Latitudinal Variation in Herbivory: Inﬂ Uences of Climatic Drivers, Herbivore Identity and Natural Enemies

Latitudinal variation in herbivory: inﬂ uences of climatic drivers, herbivore identity and natural enemies

Xoaqu í n Moreira , Luis Abdala-Roberts , V í ctor Parra-Tabla and Kailen A. Mooney

X. Moreira ([email protected]), L. Abdala-Roberts and K. A. Mooney, Dept of Ecology and Evolutionary Biology, Univ. of California, Irvine, CA 92697, USA. XM also at: Inst. of Biology, Laboratory of Evolutive Entomology, Univ. of Neuch â tel, Rue Emile-Argand 11, CH-2000 Neuch â tel, Switzerland. – V. Parra-Tabla, Depto de Ecolog í a Tropical, Campus de Ciencias Biol ó gicas y Agropecuarias, Univ. Aut ó noma de Yucat á n, Apartado Postal 4-116, Itzimn á , 97000 M é rida, Yucat á n, M é xico.

Although a number of investigations have concluded that lower latitudes are associated with increases in herbivore abundance and plant damage, the generality of this pattern is still under debate. Multiple factors may explain the lack of consistency in latitude– herbivory relationships. For instance, latitudinal variation in herbivore pressure may be shaped entirely or not by climatic variables, or vary among herbivore guilds with diff ering life-history traits. Addition- ally, the strength of top– down eff ects from natural enemies on herbivores might also vary geographically and infl uence latitude– herbivory patterns. We carried out a fi eld study where we investigated the eff ects of latitude and climate on herbivory by a seed-eating caterpillar and leaf chewers, as well as parasitism associated to the former across 30 populations of the perennial herb Ruellia nudifl ora (Acanthaceae). Th ese populations were distributed along a 5° latitudinal gradient from northern Yucatan (Mexico) to southern Belize, representing one-third of the species’ latitudinal distribution and the entirety and one-third of the precipitation and temperature gradient of this species’ distribution (respectively). We found opposing latitudinal gradients of seed herbivory and leaf herbivory, and this diff erence appeared to be mediated by contrasting eff ects of climate on each guild. Specifi cally, univariate regressions showed that seed herbivory increased at higher latitudes and with colder temperatures, while leaf herbivory increased toward the equator and with wetter conditions. Multiple regressions including temperature, precipitation and latitude only found signifi - cant eff ects of temperature for seed herbivory and latitude for leaf herbivory. Accordingly, that latitudinal variation in seed herbivory appears to be driven predominantly by variation in temperature whereas latitudinal variation in leaf herbivory was apparently driven by other unexplored correlates of latitude. Parasitism did not exhibit variation with latitude or climatic factors. Overall, these fi ndings underscore that the factors driving latitudinal clines in herbivory might vary even among herbivore species coexisting on the same host plant.

A long-standing paradigm in ecology holds that species and Westoby 2003), a positive relationship (Salgado and interactions become increasingly strong towards the trop- Pennings 2005, Garibaldi et al. 2011), or variable rela- ics, and that this has led to increased diversifi cation rates tionships depending on the herbivore group considered and thus higher species richness in the tropics (Dobzhansky (Pennings et al. 2009). Accordingly, a recent meta-analysis 1950, Janzen 1970, Schemske et al. 2009). In the case suggests that negative latitude– herbivory relationships are of plant – herbivore interactions, a number of studies have not as common as previously thought; a meta-analysis by concluded that lower latitudes are associated with increases Moles et al. (2011) found that only 37% of the published in herbivore damage (Coley and Aide 1991, Coley and studies showed higher herbivory at lower latitudes, and the Barone 1996, Hillbrand 2004, Pennings et al . 2009, average eff ect size was not signifi cantly diff erent from zero. Schemske et al. 2009, Salazar and Marquis 2012) and plants As a result, the paradigm of increased herbivory at lower should have thus evolved higher defences (Rasmann and latitudes is under debate. Agrawal 2011, Pearse and Hipp 2012, Moreira et al. 2014, Multiple factors, many of which have not been usually Pratt et al . 2014). Accordingly, Pearse and Hipp (2012) accounted for, may explain the lack of consistency in lati- found that leaf defences in 56 oak Quercus species are higher tude – herbivory relationships. First, the strength of herbivore for species growing at lower latitudes, while Rasmann and pressure or anti-herbivore defences may depend entirely or Agrawal (2011) similarly observed that milkweed Asclepias not upon abiotic drivers (e.g. temperature, precipitation, species from lower latitudes are better defended. Nonethe- seasonality). For example, recent work has shown that large- less, there are also several studies that have found no relation- scale latitudinal variation in herbivore pressure is predomi- ship between herbivore damage (or anti-herbivore defences) nantly explained by climatic variables (Adams and Zhang and latitude (Garc í a et al. 2000, Pennings et al. 2001, Moles 2009, Pearse and Hipp 2012, Moreira et al. 2014). Second,

EV-1 most studies have measured community-level patterns of latitudinal variation in herbivory or defensive traits has been herbivory by pooling data across groups of herbivores, and documented across species and comparing temperate versus the few studies that have looked at damage by individual tropical regions (Coley and Barone 1996, Schemske et al. species or groups have focused on herbivores within the same 2009, Rasmann and Agrawal 2011). However, recent work feeding guild or with a similar diet breadth (but see Andrew has demonstrated that smaller-scale latitudinal gradients in and Hughes 2005, Pennings et al. 2009, Anstett et al. 2014). climatic variables have shaped intra-specifi c clines in plant One exception is a study by Pennings et al . (2009) that traits (e.g. life history traits and defenses, Woods et al. 2012, reported damage separately from chewing, sap-feeding, and Pratt and Mooney 2013), contributing to our understanding gall-making insect herbivores in Atlantic coast salt marshes. of how clinal variation in biotic and abiotic drivers shapes Th is study showed contrasting latitudinal patterns among microevolutionary patterns in plants. In the present study, herbivore guilds, possibly due to diff erent responses to grow- we addressed the following questions: 1) is latitude corre- ing season duration and winter conditions in relation to the lated with the intensity of seed herbivory, leaf herbivory, and number of generations per year exhibited by each herbivore parasitism associated to R. nudifl ora, and do these latitudinal group. Variation in latitudinal responses among herbivore relationships vary in strength and/or direction? 2) which cli- guilds may be thus driven by not only to diff erences in matic factors explain these latitudinal patterns and do these feeding mode or tissue specialization, but also by herbivore vary among herbivore guilds or trophic levels? And 3) do lat- life-history traits. itudinal or climatic clines in parasitism contribute to explain A third factor which may infl uence latitudinal gradients latitudinal variation in seed herbivory? By addressing these in herbivore pressure is geographic variation in the strength questions, our work builds towards an understanding of the of top – down control by natural enemies (Dyer 2007, joint infl uences of climatic forcing, herbivore identity, and Gripenberg and Roslin 2007, Cornelissen and Stiling 2009, natural enemies on latitudinal variation in plant – herbivore Abdala-Roberts et al . 2010, Mooney et al . 2010). Predator interactions. and parasitoid abundance and diversity are thought to be greater at lower latitudes and in warmer climates and thus lead to stronger suppression of herbivores (Dobzhansky Material and methods 1950, Pennings and Silliman 2005, Van Bael et al. 2008, Schemske et al. 2009, Bjö rkman et al. 2011), but few studies Study system have rigorously addressed the potential eff ects of climate and latitude on the third trophic level (Bjö rkman et al. 2011). Ruellia nudifl ora is a self-compatible perennial herb dis- One exception is a study by Stireman et al. (2005) which tributed from southern United States (Texas) to Honduras found that greater climatic variability in temperate relative (Orteg ó n-Campos et al. 2012). It grows in disturbed open to tropical regions is associated with decreases in parasitism or partially shaded areas, and under a wide range of climatic across numerous caterpillar taxa (Marczak et al. 2011). In and soil conditions (Ortegó n-Campos et al. 2012). Th is contrast, a recent meta-analysis by Mooney et al. (2010) did species has a mixed reproductive system, producing chas- not fi nd latitudinal variation in bird predation on herbivores mogamous (CH) fl owers that have an open corolla and or eff ects of birds on plant damage/growth. Direct eff ects of are visited by pollinators, and cleistogamous (CL) fl owers variation in climate on plants may also feedback to infl u- that do not open, have reduced corollas, and self-pollinate ence top– down control of herbivores through diff erences in obligatorily. Fruits are dry and dehiscent, each one normally herbivore diet breadth (Bernays and Graham 1988, Mooney bearing seven to 12 seeds that typically fall within a meter et al. 2012), herbivore susceptibility to predators (Mooney of the parent plant after the fruit opens. Th e peak of CH et al. 2012), or predator abundance (Oksanen et al. 1981). fl ower production is during July and August, and CL fl ower Unfortunately, our understanding of the eff ects of broad- production usually spans over a longer period of time (from scale climatic variation on predators and parasitoids remains May to September, Mungu í a-Rosas et al. 2012). limited, and we largely ignore if the third trophic level Both CH and CL fruits are attacked by larvae of a modulates latitudinal patterns in herbivory. specialist noctuid moth Tripudia sp. (Lepidoptera: Th e aim of this paper was to examine the eff ects of Noctuidae) that feeds on seeds prior to fruit dehiscence latitude and climate on two insect herbivore guilds and (Abdala-Roberts and Mooney 2013, 2014). Because the natural enemies of one of such guilds. We carried out a species of Tripudia moth has not yet been identifi ed, it fi eld study in which we measured herbivory by a specialist is possible that more than one species is feeding on seed-eating caterpillar and a guild of generalist leaf-chewing R. nudifl ora. If so, however, these species are equivalents eco- herbivores, as well as parasitism rates associated to the former logically because they use the same resource. Adult female across 30 populations of the perennial herb Ruellia nudifl ora moths oviposit on recently pollinated fl owers and, unless (Acanthaceae). Th ese populations were distributed along a parasitized, a single larva grows inside a developing fruit, 5 o latitudinal gradient (from 16° N to 21 °N) extending from and usually consumes all the seeds (ca 95% of seeds con- northern Yucatan (Mexico) to southern Belize (mean annual sumed per fruit on average, Abdala-Roberts and Mooney precipitation from 700 to 2900 mm, mean temperature 2013). Previous work has shown that increasing fruit output from 26 to 24 ° C, and two-fold seasonality in precipitation), at the plant level reduces the proportion of fruits attacked representing one-third of the species’ latitudinal distribu- per plant (i.e. negative density-dependent attack, Abdala- tion (14° N to 29° N) and the entirety and one-third of Roberts and Mooney 2013, 2014). Seed-eating caterpil- the precipitation and temperature gradient of this species’ lars are in turn attacked by parasitic wasps belonging to distribution (respectively) (Hijmans et al. 2005). Historically, Braconidae (four species), Ichneumonidae (one species), and

EV-2 Pteromalidae (two species), as well as one fl y species belong- of the latitudinal distribution range of R. nudifl ora (14 ° N ing to Tachinidae (Abdala-Roberts et al. 2010). Parasitoids to 29° N), and covered the entire precipitation gradient and stop or reduce seed predator consumption, thus having an one-third of the temperature gradient found throughout this indirect positive eff ect on plant fi tness through this so called species ’ distribution. Th erefore, our fi ndings are of broad “seed rescue eff ect ” (Abdala-Roberts et al. 2010). Previous importance for understanding the eff ects of climatic drivers work has shown that parasitoid attack ranges from density- on latitudinal clines in herbivory and parasitism associated to independent (i.e. the proportion of parasitized caterpillars is this plant (Hijmans et al. 2005). Accompanying the sampled unrelated to the number of fruits attacked by the herbivore, climatic gradient are southward increases in primary produc- Abdala-Roberts and Mooney 2013), to weakly positive tivity and plant diversity, and a shift in vegetation type from density-dependent (Abdala-Roberts and Mooney 2014). deciduous spiny tropical forests in the north coast of Yucatan Results from a previous study that sampled 21 to tropical wet perennial forests in southern Belize (Flores R. nudifl ora populations throughout the state of Yucatan and Espejel 1994). Such clinally-varying biotic and abiotic (Mexico) showed that there was strong population-level vari- factors are in turn predicted to correlate with a southward ation in the proportion of R. nudifl ora fruits attacked by the increase in both herbivory and parasitism (Stireman et al. seed herbivore (up to 8.5-fold, Abdala-Roberts et al. 2010). 2005, Rodr í guez-Casta ñ eda 2013). Likewise, there was even greater population variation for We sampled all 30 R. nudifl ora populations throughout the proportion of seed herbivore larvae parasitized (12-fold, a two-week period starting in early July 2013. To avoid Abdala-Roberts et al. 2010), and as a result of variation in biases in seed herbivore and parasitoid attack due to popu- parasitoid attack, there was parallel spatial variation for the lation diff erences in reproductive phenology, we exclusively indirect eff ects of parasitoids on plant fi tness (13-fold, mea- sampled populations that were at the peak of fruit produc- sured as the proportion of rescued seeds, Abdala-Roberts tion (Abdala-Roberts et al. 2010). A total of 10 plants were et al. 2010). Moreover, these large-scale geographical pat- haphazardly sampled within each population, and for each terns of seed herbivore and parasitoid attack appeared to individual we visually estimated whole-plant leaf damage by remain relatively constant across years (Supplementary mate- insect leaf chewers using the following scale: 0 ϭ undamaged, rial Appendix 1 Fig. A1), suggesting that there may be con- 1 ϭ 0 – 25% of leaf area consumed, 2 ϭ 25 – 50% of leaf area sistent population variation in the structure of interactions consumed, 3 ϭ 50 – 75% of leaf area consumed, 4 ϭ more throughout the plant ’ s distribution range in Yucatan. Such than 75% of leaf area consumed. In addition, for 28 out of a scenario sets the stage for spatially divergent evolutionary the 30 populations sampled for leaf damage, we estimated outcomes from these tri-trophic interactions (i.e. selection seed herbivory and parasitism by using 10 diff erent plants to mosaics, Rudgers and Strauss 2004). those surveyed for leaf herbivory and collecting 10 mature Ruellia nudifl ora leaves are oval-shaped, arranged in pairs, CH fruits and all mature CL fruits available (usually 3– 4 and are usually found at the base of the plant. Most leaf dam- per plant during the sampling period) per plant. Fruits were age is caused by caterpillars of several species of Lepidoptera. placed in paper bags and transported to the laboratory where Surveys of leaf consumption conducted on a monthly basis they were dissected to record seed herbivore presence (direct during 2006 and 2007 for R. nudifl ora populations located observation of larva or indirect evidence from frass) and in the north, center, and south of Yucatan showed that the parasitoid presence (direct observation of adult or larva or most important leaf chewers (in terms of abundance and leaf indirect evidence from empty pupal casing) (Abdala-Roberts area consumed) are Anartia jatrophae and Siproeta stelenes et al. 2010). Based upon these observations, we calculated the (Lepidoptera: Nymphalidae), which are considered general- proportion of fruits attacked by the seed herbivore (attacked ist species that feed on several species of Acanthaceae, Scro- fruits/total number of fruits collected per plant) and the phulariaceae and Verbenaceae (DeVries 1987, Lederhouse proportion of seed herbivore-attacked fruits with parasitoid et al. 1992). Th ese two species of caterpillars are responsible (number of attacked fruits with parasitoid/number of fruits for 10– 20% of leaf area consumed of R. nudifl ora and have with seed herbivore per plant) (Abdala-Roberts et al. 2010). been recorded in most R. nudifl ora populations sampled in We calculated these proportions by pooling data across fruit the Yucatan Peninsula (Ortegó n-Campos et al. 2009). Total types as previous work has shown that the proportion of leaf area lost to insect herbivores in R. nudifl ora usually attacked fruits per plant does not diff er between CH and CL ranges from 15 – 20% (Orteg ó n-Campos et al. 2009); there- fruits (Mungu í a-Rosas et al. 2013). fore these two caterpillar species account for most of the leaf herbivory experienced by this plant. Geographic and climatic variables

Plant sampling scheme and measurements Th e latitude and longitude of each R. nudifl ora population were acquired in situ using a global positioning system We sampled a total of 30 R. nudifl ora populations distrib- navigation device. To characterize the mean climate of each uted along a north-to-south gradient from northern Yucatan R. nudifl ora population site, we used a subset of the BioClim (Mexico) to southern Belize spanning ca 900 km and fi ve climate variables (available at: Ͻ www.worldclim.org/ Ͼ ), degrees in latitude (16° N to 21 °N, Fig. 1). From north to namely: BIO1 (annual mean temperature,° C), BIO4 (tem- south along this gradient, there is a four-fold increase in pre- perature seasonality, expressed as the standard deviation of cipitation (700 to 2900 mm per year), a 20% decrease in temperature among months ϫ 100), BIO5 (maximum coeffi cient of variation in precipitation (among months), and temperature of the warmest month,° C), BIO6 (minimum a decrease of 2° C in mean annual temperature (from 26 to temperature of the coldest month,° C), BIO12 (annual pre- 24 °C). Importantly, the sampled transect spanned one-third cipitation, mm), BIO13 (precipitation of the wettest month,

EV-3 Figure 1. Spatial layout of Ruellia nudiﬂ ora populations sampled throughout the Yucatan Peninsula (Yucatan and Quintana Roo, Mexico) and Belize. Circles represent the location of each population (n ϭ 30).

mm), BIO14 (precipitation of the driest month, mm), ance in the four temperature variables across populations BIO15 (precipitation seasonality, expressed as standard devi- (‘ PC temperature’ hereafter), and was positively related ation of precipitation across months) (Moreira et al. 2014). to annual mean temperature, temperature seasonality, Th e procedures used to calculate these variables are fully and maximum temperature of the warmest month, and described in Hijmans et al. (2005). negatively related to minimum temperature of the coldest month. Similarly, the fi rst principal component explained Statistical analyses 77% of the variance in the four precipitation variables across populations (‘ PC precipitation’ hereafter), and was To investigate whether there was population variation in the positively related to annual precipitation, precipitation of levels of seed herbivory, leaf herbivory, and parasitism associ- the wettest month, and precipitation of the driest month. ated with the seed herbivore, we performed general linear Th e standardized z -scores of each PC were then used in the models (PROC GLM in SAS 9.2) for each response vari- regression analyses to examine the relationships between able testing for an eff ect of population which was treated climate and herbivory and parasitism. as a fi xed factor (R. nudifl ora populations were intentionally Finally, we performed three multiple regressions to selected in a latitudinal gradient). simultaneously evaluate the eff ects of latitude, PC To investigate whether latitude and climate were related temperature, and PC precipitation on seed herbivory, leaf to seed herbivory (proportion of attacked fruits averaged herbivory, or parasitism. Th e aim of these regressions was across plants per population), leaf herbivory (leaf damage to test whether latitude was a signifi cant predictor of her- score averaged across plants per population), and parasit- bivory and parasitism once the eff ect of climatic factors was ism intensity associated with the seed herbivore (proportion accounted for (i.e. if climate explained latitudinal patterns), of attacked fruits with parasitoid averaged across plants per as well as determine which climatic factor(s) drive(s) varia- population), we performed simple linear regressions using tion in herbivory and parasitism once correlations among population means for each variable where latitude or climate these factors are accounted for. For all the simple and mul- were predictors of each type of herbivory or parasitism using tiple regressions, we logit-transformed seed herbivory and PROC REG in SAS 9.2. To make use of the information parasitism data, and log-transformed leaf herbivory data to from all climatic variables without infl ating type I error due achieve normality of residuals. to multiple tests, prior to running the regressions we summarized climatic variables by conducting two independent principal component analyses (PCA) using PROC FAC- Results TOR (rotation ϭ varimax) in SAS 9.2, one for temperature variables and one for precipitation variables (Moreira et al. We found population variation in the proportion of attacked ϭ Ͻ 2014). In each case, climatic variables were summarized fruits by the seed herbivore (F 27,286 11.74, p 0.001), in ϭ Ͻ with the fi rst principal component (Moreira et al. 2014). leaf herbivory (F29,220 3.01, p 0.001), and in the pro- ϭ Th e fi rst principal component explained 79% of the vari- portion of attacked fruits with parasitoids (F27,261 4.35,

EV-4 2 (a) R2 = 0.12 (c) R2 = 0.16 (e) R = 0.09 P = 0.076 P = 0.040 P = 0.133 2

0.2 Proportion of attacked fruits 0.01 16 17 18 19 20 21 22 –2 –1 0 1 2 –2 –1 0 1 2

(b) R2 = 0.46 (d) R2 = 0.03 (f) R2 = 0.27 10 P < 0.001 P = 0.402 P = 0.013

1 score Leaf damage 0.1 16 17 18 19 20 21 22 –2 –1 0 1 2 –2 –1 0 1 2 3 4 5 Latitude PC temperature PC precipitation

Figure 2. Latitude, temperature, and precipitation as predictors of the proportion of fruits attacked by a Noctuid seed herbivore (a, c, e) and leaf damage by leaf chewers (b, d, f) in Ruellia nudiﬂ ora populations sampled from northern Yucatan (Mexico) to southern Belize (n ϭ 2 8 populations for seed herbivory and n ϭ 30 for leaf herbivory). We plot original values on either a logit (for seed herbivory) or log (for leaf herbivory) scale. R2 -coeﬃ cients and p-values are from simple regressions using transformed data. Circles represent population means (n ϭ 10 plants per population).

p Ͻ 0.001). Latitude was signifi cantly negatively related R. nudifl ora populations from lower latitudes and wetter to PC precipitation (R2 ϭ 0.61, p Ͻ 0.001) but not to PC regions (higher annual precipitation, higher precipitation temperature (R2 ϭ 0.04, p ϭ 0.319), and PC temperature of the wettest month, and higher precipitation of the dri- and PC precipitation were unrelated (R 2 ϭ 0.07, p ϭ 0.168) est month based upon precipitation PCA) showed higher (Supplementary material Appendix 1 Fig. A2). scores of leaf herbivory (Fig. 2b, f), and temperature did not Th e proportion of fruits attacked by the seed herbi- explain population variation in damage by this herbivore vore ranged from 0.04 Ϯ 0.02 to 1.00 Ϯ 0.04 (overall guild (Fig. 2d). Multiple regression showed that latitude was mean ϭ 0.43 Ϯ 0.06) among Ruellia nudifl ora populations. the only signifi cant predictor of leaf herbivory and that the Overall, seed predation was 2.7-fold higher at the north- climatic correlates of latitude did not explain the latitudinal ernmost population than at the southernmost population pattern (Table 1). (0.32 Ϯ 0.07 at 21 ° N versus 0.12 Ϯ 0.04 at 17 ° N). Simple Th e proportion of attacked fruits with parasitoids ranged (univariate) regressions showed that latitude and tempera- from 0.07 Ϯ 0.06 to 1.00 Ϯ 0.03 (mean ϭ 0.84 Ϯ 0.04) ture were associated with seed herbivory across R. nudifl ora among R. nudifl ora populations. Simple regressions showed populations marginally signifi cantly (latitude: R2 ϭ 0.12, that neither latitude nor any of the climatic variables were p ϭ 0.076; Fig. 2a) and signifi cantly (PC temperature associated with parasitism intensity on the seed herbivore R2 ϭ 0.16, p ϭ 0.040; Fig. 2c), respectively. Specifi cally, we (Fig. 3a – c). Th ese results remained qualitatively unchanged found that R. nudifl ora populations from higher latitudes based upon the multiple regression analysis (Table 1). and colder regions (lower mean annual temperature, lower temperature seasonality, and lower maximum temperature of the warmest month based upon the temperature PCA) showed a higher proportion of fruits attacked by the seed Table 1. Multiple regression showing the effects of latitude and herbivore (Fig. 2a, c). Latitude did not signifi cantly predict climate (based upon a principal components analysis summarizing seed herbivory once climatic factors were accounted for in the a suite of variables associated to precipitation or temperature) on multiple regression, and PC temperature was the only signifi - seed herbivory (proportion of attacked fruits), leaf herbivory (leaf cant predictor of seed herbivory in this model (Table 1). damage score), and parasitism (proportion of attacked fruits with Ϯ Ϯ parasitoids) associated with Ruellia nudifl or a . Student ’ s t-tests ( t ) Leaf herbivory scores ranged from 0.5 0.17 to 2.1 and p-values for an analyses based upon logit-transformed (seed 0.23 (mean ϭ 1.10 Ϯ 0.08) among R. nudifl ora populations. herbivory and parasitism) and log-transformed (leaf herbivory) data Overall, leaf damage was 1.8-fold higher at the southern- are shown. Signifi cant (p Ͻ 0.05) p - values in bold. most population than at the northernmost population Seed herbivory Leaf herbivory Parasitism (1.62 Ϯ 0.26 at 21 ° N versus 0.90 Ϯ 0.18 at 16 ° N). Simple Variable t p t pt p (univariate) regressions showed that latitude (R2 ϭ 0.46, 1,26 1,27 1,22 p Ͻ 0.001; Fig. 2b) and precipitation (R2 ϭ 0.27, p ϭ 0.013; Latitude 0.37 0.711 – 2.14 0.042 0.38 0.706 Fig. 2f) signifi cantly predicted leaf herbivory across PC temperature – 2.20 0.038 0.28 0.779– 1.62 0.122 PC precipitation – 0.83 0.416 0.25 0.807 – 0.38 0.707 R. nudifl ora populations. Contrarily to seed herbivory,

EV-5 (a) R2 = 0.03 (b) R2 = 0.07 (c) R2 = 0.02 P = 0.467 P = 0.120 P = 0.496 2

0.2 fruits with parasitoids Proportion of attacked Proportion 0.01 16 17 18 19 20 21 22 –2 –1 0 1 2 –2 –1 0 1 2 Latitude PC temperature PC precipitation

Figure 3. (a) Latitude, (b) temperature and (c) precipitation as predictors of the proportion of attacked fruits with parasitoids across 28 populations of the perennial herb Ruellia nudiﬂ ora sampled from northern Yucatan (Mexico) to southern Belize. We plot original values on a logit scale on the Y axis. R2 -coeﬃ cients and p-values are from simple regressions using logit-transformed data. Circles represent population means (n ϭ 10 plants per population).

Discussion multiple regression was temperature. Specifi cally, we found that R. nudifl ora populations from colder sites showed a Latitudinal variation and climatic drivers of seed higher proportion of fruits attacked by the seed herbivore. herbivory and leaf herbivory Th is result is surprising as recent paleobiological studies of the latitudinal diversity gradient have suggested that latitu- Latitude is considered one of the main geographical pre- dinal gradients only occur when there is a strong latitudinal dictors of species abundance and diversity (Dobzhansky climate gradient (reviewed by Mannion et al. 2014). In our 1950, Janzen 1970). While many studies have supported the case, however, the range in mean annual temperature from hypothesis that herbivory increases with decreasing latitude the highest to lowest latitudinal point of seed herbivore and (Coley and Aide 1991, Coley and Barone 1996, Hillbrand R. nudifl ora distributions is relatively small, spanning only 2004, Pennings et al. 2009, Schemske et al. 2009, Salazar 5 ° C (from 22° C in southern Texas to 27 ° C in Honduras). and Marquis 2012), examples of null (Garcí a et al. 2000, Although theory predicts that both herbivory consump- Pennings et al. 2001, Moles and Westoby 2003) and positive tion rates and fi tness increase exponentially with increased (Salgado and Pennings 2005, Garibaldi et al. 2011) latitude – temperature (O ’ Connor et al. 2011), insect herbivore per- herbivory relationships are equally and in some cases more formance (e.g. food-to-biomass conversion effi ciency) may common (reviewed by Moles et al. 2011). Accordingly, in drastically decline when a species encounters temperatures this study we found contrasting latitudinal patterns for seed beyond its thermal optimum (Angilletta 2009, Bauerfeind herbivory and leaf herbivory associated with the peren- and Fischer 2013). High temperatures have been also shown nial herb Ruellia nudifl ora; plant populations from lower to increase the production of plant chemical defences and latitudes showed higher rates of leaf herbivory but a lower to reduce nitrogen concentration in plant tissues, dimin- proportion of attacked fruits by the seed herbivore. Because ishing thus plant host quality for herbivores (reviewed by our estimate of leaf herbivory was of low resolution, this DeLucia et al. 2012). Collectively, our fi ndings suggest fi nding provides a conservative test for variation among that the opposing latitudinal gradients observed for seed populations and along latitudinal gradients, and it is herbivory and leaf herbivory in R. nudifl ora is explained by possible that a more powerful analysis might have also diff erences in the identity and relative importance of abiotic detected eff ects of precipitation. (climate) drivers infl uencing each guild. One factor possibly underlying the opposing latitudinal Biogeographical and historical factors may also explain gradients in seed and leaf herbivory observed in our study, the contrasting latitudinal and climatic gradients in seed as well as inconsistent latitudinal patterns from previous herbivory and leaf herbivory. Current work suggests that studies, is that geographic variation in the strength of R. nudifl ora has undergone a recent range expansion from herbivore pressure may be shaped entirely or not by climatic southern Texas into the Gulf of Mexico and southern variables. Moreover, herbivore species or entire guilds may Mexico (Ortegó n-Campos et al. unpubl.), which coincides vary in their responses to diff erent climatic drivers. In our with increasing human-induced disturbances since the end case, multiple regression showed that latitudinal variation of the nineteenth century, and start of the twentieth century in leaf herbivory was not explained by climatic correlates (Mizrahi et al. 1997). Considering that the R. nudifl ora -seed of latitude, i.e. eff ect of latitude remained signifi cant after herbivore interaction is highly specialized (as observed for accounting for climatic drivers, which themselves did not other Tripudia caterpillars specializing on the genus Ruellia , have an eff ect (Table 1). Th is suggests that there were other J. Heywood pers. comm., Pogue 2009, as well as based upon unaccounted factors that co-vary with latitude and presum- previous work in this system, Ortegó n-Campos et al. 2009), ably drove the geographical pattern in leaf herbivory. In con- it is possible that this herbivore has tracked the southward trast, and despite sampling a small range in mean annual expansion of the distribution range of its host plant. Th ere- temperature (2° C, one-third of the temperature gradient fore, the observed increase in seed herbivory with decreas- of this species ’ distribution), the only signifi cant predictor ing temperature could refl ect a signature of adaptation of of seed herbivory after accounting for all factors in the this herbivore to its original distribution range at higher

EV-6 and colder latitudes. In contrast, the most important leaf not contribute the observed latitudinal and climatic gradi- chewers feeding on R. nudifl ora are species of Lepidoptera ents in R. nudifl ora seed herbivory throughout the sampled with a wider distribution range and adapted to conditions latitudinal gradient. Th ese results run counter to previous present in both low and high latitudes (Ortegó n-Campos investigations reporting that top – down eff ects of the third et al. 2012), with warmer temperatures and greater precipi- trophic level on insect herbivores are greater in warmer, tation at lower latitudes likely favouring greater population less seasonal climates, which are characteristic of lower sizes and increased damage by these herbivores. Addition- latitudes (Coley and Aide 1991, Novotny et al. 1999, Heil ally, it is also possible that leaf damage patterns might have 2008). Th e higher temporal stability in climatic conditions been driven by unmeasured changes in species richness or at lower latitudes have been proposed to increase predator composition of generalist leaf chewers, as reported by a and parasitoid eff ects by increasing the predictability of recent study (Salazar and Marquis 2012). host traits and population dynamics (Stireman et al. 2005). Another plausible reason for the contrasting latitudi- Additionally, tropical plants may be better defended against nal patterns of seed and leaf herbivory could be that geo- herbivores than temperate species, reducing herbivore graphical clines in plant– herbivore interactions vary with performance (Salazar and Marquis 2012), which in turn herbivore identity depending on the feeding guild, diet leads to slower herbivore growth rates and thus heightened breadth, or life-history traits (Salazar and Marquis 2012). risk of predation or parasitism (slow-growth, high-mortality Accordingly, Pennings et al. (2009) observed that plant hypothesis; reviewed by Williams 1999). While these studies damage by chewing and gall-making herbivores was greater provide good evidence for the linkage between climate and at low-latitude sites than at high-latitude sites in Atlantic predation/parasitism rates as well as latitudinal variation in Coast salt marshes, whereas damage by sap-feeding herbi- natural enemy eff ects, our results highlight the importance vores was unrelated to latitude. Th e authors speculated that of considering system-to-system variation and to account univoltine populations of chewing herbivores are less abun- for specifi c factors which could be driving the observed dant and cause less damage at higher latitudes because they patterns (Zhang and Adams 2011). For example, dietary are more heavily impacted by the short growing season and herbivore specialists are predicted to be superior to general- harsh environmental conditions during the winter present in ists at exploiting their host plants for defence or refuge from more northern sites (Pennings et al. 2009). In contrast, sap- enemies, leading to weaker top– down eff ects of the latter feeding herbivores such as aphids have multiple generations (Bernays and Graham 1988; Singer et al. 2014). Accord- per year, and their populations can therefore recover more ingly, further work is needed to determine which herbivore easily from winter losses (Pennings et al. 2009). Additionally, and natural enemy traits underlie geographical variation in with respect to herbivore diet breadth, it has been predicted tri-trophic interactions as these may cause departures from that specialist and generalist herbivores are diff erentially the expected patterns. adapted to plant defences (Leimu et al. 2005, Mathur et al. 2011, Ali and Agrawal 2012). While specialist herbivores Conclusions are better-adapted to qualitative defences such as alkaloids, generalist herbivores are better-adapted to quantitative By sampling a substantial portion of the latitudinal gradient defences such as phenolic compounds (Leimu et al. 2005, and most of the climatic variation of the distribution range Mathur et al. 2011, Ali and Agrawal 2012). Because quan- of the perennial herb R. nudifl ora, we show that latitudinal titative and qualitative defences may shape variation in the clines in plant – herbivore interactions are contingent upon strength or direction of herbivory– latitude relationships climatic eff ects and herbivore identity. Th ese fi ndings call (Marquis et al. 2012, Pearse and Hipp 2012), and at the for further work examining latitudinal patterns of herbivory same time may exhibit diff erent relationships with lati- across multiple herbivore species varying in traits such as tude, it is thus reasonable to expect that damage by her- feeding guild and dietary specialization in order to gain a bivores with contrasting diet breadths might vary in their predictive understanding of how herbivore traits interact relationship with latitude. Accordingly, further research is with climate in predicting geographical gradients in her- needed to determine whether qualitative (i.e. alkaloids) and bivory. Finally, contrarily to evidence gathered thus far, our quantitative (i.e. phenolics) defences in seeds and leaves of work showed that climate and latitude do not have eff ects R. nudifl ora populations vary with the latitude and explain on parasitism intensity, in turn suggesting that natural ene- the observed latitudinal patterns for seed herbivory (special- mies do not have a predominant role in shaping geographi- ist) and leaf herbivory (generalists). cal gradients of seed herbivory associated with R. nudifl ora . Nonetheless, further work examining latitudinal variation in Geographic variation in parasitoid attack tri-trophic interactions is much needed. Overall, this study lays a foundation towards understanding the simultaneous Parasitism intensity associated with the seed herbivore infl uence of climatic forcing and natural enemies on was extremely high with 84% of the attacked fruits hav- geographic variation in plant – herbivore interactions. ing parasitoids, on average, and 24 out of 28 populations exhibiting greater than 70% parasitism. However, despite observing a large amount variation in the strength of top– Acknowledgements – Author contributions: conceived and designed the experiments: XM, LAR and KAM. Performed the experiments: down eff ects by parasitoids, and contrary to our expectations, XM and LAR. Contributed reagents/materials/analysis tools: XM, our results showed that parasitism intensity was unrelated LAR, KAM and VPT. Analysed the data: XM. Wrote the paper: to latitude or climatic conditions. Accordingly, this result XM wrote the fi rst draft of the manuscript, and LAR, KAM suggests that geographic variation in parasitism rates does and VPT contributed substantially to revisions. We thank Nicol á s

EV-7 Salinas-Peba, Luis Salinas-Peba and Jorge Omar L ó pez-Mart í nez Flores, J. S. and Espejel, I. 1994. Tipos de vegetaci ó n de la Pen í n- for their technical assistance during fi eld sampling. Nicol á s Salinas- sula de Yucat á n. – Etnofl ora Yucatanense Fasc í culo 3: UADY, Peba also kindly assisted with fruit dissections and estimation of Yucat á n, M é xico. seed herbivory and parasitism. We also thank William K. Petry for Garc í a, D. et al. 2000. Geographical variation in seed production, his advices about climate data compilation. Comments and sugges- predation and abortion in Juniperus communis throughout its tions by Angela Moles helped to improve the manuscript. Th is range in Europe. – J. Ecol. 88: 436 – 446. research was supported by a UC MEXUS-CONACyT collabora- Garibaldi, L. A. et al. 2011. Latitudinal decrease in folivory within tive grant to all the authors (UCM-55592) and a UC MEXUS Nothofagus pumilio forests: dual eff ect of climate on insect den- grant for UC postdocs to X.M. (UCM-101443). XM received sity and leaf traits? – Global Ecol. Biogeogr. 20: 609 – 619. fi nancial support from Postdoctoral Fulbright/Spanish Ministry Gripenberg, S. and Roslin, T. 2007. Up or down in space? Uniting of Education grant program, and LAR was funded by a GAANN the bottom – up versus top – down paradigm and spatial ecology. fellowship and a UC MEXUS-CONACyT scholarship. – Oikos 116: 181 – 188. Heil, M. 2008. Indirect defence via tritrophic interactions. – New Phytol. 178: 41 – 61. References Hijmans, R. J. et al. 2005. Very high resolution interpolated climate surfaces for global land areas. – Int. J. Climatol. 25: Abdala-Roberts, L. and Mooney, K. A. 2013. Environmental and 1965 – 1978. plant genetic eff ects on tri-trophic interactions. – Oikos 122: Hillbrand, H. 2004. On the generality of the latitudinal diversity 1157 – 1166. gradient. – Am. Nat. 163: 192 – 211. Abdala-Roberts, L. and Mooney, K. A. 2014. Ecological and Janzen, D. H. 1970. Herbivores and the number of tree species in evolutionary consequences of plant genotype diversity in a tropical forests. – Am. Nat. 104: 501 – 528. tri-trophic system. – Ecology 95: 2879 – 2893. Lederhouse, R. C. et al. 1992. Host plant-based territoriality in Abdala-Roberts, L. et al. 2010. Spatial variation in the strength of the white peacock butterfl y, Anartia jatrophae (Lepidoptera: a trophic cascade involving Ruellia nudifl ora (Acanthaceae), an Nymphalidae). – J. Insect Behav. 5: 721 – 728. insect seed predator and associated parasitoid fauna in Mexico. Leimu, R. et al. 2005. Food preference and performance of the – Biotropica 42: 180 – 187. larvae of a specialist herbivore: variation among and within Adams, J. M. and Zhang, Y. 2009. Is there more insect folivory in host-plant populations. – Acta Oecol. 28: 325 – 330. warmer temperate climates? A latitudinal comparison of insect Mannion, P. D. et al. 2014. Th e latitudinal biodiversity gradient folivory in eastern North America. – J. Ecol. 97: 933 – 940. through deep time. – Trends Ecol. Evol. 29: 42– 50. Ali, J. G. and Agrawal, A. A. 2012. Specialist versus generalist Marczak, L. B. et al. 2011. Latitudinal variation in top – down and insect herbivores and plant defense. – Trends Plant Sci. 17: bottom – up control of a salt marsh food web – Ecology 92: 293 – 302. 276 – 281. Andrew, N. R. and Hughes, L. 2005. Herbivore damage along a Marquis, R. J. et al. 2012. Testing the low latitude/high defense latitudinal gradient: relative impacts of diff erent feeding guilds. hypothesis for broad-leaved tree species. – Oecologia 169: – Oikos 108: 176 – 182. 811 – 820. Angilletta, M. J. 2009. Th ermal adaptation: a theoretical and Mathur, V. et al. 2011. Temporal dynamics of herbivore-induced empirical synthesis. – Oxford Univ. Press. responses in Brassica juncea and their eff ect on generalist and Anstett, D. N. et al. 2014. Latitudinal gradients in herbivory on specialist herbivores. – Entomol. Exp. Appl. 139: 215 – 225. Oenothera biennis vary according to herbivore guild and spe- Mizrahi, A. et al. 1997. Composition, structure and management cialization. – Ecology 95: 2915 – 2923. potencial of secondary dry tropical vegetation in two aban- Bauerfeind, S. S. and Fischer, K. 2013. Increased temperature doned henequé n plantations of Yucatan, Mexico. – For. Ecol. reduces herbivore host-plant quality. – Global Change Biol. Manage. 96: 273 – 282. 19: 3272 – 3282. Moles, A. T. and Westoby, M. 2003. Latitude, seed predation and Bernays, E. and Graham, M. 1988. On the evolution of host spe- seed mass. – J. Biogeogr. 30: 105 – 128. cifi city in phytophagous arthropods. – Ecology 69: 886 – 892. Moles, A. T. et al. 2011. Assesing the evidence for latitudinal gra- Bj ö rkman, C. et al. 2011. Causes behind insect folivory patterns dients in plant defence and herbivory. – Funct. Ecol. 25: in latitudinal gradients. – J. Ecol. 99: 367 – 369. 380 – 388. Coley, P. D. and Aide, T. M. 1991. Comparison of herbivory and Mooney, K. A. et al. 2010. Interactions among predators and plant defenses in temperate and tropical broad-leaved forests. the cascading eff ects of vertebrate insectivores on arthropod – In: Price, P. W. et al. (eds), Plant– animal interactions: communities and plants. – Proc. Natl Acad. Sci. USA 107: evolutionary ecology in tropical and temperate regions. Wiley, 7335 – 7340. pp. 25 –49. Mooney, K. A. et al. 2012. Th e tri-trophic interactions hypothesis: Coley, P. D. and Barone, J. A. 1996. Herbivory and plant defenses interactive eff ects of host plant quality, diet breadth and natu- in tropical forests. – Annu. Rev. Ecol. Syst. 27: 305 – 335. ral enemies on herbivores. – PLoS ONE 7: e34403. Cornelissen, T. and Stiling, P. 2009. Spatial bottom – up, and top – Moreira, X. et al. 2014. Tradeoff s between constitutive and induced down efcts on the abundance of a leaf miner. – Ecography 32: defences drive geographical and climatic clines in pine chemi- 459 – 467. cal defences. – Ecol. Lett. 17: 537 – 546. DeLucia, E. H. et al. 2012. Climate change: resetting plant– insect Mungu í a-Rosas, M. A. et al. 2012. Environmental control of interactions. – Plant Physiol. 160: 1677 – 1685. reproductive phenology and the eff ect of pollen supplementa- DeVries, P. J. 1987. Th e butterfl ies of Costa Rica and their natural tion on resource allocation in the cleistogamus weed Ruellia history: Papilionidae, Pieridae, Nymphalidae. – Princeton nudifl ora (Acanthaceae). – Ann. Bot. 109: 343 – 350. Univ. Press. Mungu í a-Rosas, M. A. et al. 2013. Eff ects of the environment, Dobzhansky, T. 1950. Evolution in the tropics. – Am. Sci. 38: pollen load and fruit heteromorphism on predispersal seed 209 – 221. predation and parasitoid incidence in the cleistogamous Dyer, L. A. 2007. Tropical tritrophic interactions: nasty hosts Ruellia nudifl ora . – Oecologia 173: 871 – 880. and ubiquitous cascades. – In: Carson, W. P. and Schnitzer, Novotny, V. et al. 1999. Predation risk for herbivorous insects S. A. (eds), Tropical forest community ecology. Blackwell, on tropical vegetation: a search for enemy-free space and time. pp. 275 – 293. – Aust. J. Ecol. 24: 477 – 483.

EV-8 O ’ Connor, M. I. et al. 2011. Th eoretical predictions for how tem- and induction following herbivory. – Ecol. Lett. 14: perature aff ects the dynamics of interacting herbivores and 476– 483 plants. – Am. Nat. 178: 626 – 638. Rodr í guez-Casta ñ eda, G. 2013. Th e world and its shades of green: Oksanen, L. et al. 1981. Exploitation ecosystems in gradients of a meta-analysis of trophic cascades across temperature and pre- primary productivity. – Am. Nat. 118: 240 – 261. cipitation gradients. – Global Ecol. Biogeogr. 22: 118– 130. Orteg ó n-Campos, I. et al. 2009. Local adaptation of Ruellia Rudgers, J. A. and Strauss, S. Y. 2004. A selection mosaic in the nudifl ora (Acanthaceae) to biotic counterparts: complex facultative mutualism between ants and wild cotton. – Proc. scenario revealed when two herbivore guilds are considered. R. Soc. B 271: 2481 – 2488. – J. Evol. Biol. 22: 2288 – 2297. Salazar, D. and Marquis, R. J. 2012. Herbivore pressure Orteg ó n-Campos, I. et al. 2012. Infl uence of multiple factors on increases toward the equator. – Proc. Natl Acad. Sci. USA 109: plant local adaptation: soil type and folivore eff ects in Ruellia 12616 – 12620. nudifl ora (Acanthaceae). – Evol. Ecol. 26: 545 – 558. Salgado, C. S. and Pennings, S. C. 2005. Latitudinal variation in Pearse, I. S. and Hipp, A. L. 2012. Global patterns of leaf defenses palatability of salt-marsh plants: are diff erences constitutive? in oak species. – Evolution 66: 2272 – 2286. – Ecology 86: 1571 – 1579. Pennings, S. C. and Silliman, B. R. 2005. Linking biogeography Schemske, D. W. et al. 2009. Is there a latitudinal gradient in the and community ecology: latitudinal variation in plant– herbiv- importance of biotic interactions? – Annu. Rev. Ecol. Evol. ore interaction strength. – Ecology 86: 2310 – 2319. Syst. 40: 245 – 269. Pennings, S. C. et al. 2001. Latitudinal diff erences in plant palatabil- Singer, M. S. et al. 2014. Herbivore diet breadth mediates the ity in Atlantic coast salt marshes. – Ecology 82: 1344– 1359. cascading eff ects of carnivores in food webs. – Proc. Natl Acad. Pennings, S. C. et al. 2009. Latitudinal variation in herbivore pres- Sci. USA 111: 9521 – 9526. sure in Atlantic coast salt marshes. – Ecology 90: 183 – 195. Stireman, J. O. et al. 2005. Climate unpredictablity and parasitism Pogue, M. G. 2009. A review of the Tripudia quadrifera (Zeller) of caterpillars: implications of global warming. – Proc. Natl (Lepidoptera: Noctuidae) species complex. – Proc. Entomol. Acad. Sci. USA 29: 17384 – 17387. Soc. Wash. 111: 68 – 97. Van Bael, S. A. et al. 2008. Birds as predators in tropical agrofor- Pratt, J. D. et al. 2014. Genetically-based latitudinal variation in Artem- estry systems. – Ecology 89: 928 – 934. isia californica secondary chemistry. – Oikos 123: 953 – 963. Williams, I. S. 1999. Slow-growth, high-mortality: a general Pratt, J. D. and Mooney, K. A. 2013. Clinal adaptation and adap- hypothesis, or is it? – Ecol. Entomol. 24: 490 – 495. tive plasticity in Artemisia californica : implications for the Woods, E. C. et al. 2012. Adaptive geographical clines in the response of a foundation species to predicted climate change. growth and defense of a native plant. – Ecol. Monogr. 82: – Global Change Biol. 19: 2454 – 2466. 149 – 168. Rasmann, S. and Agrawal, A. A. 2011. Latitudinal patterns Zhang, Y. and Adams, J. 2011. Top – down control of herbivores in plant defense: evolution of cardenolides, their toxicity, varies with ecosystem types. – J. Ecol. 99: 370 – 372.

Supplementary material (available online as Appendix oik.02040 at Ͻ www.oikosjournal.org/readers/appendixϾ ). Appendix 1.

EV-9