Performance in a novel environment subject to ghost competition

Introduction

Populations are facing a continuously changing world that they can possibly cope with in various ways, such as through phenotypic plasticity or by tracking their favoured habitat. If these solutions are not possible, evolutionary rescue by genetic adaptation may eventually allow persistence (Lindsey et al., 2013). One factor influencing this local adaptation is competition.

Interspecific competition is known to influence local adaptation in many different ways, but the effect is still largely unpredictable (Rice & Knapp, 2008; Alzate et al., 2017; Zhao et al., 2018). First, heterospecific competitors might modify the selection pressure exerted by the abiotic environment, enhancing or limiting genetic adaptation to the novel environment (Osmond & De Mazancourt, 2013). Classical examples of enhanced genetic adaptation are seen in adaptive radiations of three-spined sticklebacks or fast character displacements in Darwin finches or Myzomelid honeyeaters (Diamond et al., 1989; Schluter, 1994; Reznick & Ghalambor, 2001). Previously, we found that additional selection pressure exerted by a congeneric species (T. evansi) facilitated adaptation of the focal species (T. urticae) to a novel environment under high dispersal from a maladapted ancestral population (Alzate et al., 2017). Adaptation to the novel environment can also be reduced by interspecific competition when there is, for instance, a trade-off between traits responsible for adaptation to the competing species and to the novel environment (Chesson, 2000; Siepielski et al., 2016).

Furthermore, interspecific competition can create new niches or change the current environment for species to adapt to. Species may use waste products or adapt to plants with modified defences caused by co-occurring individuals (Sarmento et al., 2011; Lawrence et al., 2012). These new niches will subsequently create opportunities for adaptive shifts to novel environmental conditions. This illustrates that competition and facilitation can jointly shape evolution, making it difficult to study the consequences of interspecific competition alone.

As a last scenario, interspecific competition can hinder or limit the process of local adaptation by restricting resource availability and hence decrease effective population size. The resulting increased probabilities of genetic drift will then decrease the evolutionary potential and hence the chance of local adaptation (Lawrence et al., 2012; Osmond & De Mazancourt, 2013; Zhao et al., 2018). Extreme hindrance to local adaptation can even cause extinction of one of the species (Jaeger, 1970; Bengtsson, 1989; Griffis & Jaeger, 1998).

In this study we aimed to further unravel the effect of interspecific competition on adaptation to a novel food source. We therefore performed an evolutionary experiment with two related spider mite species, Tetranychus urticae and T. ludeni, adapting to a novel host. Both species were placed alone or together on a new host plant and we wanted to study how this interspecific competition affects local adaptation. More precisely, we wanted to test for differences in the rates of adaptation and the final performance on the novel food source at the end of the experiment. The two competitors were supposed to be competitively similar, but the experiment demonstrated that this was not the case: both species could only temporarily co-occur, because T. ludeni went extinct after a few generations. We nevertheless continued the experiment and conducted subsequent analyses that provided convincing evidence for the effect of the ghost competition on adaptation. More precisely, we focused on the surviving species, T. urticae, and explored whether we could detect long-term evolutionary effects on performance (measured as fecundity) of this species due to differences in early selection pressures caused by the ghost competitor, T. ludeni. We chose the average of the initial population size of the unsuccessful species during the 1st month of co-occurrence as an indication for the early competitive pressure. These differences in population sizes arose naturally and can be attributed to selection, as well as drift and founder effects.

While inferior competitors are predicted to eventually go extinct, they may co-occur with the more successful competitors for many generations (Holmes & Wilson, 1998; Lankau, 2011). These early and non-persisting interactions may leave a strong signature on the future community dynamics (Law & Daniel Morton, 1996; Miller, TerHorst & Burns, 2009; Mallon et al., 2018), because they have the possibility to induce large habitat modifications or evolutionary changes in the more successful species. Historical contingency (i.e. the influence of the arrival time of a certain species in a community; Fukami, 2015) in terms of limitations imposed by so-called ghost species (Hawkes & Keitt, 2015) may thus have a strong impact on the eco-evolutionary trajectories of populations and communities, in the same way as successful species do (Fukami, 2015). The role of competition intensity of an inferior species prior to its extinction on the ecological and evolutionary dynamics of persisting species is still largely unknown, however.

In this study we found a lower fecundity of T. ludeni on bean and cucumber than T. urticae in the control populations, which may explain their rapid extinction. Still, the ghost species T. ludeni showed an effect on the surviving species T. urticae, because the eventually achieved strength of adaptation of T. urticae increased with the initial density of T. ludeni. We therefore suggest that ghost competition may lead to differences in long-term local adaptation.

Materials and Methods

Experimental set-up

Novel host islands were created by placing two 3-weeks-old cucumber plants, Cucumis sativus Tanja, in boxes with yellow sticky paper (Pherobank) at the bottom and Vaseline at the walls to avoid contamination between islands; this method is known to work from previous research (Alzate et al., 2017; Alzate, Etienne & Bonte, 2019; Bisschop et al., 2019). These units consisting of multiple plants are referred to as ‘islands’ because they represent isolated habitat. They represent continuous habitat because the leaves were overlapping allowing easy dispersal across the plants. After the 1st week, two fresh 3-weeks-old cucumber plants were added to create the island size of the experiment. To provide enough food for the spider mites, the islands were then weekly refreshed by replacing the two oldest plants with two new 3-weeks-old cucumber plants. In this way, sufficient time was provided for a generation of spider mites to develop on the new plants, while allowing the population to move toward the fresh leaves. The removed old plants may have contained mites or unhatched eggs, but we nevertheless chose this refreshment procedure to maintain relatively natural movement dynamics. It is for instance known that especially young fertilised females disperse more (Li & Margolies, 1993) and dispersive individuals may differ in their body condition or performance compared to sedentary individuals (Bonte et al., 2014; Dahirel et al., 2019). This refreshment procedure may have caused an extra competitive pressure if one species was more dispersive or delayed its dispersal for avoiding competition, but we preferred to design the experiment in a way that it resembled more the actual life strategy of spider mites (colonisation with few founders followed by rapid growth).

The spider mite isofemale lines (as subscribed under ‘study species’) were placed on the islands with or without heterospecifics. Eight replicate islands received both T. urticae and T. ludeni. Eight replicate islands received only T. urticae and another eight replicate islands received only T. ludeni. Each island started with the same total population size of 52 adult females and as similar as possible gene pool. Therefore, the group with only T. urticae received four adult females from each of the 13 isofemale lines. The group with only T. ludeni received four adult females from the six isofemale lines and was supplemented with 28 females from its stock population, because of the lack of success of creating more isofemale lines. The group with both spider mite species received 26 adult females of T. urticae (two from each of the isofemale lines) and 26 adult females of T. ludeni (11 T. ludeni females per island came from the six isofemale lines and were supplemented with 14 mites from its stock population). Control populations on bean were created as explained under ‘study species’.

The use of the outbred stock population of T. ludeni to supplement the populations provided an unanticipated opportunity as it increased the initial genetic variation of T. ludeni among replicates and hence differences in early selection pressures on T. urticae.

We started with a relatively low population size to make it biologically relevant as natural populations usually colonise plants at small population sizes. All adult female mites were equally distributed over the plants.

We chose the same total population size and no differences among island sizes, as it is known that differences in densities change both the intra-and interspecific competitive pressure and that an increase in island size would change the adaptive potential of the treatment (Alzate, Etienne & Bonte, 2019). This necessarily meant that the initial population size for each species was not the same across the different treatments; we recognise that this may affect genetic drift and cause sampling effects, which we will come back to in the “Discussion”. A schematic overview of the different treatments is provided in Fig. 1.

The total experiment lasted for 10 months, which is approximately 25 generations and long enough to detect local adaptation (Gould, 1979; Fry, 1989; Magalhães et al., 2007, 2009; Bonte et al., 2010). For logistical reasons the experiment was performed in two blocks with 1 month difference, each block consisted of four replicate islands per treatment.

Results

Discussion

The process of genetic adaptation to novel environmental conditions is typically studied and understood from the perspective of the available genetic variation and selection pressures as imposed by the environment. Because competing species are an intrinsic part of novel experienced environmental conditions, they are known to mediate sometimes complex evolutionary processes. Here we provide empirical evidence that initial competition between two species can have a long-lasting effect on their performance in a novel environment.

The unintentional rapid extinction of T. ludeni seems a logical consequence of the higher attained fecundity of T. urticae on the novel host already at the onset of the experiment (Fig. 2C). This higher fecundity and hence higher growth rate increased the chance for better establishment or recovery after disturbance (Turcotte, Reznick & Hare, 2011, 2013). Also, populations from T. urticae were at a higher density at the plateau phase than populations from T. ludeni (Fig. 2B). The density of T. urticae on the measured surface was almost fifty percent more than the density of T. ludeni when grown alone. This suggests that T. urticae has a higher resource efficiency than T. ludeni, which could for instance arise from evolved detoxification mechanisms as often found between herbivores and their hosts (Després, David & Gallet, 2007; Dermauw et al., 2018). The population size at the plateau phase might not only be due to a signature of ecological dynamics, but may also be a consequence of adaptation itself, because the carrying capacity may be affected by organismal traits. In this case, our results suggest selection for growth rate but not for carrying capacity due to ghost competition. After 1 month the density of populations of T. urticae without heterospecific competition was higher than that of populations with heterospecific competition, but this difference vanished together with the extinction of the competitor. This probably means that the ghost competitor (when still present) decreased the available resources resulting in a lower population size for T. urticae (Fig. 2A). Another explanation is a delayed growth due to the lower initial population size in the experiment.

We have shown that the higher the competitive pressure of the ghost competitor (as measured by its average density in the 1st month), the higher the fecundity of T. urticae was on the novel host plant (Fig. 3B). We speculate that a higher selection pressure was exerted under a higher early competitive pressure of the ghost competitor, which eventually led to an increase in fecundity of the focal species. It is known that the competitor, T. ludeni, can down-regulate plant defences (Godinho et al., 2016), but this cannot explain the correlation between its density and the fecundity of the focal species even long after the ghost competitor went extinct, because plants were refreshed weekly.

Furthermore, we compared the evolved individual performance of mites from a control population (maintained on the initial host plant) with mites adapting to cucumber with or without the ghost competitor. We found that individuals from populations of T. urticae grown on cucumber plants without T. ludeni reached a higher evolved fecundity on the novel host plant than individuals from the control population (maintained on the initial host plant), implying local adaptation to the novel host (Fig. 3B). This difference in evolved individual fecundity with the control population was not found for individuals from the populations that were initially under competition with T. ludeni when considering all these populations as a single group (so not splitting them according to initial population size of T. ludeni). However, the fecundity of this group under interspecific competition was also not significantly different from the adapted T. urticae populations without interspecific competition, indicating that this group was intermediate between the adapted group without interspecific competition and the control population.

We explain this by the various degrees of early competitive pressure in this group under heterospecific competition, the main result of our study. We suggest that not finding a difference in fecundity with the control population may be due to an initial trade-off between performance on the novel host and the ability to compete with heterospecifics (Siepielski et al., 2016). It is possible that the populations under competition initially adapted to the competitor and that this lowered the amount of standing genetic variation necessary for adaptation to the novel host. A likely alternative explanation for these various degrees of early competitive pressure is variation in drift effects which is more probable in the populations under heterospecific competition because of the lower initial population size per species. Under a scenario of strong drift effects we would expect large differences in performance between replicates with and without T. ludeni (and thus with high or low initial population size for T. urticae), but they were quite similar (Fig. S1), so strong drift can be largely discarded as a driver of the observed evolutionary dynamics.

Individuals from populations where half of the population was replaced with heterospecifics could not adapt as well to the novel host plants as the populations consisting of only conspecifics. This could be due to a higher selection pressure from conspecifics compared to heterospecifics which may in turn be due to the larger initial population sizes or due to a larger niche overlap (Bolnick, 2001; Svanbäck & Bolnick, 2007).

The history of species in a community can have an impact on interspecific interactions (Fukami, 2015). The magnitudes of such historical contingencies do, however, strongly differ among species and environments (Vannette & Fukami, 2014). Differences in historical contingency may explain why some populations experience radiations, whereas others from the same clade do not achieve this under seemingly similar conditions (Seehausen, 2007). Our results suggest that increased interspecific competition leads to higher selection pressures and thus improved performance (Fig. 3B). Our results coincide in this respect with other empirical work demonstrating that increased competition with heterospecifics increased local adaptation in bunchgrasses (Rice & Knapp, 2008). Similarly, intraguild predation between lizard species increased the selection pressure and led to strong divergence in morphological adaptation as associated with niche specialisation (Stuart et al., 2014).

Nevertheless, we have to be careful with generalising our results. First, we chose small populations sizes as they are more biologically relevant, but this may limit adaptation and establishment (Del Castillo et al., 2011; Yates & Fraser, 2014). We also used populations that have been maintained in the lab for many generations, probably leading to a decrease in genetic variation compared to wild populations. The problems we encountered in creating isofemale lines for T. ludeni could be an indication of inbreeding depression. However, we are confident that our results are robust as we could still provide evidence for local adaptation in the populations of T. urticae without competitor (Fig. 3B). This suggests that the initial amount of genetic variation did not limit T. urticae in our study.

Second, it is impossible to add a competitor without changing total population sizes, population densities, or island sizes; all of these affect genetic variation and drift (Del Castillo et al., 2011; Alzate, Etienne & Bonte, 2019). As it is known that larger populations usually contain more genetic variation, we chose to standardise this by means of isofemale lines, knowing that this might create differences in drift among treatments. One way to better disentangle the effects of drift from those of selection with our small population size would have been to increase the number of replicates which was difficult for logistical reasons.

Third, our experimental design is not strictly suitable to assess adaptation in the interspecific competition treatment, as we did not keep a heterospecific control population on bean (again for logistical reasons). Hence, we cannot disentangle the effect of changes in fecundity due to competition (independent of the novel environment) from the effect of competition on adaptation to the novel environment. Although this means that we cannot detect adaptation in the treatment under ghost competition, we did find a positive influence of the density of the ghost competitor on fecundity, meaning that early competitive pressures substantially matter and providing evidence for eco-evolutionary dynamics.

Conclusions

In conclusion, we have shown the importance of early selection pressures such as ghost competition. Even when one species becomes extinct, the competition signature continues to affect the adaptation process of the successful species.

Supplemental Information