17fe5.html

<div>Imaging was done on an Ultima II 2-photon scanning microscope (Bruker, Billerica, MA) with a Vision II laser (Coherent, Santa Clara, CA). Brains were then continuously perfused in the saline used for dissection at 60mL/hour. Once the sample was placed and centered under the objective, we waited 5 minutes before starting the experiment to avoid any lingering network activation from the dissection or transmission lights. 2-photon excitation wavelength was 920nm, and powers at the sample varied between 3 and 10mW. CsChrimson was excited with trains of 2 ms 590 nm light pulses via a LED (M590L3-C1, Thorlabs, Newton, NJ) shining through the objective. The excitation light path contained a 605/55 bandpass filter and reflected to the objective with a custom dichroic (zt488-568tpc, reflecting between 568nm and 700nm). Instantaneous power measured out of the objective was roughly 50μW/mm 2 . Trains were delivered at 30Hz and the number of pulses was varied between 1,5,10, 20 and 30 -- corresponding to 2ms to 1s long stimulations. Imaging fields of view were chosen as to avoid scanning regions containing the CsChrimson while being as close as possible to the supposed connection site, &nbsp;as we observed occasional 2-photon stimulated slow activations of CsChrimson expressing cells pattern (as used in&nbsp;<cite class="ltx_cite raw v1">\citealt{Kim_2017}</cite> and also observed&nbsp;<b>cite Allan if he reported it already</b>). When this was impossible -- for example in self-activation controls or for completely overlapping cell types -- we chose a large ROI of which the CsChrimson/GCaMP6-m represented a small fraction to minimize the duty cycle. ROIs were maintained constant throughout the experiment. Each experimental run consisted of 4 repeats each approximately 16 seconds long. Runs were themselves repeated every 2 minutes. All experiments started with 5 runs corresponding to the 5 stimulation strengths, in a random order. This was sometimes followed by pharmacological testing. At the end of the experiment, a high intensity stack was acquired to insure the expression patterns were as expected, and the region imaged correct. At least 6 flies were tested for every pair considered.</div><h2 class="ltx_title_subsection">Pharmacology</h2><div>For blocking nicotinergic or inhibitory (GABAergic or glutamatergic) transmission, mecamylamine (50μM) or picrotoxin (10μM) (Sigma-Aldrich, St Louis, MO) were administered through the perfusion by switching to a different line for 3 minutes, followed by a wash period during which the perfusion was drug-free again. 30 pulses stimulations runs were repeated every 2 minutes starting 4 minutes before the drug application and throughout the wash. Prior to use, they were kept frozen in 25mM and 0.3M aliquots, respectively.</div><h2 class="ltx_title_subsection">Analysis</h2><div>All analysis was performed in&nbsp;<a href="http://julialang.org/" target="_blank">Julia</a>, using custom written routines (<b>github of that code public once it's cleaned, then link it on the website</b>).</div><h3 data-label="815715" class="ltx_title_subsubsection">Data processing</h3><div>For a given experiment, all movies were aligned to each other to compensate for slow drifts of the sample : for each run, the average image was calculated, and average images were aligned using correlation based sub-pixel registration (<b><cite class="ltx_cite raw v1">\citealt{guizar-sicairos_efficient_2008}</cite></b>,<b> </b>and<b>&nbsp; <a href="https://github.com/romainFr/subpixelRegistration.jl.git">https://github.com/romainFr/subpixelRegistration.jl</a>&nbsp;</b>for the Julia implementation used here). A region of interest (ROI) was defined for the full experiment : the average image (of all the runs) was clustered between foreground and background by k-means. It’s worth noting that the selection method relies only on average intensity and not activity. This is because we want to use the same detection method for responsive and non-responsive runs. This also relies on selecting fields of view as unambiguously containing the neuron of interest – and only the neuron of interest – during the experiment.</div><div>ΔF/F0 (<span class="math ltx_Math v1">\(\frac{\left(F-F_{0\ }\right)}{\left(F_0-B\right)}\)</span>, where F is the raw fluorescence and B the background signal (calculated as the intensity of the&nbsp;10% dimmest pixels of the average image) were then computed for each movie in the ROI. As we noticed that baseline fluorescence could vary widely over the course of an experiment (<b>ref to the part of the results or discussion where we'll cover that</b>), F0 is here defined as the median fluorescence in the ROI in the 3%  dimmest frames of the full experiment.</div><div></div><h3 data-label="786229" class="ltx_title_subsubsection">Statistics (<b>see the github repo notebook for details</b>)</h3><div>For every experimental repeat, we computed the following statistics  :</div><ul><li>F<sub>peak</sub> the peak fluorescence value, and T<sub>peak</sub> the time after stimulation at which the peak value is reached</li><li>I<sub>toPeak</sub> the integral of the signal up to the peak time</li><li>τ<sub>1/2</sub> the half-decay time</li><li>F<sub>base&nbsp;</sub>the fluorescence baseline before stimulation expressed in ΔF/F0</li></ul><div>Then for every run, which consists of 4 repeats, we computed:</div><ul><li>R<sub>repeats</sub> the average correlation between the 4 repeats of the run&nbsp;</li><li>&lt;F<sub>peak</sub>&gt;, &lt;T<sub>peak</sub>&gt;, &lt;I<sub>toPeak</sub>&gt;, &lt;F<sub>base</sub>&gt;&nbsp; and &lt;τ<sub>1/2</sub>&gt; the medians of&nbsp;F<sub>peak</sub>, T<sub>peak</sub>, I<sub>toPeak,&nbsp;</sub>F<sub>base&nbsp;</sub> and τ<sub>1/2</sub>, respectively</li></ul><div></div>