Phenotypic selection gradients in a tripartite plant interaction in southern South America

Abstract

The study of how phenotypic selection by plant mutualists and antagonists varies within ecosystems is crucial for evaluating potential plant adaptations to changing environments, such as urban areas. Here, we describe the interaction intensity of plants with their mutualistic and antagonistic partners, as well as the selective landscapes in one natural and one urban site. We measured direct phenotypic selection on plant traits relevant to three interaction partners: floral tube length, resistance to herbivores and arbuscular mycorrhizal fungi (AMF) colonization. Significant direct phenotypic selection on the three interaction traits was detected in the natural site. We observed disruptive selection on floral tube length, with the two fitness optima closely matching the proboscis peak lengths of the main pollinator species. In addition, we observed stabilizing selection by arbuscular mycorrhizal fungi, favouring plants with around 11% of arbuscular colonization. This result supports the idea that plants may autoregulate mean intraradical arbuscular colonization to maximize fitness relative to carbon allocation to arbuscular mycorrhizal fungi. Finally, we detected directional selection benefiting plants with high resistance to herbivores. Our results highlight the importance of adopting a multispecies, context-dependent approach to better understand the complex ecological interactions that drive plant evolution.

1. Introduction

In plants, the dependence of reproductive success on interactions with organisms of multiple species is particularly evident [1]. This is because plants overcome sessility by benefiting from animal pollinators for pollen dispersion services and mycorrhizae for extending their reach of nutrients in the soil [2,3]. Sessility also makes plants vulnerable by rendering them easy targets for damaging leaf-feeding herbivores [4]. The study of phenotypic selection imposed by plant mutualists and antagonists within an ecosystem is crucial to answer whether plants are adapting to environments and whether they are able to respond to rapid and profound environmental changes [5].

Phenotypic selection is defined as the covariance between relative fitness and standardized trait values. Phenotypic traits must show heritable and sufficient genetic variation to have an impact on selection [6]. These traits may fluctuate depending on the environmental context reflecting changes in both the mode of selection (directional, stabilizing or disruptive) and strength of selection (magnitude of the relationship between traits and fitness) within these ecosystems [6,7].

Under naturally stable conditions, populations are expected to become co-adapted through reciprocal selection among both mutualistic and antagonistic partners, maximizing fitness within the respective adaptive landscapes [8]. Phenotypic selection on plant traits can be driven by interactions with multiple agents of selection, both above- and belowground. Investigating the implications of these interactions is crucial for understanding how biotic partners simultaneously influence phenotypic selection processes in a plant species. However, studies on the impact of selection in the context of such multipartite interactions are lacking. For this reason, we aimed to investigate whether floral tube length, resistance to herbivores and percentage of arbuscular fungi (AMF) colonization are traits subject to selection by pollinators, herbivores and AMF, respectively. We focused on a species with long-tubed flowers, Nicotiana longiflora, which is primarily pollinated by the hawkmoths Manduca sexta and M. paphus (the South American sister species of M. sexta), both of which have matching proboscis lengths and an innate preference for Nicotiana floral volatiles [9–12]. This suggests an evolutionary co-adaptation between these plant-pollinator species, where floral tube length is shaped by selection exerted by pollinators through flower-proboscis match and mismatch. Regarding plant–herbivore interactions, resistance is the ability to reduce the probability of being eaten by herbivores. Hawkmoth larvae of M. sexta and M. paphus are specialist herbivores of N. longiflora [13]. In addition, adults of Cyrtomon inhalatus (Curculionidae) have been documented feeding on this plant species [14]. However, little is known about the overall herbivore insect community associated with N. longiflora. Resistance to herbivory is genetically associated with relative plant fitness, and plant responses to selection by herbivores generally reduce the mean damage levels, i.e. directional selection moves populations towards increased resistance [15,16]. Finally, symbiotic mycorrhizal fungi interacting with plant roots in the soil are key partners directly involved in plant performance and protection [17]. The plant’s ability to form mycorrhizal arbuscules, the intraradical structures involved in plant–fungi interchange [18], and the extent of their colonization are largely under plant control, mainly regulated by phytohormone signalling pathways [19]. A few lines of evidence suggest that symbiotic mycorrhizal traits (e.g. mycorrhizal colonization intensity) and AMF traits (e.g. arbuscule formation) possess sufficient genetic variation and heritability for the potential of phenotypic selection to act on these traits [18,20–23]. Due to the trade off in carbon allocation, it has been proposed that plants regulate mycorrhizal colonization through an autoregulatory mechanism [24]. Therefore, if such regulation applies in N. longiflora, we would expect stabilizing selection to act on arbuscular colonization intensity, favouring intermediate levels rather than extreme values.

Changes in natural ecosystems can differentially modify plant–organism interaction intensity, thereby altering the covariance between traits and relative fitness within plant populations, i.e. changing phenotypic selection patterns [6]. The absence or alteration in abundance and richness of co-evolved interaction partners could alter absolute mean fitness which, in turn, can modify the opportunity for selection, i.e. the variance in relative fitness within a population [8,25]. Therefore, the opportunity for selection could translate into altered strength in modified ecosystems, depending on the agent of selection, e.g. mutualistic or antagonistic partner [8,26]. For instance, under disturbed conditions, plants might become pollen limited, i.e. the amount and quality of pollen deposited on stigmas may become insufficient to achieve full reproductive potential [27]. As a result, selection strength exerted by pollinators would be modified [27,28]. Similar patterns would be expected for AMF, where increased root signalling has been shown to promote AMF root colonization, thereby alleviating plant stress under adverse conditions [29].

In this study, we aimed to describe interaction intensity and selective landscapes of mutualistic and antagonistic partners in one natural and one urban plant population. To this end, we estimated phenotypic selection on plant mycorrhizal, leaf and floral traits involved in the interactions with aboveground (hawkmoth pollinators and herbivores) and belowground AMF partners. We asked whether: (i) pollinators exert directional selection for longer flowers that match long proboscis lengths, (ii) herbivores exert directional selection for increased plant resistance and (iii) AMF exert stabilizing selection favouring intermediate root colonization levels. We also asked whether these selective patterns change in a modified environment, such as an urban area, which might influence co-evolutionary processes involving multiple agents of selection [30]. Although the populations were not replicated at the level of natural versus urban treatments, which prevents broad generalizations about differences in selection across environments, comparing these two populations provides an initial approach to assessing whether the observed patterns align with expectations.

2. Material and methods

(a). Plant species and study sites

We used N. longiflora to describe the phenotypic selection patterns mediated by three interaction partners in two populations, one in a natural and one in an urban site. This solanaceous plant species is a native perennial ruderal herb from Argentina that grows in diverse habitats. Plants range between 0.5 and 1.5 m in height, produce tubular flowers of 9 to 12 cm in length which open at night. It is a self-compatible plant that produces fruits through self-pollination and hawkmoth-mediated cross-pollination [9]. However, it is predominantly an outcrosser, since the growth rate of pollen tube and siring success favour outcross pollen over self-pollen [9,31].

Two sites, situated 40 km apart, each with a N. longiflora population of at least 60 individuals, were selected in central Argentina. One site was located within the city of Córdoba (hereafter ‘urban’), with 99% urban cover, and the other (hereafter ‘natural’), in an inhabited site covered by woodlands and shrublands typical of the Chaco Serrano district within the Chaco phytogeographical province [32].

(b). Interaction intensity and interaction partner identity

Interaction intensity of each N. longiflora plant with hawkmoths, herbivores and AMF was measured as hawkmoth visits (presence of hawkmoth scales on floral parts), herbivory (proportion of damaged leaf area) and AMF total root colonization (i.e. hyphal, vesicular and arbuscular colonization). In addition, for each interacting partner, we characterized the communities of hawkmoths, herbivores and AMF of the urban and natural sites.

(i). Hawkmoth pollinators

To measure hawkmoth visits, three to five flowers from each of the 60 plants per site were collected and inspected under a Leica MS5 stereomicroscope to determine the presence of scales that hawkmoths leave on floral petals while visiting them.

In addition, we characterized the guild of hawkmoths in each site. To this end, one vertical sheet light trap illuminated with two 150 W mercury vapour light bulbs was placed twice at each site for 4 h, from 20.00 to 00.00 h. Hawkmoths that settled on the sheets were temporarily immobilized using nail polish remover (impure ethyl acetate). To determine whether captured hawkmoths could act as pollinators, the proboscis length of each individual was measured with a digital caliper to the nearest 0.01 mm.

(ii). Herbivores

In each site, 60 flowering plants of N. longiflora separated by at least 2 m from each other were randomly selected. To measure herbivory, we randomly collected 10 N. longiflora leaves per plant in both sites. We then obtained digitally scaled photographs of each leaf, and the proportion of foliar damage was estimated using IMAGEJ 1.54 g software [33].

To characterize herbivores in each site, each leaf was inspected under a Leica MS5 stereomicroscope to extract herbivore adults. Additionally, leaves that showed typical leafminer marks were kept in plastic bags to identify emerging adults. Furthermore, we manually collected arthropods from six N. longiflora individuals of each site using plastic tubes, preserving them in 70% alcohol until taxonomic identification. Finally, we collected 10 N. longiflora flowers from each site to examine insects feeding on floral parts and preserved them in 70% alcohol for further identification.

(iii). Arbuscular mycorrhizal fungi

To measure the percentage of AMF total root colonization and percentage of arbuscular colonization, we harvested 60 N. longiflora plants per site. Roots were clarified with 10% KOH (15 min at 90°C), acidified with 1% HCl (1 min, room temperature) and stained in 0.05% aniline blue, following the techniques described by Phillips & Hayman [34] and Grace & Stribley [35]. The stained roots of each plant were mounted on one semi-permanent slide in a polyvinyl–lactic acid–glycerol solution and 100 line intersections were counted at 200 magnification under a compound microscope (E200, Nikon), following the McGonigle [36] magnified intersection method.

To characterize the AMF community composition of each site, i.e. to identify and quantify AMF morphospecies, we obtained a representative soil sample per site by pooling and mixing six rhizospheric soil samples of N. longiflora from each site. AMF spores were extracted by wet sieving and decanting 100 g of dry soil followed by centrifugation in 50% sucrose [37]. Each soil sample was first sieved through a 1 cm mesh to remove litter, stones and sticks. A fine sieve (38 µm) was used to collect small spores and the upper sieve (125 µm) was also checked for sporocarps and larger spores. Only apparently healthy spores were isolated by direct observation using a stereomicroscope (SMZ745T, Nikon). For quantification and taxonomic identification, fungal spores were mounted on slides using polyvinyl alcohol (PVA) with and without Melzer’s reagent [38] and examined using a compound microscope (E200, Nikon). AMF morphospecies were identified following Mycobank (http://www.mycobank.org/) and http://www.amf-phylogeny.com/ for species names and families, and Oehl [39] for clade names. Four main taxonomic groups were identified: Glomerales, Entrophosporales, Diversisporales and Gigasporales, and two small basal groups, i.e. Archaesporales and Paraglomerales, as a fifth group called ‘basal lineages’ [17].

(c). Fitness metric and phenotypic traits under selection

We used individual fruit set, i.e. number of fruits per total number of flowers, as a proxy for fitness, as it reflects the reproductive success of individuals resulting from their interaction with the three partners. Successful pollination by insects directly influences fruit production in plants like N. longiflora, which as a predominantly outcrosser, strongly relies on cross-pollination by hawkmoths for fertilization [9,31]. In the case of AMF, root colonization can increase fruit set either through a direct effect on nutrient uptake or by increasing pollination via the promotion of floral traits that attract pollinators (e.g. nectar production, volatile emission, flower number) [40]. Finally, herbivores can indirectly decrease fruit set by reducing resource availability through the loss of photosynthetic leaf area [41], or by altering pollinator preferences when chemical leaf defences that deter herbivores also affect pollinators through changes in nectar chemistry [42]. We estimated fruit set as the number of developed fruits divided by the total number of flowers, based on one to three inflorescences selected per plant. The total number of flowers could be quantified since those that do not persist leave an evident scar on the inflorescence axis.

(i). Hawkmoth pollinators, herbivores and arbuscular mycorrhizal fungi

To measure floral tube length, three to five newly opened flowers from different inflorescences of each plant were collected (60 plants per population). Scaled photographs of the flowers were obtained with a Nikon D80 camera, and measurements were subsequently extracted from the digital images using IMAGEJ software. We estimated resistance to herbivores as the phenotypic trait under selection by herbivores. We quantified root colonization by arbuscules as the phenotypic trait in plants that may be under selection by AMF.

(d). Statistical analysis: estimation of phenotypic selection

Opportunity for selection (I) acting on each N. longiflora population was estimated as the variance in female fitness metrics, specifically number of fruits per inflorescence, i.e. relative fitness [6]. Phenotypic selection was estimated for each population (n = 60) following multiple regression analyses. Floral tube length, resistance to herbivores and arbuscular colonization were used as predicting variables, and the fruit set fitness metric as the response variable. Given that the total number of flower scars varied among individuals, it was included as an offset to account for differences in total flower number per individual [43]. To assess the magnitude of natural selection acting on the phenotypic traits, we followed the standard model proposed by Lande & Arnold [44]. First, phenotypic traits were standardized to zero mean and unit variance and fitness metrics were relativized to the population mean. Multivariate selection analysis was conducted to estimate direct selection acting on floral tube length, resistance to herbivores and arbuscular colonization. To this end, we calculated directional (β_ii) and stabilizing (ɣ_ii) selection gradients on each trait or correlated traits (ɣ_ij). Statistical significance of the regression coefficients was obtained through quasi-generalized linear models, with Poisson error distributions and log link function using the lme4 package [45] in R v. 4.4.3 [46].

Significant phenotypic regression coefficients indicate that changes in plant traits affect plant fitness. In addition, the magnitude and sign of selection gradients show the intensity and direction of phenotypic selection on floral traits. When these traits are central to a mutualistic or antagonistic interaction, these analyses are key to understanding the process of natural selection in coevolutionary partners. If we consider that under natural conditions long floral tube length and high arbuscular colonization increase plant fitness, but damage by herbivores reduces plant fitness, we would expect that natural selection tends to increase the mean floral tube length and mean arbuscular colonization but decrease the mean level of damage (i.e. level of resistance to herbivory). To illustrate the selection patterns, we visualized the fitness surfaces using non-parametric regressions with cubic splines. To estimate the cubic splines, we used the MGCV package [47] in R. For each univariate spline, we fixed the covariate at its mean value. Smoothing parameters were obtained by minimizing the generalized cross-validation scores [48], and Bayesian standard errors were obtained according to Wood [47].

3. Results

(a). Interaction intensity and interacting partner’s identity

Interaction intensity of plants with the three agents of selection varied between the natural and urban sites and was higher in the former: hawkmoths (natural: 0.36 ± 0.03, urban: 0.10 ± 0.02 relative frequency of visits), herbivores (natural: 0.05 ± 0.05, urban: 0.004 ± 0.009 herbivory) and AMF (natural: 66.49 ± 2.73, urban: 44.19 ± 1.84 total root colonization).

Abundance and diversity of captured hawkmoths consisted of 46 individuals belonging to seven species (Callionima grisescens, Hyles euphorbiarum, Litneria maura, Manduca diffisa, M. rustica, M. paphus and Xylophanes tersa) of which only one individual (H. euphorbiarum) was captured in the urban site. Proboscis length of these species ranged from 13.23 mm in C. griscescens to 122.04 mm in M. rustica (figure 1). Considering that floral tube length of N. longiflora varied from 64.61 to 126.54 mm (figure 1), M. paphus and M. rustica were the only species capable of reaching the nectar contained in the floral tubes and touching fertile parts of the flowers with their proboscis. M. paphus proboscis length was bimodal with peak lengths of 86.67 and 102.46 mm (figure 1).

Figure 1.

Proportion of individuals in the natural site across the ranges of proboscis and flower tube lengths for the hawkmoth community interacting with Nicotiana longiflora. The number of individuals measured for each species is shown in brackets. Arrows indicate the peak values in the proboscis length distribution of Manduca paphus individuals.

We recorded six herbivore morphospecies insects belonging to three orders: Coleoptera (one Curculionidae), Hemiptera (two Lygaeidae, one Pentatomidae and one Aphididae) and one Thysanoptera. All these morphospecies were present in the natural site, while only two, Lygaeus sp. (Lygaeidae) and the Aphididae spp., in the urban site. In flower samples, we recorded one herbivore insect morphospecies from the order Thysanoptera in the natural site and one belonging to the Aphididae family in the urban site.

We identified a total of 10 AMF morphospecies belonging to three families (i.e. Acaulosporaceae, Entrophosporaceae and Glomeraceae) and three taxonomic groups (i.e. Diversisporales, Entrophosporales and Glomerales). Four of the AMF morphospecies were common to natural and urban sites (A. scrobiculata, Entrophospora etunicata, Funneliformis mosseae, Septoglomus constrictum) while four (i.e. Acaulospora sp., E. infrequens, Glomus brohultii and Rhizoglomus intraradices) were exclusive to the urban site and two to the natural site (i.e. E. lutea and Funneliformis geosporum).

(b). Phenotypic selection: opportunity for selection, selection strength and direction

Fruit set was similar at both sites (mean = 0.74 fruits per total number of flowers), but the standard deviation was higher in the urban site than in the natural site (0.39 and 3.22, respectively). However, the natural population of N. longiflora showed a higher mean absolute fitness than the urban population (figure 2a), i.e. mean fruit number. The standard deviation was higher in the urban site than in the natural site (0.39 and 3.22, respectively), as well as the variance in relative fitness, i.e. the opportunity for selection, was also higher in the urban site (figure 2b).

Figure 2.

(a) Mean absolute fitness (mean fruit number) and (b) opportunity for selection (variance of relative fitness) in two populations of Nicotiana longiflora (n = 60), one from a natural site and one from an urban site.

Direct (directional or stabilizing/disruptive) selection acted on the three interacting traits of the N. longiflora population and was significant only in the natural population (table 1). In the natural site, a significant disruptive selection on floral tube length, favouring 89 and 103 mm floral tube lengths was evident (table 1, figure 3). These lengths matched with the two proboscis peak lengths found for M. paphus individuals (i.e. 87 and 102 mm; figure 1). In addition, we detected directional selection for increasing resistance to herbivores and stabilizing selection favouring around 11% of arbuscular colonization (table 1, figure 3). We did not detect any significant correlational selection gradient for the three plant traits analysed (table 1).

Table 1.

Mean trait values and standardized selection gradients (with standard errors in parentheses)—directional (βi), stabilizing/disruptive (ɣii) and correlational (ɣij)—for floral tube length, resistance to herbivores and arbuscular colonization of arbuscular mycorrhizal fungi (AMF) in two Nicotiana longiflora populations one from a natural site and one from an urban site (n = 60). The coefficients of selection gradients indicate the direct change in trait mean βi and variance ɣii accounting for the indirect effects of correlated traits ɣij, in response to selection by each interaction partner. ns, non-significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001.

	natural	urban
mean trait values
floral tube length (mm)	98.67 (7.62)	96.77 (8.28)
resistance to herbivores (%)	0.95 (0.05)	0.99 (0.01)
arbuscular colonization (%)	15.03 (9.04)	8.31 (7.53)
directional selection (βii)
floral tube length (mm)	0.047 (0.055) ns	0.027 (0.130) ns
resistance to herbivores (%)	0.103 (0.075)***	−0.061 (0.280) ns
arbuscular colonization (%)	0.036 (0.045) ns	0.027 (0.122) ns
stabilizing/disruptive selection (ɣii)
floral tube length (mm)	0.087 (0.129)*	−0.130 (0.203) ns
resistance to herbivores (%)	−0.007 (0.205)*	0.477 (1.157) ns
arbuscular colonization (%)	−0.046 (0.076)*	−0.017 (0.165) ns
correlational selection (ɣij)
floral tube length (mm) × resistance to herbivores (%)	0.100 (0.141) ns	−0.382 (0.395) ns
floral tube length (mm) × arbuscular colonization (%)	−0.155 (0.065) ns	−0.219 (0.189) ns
resistance to herbivores (mm) × arbuscular colonization (%)	−0.014 (0.098) ns	0.642 (0.424) ns

Figure 3.

Phenotypic selection gradients on floral tube length, resistance to herbivores and arbuscular colonization in two Nicotiana longiflora populations (n = 60), one from a natural site (a) and one from an urban site (b). Arrows indicate the peak values of significant adaptive cubic splines for each phenotypic trait analysed.

All selection gradients in the urban site were non-significant (table 1), consistent with the lower occurrence of AMF, hawkmoths and herbivores, and their weaker interaction intensity than in the natural site.

4. Discussion

We evaluated patterns of phenotypic selection on plant traits of N. longiflora exerted by three different agents of selection. Each plant trait was involved in interactions with different above- and belowground mutualistic or antagonistic partners. In a natural site, we found significant selection gradients for floral tube length, resistance to herbivory, and for the first time, we recorded arbuscular colonization intensity as the symbiotic trait under phenotypic selection. In the population of an urban site, the absence of significant selection patterns suggests that phenotypic selection might be absent in this site. However, since populations were not replicated at the level of natural versus urban treatments, we cannot make overall generalizations about differences in the patterns of selection in relation to urbanization, but can assess whether our observations are in the expected direction.

In the natural population, we detected disruptive selection on floral tube length, closely matching the proboscis peak lengths of M. paphus individuals. This is consistent with experimental evidence showing that M. sexta (the sister North American species of M. paphus) exhibits trait co-adaptation with Nicotiana flowers [10]. However, in the urban population studied, selection on floral tube length was not significant, and a significant reduction in the mean floral tube length and correlated anther height was observed, potentially favouring self-pollination. If trait co-adaptation is lost in urban ecosystems, chemical signalling may also be altered, and thus disrupting other plant physiological mechanisms [26]. A decrease in floral herkogamy in response to increasing pollinator limitation associated with urbanization was previously evidenced for another study system [28]. Therefore, further ecological and biological changes in the reproduction of this species could be shaped by human-modified ecosystems, such as urban areas. However, a broader study is needed to confirm if the tendency observed here can be held.

The defence mechanism of resistance to herbivory (i.e. avoidance of herbivory) in N. longiflora was under selection in the natural site. Selection acting on resistance might indicate that this plant species is under attack not only by its specialist herbivore, but also by generalist herbivores, since these specialist herbivores are able to overcome plant secondary metabolites responsible for repelling generalist herbivores [49]. However, since plants simultaneously allocate resources to resistance and tolerance as defence mechanisms, further studies could include tolerance-related plant traits to elucidate the complementarity or redundancy between these two defence strategies in shaping the selective landscape of N. longiflora [50]. The absence of selection for resistance to herbivores in the urban area studied here may indicate a loss of evolved defences, potentially hindering the ability to respond to novel herbivores under changing conditions [26].

The inclusion, to our knowledge for the first time, of arbuscular colonization intensity as the symbiotic trait under phenotypic selection provides a novel approach to understanding selection exerted by these organisms [18,23]. Notably, mycorrhizal colonization in some plant species exhibits moderate heritability, suggesting potential for selection to promote optimal levels of AMF colonization in plants [21]. Since the formation of arbuscules involves structural changes in host plant root cells (e.g. vacuole fragmentation and migration of the plant nucleus from the periphery to the centre of the cell [51]) the degree of arbuscular colonization may reflect a coordinated response between plant and AMF. In line with this, we detected stabilizing selection on mean arbuscular colonization levels, supporting the idea that plants autoregulate AMF colonization [24].

Overall, AMF morphospecies observed for both sites were ruderal, which commonly grow in early-successional plant communities [52]. Since N. longiflora is also a ruderal species, it is expected to be associated with low-rich AMF communities, comprising a great fraction of ruderal taxa [53], which are also more frequent in human-modified environments [54]. In the natural site, where a higher interaction strength occurred, individuals with around 11% of arbuscules in the roots showed higher fruit sets. Two ruderal AMF morphospecies (i.e. E. lutea and F. geosporum) were unique to this natural site. Further controlled experiments should test the effects of these species on N. longiflora performance.

Although we did not find significant correlational gradients on the traits involved in this tripartite interaction, selection exerted by one partner could potentially disrupt interactions with the others. For example, plants could hinder pollinator attraction through chemical signalling when chemical leaf defences that deter floral antagonists also affect pollinators by altering nectar chemistry [42]. Additionally, AMF can influence plant interactions with both pollinators and herbivores [40,55]. For instance, the association with AMF could indirectly promote pollination by enhancing plant growth, leading to larger and more numerous flowers and inflorescences, as well as increased nectar production [56,57]. Conversely, AMF could also promote herbivory by increasing the nutritional quality of plants or reducing it by enhancing the production of defensive compounds [55], which in turn could deter pollination.

In the population of the urban site we observed a decrease in mean absolute fitness, mean fruit number, likely driven by the reduction in the interaction intensity of coevolved partners. Contrary to our expectations, these did not translate into changes in selection strength in this site. Reduction in partners’ occurrence and their interaction intensity, may have relaxed selection of interaction traits under these urban conditions. In this scenario, an increasing intraspecific phenotypic variation and a potential loss of traits that originally contributed to maximum plant fitness would be expected [26,30,58]. However, it is important to consider that the urban site had a lower average number of flowers and fruits per individual, and therefore the fruit set was similar between populations, which could reduce the resolution of fitness estimates and contribute to the lack of significant selection gradients. To effectively test whether urbanization is relaxing phenotypic selection, additional urban sites in similar study systems are necessary. The human-driven weakening or loss of original selective forces exerted by coevolved interacting partners could further lead to the loss of unique functional traits in urban communities that could disrupt interactions, and contribute to ecosystem services and functional weakening in such human-modified ecosystems. On the other hand, relaxed selection, contrary to novel adaptive responses in human-modified ecosystems, could allow the persistence of trait variation as a source of future adaptation in cases where interacting partners can be recovered. The creation of ‘islands of shade’ within urban areas could prevent some of the negative impacts of urbanization on insect diversity [59] and eventually provide the context to restart reciprocal adaptation. For instance, high tree cover in artificially lit environments seems to mitigate the decrease in macro-moth species richness and abundance [59], and favour hawkmoth pollination activity [60]. Provided that hawkmoths, as well as herbivores and AMF, can be protected in an urban environment the process of co-adaptation of plants would be reinstated.

Understanding phenotypic selection on plant traits by different agents of selection is crucial for predicting species responses to environmental change. Selection pressures from both biotic and abiotic agents can act on plant traits, sometimes in conflicting ways, thereby shaping evolutionary outcomes and species resilience [1,61]. While most studies have acknowledged phenotypic selection in pairwise animal–plant interactions [5,50], few have evaluated selection patterns within multispecies plant interactions, as was done in the present study. We found that different plant traits are targets of selection (i.e. flower tube, leaf damage and arbuscular colonization) by multiple agents of selection (i.e. pollinators, herbivores and AMF), simultaneously influencing plant fitness. Furthermore, we observed different selection patterns depending on population contexts, suggesting that environmental context can modify these patterns. These findings underscore the importance of adopting a multispecies, context-dependent approach to gain a deeper understanding of how complex ecological interactions influence plant evolution.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

All data and code associated with this manuscript, currently provided as a compiled file, will be available from Zenodo [62].

Supplementary material is available online [63].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

F.S.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing—original draft, writing—review and editing; S.L.: conceptualization, data curation, methodology, writing—review and editing; N.C.: conceptualization, data curation, funding acquisition, methodology, writing—review and editing; M.L.B.: data curation, funding acquisition, methodology, writing—review and editing; N.M.: conceptualization, data curation, methodology, writing—review and editing; J.R.: data curation; M.C.: data curation; R.P.A.: data curation; S.M.C.: data curation, writing—review and editing; A.G.: data curation, writing—review and editing; A.C.: conceptualization, data curation, formal analysis, funding acquisition, methodology, writing—review and editing; C.V.: data curation, writing—review and editing; G.G.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

The study was financially supported by FONCyT (BID 2018 PICT 3376, BID 2018 PICT 692, BID 2019 PICT 1363), CONICET (PIBAA 2022-2023, PIP 2023-2025), and Ministerio de Ciencia y Tecnología del Gobierno de Córdoba (GRFT 2019).