{"paper_id":"3fbd4967-311f-49d7-af04-3c6ea04b3ba4","body_text":"1 \n \nTitle: Phenotypic evolutionary response to temporally limited pollinator access in Brassica 1 \nrapa 2 \n 3 \nAuthors: Authier Elisabeth1,2, Aeschbacher Simon2,3, Frachon Léa1,4 4 \n 5 \nAffiliations: 6 \n1 Department of Systematic and Evolutionary Botany, University of Zurich 7 \n2 Department of Evolutionary Biology and Environmental Studies, University of Zurich 8 \n3 Swiss National Park, Division of Research and Monitoring, Zernez, Switzerland 9 \n4 Université Bourgogne Europe, Institut Agro Dijon, INRAE, Agroécologie, Dijon, France 10 \n 11 \nAuthors’ e-mail addresses: Elisabeth Authier, Elisabeth.authier@helsinki.fi 12 \nLéa Frachon, lea.frachon@inrae.fr  13 \nSimon Aeschbacher, simon.aeschbacher@uzh.ch 14 \n 15 \nShort running title: Phenotypic response to pollinator access 16 \n 17 \nKeywords: pollinator communities, experimental evolution, selection, phenotypic response, 18 \ngeneralist plant species, plant—pollinator interaction 19 \n 20 \nType of article: letter 21 \n 22 \nAuthor Contributions: LF and SA acquired the University Research Priority Program 23 \n‘Evolution in action” funding, conceptualized and designed the experiment, EA performed the 24 \nexperiment with the contribution of LF, EA performed the analyses, EA wrote the original draft, 25 \nLF and SA reviewed and edited the manuscript. 26 \n 27 \nData Availability Statement: Data and code are available in Zenodo 28 \n(10.5281/zenodo.15425805). 29 \n 30 \nMetrics: 131 words in the abstract, 4971 words in the main text, 84 references, 4 figures, and 31 \n1 table. 32 \n 33 \nCorrespondence: Dr. Léa Frachon, e -mail: lea.frachon@inrae.fr, phone: +33681504491, 34 \naddress: UMR1347 AGROECOLOGIE, 17 rue Sully, BP 86540, 21065 Dijon, Cedex, France 35 \n 36 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n2 \n \nAbstract 37 \nPlant–pollinator interactions are essential for plant reproductive success and pollinator 38 \nfood supply. However, the ongoing pollinator decline threatens many wild and cultivated 39 \nflowering plants, urgently requiring studies on its impact on plant populations and their potential 40 \nevolutionary responses. We combined an experimental evolution with a resurrection approach 41 \nto test phenotypic evolutive changes in response to artificially limited access of plants to natural 42 \npollinators in a common garden setting using Brassica rapa. After six generations, we detected 43 \nputative adaptive responses to the pollination treatments, including changes in phenology, 44 \nfloral morphology and flower volatile organic compounds, associated with changes in overall 45 \nattractiveness to hoverflies. Although the generalist plant B. rapa shows ability to rapidly 46 \nrespond to strong limitation s to its natural pollinator community, the observed decrease of 47 \nfitness could threaten the population. 48 \n 49 \n 50 \n 51 \n 52 \n 53 \n 54 \n 55 \n 56 \n 57 \n 58 \n 59 \n 60 \n 61 \n 62 \n 63 \n 64 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n3 \n \nIntroduction 65 \nAbout 88% of all species of flowering plants worldwide are pollinated by animal pollinators 66 \n(Ollerton et al. 2011) and depend on them for their reproduction. However, in the last few 67 \ndecades, a worldwide decline of domesticated and wild insect pollinators has been observed 68 \n(Kluser & Peduzzi 2007; Potts et al. 2010). This loss of pollinators has also been observed to 69 \nnegatively impact ecosystem services and food production (Aizen et al. 2009; Klein et al. 2007; 70 \nSingh & Adhikary 2021). For instance, local declines in bee and hoverfly species abundance 71 \nin Britain and the Netherlands over the past century (Biesmeijer et al. 2006; Powney et al. 72 \n2019) support growing concerns about the persistence of local pollinator communities. Indeed, 73 \na decrease in pollinator diversity and abundance leads to pollen limitation and reduces plant 74 \nreproductive success (Bennett et al. 2020; Thomann et al. 2013). For instance, lower insect 75 \npollinator diversity negatively affects fruit and seed set in Raphanus sativus (Albrecht et al. 76 \n2012). 77 \nThe extent to which a decrease in pollinator diversity and abundance affects plants may 78 \ndepend on the degree to which plants are specialised to their pollinators ( Ramos-Jiliberto et 79 \nal. 2020). Generalist -pollinated plant species, i.e., plant species interacting with a wide 80 \ndiversity of pollinator species for their reproductive success (Ollerton et al. 2007), play a central 81 \nrole in plant–pollinator networks due to their higher connectivity than other specialist-pollinated 82 \nplant species (Bascompte & Jordano 2007). Indeed, thanks to their broader interaction 83 \nnetworks and inherent resilience, generalist plants are less prone to extinction and potentially 84 \nbetter able to adapt to pollinator loss than specialist species. However, few studies have 85 \nexamined the adaptive potential of these generalist -pollinated plant specie s in response to 86 \npollinator declines or to the destabilization of their pollination networks. Understanding the 87 \nadaptive potential of generalist plants is particularly critical in crops, where pollinator declines 88 \ncould directly threaten crop yields and food security (Klein et al. 2007, Aizen et al. 2009). 89 \nMoreover, the loss of generalist plant species can destabilize plant –pollinator interactions, 90 \ndecrease pollinator abundance (Biella et al.  2019; Palacio et al. 2016), and lead  to chain  91 \nextinctions of plant and pollinator species from the ecological network (Bascompte et al. 2019). 92 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n4 \n \nUnderstanding whether and how generalist-pollinated species respond to changes in access 93 \nto pollinators is crucial for ensuring ecosystem stability. However, only a few studies have 94 \nexplored the adaptive response of generalist plant species to pollinator community 95 \ndisturbance. For instance, rapid phenotypic response s to a single pollinator ha ve been 96 \nobserved in generalists Mimulus guttatus  (Bodbyl Roels & Kelly 2011) and Brassica rapa 97 \n(Gervasi & Schiestl 2017) in greenhouse conditions. Interestingly, Schiestl et al. (2018) found 98 \nthat Brassica rapa exposed to a synthetic pollinator community with one bumblebee and one 99 \nhoverfly species under controlled conditions evolved a ‘generalised -pollination phenotype’ 100 \noverlapping with, but distinct from, the phenotypes seen under pollination by only one of the 101 \ntwo pollinator species. However, in natural ecosystems, generalist plants interact with complex 102 \npollinator communities involving more than two pollinator species. To our knowledge, the 103 \nevolutionary response of generalist plant species to changes in the extent of access to their 104 \nnatural pollinator communities under field conditions remains unknown. 105 \nIn response to limited access to pollinators, flowering plant species can adopt two opposite 106 \nstrategies: decreasing their dependency on pollinators or reinforcing their interaction with 107 \npollinators (Thomann et al. 2013). Specifically, the first strategy for plants in response to limited 108 \naccess to pollinators is to enhance selfing ability, i.e., reproducing using their own pollen, 109 \nensuring reproductive success in the absence of pollinators (Lloyd 1992) and is referred to as 110 \nreproductive assurance. By default, selfing is suppressed in many plant species by physical 111 \nand molecular barriers (Rea & Nasrallah 2008; Wang & Filatov 2023). However, these barriers 112 \ncan disappear within just a few generations in the absence of pollinators, as demonstrated by 113 \nincreased selfing ability in Mimulus guttatus experimental evolution study (Bodbyl Roels & 114 \nKelly 2011). Similarly, selfing ability increased in Brassica rapa in response to less efficient 115 \npollinators for pollination, like hoverflies (Gervasi & Schiestl 2017). In these two examples, the 116 \nincrease in selfing ability was associated with a decrease in herkogamy, a key physical barrier 117 \ndefined as the distance between the stigma and anthers in individual flowers, which promotes 118 \ngamete contact within the same flower (Opedal 2018). However, the evolution of selfing is 119 \ncomplex and involves a set of distinct traits, collectively termed the ‘selfing syndrome’ such as 120 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n5 \n \nsmaller flowers, decreased scent emission, etc. (Barrett 2003; Cutter 2019). However, this 121 \nincreased selfing capacity also leads to population -level costs, such as reduced genetic 122 \ndiversity. A second potential strategy in response to limited access to pollinators is to 123 \nstrengthen interactions with pollinators (Thomann et al.  2013) by increasing plant 124 \nattractiveness (pollinator visitation rates) and pollen transfer efficiency. Flower traits that 125 \ndetermine flower attractiveness are associated with pollinator detection and access to flowers 126 \n(Bauer et al.  2017; Gervasi & Schiestl 2017) such as plant size, flower number, flower 127 \nmorphology, flower colour, composition and amount of nectar, and floral scent produced 128 \n(Klinkhamer & De Jong 1993; Klumpers et al. 2019; Majetic et al. 2009; Wright & Schiestl 129 \n2009). However, it remains unclear whether plants preferentially adopt one strategy over the 130 \nother in response to limited pollinator access. 131 \nAn efficient method to study plant responses to environmental changes is an experimental 132 \nevolution study (Kawecki et al.  2012). Experimental evolution offers the possibility of 133 \nmanipulating a set of environmental conditions to exert selection pressures (Bennett & Lenski 134 \n1999). However, if the aim of an experimental evolution study is to quantify the evolutionary 135 \nresponse of an organism to a set of manipulated conditions under otherwise natural conditions, 136 \nit may be difficult to separate the effect of the focal conditions from those of the uncontrolled 137 \nnatural conditions. Thus, such an in situ experimental evolution approach provides ecological 138 \nrealism but also introduces the possibility of confounding environmental effects. For instance, 139 \nclimatic conditions cannot be fully controlled in in situ experiments but may elicit responses 140 \nthat enhance or attenuate the effects of the experimental manipulation. Distinguishing the 141 \neffects of the experimental manipulation from those of the climate therefore requires recording 142 \nthe climatic conditions over the course of the experiment and including them as explanatory 143 \ncovariates in the statistical analyses. Moreover, the expression of genetically determined focal 144 \ntraits may be affected by uncontrolled environmental conditions and parental effects. These 145 \nfactors may thus confound the comparison of traits measured before and after the 146 \nexperimental treatment. In plants, it is possible to mitigate both of these factors by regenerating 147 \nfrom stored seeds the generations from before and after the evolution experiment under the 148 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n6 \n \nsame controlled conditions in a so -called resurrection approach (Franks et al. 2008, 2018; 149 \nWeider et al. 2017). 150 \nThe overarching aim of our study is to better understand how generalist -pollinated 151 \nflowering plant species respond to temporal limitation in access to their natural pollinator 152 \ncommunity. To achieve this aim, we designed an  experimental evolution study to assess i) 153 \nwhether and how morphological traits and fitness components respond, and ii) how the 154 \nattractiveness of plants to pollinators evolves. 155 \n 156 \nMaterial and Methods 157 \nStudy system and experimental evolution design 158 \nWe used the fast-cycling standard variety of Brassica rapa from Wisconsin Fast Plants® 159 \n(Carolina Biological Supply, Burlington, USA) as our study system (Wendell & Pickard 2007). 160 \nWe randomly divided 108 full-sib seed families (supplementary information) into three sets of 161 \n36, hereafter referred to as “replicates”. Each replicate underwent an independent six -162 \ngeneration evolutionary experiment with three pollination treatments ( “Pollination treatments” 163 \nsubsection). Throughout the manuscript, we refer to a treatment × replicate combination as a 164 \n“replicate population”. We started the experiment with nine initial replicate populations (3 165 \ntreatments × 3 replicates) and propagated these nine replicate populations individually over 166 \nthe experimental generations for each of the three pollination treatments  (Fig. 1 , 167 \nsupplementary information). 168 \nWe grew the 324 individual plants (9 replicate populations x 36 individual plants) in a 169 \nphytotron for two weeks, potted them in standardized soil, and moved them into nine outdoor 170 \ncages surrounded by an insect -proof net that were set up at the Botanical Garden of the 171 \nUniversity of Zürich (Fig. S1). Inside each cage, we randomly positioned  36 individual plants 172 \nfrom a given replicate population (Fig. S1). 173 \n 174 \n 175 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n7 \n \n 176 \nFIGURE 1 | Experimental design. a) 108 full sib -seed families of fast -cycling B. Rapa have been generated from 400 B. rapa seeds, and b) 177 \npropagated over six generations (G1 to G6) in three pollination treatments in common garden, each comprising three parallel replicate populations 178 \nof 36 individuals. c) To remove maternal and environmental effects, seeds from the full -sib families and from the sixth generations of the 179 \nexperimental evolution were regrown in a resurrection experiment under similar greenhouse conditions for two generations (fir st and second 180 \nresurrections). Due to set up constraint, we regrow a second time the seeds of the first resurrection, leading to a third resurrection. d) Treatments 181 \nincluded one control treatment and two pollination treatments in which access of pollinators was temporally limited. e) To pr opagate plants from 182 \none generation to the next, we estimated relative fitness from the product of the total number of seeds produced times the ge rmination rate and 183 \ncorrected the number of seedlings in each pot by the estimated relative fitness of its maternal plant. 184 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n8 \n \n \nDuring the experimental evolution study, we measured 17 traits including seven 185 \narchitectural and morphological traits , and 10 fitness-related traits at each generation (see 186 \ntable S 1, supplementary information).  Climate data was recorded (temperature and light 187 \nintensity, supplementary information) using a datalogger installed in the central outdoor cage, 188 \nand principal component analysis was performed to summarize the variation of the 10 climatic 189 \nvariables (Table S2). 190 \n 191 \nPollination treatments 192 \nAfter moving plants into outdoor cages, we applied  three pollination treatments by 193 \nartificially manipulating temporal pollinator access to simulate a decrease in pollinator richness 194 \nand abundance. 1) Control treatment: we hand -pollinated plants by transferring pollen from 195 \nflowers of a given plant to one to six flowers of the nearest plant. We kept the three cages 196 \nassigned to the three replicates in this treatment constantly closed to exclude insect pollinators. 197 \n2) Full Access treatment: we let the natural surrounding insect pollinator community pollinate 198 \nthe plants. We opened each of the three cages once during the life cycle for five consecutive 199 \nhours during the highest pollinator activity (between 11 am and 4 pm). 3) Limited Access 200 \ntreatment: we temporally reduced the access of the natural pollinator community to plants 201 \ncompared to the Full-Access treatment. We opened each of the three cages assigned to the 202 \nthree replicates in this treatment once during the life cycle  for one hour during the highest 203 \npollinator activity (between 12 pm and 1 pm).  Our observations of pollinator visits during 204 \ntreatment application (supplementary information) revealed significantly lower pollinator 205 \nabundance and diversity in the Limited Access treatment compared to the Full Access one 206 \n(Supplementary information, Table S3, Dataset 1). 207 \n 208 \nResurrection experiments 209 \nTo reduce maternal and environmental effects that could confound generation effects, we 210 \nconducted two refresher generation resurrection experiments (Fig. 1c , supplementary 211 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n9 \n \n \ninformation). We sowed seeds from the first and sixth generations together to generate a total 212 \nof 12 replicate resurrection populations (three initial replicate populations plus three treatments 213 \n× three replicate populations from the sixth generation). We grew these replicate resurrection 214 \npopulations in a phytotron for two weeks, and then transferred them into a greenhouse (Irchel 215 \nCampus, University of Zürich). Two weeks later, within each replicate resurrection population, 216 \nwe cross-pollinated eight flowers per plant with pooled pollen  from the same replicate 217 \nresurrection population. At fruit maturity, we generated a second resurrection generation to 218 \ndecrease maternal effect by growing one offspring per maternal plant (n = 36) following the 219 \nsame protocol and conditions as previously described. Finally, we repeated a third resurrection 220 \nexperiment in order to estimate changes in mating system  using seeds from the first 221 \nresurrection (Fig. 1c). In the manuscript, unless contraindicated, we use the term 'resurrection 222 \nexperiment' without distinction, referring to either the second or third resurrection depending 223 \non the context. 224 \nDuring the resurrection experiments, we measured 33 traits including: flowering time, eight 225 \narchitectural and morphological traits, 13 flower volatile organic compounds (VOCs), and 11 226 \nfitness-related traits (Table S1, supplementary information). 227 \n 228 \nPollinator preference 229 \nTo investigate the effect of pollination treatments on  the overall flower attractiveness to 230 \npollinators, we performed a four-choice test. In this test, an insect pollinator was given a choice 231 \namong four focal plant individuals from the four different resurrection populations (initial 232 \ngeneration, sixth generation of Control, Full Access, and Limited Access) . We used the 233 \npollinator’s first choice as a preference proxy. During the four -choice test, we released one 234 \nbumblebee (Bombus terrestris) or one adult hoverfly (Episyrphus balteatus) inside a cage. In 235 \ntotal, we performed 106 four -choice tests for each pollinator species . On the test day, we 236 \ncounted the number of open flowers for each of the four focal plants. 237 \n 238 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n10 \n \n \nStatistical analyses 239 \nWe tracked phenotypic changes over time during the experimental evolution. Based on the 240 \nmean values of traits across plant individuals within each replicate population times generation, 241 \nwe tested separately (1) the overall effect of pollination treatment, and (2) the overall effect of 242 \ngeneration with Kruskal—Wallis test. 243 \nTo quantify both the effect of climate and that of generations, we performed two 244 \nindependent multi-way ANOVAs to explain trait changes by treatment, and either by generation 245 \nor by climate for the ten traits showing a normal distribution in R v4.4.1 according to the 246 \nfollowing equations: 247 \nTrait ~ treatment x generation + (1 | replicate population)  (anova 1) 248 \nTrait ~ treatment x climatePCs + (1 | replicate population)  (anova 2) 249 \nwhere Trait represents the mean of 10 phenotypic and fitness-related traits, treatment the three 250 \npollination treatment s, generation a categorial explanatory variable representing the six 251 \ngenerations, climatePCs the first two principal components derived from the climate PCA, and 252 \nreplicate population as random factor. 253 \nTo detect phenotypic and fitness -related trait evolutionary changes, we used arithmetic 254 \nmean for each trait from the resurrection experiment dataset. First, to understand changes in 255 \ntrait relationships across treatments, we calculated Spearman’s pairwise rank correlation 256 \ncoefficients (package Hmisc v5.0-1, Harrell Jr 2003) in R environment among traits within each 257 \nfour resurrection populations. Additionally, to assess evolutionary changes occurred over six 258 \ngenerations, we performed Kruskal—Wallis tests on arithmetic mean of 33 traits within each 259 \nreplicate resurrection population, with treatment as a categorical explanatory variable . For 260 \ntraits with significant results (p -value < 0.05), we performed multiple pairwise Kruskal -Wallis 261 \ntests (R package pgirmess v 2.0.3, Giraudoux et al. 2024). 262 \nTo quantify directional selection, we applied the Lande & Arnold framework (Lande & 263 \nArnold 1983) to explain the effect of uncorrelated (significant rhoSpearman < 0.8) phenotypic traits 264 \n(fixed-effect explanatory variables) on relative fitness estimate. We performed three directional 265 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n11 \n \n \nselection analysis based (1) on five uncorrelated phenotypic traits that passed the Shapiro test, 266 \n(2) on seven uncorrelated phenotypic traits related to phenology and floral morphology traits, 267 \nand (3) on 11 uncorrelated traits associated with VOCs. The relative fitness estimate was the 268 \nrelative number of seeds per plant within the respective replicate resurrection population. We 269 \nincluded replicate as a random slope, and the phenotypic traits were centred and scaled within 270 \ntheir replicate resurrection population for each generation. We interpreted the partial 271 \nregression coefficients pertaining to the explanatory variables as directional selection 272 \ngradients. 273 \nTo test changes in overall plant attractiveness, we performed a four-choice experiment and 274 \ntested the pollinator preference  for one of the four resurrection treatments. We performed 275 \npairwise comparison of proportions (pairwise.prop.test function) of chosen plants between the 276 \ninitial and the sixth generation for the three treatments. To see if the distribution of the 277 \nphenotypic traits differs between plants chosen and not chosen by pollinators, we performed 278 \npairwise Wilcoxon tests in R for each phenological and phenotypic trait by using the arithmetic 279 \nmeans between plants chosen and not chosen by either bumblebees or hoverflies. 280 \n 281 \nResults 282 \nTracking phenotypic changes over generations 283 \nWe assessed the overall effects of treatment  and generation on 17 traits using pairwise 284 \nKruskal-Wallis tests. Additionally, to disentangle the potential confounding effects between 285 \ngeneration and climate, we performed two multi -way ANOVAs: included treatment and 286 \ngeneration as factors  (anova 1) , or treatment and climate  as factor s (anova 2) . First, we 287 \nobserved a significant overall effect of treatment on seven out of ten fitness-related traits, but 288 \nno treatment effect on floral morphological traits (Table S4). Second, we observed a significant 289 \noverall effect of generation (considering all replicate population together) on 11 traits related 290 \nto morphological traits and fitness-related traits (Fig. 2, Table S4). We also detected significant 291 \neffects of both generation (anova 1) and climate (anova 2) for all tested traits (“within” replicate 292 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n12 \n \n \npopulations in Dataset 2), as illustrated by mean flower diameter (Fig. 2a) . These results 293 \nrevealed variations across generations , which is confounded by the distinct climate 294 \nexperienced by each generation , emphasizing the importance of combining experimental 295 \nevolution in natural condition with resurrection approach.  296 \n 297 \n 298 \n 299 \n 300 \n 301 \n 302 \n 303 \nFIGURE 2 | Tracking phenotypic trait evolution across six generations of the experimental 304 \nevolution study in common garden. Empirical distribution of (a) mean flower diameter and (b) 305 \nmean herkogamy per pollination treatment as measured directly in the experimental evolution 306 \nstudy. 307 \n 308 \nEvolutionary changes in phenotypic trait correlations  309 \nTo assess the variation in phenotypic correlations between generations and according to 310 \ntreatments, pairwise Spearman correlations among 30 phenotypic traits  and fitness-related 311 \ntraits measured in resurrection experiment  were performed. These correlations showed 312 \ndifferent patterns according to the resurrection treatments (initial generation , three evolved 313 \npollination treatments), and to the trait sets considered (Fig. S2). For instance, we observed a 314 \ndecrease in the strength of positive correlations between the flowering time and flower 315 \nmorphology traits from the initial generation to the last generation across all treatments ( Fig. 316 \nS2). Moreover, we observed a shift from negative correlation in initial generation to positive 317 \ncorrelation between some flower morphology traits and amounts of flower scent components, 318 \nespecially in Full and Limited access treatments  (Fig. S2) . In contrast, a slight change in 319 \ncorrelation from positive to negative appear ed between fitness -related traits and the floral 320 \nVOCs from initial to last generation for Control and Limited access treatments (Fig. S2). Finally, 321 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n13 \n \n \nfor certain sets of traits, the correlations within them remain overall stable during the 322 \nevolutionary process like the floral VOCs, the floral morphological traits, and the fitness-related 323 \ntraits. 324 \n 325 \nRapid phenotypic evolutionary response to pollination treatments 326 \nTo unravel the phenotypic differentiation among the replicate resurrection populations, we 327 \nperformed a linear discriminant analysis (LDA)  on different categories of traits. We observed 328 \nan overall phenotypic differentiation among the initial generation and the replicate resurrection 329 \npopulations of the last generation  (Fig. S3). We found phenotypic differentiation among 330 \ntreatment populations in LDA based on floral morphological traits and on fitness-related traits 331 \n(Fig. S3ac) highlighted by a non -overlapping of centroid. However, these phenotypic 332 \ndifferentiation within trait categories are slight, as indicated by the overlap of the ellipses in the 333 \nspace described in LDA. In contrast, i n LDA based on floral VOCs, the centroid of replicate 334 \npopulations from evolved pollination treatments clustered together but remained distinct from 335 \nthe initial generation (Fig. S3e). These results indicate d some phenotypic changes of 336 \ncombination of traits in response to different selection-mediated treatments. 337 \nTo investigate the treatment effect (initial generation and the three evolved pollination 338 \ntreatments) on phenotypic trait variations, we performed Kruskal —Wallis test.  Overall, we 339 \nobserved significant evolutionary changes across treatments for 19 out of 33 measured traits 340 \nrelated to morphological and fitness-related traits i.e. significant differences between the initial 341 \ngeneration and at least one of the evolved populations (Table 1). Interestingly, our findings 342 \nrevealed significant differences between the initial and the last generations in the Full and/or  343 \n 344 \n 345 \n 346 \n 347 \n 348 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n14 \n \n \nTABLE 1| Results of Kruskal—Wallis tests to explain the variation of 33 traits measured in the 349 \nresurrection experiments by treatment as a categorial explanatory variable. Two sets of tests 350 \nhave been performed: (1) an overall Kruskal—Wallis tests and (2) multiple pairwise Kruskal—351 \nWallis tests (between each per of treatments). For each trait, the arithmetic mean within each 352 \ntreatment has been calculated and indicated on the right of the table. The four pollination 353 \ntreatments are: Initial sib -seed families (\"G0\"), Control (\"C\"), Full access (\"FA\"), and Limited 354 \naccess (\"LA\"). Significant p-values (<0.05) are indicated in bold. 355 \n 356 \n 357 \n 358 \n 359 \n 360 \n 361 \n 362 \n 363 \n 364 \n 365 \n 366 \n 367 \n 368 \n 369 \n 370 \n 371 \n 372 \n 373 \n 374 \n 375 \n 376 \n 377 \n 378 \n 379 \nLimited Access treatment, but not in the Control, across 13 traits associated with different trait 380 \ncategories, highlighting evolution ary response to the pollination treatment, rather than 381 \nuncontrolled environmental effects that would also be present in the control treatment . For 382 \ninstance, we detected a significant delay in flowering time  for both Limited and Full access 383 \ntreatments, but not in Control treatment (Fig. 3a, Table 1). Moreover, we observed significant 384 \nevolutionary changes for 11 out of 13 floral VOCs with, for instance, a decrease in benzyl nitrile 385 \nin Limited Access and Full Access compared to the initial generation (Fig. 3b, Table 1).  Finally, 386 \nwe detected significant evolutionary changes in outcrossing ability with a decrease in fruit 387 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n15 \n \n \nlength and seed weight across generations in Limited and Full Access treatments (Fig. 3c , 388 \nTable 1). However, we did not observe significant evolutionary changes for autogamy ability 389 \nfor none of the pollination treatments (Table 1). Finally, we observed evolutionary changes in 390 \nsix traits where significant changes occurred in the Limited and/or Full Access treatments, but 391 \nalso in the Control treatment (Table 1). This introduces uncertainty in attributing these 392 \nevolutionary changes to the pollination treatment (s) rather than uncontrolled environmental 393 \neffects selecting in our populations. 394 \n 395 \n 396 \nFIGURE 3 | Evolutionary changes observed in resurrection approach for the different 397 \npollination treatments. Empirical distribution of a) the flowering time; b) the amount of benzyl 398 \nnitrile in floral VOCs; and (c) the number of seeds per fruit under outcrossing. For each plot, 399 \nthe initial populations and the three evolved populations (Control, Full Access, and Limited 400 \nAccess) are represented. Significant pairwise comparison between means of distributions 401 \n(pairwise Kruskal-Wallis test) are emphasized. 402 \n 403 \nChanges in the strength of directional selection 404 \nFinally, to quantify changes in the strength and the direction of the selection for phenotypic 405 \ntraits before and after potential selection within different pollination treatments, we investigated 406 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n16 \n \n \ndirectional selection analyses in resurrection experiment.  These analyses have been 407 \nperformed on two different set of traits: (1) with all plant architecture and flower morphology; 408 \nand (2) VOCs floral traits. In initial generation, two phenotypic traits were under positive 409 \nselection (height of the plant, number of inflorescences, Table S5a), and one under negative 410 \nselection (pistil length). The strength of directional selection increased in the control treatment 411 \nfor these traits. However, we observed different patterns for both pollination treatments. In the 412 \nFull Access, we observed the emergence of negative selection for flowering time, and positive 413 \nselection for petal width (Table S5a). In the Limited Access, we observed a strengthening of 414 \npositive selection on plant height (Table S5a). Finally, we did not observe directional selection 415 \nin traits related to scent compounds, either in the initial generation or in evolved populations  416 \n(Table S5b). 417 \n 418 \nPlant evolution driven by pollinator preferences  419 \nTo determine whether the phenotypic responses observed in our experiment is associated 420 \nwith evolutionary changes in overall plant attractiveness, we tested for potential shifts in 421 \npollinator preference among  resurrection treatments (initial generation, Control, Limited 422 \nAccess, and Full Access). In four-choice test experiment, we assessed the preference of two 423 \ncommon pollinators (bumblebees and hoverflies) in our common garden that visited our plants 424 \nduring experimental evolution (Dataset 1). We did not observe distinct pollinator preferences 425 \nbetween the initial generation and the evolved populations, except for a lower preference of 426 \nhoverflies for plants in Limited Access over the ones from initial generation (Fig. 4ad, Table 427 \nS6). 428 \n 429 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n17 \n \n \n 430 \nFIGURE 4 | Preference of Bombus terrestris and Episyrphus balteatus  for treatment and 431 \nphenotypic traits. Preference of a) bumblebees (Bombus terrestris) and d) hoverflies 432 \n(Episyrphus balteatus) for resurrection pollination treatments. Distributions of trait values of 433 \nplants chosen vs. not chosen as the first preference by bumblebees (light bars) and hoverflies 434 \n(dark bars) for b) the number of flowers, c) herkogamy, and amounts of e) nonanal and f) 435 \nbenzyl nitrile in floral VOCs. The ‘number of first choices’ refers to the number of times a plant 436 \nfrom a specific treatment was chosen as the first plant visited during the four-choice tests. For 437 \na) and b), significant pairwise comparisons of proportions are emphasized (***: 0 ≤ p ≤ 0.001, 438 \n**: 0.001 < p ≤ 0.01, *: 0.01 < p ≤ 0.05, ▪: 0.05 < p < 0.1, ns: non-significant). For c), d), e), and 439 \nf) significant pairwise Wilcoxon tests are emphasized (***: 0 ≤ p ≤ 0.001, **: 0.001 < p ≤ 0.01, 440 \n*: 0.01 < p ≤ 0.05, ns: non-significant). 441 \n 442 \n 443 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n18 \n \n \nMoreover, we tested the overall preference of bumblebee and hoverfly for specific 444 \nphenotypic traits. For instance, we observed a preference for high number of flowers for both 445 \nspecies of pollinators (Fig. 4b, Table S6), as well as a preference of bumblebees  for earlier 446 \nflowering time (Table S6). Additionally, we observed a significant preference of bumblebees  447 \nfor smaller pistil length, negative herkogamy, and lower amount of methyl benzoate and higher 448 \namount of phenylethyl alcohol in floral VOCs (Fig. 4, Table S6). Similarly, we found that 449 \nhoverflies preferred higher amount of limonene and methyl salicylate, and lower amount of 2-450 \namino benzaldehyde (Table S6). Lastly, we observed a shared preference of bumblebees and 451 \nhoverflies for higher amount of nonanal and benzyl nitrile in floral VOCs (Fig. 4e and f, Table 452 \nS6). 453 \n 454 \nDiscussion 455 \nIn response to disturbances in natural pollinator communities, generalist flowering plants 456 \nmay rapidly adapt by increasing their pollinator attractiveness or enhancing selfing ability. In a 457 \nsix-generation common-garden evolutionary experiment with three varying temporal access to 458 \nnatural pollinators, we observed evolutionary changes in morphological and fitness -related 459 \ntraits. However, distinguishing these changes from the effects of distinct environmental 460 \nconditions across generations is challenging. To minimize environmental and maternal effects, 461 \nwe conducted a resurrection experiment in a controlled greenhouse, growing together the 462 \ninitial and the last generations. We found rapid phenotypic changes in response to variation in 463 \npollinator access, including i) reduced outcrossing ability under Limited and Full Access and ii) 464 \nshifts in phenology (delay of flowering), floral morphology (decrease in petal size), and floral 465 \nscent under Limited Access . Finally, contrary to our assumptions, we detected reduced 466 \nattractiveness to hoverflies in Limited Access. 467 \n 468 \nOur resurrection approach revealed that temporally limited access to natural pollinators 469 \ncan drive rapid shifts in plant mating systems. Specifically, we observed a decrease in 470 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n19 \n \n \noutcrossing ability in both the Limited and Full Access treatments. This mating  system shift 471 \nwas associated with reduced fruit length under outcrossing , and fewer seeds per fruit under 472 \noutcrossing, with higher effect in the Limited Access. Such reduction in outcrossing ability was 473 \nalso documented in a controlled greenhouse evolutionary experiment in response to hoverfly 474 \npollination (Gervasi and Schiestl 2017, Kofler et al. 2024). These changes could result from a 475 \ndecrease in pollinator visits, which may reduce pollen quantity — suggested but not measured 476 \n— potentially leading to pollen limitation (Burd 1994).  However, while a decrease in 477 \noutcrossing ability is known as adaptive response to reduced pollinator availability, it is often 478 \nassociated with an increase in selfing (Acoca-Pidolle et al. 2024; Cheptou et al. 2022; Bodbyl 479 \nRoels & Kelly 2011). In our study, we did not observe such significant changes; however, we 480 \ndid observe a non -significant trend toward increased selfing ability. The lack of significant 481 \nchanges in selfing ability  may reflect known barrier s preventing selfing in Brassica (e.g., 482 \ngenetic self-incompatibility, floral morphology, Brugière et al. 2000, Nasrallah 2017, Murase et 483 \nal. 2020), or a latent period before any evolutionary shift in selfing ability becomes detectable. 484 \nOur findings suggested that small populations facing pollinator community disturbance are at 485 \nhigh risk of decline. In fact, the reduction in outcrossing compromises genetic diversity, while 486 \nthe absence of increased selfing, which could provide reproductive insurance, further reduces 487 \nreproductive success, creating a dual threat to population viability.  488 \nIn addition, our study highlighted rapid evolution of phenology and floral morphology in 489 \nresponse to pollination treatments. Both Limited and Full Access treatments exhibited a delay 490 \nin flowering time, with a delay more pronounced in the Limited Access. In our study, this delay 491 \nin flowering time was associated with a reduced preference of both bumblebees and hoverflies. 492 \nOur finding is in line with previous studies in which pollinators drive changes in flowering time 493 \n(Elzinga et al. 2007, Xu 2023). Although flowering time is often linked to climate adaptation 494 \n(Fournier-Level et al. 2022; Geissler et al. 2023; Preston & Fjellheim 2022), we detected no 495 \nsignificant differences between the initial population and the Control treatment, confirming that 496 \nchanges arose from pollination treatments rather than environmental factors  in our study . 497 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n20 \n \n \nHowever, while we observed this delay in flowering time, we highlighted a significant negative 498 \nselection of this phenological trait in Full Access (delay flowering time correlated with lower 499 \nfitness). This finding could be explained by several hypotheses: (1) selection acting on a trait 500 \nnegatively correlated with flowering time, (2) a recent shift in selection that has not yet 501 \nmanifested in phenotypic changes, or (3) selection favouring synchrony with pollinator peak 502 \nactivities rather than early flowering per se. Moreover, we observed shifts in floral morphology 503 \nwith a decrease in petal size in the Limited Access treatment. This flower reduction pattern is 504 \noften observed as a response to loss of pollinators (Tusuubira and Kelly 2024, Acoca-Pidolle 505 \net al.  2024). Indeed, an absence of pollinator can lead to a shift in mating -system, from 506 \noutcrossing animal-pollination to selfing pollination. This transition, called ‘selfing syndrome’, 507 \nis accompanied by phenotypic changes reducing the plant attractiveness to pollinators, 508 \nincluding decrease in flower size  (Tsuchimatsu and Fujii, 2022 ). While we did not observe 509 \nsignificant increase in selfing ability , only a non -significant trend , this floral morphological 510 \nchange may represent an early evolutionary shift toward a selfing syndrome, potentially 511 \nleading to increased selfing rates in subsequent generations. 512 \nAdditionally, we observed evolutionary changes in many floral scent compositions in only 513 \nsix generations. Pollinators often show strong preferences for a specific or a combination of 514 \nfloral VOCs , and previous studies suggested that pollinator -mediated selection can drive 515 \nchanges in floral scent composition in B. rapa (Dorey & Schiestl 2024; Gervasi & Schiestl 2017; 516 \nRamos & Schiestl 2019). Among the observed changes in VOCs, most floral VOCs are known 517 \nto influence pollinator preferences . For instance,  while methyl salicylate is associated with 518 \nhoverfly preference, limonene, p-anisaldehyde, 2-aminobenzaldehyde, benzaldehyde, methyl 519 \nsalicylate, and nonanal are linked to preferences in Hymenoptera (Dötterl & Gershenzon 520 \n2023). For example, we confirmed that plants with lower nonanal levels are less frequently 521 \nchosen by both bumblebees and hoverflies. Additionally, we observed evolutionary reductions 522 \nin nonanal production across six generations in both Limited and Full Access treatments, 523 \nfurther demonstrating that our pollination treatments decrease plant attractiveness. Due to the 524 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n21 \n \n \ninvolvement of many VOCs and their shared biosynthetic pathways (Dötterl & Gershenzon 525 \n2023), the target of selection in our study is unlikely to be a specific VOC compound, as 526 \nconfirmed by the lack of consistent directional selection for these traits. Instead, selection likely 527 \nacts on the floral scent bouquet as a whole.  528 \nLastly, we detected an overall decrease in plant attractiveness to hoverflies in evolved 529 \nplants from the Limited Access treatment , for which we observed a significantly lower 530 \nabundance of hoverflies over the six generations of selection compared to the Full Access 531 \ntreatment. The decrease in hoverfly attractiveness in Limited Access can be explained by 532 \nchanges observed in a combination of phenotypic traits, including VOCs, in response to low 533 \noccurrence of these pollinators. For instance, we documented evolutionary reductions in VOCs 534 \nsuch as limonene, nonanal, and benzyl nitrile across both Limited and Full Access treatments. 535 \nCorrespondingly, plants producing lower levels of these compounds were significantly less 536 \nattractive to hoverflies, while these compounds are known to be involved in plant attractiveness 537 \nto hoverflies (Dötterl & Gershenzon 2023). Additionally, we observed that the Limited Access 538 \ntreatment reduced petal size, which aligns with the overall decrease in plant attractiveness . 539 \nBecause producing floral scent and enhancing floral visibility to improve plant attractiveness 540 \nare costly (Spigler et al. 2020), one plausible hypothesis for this reduced hoverfly preference 541 \nfor plants evolved in Limited Access is that these plants have adapted by reducing their overall 542 \nattractiveness in response to reduced pollinator community. In contrast, we observed no such 543 \ntrend in bumblebee preferences, possibly because there was no substantial difference in 544 \nbumblebee abundance between the Limited and Full Access treatments. However, we tested 545 \nplant attractiveness using only two common pollinators under controlled conditions, which does 546 \nnot reflect the ecological realism of the environments where these plants have evolved. It would 547 \ntherefore be valuable to test the attractiveness of the evolved plants to a broader range of 548 \npollinators, including honeybees or small bees – groups for which we observed abundance 549 \ndifferences between treatment s - as well as  within the context of a complete pollinator 550 \ncommunity. Moreover, six generations are a short time period for selective processes, and the 551 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n22 \n \n \nvariation of our selective pressure (pollinator community composition) across generations, may 552 \nhave buffered the speed of selection and likely imposed diffuse selection on traits.  553 \n 554 \nOverall, our study showed that temporally limiting generalist plants’ access to their natural 555 \npollinator community can drive short-term changes in mating system and overall plant 556 \nattractiveness. It therefore appeared that flowering plants have substantial adaptive capacity, 557 \nparticularly in traits affecting pollinator attractiveness (floral morphology and scent). However, 558 \nthe short time period of our experimental evolution provides only a partial view of how plants 559 \nevolve in response to a decline in pollinators, and the long-term consequences of pollinator 560 \ndecline remain unclear. Reducing pressure on natural pollinator communities should therefore 561 \nremain a priority. 562 \n 563 \nAcknowledgement 564 \nWe thank Markus Meierhofer, Rayko Jonas, Daniel Schlagenhauf, Matthias Furler, Franz 565 \nHuber, Laura Dällenbach for their help during experiments. We thank Florian Schiestl, Tobias 566 \nZüst, Maxime Bonhomme and Carolin Kosiol for helpful discussions and inputs during the 567 \nstudy. This work was supported by the University of Zürich and the University Research Priority 568 \nProgram ‘Evolution in action’.  569 \n 570 \n 571 \n 572 \n 573 \n 574 \n 575 \n 576 \n 577 \n 578 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n23 \n \n \nSupplementary Materials 579 \nDataset 1: raw data pollinator abundance and diversity 580 \nDataset 2: result of multi-ways ANOVAs (models 1 and 2) – experimental evolution 581 \nDataset 3: raw data individual traits – experimental evolution 582 \nDataset 4: raw data flower morphology – experimental evolution 583 \nDataset 5: raw data fruit length – experimental evolution 584 \nDataset 6: temperature and light recording – experimental evolution 585 \nDataset 7: raw data individual traits – second resurrection 586 \nDataset 8: raw data flower morphology – second resurrection 587 \nDataset 9: raw data fruit length – second resurrection 588 \nDataset 10: raw data individual traits – third resurrection 589 \nDataset 11: raw data fruit length – third resurrection 590 \n 591 \n 592 \n 593 \n 594 \n 595 \n 596 \n 597 \n 598 \n 599 \n 600 \n 601 \n 602 \n 603 \n 604 \n 605 \n 606 \n 607 \n 608 \n 609 \n 610 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n24 \n \n \nReferences 611 \nAcoca-Pidolle, S., Gauthier, P., Devresse, L., Deverge Merdrignac, A., Pons, V. & Cheptou , 612 \nP.-O. (2024). Ongoing convergent evolution of a selfing syndrome threatens plant –613 \npollinator interactions. New Phytol., 242, 717–726. 614 \nAizen, M.A., Garibaldi, L.A., Cunningham, S.A. & Klein, A.M. (2009). How much does 615 \nagriculture depend on pollinators? Lessons from long -term trends in crop production. 616 \nAnn. Bot., 103, 1579–1588. 617 \nAlbrecht, M., Schmid, B., Hautier, Y. & Müller, C.B. (2012). Diverse pollinator communities 618 \nenhance plant reproductive success. Proc. R. Soc. B., 279, 4845–4852. 619 \nBarrett, S.C.H. (2003). Mating strategies in flowering plants: the outcrossing–selfing paradigm 620 \nand beyond. Phil. Trans. R. Soc. Lond. B, 358, 991–1004. 621 \nBascompte, J., García, M.B., Ortega, R., Rezende, E.L. & Pironon, S. (2019). Mutualistic 622 \ninteractions reshuffle the effects of climate change on plants across the tree of life. Sci 623 \nAdv, 5, eaav2539. 624 \nBascompte, J. & Jordano, P. (2007). Plant -Animal Mutualistic Networks: The Architecture of 625 \nBiodiversity. Annu. Rev. Ecol. Evol. Syst., 38, 567–593. 626 \nBauer, A.A., Clayton, M.K. & Brunet, J. (2017). Floral traits influencing plant attractiveness to 627 \nthree bee species: Consequences for plant reproductive success. Am J Bot, 104, 772–628 \n781. 629 \nBennett, A.F. & Lenski, R.E. (1999). Experimental Evolution and Its Role in Evolutionary 630 \nPhysiology. Am Zool, 39, 346–362. 631 \nBennett, J.M., Steets, J.A., Burns, J.H., Burkle, L.A., Vamosi, J.C., Wolowski, M., et al. (2020). 632 \nLand use and pollinator dependency drives global patterns of pollen limitation in the 633 \nAnthropocene. Nat Commun, 11, 3999. 634 \nBiella, P., Akter, A., Ollerton, J., Tarrant, S., Janeček, Š., Jersáková, J., et al.  (2019). 635 \nExperimental loss of generalist plants reveals alterations in plant-pollinator interactions 636 \nand a constrained flexibility of foraging. Sci Rep, 9, 7376. 637 \nBiesmeijer, J.C., Roberts, S.P.M., Reemer, M., Ohlemüller, R., Edwards, M., Peeters, T., et al. 638 \n(2006). Parallel Declines in Pollinators and Insect -Pollinated Plants in Britain and the 639 \nNetherlands. Science, 313, 351–354. 640 \nBodbyl Roels, S.A. & Kelly, J.K. (2011). Rapid evolution caused by pollinator loss in mimulus 641 \nguttatus: rapid evolution after pollinator loss. Evolution, 65, 2541–2552. 642 \nBoff, S., Henrique, J.A., Friedel, A. & Raizer, J. (2021). Disentangling the path of pollinator 643 \nattraction in temporarily colored flowers. Int J Trop Insect Sci, 41, 1305–1311. 644 \nBrugière, N., Rothstein, S.J., Cui, Y. (2000). Molecular mechanisms of self -recognition in 645 \nBrassica self-incompatibility. Trends Plant Sci. 5, 10, 432-438. 646 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n25 \n \n \nBurd, M. (1994). Bateman’s principle and plant reproduction: The role of pollen limitation in 647 \nfruit and seed set. Bot. Rev, 60, 83–139. 648 \nCarr, D.E. & Eubanks, M.D. (2014). Interactions between insect herbivores and plant mating 649 \nsystems. Annu Rev Entomol, 59, 185–203. 650 \nCheptou, P.-O., Imbert, E. & Thomann, M. (2022). Rapid evolution of selfing syndrome traits 651 \nin Viola arvensis revealed by resurrection ecology. Am. J. Bot., 109, 1838–1846. 652 \nChristopher, D.A., Karron, J.D., Semski, W.R., Smallwood, P.A., Trapnell, D.W. & Mitchell, 653 \nR.J. (2021). Selfing rates vary with floral display, pollinator visitation and plant density 654 \nin natural populations of Mimulus ringens. J Evol Biol, 34, 803–815. 655 \nCroissant, Y. (2020). Estimation of Random Utility Models in R: The mlogit Package. J. Stat. 656 \nSoftw., 95, 1–41. 657 \nCutter, A.D. (2019). Reproductive transitions in plants and animals: selfing syndrome, sexual 658 \nselection and speciation. New Phytol., 224, 1080–1094. 659 \nDorey, T., Frachon, L., Rieseberg, L.H., Kreiner, J.M. & Schiestl, F.P. (2024). Biotic interactions 660 \npromote local adaptation to soil in plants. Nat Commun, 15, 5186. 661 \nDorey, T. & Schiestl, F.P. (2024). Bee-pollination promotes rapid divergent evolution in plants 662 \ngrowing in different soils. Nat Commun, 15, 2703. 663 \nDötterl, S. & Gershenzon, J. (2023). Chemistry, biosynthesis and biology of floral volatiles: 664 \nroles in pollination and other functions. Nat. Prod. Rep., 40, 1901–1937. 665 \nElzinga, J.A., Atlan, A., Biere, A., Gigord, L., Weis, A.E., Bernasconi, G. (2007). Time after 666 \ntime: flowering phenology and biotic interactions. Trends Ecol. Evol., 22, 8, 432—439. 667 \nDray, S. & Dufour, A. -B. (2007). The ade4 Package: Implementing the Duality Diagram for 668 \nEcologists. J. Stat. Softw., 22, 1–20. 669 \nFournier-Level, A., Taylor, M.A., Paril, J.F., Martínez-Berdeja, A., Stitzer, M.C., Cooper, M.D., 670 \net al. (2022). Adaptive significance of flowering time variation across natural seasonal 671 \nenvironments in Arabidopsis thaliana. New Phytol., 234, 719–734. 672 \nFrachon, L. & Schiestl, F.P. (2024). Rapid genomic evolution in Brassica rapa with bumblebee 673 \nselection in experimental evolution. BMC Ecol Evo, 24, 7. 674 \nFranks, S.J., Avise, J.C., Bradshaw, W.E., Conner, J.K., Etterson, J.R., Mazer, S.J., et al. 675 \n(2008). The Resurrection Initiative: Storing Ancestral Genotypes to Capture Evolution 676 \nin Action. Biosci., 58, 870–874. 677 \nFranks, S.J., Hamann, E. & Weis, A.E. (2018). Using the resurrection approach to understand 678 \ncontemporary evolution in changing environments. Evol. appl., 11, 17–28. 679 \nGeissler, C., Davidson, A. & Niesenbaum, R.A. (2023). The influence of climate warming on 680 \nflowering phenology in relation to historical annual and seasonal temperatures and 681 \nplant functional traits. PeerJ, 11, e15188. 682 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n26 \n \n \nGervasi, D.D.L. & Schiestl, F.P. (2017). Real -time divergent evolution in plants driven by 683 \npollinators. Nat Commun, 8, 14691. 684 \nGiraudoux, P., Antonietti, J. -P., Beale, C., Groemping, U., Lancelot, R., Pleydell, D., et al. 685 \n(2024). pgirmess: Spatial Analysis and Data Mining for Field Ecologists. 686 \nGómez, J.M., Perfectti, F. & Lorite, J. (2015). The role of pollinators in floral diversification in a 687 \nclade of generalist flowers. Evolution, 69, 863–878. 688 \nGoodwin, E.K., Rader, R., Encinas -Viso, F. & Saunders, M.E. (2021). Weather Conditions 689 \nAffect the Visitation Frequency, Richness and Detectability of Insect Flower Visitors in 690 \nthe Australian Alpine Zone. Environ Entomol, 50, 348–358. 691 \nGünther, J., Irmisch, S., Lackus, N., Reichelt, M., Gershenzon, J. & Köllner, T. (2018). The 692 \nnitrilase PtNIT1 catabolizes herbivore -induced nitriles in Populus trichocarpa. BMC 693 \nPlant Biol., 18. 694 \nHarrell Jr, F.E. (2003). Hmisc: Harrell Miscellaneous. 695 \nKawecki, T.J., Lenski, R.E., Ebert, D., Hollis, B., Olivieri, I. & Whitlock, M.C. (2012). 696 \nExperimental evolution. Trends Ecol Evol, 27, 547–560. 697 \nKlein, A.-M., Vaissière, B.E., Cane, J.H., Steffan-Dewenter, I., Cunningham, S.A., Kremen, C., 698 \net al. (2007). Importance of pollinators in changing landscapes for world crops. Proc. 699 \nR. Soc. B., 274, 303–313. 700 \nKlinkhamer, P.G.L. & De Jong, T.J. (1993). Attractiveness to Pollinators: A Plant’s Dilemma. 701 \nOikos, 66, 180. 702 \nKlumpers, S.G.T., Stang, M. & Klinkhamer, P.G.L. (2019). Foraging efficiency and size 703 \nmatching in a plant–pollinator community: the importance of sugar content and tongue 704 \nlength. Ecol Lett, 22, 469–479. 705 \nKluser, S. & Peduzzi, P. (2007). Global Pollinator Decline: A Literature Review. Environment 706 \nAlert Bulletin, 8. 707 \nKnauer, A.C. & Schiestl, F.P. (2017). The effect of pollinators and herbivores on selection for 708 \nfloral signals: a case study in Brassica rapa. Evol Ecol, 31, 285–304. 709 \nKofler, X.V., Grossniklaus, U., Schiestl, F.P. & Frachon, L. (2024). Uncovering genes involved 710 \nin pollinator-driven mating system shifts and selfing syndrome evolution in Brassica 711 \nrapa. New Phytol., 243, 1220–1230. 712 \nLande, R. & Arnold, S.J. (1983). The Measurement of Selection on Correlated Characters. 713 \nEvolution, 37, 1210–1226. 714 \nLiu, Y.B., Zeng, Z.J., Barron, A.B., Ma, Y., He, Y.Z., Liu, J.F., et al. (2022). The involvement of 715 \na floral scent in plant-honeybee interaction. Sci Nat, 109, 30. 716 \nLloyd, D.G. (1992). Self- and Cross-Fertilization in Plants. II. The Selection of Self- Fertilization. 717 \nInt. J. Plant Sci., 153, 370–380. 718 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n27 \n \n \nLoria, A., Cristescu, M.E. & Gonzalez, A. (2019). Mixed evidence for adaptation to 719 \nenvironmental pollution. Evol Appl, 12, 1259–1273. 720 \nMajetic, C.J., Raguso, R.A. & Ashman, T.-L. (2009). The sweet smell of success: floral scent 721 \naffects pollinator attraction and seed fitness in Hesperis matronalis. Functional 722 \nEcology, 23, 480–487. 723 \nMurase, K., Moriwaki, Y., Mori, T. et al. (2020). Mechanism of self/nonself -discrimination in 724 \nBrassica self-incompatibility. Nat Commun., 11, 4916 725 \nNasrallah, J.B. (2017). Plant mating systems: self-incompatibility and evolutionary transitions 726 \nto self-fertility in the mustard family. Curr Opin Genet Dev, Volume 47, 54-60. 727 \nOllerton, J., Killick, A., Lamborn, E., Watts, S. & Whiston, M. (2007). Multiple Meanings and 728 \nModes: On the Many Ways to Be a Generalist Flower. Taxon, 56, 717–728. 729 \nOllerton, J., Winfree, R. & Tarrant, S. (2011). How many flowering plants are pollinated by 730 \nanimals? Oikos, 120, 321–326. 731 \nOpedal, Ø.H. (2018). Herkogamy, a Principal Functional Trait of Plant Reproductive Biology. 732 \nInt. J. Plant Sci., 179, 677–687. 733 \nPalacio, R.D., Valderrama-Ardila, C. & Kattan, G.H. (2016). Generalist Species Have a Central 734 \nRole In a Highly Diverse Plant–Frugivore Network. Biotropica, 48, 349–355. 735 \nPerkins, J., Hayashi, T., Peakall, R., Flematti, G.R. & Bohman, B. (2023). The volatile 736 \nchemistry of orchid pollination. Nat. Prod. Rep., 40, 819–839. 737 \nPotts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger, O. & Kunin, W.E. (2010). 738 \nGlobal pollinator declines: trends, impacts and drivers. Trends Ecol. Evol. , 25, 345 –739 \n353. 740 \nPowney, G.D., Carvell, C., Edwards, M., Morris, R.K.A., Roy, H.E., Woodcock, B.A., et al. 741 \n(2019). Widespread losses of pollinating insects in Britain. Nat Commun, 10, 1018. 742 \nPreston, J.C. & Fjellheim, S. (2022). Flowering time runs hot and cold. Plant Physiol., 190, 5–743 \n18. 744 \nQian, J., Liao, Y., Jian, G., Jia, Y., Zeng, L., Gu, D., et al. (2023). Light induces an increasing 745 \nrelease of benzyl nitrile against diurnal herbivore Ectropis grisescens Warren attack in 746 \ntea (Camellia sinensis) plants. Plant Cell Environ, 46, 3464–3480. 747 \nR Core Team . (2024). Available at: https://www.r -project.org/. Last accessed 10 November 748 \n2024. 749 \nRamos, S.E. & Schiestl, F.P. (2019). Rapid plant evolution driven by the interaction of 750 \npollination and herbivory. Science, 364, 193–196. 751 \nRamos-Jiliberto, R., P. Moisset de Espanés, and D. P. Vázquez. (2020). Pollinator declines 752 \nand the stability of plant –pollinator networks. Ecosphere 11(4):e03069. 753 \n10.1002/ecs2.3069 754 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n28 \n \n \nRea, A.C. & Nasrallah, J.B. (2008). Self-incompatibility systems: barriers to self-fertilization in 755 \nflowering plants. Int J Dev Biol, 52, 627–636. 756 \nSahli, H.F. & Conner, J.K. (2011). Testing for conflicting and nonadditive selection: floral 757 \nadaptation to multiple pollinators through male and female fitness. Evolution, 65, 1457–758 \n1473. 759 \nSchiestl, F.P. (2014). Correlation analyses between volatiles and glucosinolates show no 760 \nevidence for chemical defense signaling in Brassica rapa. Front. Ecol. Evol., 2. 761 \nSchiestl, F.P., Balmer, A. & Gervasi, D.D. (2018). Real -time evolution supports a unique 762 \ntrajectory for generalized pollination. Evolution, 72, 2653–2668. 763 \nSchneider, C.A., Rasband, W.S. & Eliceiri, K.W. (2012). NIH Image to ImageJ: 25 years of 764 \nimage analysis. Nat Methods, 9, 671–675. 765 \nSingh, A. & Adhikary, T. (2021). Importance of Pollinators in Fruit Production: A Review, 6. 766 \nSpigler, R., Smith-Ramesh, L. & Kalisz, S. (2020). Too attractive to self: How pollinators can 767 \ninterfere with the evolution of selfing. bioRxiv. 768 \nThomann, M., Imbert, E., Devaux, C. & Cheptou, P.-O. (2013). Flowering plants under global 769 \npollinator decline. Trends Plant Sci, 18, 353–359. 770 \nTraine, J., Rusman, Q. & Schiestl, F.P. (2024). Too hot to handle: temperature -induced 771 \nplasticity influences pollinator behaviour and plant fitness. New Phytol., 243, 1571 –772 \n1585. 773 \nTsuchimatsu, T., Fujii, S. (2022). The selfing syndrome and beyond: diverse evolutionary 774 \nconsequences of mating system transitions in plants. Philos Trans R Soc Lond B Biol 775 \nSci, 377 (1855) 776 \nTusuubira, S.K., Kelly, J.K. (2024). Experimental evolution suggests rapid assembly of the 777 \n‘selfing syndrome’ from standing variation in Mimulus guttatus. Front. Plant Sci . 778 \n15:1378568. 779 \nTwidle, A.M., Barker, D., Pilkington, L.I., Fedrizzi, B. & Suckling, D.M. (2022). Identification of 780 \nherbivore-induced plant volatiles from selected Rubus species fed upon by raspberry 781 \nbud moth (Heterocrossa rubophaga) larvae. Phytochemistry, 202, 113325. 782 \nUrbancsok, J., Bones, A.M. & Kissen, R. (2018). Benzyl Cyanide Leads to Auxin-Like Effects 783 \nThrough the Action of Nitrilases in Arabidopsis thaliana. Front Plant Sci, 9, 1240. 784 \nVenables, B. & Ripley, B. (2002). Modern Applied Statistics With S. In: Springer. 785 \nVicens, N. & Bosch, J. (2000). Weather -Dependent Pollinator Activity in an Apple Orchard, 786 \nwith Special Reference to Osmia cornuta  and Apis mellifera  (Hymenoptera: 787 \nMegachilidae and Apidae). Environ Entomol, 29, 413–420. 788 \nWang, L. & Filatov, D.A. (2023). Mechanisms of prezygotic post -pollination reproductive 789 \nbarriers in plants. Front Plant Sci, 14, 1230278. 790 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint \n\n29 \n \n \nWeider, L.J., Jeyasingh, P.D. & Frisch, D. (2017). Evolutionary aspects of resurrection ecology: 791 \nProgress, scope, and applications—An overview. Evol Appl, 11, 3–10. 792 \nWen, Y., Ye, Q., Román -Palacios, C., Liu, H. & Wu, G. (2023). Physiological cold tolerance 793 \nevolves faster than climatic niches in plants. Front. Plant Sci., 14. 794 \nWendell, D.L. & Pickard, D. (2007). Teaching human genetics with mustard: rapid cycling 795 \nBrassica rapa (fast plants type) as a model for human genetics in the classroom 796 \nlaboratory. CBE Life Sci Educ, 6, 179–185. 797 \nWright, G.A. & Schiestl, F.P. (2009). The evolution of floral scent: the influence of olfactory 798 \nlearning by insect pollinators on the honest signalling of floral rewards. Funct. Ecol., 799 \n23, 841–851. 800 \nXu, K., (2023). Evolution of flowering time due to variation in the onset of pollen dispersal 801 \namong individuals, Evolution, 78, 401–412. 802 \nZu, P., Blanckenhorn, W.U. & Schiestl, F.P. (2016). Heritability of floral volatiles and pleiotropic 803 \nresponses to artificial selection in Brassica rapa. New Phytol., 209, 1208–1219. 804 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.23.701299doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}