Quantifying Autophagy: Measuring LC3 Puncta and Autolysosome Formation in Cells Using Multispectral Imaging Flow Cytometry

The use of multispectral imaging flow cytometry has been gaining popularity due to its quantitative power, high throughput capabilities, multiplexing potential and its ability to acquire images of every cell. Autophagy is a process in which dysfunctional organelles and cellular components that accumulate during growth and differentiation are degraded via the lysosome and recycled. During autophagy, cytoplasmic LC3 is processed and recruited to the autophagosomal membranes; the autophagosome then fuses with the lysosome to form the autolysosome. Therefore, cells undergoing autophagy can be identified by visualizing fluorescently labeled LC3 puncta and/or the co-localization of fluorescently labeled LC3 and lysosomal markers. Multispectral imaging flow cytometry is able to collect imagery of large numbers of cells and assess autophagy in an objective, quantitative, and statistically robust manner. This review will examine the four predominant methods that have been used to measure autophagy via multispectral imaging flow cytometry.


Introduction
Macroautophagy, hereafter referred to as autophagy, is a catabolic pathway in which long-lived proteins and organelles that accumulate during growth and differentiation are degraded via the lysosome.
[1] Autophagy is also a survival mechanism that reallocates nutrients from unnecessary processes to more vital processes in the cell. [2] Basal levels of autophagy are generally low but can be upregulated by various stimuli including nutrient starvation, physiological stress, pharmacological agents and infections. [3] Autophagy is a dynamic multi-step process that involves the formation of autophagosomes, fusion of the autophagosome with the lysosome to form the autolysosome, and finally the degradation of the contents in the autolysosome. [4] The key biological marker to identify autophagy in mammalian systems is the microtubule associated protein 1A/1B-light chain 3 (LC3). During autophagy, cytosolic LC3-1 is conjugated to phosphatidylethanolamine to form LC3-II. LC3-II is recruited and incorporated into the autophagosomal membrane. [5] While there is no single "gold standard" to measure autophagy, [4,6] methods that have traditionally been used to measure LC3 include Western blot analysis and fluorescence microscopy. [7] In recent years, measurement of LC3 by flow cytometry and multispectral imaging flow cytometry (MIFC) has been gaining popularity due to their quantitative power, high throughput capabilities and multiplexing potential. A recent Methods review article by Warnes demonstrates that traditional flow cytometry can be used to measure various aspects of the autophagic process including the accumulation of LC3 upregulation and corresponding cell cycle distribution. [8] Phadwal et al. used MIFC to multiplex immunophenotyping, measuring LC3 (autophagosome) and LYSO-ID (lysosome) colocalization as well as to measure the levels of γH2AX. [9] Neither of these examples would have been possible using traditional methods. Furthermore, the experiment done by Phadwal et al. could not have been accomplished using traditional flow cytometry because the spatial information obtained using MIFC is required to measure colocalization.
The ImageStream ® (EMD Millipore) is a MIFC that combines the speed and statistical power of flow cytometry with the information content of fluorescent microscopy. Unlike traditional flow cytometers that use photomultiplier tubes to collect fluorescence intensities, the ImageStream ® uses a charge-coupled device camera to collect multiple high-resolution images of every cell in flow, including brightfield, darkfield (SSC), and up to 10 fluorescent markers. Measuring signal intensity alone by traditional flow cytometry can work well with reporter cell lines to distinguish the control population from the treated population. However, fluorescent dye background and nonspecific staining may complicate analysis. MIFC is better able to cope with these staining issues as well as troubleshooting experimental details since images of the cells can be examined; and therefore puncta can be observed and measured. In addition, an increase in LC3 signal alone does not provide a complete picture of what is happening in the cells and recent publications emphasize the need to examine concurrent formation of the autolysosome. [4,[9][10][11] MIFC is uniquely able to measure this formation by the co-localization of the autophagosomes and lysosomes which is not possible using traditional flow cytometry.
An examination of publications reveals four principal methods using MIFC to assess autophagy have been published: Bright Detail Intensity, Spot Count, Bright Detail Similarity and a combined method incorporating Bright Detail Similarity and Spot Count. This review describes the specific techniques that others have previously reported for thorough evaluation of autophagic events. This review will take a critical look at each method and how it performs on a single sample set. The image analysis techniques highlighted in this review are not a full list of all possible techniques that could be done by MIFC but is a comprehensive review of the most used methods to assess autophagy using MIFC.

Treatment and Labeling
Jurkat cells in the exponential growth phase were washed with Hank's Balanced Salt Solution (HBSS) without Ca ++ and Mg ++ (Invitrogen) and given fresh media or Earle's Balanced Salt Solution (EBSS; Sigma-Aldrich) for 2 hours at 37 o C and 5% CO 2 .
The samples were resuspended in 1% formalin (Polysciences). For nuclear staining DAPI (Molecular Probes) was added to the samples at 1µg/mL.

Image Acquisition
In each experiment 5,000 events for each sample were acquired using a 12 channel Amnis ® brand ImageStream X Mark II (EMD Millipore) imaging flow cytometer equipped with the 405 nm, 488 nm and 642 nm lasers. Samples were acquired at 40x magnification. Single color compensation controls were also acquired. The integrated software INSPIRE ® (EMD Millipore) was used for data collection. The IDEAS ® analysis was performed as follows. Single color controls were used to calculate a spectral crosstalk matrix that was applied to each of the files for spectral compensations in the detection channels. The resulting compensated data files were analyzed using image-based algorithms available in the IDEAS ® statistical analysis software package. Single cells were separated from debris and doublets using a bivariate plot of aspect ratio vs area of the Brightfield image. Next cells in best focus were identified using Gradient RMS of the Brightfield image. This is followed by gating on positive events for DAPI, LAMP1 and LC3. Finally apoptotic cells are gated out and Bright Detail Intensity, Spot Count, and Bright Detail Similarity values were calculated from the positive non-apoptotic cells (Supplemental Figure 1 and 2). A more detailed description of the analysis is given in the following sections.

Bright Detail Intensity
The first method described here was originally published in 2007 by Lee et al. [11] At that time, Bright Detail Intensity R7 was called "small-spot total intensity". In this study MIFC was used to quantify the LC3-GFP autophagosomes in pDCs obtained from heterogeneity of LC3 puncta in size, shape and intensity; and that counting spots accurately might be challenging, especially if the autophagosome/lysosome fusion is inhibited resulting in massive LC3 accumulation. [13] Furthermore, LC3 staining is often a combination of LC3-II puncta and diffuse cytosolic LC3-I staining, in addition to nonspecific binding of primary or secondary antibodies. The diffuse LC3-I staining and nonspecific staining can lead to a significant increase in the intensity value. This overall intensity increase is not due to autophagy, BDI removes the diffuse signal and calculates the intensity on only the bright LC3 puncta.
The IDEAS ® software has two versions of BDI, BDI R3 and BDI R7. Both features compute the intensity of localized bright spots within the masked area in the image. BDI R3 computes the intensity of bright spots that are 3 pixels in radius or less, while BDI R7 computes the intensity of bright spots in the image that are 7 pixels in radius or less. In each case, the local background surrounding the spots is removed before the intensity computation. [12] Figure 1 shows graphical representation of how the BDI R3 feature is calculated. The image is processed using a top-hat transform to produce a bright detail image, after which the total intensity of the bright detail image is then calculated. BDI increases both with the increase in the intensity of individual LC3 puncta as well as with an increase in the number of puncta in a cell.
To demonstrate BDI. Figure
When using the spot count wizard, selection of exemplary 'truth populations' is critical.
The Spot Count Wizard can assist the user in determining which Mask works best with the data. To get a good robust mask it might take a few rounds of the Wizard to get an appropriate feature. This may mean re-evaluating and refining the 'truth populations.' To demonstrate how spot count is used to measure autophagy, the same population of cells were used as described for the BDI example. The Peak Mask was determined to be the most appropriate for assessment of autophagy in this data set. For further evaluation of Mask selection, see Supplemental  Figure 4A shows the LC3 Spot Count histograms for the Basal + CQ (blue) and Starved + CQ (red) Jurkat cells using the Spot Wizard in the IDEAS ® software. Mean LC3 spot counts for the Basal + CQ and Starved + CQ Jurkat cells were found to be 0.61 and 2.82 spots, respectively. Brightfield (BF), LC3-AF647, DAPI nuclear dye and a composite of LC3-AF647 and DAPI images of representative cells for the mean spot counts are shown, Figure 4B and 4C respectively. The size/shape/brightness of LC3 puncta can vary drastically between cells.
This fact was emphasized by de la Calle et al. and cited as a rationale for choosing the Texture Feature BDI. [13] Furthermore, no Spot Count Feature will be perfect due to this large variety in size/shape/brightness of LC3 puncta but a good Spot Count Feature will work most of the time and should have the appropriate trend. Background staining and reproducibility of the experiment are other potential pitfalls with Spot Count; this could result in the same Feature for Spot Count producing differential results amongst data sets of varying quality. For large studies in particular, it is important to ensure the Spot Count Feature being used is robust and produces consistent quality amongst multiple data sets.

Bright Detail Similarity
Autophagy is a highly dynamic, multi-step process and the mere detection of the number of autophagosomes or measurement of LC3 puncta is insufficient for a comprehensive evaluation of the entire autophagic process. It has been well documented that the formation of the autophagosome as well as an increase in the lysosomal content are hallmarks of autophagy. The BDS Feature is designed specifically to compare the small bright image detail of two images and can be used to quantify the co-localization of two probes. In this case, BDS measures the co-localization of fluorescently labeled autophagosome and lysosome markers. BDS is the log transformed Pearson's correlation coefficient of the localized bright spots with a radius of 3 pixels or less within the masked area in the two input images. In other words, BDS calculates the degree of overlapping pixels from two different fluorescent channels. Since the bright spots in the two images are either correlated (in the same spatial location) or uncorrelated (in different spatial locations), the correlation coefficient varies between 0 (uncorrelated) and 1 (perfect correlation).
The coefficient is log transformed to increase the dynamic range to between zero and infinity (0, inf). For BDS to be accurate, it is essential to gate on cells that are bright for both fluorescent markers of interest. [9] Gating on positive events is needed to prevent measuring BDS on non-specific binding, imaging artifacts or noise. This is a crucial step since BDS can potentially amplify these artifacts that are not true signals.  Figure 5B and 5C, respectively. There is a shift from the Basal + CQ to the Starved + CQ sample; however, it is not as dramatic of a shift compared to other methods show in this review.
The images from the means of both the Basal + CQ and Starved + CQ samples show a greater difference than the histograms may lead you to believe; highlighting the fact that BDS does not take into account the number of autophagosomes and lysosomes that are co-localizing.

Bright Detail Similarity and Spot Count
The final method reviewed is the combined use of BDS-LC3/LAMP1 and LC3 Spot Count. There are several reports of using both BDS and Spot Count in the same study. [10,11,22,23] However, there is only one publication that added dimension by using a bivariate plot of Spot Count and BDS to assess autophagic flux. [11] Rajan et al.
found that the measurement of BDS alone is not always sufficient and that using only The high LC3 spot population was further classified into those with low co-localization (accumulation of autophagosomes) and those with high co-localization (accumulation of autolysosomes). By separating the bivariate into three populations, Rajan et a. was able to distinguish between cells with very few autophagosomes versus cells with an accumulation of autophagosomes or autolysosomes. [11] To demonstrate the benefit of combining BDS-LC3/LAMP1 and LC3 Spot Count Sample images from the three regions of the Starved + CQ sample are shown in Figure   6B. Table 1 summarizes the bivariate results for the different experimental conditions. This experiment led to a similar conclusion as that derived by Rajan et al. Under basal conditions the number of autophagosomes was low and few cells were found with high spots, while the addition of CQ increased the number of LC3 puncta. Since the lysosome is unable to break down the autophagosome, this leads to an increase in the co-localization of the autophagosomes and lysosomes. This effect is amplified under nutrient starvation which induces autophagy; however, without the addition of CQ there is not a significant increase in the number of autophagosomes, likely due to an increase in the rate of autophagic turnover. When the cells are starved in the presence of CQ there is an increase in the co-localization and number of autophagosomes, supporting the conclusion that starvation increases autophagic flux.

Discussion and Conclusions
In this review 4 methods were presented to measure autophagy using MIFC.
However, the question still remains what method does the best job to represent autophagy and does MIFC add additional information that is not obtained by traditional flow cytometry (i.e. measuring Intensity of LC3-AF647 only). To address this Table 2 has the mean and standard deviation values of Intensity-LC3, BDI-LC3, LC3 Spot Count However, there is not one single correct method that will work for all systems. Each researcher needs to examine their data and the question they are trying to answer to determine which method is most appropriate for their data set. For simple LC3 puncta accumulation purposes, BDI or Spot Count should suffice. BDS will enhance comprehensive assessment of autophagic flux by measuring the autophagosome/lysosome fusion but if there is a large difference in the number of autophagy organelles it is likely that BDS alone might not be sufficient and LC3 Spot autophagy organelles in the sample.
Autophagy is a complex process that has been implicated in a broad spectrum of mammalian diseases including cancer, neurodegenerative disorders and inflammation.
There is an increasing need to accurately detect/measure autophagy because it is critical to improve the understanding of this fundamental cellular process and its link to disease. Furthermore, MIFC has the potential for use in pharmaceutical development due to its capacity for high throughput screening of drug targets that modulate autophagy.          Figure 1: Gating strategy for Intensity-LC3, Bright Detail Intensity(BDI) LC3 and LC3 Spot Count. Single cells were separated from debris and doublets using a bivariate plot of aspect ratio vs area of the Brightfield image. Next cells in best focus were identified using Gradient RMS of the Brightfield image. Followed by gating on positive events for DAPI, LAMP1 and LC3. This is followed by gating on non-apoptotic cells "Cells" population in Area_Threshold(DAPI, 50%)_DAPI vs Contrast_M01_BF plot.
Supplemental Figure 2: Gating strategy for Bright Detail Similarity (BDS) LC3/LAMP1 and bivariate plot BDS-LC3/LAMP1 and LC3 Spot Count. Single cells were separated from debris and doublets using a bivariate plot of aspect ratio vs area of the Brightfield image. Next cells in best focus were identified using Gradient RMS of the Brightfield image. Followed by gating on positive events for DAPI, LAMP1 and LC3. This is followed by gating on non-apoptotic cells "Cells" population in Area_Threshold(DAPI, 50%)_DAPI vs Contrast_M01_BF plot. Note: BDS requires cells to be brightly positive for LC3-AF647 and LAMP1-PE; therefore, the positive gate for LAMP1 and LC3 starts at 10,000 (not 1,000 as in Supplemental Figure 1).