Incorporation of an invasive plant into a native insect herbivore food web

Introduction

The introduction and subsequent explosive spread of non-native species is seen as one of the main environmental disturbances threatening ecosystems globally (Glowka, Burhenne-Guilmin & Synge, 1994; Gurevitch & Padilla, 2004; Simberloff, 2011). Not all introduced species will eventually successfully establish themselves and spread invasively (Williamson & Fitter, 1996). For example, populations of colonists may die out due to disease or adverse environmental conditions (Rodriguez-Cabal, Williamson & Simberloff, 2013). Nonetheless, the numbers of environmentally problematic exotics are increasing worldwide (Butchart et al., 2010). This even holds for parts of the world that are traditionally seen as sources, rather than recipients of exotic species, such as Europe (Hulme et al., 2009; Van Kleunen et al., 2015).

One potential explanation for the invasiveness of an introduced species is the so-called enemy release hypothesis, ERH (Keane & Crawley, 2002; Liu & Stiling, 2006), which states that, because the introduced species has not coevolved with the native biota, release from specialized parasites and predators causes explosive population growth.

Enemy release may cause the initial spread, but the subsequent population dynamics are more complex, and influenced by evolutionary processes. Reduced selection pressure for defences against specialist herbivores may result in the evolution of changed energy investment. For example, the plant may evolve stronger allocation of resources towards growth and reproduction and/or towards defence against generalists (Blossey & Nötzold, 1995; Joshi & Vrieling, 2005; Zangerl & Berenbaum, 2005; Prentis et al., 2008; Whitney & Gabler, 2008). However, at the same time, native herbivores may evolve the ability to locate and feed on introduced species (Vellend et al., 2007; Pearse & Hipp, 2014). Therefore, the course of the establishment of an introduced species is complex, with population dynamics modified by evolution: over time, the community of natural enemies attacking an introduced species tends to expand (Brändle et al., 2008) and the adverse impact of invasive species tends to wane (Williamson, 1996; Simberloff & Gibbons, 2004; Blackburn, Lockwood & Cassey, 2009; Dostál et al., 2013). This may be due to evolution in both the invader and the species it interacts with (Vellend et al., 2007). However, a species’ invasive character is often considered static, and management policies rarely consider the possibility that it may change due to evolutionary adaptation (Whitney & Gabler, 2008).

One prominent invasive plant species in Europe is the black cherry, Prunus serotina Ehrh, native of eastern North America and considered a “forest pest” in Europe after widespread planting as auxiliary tree in pine plantations throughout the 20th century (Schütz, 1988; Bakker, 1963; Starfinger et al., 2003). Being bird-dispersed, it has been rapidly invading forested and open habitats (Deckers et al., 2005). In many European countries (Starfinger et al., 2003), it is now considered one of the most important threats to habitat quality of vegetation on dry, acidic, and/or poor soil, such as dunes and moorland (Fig. 1; Godefroid et al., 2005). In the Netherlands, for example, P. serotina has increased in distribution and abundance by at least two orders of magnitude during the second half of the 20th century (Tamis et al., 2005). Current control measures (chemical and mechanical eradication) are temporary and cosmetic (Starfinger et al., 2003). Nonetheless, they are costly: Reinhardt et al. (2003) conservatively calculated the annual cost of P. serotina control in Germany to be ca. 25 million euros.

Possibly the initial spread of P. serotina was facilitated by an absence of natural enemies; for example, Reinhart et al. (2003) found that, in the native range, soil pathogens inhibit the establishment of P. serotina seedlings near conspecifics, whereas in the invaded range, the species-specific soil community facilitates establishment. However, it is to be expected that the rich resource which P. serotina constitutes will provide adaptive opportunities for phytophagous insects to exploit. Such an evolutionary process will be even more likely if P. serotina represents an enemy-free space for herbivores (see Feder (1995) and Karolewski et al. (2014) for examples in other plants), and if it has been evolving reduced herbivore defences (Blossey & Nötzold, 1995). The changes in chemical defences may be complex. Joshi & Vrieling (2005) found that invasive plants may increase energetically cheap defences aimed at generalist herbivores, while reducing costly defences aimed at specialists when these specialists are no longer present.

Reports of native insects exploiting introduced P. serotina in Europe have been scarce throughout much of the 20th century, and have mostly concerned accidental feeding (by, e.g., moths, aphids, weevils, and leaf beetles; Korringa, 1947; Hille Ris Lambers, 1971; Moraal, 1988; Klaiber, 1999; Fotopoulos, 2000). Simultaneously, at least among nature management workers, a widespread belief has been maintained that the strong cyanogenic properties of the species, stronger than in P. padus (Poulton, 1990; Swain, Li & Poulton, 1991; Santamour, 1998; Hu & Poulton, 1999; Fitzgerald, 2008; Pimenta et al., 2014), have prevented native insect herbivores from colonizing it (Nyssen et al., 2013; Anonymous, 2014). More recently, however, studies from France, Germany, the Netherlands, and Poland are beginning to suggest that a community of native herbivores may in fact be accumulating on P. serotina (Karolewski et al., 2014; Wimmer & Winkel, 2000; Winkelman, 2005; Nowakowska & Halarewicz, 2006; Żmuda et al., 2008; Boucault, 2009; Halarewicz & Jackowski, 2011; Meijer et al., 2012; Karolewski et al., 2013).

In this paper, we investigate the composition of the insect herbivore community feeding on P. serotina in the Netherlands. Because congenerics are likely to have been an important source of colonists, we compare the P. serotina herbivore community with the one occurring locally on P. padus, its closest native relative in the Netherlands (Bortiri et al., 2001). To obtain an impression of the accumulation of herbivory in P. serotina, herbivore damage in both Prunus species is quantified on the basis of herbarium records. We then investigate the impact that two conditions may have had on herbivore presence: cyanogenic defence compounds and parasitoid attack, in both Prunus species. Finally, as an example of the adaptive evolution that specialist P. serotina herbivores may have undergone, we studied host preference and genomics in one particular P. serotina herbivore, the leaf beetle Gonioctena quinquepunctata.

Materials and Methods

Results

Sampling herbivore communities on Prunus serotina and Prunus padus

Our sampling method covered on average, per tree, 258 (±136 s.d.) and 141 (±91 s.d.) leaves of P. serotina and P. padus, respectively. Given mean fresh weights of P. serotina and P. padus leaves of 0.44 and 0.91 g, respectively, the amounts of foliage searched in 5 min were 113.5 g and 128.3 g for P. serotina and P. padus, respectively. After correction for the 1.13×more foliage searched in P. padus, we found that P. serotina harbors a 4.15-fold lower density but almost two-fold higher species diversity of herbivorous insects (Table 1; Table S1) than P. padus. The higher herbivore load on P. padus is, however, largely due to only two monophagous species, Y. evonymella (Lepidoptera: Yponomeutidae) and R. padi (Hemiptera: Aphididae), which usually occur in dense “nests” and “colonies”, respectively (Leather, 1985). These two species were found on P. serotina at much lower densities and usually only as single individuals. Almost half of the herbivore specimens found on P. padus belong to these two species. We did not find a difference in the proportions of specialists versus generalists on the native and the non-native host (Fig. 2): both species carried similar (chi-square =4.13; P = 0.38) proportions of each of the four categories of host range (G, generalists; R, Rosaceae-specialists; P, Prunus-specialists; M, P. padus monophages; and O, other—mostly Quercus—specialists).

Table 1:

Data on insect herbivores communities sampled.

	N (trees)	Leaf material searched	N (ins.)	S (ins.)	Na (ins.)
P. serotina	150	17.025 kg	794	64	748
P. padus	150	19.245 kg	3,732	39	1,113
Total	300	36.270 kg	4,523	72	1,860

DOI: 10.7717/peerj.1954/table-1

Notes:

aWithout Yponomeuta and Rhopalosiphum.

N

number of individuals (“load”)

S

number of species

History of herbivory on Prunus padus and Prunus serotina

Herbarium records (Table S2) for P. serotina (n = 96; 2,817 leaves) showed a more than two-fold increase in herbivory (proportion damaged leaves) from 18.8% to 40.6% over the past 170 years (r = 0.262; P = 0.0099, df = 94; Pearson test; Fig. 3A). For P. padus (n = 222; 6,612 leaves), herbivory has remained stable at c. 35% over the past two centuries (r = − 0.020; P = 0.766, Pearson test; Fig. 3B). In the most recent year (2013) we found no significant difference between the herbivory in P. padus (40%) and P. serotina (41%) (T-test; P = 0.53).

Determination of cyanogenic glycosides

In the NMR-analyses (Table S3), we found that the concentration of cyanogenic glycosides (prunasin and amygdalin combined) per unit leaf dry weight is similar in both Prunus species. Mean concentrations in young and old leaves differed by <5% in each plant species. In both plant species, the ratio prunasin : amygdalin was c. 3 : 1. Generalist and specialist herbivores showed different relations with cyanogenic glycoside concentrations, and the responses to prunasin differed from those to amygdalin. Specifically, we found that the generalist herbivore load was not correlated with prunasin (R = − 0.08, P = 0.39, both in P. prunus and P. serotina), but increased with amygdalin concentration (R = 0.24 and 0.36; P = 0.01 and 0.0001, respectively, in P. padus and P. serotina), whereas the specialist herbivore load increased with prunasin concentration, and decreased with amygdalin concentration, but significantly so only in P. padus (of which the amygdalin relationship would lose significance after Bonferroni correction; see statistical test results given in Fig. 4).

Genomic differentiation in host-specific subpopulations of a specialist herbivore

Illumina sequencing of a G. quinquepunctata larva from S. aucuparia gave 157,327,896 reads, and 191,340,606 reads were obtained from an adult beetle found on P. serotina. The de-novo assembly with Abyss resulted in 438,237 contigs longer than 200 bp. The data were deposited in the NCBI short read archive under BioProject accession code: PRJNA277307. A total of 729 usable SNPs were obtained from the SNP discovery. To assess genetic differentiation in both host-specific subpopulations, we genotyped 379 individuals from both hosts at each of five localities, for the selected 119 SNP loci (Table S4). Our Structure analysis (Text S1) failed to detect overall genetic differentiation between the populations on both host plants: the highest posterior probability was for K = 2, but these two groups did not correspond with host plant nor with locality. The hierarchical AMOVA with host plants nested within localities, showed significant (P < 0.01) differentiation between host plants in each locality. On a locus-by-locus basis, the AMOVA revealed 13 loci that were significantly differentiated between the two subpopulations from P. serotina and S. aucuparia, two of which remained significant after strict Bonferroni correction (Table S6). The distribution of per-locus pairwise (Prunus–Sorbus) $F_{\mathrm{ST}}$ values (Fig. 5) also shows that at least two loci are outliers. Homology searches in Genbank for these SNP loci yielded no matches with genes of known function.

Discussion

Our inventories show that the invasive P. serotina in the Netherlands harbours a surprisingly rich community of herbivores. Although the densities were lower than on native P. padus, the species diversity was greater. Also, contrary to expectations, the P. serotina herbivore community contained similar proportions of specialists versus generalists as the one on P. padus. The only species strikingly absent from P. serotina were two abundant P. padus monophages, Y. evonymellus and R. padi. Consistent with Leather (1985), both species were responsible for more than two thirds of all insects found feeding on P. padus, whereas they occurred on P. serotina only in small numbers (we found only a single Y. evonymellus caterpillar and a single R. padi colony on P. serotina). Nonetheless, laboratory data (Kooi, Van de Water & Herrebout, 1991) and field data from Poland (Karolewski et al., 2014) suggest that at least Y. evonymellus has the potential to feed on P. serotina. Karolewski et al. (2014) state that in Poland, the latter species has progressed from avoiding P. serotina altogether to feeding and developing on it massively over the past decade. The near-absence from P. serotina of this herbivore in our study area suggests that a similar colonization event may not yet have taken place, but this may change in the near future, possibly aided by long-distance gene flow from the populations in Poland. Another striking difference between both hosts is the relatively large numbers of non-Rosaceae specialists on P. serotina. While some of these may be accidental “tourists”, the high number of individuals for some of these species (e.g., the Quercus-specialist Harpocera thoracica) is noteworthy.

These results add to a body of data on insect herbivory on native versus non-native plants (reviewed in, e.g., Liu & Stiling, 2006; Colautti et al., 2004; Meijer, 2013). Although these studies tend to show that introduced plant species, especially those with powerful chemical defences, are poor hosts for native herbivores, exceptions have also been found of introduced species hosting a larger number of species than closely-related native plants (Novotny et al., 2003). The rich herbivore community on non-native P. serotina, and especially the high number of specialist species, fits with the observation that the food web supported by a non-native plant expands as time since initial introduction increases (Brändle et al., 2008). Although P. serotina was introduced into Europe earlier (Schütz, 1988), it only became common in the 20th century (Starfinger et al., 2003). Its increasing abundance in Dutch ecosystems over the past 80 years may have been the phase during which most of the herbivore community has built up. Indeed, while our study of leaf damage in herbarium specimens cannot reveal the diversity of herbivores, it does show that herbivore damage, and therefore perhaps herbivore load, has gradually doubled over this period, while that on P. padus has not changed. Today, at least based on our herbarium records, herbivory levels in both plant species appear to be similar (despite the lower herbivore load that we found in our inventory for P. serotina—see above).

In theory, the rapid assembly of this community may have been aided by the presence of an enemy-free space for the insect herbivores. If local parasitoids, for example, are not adapted to using P. serotina volatiles as a cue for attraction to a possible patch in which to find hosts, this may have helped the establishment of herbivore populations on the introduced plant (Feder, 1995; Harvey & Fortuna, 2012). Indeed, Karolewski et al. (2014) found reduced parasitization of one species, Y. evonymella on P. serotina. However, we find that current attack rates of caterpillars by parasitoids do not differ between P. padus and P. serotina.

After an initial period of reduced specialist herbivory in the non-native range, P. serotina may have shifted its investment in chemical defences in favour of those aimed at generalists (Joshi & Vrieling, 2005). Cyanogenic glycosides are generally considered to be systemic, non-inducible, and energetically cheap chemical defences aimed primarily at generalist herbivores (Gleadow & Møller, 2014). However, our phytochemical data suggest that, in P. padus (and, less clearly, in P. serotina), the Rosaceae-specific compound amygdalin has a positive relationship with generalist load but a negative one with specialist load, whereas the more widespread compound prunasin has a positive correlation with specialist herbivore load, while lacking any clear relation with generalist load. It would be tempting to compare the levels and ratios of prunasin and amygdalin in today’s P. serotina populations in the Netherlands with those reported for the native American population. However, we only have access to a single American study (Santamour, 1998), which, moreover, employed somewhat different methods (see below), so we do so with considerable hesitation. Santamour (1998) reported a summertime HCN production in native American P. serotina corresponding to 29.6 mg cyanogenic glycosides per g fresh leaf material (see Text S2). In an earlier study of 22 Dutch P. serotina trees (Pimenta et al., 2014), we found on average 30.4 mg cyanogenic glycosides per mg dry leaf material. As P. serotina dry leaf weight is 36% of fresh leaf weight (see Text S2), this might suggest that total cyanogenic glycoside content in the invaded range could be about two- to threefold lower than in North America. Also, Santamour found prunasin : amygdalin proportions of 22:1, whereas we found a ratio of 3:1. In the Dutch P. serotina, prunasin investment might therefore have decreased, with amygdalin content remaining more or less constant. Since both the absolute and relative amounts of prunasin and amygdalin content have a genetic basis (Santamour, 1998), these results might indicate that cyanogenic glycoside defence has, after the introduction into Europe, adapted to the novel herbivore communities. With a mean age at first reproduction of only 5.2 years (Deckers et al., 2005) and evidence, in general, of rapid evolution of defence in invasive plants (Felker-Quinn, Schweitzer & Bailey, 2013), such a quick evolutionary change is not implausible. However, since Santamour (1998), Pimenta et al. (2014) and the present study appear to be the only available quantifications of prunasin and amygdalin in P. serotina, and since the range of phenotypic plasticity in cyanogenic glycoside content is unknown, more data, with more comparable methods, are needed before this conclusion can be substantiated. Moreover, we stress that our results and their discussion refer only to the cyanogenic potential (HCNp), whereas the true defence potential is a combination of HCNp and HCNc, cyanogenic capacity, which is a function of glucosidase presence and activity. Since the latter is unknown in this study, we implicitly assume that HCNp is an indicator for cyanogenic defence, which may only be partly true and is known to differ between specialists and generalists (Ballhorn, Kautz & Lieberei, 2010a).

The accumulation of the herbivore community on P. serotina may also have involved evolutionary processes within the insect community itself. One possibility is that all present-day herbivores were able to feed and reproduce on P. serotina from the moment the new host was introduced. However, this would not explain the slow increase in herbivory that our herbarium data show: highly mobile insects with short generation times would have established on the new host instantaneously, rather than gradually. It is therefore likely that adaptive evolution in the herbivores played an important role in the assembly of this community over time.

As a possible example of this scenario, we performed a case study on one specialist herbivore, the leaf beetle G. quinquepunctata, which has recently colonized P. serotina from its original host, rowan (S. aucuparia). We find indications of weak differentiation in host preference and SNP-loci on Sorbus- versus Prunus-derived beetle individuals. We found that individuals collected on Sorbus retained a significant host preference for this host, whereas beetles collected from Prunus showed no preference for Prunus over Sorbus. We found the same host preference in adults and larvae, although presumably host choice is made mostly in the mobile, adult stage. While these results do not necessarily imply genetic differentiation, as learning may be involved as well (Salloum, Colson & Marion-Poll, 2011), our SNP-analysis does show indications of weak genetic differentiation, with several loci showing divergence, and potentially linked to regions that are under disruptive, host-imposed selection. In other words, the introduced P. serotina may have selected for weak, incipient divergence (Vellend et al., 2007; Nosil & Feder, 2011) in this particular herbivore. Whether such selection will allow further sympatric speciation, in this herbivore or others, depends not only on the different selection regimes imposed by the different host plants, but also on the mount of gene-flow between the populations feeding on the two hosts (Nosil & Feder, 2011).

Overall, our results indicate that, since its introduction, a rich and diverse herbivore community has accumulated on P. serotina. It is possible that evolutionary adaptations in these herbivores as well as in the plant itself have played an important role in shaping this community. Adaptation may have involved niche widening in generalist herbivores, incipient genetic divergence in specialists, as well as adjustments of chemical defences in the host plant.

These results may have implications for invasive species management. It may be expected that the gradual evolutionary integration of a novel plant species in a native herbivore food web may eventually reduce its invasive character to the point where it attains the status of non-harmful, naturalized neophyte. Whether this will happen in the case of P. serotina depends on a number of factors. In this paper, we dealt with herbivorous insects only, whereas plant demographics are affected by a much broader spectrum of natural enemies. Reinhart et al. (2003) and Van der Putten (2000) suggested that its invasiveness may be more due to an absence of belowground interactions (with the Prunus-pathogenic fungus Pythium, for example) than aboveground interactions. However, preliminary studies in the Netherlands indicate the presence of local Pythium populations that are powerful in attacking introduced P. serotina (Tamis & Van der Klugt, pers. comm., 2015). Furthermore, Ballhorn, Pietrowski & Lieberei (2010b) and Ballhorn (2011) found that in cyanogenic plants a trade-off exists between defence against herbivores and against fungal pathogens, which is an additional complication not yet considered. A final point of concern is the intensity of the regime of natural selection. Presently, manual control of mature P. serotina in many European habitats is reducing the continued exposure of the host to its potential herbivores. On the basis of the results presented here, we would like to caution that this might have the adverse effect of a consequent slowing down of processes of adaptation, and a delay in the decline of the invasive character of P. serotina.

Supplemental Information