Ecology, 97(6), 2016, pp. 1452–1462 © 2016 by the Ecological Society of America

Demographic shifts related to mycoheterotrophy and their fitness impacts in two species

Richard P. Shefferson,1,6 Mélanie Roy,2 Ülle Püttsepp,3 and Marc-André Selosse4,5 1Department of General Systems Sciences, University of Tokyo, 3-8-1 Komaba, Tokyo 153-8902, Japan 2Laboratoire Evolution et Diversité Biologique, Université Paul Sabatier – CNRS, UMR5174, 118 route de Narbonne, 31062, Toulouse Cedex, 3Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Street Kreutzwaldi 5, 51014, Tartu, Estonia 4Institut de Systématique, Évolution, Biodiversité (ISYEB - UMR 7205 – CNRS, MNHN, UPMC, EPHE), Muséum national d’Histoire naturelle, Sorbonne Universités, 57 rue Cuvier, CP50, 75005, Paris, France 5Department of and Nature Conservation, University of Gdansk, Wita Stwosza 59, 80-308, Gdansk, Poland

Abstract. Evolutionary losses of photosynthesis in terrestrial all originate in photosynthetic ancestors. The adaptive context under which this transition happens has remained elusive because of the rarity of plants in which both photosynthetic and non- photosynthetic forms exist as a polymorphism. Here, we report on demographic patterns in photosynthetic (“green”) and nonphotosynthetic (“albino”) individuals within populations of two such species, and C. longifolia, which also acquire nutrition­ from their mycorrhizal hosts (partial mycoheterotrophy). We hypothesized that demographic shifts in albinos relative to greens would include compensatory patterns with respect to fitness, such that maladaptive changes to survival or reproduction would be adaptively countered by changes to other parameters, such as growth probabilities. We tracked ­individuals in two populations of C. damasonium for 3 yr, and in one population of C. longifolia for 14 yr. We then analyzed vital rates for both phenotypes using general linear mixed models (GLMMs) and multi-­state capture mark-­recapture models (CMR), and used these models to develop size-classified,­ function-­based population projection matrices. We estimated fitness as the deterministic population growth rate λ( ) for each phenotype, and explored the impact of shifts in demographic patterns to albinism via life table response experiments (LTREs). Mortality differed between greens and albinos, but not similarly across species. Albinos generally sprouted less than greens, and flowered more when small but less at other times. Albinos typically had a higher probability of fruiting, although their lower flower numbers yielded lower numbers of fruits overall. Fitness did not differ significantly among phenotypes. Thus, we did not find significant evidence that albinism is adaptive or maladaptive; however, if in fact it is the latter, then we did find evidence of incomplete compensation for declines in survival and reproduction from growth transitions, particularly to small flowering size classes inC. damasonium, and to large vegetative size classes in C. longifolia. These patterns indicate some support for the idea that albinism may lead to the speciation of mycoheterotrophic plants. Key words: Cephalanthera; loss of photosynthesis; mixotrophy; mycoheterotrophy; .

Introduction carbon from nearby photosynthetic plants (Smith and Read 2008), or more rarely from soil organic matter Plants are most well-known­ for their photosynthetic (Martos et al. 2009). In temperate forests, mycohetero- capability. However, many species have lost the ability trophic species are particularly numerous in the to photosynthesize, or have evolved a greatly weakened Orchidaceae and Ericaceae (Leake 1994, Bidartondo ability to do so (Krause 2008). Such species include >400 2005, Merckx 2013). species of terrestrial plants that have developed other The loss of photosynthesis presents an evolutionary means to procure the carbohydrates that they need, such puzzle to plant biologists. Photosynthesis makes energy as the mycoheterotrophic plants (Leake 1994, Merckx relatively cheap, and so at least in principle it may prevent 2013). The latter use carbon from their mycorrhizal carbon from becoming limiting under many circum- partners, the soil fungi establishing symbioses with their stances (Roy et al. 2013). Thus, although many muta- roots (Smith and Read 2008), which in turn obtain tions can abolish it because it requires many genes for complex protein machineries (Gao et al. 2010), its loss Manuscript received 18 July 2015; revised 18 January 2016; accepted 2 February 2016. Corresponding Editor: L. S. Adler. may generally be selected against. However, the many 6 E-mail: [email protected] evolutionary losses of the trait suggest that some adaptive

1452 June 2016 MYCOHETEROTROPHY AND FITNESS 1453 context(s) can favor this loss. Since nonphotosynthetic, nonphotosynthetic genotypes under at least some condi- mycoheterotrophic plants often live in shady forest floors tions. However, the extreme rarity of species in which (Gebauer and Meyer 2003, Bidartondo et al. 2004, both photosynthetic and nonphotosynthetic forms exist, Selosse et al. 2004), and typically acquire carbon from particularly as a polymorphism within the same popu- mycorrhizal fungi (Selosse et al. 2002), the adaptive lation (Tranchida-Lombardo­ et al. 2010), has generally context of the loss of photosynthesis in mycohetero- prevented the examination of demographic differences trophs, and the maintenance of this trait, may be tied to between these phenotypes. Such comparisons between difficulty in acquiring carbon by photosynthesis. Thus, photosynthetic (hereafter “green”) and nonphotosyn- forest herbs may be more likely to develop heterotrophic thetic phenotypes (hereafter “albino”) within the same lifestyles than plants inhabiting other ecosystems, both species are ultimately required to infer fitness differences. due to the ecological context of interspecific plant com- Among orchids, some species phylogenetically close to petition for light in forests, and to the availability of other mycoheterotrophic species exhibit such a polymorphism. energy sources via ectomycorrhizal fungi. Indeed, the Interestingly, these green photosynthetic species are par- latter are efficiently provided through surrounding trees, tially mycoheterotrophic (they are also called ‘mix- which provide up to 20% of their photosynthates to ecto- otrophic’; Julou et al. 2005, Selosse and Roy 2009, mycorrhizal fungi (Smith and Read 2008). Selosse et al. 2016), a condition that allows pure albinos The evolutionary shift from photosynthesis to pure to survive (e.g., Selosse et al. 2004, Julou et al. 2005, mycoheterotrophy may also make growth and repro- Abadie et al. 2006, Roy et al. 2013, Gonneau et al. 2014). duction costlier in mycoheterotrophs. Growth and repro- Further, reproduction by seed appears strongly limited duction are theoretically costly to survival, and to current by energy limitations which might have to do with the and future growth and reproduction. Such costs rarely loss of photosynthesis (Bellino et al. 2014), while vege- exhibit themselves in the short-­term in long-­lived plant tative survival, and potentially clonal growth, do not species (Primack and Stacy 1998, Shefferson et al. 2006), appear to be impaired due to carbon acquisition from although high-resolution­ matrix methods like integral mycorrhizal fungi (Roy et al. 2013, Gonneau et al. 2014). projection models combined with long-term­ monitoring Here, we use two populations of C. damasonium and and high sample sizes can identify them (Jacquemyn et al. one of C. longifolia to ask: Is the evolutionary loss of 2010, Miller et al. 2012, Sletvold and Ågren 2015a,b). photosynthesis evolutionarily adaptive in mixotrophic Generally, costs are only observed when populations are species? We report data on green and albino individuals experimentally stressed or resource variability is of Cephalanthera damasonium and of C. longifolia to accounted for (Reznick 1992, Primack and Stacy 1998), address this question, and we explore demographic sim- or when large sample sizes allow the demographic effects ilarities and differences of these two phenotypes, inte- of individual histories to be dissected (Shefferson et al. grating demographic differences across the life cycle into 2014). However, the few studies of demography in myco- a comparison of fitness based on the deterministic pop- heterotrophs suggest that growth and reproduction may ulation growth rate. We ask whether the albinos and more commonly be costly. For example, the loss of pho- greens co-occurring­ in the same population have equiv- tosynthesis may favor sprouting only for reproduction, alent fitness. We further ask whether any albino traits since leaves serve no more nutritional or reproductive exhibit trends suggesting adaptive compensation for any function (Shefferson et al. 2011). Furthermore, with seemingly maladaptive differences, and test the latter carbon supply obtained heterotrophically, carbon may with life table response analysis (LTRE). be limiting much more than would be possible under a photosynthetic lifestyle (Bruns et al. 2002, Roy et al. Methods 2013), and this may even limit vegetative growth at least some of the time in such plants (Merckx 2013). If a purely Study species and field sites mycoheterotrophic lifestyle is costly, then demographic patterns should shift to deal with these trade-­offs. For We studied the long-lived­ herbaceous perennial species, example, if the production of flowers and fruits drops Cephalanthera damasonium (Mill.) Druce and due to lower carbon supply, then sprouting patterns or (L.) Fritsch (, reproductive mode might evolve to compensate, and Orchidaceae). These species are found throughout indeed mycoheterotrophic plants often have unusual and extend into (Pridgeon et al. 2006). Their underground multiplication modes (Klimešová 2007, life histories have not been well-characterized,­ but are Shefferson et al. 2011). The transition to mycohetero- probably similar (Rasmussen 1995). In both cases, the trophy may put a premium on storing energy over the plant begins life as a dust seed. Upon germination, it long-­term as much as possible to deal with potentially enters the protocorm phase, a life stage that involves the unpredictable energy supply, making slow growth rates progressive underground development of a heterotrophic and minimal sprouting patterns adaptive. and root system without any above-ground­ Regardless of mechanisms, an adaptive context to the green tissue (Rasmussen 1995, Dearnaley et al. 2012). In loss of photosynthesis and resulting speciation to myco- these species, it is not clear whether the protocorm stage heterotrophy requires that natural selection favor typically persists for multiple years, or whether seedlings 1454 RICHARD P. SHEFFERSON ET AL. Ecology, Vol. 97, No. 6 develop within a year of germination, but given the ten- fitness via sensitivity and elasticity analysis. Finally, we dency of related wild orchid species to spend years below compared the fitness of greens and albinos and assessed ground, the former is likely (Rasmussen 1995). Adults the impacts of altered demographic rates on albino are mixotrophic (see above); they may sprout or remain fitness, using life table response experiments (LTREs). vegetatively dormant, and sprouting individuals carry Linear and capture mark-recapture­ modeling of out photosynthesis when green (Julou et al. 2005, Abadie ­demographic parameters.—We characterized demo- et al. 2006) and may produce flowers or not. When they graphic rates in green and albino individuals at these flower, C. damasonium is self-­fertilized, while C. longi- three populations using generalized linear mixed models folia outcrosses, and in neither case do albinos differ from (GLMMs) and capture mark-recapture­ (CMR) models greens in the method of pollination (Tranchida-­ of vital rates in all post-protocorm­ life stages. Vital rates Lombardo et al. 2010, Roy et al. 2013). modeled included: probability of survival from year t to We monitored a total of 808 Cephalanthera dama- year t + 1, probability of sprouting in year t + 1 contin- sonium individuals, of which 693 were green and 115 were gent on survival, probability of becoming each possible albino, in a population at Montferrier, France, from 2006 size in year t + 1 contingent on survival and sprouting, to 2008 (see details in Roy et al. 2013). The site is a probability of flowering in yeart + 1 contingent of surviv- Populus nigra plantation with Quercus pubescens also al and sprouting, number of flowers in yeart contingent common, and was once a vineyard. We also monitored on sprouting and flowering in that year, probability of a total of 30 C. damasonium individuals, including 15 that fruiting in year t conditional on flowering and sprouting were green and 15 that were albino, in a population at in that year, and number of fruits in year t conditional Boigneville, France, over the same time period (Roy et al. on sprouting, flowering, and fruiting in that year. Here, 2013). The site is a calcareous, young but canopy-­closed we used CMR models to estimate survival and sprouting Quercus robur L. and Corylus avellana L. forest, with probabilities, and GLMMs to estimate all other terms. various herbaceous plants and shrubs such as Crataegus We did this because estimating survival and sprouting laevigata, Cornus sanguinea, and Hedera helix (Julou in vegetative dormancy-prone­ plant species using non-­ et al. 2005). Finally, we monitored a population of CMR methods requires making assumptions about the Cephalanthera longifolia at Pussa, Saaremaa island, length of dormancy after the final sighting, and these Estonia, from 1992 to 2005 (see details in Abadie et al. assumptions typically bias estimates of these probabilities, 2006), and this included a total of 36 individuals, of which particularly when the study period is short (Shefferson 17 were green and 19 were albino. This sample includes 2002). No such assumptions are required to deal with the all albino individuals at the site, and a roughly equal probabilities of growth, flowering, and fruiting, however. sampling of the larger green portion of the population. We estimated survival and sprouting probability in The population grows in a calcareous coastal plain in a greens and albino C. damasonium at Montferrier and site undergoing an ecological transition from grassland Boigneville, and C. longifolia at Pussa, using multi-state­ to forest stage. The vegetation is a shrubland of scattered capture mark-recapture­ (CMR) models in program Juniperus communis with some young Pinus sylvestris MARK (White and Burnham 1999). CMR modeling (Abadie et al. 2006). Tranchida-Lombardo­ et al. (2010) began with the construction of a global model, in which investigated these populations (under the names Mon, survival probability from year t to year t + 1 in sprouting Boi, and Est, respectively), and revealed that (1) no individuals varied with size in year t and phenotype spatial aggregation of albinos or green individuals (green or albino), survival in dormant individuals varied occurred, and (2) these two phenotypes were not genet- with phenotype, the probability of transition from ically aggregated (i.e., diverse genotypes occurred in both sprouting to dormant conditional upon survival varied phenotypes). Thus, the present study is carried out on with size and phenotype, and the probability of transition individuals that are not biased by spatial distribution and from dormant to sprouting conditional upon survival genotype. We also note that the two French populations varied with phenotype. C. damasonium populations were have since crashed (M.A. Selosse and M. Roy, personal analyzed together because of the data hungry nature of observation). multi-­state CMR models, and so population was also a term in each estimated parameter in the global model for that species, with the global model incorporating an Analytical methods interaction between population and all other factors. In Our analytical strategy was focused on comparisons C. damasonium, the number of sprouts per plant was of greens and albinos at each site, with the goal of com- highly variable, and so we used this metric as our estimate paring the fitness of the two groups and assessing key of plant size; however, this metric was not particularly demographic changes involved in the transition from variable among individuals in C. longifolia, leading us to green to albino plant. First, we assessed life history costs use the total number of leaves as a proxy for plant size and size-based­ trends in demographic parameters using in the Pussa population. These metrics were highly, pos- linear modeling. Next, we used these models to build itively, and significantly correlated with all other size Lefkovitch matrices for each group. Next, we explored metrics within our datasets (C. damasonium: sprout the potential for evolution in demographic rates to alter number, tallest sprout height; C. longifolia: sprout June 2016 MYCOHETEROTROPHY AND FITNESS 1455

number, leaf number, tallest sprout height; all r > 0.80 Pi is the probability of fruiting in year t conditional upon and P < 0.05). Resighting was fixed at 1.000 for sprouting flowering, pi is the mean number of flowers produced individuals and 0.000 for dormant individuals, per the conditional upon fruiting, u is the mean number of seeds two-­state (observed vs. unobserved) procedure outlined per fruit, z is the probability of seed dormancy, and y is in Schaub et al. (2004). Plant size was included as an the probability of germination into the protocorm stage. individual covariate, and standardized to range from 0 Lefkovitch matrices for both species included three to 1. All possible reduced models were created for each juvenile stages: one dormant seed stage, one protocorm species, and the model with the lowest AICc was treated stage, and one seedling stage. Additionally, there were as the best-­fit model and used for inference (Appendix one adult dormant stage, 17 and 20 nonflowering adult S1: Table S1). sprouting stages in C. damasonium and C. longifolia, The probability of growth conditional upon survival, respectively, and an equivalent number of flowering adult the probability of flowering contingent on survival and sprouting stages in each case. In both species, seeds are sprouting, the number of flowers contingent on flow- likely capable of long-­term dormancy (see Tešitelová ering, the probability of fruiting contingent on flowering, et al. 2012, for related spp.), and protocorms and the number of fruits contingent on fruiting were all are thought to be capable of existing for multiple years analyzed via GLMMs. Fixed factors tested include phe- without becoming photosynthetic seedlings, but little notype (green or albino), plant size in year t, and flow- data exist on the actual transition probabilities (Shefferson ering status in year t (binomial), and all interactions. To et al. 2012, Roy et al. 2013). Per previously published characterize demographic patterns in C. damasonium field germination studies on these populations, we with as much power as possible, we also included popu- assumed that green and albino individuals produced lation as a fixed factor for those two populations, as in 7339 ± 2371 and 2554 ± 1557 seeds per fruit (mean ± 1 CMR modeling. Year was included as a random factor. SE), respectively (Roy et al. 2013), and that green and Modeling began with the construction of a global model albino seed germinated to the protocorm stage at prob- with all terms incorporated, using package lme4 in R abilities of 0.79 ± 0.02% and 0.55 ± 0.01%, respectively 3.2.2 (Bates et al. 2012, R Core Team 2015). We then (Roy et al. 2013). In addition, we assumed that seed dor- produced all possible reduced models and selected the mancy probability was 0.25, probability of stasis as a model with the lowest AICc as the best-­fit model (Table protocorm was 0.20, and the probability of transitioning S2), using function dredge in package MuMIn (Barton to seedling from protocorm was 0.25, regardless of phe- 2014). notype and population, per a previously published study Lefkovitch matrix construction.—We used the best-fit­ on a different Estonian population of C. longifolia CMR models and GLMMs to create functions for each (Shefferson et al. 2012). We tested the impacts of these vital rate, which were then used to populate two Lefko- assumptions by redoing the sensitivity, elasticity, and vitch matrices for population projection analysis of each LTRE analyses (all described below) using altered but phenotype in each population. Survival-growth­ transi- realistic versions of fruit production rate, seed germi- tion probabilities for adult stages were estimated as: nation, and seed dormancy (±1000 seeds/fruit, ±10% a = S ×s ×G ×F seed germination, ±10% seed dormancy), but found no ij j j ij ij (1) qualitative difference with the results presented here. We where aij is the transition probability from stage j in year further assumed that seedlings could develop into adults t to stage i in year t + 1 (corresponds to the Lefkovitch with the same probabilities of survival and transition as matrix element), Sj is the survival probability of an indi- vegetatively dormant adults (Shefferson et al. 2012). vidual in stage j in year t to year t + 1, sj is the probability Adult size classes were based on the number of sprouts of sprouting in year t + 1 contingent on stage j in year t per plant in C. damasonium, and the total number of and conditional upon survival, Gij is the probability of leaves per plant in C. longifolia. growing from stage j in year t to stage i in year t + 1 Initial matrix analysis and fitness comparison.—We conditional upon survival and sprouting, and Fij is the developed two deterministic mean matrices per each probability of flowering from stagei in year t + 1 condi- population using all of the CMR models and GLMMs tional upon stage j in year t and survival and sprouting for each combination of population and phenotype, to that time (the complement of this term was used for one for green individuals and one for albino individu- vegetatively sprouting individuals). Fecundity was esti- als (because­ some CMR models and GLMMs included mated as: trophic status as a significant term; Appendix S1: Tables a = f ×P ×p ×u×z S1 and S2). We estimated the fitness of each phenotype di i i i (2) as the eigenvalue of each matrix, corresponding to the for dormant seeds (d), and as asymptotic, deterministic population growth rate as a = f ×P ×p ×u×y though each phenotype occurred independently (Metcalf ci i i i (3) and Pavard 2007). We estimated the standard errors for direct germination into the protocorm stage (c), of these λ estimates via 1000 nonparametric, hierar- where fi is the mean number of flowers produced by an chical bootstraps with resampling from each dataset individual in stage i in year t conditional upon flowering, (i.e., we bootstrapped data from each individual, rather 1456 RICHARD P. SHEFFERSON ET AL. Ecology, Vol. 97, No. 6 than from each individual-­transition). To test whether of demographic shifts between the two phenotypes in the albino phenotype exhibited significantly different each population on fitness, measured as deterministic fitness from the green phenotype, we conducted 1000 population growth rate, via associated shifts in demo- replicate hierarchical, nonparametric permutation tests graphic transitions (Caswell 1989, 2001). Green matri- of the difference between λ for greens and albinos in ces served as baseline controls for analyses, since they each population (Caswell 2001). Here, the significance represent the ancestral condition. We summed LTRE of each test was given as 1 plus the rank of each estimat- contributions to the change in fitness between greens and ed λ against the replicate permutations, divided by 1001 albinos associated with each stage, and with each of the (Caswell 2001). four demographic transitions described above. We used The impact of a shift in demographic parameters on these LTRE contributions to test our hypothesis that, to fitness can vary from parameter to parameter. To assess maintain fitness relative to greens, the LTRE contribu- the evolutionary importance of demographic parameters, tions of albinos must completely compensate for any neg- we estimated the sensitivity and elasticity of λ to each ative impacts of the shift to a nonphotosynthetic lifestyle. element in each matrix. Here, the sensitivity of fitness to All analyses were performed in R 3.2.2 (R Core Team a particular matrix element describes the additive impact 2015), with sensitivity and elasticity analyses performed of an infinitesimally small change in that element on the using package popbio (Stubben and Milligan 2007). We deterministic growth rate, while the elasticity describes the did not perform stochastic versions of these analyses proportional impact of the same (Caswell 2001). We com- because the Montferrier and Boigneville datasets only pared the importance of each stage by summing all sensi- included 3 years, and because the Boigneville dataset and tivities or elasticities associated with elements from that the much longer Pussa dataset only included 30 and 36 stage in each matrix. We also compared the importance individuals, respectively (although these sample sizes of each kind of four major demographic transitions: reflect equally low population sizes). (1) fecundity, referring to the production of seeds that either stay dormant or germinate into protocorms in the Results following year; (2) progression, referring to growth to a larger size; (3) stasis, in which individuals remain in the Above-­ground population size varied from 56 in same size class across two consecutive years; and (4) ret- 2006 to 373 in 2007 in Montferrier (C. damasonium), rogression, in referring to shrinkage to a smaller size. We from 3 in 2008 to 18 in 2007 in Boigneville (C. dama- used elasticities because of their importance in the calcu- sonium), and from 20 in 1992 to 100 in 1997 in Pussa lation of LTRE contributions (see below), but also com- (C. longifolia). Key traits of green and albino pheno- pared sensitivities due to their closer theoretical relationship types differed similarly across species. Green indi- to selection gradients (Caswell 2001). viduals were on average significantly larger than albino Life table response experiment (LTRE).—Finally, individuals in terms of the mean number of sprouts we assessed the actual impact of altered demographic (mixed model with phenotype as a fixed factor and patterns in albinos by performing a life table response population as random: P = 0.046; Table 1). Green indi- ­experiment (LTRE) comparing albino matrices against viduals did not have significantly more leaves than the green matrices within each population. Briefly, albinos (Welch t-­test: Montferrier: t = −0.187, LTREs allow shifts in the deterministic population df = 877.6, P = 0.426; Boigneville: t = 0.671, df = 36.7, growth rate to be decomposed both to the additive P = 0.253; Pussa: t = −1.15, df = 82.0, P = 0.253; impacts of specific factors, such as experimental treat- Table 1). Green individuals produced significantly ments, and to the associated changes in population pro- more flowers (mixed model as before:P < 0.0001; jection matrix elements (Caswell 2001). They involve Table 1), but did not differ significantly from albinos a comparison between two matrices, one of which is in terms of number of fruits per plant (mixed model as considered to be the experimental matrix and the other before: P = 0.153; Table 1). serves as a control or reference, and thus produce a Demographic patterns differed between greens and retrospective analysis of altered population dynamics. albinos, but not similarly across species (Appendix S1: This analysis allowed us to measure the actual impact Table S1). Albino individuals exhibited significantly

Table 1. Key traits of green vs. albino Cephalanthera damasonium and Cephalanthera longifolia. Numbers presented are mean ± 1 SE.

Species Population Phenotype Sprouts/plant Leaves/plant Flowers/plant Fruits/plant

C. damasonium Montferrier Green 1.45 ± 0.01 (n = 319) 3.2 ± 1.0 (n = 840) 9.02 ± 0.07 (n = 182) 5.75 ± 0.01 (n = 569) Albino 1.27 ± 0.01 (n = 94) 3.4 ± 0.4 (n = 76) 7.04 ± 0.24 (n = 23) 3.19 ± 0.04 (n = 95) Boigneville Green 1 ± 0 (n = 7) 3.2 ± 0.4 (n = 25) 5.00 ± 2.12 (n = 2) 6.25 ± 0.62 (n = 12) Albino 1 ± 0 (n = 11) 2.6 ± 0.8 (n = 26) 3.00 ± 0 (n = 1) 2.5 ± 0.26 (n = 8) C. longifolia Pussa Green 0.975 ± 0.01 (n = 40) 5.45 ± 0.11 (n = 40) 5.09 ± 0.40 (n = 23) 0.26 ± 0.02 (n = 27) Albino 0.678 ± 0.01 (n = 59) 4.42 ± 0.07 (n = 59) 1.25 ± 0.10 (n = 24) 0.20 ± 0.05 (n = 15) June 2016 MYCOHETEROTROPHY AND FITNESS 1457 higher mortality than green individuals in both species were more likely to flower than green individuals, par- except when dormant, when albinos exhibited slightly ticularly at smaller sizes (Fig. 1e). In C. longifolia, smaller higher survival than greens (Fig. 1a,b). Sprouting was albinos flowered more frequently than smaller green con- similar between greens and albinos in most sizes, except specifics and larger individuals showed the reverse pattern for small, sprouting C. damasonium, which were more (Fig. 1f; Appendix S1:Table S2). These trends suggest that likely to sprout if albino (Fig. 1c, d). Interestingly, pre- younger or more resource-stressed­ albinos have a higher viously flowering individuals of both phenotypes in chance of producing flowers than similarly young or C. damasonium were more likely to flower than previously resource-­stressed green individuals. The probability of non-­flowering individuals when small, but less likely fruiting was typically higher in albino C. damasonium when large (Fig. 1e; Appendix S1: Table S1). (Fig. 1g, h), although at Montferrier, smaller plants were Reproductive patterns conflicted across phenotypes more likely to fruit if they were green (Fig. 1g). but suggested that albinos were subject to greater life When estimated as the deterministic growth rate of history costs or resource stresses. Albino C. damasonium each phenotype, fitness did not differ between greens

a b ) al viv 0.6 0.6 Green Albino P(sur 0.0 0.0 0510 15 0510 15 20 Size (# sprouts) Size (# leaves)

c d 0.6 0.6 P(sprouting) 0.0 0.0 0510 15 0510 15 20 Size (# sprouts) Size (# leaves)

e f

ing) Green non−flowering Green flowering 0.6 Albino non−flowering 0.6 Albino flowering P(fl ower 0.0 0.0 0510 15 0510 15 20 Size (# sprouts) Size (# leaves)

g h 0.6 0.6 uiting) P(fr 0.0 0.0 0510 15 20 25 30 0510 15 20 25 30 # flowers # flowers

Fig. 1. The probabilities of survival from year t to t + 1 (a, b), sprouting in year t + 1 (c, d), probability of flowering in yeart + 1 (e, f), and fruiting in year t (g, h), as a function of size (a-f)­ or number of flowers (g, h) in yeart in green and albino individuals in two populations of Cephalanthera damasonium (Montferrier and Boigneville, France) and one population of C. longifolia (Pussa, Estonia). Fruiting is given in separate panels for Montferrier (g) and Boigneville (h). For other parameters, C. damasonium patterns are shown together in panels a, c, and e, while C. longifolia patterns are shown in b, d, and f. Size is given as the number of sprouts in C. damasonium and the number of leaves in C. longifolia. 1458 RICHARD P. SHEFFERSON ET AL. Ecology, Vol. 97, No. 6

Fig. 2. Fitness by phenotype in two populations of Cephalanthera damasonium (Montferrier and Boigneville, France) and one population of Cephalanthera longifolia (Pussa, Estonia). Fitness was estimated as the deterministic population growth rate of individuals in each combination of population and phenotype. Error bars indicate standard error, as estimated via 1000 random bootstraps of the monitoring datasets. and albinos at each site (Montferrier: P = 0.471, impacts on fitness, and also to transitions involving Boigneville: P = 0.331, Pussa: P = 0.850; Fig. 2). large, flowering adults. All in all, this suggests that the Although a similar pattern exists across populations, survival of flowering individuals and recruitment play with fitness in greens seemingly higher than in albinos, important evolutionary roles in the population dynamics we did not have enough populations to confirm a sig- of albino plants. nificant relationship across the entire study (sign test Demographic shifts to albinism differed in their with three out of three successes: P = 0.25). However, impacts on fitness across species. InC. damasonium, fitness differed in its responsiveness to different stages green individuals in the largest size class stayed that size and demographic transitions. In all populations, fitness but switched from flowering to non-­flowering and from was most strongly sensitive and elastic in response to non-­flowering to flowering with a probability approaching growth transitions, suggesting that any evolutionary 1.0. In comparison, although albinos in that size class change affecting growth to larger sizes would dramati- exhibited the same pattern, they did so with a reduced cally influence fitness (Appendix S1: Figure S1). Fitness survival leading to a probability of transition of approx- was next most responsive to stasis, with fecundity and imately 0.76. This decreased tendency to survive in retrogression alternating between 3rd and 4th place albinos led to a sharply negative influence on fitness rel- among populations. In terms of life stages, fitness was ative to greens in both populations, as did their reduced most elastic to transitions involving small flowering fecundity (Fig. 3a,b). This trend was incompletely coun- adults and protocorms in albino C. damasonium, but tered by increased growth from vegetative dormancy to more to protocorms in green C. damasonium (Figure small flowering sizes, and by increased retrogression S2a–d). However, fitness was most elastic to transitions from the largest size classes to small, flowering size classes involving protocorms and dormant individuals in green (Fig. 3a,b). In C. longifolia, reduced fecundity in flow- and albino C. longifolia, respectively (Figure S2e,f). ering stages strongly contributed to a drop in fitness in Sensitivity analysis suggested that fitness is most albinos, which was incompletely countered by growth responsive to changes in seed germination, suggesting from vegetative dormancy and growth to larger vege- that variability in germination would have strong tative classes (Fig. 3c). June 2016 MYCOHETEROTROPHY AND FITNESS 1459

Fig. 3. Contributions of shifts in demographic parameters of albino plants relative to green plants in two populations of Cephalanthera damasonium (a, Montferrier and b, Boigneville, France) and one population of Cephalanthera longifolia (c, Pussa, Estonia). Bars above the zero line show positive contributions to fitness in albinos, while bars below the zero line show losses to fitness in albinos. Stages are arranged in order, with dormant seeds, protocorms, and seedlings in the juvenile category, followed by vegetative dormancy as the first stage of the adult vegetative group, followed by sprouting vegetative plants from the smallest to the largest category, followed by sprouting, flowering plants from the smallest to the largest.

Discussion The conditions leading to approximately equal fitness Albinos of Cephalanthera spp. exhibit many character- between green and albino phenotypes may be linked to istics that suggest that their condition is maladaptive, in the life history context of vegetative dormancy. This con- accordance with data accumulated in earlier studies. For dition is a life history stage of both juvenile and mature example, albino C. damasonium have smaller fruits with stages of many herbaceous perennials, in which the plant seeds that germinate less efficiently (Roy et al. 2013) and lives underground as a root system for at least a full year a tendency to remain smaller and more dormant as adults and perhaps longer, without any above-ground­ sprouts (Julou et al. 2005, Roy et al. 2013). Yet, when considering (Lesica and Steele 1994, Shefferson 2009). It occurs in flowering and fruiting, and the impact of increased many species, especially in orchids, and may be adaptive ­vegetative dormancy levels on fitness, the adaptive con- as a compensation mechanism for life history costs sequences were less obvious. While the majority of indi- involved in growth (Shefferson et al. 2014), or as a bet-­ cators would suggest that whatever mutations are hedge against temporal environmental variation (Gremer involved in the production of albinos are typically et al. 2012). In particular, the high cost of growth in some ­maladaptive, some selection may favor particular com- years puts a premium on avoiding sprouting in others, binations of traits that could maintain albinism over the provided that either a store of resource reserves or diver- evolutionary long-term.­ It is not clear whether such com- sified energy supply exists and is occasionally exhausted pensatory patterns occur in other species, but they are via growth at too high levels (Shefferson et al. 2014). This nonetheless suggestive of possible adaptive explanations hypothesis implies that vegetative dormancy will be for the loss of photosynthesis and speciation of myco- favored at some optimal rate. In mixotrophic species heterotrophs that are frequent in tribe Neotieae, which particularly, extreme growth and sprouting costs in the Cephalanthera spp. belongs to (Abadie et al. 2006, Selosse presence of cheap mycorrhizal carbon may create the and Roy 2009, Selosse et al. 2016). conditions for a completely mycoheterotrophic lifestyle 1460 RICHARD P. SHEFFERSON ET AL. Ecology, Vol. 97, No. 6 during dormancy, in which sprouting only occurs to may be balanced in more wet and shaded forests (Roy reproduce (Shefferson 2009, Shefferson et al. 2011). In et al. 2013). Further, increased levels of vegetative dor- support of this hypothesis, both investigated species are mancy may help to stabilize survival in the face of envi- capable of vegetative dormancy, and the probability of ronmental stochasticity, to which albinos may be more sprouting was lower in albinos of both species (Fig. 1d, subject given the increased resource stress that they likely e; Appendix S1: Table S1). Dormant C. damasonium indi- face relative to greens (Bruns et al. 2002, Roy et al. 2013). viduals behave physiologically as full mycoheterotrophs Alternatively, greens may be more subject to temporal (Gonneau et al. 2014). Further, vegetative dormancy was stochasticity if albinos prefer dark, wet forests to which associated with some of the strongest compensatory a more photosynthetic strategy yields lower fitness, at shifts to fitness inC. longifolia (Fig. 3c). However, we least in some years. Such a situation can only be resolved note that the low sample sizes in our study (dictated by with a long-­term study followed by a stochastic, retro- low numbers of albinos present at the site) and the low spective analysis, such as a stochastic LTRE. number of years of observation in C. damasonium (dic- Traits may also be maintained nonadaptively if poten- tated by population crashes and the loss of albinos), may tially competing phenotypes become extinct for other have led to low power yielding a general inability to find reasons, as might happen due to genetic drift in small statistical evidence in support of differing fitness between populations, or if individuals with the novel trait escape phenotypes. into new habitats without the presence of competing phe- The differing demography of phenotypes in these notypes. Indeed, albinos are certainly capable of repro- species suggests some possible speciation scenarios. duction and dispersal, and orchids often form small Evolutionary explanations for speciation generally populations prone to drift (Tremblay et al. 2005). Should require a hypothesized mechanism resulting in (1) the such a situation be followed by reproductive isolation, appearance of the novel trait, (2) a mechanism main- as might happen if the loss of photosynthesis results in taining that trait, and (3) a mechanism allowing indi- pollinator shifts or a different flowering schedule, then viduals with that trait to reproduce in isolation from condition (3) would also be met and speciation would at individuals without that trait (Nowak et al. 2010). least have the potential to occur. Condition (1) is likely the result of the accumulation of Although this series of hypothetical events may cer- random but fortuitous mutations that may each individ- tainly lead to speciation, the evidence suggests that such ually involve the loss of function and fitness,i.e., a muta- a scenario should be very rare. Field surveys show that tional drift in sequence that, altogether, may yield overall albinos are rare and never exclusively form a population positive or no impacts on fitness. Some genes implicated in mixotrophic species to the best of our knowledge (e.g., in the loss of photosynthesis have been identified in other Tranchida-­Lombardo et al. 2010, Roy et al. 2013, species (de Pamphilis et al. 1997, Wolfe and de Pamphilis Suetsugu and Kato 2014). Thus, instances of either drift 1998), and it is interesting to note that mycoheterotrophic or selection yielding stand-­alone albino populations have species, mixotrphic species, and albinos of green species never been documented. But on the other hand, it is dif- do not simply stop photosynthesis, but also shift mycor- ficult to imagine that the fully mycoheterotrophic species rhizal fungal partners from their fully green ancestors in the otherwise green genus Cephalanthera in Asia and (Abadie et al. 2006, Yamato et al. 2014). Further, some America did not issue from albino variants of a green green orchids even do not produce chlorophyll when ancestor (Merckx 2013). Facing this apparent paradox, grown with abundant supplies of simple sugars (Beyrle the main take-­home message of the present study is that and Smith 1993), perhaps because they extend the nutri- the limited, potentially compensatory demographic pat- tional strategy of the heterotrophic protocorm under terns observed between albinos and green individuals in such conditions. This suggests that orchids may be par- our analysis open the door to survival, and potentially ticularly predisposed via phenotypic plasticity to occa- to speciation, in some cases. The pleiotropic dysfunctions sional loss of photosynthesis. Despite a limited number revealed in recent comparative studies of albino and of studies (cited above; see Selosse et al. In press for a green individuals (Roy et al. 2013) can be remediable by review), the latter phenomenon deserves more research, some change in genetic background and/or environment, especially in situ, for a greater understanding of the allowing speciation into mycoheterotrophy. While we context under which photosynthesis is evolutionarily can only speculate as to what environmental conditions lost. might favour nonphotosynthetic lineages, it is likely that Condition (2), maintenance of the trait, is likely the these environmental contexts may be satisfied by, at the most difficult condition, but may be met in adaptive or very least, a moist, dark forest with exploitable mycor- nonadaptive means. Traits may persist in a population rhizal fungi transporting carbon. if they increase fitness, if they are linked to adaptive traits, In this study, we observed strong differences between or if environmental variation changes the fitness context the two investigated orchid species, including flowering across space and time, making a trait advantageous behavior and relationships between demographic traits under at least some conditions (Crespi 2000, Byers 2005). and fitness. However, plant populations often exhibit For example, albinos typically exhibit greater water loss strong variability in population dynamics even within relative to greens due to stomatal dysfunction, but this species (Menges and Dolan 1998, Shefferson and Tali June 2016 MYCOHETEROTROPHY AND FITNESS 1461

2007), and so we cannot know whether the differences Byers, D. L. 2005. Evolution in heterogeneous environments we observed are due to differences between species or and the potential of maintenance of genetic variation in traits differences between populations, as might be the case if of adaptive significance. Genetica 123:107–124. Caswell, H. 1989. Analysis of life table response experiments local adaptation produced strong differences in eco- I. Decomposition of effects on population growth rate. logical response to different local conditions. Regardless, Ecological Modelling 46:221–237. further study of such species, particularly utilizing mul- Caswell, H. 2001. Matrix population models: construction, tiple populations, may yield a greater understanding of analysis, and interpretation. Second edition. Sinauer the extent to which evolutionary differences between Associates Inc, Sunderland, Massachusetts, USA. populations represent true species differences, or geo- Crespi, B. J. 2000. The evolution of maladaptation. Heredity 84:623–629. graphic differences in local adaptation. Dearnaley, J. D. W., F. Martos, and M. A. Selosse. 2012. 12 Orchid Plant species with both photosynthetic and nonphoto- mycorrhizas: molecular ecology, physiology, evolution and synthetic forms have rarely been the subject of evolu- conservation aspects. Pages 207–230 in B. Hock, editor. Fungal tionary research, and we argue that the evolutionary associations. Springer Berlin Heidelberg, Berlin, . context of photosynthesis itself cannot be understood de Pamphilis, C. W., N. D. Young, and A. D. Wolfe. 1997. without more research on these fascinating albino plants. Evolution of plastid gene rps2 in a lineage of hemiparasitic Studies of such species are inherently difficult to carry and holoparasitic plants: many losses of photosynthesis and complex patterns of rate variation. Proceedings of the out because of the need to limit the potential for observer-­ National Academy of Sciences of the United States of induced damage to populations of conservation concern, America 94:7367–7372. because such populations are rare, and because the sizes Gao, L., Y.-J. Su, and T. Wang. 2010. Plastid genome sequenc- of such population are often small. However, we believe ing, comparative genomics, and phylogenomics: current sta- that the transition to the loss of photosynthesis can only tus and prospects. Journal of Systematics and Evolution be understood through detailed, long-term­ study of such 48:77–93. Gebauer, G., and M. Meyer. 2003. 15N and 13C natural abun- populations, and we encourage more such work in the dance of autotrophic and myco-­heterotrophic orchids pro- hopes of developing robust data sets toward a general vides insight into nitrogen and carbon gain from fungal theory for the evolution of mycoheterotrophy. association. New Phytologist 160:209–224. Gonneau, C., J. Jersáková, E. de Tredern, I. Till-Bottraud, K. Saarinen, M. Sauve, M. Roy, T. Hájek, and M.-A. Selosse. Acknowledgments 2014. Photosynthesis in perennial mixotrophic Epipactis spp. RPS thanks the Dept. of General Systems Studies and the Dept. (Orchidaceae) contributes more to shoot and fruit biomass of Biology at the University of Tokyo for logistical support during than to hypogeous survival. Journal of Ecology the analysis and writing of this manuscript. Ü.P. would like to 102:1183–1194. thank K. Sepp, I. Tamu, E. Leibak, T. Gross, A. Palo, and Gremer, J. R., E. E. Crone, and P. Lesica. 2012. Are dormant U. Püttsepp. M.-A.S.­ was supported by the Laboratoire d’Excel- plants hedging their bets? Demographic consequences of pro- lence BCDiv, and M.R. by the Laboratoire d’Excellence CEBA. longed dormancy in variable environments. American Naturalist 179:315–327. Jacquemyn, H., R. Brys, and E. Jongejans. 2010. Size-dependent­ Literature Cited flowering and costs of reproduction affect population dynam- ics in a tuberous perennial woodland orchid. Journal of Abadie, J.-C., Ü. Püttsepp, G. Gebauer, A. Faccio, P. Bonfante, Ecology 98:1204–1215. and M.-A. Selosse. 2006. Cephalanthera longifolia (Neottieae, Julou, T., B. Burghardt, G. Gebauer, D. Berveiller, C. Damesin, Orchidaceae) is mixotrophic: a comparative study between and M.-A. Selosse. 2005. Mixotrophy in orchids: insights green and nonphotosynthetic individuals. Canadian Journal from a comparative study of green individuals and nonphoto- of Botany 84:1462–1477. synthetic individuals of Cephalanthera damasonium. New Barton, K. A. 2014. MuMIn: multi-model inference. Phytologist 166:639–653. Bates, D., M. Maechler, and B. Bolker. 2012. lme4: linear Klimešová, J. 2007. Root-­sprouting in myco-­heterotrophic mixed-effects models using S4 classes. plants: prepackaged symbioses or overcoming meristem limi- Bellino, A., A. Alfani, M.-A. Selosse, R. Guerrieri, M. Borghetti, tation? New Phytologist 173:8–10. and D. Baldantoni. 2014. Nutritional regulation in mix- Krause, K. 2008. From chloroplasts to “cryptic” plastids: evo- otrophic plants: new insights from Limodorum abortivum. lution of plastid genomes in parasitic plants. Current Genetics Oecologia 175:875–885. 54:111–121. Beyrle, H. F., and S. E. Smith. 1993. Excessive carbon prevents Leake, J. R. 1994. The biology of myco-heterotrophic­ (‘sapro- greening of leaves in mycorrhizal seedlings of the terrestrial phytic’) plants. New Phytologist 127:171–216. orchid Orchis morio. Lindleyana 8:97–99. Lesica, P., and B. M. Steele. 1994. Prolonged dormancy in vas- Bidartondo, M. I. 2005. Tansley review: the evolutionary ecol- cular plants and implications for monitoring studies. Natural ogy of myco-­heterotrophy. New Phytologist 167:335–352. Areas Journal 14:209–212. Bidartondo, M. I., B. Burghart, G. Gebauer, T. D. Bruns, and Martos, F., M. Dulormne, T. Pailler, P. Bonfante, A. Faccio, D. J. Read. 2004. Changing partners in the dark: isotopic and J. Fournel, M.-P. Dubois, and M.-A. Selosse. 2009. molecular evidence of ectomycorrhizal liaisons between for- Independent recruitment of saprotrophic fungi as mycor- est orchids and trees. Proceedings of the Royal Society of rhizal partners by tropical achlorophyllous orchids. New London. Series B, Biological Sciences 271:1799–1806. Phytologist 184:668–681. Bruns, T. D., M. I. Bidartondo, and D. L. Taylor. 2002. Host spec- Menges, E. S., and R. W. Dolan. 1998. Demographic viability ificity in ectomycorrhizal communities: what do the exceptions of populations of Silene regia in midwestern prairies: relation- tell us? Integrative and Comparative Biology 42:352–359. ships with fire management, genetic variation, geographic 1462 RICHARD P. SHEFFERSON ET AL. Ecology, Vol. 97, No. 6

location, population size and isolation. Journal of Ecology Shefferson, R. P. 2009. The evolutionary ecology of vegetative 86:63–78. dormancy in mature herbaceous perennial plants. Journal of Merckx, V. S. F. T. 2013. Mycoheterotrophy: an introduction. Ecology 97:1000–1009. Pages 1–17 in V. S. F. T. Merckx, editor. Mycoheterotrophy: Shefferson, R. P., and K. Tali. 2007. Dormancy is associated the biology of plants living on fungi. Springer Science and with decreased adult survival in the burnt orchid, Neotinea Business Media, New York, New York, USA. ustulata. Journal of Ecology 95:217–225. Metcalf, C. J. E., and S. Pavard. 2007. Why evolutionary biolo- Shefferson, R. P., T. Kull, and K. Tali. 2006. Demographic gists should be demographers. Trends in Ecology & Evolution ­response to shading and defoliation in two woodland orchids. 22:205–212. Folia Geobotanica 41:95–106. Miller, T. E. X., J. L. Williams, E. Jongejans, R. Brys, and Shefferson, R. P., M. McCormick, K. J. P. O’Neill, and D. F. H. Jacquemyn. 2012. Evolutionary demography of iteropa- Whigham. 2011. Life history strategy in herbaceous perenni- rous plants: incorporating non-lethal­ costs of reproduction als: inferring demographic patterns from the aboveground into integral projection models. Proceedings of the Royal dynamics of a primarily subterranean, myco-­heterotrophic Society B: Biological Sciences 279:2831–2840. orchid. Oikos 120:1291–1300. Nowak, M. A., C. E. Tarnita, and E. O. Wilson. 2010. The evo- Shefferson, R. P., T. Kull, K. Tali, and K. M. Kellett. 2012. lution of eusociality. Nature 466:1057–1062. Linking vegetative dormancy to fitness in two long-lived­ her- Pridgeon, A. M., P. J. Cribb, M. W. Chase, and F. N. Rasmussen, baceous perennials. Ecosphere 3:13. editors. 2006. Genera Orchidacearum. Volume 4: Shefferson, R. P., R. J. II Warren, and H. R. Pulliam. 2014. Life (Part 1). Oxford University Press, Oxford, UK. history costs make perfect sprouting maladaptive in two her- Primack, R., and E. Stacy. 1998. Cost of reproduction in the baceous perennials. Journal of Ecology 102:1318–1328. pink lady’s slipper orchid (Cypripedium acaule, Orchidaceae): Sletvold, N., and J. Ågren. 2015a. Climate-­dependent costs of an eleven-year­ experimental study of three populations. reproduction: survival and fecundity costs decline with length American Journal of Botany 85:1672–1679. of the growing season and summer temperature. Ecology Rasmussen, H. N. 1995. Terrestrial orchids: from seed to Letters 18:357–364. mycotrophic plant. Cambridge University Press, Cambridge, Sletvold, N., and J. Ågren. 2015b. Nonlinear costs of reproduc- UK. tion in a long-lived­ plant. Journal of Ecology 103: R Core Team. 2015. R: a language and environment for statisti- 1205–1213. cal computing. R Foundation for Statistical Computing, Smith, S. E. and D. J. Read. 2008. Mycorrhizal symbiosis. Third Vienna, Austria. edition. Academic Press, New York, New York, USA. Reznick, D. 1992. Measuring the costs of reproduction. Trends Stubben, C. J., and B. G. Milligan. 2007. Estimating and ana- in Ecology and Evolution 7:42–45. lyzing demographic models using the popbio package in R. Roy, M., C. Gonneau, A. Rocheteau, D. Berveiller, J.-C. Journal of Statistical Software 22:11. Thomas, C. Damesin, and M.-A. Selosse. 2013. Why do mix- Suetsugu, K. and M. Kato. 2014. Evolution of life history traits otrophic plants stay green? A comparison between green and suited to mycoheterotrophic life style with special reference to achlorophyllous orchid individuals in situ. Ecological pollination biology. BSJ-Review 5:93. Monographs 83:95–117. Tešitelová, T., J. Tešitel, J. Jersáková, G. Ríhová, and M.-A. Schaub, M., O. Gimenez, B. R. Schmidt, and R. Pradel. 2004. Selosse. 2012. Symbiotic germination capability of four Estimating survival and temporary emigration in the multi- Epipactis species (Orchidaceae) is broader than expected from state capture-­recapture framework. Ecology 85:2107–2113. adult ecology. American Journal of Botany 99:1020–1032. Selosse, M.-A., and M. Roy. 2009. Green plants that feed on Tranchida-Lombardo, V., M. Roy, E. Bugot, G. Santoro, fungi: facts and questions about mixotrophy. Trends in Plant Ü. Püttsepp, M. Selosse, and S. Cozzolino. 2010. Spatial Science 14:64–70. repartition and genetic relationship of green and albino indi- Selosse, M.-A., M. Weiß, J.-L. Jany, and A. Tillier. 2002. viduals in mixed populations of Cephalanthera orchids. Plant Communities and populations of sebacinoid basidiomycetes Biology 12:659–667. associated with the achlorophyllous orchid Neottia nidus-avis Tremblay, R. L., J. D. Ackerman, J. K. Zimmerman, and R. N. (L.) L.C.M. Rich. and neighbouring tree ectomycorrhizae. Calvo. 2005. Variation in sexual reproduction in orchids and Molecular Ecology 11:1831–1844. its evolutionary consequences: a spasmodic journey to diver- Selosse, M. A., A. Faccio, G. Scappaticci, and P. Bonfante. sification. Biological Journal of the Linnean Society 2004. Chlorophyllous and achlorophyllous specimens of 84:1–54. Epipactis microphylla (Neottieae, Orchidaceae) are associated White, G. C., and K. P. Burnham. 1999. Program MARK: sur- with ectomycorrhizal septomycetes, including truffles. vival estimation from populations of marked animals. Bird Microbial Ecology 47:416–426. Study 46(Supplement):120–138. Selosse, M.-A., M. F. Bocayuva, M. C. Kasuya, and P. E. Courty. Wolfe, A. D., and C. W. de Pamphilis. 1998. The effect of In press. Mixotrophy in mycorrhizal plants: extracting C from ­relaxed functional constraints on the photosynthetic gene mycorrhizal networks. F. Martin, editor. Molecular mycor- rbcL in photosynthetic and nonphotosynthetic parasitic rhizal symbiosis. Wiley-Blackwell, Hoboken, New Jersey, USA. plants. Molecular Biology and Evolution 15:1243–1258. Shefferson, R. P. 2002. Dormancy and survival in rare, terres- Yamato, M., Y. Ogura-Tsujita, H. Takahashi, and T. Yukawa. trial orchids. Pages 53–64 in P. Kindlmann, J. H. Willems, 2014. Significant difference in mycorrhizal specificity and D. F. Whigham, editors. Trends and fluctuations and ­between an autotrophic and its sister mycoheterotrophic underlying mechanisms in terrestrial orchid populations. plant species of Petrosaviaceae. Journal of Plant Research Backhuys Publishers, Leiden, the Netherlands. 127:685–693.

Supporting Information Additional supporting information may be found in the online version of this article at http://onlinelibrary.wiley.com/ doi/10.1890/15-1336.1/suppinfo