OUP user menu

Vertical and horizontal transmission of intestinal commensal bacteria in the rat model

Ryo Inoue, Kazunari Ushida
DOI: http://dx.doi.org/10.1016/S0168-6496(03)00215-0 213-219 First published online: 1 November 2003

Abstract

The intestinal microbiota of 10 pups (five from dam A and five from dam B) in the suckling stage (18 days old) and at maturity (40 days old) were compared with those of their dams to assess the mechanisms of bacterial transmission during development of intestinal microbiota in the rat model. Fecal samples were subjected to amplified ribosomal DNA restriction analysis and 65 operational taxonomic units (OTUs) were identified. The intestinal microbiota of mature pups were more complex than those of the suckling stage. Most of the OTUs present in dams were detected in their pups at maturity. These common OTUs accounted for more than 70% of the clones of libraries generated from both groups of pups. In contrast, the number of OTUs in pups that were not shared by their dams was larger than the number they had in common. These bacteria were presumably transmitted horizontally from environmental sources. However, these OTUs accounted for less than 30% of the clones generated from both groups of pups. This study suggested that vertically transmitted bacteria were the predominant component of the intestinal microbiota of pups, although the diversity of intestinal microbiota during pup growth was influenced by horizontal transmission.

Keywords
  • Rat
  • Intestinal microbiota development
  • Vertical transmission
  • Horizontal transmission
  • Amplified ribosomal DNA restriction analysis

1 Introduction

The microbial community inhabiting the intestinal tract of animals is characterized by high population density, wide diversity, and complexity of interactions. The adult human colon has been estimated to contain more than 1011 bacterial cells g−1 belonging to as many as 400 species [1,2]. The source of these intestinal microbes has continued to be of interest since the first study on the subject by Tissier [3]. It is usually assumed that a neonate is exposed to bacteria from the vagina and external genitalia of its mother and from other environmental sources with which it comes into contact at birth. The former is categorized as vertical transmission and the latter as horizontal transmission. Intestinal microbiota are, therefore, developed from these two sources of bacteria [2] under the influence of external and host factors. However, the assumption lacks scientific evidence because discriminatory tests that permit comparison of bacterial strains from maternal and infant sources have not previously been performed.

These intestinal commensal microbiota interact with the host, eliciting a wide range of physiological host reactions. Colonization by such microbiota makes it possible for a range of intestinal functions to develop, as evidenced by the maldevelopment of intestinal functions in germ-free animals. Germ-free rodents require a higher calorific intake to maintain their weight than those with intestinal microbiota [4]. Germ-free mice also show lower expression of mRNA than conventionally reared mice in the ileal crypt epithelium for colipase, which plays a critical role in lipid metabolism [5]. Moreover, suboptimal stimulation by environmental microbes of mucosal immune development, due to sanitary conditions, may contribute to increased frequency of autoimmune diseases, such as food allergies and atopic diseases, in industrialized countries [6].

Since colonization of intestinal commensal microbiota interacts strongly with host physiological functions, bacteriological studies need to be conducted to reveal when, which, and how many commensal bacteria are transferred vertically and/or horizontally in connection with the development of the host physiology. The information is required to modulate the intestinal microbiota to improve the nutritional efficiency and prevent autoimmune diseases in the host.

Numerous studies have been conducted on microbial invasion and bacterial translocation in the intestine at birth. The fluctuation of the fecal bacterial population of neonates has been well established, especially in humans [79]. However, very few studies about the source of these microbiota have been performed [10]. Consequently, so far there has been no definitive explanation of bacterial transmissions from a mother and/or any other environmental sources to a neonate intestine. This is due to the use of traditional bacteriological techniques, in particular, cultivation on selected media. Traditional cultivation has led essentially to narrow specific studies of a few cultivable bacterial groups, such as Enterobacteriaceae and Bifidobacterium, and more time and resources need to be invested to differentiate the microbial community at the subspecies level. Moreover, specific genera of bacteria, such as Bifidobacterium and Lactobacillus, tend to be overestimated by this method [11], and up to 85% of the entire microbial population in the human intestine might be uncultivable by conventional cultivation techniques, according to recent estimates [11,12]. These drawbacks demand the introduction of molecular-based techniques that can solve these technical limitations. In this study, fecal samples of dams and pups in the suckling stage and at maturity were analyzed by using a molecular fingerprinting technique, amplified ribosomal DNA restriction analysis (ARDRA), to assess the mechanisms of bacterial transmission during the development of intestinal microbiota. ARDRA has been widely used, for example, to evaluate generic diversity within fungal and bacterial species and to identify particular races and pathotypes [1315]. This method consists of polymerase chain reaction (PCR) amplification of the 16S rRNA gene fragments and subsequent restriction digestion of the amplicons to yield specific restriction patterns that define operational taxonomic units (OTUs), which enable the differentiation of most species and possibly strains. This method is considered as potent as other molecular fingerprinting techniques such as arbitrarily primed PCR (AP-PCR) and amplified fragment length polymorphism (AFLP) [16]. ARDRA permits comparisons of the compositions of intestinal microbiota between dams and pups by subspecies and/or strain level in this study because an OTU that shows a unique ARDRA pattern represents a unique subspecies and/or strain [16,17].

2 Materials and methods

2.1 Animals

Two pregnant standard (conventional) Wistar rats (dams A and B) bred under barrier circumstances (SPF circumstances) were obtained from a commercial supplier (Japan SLC, Shizuoka, Japan). Rats were housed under non-barrier systems in our animal facility and fed a standard pelleted chow (Labo MR stock, Nihon Nosan Kogyo, Tokyo, Japan). Littermates were not separated until weaning (21st day after birth), and were offered the same diet as dams after weaning.

2.2 DNA extraction

Five pups were randomly selected from each litter (pups A (A1–A5) from dam A and pups B (B1–B5) from dam B), and fresh rectal feces was aseptically collected into sterile microfuge tubes 18 and 40 days after birth. The diversity of intestinal microbiota in rats began to increase from the early stage of weaning (18 days old) and was already stabilized in rats aged 40 days [18]. Fresh rectal feces of dams was also collected when their pups reached 40 days of age. Extraction of whole bacterial DNA from feces was performed as described by Godon et al. [19]

2.3 Amplification of 16S rRNA genes (16S rDNA)

Primers 907f (5′-AAA CTC AAA TGA ATT GAC GGG-3′) [20] and L1401 (5′-GCG TGT GTA CAA GAC CC-3′) [21] were used to amplify the V6–V8 regions of the bacterial 16S rDNA. PCR reactions were performed with rTaq DNA polymerase (Toyobo, Osaka, Japan). Each 50 μl PCR mixture contained 50 mM KCl, 10 mM Tris–HCl (pH 8.3), 0.1% Triton X-100, 1.5 mM MgCl2, 0.16 mM each deoxynucleoside triphosphate (dNTP), 5% acetamide, 1.25 U rTaq DNA polymerase, 0.4 μM of each primer, and 1 μl extracted bacterial DNA. The samples were amplified in a Gene Amp System 2400 Thermal Cycler (Perkin-Elmer, Branchburg, NJ, USA) by using the following program: initial denaturation at 95°C for 3 min; 12 thermal cycles of 94°C for 30 s, 56°C for 30 s, and 72°C for 30 s; and final elongation at 72°C for 3 min.

2.4 Cloning of PCR products

PCR products from the amplification reactions were purified and concentrated with a QIAquick gel extraction kit (Qiagen, Tokyo, Japan) and then cloned into a pGEM-T vector plasmid (Promega, Tokyo, Japan). Ligation was done at 4°C overnight followed by transformation into Escherichia coli JM109-competent cells (TaKaRa, Kyoto, Japan) according to the manufacturer's instructions. The clones were screened for α-complementation of β-galactosidase using X-Gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside)–IPTG (isopropyl-β-D-thiogalactopyranoside) indicator Luria–Bertani agar plates supplemented with 100 μg ml−1 ampicillin. Clone libraries were constructed for the two dams and each of their pups, aged 18 and 40 days, respectively.

2.5 ARDRA screening of clones

Plasmids of positive clones were reamplified by colony PCR. Primers T7 (5′-GTA ATA CGA CTC ACT ATA GGG C-3′) and SP6 (5′-ATT TAG GTG ACA CTA TAG-3′) were used to amplify the insert region of the plasmid vector. PCRs were performed with rTaq DNA polymerase (Toyobo). Each 24 μl PCR mixture contained 50 mM KCl, 10 mM Tris–HCl (pH 8.3), 0.1% Triton X-100, 1.5 mM MgCl2, 0.16 mM each dNTP, 0.63 U rTaq DNA polymerase, 0.4 μM of each primer and a toothpick full of colonies, which were added as the template. The samples were amplified in a Gene Amp System 2400 Thermal Cycler (Perkin-Elmer) using the following program: initial denaturation at 95°C for 3 min; 25 thermal cycles of 94°C for 30 s, 48°C for 30 s, and 72°C for 30 s; and final elongation at 72°C for 3 min. The size and amounts of PCR products were confirmed by analyzing 5 μl samples using 1% agarose (w/v) gel electrophoresis in a 1× Tris–(hydroxymethyl)aminomethane-acetate–EDTA pH 8.0 buffer (TAE) followed by ethidium bromide staining and visualization with UV light.

PCR products were digested to completion with the enzymes HaeIII, HhaI, and Sau3AI (Toyobo). Digestion was carried out overnight at 37°C in 10 μl commercially supplied incubation buffer containing 2 U of restriction enzymes, respectively. Restriction fragments were analyzed by electrophoresis on a 2.2% agarose gel (w/v) in a 1× Tris–(hydroxymethyl)aminomethane-borate–EDTA pH 8.0 buffer (TBE) followed by ethidium bromide staining and visualization with UV light. Gels were photographed and restriction patterns were compared visually. An OTU was defined as a unique combination of the restriction patterns obtained with respective enzymes, interpreted according to the scheme of Vannechoutte et al. [22]

2.6 Sequence analysis

The plasmid DNA of clones representing each OTU was prepared by using a plasmid Mini-Prep Kit (Bio-Rad, Tokyo, Japan). The cloned DNA fragments were sequenced by Shimadzu Genomic Research Laboratory (Shimadzu, Kyoto, Japan). Homology searches of the GenBank DNA database for these sequences were performed with BLAST N. In this study, three to five E. coli clones were randomly selected and sequenced in the case of the major OTUs to which belonged more than 50 clones such as OTU-1, -2, -10, -39, -44 and -57. Since the sequence data of these multiple clones were very similar and had closest homology to the 16S rRNA gene of the same bacterium in the database, the sequencing of minor OTUs was performed on one representative clone selected randomly.

Chemicals were obtained from Nacalai Tesque (Kyoto, Japan) and Wako Pure Chemicals (Osaka, Japan) unless otherwise stated. Animals were handled in accordance with the guidelines for research with laboratory animals of the Kyoto Prefectural University Experimental Animal Committee.

3 Results

3.1 ARDRA and sequence analysis

A total of 22 clone libraries of 16S rDNA were constructed, two for each dam, one for each of 10 pups A and B in the suckling stage (18 days after birth), and one for each 10 for pups A and B at maturity (40 days after birth). A total of 2102 clones were analyzed and five restriction patterns were obtained from HaeIII and Sau3AI digestion, respectively, and six restriction patterns were obtained from HhaI digestion. Sixty-five OTUs were identified by ARDRA and the results from the BLAST search on the representative clone of each OTU are shown in Table 1. The nearest known species are given for the sequences showing less than 90% similarity to the database sequences.

View this table:
1

Homology search results from the BLAST search on representative clones of each OTU

OTURelated known speciesGenBankIdentity (%)
1Escherichia coliE0513395
2Escherichia coliA14565100
3Escherichia coliZ8320596
4Escherichia coliZ8320599
5Escherichia coliA1456598
6Escherichia coliZ8320595
7Bacteroides eggerthiiAB05010781
8Bacteroides forsythusAB05394288
9Bacteroides acidofaciensAB02115789
10Bacteroides vulgatusAB05011193
11Bacteroides distasonisM8669591
12Bacteroides eggerthiiAB05010786
13Bacteroides sp.X8921792
14Bacteroides eggerthiiAB05010784
15Bacteroides distasonisM8669582
16Bacteroides forsythusAB05394290
17Bacteroides sp.X8921796
18Bacteroides sp.X8921795
19Bacteroides sp. AR20AF13952499
20Bacteroides forsythusAB05393882
21Prevotella brevisAJ01168279
22Prevotella ruminicola strain 223/M2/7AF21861889
23Prevotella buccae ATCC 33690L1647890
24Prevotella brevisAJ01168290
25Prevotella sp. oral clone CY006AY00506391
26Prevotella sp.AB00338691
27Prevotella brevisAJ01168290
28Prevotella sp. oral clone CY006AY00506391
29Prevotella heparinolytica ATCC 35895L1648792
30Prevotella dentalisX8187689
31Prevotella brevisAJ01168290
32Prevotella brevisAJ01168289
33Prevotella brevisAJ01168290
34Prevotella brevisAJ01168289
35Ruminococcus hydrogenotrophicusX9562492
36Ruminococcus hydrogenotrophicusX9562491
37Ruminococcus torquesL7660490
38Ruminococcus torquesL7660492
39Ruminococcus torquesL7660490
40Ruminococcus hydrogenotrophicusX9562491
41Ruminococcus hydrogenotrophicusX9562491
42Ruminococcus torquesL7660492
43Enterococcus faeciumAF07022396
44Clostridium sp.AF15705385
45Clostridium clostridiiformesM5908986
46Clostridium orbiscindensY1818794
47Clostridium sp. ASF502AF15705392
48Clostridium sp. DSM 6877 (FS 41)X7674789
49Clostridium saccharolyticumY1818585
50Clostridium leptumAJ30523889
51Clostridium leptumAF26223992
52Clostridium orbiscindensY1818793
53Clostridium orbiscindensY1818795
54Lactobacillus crispatus strain BLB2AF24314190
55Lactobacillus gasseri strain KC5aAF24316597
56Lactobacillus gasseri strain KC5aAF24316596
57Lactobacillus murinusAF15704995
58Lactobacillus murinusAF15704996
59Lacotobacillus reuteri (DSM 20016 T)X7632898
60Lacotobacillus reuteri (DSM 20017 T)X7632895
61Lacotobacillus reuteri (DSM 20018 T)X7632896
62Corynebacterium mastitidisY0980695
63Eubacterium lentumAB01181789
64Eubacterium desmolansL3461889
65Actinomyces viscosusAJ23405395
  • Identity (%) of OTU in which multiple clones were sequenced (OTU-1, -2, -10, -39, -43, -44 and -57) is that of the clone having shown the highest identity.

  • OTUs whose similarities were less than 90% were still unknown taxa at species level, therefore related species are presented for convenience.

Twenty-nine OTUs were identified from the 16S rDNA clonal library of dam A and 25 OTUs from dam B (Figs. 1 and 2). Approximately one-third of the 16S rDNA clones generated from both dams A and B belonged to either OTU-10, -52 or -57. OTU-10 accounted for 11.9% and 21.3% of the 16S rDNA clones of dams A (total 110 clones) and B (total 80 clones), respectively. OTU-52 accounted for 12.7% and 12.5%, and OTU -57 for 9.1% and 10%, respectively. OTU-10, -52 and -57 showed high sequence similarities to the 16S rRNA genes of Bacteroides vulgatus (GenBank: AB050111, identity: 93%), Clostridium orbiscindens (Y18187, 93%), and Lactobacillus murinus (AF157049, 95%), respectively. OTU-7 and -44, respectively, accounted for 6.4% and 10.9% of the 16S rDNA clones of dam A, but neither OTU was detected in the library of dam B. Approximately one-tenth of the 16S rDNA clones generated from dam B belonged to OTU-22, while only 3.6% of the clones belonged to this OTU in the case of dam A. OTU-7, -22 and -44, respectively, showed sequence similarities of less than 90% to the 16S rRNA genes of Bacteroides eggerthii (AB050107, 81%), Prevotella ruminicola strain 223/M2/7 (AF218618, 89%), and Clostridium sp. (AF157053, 85%), respectively. Other OTUs accounted for less than 5% of the total clones of both dams A and B.

1

Distribution of OTUs identified from the clonal libraries of dam A (a), pups A in the suckling stage (18 days old) (b) and at maturity (40 days old) (c). Each OTU was defined by a unique ARDRA pattern. Filled bars: OTUs common to dam and pups. Open bars: OTUs detected from either dam or pups. Data of pups are shown by mean value for five pups. Homology search results for each OTU are presented in Table 1.

2

Distribution of OTUs identified from the clonal libraries of dam B (a) and pups B in the suckling stage (18 days old) (b) and at maturity (40 days old) (c). Each OTU was defined by a unique ARDRA pattern. Filled bars: OTUs common to dam and pups. Open bars: OTUs detected from either dam or pups. Data of pups are shown by mean value for five pups. Homology search results for each OTU are presented in Table 1.

A total of 23 and 34 OTUs were identified from all 10 16S rDNA clonal libraries generated from pups A and B, respectively, in the suckling stage. The mean value for five animals was 10.2 and 16.4 OTUs (Figs. 1 and 2). Fifteen OTUs were common to dam A and pups A in the suckling stage, and 17 OTUs were common between dam B and pups B in the suckling stage. OTU-1 and -2 were predominant in the 16S rDNA clonal libraries of both pups A and B in the suckling stage. OTU-1 accounted for 45.2% and 33.4% (mean value for five animals) of the 16S rDNA clones from pups A and B, respectively. OTU-2 accounted for 34.5% and 15.1% of the 16S rDNA clones from pups A and B, respectively. OTU-57 was detected next to OTU-1 and -2, which accounted for 7.8% and 8.3% of clones, respectively, for pups A and B (mean value for five animals) in the suckling stage. OTU-1 and -2 both showed high sequence similarities to the 16S rRNA genes of E. coli (E05133, 95% and A14565, 100%, respectively) (Table 1).

Fifty-four OTUs were identified from all five 16S rDNA clonal libraries of pups A at maturity (40 days old) (mean value 29.8 OTUs) (Fig. 1). Twenty-five OTUs were common with those of their dam (dam A), but 29 OTUs were not detected in their dam. Each one of all OTUs except 4 (OTU-10, -37, -38, and -57) accounted for less than 5% of the 16S rDNA clones. OTU-10, -37, -38, and -57 accounted for 7.5%, 13.6%, 5.3%, and 6.1% of the clones, respectively. Both OTU-37 and -38 showed more than 90% sequence similarity to the 16S rRNA gene of Ruminococcus torques (L7664, 92% and 90%, respectively) (Table 1).

Forty-seven OTUs were identified from all five 16S rDNA clonal libraries of pups B at maturity (40 days old) (mean value 23 OTUs) (Fig. 2). Twenty-one OTUs were common to those of their dam (dam B), but 26 OTUs were not detected in their dam. Each one of seven OTUs (OTU-10, -35, -37, -41, -43, -47, and -57) accounted for more than 5% of the clones. OTU-10, -35, -37, -41, -43, -47, and -57 accounted for 7.7%, 6.1%, 6.8%, 5.3%, 10.4%, 6.8%, and 7% of the clones, respectively. Both OTU-35 and -41 showed more than 90% sequence similarity to the 16S rRNA gene of Ruminococcus hydrogenotrophicus (X95624, 92%, and 91%, respectively). OTU-43 showed 96% similarity to Enterococcus faecium (AF070223) and OTU-47 showed 92% similarity to Clostridium sp. ASF502 (AF157053) (Table 1).

4 Discussion

In this study, ARDRA was applied to analyze intestinal microbiota of the dams and their pups. This technique has been widely used to compare the microbial community by defining OTUs [15,23,24] as it can detect uncultivable bacteria and differentiate most species and possibly strains with less time and resources. However, due to some methodological limitations, such as PCR bias and cloning bias, minor populations in the microbial community may not be detected [25]. Therefore, we primarily considered that the OTUs detected only from pups that were horizontally transmitted from the surrounding environment, as discussed below, but this does not completely exclude the possibility of the presence of these OTUs as the minor undetectable population in dams.

Intestinal microbiota were simple shortly after birth and became complex from the suckling stage (18 days old) to maturity (40 days old), as demonstrated in a previous study [18]. OTU-1 and -2 represented the predominant component in the intestinal microbiota of pups A and B in the suckling stage (Figs. 1 and 2), and the sequences of both OTUs showed more than 95% similarity to that of E. coli (Table 1). This result confirms the study of Yajima et al. [26], which showed that Enterobacteriaceae were the predominant bacteria, representing 45% of the total bacterial count (CFU) in the intestine of 14-day-old rats. OTU-1 and -2 accounted for approximately 80% of the total clones in pups A but approximately 49% of the total clones in pups B (Figs. 1 and 2). The numbers of these bacteria in their dam may explain the difference between pups A and B. Indeed, OTU-1 and -2 accounted for approximately 8.2% of the total clones of dam A, but were undetectable in dam B (Figs. 1 and 2). E. coli may have become predominant in the intestinal microbiota during the suckling stage, but its level of predominance apparently depends on how many of these bacteria are in the intestine of their dam.

Most of the OTUs in dams, 25 of 29 and 21 of 25, respectively, for dams A and B, were detected in their pups at maturity and belonged to various genera, such as Bacteroides, Clostridium, and Lactobacillus. Only a small number of the OTUs detected in dams were undetectable in their pups. OTUs common to dams and their pups accounted for 71.1% and 76.9% of the total E. coli clones generated from pups A and B, respectively. This suggests that the bacteria transmitted vertically from the dams to their pups become predominant components in the pups’ intestines. The limitation of methodology as pointed out above should also be considered when the implications of OTUs commonly shared by the dams and their pups are discussed. It is less plausible that these common OTUs were transmitted horizontally from the surrounding environment both to the dams and their pups. Since the intestinal microbiota of the dams had been constructed in the breeder's facility before the experiment, horizontal transmission of bacteria to the completed intestinal microbiota of the matured animal would be difficult [18]. In contrast to these commonly shared OTUs, the remaining OTUs in pups, 29 of 54 and 26 of 47, respectively, for pups A and B, were not in common with those of their dams. These OTUs also belonged to various genera, such as Bacteroides, Lactobacillus, Clostridium, and Eubacterium. These bacteria were presumably transmitted horizontally from environmental sources. Horizontal transmission has been suggested in humans in whom several strains of Bifidobacterium were uniformly detected from the feces of human infants born in the same hospital [27]. These authors suggested the presence of a site-specific (i.e. hospital-specific) strain of Bifidobacterium and a site-specific horizontal transmission of the strains.

The transmission of these bacteria from environmental sources may occur from the early weaning stage to 27 days of age in rats. Evidence for this may be that the amount of luminal secretory immunoglobulin A (SIgA), which mediates a primary defense line to exclude foreign antigens, decreased rapidly from the early weaning period (18 days old) and showed a lag period until recovery (27 days old) [18]. This lag period, which is required for the pups to develop their own SIgA secretion, allows transmission and colonization of various bacteria, including those of non-maternal origin. In fact, the diversity of the intestinal microbiota in rats increased from 18 days of age, and the increase ceased at 27 days of age.

More OTUs seem to be transmitted horizontally than vertically in our rat model. Diversification of intestinal microbiota is actually expressed by the number of OTUs, more than half of which were due to horizontally transmitted bacteria. Interestingly, the bacteria that were transmitted horizontally into pups did not become the predominant components in the intestinal microbiota because OTUs in pups that were not shared by their dams accounted for less than 30% of the clones generated from pups of both groups. Dietary condition could have an impact on the bacterial transmission. This may be considered in the human case, where various feeds are offered as baby food, although 5% (w/w) fructo-oligosaccharide supplementation to standard pelleted chow had little effect on the ratio of vertical and horizontal transmissions in our preliminary study using the rat model (data not shown). This point needs further elucidation.

The relatively low diversity of intestinal microbiota of both dams compared to that of their pups at maturity suggested that, once the intestinal microbiota were established and the level of luminal SIgA was recovered, intensive horizontal transmission of bacteria from the environment seems not to be allowed. The artificially and strictly controlled environment in which both of these dams grew explains the poor diversity of their intestinal microbiota. As commercially available experimental animals, they were born and reared in barrier systems [28] that limit the horizontal transmission of bacteria. When they were introduced into the non-barrier systems employed in this study, horizontal transmission had already been prevented since their immune systems had already developed, in particular SIgA production, and competition with indigenous bacteria was also limited in the intestinal tract.

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]
View Abstract