ORIGINAL ARTICLE

doi:10.1111/evo.13683

Cascading reproductive isolation: phenology drives temporal isolation among populations of a host-specific herbivore

Glen R. Hood,1,2,3 Linyi Zhang,1 Elaine G. Hu,1 James R. Ott,4 and Scott P. Egan1 1Department of Biosciences, Anderson Biological Laboratories, Rice University, Houston, Texas 77005 2Department of Biological Sciences, Wayne State University, Detroit, Michigan 48202 3E-mail: [email protected] 4Population and Conservation Biology Program, Department of Biology, Texas State University, San Marcos, Texas 78666

Received November 5, 2018 Accepted January 7, 2019

All organisms exist within a complex network of interacting species, thus evolutionary change may have reciprocal effects on multiple taxa. Here, we demonstrate “cascading reproductive isolation,” whereby ecological differences that reduce gene flow between populations at one trophic level affect reproductive isolation (RI) among interacting species at the next trophic level. Using a combination of field, laboratory and common-garden studies and long-term herbaria records, we estimate and evaluate the relative contribution of temporal RI to overall prezygotic RI between populations of Belonocnema treatae, a specialist gall- forming wasp adapted to sister species of live (Quercus virginiana and Q. geminata). We link strong temporal RI between host-associated populations to differences between host plant budbreak phenology. Budbreak initiates flowering and the production of new , which are an ephemeral resource critical to insect reproduction. As flowering time is implicated in RI between plant species, budbreak acts as a “multitrophic multi-effect trait,” whereby differences in budbreak phenology contribute to RI in and . These sister oak species share a diverse community of host-specific gall-formers and insect natural enemies similarly dependent on ephemeral plant tissues. Thus, our results set the stage for testing for parallelism in a role of plant phenology in driving temporal cascading RI across multiple species and trophic levels.

KEY WORDS: Belonocnema treatae, live oak, multitrophic multieffect trait, Quercus, reproductive isolation.

Ecology plays an essential role in the speciation process when (Talley et al. 2001). However, a synthetic view of how ecologically barriers to gene flow evolve between populations as a result of driven RI arises requires understanding (1) if and how individual ecologically based divergent natural selection (Rundle and Nosil phenotypes are affected by divergent selection between popula- 2005). Thus, knowledge of how such barriers arise is necessary tions experiencing different environments, (2) if those effected for understanding how divergence among populations is initiated phenotypes reduce gene flow between diverging populations, and and maintained. In the last 35 years, the role that ecology plays (3) how multiple barriers accumulate to contribute to RI. during population divergence and the evolution of reproductive To date, a small but growing number of studies have esti- isolation (RI) has been the subject of intensified research (Nosil mated the combined effect of multiple barriers to RI (e.g., Ramsey 2012). Consequently, the study of ecologically based RI has in- et al. 2003; Martin and Willis 2007; Matsubayashi and Katakura creased our understanding of how barriers to gene flow evolve in a 2009; Dopman et al. 2010; Sanchez-Guillen et al. 2012; Lackey diversity of taxa including plants (Richards and Ortiz-Barrientos and Boughman 2017; Paudel et al. 2018; Sambatti et al. 2012). 2016), fishes (Rundle 2002), insects (Feder et al. 1994; Egan However, even for well-studied systems, biologists frequently and Funk 2009), birds (Huber et al. 2007), amphibians (Twomey lack a detailed understanding of the relative contributions of the et al. 2014), reptiles (Rosenblum et al. 2010), and mammals individual components of RI to total RI and/or the chronological

C 2019 The Author(s). Evolution C 2019 The Society for the Study of Evolution. 554 Evolution 73-3: 554–568 TEMPORAL ISOLATION IN POPULATIONS

order in which components of RI evolve. Consequently, the host plants and/or the probability that immigrants successfully forms of RI most common during ecological speciation remain reproduce. Thus, for short-lived insect species, temporal RI may unclear (see “Twenty-Five Major, Yet Unresolved, Questions in represent an ecological adaptation of a population to its environ- Ecological Speciation,” Nosil 2012). ment that can also function as a critical barrier to gene flow and Here, we investigate an understudied driver of RI, which promote population divergence. we describe as “cascading RI,” whereby ecological differences in In their comprehensive review of the role of allochrony dur- traits that reduce gene flow between populations at one tropic level ing speciation, Taylor and Friesen (2017) conclude that (1) dif- “cascade” upward to similarly affect interacting organisms at the ferences in the timing of key life-cycle events can contribute to next trophic level. To test for cascading temporal RI in a special- RI, (2) the effect of allochrony can substantially reduce gene flow ized insect herbivore, we first document phenological differences when it disrupts the overlap of breeding times between popula- in the trait budbreak, a hypothesized driver of RI between sister tions, and (3) allochrony may often be the initial or key driver plant taxa occupying different microhabitats. Second, we test the of speciation. The last point is important owing to the sequen- role of ecology (differences in the timing of availability) in tial nature of RI—those barriers that act earliest in the life cycle generating temporal RI between host-plant-associated insect pop- contribute more to total RI when multiple barriers are present ulations dependent on this ephemeral resource. Third, we combine (Hood et al. 2012; Ramsey et al. 2003). Thus, the accumulation estimates of temporal RI with two established prezygotic barri- of barriers to gene flow can act as sieves, with the earliest acting ers to estimate total RI among host-associated insect populations. barriers allowing fewer individuals to pass genes to the next gen- Cascading RI differs from “cascading divergence” whereby both eration. Despite the orthodoxy of the perceived role of allochrony ecological and genetic divergence at one trophic level results in in insect speciation, Forbes et al. (2017) found that allochronic concordant divergence of species at the next trophic level that isolation initiated speciation via host plant shifting in only 13 adapt to the newly created niche. For example, the shift and sub- out of 85 (15%) herbivorous insect systems, while Taylor and sequent divergence of Rhagoletis pomonella from native plant Friesen (2017) documented only nine cases of allochrony driv- hawthorn to apple resulted in the formation of a new niche that ing ecological speciation (five of which were in insect systems) was colonized by multiple species that have diverged and an additional 56 systems in which “further study is needed.” in tandem with R. pomenella (Hood et al. 2015). Moreover, while Thus, the assertion that “allochronic isolation is common . . . for cascading RI is a necessary component of cascading divergence, phytophagous insects” (Berlocher and Feder 2002) may be less RI at both trophic levels must be under (partial) genetic control. substantiated by available data than models and theory alone Cascading RI, as defined herein, need not necessarily be con- suggest. trolled genetically. Herein, we postulate that the ecological driver (divergence Herbivorous insects and their host plants are excellent sys- in the timing of the trait budbreak) hypothesized to generate tems to investigate temporal RI as many species are highly spe- RI between sister species of live Quercus virginiana and cialized on one or a few closely related plant species and depen- Q. geminata (Cavender-Bares and Pahlich 2009), cascades dent on specific tissues as sites for courtship, mating, feeding, across trophic levels to generate temporal RI between host- oviposition, and development (Bernays 1998; Funk et al. 2002). plant-associated populations of the gall-former, Belonocnema Host plant specialization often has a temporal component as plant treatae. Here, we combine field and laboratory observations, tissues can be used for only short windows of time. Thus, the tim- common-garden experiments and long-term herbaria records to ing of insect life-cycle events becomes critical when reproductive link phenological differences in host plant flowering and leaf success is intrinsically tied to the acquisition of ephemeral re- flush to temporal RI among associated gall-former populations. sources. As a result, host-tissue-specific populations of insects Importantly, two additional phenotypes, adult longevity and adapting to different plant environments that vary in the produc- breeding time, interact to affect temporal RI. If populations tion of a critical plant tissue may experience a degree of “tempo- emerge at different times on alternative host plants, but earlier ral” or “allochronic” RI as a consequence of ecologically based emerging individuals survive long enough to partially/completely selection (Mopper 2005). Evidence for allochrony among host- overlap with the later emerging population during reproduction, plant-associated insect populations exists for a number of taxa, for RI will be reduced/eliminated. To date, few studies have com- example treehoppers (Wood 1993; Wood et al. 1999), and Rhago- bined these phenotypes to estimate temporal RI between insect letis fruit flies (Filchak et al. 2000). In cases in which insects populations residing on different hosts (Feder et al. 1993; Forbes reside on multiple plant species whose ranges overlap, temporal et al. 2009; Hood et al. 2015; Mattsson et al. 2015). Thus, we use RI between plant species can disrupt gene flow between herbivore this approach to quantify temporal RI between host-associated populations by reducing the chance that individuals from alterna- populations. We then combine published estimates of habitat tive host plants will mate when migration occurs to “nonnatal” isolation (Egan et al. 2012a) and sexual isolation (Egan et al.

EVOLUTION MARCH 2019 555 G. R. HOOD ET AL.

2012b) with estimates of temporal isolation to estimate total suring temporal RI between populations of the gall former due to prezygotic RI between host-associated populations of B. treatae. differences in the timing of newly flushed leaves between plant species, thus we focus on the emergence of the sexual generation. Prior research demonstrates that B. treatae from parapatric Qv and Qg in Florida exhibit ecological, morphological, and be- Methods havioral differences. First, B. treatae produce larger and leaf STUDY SYSTEM galls and adults of both generations are larger on Qg than Qv Belonocnema treatae (: Cynipidae: ) is a (Egan et al. 2012a, 2013). Second, host-associated populations gall-forming wasp, distributed across the southern and southeast- of B. treatae exhibit habitat isolation, showing strong preference ern United States, that exhibits regional host plant specialization for their natal host plant (Egan et al. 2012b), and sexual isola- on species of live oak, section Virentes (Lund et al. 1998; Schuler tion, showing mate preference for individuals from the same host et al. 2018). In southern Florida, where fieldwork for this study plant species. Third, B. treatae can form galls that generate viable was conducted, host-associated populations of B. treatae develop adults on alternative host plants but experience a high degree of on the sister species, Quercus virginiana (Qv)andQ. geminata immigrant inviability on nonnatal host plants (Zhang et al. 2017). (Qg). These species occupy different habitats with Qv commonly Last, hybrid B. treatae can induce leaf galls on both Qv and Qg found in mesic soils and Qg found in xeric sandy soils (Cavender- (unpubl. data; G. R. Hood and L. Zhang). Taken collectively, these Bares et al. 2015). Therefore, the scale at which the size and lines of evidence suggest that divergent host plant use promotes distribution of habitats vary determines the scale at which alter- ecological divergence. nate host plants occur. Hence, the two host plant species co-occur regionally as a mosaic spanning sympatry to parapatry. More- over, these species differ in leaf morphology and physiological TESTING FOR CASCADING REPRODUCTIVE traits (Cavender-Bares and Pahlich 2009), likely presenting differ- ISOLATION ing challenges to gall-forming insects. On average, Qv produces To test whether temporal RI between sister host plant species cas- larger leaves and assimilates CO2 at a higher rate and, therefore, cades across trophic levels to affect temporal RI in the specialized experiences greater water loss per unit biomass compared with insect herbivore, we investigated differences in life-history traits at Qg. As well, Qv has been reported to earlier (Sargent both the plant and insect trophic levels across a broad spatiotempo- 1918; Nixon and Muller 1997), and importantly, this difference ral scale (Fig. 1). We first tested whether the differences in the tim- is hypothesized to be a barrier to gene flow between these sister ing of resource availability (i.e., flowering time, which is a proxy species (Cavender-Bares and Pahlich 2009). for leaf flush) documented most recently by Cavender-Barres and Throughout Florida, Qv and Qg display frequency-based dif- Pahlich (2009) for seven individual Qv and Qg at one site in ferences in microsatellite markers and can occasionally form hy- northern Florida, represents a general phenomenon throughout brids but overall gene flow is limited (divergence is estimated as our study region. To do this, we compared the flowering phenol- 5–7 mya; Cavender-Bares et al. 2015; Hipp et al. 2018). Whether ogy of Qv and Qg based on herbaria records across a noncontin- host-associated populations of B. treatae represent a host shift or uous 86-year period from 1925 to 2011. We then compared the codiversification with the oak species is unknown. However, the phenology of leaf flush across seven sympatric/parapatric pop- presence of B. treatae populations on Q. fusiformis, the most dis- ulations of the host plants in southern Florida known to harbor tantly related of the live oak clade to sister plant species Qv and high gall densities. These two approaches estimate the timing of Qg (Cavender-Bares et al. 2009), is compatible with the scenario resource availability for B. treatae.Inacommongardenenvi- of plant–insect codiversification. ronment, we then compared the timing of leaf flush of Qg and Many cynipids, including B. treatae (Lund et al. 1998), ex- Qv to test whether the between-species difference in the onset of hibit cyclical parthenogenesis, in which temporally and spatially leaf flush was environmentally mediated. To test the hypothesis segregated sexual and asexual generations alternate to complete that differences in host plant phenology result in parallel differ- the life cycle (Stone et al. 2002). Sexual B. treatae emerge from ences in the phenology of emergence of sexual B. treatae,we multichambered root galls commensurate with spring budbreak compared the timing of sexual emergence across multiple parap- and leaf flush. Upon emergence, sexual females immediately atric populations of B. treatae developing on the same Qv and Qg mate on or near the host plant and begin ovipositing into newly plants that were monitored for leaf flush phenology under both flushed leaves. The leaf galls that are induced each house a single field and laboratory conditions. Last, we measured longevity of asexual. Asexuals emerge in the fall, oviposit into and adults from both host plant species, which we combine with esti- produce either all male or all female sexual generation broods mates of emergence phenology, to calculate temporal RI between within root galls. In the current study, we were interested in mea- host-associated populations of B. treatae.

556 EVOLUTION MARCH 2019 TEMPORAL ISOLATION IN GALL WASP POPULATIONS

A B

Figure 1. (A) Trees monitored for flowering time from herbaria records. Qv flowering = gray (n = 28); Qv not flowering = black (n = 100); Qg flowering = pink (n = 24); Qg not flowering = red (n = 86). Resolution is at the county level. On occasion, when multiple trees of a single species are present in a single county, the number is shown in a box preceded by an “×.” (B) Sites monitored for leaf flush phenology and/or B. treatae emergence in southern Florida in 2017. Numbers in (B) correspond to the locations described in Table S1.

HERBARIUM RECORDS OF FLOWERING TIME with catkins present (n = 24 Qg; n = 28 Qv) and conducted We compiled records of Qv and Qg flowering tim- an analysis of covariance (ANCOVA) with the factors “plant ing from three digital herbaria: Florida State University species” as a fixed effect and the covariate “geography” as a (http://herbarium.bio.fsu.edu), University of Florida (https:// random effect. For herbaria records missing latitude, we used www.floridamuseum.ufl.edu/herbarium), and Louisiana State the geographic center of the county in which the record was University (http://www.herbarium.lsu.edu). In general, for oaks, recorded. These and subsequent analyses were performed in R as buds break, leaves unfurl in tandem with shoot elongation, v3.2.3. Means and standard errors (SE) are reported throughout. male flower (catkin) expansion, and bloom (Hood et al. 2018). Shoots with catkins represent new flower production as mature PHENOLOGY OF LEAF FLUSH catkins abscise within one to two weeks, thus flowering timing is We compared leaf flush phenology of Qv and Qg in the eco- commensurate with leaf flush. Each herbarium record included logical context that is directly relevant to determining RI in photographs of collected branches that we scored for the B. treatae: by monitoring the production of new leaves at seven presence/absence of , collection date, and geographic sympatric/parapatric sites where B. treatae was also present location (county or GPS coordinates). We included data from 50 (Fig. 1, Table S1). The majority of sites (and trees within sites) counties in Florida, 2 in Mississippi, 2 in Georgia, and 1 county exhibited high leaf and root gall densities. At three sites (Arch- in Alabama. In total, 238 records were examined (n = 110 Qg; bold, Kissimmee River, Lake Lizzie), we located and monitored n = 128 Qv). We note that almost all branches with flowers were leaf flush for trees of the alternative live oak species that were collected during or near typical timing of live oak leaf flush from not galled. Additionally, we monitored leaf flush phenology at a February to May, while those branches without flowers were sympatric site (Gatorama) that historically harbored low leaf galls largely collected outside this window. Furthermore, plants scored densities on Qv but was devoid of galls in 2017. We monitored for the presence/absence of flowers were distributed across the trees 14 times (every 2 to 4 days) across a 45-day period from calendar year, and sampling was similar across host plant species March 8 to April 22 2017. Not all sites were monitored on the (Fig. S1). Thus, to compare flowering time between species we same day, as sites were separated by hundreds of kilometers. At pooled observations across years and herbaria for the records each sampling event and site, we randomly chose 3–15 branches

EVOLUTION MARCH 2019 557 G. R. HOOD ET AL.

per tree and counted the number of broken and unbroken ax- to avoid signs of stress and were fertilized yearly in March and illary leaf bud scales (undeveloped embryonic shoot tissue that November with general purpose 20-20-20 (nitrogen-phosphorus- gives rise to new leaves) along a 30–120 cm length of branch potassium) water-soluble fertilizer at a concentration of 750 ppm, beginning at the terminal end. As bud density varied among trees, and Florikan 19-5-7 270 Nutricote Advantage slow-release fer- the number and length of branches measured per tree was adjusted tilizer (Floridan E.S.A., LLC, Sarasota, FL) at 15 grams per pot. to ensure adequate sample sizes. To capture the tree-wide phenol- All plants were maintained outdoors at the Texas State University ogy and control for within-tree variation, the same branch–bud greenhouse (San Marcos, TX). combination was not measured in successive surveys. In total, we We monitored the date of first budbreak for each five-year-old scored the status of 74,744 buds distributed across seven sites and sapling (n = 67 Qv and n = 61 Qg) daily beginning March 2, 2017 41 trees. Inspection of our data and herbarium records suggest and continuing until budbreak was observed for all trees (April that Qv had already begun flushing leaves prior to the start of 16, 2017). While the methods for monitoring leaf flush differ monitoring; thus, our estimates of the differences in leaf flush between field studies in Florida and the common garden study in between Qv and Qg are conservative. Texas, the latter represent the more conservative estimate of the To test the hypothesis that leaf flush differs between Qv and phenological differences in the timing of leaf flush because initial Qg, we compared the distributions of the percentage of buds bro- budbreak represents the earliest date that a wasp could oviposit. ken through time (i.e., number of buds broken at each interval/total For analysis, we performed an unpaired t-test of Julian dates of number of buds measured across all intervals) by means of a series first budbreak to compare phenology between plant species. of multifactor ANOVA. In each ANOVA, we logit transformed the percent of buds broken following Collett (2002) and Warton PHENOLOGY OF GALL-FORMER EMERGENCE and Hui (2011), whereby 0.01 × 10−10 was added to each ob- To test the hypothesis that differences in host plant phenology servation when no buds were broken. We pooled the percent of translate into parallel differences in the phenology of emergence buds broken across branches for each tree and the factors “site,” of sexual B. treatae, we compared the distribution of emergence “plant species,” and “individual tree” were considered as fixed times of adult gall wasp across the populations of Qv and Qg that effects. A series of preliminary two-way ANOVA conducted per were simultaneously being monitored for leaf flush. We monitored site showed the phenology of leaf flush did not differ among trees emergence at six high gall density sites by placing yellow sticky within a site for either species (P > 0.05 in all cases) with one traps under trees and also by rearing detached root galls in a exception: a post-hoc analysis revealed that one of four Qv trees at laboratory setting (Fig. 1, Table S1). Every two to four days from Lake Lizzie flushed leaves significantly later (e.g., reached 50% March 3 through April 21 over 14 monitoring intervals, we placed and 90% leaf flush 20 and 13 days later, respectively). Thus, to a 3.5 cm × 4.5 cm trap atop a 1-meter tall bamboo stick driven test for differences in the timing of leaf flush between host plants, vertically into the ground under each tree. At each interval, we we averaged cumulative emergence phenology across individual deployed one to five new traps (dependent on the number of trees trees of the same species within a site (including all four Qv at monitored) and collected all old traps. For analysis, we counted Lake Lizzie) and performed a fully crossed two-way ANOVA the total number of gall wasps per trap per sample interval per site with “site” and “plant species” as factors. The average dates by (pooled across multiple traps when applicable). We also collected which 25%, 50%, and 90% of buds were broken (no tree breaks and then monitored emergence from root galls from the trees for all buds) for both Qv and Qg are reported. which we deployed traps. Informed by when adults were initially captured by sticky traps, samples were collected two to four times COMMON GARDEN EXPERIMENT between March 11 and March 30, 2017 for Qv and between March To determine whether differences in leaf flush between Qv and Qg 18 and April 15, 2017 for Qg. Galls were housed individually reflect environmental conditions characterizing the habitats occu- in clear plastic Drosophila vials or in bulk in 2-L clear plastic pied by each plant species, we monitored budbreak for Florida- bottles placed outdoors in a shaded alcove (i.e., the laboratory) derived populations of each species after five years of growth in under ambient environmental conditions at Archbold Biological a common environment. In fall of 2013, we purchased one-year- Station (Venus, FL), and misted daily. Galls were monitored for 50 old seedlings of Qv and Qg from Superior Trees, Inc. (Lee, FL) consecutive days, at which point emergence had ceased. In total, derived from acorns collected from at least four sites in southern we recorded emergence dates for 15,252 individuals in nature and Florida to ensure a mixture of genotypes. Plants were grown in 5,205 individuals in the laboratory. individual pots in a standardized soil mix consisting of Sphag- We compared the distributions of the logit-transformed pro- num Peat Moss, potting soil, pulverized granite, and Perlite. portion of adults that emerged through time from each host plant Saplings were repotted as needed based on size and water re- using a series of multifactor ANOVA with the factors “site” and quirements into 4-gallon pots by fall 2015. Trees were watered “host plant species” as fixed effects for each method. Preliminary

558 EVOLUTION MARCH 2019 TEMPORAL ISOLATION IN GALL WASP POPULATIONS

two-way ANOVA revealed that emergence timing did not differ TOTAL PREZYGOTIC REPRODUCTIVE ISOLATION among trees within or between sites on either host plant (P > 0.05 We estimated total RI (RItotal) between B. treatae populations in all cases) with one exception: a post-hoc analysis showed that developing on Qg versus Qv due to the combined effects of three B. treatae from Archbold reared in the laboratory emerged later sequential components of prezygotic RI (temporal isolation [TI], than at other Qg sites. Thus, we performed a one-way ANOVA habitat isolation [HI] and sexual isolation [SI]). Habitat isolation pooling emergence phenologies across all trees within each was calculated from preference experiments of B. treatae reared species for each method. We also report the average dates by from Qv and Qg for natal and nonnatal host plants (Egan et al. which 25%, 50%, and 95% of adults emerged for both Qv- and 2012a). We calculated HI, which measures the behavioral overlap Qg-associated populations. between populations x and y to each other’s environment (leaf material), following Powell et al. (2014): ADULT LONGEVITY p + p HI = 1 − x y To estimate longevity of male and female B. treatae from each host 2 association, we placed newly emerged wasps individually into where px is the proportion of a gall wasp population from host 500-mL clear plastic cups stocked with fresh clippings of the natal plant x displaying a preference for the nonnatal host plant y,and host plants (to promote insect activity) housed outside. Five to six py is the proportion from host plant y displaying preference for times daily, we misted cups to maintain a humid environment the nonnatal host plant x.WeusedestimatesofSI from Egan et al. and recorded whether each wasp was alive or dead. Longevity (2012b) that were based on pairwise comparisons of mating fre- was calculated as the number of days between emergence and quencies among populations using the standard metric IPSI,which death. Preliminary analysis revealed that longevity did not differ ranges from −1 (complete disassortative mating) to 0 (random = = = between males and females from Qg (t 0.67, P 0.51; n♂ mating) to 1 (complete assortative mating). To estimate RI ,we = ± = = ± total 52, mean 2.98 0.17; n♀ 56, mean 3.14 0.17), but was combined HI, SI, and TI via Ramsey et al. (2003). Estimates of = = = = ± different for Qv (t 2.47, P 0.02; n♂ 52, mean 2.42 0.13; RI take into account the contribution of each barrier consid- = = ± total n♀ 59, mean 2.00 0.12). Furthermore, longevity differed ering the reduction in gene flow imposed by the barriers acting = = = between sites for Qg (F 4.54, P 0.01) but not Qv (F previously in the life cycle. Total prezygotic RI and the contri- 1.97, P = 0.12). Due to a lack of ubiquitous site- and sex-specific bution of each barrier at three stages (R1, R2, and R3) of the life differences across host plant species, we pooled individuals across cycle were calculated as: sex and populations within each host association for analysis.

RI1 = TI, (1) TEMPORAL REPRODUCTIVE ISOLATION = × − , To test the hypothesis that difference in the timing of leaf flush RI2 HI (1 RI1) (2) generates temporal RI, we compared RI values between B. treatae RI3 = SI × [1 − (TI+ HI)] , and (3) populations developing on different host plants (Qv vs Qg)toRI values between populations developing on the same host plant RItotal = RI1 + RI2 + RI3. (4) (Qv vs Qv, Qg vs Qg). We combined estimates of the timing of B. treatae emergence and adult longevity to calculate temporal RI Our estimates of TI, HI,andSI provide a range of values between adults from all six populations. This approach generated based on multiple pairwise combinations of B. treatae populations nine pairwise comparisons of populations of wasps developing developing on Qv and Qg; thus,weestimatedRItotal using the on different host plant species and six pairwise comparisons of single lowest, the average, and the single largest pairwise estimate wasps developing on the same host plant. We quantified temporal of each barrier. In addition, for each method, we calculated the isolation (TI) between population pairs using the equation from relative contribution of each barrier to RItotal. Feder et al. (1993):  xi yi TI = 1 −   Results 2 2 xi yi HERBARIUM RECORDS OF FLOWERING TIME Collectively, herbarium records of flowering time, comparison of where xi and yi are the proportion of wasps from populations x and leaf flush phenology, and common garden studies confirmed that y alive on day i based on emergence phenology and probabilities differences in the timing of leaf flush between Qg and Qv rep- of survival to day i calculated from average longevity estimates. resents a general phenomenon throughout the region. Herbarium We performed this analysis separately for emergence phenology records show that across years, on average, Qv flowered (mean data collected from nature and the laboratory. = 29 March ± 8 days) 38 days earlier than Qg (mean = 6 May

EVOLUTION MARCH 2019 559 G. R. HOOD ET AL.

Figure 2. Box plot of the average flowering date (a proxy for the timing of leaf flush) of Qv (n = 24) and Qg (n = 28) from herbaria records spread across an 86-year period from 1925 to 2011.

± 12 days) (Fig. 2). ANCOVA confirmed that flowering time plants were reared in a common garden in Texas (t126 = 9.03; differed significantly between plant species (F1,48 = 8.30, P = P < 0.0001). The average day of first leaf flush was March 8, 0.006). In addition, ANCOVA revealed a modest effect of latitude 2017 ± 5 days and March 18, 2017 ± 7daysforQv and Qg,re-

(F1,48 = 5.16, P = 0.03), with plants at higher latitudes flower- spectively. This difference equates to approximately one-half the ing later and a nonsignificant plant species × latitude interaction magnitude of the difference in phenology of leaf flush observed (F = 0.71, P = 0.40), indicating that differences observed across between host plant species in naturally occurring populations in latitude scale equivalently between host plant species. Florida, again supporting intrinsic phenological differences in the timing of leaf flush between the two oak species. PHENOLOGY OF LEAF FLUSH An ANOVA of the cumulative number of buds broken across the PHENOLOGY OF GALL-FORMER EMERGENCE season showed that the phenology of leaf flush differed between The timing of emergence of sexual generation B. treatae dif- natural populations of Qv and Qg as evidenced by the significant fered between natural populations developing on Qv and Qg −7 host plant effect (F1,143 = 30.51, P = 1.52 × 10 ). This difference (F1,82 = 14.55, P = 0.0003). On average, Qv-associated popu- was equivalent among sites across southern Florida as indicated by lations emerged earlier, with the differences between when 25%, the nonsignificant site (F6,143 = 1.63, P = 0.14) and plant × site in- 50%, and 95% of adults emerged being 14, 17, and 24 days, re- teraction effects (F3,143 = 0.85, P = 0.47). On average, the differ- spectively (Fig. 4A). For wasps that emerged from galls that had ence between when host plant species reached 25%, 50% and 90% been excised from plants and reared in the laboratory, the direction of leaves flushed was 20, 22, and 19 days, respectively. This pat- of the results was similar, with Qv derived populations emerging −9 tern is evident in comparisons within the four sympatric sites that earlier (F1,304 = 37.61, P = 2.68 × 10 ). The average difference capture the temporal displacement in resource availability con- between when 25%, 50% and 95% of adults emerged was 12, 14, fronted by B. treatae as they emerge (Fig. 3), and strongly suggests and 10 days, respectively (Fig. 4B). that the observed difference in leaf flush phenology is not due to variation in environmental conditions among sampling locations. TEMPORAL REPRODUCTIVE ISOLATION Estimates of longevity and the timing of emergence allowed us to COMMON GARDEN EXPERIMENT calculate temporal RI between host-associated populations of B. On average the difference in the timing of first budbreak be- treatae on Qv and Qg. Longevity of adult B. treatae from both Qv tween Qv and Qg from southern Florida was maintained when (mean = 2.20 ± 0.09 days, range = 1–5 days) and Qg (mean =

560 EVOLUTION MARCH 2019 TEMPORAL ISOLATION IN GALL WASP POPULATIONS

1.00 A Archbold (1) B Lake Lizzie (2) 0.80

0.60

0.40

0.20

0.00 1.00 C Kissimmee River (4) D Gatorama (7)

0.80

0.60

0.40

0.20

0.00 3/6 3/11 3/16 3/21 3/26 3/31 4/5 4/10 4/15 4/20 4/25 1.00 E Pooled

0.80

0.60

0.40

0.20

0.00 3/6 3/11 3/16 3/21 3/26 3/31 4/5 4/10 4/15 4/20 4/25

Figure 3. Cumulative curves of the frequency of new leaves flushed on individual trees, monitored from early March until late April, 2017 at four sympatric sites in southern Florida: (A) Archbold Biological Research Station, (B) Lake Lizzie, (C) Kissimmee River, and (D) Gatorama, and (E) 24 individual Qv (gray circles) and 22 individual Qg (maroon squares) pooled across all trees at all seven sites. Site number designations are given after site names in parentheses. The black and red curves in (E) represent the average across Qv and Qg, respectively. Note that we began monitoring most Qv trees after the date of initial leaf flush.

3.06 ± 0.12 days, range = 1–7 days) differed significantly be- with two exceptions. First, same-host comparisons involving Qg tween host plant species (t217 = 5.83; P < 0.0001). Based on at Archbold, the site with the latest-emerging B. treatae,were emergence from galls reared in the laboratory, temporal RI be- relatively large, but still smaller than different-host comparisons tween B. treatae developing on Qv versus Qg ranged from 26.03% involving Archbold. Second, different-host comparisons involv- to 92.06% with an average of 55.54 ± 7.75% across the nine pair- ing Qv at Kissimmee, the site with the latest-emerging B. treatae wise comparisons (Table 1). Estimates of temporal RI between from Qv, were comparable to control values for comparisons in- B. treatae populations affiliated with different hosts based on the volving Kissimmee (Table 1). These results suggest that while emergence phenology in nature exceeded the above estimates in temporal isolation is substantial between Qv- and Qg-associated all nine cases, with temporal RI ranging from 65.10% to 95.34% B. treatae populations, levels of RI also may be considerable with an average of 82.06 ± 4.71% (Table 1). Despite a mean between gall-former populations on the same plant species. difference of 23.7 ± 2.4%, the two datasets were significantly correlated (R = 0.64, P = 0.01). TOTAL PREZYGOTIC REPRODUCTIVE ISOLATION Inspection of estimated temporal isolation between B. treatae Using our estimates of temporal RI coupled with published esti- residing on different host plants (Qv × Qg) versus the same host mates of habitat isolation (Egan et al. 2013) and sexual isolation plants (Qv × Qv and Qg × Qg), shows that same-host compar- (Egan et al. 2012a), we estimated total prezygotic RI (RItotal) isons were almost always lower than different-host comparisons between populations of B. treatae developing on Qv and Qg.

EVOLUTION MARCH 2019 561 G. R. HOOD ET AL.

1.00 A 0.80

0.60

emerged

0.40 Q. virginiana Q. geminata

treatae 0.20 Alva(6) ABS (1)

B. KRE(4)KRE DSP (3) Okee (5) LL (2) 0.00 adult 3/4 3/14 3/24 4/3 4/13 4/23 5/3

of 1.00 B 0.80

frequency 0.60

0.40

0.20 Cumulave

0.00 3/4 3/14 3/24 4/3 4/13 4/23 5/3

Figure 4. Cumulative frequency of emergence of B. treatae from root galls formed on Qv (circles) and Qg (squares) monitored between early March and early May, 2017 in southern Florida (A) caught on yellow sticky traps in nature and (B) monitored in the laboratory. Qv sites = Alva (red), Kissimmee River (KRE; orange), Okeechobee (Okee; gray); Qg sites = Archbold Biological Station (ABS; purple), Jonathan Dickinson State Park (DSP; blue), Lake Lizzie Nature Preserve (LL; green). Site number designations are given after site names in parentheses.

Table 1. Estimates of temporal RI between B. treatae populations developing on the same and different host plants in the laboratory and in nature.

Pairwise comparison Comparison Laboratory RI (%) Nature RI (%)

Kissimmee (4) × Alva (6) Same host 32.32 32.03 Kissimmee (4) × Okeechobee (5) 26.58 46.47 Alva (6) × Okeechobee (5) 32.20 39.37 Archbold (1) × Lake Lizzie (2) 43.54 68.23 Archbold (1) × Dickinson (3) 61.70 75.82 Lake Lizzie (2) × Dickinson (3) 30.69 4.62 Average 37.84 (5.30) 44.42 (10.52) Kissimmee (4) × Archbold (1) Different host 80.67 95.10 Kissimmee (4) × Lake Lizzie (2) 35.01 65.88 Kissimmee (4) × Dickinson (3) 26.03 65.10 Alva (6) × Archbold (1) 75.36 93.78 Alva (6) × Lake Lizzie (2) 55.72 69.90 Alva (6) × Dickinson (3) 59.94 68.17 Okeechobee (5) × Archbold (1) 92.06 95.34 Okeechobee (5) × Lake Lizzie (2) 39.77 92.5 Okeechobee (5) × Dickinson (3) 35.34 92.77 Average 55.54 (7.75) 82.06 (4.71)

Site number designations are given after site names in parentheses. In bold are the average RI values ± SE in parentheses.

562 EVOLUTION MARCH 2019 TEMPORAL ISOLATION IN GALL WASP POPULATIONS

Table 2. The lowest, average, and highest values of temporal isolation (TI), habitat isolation (HI) and sexual isolation (SI) and total RI for populations of B. treatae developing on different host plants.

Type of RI (Percent contribution)

Estimate of RI TI HI SI RItotal

Low 26.03 (32.01) 73.40 (66.78) 5.0 (0.98) 81.31 Average 68.80 (72.30) 77.48 (25.41) 31.0 (2.29) 95.15 High 95.34 (95.56) 83.25 (3.89) 70.0 (0.55) 99.77

Values in parentheses represent the percent contribution of each barrier to the total.

Estimates of RItotal ranged from 81.31% to 99.80%, with an aver- demonstrated that the difference in the direction and magnitude age of 95.15% between B. treatae populations (Table 2). Whether of phenology between the host plants, which is a hypothesized calculated using the average or largest values of each component reproductive barrier between the two live oak species (Cavender- of prezygotic RI, temporal RI disproportionately contributed to Bares and Pahlich 2009), accurately predicts the direction and

RItotal (72% and 96%, respectively). However, when estimated magnitude of the difference in emergence phenology of popula- using the lowest value of each component of prezygotic RI, tem- tions of the gall former developing on these host plant species. poral RI contributed 32% compared to habitat isolation, which In fact, a correlation between average date of live oak leaf flush contributed 66% (Table 2). and B. treatae emergence across the six sites for which phenol- ogy was measured for both plant and insect is significant in nature (r = 0.943, P = 0.005) and in the laboratory (r = 0.937, P = 0.006) Discussion (Fig. 5). As a consequence, in combination with the ephemeral The evolution of temporal RI is a critical consideration in mod- nature of newly flushed leaves, and the short adult life span of els of population divergence and speciation (Taylor and Friesen sexual B. treatae, strong temporal RI is generated between pop- 2017), especially during the early stages of host race formation in ulations of the gall wasps associated with these alternative host host-plant-specific insects (Berlocher and Feder 2002; Dres and plant species. This temporal RI differentially contributes to total Mallet 2002). As acknowledged by Coyne and Orr (2004), few prezygotic RI and, in combination with habitat and sexual isola- studies have addressed the hypothesis that temporal RI exist when tion, is predicted to result in strong total prezygotic RI between the feeding ecology and/or life cycle of an is intrinsically populations of B. treatae developing on the two host plants. Yet, linked to host plants with different phenologies. The research pre- despite the strength of the effect we measured, our methods likely sented herein contributes to knowledge gained from a small but underestimate both the difference in the timing of leaf flush and growing number of systems, including Rhagoletis pomonella fruit gall-former emergence. We began monitoring leaf flush and adult flies (Smith 1988; Filchak et al. 2000), Goldenrod gall flies (Craig emergence of B. treatae developing on Qv an estimated one to et al. 1993) and Enchenopa treehoppers (Wood et al. 1999), that two weeks after they began. In addition, our estimates of tempo- connect differences in host plant phenology to RI between closely ral RI do not take into account that B. treatae can induce galls related populations or species of insects developing on alternate only into newly flushed leaves. Thus, a more complete measure hosts. Moreover, our study highlights a role of cascading (tem- of temporal RI in this and similar systems where insects rely on poral) RI and suggests that in certain scenarios (i.e., short-lived ephemeral host plant tissues would incorporate not only plant and insects reliant on ephemeral plant tissue) this process be may insect phenologies but also the age at which new plant tissues underappreciated. can be utilized. This caveat represents a universally overlooked We found that the phenology of a single plant trait, budbreak, nuance associated with calculating temporal RI and should be the which includes the production of a resources critical for plant re- focus of future studies. production (male flowers), and herbivore reproductive success In addition to finding substantial differences in the timing of (new leaves), is a “multitrophic multieffect trait, contributing to adult emergence between populations of B. treatae developing on RI in both plants and the specialized herbivores that feed upon different host plants, in some cases we also found differences in these plants. Our results indicate that the difference in flowering emergence phenology among populations developing on the same time and leaf flush phenology of Qg and Qv is a widespread phe- host plant species (Table 1). For example, temporal RI ranged nomenon throughout their shared range, with the phenology of from 5% to 76% between populations developing on the earliest Qg lagging two to three weeks behind that of Qv. Moreover, we and latest flushing Qg. This is interesting in light of studies of

EVOLUTION MARCH 2019 563 G. R. HOOD ET AL.

4/15 Q. virginiana Q. geminata Alva (6) ABS (1) KRE(4) DSP (3) 4/10 Okee (5) LL (2) Laboratory r = 0.937; P = 0.006 4/5

3/31 emergence

3/26

treatae B.

of 3/21 date Nature 3/16 r = 0.943; P = 0.005 Mean

3/11

3/6 3/6 3/11 3/16 3/21 3/26 3/31 4/5 4/10 4/15 Mean date of bud break

Figure 5. Correlation between mean date of B. treatae emergence from galls and in nature and mean date of leaf flush for Q. virginiana (circles) and Q. geminata (squares) across six sampling sites in southern Florida. Dashed and solid lines are superimposed to highlight the difference between datasets. Site number designations are given after site names in parentheses. local adaptation involving the asexual B. treatae in central Texas to speciation in every situation; instead, temporal differences may that develop on a closely related live oak, Q. fusiformis. Asexuals be the “ghost of reinforcement past,” evolving before or com- exhibit local adaptation to individual host plants and experience mensurate with other isolating barriers (Coyne and Orr 2004). selection at small spatial scales (Egan and Ott 2007; Egan et al. This may be particularly true for B. treatae and other herbivorous 2011). If this pattern of localized selection and adaptation also insects for which habitat isolation (host choice) and sexual isola- holds true for sexual B. treatae on Qv and Qg, then the emergence tion (mate choice) are linked, for example, when individuals use phenologies specific to individual host plants may promote in- chemical cues to find and differentiate between natal and nonnatal traspecific RI. Future studies in this and other systems that display host plant tissue in sympatry/parapatry and mate on or near that characteristics of local adaptation should test the hypothesis that same plant tissue (Jaenike 1990; Bush 1993; Berlocher and Feder temporal isolation contributes to local genetic structure among in- 2002). Regardless, the absence of B. treatae hybrids developing sect “populations” presumably adapted to individual host plants on Qv and Qg in population genetics studies (Driscoe 2018) cou- at small spatial scales. pled with hybrid formation when temporal RI is experimentally As noted by Coyne and Orr (2004), “to show that temporal removed suggests that allochrony may be important in produc- isolation currently impedes gene flow, one must show that re- ing and/or maintaining population divergence between gall wasp ducing or eliminating the barrier would lead to the production populations in sympatry or parapatry (Lamont et al. 2003). of viable and fertile hybrids.” Preliminary data from field-based The method we employed to estimate total prezygotic RI experiments where the timing of B. treatae emergence has been takes into account three components of RI that each act indepen- manipulated have revealed that hybrid offspring can be produced dently at different times during the insects’ life cycle. However, following matings of Qv males × Qg females and Qv females × the method may underestimate total RI as a result of the corre- Qg males on both host plant species, albeit with a negative fitness lated effects of these individual components. We illustrate this by consequence (personal observation; unpubl. data; G.R. Hood and examining results from Egan et al. (2012a) that found significant L. Zhang). In cases involving plants, amphibians, and insects, sexual isolation between host-associated populations of B. treatae. temporal RI has evolved by reinforcement to avoid maladaptive The average differences between heterospecific and conspecific hybridization (McNeilly 1968; Hillis 1988; Noor 1999). In these copulation frequencies in the absence of the host plant was 0.25; examples, the evolution of temporal isolation may be a postspeci- however, this difference almost doubled to 0.48 when cuttings ation phenomenon, but reinforcement is not necessarily irrelevant of the host plant were present in the mating assays. Since

564 EVOLUTION MARCH 2019 TEMPORAL ISOLATION IN GALL WASP POPULATIONS

B. treatae mate where they oviposit, host choice and mate choice by developmental plasticity triggered by differences in plant phe- are necessarily linked, similar to other herbivorous insect systems nology (Wood 1993; Wood et al. 1999). Previous studies of (Berlocher and Feder 2002). Moreover, the magnitude of host B. treatae showed that a six- to eight-week difference in the timing choice, and mate choice and the interaction of these barriers may of oviposition by the sexual generation into leaves of Q. fusiformis differ across the phenology of emergence. For example, later- did not delay the timing of emergence of asexuals from leaf galls emerging B. treatae from Qv and the earliest-emerging individu- in Texas. Instead, developmental time (the time between sexual als from Qg are most likely to overlap, and thus may experience generation oviposition and subsequent asexual adult emergence) increased selection for habitat and/or sexual isolation compared was plastic (Hood and Ott 2009). Thus, we cannot rule out the with other individuals across the emergence distributions. In sum, role of environmental factors in driving the difference between our results highlight the interdependent nature of prezygotic RI, emergence phenologies. Importantly, regardless of the basis of and suggest that more robust estimates of total RI should account trait variation, the striking correlation between host plant phenol- for the interaction between individual components of RI acting ogy and insect emergence coupled with the short adult life cycle simultaneously as opposed to multiplicatively combining barriers suggests emergence timing may facilitate population divergence. measured alone (Craig et al. 1993; Feder et al. 1994; Funk 1998; While the timing of gall-former emergence does not meet the Bolnick et al. 2009). full criteria of a “magic” trait, the trait may still have a “multiple In insects, differences in the seasonal timing of adult activ- effect,” albeit in a different manner. We have investigated RI in ity, particularly in breeding times, is often cited as a “magic” an insect that resides on sister plant species that occupy different or “multiple-effect” trait (i.e., under divergent selection and con- microhabitats. While we are unsure of the specific role of RI in tributes to assortative mating) that maintains population diver- the evolution of Qv and Qg, a difference in budbreak is likely a gence and promotes speciation (Gavrilets 2004; Smadja and But- result of divergent habitat use (Nixon and Muller 1997) that may lin 2011). Our results suggest that emergence timing of sexual have driven an adaptive (potentially sympatric) speciation event B. treatae meets the second criteria: when populations geograph- between the two sister plant species (Cavender-Bares and Pahlich ically overlap, temporal RI increases assortative mating by reduc- 2004; Cavender-Bares et al. 2009; Hipp et al. 2018). In this re- ing the probability that individuals from alternative host plants gard, the live oak–B. treatae system may represent an example encounter each other. The biology of B. treatae is such that tempo- of “sequential” or “cascading” divergence whereby the effects of ral isolation likely results in assortative mating given that females RI (differences in the timing of leaf flush) at the primary pro- are proovigenic (born with all egg matured; Hood and Ott 2011), ducer trophic level (the plant) cascade upward to similarly affect mating and oviposition occur immediately upon emergence, and herbivores (the gall wasp) at the adjacent tropic level (driving individuals are short-lived. Additionally, observations of assor- differences in the timing of adult emergence) (Stireman et al. tative mating between populations of B. treatae developing on 2006; Forbes et al. 2009; Hood et al. 2015; Bracewell et al. 2018; alternative host plants (sexual isolation) demonstrated that this Brodersen et al. 2018). Thus, temporal RI may be thought of pattern was magnified in the presence of the host plant (habitat as a multiple effect trait that transcends trophic levels. By what isolation) (Egan et al. 2012a, b). Manipulative studies will be re- mechanism may these differences arise? Abiotic conditions such quired to estimate the form and magnitude of divergent selection as soil type can affect the timing of tissue availability in plants on the timing of emergence and/or mating. (Vasek and Sauer 1971; Macnair and Gardner 1998). Soil types We do not know whether host-associated differences in the and conditions differ between Qv and Qg (Cavender-Bares et al. timing of gall-former emergence are genetically or environmen- 2004), and B. treatae developing within root galls are exposed tally controlled (by the plant and/or the insect) or represent a to these differences acting through the plant during development. genetic × environmental interaction. It is tempting to speculate We hypothesize that divergent conditions in soil moisture and nu- that RI in the gall former–live oak system is a product of adaption trient availability that contribute to differences in flowering time to divergent host plant environments and thus has a genetic basis. between the two live oaks species may also affect emergence A number of studies of plant–insect systems, including Rhagoletis of host-associated populations of B. treatae from root galls in a pomonella fruit flies and European corn borer, have revealed that similar fashion. allochrony during population divergence is linked to differences Finally, the effects of a sequential cascade on population di- in the timing of diapause life history, is genetically controlled, and vergence may not be confined to one insect across two trophic highly polygenic (Egan et al. 2015; Kozak et al. 2017; Ragland levels. At least eight species of host-specific gall wasps, each et al. 2017; Doellman et al. 2018). However, temporal RI also may restricted to forming galls on ephemeral tissues, co-occur on Qv represent plasticity. For example, allochronic isolation between and Qg (Egan et al. 2013; Hood et al. 2018). Therefore, the populations of the treehopper, Enchenopa binotata, developing live oak–gall wasp community provides the opportunity to test on different species of Viburnum is nongenetic, instead produced for parallelism in the role of host plant phenology in driving

EVOLUTION MARCH 2019 565 G. R. HOOD ET AL.

sequential temporal RI across multiple gall-former taxa. More- Cavender-Bares, J., K. Kitajima, and F. A. Bazzaz. 2004. Multiple trait as- over, the effects of RI between the host plants may cascade across sociations in relation to habitat differentiation among 17 Floridan oak further tropic levels as B. treatae developing in leaf galls on Qv and species. Ecol. Monogr. 74:635–662. Cavender-Bares, J., and A. Pahlich. 2009. Molecular, morphological, and Qg are attacked by a diverse community of 19 shared natural en- ecological niche differentiation of sympatric sister oak species, Quer- emies consisting of , hyperparasitoids, and inquilines cus virginiana and Q. geminata (Fagaceae). Am. J. Bot. 96:1690– whose phenologies are tied to that of their insect hosts (Hall 2001; 1702. Forbes et al. 2015; Busbee 2018). Thus, differences in host plant Cavender-Bares, J., A. Gonzalez-Rodriguez, D. A. R. Eaton, A. A. L. Hipp, A. Beulke, and P. S. Manos. 2015. Phylogeny and biogeography of phenology may lead to “starbursts” of adaptive radiation (Forbes the American live oak (Quercus subsection Virentes): a genomic and et al. 2009; Hood et al. 2015) in phylogenetically distinct, but eco- population genetics approach. Mol. Ecol. 24:3668–3687. logically interacting species across multiple trophic levels. This Collett, D. 2002. Modelling binary data. 2nd ed. Chapman and Hall, Boca scenario represents an intriguing possibility in gall-former com- Raton, FL. Coyne, J. A., and H. A. Orr. 2004. Speciation. Sinauer, Sunderland, MA. munities specifically, and more broadly insect herbivores reliant Craig, T. P.,J. K. Itami, W. G. Abrahamson, and J. D. Horner. 1993. Behavioral on ephemeral plant tissues. evidence for host race formation in Eurosta solidaginis: implications for sympatric speciation. Evolution 51:1552–1560. Doellman, M. D., G. J. Ragland, G. R. Hood, P. J. Meyers, S. P. Egan, T. H. AUTHOR CONTRIBUTIONS Q. Powell, P. Lazorchak, M. M. Glover, C. Tait, H. Schuler, et al. 2018. GRH, LZ, JRO, and SPE designed the study and all authors collected Genomic differentiation during speciation-with-gene-flow: comparing data. GRH analyzed the data and GRH and JRO wrote the manuscript. geographic and host-related variation in divergent life history adaptation All authors edited and approved the final version. in Rhagoletis pomonella. Genes 9:262. Dopman, E. B., P. S. Robbins, and A. Seaman. 2010. Components of re- productive isolation between North American pheromone strains of the ACKNOWLEDGMENTS European corn borer. Evolution 64:881–902. We thank the staff and volunteers at Archbold Biological Research Sta- Dres, M., and J. Mallet. 2002. Host races in plant-feeding insects and their tion, especially Mark Deyrup and Hilary Swain, for field assistance in importance in sympatric speciation. Philos. Trans. R Soc. B 357:471– Florida and Robert Busbee, Amanda Driscoe, and Chelsea Smith for plant 492. husbandry. The authors acknowledge Lake Lizzie Conservation Area in Driscoe, A. L. 2018. Host plant affiliation and spatial autocorrelation as drivers Osceola County and South Florida Water Management District at Kissim- of genetic differentiation among populations of a regionally host-specific mee River for access to experimental sites. Support was provided to GRH insect herbiovore. M.S. Thesis, Texas State University, San Marcos, TX. by the Rice Academy of Fellows, to SPE by Rice University and to JRO Egan, S. P., and J. R. Ott, J.R. 2007. Host plant quality and local adaptation from Texas State University Research Enhancement Grants. determine the distribution of a gall-forming herbivore. Ecology 88:2868– 2879. Egan, S. P., and D. J. Funk. 2009. Ecologically dependent postmating isolation DATA ARCHIVING between sympatric host forms of Neochlamisus bebbianae leaf beetles. Data will be deposited in DRYAD. The doi for our data is 10.5061/ Proc. Natl. Acad. Sci. 106:19426–19431. dryad.82j677c. Egan, S. P., G. R. Hood, and J. R. Ott. 2011. Natural selection on gall size: vari- able contributions of individual host plants to population wide patterns. Evolution 65:3543–3557. LITERATURE CITED Egan, S. P., G. R. Hood, and J. R. Ott. 2012a. Testing the role of habitat Berlocher, S. H., and J. L. Feder. 2002. Sympatric speciation in phytophagous isolation among ecologically divergent gall wasp populations. Int. J. insects: moving beyond controversy. Annu. Rev. Entomol. 47:773–815. Ecol. 2012:809897. Bernays, E. A. 1998. Evolution of feeding behavior in insect herbivores. Egan, S. P., G. R. Hood, J. L. Feder, J. L., and J. R. Ott. 2012b. Divergent Bioscience 48:35–44. host-plant use promotes reproductive isolation among cynipid gall wasp Bolnick, D. I., L. K. Snowberg, C. Patenia, W. E. Stutz, T. Ingram, and O. populations. Biol. Lett. 8:605–608. L. Lau. 2009. Phenotypic-dependent native habitat preference facilitates Egan, S. P., G. R. Hood, G. DeVela, and J. R. Ott. 2013. Parallel patterns of divergence between parapatric lake and stream stickleback. Evolution morphological and behavioral variation among host-associated popula- 63:2004–2016. tions of two gall wasp species. PLoS One 8:e54690. Bracewell, R. R., D. Vanderpool, J. M. Good, and D. L. Six. 2018. Cascad- Egan, S. P., G. J. Ragland, L. Assour, T. H. Q. Powell, G. R. Hood, S. Emrich, ing speciation among mutualists and antagonists in a tree-beetle-fungi P. Nosil, and J. L. Feder. 2015. Experimental evidence of genome-wide interaction. Proc. R Soc. B Biol. Sci. 285:20180694. impact of ecological selection during early stages of speciation-with- Busbee, R. 2018. Host plant and spatial influences on the natural enemy gene-flow. Ecol. Lett. 18:817–825. community structure of a host specific insect herbivore. M.S. Thesis. Feder, J. L., T. A. Hunt, and G. L. Bush. 1993. The effects of climate, host Texas State University, San Marcos, TX. plant phenology and host fidelity on the genetics of apple and hawthorn Bush, G. L. 1993. A reaffirmation of Santa Rosalia, or why there are so many infesting races of Rhagoletis pomonella. Entomol. Exp. Appl. 69:117– kinds of small . Pp. 229–249 in D. Edwards and D. R. Less, eds. 135. Evolutionary patterns and processes. Academic, New York. Feder, J. L., S. B. Opp, B. Wlazlo, K. Reynolds, W. Go, and S. Spisak. Brodersen, J., D. M Post, and O. Seehausen 2018. Upward adaptive radiation 1994. Host fidelity is an effective premating barrier between sym- cascades: predator diversification induced by prey diversification. Trend patric races of the apple maggot fly. Proc. Natl. Acad. Sci. 17:7990– Ecol. Evol. 22:59–70. 7994.

566 EVOLUTION MARCH 2019 TEMPORAL ISOLATION IN GALL WASP POPULATIONS

Feder, K. E., Roethele, J. B., and J. L. Feder. 2000. Natural selection and Lund, J. N., J. R. Ott, and R. Lyons. 1998. Heterogony in Belonocnema sympatric divergence in the apple maggot Rhagoletis pomonella. Nature treatae Mayr (Hymenoptera: Cynipidae). Proc. Entomol. Soc. Washing- 407:739–741. ton 100:755–763. Forbes, A. A., T. H. Q. Powell, L. L. Stelinski, J. L. Smith, and J. L. Feder. Macnair, M. R., and M. Gardner. 1998. The evolution of edaphic endemics. 2009. Sequential sympatric speciation across trophic levels. Science Pp. 157–171 in D. J. Howard and S. H. Berlocher, eds. Endless forms: 6:776–779. Species and speciation. Oxford Univ. Press, New York. Forbes, A. A., M. C. Hall, J. Lund, G. R. Hood, R. Izen, S. P. Egan, and Martin, N. H., and J. H. Willis. 2007. Ecological divergence associated with J. R. Ott. 2015. Parasitoids, hyperparasitoids, and inquilines associated mating system causes nearly complete reproductive isolation between with the sexual and asexual generations of the gall former, Belonocnema sympatric Mimulus species. Evolution 61:68–82. treatae (Hymenoptera: Cynipidae). Ann. Entomol. Soc. Am. 109:49–63. Mattsson, M., G. R. Hood, J. L. Feder, and L. A. Ruedas, L. A. 2015. Rapid and Forbes, A. A., S. N. Devin, A. C. Hippee, E. S. Tvedte, A. K. G. Ward, H. A. repeatable shifts in life history timing of Rhagoletis pomonella (Diptera: Widmayer, and C. J. Wilson. 2017. Revisiting the particular role of host Tephritidae) following colonization of novel host plants in the Pacific shifts in initiating insect speciation. Evolution 71:1126–1137. Northwestern United States. Ecol. Evol. 5:5823–5837. Funk, D. J. 1998. Isolating a role of natural selection in speciation: host Matsubayashi, K. W., and H. Katakura. 2009. Contribution of multiple iso- adaptation and sexual isolation in Neochlamisus bebianea leaf beetles. lating barriers to reproductive isolation between a pair of phytophagous Evolution 52:1744–1759. ladybird beetles. Evolution 63:2653–2680. Funk, D. J., K. E. Filchak, and J. L. Feder. 2002. Herbivorous insects: McNeilly, T. 1968. Evolution in closely adjacent plant populations. III. model systems for the comparative study of speciation ecology. Ge- Agrostis tenuis on a small copper mine. Heredity 23:99–108. netica 116:251–267. Mopper, S. 2005. Phenology—the role of time in structuring space for her- Gavrilets, S. 2004. Fitness landscapes and the origin of species. Princeton bivorous insect populations. Annales Zoologici Fennici 42:327–333. Univ. Press, Princeton, NJ. Nixon, K. C., and Muller, C. H. 1997. Quercus Linnaeus sect. Quercus white Hall, M. C. 2001. Community structure of parasitoids attacking leaf galls of oaks. Pp. 436–506 in Flora of North America North of Mexico. Oxford Belonocnema treatae on Quercus fusiformis. M.S. Thesis. Texas State Univ. Press, New York, NY. University, San Marcos, TX. Noor, M. A. F. 1999. Reinforcement and other consequences of sympatry. Hillis, D. 1988. Systematics of the Rana pipiens complex: puzzle and Heredity 83:503–508. paradigm. Ann. Rev. Ecol. Syst. 19:39–63. Nosil, P. 2012. Ecological speciation. Oxford Univ. Press, New York, NY. Hipp, A. L, P. S. Manos, A. Gonzalez-Rodriguez, M. Hahn, M. Kaproth, J. D. Paudel, B. R., M. Burd, M. Shrestha, A. G. Dyer, and Q- J. Li. 2018. Reproduc- McVay, S. V. Avalos, and J. Cavender-Bares. 2018. Sympatric parallel tive isolation in alpine gingers: how do co-existing Roscoea (R. purpurea diversification of major oak clades in the Americas and the origins of and R. tumjensis) conserve species integrity. Evolution 72:1840–1850. Mexican species diversity. New Phytol. 217:439–452. Powell, T. H. Q., A. A. Forbes, G. R. Hood, and J. L. Feder. 2014. Ecological Hood, G. R., and J. R. Ott. 2009. Developmental plasticity and reduced sus- adaptation and reproductive isolation in sympatry: genetic and pheno- ceptibility to natural enemies following host plant defoliation in a spe- typic evidence for native host races of Rhagoletis pomonella. Mol. Ecol. cialized herbivore. Oecologia 162:673–683. 23:688–704. Hood, G. R., and J. R. Ott. 2011. Generational shape shifting: changes in egg Ragland, G. J., M. D. Doellman, P. J. Meyers, G. R. Hood, S. P. Egan, T. H. Q. shape and size between sexual and asexual generations of a cyclically Powell, D. A. Hahn, P. Nosil, and J. L. Feder. 2017. A test of genomic parthenogenic gall former. Entomol. Exp. Appl. 141:88–96. modularity among life-history adaptations promoting speciation with Hood, G. R., S. P. Egan, and J. L. Feder. 2012. Evidence for sexual isolation gene flow. Mol. Ecol. 26:3926–3942. as a prezygotic barrier to gene flow between morphologically divergent Ramsey, J., H. D. Bradshaw, and D. W. Schemske. 2003. Components of species of Rhagoletis fruit flies. Ecol. Entomol. 37:219–230. reproductive isolation between monkeyflowers Mimulus lewisii and M. Hood, G. R., A. A. Forbes, T. H. Q. Powell, S. P. Egan, G. Hamerlinck, J. J. cardinalis (Phrymaceae). Evolution 57:1520–1534. Smith, and J. L. Feder. 2015. Sequential divergence and the multiplica- Richards, T. J., and D. Ortiz-Barrientos. 2016. Immigrant inviability produces tive origin of community diversity. Proc. Natl. Acad. Sci. 112:E5980– a strong barrier to gene flow between parapatric ecotypes of Senecio E5989. lautus. Evolution 70:1239–1248. Hood, G. R., L. Zhang, L. Topper, P. F. P. B- D. Pinto, G. A. D. Pino, M. S. Rosenblum, E. B., H. Rompler, T. Schoneberg, and H. E. Hoekstra. 2010. Comerford, and S. P. Egan. 2018. ‘Closing the life cycle’ of Andricus Molecular and functional basis of phenotypic convergence in white quercuslanigera (Hymenoptera: Cynipidae). Ann. Entomol. Soc. Am. lizards at White Sands. Proc. Natl. Acad. Sci. 107:2113–2117. 111:103–111. Rundle, H. D. 2002. A test of ecologically dependent postmating isolation Huber, S. K., L. F. De Leon, A. P. Hendry, E. Bermingham, and J. Podos. between sympatric sticklebacks. Evolution 56:322–329. 2007. Reproductive isolation of sympatric morphs in a population of Rundle, H. D., and P. Nosil. 2005. Ecological speciation. Ecol. Lett. 8:336– Darwin’s finches. Proc. Roy. Soc. B 274:1709–1714. 552. Jaenike, J. 1990. Host specialization in phytophagous insects. Annu. Rev. Sambatti, J. B. M., J. L. Strasburg, D. Ortiz-Barrientos, E. J. Baack, and L. Ecol. Syst. 21:243–273. H. Rieseberg. 2012. Reconciling extremely strong barriers with high Kozak, G. M., C. B. Wadsworth, S. C. Kahne, S. M. Bogdanowicz, R. G. levels of gene exchange in annual sunflowers. Evolution 66:1459– Harrison, B. S. Coates, and E. B. Dopman. 2017. A combination of 1473. sexual and ecological divergence contributes to rearrangement spread Sanchez-Buillen, R. A., M. Wullenreuther, and A. Rivera. 2012. Strong asym- during initial stages of speciation. Mol. Ecol. 26:2331–2347. metry in the relative strengths of prezygotic and postzygotic barriers Lackey, A. C. R., and J. W. Boughman. 2017. Evolution of reproductive between two damselfly sister species. Evolution 66:690–707. isolation in stickleback fish. Evolution 71:357–372. Sargent, C. S. 1918. Notes on North American trees 1. Quercus. Bot. Gazette Lamont, B. B., T. He, N. J. Enright, S. L. Krauss, and B. P. Miller. 2003. 65:423–459. Anthropongenic disturbance promotes hybridization between Banksia Schuler, H., S. P. Egan, G. R. Hood, R. W. Busbee, A. L. Driscoe, and J. R. Ott. species by altering their biology. J. Evol. Biol. 16:603–609. 2018. Diversity and distribution of Wolbachia in relation to geography,

EVOLUTION MARCH 2019 567 G. R. HOOD ET AL.

host plant affiliation and life cycle of a heterogonic gall wasp. BMC Vasek, F. C., and R. H. Sauer. 1971. Seasonal progression of flowering in Evol. Biol. 18:37. Clarkia. Ecology 52:1038–1045. Smadja, C. M., R. K. Butlin. 2011. A framework for comparing processes of Warton, D. I., and F. K. C. Hui. 2011. The arcsine is asinine: the analysis of speciation in the presence of gene flow. Mol. Ecol. 20:5123–5140. proportions in ecology. Ecology 92:3–11. Smith, D. C. 1988. Heritable divergence of Rhagoletis pomonella host races Wood, T. K. 1993. Speciation of the Enchenopa binotata complex (Insect: by seasonal asynchrony. Nature 336:66–67. Domoptera: Membracidae). Pp. 299–319 in D. R. Lees and D. Edward, Stireman, J. O. III, J. D. Nason, S. B. Heard, and J. M Seehawer. 2006. eds. Evolutionary patterns and processes. Academic Press, New York, Cascading host-associated genetic differentiation in parasitoids of phy- NY. tophagous insects. Proc. R Soc. B Biol. Sci. 273:523–530. Wood, T. K., K. J. Tilmon, A. B. Chantz, C. K. Harris, and J. Pesek. 1999. Stone, G. N., K. Schonrogge, R. J. Atkinson, D. Bellido, and J. Pujade- The role of host-plant fidelity in initiating insect race formation. Evol. Villar. 2002. The population biology of oak gall wasps (Hymenoptera: Ecol. Res. 1:317–332. Cynipidae). Annu. Rev. Entomol. 47:633–668. Zhang, L., A. Driscoe, R. Izen, C. Toussaint, J. R. Ott, and S. P. Egan. Talley, H. M., C. M. Laukaitis, and R. C. Karn. 2001. Female preference 2017. Immigrant inviability promotes reproductive isolation among host- for male saliva: implications for sexual isolation for Mus musculus sub- associated populations of the gall wasp Belonocnema treatae. Entomol. species. Evolution 55:631–634. Exp. Appl. 162:379–388. Taylor, R. S., and V. L. Friesen. 2017. The role of allochrony in speciation. Mol. Ecol. 26:3330–3342. Twomey, E., J. S. Vestergaard, and K. Summers. 2014. Reproductive isolation Associate Editor: J. Light related to mimetic divergence in the poison frog Ranitomeya imitator. Nat. Comm. 5:4749. Handling Editor: Mohamed A. F. Noor

Supporting Information Additional supporting information may be found online in the Supporting Information section at the end of the article.

Table S1. Collection sites analyzed in the study (Fig. 1). Figure S1. Box plot depicting variation in survey dates of Qv and Qg that were nonflowering (No) and flowering (Yes) spread across an 86-year period from 1925 to 2011.

568 EVOLUTION MARCH 2019