Molecular and Morphological Phylogenetic Analyses of New World Cycad Beetles: What They Reveal about Cycad Evolution in the New World

Two major lineages of beetles inhabit cycad cones in the New World: weevils (Curculionoidea) in the subtribe Allocorynina, including the genera Notorhopalotria Tang and O’Brien, Parallocorynus Voss, Protocorynus O’Brien and Tang and Rhopalotria Chevrolat, and beetles in the family Erotylidae, including the genus Pharaxonotha Reitter. Analysis of the 16S ribosomal RNA (rRNA) mitochondrial gene as well as cladistic analysis of morphological characters of the weevils indicate four major radiations, with a probable origin on the cycad genus Dioon Lindl. and comparatively recent host shifts onto Zamia L. Analysis of the 16S rRNA gene for erotylid beetles indicates that an undescribed genus restricted to New World Ceratozamia Brongn. is the most early-diverging clade, and this lineage is sister to a large radiation of the genus Pharaxonotha onto Zamia, with apparent host shifts onto Dioon and Ceratozamia. Analysis of beetles are in accord with current models of continental drift in the Caribbean basin, support some proposed species groupings of cycads, but not others, and suggest that pollinator type may impact population genetic structure in their host cycads. Diversity 2018, 10, 38; doi:10.3390/d10020038 www.mdpi.com/journal/diversity Diversity 2018, 10, 38 2 of 26


Introduction
Over the last three decades, evidence has accumulated that insect pollination is widespread in New World cycads. This evidence includes wind and insect exclusion experiments on ovulate cones of three species in the cycad genus Zamia, detailed observations of the life cycle and behavior of the beetles that inhabit them [1][2][3][4][5], as well as observations on other New World cycad genera Ceratozamia, Dioon, and Microcycas (Miq.) A. DC. [6][7][8]. Similar experiments and observations on other continents indicate the same for other genera of cycads [9][10][11][12][13][14][15][16][17][18][19]. In a recent seminal work on guidelines for cycad classification, insect symbionts of cycads were identified as having a potentially important impact on cycad classification: "Insects appear to be the primary vectors for pollination [ . . . ] evidence is accumulating to suggest coevolutionary processes between cycads and their pollinators. Once these processes are uncovered, resulting data will probably have a significant impact on how cycad taxa are classified" [20]. During the 6th International Conference on Cycad Biology, a coordinated global effort was organized to collect and study the insect pollinators of cycads [21]. One of the explicit goals was to use this information to understand cycad evolution. In this paper, we report on results of this insect survey effort in the New World, present phylogenetic analyses of the insects found, and discuss implications for cycad taxonomy.
O'Brien and Tang [8] recently described or reviewed all known species of Allocorynina, but they did not present a detailed phylogenetic analysis of the species. All known species inhabit and develop in cones of New World cycads. Six species of New World Pharaxonothinae have been described, but many remain undescribed [7,27]. New World forms are closely related to the recently described genus Cycadophila found on the Asian cycad genus Cycas [26,28]. The lack of phylogenetic frameworks for the New World groups of cycad beetles hinders the proper allocation of biological information and host/beetle associations and limits what can be interpreted from them. For instance, Maldonado-Ruiz and Flores-Vazquez [29] catalogued beetles found with a species of Dioon in Mexico, however, due to lack of keys for identification or prior phylogenetic work, they were not able to assess how many species of Allocorynina and Pharaxonothinae they were dealing with. In this paper, phylogenetic analyses were conducted on both morphological characters and DNA data of Allocorynina and DNA data of Pharaxonothinae beetles collected from cycad cones. The resulting trees from these analyses are used to generate hypotheses about cycad biogeography and evolutionary patterns at genus, species, and population levels [30,31].

Materials and Methods
Beetles were collected from cycad cones in habitat by the authors and other cooperators and include previously described cycad-associated beetle species, as well as many undescribed beetle taxa [8,27]. Total number of cycad taxa sampled include 3 of the 31 recognized species of Ceratozamia, 13 of the 15 species of Dioon, and 29 of the 77 currently recognized species of Zamia [24]. In total, 89 cycad populations or localities yielded beetles with useable DNA for this study. The monotypic Cuban cycad genus Microcycas was not sampled, however, a species of Pharaxonotha has been described from this host [7]. For Allocorynina, institutions for deposition of specimens are listed in O'Brien and Tang [8]. Pharaxonothinae used are deposited in the Florida State Collection of Arthropods, Division of Plant Industry, Gainesville, FL, USA.

Morphological Analysis
For the Allocorynina weevils, external morphology and morphology of genitalia were studied and photographed using Nikon ® SMZ1500 stereoscopic and Eclipse 80i compound microscopes mounted with Nikon ® DS-Fi1 digital cameras. For morphological characters, taxonomic description of species and specimens used, see Appendix A and O'Brien and Tang [8]. For the outgroup, Oxycraspedus cornutus Kuschel (Belidae: Oxycorininae: Oxycornini: Oxycraspedina), currently placed in the same subfamily and tribe as the Allocorynina [32], was chosen. A matrix of 89 characters based mainly on morphology, but also host-associations and behavior such as diet and pupation sites, was built (see Figure A1 in Appendix B). A phylogenetic tree was generated using maximum parsimony (MP) implemented by TNT [33] using default settings, and a strict consensus tree was generated from the ten most parsimonious trees found. Bootstrap support values were generated based on 1000

Materials and Methods
Beetles were collected from cycad cones in habitat by the authors and other cooperators and include previously described cycad-associated beetle species, as well as many undescribed beetle taxa [8,27]. Total number of cycad taxa sampled include 3 of the 31 recognized species of Ceratozamia, 13 of the 15 species of Dioon, and 29 of the 77 currently recognized species of Zamia [24]. In total, 89 cycad populations or localities yielded beetles with useable DNA for this study. The monotypic Cuban cycad genus Microcycas was not sampled, however, a species of Pharaxonotha has been described from this host [7]. For Allocorynina, institutions for deposition of specimens are listed in O'Brien and Tang [8]. Pharaxonothinae used are deposited in the Florida State Collection of Arthropods, Division of Plant Industry, Gainesville, FL, USA.

Morphological Analysis
For the Allocorynina weevils, external morphology and morphology of genitalia were studied and photographed using Nikon ® SMZ1500 stereoscopic and Eclipse 80i compound microscopes mounted with Nikon ® DS-Fi1 digital cameras. For morphological characters, taxonomic description of species and specimens used, see Appendix A and O'Brien and Tang [8]. For the outgroup, Oxycraspedus cornutus Kuschel (Belidae: Oxycorininae: Oxycornini: Oxycraspedina), currently placed in the same subfamily and tribe as the Allocorynina [32], was chosen. A matrix of 89 characters based mainly on morphology, but also host-associations and behavior such as diet and pupation sites, was built (see Figure A1 in Appendix B). A phylogenetic tree was generated using maximum parsimony (MP) implemented by TNT [33] using default settings, and a strict consensus tree was generated from the ten most parsimonious trees found. Bootstrap support values were generated based on 1000 replicates. Multi-state characters were treated as non-additive; and all characters were weighed equally in the analysis. Morphological analyses of cycad-associated erotylid taxa are under way [27], but are not presented here.

DNA Analysis
The quality of DNA preservation of the beetles available for this study varied and was often poor, therefore, only short sections of DNA could be consistently sequenced in the samples available. We selected a fragment of the mitochondrial 16S ribosomal RNA (rRNA) gene with a combined sequence length ranging from 311-316, varying with additions and deletions of sections. The aligned data set contained 318 sites, with 222 constant, 93 variable, and 68 parsimony informative sites. As seen in other arthropods, the 16S rRNA gene is highly AT-rich with average nucleotide frequencies of thymine (T) 42.9%, cytosine (C) 7.9%, adenine (A) 35.0%, and guanine (G) 14.3%. The 16S rRNA gene has been used widely in insect molecular systematics and its utility in discerning species groups and deeper divisions in beetles and other holometabolous insects is well-founded [34][35][36][37][38]. It has been proposed for use as a standard for insect phylogenetics [39]. Total DNA was extracted from individual beetles, either adults or larvae, using Epicenter Master Complete DNA and RNA Purification Kits (Epicenter Technologies, Madison, WI, USA) following the manufacturer protocols and dissolved in 30 µL H 2 O. The mitochondrial 16S rRNA was amplified using the following primers: 73Forward-AGATAGAAACCARCCTGGCT, 98Forward-CGGTYTRAACTCAGATCATGTA, and 430Reverse-AAGACGAGAAGACCCTATAG [26,28]. Reactions were carried out in 25 µL volumes containing 1 µL DNA, 5 µL 5X buffer, 4 µL of 25 µM MgCl 2 , 1 µL of 10 mM dNTPs, 1 µL of 10 µM of each primer, and 0.2 µL 5 U/µL of Taq polymerase (Promega, Madison, WI, USA). PCR was performed using an Eppendorf ep mastercycler (Eppendorf, Westbury, NY, USA) using the following DNA denaturation, annealing and replication protocol: 94 • C for 1 min, then 40 cycles of 94 • C for 15 s, 50 • C for 15 s, and 72 • C for 40 s. Amplified products were cleaned up with the ExoSAP-IT kit (USB, Cleveland, OH, USA) and sequenced bidirectionally on an ABI 3130XL Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).
For weevils, nucleotide sequences were aligned using MAFFT version 7 [40]. The aligned sequences were then phylogenetically analyzed via: (1) maximum parsimony (MP) implemented by TNT [33] using default settings: a strict consensus tree was generated from the ten most parsimonious trees found; (2) maximum likelihood (ML) implemented in RAxML version 8 [41]: the best tree from 20 independent searches was selected and bipartition bootstrap values were written from 500 bootstrap replicates; (3) bayesian inference (BI) implemented in Mr Bayes version 3.2 [42], which was run with two simultaneous searches each using four chains for 1 million generations: trees were sampled every 100 generations, and the first 25% were discarded as burnin.
For erotylid beetles, two Asian species of Pharaxonothinae that inhabit the genus Cycas were used as outgroups: Cycadophila (C.) debaonica and C. (Strobilophila) tansachai [26,28], GenBank accession numbers KR005715, KY365223, respectively. Multiple-sequence alignments were conducted with CLUSTAL W [43]. Phylogenetic trees were generated using maximum parsimony (MP), neighbor joining (NJ), and maximum likelihood (ML) methods as implemented in PAUP 4.0b10 [44] and MEGA5 [45]. Bootstrap support values were generated based on 1000 replicates. The best fit model of sequence evolution employed in the ML analysis was the Tamura 3-parameter+G model with log likelihood −1448.98 and Gamma distribution 0.2653.

Allocorynina Trees
The phylogenetic tree based on MP analysis of the morphological, behavioral, and host-association data set is displayed in Figure 2. This tree generally supports the genera and subgenera recognized by O'Brien and Tang [8]. Genera Protocorynus, Neocorynus, and Parallocorynus and subgenera Rhopalotria, Dysicorynus, Neocorynus, Eocorynus, and Parallocorynus are monophyletic. Only the monophyly of genus Rhopalotria and its subgenus Allocorynus is not supported. The three molecular trees for the Allocorynina generated from MP, ML, and BI analyses of the 16S rRNA data set were almost identical, and any slightly conflicting clades are linked to very low support values. Therefore, in Figure 3, only the molecular tree from the BI analysis is presented, with ML support values annotated underneath the branches. Also, in this tree, each analyzed sample displays corresponding host cycad species and the geographic region where they were collected. All genera (and most subgenera) are well supported, but the relationships between them are not. The main difference between the two trees is that in the molecular tree subgenus Rhopalotria is paraphyletic with respect to Allocorynus, whereas in the morphology tree, the situation is reversed and subgenus Allocorynus is paraphyletic, but neither of them refute the monophyly of the other analysis. Another major difference in the molecular tree is the strong support for Protocorynus to be sister to the remaining three genera. These differences between the two trees may, in part, be attributed to different choices in outgroups.

Allocorynina Trees
The phylogenetic tree based on MP analysis of the morphological, behavioral, and hostassociation data set is displayed in Figure 2. This tree generally supports the genera and subgenera recognized by O'Brien and Tang [8]. Genera Protocorynus, Neocorynus, and Parallocorynus and subgenera Rhopalotria, Dysicorynus, Neocorynus, Eocorynus, and Parallocorynus are monophyletic. Only the monophyly of genus Rhopalotria and its subgenus Allocorynus is not supported. The three molecular trees for the Allocorynina generated from MP, ML, and BI analyses of the 16S rRNA data set were almost identical, and any slightly conflicting clades are linked to very low support values. Therefore, in Figure 3, only the molecular tree from the BI analysis is presented, with ML support values annotated underneath the branches. Also, in this tree, each analyzed sample displays corresponding host cycad species and the geographic region where they were collected. All genera (and most subgenera) are well supported, but the relationships between them are not. The main difference between the two trees is that in the molecular tree subgenus Rhopalotria is paraphyletic with respect to Allocorynus, whereas in the morphology tree, the situation is reversed and subgenus Allocorynus is paraphyletic, but neither of them refute the monophyly of the other analysis. Another major difference in the molecular tree is the strong support for Protocorynus to be sister to the remaining three genera. These differences between the two trees may, in part, be attributed to different choices in outgroups.  The morphological analysis and the resultant tree are valuable in answering how each of these weevil lineages differ and what may be driving evolution in this group. Each recognized genus and subgenus is distinguished by differences in their genitalia and also in the spination of their profemora. Behavioral observations [3,4] indicate that the profemora are used during courtship battles between males. The spination on their profemora appear to function in grasping an opposing male during these mating struggles. Sexual selection, larval feeding sites, pupation sites, as well as host genus (Dioon vs. Zamia) appear to account for many of the morphological differences between the genera and subgenera. For details on synapomorphic characters for major clades see Appendix C.
This phylogenetic analysis of the Allocorynina is the most extensive to date and greatly expands upon previous efforts [31,32], while generally reaffirming the sub-/generic concepts erected by O'Brien and Tang [8]. The morphological analysis and the resultant tree are valuable in answering how each of these weevil lineages differ and what may be driving evolution in this group. Each recognized genus and subgenus is distinguished by differences in their genitalia and also in the spination of their profemora. Behavioral observations [3,4] indicate that the profemora are used during courtship battles between males. The spination on their profemora appear to function in grasping an opposing male during these mating struggles. Sexual selection, larval feeding sites, pupation sites, as well as host genus (Dioon vs. Zamia) appear to account for many of the morphological differences between the genera and subgenera. For details on synapomorphic characters for major clades see Appendix C.
This phylogenetic analysis of the Allocorynina is the most extensive to date and greatly expands upon previous efforts [31,32], while generally reaffirming the sub-/generic concepts erected by O'Brien and Tang [8].

Erotylidae Tree
The trees produced from MP, NJ, and ML analyses of 16S rRNA sequences for Erotylidae were similar, therefore, only the ML tree is displayed in Figure 4. The MP and NJ trees are provided in Appendix D as Figures A2 and A3. The ML tree, as well as the MP and NJ trees, can be divided into four sections: (1) The most early-diverging clade is an undescribed genus residing in cones of Ceratozamia (purple color in Figure 4); the remainder of the New World taxa are sister to this clade and are tentatively assigned to the genus Pharaxonotha. (2) This Pharaxonotha clade can be broadly assigned to three sections; the first of these are labeled "Early-diverging lineages Mexico to South America" (blue color in Figure 4). Within this division, the most early-diverging branches reside in cones of Zamia onan-reyesii C. Nelson and Sandoval, an aerial-stemmed species from Honduras and subterranean-stemmed Z. cunaria Dressler and D. W. Stev. and Z. pyrophylla Calonje, D. W. Stev. and A. Lindstr., in the eastern Panama-northern Colombia region. Also among these early diverging lineages are species of Pharaxonotha that reside in other Ceratozamia, Dioon, and Zamia hosts. (3) A third broad division consist of Pharaxonotha inhabiting cones of the Zamia pumila L. species complex in the Caribbean, specifically on islands of the Greater Antilles, the Bahamas archipelago, and the Florida peninsula (green color in Figure 4). This division consists of two clades; one in the easternmost section of the Greater Antilles on Puerto Rico and Hispaniola and the other clade in the western Greater Antilles, Florida, and the Bahamas. (4) Lastly, there is a fourth major division, consisting of a more recent radiation of Pharaxonotha that also extends from Mexico to South America and inhabits Zamia and Dioon (red color in Figure 4). Although bootstrap support for many of these clades are weak, they fall consistently into these four broad categories in the ML, MP, and NJ trees, with only one exception: the P. kirschii Reitter lineage, discussed below.
Although no morphological analyses are presented in this paper for the Erotylidae, Xu et al. [28], Skelley et al. [26], and Tang et al. [27] have identified a number of external morphological and genital characters which support many of the clades revealed here (Figure 4) through phylogenetic analysis of the 16S rRNA gene. All species discussed herein are assigned to a single subfamily, the Pharaxonothinae, of the Erotylidae. Of special interest is the taxon Pharaxonotha kirschii, the type species for the genus and the only known Pharaxonothinae in the New World that is a generalist feeder. Although it has been found with Zamia, it is the only known member of the New World Pharaxonothinae that is not an obligate inhabitant of a cycad during its life cycle and is widely distributed in Central America [26]. Our DNA analysis indicates that the Pharaxonotha associated with Z. inermis (specimen D0057 in Figure 4), while a distinct species, is related to P. kirschii. The specimen used for DNA analysis was a larva and an adult female associated with that larva matches the morphology of P. kirschii. This specimen and Pharaxonotha kirschii may be part of a species complex, with different branches of this complex inhabiting either cycads or non-cycad hosts. In the ML tree, this P. kirschii clade is associated with the Caribbean group, however, in the MP and NJ trees ( Figures A2 and A3), these two taxa are associated with the early diverging lineages from Mexico to South America.
While all Zamia that have been closely sampled have yielded Pharaxonotha beetles from their cones, this is not so for all Dioon and Ceratozamia populations examined. The cycad genus Dioon was especially well-sampled in this study and in one species, Dioon mejiae Standl. and L.O.Williams, no erotylids were detected in three separate samples consisting of a total of many hundreds of beetles (only Allocorynina weevils found). Furthermore, the erotylid beetles sampled here from seven other Dioon species cluster in three disparate branches in the erotylid trees, suggesting separate colonization events of Dioon by this group of beetles.

Using Beetle Trees to Generate Hypotheses of Cycad Evolution
While the 16S rRNA gene is considered to provide fairly accurate phylogenies for beetles and other insects, the search for plastid and nuclear genes of similar utility in the construction of

Using Beetle Trees to Generate Hypotheses of Cycad Evolution
While the 16S rRNA gene is considered to provide fairly accurate phylogenies for beetles and other insects, the search for plastid and nuclear genes of similar utility in the construction of phylogenies for New World cycads has had equivocal results [46][47][48][49]. Genetic analysis of beetles that have coevolved with cycads may reveal evolutionary patterns in cycads that are much more difficult to discern from direct examination of cycad genomes. The following discussions will be confined mainly to the generation of hypotheses about cycad evolution in the New World based on the interpretation of the beetle trees produced in this study.
Before proceeding, we offer some caveats and limitations as to what can or cannot be deduced about cycad evolution from coevolving insect pollinators. Cycads typically form colonies or groves widely dispersed from other populations of conspecific or closely related species [50], and in such situations, flight distances of seed dispersers or pollinators between populations may be typically measured in kilometers or tens of kilometers. Cycad seeds are relatively large and heavy compared to available animal dispersers, and field studies indicate that the great majority of cycad seeds fall within two meters of their mother [51][52][53], suggesting that gene flow between distant populations via seed dispersal is rare. These population structures and seed dispersal constraints may configure insect pollinators as the primary gatekeepers of gene flow in most cycad populations. Thus, when two cycad populations host different species of beetles believed to be involved in pollination, we can use this to infer that the cycads in question may be genetically isolated, as little or no gene flow is likely to occur via exchange of pollinators. When the scenario is reversed and cycad populations share the same pollinator, we cannot automatically assume gene flow is occurring between the cycad populations via pollinators. Studies by Donaldson et al. [9] and Terry [13] show that pollen loads on insects that leave cycad cones drop rapidly with distance and time from the source cones. Crowson [54,55] illustrated deep antennal pockets in the antennae of Allocoynina weevils and hypothesized that these might be structures that function to hold pollen. Closer examination of these pockets with SEM [8] indicate that their entrances are occluded with hairs with a probable sensory function and that these pockets are unlikely to facilitate pollen transport. While a cycad pollinator may be able to fly between populations of different host cycad species and interbreed with other populations of its own species, during this process, it may not be transferring any cycad pollen between distant host cycad populations. A phylogenetic study of thrips pollinators of the Australia cycad Macrozamia suggests that both cycad and pollinator populations fragment together as a result of climate change and aridification [56] and that co-diversification of the host and pollinator in allopatry appears to be the process affecting diversity. Furthermore, that study concluded that distinct cycad species separated by short distances may continue to share the same pollinator species. Similarly, in New World cycads, O'Brien and Tang [8] recognized that some cycad beetles inhabit more than one host cycad species: Examination of these DNA-based trees indicates that host-shifts of pollinating beetles have occurred during the evolution of cycads. We cannot assume strict co-speciation of beetle lineages with their corresponding host cycad lineage. Host-shifts of cone beetles between cycad genera and between cycad species within a genus can occur. Although this limits our ability to make inferences about coevolution of cycads and their beetles, the host-shifts in themselves are interesting and informative about evolutionary processes in cycads [57].

Cycad Hypotheses Based on the Allocorynina Trees
From the DNA-based tree for Allocorynina (Figure 3) in combination with morphological data (Appendix A, [8]) we generate these hypotheses about cycads: (1) The presence of one or two species of Allocorynina in all species of Dioon sampled compared with its absence from many Zamia species and its complete absence in the other New World cycad genera Ceratozamia and Microcycas suggests that Dioon is the earliest host lineage colonized by Allocorynina weevils, with one or possibly two host-shifts onto Zamia. In this hypothesis, Allocorynina are the original pollinators in the genus Dioon, while erotylid beetles are later colonists in Dioon.
(2) Based on the morphological and genetic analyses of its pollinator Dioon mejiae, located on the Chortis block, a tectonic terrane roughly corresponding to the country of Honduras [58], is hypothesized to be one of the earliest bifurcating lineages within Dioon.

Cycad Hypotheses Based on the Erotylidae Tree
From the DNA-based tree of the Erotyidae (Figure 4) we generate these hypotheses: (I) The most early-diverging lineage of cycad-associated erotylid pharaxonothine beetles in the New World is confined to the genus Ceratozamia. This branching pattern is consistent with fossil evidence indicating that the Ceratozamia lineage may have first evolved in Europe in the mid Cenozoic and then migrated to North America prior to the complete separation of these continents [61]. In addition, the apparent close relation between cycad-associated erotylids of the New World with those found on Cycas in Asia, suggest that these beetles may have an ancient Laurasian association with cycads that predates the breakup of Laurasia. (II) Two early-diverging erotylid beetle lineages associated with Zamia are located in: (a) Honduras on Z. onan-reyesii and (b) The northern South America-Darien region on Zamia cunaria and Z. pyrophylla. We may hypothesize that these host lineages of Zamia are among the earliest to diverge for the genus and are likely relicts from an earlier radiation of Zamia throughout these regions. (III) The presence of erotylid beetles in the cones of all species of Zamia that have been carefully sampled suggests that these were the original pollinators of Zamia. In this hypothesis, the Allocorynina weevils are later colonists of Zamia. (IV) In addition to old relictual clades in hypothesis II, the existence of three separate derived clades of Pharaxonotha beetles on Zamia suggests that at least three recent and separate radiations of Zamia have occurred in the following regions: (a) A radiation into the eastern islands of the Greater Antilles, which includes Hispaniola and Puerto Rico, probably beginning when these landmasses were more closely associated with Central America [60]; (b) A more recent radiation into the western islands of the Greater Antilles, including Cuba, Cayman Islands and Jamaica and neighboring landmasses of the Bahamas and Florida; and (c) Sister to these two Caribbean lineages, a recent radiation onto Zamia in Mesoamerica, Central America, and northern South America. (V) The separation of Pharaxonotha beetles, that inhabit Dioon cones, into three distinct and not closely related clades and the absence of erotylids on one species, D. mejiae in Honduras, at the periphery of the range of Dioon, suggests that erotylids colonized Dioon from the Zamia lineage in three separate host shift events and that these host shifts have occurred relatively recently compared to the radiation of Allocorynina in Dioon. (VI) At least one recent host-shift of Pharaxonotha, originating from Zamia, have occurred onto Ceratozamia. A larger and wider sampling of Ceratozamia beetles may reveal more than one host shift. These coexist within Ceratozamia cones with the more ancient erotylids beetles discussed in hypothesis I, so that now two disparate lineages of Pharaxonothinae coinhabit Ceratozamia cones. This host-shift radiation is allied with those in Dioon, suggesting important watershed periods in cycad evolution when exchange of pollinators occurred among cycad genera in the New World.
Deeper study of these periods may be crucial in understanding the relatively recent resurgence of cycad evolution that have been proposed [62][63][64]. (VII) The pattern of population genetic variation of Pharaxonotha beetles that presumably pollinate Zamia in Puerto Rico and Hispaniola mirrors to a great extent the population genetic variation exhibited by the Zamia on those islands [65]. These mirroring patterns suggest that the mobility and/or abundance of a cycad's pollinator may influence gene flow in its host cycad and consequently the speciation pattern of its host. For example, observations [66] suggests that unlike other cycad beetles, this lineage of Pharaxonotha is highly sensitive to human disturbance of its vegetative habitat and easily becomes rare or locally extinct as a result. This susceptibility to disturbance and the low ability to recolonize its host from nearby populations suggests low mobility and low ability to mediate gene flow in its host Zamia. The resulting effect is to produce local reproductive isolation of cycad populations that appear on casual observation to have continuous distributions. (VIII)Based on population genetic variation of beetles discussed in hypothesis VII, the Zamia populations near Bayamon and Toa Baja, Puerto Rico may be conspecific with Zamia pumila in Hispaniola; furthermore, Z. pumila populations in Hispaniola may be recent colonists from an ocean dispersal event originating from Bayamon and Toa Baja.

Independent Tests of Beetle Generated Hypotheses
Many of these beetle-generated hypotheses about cycad evolutionary patterns in the New World may be tested with data sets of genetic and/or morphological characters from their host Ceratozamia, Dioon, or Zamia. Our tree for erotylid beetles (Figure 4) contains samples from 61 populations of Zamia along with collection localities. A phylogenetic tree for Zamia based on sequences of one gene combined with morphological data published by Caputo et al. [67] is available for comparison. Their tree contains 22 species of Zamia, of which only ten correspond with populations in our tree, nevertheless, some broad tests can be made of our erotylid hypotheses II and IV. Their tree shows some congruence with the erotylid trees presented here. For instance, in their tree as in ours, the Caribbean Z. pumila group is sister to many Mesoamerican and South American Zamia. Their tree differs significantly from our erotylid tree, however, in that a large branch of the Central American Zamia forms a clade that is sister to the previously described clades. Our erotylid tree suggests that their Central American clade should form part of their Mesoamerican and South American Zamia clade. In addition, our erotylid tree suggests that Z. soconuscensis Schutzman, Vovides and Dehgan belongs in a clade that is an early diverging relative to the Caribbean clade, however, Z. soconuscensis appears as a member of their Mesoamerican and South American clade that is sister to the Caribbean clade. If both beetle and Zamia host phylogenies are accurate, we would conclude that host and symbiont beetles are not co-speciating in a perfectly parallel fashion and that host shifts have occurred at important junctures in both Zamia and beetle radiations. The converse may be true, that the beetle phylogenies do reflect the evolutionary patterns of their host Zamia accurately, and that a more extensive genetic data set for Zamia may be required for a more accurate comparison.
A maximum likelihood tree produced by Nagalingum et al. [68] using three genes is also available for testing of hypothesis IV. Their tree contains 35 species of Zamia, of which 17 correspond with host populations in our tree and is broadly congruent with ours. In their tree, the Caribbean clade is sister to other Mesoamerican and Central and South America clades. None of our identified early diverging host Zamia lineages, however, were sampled in their study, so hypothesis I cannot be tested with their dataset.
Hypotheses may be tested by comparing trees from different beetle groups that are cohabiting in the same hosts. For example, we can test hypotheses 3A-D generated using Allocorynina weevils that there are four species groups of narrow-leafed Dioon that inhabit four distinct geographic regions. Our erotylid tree, based on less extensive samples from Dioon than that for Allocorynina, exhibits three distinct clades for Dioon erotylids that largely correspond to the three Allocorynina clades in hypotheses 3B-D. The only exception is the erotylid beetle on D. caputoi, which suggest a single host shift has occurred between our proposed regions 3C and 3D. Also, no usable DNA was extracted for erotylids from region 3A, so no comparison can be made for hypothesized region 3A. A recently published phylogeny for Dioon based on DNA by Gutiérrez-Ortega et al. [49] also provides a test of some of the Allocorynina generated hypotheses. Their tree [49] supports Allocorynina hypothesis 1 that D. mejiae is one of the early diverging lineages of the genus and the geographic regions proposed in hypotheses 3A-B. Their tree, however, does not support the distinction between hypothesized Dioon regions 3C and 3D, and in their tree, the Dioon species of 3C and 3D form an integrated clade. Their tree also does not support our hypothesis 4, that D. edule and D. angustifolium north of the Trans-Mexican Volcanic Belt form a distinct clade from D. edule south of the belt. Another published phylogeny for Dioon based on combined molecular and morphological data by Moynihan et al. [48] supports our hypotheses 3A-D, except their tree shows D. caputoi as a distinct branch separate from the other four.
A test of erotylid hypothesis VII, that mobility of a pollinator may influence the population genetic structure of its host, can be partly tested with the beetles inhabiting Zamia in the six island groups of the Bahamas. Although the Allocorynina beetles sampled from these islands all display the same 16S rRNA haplotype, suggesting panmixia or alternatively recent introduction to the islands from a single source, the Pharaxonotha beetles on three of these islands, Eleuthera, Long Island, and Tiloo Cay (near Abaco), exhibit haplotypes distinct from the rest. The hosts of these three Pharaxonotha haplotypes, Z. angustifolia Jacq., Z. lucayana Britton, and Z. sp. "Tiloo Cay", possess distinctly narrower, broader, or more coriaceous leaflet phenotypes than Zamia on other Bahamian islands. Genetic analysis of Bahamian Zamia populations [69] supports the possibility of extended genetic isolation of the Long Island Z. lucayana population, but not of the Eleuthera and Tiloo Cay Zamia populations. In this case, genetic patterns in Pharaxonotha mirror the phenoptypic traits displayed by their host Zamia better than the genetic analysis of the Zamia populations themselves. Possibly, restricted gene flow mediated by the Pharaxonotha may have been masked by gene flow mediated by Allocorynina pollinators at a later stage, since both pollinators are present in these Zamia populations.
Hypothesis VII also predicts that low mobility of cycad pollinators may result in high genetic variation among cycad populations that are in relatively close proximity. Lazcano-Lara (66) demonstrated that successful pollination of Z. portoricensis Urb. in Puerto Rico decreases dramatically when a female plant is beyond 1.9 m of a male plant, suggesting that its sole beetle pollinator, Pharaxonotha portophylla Franz and Skelley, provides relatively ineffective long range pollination, and this appears to contribute to the high degree of genetic differentiation exhibited between neighboring Zamia populations on this island [65,66].

Conclusions
It is widely understood that insect pollination is a critical facet of cycad biology and conservation [30,31,70,71]. In addition to their importance in cycad reproductive biology, the morphological and genetic analysis of pollinators can provide evidence for supporting or refuting hypotheses about cycad taxonomy and biogeography. As more extensive phylogenetic studies of New World cycads become available, the hypotheses presented here can be tested in more depth. Our hypotheses and analyses can also be refined in future work with a broader sampling of mitochondrial as well as nuclear genes from cycad beetles and a wider geographic sample of beetles that corresponds more closely with the host cycads on which phylogenetic studies have been conducted. Also, analyses of beetles can be improved with divergence time estimates, which require fossils. Recently, putative fossil relatives in both Allocorynini and Pharaxonothinae have been described from amber deposits [72,73], however, with only crude age estimates ranging from Eocene to Miocene. Hypotheses about cycad evolution generated through study of their associated insects can be novel and unexpected, with examples above, and provide new insights into the evolutionary history of New World cycads. This first attempt at reciprocal illumination of New World cycads through study of their pollinators indicates that this is a fruitful avenue of endeavor.  (2) Elongate oval with irregular outline. Character state 2 is a synapomorphy for Rhopalotria subgenus Rhopalotria. (2) Away from margin. Spine location away from pit is a synapomorphy for Notorhopalotria. Character state 1 is a synapomorphy for the Notorhopalotria.