2.4. Magnetic resonance and X-ray imaging

For magnetic resonance imaging (MRI), newly eclosed and 25 day old bee abdomens were fixed using cold 4% PFA buffered in PBS (pH 7.4) overnight. Samples were placed under vacuum to minimize trapped air pockets and then mounted in 5 mm diameter tubing filled with PBS. Three newly eclosed and three 25 day abdomens were imaged using a 16.4T MRI (Ultrashield Plus 700 WB Advance NMR spectrometer, Bruker Optics, Billerica, MA, USA) fitted with a quadrature micro-imaging coil. A gradient echo FLASH sequence was used to obtain high-resolution (15–20 μm isotropic) three-dimensional images of the bee parts suitable for qualitative anatomic characterization and segmentation. A multi-echo RAREVTR sequence was used to acquire spin echo data in two-dimensional slices (in-plane resolution 50 µm) from which R2 relaxation rates were calculated. R2 values were calculated for each voxel using the MRI Analysis Calculator plug-in for ImageJ (v. 1.42) (for detailed acquisition parameters see electronic supplementary material C). The software fits a monoexponential decay function of the form, S(TE) = S(0)e2−R∗TE to the data, where S(0) is the initial amplitude and TE is the echo time. Voxels were excluded where the goodness of fit parameter was less than 0.9.

For X-ray imaging, adult forager abdomens (n = 3) were fixed in cold 4% PFA buffered in PBS (pH 7.4) overnight before dehydrating in a graded series of ethanol to 70% (30, 50, 70%). To increase X-ray absorption in soft tissues, abdomens were heavy metal stained for 24 h using 1% iodine in 70% ethanol (I2E). Samples were rinsed (×3) and imaged in 70% ethanol prior to imaging. Abdomens were scanned at 40 kV and 74 µA using an X-ray micro-computed tomography (X-ray µCT) system (Versa 520, Zeiss, Pleasanton, CA, USA) running Scout and Scan software (v. 10.6.2005.12038, Zeiss). A total of 3201 projections were collected over 360°, each with a 5 s exposure. 2× binning was used to achieve a suitable signal to noise ratio and 0.4× optical magnification was used to achieve an isotropic voxel resolution of 7.9 µm. Raw data were reconstructed using XMReconstructor software (v. 10.7.3679.13921, Zeiss) following a standard centre shift and beam hardening correction of −0.25 and 0, respectively. The standard 0.7 kernel size recon filter setting was also used.

The visualization and analysis of data generated from both MRI and X-ray µCT scans was performed following manual segmentation using Avizo (v. 8.1.1, FEI) software. All measurements were conducted using standard Avizo tools (for workflow see electronic supplementary material D).