Wim H. van der Putten, NIOO-CTO, PO Box 40, 6666 ZG Heteren, the Netherlands (fax +31 26 4723227; e-mail email@example.com).
1The feedback between individual plants and their soil communities is a major driver of plant community processes. We analyse the rate at which plant–soil feedback develops in the root zone of the clonal dune grass Ammophila arenaria.
2Ammophila arenaria grows vigorously when it can form new roots in newly deposited windblown beach sand. The colonization zone is hypothesized to provide an enemy-free space, with root pathogens and parasites possibly contributing to degeneration of A. arenaria when deposition stops.
3We quantified root biomass and plant parasitic nematode densities in new and 1-year-old root zones, as well as in degenerate stands, in the field at monthly intervals. Each month, we studied the biomass production of Ammophila seedlings in controlled conditions in sterilized and non-sterilized soil from the different sampling sites, with and without nematicides.
4Colonization of newly deposited sand by the endoparasitic nematodes Heterodera arenaria and Pratylenchus spp. within 1 month of root formation coincided with a negative feedback in the bioassays. Nematicides counteracted growth reduction significantly, but their effectiveness decreased in soil samples collected later in the growing season.
5In older layers, roots and nematodes were already present before the sampling started. Growth reduction in unsterilized sand was observed at all sampling events, but was not counteracted by nematicides.
6In the field, root biomass of A. arenaria in the newly colonized sand increased throughout the growing season, despite the development of a negative plant-soil feedback.
7We conclude that the development of negative plant–soil feedback in the root zone of A. arenaria closely follows the colonization of newly deposited sand by roots and then by the endoparasitic nematode species Heterodera arenaria. There seems to be a shift in organisms causing this feedback during the first growing season, but root biomass nevertheless increases. However, in older root layers and in the degenerate stand, ongoing negative plant–soil feedback may contribute to a gradual decrease in root biomass.
Seed dispersal is an effective trait for escaping to non-infested sites when the root zone of parent plants has become severely infested by pathogens or parasites (Packer & Clay 2000). In the case of facultative saprotrophic pathogens, in particular there is little opportunity for selection of resistant genotypes, as an ability to continue to feed on organic matter even when entire seedling cohorts have been killed, make these pathogens highly aggressive (Jarosz & Davelos 1995). Clonal plants can also escape from pathogens through vegetative expansion, as demonstrated for systemic pathogens (Wennström & Ericson 1992; Frantzen 1994; Wennström 1994, 1999). The patterns of rhizome branching of clonal plants are known to differ under the influence of pathogens or mutualistic symbionts in the root zone. Thus, exposure to soil pathogens stimu-lates unidirectional rhizome development, resulting in an avoidance of pathogen-infested patches (D’Hertefeldt & Van der Putten 1998), whereas arbuscular mycorrhizal fungi cause intensified rhizome branching and patch exploitation (Streitwolf-Engel et al. 1997).
The escape of plants from root pathogens and parasites is, however, only temporary, as the soil organisms will eventually colonize any newly formed roots and for clonal plants it may last less than one growing season (De Rooij-Van der Goes et al. 1998). In the case of white clover (Trifolium repens), plant-parasitic nematodes were capable of tracking their host plants, although they were most prevalent near the centre of circular plant patches (Ennik et al. 1964). The dune grass Ammophila arenaria is sensitive to the accumulation of soil pathogens in its root zone. Plants are highly vigorous in freshly wind-blown beach sand, but growth conditions deteriorate within one growing season due to development of a negative plant–soil feedback (Van der Putten et al. 1988, 1989). When A. arenaria plants were covered with sand collected from a 1-year-old-root layer, fewer shoots were formed than when sterilized sand or fresh beach sand were used (De Rooij-Van der Goes et al. 1995b). These results suggested that a time-lag might occur in the development of a negative plant–soil feedback if vigour of A. arenaria is to be maintained. Therefore, in order to understand interactions between clonal plants and their root pathogens and root herbivores, more information is required on the timing of the development of pathogenicity following expansion of rhizomes towards previously un-colonized soil.
The benefit to clonal plants of expanding towards non-pathogen infested soil most likely depends on how long an enemy-free space exists. However, previous studies have focused only on the rate of root colonization by soil pathogens and parasites over the growing season without linking this to the development of a feedback from the soil community (De Rooij-Van der Goes et al. 1998). We linked a sequential field survey of the root zone of the clonal grass A. arenaria to plant–soil feedback trials in controlled conditions. As the role of nematodes in the plant–soil feedback of A. arenaria is not yet clear (Van der Putten & Troelstra 1990; Van der Putten et al. 1990; De Rooij-Van der Goes 1995; Little & Maun 1996) we first quantified numbers of plant parasitic nematodes and root mass present in samples of field soil. These samples were then used for feedback trials in controlled conditions, measuring biomass production of planted seedlings in sterilized and non-sterilized soil, with and without nematicides. This approach enabled us to link colonization patterns of plants and nematodes in the field to the development of a negative plant–soil feedback in both space (different stands and root zones of different ages) and time (different phases of the growth season).
We tested the hypothesis that clonal expansion of A. arenaria is followed, after a time-lag, by the development of a negative plant–soil feedback. We therefore examined the duration of the lag effect to obtain an indication of the effectiveness of clonal expansion in preventing colonization by plant parasitic nematodes from exerting a negative feedback. Immediate development of pathogenicity would suggest that colonizing freshly deposited wind-blown beach sand might provide A. arenaria with only a narrow ‘window’ for escape from its soil pathogens, while a time-lag implies that a longer period of the growing season could be used for root development and growth.
Material and methods
the study system
The sand dune grass Ammophila arenaria (L.) Link (marram grass) was used as a model, because plant parasitic nematodes may affect the spatio-temporal dynamics of this and other dominant dune species (e.g. Oremus & Otten 1981; Van der Putten & Troelstra 1990; Seliskar & Huettel 1993; Zoon et al. 1993; De Rooij-Van der Goes 1995; Little & Maun 1996). Ammophila arenaria, as well as its congener A. breviligulata, are vigorous as long as they are regularly buried by fresh wind-blown sand from the beach (Huiskes 1979; Maun 1998) but degenerate when sand deposition ceases further inland.
collecting and processing soil samples from the field
Soil samples were collected from both vigorous and degenerating stands at the coastal foredunes of Voorne, the Netherlands (51°52′ N, 4°04′ E) for nematode analysis and feedback trials. Samples were collected at monthly intervals from April 1997 up to November 1997 and again in April 1998. The seaward slope of the first dune ridge along the beach has vigorous A. arenaria, and here roots and root zone sand were collected from a layer deposited in autumn/winter 1995/1996 and colonized in spring/summer 1996 (the 1-year-old layer), as well as from a layer with sand deposited in autumn/winter 1996/1997 (the newly deposited layer, Fig. 1). Degenerating A. arenaria occurred on the landward slope and in the slack behind the first dune ridge, and roots and sand were collected from the top 20-cm layer, where active roots occur (Van der Putten et al. 1989).
For both sites, soil was collected from a 10-m-wide strip running parallel to the coastline for 100 m. At each sampling date, five random samples of 20 × 20 × 20 cm3 were collected (separately from the two layers in the vigorous stand), with additionally five random samples in April 1998 from sand deposited in autumn/winter 1997/1998 (the newest deposited layer). Each replicate sample was obtained from close to a randomly chosen tussock and once sampled, a tussock was not re-visited, so that samples from different layers and sampling dates were independent. A pit was dug near each tussock and the root layers (usually 20–40 cm deep) were sampled horizontally using the branching pattern and node staples of the buried shoots to identify the subsequent root layers (Van der Putten et al. 1989).
Roots were collected from every soil sample by sieving (mesh size 0.5 cm) and weighed before dividing into three weighed subsamples. From one subsample of roots, free-living nematodes were extracted by the funnel-spray method (Oostenbrink 1960), identified and counted: Tylenchus, Filenchus and Ditylenchus were considered as potential plant parasites (Yeates et al. 1993). Nematode genera that occurred in less than 10% of the samples from a specific layer (Boleodorus, Ecphyadophora, Telotylenchus, Helicotylenchus, Rotylenchus, Hemicycliophora and Criconema) were discarded from the analysis. Heterodera cysts and Meloidogyne root-knots were collected from the second subsample and counted, using a binocular microscope (10–15 × magnification). After the nematodes had been separated from the roots, both these subsamples were dried at 70 °C for 48 h, and weighed. The third subsample was cut into 1-cm pieces and mixed with a subsample of the sieved soil to serve as a source of inoculum for rhizosphere-inhabiting soil organisms. The proportion of roots in this inoculum was made equal to that in the original soil sample collected from the field.
The soil from every sieved field sample was gently homogenized. Free-living root-feeding nematodes were extracted from a 250-mL subsample by an Oostenbrink elutriator (Oostenbrink 1960), identified and counted as for the roots. Cysts of Heterodera spp. were extracted from a further (weighed) subsample of 1 L, by adding 4 L water, stirring the suspension and decanting the water plus the floating cysts over a 180-µm mesh sieve. This procedure of adding and decanting was repeated five times. About 50 g soil was weighed, dried at 70 °C for 48 h, and re-weighed to determine soil moisture content. Finally, about 6 kg of the homogenized soil was combined with a standardized amount of inoculum (see above) for use in growth experiments designed to measure plant soil feedback.
About 100 kg of ‘standard’ soil was collected from the succession stages before and succeeding degenerating A. arenaria on the first sampling date. This soil was sieved, homogenized, sterilized by means of gamma-irradiation (25 kGray), subdivided in 10 plastic bags and stored in the dark at 4 °C until further use.
monthly bioassay with three soil origins
Every month, a bioassay was carried out according to a fully factorial design with three factors, soil origins (SO; newly deposited and 1-year-old layers from vigorous A. arenaria, and degenerating A. arenaria), and with or without soil sterilization (s, ns), and nematicide addition (+, –). The five replicates of field soil were processed separately to give five independent samples for the bioassay. Half (3 kg) of each replicate was stored at 4 °C, while the remainder was sterilized by autoclaving at 120 °C (twice, for 1 hour, with a 48-h interval) and then stored at 4 °C until use. The period between sampling and starting the bioassay was always less than 1 week.
Half of each sterilized and non-sterilized soil sample was treated with the nematicide oxamyl at a dosage of 100 mg Vydate (10% active ingredient) kg−1 dry soil (Van der Putten et al. 1990), which was mixed with the soil. Each of the 1.5-kg soil samples was placed in a 1.5-L pot and moistened to 10% (w·w−1). Five extra pots of standard soil were included in each monthly trial to check for temporal variation in plant growth.
Prior to the start of each monthly bioassay, seeds of A. arenaria from the same foredune were germinated for 3 weeks on glass beads at a 16/8 h light/dark regime with corresponding temperatures of 25/15 °C. Four seedlings were planted in each pot, the soil surface was covered with tin foil to prevent desiccation and the pots were randomly placed in a climate-controlled growth chamber at a 16/8 h light/dark regime of 21/18 °C. Twice a week, the soil moisture content was reset at 10% (w·w−1) with demineralized water and a full strength Hoagland nutrient solution was added to all pots at a rate of 12.5 mL per week during the first 6 weeks and 25 mL in the final 2 weeks. All pots were harvested 8 weeks after planting. The soil was carefully washed from the roots, and debris was removed. Both shoot and root biomass were dried at 70 °C for 48 h and weighed.
Biomass production in the standard soil was significantly different between months, suggesting differences in conditions despite use of a controlled growth chamber. We therefore decided to use biomass production as a proportion of that in sterilized soil without nematicide for that bioassay. The data were tested for normality and homogeneity of variances and, when necessary, log-transformed to obtain homogeneity: data were analysed in a non-parametric Kruskal–Wallis test when no homogeneity was obtained. As each sampling event delivered independent samples, we carried out a three-way anova for each monthly bioassay testing the main effects of soil origin, soil sterilization and nematicide addition. Lack of homogeneity meant that interactions could not be calculated, and the effects of soil sterilization and nematicide addition were therefore analysed in two-way anovas, separately for each layer and month. Treatment means were compared using Tukey's HSD test (P < 0.05).
Root biomass in the field samples (g kg−1 dry soil) was analysed by one-way anova for each soil origin using month as independent factor, and means were compared using the LSD test (P < 0.05). Then, in order to determine if the degree of growth inhibition in the greenhouse experiment correlated with root mass in the field (as a measure of the amount of inoculum), a regression analysis was performed using the root mass in the field samples as X and total biomass in non-sterilized soil divided by total biomass in the sterilized soil as Y.
The nematode data were all expressed as numbers per 100 g dry soil, log(x + 1)-transformed when required to obtain homogeneity of variances, and analysed in a one-way analysis of variance with ‘month’ as factor. As all samples were independently collected, there was no need to use manova. Treatment means were compared using Tukey's HSD test (P < 0.05). To test for differences between the layers, Wilcoxon's matched pairs tests were performed, comparing the monthly average values of two soil origins. Furthermore, increases or decreases in the abundance of individuals of a species or a genus within a soil origin over time were compared by one-way anovas with month as independent variable. When necessary, data were log(x + 1)-transformed and means were compared using LSD tests (P < 0.05).
The biomass of A. arenaria plants grown in standard soil varied significantly between months (data not shown), indicating the difficulty of maintaining standard growth conditions. Time was not therefore included as a factor in the analyses. In a three-way analysis, each of the factors soil origin, soil sterilization and nematicide addition had a significant effect (P < 0.05) on biomass, but interaction effects could not be tested (see methods). The results of two-way anova's for individual months and soil origins are shown in Table 1.
Table 1. The results of a two-way analysis of variance with the factors soil sterilization (S) and nematicide addition (N) and the mean values of the total biomass (g pot−1) per treatment for the newly deposited and 1-year-old layers of vigorous A. arenaria and for degenerate stands. *P < 0.05; **P < 0.01; *** P < 0.001; – no significant difference. For each soil origin and each month the F- and P-value, and the results of Tukey's HSD test are presented. Different letters indicate significant differences
In the newly deposited layer soil, neither sterilization nor nematicide addition had a significant effect on plant biomass in April, May or June (Table 1, Fig. 2a), but from July onwards, the biomass of A. arenaria was significantly lower in unsterilized than in sterilized soil. From July through to September, the addition of nematicide to unsterilized soil counteracted this growth reduction, such that biomasses in ns+ and s+ were not significantly different, but this effect diminished in October and November (Table 1, Fig. 2a).
The 1-year-old layer showed a continuation of the growth reduction patterns observed at the end of the first growing season in the newly deposited layer; biomass of A. arenaria was generally higher in sterilized than in unsterilized soils, irrespective of nematicide addition or month (Fig. 2b). The pattern differed between months, but biomass differences between sterilized and unsterilized soils were significant in most cases (Table 1). Although not significant for every month, addition of nematicide always increased biomass when unsterilized soil without nematicide was considered, but only to a level significantly below that in sterilized soils with nematicide.
Biomass was again lower in soil collected from the root zone of degenerating A. arenaria, when it was unsterilized (Table 1, Fig. 2c). Addition of nematicide significantly increased biomass in unsterilized soil in June, August and September, whereas in April, July and August there were no significant differences between unsterilized and sterilized soils with nematicide added.
In April 1998, the biomass of A. arenaria was significantly lower in unsterilized than in sterilized soil from all origins (Table 2). However, in the newest layer, deposited in winter 1997–98, the significance of the sterilization effect was relatively weak and nematicide addition to unsterilized soil increased biomass to a level that was not significantly different from any other treatment. Nematicide addition significantly enhanced biomass production in the soil from the degenerate stand, but not in any soil layer from the vigorous stage (Table 2).
Table 2. Effects of soil sterilization (S) and nematicide addition (N) and the mean values of the total biomass (g pot−1) per treatment for the April 1998 sampling. All data were analysed by two-way anova, but data of the newly deposited and the 1-year-old layers were analysed by a Kruskal–Wallis non-parametric test, which does not allow testing for more than one factor at a time. *P < 0.05; **P < 0.01; ***P < 0.001;– no significant difference. For each soil origin the F- and P-value, and the results of Tukey's HSD test are presented. Different letters indicate significant differences
Root formation in the new sand layers differed considerably between years. In 1997, the first roots were formed in May (Fig. 3a), but in 1998 the first roots (0.054 g kg−1) were formed before April. In both years, the quantity of roots developed progressively. In 1997 root biomass in the new layer peaked in November (Fig. 3a), but then strongly declined, whereas root mass in the 1-year-old layer and in the degenerate site showed a more gradual decline (Fig. 3b,c). However, regression of root mass against growth reduction did not reveal any significant correlation (data not shown). The amount of root mass present in the soil does not therefore appear to affect plant performance in the bioassays, either directly by the effect of root decomposition on nutrient availability or indirectly by the accumulation of growth-inhibiting root pathogens.
In the newly formed layer, the numbers of nematodes differed significantly between months (P = 0.0011), owing to increasing nematode numbers towards the end of the growing season (Fig. 4). Individuals of Heterodera arenaria were first observed in July (Fig. 5a) and the females increased to a maximum of 2.5 per 100 g dry soil in November. The number of Heterodera juveniles was highest in November (P < 0.001). Juveniles of Meloidogyne were only present in April and from August onwards, and females were hardly present until the second half of the growing season (Fig. 5b). The density of Pratylenchus was low and variable, but did not vary significantly (Fig. 5c), whilst ectoparasitic nematodes tended to increase, especially between October and November (Fig. 5d).
In the 1-year-old layer formed in 1996, plant-parasitic nematodes were present throughout the whole year, and their total numbers did not differ significantly between months (Fig. 4). Heterodera females were present throughout the whole year, although new cysts were only formed between July and August (data not shown), whereas numbers of Heterodera males were significantly highest in August (Fig. 5e). Meloidogyne juveniles were already present in April and a second peak occurred in August, but this did not result in an increase in the amount of root knots (Fig. 5f). Both Pratylenchus and Paratylenchus were present early in the growing season, and their numbers tended to be highest in summer months (Fig. 5g). The population density development of Pratylenchus tended to lag behind that of Paratylenchus. Three species of ectoparasites occurred regularly in the soil samples: Filenchus was the most abundant, Ditylenchus was variable, and Tylenchus numbers were relatively low (Fig. 5h).
Significantly more plant-parasitic nematodes per unit of soil mass were collected from the root zone of degenerate A. arenaria than from the upper root layers in vigorous stands (Fig. 4) (P = 0.017 and 0.012 in comparison with newly formed and 1-year-old layers, respectively). The difference was mainly due to the relatively high numbers of ectoparasites (Fig. 5l). Although cysts of Heterodera did occur, these contained hardly any viable eggs (Fig. 5i), and no juveniles or males were present. Meloidogyne juveniles were collected, but no root knots were found (Fig. 5j), suggesting that juveniles did not develop into adult females. Pratylenchus numbers tended to be similar to those in the other soils, but Paratylenchus was not present (Fig. 5k). The increased density of ectoparasites was again mainly due to Filenchus, with variable numbers of Ditylenchus and Tylenchus (Fig. 5l vs. 5d).
The negative feedback on Ammophila arenaria from its soil appeared within 1 month of the formation of new roots in a freshly deposited layer of wind-blown beach sand. Initially, both soil sterilization and nematicide addition enhanced biomass production in unsterilized soil, suggesting that plant-parasitic nematodes were contributing to the growth reduction observed in the bioassays. In the newly deposited sand layer, the nematode density gradually increased and, by the end of the growing season, densities were comparable with those in previous studies (Yeates 1968; De Rooij-Van der Goes et al. 1995a). Growth reduction corresponded best with the occurrence and abundance of the cyst-forming nematode Heterodera arenaria and, less strongly, with the root-lesion nematode Pratylenchus spp. The ectoparasites Filenchus spp. and Ditylenchus spp. were already present at the start of the field survey when little growth reduction was measured in the bioassays. The root-knot nematode Meloidogyne maritima became more prominent later in the growing season, although by then the effectiveness of the nematicides had decreased. The decreased effectiveness of the nematicides continued in the 1-year-old layer and was also observed in soil from the degenerate A. arenaria stand. This suggests that certain plant parasitic nematodes play a key role in the development of a negative plant–soil feedback in the early colonization of a newly deposited sand layer by A. arenaria roots, while other soil organisms may take over as soon as the layer becomes more completely colonized and, finally, as the stands degenerate.
The bioassays indicated that clonal expansion of A. arenaria into newly deposited sand did not provide a long-term escape from pathogenicity. Nevertheless, in the field, root biomass in the newly colonized layer continued to increase throughout most of the growing season. Plants in the field must therefore be able to overcome the proposed negative feedback. The mass of shoots and below-ground stems (rhizomes) were not determined, but previous studies have demonstrated a mass gain in these plant organs over the growing season (De Rooij-Van der Goes et al. 1995b). The results of the bioassays therefore indicated the development of pathogenicity in the newly colonized sand layer, but they do not explain the ability of A. arenaria to continue producing new roots even when an inhibitory soil community has started to develop.
The hypothesis that colonization of newly deposited sand enables Ammophila to escape from soil pathogens has been challenged by Little & Maun (1996), who contended that arbuscular mycorrhizal fungi may counteract the effects of plant parasitic nematodes on the North American A. breviligulata. Indeed, arbuscular mycorrhizal fungi are known to protect plants from pathogenic fungi (Carey et al. 1992; Newsham et al. 1994, 1995a, b), as well as from plant parasitic nematodes (Francl 1993). Moreover, soil homogenization during the sieving, which disrupts hyphal networks (Helgason et al. 1998), may have had a negative effect on the infection potential of mycorrhizal fungi in our bioassays. We also supplied a nutrient solution that may suppress arbuscular mycorrhizal associations (Smith & Read 1997), and carried out the experiments for a relatively short period of 8 weeks, which may be only just enough (Little & Maun 1996) or just too short (J.P. Clapp, personal communication) for arbuscular mycorrhizal fungi to have an impact on plant growth.
Most likely, our results do not account for the full net plant–soil feedback, but focus more on its negative component, as caused by plant parasitic nematodes and plant pathogens (De Rooij-Van der Goes 1995; Kowalchuk et al. 1997). We may have included the effects of natural non-symbiotic antagonists, such as antibiotic-producing or chitinolytic microorganisms (De Boer et al. 1998a, b), but effects of arbuscular mycorrhizal fungi may have been excluded. In the field, these fungi can improve nutrient acquisition by buried plants (Perumal & Maun 1999) and both mycorrhizal diversity and infection levels are greater in vigorous than in degenerating A. arenaria (Kowalchuk et al. 2002). Therefore, our results complement, rather than contradict, those of Little & Maun (1996).
Previous studies have assumed that plant parasitic nematodes are involved in the negative plant–soil feedback in A. arenaria (Van der Putten et al. 1990; Van der Putten & Van der Stoel 1998), although some results have suggested that nematode densities are too low to allow them to make a major contribution to direct growth reduction (Van der Putten & Troelstra 1990; De Rooij-Van der Goes 1995). Similar conclusions have been drawn for the later successional dune shrub Hippophaë rhamnoides (Oremus & Otten 1981; Maas et al. 1983; Zoon et al. 1993). In the present study, the results of the nematicide treatment correlated with the colonization of the new root zone by endoparasitic nematodes (Heterodera arenaria and, to a lesser extent, Pratylenchus spp.). Glasshouse inoculation trials, however, showed that these nematodes are not very aggressive towards their host plant (C.D. Van der Stoel and E.P. Brinkman, unpublished results), as expected for biotrophic parasites (Lenski & May 1994).
The nematicide could be exerting indirect or non-target effects. As nematicides did not account for the same proportional growth increase in sterilized and non-sterilized soil, effects via the availability of soil nutrients can be excluded, but non-target effects remain possible. Alternatively, it could be that nematodes are involved indirectly in the development of a soil pathogen complex, e.g. as a vector for soil pathogens (Khan & Pathak 1993). This aspect of nematode ecology requires further investigation.
It is necessary to recognize the possible effects of sterilization on nutrient availability and thus plant growth (Troelstra et al. 2001). In the present study, soil sterilization tended to lead to more biomass of the planted seedlings, even in samples collected before any new roots had been formed. Interestingly, as soon as the first new roots were formed, performance in non-sterilized soil decreased significantly, although the magnitude of this growth reduction was not correlated with root biomass in the field samples. In newly deposited sand, the trend for a soil sterilization effect prior to the formation of new roots might have been a background effect, whereas the additional growth reduction recorded in the months after the first new roots had been formed may be attributable to the activity of soil pathogens. To date, however, there are no techniques to separate a possible side-effect of soil sterilization on the soil nutrient availability from that on soil pathogens, parasites, and symbiotic and free-living mutualists.
Singling out effects of nutrients, pathogens, parasites, mutualists and free-living antagonists complicates plant–soil feedback studies (Troelstra et al. 2001). We have demonstrated that a negative plant–soil feedback in the colonization zone of A. arenaria roots develops within 1 month of the formation of the first new roots. This leaves A. arenaria with very little enemy-free space, unless plants are protected against pathogens by other means, such as mutualistic symbionts (Little & Maun 1996) or other beneficial free-living soil organisms (De Boer et al. 1998b). Our results suggest that soil pathogens of A. arenaria, unlike some systemic pathogens above ground (Wennström & Ericson 1992; Wennström 1994, 1999), do not lose track of (part of) their host plant. The negative plant–soil feedback closely follows the colonization of the root zone by endoparasitic nematodes, although their role appears to be indirect.
We thank Lijbert Brussaard, Jan Woldendorp, Jeff Harvey, George Kowalchuk, Wietse de Boer and an anonymous referee for critical reading and for helpful comments on earlier drafts of the manuscript and Lindsay Haddon for editing. This is publication 3037 of NIOO-KNAW, Netherlands Institute of Ecology.