Automated Functional Analysis of Astrocytes from Chronic Time-Lapse Calcium Imaging Data

Recent discoveries that astrocytes exert proactive regulatory effects on neural information processing and that they are deeply involved in normal brain development and disease pathology have stimulated broad interest in understanding astrocyte functional roles in brain circuit. Measuring astrocyte functional status is now technically feasible, due to recent advances in modern microscopy and ultrasensitive cell-type specific genetically encoded Ca2+ indicators for chronic imaging. However, there is a big gap between the capability of generating large dataset via calcium imaging and the availability of sophisticated analytical tools for decoding the astrocyte function. Current practice is essentially manual, which not only limits analysis throughput but also risks introducing bias and missing important information latent in complex, dynamic big data. Here, we report a suite of computational tools, called Functional AStrocyte Phenotyping (FASP), for automatically quantifying the functional status of astrocytes. Considering the complex nature of Ca2+ signaling in astrocytes and low signal to noise ratio, FASP is designed with data-driven and probabilistic principles, to flexibly account for various patterns and to perform robustly with noisy data. In particular, FASP explicitly models signal propagation, which rules out the applicability of tools designed for other types of data. We demonstrate the effectiveness of FASP using extensive synthetic and real data sets. The findings by FASP were verified by manual inspection. FASP also detected signals that were missed by purely manual analysis but could be confirmed by more careful manual examination under the guidance of automatic analysis. All algorithms and the analysis pipeline are packaged into a plugin for Fiji (ImageJ), with the source code freely available online at https://github.com/VTcbil/FASP.

data. Different ROIs were labelled using different colors to show the overlap between ROIs. The video of Figure 6D2.  Figure 6E. FASP was applied to the Ca 2+ time-lapse images of rat hippocampal astrocytes, monitoring their calcium dynamics as responses to quantitatively controlled ATP stimulation. The control sample only contained one FIU with spontaneous activity. Left: original input imaging data of a control sample with no ATP was added in; spatial dimension 512×512; 100 frames. Right: analytical results by FASP, with FIUs labeled using carmine contours on the input data.  Figure 6E. A sample of rat hippocampal astrocytes, with 100uM ATP applied. Panels similar as in Supplementary Video 4.
The plugin has been tested and successfully run on ImageJ version ≥ 1.49q. If users encounter problems using older versions of ImageJ, please update your ImageJ to a newer version ≥ 1.49q.

Input
FASP is designed for single-channel gray-scale time lapse image stack. So far we only support 8-bit or 16-bit images of any file format supported by Fiji/ImageJ. Open the image stack of interest in Fiji/ImageJ.
(Note 1: Fiji/ImageJ typically provides users with "Lookup Tables" under "Image" menu that can be applied to single-channel grayscale images or image stacks to produce pseudo-color images. Some colors may make manual proofreading easier, such as green color. FASP can be applied to singlechannel grayscale image stacks of any pseudo color, and the results will be shown as overlay on the input pseudo-color image stack.) (Note 2: FASP works on single-channel image stacks. If the input image stack contains more than one channel, FASP will report an alert and exit automatically. Users can resort to other tools, such as Fiji/ImageJ function "Split Channels" under "Image\Color", to make sure the input data for FASP is single-channel.)

Parameter Setting
With the input image stack loaded, run FASP from the plugins menu. A dialogue window will occur, allowing users to adjust the input parameters. Suggested values of parameters are given as default in the textboxes. After the parameter setting is done, click "OK".

Processing Procedure
Since FASP is a completely automated algorithm, NO interactive user input is needed while FASP is running. The progress of the program will be shown using a progress bar in the bottom right corner of Fiji/ImageJ toolbar. Besides computing platform, the running time of FASP also depends on the activity level of the imaged cells, the spatiotemporal scales of input image stack, and the minimum allowed size of FIUs.

Outputs
The results are visualized using tables and the ROI manager of Fiji/ImageJ. All FIUs are circled and labeled on the original image stack. The edges of FIUs are indicated as yellow, and a label is put on each FIU to show the ID of this FIU. If the yellow edges disappear due to some reason/operation, users may reset the edge overlay by checking the checkbox "show all" on the bottom right.
If a user is particularly interested in one FIU, she/he can click on the label of the FIU of interest to inspect its characteristic curve (in Δ / 0 ). The selected FIU will be highlighted with blue edges on the image stack.