https://orcid.org/0000-0002-1199-689X
Figures
Abstract
Modern agricultural biotechnologies, such as those derived from genetic modification, are solutions that can enable an increase in food production, lead to more efficient use of natural resources, and promote environmental impact reduction. Crops with altered genetic materials have been extensively subjected to safety assessments to fulfill regulatory requirements prior to commercialization. The Brazilian National Technical Biosafety Commission (CTNBio) provides provisions for commercial release of transgenic crops in Brazil, including requiring information on pollen dispersion ability as part of environmental risk assessment, which includes pollen viability and morphology studies. Here we present the pollen viability and morphology of non-transgenic conventional materials, single-event genetically modified (GM) products, and stacked GM products from soybean, maize and cotton cultivated in Brazil. Microscopical observation of stained pollen grain was conducted to determine the percentage of pollen viability as well as pollen morphology, which is assessed by measuring pollen grain diameter. The pollen viability and diameter of GM soybean, maize and cotton, evaluated across a number of GM events in each crop, were similar to the conventional non-GM counterparts. Pollen characterization data contributed to the detailed phenotypic description of GM crops, supporting the conclusion that the studied events were not fundamentally different from the conventional control.
Citation: Soares D, Vertuan H, Bacalhau F, José M, Crivellari A, Belchior GG, et al. (2023) Genetically modified crops do not present variations in pollen viability and morphology when compared to their conventional counterparts. PLoS ONE 18(5): e0285079. https://doi.org/10.1371/journal.pone.0285079
Editor: Karthikeyan Adhimoolam, Jeju National University, REPUBLIC OF KOREA
Received: February 22, 2022; Accepted: April 16, 2023; Published: May 1, 2023
Copyright: © 2023 Soares et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper. If there is any need for additional data, the corresponding author should be consulted to provide the required data.
Funding: The funder provided support in the form of salaries for authors DS, HV, FB, MJ, AC, GGB and GUB, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section. Funding source: Monsanto Legacy/Bayer Crop Science.
Competing interests: The authors have declared that no competing interests exist. Daniel Soares, Hallison Vertuan, Fabiana Bacalhau, Marcia Jose, Augusto Crivellari, Gustavo Belchior and Geraldo Berger are employed by Monsanto Legacy / Bayer Crop Science and were provided financial support in the form of author´s salaries and research materials.
Introduction
Plants derived from modern biotechnology have been one of the fastest adopted crop technologies in history, with approximately 70 countries utilizing biotechnology-derived crops for cultivation and importation for food, feed, and processing. Genetically modified (GM) crops offer sustainable solutions to agriculture, especially important with increasing challenges brought by climate change and a growing population [1]. Conventional plant breeding has successfully improved crops through selection practices that capture genetic gains, even though information on the specific genes and genetic networks that contribute to the desired agronomic characteristics has historically been limited. GM crops providing agronomic traits, such as herbicide tolerance (HT) and insect resistance (IR), have shown significant commercial relevance due to yield protection, environmental benefits through reduced pesticide use, greater flexibility in crop management practices, and increased crop performance, allowing farmers to overcome challenges and meet their needs [2, 3].
Biotechnology-derived crops have been adopted for more than 25 years, and their extensive biosafety assessment has consistently concluded that they do not pose an increased food/feed or environmental risk when compared to their conventional counterparts [4]. Additionally, commercially available GM crops have been determined to pose negligible ecological risk via invasion of non-agricultural habitats based on the absence of phenotypic differences with non-transgenic comparators that would indicate an increase in invasiveness potential [5, 6]. Environmental Risk Assessment (ERA) field trials are designed to test whether the invasiveness potential of the transgenic crop has increased due to unintended effects of transformation and introgression of the biotechnology-derived trait [7].
Evidence in previous environmental risk assessment studies have shown that both single-event GM and stacked-event GM products do not represent risks for human, animal, and environment. [8].Trials performed across Brazil demonstrated there were no differences in agronomic and phenotypic plant characteristics for both single-event and stacked products for soybean, maize and cotton when compared to their conventional counterparts [9]. Moreover, there is scientific consensus that genetic engineering methods are safe [10–19]. Even though years of research and decades of commercialization of GM crops show that these products are safe and raise no additional biosafety concerns, risk assessments for these products are still required in many contexts, including ERA.
The Brazilian National Technical Biosafety Commission (CTNBio) is the regulatory agency for technical and advisory matters pertaining to GM crops in Brazil. Applicants seeking to commercialize biotechnology-derived crops should provide CTNBio with molecular characterization, product biosafety data to human and animal health, and environmental data from an ERA perspective.
As part of the ERA to evaluate the biosafety of GM crops, the Brazilian regulatory agency requests pollen evaluation of GM plants compared to their conventional counterparts to be presented as part of the risk assessment evaluation for biotechnology-derived crops. Here, we present pollen viability and morphology (diameter) data collected from non-GM conventional, single-event GM and stacked-event GM products from soybean (Glycine max), maize (Zea mays) and cotton (Gossypium hirsutum) cultivated in different, representative agricultural regions in Brazil.
Materials and methods
Field trials were conducted in five different seasons (from 2013 to 2017) in Brazil at Bayer Experimental Stations in Santa Cruz das Palmeiras, São Paulo (SP) State and Rolândia, Paraná (PR) State for field pollen collection and analysis. All trials were conducted in a randomized block design with four replications. Crop materials included GM soybean (Glycine max L.), GM maize (Zea mays L.) and GM cotton (Gossypium hirsutum L.) products, described in Table 1. Along with each crop and event combination, a conventional control (with the same genetic background as the GM event) and three to four commercial references available in the market were cultivated. The materials (GM, conventional control and references), cultivation season and location are presented in S1 Table.
After event transformation and selection in soybean, maize and cotton, conventional breeding techniques were used to obtain the introgressed variety (soybean and cotton) or hybrid (maize). Backcrossing was utilized to obtain the event in the desired variety or inbred background prior to field testing. Generally, the donor parent contains the gene of interest that can be introgressed via backcrossing into a variety or inbred. Backcross breeding allows the transfer of one or more genes of interest from a donor parent into the background of an improved variety/inbred to recover the recurrent parent genome by eliminating the undesirable genes of the transformation line. The backcross populations are made by crossing the recurrent parent with the donor parent to produce the F1 generation, crossing the F1 generation again with the recurrent parent (BC1F1) [20] and then conducting successive backcrosses to recover the recurrent parent genome while retaining the desired characteristic (e.g., insect resistance or herbicide tolerance). To obtain the stack products, classical breeding crosses and selection were made, where the transgenes of interest came from paternal, maternal or both parents, depending on the stack product. Specifically for maize, after inbred introgression, the single cross hybrid (F1) was produced for planting in the trials. In the cultivation of GM plants in the field, both the event and its respective conventional control were properly evaluated.
The events and corresponding conventional controls in these trials were introgressed and cultivated in a 5.9 maturity group soybean variety, a 130 relative maturity maize hybrid, and a mid-maturity cotton variety. These selected genetic backgrounds were appropriate for the geographic regions in which the trials were conducted. Commercial reference materials for each trial were selected to represent a range of maturities and genetic backgrounds available on the market and to be suitable to the geographic regions where they were cultivated.
Sample collection
Pollen grains from GM, conventional counterpart and commercial reference plants were collected randomly in each plot. Plots of each material were replicated four times in a randomized complete block design.
For soybean sampling, flowers were collected at the R2 growth stage. Newly opened individual flowers were selected for pollen collection. Soybean flowers were kept cool with ice packs upon collection to avoid decrease in viability. One sample of approximately 20 flowers was collected per plot. Fresh flowers were transferred to the laboratory where pollen was extracted from anthers using forceps and stained prior to microscopic observation.
For maize plants, pollen was collected between the VT and R1 growth stages. Pollen samples were collected from three different tassels per plot and fixed in the stain solution as soon as possible after collection, as viability decreases rapidly when removed from the plant. Three subsamples were collected per plot, each subsample consisting of pollen from the tassel of a single plant. In the field, pollen samples were sieved through a wire mesh screen (#40 mesh sieve) to remove excess debris and anthers and were stained within 30 minutes of collection.
Cotton plants were sampled at approximately the peak bloom growth stage. Newly opened white cotton flowers were selected to ensure fresh pollen was collected. Three subsamples were collected per plot, each subsample consisting of a single flower from a randomly selected plant. Pollen grains were collected from each subsample and fixed in the stain solution immediately after collection.
Staining procedure and laboratory analyses
The pollen viability and morphology analyses were carried out in the laboratory in Santa Cruz das Palmeiras, SP, Brazil and Rolândia, PR, Brazil. The evaluation was performed using a staining protocol [21]. For soybean, viable pollen grains were stained dark blue and light blue for non-viable grains. For maize, viable pollen grains were stained purple and greenish for non-viable grains. For cotton, viable pollen grains were stained dark blue and greenish for non-viable grains. After solution preparation and pollen collection, pollen grains were placed in the staining solution, diluted to 1:5 (1 part stain:5 parts distilled water) for maize and cotton and undiluted for soybean, and mixed thoroughly.
Staining solution-to-pollen volume ratio (v/v) was of at least 2:1 for maize and cotton. Soybean and maize pollen were stained for at least 10 minutes before microscope observation; cotton pollen was stained for at least 20 hours at room temperature due to pollen wall thickness. Approximately 20–30 μL of each stained pollen subsample was added to a glass slide for observation under the microscope. Fig 1 represents pollen samples from soybean, maize, and cotton, respectively, analyzed under an optical microscope.
Optical microscope observation for analysis of pollen viability for soybean (A), maize (B), and cotton (C) at 100x magnification.
Pollen viability and diameter evaluation
The number of viable and non-viable pollen grains was determined by visual counting, from a single field or multiple fields of view, with a microscope field of view (FOV) showing the pollen grains at a total magnification of 100x (10x objective lens magnification × 10x eyepiece lens magnification). Approximately 75–100 pollen grains were counted for viability per subsample. All pollen grains within the FOV were counted, even if the number of pollen grains exceeded 100. Viable and non-viable pollen grains were counted, and the percentage of viable pollen grains was estimated.
Microscopic Leica IC80 HD images were used for pollen diameter measurements. Microscopic images were taken of at least 1 subsample per replicate. Using the saved pollen sample image(s) from 1 randomly selected subsample from each plot, the diameter of 20 viable pollen grains were determined.
Variance components analyses
Analysis of variance (ANOVA) was performed and pairwise differences between the GM events and their conventional counterparts were tested by t-test at the 5% level of significance (α = 0.05). The analysis was conducted according to the following model for a randomized complete block design, using JMP® 12 software:
where:
Yij is the observed response for the jth material (GM and non-GM) in the ith replicate;
μ is the overall mean;
Ri is the random effect of the ith replicate;
Mj is the fixed effect of the jth material (GM and non-GM);
εij is the residual error.
Reference ranges were determined from the minimum and maximum mean values observed for commercial references materials. When significant differences between the GM event and conventional control were detected, the GM mean value was compared to the reference range and determined if the value was within the range. The commercial references cultivated in these studies represented the natural variability for pollen viability and morphology (diameter) in each crop.
ERA studies on GM crops typically make use of pairwise comparisons between the GM product and its conventional counterpart to evaluate mean values. This statistical analysis approach used to compare the means between GM crops versus conventional control has been demonstrated in previous publications [8, 9, 22].
Results and discussion
For soybean, maize and cotton crops, the results were expressed as averages of the percentage of pollen viability and the diameter of pollen grains (morphology) in an evaluation between GM crops (single events and stacked products) and their respective conventional counterparts. Commercial references were included to establish a natural range of pollen viability and morphology of the crop within the marketplace. The results are presented in Figs 2–5.
Soybean pollen viability (%) (A, C, E) and diameter (μm) (B, D, F) evaluation in single and stacked event products. Test materials (single and stacked event products) were compared to conventional controls (conventional counterparts). Commercial references provided the minimum and maximum mean values used to create a crop-specific range. Data are presented as box and whisker plots, where the plus symbol (+) represents the mean value, the upper limit is the maximum value, and the lower limit is the minimum value for a given material. Commercial reference material names, trial locations and years can be found in S1 Table. “Soybean stack” refers to the stacked event MON 87751 × MON 87701 × MON 87708 × MON 89788. No significant differences (α = 0.05) were detected between any test and conventional control comparisons.
Maize pollen viability (%) (A, C, E) and diameter (μm) (B, D, F) evaluation in single and stacked event products. Test materials (single and stacked event products) were compared to conventional controls (conventional counterparts). Commercial references provided the minimum and maximum mean values used to create a crop-specific range. Data are presented as box and whisker plots, where the plus symbol (+) represents the mean value, the upper limit is the maximum value, and the lower limit is the minimum value for a given material. Commercial reference material names, trial locations and years can be found in S1 Table. No significant differences (α = 0.05) were detected between any test and conventional control comparisons.
Maize pollen viability (%) (A, C, E) and diameter (μm) (B, D, F) evaluation in single and stacked event products. Test materials (single and stacked event products) were compared to conventional controls (conventional counterparts). Commercial references provided the minimum and maximum mean values used to create a crop-specific range. Data are presented as box and whisker plots, where the plus symbol (+) represents the mean value, the upper limit is the maximum value, and the lower limit is the minimum value for a given material. Commercial reference material names, trial locations and years can be found in S1 Table. “Maize stack” refers to the stacked event MON 87427 × MON 89034 × MIR162 × MON 87411. No significant differences (α = 0.05) were detected between any test and conventional control comparisons.
Cotton pollen viability (%) (A, C) and diameter (μm) (B, D) evaluation in single and stacked event products. Test materials (single and stacked event products) were compared to conventional controls (conventional counterparts). Commercial references provided the minimum and maximum mean values used to create a crop-specific range. Data are presented as box and whisker plots, where the plus symbol (+) represents the mean value, the upper limit is the maximum value, and the lower limit is the minimum value for a given material. Commercial reference material names, trial locations and years can be found in S1 Table. “Cotton stack” refers to the stacked event COT102 × MON 15985 × MON 88913 × MON 88701. *Indicates significant difference (p≤0.05) between GM and conventional control.
There were no significant differences in pollen viability (%) and morphology (diameter) across the three GM crops tested compared to their conventional controls, except for the viability percentage of MON 88701 cotton, which presented a lower mean value when compared to the conventional control (96.85 vs. 98.55%). However, the MON 88701 mean value was within the range of commercial reference means tested in the same trial (Fig 5A). It is important to note that commercial references (varieties/hybrids) are included in ERA studies to observe the variability that already occurs for each agronomic characteristic in each particular crop [5, 23, 24].
Agronomic endpoints related to reproduction (pollen viability, pollination rates, etc.) can be used to inform whether a GM trait has increased the potential for outcrossing, and provides information on the crop biology [25] and the potential to increase weediness or invasiveness due to reproduction rates [26].
Using a similar methodology and working with different soybean events, a previous study also observed no significant differences in pollen characteristics between GM soybean and their respective non-GM conventional controls [27]. An ERA study performed in Brazil assessed agronomic characteristics, including pollen viability, for a single-event and a stacked maize product. Pollen viability, as well the agronomic characteristics for all GM maize materials were statistically indistinguishable from the non-GM used as a conventional control [13]. A study with a herbicide tolerant GM cotton demonstrated that no significant differences were detected between the HT event compared to the conventional control when the floral morphology and pollen viability were assessed [28]. These results corroborate the results found in this paper, demonstrating that single events and stacked products across multiple crops do not change pollen viability and diameter when compared to a conventional control, and do not pose relevant pathways to impact traits related to reproduction characteristics.
Conclusions
These results show that the genetic modification of soybean, maize, and cotton by the introduction of genes that confer insect resistance and/or herbicide tolerance does not affect the pollen viability and diameter of these crops studied in Brazil. These results corroborate the current weight of evidence from the previously published literature which support the biosafety profile of GM crops from an ERA perspective. The pollen characterization data contribute to the detailed phenotypic description of GM crops and support the conclusion that the studied events were not fundamentally different from the conventional control and do not pose relevant pathways to impact traits related to reproduction characteristics.
Supporting information
S1 Table. Single and stacked events, conventional control and commercial references planted across different growing seasons and locations for pollen collection.
https://doi.org/10.1371/journal.pone.0285079.s001
(PDF)
Acknowledgments
The authors would like to thank the Field Team across sites in Brazil for their scientific contribution and the Bayer Regulatory Science LATAM department for all the support and incentive. We are grateful for manuscript review by Dr. John Vicini, Dr. Timothy Ball, Dr. Cody Postin and the Bayer Scientific Council.
References
- 1. ISAAA. Biotech Crops Drive Socio-Economic Development and Sustainable Environment in the New Frontier 2019 Available from: https://www.isaaa.org/resources/publications/briefs/55/executivesummary/default.asp.
- 2. Raman R. The impact of genetically modified (GM) crops in modern agriculture: A review. Gm Crops & Food-Biotechnology in Agriculture and the Food Chain. 2017;8(4):195–208. pmid:29235937
- 3. Smyth SJ. The human health benefits from GM crops. Plant Biotechnology Journal. 2020;18(4):887–8. pmid:31544299
- 4. Waters S, Ramos A, Culler AH, Hunst P, Zeph L, Gast R, et al. Recommendations for science-based safety assessment of genetically modified (GM) plants for food and feed uses. Journal of Regulatory Science. 2021;9(1):16–21. https://doi.org/10.21423/jrs-v09i1waters
- 5. Horak MJ, Rosenbaum EW, Phillips SL, Kendrick DL, Carson D, Clark PL, et al. Characterization of the ecological interactions of Roundup Ready 2 Yield (R) soybean, MON 89788, for use in ecological risk assessment. Gm Crops & Food-Biotechnology in Agriculture and the Food Chain. 2015;6(3):167–82. https://doi.org/10.1080/21645698.2015.1067365
- 6. Horak MJ, Rosenbaum EW, Woodrum CL, Martens AB, Mery RF, Cothren JT, et al. Characterization of roundup ready flex cotton,‘MON 88913’, for use in ecological risk assessment: evaluation of seed germination, vegetative and reproductive growth, and ecological interactions. Crop science. 2007;47(1):268–77. https://doi.org/10.2135/cropsci2006.02.0063
- 7. Craig W, Tepfer M, Degrassi G, Ripandelli D. An overview of general features of risk assessments of genetically modified crops. Euphytica. 2008;164(3):853–80. https://doi.org/10.1007/s10681-007-9643-8
- 8. Clawson EL, Perrett JJ, Cheng LL, Ahmad A, Stojsin D, McGowan Y, et al. Consistent risk assessment outcomes from agronomic characterization of GE maize in diverse regions and as single-event and stacked products. Crop Science. 2019;59(4):1681–91. https://doi.org/10.2135/cropsci2018.12.0758
- 9. Jose M, Vertuan H, Soares D, Sordi D, Bellini LF, Kotsubo R, et al. Comparing agronomic and phenotypic plant characteristics between single and stacked events in soybean, maize, and cotton. Plos One. 2020;15(4). pmid:32339186
- 10. Herman RA, Gaffney J, Storer NP. Enlightened oversight of genetically engineered crops for the next generation. Agricultural & Environmental Letters. 2020;5(1):e20004. https://doi.org/10.1002/ael2.20004
- 11. Bell E, Nakai S, Burzio LA. Stacked genetically engineered trait products produced by conventional breeding reflect the compositional profiles of their component single trait products. Journal of Agricultural and Food Chemistry. 2018;66(29):7794–804. pmid:29953223
- 12. Berman KH, Harrigan GG, Nemeth MA, Oliveira WS, Berger GU, Tagliaferro FS. Compositional equivalence of insect-protected glyphosate-tolerant soybean MON 87701 x MON 89788 to conventional soybean extends across different world regions and multiple growing seasons. Journal of Agricultural and Food Chemistry. 2011;59(21):11643–51. https://doi.org/10.1021/jf202782z
- 13. de Cerqueira DTR, Schafer AC, Fast BJ, Herman RA. Agronomic performance of insect-protected and herbicide-tolerant MON 89034 x TC1507 x NK603 x DAS-40278-9 corn is equivalent to that of conventional corn. Gm Crops & Food-Biotechnology in Agriculture and the Food Chain. 2017;8(3):149–55. https://doi.org/10.1080/21645698.2017.1301331
- 14. Gampala SS, Fast BJ, Richey KA, Gao ZF, Hill R, Wulfkuhle B, et al. Single-event transgene product levels predict levels in genetically modified breeding stacks. Journal of Agricultural and Food Chemistry. 2017;65(36):7885–92. pmid:28825812
- 15. Herman RA, Fast BJ, Scherer PN, Brune AM, de Cerqueira DT, Schafer BW, et al. Stacking transgenic event DAS‐Ø15Ø7‐1 alters maize composition less than traditional breeding. Plant biotechnology journal. 2017;15(10):1264–72. https://doi.org/10.1111/pbi.12713
- 16. Marques LH, Santos AC, Castro BA, Storer NP, Babcock JM, Lepping MD, et al. Impact of transgenic soybean expressing Cry1Ac and Cry1F proteins on the non-target arthropod community associated with soybean in Brazil. Plos One. 2018;13(2). pmid:29394266
- 17. Raybould A, Graser G, Hill K, Ward K. Ecological risk assessments for transgenic crops with combined insect-resistance traits: the example of Bt11 x MIR604 maize. Journal of Applied Entomology. 2012;136(1–2):27–37. https://doi.org/10.1111/j.1439-0418.2010.01601.x
- 18. Ridley WP, Harrigan GG, Breeze ML, Nemeth MA, Sidhu RS, Glenn KC. Evaluation of compositional equivalence for multitrait biotechnology crops. Journal of Agricultural and Food Chemistry. 2011;59(11):5865–76. pmid:21280600
- 19. Venkatesh TV, Breeze ML, Liu K, Harrigan GG, Culler AH. Compositional analysis of grain and forage from MON 87427, an inducible male sterile and tissue selective glyphosate-tolerant maize product for hybrid seed production. Journal of Agricultural and Food Chemistry. 2014;62(8):1964–73. pmid:24397242
- 20. Hasan MM, Rafii MY, Ismail MR, Mahmood M, Rahim HA, Alam MA, et al. Marker-assisted backcrossing: a useful method for rice improvement. Biotechnology & Biotechnological Equipment. 2015;29(2):237–54. pmid:26019637
- 21. Alexander M. A versatile stain for pollen fungi, yeast and bacteria. Stain technology. 1980;55(1):13–8. pmid:6158141
- 22. Díaz OH, Meza JLA, Baltazar BM, Bojórquez GB, Espinoza LC, Madrid JLC, et al. Plant characterization of genetically modified maize hybrids MON-89Ø34-3× MON-88Ø17-3, MON-89Ø34-3× MON-ØØ6Ø3-6, and MON-ØØ6Ø3-6: alternatives for maize production in Mexico. Transgenic Research. 2017;26(1):135–51. https://doi.org/10.1007/s11248-016-9991-z
- 23. Nakai S, Hoshikawa K, Shimono A, Ohsawa R. Transportability of confined field trial data from cultivation to import countries for environmental risk assessment of genetically modified crops. Transgenic research. 2015;24(6):929–44. pmid:26138875
- 24. Sammons B, Whitsel J, Stork LG, Reeves W, Horak M. Characterization of drought‐tolerant maize MON 87460 for use in environmental risk assessment. Crop science. 2014;54(2):719–29. https://doi.org/10.2135/cropsci2013.07.0452
- 25. Bachman PM, Anderson J, Burns A, Chakravarthy S, Goodwin L, Privalle L, et al. Data Transportability for Studies Performed to Support an Environmental Risk Assessment for Genetically Modified (GM) Crops. Journal of Regulatory Science. 2021;9(1):38–44. https://doi.org/10.21423/jrs-v09i1bachman
- 26. Anderson J, Bachman P, Burns A, Chakravarthy S, Goodwin L, Privalle L, et al. Streamlining data requirements for the environmental risk assessment of genetically modified (GM) crops for cultivation approvals. Journal of Regulatory Science. 2021;9(1):26–37. https://doi.org/10.21423/jrs-v09i1anderson
- 27. Matsushita A, Goto H, Takahashi Y, Tsuda M, Ohsawa R. Consideration of familiarity accumulated in the confined field trials for environmental risk assessment of genetically modified soybean (Glycine max) in Japan. Transgenic research. 2020;29(2):229–42. pmid:31997144
- 28. Thomas WE, Pline WA, Wilcut JW, Edmisten KL, Wells R, Viator RP, et al. Glufosinate does not affect floral morphology and pollen viability in glufosinate-resistant cotton. Weed technology. 2004;18(2):258–62. https://doi.org/10.1614/WT-03-032R1