A critique of studies evaluating glyphosate effects on diseases associated with Fusarium spp. J R POWELL* & C J SWANTON! Departments of *Integrative Biology and !Plant Agriculture, University of Guelph, Guelph, ON, Canada Received 3 February 2007 Revised version accepted 18 February 2008

Summary With the large-scale adoption of glyphosate-resistant crops in North America, there are concerns that nontarget microbial populations might be affected by increased frequency of glyphosate use. Stimulation of fungal species associated with crop diseases, including Fusarium spp., has been observed in laboratory and glasshouse experiments. Although field surveys in Saskatchewan detected positive associations between the incidence of Fusarium head blight and application of glyphosate formulations, few field experiments have been successful at demonstrating a stimulatory effect of glyphosate on crop diseases, including diseases associated with Fusarium spp. Taken at face value, there is little evidence from experimental field trials to support a causative link between glyphosate and crop diseases

associated with Fusarium spp. However, we are concerned that the experimental field trials investigating links between glyphosate and Fusarium spp. are not representative of interactions that occur under actual farming conditions. In addition, inadequate consideration may have been given to microbial ecology during the design and maintenance of these experimental field trials. At this time, there is insufficient evidence to prove or disprove a link between glyphosate and crop diseases associated with Fusarium spp. and this area should receive high research priority, given the rapid and widespread increase in glyphosate use. Keywords: fungal pathogen, Fusarium head blight, glyphosate, herbicide-resistant crops, microbial ecology, sudden death syndrome, weed–pathogen interactions.

POWELL JR & SWANTON CJ (2008). A critique of studies evaluating glyphosate effects on diseases associated with Fusarium spp. Weed Research 48, 307–318.

Introduction Herbicide use, in general, can have positive, negative or no effects on pathogenic organisms and associated crop diseases, through direct effects on inoculum levels or indirect effects on host susceptibility (Wisler & Norris, 2005). Many examples of plant disease outbreaks associated with herbicides can be found in the literature, as can many discrepancies with these examples (Le´vesque & Rahe, 1992; Altman, 1993; Wisler & Norris, 2005). Herbicide mode of action, crop genotype, differences among pathogen isolates and species, environmental conditions and lack of standardised techniques are among several factors that are likely sources of these discrepancies (Le´vesque & Rahe, 1992; Altman, 1993). As a result, it is difficult to generalise pathogen responses

to herbicides and interactions must be considered on a case-by-case basis. Recently, several studies have focused on glyphosate use in particular and its effects on fungal pathogens. Use of glyphosate (N-(phosphonomethyl)glycine) in North America has increased in the last decade, as a result of the large-scale adoption of genetically modified (GM), glyphosate-resistant (GR) crops by growers. GM herbicide-resistant varieties, the majority of which are GR varieties, were planted to 89%, 65% and 36% of US soya bean [Glycine max (L.) Merr.], cotton (Gossypium hirsutum L.) and maize (Zea mays L.) acreage, respectively, in 2006 (NASS, 2006a); in turn, glyphosate was applied (pre-plant and ⁄ or in-crop) to 88%, 71% and 31% of total soya bean, cotton and maize acreage, respectively, in 2005 (NASS, 2006b). Similar increases in

Correspondence: Jeff R Powell, Freie Universitaet Berlin, Institut fuer Biologie, Oekologie der Pflanzen, Altensteinstr 6, 14195 Berlin, Germany. Tel: +49 (0) 30 838 53172; Fax: +49 (0) 30 838 53886; E-mail: [email protected] ! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

308 J R Powell & C J Swanton

glyphosate use are likely as GR varieties are adopted in Europe. Several studies have demonstrated that microbial activity and ⁄ or biomass can be stimulated following application of glyphosate or its formulations (hereafter referred to as "glyphosate#) to field soil (e.g. Wardle & Parkinson, 1990a,b; Stratton & Stewart, 1991). Responses of individual fungal species can vary depending on susceptibility to the herbicide; some fungal species, for example, express glyphosate-sensitive forms of 5-enolpyruvylshikimic acid-3-phosphate synthase (EPSPS) (Anderson & Kolmer, 2005; Feng et al., 2005). Glyphosate-tolerant species of fungi may metabolise glyphosate, an amino acid-analogue, if they are able to utilise available phosphate or amine structures. Several fungal species, including some rust fungi and blight fungi, show enhanced growth on glyphosate-amended media (Wardle & Parkinson, 1990b; Hanson & Fernandez, 2003). Others show enhanced disease responses associated with glyphosate in the absence of increased fungal growth (Larson et al., 2006). In addition, root exudates of treated weed and crop plants can have stimulatory effects on fungal species, including some causative agents of damping off and root rot diseases (Liu et al., 1997; Kremer et al., 2005). These responses, observed under controlled laboratory conditions, have led some researchers to suggest that there is a link between glyphosate use and outbreaks of fungal disease (Termorshuizen & Lotz, 2002; Hanson & Fernandez, 2003).

Interactions between glyphosate and Fusarium spp. Glyphosate has been found to stimulate Fusarium spp. in glasshouse studies. For example, Sanogo et al. (2000) observed greater disease severity and increased isolation frequency of Fusarium solani f. sp. glycines on glyphosate-treated GR soya bean relative to untreated GR soya bean. Kawate et al. (1997) observed increased levels of F. solani f. sp. pisi in the rhizosphere of glyphosate-treated Lamium amplexicaule L. (henbit dead-nettle) relative to untreated henbit. Sanogo et al. (2001) built on their study by conducting field trials (discussed below); however, Kawate et al. did not expand their research to field trials. Despite these stimulatory effects observed in glasshouse studies, few field studies have demonstrated a link between glyphosate and crop diseases associated with Fusarium spp. Field survey data from Saskatchewan suggested that glyphosate can promote Fusarium head blight (FHB) of wheat and barley. In each of 4 years, increased FHB in spring wheat was positively correlated with the application of glyphosate in the previous 18 months (Fernandez et al., 2005). FHB severity (for

conventional- and zero- tillage systems) and incidence of Fusarium spp. infecting barley sub-crown internodes (for minimum-tillage systems) was significantly associated with glyphosate applied in the previous 18 months over 4 years (Fernandez et al., 2007a,b); number of glyphosate applications was significantly positively correlated with FHB severity associated with Fusarium graminearum and Fusarium avenaceum, particularly for cultivars with intermediate FHB resistance (Fernandez et al., 2007a). In addition, Le´vesque et al. (1987) observed that glyphosate applied to a number of weedy plant species in field plots enhanced populations of Fusarium spp., although there was no increase in disease in the subsequent barley (Hordeum vulgare L.) crop. Sanogo et al. (2001) observed that glyphosate resulted in higher levels of sudden death syndrome (SDS), caused by F. solani f. sp. glycines, in GR soya bean relative to untreated and lactofen- (but not acifluorfen- and imazethapyr-) treated soya beans, regardless of cultivar susceptibility to the disease; significant differences in root infection levels were not detected, suggesting that the response might have been manifest via means other than enhanced fungal population growth. Other field experiments did not show effects of glyphosate on either Fusarium spp. abundance or Fusarium-related disease levels (Njiti et al., 2003; Henriksen & Elen, 2005). One experiment observed reduced FHB severity associated with a sublethal dose of glyphosate to spring and durum wheat (Hansen et al., 2004). However, the glyphosate treatment in this experiment was due to accidental drift from a GR soya bean field and was not randomised across the trial. Researchers compared nursery wheat plots adjacent to and distant from the soya bean field and were, thus, unable to rule out effects of confounding variables. Few field experiments involving other fungal species have linked fungal diseases with glyphosate. Monosporascus cannonballus, the causative agent of root rot and vine decline of melons, increased following glyphosate when applied as a post-harvest crop destruction method; however, effects on disease development on subsequent crops were not reported (Stanghellini et al., 2004). Similarly, glyphosate applied as a spring burndown enhanced Rhizoctonia root rot of barley (Smiley et al., 1992). Of the few field studies examining effects of glyphosate applied in-crop on Rhizoctonia and Sclerotinia rots in GR crops, none demonstrated increased disease levels relative to untreated controls (Bradley et al., 2002; Harikrishnan & Yang, 2002; Nelson et al., 2002; Baird et al., 2005; Pankey et al., 2005). Based on the studies mentioned above, there is clearly disagreement in the literature when it comes to glyphosate use and enhanced plant disease associated with tolerant fungal species. The discordance between

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

Glyphosate–Fusarium interactions 309

glasshouse and many field studies might indicate that glyphosate stimulation of fungal diseases is largely a glasshouse phenomenon, where environmental conditions can be easily controlled to promote infection and advancement of the disease. Alternatively, differences among field studies in how error variances are partitioned and overall levels of variance and replication could explain why some studies observe significant treatment effects while others do not. For example, using data from table 1 and figs 1–3 in Njiti et al. (2003), the authors had sufficient power (assuming a = 0.10) to detect a 25% reduction in yield in only two of five experiments and did not have sufficient power to detect a 25% increase in Fusarium abundance or disease level in any of the experiments. However, in many cases, it is difficult to assess the statistical power of experimental designs, as authors do not report estimates of variation in the absence of statistically significant effects. Another (not necessarily exclusive) possibility for the discordance among field studies might be related to differing assumptions among these studies relating to how these glyphosate–pathogen interactions are mediated by other factors (e.g. cropping practices, host susceptibility, inoculum levels). As seen for other herbicides (Le´vesque & Rahe, 1992; Altman, 1993; Wisler & Norris, 2005), the context in which glyphosate is applied in relation to other factors might be an important determinant of whether effects are observed. Given the widespread adoption of GR crops in North America and continued market growth in other parts of the world, it is valid to consider how mediating factors in field experiments might contribute to glyphosate-associated responses. In a review of interactions between herbicides and fungal root pathogens, Le´vesque and Rahe (1992) provide a conceptual framework for the ways that glyphosate and other herbicides can influence crop diseases, via direct (e.g. stimulatory effects on pathogen growth and activity, negative effects on host defences) and indirect (e.g. increasing substrate available for pathogen growth, negative effects on antagonists of pathogens) pathways. Here, we summarise this framework and expand it to include interactions in GR cropping systems. We then use this framework to address how assumptions made in previous studies might have biased their outcomes and suggest how future studies could be conducted to effectively address the extent to which glyphosate can enhance crop diseases associated with Fusarium spp.

Potential mechanisms Figure 1 depicts a diagrammatic representation of the pathways and predicted effects by which glyphosate

could enhance crop diseases associated with Fusarium spp. In the absence of other factors, glyphosate indirectly enhances crop productivity by reducing the negative effects caused by crop–weed competition (Fig. 1A). However, this beneficial effect of glyphosate, relative to other herbicides, on crop productivity might be reduced or possibly reversed in the presence of pathogenic Fusarium spp. (Fig. 1B–E). Glyphosate might increase crop disease severity directly via stimulatory effects on plant pathogens (Fig. 1B). As shown in laboratory experiments, glyphosate itself can have stimulatory effects on growth of selected fungi when amended to growth media or applied to soil (Wardle & Parkinson, 1990b; Kawate et al., 1992; Hanson & Fernandez, 2003). Glyphosate or plant metabolites in root exudates can stimulate pathogen activity in the rhizosphere of treated plants, whether glyphosate-sensitive (following a non-lethal dose) or resistant (Liu et al., 1997; Kremer et al., 2005). Stimulatory effects on plant pathogens might arise indirectly as a result of glyphosate increasing host susceptibility to pathogenic fungi (Fig. 1C). Glyphosate inhibits synthesis of aromatic amino acids such as phenylalanine and, thus, inhibits the production of phenolics involved in disease resistance (e.g. phytoalexins) and the synthesis of structural compounds (e.g. lignin), resulting in increased pathogen susceptibility. While GR crops produce a form of EPSPS that is tolerant of glyphosate, the enzyme may not be as efficient as the wild-type EPSPS (in the absence of glyphosate), resulting in the potential for glyphosate ⁄ genotype or glyphosate ⁄ environment interactions affecting disease responses in GR crops (Pline-Srnic, 2005). Pathogen inoculum levels might be enhanced as a result of interactions among pathogens and weedy plants (Fig. 1D). Weed control, especially at times of high weed biomass (e.g. late in-crop applications, spring pre-plant burndowns), will result in a flush of organic matter substrate that might serve to increase inoculum levels of saprotrophic fungi, including many pathogenic species. This type of effect is not specific to glyphosate itself, but glyphosate effects on pathogen activity (Liu et al., 1997; Hanson & Fernandez, 2003) could interact synergistically with a flush of new organic matter to result in greater pathogenic fungal population growth (e.g. Smiley et al., 1992). Glyphosate could exacerbate fungal diseases if pathogens are stimulated to a greater extent than their antagonists, or if antagonists are susceptible to its toxic properties (Fig. 1E). For example, variable outcomes of competitive interactions among soil fungi were observed under a range of glyphosate concentrations (Wardle & Parkinson, 1992). We know little about the effects of

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

310 J R Powell & C J Swanton

A

Glyphosate – Antagonists

+

Weeds

Fusarium

– D

Crop B

+

+

Antagonists

Glyphosate

+

Fusarium –

Antagonists

Weeds

Fusarium

C

+

– E

–

Crop

– Weeds

–

Crop

Glyphosate – Antagonists

+ –

Weeds

Fusarium

–

Glyphosate –

Antagonists

Glyphosate

Weeds

Fusarium –

Crop

–

Crop

Fig. 1 (A) In the absence of other interactions, glyphosate indirectly benefits crop productivity by reducing the negative effects of crop–weed competition. (B–E) Hypothesised mechanisms by which glyphosate enhances crop diseases associated with Fusarium spp. (see text for detailed explanations). Solid arrows indicate direct interactions, while dashed arrows indicate indirect interactions. Signs indicate the type of effect (beneficial ⁄ antagonistic) that one interactant has on another. Except for (C), which is most likely for glyphosate-resistant cropping systems, the depicted mechanisms are possible in both conventional and glyphosate-resistant cropping systems.

glyphosate on Fusarium antagonists or on the outcomes of interactions between Fusarium spp. and their antagonists, especially under field conditions. Pseudomonas fluorescens was not negatively affected in glyphosateamended medium (Zboinska et al., 1992). Trichoderma harzianum and Streptomyces spp. increased when growth media were amended with glyphosate (Wardle & Parkinson, 1990b; Obojska et al., 1999). Glyphosate reduced root colonisation by arbuscular mycorrhizal (AM) fungi when applied in vitro to glyphosate-sensitive host plants (Wan et al., 1998), but effects following treatment of GR plants have not been observed (JR Powell unpubl. obs.). Clearly there are a number of pathways by which glyphosate could result in enhanced crop diseases associated with Fusarium spp. Below, we discuss the outcomes of field studies with regard to predictions derived from the conceptual model presented above.

Reconciling variable outcomes in glyphosate–Fusarium studies We described, in previous paragraphs, the variability that exists among field surveys and experiments in the responses of pathogenic fungi to glyphosate. Surveys, or

comparative studies, infer the relationship among two factors from the similarity of their responses to another, sometimes unknown, factor, while experimental studies infer the relationship among two factors from the response of one (dependent) factor to the manipulation of the second (independent) factor. Comparative studies are often looked upon as less desirable than experimental studies, as only the latter can demonstrate a causative relationship between two factors. The benefit of comparative studies is that they are conducted under realistic field conditions. Taken at face value, there is little evidence from the experimental field trials mentioned above to support a causative link between glyphosate and crop diseases associated with Fusarium spp. This suggests, by inference, that observational evidence for this link from field surveys might be due to confounding variables. However, it is not clear from these experimental studies how representative they are of the interactions that occur under realistic conditions, especially concerning ecological interactions occurring among microorganisms beyond the direct effect of glyphosate on host susceptibility and pathogen levels. Potential criticisms of these experimental studies, in relation to the ecology of Fusarium spp., can be grouped under two general categories: (i) lack of consideration

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

Glyphosate–Fusarium interactions 311

B

Disease severity

towards how initial inoculum levels influence glyphosate–pathogen and pathogen–antagonist interactions, and (ii) lack of consideration towards how weed cover and the composition of the weed community influence these interactions. In addition, the timing at which glyphosate is applied is an additional variable that can interact with weeds and inoculum to influence pathogen responses.

+ glyphosate

– glyphosate

A

Initial inoculum level

Associations between disease levels and the population dynamics of the disease-causing agent need not be linear. For example, disease symptoms might not be observed in a host population until the pathogenic organism reaches a population threshold whereby the defence response of the host is compromised. Similarly, the severity or incidence of the disease will not continue to increase at a constant rate, as the proportion of uninfected hosts in the population will decline and disease severity for individual hosts will reach an upper bound. Nonlinear relationships between disease levels and pathogen population dynamics suggest that herbicide effects on crop diseases are likely dependent on the initial status of the pathogen population. Stack (1989) observed that FHB in wheat increased in a sigmoidal fashion as the initial level of F. graminearum inoculum increased. Under these conditions, an effect of glyphosate would more likely exist with low inoculum levels, where stimulation can result in actual increased infection rates, compared with at high inoculum levels, where infection rates might already have reached a maximum or any increase might be smaller than the detection limits associated with high variability (Fig. 2). While Henriksen and Elen (2005) did not detect an effect of glyphosate on FHB, disease levels in all treatments were very high: mean per cent Fusarium diseased kernels (FDK) for oats was 77.9% across all treatments, ranging from 70.9% to 85.4% at each treatment level. In the study by Fernandez et al. (2005), where application of glyphosate was correlated with increased FHB, the authors state that FHB disease levels detected in Saskatchewan were low when compared with surveys of Minnesota, New Zealand and Ontario. FDK levels ranged from 0% to 7.6% in Saskatchewan (Fernandez et al., 2005), compared with a range from 0% to !25% in Ontario (Schaafsma et al., 2001). Pathogen inoculum level and activity varies due to climatological, meteorological and geological factors; further research is required to determine whether heterogeneity in these factors might act as a predictor of the effect of glyphosate on Fusarium-associated diseases. The selection of sites that already possess sizable disease levels and the manipulation of factors favour-

Initial inoculum level Fig. 2 Dependence of initial inoculum level on the ability to detect an effect of glyphosate on plant-pathogenic fungi, assuming a sigmoidal relationship between initial inoculum level and disease severity. The solid and dashed lines represent the relationship between initial inoculum level and disease severity in the absence and presence of glyphosate respectively. Potential stimulation of crop diseases by glyphosate at low inoculum levels (A) is easier to detect than at high inoculum levels (B), where disease severity is approaching capacity.

ing disease levels for experimental studies of glyphosate impacts on disease and pathogen populations might prevent researchers from detecting these effects. In the study by Njiti et al. (2003), field trials were only conducted at locations that displayed uniform distributions of SDS symptoms and, by the end of the experiment, abundance of F. solani f. sp. glycines was sizable, ranging from !200 to 1000 colony forming units (CFU) g)1 root. SDS occurrence and distribution is patchy at the field level in affected regions (Yang & Lundeen, 1997). While the patchy abundance of F. solani f. sp. glycines (0–3000 CFU g)1 root) and its relation to SDS levels in affected fields has been studied (Luo et al., 2000, 2001), its abundance in fields not demonstrating SDS symptoms has not. Site selection could act as a source of bias against detecting effects of glyphosate on Fusarium spp., especially if responses to glyphosate are stronger or more easily detected when the initial inoculum level is low. Fusarium spp. persist in agricultural soils in various forms, including as resting structures (such as perithecia, chlamydospores and conidia) or as hyphae in crop and weed residue. It is not known as to whether responses to glyphosate, whether direct or indirect, differ among the inoculum types. For example, actively growing hyphal structures might be better able to respond to increased litter input than resting structures, but this hypothesis needs to be tested to determine whether the type(s) of inoculum present has any power for predicting disease responses to glyphosate.

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

312 J R Powell & C J Swanton

Pathogen–weed interactions

Weedy plants can serve as hosts to disease-causing agents of crop species and various forms of weed control can result in altered behaviour of these agents. For example, the incidence of viral diseases is enhanced following post-emergence herbicide (glyphosate or nicosulfuron) application, as insect vectors move to crop plants from dying weeds (King & Hagood, 2003). Some weedy plants can serve as alternative hosts to pathogenic Fusarium spp. and saprotrophic behaviour by these pathogenic fungi on recently killed weedy plants might serve to increase the inoculum level in the current or subsequent crop. For example, growth and mortality of barley seedlings increased in the presence of glyphosatetreated Elytrigia repens (L.) Desv. ex Nevski (couch grass); the effect was attributed to the promotion of a barley pathogen, Fusarium culmorum, on decaying E. repens residues, especially under drier soil conditions (Lynch & Penn, 1980; Penn & Lynch, 1981). While glyphosate is not necessarily required to observe some of these effects (Lynch & Penn, 1980), the stimulatory effects of glyphosate on pathogen activity (Liu et al., 1997; Hanson & Fernandez, 2003) could interact synergistically with saprotrophic behaviour by pathogens on weed litter to result in enhanced pathogenic fungal population growth (e.g. Smiley et al., 1992). Despite the potential interactive effect of glyphosate and weed litter inputs, methods used to manage experimental plots might result in bias against detecting this kind of effect. In the studies by Sanogo et al. (2001), Bradley et al. (2002) and Nelson et al. (2002), the authors state that all field plots were maintained weedfree by manual weeding to reduce weed–crop competition. Most studies mention weed control strategies for establishing experimental plots, but do not report the extent of weed pressure at the time of herbicide application. For example, Njiti et al. (2003) reported that experimental plots were subjected to conventional tillage prior to the experiment and that all plots, including those receiving post-emergent applications of glyphosate, were treated with pre-emergence herbicides. We predict that investigators are underestimating the effect of glyphosate on crop disease, if such an effect exists, by reducing weed pressure to below a level that it would normally occur. In the one study where weed pressure was reported at the time of glyphosate application, glyphosate enhanced abundance of Fusarium spp. (Le´vesque et al., 1987). The presence of weeds in field trials is not a requirement to enhance Fusariumassociated diseases (Sanogo et al., 2001); however, positive responses might still underestimate responses under actual farming conditions. More research is required to determine whether this is a phenomenon of

the two particularly weedy fields sampled by Le´vesque et al. (1987) or whether this response can be generalised to other systems. The structure of the weed community is rarely taken into consideration in studies looking at glyphosate– Fusarium interactions, even though the indirect effect of glyphosate on crop diseases via inoculum build-up on weedy hosts might be dependent on the host ranges of those pathogenic fungi. In one exception, Le´vesque et al. (1987) observed plant species-dependent effects on the abundance of Fusarium spp. colonising various glyphosate-treated crop and weed species in a field survey. Fusarium spp., especially crop pathogens, have been observed infecting a broad range of weeds, especially weedy grasses and legumes (Farr et al., 2006). Based on observations of the host range of F. graminearum, incrop (for RR maize) or burndown applications of glyphosate might not be desirable for preventing higher levels of FHB in field plots dominated by grassy weeds, as these are more likely to support populations of F. graminearum than other herbaceous weeds (Farr et al., 2006). Fusarium solani var. glycines has been isolated only from soya bean, so its weedy host range is unclear; however, F. solani has a broad host range, isolated from a wide variety of grassy and herbaceous plants (Farr et al., 2006). Weed communities might further mediate these interactions by affecting the abundance of Fusarium antagonists. AM fungi have shown potential as antagonists of pathogenic Fusarium spp. (e.g. Filion et al., 2003); however, due to their obligately symbiotic nature and limits to their host range, they might be affected by the abundance and composition of the weed community (e.g. Harley & Harley, 1987). Herbicide timing

The timing of glyphosate burndown is an important determinant of pathogen inoculum at the time of planting for some plant pathogens (Fig. 3). Plantpathogenic fungi are frequently recovered from plant tissue in the early stages of decomposition and following inoculation of sterile plant tissue with field soil, but are eventually replaced by species that are better suited to decomposing recalcitrant organic matter (Garrett, 1981). Smiley et al. (1992) proposed that weedy plants and volunteer cereals act as a "green bridge# for Rhizoctonia solani, supporting the pathogen population during the time from the harvest of the prior crop to the sowing of barley. Glyphosate applied in the spring was associated with greater levels of Rhizoctonia root rot and lower yields relative to autumn application, especially when the spring application was delayed until 1–3 days prior to sowing barley. The authors

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

Glyphosate–Fusarium interactions 313

Spring planting

Inoculum potential

Disease threshold

A

Jul

Aug

Sept

Oct

Nov

Dec

Jan

Feb

Mar

B

C

Apr

May

Jun

Jul

Fig. 3 Hypothetical potential of Fusarium graminearum inoculum to initiate disease in spring wheat following maize depending on the timing of glyphosate burndown (A: autumn, B: early spring, C: mid-spring near planting), based on the "green bridge# hypothesis of Smiley et al. (1992). Inoculum potential increases gradually as F. graminearum colonises and occupies maize stalks. Rapid increases in inoculum potential are due to saprotrophic growth on weed biomass following burndown while declines are due to antagonistic interactions with other microorganisms.

hypothesised that, under zero tillage, dead plant material was quickly colonised by R. solani, allowing a buildup of inoculum, but that microbial succession did not occur in time to reduce inoculum levels under the late spring glyphosate application. In some cases, a 3-week window between burndown in the spring and sowing was enough to negate the stimulatory effect on root rot. In another system, 7 weeks between glyphosate-treatment of E. repens and sowing of barley was necessary to avoid negative effects associated with F. culmorum under particular laboratory conditions (Penn & Lynch, 1981). Fernandez et al. (2005) observed a positive correlation between FHB and application of glyphosate formulations during an 18-month window prior to measurement, much longer than the 4-month window between burndown and sowing required to completely negate stimulation of Rhizoctonia root rot (Smiley et al., 1992) and the 7-week window required to avoid F. culmorum damage (Penn & Lynch, 1981). This apparent difference might be due to environmental variables or differences in fungal biology between the pathogen species. It might also be an artefact of the survey design by Fernandez et al. (2005), as the survey results presented did not indicate how late in that 18-month period glyphosate was applied. In reality, a large proportion of fields sampled were probably treated with glyphosate in the spring, shortly prior to data collection and late in the stated 18-month period. Using data from table 6 in Fernandez et al. (2005), 67% of fields (395 of 589) had at least one glyphosate formulation applied in the 18 months prior to data collection.

The authors state "Most fields (average of 61% for the 4 years) did not receive any herbicide in the spring before planting, while the rest generally received a glyphosate formulation#. Therefore, it is likely that 230 fields (39% of 589), or 58% of the 395 glyphosatetreated fields, had glyphosate applied in the spring prior to planting, making it difficult to interpret how FHB responded to glyphosate when applied earlier. The extensive adoption of GR soya bean and, to a lesser extent, GR maize has led to the possibility of more frequent glyphosate applications in crop rotations that include GR varieties. FHB levels in wheat are enhanced following maize and reduced following soya bean (Dill-Macky & Jones, 2000). Fusarium graminearum colonises pith tissue within maize stalks late in the growing season and then proliferates within dead stalks, allowing for overwinter survival (Windels & Kommedahl, 1984). To determine whether glyphosate applied to GR maize further enhances FHB inoculum level, it must be determined whether Fusarium spp. associated with FHB are better able to colonise maize stalks or ward off attack by microbial antagonists following in-crop glyphosate application. In addition, timing of herbicide application may be an important factor. Fusarium graminearum is not readily recovered from pith tissue within stalks until later in the growing season, soon after senescence (Windels & Kommedahl, 1984). An in-crop treatment with glyphosate allows for greater weed biomass at the time of application than a residual pre-plant application, increasing potential for pathogen proliferation. This might result in greater Fusarium colonisation of maize stalks and a reduction in the amount of time for antagonistic microbial species to overcome pathogenic Fusarium spp. Fusarium graminearum might also be stimulated if glyphosate enhances colonisation by other Fusarium spp., such as F. verticillioides (formerly F. moniliforme), that are speculated to facilitate growth of F. graminearum by hastening decay within maize stalks (Windels & Kommedahl, 1984).

Recommendations for future studies At the present time, there is insufficient evidence to prove or disprove a link between glyphosate and crop diseases associated with Fusarium spp., so this area should receive high research priority, given the rapid and widespread increase in glyphosate use. We propose that researchers pay particular attention to aspects of microbial ecology and experimental design, as discussed below, when studying this potential linkage to accurately characterise the potential for glyphosate to influence crop disease and to make inferences beyond the parameters present in individual studies.

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

314 J R Powell & C J Swanton

Adequate consideration of microbial ecology

The abundance and activities of Fusarium spp. are determined by a variety of biotic and abiotic factors and these factors might adequately predict how crop diseases associated with Fusarium spp. are likely to respond to glyphosate at agronomically relevant spatial scales (regional levels to field levels). Further research is necessary to identify such predictors, as these would benefit growers by aiding decisions regarding weed management practices. Meta-analyses would be useful for this purpose, but greater consideration of these factors in individual studies is required. We have already discussed the potential importance of inoculum level, weed community abundance and competition, and herbicide timing in mediating glyphosate–Fusarium interactions. Researchers should also note the following additional parameters that have well-documented effects on Fusarium spp. and other pathogenic fungi, as these will probably be important for evaluating the contextdependent nature of glyphosate–Fusarium interactions. Soil biota are important components of pathogensuppressive soils. Microbial antagonists can reduce the severity of Fusarium-associated diseases and glyphosate might have positive or negative effects on the abundance and activity of these antagonists. However, predicting the effects of glyphosate on Fusarium-antagonist interactions is difficult, as the abundance and identities of the antagonistic biota can vary spatially and temporally and we know little about the responses of many of these antagonists to glyphosate. Characterisation and comparison of microbial communities from fields eliciting weak and strong responses to glyphosate might indicate components of these communities that inform the likelihood of glyphosate enhancing a particular Fusarium-associated disease. In addition to microbiological factors, a variety of physical and chemical factors are important in determining the conduciveness of soils to plant diseases (Ho¨per & Alabouvette, 1996), which might mediate whether and how glyphosate affects Fusarium. Soil texture and organic matter interact with climatic variables to influence water potential, which affects the development of plant diseases associated with Fusarium spp. (Ho¨per & Alabouvette, 1996). Glyphosate adsorbs to clay particles, soil organic matter and soil oxides and hydroxides (Vereecken, 2005), suggesting that the stimulating effects of glyphosate on Fusarium spp., via direct effects (Fig. 1B) and via effects on antagonists (Fig. 1E), would be reduced in soils with high levels of these glyphosate-adsorbing substances. Nutrient availability in soil, the rhizosphere and plant tissue might affect whether and the extent to which glyphosate stimulates Fusarium spp. For example,

germination of F. graminearum macroconidia is dependent on nutrient availability (Beyer et al., 2004). Effects of glyphosate on fungal growth and competitive outcomes differed among the type of nutrient medium that was used in each experiment (Wardle & Parkinson, 1990b, 1992). Busse et al. (2001) suggested that increased microbial respiration following glyphosate applied to field soil was the result of carbon utilization, because the contribution of the glyphosate molecule to N and P soil pools was inconsequential. However, P freed from glyphosate exuded into the rhizosphere might be significant given the rapid uptake of available P by plants and poor mobility in soil (Karandashov & Bucher, 2005). Some micronutrients are required by Fusarium spp. to initiate plant infection and disease development, with micronutrient concentration often interacting with other soil factors, such as pH and redox potential, to have seemingly idiosyncratic effects on disease responses (Ho¨per & Alabouvette, 1996). The chelating effects of glyphosate can reduce micronutrient availability (Glass, 1984), potentially reducing plant disease levels and suggesting that the stoichiometry of available nutrients could be more important than absolute nutrient availability for predicting disease responses to glyphosate. These physical and chemical soil variables could be measured at the time of the study or after the fact, from stored soil and plant tissue samples, then incorporated into multivariate analyses, along with the biotic variables, to identify predictors of glyphosate-associated pathogen responses. Tillage and residue management also influence inoculum levels of Fusarium spp., which then influences levels of disease and mycotoxins in wheat, but not necessarily maize crops (Munkvold, 2003). Therefore, disease responses to glyphosate might be dependent on tillage practices in any particular cropping system. In the study by Fernandez et al. (2005), glyphosate was only a significant factor for FHB in fields under minimum tillage. However, the lack of significance for conventional- and zero-tillage might be an artefact due to low sample size (reflecting the strong interdependence of glyphosate and tillage factors in such surveys), especially for zero-tillage where very few fields were categorised as not receiving glyphosate in the previous 18 months. Tillage might also influence pathogen responses indirectly by altering the phenology and composition of weedy host communities (Smith, 2006). Tillage also alters dynamics of plant residues and organic matter, affecting the biotic and abiotic factors discussed in previous paragraphs. Measuring responses under various tillage regimes at any particular site represents an opportunity for experimentally manipulating these biotic and abiotic factors under agronomically realistic conditions.

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

Glyphosate–Fusarium interactions 315

Careful consideration should also be given to the choice of pathogen response variables, as variables will differ in their response to additional factors, such as crop genetics for disease severity or litter incorporation depth for pathogen abundance. Spatial and temporal variability within fields also require consideration. For example, the FHB index based on the incidence and severity of diseased heads and percentage of kernels infected with Fusarium spp. (FDK) were weakly correlated in Fernandez et al. (2005). For the FHB index, sampling took place at the mid-milk to early-dough stage of crop development and subsamples from multiple points within the fields accounted for spatial variability. However, for percentage FDK, grain was sampled post-harvest and it is not clear whether spatial variability could have been controlled for to the same extent as for the FHB index. In addition, culture-based techniques for determining fungal abundance can result in highly variable estimates when compared with molecular approaches (e.g. Filion et al., 2003). Experimental design

Comparative studies will be valuable for the purposes of identifying factors that are correlated with, and likely indicators of, the potential for glyphosate to enhance Fusarium-associated diseases; however, experimental studies will be required to determine whether a causative link exists. The optimal experimental design will depend on the ultimate goal of the study. If the goal is to estimate the effect of specific weed management systems on disease levels or crop yields, the researcher would be best served by a design that compares a system that incorporates burndown and ⁄ or in-crop glyphosate

applications to other modes of weed management that are typical of the cropping system under study (e.g. Liphadzi et al., 2005). However, if the goal is to characterise the contextspecific nature by which glyphosate might affect diseases associated with Fusarium spp., studies will need to be conducted in a variety of environments and specific experimental treatments and controls will be necessary (Fig. 4). To address the role of pathogen inoculum level in glyphosate–Fusarium interactions, disease responses to glyphosate should be measured in field soil with little or no pathogen pressure (e.g. following a long rotation free of host crop species) and compared with those in the same field soil that had been moderately and heavily inoculated with pathogenic Fusarium spp. Inoculating soil at the main-plot level, with fewer independent replicates than the other factors, makes the confinement of pathogens to the inoculated soil more manageable. The design is also practical in that the rows are allocated to the "herbicide# factor, facilitating spraying, while the "weed# factor is manipulated across rows, reducing the total number of rows and the amount of space requiring hand weeding. Comparing pathogen responses to glyphosate in plots experiencing normal weed pressure, elevated weed pressure and hand-weeded checks should reveal the role of weed litter in mediating these pathogen responses. Additional treatments with other herbicides should be used to determine whether the observed phenomena are specific to glyphosate or can be generalised to other types of chemical weed control. Responses in the untreated control will indicate effects of weeds and inoculum level independent of herbicides. This design is appropriate to evaluate the effects of both burndown and in-crop glyphosate applications.

Crop rows

Handweeded

Fig. 4 Suggested split-plot layout to test for interactive effects among inoculum level, weed pressure and glyphosate on pathogenic Fusarium spp. Inoculum level is predicted to have the largest effect on Fusarium spp. abundance and disease levels and is manipulated at the main plot level. Manipulating weed pressure and herbicide factors at the subplot level increases the degrees of freedom associated with these factors.

Normal weeds Increased weeds

Low/no Fusarium Glyphosate

Moderate Fusarium

High Fusarium

Other herbicides

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

Untreated check

316 J R Powell & C J Swanton

The design is open to modification in order to answer additional research questions posed by the investigators; for example, both disease-resistant and susceptible varieties could be included to determine the extent that crop genotype mediates the responses. However, the core treatments presented here are necessary for the purposes we have described. Researchers should also record the variables described in the previous section (or retain material so that these variables can be estimated at a later date) and report estimates of variation along with treatment means, so that studies can be compared and combined into meta-analyses. Regardless of the study goal, considerable thought needs to be given to maximising the statistical power of the data analysis. Soils contain significant biotic and abiotic heterogeneity (Ettema & Wardle, 2002). Experiments should use all the tools available to them to take into account additional sources of variation, including blocking across environmental gradients, measurement of potentially important covariables, subsampling within plots and maximising replication within treatments. Our split-plot design allocates more replication to the "herbicide# and "weed# sub-plot factors, the effects of which are manifest indirectly and are likely to be weaker and more variable than the "inoculum# main-plot factor. However, even the most statistically powerful design will fail to measure the true effect if assumptions made during the management of research plots bias responses such that effects are over- or underestimated.

Conclusions There is little evidence from experimental field trials to support a causative link between glyphosate and elevated crop diseases associated with Fusarium spp. However, at the present time, we cannot disprove this linkage either. If a link between glyphosate and fungal diseases actually exists, observations of microbial ecology suggest that this link might be context dependent. Based on observations of fungal and disease ecology, future research needs to take into account additional factors in order to properly address whether (and when) concerns regarding glyphosate-pathogen interactions are appropriate. Initial inoculum level, weed abundance and community composition, fertility, agronomical practices, climate and soil chemical and physical properties are all potentially important factors that interact with glyphosate and need to be considered when designing field experiments to test for glyphosate effects on fungal diseases.

Acknowledgements We thank Rob Gulden, Mike Cowbrough and four anonymous reviewers for their suggestions on an earlier

draft of this paper. We gratefully acknowledge financial support from the Natural Sciences Research Council of Canada, as a postgraduate scholarship to J.R.P., from The Ontario Wheat Producers# Marketing Board, and from the Ontario Ministry of Agriculture, Food and Rural Affairs.

References ALTMAN J (1993) Pesticide–pathogen interactions in plant disease. In: Pesticide Interactions in Crop Production: Beneficial and Deleterious Effects (ed J ALTMAN), 315–332. CRC Press, Boca Raton, FL. ANDERSON JA & KOLMER JA (2005) Rust control in glyphosate tolerant wheat following application of the herbicide glyphosate. Plant Disease 89, 1136–1142. BAIRD R, BATSON W, WATSON C & HIGHTOWER P (2005) Evaluation of transgenic cotton varieties and a glyphosate application on seedling disease incidence. Mycopathologia 159, 481–486. BEYER M, RODING S, LUDEWIG S & VERREET JA (2004) Germination and survival of Fusarium graminearum macroconidia as affected by environmental factors. Journal of Phytopathology 152, 92–97. BRADLEY CA, HARTMAN GL, WAX LM & PEDERSEN WL (2002) Influence of herbicides on Rhizoctonia root and hypocotyl rot of soyabean. Crop Protection 21, 679–687. BUSSE MD, RATCLIFF AW, SHESTAK CJ & POWERS RF (2001) Glyphosate toxicity and the effects of long-term vegetation control on soil microbial communities. Soil Biology & Biochemistry 33, 1777–1789. DILL-MACKY R & JONES RK (2000) The effect of previous crop residues and tillage on Fusarium head blight of wheat. Plant Disease 84, 71–76. ETTEMA C & WARDLE DA (2002) Spatial soil ecology. Trends in Ecology & Evolution 17, 177–183. FARR DF, ROSSMAN AY, PALM ME & MCCRAY EB (2006) Fungal Databases [WWW document]. URL http://nt.arsgrin.gov/fungaldatabases/ [accessed on 26 February 2006]. FENG PCC, BALEY GJ, CLINTON WP et al. (2005) Glyphosate inhibits rust diseases in glyphosate-resistant wheat and soyabean. Proceedings of the National Academy of Sciences of the United States of America 102, 17290–17295. FERNANDEZ MR, SELLES F, GEHL D, DEPAUW RM & ZENTNER RP (2005) Crop production factors associated with Fusarium head blight in spring wheat in eastern Saskatchewan. Crop Science 45, 1908–1916. FERNANDEZ MR, ZENTNER RP, DEPAUW RM, GEHL D & STEVENSON FC (2007a) Impacts of crop production factors on Fusarium head blight in barley in eastern Saskatchewan. Crop Science 47, 1574–1584. FERNANDEZ MR, ZENTNER RP, DEPAUW RM, GEHL D & STEVENSON FC (2007b) Impacts of crop production factors on common root rot of barley in eastern Saskatchewan. Crop Science 47, 1585–1595. FILION M, ST-ARNAUD M & JABAJI-HARE SH (2003) Quantification of Fusarium solani f. sp. phaseoli in mycorrhizal bean plants and surrounding mycorrhizosphere soil using realtime polymerase chain reaction and direct isolations on selective media. Phytopathology 93, 229–235.

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

Glyphosate–Fusarium interactions 317

GARRETT SD (1981) Soil Fungi and Soil Fertility: An Introduction to Soil Mycology. Oxford University Press, New York. GLASS RL (1984) Metal complex formation by glyphosate. Journal of Agricultural and Food Chemistry 32, 1249–1253. HANSEN JM, STACK RW, ELISA E & MERGOUM M (2004) Effect of application of a sublethal dosage of glyphosate on FHB severity in spring wheat and durum. In: The 2nd International Symposium on Fusarium Head Blight (eds SM CANTY, T BORING, K VERSDAHL, J WARDWELL & RW WARD), 311– 314. Michigan State University, Lansing, MI. HANSON KG & FERNANDEZ MR (2003) Glyphosate herbicides affect plant pathogenic fungi. Canadian Journal of Plant Pathology 25, 120. HARIKRISHNAN R & YANG XB (2002) Effects of herbicides on root rot and damping-off caused by Rhizoctonia solani in glyphosate-tolerant soyabean. Plant Disease 86, 1369–1373. HARLEY JL & HARLEY EL (1987) A check-list of mycorrhiza in the British flora. New Phytologist 105, 1–102. HENRIKSEN B & ELEN O (2005) Natural Fusarium grain infection level in wheat, barley and oat after early application of fungicides and herbicides. Journal of Phytopathology 153, 214–220. HO¨PER H & ALABOUVETTE C (1996) Importance of physical and chemical soil properties in the suppressiveness of soil to plant diseases. European Journal of Soil Biology 32, 41–58. KARANDASHOV V & BUCHER M (2005) Symbiotic phosphate transport in arbuscular mycorrhizas. Trends in Plant Science 10, 22–29. KAWATE MK, KAWATE SC, OGG AGJ & KRAFT JM (1992) Response of Fusarium solani f.sp. pisi and Pythium ultimum to glyphosate. Weed Science 40, 497–502. KAWATE MK, COLWELL SG, OGG AG & KRAFT JM (1997) Effect of glyphosate-treated henbit (Lamium amplexicaule) and downy brome (Bromus tectorum) on Fusarium solani f. sp. pisi and Pythium ultimum. Weed Science 45, 739–743. KING SR & HAGOOD ES (2003) The effect of johnsongrass (Sorghum halepense) control method on the incidence and severity of virus diseases in glyphosate-tolerant maize (Zea mays). Weed Technology 17, 503–508. KREMER RJ, MEANS NE & KIM S (2005) Glyphosate affects soyabean root exudation and rhizosphere micro-organisms. International Journal of Environmental and Analytical Chemistry 85, 1165–1174. LARSON RL, HILL AL, FENWICK A, KNISS AR, HANSON LE & MILLER SD (2006) Influence of glyphosate on Rhizoctonia and Fusarium root rot in sugar beet. Pest Management Science 62, 1182–1192. LE´VESQUE CA & RAHE JE (1992) Herbicide interactions with fungal root pathogens, with special reference to glyphosate. Annual Review of Phytopathology 30, 579–602. LE´VESQUE CA, RAHE JE & EAVES DM (1987) Effects of glyphosate on Fusarium spp. – its influence on root colonization of weeds, propagule density in the soil, and crop emergence. Canadian Journal of Microbiology 33, 354–360. LIPHADZI KB, AL-KHATIB K, BENSCH CN et al. (2005) Soil microbial and nematode communities as affected by glyphosate and tillage practices in a glyphosate-resistant cropping system. Weed Science 53, 536–545. LIU L, PUNJA ZK & RAHE JE (1997) Altered root exudation and suppression of induced lignification as mechanisms of predisposition by glyphosate of bean roots (Phaseolus

vulgaris L.) to colonization by Pythium spp. Physiological and Molecular Plant Pathology 51, 111–127. LUO Y, HILDEBRAND K, CHONG SK, MYERS O & RUSSIN JS (2000) Soyabean yield loss to sudden death syndrome in relation to symptom expression and root colonization by Fusarium solani f. sp glycines. Plant Disease 84, 914–920. LUO Y, CHONG SK & MYERS O (2001) Spatio-temporal analysis of soyabean root colonization by Fusarium solani f. sp glycines in fields. Plant Disease 85, 303–310. LYNCH JM & PENN DJ (1980) Damage to cereals caused by decaying weed residues. Journal of the Science of Food and Agriculture 31, 321–324. MUNKVOLD GP (2003) Cultural and genetic approaches to managing mycotoxins in maize. Annual Review of Phytopathology 41, 99–116. NASS (2006a) Acreage. US Department of Agriculture, Washington, 43 pp. NASS (2006b) Agricultural Chemical Usage 2005 Field Crops Summary. US Department of Agriculture, Washington. 165 pp. NELSON KA, RENNER KA & HAMMERSCHMIDT R (2002) Cultivar and herbicide selection affects soyabean development and the incidence of Sclerotinia stem rot. Agronomy Journal 94, 1270–1281. NJITI VN, MYERS O, SCHROEDER D & LIGHTFOOT DA (2003) Roundup ready soyabean: Glyphosate effects on Fusarium solani root colonization and sudden death syndrome. Agronomy Journal 95, 1140–1145. OBOJSKA A, LEJCZAK B & KUBRAK M (1999) Degradation of phosphonates by streptomycete isolates. Applied Microbiology and Biotechnology 51, 872–876. PANKEY JH, GRIFFIN JL, COLYER PD, SCHNEIDER RW & MILLER DK (2005) Preemergence herbicide and glyphosate effects on seedling diseases in glyphosate-resistant cotton. Weed Technology 19, 312–318. PENN DJ & LYNCH JM (1981) Effect of decaying couch grass rhizomes on the growth of barley. Journal of Applied Ecology 18, 669–674. PLINE-SRNIC W (2005) Technical performance of some commercial glyphosate-resistant crops. Pest Management Science 61, 225–234. SANOGO S, YANG XB & SCHERM H (2000) Effects of herbicides on Fusarium solani f. sp. glycines and development of sudden death syndrome in glyphosate-tolerant soyabean. Phytopathology 90, 57–66. SANOGO S, YANG XB & LUNDEEN P (2001) Field response of glyphosate-tolerant soyabean to herbicides and sudden death syndrome. Plant Disease 85, 773–779. SCHAAFSMA AW, TAMBURIC-ILINIC L, MILLER JD & HOOKER DC (2001) Agronomic considerations for reducing deoxynivalenol in wheat grain. Canadian Journal of Plant Pathology 23, 279–285. SMILEY RW, OGG AG & COOK RJ (1992) Influence of glyphosate on Rhizoctonia root rot, growth, and yield of barley. Plant Disease 76, 937–942. SMITH RG (2006) Timing of tillage is an important filter on the assembly of weed community assembly. Weed Science 54, 705–712. STACK RW (1989) A comparison of the inoculum potential of ascospores and conidia of Gibberella zeae. Canadian Journal of Plant Pathology 11, 137–142.

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318

318 J R Powell & C J Swanton

STANGHELLINI ME, WAUGH MM, RADEWALD KC, KIM DH, FERRIN DM & TURINI T (2004) Crop residue destruction strategies that enhance rather than inhibit reproduction of Monosporascus cannonballus. Plant Pathology 53, 50–53. STRATTON GW & STEWART KE (1991) Effects of the herbicide glyphosate on nitrogen cycling in an acid forest soil. Water, Air & Soil Pollution 60, 231–247. TERMORSHUIZEN AJ & LOTZ LAP (2002) Does large-scale cropping of herbicide-resistant cultivars increase the incidence of polyphagous soil-borne plant pathogens? Outlook on Agriculture 31, 51–54. VEREECKEN H (2005) Mobility and leaching of glyphosate: a review. Pest Management Science 61, 1139–1151. WAN MT, RAHE JE & WATTS RG (1998) A new technique for determining the sublethal toxicity of pesticides to the vesicular-arbuscular mycorrhizal fungus Glomus intraradices. Environmental Toxicology and Chemistry 17, 1421–1428. WARDLE DA & PARKINSON D (1990a) Effects of three herbicides on soil microbial biomass and activity. Plant and Soil 122, 21–28.

WARDLE DA & PARKINSON D (1990b) Influence of the herbicide glyphosate on soil microbial community structure. Plant and Soil 122, 29–37. WARDLE DA & PARKINSON D (1992) The influence of the herbicide glyphosate on interspecific interactions between four soil fungal species. Mycological Research 96, 180–186. WINDELS CE & KOMMEDAHL T (1984) Late-season colonization and survival of Fusarium graminearum group II in maize stalks in Minnesota, USA. Plant Disease 68, 791–793. WISLER GC & NORRIS RE (2005) Interactions between weeds and cultivated plants as related to management of plant pathogens. Weed Science 53, 914–917. YANG XB & LUNDEEN P (1997) Occurrence and distribution of soyabean sudden death syndrome in Iowa. Plant Disease 81, 719–722. ZBOINSKA E, LEJCZAK B & KAFARSKI P (1992) Organophosphonate utilization by the wild-type strain of Pseudomonas fluorescens. Applied and Environmental Microbiology 58, 2993–2999.

! 2008 The Authors Journal Compilation ! 2008 European Weed Research Society Weed Research 48, 307–318