SEARCH

SEARCH BY CITATION

Keywords:

  • biodiversity;
  • biomass;
  • equalizing mechanism;
  • negative frequency dependency;
  • Pseudomonas fluorescens;
  • stabilizing mechanism

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1. Ecologists have identified two types of processes promoting species coexistence: stabilizing mechanisms (niche differentiation and related processes) that increase negative intraspecific interactions relative to negative interspecific interactions, and equalizing mechanisms (neutrality) that minimize the differences in species’ demographic parameters. It has been theoretically and empirically shown that the two types of mechanisms can operate simultaneously; however, their relative importance remains unstudied although this is a key question in the synthesis of niche and neutral theories.

2. We experimentally quantified the relative importance of niche and neutral mechanisms in promoting phenotypic diversity in a model microbial system involving different phenotypes of the bacterium Pseudomonas fluorescens. Initially isogenic populations of the bacterium can diversify into a series of major and minor classes of phenotypes that can be treated as analogues of species. We estimated the relative population growth rate when rare of 32 phenotypes from six replicate microcosms. Each phenotype was assessed in a re-assembled microcosm in which the relative densities of all phenotypes remained the same except for the focal one which was reduced in frequency. A growth rate advantage when rare was considered evidence of non-neutral processes.

3. Approximately one-third of the phenotypes had a growth rate advantage when rare while the remaining two-thirds showed neutral or near-neutral dynamics. Furthermore, there was overall little evidence that productivity increased with phenotypic diversity.

4. Our results suggest that niche and neutral processes may simultaneously contribute to the maintenance of biodiversity, with the latter playing a more important role in our system, and that the operation of niche mechanisms does not necessarily lead to a positive biodiversity effect on ecosystem properties.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

One of the oldest problems in ecology is how species coexist in nature (Hutchinson 1961). Traditional explanations based solely on niche differences face difficulties in explaining persistent communities containing numerous similar species in seemingly homogeneous environments with a small number of niche dimensions (Hutchinson 1961; Hubbell 2001). Ecologists have identified two classes of processes that promote species coexistence: (i) stabilizing mechanisms (niche differentiation) that increase negative intraspecific interactions relative to negative interspecific interactions and (ii) equalizing mechanisms (neutrality) that minimize differences among species in population growth rate and other demographic parameters (Chesson 2000). Stabilizing mechanisms result in negative frequency-dependent selection: each species enjoys an advantage in population growth when rare. They are essential for long-term stable coexistence among species, and include processes such as resource partitioning (Schoener 1974; Tilman 1982; Tilman & Pacala 1993), frequency-dependent predation (Janzen 1970; Armstrong 1989; Wills et al. 1997; Chase et al. 2002) and those dependent on fluctuations in population density and environmental factors in space and time (Hutchinson 1961; Grubb 1977; Chesson & Warner 1981; Chesson 2000). Equalizing mechanisms contribute to species coexistence by reducing the speed of competitive exclusion and do not lead to increased population growth rates when rare. The neutral model of biodiversity (Hubbell 2001) is a special case of an equalizing mechanism where all species have identical ecologies and hence identical population growth rates, and niche-based stabilizing processes are completely absent. It has been demonstrated theoretically that either perfect neutrality (Hubbell 2001) or near neutrality (Zhou & Zhang 2008) can allow species to co-occur for long periods of time in the absence of any stabilizing processes.

Stabilizing and equalizing mechanisms can operate simultaneously to structure communities (Chesson 2000), and a synthesis of neutral theory with classical niche theory may provide a better understanding of the maintenance and patterns of biodiversity (Bell 2001; Hubbell 2001; Tilman 2004; Chase 2005; Purves & Pacala 2005; Gravel et al. 2006; Holt 2006; Leibold & McPeek 2006; Scheffer & van Nes 2006; Adler et al. 2007; Cadotte 2007). To create such a synthesis it is important to understand the relative importance of niche and neutrality mechanisms in promoting species coexistence (Bell et al. 2006; Gravel et al. 2006; Adler et al. 2007). An understanding of that relative importance is also required by studies of other questions related to species coexistence; for instance, we need to know the importance of neutral coexistence before being capable of predicting the extent of functional redundancy of species diversity in communities (Loreau 2004). Ecologists have realized that the niche/neutrality relative importance cannot be examined simply by analysing species abundance patterns (Adler et al. 2007) because frequently the same data can be explained by either neutral or niche-based processes (Chave et al. 2002; McGill 2003; Sugihara et al. 2003; Tilman 2004; Purves & Pacala 2005). Direct examination of the underlying stabilizing and equalizing processes seems essential to resolve this issue (Adler et al. 2007). However, we know of little empirical research to date that quantitatively estimates the relative importance of the two categories of mechanisms.

A testable difference between stabilizing and equalizing mechanisms is that the former assumes negative frequency dependency in population growth rate of species while the latter does not. Here we experimentally estimate the relative importance of the two types of mechanisms by examining how many ‘species’ in a stable community tend to increase in relative abundance when rare. We used communities made up of diverse phenotypes of the bacterium Pseudomonas fluorescens, a system that has become an important model in experimental ecology (Rainey & Travisano 1998; Rainey et al. 2000; Travisano & Rainey 2000; Hodgson et al. 2002; MacLean 2005; Brockhurst et al. 2006, 2007; Venail et al. 2008). When propagated in spatially heterogeneous environments (static tubes of growth media), initially isogenic P. fluorescens populations rapidly diversify, generating by mutation various phenotypes that are distinguishable by their heritable colony morphologies on agar plates and whose genetic bases have been partially identified (Spiers & Rainey 2005; Goymer et al. 2006; Bantinaki et al. 2007). These phenotypes fall into three main categories based on colony morphology and niche occupancy, and there are multiple variants within each category. Smooth morphs (SM) resemble the ancestral phenotype and inhabit the liquid phase, wrinkly spreaders (WS) form a biofilm at the air–liquid interface and fuzzy spreaders (FS) colonize the bottoms of the tubes (Rainey & Travisano 1998; Buckling et al. 2000; Kassen et al. 2000). The different phenotypes also show substantial variation in metabolic activity due to pleiotropic effects of the mutations underlying the morphological differentiation (MacLean et al. 2004). The bacterial populations reproduce asexually, therefore these phenotypes can be considered as analogous to species. Previous studies have reported the operation of negative frequency-dependent selection among the three major categories of phenotypes (Rainey & Travisano 1998), as well as both neutral (Fukami et al. 2007) and stable coexistence (Brockhurst et al. 2006) of several WS phenotypes. It thus seems likely that both niche mechanisms and neutrality contribute to the maintenance of diversity in this system, but their relative importance has not been quantitatively examined. We address this question by determining how many phenotypes increase in relative abundance from rare.

The mechanisms responsible for the maintenance of species diversity are also believed to influence the relationship between biodiversity and ecosystem functioning (Kinzig & Pacala 2001; Mouquet et al. 2002; Fox 2003; Loreau 2004). Models based on niche processes often predict a positive relationship between diversity and ecosystem properties such as biomass production as a result of more efficient total resource use, i.e. niche complementarity (Tilman 1999; Loreau 2000). In contrast, neutral theory assumes that all species are functionally equivalent and hence diversity has no effect on ecosystem properties (but see Loreau 2004; Pueyo et al. 2007; Zhou & Zhang 2008). We look at whether the relationship between phenotypic diversity and biomass production in our system could be predicted by the estimated importance of niche processes in maintaining phenotypic diversity.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Bacterial phenotypes

Six replicate lines of Pseudomonas fluorescens SBW25 (Rainey & Bailey 1996) were propagated in static 30 mL glass universal bottles (referred to as microcosms) containing 6 mL of M9KB media (glycerol, 10 g L−1; proteose peptone no. 3, 20 g L−1; Na2HPO4, 6 g L−1; KH2PO4, 3 g L−1 NH4Cl, 1 g L−1; NaCl, 0·5 g L−1) at 28 °C. Every microcosm was initially inoculated with c. 108 cells of an ancestral isolate of SBW25. Sixty microlitres (1%) of each culture was transferred to fresh media every 2 days. At day 6 (when the phenotypic diversity of the populations had approached an equilibrium) the diversified bacterial populations (referred to as ‘base communities’ hereafter) were vortexed and an aliquot from each microcosm was diluted and plated onto M9KB agar. We counted ∼400 random colonies for each base community, and isolated a colony for every distinguishable phenotype with a proportional abundance ≥0·02 (for the detailed composition of every base community, see Table S1 in Supporting Information). Of the different phenotypes identified two fell within the broader SM category of phenotypes while five were members of the WS category. A total of 32 isolates from the six base communities were obtained, propagated and stored at −80 °C in 50% glycerol. No FS phenotype was observed in our experiment. In subsequent analyses, we consider the 32 isolates (rather than the seven morphotypes) as species analogues. We do this because the six base communities have evolved independently from a common ancestor and two isolates that show the same morphology but come from different communities may be distinct in ecology (Brockhurst et al. 2006).

Invasion experiment

For each phenotype from each of the six base communities, we performed an invasion experiment to determine whether it could increase in relative abundance from rare within its own community. To test whether a particular phenotype, A, can invade its community, we initiated new communities where the initial density of every phenotype except A was 100-fold lower than that observed in the base community, while phenotype A was introduced at an even lower density, 1000-fold lower than that observed in the base community. The new microcosm was grown in a static incubator for 2 days at 28 °C. Each invasion experiment was replicated three times.

We calculated relative growth rate when rare (W) for each phenotype within its community. For a phenotype A in a given community, let mA be the estimated Malthusian parameter of A and mA the equivalent compound growth rate for the rest of the community excluding A. The relative population growth rate of A is WA = mA/mA where the two parameters mA and mA are calculated as ln (Nf/N0) with N0 and Nf the relevant initial and final densities respectively (Lenski et al. 1991). W > 1 suggests that the focal phenotype can invade from rare (and hence persist stably in the community), with = 1 indicating neutrality and < 1 suggesting transient persistence (i.e. the phenotype will be competitively excluded).

We used two methods to assess the numbers of neutral and non-neutral phenotypes. First, difference between every phenotype’s W value and one was tested using one-sample t-test. Second, we compared the observed frequency distribution of W values with that expected under the assumption of neutrality. We obtained the latter by assuming all 32 phenotypes had no low-frequency advantage (mean W = 1) and that the variance around the mean was equal to the average observed among-replicate (within-phenotype) variance.

Ecosystem function experiment

We performed a ‘phenotype removal experiment’ to investigate the relationship between phenotypic richness and biomass (cell density). Phenotypes isolated from each base community were used to initiate new assemblages with one or more component phenotype(s) omitted. The initial density of each phenotype in the new assemblage was 100-fold lower than that in the base community. All possible phenotype removal combinations were carried out for each base community. The new assemblages were incubated in a static incubator for 2 days at 28 °C, and the final density of each phenotype was measured by plating the cultures onto M9KB agar. Each assemblage was replicated three times, and the average across the three replicates was used in analyses. Biomass data (colony forming units, per mL) were log-transformed before ancova analysis in which phenotypic richness was included as a continuous variable, and base community as a random factor.

Biodiversity can influence ecosystem processes through two types of mechanisms: niche complementarity and the selection effect (Aarssen 1997; Huston 1997; Tilman 1999; Wardle 1999; Leps et al. 2001; Loreau et al. 2001). Complementarity relies on species occupying different niches or interacting positively so that diverse communities are more efficient in carrying out a specific process such as biomass production. The selection effect explains the superiority of diverse communities by the higher chance that species-rich assemblages include particularly efficient species that come to govern the community’s property. For example, diverse communities are more likely to include the most productive species and hence produce more biomass.

We estimated the complementarity effect by relative yield (RY) analysis (Trenbath 1974; Harper 1977; Vandermeer 1989; Hector 1998; Hooper 1998). The RY of a phenotype in a mixed assemblage is the ratio of its biomass in that mixture to its yield in monoculture. The sum of the RYs of all the phenotypes in a mixture is the relative yield total (RYT). RYT = 1 indicates perfect niche overlap (no niche differentiation among phenotypes). RYT > 1 indicates some degree of niche partitioning. Finally, RYT < 1 suggests interference among phenotypes. RYT was analysed using mixed-model anova with phenotype morphology as a categorical explanatory variable and base community as a random factor.

Evidence for a selection effect can be found by looking for a positive correlation between the (normalized) RY of phenotypes in diverse communities and their monoculture yields (for details, see Zhang & Zhang 2007). We used this method rather than ‘additive partitioning’ (Loreau & Hector 2001) or ‘tri-partitioning’ (Fox 2005, 2006) because the latter apply to experiments with substitutive design (mixtures at different diversity levels have the same initial total abundance) while our experiment had an additive design (diverse mixtures have higher initial total abundance than species-poor ones).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Across all six base communities two SM and five WS phenotypes could be distinguished based on consistent heritable differences in colony size and morphology. They are referred to as SM1, SM2, WS1, WS2, WS3, WS4 and WS5. We observed SM1, SM2 and WS1 in every base community while WS2, WS3, WS4 and WS5 occurred in five, two, four and three base communities respectively (Table S1). All base communities contained either five or six phenotypes, and in total 32 phenotypes isolated from the six communities were analysed.

Stable and neutral persistence

The relative population growth rate when rare (W) was calculated for 32 phenotypes derived from six base communities (Fig. 1). Eleven W values were significantly greater than 1 (one-sample t-test, < 0·05) while a twelfth approached significance (0·05 < < 0·10; Figs 1 and 2). No case of relative growth rates significantly less than 1 was found. The remaining 20 phenotypes formed a near-symmetrical distribution centred on = 1 that had a variance similar to that predicted by the null hypothesis of neutrality with the observed among-replicate sampling variance. Therefore, our results suggest that approximately one-third of the phenotypes are non-neutral and two-thirds are neutral or nearly neutral.

image

Figure 1.  Frequency distribution of estimated relative growth rates when rare (W) of the 32 phenotypes identified in the study. The histograms show the observed frequency distribution (1·11 ± 0·026 as mean ± SE), with the shaded parts of the bars representing the 12 W values that were significantly greater than 1 (one-sample t-test, < 0·05 for 11 values and 0·05 < < 0·10 for the remaining one). The curve shows the expected distribution of W values for ‘truly neutral’ phenotypes (= 1) with the observed sampling variance (0·0052).

Download figure to PowerPoint

image

Figure 2.  Relative growth rate when rare (W) of each phenotype in each base community. In each panel a horizontal line through 1 shows the null hypothesis that a phenotype, when rare, has the same growth rate as the other members of its community. Data show mean ± SE (= 3). Asterisks above the bars indicate significant difference from 1 (based on one-sample t-test). *< 0·05; **< 0·01, ms (marginally significant), < 0·10.

Download figure to PowerPoint

W values did not differ between the two major categories of phenotypes (SM vs. WS; F1,25 = 1·058, = 0·31; base community, F5,25 = 1·09, = 0·39), nor across all the seven morphotypes we identified (F6,20 = 1·14, = 0·38; base community, F5,20 = 1·08, = 0·40).

Diversity–biomass relationship

Biomass production (cell density) was measured in a total of 250 assemblages made up of one to six phenotypes. Overall, there was no significant relationship between biomass production and phenotypic richness (F1, 238 = 0·078, = 0·78) but biomass differed among assemblages constructed from different base communities (F5, 238 = 9·98, < 0·001) and there was a significant richness × base community interaction effect (F5, 238 = 3·75, = 0·003). Examination of the assemblages derived from individual base communities revealed a positive relationship for base community 1, a negative relationship for community 4 and no significant relationship for the remaining four (Fig. 3).

image

Figure 3.  The biomass of assemblages created from the six different base communities as a function of the number of phenotypes each includes. Fitted regression lines are drawn for the two significant relationships (based on linear regression model).

Download figure to PowerPoint

An RYT index equal to 1 indicates perfect niche overlap while RYT > 1 indicates a degree of niche complementarity. The grand mean RYT across all non-monoculture assemblages was just higher than 1 (1·025 ± 0·012, mean ± SE; = 0·035). Comparing assemblages constructed from different base communities we found a significant richness × base community interaction (F5,206 = 4·33, = 0·001) and a significant main effect of base community (F5,206 = 3·85, P = 0·002), but the effect of phenotypic richness was not significant (F1,206 = 0·039, = 0·88). RYT was significantly higher than 1 in assemblages from base communities 2 and 3 but significantly less than 1 for base community 4 (and marginally significantly lower than 1 for base community 5; Fig. 4).

image

Figure 4.  Relative yield total (RYT) of mixtures as a function of the number of phenotypes they include. A horizontal line through 1 in each panel indicates the prediction of the null hypothesis that no niche complementarity occurs. In each panel ‘Grand mean*’ indicates a grand mean significantly different from 1 (based on one-sample t-test) while ‘Richness*’ indicates a significant relationship between RYT and phenotypic richness (based on linear regression model). Bars show mean ± SE with significance indicated by: *< 0·050; **< 0·010; ***< 0·001; ms (marginally significant, 0·50 < < 0·10).

Download figure to PowerPoint

A negative selection effect, measured as the correlation between normalized RYs and monoculture yields, was found when all the data were analysed together (Pearson’s r = −0·19, < 0·001). When the analyses were performed separately for assemblages derived from different base communities, a negative selection effect was found only for assemblages from base communities 3, 4 and 6 (< 0·001), while assemblages from the remaining three base communities had no significant selection effect (> 0·05).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Niches and neutrality

Of the phenotypes we scored we estimated that approximately two-thirds were neutral and one-third had the ability to increase in frequency from rare. It is important to state exactly what we mean by ‘neutral’ here because these figures depend to a large extent on sampling intensity (W is a continuous trait and given a sufficiently large sample it will always be possible to statistically distinguish W ≠ 1). Our estimate of sampling variance indicates that we can distinguish from neutral most phenotypes with W > ∼1·1 (Fig. 1). We understand that the t-tests for the difference between W values and 1 (our first method to estimate the relative frequency of stably and neutrally persisting phenotypes) could have produced both false-positive and false-negative results. Correction for multiple testing was not performed here because (i) our main goal was to get an estimate of the relative frequency of values of W rather than a conservative estimate of W values greater than, or equal to, 1, and (ii) there is not a straightforward way to calculate the false-negatives rate for all the 32 t-tests (which depends on the variable statistical power of different tests), although the false-positive rate (which would be quite small, of the order of 5–10%) can be calculated using a Bonferroni or similar correction method.

Our results suggest that both stabilizing and equalizing mechanisms may have played significant roles in the maintenance of phenotypic diversity in our experimental microbial system. The coexisting phenotypes in the microcosms have diversified from one common ancestral form, so explanations are required for not only their maintenance but also how they initially increased in frequency. The emergence and maintenance of a phenotype which has a selective advantage when rare can both be explained by adaptation to a distinct niche. Niche differentiation in this system may reflect spatial specialization (Rainey & Travisano 1998) or nutrient partitioning (MacLean et al. 2005; Brockhurst et al. 2006). It is harder to explain the emergence of neutral phenotypes: every new phenotype that arises through mutation has a very low initial abundance and no advantage in growth rate when rare. Our experimental microcosms contained large bacterial populations (∼108 cells at bottleneck and ∼1010 cells at stationary phase) and so neutral mutants were unlikely to spread by drift. The possibility that neutral or even deleterious mutants emerge and reach a detectable frequency cannot be excluded, but the chance of such events occurring is likely to be very small in populations of this size, and hence such processes are unlikely to contribute significantly to the neutral diversity we observed in our system. Two other explanations are plausible. First, recurrent weak beneficial mutations may lead to the coexistence of several phenotypes. The mutants will ultimately exclude the ancestor, but very slowly. Our experimental assays may not be able to detect the advantage when rare for such mutants. The possibility that weak beneficial mutations allow near-neutral coexistence has been shown in a previous study with a chemostat Escherichia coli population (Maharjan et al. 2006): an ancestral isolate diversified into more than five phenotypic clusters within 26 days, but none of the evolved phenotypes showed negative frequency-dependent growth rates. Second, more than one ecologically equivalent mutant may appear at or around the same time and compete for fixation (Gerrish & Lenski 1998; de Visser et al. 1999; Rozen et al. 2002; Brockhurst et al. 2007); the competing mutants drive their ancestor extinct, but then persist with nearly neutral dynamics for a long period of time. This type of interaction may be especially common in large bacterial populations (Cooper 2007; Perfeito et al. 2007) and we believe it is the most likely explanation for the establishment of neutral phenotypes in our experimental system. This idea is also consistent with an earlier finding that most beneficial mutations in P. fluorescens populations evolving in a novel environment with serine as the single carbon source exhibit effects of similar sizes (Barrett et al. 2006).

Our estimate of the fraction of neutral phenotypes is subject to constraints imposed by our experimental methods. First, we only examined phenotypes with a proportional abundance ≥0·02 in our base communities. It is possible that these phenotypes are more likely to have experienced an advantage when very rare compared with a typical new mutation; this would lead to the overall frequency of neutral phenotypes being underestimated. Our estimate thus strictly applies only to phenotypes above this frequency threshold (which, of course, are likely to include the most ecologically significant). Second, it is possible that phenotypes exist that have an advantage when very rare but which show neutral dynamics over the range of frequencies used in assays (we measured the relative growth rate of every phenotype when its relative abundance was reduced by only 10-fold). This would lead to an overestimate of the frequency of neutral phenotypes. However, were such phenotypes to occur in natural systems, whether they would appear as neutral or not would depend on the details of the system’s biology. In particular, the dynamics of a very low-frequency phenotype will be dominated by the risk of stochastic extinction, and if the advantage when rare was weak, or only came into effect at densities below that at which stochastic processes predominated, then the phenotype’s dynamics would appear neutral.

Our work confirms that niche and neutral processes can function together in structuring ecological communities. However, the finding that neutrality plays a more important role in explaining diversity may or may not hold in other systems, and more work is needed before any generalizations can be made. Our experimental microcosms differ from natural communities in many ways, for example their large community sizes, the absence of immigration, restriction to a single trophic level and the generation of new ‘species’ through mutation. As mechanisms underlying neutral coexistence are more sensitive to stochastic processes than those responsible for stable coexistence (Adler et al. 2007), small communities may allow fewer neutral or nearly neutral species to co-occur (Hubbell 2001; Zhou & Zhang 2008). Immigration, however, may introduce more neutral species into a local community (Zhou & Zhang 2008). Complex abiotic and biotic interactions involving multiple trophic levels may increase the number of niche dimensions, which in turn, allows more opportunities for stable coexistence (Harpole & Tilman 2007). A recent field experiment with an asexual mite species found that every genotype collected from a local population enjoyed an advantage in population growth rate when rare (Weeks & Hoffmann 2008); it would be interesting to know whether neutral lineages arise but are lost because of stochastic drift processes, perhaps associated with small effective population size. Note also that the influence of ecological context on neutral or near-neutral species diversity could be scale-dependent. For instance, small local communities may each contain few neutral or near-neutral species, but they may experience greater turnover in species composition at the regional scale (β-diversity) because the outcomes of competition among nearly neutral species in small communities are unpredictable and different local communities are likely to be dominated by different species. Thus niche and neutral mechanisms may both contribute to species coexistence in nature, but their relative importance may vary dramatically across habitats. Our results and those of other groups (Bell 2001; Hubbell 2001; Tilman 2004; Chase 2005; Purves & Pacala 2005; Gravel et al. 2006; Holt 2006; Leibold & McPeek 2006; Scheffer & van Nes 2006; Adler et al. 2007; Cadotte 2007; Gilbert et al. 2008) support the current efforts to integrate niche and neutral theory in the same theoretical framework.

Diversity and ecosystem function

In our experimental bacterial communities the overall effect of phenotypic diversity on biomass production was negligible. We also found little evidence for complementarity: though the RYT was significantly greater than 1, the biological significance of an RYT value of 1·025 is likely to be minor. Our results do not support the prediction that the operation of niche processes leads to a positive relationship between biodiversity and ecosystem functioning (Kinzig & Pacala 2001; Mouquet et al. 2002; Fox 2003; Loreau 2004).

The absence of a stronger complementarity effect is surprising as we know that the different major phenotypic categories inhabit distinct spatial niches in static microcosms. Phenotypes within a major category may also show differential nutrient use (MacLean et al. 2005; Brockhurst et al. 2006). A priori we expected mixed assemblages to be more efficient at using resources and hence have higher total biomass, but this prediction is rejected by our data. Bacterial phenotypes may interact in other ways than through resource competition. For example, there can be complex interference interactions (via allelopathic effect) among the phenotypes in this experimental system (Day & Young 2004; Rainey 2005) which may explain the lack of complementarity (or even a ‘negative’ complementarity effect as found in community 4; Fig. 4). A previous study (Hodgson et al. 2002) including all three major categories of phenotype (SM, WS and FS) found a positive diversity effect on biomass production mainly driven by a selection effect. Significant niche complementarity was also detected in certain combinations of phenotype (WS & FS and WS & FS & SM). The present study found consistently very weak complementarity and selection effects in different phenotypic combinations (SM assemblages, WS assemblages or SM & WS assemblages). It is unclear whether the differences between the two studies were due to the absence in our experiments of the FS phenotype or the presence of a greater number of WS phenotypes. It is also surprising that in the only base community that showed a positive diversity–biomass relationship (community 1; Fig. 3) there is no evidence for significant negative frequency dependency in phenotype growth rates (Fig. 2). However, this positive diversity–biomass relationship is mainly driven by the high productivity of the four-phenotype assemblages: the other multiple-phenotype assemblages from this base community displayed little complementarity (Figs 3 and 4). We therefore do not think there is strong evidence for pervasive complementarity effects among phenotypes in community 1.

Biodiversity-ecosystem functioning studies may involve artificially constructed assemblages of species or naturally self-organized communities. With the caveat that our community developed in an artificial microcosm from an isogenic ancestor, our experiment was of the second type. We found little effect of biodiversity on productivity and it is interesting that this is in accord with most other studies of naturally assembled communities (e.g. Wardle et al. 1997; Grime 1998; Loreau et al. 2001; Wardle 2001; Mokany et al. 2008), but contrasts with experiments that artificially create new communities where positive effects are much more common (reviewed by Hooper et al. 2005; Spehn et al. 2005; Balvanera et al. 2006; Cardinale et al. 2006, 2007). This suggests that the two different experimental approaches may be answering subtly different questions; for example ‘species removal’ experiments may be a better approach to understanding the consequences of biodiversity declines in nature (Diaz et al. 2003) while the artificial assembly experiments may be more appropriate in studies of the consequences of invasions.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

This work was supported by a grant from the Leverhulme Trust to A.B.; Q.G.Z. was supported by a GSK-NERC Dorothy Hodgkin Postgraduate Scholarship from Research Councils UK.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Table S1. The detailed composition of every base community.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
FEC_1579_sm_TableS1.pdf16KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.