharomyces cerevisiae have diesterified phosphate groups attached to position 6 of particular mannoses in the molecule [13]. The phosphate groups confer a net negative charge to the cell surface which might represent an advantage in some particular environments. However, neither the precise function of these groups nor the complete biosynthetic pathway are well understood yet (reviewed in [4]). Pioneer studies on the transfer of mannosylphosphate (MP) groups into N-linked oligosaccharides were carried out by Ballou and co-workers some 30 years ago. They isolated two non-conditional mannan-defective mutants –mnn4[5] and mnn6[6]– which showed almost total absence of phosphate groups in the mannoprotein-linked oligosaccharides. They also set up an assay for mannosylphosphate transferase activity [6] and established a direct correlation between the phosphate content and the affinity of the whole yeast cells for the cationic dye alcian blue [5]. The identification of the genes revealed that both MNN4 and MNN6 are directly involved in the phosphorylation process [7,8].

In the last few years we have been studying the phosphorylation of mannoprotein-linked oligosaccharides through the selection and characterization of mutants generated in random mutagenesis experiments. We isolated 15 S. cerevisiae low-dye-binding (ldb) mutants which show a reduction in the level of phosphate groups attached to the N-linked oligosaccharides of the mannoproteins [911]. LDB1, LDB3, and LDB8 were already identified as aliases of PMR1[12], VRG4, and MNN2[9], respectively. PMR1 encodes a Golgi-located Ca2+/Mn2+-ATPase required for the normal function of the Golgi apparatus [13], Vrg4p is involved in nucleotide-sugar transport in the Golgi [14], and Mnn2p is an α(1,2)-mannosyltransferase responsible for the initiation of branching in the outer chain of N-linked oligosaccharides [15]. In all three cases, the reduction in MP groups attached to N-linked oligosaccharides seems to be a consequence of the malfunction of the Golgi or a reduction in size of the N-linked oligosaccharides, rather than a defect in the biosynthetic pathway and/or transfer of the MP groups. In the present work we show the identification of the genes that are defective in the remaining unidentified ldb mutants. The functions of the genes are diverse and most of them are involved in such general processes as secretion–glycosylation, as well as in cell wall organization or stress response. It is remarkable that ldb6 and ldb7 are defective in hypothetical open reading frames (ORFs) of unknown function and that in other cases, the function has not clearly been established yet. In this sense, our results may contribute to the functional characterization of the S. cerevisiae genome.

2 Materials and methods

2.1 Strains, sporulation, and growth conditions

S. cerevisiae: BY4741 (MATa; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0); BY4742 (MATα; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0); and the complete collection of haploid deletion strains (Mata) were obtained from EUROSCARF (European Saccharomyces cerevisiae Archive for Functional Analysis). The mutants ldb2, ldb4, ldb5, ldb6, ldb7, ldb9, ldb10, ldb11, ldb12, ldb13, ldb14, and ldb15 were from previous studies [9,11]. The ura3 marker was introduced into all of the strains by standard genetic techniques.

Cells were grown at 30°C in solid or liquid YEPD (1% yeast extract, 2% peptone, 2%d-glucose) or a synthetic medium without uracil (SD−ura). When indicated, the YEPD medium was supplemented with sodium orthovanadate. To derepress external invertase synthesis, the amount of glucose was lowered to 0.1%. Sporulation was done following standard protocols [16].

Escherichia coli DH5α was obtained from Gibco BRL (Carlsbad, CA, USA) and grown at 37°C in commonly used bacterial media: LB, SOC, or TB supplemented with 50 mg ml−1 of ampicillin when indicated.

2.2 Dye-binding assays

The staining of the yeast cells with alcian blue was done in duplicate in microtiter plates as in [17]. The method is a modification of that described previously [3,9].

The yeast strains were inoculated into the wells of sterile microtiter plates loaded with 0.2 ml per well of liquid YEPD. As control, strains of known dye-binding behavior were also inoculated on each plate: wild-type, ΔYPL050C (Δmnn9) and ΔYPL053C (Δmnn6). These were taken as reference for the evaluation of the intensity of the staining reaction. The values assigned to these strains were: 0 for Δmnn6, 3 for Δmnn9, and 5 for wild-type [9].

After inoculation, the plates were covered with a sterile lid, allowed to grow for 48 h at 30°C, and then kept at 4°C for 48 h before starting the staining procedure.

For staining, the plates were centrifuged using an appropriate rotor, discarding the culture medium. The cells in the pellet were washed with 0.2 ml of 50 mM acetic acid, pH 3, using multichannel pipettes. After a new centrifugation, the pellets were resuspended in 0.2 ml of a solution of 0.1% alcian blue in 50 mM acetic acid. Finally, the cells were washed twice with 50 mM acetic acid, pH 3, without removing the supernatant the second time.

2.3 Transformation and selection of transformants

For the transformation of competent cells of E. coli DH5α, we followed the instructions of the supplier. S. cerevisiae was transformed by the lithium acetate method [18]. Most ldb mutants were selected and identified using QAE-Sephadex beads and alcian blue dye by a method based on the decrease in net negative charge of the cell surface in cells with reduced amounts of phosphate attached to the manno-oligosaccharides [9]. The mutants did not bind the beads and showed a decrease in the affinity for the dye. In this work we used the QAE-Sephadex beads to isolate the transformed cells that recovered the wild-type phenotype. The ldb mutants were transformed with the genomic library CEN BANK constructed in plasmid YCp50 which carries the URA3 marker [19]. The transformants were mixed with the QAE-Sephadex beads at pH 3. Cells that have recovered the negative charge on the cell surface have also recovered the ability to bind to the beads. The unbound cells were discarded and the bound cells were released by washing the beads with 2 M NaCl. The released cells were washed with 0.9% NaCl, and spread onto SD−ura plates to obtain individual colonies. Finally, the colonies were stained with alcian blue to assess their recovery of the wild-type phenotype. This method had been used successfully in the identification of the LDB1 gene [12].

2.4 Other methods

Native gel electrophoresis of external invertase was done as described previously [3,9]. Fig. 1 shows the invertase migration pattern of representative strains.

Figure 1

Native gel electrophoresis of secreted invertase in representative S. cerevisiae strains. Lane 1, wild-type; lane 2, mnn9; lane 3, mnn2; lane 4, ldb11. The narrow band at the bottom of all lanes corresponds to the constitutive non-glycosylated form of the enzyme.

Plasmids were recovered from S. cerevisiae by breaking the cells with glass beads [20] and from E. coli by using commercial plasmid purification kits from Promega (Madison, WI, USA). DNA sequencing was done in an ABI Prism 377 DNA Sequencer (Centro de Secuenciación de DNA, Facultad de Farmacía, Universidad Complutense de Madrid, Spain) using the BigDye Terminator Cycle Sequencing Ready Reaction Kit, from PE Biosystems (Foster City, CA, USA). Two primers were designed around the BamHI site of YCp50 to sequence a fragment on each side of the insert. Sequence comparisons against GenBank were performed using the FASTA program on the S. cerevisiae Genome Database (SGD) web page (http://genome-www.stanford.edu/Saccharomyces/). This page was also routinely used to obtain general information about the genes.

Micromanipulation and dissection of yeast asci was done with the aid of a Tetrad dissection system from Micro Video Instruments (Avon, MA, USA). Microphotographs were taken at 600× with a Nikon Eclipse E600 equipped with Nomarski optics and a Zeiss AxioCam digital camera.

3 Results and discussion

3.1 Identification of LDB genes by complementation of the mutant phenotypes

Table 1 summarizes some phenotype characteristics of the ldb mutants that are relevant for this study. The alcian blue affinity defines the ldb phenotype. The size of external invertase is indicative of the function of the glycosylation machinery ranging from wild-type to mnn9 (Fig. 1). The formation of clumps might reflect defects in cell wall organization and/or biosynthesis as well as problems in cell cycle completion. Fig. 2 shows microphotographs of ldb mutants with clumpy phenotype and/or significant alterations in cell shape. The wild-type and mnn9 strains are shown as controls of known behavior. Vanadate resistance has also been linked to glycosylation defects. The wild-type can grow at 2 mM sodium orthovanadate but not at 10 or 15 mM, while the mnn9 strain is resistant to all three concentrations [21].

View this table:
Table 1

Phenotype characteristics of ldb mutants

MutantAlcian blueaIMPbClumpscGrowth in vanadatedMutated gene
2 mM10 mM15 mM
  • aThe values range from 0 (white, like mnn6 or ldb1) to 5 (dark blue, like wild-type).

  • bInvertase migration pattern (see Fig. 1); wt: wild-type.

  • cThe number of + symbols indicates relative degree of aggregation. ldb9 also shows alterations in cell shape (see Fig. 2C).

  • dFour + symbols indicate normal growth.

Figure 2

Microphotographs of strains showing some alterations in cell shape and/or aggregation. A: Wild-type; B, mnn9; C, ldb9; D, ldb11; E, ldb12; F, ldb13.

The ldb mutants, all of them bearing the ura3 marker, were transformed with the CEN BANK genomic library. The clones that recovered the wild-type phenotype were selected with the aid of QAE-Sephadex beads as described in Section 2.3. With this method, we succeeded in the selection of wild-type transformants in six cases: ldb2, ldb6, ldb11, ldb12, ldb13, and ldb15. The plasmids responsible for the complementation of the mutant phenotypes were recovered, amplified in E. coli, and the inserts were identified by sequencing a few hundred base pairs on each end. Fig. 3 shows the DNA fragments that cause complementation of all six mutations. As shown, each of the inserts includes several ORFs. To identify the particular gene that is responsible for complementation of the mutant phenotype, we followed the same protocol as in the identification of LDB1[12]. Although all the ldb mutants are non-conditional, the possibility exists that an ldb phenotype might be the result of a mutation that causes residual activity in an otherwise essential gene. However, it seems more likely that the mutations lie in non-essential genes, so that the non-essential genes were considered first. Taking advantage of the availability of the complete collection of S. cerevisiae deletion strains, we ran alcian blue dye-binding tests on all strains lacking each of the non-essential genes included in the fragments shown in Fig. 3, and, when indicated, native gel electrophoresis of secreted invertase. Then the phenotypes of the deletion mutants corresponding to the genes of a particular fragment were compared to the ldb mutant whose defect was complemented by that fragment. Finally, the deletion mutants that showed a similar phenotype to the corresponding ldb were crossed to it to check that they did not complement each other in the zygote. Additionally, the diploids were sporulated to prove that all spores of each tetrad showed the ldb phenotype.

Figure 3

Schematic representation of the inserts of YCp50 that cause complementation of several ldb phenotypes. A: ldb2; B, ldb 6; C, ldb11; D, ldb12; E, ldb13; F, ldb15. The essential genes are represented as white boxes. On each panel, the chromosome in which the fragment is included and the coordinates of both ends are indicated.

3.1.1 Identification of ldb2

As shown in Table 1, ldb2 exhibits a dye-binding value of 0, which indicates a total lack of affinity for alcian blue. Fig. 3A shows that there are five complete non-essential genes included in the insert that are candidates to complement the ldb2 defect: ADE8, SIZ1, STE14, YDR411C, and ERD1. The dye-binding test of the corresponding deletion strains showed that only ΔERD1 has an ldb phenotype identical to ldb2, while ΔADE8, ΔSIZ1, ΔSTE14, and ΔYDR411C stained blue as did the wild-type which was used as a control. ERD1 is involved in the retention of proteins in the endoplasmic reticulum as well as in Golgi-dependent modification of proteins [22]. Since the transfer of phosphate groups to N-linked oligosaccharides is a Golgi-dependent function, it seems that the ldb2 phenotype is just a consequence of the malfunction of the Golgi rather than a defect in a protein directly involved in MP transfer. To definitely prove that LDB2 and ERD1 are allelic, we did a standard test by crossing ldb2×ΔERD1. The zygotes showed an ldb phenotype and, when sporulated, all the spores in each tetrad also behaved as low-dye-binding. In consequence LDB2 must be considered to be an alias of ERD1.

3.1.2 Identification of ldb6

Fig. 3B shows that there are three complete non-essential genes in the insert that complement the ldb6 mutation. The low-dye-binding test of the corresponding deletion mutants showed that ΔMNN5 and ΔGON7 both have an ldb phenotype with values of 2–3, very similar to ldb6 (see Table 1). In addition, the electrophoresis of secreted invertase also showed an mnn2-type migration pattern in both cases (not shown). MNN5 is a well-known gene that encodes an α(1,2)-mannosyltransferase involved in the extension of branches of the outer chain on N-linked oligosaccharides [15]. In the initial isolation of the mutant mnn5[23] it was described as synthesizing smaller N-linked oligosaccharides. In consequence it seems reasonable that in the present work it also showed a reduction in the number of phosphate groups. GON7 is at this moment a reserved name for YJL184W, a hypothetical ORF of unknown function. The standard test of allelism revealed that the ΔMNN5 strain complemented the ldb6 phenotype but ΔGON7 did not. We therefore propose LDB6 as the standard name for YJL184W, or an alias of GON7 if this name is published first. These results suggest that YJL184C might be involved in synthesis of N-linked oligosaccharides whether directly or as a consequence of a more general defect in the Golgi function. The predicted Ldb6p has 123 amino acids, a molecular weight of 13 605 Da and has no leader peptide or transmembrane domains. The sequence is conserved in other yeasts (for additional information see the SGD web page at http://genome-www.stanford.edu/Saccharomyces/ or other S. cerevisiae databases).

3.1.3 Identification of ldb11

Fig. 3C shows that the genes STD1, COG1, and EDC1, located in chromosome VII, are the candidates to complement the ldb11 defect. The alcian blue test of the corresponding deleted strains revealed that only COG1 exhibits the ldb phenotype. In addition, native gel electrophoresis of the secreted invertase of ΔCOG1 also showed a heterogeneous migration pattern similar to that shown by the ldb11 strain (Fig. 1, lane 4). The lack of complementation in the zygote obtained by crossing ldb11×ΔCOG1 confirmed that LDB11 is allelic to COG1. COG1 is a member of a gene family involved in Golgi transport [24], so a transport defect in the Golgi apparatus seems to be the reason for the ldb11 phenotype. The certain degree of vanadate resistance shown by ldb11 (see Table 1) as well as the clumpy phenotype (Table 1 and Fig. 2D) is also shown by COG1.

3.1.4 Identification of ldb12

In this case OCH1, PNC1, YGL036W, MIG1, and YGL034C located in chromosome VII are the candidates to bear the ldb12 mutation (Fig. 3D). It should be noted that OCH1 is a well-defined gene encoding an α(1,6)-mannosyltransferase responsible for the initiation of the outer chain in N-linked oligosaccharides [25]. The migration pattern of external invertase synthesized by ΔOCH1 is almost indistinguishable from that of the mnn9[26] and ldb12 (Table 1) strains. These observations make OCH1 a good candidate to be allelic to LDB12. The dye-binding tests proved that ΔOCH1 showed a clear ldb phenotype while ΔPNC1, ΔYGL036W, ΔMIG1, and ΔYGL034C all stained as the wild-type. In addition, vanadate resistance (see Table 1), the clumpy phenotype (Table 1 and Fig. 2E), and the standard test of allelism confirmed that LDB12 and OCH1 are allelic.

3.1.5 Identification of ldb13

The mutant ldb13 also shows an mnn9-type invertase migration pattern and is able to grow in 10 and 15 mM sodium orthovanadate (see Table 1). The fragment that shows complementation of the ldb13 defect lies in chromosome XIII and includes the non-essential genes VAN1, DAT1, and CTK3 (Fig. 3E). The name VAN1 is a synonym of VRG7 and VRG8 which were so named because of the ability of the corresponding mutants to grow in vanadate (vanadate-resistant glycosylation mutants) [21]. In addition, van1 mutants also exhibit an invertase migration pattern of the mnn9 type. These data suggest that VAN1 could be the candidate to complement the ldb13 defect. The dye-binding test, the clumpy phenotype (Table 1 and Fig. 2F), and the standard test of allelism confirmed that LDB13 must be considered an alias of VAN1. Van1p is a protein located in cis-Golgi, and involved in elongation of N-linked oligosaccharides [27]. Van1p and Mnn9p form the mannan polymerase I complex responsible for the initial elongation of the outer chain in cis-Golgi [28], so the reason for the ldb phenotype shown by ldb13 seems to be just the reduction in size of the N-linked oligosaccharides.

3.1.6 Identification of ldb15

The ldb15 phenotype was complemented by a fragment of chromosome XVI including the following non-essential genes: PHO85, YPL030W, SUV3, and SMA1 (Fig. 3F). The dye-binding test of the corresponding deletion strains showed that only ΔPHO85 has a reduction in affinity for the alcian blue dye. We assigned a value of 1 (very light blue color) to ΔPHO85 which is identical to that shown in Table 1 for ldb15. The rest of the deletion strains –ΔYPL030W, ΔSUV3, and ΔSMA1– all stained like the wild-type. In addition, when we crossed ldb15×ΔPHO85 we obtained a zygote with the ldb phenotype which was unable to sporulate. Consequently, LDB15 must be considered an alias of PHO85. PHO85 encodes a multifunctional cyclin-dependent protein kinase involved in stress adaptation and cell integrity. The deletion of PHO85 causes a pleiotropic phenotype related to different aspects of metabolism, the cell cycle, cell polarity, and gene expression [29]. Our results add one more characteristic to this phenotype – the severe reduction in the incorporation of phosphate groups into the N-linked oligosaccharides. The way PHO85 participates in the transfer of MP remains obscure, although it has already been suggested that the transfer of phosphate into the N-linked oligosaccharides might be part of the general cell response to stress situations [4]. The involvement of PHO85 in the process might reinforce this suggestion. In addition, it is also possible that in pho85 mutants the cells behave as starved for phosphate [30] and for that reason they cannot phosphorylate the N-linked oligosaccharides. We must point out that ΔPHO85 shows an invertase migration pattern indistinguishable from that of the wild-type, which suggests that the defect in the PHO85 specifically affects the incorporation of the MP groups into the N-linked oligosaccharides.

3.2 Identification of LDB4, LDB5, LDB7, LDB9, LDB10 and LDB14

As shown in Section 2.2, when we transformed the mutants ldb4, ldb5, ldb7, ldb9, ldb10, and ldb14 with the CEN BANK library, we were not able to detect the clones that recovered the wild-type phenotype in two independent experiments. The reason for this result might be related to the amount of phosphate present in the mutants since none of them showed a too severe reduction in their affinity for alcian blue dye (see Table 1). This would make it more difficult for the QAE-Sephadex beads to distinguish between wild-type and mutant cells and consequently would reduce the probability of binding for the cells that have recovered the wild-type phenotype. However, before analyzing further the reason for such negative results, we took advantage of the situation that, simultaneously with the last part of this study, we were running a genome-wide search for ldb phenotypes in the complete collection of S. cerevisiae deletion strains. We found around 200 single deletion strains with different degrees of reduction in the affinity for alcian blue dye (I. Corbacho, I. Olivero and L.M. Hernández, in preparation).

Each of the aforementioned ldb mutants was crossed to several deletion strains of similar phenotype (alcian blue stain intensity, invertase migration pattern, cell morphology, and cell arrangements). The results of these standard tests of allelism were as follows: LDB4, allelic to SPC72; LDB5, allelic to RLR1; LDB7, allelic to YBL006C (hypothetical ORF); LDB9, allelic to ELM1; LDB10, allelic to CWH36; and LDB14, allelic to BUD32.

Spc72p is an essential component of the spindle body. Mutants in SCP72 have very few cytoplasmic microtubules [31], and the transit of glycoproteins through the Golgi might be affected as a consequence of the defect in the cytoskeleton. This could explain the finding that MP transfer into the N-linked oligosaccharides was slightly less effective in the ldb4 strain.

RLR1 encodes a protein with nucleic-acid-binding activity which is involved in RNA elongation from RNA polymerase II promoter. Mutants in RLR1 cause a wide range of growth defects and are unable to express LacZ fusions in yeasts [32]. The ldb phenotype might be the result of a general transcription defect that affects the glycosylation machinery and probably the gene(s) involved in the transfer of MP groups.

YBL006C is a hypothetical ORF of unknown function. The deletion mutant ΔYBL006C shows an invertase migration pattern that is very similar, if not identical, to the wild-type. The cell shape is regular and it shows neither clumpy phenotype nor vanadate resistance. We can now only state that the transfer of MP groups into the N-linked oligosaccharides is reduced in the mutant, so the gene must in some way be involved in the process. The predicted protein encoded by LDB7/YBL006C has 145 amino acids, a molecular mass of 15 869 Da and has no leader peptide or transmembrane domains. The sequence is conserved in other yeasts although it is remarkable that Ldb7p is 35–38 residues shorter than its orthologs in the closely related yeasts (for additional information see the SGD web page at http://genome-www.stanford.edu/Saccharomyces/ or other S. cerevisiae databases).

ELM1, CWH36 and BUD32 are genes involved in morphogenetic events affecting cell wall organization. Both ELM1 and BUD32 encode proteins with serine/threonine kinase activities implicated in bud growth and bud site selection. ELM1 is a particular type of protein kinase whose deletion causes pseudohyphal growth, suggesting that Elm1p acts as a repressor of mycelial growth under non-inducing conditions [33] (see Fig. 2C). BUD32 was detected in a genome-wide study for genes involved in bud site selection [34]. Deletion of BUD32 causes a pleiotropic phenotype including slow growth, alterations in cell wall structure, and random budding [34,35]. The ldb14 mutant (see Table 1) and ΔBUD32 (not shown) also show defects in the glycosylation machinery as deduced by the reduction in invertase size. The participation of protein kinases in several cell processes is a quite common feature of these enzymes, so it is not surprising that they also influence the mannosylphosphorylation of N-linked oligosaccharides. Finally, CWH36 was detected in a screen for S. cerevisiae mutants that showed calcofluor white hypersensitivity (CWH) because of alterations in cell wall organization and/or biosynthesis [36]. The mutant ldb10 (Table 1) and ΔCWH36 also synthesize external invertase of reduced size. Although the molecular function of CWH36 is not yet known, the reduced amount of N-linked carbohydrate in the mannoproteins could account for the ldb phenotype.

In this work we have identified 12 genes which when mutated/deleted result in an ldb phenotype (see Table 1). The functions of the genes are very diverse: secretion (ERD1, COG1), microtubule organization (SPC2), transcription (RLR1), cell wall organization (CWH36), protein glycosylation (OCH1, VAN1), and protein phosphorylation (ELM1, BUD32, PHO85). In addition, LDB6 (GON7/YJL184W) and LDB7 (YBL006C) are hypothetical ORFs of unknown function. We propose LDB6 and LDB7, respectively, as standard names for them. From the present data, it seems clear that the transfer of MP into the N-linked oligosaccharides of yeast mannoproteins depends on many genes and is linked to important cell processes. The new phenotype characteristics of the deletion strains that we report here may contribute to the functional characterization of the S. cerevisiae genome.


We thank Dr. M. Ramirez of this laboratory for his generous gift of the CEN BANK genomic library. I.C. is the recipient of a grant from the University of Extremadura. This work was supported by Grant 2PR01A099 from the Junta de Extremadura.


  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].
View Abstract