On the factors that promote the diversity of herbivorous and in tropical forests

Judith X. Becerra1

Department of Biosphere 2, University of Arizona, Tucson, AZ 85721

Edited by Paul R. Ehrlich, Stanford University, Stanford, CA, and approved February 27, 2015 (received for review November 13, 2014)

Some of the most fascinating and challenging questions in relationship and potential feedback loop between divergent are why biodiversity is highest in tropical forests and whether the chemical defenses in a and speciali- factors involved are unique to these habitats. I did a worldwide test zation that could have effects on insect and plant diversities (12, of the hypotheses that plant community divergence in antiherbi- 13). In places with mostly highly specialized herbivore–plant as- vore traits results in higher insect herbivore diversity, and that sociations, plant communities tended to have a strong pattern of predominant attack by specialized promotes plant chemical disparity. This result suggests that disparate defensive richness. I found strong correlative support for both ideas. Butterfly systems in a plant community might promote insect specialization diversity was greatest in regions where the community average because it is difficult to attack other plants that have very different species-pairwise dissimilarity in antiherbivore traits among plant antiherbivore traits relative to those of the current host (14). At species was highest. There was also a strong positive relationship the same time, the tendency of insects to specialize on plants between specialized (insect) vs. generalized () herbivores species sharing the same defenses would favor plant communities and plant richness. Regions where herbivory impact by with species that have divergent defenses. Similar patterns of was higher than that of insects tended to have lower plant nonrandom defense overdispersion have been reported for other diversities. In contrast, regions in which insects are the main plant genera as well (15, 16). consumers, particularly in the Central and South American tropics, The existence of a feedback between herbivore specialization had the highest plant richness. Latitude did not explain any and plant defense community structure is important because it residual variance in insect or plant richness. The strong connec- tions found between insect specialization, plant defense diver- could potentially reduce niche overlap for both plants and in- gence, and plant and insect diversities suggest that increasing our sects and promote their diversities: specialized herbivores can understanding of the ecology of biological communities can aid in attack conspecific plants or species with similar morphological/ considerations of how to preserve biodiversity in the future. chemical antiherbivore traits keeping plants rare (4, 5, 14). Consequently, communities will tend to assemble or evolve plant diversity | insect diversity | tropical forests | latitudinal diversity plant species with disparate defensive systems. In turn, defensive gradient | insect specialization disparity would make it difficult for insect species to use addi- tional hosts, thus reinforcing insect specialization (13). This nderstanding the evolutionary and ecological processes that feedback mechanism has not been previously investigated at a Umaintain diversity in tropical forests has been one of the global scale but its predictions can be tested. If plant diversity is central goals in biology for many years. Plants and insect herbi- driven by herbivore specialization, we expect communities to ex- vores comprise the leading percentage of species in these eco- hibit a positive relationship between plant species richness and systems, making it of particular importance to know what is driving predominance of attack by specialized herbivores. Additionally, if their diversity (1, 2). Plant and plant-feeding insects also constitute insect diversity is determined by divergent plant defensive traits in an abundant resource base for a variety of consumers, including a community, we expect to find a positive correlation between the , mammals, reptiles, , fish, other insects, and ar- divergence of antiherbivore traits in a plant community and the thropods, and can thus in turn impact their consumer’s diversity. richness of its insect herbivore species. Here I present tests of Increased niche differentiation for both insects and plants these two predictions. could potentially promote coexistence of large numbers of species, and thus many ideas are directed at identifying the potential fac- Significance tors that can lead to a decrease of niche overlap in these organisms (3). A prominent framework for plant diversity is niche differen- A central mission in ecology has been to explain differences in tiation by natural enemy partitioning. Several variants of this mechanism have been published, like the Janzen-Connell hy- biodiversity across the world and across groups of organisms. pothesis, and the Chesson and Kuang model: they concur in This knowledge is not only fundamental as a scientific objec- predicting that natural enemies, when specialized, can keep spe- tive but also for practical issues of major concern to human- cies rare and promote plant coexistence and diversity (4–6). kind. Plants and phytophagous insects comprise the highest Similarly, higher host specificity in tropical insects, and therefore macroscopic diversity on Earth, making it particularly impor- increased niche differentiation through resource partitioning, has tant to know the mechanisms that allow their species to co- often been associated with insect diversity (7, 8). Although these exist. This study suggests that without insect specialization the mechanisms have considerable evidence to support them, there is maintenance of tropical plant diversity might not be possible. still controversy because, among other things, the Janzen–Connell Also, without high divergence of defenses among plants, insect mechanism for high tropical richness appears to operate not only tropical diversity may not persist either. in tropical forests, but also in temperate habitats where plant di- versity is lower (9). Similarly, many temperate insects may be as Author contributions: J.X.B. designed research, performed research, analyzed data, and specialized as ones living in the tropics (10). Another challenge is wrote the paper. that, although it is evident that partitioning resources might lead The author declares no conflict of interest. to narrower insect niches, it is still unclear what drives insect This article is a PNAS Direct Submission. specialization (11). 1Email: [email protected]. Recently, herbivory studies performed in the dry tropical for- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. ests of Mexico with the plant Bursera found a positive 1073/pnas.1418643112/-/DCSupplemental.

6098–6103 | PNAS | May 12, 2015 | vol. 112 | no. 19 www.pnas.org/cgi/doi/10.1073/pnas.1418643112 Downloaded by guest on October 2, 2021 Results defenses (spines and hairs). Additionally, although they produce Relating Butterfly Diversity to Community Divergence of Plant many chemical compounds, these compounds tended to belong Antiherbivore Traits. To determine whether community di- to only a few classes of chemical defense compounds. As a con- vergence in antiherbivore traits is driving insect diversity, I mea- sequence, the values of the indices of defense evenness were the sured the average plant species-pairwise dissimilarity in chemical, lowest in these areas (Table S3). biotic (-attracting traits), and mechanical/morphological de- Butterfly diversity was greatest in regions such as Costa Rica, fenses (e.g., hairs, spines). This involved comparing each plant PNG, and Ecuador, where the community average dissimilarity in species to each other plant species, then averaging over all pair- antiherbivore traits among plant species was highest. Here, some wise dissimilarities. I did this for seven areas of the world, then mechanical/morphological plant defenses, such as hairs, glandular related this average pairwise dissimilarity to the richness of insect trichomes, and laticifers, occurred frequently, but plants were also herbivores that feed on those plants in those areas. I selected often armed with ant-attracting traits, such as extrafloral nectaries, butterflies as the herbivores to examine because their diversities food bodies, or domatia. Additionally, the plant taxa produced a and hosts are well documented in these seven regions. These wide array of chemical compounds, as illustrated by their high trait places have similar physical dimensions (mean = 54,009 km2, evenness index. Some classes of compounds that are infrequent in other regions studied (e.g., alkaloids, proteins, peptides, iso- SD = 2,089) and are: (i) the joint areas of Burundi and Rwanda; flavones, and triterpenoids) are well represented in these tropical (ii)CostaRica;(iii) Morobe and Northern provinces in Papua New areas. All this resulted in communities where species pairs have Guinea (PNG); (iv) Israel, Mount Hermon, and a portion in the highly disparate plant chemistries. Arizona had intermediate levels East of the Sinai Peninsula (Middle East); (v) the provinces of vi of community chemical dissimilarity and it also had intermediate Napo, Orellana, and Sucumbíos in Ecuador (Ecuador); ( )the values of butterfly diversity. southern half of Louisiana in the United States (Louisiana); and vii ( ) the South East corner of Arizona in the United States Testing for the Influence of the Number of Host Plants Studied on the (Arizona). After compiling a list of butterflies and their known Degree of Community Average Antiherbivore Trait Divergence. It is hosts for each area, each plant host species was researched for the possible that the number of host plants in a community influences known presence of biotic, morphological/mechanical, and chem- the average divergence of antiherbivore traits; for example, larger ical defense traits (60 different kinds of compounds belonging communities might exhibit higher or lower average divergence. To to seven classes, such as phenolics, alkaloids and terpenoids) ensure that this was not confounding my results, I correlated the (Table S1). average divergence of antiherbivore traits with the number of plant I correlated the presence/absence (coded as 1 vs. 0) of each species. I also did randomization tests to determine whether anti- antiherbivore trait for each species pair combination for a re- herbivore traits in these communities were more or less over- gion. From this, I calculated the average degree of dissimilarity dispersed than communities constructed by randomly choosing the (= 1 − Pearson r) from all possible pairs of host plant species for J′ same number of plant hosts from the full list of hosts analyzed for each of the seven areas studied. An evenness index ( ; ref. 17) chemistry. However, average community divergence of plant de- was also calculated for each region to measure how evenly the fensive traits was not correlated with the number of plant species species were spread among defenses (used using the proportion analyzed (P = 0.28) (Table S2). Randomization tests confirmed this of plant species with each trait). pattern, showing that real communities had defense dissimilarity There was a strong positive correlation between average com- that was statistically different from that of random communities of munity divergence in host plant antiherbivore traits and butterfly the same size. Specifically, in places where butterfly richness was R2 = P = species richness ( 0.91 0.0008) (Fig. 1 and Table S2). lower (Middle East, Rwanda-Burundi, and Louisiana), communi- Latitude had limited impact on these butterfly-richness patterns ties had significantly lower divergence of defensive traits than would R2 = P = P = ( 0.45, 0.10) and its effects became less significant ( be expected for a random community with the same number of 0.47) when the independent effects of divergence in antiherbivore plant hosts (P < 0.0001 for each area) (Fig. 2). Arizona, was less and latitude are determined with partial regression. Divergence of divergent than expected but only marginally significant (P = 0.06). defensive traits, however, remained highly significant when the Ecuador, Costa Rica, and PNG were more overdispersed than effect caused by the latitude was statistically removed (P = 0.007). random but the differences from random communities are not Of the sites tested, the Middle East, Rwanda-Burundi, and Loui- significant. These results clearly indicate that the particular con- siana had the smallest numbers of butterfly species, as well as the figurations of defensive traits in these communities are not de- lowest mean dissimilarities of host-plant defenses. Host plants from termined by the number of plant species present. these areas tended to be armed with the same kind of morphological Relating Herbivore Resource Partitioning to Plant Diversity in Communities Around the World. I compiled results of 52 pub- lished herbivory studies and related them to the diversity of plant species estimated for the area where each of these studies ECOLOGY took place (Table S4). These studies were performed in different locations around the world and were selected because they in- cluded data on the amount of biomass lost to herbivory as well as describing the herbivore agents responsible for the attack. To determine if herbivore specialization in communities promotes plant diversity, I took advantage of the widely accepted dichotomy in specialization between the two major groups of plant herbi- vores: mammals and insects. Insects and mammals are very dif- ferent morphologically and also distantly related, but their ecological similarity has been recognized (18–22). In ecological terms, mammalian herbivores and insect herbivores relate to plants in the same way: they consume plants and decrease their fitness. However, in terms of the number of plant species they feed on, most phytophagous insects are relatively specific, feeding on Fig. 1. Butterfly species richness is positively correlated with average com- only one or a few genera or on a single plant family (11, 23–26). It munity dissimilarity in antiherbivore traits (R2 = 0.91, 0.0008). Average defense has been estimated that less than 10% of all phytophagous insects dissimilarity was highest in Costa Rica, PNG, and Ecuador. The latitudinal effect feed on plants from more than three different plant families (26). on butterfly diversity was nonsignificant. Furthermore, some groups of insect herbivores (some lepidopterans,

Becerra PNAS | May 12, 2015 | vol. 112 | no. 19 | 6099 Downloaded by guest on October 2, 2021 forests are more homogeneous in secondary chemistry and other physical traits, with many plant species coinciding in their de- fensive repertoires, and this is independent of the number of plant species living there. In comparison, plants in tropical forests tend to have more disparate defensive traits, making many plant species highly unique in their communities in terms of chemistry and other defenses, a characteristic that might potentially encourage herbi- vore specialization (8). The impact of community plant traits on insect diversity has increasingly been documented. Dinnage et al. (33), for example, found that more phylogenetically diverse plant communities (a potential proxy for communities with more diverse defensive and nondefensive traits) had a strong positive influence on the species richness of herbivorous insects in Minnesota. Similarly, Pellissier Fig. 2. The particular configuration of antiherbivore traits in a community et al. (34) provided evidence that plant phylogenetic diversity = is not correlated with the number of plant species (P 0.34). The dashed line correlated with diversity of lepidopteran assemblages in the Swiss indicates the levels of divergence in plant defense of communities con- Western Alps. As more plant communities are studied in the future structed by randomly choosing the same number of plant hosts from the full it will be interesting to see whether dissimilarity of nonresistance list of hosts analyzed for chemistry. Stars indicate communities that are significantly less divergent than expected (P < 0.0001; Middle East, Rwanda- plant traits can also contribute to increased herbivore specialization. Burundi, and Louisiana) and the filled circle represents the community of Because of their intense defense-mediated ecological interactions, Arizona, which is less divergent than expected but only marginally signifi- it is not surprising that plants and herbivores can have reciprocal cant (P = 0.06). Ecuador, Costa Rica, and PNG (rectangles) are more over- effects on each other’s diversities. Herbivory is a considerable source dispersed than random but the differences are not significant. of fitness loss for plants. It has been estimated that typical levels of herbivory often result in roughly 50% decrease in plant growth and reproduction, more if it is inflicted on seedlings or flowers (35–39). , aphids, and leaf miners,) are dominated (between 40% and This aspect can impose selective pressures on plants to decrease or 75%) by monophages [feeding in only one or two congeneric species prevent herbivory (40, 41). Because herbivores tend to attack hosts (8, 24)]. In contrast, dietary specialization is considered the excep- with similar defensive traits (12), it might be adaptive for plants in a tion rather than the rule in mammalian herbivores. Most mammals community to have dissimilar defensive strategies. maintain a broad diet often comprising dozens to hundreds of often This analysis also found significant support for the idea that unrelated plant species (23, 27, 28). Monophagy is absent, and even the higher the herbivory impact of specialists compared with the koala (Phascolarctos cinereus), one of the most specialized generalists in a community, the greater the plant diversity in the mammalian herbivores, consumes 120 plant species in the genus community. This outcome concurs with previous hypotheses and Eucalyptus and 30 noneucalypts from three plant families (29, 30). research supporting the view that specialized herbivores may Thus, comparing herbivory of mammals and insects provides one promote plant coexistence and diversity (4–6). It also agrees with way to measure the consequences of generalization vs. specializa- other studies that indicate that this mechanism is not unique to tion. Furthermore, although other consume plants, mam- the tropics, but occurs across different latitudes (9). mals and insects are by far the herbivore groups that impact plants The latitudinal diversity gradient is one of the most striking the most (31). I anticipated that generalized (mammal) herbivory biogeographic patterns on Earth and the claim has been made that should not favor plant diversity as much as specialized (insect) species interactions, including herbivory, are stronger in the tropics herbivory, because with generalists there is reduced enemy parti- and that this promotes species coexistence and in turn leads to an tioning in plants and defense niche overlap should increase (6, 32). increment in species diversity. Following this reasoning, a number of Results showed a strong positive relationship between relative studies have examined latitudinal patterns in herbivory as well as damage by specialized (insect) vs. generalized (mammal) herbi- plant defenses, looking for evidence of greater investment in de- vores and plant species richness (Box-Cox transformed species 2 fenses in tropical plants and higher herbivore pressure in the tropics. richness; R = 0.61, P < 0.0001) (Fig. 3 and Table S5). Regions However, their results have been contradictory. Some have found where damage by mammals relative to insects is higher tended to evidence of larger amounts of tropical plant tissues lost to herbivory have lower plant diversities. In comparison, regions in which in- or plants better defended at lower latitudes (42), but others, par- sects are the main consumers of plants, particularly in the Central ticularly the ones using more extensive datasets and meta-analysis, and South American tropics, tended to support the highest plant richness. There was also a negative relationship between latitude of the study sites and plant species richness (R2 = 0.41, P < 0.0001). However, when relative damage by specialized vs. gen- eralized herbivores and latitude are both included in the model, latitude does not explain any additional variance in species rich- ness beyond that explained by relative damage by specialized vs. generalized herbivores (P = 0.24). Nonetheless, relative damage by specialized vs. generalized herbivores does have additional explanatory power when added to a model already containing latitude (P < 0.0001). Neither the relationship between total herbivory and plant richness nor the one between latitude and total herbivory were significant (P < 0.49 and P < 0.17). Discussion Results from this study support the hypothesis that the more overdispersed a plant community is in terms of antiherbivore traits, Fig. 3. There was a positive relationship between plant species richness and the higher the diversity of phytophagous insects (8, 13). Latitude relative damage by specialized (insect) vs. generalized (mammal) herbivores did not explain any residual variance in insect species richness. across 42 sites around the world (R2 = 0.60, P < 0.0001). Latitudinal effects on Furthermore, my results suggest that the particular configuration plant species richness were smaller (R2 = 0.39, P < 0.0001) and became of defenses in plant communities is not random and does not nonsignificant (P = 0.21) when effects of relative damage by specialized vs. depend on plant species richness. In my analysis, nontropical generalized herbivores was taken into account.

6100 | www.pnas.org/cgi/doi/10.1073/pnas.1418643112 Becerra Downloaded by guest on October 2, 2021 have found no consistent relationship between latitude and either the amount of herbivory on plants or the degree of resistance in plants (43, 44). According to this study, it is not higher herbivore pressure that is maintaining plant diversity, but differences in how the herbivore pressure is distributed between relatively specialized or generalized herbivores. Thus, rather than latitude, the fundamental difference between tropical forests and other tropical or temperate biomes that affects plant diversity is that insect herbivory is rampant in tropical forests. In other insect damage is often lower, with mammalian herbivores having a larger impact. My in- terpretation is that because mammalian herbivores tend to feed on a wider variety of plants than insects do [specialization is rare among mammals and monophagy is absent (27)], their herbivory will not favor divergent plant defenses in nontropical forests. This Fig. 4. A positive feedback may exist between insect herbivore specializa- is because the impact of divergence will be weaker on feeders that tion and divergent plant defenses in a community. Disparate defensive can deal with a variety of plant defensive traits. Thus, the impact of systems in a plant community may promote insect specialization because it is specialized insects in nontropical forests should be reduced and difficult to adapt to multiple plant antiherbivore traits. Insect tendencies to diluted by the pressures of generalized herbivory (45). Without the specialize on plants with same defenses would favor plant communities with selective pressures keeping similarly defended plants rare, their species that have divergent defenses. Over time this cycle narrows niches in insects and plants. In insects it is because of resource partitioning and in degree of niche differentiation will be lower (6). Many other plants it is because of natural enemy partitioning. Narrower niches favor studies have previously tested the importance of herbivores pro- both plant and insect stable coexistence and diversity. moting plant diversity, but to my knowledge this the first study that has been done in communities across the world and is also the first time that the effects of insects and mammals have been compared biology. My analysis provides evidence consistent with the idea that across plant communities at different latitudes. this phenomenon is the result of a higher impact of insect herbivory Other reports have indicated that, like insects, fungal patho- relative to mammal herbivory, as well as the higher heterogeneity in gens can also have strong selective pressures on plants that could – plant defenses that predominate in these ecosystems. This study affect plant community structure and diversity (46 48). Although also suggests that decisions on how to preserve biodiversity for the evidence is still scarce, current information indicates that most future will require a deep understanding of the ecology of biological pathogenic fungi are generalists, and thus theoretically their role communities. in community plant diversity over time would be similar to that of mammals, decreasing the impact of specialized insect her- Methods bivory. However, studies also suggest that some fungal species might be specialized, and thus could also favor higher community Compiling Host-Plant Antiherbivore Trait Information. For each butterfly host plant richness (49). As we get more information on host ranges plant I compiled the presence (1)/absence (0) of biotic (ant-attracting traits), of plant pathogenic fungi it will be very interesting to see whether mechanical/morphological (e.g., hairs, spines), and chemical traits potentially involved in defense. In total, the presence of 3 biotic, 12 mechanical/mor- their effects antagonize or synergize with those of insects to phological, and 60 chemical traits were investigated for each reported butterfly make plant communities more or less diverse. host. Biotic and morphological traits were researched mostly at the level of The reasons behind phytophagous insects’ specialization have family and genus, but when possible, at the level of species using botanical been of great interest in ecology for many decades. Although it is descriptions and research articles (56–67). Chemical traits belonged to five thought that most insect species can circumvent only a limited set classes of compounds [following the classification of Harborne et al. (68)]. of plant resistance traits because of genetic, physiological, and – These traits were: biologically active carbohydrates and lipids, nitrogen-con- behavioral constraints (50 52), we still know little about the taining compounds (excluding alkaloids), alkaloids, phenolics, and terpenes. impact of plant community assembly in insect herbivore habits. Each of these classes were subdivided further into a total of 60 subclasses Interpretation of the causal basis of the correlations in this re- following Harborne et al. (68). For each host plant, I investigated the presence port is limited in the absence of analyses of the other potential of these compounds by searching the ISI Web of Knowledge database using ecologically relevant correlations. However, results presented the taxon name as a key topic word. Every host plant was first investigated here do support a previous proposal that there might be a using the species and then the genus name. To ensure a thorough search for feedback loop between herbivore specialization and plant de- each taxon, all ISI records listed up to August 2013 were reviewed for in- fense community structure that might further promote insect formation on chemical constituents. Many references cited by the ISI records specialization by limiting colonization of new hosts (13) (Fig. 4). were also examined. When available, the chemical search also included books ECOLOGY As they go on attacking plants in a community, every now and written on chemical characterizations of different families, genera, and species then insects will land on a nonhost plant and, depending on how (68–71). Some deficiencies in published data can affect the precision of esti- unique its defensive traits are, they may or may not be able to start mates of defense divergence, but again, I attempted to ensure that any errors using this additional host (12). This constant herbivore pressure made comparisons between regions more conservative. For example, the will favor plants that produce novel defenses to which local her- chemistry of plants from some regions has been more thoroughly investigated bivores are not adapted (14, 53, 54). However, as plant species by researchers and this level of effort can affect comparisons. However, re- diverge in defenses insects will increasingly tend toward feeding gions where plants are chemically better known are the ones with a lower only on those hosts they can still attack (55). The fewer hosts an diversity of butterflies, (Middle East, Louisiana, and Arizona) and we would insect has to contend with, the fewer conflicting selective pressures expect this to work against the thesis that higher community chemical di- it experiences and the more effectively it may evolve to exploit its vergence in plants leads to higher butterfly diversity. hosts. Any such increased insect virulence only strengthens se- lection on plants to express divergent defensive systems, thus Testing for the Influence of the Number of Host Plants Studied on the Degree perpetuating the feedback cycle of increasing plant dissimilarity of Average Pairwise Antiherbivore Trait Divergence of Communities. If the and increasing insect specialization. Although there is evidence magnitude of average pairwise defense dissimilarity was dependent only on for these premises, it will require more research to ascertain the the number of species of plants, real communities should have an average extent at which herbivore specialization and plant community degree of dissimilarity equal to that of a random community of same size (the defenses feed back upon each other (13). null model). The null distributions of average chemical distances were gen- Why insect herbivores and plants are more diverse in tropical erated with 10,000 randomizations for each test. P values were for two-tailed forests has long been one of the most intriguing questions in tests: that is, the observed defense distance was more different from all but

Becerra PNAS | May 12, 2015 | vol. 112 | no. 19 | 6101 Downloaded by guest on October 2, 2021 the proportion P of the 10,000 null communities. Randomizations were done Relating Herbivory Studies to Plant Diversities. After the geographic location with a program written in R (72). of each study was determined, I recorded the plant species richness for each site as the number of species of vascular plants per 10,000 km2, following Compiling Herbivory Studies and Calculating Relative Damage by Specialized estimates on a map from Global Biodiversity: Species Numbers of Vascular vs. Generalized Herbivores. To locate herbivory studies, I searched the ISI Web Plants (https://www.biologie.uni-hamburg.de/b-online/bonn/Biodiv_mapping/ of Knowledge database (v4.2, Thompson Corporation) using a key word phytodiv.htm) (75). Given that the studies of herbivory were done at a lo- search for any of the following terms: herbivor*, folivor*, defoliat*, grazer. cal scale but the map of global biodiversity is reported at a regional scale, I conducted the review during 2011–2013 and cataloged all records published I tested how well the map reflected local diversities by comparing plant before November of 2013. I also reviewed books and other references that species richness from the map with species counts available for seven sites were cited by these articles. The database included studies that provided where herbivory studies took place. The seven sites were: (i) Monterey – information on herbivory (folivory) and also described the herbivore groups County, California (1,500 2,000 spp. in the map; 1,750 spp. recorded for the responsible for the attack (insect/mammal), so that the relative amount present study; 1,776 spp. according to www.calflora.org); (ii) Tehuacán, – of damage inflicted by each group was available or could be inferred. Puebla, Mexico (2,000 3,000 spp. in the map; 2,500 spp. recorded for the I recorded standing levels of damage (percent of plant biomass consumed). present study; 2,621 spp. according to whc.unesco.org/en/tentativelists/ > Considerable effort was taken to have the most precise information on es- 5783); (iii) La Selva Biological Station, Costa Rica ( 5,000 spp. in the map; > timates of herbivory. When the article only mentioned insect or mammal 5,100 spp. recorded for the presentent study; 5,000 spp. according to www. – herbivory, the authors were contacted to ask if the reason was that herbivory ots.ac.cr); (iv) Monteverde, Costa Rica (2,000 3,000 spp. in the map; 2,500 > by the unreported agents was absent or if they had unreported data. If they spp. recorded for the present study; 2,500 spp. according to www.westnet. did, I included it. Some unavoidable problems can affect calculations, but I com/costarica/foundation/cct.html); (v) Dona Ana Mountain Range, New Mexico (1,000–1,500 spp. in the map; 1,250 spp. recorded for the present study; tried to remedy them by using estimates that make testing of the hypothesis 1,143 spp. according to jornada-www.nmsu.edu); (vi) Chamela, Jalisco, Mexico more conservative. Thus, because a number of tropical mammals have been (1,500–2,000 spp. in the map; 1,750 spp. recorded for the present study; >1,500 extirpated by human activities from some neotropical regions in the last spp. according to www.ibiologia.unam.mx/ebchamela); and (vii) Flora sur- couple of centuries, I used reports on herbivory fractions by Leigh and Smythe rounding Iquitos, Peru (2,000–3,000 spp. in the map; 2,500 spp. recorded for the (73) for Barro Colorado Island that indicated that mammals (monkeys and present study; >2,265 spp. according to www.botany.si.edu/projects/cpd/sa/sa9. sloths) consume about 14% as much foliage as is consumed by insects. For the htm). Because there were no major discrepancies between the map and tropical rain forests of Peru, where monkeys tend to be scarcer, this value was reported species counts, I considered the richness values provided by the 10% (74). Some of these herbivory estimates cannot be separated from fru- map to be adequate for the purposes of this investigation. givory that can lead to . However, because I was testing whether A Pearson product-moment correlation was calculated for relative damage generalized herbivory would result in decreased plant diversity and the neo- by specialized vs. generalized herbivores and plant species richness. Plant tropics have high plant richness, overestimating mammalian herbivory was species richness was transformed following the Box-Cox procedure with the conservative. Similarly, mammal herbivory is sometimes underestimated be- parameter λ = 0.4 to stabilize the variance and make the residuals more cause mammals are more likely to eat whole leaves or branches and these symmetrically distributed. In total, 53 studies were compiled, but when sites leaves might not be selected for quantification in herbivory studies. This can be were too close (within 1° latitude and longitude) and had the same plant especially important in regions outside tropical forests, where mammals are species richness, they were averaged. I recorded the latitude of each study and more abundant. However, this also contributed to a more conservative testing. examined its correlation with Box-Cox transformed plant diversity using JMP A measure of relative damage by specialized (insect) vs. generalized (76). I also examined the independent effects of specialized herbivory, latitude, (mammal) herbivores was calculated as the natural log of the percent spe- and total amount of herbivore damage (percent specialized + percent gen- cialized (insect) herbivory damage divided by the percent generalized eralized) with multiple regression analysis. Additionally, I analyzed the (mammal) herbivory damage. To avoid undefined logs, I gave a value of 1% relationship between total herbivory and latitude. to herbivory caused by either group when it was reported to be absent or minimal. I previously tried other measures of relative damage, such as the ACKNOWLEDGMENTS. Thanks to Peter Chesson, Brian Enquist, Larry Venable, difference between percent specialized and percent generalized herbivory John Wiens, Ursula Basinger-Walholm, Yue Li, Simon Stump, and two anony- and vice versa, and results were highly similar. mous reviewers for providing comments on an earlier manuscript.

1. Novotny V, et al. (2004) No tree an island: The plant-caterpillar food web of secondary 16. Gilbert LE (1980) Coevolution of animals and plants. Coevolution of Animals and rainforest in New Guinea. Ecol Lett 7(11):1090–1100. Plants, eds Gilbert LE, Raven PH (Univ of Texas Press, Austin, TX), pp 247–263. 2. Price PW (2002) Resource-driven terrestrial interaction webs. Ecol Res 17(2):241–247. 17. Shannon CE (1948) A mathematical theory of communications. The Bell System 3. Adler PB, Fajardo A, Kleinhesselink AR, Kraft NJB (2013) Trait-based tests of co- Technical Journal 27:379–423. existence mechanisms. Ecol Lett 16(10):1294–1306. 18. Hansen RM, Ueckert DN (1970) Dietary similarity of some primary consumers. Ecology 4. Connell JH (1971) On the role of natural enemies in preventing competitive exclusion 51(4):640–648. in some marine animals and in rain forest trees. Dynamics of Populations, eds Den 19. Brown JH, Davidson DW (1977) Competition between seed-eating rodents and in Boer PJ, Gradwell GR (Center for Agricultural and Documentation, Wageningen, The desert ecosystems. Science 196(4292):880–882. Netherlands), pp 298–312. 20. Steneck R (2001) Functional groups. Encyclopedia of Biodiversity, ed Levin SA (Aca- 5. Janzen DH (1970) Herbivores and the number of tree species in tropical forests. Am demic, New York), pp 121–139. Nat 104(940):501–517. 21. Huntzinger M, Karban R, Cushman JH (2008) Negative effects of herbi- 6. Chesson P, Kuang JJ (2008) The interaction between and competition. vores on invertebrates in a coastal dune community. Ecology 89(7):1972–1980. Nature 456(7219):235–238. 22. Ritchie ME (2000) Nitrogen limitation and trophic vs. abiotic influences on insect 7. Basset Y (1992) Host specificity of arboreal and free-living insect herbivores in rain herbivores in a temperate grassland. Ecology 81(6):1601–1612. forests. Biol J Linn Soc Lond 47(2):115–133. 23. Crawley M (1983) Herbivory (Univ of California Press, Berkeley, CA). 8. Dyer LA, et al. (2007) Host specificity of in tropical and temperate forests. 24. Schoonhoven LM, Van Loon JJA, Dicke M (2005) Insect-Plant Biology (Oxford Univ Nature 448(7154):696–699. Press, New York), p 543. 9. Johnson DJ, Beaulieu WT, Bever JD, Clay K (2012) Conspecific negative density de- 25. Ali JG, Agrawal AA (2012) Specialist versus generalist insect herbivores and plant pendence and forest diversity. Science 336(6083):904–907. defense. Trends Plant Sci 17(5):293–302. 10. Novotny V, et al. (2002) Low host specificity of herbivorous insects in a tropical forest. 26. Bernays E, Graham M (1988) On the evolution of host specificity in phytophagous Nature 416(6883):841–844. . Ecology 69(4):886–892. 11. Bernays EA, Chapman RF (1994) Host-Plant Selection by Phytophagous Insects 27. Shipley LA, Forbey JS, Moore BD (2009) Revisiting the dietary niche: When is a (Chapman and Hall, New York), p 305. mammalian herbivore a specialist? Integr Comp Biol 49(3):274–290. 12. Becerra JX (1997) Insects on plants: Macroevolutionary chemical trends in host use. 28. Dearing MD, Mangione AM, Karasov WH (2000) Diet breadth of mammalian herbi- Science 276(5310):253–256. vores: Nutrient vs. detoxification constraints. Oecologia 123:397–405. 13. Becerra JX (2007) The impact of plant–herbivore coevolution on plant community 29. Lindroth RL (1988) Adaptations of mammalian herbivores to plant chemical defenses. structure. Proc Natl Acad Sci USA 104(18):7483–7488. Chemical Mediation of Coevolution, ed Spencer KC (Academic, London), pp 415–445. 14. Ehrlich PR, Raven PH (1964) Butterflies and plants: A study in coevolution. Evolution 30. Moore BD, Foley WJ (2000) A review of feeding and diet selection in koalas (Phas- 18(4):586–608. colarctos cinereus). Aust J Zool 48(3):317–333. 15. Kursar TA, et al. (2009) The evolution of antiherbivore defenses and their contribu- 31. Schowalter TD (2000) Insect Ecology (Academic, London), p 483. tion to species coexistence in the tropical tree genus Inga. Proc Natl Acad Sci USA 32. Sedio BE, Ostling AM (2013) How specialised must natural enemies be to facilitate 106(43):18073–18078. coexistence among plants? Ecol Lett 16(8):995–1003.

6102 | www.pnas.org/cgi/doi/10.1073/pnas.1418643112 Becerra Downloaded by guest on October 2, 2021 33. Dinnage R, Cadotte MW, Haddad NM, Crutsinger GM, Tilman D (2012) Diversity of Advances in Phytochemistry, eds Romeo JT, Saunders JA, Barbosa P (Plenum Press, plant evolutionary lineages promotes diversity. Ecol Lett 15(11):1308–1317. New York), Vol 30, pp 1–24. 34. Pellissier L, et al. (2013) Turnover of plant lineages shapes herbivore phylogenetic 54. Becerra JX (2015) Macroevolutionary and geographical intensification of chemical beta diversity along ecological gradients. Ecol Lett 16(5):600–608. defense in plants driven by insect herbivore selection pressure. Current Opinion in 35. Louda SM, Potvin MA (1995) Effects of inflorescence-feeding insects on the de- Insect Science, 10.1016/j.cois.2015.01.012. mography and lifetime fitness of a native plant. Ecology 76(1):229–245. 55. Cornell HV, Hawkins BA (2003) Herbivore responses to plant secondary compounds: A 36. Maron JL, Kauffman MJ (2006) Habitat-specific impacts of multiple consumers on test of phytochemical coevolution theory. Am Nat 161(4):507–522. plant population dynamics. Ecology 87(1):113–124. 56. Dahlgren RMT, Clifford HT, Yeo PF (1985) The Families of the Monocotyledons: 37. Rausher MD, Feeny P (1980) Herbivory, plant density, and plant reproductive success: Structure, Evolution, and (Springer, New York), p 520. The effect of on reticulata. Ecology 61:905–917. 57. Kubitzki K (2011) Eudicots: Sapindales, Cucurbitales, Myrtaceae (Springer, New York), 38. Marquis RJ (1984) Leaf herbivores decrease fitness of a tropical plant. Science p 446. 226(4674):537–539. 58. Kubitzki K (1990) Pteridophytes and Gymnosperms (Springer, New York), p 417. 39. Cascante-Marin A, Wolf JHD, Oostermeijer JGB (2009) Wasp florivory decreases re- 59. Kubitzki K (1993) Flowering Plants. Dicotyledons: Magnoliid, Hamamelid and Car- productive success in an epiphytic bromeliad. Plant Ecol 302:149–153. yophyllid Families (Springer, New York), p 663. 40. Voelckl C, Baldwin IT (2004) Generalist and specialist lepidopteran larvae elicit dif- 60. Kubitzki K (1998) Flowering Plants. Monocotyledons: Lilianae (Except Orchidaceae) ferent transcriptional responses in Nicotiana attenuata, which correlate with larval FA (Springer, New York), p 488. profiles. Ecol Lett 7(6):770–775. 61. Kubitzki K (2004) Flowering Plants. Dicotyledons: Celastrales, Oxalidales, Rosales, 41. Bingham RA, Agrawal AA (2010) Specificity and trade-offs in the induced plant de- Cornales, Ericales (Springer, New York), p 501. fence of common milkweed Asclepias syriaca to two lepidopteran herbivores. J Ecol 62. Kubitzki K (2003) Flowering Plants. Dicotyledons: Capparales, Malvales and Non- 98(5):1014–1022. Betalain Caryophyllales (Springer, New York), p 427. 42. Schemske DW, Mittelbach GG, Cornell HV, Sobel JM, Roy K (2009) Is there a latitudinal 63. Kubitzki K (1998) Flowering Plants. Monocotyledons: Alismatanae and Commelinanae gradient in the importance of biotic interactions? Annu Rev Ecol Syst 40:245–269. (Except Gramineae) (Springer, New York), p 521. 43. Moles AT, et al. (2011) Putting plant resistance traits on the map: A test of the idea 64. Kubitzki K (2004) Flowering Plants. Dicotyledons (Springer, New York), p 501. that plants are better defended at lower latitudes. New Phytol 191(3):777–788. 65. Kubitzki K (2007) Flowering Plants. Eudicots: Berberidopsidales, Buxales, Cross- 44. Moles TA, Bonser SP, Poore GB, Wallis IR, Foley WJ (2011) Assessing the evidence for osomatales, Fabales P.P., Geraniales, Gunnerales, Myrtales P.P., Proteales, Saxifraga latitudinal gradients in plant defence and herbivory. Funct Ecol 25(2):380–388. (Springer, New York), p 524. 45. Stinchcombe JR, Rausher MD (2002) The evolution of tolerance to deer herbivory: 66. Kubitzki K (2007) Flowering Plants. Eudicots: Asterales (Springer, New York), p 647. Modifications caused by the abundance of insect herbivores. Proc Biol Sci 269(1497): 67. Cronquist A (1981) An Integrated System of Classification of Flowering Plants 1241–1246. (Columbia Univ Press, New York). 46. Comita LS, Muller-Landau HC, Aguilar S, Hubbell SP (2010) Asymmetric density de- 68. Harborne JB, Baxter H, Moss GP (1999) Phytochemical Dictionary, a Handbook of pendence shapes species abundances in a tropical tree community. Science 329(5989): Bioactive Compounds from Plants (Taylor & Francis, London). 330–332. 69. Bisby FA, Buckingham J, Harborne JB (1994) Phytochemical Dictionary of the Legu- 47. Terborgh J (2012) Enemies maintain hyperdiverse tropical forests. Am Nat 179(3): minosae (Chapman & Hall, New York). 303–314. 70. Bussing A (2000) Mistletoe: The Genus Viscum (Harwood Academic, Amsterdam). 48. Bagchi R, et al. (2014) Pathogens and insect herbivores drive rainforest plant diversity 71. Jury SL (1987) The Euphorbiales: Chemistry, Taxonomy and Economic Botany (Aca- and composition. Nature 506(7486):85–88. demic, London), p 325. 49. Bagchi R, et al. (2010) Testing the Janzen-Connell mechanism: Pathogens cause 72. R Development Core Team (2004) R: A Language and Environment for Statistical overcompensating density dependence in a tropical tree. Ecol Lett 13(10):1262–1269. Computing (Univ of California, Berkeley, CA), 2.01. 50. Becerra JX (1994) Squirt-gun defense in Bursera and the chrysomelid counterploy. 73. Leigh EG, Smythe N (1978) Leaf production, leaf consumption, and the regulation Ecology 75(7):1991–1996. of folivory on Barro Colorado Island. The Ecology of Arboreal Folivores,ed 51. Whiteman NK, Pierce NE (2008) Delicious poison: Genetics of Drosophila host plant Montgomery GG (Smithsonian Institute Press, Washington, DC), pp 33–50. preference. Trends Ecol Evol 23(9):473–478. 74. Coley PD (1982) Rates of herbivory on different tropical trees. The Ecology of Tropical 52. Rasmann S, Agrawal AA (2011) Evolution of specialization: A phylogenetic study of Forests, eds Leight EG, Stanley A, Windsor DM (Smithsonian Institution Press, Wash- host range in the red milkweed (Tetraopes tetraophthalmus). Am Nat 177(6): ington, DC), pp 123–132. 728–737. 75. Barthlott W, et al. (1999) Terminological and methodological aspects of the mapping 53. Berenbaum M, Zangerl AR (1995) Phytochemical diversity: Adaptation or random and analysis of global biodiversity. Acta Botanica Fennica 162(0):103–110. variation? Phytochemical Diversity and Redundancy in Ecological Interactions. Recent 76. SAS Institute Inc. (2012) Using JMP 10 (SAS Publishing, Cary, NC). ECOLOGY

Becerra PNAS | May 12, 2015 | vol. 112 | no. 19 | 6103 Downloaded by guest on October 2, 2021