Plant Ecology

, Volume 193, Issue 2, pp 211–222

Seed characteristics and susceptibility to pathogen attack in tree seeds of the Peruvian Amazon

Authors

    • Center for Tropical ConservationDuke University
    • Department of Biological SciencesStanford University
  • Patricia Álvarez-Loayza
    • Center for Tropical ConservationDuke University
    • Department of Plant Biology and PathologyRutgers University
  • John Terborgh
    • Center for Tropical ConservationDuke University
Original Article

DOI: 10.1007/s11258-006-9259-4

Cite this article as:
Pringle, E.G., Álvarez-Loayza, P. & Terborgh, J. Plant Ecol (2007) 193: 211. doi:10.1007/s11258-006-9259-4

Abstract

Many studies now suggest that pathogens can cause high levels of mortality in seeds and seedlings. Recruitment from seed to sapling is an important bottleneck for many tree species, and if specialist or generalist pathogens have differential negative effects among species of juvenile trees, then they may have a significant impact on forest community structure. To explore the effects of differential pathogen attack among tropical tree species, we quantified pathogen attack on the seeds of 16 tree species from the southeastern Peruvian Amazon and asked which seed characteristics, including size, hardness, germination time and mode, shade tolerance, and fruit type, were most closely correlated with susceptibility to pathogens. Shade tolerance and seed weight were positively and significantly correlated with susceptibility to pathogen attack by ecological trait regressions (ETRs), and correspondence analysis indicated that there might be increased attack rates in species with brightly colored, pulpy fruits (often dispersed by primates). Only shade tolerance was significantly correlated with pathogen attack when the analyses accounted for phylogenetic relatedness between species. Thus, contrary to standard predictions of size-defense ratios, our results suggest that larger, shade-tolerant seeds tend to be more susceptible to pathogen attack than smaller, light-dependent seeds. Moreover, differential pathogen attack may shape seed community composition, which may affect the successful recruitment of adults.

Keywords

Light dependenceFungiPlant pathogensRecruitment limitationSeed dispersalSeed weight

Introduction

The role that plant pathogens play in natural forest communities is the subject of a growing body of research but remains poorly understood (Coley and Barone 1996; Gilbert 2002). Janzen (1970) and Connell (1971) proposed that host-specific seed and seedling predators could have a major effect on forest community composition and structure. Indeed, recruitment limitation as a result of high mortality among seeds and seedlings appears be an important phenomenon for many tree species (Clark et al. 1999), and these bottlenecks may play a particularly important role in tropical forests, contributing to their unique floral diversity (Hubbell et al. 1999). Although both animal and microbial predators could be important mortality agents for juvenile plants, knowledge of the specific effects of pathogens has lagged behind that of animals, partly because of ecologists’ limited understanding of pathogen dispersal and host-specificity in natural systems (Burdon 1987; Gilbert 2002). Previous studies that have looked at soil-borne pathogens in juvenile plants have generally shown significantly negative effects of both true fungi and oomycetes in a variety of ecosystems (Augspurger 1984; Crist and Friese 1993; Lonsdale 1993; Dalling et al. 1998; Packer and Clay 2000; Hood et al. 2004), and the few reports that have examined soil-borne pathogen effects on several species in parallel suggest that the severity of these effects also varies among species (Augspurger 1984; Augspurger and Kelly 1984; Burdon 1987; Dalling et al. 1998). Thus, host-specific pathogens could potentially play an important role in recruitment limitation and therefore in shaping forest communities. Moreover, if susceptibility to generalist pathogens, which may include Aspergillus spp., Penicillum spp., and Cephaleuros virescens (P. Álvarez-Loayza unpublished data), varies among species, then non-host-specific pathogens could also affect community composition.

Theory predicts that there will be trade-offs among offspring traits when a given amount of parental energy is expended in reproduction, and further, that these trade-offs will result in the correlation of certain traits (Moles and Westoby 2004). For example, small seeds generally have less physical protection and are shade intolerant, lacking the nutrient reserves needed to germinate successfully from deep in the soil or emerge above existing ground vegetation (Foster and Janson 1985; Mazer 1989; Pons 1992). In contrast, larger seeds usually have both thicker seed coats and greater reserves (Howe and Richter 1982; Westoby et al. 1996; Pearson et al. 2002; Moles and Westoby 2004). However, the production of larger seeds generally results in a smaller number of seeds per unit of parental investment (Smith and Fretwell 1974; Westoby et al. 1996; Henery and Westoby 2001), and host-specific predators may be more prevalent in larger-seeded species (Janzen 1969; Hubbell 1979).

Among the few studies that have addressed the effects of disease in natural plant communities, fewer still have looked at the effects of pathogens specifically on seeds (but see Crist and Friese 1993; Lonsdale 1993; Dalling et al. 1998; Schafer and Kotanen 2004). In addition, little is known about the relationships between seed traits and susceptibility to disease because these previous investigations have typically involved only one or two species. In one of the few exceptions, Augspurger and Kelly (1984) found no significant relationship between seed size and pathogen attack on new seedlings of 18 tree species in a wet forest in Panama. However, that study was confined to wind-dispersed species, all of which have relatively small seeds for tropical trees (Westoby et al. 1996).

In this study, we tested various predictions about the relationships between seed traits and susceptibility to disease, including: (i) seeds with longer germination times should be less susceptible because they must withstand longer exposure to pathogens; (ii) seeds with harder and/or thicker coats should be less susceptible; (iii) seeds whose germination modes result in retention of greater energy reserves should be less susceptible because of their greater ability to recover from attack; (iv) dispersal mode, which presumably selects for a variety of seed traits (Gautier-Hion et al. 1985; Hammond and Brown 1995), should be correlated with susceptibility; (v) shade-tolerant species should be less susceptible than light-dependent species because pathogens tend to be more prevalent in shaded, moister soil (Augspurger 1984); (vi) large-seeded species should be less susceptible than small-seeded species because large seeds typically have both greater energy reserves and harder and/or thicker coats; and (vii) as suggested by previous studies of the effects of light variation and fungal pathogens on the success of seedlings (Augspurger 1984; Augspurger and Kelly 1984; Hood et al. 2004), ungerminated seeds should also be more susceptible to pathogens under greater shade and be attacked less frequently in the presence of a broad-acting fungicide.

We designed a series of experiments to evaluate these relationships and potential trade-offs using seeds of 16 tree species from the southeastern Peruvian Amazon, analyzing the data with inter-specific ecological trait regressions (ETRs) and phylogenetically independent contrasts (PICs) (Felsenstein 1985). Our results represent a first step towards making predictions about the effect of pathogens on seed communities and how these effects may influence long-term forest composition.

Methods

Study site and species

The experiments were performed over a period of 7 months (September 2004–March 2005) in a greenhouse in a clearing at Cocha Cashu Biological Station, located in Manu National Park, Perú (11°51′ S, 71°19′ W; see Terborgh 1990 for description). Seeds of 16 species were collected and planted while the parent tree was still fruiting, and thus within about 2 weeks of landing on the forest floor. Species were chosen to represent a variety of taxonomic groups, dispersal modes, life histories, fruiting phenology, and abundance (Fig. 1). We included only species represented by at least five fruiting adults within the station’s trail system, with the exception of Swietenia macrophylla (mahogany), which was included with only two fruiting adults due to its distinction as a valuable, highly endangered timber species.
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-006-9259-4/MediaObjects/11258_2006_9259_Fig1_HTML.gif
Fig. 1

Tree species and characteristics, with diagram of tentative phylogenetic relationships. Seed weight represents a species average ± SD; shade tolerance is measured as number of saplings per adult per hectare; fruit type is a list of characteristics with the order: fruit color, pulp attachment score (0–3, None–High), dehiscence (I or D); and pathogen susceptibility is measured as the number of seeds observed with pathogens divided by the total number of seeds (n = 150 for each species). Note that in the case of S. mombin, seeds that suffered from pathogen attack frequently survived. For details see “Methods”

Seed preparation and experimental procedure

Immediately after collection, seeds were scrubbed clean of pulp and allowed to air-dry for 1–3 days before weighing or planting, a procedure that yields germination success similar to that of naturally dispersed seeds (Leiberman and Leiberman 1986; Palmeirim et al. 1989). In order to begin with the highest possible percentage of viable seeds, all seeds were examined visually for damage, and non-floating species were checked for pre-experiment insect infestation by placing them in water; only seeds appearing healthy and un-infested were planted.

Soil was collected from within tree plots surrounding the station (Terborgh et al. 2002). Seeds were placed individually in the soil so that approximately half of each seed remained exposed to the surface for easier observation, and 10 seeds from a single adult were planted per 9 × 9 × 9-cm open-topped bag of soil; a total of 150 seeds was planted for each species. Soil bags had holes to allow drainage and were watered every other day or with enough frequency to keep the soil consistently moist. Each species was maintained for 3 months, and observations of all seeds were made every 6 d, recording germination, seedling establishment, and presence or absence of pathogens as assessed by visual inspection.

After 3 months, seedlings were measured, and all seeds that had not germinated and could be found (>90% of the ungerminated seeds for each species) were inspected, scored as alive or dead according to a visual examination of embryonic tissue, and, if dead, assigned an apparent cause of death: externally visible pathogen attack; internal decomposition; failed germination; or insect damage that had occurred since planting. Seeds that were scored as “with pathogens” included seeds that had visible fatal or non-fatal attack before germination during the 3-month observation period and those that never germinated and were scored as decomposed at the end of the experiment. No distinction was made between true fungi, oomycetes, and bacterial pathogens for these experiments. The few seeds that could not be found at the end of the experiment were not included in the analyses. We considered the proportion of seeds attacked by pathogens to be a measure of susceptibility to pathogens.

Seed characteristics

We determined several seed characteristics for each species, including average fresh seed weight of ∼30 seeds, median germination time, germination mode, seed hardness, fruit type, and shade tolerance (Fig. 1, Appendix 1, and see below). The median number of days to germination was calculated for the seeds that germinated in each soil bag replicate and averaged for each species. Germination mode is described as either hypogeal, when the germinating seed retains its cotyledons below ground as stored energy, or epigeal, when the germinating seed is elevated above ground and sheds its seed coat to expose photosynthetic cotyledons. Seed hardness was used to estimate seed-coat thickness and evaluated by biting a subset of seeds of each species and ranking them from 1–4: 1 = falls apart easily, disintegrates; 2 = possible to break but pieces stay firm; 3 = possible to leave marks or crack with a lot of force; 4 = nearly impossible to crack.

Gautier-Hion et al. (1985) found that fruit color, dehiscence, and pulp type were significantly related to disperser choice in an Old World tropical forest, and these findings are supported by observations of dispersers in Manu (P. Álvarez-Loayza unpublished data). Thus, due to the difficulty of objectively identifying seed dispersers in a fauna-rich tropical forest, we used these three fruit traits to approximate mode of dispersal. According to Gautier-Hion et al. (1985), orange and yellow indehiscent fruits with juicy, sticky pulp tend to be dispersed by primates; red and purple dehiscent fruits with more easily detached aril pulp tend to be dispersed by birds; and brown indehiscent fruits with fibrous pulp tend to be dispersed abiotically or by terrestrial mammals. Each type of fruit was given a binary presence/absence score for each nominal category of color, dehiscence, and ranked score corresponding to the strength of attachment of fruit pulp to seeds during the efforts to completely clean them (0–3, None–High), and species were then treated as replicates in a design matrix that was analyzed by ordination (see “Data analysis”).

Finally, a continuous scale of juvenile shade tolerance was approximated by the number of tagged and identified saplings per adult in a 1 hectare plot of mature forest free of large gaps and adjacent to the station (Appendix 1); the tree and sapling plots were established in 1974–1975 and 1993, respectively (Terborgh et al. 2002). The number of saplings of each species growing under the forest canopy serves as an indication of the ability of juveniles to establish in the shade, and that number was divided by the number of tagged adults in the plot to correct for species abundance. The total number of saplings should be affected little, if at all, by total seed input because of the severity of recruitment limitation in tropical forests (Hubbell 1979; Hubbell et al. 1999).

Shade and fungicide experiments

Additional experiments to test the effects of light levels and fungicide on pathogen success were conducted with two of the study species, Spondias mombin and Sapium marmieri. Shade consisted of two layers of 2-mm mesh hung 10 cm above germinating seeds. Fungicide treatments were conducted with Captan (N-trichloromethylthio-4-cyclahexene-1,2-dicarboximide), which is active against both true fungi and oomycetes (G.S. Gilbert personal communication). Seeds were soaked in 1 g/l Captan solution for 3 min before planting, and seeds and soil were sprayed with the same solution every 2 weeks thereafter. Shaded and unshaded control seeds were mock-treated with water.

Data analysis

Statistical analyses were performed using the statistical software JMP IN 5.1.2 (SAS Institute 2004), except as described below. Seed weights are reported ± SD, and all other experimental means are reported ± SE. Nonparametric tests were used when soil bags were considered separately instead of being averaged by species in order to handle the large numbers of zeros without violating the assumptions of ANOVA; ANOVA was used for all other tests of significance except as noted. In all tests that assume normality and homogeneity of variance, seed weights, and the fractions of seeds attacked by pathogens were log-transformed, and the data from the additional experiments on S. mombin and S. marmieri were arcsine-transformed.

Explained variance of multi-variate species comparisons was calculated using step regression in a general linear model, but because both seed weight and light dependence, which we treat as variables in the x dimension, are measured with error, the assumptions of model I-type linear regression are not fully satisfied, and the slopes of the predictive equations are inaccurate. Thus, to minimize the residual variance in both x and y and calculate the proportional relationship between them (Sokal and Rohlf 1995), we also performed model II-type Standardized Major Axis (SMA) regressions using the program (S)MATR (Falster et al. 2003).

Multiple correspondence analysis (MCA) ordination was performed on a binary design matrix of fruit traits using CANOCO 4.5 (Ter Braak and Smilauer 2002), fitting a unimodal model and biplot scaling with a focus on inter-species distances. Fruit traits were designated “species” in the CANOCO matrix, and species’ fruits were designated as “samples”, i.e., replicates. Frequency of pathogen attack was then described as one of three unique categorical variables: low (<mean − SD/2), medium (mean ± SD/2), or high (>mean + SD/2). Pathogen attack was then added as a supplementary variable in the design matrix and analyzed as part of the MCA in order to obtain an estimate of the relationship between pathogen attack and fruit traits, a technique known as predictive mapping (StatSoft 2006). The ordination diagram of the first two axes produced from the MCA output illustrates both species values, which approximate inter-species covariances, and the χ2 distances between these values, which approximate the similarity of each value to the next. Apeiba membranacea was removed as a color replicate because it was the only black fruit.

Finally, in addition to ETRs, in which pairs of traits were assigned coordinates corresponding to each species, we also analyzed PICs of traits of interest to correct for phylogenetic relatedness. A phylogeny of the species was assembled in the program Phylomatic (Webb and Donoghue 2005) using data from the Angiosperm Phylogeny Group tree (Fig. 1). Species that were not recognized by the program were added as polytomies (unresolved relationships) by hand according to Gentry (1996); three of these species, which are members of the family Bombacaceae (Matisia spp. and Quararibea witti), were identified as congenerics by Gentry and are treated as such in our analysis (Fig. 1). PICs of log-transformed data were calculated using the software Phylocom (Webb et al. 2004; see Moles et al. 2005 for description of methods), which can analyze trees with polytomies; branch lengths were designated equal. Contrasts were then analyzed as linear regressions, except that lines of fit were forced through the origin (Garland et al. 1992).

Results

Germination success averaged 43% over all species. The proportion of seeds attacked by pathogens ranged from 0 to 0.75 (Fig. 1) and varied significantly among the species analyzed (Wilcoxon Rank Test, P < 0.0001). Among soil bags in which at least one seed was attacked by pathogens, the total number of seeds attacked per bag exhibited a downward-sloping distribution (data not shown), suggesting that the probability of attack for each seed was independent of whether a pathogen was attacking another seed in the same bag. Visible pathogen attack led to mortality in ∼90% of all cases, with the exception of seeds of S. mombin, which often suffered a brief period of attack by a locally common saprophyte but later produced healthy seedlings.

Relationships between seed characteristics and pathogen susceptibility

Several of the seed traits that we tested were not significantly correlated by ETRs with pathogen susceptibility. These included median germination time (P = 0.13), seed hardness (P = 0.3), and germination mode (P = 0.08), although it should be noted that the mean proportion of seeds attacked was lower for seeds with hypogeal germination (0.09 ± 0.04) than for those with epigeal germination (0.33 ± 0.14).

However, fruit traits were related to susceptibility to pathogens by MCA ordination. The first four axes of the MCA explained 77% of the variance in our trait data (Table 1), which contributed to a total inertia, or Pearson χ2 for the design matrix divided by the total number of observations, of 2.5. Eigenvalues for the first two axes that are closer to 1 than to 0 indicate a relatively high degree of correspondence between traits (Leps and Smilauer 2003); these values especially reflect traits that are distant from the origin and cluster together. A plot of the first two axes (Fig. 2) showed that high indices of pathogen attack clustered around the negative end of the first axis with fruit traits that are associated with dispersal by primates: orange/yellow fruits; indehiscence; and juicy pulp, as approximated by high adherence of pulp to the seed (categories 2 and 3). In addition, red/purple fruits, dehiscent fruits, and intermediate pulp adhesion (fruit traits associated with dispersal by birds) clustered around the positive end of the first axis, and, although slightly less clustered, brown fruits and zero pulp adhesion (fruit traits associated with dispersal by wind or terrestrial mammals) fell along the second axis near low pathogen attack. The first two axes for fruit traits explained 26% of the variation in high pathogen attack, 58% of the variation in medium pathogen attack, and 55% of the variation in low pathogen attack; these percentages reflect the fractions of the weighted regression sums of squares of each variable over the total sums of squares for all the traits.
Table 1

Summarized results of the MCA of fruit traits and pathogen susceptibility

 

Axis 1

Axis 2

Axis 3

Axis 4

Eigenvalues

0.657

0.522

0.416

0.34

Cumulative percentage variance of trait data

26.3

47.1

63.8

77.4

https://static-content.springer.com/image/art%3A10.1007%2Fs11258-006-9259-4/MediaObjects/11258_2006_9259_Fig2_HTML.gif
Fig. 2

Trait values and inter-trait distances for the first two axes obtained in an MCA. The first axis is horizontal; the second axis is vertical. Gray circles indicate groups of traits discussed in the text. 0–3 pulp = strength of attachment of fruit pulp to seeds, see “Methods”; colors (brown, green, orange/yellow, red/purple) = fruit color; D = dehiscent; HPA = high pathogen attack; I = indehiscent; LPA = low pathogen attack; MPA = medium pathogen attack

Interestingly, both shade tolerance and seed weight were positively and significantly correlated with the frequency of detectable pathogen infection over the 16 species (Fig. 3). SMA regression led to steeper slopes for the ETRs between pathogen susceptibility and shade tolerance or seed weight (0.20 ± 0.14 and 0.38 ± 0.18, respectively) than were calculated through linear regression (0.11 and 0.23, respectively). Stepwise regression (Forward direction, P to enter = 0.15) showed that these two variables, along with a non-significant interaction effect (P = 0.11), accounted for 66% of the variation in the frequency of pathogen attack (Table 2). In order to investigate whether these relationships were conflated by evolutionary relationships between species, we also calculated PICs for susceptibility to pathogens, seed weight, and shade tolerance, and regressed contrasts for susceptibility against those for seed weight and shade tolerance. There was a significant, positive correlation between PICs for shade tolerance and pathogen attack (P < 0.02), but the positive correlation between PICs for seed weight and pathogen attack was not significant (P = 0.08).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-006-9259-4/MediaObjects/11258_2006_9259_Fig3_HTML.gif
Fig. 3

SMA regressions of the positive ETR relationships between susceptibility to fungal attack and (a) seed weight (P < 0.005) and (b) shade tolerance, estimated by the number of saplings per adult (P < 0.01). Arrows indicate outlier species, which were included in the regression analysis, but which may have special properties (see “Discussion”)

Table 2

Full model statistics for the effect of seed weight and saplings per adult as a proxy for shade tolerance on the proportion of seeds attacked by pathogens (data grouped by species)

Fixed effect

df

F

P

Explained variation (R2)a

Seed weight

1

12.37

0.0042

0.35

Saplings/adult

1

10.69

0.0067

0.22

Seed weight × saplings/adult

1

3.06

0.1059

0.09

aR2 values indicate the contribution of each effect to the model’s total explained variation (R2 = 0.66)

Effects of light levels and fungicide

Consistent with previous experiments on seedlings (Augspurger 1984; Augspurger and Kelly 1984; Hood et al. 2004), we found a higher frequency of pathogen infection among ungerminated seeds in the shade than in the higher-light control and fungicide treatments for both species tested (Fig. 4). Approximately 6.0 ± 2% of S. marmieri seeds were attacked in the shade treatment, which was significantly higher than the zero seeds attacked by pathogens in both the control and fungicide treatments (Student’s t-test, P < 0.001). Similarly, in S. mombin, 26 ± 6% of seeds were attacked by pathogens in the shade as opposed to 21 ± 4% in the higher-light control, and an even smaller proportion of seeds was attacked in the fungicide treatment (11 ± 3%). However, the only statistically significant difference for S. mombin was between the mean frequency of attack in the shade and fungicide treatments (Student’s t-test, P < 0.05).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-006-9259-4/MediaObjects/11258_2006_9259_Fig4_HTML.gif
Fig. 4

Proportion + SE of Spondias mombin (black) and Sapium marmieri (hatched) seeds with pathogens under shade cloth, control greenhouse conditions, or fungicide treatment (for each treatment, n = 15 soil bags, each with 10 seeds). Letters indicate statistically significant differences among means of arcsine-transformed data by Student’s t-test comparisons (P < 0.05 for S. mombin; P < 0.001 for S. marmieri). Note that in the case of S. mombin, seeds that suffered from pathogen attack frequently survived (see “Results”)

Discussion

In this study, we examined the relationships between seed traits and susceptibility to pathogens. Contrary to prediction, we found that there was no significant relationship between either germination time or seed hardness and susceptibility to pathogens, that shade-tolerant seeds were generally more susceptible than light-dependent seeds, and that larger seeds were generally more susceptible than smaller seeds. However, in accord with prediction, we found that hypogeal germinators had a lower mean susceptibility than epigeal germinators, that certain fruit traits were related to pathogen susceptibility, and that seeds of both S. mombin and S. marmieri were more susceptible to pathogens in the shade.

Germination time, seed hardness, and germination mode

Overall germination success was relatively high, perhaps because of the favorable growth conditions (relatively high-light levels, regular watering) provided by the greenhouse. We expected to find that seeds that germinate more slowly and have harder endocarps were better defended against pathogen attack, but we found no correlation between these traits. Although the correlation between germination mode and pathogen attack was not statistically significant, the mean susceptibility of hypogeal germinators was less than that of epigeal germinators. Epigeal species tend to germinate quickly and may thus invest little in immediate pre- and post-germination defense (Coley et al. 1985), whereas hypogeal germinators have substantial ability to recover from early seedling damage due to their withheld energetic reserves (Green and Juniper 2004). Our data suggest that hypogeal seeds themselves may also be more resistant to attack. Additional experiments with higher sample sizes will be necessary to determine whether this relationship is indeed significant.

Fruit type

The relationship between fruit type (and thereby presumed primary disperser) and susceptibility to pathogen attack may be caused by relationships between unidentified common qualities of these seeds. For example, dispersers may select for seed shape (Howe and Vandekerckhove 1981), and differences in surface:volume ratios or in surface texture may affect pathogen success. However, Lambert (2001) found that a primate-dispersed tree species benefited from dispersal in part because seeds that fell directly from the parent tree with sticky pulp, typical of primate-dispersed species, had much higher rates of fungal attack than seeds that had been spat out by dispersers. Thus, our result that primate-dispersed seeds tended to be more susceptible to pathogen attack may simply indicate differences between natural dispersal and even the most careful experimentally imposed dispersal, in which some pulp may have remained.

Light dependence and seed weight

Although we predicted that shade-tolerant seeds would need to be better defended because they germinate in shaded soil where pathogens may be more prevalent, the positive ecological and evolutionary correlation between high-light dependence and low susceptibility to pathogen attack is probably advantageous to gap-colonizing species, as their seeds may need to persist for long periods in shaded soil before they encounter high-light conditions (Pons 1992). In addition, the positively correlated ETRs between seed weight and pathogen attack suggests either that seed weight itself plays a role in pathogen susceptibility or, perhaps, that there are selective pressures for gap colonizers to produce seeds that are small as well as pathogen resistant. The relationship between seed weight and susceptibility remained positive but lost statistical significance when the relationships were analyzed using PICs. However, the low baseline rates of pathogen attack may mean that this correlation would be difficult to detect reliably with PICs without larger sample sizes and a better resolved phylogenetic tree.

Previous observations of soil seed-bank composition have shown that light-dependent, pioneer species and, to a lesser extent, smaller-seeded species are over-represented in comparison to shade-tolerant, primary-forest, larger-seeded species (Hopkins and Graham 1987; Garwood 1989; Thompson 1992; Dalling et al. 1997; Bekker et al. 1998; Hulme 1998; Moles et al. 2000; Pearson et al. 2002). Although this phenomenon can be explained in part by the greater numbers of smaller, light-dependent seeds produced, it also seems possible that the greater resistance of many light-dependent seeds to pathogen attack and decomposition contributes to this effect.

Pearson et al. (2002) proposed that larger seeds are better defended than smaller seeds against both seed predators and pathogens because they have thicker seed coats. However, this proposal was based on a study exclusively of pioneer species. Among groups of species that span a wider range of light dependence, there is equivocal evidence on the relationship between seed size and predation (Moles et al. 2003), and we have shown here that seed size may be positively correlated with susceptibility to pathogen attack. Also in contrast to our results, Dalling et al. (1998) found a >90% mortality rate in two species of pioneers, which was attributable in large part to attack by fungal pathogens. However, these experiments were carried out with seeds that were 80- and 500-fold smaller, respectively, than the smallest seeds in our experiments, and were also completely buried in soil. There is preliminary evidence that there may be different frequencies of pathogen attack on buried seeds than on seeds exposed to the surface among the species tested in our study (P. Álvarez-Loayza unpublished data). Further experiments that include an even larger range of seed sizes and comparison of attack on seeds at different depths in the soil are required to begin to understand the full complexity of seed–pathogen interactions.

Effects of light levels and fungicide

As expected, we found higher frequencies of pathogen attack in the shade, suggesting the interesting possibility that there may be spatial heterogeneity in the importance of seed pathogens in the forest. In addition, the marked seasonality in Manu, with ∼95% of the yearly 2000 mm of rain falling between November and May (Terborgh 1990), and a concomitant increase in cloudy days during these months, may also contribute to temporal variation in pathogen effects. The reduction of pathogen attack on S. mombin seeds through the application of fungicide suggests that a significant proportion of the pathogens that attacked our seeds were either fungi or oomycetes.

Generalist vs. specialist pathogens

Schafer and Kotanen (2004) found that the lethality of seed fungal pathogens was highly dependent on the specific plant–pathogen combination because only some seeds were hosts to specialist pathogens. In our experiments, we were unable to isolate the pathogens, and we were thus unable to differentiate definitively between specialist and generalist pathogens. However, there were two sets of unique-looking disease symptoms consisting of pinkish, thick-stemmed hyphae, resembling Rhizostilbella hibisci, and yellow, fuzzy hyphae, resembling Aspergillus spp., that appeared solely on Leonia glycycarpa, or Oxandra acuminata, respectively. The fact that these symptoms were unique to these two species could indicate that these pathogens were more specialized, and it is interesting to note that L. glycycarpa and O. acuminata had some of the highest overall rates of seed mortality due to pathogens (Fig. 1) and were in fact outliers in the regression of pathogen attack by seed weight (Fig. 3a). It is also possible that our experiments may have underestimated levels of pathogen attack in the field because the greenhouse provided a refuge from specialist pathogens that would have attacked in situ.

Conclusions

In this study, detectable pathogen attack on ungerminated seeds of 16 tree species almost always resulted in mortality; in fact, S. mombin was the only species whose seeds regularly managed to escape mortality after apparent attack. Thus, we conclude that pathogens are potentially important agents of seed mortality in forest communities. Unlike the effects of specialist predators and pathogens, which may contribute to spatial separation among conspecifics by intense effects near adult trees (Janzen 1970; Connell 1971), especially of more common species (Leigh et al. 2004), generalist predators and pathogens with differential effects may increase spatial and temporal heterogeneity in forest species composition in less predictable or ordered ways. Future experiments addressing the relationships between physical and chemical plant defenses and pathogen attack, as well as the specificity and distribution of pathogens in the field, especially when phylogenetic relationships are considered in the experimental design, should eventually shed more light on the impact of soil pathogens on plant recruitment and survival.

Acknowledgments

The Instituto Nacional de Recursos Naturales in Perú kindly granted the permits to conduct this research. We are grateful to N. Quinteros, D. Osorio, and V. Swamy for helpful conversations in the field. The manuscript benefited from comments by J. Hille Ris Lambers, B.S. Mitchell, J.R. and R.M. Pringle, and S. Ramírez. We also thank S. Ramírez and S. Russo for help with the independent contrast analyses and J. Lancaster for help with the ordination analysis. Thanks to J. White, K. Seifert, and B. Gerald for help with fungal identifications. This work was funded in part by the Andrew Mellon Foundation.

Copyright information

© Springer Science+Business Media, Inc. 2007