Journal list menu

Volume 46, Issue 2 p. 388-396
Free Access

Effects of genetically modified, herbicide-tolerant crops and their management on soil food web properties and crop litter decomposition

Jeff R. Powell

Corresponding Author

Jeff R. Powell

Department of Integrative Biology,

Current address: Freie Universität Berlin, Institut für Biologie –Ökologie der Pflanzen, Altensteinstraße 6, 14195 Berlin, Germany

*Corresponding author. E-mail: [email protected], Skype: jeff-powellSearch for more papers by this author
David J. Levy-Booth

David J. Levy-Booth

Department of Environmental Biology,

Search for more papers by this author
Robert H. Gulden

Robert H. Gulden

Department of Plant Agriculture, and

Search for more papers by this author
Wendy L. Asbil

Wendy L. Asbil

University of Guelph Kemptville Campus, Kemptville, Ontario, Canada K0G 1J0;

Search for more papers by this author
Rachel G. Campbell

Rachel G. Campbell

Department of Plant Agriculture, and

Search for more papers by this author
Kari E. Dunfield

Kari E. Dunfield

Department of Land Resource Science, University of Guelph, Guelph, Ontario, Canada N1G 2W1;

Search for more papers by this author
Allan S. Hamill

Allan S. Hamill

Agriculture and Agri-Food Canada, Greenhouse and Processing Crops Research Centre, Harrow, Ontario, Canada N0R 1G0; and

Search for more papers by this author
Miranda M. Hart

Miranda M. Hart

Department of Environmental Biology,

Search for more papers by this author
Sylvain Lerat

Sylvain Lerat

Department of Environmental Biology,

Search for more papers by this author
Robert E. Nurse

Robert E. Nurse

Agriculture and Agri-Food Canada, Greenhouse and Processing Crops Research Centre, Harrow, Ontario, Canada N0R 1G0; and

Search for more papers by this author
K. Peter Pauls

K. Peter Pauls

Department of Plant Agriculture, and

Search for more papers by this author
Peter H. Sikkema

Peter H. Sikkema

University of Guelph Ridgetown Campus, Ridgetown, Ontario, Canada N0P 2C0

Search for more papers by this author
Clarence J. Swanton

Clarence J. Swanton

Department of Plant Agriculture, and

Search for more papers by this author
Jack T. Trevors

Jack T. Trevors

Department of Environmental Biology,

Search for more papers by this author
John N. Klironomos

John N. Klironomos

Department of Integrative Biology,

Search for more papers by this author
First published: 03 March 2009
Citations: 48

Summary

  • 1

    Genetically modified (GM), herbicide-tolerant crops have been adopted extensively worldwide, resulting in increased homogenization of agricultural practices. However, several countries still view GM crops with trepidation, citing potential risks to human health and the environment. We currently know little about how non-target biota responds to cultivation of GM crops under field conditions.

  • 2

    This study describes a series of microcosm and field experiments in Ontario, Canada, that estimated the effects of transgenic, glyphosate-tolerant (GT) crops and their management on the abundances of detritivorous soil biota and crop litter decomposition.

  • 3

    Absolute abundance of few of the measured biotic groups were affected by either the herbicide or variety treatments and, where significant effects were observed, the responses were not consistent across all years or for all sample dates within a year. More frequently, but not consistently, the GT herbicide system was associated with increased fungal : bacterial biomass ratios, suggesting a state of reduced enrichment.

  • 4

    Although the conventional and GT varieties studied differed in composition, we observed few effects of the modification for glyphosate-tolerance on maize Zea mays and soybean Glycine max litter decomposition. Overall, the herbicide system associated with GT crops reduced soybean- and corn-litter decomposition, but responses were inconsistent across Ontario, with many trials demonstrating no effect. Effects were probably underrepresented in this study as average daily precipitation was positively correlated with the magnitude of this system effect and many sites received well-below average levels of precipitation.

  • 5

    Synthesis and applications. Most concerns regarding the potential impacts of GM crops on non-target biota have targeted traits associated with the biotechnology itself. However, shifts in management practices associated with biotechnology are also widespread and have the same, if not greater, potential to alter the structure and functioning of agroecosystem biodiversity. The lack of observed permanent negative effects on soil biota in this study is heartening; however, more research is required to determine the functional consequences of observed transient effects and effects on other biota, as well as how altered crop litter decomposition affects agroecosystem nutrient cycling and carbon sequestration.

Introduction

Genetically modified (GM) crops are among the most successful of the first generation of transgenic products, and those engineered to tolerate certain herbicides have the greatest market share (Duke 2005). There exist potential benefits of GM, herbicide-tolerant (HT) crops to growers and to society: herbicides can be applied within the crop, resulting in lowered fuel costs and soil erosion due to the need for fewer herbicide applications and less tillage (Marra, Piggott & Carlson 2004). Clearly, growers recognize these benefits since GMHT crops are exceedingly popular in North America; in the USA, 92% of soybeans (27·8 million ha), 68% of cotton Gossypium hirsutum (2·5 million ha), and 63% of maize (22·3 million ha) were seeded in 2008 with GMHT varieties (NASS 2008). However, many countries, especially in Europe, still view GM crops with trepidation, citing potential risks to human health and the environment, and limit their cultivation to only a few approved varieties and only under stringent conditions (Annerberg 2003; James 2007).

The most frequently used varieties of GMHT crops are engineered to express cp4-epsps, the product of which is not inhibited by a non-selective herbicide, glyphosate (Duke 2005). EPSPS catalyses an intermediate step in the shikimic acid pathway, responsible for production of aromatic amino acids, an important component of structural phenolic compounds (Herrmann & Weaver 1999). Unintended phenotypes in glyphosate-tolerant (GT) crops may directly or indirectly influence resource quality for herbivores and detritivores, with potential consequences for element cycling in agroecosystems.

Glyphosate may have direct toxicity on some bacteria, fungi, and protists as these organisms also utilize the shikimic acid pathway (Roberts et al. 1998; Zablotowicz & Reddy 2004; Feng et al. 2005). Glyphosate has low toxicity to animals since they do not possess the shikimic acid pathway; for these organisms, toxicity is usually linked to additional products in the formulation (Atkinson 1985). However, effects on biota may arise due to indirect effects associated with microbial metabolism of the glyphosate molecule (Franz 1985; Wardle & Parkinson 1992), trophic cascades (Wardle 2002), changes in weed communities and litter input (Liphadzi et al. 2005; Westra et al. 2008) and effects related to litter decomposition processes (House et al. 1987).

Studies estimating the non-target effects of GMHT genotypes and cropping practices have focused primarily on above-ground organisms, including pollinators, herbivores, and predators (e.g. Hawes et al. 2003; Morandin & Winston 2005). Studies of soil organisms are less common, despite the important roles that many soil organisms play in key ecosystem processes such as nutrient cycling, organic matter turnover, soil physical structure and plant growth promotion (Lilley et al. 2006). To date, a major below-ground focus has been to study effects of GMHT genotypes on the abundance and diversity of rhizosphere and endophytic microorganisms (e.g. Dunfield & Germida 2004; Powell et al. 2007); less attention has been directed at management practices associated with GMHT cropping systems (Haughton et al. 2003; Liphadzi et al. 2005; Powell & Dunfield 2007). No studies of which we are aware have investigated, concurrently, the relative importance of the genetic modification itself and associated shifts in management practices.

Much of the literature on the effects of GM plants on litter decomposition has focused on modifications associated with insect resistance (Cortet et al. 2006), altered lignin synthesis (Seppanen et al. 2007), and chitinase production (Vauramo et al. 2006). To our knowledge, there have been no studies evaluating effects of GMHT crops on litter decomposition in soil. The few studies that have estimated the effects of glyphosate on rates of litter decomposition have been conducted in the context of pre-plant and post-harvest herbicide applications (Hendrix & Parmelee 1985; House et al. 1987; Eijsackers 1991). Given the prevalent shift towards in-crop use of glyphosate in GT cropping systems, it is prudent to re-examine these effects in this new context.

The objectives of this study were to estimate and compare the relative effects of GT varieties and weed management using glyphosate on (i) the abundance of various trophic groups in soil, (ii) food web structure via relative abundance of biomass C in food web compartments, and (iii) crop litter decomposition. These objectives were addressed in three complementary studies consisting of a series of field experiments across southern and eastern Ontario, Canada, and a laboratory microcosm experiment.

Materials and methods

microcosm study: variety effects on soybean litter decomposition and soil microorganisms

Near-isoline soybean varieties were compared in this study (GT: OAC Raptor; conventional: OAC Bayfield). Soybean plants were grown for 60 days prior to incorporation into litterbags, constructed of fibreglass screen (30 mm width by 50 mm length, with 1 mm openings). Senesced leaves, pods (seeds removed), and stems were used; litter was cut into squares of approximately 1 cm2 (1 cm length for stems). Approximately 500 mg (dry weight) of litter was added to each litterbag; the litterbags and litter were dried at 60 °C to constant mass, which was then recorded. Concentrations of N (Kjeldahl method), C (combustion method), and cellulo-lignin (acid detergent fibre digestion) and lignin fractions were determined by Agri-Food Laboratories, Guelph, ON, Canada, according to AOAC International guidelines (Horwitz 2000).

We estimated litter decomposition in soil-filled plastic microcosms (40 × 28 × 23 cm), maintained under no light at 15 °C (16 h) and 10 °C (8 h) and 60% WHC. Four litterbags of each variety were placed in each of five microcosms (blocks); one litterbag of each variety (experimental units) was removed from each microcosm after 10, 19, 42, and 80 days. Upon collection, the litterbags were dried at 60 °C to constant mass, which was recorded. Percentage mass remaining was calculated as [(massremaining litter)/(massinitial litter+bag– massbag)] × 100. In addition, 10 g of soil were collected from beneath each litterbag. We used dilution-plating to estimate abundance of fungi and heterotrophic, proteolytic, and cellulolytic bacteria (Supporting Information, Appendix S1).

field study no. 1: variety and herbicide effects on soil biota and crop litter decomposition

Site description and experimental design

The experiment was conducted at the Elora Research Station (Elora, ON, Canada; 43°38′44″N, 80°23′42″W) on an area of the station that had not been cropped for at least 60 years. The soil was classified as a silty loam (26·1% sand, 60·1% silt, 13·8% clay; 5·0% organic matter; pH 7·3; 27·1 cm kg−1 cation exchange capacity). A maize–soybean crop rotation was established in 2003. Plots (14 × 12 m) were ploughed (15 cm) in autumn 2002, cultivated (8 cm) and fertilized (120 kg N ha−1, 42 kg P ha−1) in spring 2003, then managed with no-tillage for the remainder of the experiment. Plots were fertilized, as above, in spring 2005. Crops were planted at recommended seeding rates (c. 7·5 plants m−2 for maize; c. 40 plants m−2 for soybean).

The experiment was set up in a single factor design with seven treatment levels; the treatments differed in terms of the variety of crop used, the types of herbicides used, and the frequency of GT variety planting and glyphosate use over the course of the study (Table 1). Near-isoline varieties were not always available for planting. The maize varieties planted in 2003 (GT: DK35-51; conventional: DK355) and 2005 (GT: 39K39; conventional: 39K40) were derived from the same parent material. DK35-51 also expresses a cry transgene, derived from a bacterium (Bacillus thuringiensis), with toxicity against herbivorous lepidopteran, coleopteran, and dipteran insects; therefore, effects in 2003 may not have been entirely or in part due to cp4-epsps incorporation. However, Saxena & Stotzky (2001) and Griffiths et al. (2006) found few effects of Cry proteins from GM crop plants on soil microorganisms and soil mesofauna. Soybean varieties planted in 2004 and 2006 (GT: Monsanto DKB06–52; conventional: OAC Bayfield) were not isoline varieties and differences among varieties are interpreted as such.

Table 1. Experimental design of the field experiment at Elora, ON, Canada. Maize was planted in 2003 and 2005, and soybean was planted in 2004 and 2006
Treatment Variety* Herbicide
2003 2004 2005 2006 2003 2004 2005 2006
1 GT GT GT GT Glyphosate Glyphosate Glyphosate Glyphosate
2 GT GT GT GT
3 Conv Conv Conv Conv
4 GT Conv Conv Conv Glyphosate
5 GT Conv Conv Conv
6 Conv GT Conv Conv Glyphosate
7 Conv GT Conv Conv
  • * GT, glyphosate-tolerant; Conv, conventional.
  • All treatments received conventional herbicides; see Table 2.

The details regarding herbicide systems are given in Table 2. Glyphosate was applied using the same formulation (Roundup Transorb®) at all dates. Conventional herbicides were applied to all crops. Thus, the comparison of GT weed management to conventional weed management is an approximation and the data are interpreted in the context of glyphosate use.

Table 2. Herbicide management for the field experiment at Elora, ON, Canada (g a.i., g active ingredient; g a.e., g acid equivalent)
Year Date Herbicides applied Treatments*
2003 26 May Isoxaflutole (79 gai/ha), atrazine (800 gai/ha) All
24 June Nicosulfuron (25 gai/ha), Agral 90 All
21 June Glyphosate (1·8 gae/ha) 1,4
30 June Glyphosate (1·8 gae/ha) 1,4
2004 22 June Quizalofop-p-ethyl (72 gai/ha), imazethapyr (75 gai/ha), bentazon (840 g gai/ha), Sure-Mix (0·5% v/v) All
22 June Glyphosate (1·8 gae/ha) 1,6
9 July Glyphosate (1·8 gae/ha) 1,6
2005 10 May Dicamba/atrazine (1800 gai/ha), nicosulfuron (25 gai/ha), S-metolachlor/ benoxacor (1600 gai/ha), Agral 90 (0·2%) All
9 June Glyphosate (1·8 gae/ha) 1
17 June Glyphosate (1·8 gae/ha) 1
2006 30 May Quizalofop-p-ethyl (72 gai/ha), imazethapyr (75 gai/ha), bentazon (840 gai/ha), Sure-Mix (0·5% v/v) All
30 May Glyphosate (1·8 gae/ha) 1
28 June Glyphosate (1·8 gae/ha) 1

Sampling and biomass estimation

Soil samples were collected four times each year: once prior to planting, twice in-crop, and once following harvest. Five soil cores (10 cm depth, 6 cm diameter) were collected within each plot; all samples were taken at least 2 m from the plot boundary and 10 cm from the crop rows. Responses were measured independently for each core. Soil moisture was recorded for each sample and estimates are reported in terms of biomass C g−1 soil (dry mass).

Bacterial and fungal biomass were quantified using a differential fluorescent staining (DFS) procedure that identifies active cells and hyphae (Anderson & Slinger 1975; Morris et al. 1997). Fungal biomass C was estimated from hyphal length estimates (Allen & MacMahon 1985) using published estimates of hyphal diameter (1·65 µm) (Kjøller & Struwe 1982), density (0·33 g cm−3) (Van Veen & Paul 1979), and C content (45%) (Swift, Cook & Perfect 1979). Bacterial biomass C was estimated assuming 6·4 × 10−14 g C cell−1 (Hunt & Fogel 1983).

Protozoan biomass was estimated by incubating soil and a bacterized soil extract and counting the number of colonies after 7 days (Bamforth 1991). Biomass C of protozoans was estimated based on volume and assuming 50% C (Janssen & Heijmans 1998). Nematodes were extracted via centrifugal-floatation (Kimpinski 1993). In 2004 and 2005, stoma and esophageal morphology were used to divide nematodes into bacterial-feeding (open, unarmed stoma; prominent median bulb), fungal/root-feeding (stylet-bearing), and predatory types (large onchium or axial tooth) (Freckman & Baldwin 1990) prior to quantification. Biomass C for nematodes was estimated based on body width and assuming 50% C (Berg et al. 2001). Collembolans and mites were extracted onto dishes containing ethylene glycol, using a high efficiency canister-type soil-arthropod extractor (Lussenhop 1971), then sorted and counted. Biomass C for mites and collembolans was estimated based on body length and assuming 47·7 and 47·5% C, respectively (Tueben 1991).

Fungal-to-bacterial (F : B) biomass ratios were calculated in all years as biomass CF/(biomass CF + biomass CB). Ratios of bacterial-feeding : stylet-bearing nematodes (BF : SB), were calculated in 2004–2005 as biomass CBF/(biomass CSB + biomass CBF + 0·01); the small constant prevents undefined values for samples where no nematodes were found. This ratio does not differentiate between hyphal-feeding and root-feeding consumers but does detect changes in the relative amount of biomass present in bacterial-feeding consumers, many of which are responsive to changes in short-term nutrient enrichment (Ferris, Bongers & de Goede 2001).

Biomass estimates were used to estimate effects associated with continuous use of glyphosate (treatment 1 vs. treatment 2) and planting of GT varieties (treatment 2 vs. treatment 3).

Effects on crop litter decomposition

We estimated effects of glyphosate on soybean and maize litter decomposition in 2005 and 2006, respectively. Soybean and maize litter was collected from research plots at the Elora Research Station in the autumn of 2004 and 2005, respectively, and litterbags were prepared as described above (senesced leaves were used for maize litter).

A split-plot experimental design was used. The main plot factor, ‘herbicide system’ (treatments 1 + 2 in Table 1) was randomized in four blocks. Litter ‘location’ (two levels: soil surface, buried 5–10 cm below soil surface) was included as a subplot factor to estimate effects of litter profile on decomposition responses to herbicide effects. Bags were constructed and either placed on the soil surface or buried in the field plots. Concentrations of N, C, and cellulo-lignin and lignin fractions were determined as described above. Litterbags were placed in the field in May and collected periodically over the course of 12 months (‘date’ factor). Between one and six litterbags (usually three or more) of each factor combination were recovered from each plot at each sample date.

In addition, in 2006, decomposition was estimated for maize litter derived from GT (Pioneer 39K39) and conventional (Pioneer 39K40) varieties, which are near-isogenic hybrids. Litter was collected following maize senescence in 2005 from replicate research plots. Litterbags were constructed, placed in field plots (treatments 2 + 3 in Table 1), and collected as described above.

field study no. 2: herbicide effects across southern and eastern ontario

We also estimated litter decomposition in eight additional, established field experiments along a 675-km transect in southern and eastern Ontario; decomposition was estimated in all eight experiments in 2005 and in five experiments in 2006. The experiments were established in 2000, 2001, or 2002 and each was managed either under conventional tillage (CT) or no-tillage (NT). Fertilizer was applied and crops were seeded as described for study no. 1.

Each field experiment was initially established in a complete block design with two factors: ‘sequence’ (maize–soybean or soybean–maize in a 2-year rotation) and ‘herbicide system’ (two levels: ‘conventional’ or ‘GT’). Each factorial combination of litter species, year, tillage, and site for which data were collected was treated as an independent treatment level within the ‘experiment’ factor in the model, with each ‘herbicide system’ treatment replicated in four blocks within each ‘experiment’. All maize and soybean seeds and litter used in these trials were derived from GT varieties. Litterbags were placed at the soil surface in the no-tillage experiments and buried 5–10 cm below the soil surface in the CT experiments. Maize litter was added to soybean plots, and vice versa, to mimic litter type present in the actual rotation. After 4 months in the field, litterbags were collected (between one and eight per plot, usually three or more) and dried at 60 °C to constant mass, which was recorded. Percentage mass remaining was calculated as described above.

Herbicides and rates corresponded to recommendations (OMAFRA 2006). All treatments in the NT trials received a pre-plant burndown with glyphosate (Roundup Transorb®; 1·80 kg a.e. ha−1). Herbicides were applied prior to crop emergence in the conventional herbicide system treatment; maize plots received an application of S-metolachlor/benoxacor (Dual II Magnum®; 1·37 kg a.i. ha−1) plus dicamba/atrazine (Marksman®; 1·60 kg a.i. ha−1) and soybean plots received an application of flumetsulam/S-metolachor (Broadstrike Dual II®; 1·44 kg a.i. ha−1). For both crop species and tillage types, plots in the GT herbicide system treatment received one or two in-crop applications (late June, early July) of glyphosate (Roundup Transorb®; 0·90 kg a.e. ha−1 application−1), as deemed necessary by the site manager.

Statistical analyses

Responses were analysed with linear mixed effects models using the ‘nlme’ package (Pinheiro et al. 2006) in r version 2·3 (r Development Core Team 2006). Factors associated with the treatment structure in the models (year, block) were treated as random effects. Models with different variance assumptions were compared using likelihood ratio tests (Pinheiro & Bates 2000). The Shapiro–Wilks’ test was used to evaluate normality assumptions of error distributions; biomass estimates, when necessary, were square root- or log10- (plus a small constant) transformed, and biomass ratio estimates were arcsine (square root)-transformed prior to analyses. For these responses, each sample period [in-crop (July), in-crop (August), post-harvest, and following spring] was analysed independently to remove higher-order interactions.

Recovered litterbags occasionally had greater mass than when placed in the field, due to colonization by microbial biomass and an inability to completely remove soil that had become attached to litter, resulting in values exceeding 1·00 for proportion mass remaining. Therefore, models were fit to the square root of (the proportion of dry mass remaining in the litter bag + 0·5); this ensured that the model was fit with positive data values and that assumptions of normality for residual error distributions were realized. Means were calculated for the lowest nested level of experimental unit in each analysis [subplot (study no. 1) or plot (study no. 2) level for herbicide system contrasts; each litterbag was considered an experimental unit for the variety contrast (study no. 1)].

Fixed effects (herbicide, variety, profile) and their interactions with other fixed or random effects were evaluated using conditional F-tests in anova (when the number of experimental units per treatment was balanced) or by testing the likelihood ratio observed when comparing the null and alternative models against a random distribution of 999 simulated likelihood ratios (when the number of experimental units was unbalanced) (Pinheiro & Bates 2000; Faraway 2006). In addition, for the second field study, a backward elimination multiple regression analysis of a linear mixed effect model was performed to determine if differences among certain abiotic parameters (average daily minimum and maximum temperature, average daily precipitation, soil texture, and soil pH) and management factors [tillage system (for sites evaluating both NT and CT tillage systems), weed biomass at time of glyphosate application (for trials where these data were available; J.R. Powell, unpublished data)] at the sites may have been associated with the effect size due to herbicide system. Likelihood ratio tests were used to compare the fit of models before and after dropping each parameter; statistical significance was determined against a random distribution of 999 simulated likelihood ratios (Faraway 2006). Parameters were removed in order of decreasing significance, initially determined using conditional F-tests in an anova on the whole model.

P values of significance (< 0·05) and marginal significance (< 0·10) are reported for interpretive purposes. P values greater than 0·10 are reported as such.

Results

microcosm study

The GT soybean litter analysed was elevated in N (2·72 vs. 2·61%) and had greater levels of cellulo-lignin (42·1 vs. 40·5%) and lignin (11·5 vs. 9·1%) than the conventional litter (variability was not estimated). Organic C was higher (45·4 vs. 44·8%) but the C : N ratio was reduced (16·5 : 1 vs. 17·4 : 1) in the GT soybean litter. Changes in resource quality had little influence on decomposition, which only differed at day 19 (Fig. 1). Of the estimated microbial groups (Supporting Information, Table S1), only proteolytic bacteria responded to the variety manipulation, with a greater abundance under the GT litter at 10 days following litter incorporation (7·55 vs. 6·96 log CFU g−1 soil; Pvariety*date = 0·02). We also observed a (marginal) decrease in the ratio of fungal relative to total microbial counts under the GT litter by the end of the experiment (0·02 vs. 0·16%; Pvariety*date = 0·05).

Details are in the caption following the image

Mass of soybean litter remaining for a glyphosate-tolerant (GT) and a near-isoline conventional variety in the microcosm experiment (mean ± SE). Decomposition was reduced for GT soybean only on day 19 (P < 0·02).

field study no. 1

Few of the measured trophic groups were affected by either the herbicide or the variety treatments and, where significant effects were observed, the responses were only significant for one sample date (Supporting Information, Tables S2 and S3 ). In addition, treatment effects were often inconsistent across years.

Glyphosate increased fungal biomass C in the spring following the first year of the study (Ptrt*year = 0·086) and increased predatory nematode biomass C in July of the third year (Ptrt*year = 0·029).

In the first year of the study, mite biomass C was greater (July in-crop; Ptrt*year = 0·008) and predatory nematode biomass C was less (spring following the 2003 growing season; Ptrt*year = 0·090) in the presence of the GT maize variety. Biomass C of stylet-bearing nematodes for the post-harvest sampling was reduced in the presence of the GT variety in both 2004 and 2005 (Ptrt = 0·002), particularly in 2004 (Ptrt*year = 0·090). With the exception of the significant effect on stylet-bearing nematode biomass C in 2005, the variety effects were only detected in years when isoline crop varieties were not available; therefore, the responses may have been independent of the glyphosate-tolerance trait.

In the first year of the study, but not in subsequent years, the F : B ratio increased in response to glyphosate for the sample following application (marginally) and again in the following spring (Fig. 2). A significant, positive main effect on the post-harvest F : B ratio was associated with the glyphosate treatment; however, a significant interaction was also detected and, when the treatment contrasts were estimated within each year, the glyphosate treatment was only significant in 2006 (Fig. 2).

Details are in the caption following the image

Effects of glyphosate (gly) and glyphosate-tolerant (GT) maize (2003, 2005) and soybean (2004, 2006) varieties on the relative abundance of fungal biomass C compared to that of bacterial biomass C (backtransformed mean ± 95% CI). Asterisks indicate significant within-year glyphosate or variety contrasts.

No significant treatment effects on the ratio of bacterial-feeding to stylet-bearing nematodes were observed (Ptrt > 0·10 for all sample dates).

Litter decomposition responses

The GT maize litter used in the field study in 2006, relative to the conventional litter, was significantly depleted in N (0·93 vs. 1·13%; P = 0·03) and had greater levels of cellulo-lignin (38·9 vs. 36·6%; P = 0·02) and lignin (3·3 vs. 2·4%; P = 0·01). No difference in organic C was observed (45·5%; P = 0·61); the C : N ratio was significantly greater for the GT maize variety (49·3 : 1 vs. 40·4 : 1; P = 0·02). However, no effect of maize variety was observed in the field (Fig. 3).

Details are in the caption following the image

Mass of maize litter remaining for a glyphosate-tolerant (GT) and a conventional near-isoline variety at Elora, ON, Canada (mean ± SE). No effect on litter decomposition was detected in association with the GT maize variety (P > 0·10); this result was consistent for both buried and surface litter (P > 0·10).

Glyphosate use significantly reduced litter decomposition in both years and this response was consistent for the duration of the study and the two litter types (Fig. 4). The glyphosate effect was dependent on the location of litter placement, reducing decomposition of surface litter but not buried litter (Fig. 4).

Details are in the caption following the image

Mass of soybean (a; 2005–2006) and maize (b; 2006–2007) litter remaining in glyphosate-treated and untreated plots at Elora, ON, Canada (mean ± SE). Decomposition was reduced in the glyphosate-treated plots for surface (P < 0·001), but not buried (P > 0·10), litter. No significant interaction with sample date was detected (P > 0·10).

field study no. 2

The GT herbicide system significantly reduced litter decomposition when effects were averaged across the different experiments (Pherbicide = 0·02). However, the herbicide effect was dependent on the experiment in which it was measured (Pherbicide*experiment = 0·007; Supporting Information, Table S4). The multiple regression analysis indicated that, of the measured abiotic parameters, only average daily precipitation was significantly correlated with effect size (Fig. 5). No effect of tillage system on the response to herbicide system was detected (Fig. 5) and no correlation was observed between weed biomass at the time of glyphosate application and response to herbicide system (Pweed > 0·10).

Details are in the caption following the image

Positive correlation between the effect on crop litter decomposition for glyphosate-tolerant (glyphosate) and conventional (conv) cropping systems and the average daily precipitation at the site and in the years that decomposition was measured. CT and NT refer to sites managed using conventional- and no-tillage, respectively (Ptillage > 0·10). The dashed line shows the model prediction [logLik = 32·3, Pprecip = 0·028, slope = 0·079 ± 0·032 (SE), intercept = −0·132 ± 0·072 (SE)].

Discussion

Permanent responses on soil biota were not observed, suggesting a high level of resilience in the soil biota and a lack of a persistent effect resulting from the GM cropping system. In most cases where responses were detected, the responses were observed only in the first year of the study. This suggests that components of the agroecosystem were initially perturbed by the different cropping practices, but that they were less sensitive to these practices in subsequent years. In addition, when treatments were applied in only 1 year (treatments 4–7 in Table 1), responses were not detected beyond the spring of the following year (data not shown). While the transient nature of these effects is heartening, it would be premature to conclude that long-term responses are unlikely to occur. When significant effects on the fungal-bacterial biomass ratios were detected in our analyses of the soil food web, glyphosate always resulted in a shift toward more fungal biomass relative to bacterial biomass, suggesting transient shifts in decomposer communities toward conditions of slower nutrient turnover and declining enrichment (e.g., resources of increasing C : N ratio). The functional consequences of such shifts are not known, but may relate to nutrient cycling and food web condition (Coleman, Reid & Cole 1983; Ferris et al. 2001) or ecosystem stability (Moore, McCann & de Ruiter 2005; Rooney et al. 2006).

Litter decomposition was sometimes reduced in plots managed according to protocols used for GT varieties (herbicides and timing). Reduced decomposition may enhance soil organic matter accumulation (or decrease soil organic matter loss), but whether the reductions observed here are likely to influence carbon cycling remains to be determined. The magnitude of the GT herbicide effect was significantly correlated with average daily precipitation during the course of the second study, possibly due to the limiting effects of moisture availability on herbicide–microbial interactions (Malkomes 1992; Han & New 1994). Historical averages for the locations and months studied here range from 2·9 to 3·1 mm day−1. No effects were observed in trials receiving less than 2·5 mm precipitation day−1, while significant effects were observed for half of the trials receiving precipitation levels closer to or greater than historical averages and in both years of the Elora experiment (2·5 and 2·7 mm precipitation day−1, on average, in 2005–2006 and 2006–2007, respectively). Therefore, GT weed management may affect decomposition processes with greater frequency than observed here.

The responses observed here may be similar in other crops engineered to express cp4-epsps (cotton, canola Brassica napus, alfalfa Medicago sativa) or for other types of herbicide-tolerance (Duke 2005, Behrens et al. 2007), unless specific transformation events result in extreme pleiotropies. However, efforts to study soil system responses to herbicide-tolerance have been relatively minor when compared to other types of GM crops, for instance, those engineered to express Cry proteins (Lilley et al. 2006; Powell & Dunfield 2007). Other modifications specifically targeting lignin production in birch Betula pendula (Seppanen et al. 2007) or poplar Populus tremula × Populus alba (Tilston, Halpin & Hopkins 2004) have resulted in no detectable effects on decomposition processes, while few in situ effects of Cry proteins released from Bt-crops have been observed for soil biota (e.g. Saxena & Stotzky 2001). Therefore, changes in resource quality associated with transgene expression often may be too small to affect the abundance or functioning of soil biota.

While conclusions regarding the direct effects of transgene incorporation and expression are based on a limited number of studies, they suggest that greater focus instead should be placed on management practices associated with GMHT cropping systems. Granted, these studies did not evaluate the potential for horizontal transfer of herbicide-resistance genes, which may be a problem in locations where close relatives of crop species are present (Hüsken & Dietz-Pfeilstetter 2007) or under conditions that enhance transformation and selection of competent soil bacteria (de Vries & Wackernagel 2004; Levy-Booth et al. 2007). Nevertheless, a more pressing issue is likely to be weed species shifts and the development of herbicide resistance, which is already occurring with glyphosate (Owen 2008). The lack of observed permanent negative effects on soil biota in the present study is encouraging, but GMHT cropping systems should continue to be monitored to evaluate their impacts on decomposition processes. Future research should also focus on the effects of GMHT cropping systems that include tolerance to multiple herbicides (Behrens et al. 2007) and how these widespread weed management systems may be tailored to offset or mediate effects of increased weed control efficacy (Pidgeon et al. 2007).

Acknowledgements

This work was supported by a strategic grant (STPGP 258065-02) and a graduate scholarship from the Natural Sciences and Engineering Research Council of Canada and by graduate scholarships from the province of Ontario and the University of Guelph. Monsanto Canada provided funding for the management of sampling locations in the second field study. We thank Dr Mario Tenuta and three anonymous reviewers for comments.