Size asymmetry in intraspecific competition and the density-dependence of inbreeding depression in a natural plant population: a case study in cassava (Manihot esculenta Crantz, Euphorbiaceae)

Authors

  • B. PUJOL,

    1. Department of Population Biology, Centre d‘Ecologie Fonctionnelle et Evolutive (CEFE), Montpellier Cedex, France
    2. Department of Plant Sciences, University of Oxford, South Parks Road, Oxford, UK
    Search for more papers by this author
  • D. MCKEY

    1. Department of Population Biology, Centre d‘Ecologie Fonctionnelle et Evolutive (CEFE), Montpellier Cedex, France
    Search for more papers by this author

Benoît Pujol, Department of Plant Sciences, University of Oxford, South Parks Road, Oxford, Ox1 3RB, UK.
Tel.: +44 (0) 1865 2750 75; fax: +44 (0) 1865 2750 74;
e-mail: benoit.pujol@plant-sciences.oxford.ac.uk

Abstract

The effects of competition on the genetic composition of natural populations are not well understood. We combined demography and molecular genetics to study how intraspecific competition affects microevolution in cohorts of volunteer plants of cassava (Manihot esculenta) originating from seeds in slash-and-burn fields of Palikur Amerindians in French Guiana. In this clonally propagated crop, genotypic diversity is enhanced by the incorporation of volunteer plants into farmers’ stocks of clonal propagules. Mortality of volunteer plants was density-dependent. Furthermore, the size asymmetry of intraspecific competition increased with local clustering of plants. Size of plants was correlated with their multilocus heterozygosity, and stronger size-dependence of survival in clusters of plants, compared with solitary plants, increased the magnitude of inbreeding depression when competition was severe. The density-dependence of inbreeding depression of volunteer plants helps explain the high heterozygosity of volunteers that survive to harvest time and thus become candidates for clonal propagation. This effect could help favour the maintenance of sex in this ‘vegetatively’ propagated crop plant.

Introduction

Variation in the spatial distribution of conspecific individual plants leads to variation in the intensity of density-dependent processes, such as intraspecific competition. Individuals often vary in their ability to capture resources, which are thus unequally shared (Firbank & Watkinson, 1987; Freckleton & Watkinson, 2001). Larger plants usually have an advantage relative to neighbouring smaller plants (Schwinning & Weiner, 1998; Berntson & Wayne, 2000), leading to asymmetric competition. Environmental heterogeneity, disturbance, age differences and variation in the degree of inbreeding, all increase initial variation in sizes of neighbouring plants, which will be further increased by competitive asymmetry (Suzuki et al., 2003). Inbreeding depression, which can reflect heterosis involving overdominant loci or be due to recessive deleterious mutations, is widespread in natural populations (Byers & Waller, 1999; Keller & Waller, 2002) and reduces plant growth rates. Competition should therefore strongly influence the relative fitness of inbred and outbred plants and thereby play a role in the evolution of mating systems (Carr & Dudash, 1996; Goodwillie, 2000; Cheptou & Dieckmann, 2002).

Another interaction between population genetics and competition arises when density-dependent mortality is combined with genetic structuring in space. Neighbouring conspecifics, which often compete intensely, are frequently closely related, owing to limited seed dispersal (Cheplick, 1992, 1993a,b; Ronsheim, 1996; Koelewijn, 2004). Competition between related plants lowers parental fitness and increases the advantage of producing genetically variable offspring (Ellstrand & Antonovics, 1985).

Asymmetric competition in natural populations is often inferred indirectly from changes in distribution of plant performance among individuals in a population in response to the density of neighbouring plants (Weiner, 1984, 1988; Penridge & Walker, 1986; Weiner & Whigham, 1988; Peterson & Squiers, 1995; Falster & Westoby, 2003). Surprisingly, asymmetric competition has rarely been studied directly at the local level by following the performance of individual plants over time. Equally surprisingly, few studies have examined the impact of asymmetric competition on survival as well as on growth, even though death is a key event in population dynamics (Thomas & Weiner, 1989).

Whether inbreeding depression, or heterosis, is magnified by environmental constraints, such as the intensity of intraspecific competition, is an important question in the microevolution of natural plant populations. Density could potentially affect the relative fitness of inbred plants (Uyenoyama et al., 1993), a hypothesis supported by empirical studies (Wolfe, 1993; Eckert & Barrett, 1994; Koelewijn et al., 1999; Cheptou & Schoen, 2003) comparing inbred and outbred progeny that were obtained artificially and planted in experimental populations. Although stress levels in these experimental tests were sometimes set to mimic those under natural conditions (Norman et al., 1995; Hauser & Loeschcke, 1996; Cheptou et al., 2000), we know of no studies that have demonstrated an interaction between competition and inbreeding depression in natural plant populations. Does competition have detectable effects on inbreeding depression in natural populations?

Detecting competition within natural populations and determining its relationship to inbreeding depression pose a significant challenge, because several factors influencing plant performance can vary simultaneously. Spatial environmental heterogeneity influences plant performance, and a broad age distribution leads to confounding of age and growth rate. Both phenomena can mask or confound density-dependent processes (Wilson, 1991; Duncan, 1995) and the effect of genotype on plant performance. Variation in these factors must be controlled statistically (Wilson, 1991; David et al., 1997), requiring increased sample sizes, or by selecting sites that are environmentally homogeneous (Biondi et al., 1992; Peterson & Squiers, 1995).

We examine asymmetric competition and inbreeding depression in a novel study system that not only provides great scope for the action of both these processes, but simultaneously allows us to minimize many of the problems that confront attempts to detect these processes in natural plant populations. We studied populations of Manihot esculenta Crantz (cassava; Euphorbiaceae) dynamically managed by Palikur Amerindian farmers in traditional slash-and-burn fields. Cassava is vegetatively propagated by stem cuttings. In addition to these individuals planted by farmers, populations in fields also include ‘volunteer’ plants originating from seeds. The latter were the focus of our study. Farmers incorporate volunteer plants regularly into the stock of vegetative propagules, enhancing the genotypic diversity of the local landraces, which are highly heterozygous (Elias et al., 2001; Olsen & Schaal, 2001). Because the mating system is partially inbred (s = 0.18, B. Pujol, unpublished data), incorporation of volunteer seedlings should lower the level of heterozygosity (Elias et al., 2001). However, this is not the case (Elias et al., 2004; Pujol et al., 2005). The paradoxical maintenance of high heterozygosity of landraces despite the regular incorporation of products of partially inbred sex is explained in part by the selective survival of larger and more heterozygous volunteer plants during the manual weeding of young fields (Pujol et al., 2005).

Nevertheless, even after weeding, which constitutes a short period of human-imposed selection experienced by volunteer plants, there is still a huge gap between the level of heterozygosity in surviving seedlings and that of landraces. Selection favouring vigour, which is related to the heterozygosity of volunteer plants (Pujol et al., 2005), must also occur at stages in the life cycle of volunteer plants following weeding. We conducted a longitudinal survey of volunteer plants to investigate whether natural selection acting on volunteer plants after the episode of weeding can contribute to the hypothesized increase in average heterozygosity of surviving volunteer plants between weeding and harvest time.

We examined the impact of density, asymmetric competition and inbreeding depression on plant performance (survival) by studying the demography and genetics of cohorts of volunteer plants of M. esculenta in two Palikur fields over a period of 10 months, and in one of these fields for a further 9 months. We addressed the following questions: (i) Does the probability of mortality of volunteer plants depend on their local density? (ii) Does death of volunteer plants of M. esculenta lead to disappearance of clusters, and/or reduction in the number of plants per cluster? (iii) If so, are initially larger volunteer plants favoured, consistent with competitive size-asymmetry? (iv) Is probability of survival positively correlated with heterozygosity, and is inbreeding depression magnified in high-density patches of volunteer plants? Finally, we discuss the impact of density-dependent processes in the volunteer seedling compartment on the genotypic composition of Palikur cassava landraces.

Material and methods

Study system

Our study was conducted in traditional slash-and-burn agroecosystems of Palikur Amerindians located near Saint Georges de l'Oyapock (03°54′N, 51°48′W) in north-eastern French Guiana. Cassava is the staple food crop, and the only plant cultivated in most fields. In this farming system, secondary vegetation in previously cultivated plots is cleared and burnt in order to prepare a new field. This triggers the synchronous germination of dormant M. esculenta seeds produced during the previous cycle(s) of cultivation (Pujol et al., 2002). Thus, an even-aged cohort of spontaneously-produced seedlings appears in the field about the same time that it is planted with stem cuttings. Farmers allow these volunteer plants to grow. Whereas plants originating from stem cuttings are regularly spaced, volunteer plants vary widely in their local density. Some are far from other individuals, whereas others occur in compact clumps of two to several individuals. This is likely to be a consequence of their system of seed dispersal, which is diplochorous, with ballistic dispersal followed by myrmecochory, each seed bearing an ant-attractant elaiosome (Elias & McKey, 2000). Indeed, seeds buried by ants are sometimes grouped in caches in nest chambers of the principal dispersers, Ectatomma spp. (Ponerinae, F. Lançon and D. McKey, unpublished data). Variation in the size of these seed caches, and in germination success, is presumably responsible for the great range of variation in local density of volunteer plants, which offers great scope for studying density-dependent processes.

Density and local-scale clumping of volunteer plants in fields

In 20 fields, 11 of them classed as ‘young’ fields (about 3–6 months after field clearing) and 9 as ‘old’ fields (≥14 months after clearing), we censused volunteer plants in a series of transects (number depending on the size of the field) of 5 × 50 m (250 m2). We noted for each plant whether or not it formed part of a cluster (i.e. was ≤15 cm distant from at least one other volunteer plant). This distance corresponds to the maximum diameter we observed for nest chambers of Ectatomma spp., the ants that most frequently disperse and bury Manihot seeds (F. Lançon and D. McKey, unpublished data). We then conducted an anova [GLM procedure of SAS software (SAS institute, 1996)] on the data from these 20 fields to test for the effect of time on the mean density of volunteer plants and on the proportion of volunteer plants that were in clusters.

Demography and genotypic composition of volunteer plants

In two newly opened fields, we studied the demography and genotypic composition of exhaustively censused cohorts of volunteer plants through a cycle of cultivation. Fields 1 and 2 were cleared and burnt in November and October 2000, respectively. In the two fields, the entire cohort of volunteer plants was censused and mapped with a precision of ±5 cm. Height of each volunteer plant was measured and we noted whether each plant was ‘clustered’ (with at least one other volunteer plant ≤15 cm distant) or ‘solitary’ (all others) at the local scale. A small quantity of young leaf material was collected from each volunteer plant in the two fields. Leaf samples were immediately dried in silica gel. DNA was extracted using DNeasy® Plant Kit (Qiagen, Hilden, Germany) following the DNeasyTM Tissue Kit Handbook protocol. Eight (GAGG5, GA12, GA21, GA126, GA131, GA134, GA136, GA140) and six (idem, without GAGG5 and GA134) microsatellite loci were amplified in samples of volunteer plants and in landraces respectively, following methods of Chavarriaga-Aguirre et al. (1998). From 28 to 35 plants were sampled of each of the five most frequently cultivated landraces in the Palikur agrosystem, Burink, Kutakwa, Noswa, Sansan and Wauviye (Ouhoud-Renoux, 2000), hereinafter referred to as B, K, N, S and W. Loci were scored using an automatic monocapillary sequencer (AbiPrismTM 310 Genetic Analyzer; Applied Biosystems, Foster City, CA, USA) and the Genotyper and Genescan AbiPrism softwares. Of the plants present in the two fields at the beginning of the study (n = 240 for Field 1 and n = 69 for Field 2), 233 and 67 plants, respectively in Field 1 and Field 2, were successfully genotyped. Statistical analyses were done on these latter plants.

A previous study (Pujol et al., 2005) dealt with the effects of the first manual weeding, which occurred about 10 and 14 weeks, respectively, after clearing and burning of Fields 1 and 2. The present study concerns natural mortality after the first weeding, from April 2001 (T0) to February 2002 (T1) in Fields 1 and 2, and in Field 1 an additional 9 months from February to November 2002 (T2), when this field was harvested. Weeding had reduced the size of the cohort of volunteer plants in each field (from an initial number of 302 to 240 in Field 1 and from 97 to 69 in Field 2) and was selective in its impact, removing smaller individuals that were also among the more highly inbred (Pujol et al., 2005). However, even after weeding, values of FIS were still significantly positive for the cohort of volunteers in each field (for Field 1 and Field 2 respectively, FIS = 0.10, P < 0.0001 and FIS =0.08, P < 0.0003 before weeding and FIS = 0.08, P < 0.0001 and FIS = 0.08, P = 0.08 after weeding).

During the census of February 2002 in Field 1, the height of volunteer plants (then 1 year and 2 weeks old) was measured (T1). In this field, volunteer plants were censused again 9 months later in November 2002 (T2), giving a temporal replicate for this field (Field 1 T1 − T2). Thus data on height, heterozygosity, clustering and survival of volunteer plants were available for two spatial replicates (Fields 1 and 2, T0 − T1) and a temporal replicate (Field 1, T0 − T1 and T1 − T2).

When only nondestructive sampling can be done, the product height multiplied by the square of stem basal diameter is considered a good estimator of above-ground biomass (Weiner & Thomas, 1992). However, at T0, plants were too small to allow accurate measurement of stem diameter; squaring this variable would have magnified measurement error. We thus used height as an indicator of size. It was strongly correlated with another measure of size recorded for each plant at T0, maximum diameter of the crown [r = 0.70 (n = 240, F1,238 = 224.18, P < 0.0001) in Field 1; r = 0.82 (n = 69, F1,67 = 135.45, P < 0.0001) in Field 2], and with stem basal diameter at both T1 [r = 0.62 (n = 155, F1,153 = 94.86, P < 0.0001) in Field 1; r = 0.52 (n = 36, F1,34 = 12.45, P = 0.0012) in Field 2] and T2 [r = 0.57 (n = 123, F1,121 = 57.29, P < 0.0001) in Field 1]. At the stages of development studied here, plants were little-branched, and cylindrical, so that height is more informative about size than if plants were highly branched and bushy. Height has already been shown to be a pertinent indicator of size in a study of mortality during the preceding stages of plant development in these two fields (Pujol et al., 2005).

Effect of density on distribution of size and growth rate

In each replicate, we tested for a difference between the size of clustered and solitary volunteer plants by an anova using the GLM Procedure of SAS (SAS institute, 1996).

For each census period, we estimated the growth rate of all surviving volunteer plants by the difference in height between the beginning and the end of the period (for Field 1 T1, n = 151; Field 1 T2, n = 119; and Field 2, n = 34). We tested for differences in growth rates among the three plant categories, solitary volunteer plants, clustered volunteer plants, and initially clustered volunteer plants which became solitary because they were the last survivor in a cluster, by an anova using the GLM Procedure of SAS (SAS institute, 1996) when residuals were normally distributed and by Kruskal–Wallis tests on Wilcoxon rank scores using the Npar1way procedure of the SAS software (SAS institute, 1996) when residuals were not normally distributed.

Spatial heterogeneity within exhaustively censused fields

Spatial heterogeneity in the distribution of size, of clusters, and of the probability of mortality of volunteer plants, was estimated in Fields 1 and 2 by the correlation coefficient between the matrix of pairwise autocorrelation coefficients and the matrix of pairwise geographical distance computed on a logarithmic (base e) scale, and tested by a two tailed test established after 104 random permutations of data. Calculations were done using the software AutocorrQ (Hardy et al., 2000).

From density dependence to analysis of neighbourhood effects

We conducted two types of analyses to examine density-dependent effects on survival and growth. First, for each replicate we conducted a logistic regression to test whether the probability of survival of volunteer plants depended on their size, on their proximity to other plants, and on the interaction between these two factors. We categorized proximity to other plants in two ways. First, we defined a plant as part of a cluster or solitary and secondly, we compared plants growing in clusters comprised of a varying number of individuals, from one (solitary plants) to many (up to 18 plants in the two fields studied). Calculation of the odds ratio estimate enabled assessment of the direct effect of the presence of one additional clustered volunteer plant in a cluster, and of a 10-cm increase in height, on the probability of survival. Second, within each replicate we conducted separate logistic regressions for solitary and clustered volunteer plants to estimate the relationship between size and probability of survival in each class. Magnitude of the relationship was estimated by the odds ratio. All logistic regression analyses were performed using the Logistic procedure of SAS software (SAS institute, 1996).

The relationship between heterozygosity and survival

On the basis of the eight microsatellite loci scored in volunteer plants, we tested the impact of multiple-locus heterozygosity (MLH) on the probability of survival using a logistic regression model. We conducted this analysis over all successfully genotyped volunteer plants of the two fields (n = 300). For Field 1, we used the data originating from the entire survey (T0 to T2). Calculation of the odds ratio permitted estimation of the increase in probability of survival for an increase in MLH of one unit (one additional heterozygous locus over the 8 loci scored).

In each replicate, we analysed the relationship between MLH (on the basis of 8 loci) of volunteer plants and their size using linear regression analysis. We then compared the relationship between MLH and probability of survival in solitary and clustered volunteer plants by conducting logistic regressions and estimating odds ratios in the two groups of plants separately. Linear and logistic regression analyses were performed using the GLM and the Logistic procedures, respectively, of SAS (SAS institute, 1996).

To depict the consequence of the relationship tested for above, we estimated mean values of MLH before and after mortality in solitary and clustered volunteer plants. To enable comparison with MLH of landraces, we took into account only the six microsatellite loci scored in both volunteer plants and landraces.

Degree of relatedness of competitors

To assess whether competition is often between relatives, we compared genetic relatedness between pairs of individuals in the same cluster with that between pairs of solitary individuals. We also tested for a correlation between values of relatedness between pairs of individuals and the distance between them. Because we were interested only in local-scale patterns and processes, we only examined pairs of solitary volunteer plants separated by a distance of <5 m. This minimized the degree to which this effect was confounded with isolation by distance at larger scales, even though both local and larger-scale patterns are partly due to limited dispersal. We estimated genetic relatedness on the basis of the eight microsatellite loci scored in volunteer plants, using Moran's I statistic (Hardy & Vekemans, 1999) between pairs of volunteer plants of the same cluster and between pairs separated by <5 m. We tested for differences of genetic relatedness between solitary and clustered plants at initial times in the two fields using a general linear model [Proc GLM of SAS (SAS institute, 1996)]. Data from Field 1 T0 were linearized by a logarithm function.

For each pair of volunteer plants defined above, we estimated the probability of mortality of a plant belonging to this pair and tested whether this pairwise probability of mortality was dependent on pairwise genetic relatedness, taking into account the density of plants (solitary and clustered), using a logistic regression model (SAS institute, 1996).

We also estimated genetic relatedness on the basis of the eight microsatellite loci scored in volunteer plants, using the kinship coefficient of Loiselle (Hardy & Vekemans, 1999), which can be considered as the average relatedness in populations of volunteer plants at the different stages of the survey.

Results

Demography and clustering of volunteer plants in Palikur Manihot esculenta fields

In the 20 Palikur fields in which transects were surveyed, clusters of two or more volunteer plants of M. esculenta in an area of a few square centimetres were widespread and frequent. We found them in 12 of these 20 fields, and in the two other fields (Fields 1 and 2) in which exhaustively censused cohorts of volunteer plants were studied over time. These clusters contained on average four plants (up to 18). Clusters of up to 32 volunteer plants have been observed in other fields.

Mean density of volunteer plants decreased dramatically with age of the field, from 0.034 plants m−2 in the 11 young fields in which transects were conducted to 0.005 plants m−2 in the nine old fields. Mean proportion of plants that were in clusters also decreased dramatically with age of the field (from 16.33% in the young fields to 3.17% in the old fields). Clusters of volunteer plants were observed in all 11 young fields, whereas only a single cluster (of two plants) was observed in the nine old fields sampled.

In both the exhaustively censused fields, a large proportion of volunteer plants died: 35.2% in Field 1 during the period T0 − T1, and 21.2% during the period T1 − T2, for a total mortality rate (period T0 − T2) of 48.9%. In Field 2, 49.3% of volunteer plants present at T0 had died by T1. In both fields, mortality was proportionally higher for volunteer plants in clusters than for solitary plants (see detailed demography in Table 1). In consequence, clusters disappeared over time, up to two-thirds of them over 10 months, and the mean number of plants in the clusters that remained was reduced (Table 1).

Table 1.  Mortality rates of solitary and clustered plants in two exhaustively surveyed cohorts of Manihot esculenta volunteer plants, and over two different periods in Field 1.
Field (area)TimeNo. plants solitaryNo. plants clusteredNo. clustersNo. plants per clusterTotal no. plantsMean MLH
  1. MR is the mortality rate of volunteer plants during a period, the disappearance rate of clusters, or the reduction in the number of plants per cluster. Some initially clustered plants became solitary, and are thus added (+) to the number of solitary plants, and subtracted (−) from the number of clustered plants, at the next T. MLH is multi-locus heterozygosity.

Field 1 (0.08 ha)T014588194.62330.43
MR T0 − T1 (%)(33.1)(38.6)(26.3)(23.9)(35.2) 
T1(97 + 5) 102(54 − 5) 49143.51510.45
MR T1 − T2 (%)(16.7)(30.6)(14.3)(22.9)(21.2) 
T2(85 + 2) 87(34 − 2) 32122.71190.47
Field 2 (0.21 ha)T0452292.44670.48
MR T0 − T1 (%)(46.7)(54.6)(66.7)(5.7)(49.3) 
 T1(24 + 3) 27(10 − 3) 732.3340.51

Size and growth rates of solitary and clustered volunteer plants in exhaustively censused fields

In all three replicates, size of solitary and clustered volunteer plants was similar at each survey time (Table 2). Growth rate, the increase in size of plants during a survey, was similar (considering only plants that survived the time interval in question) between clustered and solitary volunteer plants (for Field 1 T0 − T1, n =146, inline image = 0.09, P = 0.77; for Field 1 T1 − T2, n = 113, F1,111 = 0.21, P = 0.65; for Field 2 T0 − T1, n = 31, F1,29 = 11.18, P < 0.01); between clustered volunteer plants and newly solitary (initially clustered) volunteer plants (for Field 1 T0 − T1, n = 54, F1,52 = 0.8, P = 0.37; for Field 1 T1 − T2, n = 38, F1,36 = 0.02, P = 0.88; for Field 2 T0 − T1, n = 10, F1,8 = 1.73, P = 0.23); and between newly and initially solitary volunteer plants (for Field 1 T0 − T1, n = 102, inline image = 0.35, P = 0.56; for Field 1 T1 − T2, n = 87, F1,85 = 0.002, P = 0.96; for Field 2 T0 − T1, n = 27, F1,85 = 0.04, P = 0.85).

Table 2.  Height (cm) of solitary and clustered volunteer plants of Manihot esculenta in the two fields at each survey time, and tests for size differences between the two categories of plants.
FieldTimeSolitary plantsClustered plantsDifference
NPlant heightNPlant heightFd.f.P-value
  1. Mean size of plants and SEM is given in cm.

  2. N, population size.

Field 1T014528.9 ± 0.98829.9 ± 1.40.391,2310.53
T1102212.3 ± 6.549215.8 ± 8.80.101,1490.75
T287270.7 ± 7.932286.6 ± 131.091,1170.3
Field 2T04559.8 ± 4.12249.3 ± 5.42.251,650.14
T127222.3 ± 12.17160.7 ± 23.95.291,320.03

Spatial heterogeneity within exhaustively censused fields

At each sample time in both fields, clusters of volunteer plants were distributed homogeneously over the field. The null hypothesis of a random spatial distribution of high density patches of volunteer plants within each field was accepted (probabilities of the bilateral test obtained after 104 random permutations of data; for Field 1 T0, P = 0.48; for Field 1 T1, P = 0.22; for Field 2 T0, P = 0.2).

Probability of mortality of volunteer plants was unlinked to geographical location of plants over the whole field, in either of the two fields, during any sample period. The spatial distribution of probability of mortality did not differ from that expected if death occurred spatially at random, as shown by the absence of spatial autocorrelation of the probability of survival (probabilities of the bilateral test obtained after 10 000 random permutations of data for Field 1 T0 − T1, P = 0.58; for Field 1 T1 − T2, P = 0.22; for Field 2 T0 − T1, P = 0.81).

The distribution of size of volunteer plants was spatially heterogeneous in Field 1 at T1 but not at T0 or in Field 2 at T0. Distribution of size of volunteer plants in space differed from that expected at random only in Field 1 T1 and Field 2 T0, as shown by the significance of spatial autocorrelation of size (probabilities of the bilateral test obtained after 104 random permutations of data for Field 1 T0, P = 0.09; for Field 1 T1, P = 0.0003; for Field 2 T0, P < 0.01). This relationship, although statistically significant, was not marked, explaining only 0.3% of size variance in Field 1 T1 and 0.7% of size variance in Field 2 T0. In summary, spatial heterogeneity of the environment was not sufficient to generate any spurious correlations between size, local density and survival.

Size of volunteer plants and density-dependence of survival

Probability of survival of volunteer plants increased significantly with their size (height) in all three replicates: Field 1 for the period T0 − T1 (inline image = 24.11, P < 0.0001), Field 1 for the period T1 − T2 (inline image = 11.37, P < 0.001) and Field 2 for the period T0 − T1 (inline image = 4.66, P < 0.05). Logistic regressions indicated that a 10-cm increase in size increased the odds of survival in these three replicates by a factor of 1.995, 1.29 and 1.278, respectively.

Volunteer plants that grew in a cluster with at least one other plant ≤15 cm distant had a significantly lower probability of survival than did solitary volunteer plants in all three replicates: Field 1 for the period T0 − T1 (inline image = 5.58, P ≤ 0.05), Field 1 for the period T1 − T2 (inline image = 7.12, P < 0.01) and Field 2 for the period T0 − T1 (inline image = 4.30, P < 0.05). Probability of survival of volunteer plants that were part of a cluster (for Field 1 T0 − T1: 0.61; for Field 1 T1 − T2: 0.69; for Field 2 T0 − T1: 0.46) was always lower than that of solitary plants (for Field 1 T0 − T1: 0.67; for Field 1 T1 − T2: 0.83; for Field 2 T0 − T1: 0.53). Furthermore, probability of survival of volunteer plants decreased with increasing number of clustered plants in both fields and all time periods (from 6% to 22% decrease in the probability of survival with one additional plant, see Table 3).

Table 3.  Effect of increasing clustering of plants, and of height, on survival in two cohorts of volunteer Manihot esculenta plants.
Volunteer plantsClusteringHeight
χ2 (d.f. = 1)P-valueOREχ2 (d.f. = 1)P-valueORE
  1. A logistic regression model provided the odds ratio estimates (ORE), which represent the change (expressed in percentage) in odds of survival for an increase of one additional plant in the number of plants present in a cluster, or for an increase of one centimetre in the height of plants.

Field 1 T0 − T111.550.0006−7%4.200.04+7%
Field 1 T1 − T27.720.006−22%43.31<0.0001+3%
Field 2 T0 − T18.990.003−9%8.990.003+3%

The interaction of these two factors – size and whether a plant grew as part of a cluster or solitarily – also had a significant effect on probability of survival in all three replicates: Field 1 for the period T0 − T1 (inline image = 4.63, P < 0.05), Field 1 for the period T1 − T2 (inline image = 6.18, P < 0.05) and Field 2 for the period T0 − T1 (inline image = 4.18, P < 0.05). The relationship between plant size and probability of survival thus depended on local-scale density (Table 4). In each of the three replicates, the strength of the dependence of survival probability on size was higher for plants in clusters (significant in all three replicates) than for plants growing solitarily (significant in Field 1 in both sample periods, not significant in Field 2). For plants growing solitarily, an increase in height of 10 cm increased the odds of survival of volunteer plants by a factor of 1.08–1.52. For plants in clusters, however, the odds ratio estimate was much higher, from 2.93 to 14.17. Depending on the field and period, the survival advantage associated with larger size was 1.93–13.12 times greater for plants in clusters than for solitary plants in the same field and period (Table 4). The size advantage thus greatly increased in clusters.

Table 4.  Effect of density on the magnitude of the relationship between height and probability of survival in two cohorts of volunteer Manihot esculenta plants.
Volunteer plantsSolitary plantsClustered plants
χ2 (d.f. = 1)P-valueOREχ2 (d.f. = 1)P-valueORE
  1. A logistic regression model provided the odds ratio estimates (ORE), which represent the increase in odds of survival for an increase in height of 10 cm.

Field 1 T0 − T15.440.021.5219.17<0.00012.93
Field 1 T1 − T212.960.00031.28.710.0033.27
Field 2 T0 − T10.420.521.084.450.03514.17

Effect of degree of inbreeding on survival

Within each replicate, at the initial time (i.e. just before the potential inbreeding depression our study was designed to detect), height of volunteer plants was significantly positively correlated with MLH (on the basis of 8 loci). Slopes of the regressions [S] (calculated as height in centimetres as a function of the proportion of loci heterozygous) were as follows: for Field 1 T0, n =233, S = 1.5, F1,231 = 11.21, P < 0.001; for Field 1 T1, n = 151, S = 10.3, F1,149 = 11.57, P < 0.001; and for Field 2 T0, n = 67, S = 5.2, F1,65 = 6.74, P < 0.05.

A significant positive correlation between MLH (8 loci) and the probability of survival was detected over all individuals of Fields 1 and 2 (n = 300, inline image = 8.66, P < 0.01). Calculation of the odds ratio, an estimator of the magnitude of the relationship, showed that an increase of one MLH unit (one additional heterozygous locus) increased the odds of survival of volunteer plants by a factor of 1.246. Effects of the field and of the interaction between the field and MLH were nonsignificant, allowing us to produce a single estimate of the effect of MLH.

Furthermore, the relationship between MLH (8 loci) of volunteer plants and their probability of survival, like that between size and survival probability, also depended on local-scale density. The magnitude of the effect of heterosis (or, equivalently, its inverse, inbreeding depression) on survival was higher for plants in clusters (significant in Field 1 in both sample periods, a nonsignificant trend in Field 2) than for plants growing solitarily (no significant relationship between heterozygosity and survival in any of the three replicates). For plants growing solitarily, an increase in MLH of one additional locus led to an apparent increase in the odds of survival by a factor of 1.03–1.30, but for plants in clusters the odds ratio estimate was higher, from 1.44 to 1.70. Depending on the field and period, the survival advantage associated with greater heterozygosity was 1.17–1.40 times greater for plants in clusters than for solitary plants in the same field and period (Table 5).

Table 5.  Differences between solitary and clustered plants in the relationship between multiple-locus heterozygosity and survival in two cohorts of volunteer Manihot esculenta plants.
Volunteer plantsSolitary plantsClustered plants
χ2 (d.f. = 1)P-valueOREχ2 (d.f. = 1)P-valueORE
  1. Odds ratio estimates (ORE) represent the increase in odds of survival if one additional locus of the eight loci analysed is heterozygous.

Field 1 T0 − T10.090.761.035.240.0221.44
Field 1 T1 − T21.370.241.214.880.0271.70
Field 2 T0 − T11.900.171.301.410.241.52

The effect of mortality, which struck smaller, more inbred plants with greatest severity, was to increase the mean MLH of surviving plants over time in each field. This effect was stronger in clustered than in solitary plants. Figure 1 shows that the mean MLH of clustered volunteer plants was always higher than that of solitary plants. As a result of the greater survival of highly heterozygous plants – most marked for clustered plants – mean heterozygosity of surviving plants increased over time, approaching the level typical of the clones constituting the vegetatively propagated landraces (Fig. 2). However, in both fields mean MLH of surviving volunteer plants still remained significantly lower than that of the three most heterozygous landraces (P < 0.0001 in comparisons of surviving plants in Field 1 with K, W and S; P < 0.05, P < 0.0001 and P < 0.001 in comparisons of surviving plants in Field 2 with K, W and S, respectively).

Figure 1.

Multilocus heterozygosity of clustered and solitary Manihot esculenta volunteer plants in two exhaustively censused Palikur fields, before and after competition. Grey bars represent clustered volunteer plants and white bars, solitary volunteer plants. Error bars represent SEM. The solid vertical line separates the two fields, and dashed vertical lines separate census times. This comparison considers only the six microsatellite loci examined in both volunteer plants and landraces.

Figure 2.

Multilocus heterozygosity in Manihot esculenta volunteer plants in two exhaustively censused Palikur fields, (grey bars; F1, in Field 1; F2, in Field 2) before weeding (BW), and at times T0, T1 and T2 compared with that of traditional landraces (white bars; B, Burink; K, Kutakwa; N, Noswa; S, Sansan; W, Wauviye). Error bars represent SEM. The dashed horizontal line indicates mean MLH for the five landraces studied. This comparison considers only the six microsatellite loci examined in both volunteer plants and landraces.

Degree of relatedness of potential competitors

Genetic relatedness between pairs of volunteer plants was significantly negatively correlated with the distance between them (considering all distance classes up to 5 m) in Field 1 T0 (F1,1945 = 10.2, P < 0.01) and the correlation approached significance in Field 2 at T0 (F1,55 = 3.83, P = 0.055). Nevertheless, this effect explained less variance (for Field 1 at T0, R2 = 0.005; for Field 2 at T0, R2 = 0.013) than did the clustered or solitary status of volunteer plants. Pairs of plants in a cluster were more closely related to each other than pairs of solitary individuals (for Field 1 at T0, R2 = 0.006; for Field 2 at T0, R2 = 0.1).

At all initial times of the surveys, in Fields 1 and 2, mean pairwise genetic relatedness, estimated on the basis of 8 loci by Moran's Index (I), was significantly higher (for Field 1 T0, F1,1945 = 11.9, P < 0.001; for Field 2 at T0, F1,55 = 6.1, P = < 0.05) among volunteer plants of the same cluster (mean ± SE for Field 1 T0, I = 0.102 ± 0.022, n = 309; mean ± SE for Field 2 T0, I = 0.244 ± 0.084, n = 18) than between solitary plants (mean ± SE for Field 1 T0, I = 0.019 ± 0.001, n = 1638; mean ± SE for Field 2 T0, I = −0.02 ± 0.031, n = 39), especially in Field 2 at T0, where average relatedness within clumps was about one-fourth, which corresponds to half-sibs.

No significant relationship between pairwise genetic relatedness and pairwise probability of mortality of volunteer plants was detected (for Field 1, T0 − T1, inline image = 0.74, P = 0.39; for Field 1 T1 − T2, inline image = 3.32, P = 0.07; and for Field 2 T0 − T1, inline image = 0.38, P = 0.54). Nevertheless, overall average relatedness of volunteer plants decreased in the two fields over time as a result of mortality. Average relatedness of surviving volunteer plants estimated by the Loiselle kinship coefficient decreased in Field 1 from 0.043 (P < 0.0001) at T0 to 0.026 (P = 0.27) at T1 and −0.006 (P = 0.81) at T2, the last two values being not significantly different from zero. In Field 2, the Loiselle kinship coefficient decreased from 0.07 (P < 0.05) at T0 to −0.019 (P = 0.71) at T1, the latter value being not significantly different from zero.

Discussion

Mortality, size asymmetry of mortality, genetic relatedness and inbreeding depression all were greater in M. esculenta volunteer plants that were members of a cluster than in those that grew with no conspecific plants nearby. Furthermore, mortality also increased with the intensity of competition, as estimated by the number of plants in a cluster, from one (solitary) to 18. The impact of size asymmetry on mortality has been most often studied by comparing size distributions at different densities or at different times; longitudinal follows of individual plants (Schmitt et al., 1987; Thomas & Weiner, 1989) are rare. To our knowledge, our results also provide the first demonstration that the magnitude of inbreeding depression depends on the intensity of competition in situ, not in an experimental but in a real plant population. Furthermore, our results provide indirect evidence that competition is often between close relatives. We will first examine the robustness of our analyses with respect to potentially confounding factors.

Exclusion of confounding factors related to an in situ approach

Many factors can potentially confound in situ studies of density-dependence (Wilson, 1991; Duncan, 1995; Fangliang & Duncan, 2000). We focus here on the features of our study system that minimize these sources of confusion, and thereby allowed us to detect the response of inbreeding depression in M. esculenta volunteer plants in an actual environment and with small sample sizes. First, confusion between growth and age effects was avoided in our study because all the M. esculenta volunteer plants in a field belonged to a single even-aged cohort. Second, several potential sources of spatial heterogeneity appear to have been minimal: (i) During the demographic survey, we did not observe obvious attacks of M. esculenta volunteer plants by phytophagous insects and/or pathogens, which could have been a cause of density-dependent mortality unlinked to plant variability in resource exploitation (Augspurger, 1983). (ii) Spatial heterogeneity was sufficiently small that it did not affect our results. Topography, soil composition, location of clusters, and probability of mortality of volunteer plants were all spatially homogeneous in the whole area in each of the two fields in which volunteer plants were exhaustively censused. (iii) Variation in the density and composition of heterospecific neighbours, and thus in the intensity and nature of interspecific competition, was also eliminated, owing to the removal by farmers of all adventitious plants other than M. esculenta volunteers during the weeding of both fields that preceded this study. Although some uncontrolled microhabitat variation undoubtedly remained, it was not large enough to confound our results or to lead to spurious correlation, as shown by the similar initial mean size of clustered and solitary volunteer plants, which is consistent with the hypothesis of similar growth rates of solitary and clustered volunteer plants before they reach sizes at which competition became intense. This result indicates that resource availability was similar in all parts of each field. Of course, our analysis of natural populations in real environmental conditions cannot totally exclude the possibility that undetected environmental heterogeneity affected the results. However, such uncontrolled variation would be more likely to obscure patterns, rather than to create patterns, such as those we observed in both fields studied. Collectively, our findings show that we successfully avoided the main factors that could have confounded our analyses, and that our demonstration of density-dependent effects on the demography and genetics of cohorts of volunteer plants of M. esculenta is robust.

Density and clustering of volunteer plants in young and old Palikur cassava fields

As expected based on their germination ecology (Pujol et al., 2002) and documented by our transect data, large numbers of volunteer plants appeared in newly opened fields. Density at the highly local scale varied greatly within each of the 11 young fields sampled. Density of volunteer plants also decreased seven-fold from young to old fields, showing that during the 2–3 years of cultivation of a Palikur cassava field, the number of volunteer plants continually decreases through mortality. Furthermore, the proportion of volunteer plants that were in clusters was five times lower in old fields, and the number and size of clusters also decreased over time. These patterns strongly suggest that the density-dependent mortality we demonstrated in the two fields in which cohorts of volunteer plants were exhaustively censused is a prominent and general feature in the demography of these plants.

Density-dependence of mortality: direct evidence for intraspecific competition

Close proximity of conspecific individuals generated competition sufficiently intense to lower the probability of survival in clustered volunteer plants. This density-dependent mortality led to decreased clustering of surviving individuals over time, as found in our transect samples from many fields. Furthermore, the size asymmetry of probability of survival, detected directly at the level of individuals within each of the two populations studied, also increased with density, as expected in asymmetric competition (Schwinning & Weiner, 1998). Greater height allows pre-emptive access to light and other resources (Berntson & Wayne, 2000). Both growth and survival are affected by how plants vary in the efficiency of resource capture and use (Schwinning & Weiner, 1998; Freckleton & Watkinson, 2001), but differential survival is arguably the more pertinent indicator of fitness.

Monitoring of populations of volunteer plants in Field 1 over a 2-year period showed that mortality rate decreased over time; volunteer plants that survived the first several months subsequently had a higher probability of survival. This pattern emphasizes that studies of competition should consider not only variability in space but also variability in time.

Inbreeding depression, genetic relatedness and density-dependence

Our results revealed the importance of environmental conditions, as influenced by the presence or absence of neighbours, in determining the effect of inbreeding on survival. The magnitude of inbreeding depression on survival of volunteer plants in M. esculenta was higher in clusters, where there was higher mortality than for solitary plants. Furthermore, mortality increased with the number of plants growing together, from one (solitary plants) to many.

Caution must be taken in generalizing this pattern, as the trend to increased inbreeding depression with increasing density was not significant in Field 2 (where lower sample size also reduced power of the analysis). Because of the variation in the degree of clustering across fields and its effects, density-dependence of inbreeding depression should not be considered as a general process affecting all fields but as the result of the interaction between the environment experienced by plants and their degree of inbreeding. Both environment and the degree of inbreeding are likely to be highly variable. Nevertheless, our results in Field 1 showed that competition intensified small initial differences between the most homozygous and inbred plants, and the most heterozygous, outbred plants, whose larger size gave them a competitive advantage over the former. Our empirical results from this natural population can be compared with results of experimental studies in which harsher environmental conditions sometimes lead to higher inbreeding depression. Results from our study in a natural population are in agreement with those obtained in experimental studies of other plant species that focus on the effects of intraspecific competition (Wolfe, 1993; Koelewijn, 1998) or both intraspecific and interspecific competition (Cheptou et al., 2000) on inbreeding depression. Our results confirm the expected increase of inbreeding depression experienced by plants growing in competitive conditions. Our results also add further weight to the growing conceptual framework that treats inbreeding depression as a property affected not only by traits of individual plants but also by the history and genetic structure of the population and of the environmental circumstances to which individuals are subjected.

The demographic, population-genetic and evolutionary consequences of competition are all affected by whether competitors are closely related or not (Koelewijn, 2004). The high average genetic relatedness between pairs of volunteer plants in the same cluster shows that competition was often between close relatives (up to halfsibs in Field 2) and led to a reduction in the average relatedness of volunteer plants in fields over time.

Evolutionary consequences of the clustered distribution of volunteer plants in M. esculenta

Our results indicate that intraspecific competition has consequences for both demography and evolutionary dynamics of these natural plant populations. The clustered distribution of volunteer plants, which generates competition-mediated, density-dependent inbreeding depression, has direct consequences for evolution in cassava populations. At harvest time, farmers prepare clonal propagules (stem cuttings) for their next field. Whereas most cuttings are prepared from individuals that were planted by farmers, which strongly predominate in the population, volunteer plants that survive to this time are also candidates for incorporation into landraces as new clones. Volunteers that survive to harvest time have survived 1–3 years of natural mortality. They are thus a highly selective subsample of the initial cohort, likely to present traits conferring vigour – and productivity – in these environments. Intraspecific competition between volunteer plants thus contributes to producing the high heterozygosity widespread in landraces of M. esculenta (Elias et al., 2001, 2004; Olsen & Schaal, 2001; Pujol et al., 2005).

In cassava fields, some clones are present at high frequency. Because many matings are between clonemates (geitonogamy), the mating system is partially inbred. As cassava suffers from inbreeding depression, progeny from inbred matings will experience lower growth performance compared with plants from cuttings prepared by multiplication of pre-existing clones, which are highly heterozygous. In the long term, a highly inbred mating system could thus negate any benefit of sexual reproduction to plant performance (and thus to farmers) and could lead to its evolutionary loss. However, sex persists. Our results suggest that by strengthening counter-selection against inbred plants, asymmetric competition between volunteer plants maintains the advantage of sex by ensuring that recombinant genotypes that survive to harvest, and thus become candidates for clonal multiplication, are highly outbred. Thus, by increasing the magnitude of inbreeding depression in the compartment of volunteer plants, and by leading to expression of inbreeding depression early in plant ontogeny, intraspecific competition could favour the maintenance of sex in the mixed clonal/sexual reproductive system of cassava.

Acknowledgments

We thank the Palikur Amerindians of Saint Georges de l'Oyapock (French Guiana), especially field owners. This work was funded by a doctoral thesis grant to B. Pujol from the French Ministry of Research and Technology and by grants to D. McKey from the CNRS (program ‘Impact des Biotechnologies dans les Agroécosystèmes’), the Bureau des Ressources Génétiques, the Ministry of Ecology and Sustainable Development (program ‘Ecosystèmes Tropicaux’), and the Contrat Plan Etat-Region (Guyane). We thank P.-O. Cheptou and Prof. J. Weiner for their helpful comments on the manuscript. We also thank two anonymous referees for invaluable suggestions.

Ancillary