Methods 75 (2015) 87–95

Contents lists available at ScienceDirect

Methods journal homepage: www.elsevier.com/locate/ymeth

Applications of flow cytometry for measurement of autophagy Alik Demishtein a, Ziv Porat b, Zvulun Elazar a,⇑, Elena Shvets c,⇑ a

Department of Biological Chemistry, The Weizmann Institute of Science, 76100 Rehovot, Israel Department of Biological Services, The Weizmann Institute of Science, 76100 Rehovot, Israel c Department of Cell Biology, MRC Laboratory of Molecular Biology, Cambridge CB2 0QH, UK b

a r t i c l e

i n f o

Article history: Received 13 October 2014 Received in revised form 22 December 2014 Accepted 23 December 2014 Available online 3 January 2015 Keywords: Autophagy Flow cytometry GFP–LC3 Lysosome P62/SQSTM1

a b s t r a c t Autophagy is a dynamic catabolic process that plays a major role in sequestering and recycling cellular components in multiple physiological and pathophysiological conditions. Despite recent progress in our understanding of the autophagic process there is still a shortage of robust methods for monitoring autophagy in live cells. Flow cytometry, a reliable and unbiased method for quantitative collection of data in a high-throughput manner, was recently utilized to monitor autophagic activity in live and fixed mammalian cells. In this article we summarize the advantages and potential pitfalls of the use of flow cytometry to study autophagy. Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction Autophagy is a dynamic intracellular trafficking pathway by which cells digest their own components [1]. Although constitutively active in mammalian cells it is significantly increased in response to nutrient starvation, thereby providing fuel for cellular metabolism [2]. Autophagy participates in diverse physiological and pathophysiological functions, such as programmed cell death, cancer, pathogen infection and degradation of ubiquitinated protein aggregates formed under pathological conditions [3,4]. The autophagic process starts with the sequestration of cytoplasmic constituents, including long-lived proteins and cytoplasmic organelles (mitochondria, peroxisomes), by a membrane sac known as the phagophore or isolation membrane. The phagophore matures through the activity of autophagy-related (ATG) proteins into double-membrane vesicles termed autophagosomes. More than 34 ATG proteins have been identified, many of which are conserved from yeast to humans [5–7]. Abbreviations: ATG, autophagy-related; LC3, microtubule-associated protein (MAP) 1A/1B-light chain 3; BafA, Bafilomycin A; BDI, bright detail intensity; BDS, bright detail similarity; mCherry, monomeric cherry; EGFP, enhanced GFP; FCS, fetal calf serum; MEF, mouse embryonic fibroblasts; TDI, time delay integration; PBMC, peripheral blood mononuclear cells; PE, phosphatidylethanolamine. ⇑ Corresponding authors. Fax: +972 8 9344112 (Z. Elazar). E-mail addresses: [email protected] (Z. Elazar), eshvets@mrc-lmb. cam.ac.uk (E. Shvets). http://dx.doi.org/10.1016/j.ymeth.2014.12.020 1046-2023/Ó 2014 Elsevier Inc. All rights reserved.

First described more than 50 years ago as a system that degrades cytoplasmic components and cell organelles via the lysosomes [8,9], autophagy was considered to be a nonspecific bulk degradation process. However, with the growing body of evidence for and recognition of specific cargos, autophagy is now divided into two types: selective and nonselective [10]. Selectivity of autophagy is achieved by distinct interactions between adaptor proteins and their cargos, which are then recruited specifically to the autophagosomes. Despite the recent advances in our understanding of the autophagic process there is still a shortage in experimental techniques for studying it, especially in live cells and whole organisms. There has been extensive discussion of autophagy-monitoring assays [11,12] among them electron microscopy, immunofluorescence, western-blot analysis of microtubule-associated protein (MAP) 1A/1B-light chain 3 (LC3) and ATG8 turnover, and flow cytometry. Flow cytometry is a technology that simultaneously measures and analyzes multiple physical characteristics of single particles, usually cells, as they flow in a fluid stream through a beam of light. The measured properties include a particle’s relative size, relative granularity or internal complexity, and relative fluorescence intensity. Given these features, flow cytometry is well adapted for quantitative analysis in a high-throughput manner of individual cells or cell populations. In this review we focus on recent methods by which flow cytometric analysis is utilized to evaluate autophagic activity in live and in fixed mammalian cells, with specific emphases on the advantages and the potential pitfalls of these techniques.

88

A. Demishtein et al. / Methods 75 (2015) 87–95

2. Markers for monitoring of autophagic flux by flow cytometry ATG8 proteins in mammals are divided according to their sequence similarities into three subfamilies: MAP1 LC3 (or LC3), c-aminobutyric acid receptor-associated protein (GABARAP), and the Golgi-associated ATPase enhancer of 16 kDa (GATE-16) [13]. The three subfamilies have similar structures comprising an ubiquitin core and two additional a-helices at the N-terminus [14–16], and both the core and the a-helices are essential for autophagosome biogenesis [17]. The ATG8 proteins are processed at the Cterminus by ATG4, revealing a glycine residue that can then be conjugated to phosphatidylethanolamine found on the autophagosomal membrane. Thus, this lipid-conjugated form of the mammalian ATG8 protein (LC3-II, or GATE-II, or GABARAP-II) is widely used as a marker of autophagosomes in mammals [18,19]. Following fusion of autophagosomes with lysosomes to form autolysosomes, the lipid-conjugated pool of ATG8 proteins found on the inner autophagosomal membranes is delivered into the lysosomes and is consequentially degraded (Fig. 1A). The degradation of the ATG8 proteins can be utilized to measure the autophagic flux (i.e. the level of autophagic activity). Thus, a fluorophore-tagged version of the ATG8 proteins (GFP–LC3, for example) can be used to quantitatively measure autophagic flux by flow cytometry [20– 23]. In addition to GFP–LC3 (or other ATG8 isoforms) that serve as structural markers for autophagic activity, the cargo receptors p62/SQSTM1 and NBR1 are widely used to measure autophagic flux [21,24]. However, not only the autophagic flux, but also the mere translocation of GFP–LC3 and its association with the autophagosomal membrane can serve as a readout in flow cytometric analysis [25,26]. In the next section we describe in more detail how these markers have been utilized to quantitatively evaluate autophagy by flow cytometry.

2.1. Use of GFP–LC3 to monitor autophagic flux by flow cytometry 2.1.1. Rationale In a number of reported studies, a fluorescent version of LC3 has been used to measure autophagy by flow cytometry. These techniques exploit the fact that LC3 protein, once associated with the inner autophagosomal membrane, is delivered into the lysosome and degraded there. Importantly, the GFP-tagged (but not the RFP-tagged) version is quenched prior to its degradation, owing to the sensitivity of this fluorescent protein to the acidic lysosomal environment. Accordingly, the reduction in GFP intensity in autophagy-induced samples relative to untreated samples serves as a readout for autophagic activity (Fig. 1).

2.1.2. Studies utilizing flow cytometry to measure turnover of tagged LC3 A report by Shvets et al. [23] provided the first description of the use of flow cytometry to measure autophagic activity in live cells. In that work, CHO and HeLa cells stably expressing GFP–LC3 were used to quantify the turnover of autophagosomes. Autophagy was induced by several means, including the use of rapamycin or amino-acid starvation, and flow cytometry was utilized to measure the fluorescence intensity of GFP–LC3. Induction of autophagy was followed by a marked reduction in the fluorescence intensity of GFP–LC3 but not of GFP–LC3G120A, a mutant unable to undergo lipidation and the consequent degradation. Furthermore, the reduction was blocked by autophagic or lysosomal inhibitors, confirming its selectivity for the autolysosomal pathway. Thus, the changes in GFP–LC3 intensity was shown to be a reliable and specific readout that can be measured by flow cytometry to study the autophagic activity. This technique was further used to determine

the specific amino acids required for suppression of autophagic activity. Advantages. Measurement of autophagic activity by flow cytometric analysis has many advantages, both of a general nature (relating to the technique itself) and more specifically (for autophagy). In general, this quantitative technique is simple to apply and can be used for high-content analyses with simultaneous collection of numerous parameters. Notably, as living cells are used for analysis by this method, flow cytometry may be also utilized to sort specific subpopulations for further characterization. Most importantly, this assay provides a functional readout of autophagic activity (as explained in Fig. 1A), because the reduction in GFP intensity is an outcome of the proper formation of autophagosomes, and their targeting to and fusion with the lysosomes. Therefore, conditions that affect any step in the autophagic pathway will result in changes in the GFP-signal readout. Since GFP is rapidly quenched in the acidic lysosomal lumen, this method is more sensitive than a biochemical approach as it measures the delivery and fusion of autophagosomes with lysosomes rather than proteolytic degradation inside lysosomes. Disadvantages. A major disadvantage of this technique is that it requires the ectopic expression of fusion proteins, and this is usually an overexpression relative to endogenous levels. Although the technique can be performed with transiently transfected cells, problems arise from artifacts of nonspecific aggregation and from the heterogeneity of protein expression. To improve sensitivity and reproducibility, it is therefore advisable to develop stable isogenic cell lines. However, even when protein is stably expressed, ectopic expression may result in different kinetics of GFP reduction. As this assay is restricted mostly to living cells, samples cannot be stored or combined with immunolabeling of intracellular markers. Importantly, since it relies on reduction in GFP intensity, care must be taken to exclude the possibility of nonspecific leakage or quenching of GFP, or artificial effects on general protein synthesis. Similarly, any condition affecting lysosomal function or acidification will result in blockage of GFP reduction, which might be misinterpreted as blockage of autophagosome formation. Another disadvantage is that this assay detects the total level of LC3 protein and does not differentiate between LC3-I and LC3-II. Finally, most methods measuring autophagy by flow cytometry suffer from the limitation that they focus mainly on one ATG8 protein, namely LC3B. The mammalian family of ATG8 proteins, however, contains three homologues, each having several isoforms that should also be tested, as they might be responsible for different subtypes of autophagy [27]. Sheen et al. [22] also utilized flow cytometry to monitor autophagic activity after amino-acid deprivation. In their study, however, a double-tagged version of LC3, namely DsRed–LC3–GFP, was used to monitor autophagy. Since the GFP tag was located Cterminally to the ATG4 recognition site, it was cleaved upon activation of autophagy, and this loss in GFP fluorescence was monitored by flow cytometry. The authors introduced an autophagic index to normalize the GFP fluorescence level to that of the DsRed–LC3, which was relatively stable. This index accounted for possible changes in the synthesis level of the reporter after different treatments. As a control they used mutated DsRed–LC3DGFP, in which the ATG4-recognition sequence was deleted. This autophagic index was further utilized in that study to examine the effect of leucine deprivation in human melanoma cells with hyper-activated RASMEK pathway (Mel-ST cells). Advantages. The advantage of this method is that the location of the GFP tag downstream of the ATG4 recognition site enables selective measurement of ATG4 activity. Disadvantages. Tagging of the C-terminus of LC3 has certain disadvantages for the evaluation of autophagic activity (i.e. auto-

A. Demishtein et al. / Methods 75 (2015) 87–95

89

Fig. 1. Diagrammatic depiction of the principle underlying the main flow-cytometry-based assays described here. (A) Left: GFP–LC3 is found in the cytosol (LC3-I), while the PE-bound form (LC3-II) is associated with autophagosomes and is later delivered into autolysosomes, where it is degraded. Induction of autophagy by nutrient deprivation leads to an increased number of GFP–LC3 molecules associated with autophagosomes and autolysosomes, resulting in reduction of the GFP signal under these conditions. Right: schematic cartoon of the expected diagram measuring the GFP intensities of GFP–LC3-expressing cells incubated in nutrient-rich (grey histogram) and starvation (green histogram) media. (B) Mild detergent washout leads to loss of the cytosolic LC3-I and detection of only the membrane-bound LC3-II. Since induction of autophagy by nutrient deprivation leads to an increase in the membrane-bound fraction, and since treatment with lysosomal inhibitor (chloroquine) prevents loss of the GFP signal in the lysosome, the expected values of GFP intensity will increase (shift rightward) in starved cells and will have the highest signal after inhibiting lysosomal activity, as depicted in the schematic cartoon on the right. The basic parameters acquired during flow cytometric analysis and their suitability for different cell populations are listed in Appendix 1. (C) Autophagosomes labeled by mCherry–GFP–LC3 emit signals from both fluorophores, whereas autolysosomes emit predominantly mCherry signals because of loss of the GFP signal in the acidic environment. Therefore, after induction of autophagy flow cytometric analysis will show a strong decrease in the GFP signal but only a mild decrease in the mCherry signal, as depicted in the schematic cartoon on the right.

phagosome formation and autophagic flux). It is not clear whether C-terminal cleavage is dependent on autophagy induction or happens post-translationally. Therefore, GFP reduction in this case does not necessarily reflect autophagic activity. Moreover, the sensitivity of this assay will be largely compromised, as following cleavage the GFP will be predominantly cytosolic and is expected to be degraded via a nonselective pathway. Finally, ATG4 activity might be affected by various conditions (such as redox [28]), and changes in GFP intensity might be misinterpreted as induction or inhibition of autophagosome formation. Hundeshagen et al. [20] monitored autolysosomal formation and subsequent lysosomal degradation in a similar manner to that of the abovementioned studies but utilizing a differently tagged version of LC3, namely the tandem mCherry–GFP–LC3. GFP protein is more sensitive than mCherry to acidic pH and is quenched in the acidic lysosomal lumen; therefore, whereas the autophagosome

has both mCherry and GFP signals, the autolysosome has predominantly mCherry signals [29] (explained in Fig. 1C). In this study, after autophagy was induced the fluorescence intensities of both GFP and mCherry were monitored using FACS. As expected, after 16 h of nutrient deprivation GFP intensity was reduced by 50%, while mCherry intensity was reduced by only 20%. Addition of Bafilomycin A (BafA), an inhibitor of lysosomal acidification, prevented this reduction in both fluorophores, demonstrating that flow cytometric analysis can be used to monitor autolysosome formation labeled by tandem–LC3. Notably, to monitor the effects of different autophagic inhibitors or inducers on general endolysosomal function in this study, the authors also used GFP–Rab7, a marker of late endosomes. Unlike the effect of mCherry–GFP–LC3, the effect on GFP–Rab7 intensity was evident only after a long period of treatment (16 h), whereas after treatment for 6 h the differences were only mild. Using this technique the authors showed that the

90

A. Demishtein et al. / Methods 75 (2015) 87–95

autophagic and endolysosomal responses to either inducers or inhibitors of autophagy are affected by the nutritional conditions of the cells (full or deprived medium). Next, utilizing the Prestwick Chemical Library of FDA-approved compounds the authors used both GFP–LC3 and the tandem–LC3 in a FACS 96-well format to screen for autophagy modulators. More recently, Gump et al. [30] utilized mCherry–GFP–LC3 to study the effect of variation in autophagic activity on the cellular response to apoptotic stimuli. BJAB and Jurkat cells were transduced with retroviruses to stably express mCherry–GFP–LC3 and were sorted by ratiometric flow cytometry according to their autophagic flux level, calculated as the ratio of mCherry to GFP signal. High and low ratios between the two proteins defined high and low autophagic flux, respectively. Using this technique, cell-to-cell variability in autophagic activity under basal growth conditions was found in clonal populations of these cell lines. Furthermore, testing the response of the different populations to FAS or TRAIL stimulus showed that BJAB cells with high autophagic flux were more prone to cell death after FAS stimulus than cells with low autophagic flux. This behavior was specific for FAS treatment in BJAB cells but not in Jurkat cells, suggesting its dependence on cell type and stimulus. This finding was confirmed by use of the more traditional methods of western-blot analysis and both fluorescence and electron microscopy. Ratiometric flow cytometry was then used to test variations in autophagic activity in different cell lines [31]. Advantages. Hundeshagen et al. [20] developed the idea of measuring GFP–LC3 reduction as a readout of autophagic flux and overcame two pitfalls of this system. First, the addition of mCherry allowed better normalization to overall changes in protein synthesis, and secondly the authors utilized GFP-Rab7 in their screening as a control for general effects on the endolysosomal pathway. By gating the flow cytometer on the basis of the ratio between mCherry and GFP, Gump et al. were able not only to normalize changes in LC3 synthesis but also to accurately measure autophagic flux in individual cells. This is the major advantage of flow cytometry, as it enables users to evaluate characteristics of single cells rather than the means of whole populations, and accordingly to sort for subpopulations with high or low autophagic flux for further study. Disadvantages. Despite the significant progress made by the utilization of tandem mCherry–GFP–LC3 to monitor autophagic flux by the method of flow cytometry, thereby resolving several drawbacks of this method, the need to use ectopically expressed protein remains a major disadvantage of this experimental system. 2.1.3. Experimental guidelines Live cells expressing fluorescent LC3 are plated 24 h prior to the planned experiment. To minimize artifacts such as heterogeneity of protein expression or protein aggregation caused by overexpression of proteins, it is important to work either with stable cell lines having similar levels of protein expression or with endogenous proteins. Notably, since basal autophagy and general LC3 expression may vary in populations with different confluences or cell-cycle state, it is advisable to plan an appropriate control sample for each treated sample. In addition, planning control samples with cells expressing only GFP and the mutant LC3G120A, which cannot be conjugated to the autophagosomal membrane, should be considered as well. Autophagic activity in live cells stably expressing fluorescent LC3 can be measured in several ways. Treatment of the cells with rapamycin or by amino-acid starvation will induce autophagy, with resulting reduction in the fluorescence intensity of GFP–LC3 (Fig. 1). Addition of the lysosomal inhibitor BafA or knockdown of any essential gene (e.g. ATG5, ATG7/3) will block the reduction in the signal. These dynamic changes demonstrate that autophagic

flux is dependent on proper functioning of the lysosomes and the functional autophagic machinery. For the experiment, cells are treated by different autophagy inducers (such as nutrient starvation, rapamycin), autophagyblocking drugs (wortmannin, 3-methyladenine), or lysosomal acidification inhibitors (BafA, chloroquine) for different periods of time (usually 2–24 h). For evaluation of basal autophagy, unstarved cells are treated with drugs that block autophagy or lysosomal degradation. The cells are then washed, trypsinized, and resuspended in appropriate buffer (usually PBS/4% FCS) and subjected to flow cytometric analysis. In the analysis the GFP intensity of the treated cells is normalized to that of the control cells. In the case of mCherry– GFP–LC3, mCherry can serve as an internal control for such normalization. Because many different parameters are measured during the experiment it is important to choose suitable parameters for the proper analysis of GFP intensity (see Appendix 1). 2.2. Measurement of membrane-conjugated LC3-II using flow cytometry 2.2.1. Rationale The approaches described above measure the total amount of LC3 (GFP–LC3), including both the cytosolic (LC3-I) and the membrane-associated (LC3-II) proteins. To focus on the membraneassociated fraction of LC3, the cells can be permeabilized with mild detergents. Low concentrations of saponin or digitonin are used to release the cytosolic fraction, making it possible to measure only the autophagosome-related LC3. In contrast to the measurement of GFP–LC3 in intact cells, in cells prewashed with mild detergents autophagy activation results in an increase in the fluorescent signal. Since the number of autophagosomes increases after treatment with rapamycin or by amino-acid starvation, the cytosolic fraction of LC3 will decrease and the membrane-bound fraction will increase. Addition of BafA or chloroquine leads to further increase in the fluorescent signal. 2.2.2. Studies utilizing flow cytometry to measure LC3-II In two studies, one by Eng et al. [25] and the other by Kaminskyy et al. [26], mild detergent washouts with saponin and digitonin, respectively, were used to release the cytosolic LC3-I, thereby differentiating between the two forms (LC3-I and LC3-II). The membrane-bound LC3-II (either GFP–LC3 or endogenous LC3) was then measured by flow cytometry (Fig. 1B). Eng et al. [25] separated the cytosolic (LC3-I) from the membrane-associated (LC3-II) fraction of EGFP–LC3 by washing human osteosarcoma cells stably expressing EGFP–LC3 with mild detergent saponin. As measured by flow cytometry, starvation treatment of this cell line did not lead to significant changes in EGFP– LC3 levels, unlike in COS and HeLa cells, in which starvation led to marked reductions in GFP–LC3 [23]. Prewashing with saponin released the cytosolic fraction of LC3, leaving the membrane-associated fraction in the cells, which was quantified by flow cytometric analysis (Fig. 1B). This method was comparable with westernblot analysis of LC3 lipidation and microscopic analysis of puncta formation. Notably, differences in endogenous levels of LC3-II were detected after treatment with chloroquine in wild-type or Atg5knockout mouse embryonic fibroblasts (MEF), suggesting that this method is applicable also in primary cells and other cell lines. Kaminskyy et al. [26] also utilized a mild detergent washout with digitonin to measure the membrane fraction of LC3. These authors also determined the optimal concentration of digitonin required for detection of the membranal fraction of LC3. This technique was then used to monitor basal autophagic activity during different stages of the cell cycle in MEF and in human A549 cells, as well as to determine whether induced autophagosome formation is blocked at different cell-cycle stages. These authors thus

A. Demishtein et al. / Methods 75 (2015) 87–95

took advantage of flow cytometric analysis for simultaneous quantification of endogenous LC3-II/ectopically expressed GFP–LC3-II levels and detection of cell-cycle stage using propidium iodide. Interestingly, in that report basal autophagy was detected at all phases of the cell cycle with no blocking of its induction at any stage, in contrast to a previously reported finding [32]. Advantages. Unlike other methods, this approach makes it possible to specifically quantify lipid-conjugated LC3. Because the procedure utilizes cell permeabilization it can be used to label endogenous LC3 as well as other intracellular structures or markers [33], giving it a massive advantage over previously described methods. Disadvantages. This assay is less functional than previously described methods because, like microscopy, it detects only the translocation of LC3 to the autophagosomal membrane, rather than autophagic flux. The level of lipidation does not necessarily increase upon autophagy induction, as it depends on the lifespan of the autophagosome. To overcome these drawbacks, Eng et al. had to use appropriate controls, in the form of inhibitors of lysosomal degradation (ammonium chloride) for evaluation of autophagic flux. As this method relies on proper permeabilization and washout of the cytosolic fraction of LC3, the system is prone to relatively more noise and inconsistency. In addition, the advantage of livecell techniques is lost here, as during the assay the cells are permeabilized and fixed and cannot be sorted or re-collected. Finally, previous reports mention that after treatment with saponin and other detergents nonautophagic GFP–LC3 puncta are formed [34]. These detergents should therefore be used with caution and only after prior examination. 2.2.3. Experimental guidelines Wild-type or GFP–LC3-expressing cells are seeded 24 h prior to experiments planned to include treatment with autophagy inducers (amino-acid starvation, rapamycin) or inhibition of lysosomal acidification (chloroquine). After treatment the cells are washed, harvested, permeabilized with detergent (saponin or digitonin), and fixed with 4% paraformaldehyde. Cells are then either analyzed by flow cytometry (GFP–LC3-expressing cells) or subjected to indirect immunostaining with anti-LC3 antibody and then analyzed by flow cytometry (wild-type cells). For analysis of autophagosomes at different phases of the cell cycle, the cells are stained with propidium iodide following their incubation with secondary antibody. 2.3. Use of P62/SQSTM1 to monitor autophagic flux by flow cytometry 2.3.1. Rationale Autophagic flux can also be examined by measuring the degradation of autophagic cargo receptors found inside the autophagosomal lumen. P62/SQSTM1 and NBR1 are well-characterized autophagic cargo receptors that are recruited to autophagosomes, and their degradation rate depends on the autophagic flux. 2.3.2. Studies utilizing flow cytometry for detection of tagged p62/ SQSTM1 and NBR1turnover Larsen et al. [21] described an inducible expression system of GFP-p62, GFP-NBR1, and GFP–LC3B. Using the Flp-In T-REx (Invitrogen) system, these authors generated human embryonic kidney (HEK293) cell lines expressing GFP tagged reporter proteins. In this system such expression was inhibited by the tetR repressor protein, which could be released by the addition of tetracycline to allow transcription. After a period during which expression of the reporter proteins was induced, their expression was shut off and changes in fluorescent intensity (reduction or increase) were measured in single cells by flow cytometry. This made it possible to

91

estimate the autophagic degradation without the use of lysosomal inhibitors, which might exert effects on other cellular pathways besides autophagy. Degradation of the reporter proteins was evaluated in the presence of lysosomal, proteasomal, and autophagic inhibitors as well as under conditions of serum and amino-acid starvation. Of the three reporters GFP-p62 was found to be the most sensitive; the expression of GFP-NBR1 was too low and that of GFP–LC3B was too high, the latter leading to non-autophagic degradation as evidenced by the similar responses to lysosomal and proteasomal inhibitors. GFP-p62 was therefore used to detect the inhibition of autophagic flux in heterogenous cell populations also after siRNA-mediated downregulation (by LC3B siRNA) or by transient overexpression of a dominant-negative mutant of the kinase ULK1. Advantages. p62/SQSTM1 is a widely used and well-characterized autophagic receptor that serves for cargo recognition, and is therefore well adapted for measurement of autophagic flux. A major advantage of using this marker is the increased sensitivity of the assay. Because this protein can polymerize, every molecule of LC3 (or other ATG8 protein) may incorporate several molecules of p62/SQSTM1, resulting in amplification of the signal. Notably, this approach is sensitive enough to measure changes resulting from perturbations in protein levels; GFP-p62 was found to accumulate following RNAi of LC3B or overexpression of a dominantnegative mutant of ULK1. In addition to the bulk autophagic flux, GFP-p62 reduction is also a reflection of proper cargo recognition and incorporation into autophagosomes. More specifically, in this work these authors produced cell lines with controlled expression, allowing their use in high throughput screens to search for inducers or inhibitors of the autophagic pathway due to the accurate and reproducible results provided as early as 6 h after shutoff of the expression. Disadvantages. This system is confined to HEK293 cells, and intensive labor would be required for the creation of such a system in other cell types. Furthermore it might be unsuitable for primary cells. Overexpression of p62/SQSQM1 may induce formation of ubiquitinated protein aggregates, which may lead to overall cellular stress and affect the proteosomal degradation pathway [35]. Finally, although this protein is efficiently recruited into autophagosomes, p62 cannot serve for the readout of general cargo incorporation since it is specific for ubiquitinated cargos, and several other autophagic receptors have also been reported [24,36].

2.3.3. Experimental guidelines Stable HEK-293 cell lines expressing p62, NBR1, and LC3 under a tetracycline-inducible promoter are used to study the autophagic flux in single cells by flow cytometry. In this system protein expression is induced for 12 h by tetracycline, after which the inducer is removed and the cells are treated for 6 or 12 h by BafA 3-methyladenine, rapamycin, or the proteosomal inhibitor lactacystin in nutrient-rich or amino acid-free medium. In case of silencing by siRNA, RNAi mix is introduced into cells 48 h before analysis. Finally, the cells are trypsinized and resuspended, and degradation of the tagged proteins is measured by flow cytometry.

2.4. Detection of single autophagic vacuoles by flow cytometry 2.4.1. Rationale All of the techniques described so far are based on quantitative analysis of the autophagic flux or the autophagosome level in live or fixed cells. The following method, however, focuses on detection of single autophagic vesicles released from their cellular milieu and their quantification by flow cytometry.

92

A. Demishtein et al. / Methods 75 (2015) 87–95

2.4.2. Studies utilizing flow cytometry for detection of single autophagic vacuoles Degtyarev et al. [37] applied mild sonication to release fluorescently labeled autophagic vacuoles [AV] from cells. The integrity of the autophagic vacuoles following sonication was confirmed by light and electron microscopy. These autophagic vacuoles were detectable by flow cytometry and the authors named the technique Organelle Flow After Cell Sonication (OFACS). This method could also be used in a 96-well-plate format, emphasizing its feasibility for high-throughput screening. Earlier publications reported that endosomes, mitochondria, phagosomes, autophagosomes and lysosomes were successfully purified and analyzed as single organelles using flow cytometry [38–40]. Advantages. This system can be specifically adapted for the analysis, purification, and further characterization of single autophagosomes. Disadvantages. Similarly to the methods described by Eng et al. [25] and Kaminskyy et al. [26], this assay is less functional and more complicated technically than previously described methods and may therefore be accompanied by increased noise and variation in the system. Although the authors showed that it could be adapted for use in 96-well plates and was therefore suitable for high-throughput screening, the sample preparation with this method is more time-consuming. Again, this method relies on specific detection of lipid-conjugated LC3, which may vary significantly between different cell lines, and might be low in cells with a short autophagosomal lifespan. 2.4.3. Experimental guidelines GFP–LC3 expressing cells are seeded 24 h prior to the planned experiment, treated with drugs (PI3K inhibitor GDC-0941 and the lysosomal acidification inhibitor chloroquine) to accumulate autophagic vacuoles, then sonicated in their media in-well, on ice, and subjected to flow cytometric analysis. The assay can be carried out in 96-well or 384-well high-throughput format or by using a higher volume in a standard format. 3. Autophagic activity measurement by imaging flow cytometry Multispectral imaging flow cytometry integrates the features of flow cytometry and fluorescence microscopy combined with an image analysis capability [41,42]. ImageStream is an imaging flow cytometer that can acquire up to 12 channel images (including 10 fluorescent channels) at a rate of up to 5000 cells per second. As in conventional flow cytometry, cells pass in single file through the instrument and are illuminated by several lasers. The light is gathered, however, by a high-quality optical lens and captured on a charge-coupled device in ‘‘time delay integration’’ (TDI) mode. TDI enables rapid high-quality images to be captured by tracking the cells in motion, thereby increasing signal collection time and sensitivity [43]. As with flow cytometry, ImageStream can analyze large numbers of cells according to their fluorescence features and provide statistical analyses of the parameters. In addition, because it acquires cell images, the IDEAS software (AMNIS, a dedicated analysis software), can calculate many features of the morphology of the analyzed cells and the distribution of the fluorescent signal thereby broadening the multi-parametric analysis of the examined cells. 3.1. Approaches for autophagic activity analysis using imaging flow cytometry 3.1.1. Rationale Imaging flow cytometry makes it possible to analyze large numbers of cells on the basis of their fluorescence intensity and

distribution, as well as to perform statistical evaluation of different morphometric parameters. Autophagy is a highly dynamic process that induces changes in the staining patterns and the subcellular distribution and localization of the autophagic machinery proteins. For quantifying these changes, morphological features demonstrate an advantage over the intensity features used in conventional flow cytometry. One of the many features that can be calculated for each cell is bright detail intensity (BDI), which can detect induction of autophagy in cells (see detailed explanation below). Another feature, bright detail similarity (BDS), can be used to examine colocalization between LC3 and lysosomes as a means of detecting autophagic flux. 3.1.2. Studies utilizing imaging flow cytometry to study autophagy de la Calle et al. [44] utilized a multispectral imaging flow cytometer (ImageStream-X) to detect populations exhibiting high LC3 staining and cleaved caspase 3. MEFs or human peripheral blood mononuclear cells (PBMCs) were treated with rapamycin and then fixed and stained for LC3 and for cleaved caspase-3 proteins, and thousands of cells were acquired using ImageStream-X. In searching for the features that optimally discriminate between cells with strong and weak LC3 staining intensity, the authors selected representative cell populations and calculated more than 40 features screening local intensity variations within their images. These features were ranked according to Fisher’s discriminant ratio, which represents the statistical separation power of each feature. One of the highest ranked features was the BDI, which calculates the intensity of localized bright spots that are not more than 3 pixels in radius within the masked area in the image. Local background around the spots was removed prior to the intensity computation. Since BDI combines a texture and an intensity feature and also yields a high discrimination value between low and high levels of LC3 puncta, it was chosen for further analysis. This feature was ranked higher than the general intensity feature (which corresponds to conventional flow cytometry measurements), demonstrating the advantage of using image-based data analysis. By utilizing a combination of the BDI feature (to accurately measure the autophagy rate) and the cleaved caspase-3 staining (for apoptosis quantification), distinct populations of cells responding to either autophagic stimuli or apoptosis-inducing treatments were detected. Notably, use of this method made it possible to identify another very rare population, of cells positive for both LC3 and cleaved caspase 3. Imaging flow cytometry can be adapted for GFP-tagged proteins and further extended using a 96-well-plate sampler for siRNA or drug screening. Moreover, other cell characteristics, such as nuclear morphology, could be used to discriminate cells undergoing apoptosis [45], as well as labeling other marker proteins in the pathways under study. Phadwal et al. [46] utilized imaging flow cytometry to measure autophagic activity in PBMCs. Autophagic degradation was inhibited by E64d and pepstatin A (cathepsin inhibitors) in PBMCs and HEK293 cells, and was induced by lipolysaccharide in monocytederived macrophages. Cells were stained for LC3 and the lysosomal marker Lyso-ID, as well as for additional markers to identify specific cell populations and to quantify c-H2AX foci (a marker of DNA double-strand breaks). In this work, autophagy was quantified by the colocalization between endogenous LC3 and lysosomal marker Lyso-ID, as calculated by the BDS feature. BDS compares the small bright details of two images and quantifies the colocalization of two probes in a defined region. It is calculated as the log-transformed Pearson’s correlation coefficient of the localized bright spots with a radius of three pixels or less within the masked area in the two input images. High autophagic activity was characterized by high LC3 intensity and pronounced colocalization with Lyso-ID, as determined by the BDS index calculated by ImageStream IDEAS software. The BDS parameter yielded significant

A. Demishtein et al. / Methods 75 (2015) 87–95

results when applied on a population of cells that was double-positive for the fluorescent markers. In addition, LC3 puncta were quantified using the Spot Count feature, based on a calibrated spot mask that defined size and intensity against a background of the spots. This feature, however, was not as good as BDS in distinguishing between the positive autophagic population and the background. Following validation of this method on HEK293 cells, it was applied on PBMCs and several cell populations were examined. By combining the autophagy measurement with cell classification and labeling of additional markers, the authors were able to show that autophagy occurred to a greater extent in T cells than in B cells, and that autophagy levels in PBMCs from aged

93

donors were lower than in young donors. It was shown, moreover, that in senescent T cells (characterized by a decrease in expression of CD28 and an increase in CD57 markers), autophagy levels were lower than in their nonsenescent counterparts. This finding was further corroborated by use of the ImageStream ability to count c-H2AX foci simultaneously with the other measurements, demonstrating that the high mean value of c-H2AX foci/cell in CD57+/CD8+ cells correlates with low autophagy levels. To determine which ImageStream features are most suitable for autophagy detection we treated HeLa cells stably expressing GFP–LC3 with the lysosomal inhibitor BafA, then fixed them and either directly imaged or additionally immunostained them for

Fig. 2. Analysis of autophagy by imaging flow cytometry. (A) HeLa cells stably expressing GFP–LC3 were incubated overnight in the presence (transparent histogram) or absence (gray histogram) of the lysosomal inhibitor Bafilomycin A (BafA), and were analyzed by ImageStream-X. Representative images of cells with median bright detail intensity (BDI) values are shown for control (DMSO) and BafA-treated cells. (B) Data collected in (A) were analyzed using the area parameter applied on the top 85% of pixels (threshold 85% mask), where the transparent histogram depicts BafA-treated and the gray histogram depicts control (DMSO-treated) cells. Representative images of cells with median area values are shown for control (DMSO) and for BafA-treated cells. (C) Cells treated as in (A) were fixed, stained with anti-p62 antibodies, and analyzed using the Bright Detail Similarity (BDS) parameter to measure the colocalization of GFP–LC3 and p62 signals (see detailed explanation in text). Shown here is a BDS histogram overlying a histogram of cells incubated with or without BafA (gated on double-positive cells). Representative images of cells with median BDS values are shown for control and for BafA-treated cells.

94

A. Demishtein et al. / Methods 75 (2015) 87–95

p62. We then imaged the cells using ImageStream-X and examined several features for autophagy detection, including BDS and BDI (Demishtein et al., unpublished observations, Fig 2). We found that the BDS feature yielded better identification of cells with high autophagic activity than the BDI feature (Fig. 2A and C). However, since an increase in autophagic activity eventually leads to degradation of LC3 in the lysosomes, high activity would result in reduction of the BDS index. Accordingly, it is essential to compare this index between cells incubated with and without lysosomal protease inhibitors (Fig. 2). In addition, given the large number of features available on the IDEAS analysis software, choice of the feature that best quantifies autophagic activity may differ under different experimental conditions. We noticed that in microscopic images, inhibition of lysosomal function results in a more condensed localization of the autophagosomes. We therefore used the area feature of the threshold mask for the 85% highest intensity pixels to quantify the differences between control and treated cells. In our experimental setup, this feature was found to demonstrate better statistical separation between the populations than the other two features (Fig. 2B). Advantages. One of the disadvantages of flow cytometry is its failure to distinguish between blockage in autophagosome formation and conditions affecting lysosomal function, that block both quenching and degradation of GFP. Also, that assay fails to detect conditions that affect the morphology, the size, or the distribution of autophagosomes. Image-based flow cytometry (ImageStream) overcomes all of these pitfalls and therefore appears to have great potential for improving the existing techniques. Another great advantage of imaging flow cytometry is its ability to stain for endogenous protein as well as to quantitatively analyze colocalization between different markers. By exploiting its capacity to image several fluorescent channels simultaneously in every individual cell, we can carry out multiparametric analyses of several additional factors in combination (Fig. 2). This demonstrates that with regard to the collection of statistically robust, multiparametric morphometric data from a large number of cells this system has an advantage over conventional flow cytometry, which is confined to measurements of fluorescence intensity. In addition, imagebased flow cytometry can reliably analyze different subpopulations in a complex sample, which is a challenge in conventional microscopy, especially in nonadherent cells. Finally, as demonstrated for mass cytometry, the collection of large multiparametic data sets may require more comprehensive data analysis methodologies for data reduction [47]. Disadvantages. It appears that ImageStream is not suitable for counting specific numbers of autophagosomes. When this technique is used, adherent cells need to be trypsinized prior to fixation, and this may adversely affect both the resolution and the distribution of internal structures. The round shape of the trypsinized cells, as well as limitations of its optical features, indeed yield a condensed cytosol with only a few bright spots, unlike the numerous autophagosomes typically observed by microscopy. Thus, because of the low resolution and the altered cell morphology, ImageStream analysis software could not identify an increase in LC3 puncta following lysosomal inhibition. Therefore, proper control and careful analysis of the many features available in ImageStream analysis software will be needed in order to find the best way to identify cells with pronounced autophagic activity. It is also possible that these features differ in different experimental systems or conditions (Fig. 2). Finally, ImageStream does not have the ability to sort the examined cells, which would no doubt confer an important advantage for the study of populations with unique characteristics. Future improvements to this system will therefore be needed before its full potential can be realized.

3.1.3. Experimental guidelines Both cell lines and primary cells have been used for experiments with different inducers of autophagy such as rapamycin or amino-acid deprivation, or inducers of cell death such as staurosporine or UV light. Adherent cells are trypsinized and washed prior to immunofluorescence staining. First, the lysosomal marker Lyso-ID is used to stain live cells and then cellular surface markers are labeled. Finally the cells are fixed, permeabilized, and stained with anti-LC3, anti-cleaved caspase-3 or other antibodies. Potential controls include autophagy-deficient Atg7 knockout cells, apoptosis-regulator Bax-deficient cells (Bax / /Bak / ), and cells transfected with siRNA against ULK1, a key autophagic protein. Unstained and single-stained samples have also been used as controls and the latter were used for compensation analysis. Data were collected with the ImageStream system and analyzed using the IDEAS software. 4. Conclusions In this article we discussed the growing number of existing methods for evaluating autophagic activity by flow cytometry. As cell populations vary considerably it is necessary to develop more robust methods to examine autophagic activity. There is also a growing need to monitor autophagic activity in live cells and organisms. Flow cytometry is a robust tool that can handle large numbers of samples and provide quantitative, sensitive, and reproducible ways to measure the autophagic activity of single cells or a whole cell population. Given these features, flow cytometry is one of the optimal methods existing in the field. A major disadvantage of this method has been the requirement to use ectopically expressed protein; however, methods utilizing endogenously labeled protein have recently emerged [25,26]. Moreover, the recent development of state-of-the art techniques for genome editing [48] has opened up exciting new opportunities for labeling endogenous proteins. While the advantages of flow cytometry are undisputable, a significant drawback of this technique is the inability to achieve adequate morphological analysis by microscopic observation. Methods such as imaging flow cytometry, which combine flow cytometry and microscopy, therefore offer great promise in the field of autophagy. We will certainly see more such applications developed in the future. Acknowledgements Z.E. is the incumbent of the Harold Korda Chair of Biology. We are grateful for funding from the Israel Science Foundation ISF (Grant #535/11), the German–Israeli Foundation GIF (Grant #1129/157), and the ISF Legacy Heritage Fund (Grant #1309/13). Appendix 1. As elaborated in the text, a major advantage of flow cytometry is its ability to yield multiparametric datasets of single-cell measurements. Therefore, proper analysis and handling of statistical data are mandatory for users to derive maximum benefit from applying the methods based on flow cytometry. For example, several parameters such as fluorescence intensity mean, median and mode are denoted as standard single-cell measurements. In a normally distributed cell population, the mean, median, and mode will be equal. The mean value of fluorescence intensity provides a value for the level of a protein of interest (for example, LC3 or p62/SQSTM1). It is preferable, however, to use the geometric mean (Geo Mean) rather than the arithmetic mean with log-amplified data. Whereas the geometric mean takes into account the weighting of data

A. Demishtein et al. / Methods 75 (2015) 87–95

distribution, the arithmetic mean should be used only for linear data or data displayed on a linear scale as it is overly sensitive to the ‘‘tail’’ of the distribution. The mode is defined as the relative intensity value that is most frequently found for a given parameter. This is the same intensity value as that of the highest point on a histogram. In normally distributed samples the mode (the highest point) will also be the ‘‘middle’’ of the histogram, but it is subject to sampling errors. Thus, in a skewed distribution, especially if there is a buildup of off-scale cells at either end of the histogram, the mode can be distant from the peak, and is therefore rarely used for data analysis of a cell population. The median can also be used as measure of quantitative cellular fluorescence. It is the central value, i.e., the 50th percentile, with half of the values above it and half below. If the population is not distributed normally but is skewed to the right or the left, the median will provide better information about the central tendency of the population. Therefore, the median is preferred as the most robust estimate of the intensity characterizing a particular cell population. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

B. Levine, D.J. Klionsky, Dev. Cell 6 (2004) 463–477. J.D. Rabinowitz, E. White, Science 330 (2010) 1344–1348. A.M. Choi, S.W. Ryter, B. Levine, N. Engl. J. Med. 368 (2013) 651–662. C. Lerena, S.D. Calligaris, M.I. Colombo, Curr. Mol. Med. 8 (2008) 92–101. T.M. Harding, K.A. Morano, S.V. Scott, D.J. Klionsky, J. Cell Biol. 131 (1995) 591–602. D.J. Klionsky et al., Dev. Cell 5 (2003) 539–545. M. Thumm, R. Egner, B. Koch, M. Schlumpberger, M. Straub, M. Veenhuis, D.H. Wolf, FEBS Lett. 349 (1994) 275–280. S.L. Clark Jr., J. Biophys. Biochem. Cytol. 3 (1957) 349–362. R.L. Deter, P. Baudhuin, C. De Duve, J. Cell Biol. 35 (1967) C11–C16. H. Nakatogawa, K. Suzuki, Y. Kamada, Y. Ohsumi, Nat. Rev. Mol. Cell Biol. 10 (2009) 458–467. D.J. Klionsky et al., Autophagy 8 (2012) 445–544. N. Mizushima, T. Yoshimori, B. Levine, Cell 140 (2010) 313–326. T. Shpilka, H. Weidberg, S. Pietrokovski, Z. Elazar, Genome Biol. 12 (2011) 226. J.E. Coyle, S. Qamar, K.R. Rajashankar, D.B. Nikolov, Neuron 33 (2002) 63–74. Y. Paz, Z. Elazar, D. Fass, J. Biol. Chem. 275 (2000) 25445–25450.

95

[16] K. Sugawara, N.N. Suzuki, Y. Fujioka, N. Mizushima, Y. Ohsumi, F. Inagaki, Acta Crystallogr. D Biol. Crystallogr. 59 (2003) 1464–1465. [17] H. Weidberg, E. Shvets, T. Shpilka, F. Shimron, V. Shinder, Z. Elazar, EMBO J. 29 (2010) 1792–1802. [18] Y. Kabeya et al., EMBO J. 19 (2000) 5720–5728. [19] Y. Kabeya, N. Mizushima, A. Yamamoto, S. Oshitani-Okamoto, Y. Ohsumi, T. Yoshimori, J. Cell Sci. 117 (2004) 2805–2812. [20] P. Hundeshagen, A. Hamacher-Brady, R. Eils, N.R. Brady, BMC Biol. 9 (2011) 38. [21] K.B. Larsen, T. Lamark, A. Overvatn, I. Harneshaug, T. Johansen, G. Bjorkoy, Autophagy 6 (2010) 784–793. [22] J.H. Sheen, R. Zoncu, D. Kim, D.M. Sabatini, Cancer Cell 19 (2011) 613–628. [23] E. Shvets, E. Fass, Z. Elazar, Autophagy 4 (2008) 621–628. [24] V. Kirkin et al., Mol. Cell 33 (2009) 505–516. [25] K.E. Eng, M.D. Panas, G.B. Karlsson Hedestam, G.M. McInerney, Autophagy 6 (2010) 634–641. [26] V. Kaminskyy, A. Abdi, B. Zhivotovsky, Autophagy 7 (2011) 83–90. [27] N. von Muhlinen et al., Mol. Cell 48 (2012) 329–342. [28] R. Scherz-Shouval, E. Shvets, E. Fass, H. Shorer, L. Gil, Z. Elazar, EMBO J. 26 (2007) 1749–1760. [29] S. Kimura, T. Noda, T. Yoshimori, Autophagy 3 (2007) 452–460. [30] J.M. Gump, L. Staskiewicz, M.J. Morgan, A. Bamberg, D.W. Riches, A. Thorburn, Nat. Cell Biol. 16 (2014) 47–54. [31] J.M. Gump, A. Thorburn, Autophagy 10 (2014) 1327–1334. [32] E. Tasdemir et al., Cell Cycle 6 (2007) 2263–2267. [33] K.E. Eng, M.D. Panas, D. Murphy, G.B. Karlsson Hedestam, G.M. McInerney, J. Virol. 86 (2012) 5674–5685. [34] I.A. Ciechomska, A.M. Tolkovsky, Autophagy 3 (2007) 586–590. [35] V.I. Korolchuk, A. Mansilla, F.M. Menzies, D.C. Rubinsztein, Mol. Cell 33 (2009) 517–527. [36] S. Mostowy, V. Sancho-Shimizu, M.A. Hamon, R. Simeone, R. Brosch, T. Johansen, P. Cossart, J. Biol. Chem. 286 (2011) 26987–26995. [37] M. Degtyarev, M. Reichelt, K. Lin, PLoS One 9 (2014) e87707. [38] R.B. Wilson, R.F. Murphy, Methods Cell Biol. 31 (1989) 293–317. [39] A. Cossarizza, D. Ceccarelli, A. Masini, Exp. Cell Res. 222 (1996) 84–94. [40] H. Koga, S. Kaushik, A.M. Cuervo, FASEB J. 24 (2010) 3052–3065. [41] E.K. Zuba-Surma, M.Z. Ratajczak, Methods Cell Biol. 102 (2011) 207–230. [42] E.K. Zuba-Surma, M. Kucia, A. Abdel-Latif, J.W. Lillard Jr., M.Z. Ratajczak, Folia Histochem. Cytobiol. 45 (2007) 279–290. [43] W.E. Ortyn et al., Cytometry A 69 (2006) 852–862. [44] C. de la Calle, P.E. Joubert, H.K. Law, M. Hasan, M.L. Albert, Autophagy 7 (2011) 1045–1051. [45] O. Ben-Ami et al., Cell Rep. 4 (2013) 1131–1143. [46] K. Phadwal et al., Autophagy 8 (2012) 677–689. [47] S.C. Bendall et al., Science 332 (2011) 687–696. [48] F.A. Ran, P.D. Hsu, J. Wright, V. Agarwala, D.A. Scott, F. Zhang, Nat. Protoc. 8 (2013) 2281–2308.