Wild-type mice and transgenic BAC(GlyT2-EGFP) mice expressing the green fluorescent protein EGFP in glycinergic neurons  were bred in the animal facility of the University Hospital Göttingen and treated in accordance with the guidelines of the German Physiological Society as well as the regulations of the State of Lower Saxony and the Federal Republic of Germany. Both wild-type mice and transgenic BAC(GlyT2-EGFP) mice were used for two-photon excitation microscopy in slices, while only BAC(GlyT2-EGFP) mice were used for whole-cell recordings and immunohistochemistry.
Preparation of respiratory slices containing the pre-Bötzinger complex
The preparation of living acute slices (“rhythmic slice preparation”) has been described previously [7, 11, 24, 36]. In brief, mice at postnatal day 4–9 (P4–P9) were rapidly decapitated and the brain was removed. The brainstem was isolated in ice-cold saline and mounted on an agar block and transferred to a vibratome (VT1000S, Leica, Bensheim, Germany). Transverse slices were sectioned in a rostro-caudal direction using ~200-µm steps until the lower brainstem was reached and the inferior olive was visible. Then, a 600–650-µm-thick slice containing the preBötC was prepared. Using this procedure, the rhythmic population of neurons was found within the rostral part of the slice. This slice was transferred to the recording chamber, with the rostral side on top, to be continuously superfused with an oxygenated (95% O2 and 5% CO2) artificial cerebrospinal fluid containing (mM) 118 NaCl, 3 KCl, 1.5 CaCl2, 1 MgCl2, 1 NaH2PO4, 25 NaHCO3, and 30 D-glucose (~330 mosM/l; pH 7.4). Temperature was slowly raised to 30°C. For long-term rhythmic activity, potassium concentration was increased to 8 mM.
To identify EGFP-fluorescent glycinergic neurons for electrophysiological analysis, slices were illuminated with a monochromator (excitation 467 nm; Polychrome II, TILL Photonics, Gräfelfing Germany), fiber-coupled to an upright microscope (Axioscope FS, Zeiss, Oberkochen, Germany) equipped with changeable filter sets (dichroic mirror 505 nm, band pass emission filter 545/50 nm). To optimize cell visibility for patch clamping, epifluorescence illumination was combined with infrared light illumination using a gradient contrast  and a CCD camera (Vx45, Optronics; Goleta, CA, USA, or Sensicam QE, PCO, Kehlheim, Germany). Images of fluorescent cells were captured and stored on a personal computer using ImagingWorkbench software (Indec BioSystems, Santa Clara, CA, USA).
Whole-cell recordings were performed with a Multiclamp 700A (Axon Instruments, Inc., Foster City, CA, USA) or an L/M-PCA patch clamp amplifier (E.S.F, Friedland, Germany). Recording electrodes were prepared from borosilicate capillaries (Biomedical Instruments, Zöllnitz, Germany) using a horizontal, programmable puller (Zeitz-Instrumente, Munich, Germany). Patch pipettes were filled with an intracellular solution containing (mM) 140 potassium gluconate, 1 CaCl2, 2 MgCl2, 4 Na2ATP, 10 HEPES, 10 EGTA (pH, 7.2 (adjusted with KOH)) leading to a resistance between 3 and 6 MΩ. The calculated equilibrium potential E
[C1] for chloride was −87 mV.
Inspiratory-related field potentials were recorded in the ventrolateral area of the slice that contained the ventral respiratory column including the preBötC. Field potentials were recorded with custom-made, thick-walled single-barreled microelectrodes (~1 MΩ, 50–100 µm outer diameter) that were placed either ipsilateral or contralateral to the site of imaging or whole-cell recordings.
Neuronal population activity was amplified by a custom-built AC amplifier (5,000–10,000 times), band-pass-filtered (0.25–1.5 kHz), rectified, and integrated (Paynter filter; time constant 40–70 ms). Recordings were digitized at 10 kHz and stored using Axograph software (Axon Instruments, Inc., Foster City, CA, USA) or Chart software (ADInstruments Pty. Ltd., Mountain View, CA, USA).
Calcium imaging using two-photon excitation microscopy
Calcium imaging in the ventral respiratory column was performed with a custom-made two-photon microscope that is based on a commercial scan head (TriMScope, LaVision BioTec, Bielefeld, Germany). In this scan head, the pulsed infrared-laser beam can be divided into a line of up to 64 foci, increasing the amount of emitted light per time without increase of photo damage . The scan head was coupled to a fixed-stage, upright microscope (Axioscope FS2, Zeiss, Oberkochen, Germany), using ×40 (0.8 NA) or ×63 (1.0 NA) water immersion objectives (Zeiss, Oberkochen, Germany). Two-photon excitation was achieved by a titanium sapphire laser equipped with broadband optics (MaiTai BB, Spectra Physics, Darmstadt, Germany). For field detection of emitted light, we used CCD cameras (Ixon 885 Andor Technology, Belfast, Northern Ireland, or PCO; Sensicam QE; Kehlheim, Germany). Fast calcium imaging was performed using the multi-beam mode (16 or 32 beams) with scan fields of 200 × 200 µm (Ixon camera) or 170 × 220 µm (Sensicam QE). Exposure time was 25–40 ms (4 × 4 binning = 255 × 256 pixel) leading to sampling rates of 10 to 30 Hz. For higher spatial resolution, the number of beams was increased to 64 and images were taken with no binning at longer integration times. Laser power was controlled by a λ/2 plate and a sheet polarizer in the scan head, controlled by “ImSpector” imaging software (LaVision BioTec, Bielefeld, Germany), and 3D-Stacks with 10 µm per step were taken using a piezo-focus (Physik Instrumente, Karlsruhe, Germany) to determine the spatial distribution of the respiratory neurons up to a depth of 100 µm. Deeper in the tissue, identification of cell borders was not possible. The microscope was also equipped with two photomultipliers (Hamamatsu Photonics, Hamamatsu, Japan) for non-descanned fluorescence detection. Non-descanned detection was used for high-resolution scanning together with 1 µm z-stacks (Fig. 2).
Cell loading for calcium imaging
Multi-cell bolus loading was performed as described in detail earlier . Briefly, 50 µg Oregon green BAPTA-1AM (OGB-1-AM, Molecular Probes, Eugene, OR, USA) was dissolved in dimethyl sulfoxide (5 µl) containing 20% Pluronic F-127 (Molecular Probes, Karlsruhe, Germany) and stored at −20°C in 0.5-µl aliquots until used. For injection, one aliquot of this stock solution was dissolved in 5–7 µl of an extracellular solution containing (mM) 150 NaCl, 2.5 KCl, and 10 HEPES (pH adjusted to 7.4). The final concentration of OGB-1-AM was between 0.6 and 0.8 mM. A small amount of the solution was injected (2 bar; 2 min) 50–100 µm below the slice surface into the preBötC region using a patch pipette . After injection, an incubation period of 30 min was allowed for sufficient dye loading.
Identification of EGFP-labeled neurons for calcium imaging
Calcium-signals in GlyT2-EGFP expressing neurons were analyzed using OGB-1-AM. The following method was adapted from Wilson and colleagues . To determine the optimal parameters for discrimination of EGFP and OGB-1-AM fluorescence, baseline fluorescence of six individual neurons stained with either of the fluorophores was determined for CFP (480/30 BP) and YFP (525/50 BP) emission filters, respectively, while the excitation wavelength of the laser was tuned from 780 to 920 nm (10-nm steps) (Supplemental Fig. 2). Similar results were obtained with the 511–551-nm (OGB-1-AM) and 450–500-nm (EGFP) band pass filters, respectively, which were used for the multi-beam mode. From these experiments, we concluded that changes of OGB-1-AM fluorescence can be faithfully detected at 800-nm excitation wavelength through a YFP-Filter, whereas EGFP-fluorescence was specifically detected at 900-nm excitation wavelength through the CFP filter. Optical filters were obtained from AHF Analysentechnik AG (Tübingen, Germany).
For choline acetyltransferase (ChAT) immunohistochemistry, adult mice (BAC(GlyT2-EGFP) mice, n = 3) were deeply anesthetized with diethyl ether. Depth of anesthesia was confirmed by the absence of nociceptive reflexes. Animals were then cardio-perfused with Hanks' balanced salt solution followed by 4% paraformaldehyde in phosphate-buffered saline (PBS) as described before . The brain was removed and placed for 24 h in the same fixative and stored thereafter in PBS with 0.01% sodium azide. Transverse or sagittal slices (50–100 µm) were cut using a vibratome (VT1000S, Leica, Bensheim, Germany). Slices were permeabilized in 0.4% Triton in PBS for 30 min, and non-specific antibody binding was minimized by adding 4% normal donkey serum in PBS with 0.2% Triton. Sections were incubated with the primary anti-ChAT-antibody (Chemicon, Temecula, CA, USA) overnight in 1% serum in PBS with 0.05% Triton (4°C) and labeled with Cy5-conjugated secondary antibodies the following day (room temperature, 2 h, anti-goat from donkey, 1:250 dilution; Jackson IR, Newmarket, UK).
Fluorescence microscopy of immunostained brainstem sections was performed using a confocal laser scanning microscope (LSM 510 Meta, Zeiss, Oberkochen, Germany). EGFP was excited at 488 nm (argon laser). Fluorescence was detected through a 505–530-nm band pass filter. Cy5-fluorescence of secondary antibodies was visualized at 633 nm excitation (HeNe laser) through a 650-nm-long pass filter. Tile scans (for overview) were acquired using a motorized x-y table attached to the confocal microscope and multi-time series software extension (Zeiss, Oberkochen, Germany).
The two-photon excitation microscope was controlled by ImSpector imaging software (LaVision BioTec, Bielefeld, Germany). For offline analyses, images were exported to TIFF format and processed with ImageJ software (http://rsb.info.nih.gov/ij/) using public macro routines provided by the Wright Cell Imaging Facility (http://www.uhnres.utoronto.ca/facilities/wcif/download.php). To correlate field potential recordings with optical signals, trigger pulses (TTL) for each image were recorded simultaneously with the field potential.
To perform cycle-triggered averaging, we developed a MATLAB (Mathworks Inc., Natick, MA, USA) routine. Optical signals from consecutive respiratory cycles were processed as follows. First, linear trends were removed, and signals were spatially averaged by 3 × 3 pixels. Then, a 20 × 20-pixels region of interest (ROI) was set on a strongly stained inspiratory neuron. Optical signals within the ROI were moving-time averaged (bin width = 5 images) and high-pass filtered (time constant = 50 images) to detect peaks of calcium transients associated with inspiratory activity. Using the frame position of inspiratory peaks as the reference, optical signals of 50 preceding and 100 following frames were averaged five to 20 times. Additionally, the cross-correlations between each pixel and the ROI were calculated.
Calcium changes were calculated as relative changes (ΔF/F
0) from the averaged intensity of a ROI drawn with the “multi-measure” ROI macro of ImageJ. For quantitative analyses of glycinergic (EGFP-positive) neurons using two-photon imaging, we counted (a) the number of inspiratory glycinergic neurons, (b) the total number of glycinergic neurons, and (c) the total number of inspiratory neurons from five experiments using at least three (3–7) consecutive recordings at different depths of the slice (10 µm z-distance). A prerequisite for a recording to be analyzed was the identification of at least one respiratory neuron in the respective optical plane.
Quantification of inhibitory postsynaptic currents (IPSCs) was performed with MiniAnalysis program (Jaejin Software, Leonia, NJ, USA) using threshold-based event detection. The IPSC frequency was averaged from the reciprocal value of the inter-event interval.
Data were further analyzed with IGOR Pro (WaveMetrics, OR, USA) and SigmaPlot/SigmaStat (Systat Software Inc., San Jose, CA, USA). Student's t tests were used to determine the significance of changes of IPSC amplitude and frequency. Results were expressed as mean ± SEM and differences were considered significant if P < 0.05.