Recent reviews and meta-analyses have made considerable inroads into understanding patterns in the strength and prevalence of trophic cascades (Schmitz et al., 2000; Halaj and Wise, 2001; Shurin et al., 2002; Borer et al., 2005). These syntheses lend support to the conjecture (Strong, 1992) that trophic cascades are stronger and more prevalent in aquatic than in terrestrial systems (Shurin et al., 2006). Moreover, systems in which the size differences between consumers and their resources are great (Loeuille and Loreau, 2005) and/or where resources have a much shorter generation time than their consumers experience stronger cascades (Borer et al., 2005). Typical to studies in these reviews, the top predator is much larger than its prey and consumes more than one prey individual over the course of its lifespan (Lafferty and Kuris, 2002). Smaller organisms which don't fit this framework, such as parasites, parasitoids and diseases, also can induce cascades, but empirical data are limited primarily to a small set of cases in which introduced pathogens decimated mammalian herbivore populations with cascading consequences for vegetation (Sinclair, 1979; Dwyer et al., 1990). The long-term outcomes in such cases may differ from ‘typical’ trophic cascades because i) the consumers are typically smaller than prey, making them highly susceptible to vagaries of the abiotic environment (Gubbins and Gilligan, 1997), and ii) smaller organisms are limited in their spatial extent (Kotliar and Wiens, 1990).

Entomopathogenic nematodes (EPN) are widespread in soil on all continents except Antarctica (Hominick, 2002) and are potent natural enemies of many insects with soil-dwelling stages (Klein, 1990). EPN can protect vegetation (Strong et al., 1999; Denno et al., in press), are integral to rhizosphere food webs (Duncan et al., 2007) and are increasingly important as innoculative agents in integrated pest management (Lewis et al., 1998). Their broad host-range and ease of propagation and application make EPN promising candidates for classical or conservation biological control (Ehlers and Hokkanen, 1996; Lewis et al., 1998; Shapiro-Ilan et al., 2006). However, successful examples of persistent EPN-mediated control of insect pests in the field remain rare (Gaugler et al., 1997; Georgis et al., 2006). Biocontrol failures can result from a combination of biotic and abiotic factors that limit establishment, persistence and/or recycling. On the abiotic end, soil structure and moisture availability (often mediated by vegetation structure) affect EPN persistence and their subsequent impact. At the biotic end, variation in quality and abundance of hosts affects EPN reproduction and thus long-term persistence.

Highly variable abiotic conditions impose stresses that can drastically alter the ability of natural enemies to suppress prey in freshwater, marine and terrestrial systems (Briggs and Godfray, 1996). Spatial and temporal variability in environmental stressors can decouple the strong consumer-resource interactions that yield trophic cascades. In the soil environment, the primary abiotic stressors are extremes of temperature and moisture. Critical thresholds of soil moisture are essential for EPN movement, dispersal and ultimately survival; unlike many nematode clades, EPN lack an anhydro-biotic stage and are desiccation-intolerant (Womersley, 1990; Kaya and Gaugler, 1993; Grant and Villani, 2003b). During seasonally warm and dry periods, EPN populations can experience local extirpation in surface soils. Although soil aggregates that remain moist can serve as refugia in which EPN can persist under harsh conditions, EPN movement into such refuges is impeded as temperatures rise and soil moisture declines.

Biotic factors also play an important role in regulating EPN populations. Much like human and animal diseases, EPN can go locally extinct despite their strong impacts on host populations (Strong, 2002). Certain viral diseases, for example, can disappear when all susceptible individuals are either killed or immunized (Holmes et al., 1997). Field populations of EPN also can be extirpated if hosts are inaccessible or insufficient in density and/or size for EPN reproduction and population persistence. These dual risks of over- and under-exploitation, which are characteristic of many host-pathogen systems, combine to narrow the window of EPN persistence (Smith et al., 2003; Dugaw et al., 2005).

In this review, we summarize findings from 13 years of research on a naturally occurring, EPN-mediated subterranean trophic cascade in a California coastal prairie ecosystem. Through laboratory and field experiments and dynamic models, we have identified mechanisms that allow EPN to persist in heterogeneous habitats where abiotic conditions and host availabilities vary.

The Bodega Marine Reserve (BMR, Sonoma County, CA; 38.32°N, 123.07°W) has a Mediterranean climate with much of the rainfall restricted primarily to the winter months. With only 147 ha of land area, the reserve spans a range of soil types, from sand dunes to grassland areas in which loam and sand are mixed with organic matter ((Barbour et al., 1973). The dominant vegetative habitat on the reserve is coastal prairie composed primarily of a mix of native and exotic annual herbs. Naturally treeless, the site contains only four woody species: Baccharus pilularis, Ericameria ericoides, Lupinus chamassonis and, the most abundant in this coastal prairie, the yellow bush lupine Lupinus arboreus (Fabaceae). Yellow bush lupines are fast-growing nitrogen fixers that mature in two to three years and flower from early spring through summer, with seedset in the fall. Shrubs grow up to a height of three meters with an average life span of about seven years (Davidson and Barbour, 1977).

Foliar-feeding western tussock moths (Orgyia vetusta; Lepidoptera: Lymantriidae) can defoliate adult lupine bushes in outbreak years; however, such occurrences are localized, uncommon and do not ordinarily kill mature plants (Harrison and Maron, 1995). Numerous other insects attack lupine flowers and seed pods, thereby affecting total seedset; however, stem- and root-boring ghost moth caterpillars (Hepialus californicus; Lepidoptera: Hepialidae) are probably the most important cause of adult lupine mortality in this system (Strong et al., 1995; Maron, 1998). Because they damage important plant transport tissues such as xylem and phloem, even a few ghost moths can have a disproportionately large effect on an individual plant (Preisser and Bastow, 2006). Widespread lupine die-offs can lead to colonization of nitrogen-enriched soil by invasive grasses and forbs and result in species turnover and regime shifts of the plant community (Maron and Connors, 1996; Maron and Jefferies, 2001).

Although the potentially devastating impact of underground herbivory by Hepialus californicus larvae on Lupinus arboreus had long been recognized (Barbour et al., 1973), the existence and potential impacts of EPN on this interaction remained unknown until the early 1990s. EPN were discovered by chance at BMR when an undergraduate student in a field ecology course discovered a reddish-orange ghost moth larva in a lupine root that, although dead, did not appear to be decaying (D. Strong, pers. comm.). Closer examination revealed that the ghost moth larva had been infected by a previously undescribed species of heterorhabditid nematode. This species was named Heterorhabditis hepialus to signify its primary host at BMR (Stock et al., 1996). This name was preempted by Liu and Berry (1996), who, working without knowledge of the BMR discovery, isolated the same species from coastal marshes in Oregon and named it Heterorhabditis marelatus. Although two species of EPN, H. marelatus and Steinernema feltae (Filip-jev) (Steinernematidae) have been isolated from BMR soils (Gruner et al., 2007), S. feltiae is widespread but locally uncommon on the reserve. No infections of Hepialus by S. feltiae have been documented from the field, and competition experiments in the lab suggest H. californicus is not a favored host for S. feltiae (Gruner et al., unpublished data).

Early research linked widespread bush lupine die-offs (>10,000 mature individuals) to periodic outbreaks of ghost moth herbivory that removed root tissue and effectively girdled the plants (Strong et al., 1995). Strong and colleagues also noted that ghost moth caterpillars suffered high mortality from Heterorhabditis marelatus (=unnamed EPN at that time) and speculated that this species might be responsible for the patchily distributed nature of lupine mortality at BMR. A follow-up study (Strong et al., 1996) used a combination of lab work and field surveys to demonstrate that i) H. marelatus is found disproportionately in the soil surrounding lupine roots (the ‘rhizosphere’) at BMR, and ii) H. marelatus is an effective ghost moth predator capable of producing >420,000 offspring from a single infected H. californicus cadaver. They also documented an inverse correlation between EPN and H. californicus densities across the reserve and noted that areas with high EPN densities did not experience H. californicus outbreaks (Strong et al., 1996).

Subsequent field experiments demonstrated the causal mechanism of trophic cascades, confirming that H. marelatus could indirectly benefit L. arboreus through its suppression of underground H. californicus herbivory (Strong et al., 1999; Preisser, 2003; Preisser and Strong, 2004). Strong et al. (1999) exposed potted lupine seedlings to a range of H. californicus densities (0, 8, 16 or 32 caterpillars) in the presence or absence of H. marelatus. In the absence of H. marelatus, lupine mortality sharply increased as a function of H. californicus density; however, even high H. californicus densities did not affect lupine survival in the presence of H. marelatus. This was the first published research to confirm that trophic cascades occurred in belowground systems.

In a larger-scale field experiment, Preisser (2003) manipulated the presence/absence of H. marelatus underneath mature adult lupines exposed to a standard density of 24 H. californicus larvae/bush. The effect of H. marelatus on lupine fitness was both rapid and dramatic: H. marelatus presence increased lupine seed set by 44% in three months and trunk growth by 67% over an eight-month period. The large effect of H. marelatus documented by this research suggested that this cascade was capable of playing an ecologically important role in the BMR coastal prairie system, a result confirmed by dynamical models (Dugaw et al., 2004).

Although this work was generally interpreted as supporting a ‘food-chain’ approach to soil food webs, other research found no evidence that the densities of nematode-trapping fungi and H. marelatus were inversely correlated and therefore rejected the hypothesis of a four-level trophic cascade (Jaffee et al., 1996; Koppenhöfer et al., 1996). Subsequent lab microcosm studies showed that even large numerical responses of nematode-trapping fungi (Arthrobotrys oligospora and Myzocytiopsis glutinospora) were insufficient to regulate H. marelatus swarms issuing from cadavers (Jaffee and Strong, 2005; Jaffee et al., 2007).

EPN are patchy by nature (Stuart and Gaugler, 1994; Wilson et al., 2003; Stuart et al., 2006) and are likely to exist as metapopulations. To establish the patterns of occurrence of H. marelatus at BMR, we surveyed lupine bushes at six pre-established sites across the reserve, representing the range of habitats in which EPN could occur. Since 1993, we have continuously monitored presence/absence (hereafter incidence, sensu (Boag, 1993)) in a cluster of marked bushes at each site. The bushes were dispersed throughout each site to capture inherent variation. We assayed soil samples from each rhizosphere using standard Galleria mellonella baiting techniques (Bedding and Akhurst, 1975).

The study found that incidence of H. marelatus varied considerably among sites (Ram et al., 2008). Three of the sites (‘high sites’: Mussel Point, Cove and Dune) had high long-term EPN incidence while the remaining three (‘low sites’: Lower Draw, Upper Draw and Bayshore) had low long-term incidence. Incidence between the highest and lowest sites varied by an order of magnitude. These differences arose despite the lack of any apparent differences in lupine cover (Strong et al., 1996) or variation in ghost moth abundance (Strong, unpublished data). The surveys clearly indicated that H. marelatus populations wax and wane dramatically. They become locally extinct in lupine rhizospheres, colonize new rhizospheres and re-colonize rhizospheres, suggesting that local movement and dispersal are important for long-term persistence.

Several mechanisms could generate such patterns of dynamic persistence in the field. Here we review a set of mechanisms that could explain patterns of long-term occurrence observed at BMR.

Later research confirmed the primacy of soil moisture conditions in facilitating H. marelatus survival (Preisser et al., 2006). Heterorhabditis marelatus IJ buried under lupines experienced higher survival over the course of a typical BMR summer than those buried in nearby grasslands, a difference that disappeared under wet winter conditions. The seasonally varying protection afforded these IJ appears linked to soil moisture conditions; mature bush lupines possess a dense canopy and thick detrital layer that reduces desiccation and helps maintain a zone of relatively moist soil around their taproots. Models parameterized using this data showed that the moisture-driven increases in IJ survival in the lupine rhizospheres have far-reaching effects: some fraction of even a small cohort of H. marelatus IJ emerging into a lupine rhizospheres are likely to survive dry summer conditions, while the extinction of even large IJ cohorts in grassland soils is virtually guaranteed (Dugaw et al., 2005).

Ram et al. (2008) examined patterns of mortality at larger spatial scales by following a fixed cohort of H. marelatus in host-free tubes over one year at all six survey sites. This research enumerated surviving H. marelatus at three different time intervals (2, 6 and 12 months in the ground) to calculate mortality rates. Survival was greater under lupines than surrounding grasslands, corroborating previous work (Preisser et al., 2005, 2006). Nearly all of the sites studied had identical mortality rates despite considerable variation in incidence. Dune was the only outlier, with significantly higher mortality and the lowest moisture levels. Despite apparently harsh conditions for EPN survival, Dune paradoxically had the second highest long-term incidence of all the sites. The three high sites (Mussel Point, Cove and Dune) spanned the entire range of moistures from low to high. The low sites spanned a narrower range of moistures. These data present a paradox of persistence, where neither local abiotic conditions nor mortality predict local persistence.

The inherent patchiness and metapopulation structure of naturally occurring EPN populations may be a key to the persistence and stability of EPN-prey interactions (Stuart and Gaugler, 1994; Holyoak and Lawler, 1996; Wilson et al., 2003). The limited mobility of both EPN and host caterpillars, each interacting strongly and uniquely with abiotic and biotic properties of individual microsites, tends to reduce correlations among microsites within a landscape matrix. With increasing spatial scale, the signals from these microsites can be superimposed to yield phenomena with diminishing resemblance to the dynamics at microsite scales (De Roos et al., 1991). Since 2004, we have been monitoring the H. marelatus—H. californicus—L. arboreus interaction on a landscape matrix of 800+ lupine shrubs. Within a landscape where more than 70% of all monitored lupine rhizospheres were consistently occupied by H. marelatus, in excess of 50% of lupines died within the first two years after sustaining heavy root and stem herbivory by Hepialus californicus. Spatial autocorrelation functions showed that characteristic patch sizes of H. marelatus were larger than the scale of individual lupine rhizospheres, but that these dynamic ‘clouds’ of EPN swelled into grasslands during the wet season and retreated to lupine rhizospheres following dry summers (Gruner et al., unpublished data). The presence of H. marelatus was a poor predictor of lupine survival on the spatial scale of this landscape, where abiotic and biotic variables were not uniformly homogenous as in smaller-scale experiments.

Soil drying during the warm summer months may largely extirpate H. marelatus populations from surface soils and grasslands. Pockets of soil that remain moist may serve as refugia where EPN can persist under harsh summer conditions. Wet winter conditions increase soil moisture levels, making it favorable for EPN to disperse and seek hosts. Given their strong dependence on soil moisture for movement, EPN are able to seek hosts and reproduce only in the wet winter, while remaining inactive in dry summer months (Grant and Villani, 2003a, 2003b). As EPN cannot move and infect hosts in dry summer soil, they must survive through a dry summer in the absence of hosts. Given that EPN have limited dispersal (Ram, unpublished, McLaughlin and Strong, unpublished) and unknown phoretic range (Eng et al., 2005), a reasonable hypothesis is that EPN retreat into deeper, moist soil refugia during dry summer months. In one study, Heterorhabditis species occurred more frequently in deeper soil layers (>20 cm) than sympatric Steinernema species (Ferguson et al., 1995), but few field studies have tested soils deeper than 40 cm (but see Mauleon et al., 2006).

To test the plausibility of the hypothesis that EPN undergo vertical local migration in summer months, we surveyed EPN at different depths at each of our six study sites in summer 2006. Using a 15-cm diam. PVC pipe to prevent collapse and integration of shallow soils with deeper layers, we cored with vacuum extensions to 80 cm below the surface. We collected soil samples from the surface, and at 50 cm and 80 cm depths, at 10 replicate locations along a transect and measured incidence using standard baiting techniques (Bedding and Akhurst, 1975). When analyzed categorically by site groups (high- and low-incidence sites), the risk of infection at the surface was significantly higher in high sites as opposed to low sites (two sample, two-tailed binomial test: n = 180, df = 1, p < 0.0001, Fig. 1). High sites also had a significantly higher risk of infection at 50 cm and 80 cm below ground (two sample, two-tailed binomial test: n = 360, df = 1, p < 0.0001, Fig. 1). This difference remains significant even when the site with the highest historical population, Mussel Point, is removed from the analysis (two sample, two-tailed binomial test: n = 240/n = 360, df = 1, p < 0.0001; Fig. 2).

Fig. 1 Proportion of Galleria larvae infected at each depth (n = 120). No infections were recovered from 80 cm at any of the low sites.

Fig. 2 Map of long-term sites at the Bodega Marine Reserve (BMR). Incidence rates reported per site are from long-term surveys carried out from 1993–2006 (redrawn from Ram et al., in press). Lines point to the approximate center of each long-term site. (more ...)

Only one infected bioassay was recovered belowground from any of the low sites. At high sites, however, infection risk by deep EPN was proportional to or exceeded that at the surface. The presence of these deep EPN populations in the summer suggests that they may undergo seasonal vertical migration in the soil column. This migration may aid local population persistence (Mauleon et al., 2006). The cooler temperatures and increased moisture content of deep soils may allow EPN to conserve their lipid to later ascend up to the host-rich upper layers (‘A’ and ‘O’ horizons) with the onset of winter rains. We are currently testing this hypothesis, particularly how variation in lipid levels affect survivorship.

Entomopathogenic nematodes clearly play important roles in both managed and natural soil ecosystems. Within a single rhizosphere, favorable abiotic conditions facilitate EPN involvement in powerful trophic cascades. This effect, however, becomes much less clear at much larger spatial scales. Metapopulation processes such as colonization and extinction, facilitated by phoretic or individual movement and patchy abiotic refugia, sometimes override local limitations and may enable EPN to persist despite seasonal and variable conditions at micro sites. Even sites with favorable local conditions may not have strong EPN-driven impacts on hosts if at low levels of EPN colonization. Moreover, our longer spatially explicit studies have demonstrated that the mere presence of H. marelatus does not guarantee a strong effect on hosts or subsequent protection of lupines. Future empirical work and models on local and long-distance movement will provide improved mechanistic explanations of how populations—and their cascading impacts on insect hosts and plants—vary in space and time.