Size constraints and sensory adaptations affect mosaic brain evolution in paper wasps (Vespidae: Epiponini)

Abstract

Body size limits brain volume, but size may also differentially constrain the volumes of brain regions. Size variation and compartmentalized brains of swarm-founding paper wasps (Epiponini) make them good models for the study of size-related brain allometry. We analysed the relative volumes of brain regions that process different sensory inputs: the optic lobes (OL, vision), the antennal lobes (AL, chemosensation) and insect forebrain neuropils (MB, mushroom bodies). Species varied 25-fold in head capsule volume. Total brain volume increased with body size. This brain size increase was a non-linear decelerating function of size: relative brain volume was highest for the smallest-bodied species. In contrast, relative MB and AL volumes declined sharply at the smallest wasp body sizes, but relative OL size did not. Different brain regions showed distinct size-related allometries among Epiponinae. Such brain allometry differences should be accounted for in comparative analyses of brain evolution. However, there was significant deviation from size-typical brain architecture in the nocturnal genus Apoica: A. pallens had reduced relative investment in OL but greater relative investment in MB visual processing regions (calyx collars). This suggests that body size effects on brain region investment can be overcome by selection on particular sensory systems when species evolve to occupy novel sensory environments.

brain evolution, comparative analysis, mosaic evolution, mushroom bodies

INTRODUCTION

Investment in central nervous system tissue, particularly in the brain, is probably constrained by a balance between strong positive and negative natural selection. Increases in brain investment can be favoured by selection for enhanced cognitive and behavioural capacities because greater investment in brain tissue is related to neural processing power (Roth & Dicke, 2005; Bullmore & Sporns, 2012; Kotrschal et al., 2013; Herculano-Houzel, 2017). However, tissue costs prohibit excess brain growth because neural tissue has some of the highest developmental and metabolic costs throughout the animal kingdom (Niven & Laughlin, 2008).

Although total brain volume generally increases with body size within animal clades, as the lower body size limits are approached, the rate of decrease in brain volume often slows sharply (Polivov & Makarova, 2017). This pattern, known as Haller’s rule, suggests there are clade-specific minimum brain sizes necessary for basic cognitive function (Wehner, Fukushi & Isler, 2007; Seid, Castillo & Wcislo, 2011; Feinerman & Traniello, 2016; O’Donnell & Bulova, 2017).

Here we suggest that total brain volume per se may be an incomplete measure of body size-related allometry of the central nervous system. When analysed in isolation, total brain volume may mask important patterns of brain allometry. Brains do not function as homogeneous units: particular cognitive neural functions are compartmentalized among anatomically distinct brain regions. The relative volumes of distinct brain regions can evolve independently, a pattern known as mosaic brain evolution (Barton & Harvey, 2000). Numerous vertebrate and invertebrate examples demonstrate matches between the relative volume of brain regions and their importance to cognitive function, given species’ ecology (Smaers & Soligo, 2013; O’Donnell et al., 2013; Gutiérrez-Ibáñez et al., 2014). Evolutionary shifts in brain region volume can involve both changes in cell (neuron) numbers and changes in cell size, both of which may affect cognitive processing power (Herculano-Houzel, Manger & Hass, 2014).

An important but largely unaddressed question is whether body size evolution differentially affects functionally distinct brain regions. To address this, we investigated whether species brain region size relationships covary with body size in a clade of eusocial wasps: the tribe Epiponini (Noll, Wenzel & Zucchi, 2004). We took advantage of the wide range of body sizes in this tribe of eusocial wasps; the subject species we sampled ranged from among the smallest to some of the largest-bodied epiponines (Richards, 1978). Our comparative analysis tested whether species-typical brain architecture covaried with body size. We used epiponine phylogenies (Wenzel & Carpenter, 1994; Carpenter, Kojima & Wenzel, 2000) to estimate the possible confounding effects of species relatedness on brain allometry. We measured the volumes of the wasps’ head capsules (as an index of body size) and the volumes of functionally distinct brain regions (Gronenberg, 2001; O’Donnell & Bulova, 2017), focusing on brain regions that process visual information (optic lobes, henceforth OL) and chemosensory information (antennal lobes, henceforth AL). We also analysed allometry of the mushroom bodies (henceforth, MB). The MB are neuropils of the insect forebrain involved in sensory integration, learning and memory (Fahrbach, 2006; Menzel, 2014). MB neuropil volumes are developmentally plastic in paper wasp adults and respond to changes in behavioural experience (Ehmer, Reeve & Hoy, 2001; O’Donnell et al., 2014). The MB include distinct neuropils for processing vision (the collar region of the MB calyx) and olfaction (the lip region of the MB calyx). Volumes of the MB calyx neuropils vary with species-typical behaviour and ecology in paper wasps (Gronenberg, Ash & Tibbetts, 2007; O’Donnell, Clifford & Molina, 2011; O’Donnell et al., 2015). However, previous studies did not include measures of species body size differences, and therefore could not assess the possible effects of size-related allometry on brain structure.

We also investigated whether major shifts in species ecology were associated with deviations from size-related brain allometry. The evolution of nocturnal foraging activity in the epiponine genus Apoica (Pickett & Wenzel, 2007) was associated with decreased investment in peripheral visual-processing brain regions (OL; O’Donnell et al., 2013). We tested whether the relative sizes of Apoica peripheral and central (MB) processing regions differed from expectations based on body size.

MATERIAL AND METHODS

Taxonomic coverage

Data were collected on 19 species of Neotropical swarm-founding wasps (Epiponini). We sampled one species per genus from 11 of the 19 epiponine genera (Fig. 1; Wenzel & Carpenter, 1994) except in the case of Polybia which has been resolved at the subgenus level; we sampled nine species of Polybia with a maximum of two species in each of seven of the 11 Polybia subgenera (Carpenter et al., 2000).

Figure 1.

Phylogenetic relationships among the 19 epiponine wasp species analysed, with generic relationships from Wenzel & Carpenter (1994) and Polybia subgeneric relationships from Carpenter et al. (2000). For the genus Polybia (P.), sub-generic names are indicated. The sizes of circles above each species are proportional to the mean log head capsule volume for that species (the symbol for Protopolybia exigua is too small to be visible).

Specimen collection

Wasps were field-collected into and stored in buffered aldehyde-based fixative (Prefer fixative, Anatech Ltd) for at least 2 months until histological processing. Species (number of individuals quantified), collection dates and locations were: Polybia (Apopolybia) jurinei (N = 9): November 1994, Ecuador, 0°40.5′S, 76°25.8′W; Agelaia xanthopus (N = 2), Polybia (Pedothoeca) emaciata (N = 4): August 2006, Costa Rica, 10°18.1′N, 84°47.9′W; Nectarinella championi (N = 5), Polybia (Trichinothorax) raui (N = 7), Synoeca septentrionalis (N = 5): August 2006, Costa Rica, 10°14.4′N, 84°54.3′W; Apoica pallens (N = 3), Angiopolybia zischkai (N = 3), Charterginus fulvus (N = 5), Leipomeles dorsata (N = 3), Parachartergus smithii (N = 5), Polybia (Cylindroeca) dimidiata (N = 3), Polybia (Furnariana) richardsi (N = 10), Protopolybia exigua (N = 3): June 2007, Ecuador, 0°40.3′S, 76°24.0′W; Polybia (Trichinothorax) flavitincta (N = 6): March 2012, Costa Rica: 10°25.6′N, 84°1.2′W; Brachygastra smithii (N = 5), Polybia (Myrapetra) aequatorialis (N = 6), Polybia (Myrapetra) plebeja (N = 5), Polybia (Formicicola) rejecta (N = 5): July 2012, Costa Rica 10°16.3′N, 84°49.4′W. In total, 94 wasps were collected.

Histology and neuroanatomy

We cut the wasps’ head capsules from the thorax and removed the mandibles and antennae to improve resin penetration of the tissues. For all species except Synoeca septentrionalis, we embedded and sectioned the head capsule (Fig. 2). For Synoeca, we dissected the brains out of the head capsules then processed them using the same protocol as above (Fig. 2). We dehydrated the heads through a series of increasing ethanol concentrations, 100% acetone and then through increasing concentrations of plastic resin. Resin composition was as follows: 5.5 g EMbed 812 [a mixture of bisphenolA/epichchlorohydrin epoxy resin (CAS #25068–38-6)] and epoxy modifier (CAS #2425–79-8)), 5.7 g dodecenyl succinic anhydride, 0.65 g dibutyl phthalate and 0.31 g 2,4,6-tri(dimethylaminoethyl)phenol. We incubated individual wasp heads in 0.1 mL resin in pyramid moulds at 60 °C for 72 h, then glued the hardened resin to 0.5 mL acrylic cylinders with cyanoacrylate adhesive. We cut each head along the frontal plane into 12- to 16-μm-thick sections (depending on species) using a rotary microtome with disposable steel histology blades. Sections were mounted on gelatin-coated microscope slides and the tissue was stained with toluidine blue. We cleared the stained sections in a series of increasing ethanol concentrations and cover slipped them under transparent mounting medium.

Figure 2.

Same-scale light photomicrographs of stained and sectioned brain tissue from three species of swarm-founding paper wasps (Epiponinae). Brains are shown in frontal view. The 1-mm scale bar refers to all three species. A, Protopolybia exigua (smallest species sampled). B, Polybia raui. C, Synoeca septentrionalis (largest species sampled). In C (S. septentrionalis), some major brain regions are indicated (AL, antennal lobes; MB, mushroom bodies; OL, optic lobes).

We photographed the tissue sections at 2560 × 1920-pixel resolution with a compound microscope-mounted digital camera, using 2.5× or 5× microscope objectives (depending on species; Fig. 2). For each wasp we began photographing every other section at the section where brain tissue first became visible. The digital imaging analysis software ImageJ version 1.46 (http://rsbweb.nih.gov/ij/) was used to quantify the volumes of brain structures. We outlined the target brain regions and quantified the number of image pixels in the structure using ImageJ, and then converted the pixel counts to area using a photograph of a stage micrometer taken at the same resolution and magnification as a size reference. We multiplied the areas by section thickness to yield estimated volume in mm³.

Neuroanatomical variables for analysis

‘Total brain volume’ was the sum of volumes of the following regions: OL (medulla and lobula layers only; the lamina layer was difficult to distinguish from the compound eyes in the histological sections and could not be scored reliably), AL glomeruli, MB, central complex, protocerebral mass and suboesophageal ganglion (the suboesophageal ganglion was included because its boundaries with the protocerebral mass are unclear; it is relatively small in volume). For some analyses the central complex, protocerebral mass and suboesophageal ganglion were pooled into ‘central brain volume’ and used as an index of overall brain size that was independent of (i.e. did not include) other structures of interest (such as the MB). We separately measured the volumes of the subregions of the MB: the olfactory-processing calyx lip neuropils, the visual-processing calyx collar neuropils and the axonal bundles (peduncles and lobes). In most analyses the total MB volume was analysed, but for some analyses the MB calyx collar and lip volumes were analysed separately. Neuroanatomical data are presented in Supporting Information Table S1.

Head capsule measurements as an index of body size

Histological processing prevented us from collecting head-size measurements on the neuroanatomy subjects. We measured head capsule sizes for the subject species on different individuals taken from the same colonies as the neuroanatomy subjects. We measured head sizes on six individuals of each species (except P. flavitincta, N = 1; P. dimidiate, N = 12); whenever possible we sampled equal numbers of queens and workers. We dissected the wasp’s head capsule from the body at the foramen (narrow attachment point to the alitrunk, or ‘neck’). We photographed each head using a digital camera mounted on a dissecting scope, and used the ruler tool in ImageJ and photographs of a stage micrometer to convert pixels to mmmillimetres. Heads were photographed in frontal view with the foramen area facing away from the camera and resting against a horizontal glass surface. We measured head width at the widest point and head height from the centre of the clypeus to the vertex, and we used ½ head width as an approximation of head depth. We then estimated head capsule volume for each individual using the formula for an ellipsoid:

$$\text{4}/\text{3}\pi \times \mathrm{½}\text{ head width}\left(\text{mm}\right)\times \mathrm{½}\text{ head height}\left(\text{mm}\right)\times \mathrm{½}\text{ head depth}\left(\text{mm}\right).$$

We used mean estimated head volumes of each species as an index of species-typical body size in the statistical analyses.

We collected wing length data on eight of the subject species as an alternative index of body size. We measured the length of the rigid vein along the leading edge of the front pair of wings (costa vein) for five or six individuals per species. We dissected wings from the thorax and mounted them flat on microscope slides with transparent tape, and photographed the wings through a dissecting scope with a digital camera. We used the ruler tool in ImageJ, calibrated to a photograph of a 2-mm stage micrometer, to measure costa vein length. We measured both forewings and averaged their lengths as a body size index for each wasp.

Statistical analyses

Analyses were performed on species mean head capsule and brain volume data. Parametric regression analyses were performed in Sigmaplot v.12.5 software unless otherwise noted. For the non-linear allometric relationships of relative brain region volume with head capsule volume we calculated the best-fit inverse first-order regressions (y = y₀ + (a/x)). We tested whether Apoica were significantly different from other taxa by first calculating the residuals from the linear regression of the target brain region (optic lobes or MB calyx collar) against central brain volume for all species. We then analysed the distributions of the residuals for significant outliers using the non-parametric stem-and-leaf analysis in SPSS v.24.

We tested for possible effects of storage time in fixative on tissue structure by calculating the best-fit linear regression relationship of time (in months) elapsed from specimen collection to specimen embedding with two key brain structure ratios: the ratio of total MB volume to head capsule volume, and the ratio of total brain volume to head capsule volume. To test for size-related differences in tissue shrinkage or degradation, we analysed whether within-species (individual) differences in relative MB size were associated with overall size variation. We used total brain volume as a size surrogate for this analysis because were not able to obtain individual head-size measurements for the neuroanatomy subjects. First, we confirmed that total brain volume was a good surrogate for head-capsule size by testing whether species mean relative MB volumes had a similar relationship with both head volume and total brain volume. We then used multiple regression (in SPSS) to test whether within-species individual size differences predicted relative MB volume, and we used the species × brain size interaction term to test whether species varied in the strength of this relationship.

We used generic and subgeneric phylogenies of the Neotropical swarm-founding Epiponini for analyses estimating the effects of species relatedness on brain architecture (Fig. 1; Wenzel and Carpenter, 1994; Carpenter et al., 2000). We sampled one species per genus except in the case of Polybia which has been resolved at the subgenus level; we sampled nine species of Polybia with a maximum of two species in each of seven of the subgenera. Phylogenetic generalized least squares (PGLS) analyses of species mean data were conducted with Compare v.4.6b software (Martins, 2004). PGLS analyses require normally distributed data; the allometric species mean data were log-transformed prior to performing PGLS regression analyses in Compare.

RESULTS

Testing for methodological confounds and confirming validity of size measures

No effects of duration of fixation

There was no evidence for changes in brain structure related to tissue degradation or fixation time. Time of fixation had no measurable effect on tissue structure or brain architecture. The number of months elapsed from initial fixation to embedding for histology was not related to the ratio of MB to central brain volume nor to the ratio of total brain volume to head capsule volume (in both cases r² = 0.003, d.f. = 17, P = 0.82; see Supporting Information Figure S1).

No size-related biased tissue shrinkage

There was no evidence for size-related tissue shrinkage or degradation. Relative MB volume showed a similar pattern of positive but decelerating increase with total brain volume (see Supporting Information Figure S2; inverse first-order regression, r² = 0.66, d.f. = 17, P < 0.001) and with head capsule volume (Fig. 3), suggesting total brain volume was a reasonable surrogate for head capsule volume. Within-species individual relative MB volumes were not significantly related to brain size (F_1,114 = 0.20, P = 0.66). The non-significant brain size effect on relative MB volume was consistent across species (interaction term, F_18,114 = 1.19, P = 0.28).

Figure 3.

Species mean relative volumes of peripheral sensory brain lobes (top panel) and mushroom body calyx neuropils (bottom panel) plotted against head capsule volume for 19 species of swarm-founding paper wasps (Epiponinae). The linear regression best-fit line is indicated for the optic lobes plot in the top panel; all other regression lines are best-fit lines for inverse first-order regressions. The significant outlier data points for the nocturnal wasp genus Apoica are indicated on the optic lobe and mushroom body collar plots. Black symbols: Apoica. Blue symbols: Polybia species. Open symbols: all other genera. Note the break in the y-axis scale in the top panel.

Head capsule size reflects body size

Head capsule volume was strongly correlated with wing (costa vein) length for eight of the subject species (r = 0.95, N = 8, P < 0.001; see Supporting Information Figure S3). Subject species varied widely in head capsule volume, from a minimum species mean volume of 0.88 mm³ (Protopolybia exigua) to a maximum of 22.34 mm³ (Synoeca septentrionalis), a greater than 25-fold volume difference (Figs 1 and 2). Head capsule volume showed little apparent correspondence to phylogenetic relatedness (Fig. 1).

Size-related mosaic brain allometry

Species mean total brain volume increased with head capsule size (Fig. 4; linear regression: r² = 0.86, d.f. = 17, P < 0.001), but the relationship was not isometric: size-related brain volume (brain volume/head capsule volume ratio) increased with decreasing body size, in accordance with Haller’s rule (Fig. 4; inverse first-order regression r² = 0.77, d.f. = 17, P < 0.001). Relative MB volume had a different, converse relationship with head size: relative MB size decreased rapidly with decreasing head size. This pattern held for total MB volume (inverse first-order regression r² = 0.67, d.f. = 17, P < 0.001) and for both the visual processing collar neuropil (Fig. 3; inverse first-order regression r² = 0.46, d.f. = 17, P < 0.005) and the olfactory processing lip neuropil (Fig. 3; inverse first-order regression r² = 0.65, d.f. = 17, P < 0.001). The positive relationship between relative MB volume and head capsule volume held after accounting for phylogenetic effects (PGLS regression of log-transformed data, r² = 0.75, d.f. = 16, P < 0.001). The relative size of the olfactory processing AL also decreased at smaller head sizes (Fig. 3; inverse first-order regression r² = 0.32, d.f. = 17, P = 0.01), but this pattern did not hold for the visual processing OL (inverse first-order regression NS, r² = 0.07, d.f. = 17, P = 0.29).

Figure 4.

Absolute total brain volume (top panel) and body-size relative total brain volume (bottom panel) plotted against head capsule volume for 19 species of swarm-founding paper wasps. Top panel: the best-fit linear regression relationship is indicated by the dashed line. Bottom panel: the best-fit inverse first-order regression relationship is indicated by the dashed line. For both plots the data points for nine Polybia species are indicated in blue, the data point for nocturnal Apoica pallens is shaded black and other species are indicated by open symbols.

Deviation from size allometry in a nocturnal species

The relationship between relative MB volume and head capsule volume suggested the nocturnal species Apoica pallens had an exceptionally high relative MB volume for its body size, and an exceptionally low relative OL volume for its size (Fig. 3). For both of these the regression residual for Apoica was identified as the single significant outlier by non-parametric stem-and-leaf analyses.

DISCUSSION

Body size constrained total brain size in epiponine wasps: brain size increased with body size, but the increase was not isometric. The pattern we observed of a steep increase in relative brain size as the lowest body sizes were approached is common to many animal taxa, and is referred to as Haller’s rule. Although the mechanisms behind this pattern are not known, it suggests there is a minimal brain size necessary to support basic cognitive and behavioural functions within a taxon (Eberhard & Wcislo, 2011; O’Donnell & Bulova, 2017).

In this study we went beyond examination of total brain volume and investigated whether particular brain regions showed distinct body size-related allometries, as might be expected under mosaic brain evolution (Barton & Harvey, 2000). In marked contrast to total brain volume, brain-relative MB and AL size decreased rapidly at the smallest body sizes; in other words, MB and AL volumes continued to decline with body size in small wasp species, while the absolute volume of the rest of the brain remained relatively constant. This pattern was robust to phylogenetic correction.

The cognitive–functional implications of mosaic brain allometry in epiponines are unknown. The MB are major centres of complex cognition in insects, involved in sensory integration, learning and memory, and spatial orientation (Fahrbach, 2006; Menzel, 2014; Aso et al., 2014). MB size increases in social insects at the individual level (due to developmental plasticity) and evolutionarily among species, corresponding with cognitive challenges that include diet breadth, foraging behaviour and social interactions (Ehmer et al., 2001; Farris, Robinson & Fahrbach, 2001; O’Donnell, Donlan & Jones, 2004; Farris & Roberts, 2005; Molina & O’Donnell, 2007). Limitations on neuron size and connectivity suggest the evolution of small brains may reduce the cognitive or behavioural capacities of small- bodied species (Niven & Farris, 2012), but there is little empirical support for this idea (Eberhard, 2011; Seid et al., 2011). While it is possible that smaller-bodied epiponine wasps possess simpler behavioural repertoires than larger-bodied species, we know of no evidence for such differences. The large colonies, complex caste differentiation and sophisticated nest architecture of small-bodied Polybia and Protopolybia species suggest this is not the case (Wenzel, 1991; Kelstrup et al., 2014). Potential cognitive deficits related to reduced MB and AL investment in small-bodied epiponine wasps await empirical testing.

Our findings suggest it is important to take body size into account in interspecific comparisons of investment in MB and other brain regions. Size-related evolutionary shifts in mosaic brain structure can complicate the interpretation of comparative analyses of brain allometry, and this problem may apply more generally to the comparative analysis of physiological allometries (Uyeda et al., 2017). Other insect clades should be examined to test the generality of the decrease in relative MB volume with body size. A recent survey across a wide range of social hymenopteran taxa concluded relative MB investment does not decrease with body size universally in the order (Muscedere et al., 2014). However, size-related brain allometries could occur in other brain regions in different hymenopteran taxa.

Our data suggest extreme changes in sensory/cognitive environments can overcome body-size constraints on brain structure. We found major changes in brain structure were associated with the evolution of nocturnality in the genus Apoica, and Apoica brain architecture deviated strongly from body-size-related allometry predictions. As suggested previously (O’Donnell et al., 2013), Apoica peripheral visual-processing OL were significantly smaller than expected. In marked contrast, the central-brain visual-processing MB calyx collar neuropils of Apoica were significantly larger than expected based on body-size allometry. The oversized MB calyx collars of Apoica indicate specialized neural mechanisms for visual processing under low-light conditions may be housed in the MB (Warrant, 2008). These findings suggest strong natural selection on brain function related to species shifts into novel ecological niches, with new cognitive challenges, can drive differential investments in brain regions that break the constraints imposed by body-size allometry (Moore & DeVoogd, 2017).

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article at the publisher’s website:

Figure S1. Species mean relative MB volume plotted against time (months) elapsed from collection to embedding of neuroanatomy specimens from 19 species of swarm-founding paper wasps. Dashed line indicates linear regression best-fit relationship.

Figure S2. Species mean relative MB size plotted against total brain volume for 19 species of swarm-founding paper wasps. Data points for nine Polybia species are indicated in blue, data point for nocturnal Apoica pallens is shaded black, and other species are indicated by open symbols.

Figure S3. Species mean head capsule volume plotted against sepecies mean wing length for species of swarm-founding paper wasps. Dashed line shows linear regression best-fit.

Table S1. Volumes of brain regions for individual epiponine paper wasps.

ACKNOWLEDGEMENTS

We thank Marie Clifford, Sara DeLeon, Robert Driver, Yamile Molina and William Wood for assistance in the field. Paulina Khodak, Skye Miller and Elisabeth Sulger helped collect neuroanatomical data. Specimens were collected under research permits from the Republics of Ecuador and Costa Rica. Funding was provided by NSF grant IOS 1209072, and by Drexel University College of Arts & Sciences startup funds, both to S.O’D.