Cm1950 cryomicrotome

Samples were held at −20°C for 24 h to bring samples to an appropriate temperature for cryosectioning. After being embedded in frozen section compound (VWR International, Westchester, PA), 5-μm thick cryosections were collected from the intestinal samples, transverse to the length of the intestine, using a Leica CM 1950 cryomicrotome. The cryosections were mounted on positively charged glass slides (VWR International; 5 cryosections per slide) and stored at 4°C for no more than 48 h before immunofluorescence staining.

The three fish selected for electron microscopy were also chosen for FTIR analyses. Muscle samples cryofixed in cooled isopentane (−160 °C) were cut in transverse orientation at −25 °C using a Leica CM1950 cryomicrotome (Nussloch, Germany). Tissue cryosections (6 µm depth) were placed onto BaF₂ windows and analysed with a FTIR microscope (Thermo Scientific iN10, Thermo Fisher Scientific, Madison, WI, USA) equipped with a liquid nitrogen-cooled detector. After the acquisition of a mosaic image of the tissue, we targeted the centre of a single-fibre cell to measure it with a 30 µm aperture. At least 35 cells were measured per fibre. Each cell signal was acquired from the average of 256 spectra. The background signal from the BaF₂ plates was obtained with 256 scans near the tissue sample. Spectra were acquired at a resolution of 2 ${\mathrm{c}\mathrm{m}}^{-1}$ in the range of 600–2500 ${\mathrm{c}\mathrm{m}}^{-1}$ . Background signal from BaF₂ was subtracted. Spectra were then further processed as described in Germond et al., 2018. Specifically, we performed a baseline correction using a polynomial fitting (5th order, 200 iterations) and vector normalisation. The area from 900 to 1780 ${\mathrm{c}\mathrm{m}}^{-1}$ was considered for subsequent machine learning analyses.