OUP user menu

Spatial variation in Bacillus thuringiensis/cereus populations within the phyllosphere of broad-leaved dock (Rumex obtusifolius) and surrounding habitats

Fay A. Collier, Sam L. Elliot, Richard J. Ellis
DOI: http://dx.doi.org/10.1016/j.femsec.2005.05.005 417-425 First published online: 1 November 2005

Abstract

The aim of this study was to determine the prevalence and toxin gene diversity of Bacillus thuringiensis/B. cereus in the phyllosphere of broad-leaved dock (Rumex obtusifolius) at a small spatial scale. B. thuringiensis/cereus populations were isolated from the phyllosphere of dock and neighbouring grass and in neighbouring soil using commercially available selective media which avoided the disadvantageous heat-shock selection procedure. The maximum density of B. thuringiensis/cereus in the dock phyllosphere was 1.9 × 104 CFU g−1 but the between-leaf variation in numbers was found to follow a lognormal distribution. B. thuringiensis/cereus was also found at significant densities in soil and the phyllosphere of grass adjacent to the dock plants. PCR screening indicated that genes encoding cry1 toxin were present in the plasmids of 36.9% of B. thuringiensis/cereus isolates tested, 11.9% contained cry2, and none of the dock leaf isolates tested contained cry3, cry4, cry7 or cry8 genes. The diversity of cry genes is similar to that found from other studies focused on other parts of the world. This work is the first concerning the prevalence of B. thuringiensis/cereus on leaves in the UK, finding population sizes of previously unrecorded levels and a greater relative proportion of B. thuringiensis. We have also illustrated that before any ecological function can be investigated, suitable sampling scales need to be considered – here we have shown that the minimum sampling unit should be individual leaves, to account for the log-normal distribution.

Keywords
  • Bacillus thuringiensis
  • cry Genes
  • Spatial variation
  • Lognormal distribution

1 Introduction

Bacillus thuringiensis is a spore forming, Gram-positive bacterium that produces a range of toxins, most noticeably a parasporal crystal inclusion or δ-endotoxin encoded by cry genes. There are many cry genes, each encoding a toxin that is active against a specific group of invertebrates (predominantly insects). This insecticidal activity has led to B. thuringiensis being used as a biological control agent whilst the cry genes have been transgenically implanted into several crop species. The parasporal crystal and cry genes have been the focus of many studies due to their commercial uses but this has left the ecology of B. thuringiensis relatively unexplored.

B. thuringiensis and the closely related (but non-entomopathogenic) B. cereus are frequently isolated from a wide variety of habitats such as soil, water, plant surfaces and stored food products [1,2]. However, the soil and phyllosphere are of particular interest as these appear to be the natural habitats of the organisms and there is some work to suggest that different populations occupy these two environments [3], despite their close proximity. Smith and Couche [4] suggested that B. thuringiensis may be present in the phyllosphere in order to protect the plant from insect herbivores. Elliot et al. [5] expanded on this “bodyguard hypothesis”, suggesting plants encourage the growth of entomopathogenic micro-organisms to protect the leaf surface from herbivores. The distribution of B. thuringiensis in the phyllosphere has been the subject of several studies [3,4,610] each of which has focused on different plant species in different locations and using a fairly rough and ready sampling strategy. The possibility of a link between the presence of insect herbivores and pathogenic B. thuringiensis has yet to be resolved, but to investigate ecological questions such as this, the distribution of B. thuringiensis/cereus isolates in natural habitats needs to be focused at the appropriate spatial scale and with appropriate isolation techniques.

The previous studies have all relied upon pasteurization of samples to provide selectivity for Bacillus species, but this technique will effectively remove any active, vegetative cells from the population. This may adversely affect the size and diversity of the population recovered and so it is important to improve the recovery methods to achieve a greater insight into the ecology of these organisms [11]. Furthermore, it is important to ascertain the importance of immigration, emigration, growth and death processes, in order to understand the dynamics of bacterial populations on the phylloplane [12]. We have used a commercially available selective media to remove the need for heat treatment, thereby maximizing the number of isolates that can be recovered from environmental samples. To determine whether populations from different plants or sample types differed in terms of their potential entomopathogenic specificity, PCR methods were used to assess the distribution of some major classes of δ-endotoxin genes.

2 Materials and methods

2.1 Isolation and storage of B. thuringiensis/cereus

In this study, the B. thuringiensis/cereus population densities on leaves of broad-leaved dock (R. obtusifolius) plants were determined. The site, which had a large natural population of broad-leaved dock (R. obtusifolius), was part of the walled garden in Silwood Park, Berkshire, UK (OS map ref: 175 944684) and there is no history of B. thuringiensis application on or near the site. Dock plants covered approximately 75% of the study site. Five randomly placed 1 m2 quadrats were marked out within the study site. Within each quadrat, five dock plants were randomly selected. The quadrats were labelled A–E and the plants 1–5.

From each dock plant an immature leaf, a mature leaf, a senescent leaf, and a grass and soil sample from next to the plant were collected on 3rd September 2002. The dock and grass samples were placed in individual plastic bags and the soil was collected in 30 ml universal tubes. The mass of each sample was recorded.

Samples were placed into a 30 ml universal tube with 5 ml 0.85% sterile saline and 0.5 g sterile sand and then vortexed for 30 s. Duplicate 100 μl aliquots of the homogenised sample were pipetted and then spread out on two plates of B. cereus Selective Agar (BcSA; Oxoid, UK) supplemented with egg yolk to aid identification. Selection is provided by the use of polymyxin B (100 IU ml−1). The plates were incubated overnight at 30°C and the number of colonies counted. Colonies of B. cereus and B. thuringiensis are crenated, about 5 mm in diameter and have a distinctive turquoise to peacock blue colour surrounded by egg yolk precipitate of the same colour. The colony morphologies of B. thuringiensis and B. cereus are identical, but can be distinguished from other Bacillus isolates.

Up to 54 isolates from each sampled dock leaf (depending on the number of colonies grown) and up to 10 isolates from each grass and soil sample were re-streaked to ensure purity and then stored in glycerol at −80°C for further analysis.

2.2 DNA extraction and molecular analysis of B. thuringiensis/cereus isolates

Total DNA was extracted from overnight broth cultures of 19 isolates from sample A4 using the CTAB method [13]. To ensure the selectivity of the isolation methods 16S rRNA gene fragments were amplified using the universal primers 63F (CAG GCC TAA CAC ATG CAA GTC) and 1387R (GGG CGG WGT GTA CAA GGC) [14]. These fragments were purified using QIAquick PCR purification kits (QIAGEN Ltd, West Sussex, UK) and subsequently sequenced using Big Dye reaction chemistry (Applied Biosystems, CA, USA) and the 63F primer. Reactions were run in-house on an ABI 3700 Analyzer (Applied Biosystems, CA, USA). These sequences were deposited in GenBank with accession numbers DQ058151DQ058170 and aligned with randomly selected Bacillus sequences from the Ribosomal Database Project (RDP) [15] using the ClustalW function of the BioEdit package [16]. Neighbor-joining phylogenetic trees were constructed using Molecular Evolutionary Genetics Analysis package (MEGA v3) [17] with the Kimura 2-parameter algorithm and the robustness of the phylogeny was tested by bootstrap analysis using 1000 iterations.

Plasmid DNA was extracted from overnight cultures of each of the B. thuringiensis/cereus isolates from dock leaves, grass and soil samples of or adjacent to three randomly selected plants A4, C3 and E3 using plasmid purification reagents from QIAGEN Ltd (West Sussex, UK) to lyse and neutralize the cells, followed by isopropanol precipitation of the lysate. Each plasmid extract was screened for the presence of cry1, cry 2, cry3, cry4, cry7 and cry8 using PCR methods (with primers listed in Table 1) to give some indication of which cry genes are present in B. thuringiensis/cereus populations from the phyllosphere of broad-leaved dock and adjacent habitats. PCR reactions were performed as follows: 50 ng target plasmid DNA, 1 × Immobuffer (MgCl2 free; Bioline Ltd, London, UK), 2 mM MgCl2, 0.25 μM each primer as required (see Table 1), 200 μM each dNTP and 1 unit Immolase DNA polymerase (Bioline Ltd, London, UK). The following cycle conditions were used for all primer pairs: 95°C for 8 min (for enzyme activation and target denaturation); followed by 30 cycles of 95°C for 1 min, 52°C for 45 s, 72°C for 1 min; and a final extension step of 72°C for 5 min. Products were visualized by agarose gel electrophoresis using standard procedures [13].

View this table:
1

Oligonucleotide primers used for PCR amplification of cry gene families [27]

Target genePrimer nameSequence (5′–3′)
cry1Un-1(d)CATGATTCATGCGGCAGATAAAC
Un-1(r)TTGTGACACTTCTGCTTCCCATT
cry2Un-2(d)GTTATTCTTAATGCAGATGAATGGG
Un-2(r)CGGATAAAATAATCTGGGAAATAGT
cry3Un-3(d)CGTTATCGCAGAGAGATGACATTAA
Un-3(r)CATCTGTTGTTTCTGGAGGCAAT
cry4Un-4(d)GCATATGATGTAGCGAAACAAGCC
Un-4(r)GCGTGACATACCCATTTCCAGGTCC
cry7 and cry8Un-7,8(d)AAGCAGTGAATGCCTTGTTTAC
Un-7,8(r)CTTCTAAACCTTGACTACTT

Plasmid DNA was also extracted using the method above from B. thuringiensis var. kurstaki isolated from the commercially available insecticide formulation, DiPel WP (Valent Biosciences, CA, USA). This plasmid DNA, together with extracts from seven isolates from scenescent leaves of plant A4, were subjected to restriction digestion using both Eco RI and Pst I (Promega, Southampton, UK) according to the manufacturer's instructions. DNA fragments were separated on a 0.6% agarose gel run at 20 V overnight and visualized by ethidium bromide staining and UV illumination [13].

2.3 Data analysis

All bacterial counts were expressed as CFU g−1. Data analysis was performed using Microsoft Excel with Analyse-It add-in software (Analyse-It Software Ltd, Leeds, UK) as required. The Shapiro–Wilk test of normal distribution was performed on both raw and log-transformed count data. Two-way analysis of variance was undertaken to determine significant differences between sample types and plots.

3 Results

3.1 Occurrence of B. thuringiensis/cereus in the dock phyllosphere, on associated grasses and in the surrounding soil

The use of commercially available B. cereus selective agar (BcSA) provided a good selective medium for the isolation and tentative identification of B. cereus group bacteria, including B. thuringiensis. Isolates with the correct colony morphology were detected in 117 of 125 samples (93.6%) taken in September 2002. Shapiro–Wilk normality tests indicated that the B. thuringiensis/cereus populations on leaves could be approximated to the lognormal distribution as the coefficient for untransformed data was low whilst for log10 (x+ 1) transformed data it was approaching unity, which indicates a near-perfect fit to normality (Table 2). Therefore all further analysis was conducted using log transformed data.

View this table:
2

Results from Shapiro–Wilk test of normality for B. thuringiensis/cereus distributions on dock leaves

QuadratNumber of leavesRawalog10b
WcPWP
A150.574<0.00010.9210.198
B150.549<0.00010.9440.439
C150.7550.00100.9720.887
D150.6710.00010.9580.661
E150.460<0.00010.9590.676
  • a Analysis of untransformed plate count data.

  • b Analysis of log10 (x+ 1) transformed plate count data.

  • c Shapiro–Wilk statistic.

The theoretical limit of detection was approximately 5 CFU g−1 on leaves and 50 CFU g−1 soil. However, the B. thuringiensis/cereus densities were in excess of 104 CFU g−1 in all soil samples. Actual numbers in soil could not be calculated as samples were not dilute enough to permit accurate counting of colonies, so conservative estimates of 600 colonies per plate were used. B. thuringiensis/cereus counts were above detectable limits in 68 of 75 (91%) dock leaf samples and 24 of 25 (96%) grass samples. However, the average density on the dock leaves was 1.88 log10 CFU g−1 (n= 75; SEM = 0.08 log10 CFU g−1. The average density of B. thuringiensis/cereus isolates from each grass sample was 2.69 log10 CFU g−1 (n= 25, SEM = 0.14 log10 CFU g−1. A summary for the data for each quadrat is shown in Fig. 1. A total of 850 isolates were stored in the freezer (58% of which were isolates from dock leaves).

1

Log10 transformed density of B. thuringiensis/cereus in the phyllosphere of dock leaves of different ages, in the phyllosphere of neighbouring grass, and in neighbouring soil (CFU g−1 collected on 3rd September 2002 for each of 5 independent quadrats (A–E). Bars indicate standard error (n= 5).

Analysis of variance of the log-transformed data indicated that there were significant differences between the B. thuringiensis/cereus population density from different quadrats or sample types. Counts in soil were in excess of 4 log10 CFU g−1, which were significantly higher than in all phylloplane samples (p < 0.0001). Furthermore, mean counts for grass phylloplane (2.69 log10 CFU g−1 were significantly greater (p < 0.0001) than the counts for the individual dock leaf types. Of the dock leaf types, mature leaves supported a significantly (p < 0.0380) lower density (1.60 log10 CFU g−1 than either immature or senescent leaves (2.01 log10 CFU g−1 and 2.03 log10 CFU g−1, respectively). It is also interesting to note that the mean population density of B. thuringiensis/cereus in quadrat E was significantly greater than in any of the other quadrats (p < 0.0046). The interaction between quadrat and sample type was not significant.

3.2 Taxonomy and plasmid profiles of isolates

Phylogenetic analysis of nucleotide sequences from amplified 16S rRNA gene fragments indicated that all selected isolates were closely related to the B. thuringiensis/cereus/anthracis/mycoides cluster of stains downloaded from Ribosomal Database Project (Fig. 2). This group was obviously distinct from the other Bacillus species included in the analysis. The phylogenetic analysis also demonstrated the diversity within the isolates obtained in this study.

2

Neighbour-joining tree showing phylogeny of 16S rRNA genes amplified from a selection of isolates from sample A4. Bar indicates 1% sequence divergence and bootstrap values above 50% are shown. Aligned sequences were 650 nucleotides in length.

Restriction analysis of extracted plasmids revealed that all of the isolates examined were distinct from B. thuringiensis var. kurstaki (Fig. 3). B. thuringiensis var. kurstaki was isolated from DiPel, a commercial formulation which is licensed for use in the UK and as such represents the most likely anthropogenic contaminant, although no commercial B. thuringiensis sprays have been used in the vicinity of our study site. This also illustrated the variation in plasmid profiles of strains isolated from a single leaf. Of the 7 isolates, there were 5 distinct profiles, but lanes 4 and 8 were identical to one another, as were lanes 6 and 7 (Fig. 3). Estimations of plasmid DNA content from the dock leaf isolates vary from approximately 50 kb (lane 3) to in excess of 150 kb (lane 6).

3

Agarose gel electrophoresis of Eco RI and Pst I restricted plasmids extracted from B. cereus/thuringiensis isolates. Lane 1 –B. thuringiensis var. kurstaki isolated from DiPel; Lanes 2–8 – isolates from senescent leaf from plant A4; Lanes Ma and Mb – size markers (Hyperladder VI [Bioline, London, UK] and Lambda DNA Eco RI +Hin dIII [Promega, Southampton, UK], respectively. Sizes indicated are bp.

3.3 cry Gene diversity of B. thuringiensis/cereus isolates in the phyllosphere of dock

PCR and gel electrophoresis was used to confirm the presence of a variety of the most common cry genes (with potential activity against coleopteran, dipteran and lepidopteran herbivores known to be active on dock plants [18]) in the B. thuringiensis/cereus isolates isolated from phylloplane and soil samples taken from our study site. The results of the examination of a subset (n= 160) of the isolates are summarized in Table 3. As there are many more cry genes that could be tested for, this only provides a small indication of the total cry gene diversity and thus it is probably an underestimate estimate of the proportion of toxin-bearing isolates in the collection. However, we can use these data to provide a conservative value for the Bt index (ratio of B. thuringiensis:B. thuringiensis/cereus[10]) of 0.425 in the 160 isolates that were examined. This value was not habitat-dependent as the Bt indices for soil, grass and dock were 0.400, 0.440 and 0.427, respectively. However, there was a great variation in individual sample types.

View this table:
3

Number of isolates from A4, C3 and E3 containing cry genes 1, 2, 3, 4, 7 and 8

PlantSample(n)acry1cry2cry3cry4cry7/cry8
A4Immature (6)01000
Mature (4)00000
Senescent (7)30000
Grass (9)60000
Soil (9)70000
C3Immature (11)00000
Mature (5)00000
Senescent (38)2717000
Grass (9)00400
Soil (10)30000
E3Immature (15)120000
Mature (2)00000
Senescent (22)11000
Grass (7)00100
Soil (6)00000
Total59 [36.9]b19 [11.9]b5 [3.1]b00
  • Total shows combined number of isolates from all samples examined which tested positive for each of the cry genes.

  • a Number in parentheses indicates the number of isolates tested for each sample.

  • b Number in square parentheses indicates the percentage of total isolates.

The most common gene identified was cry1 (36.9% of isolates examined), followed by cry2 (11.9%). Fourteen isolates from C3 senescent leaves and one isolate from an E3 senescent leaf were positive for the presence of both cry1 and cry2 genes. No isolates tested were positive for cry genes 4,7 or 8 and all of the isolates from dock leaves were negative for cry3 (see Table 3). Isolates from mature dock leaves did not contain any of the cry genes tested for. PCR products corresponding to cry3 were only observed in grass isolates, whilst soil isolates were positive for cry1 only. The predominance of the cry1 gene was consistent for each of the three plants examined, but there was a high prevalence of cry2 in isolates from plant C3.

4 Discussion

In this study, as in previous work [3,4,610], B. thuringiensis/cereus has been found in significant numbers in the phyllosphere. B. thuringiensis/cereus was isolated from 91% of the dock leaves sampled. However, the PCR screening data indicate the probability that a lower proportion than this carried entomopathogenic B. thuringiensis strains. This means that our values are similar to those obtained by other authors. For example, Smith and Couche [4] isolated B. thuringiensis from between 50% and 70% of leaf samples analysed, whilst Ohba [3] found B. thuringiensis on 96% of mulberry trees examined, but only on 11% of leaves tested. The highest proportion of positive samples to date (74%) was attributed to the location of sampling sites in the tropics [8]. However, we found similar proportions of positive samples of leaves from a temperate region. There may be several explanations for this. Dock may produce plant exudates that encourage B. thuringiensis/cereus survival or the unusually high pH of the dock phyllosphere [18] may be particularly suitable to B. thuringiensis/cereus. This is particularly important as it is known that germination of spores is induced by alkaline conditions in the insect gut [2]. Alternatively, it is possible that the large, thick leaves protect the phyllosphere bacteria from UV light, desiccation and wash-off which are the main factors that regulate epiphytic population densities [19]. However, it is most likely that the technique of isolating B. thuringiensis/cereus using B. cereus specific agar is more successful than previous methods used which have included heating samples to remove non-spore forming bacteria and thereby killing any vegetative Bacillus cells. This is of particular relevance as the previous method only allows isolation from spores, which by their nature are not metabolically active and therefore not of immediate functional importance in their habitat [20], whereas direct isolation will also capture those cells which were metabolically active at the time of sampling.

The density of B. thuringiensis/cereus in the dock phyllosphere varied between undetectable and 4.29 log10 CFU g−1. Using the Shapiro–Wilk test it was possible to demonstrate that the populations were lognormally distributed among leaves (Table 2). Lognormal microbial population distributions are common in the phylloplane [21,22] and appear to be independent of density. Whether such distributions are a result of dynamic processes such as immigration, emigration, birth and death remains to be determined. However, it has been suggested that leaf quality is primarily responsible for the variable distribution, indicating that growth plays a vital role in maintaining populations on leaves [21]. In much the same way as the lognormal distribution of epiphytic populations of phytopathogens has been put forward as an explanation of variation in the onset of plant disease [12], the lognormal distribution of B. thuringiensis may explain, at least in part, the lack of natural epizootics observed in nature [2].

Evidence is mounting to suggest the merging of B. cereus and B. thuringiensis into a single species [23], which is supported by the phylogenetic analysis performed here (Fig. 2). We have effectively demonstrated the accurate selectivity of the isolation media and showed that the isolates examined belong to the B. thuringiensis/cereus species cluster. Although the traditional method for differentiating between B. cereus and B. thuringiensis has been to microscopically examine sporulating cells for the presence of toxin crystals, it is regarded as too narrow a criterion for taxonomic purposes [23] and the subjective nature of the test precluded its use here. Furthermore, there has been evidence that not all isolates with insecticidal activity produce visible crystal inclusions [24] and isolates without visible crystal inclusions can test positive for cry genes [25]. As a result, PCR was used to assess the presence of a selection of the most common toxin genes in our collection. PCR analysis of a sub-sample of the isolates indicated that over 42% contained one or more cry genes (Table 3). As with the majority of previous studies of both phyllosphere and soil samples [8,2628], cry 1 genes followed by cry 2 were the most prevalent in our collection of isolates (Table 3). The only other cry type that was detected was cry 3 but was restricted to isolates from grass. None of the isolates from the dock plants tested had cry genes 3, 4, 7 or 8 (Table 3). Although it is commonly thought that cry genes are located on plasmids [23], it is possible that some cry genes may be found on the chromosome and thus have gone undetected because of the focus on plasmid DNA in this study. It has also been suggested that hybridization methods could have detected more toxin genes than PCR [29], but the converse result has also been noted [30]. It is apparent that no method will detect all possible cry genes but regardless of this, our results indicated some interesting trends in the distribution of B. thuringiensis/cereus populations and of the cry genes within those populations.

There were significant differences between the densities of B. thuringiensis/cereus on the immature, mature and senescent leaves of dock and grass (Fig. 1). The average density of B. thuringiensis/cereus in the grass samples was higher than found in dock but the significance of this is unknown. However, individual grass leaves were not examined and instead bulk samples were used for enumeration. If the distribution is lognormal as on dock leaves, bulking of samples will lead to an overestimation of the density [22]. The only other record of B. thuringiensis isolation from grass was not quantitative [9] and therefore cannot be used for comparison of densities. Nevertheless, the authors found that the majority of isolates were serovar israelensis with toxicity towards Diptera, but we found no evidence of dipteran-specific cry4 or lepidopteran- and dipteran-specific cry2 toxin genes in our isolates from grass. We did find that the presence of coleopteran-specific cry3 genes was restricted to isolates from grass samples (Table 3), but the common coleopteran herbivore Gastrophysa viridula[18] was seen to be active on the dock plants (but not grass) at our study site.

Mature dock leaves had significant lower densities of B. thuringiensis/cereus than either immature or senescent leaves. One explanation for this could be that due to the decumbent growth habit of dock, mature leaves are the most exposed to fluctuations in humidity and UV exposure, whilst immature leaves are protected at the centre of the plant and senescent leaves leak nutrients [12] and are close to the soil. Another could be that the plant actively promotes bacterial growth on newer leaves. These differences in density were also reflected in the distribution of cry genes within the isolates (Table 3). For example, PCR analysis failed to detect any cry genes in the isolates from mature leaves, in contrast to immature and senescent leaves where at least 40% were positive for cry1 or cry2 genes (Table 3). The reasons behind these differences are not known but previous hypotheses suggest that it could be linked to the prevalent herbivores on each plant, or on different leaves within each plant [4,8].

As the majority of our isolates that were positive for cry gene PCR showed similarity to cry gene profiles of strains used in commercial B. thuringiensis sprays, it was deemed necessary to ensure that they were not identical. This was done by comparison of plasmid restriction digests from a selection of isolates from this study with that from B. thuringiensis var. kurstaki isolated from DiPel (Fig. 3). In addition to demonstrating that isolates from this study were not the same as those used in a commercial formulation, the analysis showed considerable variation between isolates taken from the same leaf. Genetic variation and the possible role of horizontal gene transfer within the B. thuringiensis/cereus population will have considerable bearing on the ecology of these organisms so will be an interesting area for future study.

In conclusion, it has been demonstrated that commercially available selective media can be effectively used to enumerate B. thuringiensis/cereus populations from environmental samples and that a relatively high proportion of the isolates are potentially entomopathogenic, despite the genetic variation within the population. Furthermore, by sampling a large number of individual leaves from a relatively small geographical area we illustrated the lognormal distribution of the population and the effect of leaf type on average density and cry gene distribution. Aside from demonstrating the critical importance of spatial scale in sampling plans for bacterial populations on the phylloplane, this evidence adds further weight to the hypothesis first put forward in 1991 [4] and echoed more recently [5], that the relationship between B. thuringiensis, plants and herbivorous insects may show a degree of specificity. Further elucidation of potential ecological functions of populations of bacteria in natural habitats will depend on rigorous sampling plans and carefully chosen isolation procedures, but also on the estimation of growth of populations in such local habitats.

Acknowledgements

We thank Natural Environment Research Council for continuing funding and Kelly Houston for technical assistance.

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]
View Abstract