Although all insect mushroom bodies share a basic ground-plan, substantial morphological and presumably functional modifications are readily observed (9). Morphological variability of mushroom bodies often reflects species-specific behavioral ecology. In the social Hymenoptera (including the ants, bees, and wasps), the calyx is partitioned into afferent zones characterized by the dendrites of Kenyon cell subpopulations and by sensory input modality (6, 10, 11). Two of these calyx zones, the lip and the collar, receive olfactory and visual input, respectively. In ants, the relative size of the lip and collar vary with the size of their primary sensory input neuropil, the antennal and optic lobes, and also with the importance of each sensory modality in the life of the animal (12). Similar correlations between the periphery and higher processing centers have been observed in mammals. In the star-nosed mole, deemphasis of vision as a primary sensory modality has led to reduction of the eyes and visual cortex, whereas somatosensory cortex has greatly expanded to accommodate the processing needs of highly specialized somatosensory processing structures on the snout (13).

Distribution of anatomical features across more phylogenetically diverse insect taxa suggests independent origins, perhaps in response to similar adaptive pressures, in these lineages. This is dramatically illustrated by the distribution of single and double primary calyx neuropils. Single ovoid calyces are found in the firebrats, representative of the most basal extant insect order, and the flies, one of the most derived groups (14–17). Doubled, deeply cup-shaped calyces are observed in widely divergent groups, such as the hemimetabolous cockroaches and the holometabolous social Hymenoptera (9).

What are some potential causal mechanisms and adaptive significances of calyx doubling? At the cellular level, mushroom bodies of cockroaches and social Hymenoptera are the largest known, with ≈170,000–175,000 Kenyon cells per hemisphere (18, 19). Calyx volume is likely to be large relative to the rest of the brain in these insects, compared with those possessing fewer Kenyon cells. Spreading of the calyx into two separate neuropils and convolution of the synaptic neuropil may be a consequence of maintaining optimal surface area to volume ratios in large mushroom bodies. Such a mechanism has been proposed for the formation of cortical gyri and sulci (gyrification) in mammals, which is pronounced in groups with relatively large cortices such as primates and cetaceans (20, 21). Developmental studies provide additional support for a positive correlation among cortex volume, neuron number, and gyrencephaly in mammals. For example, decreased neuronal number is observed in human patients and mouse mutants with lissencephalic (smooth cortex) mutations (22), whereas increasing the number of neural precursors can generate an abnormally large number of gyri and sulci in the cortex of the mouse (23).

We tested the hypothesis that double mushroom body calyces, like cortical gyri and sulci, are associated with an increase in mushroom body size relative to single calyces in terms of Kenyon cell number, calyx neuropil surface area and calyx volume. We accomplished this by quantifying mushroom body architecture in scarab beetles (Coleoptera: Scarabaeidae), species of which may possess either single or double calyces (24), as illustrated in Fig. 1 A and B. We also asked whether calyx morphology in scarab beetles was associated with changes in its most prominent source of synaptic input, the antennal lobes (25). Finally, we mapped calyx structure according to phylogeny and feeding ecology in other Coleoptera to determine potential behavioral correlates of calyx diversity in insects.

Fig. 1. Single and double calyces of scarabaeid mushroom bodies. (A) Single calyx of Onthophagus hecate (Scarabaeinae). (B) Double calyx of Maladera castanea (Melolonthinae) (Scale bars: 20 μm.) (C–F) Average brain measurement ratios for beetles (more ...)

Insects and Histology. A total of 48 individual beetles representing six scarabaeid subfamilies and 11 different species (see Table 1, which is published as supporting information on the PNAS web site, for summary information on species identity, feeding ecology, calyx morphology, and sample sizes) were collected as adults in the Morgantown, WV, area or ordered from Hatari Invertebrates (Portal, AZ). The nonscarabaeids Geotrupes sp. (Geotrupidae), Trox sp. (Trogidae), Odontotaenius disjunctus (Passalidae), Sphaeridium sp. (Hydrophilidae) and Silpha sp. (Silphidae), Scarites subterraneus (Carabidae), Scaphinotus elevatus (Carabidae), and Harmonia axyridis (Coccinellidae) were also collected in or around Morgantown or obtained from Carolina Biological Supply Company (Burlington, NC). Insects were anesthetized with cold and brains dissected in physiological saline, followed by fixation in Carnoy's fixative at 20° C. Fixation was allowed to proceed for a duration ranging from 1 h 15 min to 1 h 45 min to keep tissue shrinkage between brains relatively constant. After fixation, brains were stored in 70% ethanol at 4°C overnight. The next day, dehydration through a graded series of ethanols (10 min each) was followed by clearing in xylene (2 × 10 min) and paraffin embedding in a frontal orientation. Embedded brains were sectioned at 10 μm on a rotary microtome and stained by using Cason's stain according to the protocol of Kiernan (26) to reveal brain architecture. Brains of both male and female beetles were analyzed, with the exception of antennal lobe glomerulus counts in which only male brains were used. The age of beetles sampled was not known, but adult mushroom body neurogenesis appears to be negligible or absent for these species (S.M.F. and J. Miller, unpublished data) and is thus unlikely to be a source of significant variation.

Volume and Cell Number Estimates. Measurements were made in one randomly selected hemisphere of each stained brain. Brain region volumes were calculated according to the Cavalieri method (27–29), which has been demonstrated to provide accurate estimations of volume from sectioned material by taking into account sampling frequency and section thickness. First, areal measurements of brain regions were directly measured from photomicrographs by using Zeiss axiovision 4 software (Zeiss, Oberkochen, Germany). Sampling frequency for these measurements was determined for each brain region [Kenyon cell body region, mushroom body calyx neuropil (excluding the pedunculus), antennal lobe neuropil, and total central brain] for each species to provide a volume estimate within 5% of that generated by measurements by using every 10-μm section. Typical sampling frequencies ranged from every second section (calyx neuropil surface areas and volumes in the smallest species) to every sixth section (total central brain volume in the largest species). Once sampling frequency was determined, the first section to be sampled in each brain was chosen by tossing a number of coins equal to the sampling frequency; the number of “heads” was the first section to be sampled after the appearance of the structure of interest.

Perimeter measurements of the calyx neuropil and diameters of individual Kenyon cell bodies were also measured from photomicrographs by using the Zeiss axiovision 4 software. Surface area of the synaptic neuropil of the calyx was calculated from perimeter measurements multiplied by section thickness. Sampling frequency for perimeter measurements was determined in the same manner as for volume estimations.

To calculate Kenyon cell number, 50–100 individual Kenyon cell soma were selected from throughout the Kenyon cell body region and their diameters measured by using the axiovision 4 software. An average Kenyon cell body diameter was calculated from these data and subsequently used to calculate an average Kenyon cell body volume for that brain. The total volume of the Kenyon cell body region was then divided by the average individual Kenyon cell body volume to provide an estimate of Kenyon cell number for that brain.

Finally, the total number of antennal lobe glomeruli was counted from serial sections of male brains only, to control for the likely presence of sex-specific glomeruli.

Measurement calculations and graphs were generated by using Microsoft excel x for Mac (Microsoft, Redmond, WA), then exported to jmp software (SAS Institute, Cary, NC) for statistical analyses. Measured calyx surface areas were compared with calculated calyx surface areas, the latter derived from the measured calyx volume and the equation for the surface area of a sphere (= 4π r²). Brain measurement ratios for species with single vs. double calyces were tested for significant differences by using one-way ANOVA. All statistical comparisons were considered significant at P ≤ 0.05.

Double (“gyrencephalic”) calyces in scarab beetles were positively associated with a significant increase in calyx volume relative to total central brain size when compared with single (“lissencephalic”) calyces (Fig. 1C; one-way ANOVA F = 124.999, P < 0.0001). Kenyon cell number was also dramatically increased in gyrencephalic mushroom bodies (Fig. 1D; one-way ANOVA F = 79.520, P < 0.0001). The measured surface areas of lissencephalic calyces were nearly identical to those predicted for spheres of the same volume, resulting in a measured to predicted surface area ratio of ≈1 (black bar, Fig. 1E). In contrast, the ratio of measured calyx surface area to predicted calyx surface area was significantly larger for gyrencephalic calyces (white bar, Fig. 1E; one-way ANOVA F = 46.684, P < 0.0001). Surface area to volume ratios, however, were significantly higher in lissencephalic calyces (Fig. 1F; one-way ANOVA F = 51.556; P < 0.0001). This suggests that, although gyrencephaly increases the surface area of the calyx neuropil, it has not kept pace with the expansion in calyx volume in scarabaeids.

We discovered that calyx morphology was tightly linked to feeding ecology; lissencephalic calyces were always observed in scarabaeid species belonging to dung-feeding (coprophagous) subfamilies, whereas gyrencephalic calyces were always observed in plant-feeding (phytophagous) subfamilies (Fig. 2; also see supporting information) (30). Phytophagous scarabs are highly generalist, feeding on the foliage, flowers, and fruits of a wide range of host plants. By contrast, coprophagous scarabs and species belonging to related nonphytophagous taxa are relatively more specialized in their feeding habits and exhibited lissencephalic calyces. The wood-feeding Passalidae, however, were exceptional in possessing somewhat enlarged and flattened calyces, perhaps reminiscent of an early stage in calyx splitting. Nevertheless, the overall distribution of calyx morphologies across the taxa sampled strongly supports the inference that robustly gyrencephalic calyces evolved in accompaniment with a derived and highly generalist feeding ecology in the Scarabaeidae.

Fig. 2. Distribution of calyx morphologies and feeding ecologies across the Coleoptera. The phylogenetic tree of the Scarabaeoidea is after ref. 30; the remainder of the coleopteran tree is after ref. 49. The family Scarabaeidae is considered a monophyletic grouping. (more ...)

Additional comparisons of scarab calyces and their major sensory input source, the antennal lobes, provided further insight into the potential behavioral significance of large mushroom bodies in generalist scarabs. We found that antennal lobe volume relative to total central brain volume was actually larger in beetles with lissencephalic calyces than in those with gyrencephalic calyces (Fig. 3A; one-way ANOVA F = 12.306, P = 0.001), and there was no difference in the number of antennal lobe glomeruli in males (Fig. 3B; one-way ANOVA F = 0.048, P = 0.829). This suggests that calyx gyrencephaly is unlikely to be associated with greatly increased numbers of antennal lobe neurons projecting to the calyx.

Fig. 3. Differences in calycal circuitry in single and double calyces. (A–D) Average brain measurement ratios for beetles with single calyces (black bars) versus double calyces (white bars). (A) Antennal lobe volume relative to central brain volume. ( (more ...)

Instead, our data suggested that individual afferent neurons might form synapses with more Kenyon cells in the mushroom bodies of generalist species. Generalist scarabs had nearly four times as much calyx volume per unit of antennal lobe volume compared with coprophagous scarabs (Fig. 3C; one-way ANOVA F = 133.017, P < 0.0001) and significantly more Kenyon cells per unit of calyx volume (Fig. 3D; one-way ANOVA F = 4.883, P = 0.032). This latter measurement was supported by high-magnification images of calycal microglomeruli, sites of convergence of Kenyon cell dendrites, and extrinsic inhibitory processes about a single antennal projection neuron terminal (31). Microglomeruli were smaller and more densely packed in the calyces of generalist scarabs (Fig. 3 E and F). Calycal subcompartments were also observed in gyrencephalic mushroom bodies (Fig. 3E), perhaps indicating the segregation of olfactory processing inputs or acquisition of other afferent modalities such as vision or gustation.

Our findings suggest that gyrencephalic calyces in scarabaeid beetles, and perhaps other lineages such as the cockroaches and the social Hymenoptera, have arisen in conjunction with an evolutionary increase in Kenyon cell number, neuropil volume, and neuropil surface area. Similar trends are observed in mammals and are particularly pronounced in lineages such as primates, in which increased size of the cerebral cortex relative to total brain size is strongly correlated with an increase in gyral complexity (20, 21). Nonhomologous brain centers like the mushroom bodies and the cerebral cortex can thus adopt strikingly similar architectures when shared principles of cellular organization are acted on by similar selective pressures.

Gyrencephalic calyces are not associated with a concomitant increase in the volume of the antennal lobes, suggesting that the evolution of large mushroom bodies in highly generalist phytophagous beetles was not driven by the acquisition of inputs from more olfactory afferent neurons. Instead, gyrencephalic calyces may represent an adaptation for enhancing the computational capacity of the mushroom bodies in three ways: by increasing the number of inputs from individual afferent neurons, by subcompartmentalization for segregation of specialized functions, and by providing additional substrate for complex learning computations and the storage of memories. Regarding the first possibility, modeling studies have proposed that small groups of antennal projection neurons converge combinatorially on populations of Kenyon cells to form odor-discriminating “functional subunits” (32). Increasing the number of Kenyon cells would thus generate a greater capacity for combinatorial synaptic inputs from the antennal lobe and the potential to discriminate among larger numbers of odors. Regarding the second possibility, calycal subcompartments like those observed in generalist scarabs are found in the social Hymenoptera (6, 11), where they are characterized by distinct afferent modalities and Kenyon cell populations. This organization suggests specialization of functions that may than be selectively integrated via combinatorial outputs from the mushroom body lobes (11). Subcompartmentalization coincident with evolutionary expansion of higher brain regions is also evident in mammals, as exemplified by the elaborate functional subpartitions in the enlarged visual cortex of primates (33). Regarding the third possibility, unique cognitive capabilities of honey bees, such as perception of generalized and abstract rules about multiple stimuli in associative learning paradigms (34, 35), suggest a role for the gyrencephalic mushroom bodies of these and perhaps other insects in enabling complex learning and memory functions.

Sociality has been proposed to drive the evolution of large higher brain centers in both insects and vertebrates (36, 37). In insects, gyrencephalic calyces and large mushroom bodies are observed in cockroaches, scarab beetles, and social hymenopterans. Cockroaches, although gregarious, are not considered social at the level of ants, bees, and wasps. Elaborated mushroom body lobes are found in the eusocial termites, but the calyces appear to be reduced and fused relative to those of their sister taxon, the cockroaches (38, 39). Similarly, gyrencephalic calyces are observed in basal nonsocial Hymenoptera such as sawflies (40). Sociality, therefore, does not appear to be a robust predictor of calyx morphology.

In scarabaeids, calyx morphology was uniformly associated with feeding ecology. We propose that large gyrencephalic calyces have evolved in response to selective pressure for behavioral flexibility that would be imparted by the above neural properties and that would be beneficial to a generalist feeding ecology in a long-lived mobile insect. For example, the adult Japanese beetle Popillia japonica is long-lived (4–6 weeks) and feeds on the tissues of >300 different host plant species (41), which it locates by using olfactory cues such as host plant volatiles induced by feeding damage (42) and visual cues (43). Despite this wide host range, Popillia clearly prefers certain plant species while rarely feeding on others (44), suggesting that these insects are actively discriminating and choosing among many potential food sources by using additional sensory cues. A long-lived insect may also need to integrate this information with that regarding food source handling, location, and availability, storing this knowledge for future encounters. Cockroaches and social Hymenoptera are similarly long-lived and flexible in their feeding behaviors and have arrived at a similar neural solution.

In contrast, dietary specialists using a few highly specific cues to rapidly and unambiguously detect one or a closely related group of host species would be expected to have fewer requirements for sensory discrimination and integration and to have a more hard-wired capacity for handling the host plant. These insects would not be predicted to undergo such strong selection for large mushroom bodies. Our results support this prediction, and we suggest that the mushroom bodies are an important component of the “larger and more sophisticated nervous system” (45) that is a hypothesized requirement for decision-making efficiency in insects with generalist feeding ecologies (46).

The hypothesis that cognitive adaptations for generalist phytophagy played a key role in the evolution of gyrencephalic calyces would equally predict that long-lived mobile generalist feeders of any kind might display a similar evolutionary trend. Generalist and specialist carnivorous beetles illustrate that this is indeed the case (Fig. 4). The generalist predator Scarites subterraneus possesses a gyrencephalic tripartite calyx capped by a dorsally protruding Kenyon cell body region (Fig. 4A). In contrast, two specialist predators, Scaphinotus elevatus (Fig. 4B) and Harmonia axyridis (Fig. 4C), possess lissencephalic calyces topped by smaller aggregates of Kenyon cell bodies.

Fig. 4. Differences in calyx structure in generalist and specialist predators. (A) Tripartite gyrencephalic calyx (brackets) of the generalist predator Scarites subterraneus (Carabidae). (B) Lissencephalic calyx of the carabid Scaphinotus elevatus, a specialist (more ...)

Despite these robust results, feeding ecology is unlikely to be equally predictive of mushroom body morphology across the highly diverse insects. For example, it is unknown how quickly changes in the brain can track changes in behavior. At the gross anatomical level, it is possible that an evolutionarily recent switch in feeding ecology may not produce measurable differences in mushroom body morphology when compared with related species retaining the ancestral feeding ecology. Developmental constraints may need to be overcome before significant enlargement or diminishment of the mushroom bodies could occur. Additionally, other complex behaviors such as social interaction and spatial navigation, also require advanced capabilities for sensory integration, discrimination, learning, and memory. Insects displaying these behaviors might be expected to exhibit calyx gyrencephaly regardless of feeding ecology.

In vertebrates such as primates and birds, large and typically gyri-rich cerebral cortices are correlated with the capacity for behavioral innovation (particularly in terms of novel food source usage) and success in invading novel environments (36, 47). Polyphagous insects are also more likely to successfully establish and maintain populations when colonizing new environments (48). The worldwide ubiquity of cockroaches, social Hymenoptera, and phytophagous scarabs, many of which are economically important invasive pests due to their colonization success, speaks to a similar role for large gyrencephalic mushroom bodies. Behavioral innovation is therefore likely to be a driving factor in the evolution of large higher processing centers in the brains of both invertebrates and vertebrates.

We thank Ms. Marissa Smith for assisting with data analysis, Drs. Ronald Bayline and Kevin Daly for fruitful discussions, and Drs. Susan Fahrbach and Gene Robinson for comments that greatly improved this manuscript. N.S.R. was supported by Howard Hughes Medical Institute Undergraduate Science Education Program Grant 52002683, awarded to Washington and Jefferson College. S.M.F. was supported by West Virginia University.