Autophagy can promote but is not required for epithelial cell extrusion in the amnioserosa of the Drosophila embryo

Abstract

During Drosophila embryogenesis the majority of the extra-embryonic epithelium known as the amnioserosa (AS) undergoes programmed cell death (PCD) following the completion of the morphogenetic process of dorsal closure. Approximately ten percent of AS cells, however, are eliminated during dorsal closure by extrusion from the epithelium. Using biosensors that report autophagy and caspase activity in vivo, we demonstrate that AS cell extrusion occurs in the context of elevated autophagy and caspase activation. Furthermore, we evaluate AS extrusion rates, autophagy, and caspase activation in embryos in which caspase activity or autophagy are altered by genetic manipulation. This includes using the GAL4/UAS system to drive expression of p35, reaper, dINR^ACT and Atg1 in the AS; we also analyze embryos lacking both maternal and zygotic expression of Atg1. Based on our results we suggest that autophagy can promote, but is not required for, epithelial extrusion and caspase activation in the amnioserosa.

Introduction

Autophagy (here used to mean macroautophagy) is, in the simplest of terms, most frequently described as a mechanism that delivers cargo originating from within the cytoplasm to the lysosome. A double-membrane structure envelopes such cargo, including organelles, long-lived proteins, and protein aggregates, to form the autophagosome; fusion of the autophagosome with the lysosome forms the autolysosome and the subsequent degradation of the autolysosome content allows nutrients to be recycled within the cell.¹ Autophagy is well known as a response to cellular stress—particularly nutrient deprivation, but is also implicated in some examples of programmed cell death (PCD).² Intriguingly, autophagy's role in PCD is not universal; it may function to promote either cell survival or PCD, although a role in promoting survival seems overall more prevalent in the scientific literature.³ The ambiguity regarding autophagy's role in PCD is often explained in terms of cellular context but the underlying mechanism that determines whether autophagy contributes to, or acts as a safeguard against, PCD remains elusive.

The terms “apoptosis” and “autophagic cell death” refer to the morphology of dying cells. In this sense, apoptosis denotes cell death that is associated with characteristic morphological features including chromatin condensation, nuclear fragmentation, reduction of cell volume, and plasma membrane blebbing.^2,4 On the other hand, dying cells that display prominent autophagosomes and autolysosomes, hallmark morphological characteristics of autophagy, are often described as undergoing autophagic cell death. While it is generally assumed that autophagy acts to promote death in cells undergoing autophagic cell death, a purely morphology-based definition of cell death does not make a distinction as to whether autophagy is operating in a pro-death or pro-survival capacity.^2,4 For example, the situation is particularly muddled when autophagic cells adopt apoptotic morphology; here the role of autophagy must be considered on a case-by-case basis.

Ultimately, the manifestation of apoptotic morphology is associated with the activation and subsequent activity of the caspase family of proteases.^2,5 Autophagic cell death can also be associated with caspase activity and, in a few instances, such as during the regression of the Drosophila larval salivary gland and the degeneration of the larval midgut, autophagic cell death may proceed in the absence of caspase activity.^6–9 In general, a growing body of evidence supports the notion that autophagy can, in some cellular contexts, play a key role in the regulation and execution of PCD.²

The Drosophila extra-embryonic epithelium known as the amnioserosa (AS) is a useful system for addressing autophagy and its role in PCD. During the course of normal embryonic development the AS, being extra-embryonic, is eliminated. The notably dorsal location of the AS during its degeneration, and its large conspicuous cells make live-imaging analysis of this tissue highly tractable. We previously suggested that the bulk of the AS (approximately 90%) degenerates via a form of PCD whereby autophagy is a prerequisite to caspase activity.¹⁰

Here we address the fate of the other ten percent of AS cells; these are removed from the AS epithelium earlier in development during dorsal closure. This removal occurs through extrusion to the basal side of the epithelium and is readily observed by time-lapse confocal microscopy of embryos expressing cadherin-GFP, or other epithelial membrane markers. Epithelial extrusions do not occur in Df(3L)ED225 homozygous mutant embryos which lack the pro-apoptotic genes head involution defective (hid), grim, reaper, and sickle (Df(3L)ED225 is described in detail in Mohseni et al.; extrusion data: Mohseni N, Reed BH, unpublished); nor do extrusions occur in embryos that express the baculovirus caspase inhibitor p35.¹¹ Overall, the extrusion of AS cells contributes between one third and one half of the net force required to drive dorsal closure and embryos lacking extrusion events are slow to complete dorsal closure.¹¹ The extrusion of damaged, dying, or superfluous cells from epithelia is also a general biological function that remains largely unexplored, especially in terms of genetic regulation. As a follow-up to our previous study of PCD in the AS, we are interested in determining how PCD relates to AS cell extrusion and what role, if any, of autophagy plays during this process.

Here we use time-lapse confocal microscopy to measure AS epithelial extrusion rates during dorsal closure in wild-type embryos and various genotypes associated with changes in autophagy and caspase activity. In addition, we monitor autophagy and caspase activity in the AS using expression of GFP-mCherry-Atg8a and the in vivo caspase sensor Apoliner, respectively. We show that driving expression of Atg1 in the AS using the GAL4/UAS system results in robust autophagy, elevated extrusion rates, and increased caspase activation. While enhancing autophagy increases extrusion rates, we show that caspase inhibition, which has little effect on the onset and progression of autophagy, dramatically decreases the extrusion rate so as to be negligible. Our previous study relied on expression of an activated form of the Drosophila insulin receptor to downregulate autophagy in the AS. In this study we also use the germline clone technique to recover mutant embryos lacking both maternal and zygotic Atg1 expression (Atg1 M/Z mutants). The attenuation of autophagy in Atg1 M/Z mutant embryos is confirmed by GFP-mCherry-DrAtg8a subcellular localization, reduced degradation of the autophagy substrate GFP-Ref(2)P, as well as TEM ultrastructure. In contrast to expression of activated INR, which results in reduced autophagy as well as reduced caspase activity and no extrusions, Atg1 M/Z mutants display caspase activation and execute AS extrusions. Our interpretation of these observations is that Atg1 kinase, a key regulator of autophagy, can promote caspase activation and associated AS extrusions, but is not required for either process.

Results

Extruded AS cells adopt an apoptotic morphology following the loss of apical-lateral contact with neighboring cells.

Our main approach toward better understanding the nature of AS epithelial cell extrusion and its relationship to PCD during dorsal closure was centered on live-imaging based confocal microscopy. In our first series of experiments we examined transgenic embryos carrying Ubi-DEcadherin-GFP (DE-cadherin-GFP expressed under the control of the ubiquitin promoter, hereafter referred to as cadherin-GFP) and/or UAS-GFP^nls (GFP containing an NLS under the control of the GAL4 upstream activating sequence, hereafter referred to as nuclear-GFP) in the AS (Fig. 1A and B; see Materials and Methods for full stock descriptions). The ability to visualize the apical-lateral membrane via cadherin-GFP expression permitted us to identify AS cells in the process of extrusion, while nuclear-GFP expression allowed us to observe the fate of the nuclear compartment during extrusion. A typical extrusion event was recognized in time-lapse movies by a reduction in the apical surface area of one particular cell relative to its neighbors, which form a rosette structure of five or six cells surrounding the extruding cell. Most extrusion events involved a single AS cell, although twin extrusions were occasionally observed with a rosette of seven neighboring cells. Most extrusions were found to occur in the anterior half of the AS (Movie 1).

Figure 1.

The extrusion of AS cells during dorsal closure is associated with apoptotic morphology, caspase activity, and autophagy. (A) Montage of a time-lapse confocal live imaging sequence showing AS extrusions in an embryo expressing cadherin-GFP. Extruding cells are shaded yellow and minutes of elapsed time are indicated in the lower left of each frame. (B) Montage of a time-lapse confocal live imaging sequence showing an AS extrusion event in an embryo expressing cadherin-GFP and nuclear-GFP. The upper and lower frames represent separate slices of the Z-stack separated by a depth of 1.75 µm. Minutes of elapsed time are indicated in the lower left of each frame. The apical surface of the extruding cell disappears between minutes 8 and 10 (upper frames) and apoptotic morphology is apparent between minutes 10 and 12 (lower frames). The width of each frame is approximately 20 µm. (C) Montage of a time-lapse confocal live imaging sequence of the AS in an embryo expressing the caspase sensor Apoliner under control of LP1-GAL4. Each frame is a Z-stack projection and elapsed time in minutes is shown in the lower left of each frame. The Apoliner-RFP signal is blue and the Apoliner-GFP signal is yellow. At the onset of dorsal closure (0 min) RFP and GFP are mostly colocalized (Apoliner-negative). As dorsal closure proceeds there is a gradual separation of the RFP and GFP signals as GFP becomes nuclear (Apoliner-positive); strong Apoliner-positive cells are clearly seen during mid-dorsal closure stages (80 min). A cell that is extruded is marked by a white arrowhead and shows apoptotic morphology at 140 min. The scale bar in the lower right of the panel represents 20 µm. (D) Montage of a time-lapse imaging sequence showing the accumulation of Atg8a puncta in a cell that is undergoing epithelial extrusion in an embryo expressing DT-Atg8a under NP3312-GAL4 control. The mCherry-Atg8a signal is blue and the GFP-Atg8a signal is yellow. Elapsed time in minutes is shown in the lower left of each frame and the zero time-point corresponds to the onset of dorsal closure. The upper two rows show projections of the entire Z-stack, while the lowermost row shows projections of the basal half of the Z-stack in which the apoptotic morphology of the extruding cell is more apparent (frames labeled as time b). The scale bar in the lower right corner represents 10 µm.

Our imaging results revealed changes in the nuclei of extruded cells consistent with canonical apoptotic morphology. In particular, leakage of nuclear-GFP from the nucleus was observed and immediately preceded nuclear fragmentation. Given that nuclear-GFP was no longer contained within the nucleus we were serendipitously able to observe cytoplasmic blebbing and cellular fragmentation—also hallmarks of apoptotic morphology. Nuclear fragmentation was examined with respect to the timing of the loss of apical surface area. Interestingly, a full-blown apoptotic morphology was never seen prior to the loss of apical surface of the extruding cell. The onset of subsequent nuclear fragmentation was rapid, occurring less than four minutes following the loss of apical-lateral contact with neighboring cells (Fig. 1B; Movie 2). Consistent with an apoptotic morphology, we found the ultimate fate of extruded AS cells to be engulfment by phagocytic macro-phages (data not shown) or alternatively, if extruded prior to the completion of midgut closure, AS corpses were also observed to be engulfed by the yolk. The engulfment of extruded cells by the underlying membrane-bound yolk (also known as the yolk sac) was readily observed in embryos expressing basigin-GFP (Movie 3). The phagocytic capability of the yolk has not been reported previously but is consistent with the localization of the macrophage receptor croquemort on the yolk membrane as well as expression of croquemort-GAL4 in the yolk (data not shown; see Fig. 5d of Franc et al.¹²).

AS epithelial extrusion events are associated with enhanced caspase activity and autophagy.

To demonstrate that epithelial extrusion events are associated with caspase activity, we used the recently described in vivo caspase sensor known as Apoliner.¹³ Briefly, Apoliner comprises a monomeric red fluorescent protein (mRFP) and enhanced green fluorescent protein (eGFP) tethered by an efficient and specific caspase-sensitive site. In addition, a transmembrane domain precedes the RFP moiety and the GFP moiety contains a nuclear localization signal. Thus, expression of UAS-Apoliner in live cells lacking caspase activity results in colocalization of RFP and GFP to membranes, a pattern that we refer to as Apoliner-negative. In live cells containing active caspases, however, the Apoliner-RFP and Apoliner-GFP fluorescence signals are separated; Apoliner-RFP remains bound to membranes while the untethered Apoliner-GFP component is translocated to the nucleus by virtue of its nuclear localization signal sequence (an outcome we describe as Apoliner-positive).

Using the UAS-Apoliner reporter in combination with an AS-specific GAL4 driver (LP1-GAL4) we were able to evaluate caspase activity in the AS throughout the process of dorsal closure. At the completion of germ band retraction and the start of dorsal closure the AS was uniformly Apoliner-negative (Fig. 1C, time 0 min). As dorsal closure progressed AS cells became Apolinerpositive, indicating caspase activation (Fig. 1C). AS cells that extruded during dorsal closure displayed intense nuclear Apoliner-GFP fluorescence relative to neighboring cells (Fig. 1C; Movie 4). Unlike the precipitous onset of apoptotic morphology following the loss of apical surface area, individual AS cells were frequently Apoliner-positive for over one hour without undergoing extrusion or acquiring apoptotic morphology. Interestingly, Apoliner-positive signal was observed to appear first in the anterior half of the AS, consistent with the location of most extrusion events (data not shown).

We next wished to examine the onset or possible upregulation of autophagy in the AS with respect to extrusion events. Autophagy can be detected via the subcellular localization of the Atg8 protein (the mammalian homolog of which is known as LC3). Briefly, Atg8 is uniformly distributed within the cytoplasm of cells that are not highly autophagic, but becomes punctate as it is processed, lipidated, and recruited the membranes of the phagophore, which matures to form the autophagosome, and subsequently the autolysosome (reviewed by He C, Klionsky DJ¹). In this study we initially tried using the RFP reporter UASp-mCherry-DrAtg8a (hereafter referred to as Cherry-Atg8a) to evaluate autophagy by live imaging confocal microscopy. While the Cherry-Atg8a fusion protein is less sensitive to the decreased pH of the autolysosome/lysosome and provides a more robust signal than its GFP counterpart,¹⁴ we found Cherry-Atg8a localization in the AS to be punctate throughout development. Moreover, while it was possible to perceive an increase in Cherry-Atg8a puncta throughout the AS as dorsal closure progressed, localized increases in Cherry-Atg8a puncta associated with extrusion events were not immediately obvious (data not shown).

An improvement on Cherry-Atg8a has been the addition of GFP to create UASp-GFP-mCherry-DrAtg8a (hereafter referred to as DT-Atg8a, an abbreviation for dual-tagged-DrAtg8a).¹⁵ Since GFP fluorescence is attenuated by the low pH environment of the autophagosome/autolysosome, and because mCherry is resistant to low pH, newly formed autophagic structures are associated with colocalization of the tandem fluorescent proteins, whereas older autophagic structures are expected to appear as mCherry puncta without GFP colocalization. Thus, DT-Atg8a is preferable for detecting increases in autophagy and facilitates visualization of autophagic flux.^15,16

Using DT-Atg8a and the GAL4 driver NP3312 (see Material and Methods for stock descriptions) we were able to observe the accumulation of Atg8a puncta in the AS throughout dorsal closure. During the earliest stages of dorsal closure, immediately following the completion of germ band retraction, a few AS cells were observed to accumulate Atg8a puncta while most neighboring cells maintained uniform DT-Atg8a localization. These cells, conspicuous by their Atg8a puncta, were found to extrude from the epithelium and subsequently execute apoptosis (based on morphology) (Fig. 1D; Movie 5). As dorsal closure progressed, all AS cells displayed punctate DT-Atg8a localization; extrusion events that could be discerned were associated with an increase in punctate DT-Atg8a relative to neighboring cells. In general, however, mCherry-puncta were more plentiful but not as distinct as GFP-puncta; in addition, colocalization of GFP and mCherry puncta, which we interpret to represent autophagosomes and newly formed autolysosomes, was also observed. Intriguingly, we also frequently observed GFP puncta that were not colocalized with mCherry; this observation is explained by the rapid movement of the associated vesicular structures and the delay between image acquisitions in the GFP and RFP channels of our confocal imaging system. Unfortunately, this precludes quantitative analysis of autophagic flux in our system.

Given that we were able to recognize the apoptotic morphology of extruded cells in embryos expressing either Apoliner or DT-Atg8a, it was possible to compare the timing of the onset of caspase activation (as evaluated by the appearance of Apoliner-positive cells) as well as autophagy (evaluated by the appearance of colocalized DT-Atg8a or GFP-only Atg8a puncta) relative to apoptosis. The appearance of apoptotic morphology, which includes membrane blebbing and cell fragmentation, was more readily observed in animated projections of only the basal half of the acquired Z-stacks (Fig. 1D, timepoints 20–35b). Caspase activation was first apparent approximately 120 min prior to apoptosis, and AS cells were found to remain strongly Apoliner-positive for up to 60 min before extrusion and apoptosis (Fig. 1C). Autophagy, however, was found to be upregulated in cells destined to extrude only 20 min prior to the appearance of apoptotic morphology (Fig. 1D). Overall, these observations suggest that caspase activation precedes the upregulation of autophagy associated with AS epithelial cell extrusion.

Changes in caspase activation and autophagy by genetic manipulation can be detected in vivo.

Our live imaging analysis suggests that AS extrusion events are associated with caspase activation and elevated autophagy. We were next interested in determining if the onset and upregulation of autophagy is required for epithelial extrusion and the subsequent execution of apoptosis. Using the GAL4/UAS system we constructed several genetic stocks that permit the recovery of embryos in which UAS-reporters lead to decreased or increased caspase activity (using UAS-p35 and UAS-reaper, respectively). We also used UAS-reporters that are expected to result in decreased or increased autophagy (UAS-dINR^ACT and UAS-Atg1, respectively). Before proceeding to measuring extrusion rates, however, we wished to confirm any disruption of caspase activity and/or autophagy associated with these expression backgrounds, and embryos of all genotypes were imaged in the context of co-expression of either Apoliner or DT-Atg8a (see Materials and Methods for full descriptions of genetic stocks and crosses).

Examining late dorsal closure stages embryos permitted a qualitative comparison of caspase activation and autophagy. In addition, an index of GFP/RFP colocalization relative to control embryos was measured quantitatively for all genetic backgrounds. All observations in the following sections regarding caspase activity as indicated by Apoliner signal are supported by this method of quantitation (Fig. S3). At the late dorsal closure stage of developmental most AS cells in control embryos were Apoliner-positive (Fig. 2A). At the same developmental stage DT-Atg8a protein was punctate throughout the AS; colocalized GFP/mCherry puncta, as well as GFP-only puncta were evident (Fig. 2B), both of which we assume (as discussed above) indicate the presence of autophagosomes and newly formed autolysosomes. The inhibition of caspases by p35 expression resulted in a dramatic reduction of Apoliner-positive cells, but DT-Atg8a puncta were unchanged and possibly more prevalent than in control embryos (Fig. 2C and D). This observation supports the view that caspase activation is not required for autophagy onset or upregulation in the AS.

Figure 2.

Caspase activity and autophagy in the AS can be detected using Apoliner and DT-Atg8a biosensors. Representative Z-stack projections of embryos expressing either the caspase sensor Apoliner under the control of LP1-GAL4 (A, C, E, G, I, K) or the autophagy reporter DT-Atg8a under the control of NP3312-GAL4 (B, D, F, H, J, L) Dorsal views of late dorsal closure stage embryos are shown in all panels (with the exception of panel F which is an early dorsal closure stage;see text) with anterior oriented to the left. The scale bars in the lower right of each panel represent 20 µm. In all panels GFP signal is displayed as yellow and RFP signal is displayed as blue. Control embryos carrying only LP1-GAL4 and UAS-Apoliner show a mix of Apoliner-positive (white arrows) and Apoliner-negative (white arrowheads) cells (A). The co-expression of UAS-Apoliner with UAS-reporters encoding the caspase inhibitor p35 (C) or dINR^ACT (G) increases the proportion of Apoliner-negative cells. Similar expression of reaper (E) or Atg1 (I) results in Apoliner-positive cells throughout the AS. Atg1 M/Z mutants (K) also show Apoliner-positive cells and homogeneous localization of Apoliner-RFP signal. (The Apoliner-RFP channel is shown separately in Fig. S1.) Control embryos carrying NP3312-GAL4 and DT-Atg8a show punctate subcellular localization of mCherry and GFP signal throughout the AS (B). As discussed in the Results section, newly formed autophagic vesicles appear as yellow (GFP-only) or white puncta (GFP/mCherry co-localized). Co-expression of p35 (D) with DT-Atg8a is similar to the control. Co-expression of reaper in an early dorsal closure stage embryo results in DT-Atg8a puncta in most cells (compare with Fig. 1D at time-point 0). Embryos co-expressing dINR^ACT (H) show no DT-Atg8a puncta. Embryos co-expressing Atg1 (J) show DT-Atg8a puncta with accumulations of mCherry puncta. Atg1 M/Z mutant embryos show diffuse DT-Atg8a signal, but do display some GFP/mCherry co-localized puncta (black arrowheads).

Increased caspase activity was achieved by expression of the pro-apoptotic gene reaper. This led to robust Apoliner-positive cells throughout the AS (Fig. 2E). The Apoliner assay used the GAL4 driver LP1 whereas the DT-Atg8a assay was only effective using the stronger GAL4 driver NP3312 (see Materials and Methods for full stock descriptions). Most embryos carrying NP3312 + UAS-reaper, however, underwent premature AS death and tissue degeneration, but some were observed in early dorsal closure stages with an intact AS, probably due to the excessive patchiness of the GAL4 expression in these particular embryos. At this early dorsal closure stage DT-Atg8a does not normally show any appreciable GFP/mCherry or GFP-only puncta (see neighboring cells at zero timepoint in Fig. 1D). The NP3312 + UAS-reaper early dorsal closure stage embryos, however, show robust DT-Atg8a puncta in most cells (Fig. 2F). These observations lead us to suggest that premature caspase activation in the AS can bring about premature upregulation of autophagy.

Similar analysis was performed on embryos expressing dINR^ACT, which is expected to inhibit autophagy through activation of the PI3K/TOR signaling pathway. In addition, embryos overexpressing Atg1 kinase, a key activator of autophagy acting downstream of TOR,¹⁷ were examined for enhanced autophagy. Expression of dINR^ACT resulted in a reduction of Apoliner-positive cells (Fig. 2G) and a striking lack of DT-Atg8a puncta of any kind (Fig. 2H). Conversely, Atg1 expression resulted in Apoliner-positive cells throughout the AS (Fig. 2I) and robust DT-Atg8a puncta (Fig. 2J). DT-Atg8a puncta associated with Atg1 expression were unusual in that the mCherry-Atg8a puncta were extremely large (Fig. S1O). Live imaging of these puncta at higher magnification and shorter intervals revealed them to be dynamic aggregates of many individual mCherry-Atg8a puncta (data not shown). We interpret these structures to be accumulations of autolysosomes and lysosomes indicative of an increased amount of cargo processed by autophagy and delivered to the lysosome.

Overall, the above data confirmed that we were able to effectively upregulate and downregulate caspase activation and autophagy in the AS during dorsal closure. Before measuring extrusion rates in these backgrounds, however, we wished to revisit the apparent lack of caspase activation in embryos expressing dINR^ACT—a surprising result we previously interpreted as supporting the view that autophagy is a prerequisite for caspase activation in the AS.¹⁰ A caveat of this interpretation, however, was that dINR^ACT could inhibit caspase activation independent of autophagy inhibition. As mentioned above, in the context of epithelial extrusions, and using the biosensors Apoliner and DT-Atg8a, caspase activation appears to precede autophagy upregulation. To address this apparent incongruity we turned to loss-of-function mutants for Atg1 kinase.

The Atg1^Δ3D allele is a purported genetic null, and homozygous mutants survive embryogenesis and die during the pupal stage of development.¹⁸ The product of Atg1 is also required during oogenesis for egg development—although females carrying Atg1^Δ3D germline clones do produce a reduced number of eggs.¹⁹ The GAL4 driver NP5328, which is inserted within a large intron of the Atg1 gene,²⁰ is associated with AS expression starting at the late stages of germ band retraction and increasing throughout dorsal closure (Movie 6). This suggests that there is zygotic transcription of Atg1 in the AS during dorsal closure. Moreover, strong maternal loading of Atg1 mRNA (from supplemental data of Tadros et al.²¹) further suggests that analysis of the loss-of-function phenotype of Atg1 mutant embryos requires the construction of genetic stocks that permit the recovery and unambiguous identification of maternal/zygotic mutant embryos also carrying appropriate GAL4 drivers and the desired UAS-reporters.

Atg1^Δ3D maternal and zygotic mutants (Atg1 M/Z) carrying the LP1 driver with either Apoliner or DT-Atg8a were recovered using the hsFLP/FRT ovo^D germline clone technique (see Materials and Methods for complete stock descriptions and crosses). Apoliner-positive cells were clearly identified in Atg1 M/Z mutant embryos (Fig. 2K); DT-Atg8a localization was very homogeneous but did display discrete GFP/mCherry co-localized puncta (Fig. 2L; Movie 7). Interestingly, colocalized GFP/mCherry puncta of the Atg1 M/Z mutants were long lived (some puncta persisted for over two hours) and mCherry-only puncta were infrequent. This suggests that, regardless of the mechanism by which puncta arise in Atg1 M/Z mutants, there is little dissipation of GFP fluorescence and associated conversion of colocalized signal to RFP-only signal. This strongly suggests that there is little or no flux through the autophagic pathway in Atg1 M/Z mutants. An additional interpretation is that DT-Atg8a puncta are not as mobile as they are in control embryos. The attenuation of autophagy in Atg1 M/Z mutants was also confirmed by live imaging embryos that express the autophagy substrate GFP-Ref (2)P in the AS. Ref(2)P is the Drosophila ortholog of the mammalian protein p62 and, as a protein that is selectively degraded via autophagy, can be used to directly evaluate autophagic degradation.²² Using identical imaging conditions and staging, Atg1 M/Z mutants were found to be deficient in their ability to degrade GFP-Ref(2)P relative to wild type (Fig. 3).

Figure 3.

The degradation of GFP-Ref(2)P is impaired in the AS of Atg1 M/Z (C and D) mutants relative to wild-type (A and B). Mid dorsal closure stages (A and C) and late dorsal closure stages (B and D) are shown. The scales bars represent 20 µm.

The appearance of Apoliner in Atg1 M/Z mutants was also striking in that its localization was uniform and distinctly compartmentalized, and this was especially evident during early stages of dorsal closure. In comparing all Apoliner-RFP images (in which membranes are labeled by the transmembrane containing RFP moiety of Apoliner) Atg1 M/Z mutants displayed uniform localization whereas UAS-Atg1 expressing embryos displayed dramatically punctate Apoliner-RFP (Fig. S1). This suggested two possibilities: first, Atg1 M/Z mutants could be generally deficient in membrane trafficking; and/or second, given that the Apoliner-RFP signal resembles the pattern associated with the endoplasmic reticulum (ER) marker YFP-ER (Fig. S2), AS autophagy could be associated with prominent ER degradation. Ultrastructural analysis by TEM revealed that dorsal closure stage wild-type controls contain autophagic structures, numerous vesicles, and multivesicular bodies (Fig. 4A). Atg1 M/Z mutant embryos of the same stage were remarkably devoid of these structures (Fig. 4B). Interestingly, autophagic structures observed in wild-type control embryos were uniformly associated with ribosomes and whorls of ER decorated with ribosomes (Fig. 4D). Consistent with our live imaging observations concerning accumulations of mCherry-Atg8a puncta, numerous lysosomes and autolysosomes, some containing rER, were observed in late dorsal closure stage embryos expressing Atg1 (Fig. 4C). Autophagic structures were especially evident in late-stage embryos expressing p35, where distinct double-membrane autophagosomes containing rER whorls were observed (Fig. 4E). Our live imaging analysis suggested that Atg8a puncta are present but less frequent in Atg1 M/Z mutants. This observation is supported by TEM analysis as examples of autophagic structures, although infrequent, were observed in Atg1 M/Z mutants; interestingly, those autophagic structures that were observed did not contain rER whorls as in wild type (Fig. 4F). Overall, the analysis of Apoliner, DT-Atg8a puncta, GFP-Ref(2)P degradation, and TEM ultrastructure supports the interpretation that autophagy is disrupted but not completely abolished in Atg1 M/Z mutants; furthermore, caspase activation occurs in the AS in the absence of Atg1 kinase and quantitative measurement of Apoliner GFP/RFP colocalization did not indicate any statistically significant difference between Atg1 M/Z mutants and control embryos (Fig. S3).

Figure 4.

TEM analysis shows a lack of vesicular structures in Atg1 M/Z mutants. Cross sections through wild-type (A) and Atg1 M/Z mutant (B) dorsal closure stage embryos. Wild-type embryos display autophagic structures containing ribosomes and rER (black arrowheads) as well as multivesicular bodies (upper white arrowheads). Multivesicular bodies are also evident in the underlying yolk (two lower white arrowheads). Embryos expressing UAS-Atg1 under LP1-GAL4 control contain clusters of lysosomes and autolysosomes (C), some of which contain ER (black arrowheads). Multivesicular bodies are also seen in these embryos (white arrowhead). Higher magnification micrographs show details of autophagic structures observed in wild-type (D) that were also abundant in embryos expressing p35 where definitive double membrane autophagosomes containing rER were observed (E). Atg1 M/Z mutants were largely devoid of any AS vesicular structures but rare autophagic structures observed in the AS did not contain rER (F). All scales bars represent 500 µm.

AS cell extrusion rates reflect caspase activation.

To determine if AS extrusions are associated with the regulation of PCD, and to determine if this is influenced by autophagy, we measured extrusion rates in embryos associated with increased or decreased caspase activity as well as increased or decreased autophagy. All extrusion rates are reported in Table 1. Control embryos carrying the LP1-GAL4 driver and cadherin-GFP, but without a UAS-reporter, were found to have an average extrusion rate of 0.133 extrusions per minute (corresponding to 10.2% of AS cells). The AS cell extrusion rate increased to 0.287 extrusions per minute when UAS-Atg1 was introduced into the same genetic background. The AS epithelium of embryos expressing UAS-dINR^ACT was slightly disorganized (data not shown) but displayed no extrusions. To further validate these results we performed identical quantitative analysis on embryos carrying either the pro-apoptotic UAS-reaper or the caspase inhibitor UAS-p35. As expected, extrusion rates were elevated by the expression of the pro-apoptotic gene reaper (0.283 extrusions per minute) and were reduced to negligible levels (0.008 extrusions per minute) by the expression of p35. Other pro-apoptotic UAS-reporters were also tested, including other UAS-reaper insertions, UAS-hid, and UAS-grim; in all cases a massive premature death of the AS and a failure in germ band retraction was observed and extrusions could not be scored—presumably due to higher levels of reporter expression associated with these different insertion lines (data not shown). Overall, this analysis supports the view that AS extrusion rates can be used to evaluate caspase activity in the AS.

Table 1.

Extrusion measures corresponding to genotypes examined

Parental Cross (female X male genotype*)	Extrusion Rate (events/min)	Embryos Examined (n)
Ubi-DEcadherin-GFP; LP1 X y w1118	0.133 (SD 0.03)†	6
Ubi-DEcadherin-GFP; LP1 X UAS-rprC14	0.283 (SD 0.05)	8
UAS-p35 (2 copy) X Ubi-DEcadherin-GFP; LP1‡	0.008 (SD 0.001)	4
Ubi-DEcadherin-GFP;LP1 X UAS-Atg1 (2 copy)	0.287 (SD 0.11)	4
Ubi-DEcadherin-GFP;LP1 X UAS-dINRACT	0.000 (SD 0.00)	3
hsFLP /+; Atg1Δ3D FRT-2A / ovoD-3L FRT-2A X ovoD-3L FRT-2A; Atg1Δ3D LP1/ TM3, twiG	0.126 (SD 0.07)	4

^†Standard deviation in parentheses.

^‡Reciprocal cross performed as one of UAS-p35 insertions maps to X chromosome.

^*See Materials and Methods for complete stock description.

The analysis of extrusion rates in Atg1 M/Z embryos was complicated by a variable phenotype; approximately half of all Atg1 M/Z mutants examined were morphologically abnormal, showing deep involutions in the lateral epidermis and defects in germ band retraction and head involution. In some embryos germ band re-extension was observed following successful germ band retraction (Movie 8). In these morphologically abnormal embryos it was not possible to measure extrusion rates although extrusions were observed. The extrusion rate of the morphologically normal Atg1 M/Z mutants (0.126 extrusions per minute) was close to the wild-type control value. In either case, morphologically normal or abnormal, Atg1 M/Z mutants were uniformly embryonic lethal.

Discussion

The degeneration of the extra-embryonic AS epithelium during Drosophila embryogenesis is a convenient model for live imaging PCD. Recently developed tools such as transgenic lines carrying UAS-Apoliner and UAS-GFP-mCherry-Dr-Atg8a represent effective biosensors of caspase activation and autophagy, respectively. In addition, scanning confocal live imaging techniques can be used to produce a high resolution of the behavior of the relatively large AS cells during the process of dorsal closure. AS death can be thought of as occurring in two phases: the first phase during dorsal closure involves the removal of approximately 10% of AS cells by epithelial extrusion; the second phase occurs after the completion of dorsal closure and involves the elimination of the remaining 90% of the tissue. Here we describe the process of AS cell extrusion and its relationship to caspase activity and autophagy. This study also provides further insight into the complex relationship between the regulation of apoptosis and autophagy.

A significant conclusion of this study is that epithelial extrusion events in the AS are caspase dependent and autophagy independent. The supporting data for this interpretation are the virtual absence of extrusions in embryos expressing the caspase inhibitor p35, combined with the observation that the average extrusion rate in Atg1 M/Z mutants is not appreciably different from the wild-type control value. A second finding was that caspase activation precedes autophagy during extrusion. This second point was based on the timing of the onset of caspase activation or autophagy upregulation in retrospect to the appearance of apoptotic morphology of a cell destined to extrude from the epithelium. Last, a third interpretation of our data is that autophagy can promote extrusions, while caspase activation can promote autophagy. If, however, extrusions are considered a reliable indicator of caspase activity, an interpretation supported by our observations, then this third point could be amended to “autophagy can promote caspase activation, while caspase activation can promote autophagy.” Data supporting such mutual positive crosstalk are the increased extrusion rate associated with Atg1 expression and the increase in autophagy of cells following expression of the proapototic gene reaper. Overall, however, our interpretation is that autophagy is, in fact, not required for epithelial extrusion or caspase activation in the AS. This, of course, is subject to the caveat that Atg1 M/Z mutants may not be entirely autophagy-deficient.

How can autophagy promote caspase activation? In Drosophila, products of proapoptotic genes such as grim, reaper, hid and sickle function to either inhibit or promote the degradation of inhibitor-of-apoptosis proteins (IAPs).²³ The IAPs in turn bind to caspases, leading to caspase inhibition as well as caspase degradation by ubiquitin-mediated proteolysis.²⁴ IAPs can, therefore, be thought of as setting a threshold for activation of the caspase cascade. Thus, any specific degradation of IAPs would lower this threshold and promote caspase activation. In the Drosophila ovary such a link has been described where the IAP protein dBruce is degraded by autophagy, leading to activation of the caspase drICE and subsequent DNA fragmentation mediated by dCAD.¹⁵

How can caspase activation promote autophagy? There are at present no clear data to establish a regulatory mechanism by which caspase activation leads to autophagy induction in Drosophila. However, experimental evidence, again from analysis of PCD in the Drosophila ovary, demonstrates that expression of the caspase Dcp-1 in the germline of well-fed flies is capable of inducing autophagy.¹⁹ The induction of autophagy in this context occurs without starvation and represents a possible role for the caspase Dcp-1 in coordinating autophagy and apoptosis. An emerging theme from these studies is that the relationship between autophagy and apoptosis is highly dependent on cellular context, such as tissue type, developmental stage, or nutrient status of the organism.

The regulation of autophagy and apoptosis in Drosophila may also be linked through the activity of the proapoptotic gene hid. One of the founding members of the proapoptotic genes, all of which harbor a small N-terminal motif known as the IAP-binding motif (or IBM), the HID protein stands out as being subject to inhibitory phosphorylation through the RAS/MAPK pathway.²⁵ Interestingly, overexpression of hid in several larval tissues has been reported not to induce apoptosis, as might be expected, but to induce autophagy.²⁶ In addition, hid was also identified as a positive regulator of autophagy in Drosophila l(2)mbn cells.¹⁹ Thus, an attractive model relating to the context of PCD in the AS is that hid is a key regulator that can function to promote both caspase activation as well as autophagy. In the context of the AS, however, there must be redundant pathways to promote autophagy because mutant embryos deleted for the proapototic genes hid, grim, reaper and sickle display a persistent AS, do not execute extrusions, but do display an autophagic AS.¹⁰

Our analysis of Atg1 M/Z mutants revealed a variable phenotype, the basis of which is unknown. One explanation, however, could relate to our method of inducing Atg1 germline clones, which involved expression of FLPase under the control of a heat shock promoter. In addition to producing clones in the germline, this approach also produces clones in all other mitotically active tissues, including the somatic follicle cells that surround the developing egg chamber. An unusual interaction between the follicle cells and germline was recently described for Atg1 mutants, wherein Atg1 mutants show normal oogenesis, but Atg1 mutant follicle cells in the context of a wild-type germline result in abnormal oogenesis.²⁷ The interpretation of this study was that autophagy is required within the follicular epithelium for normal egg development as well as for proper communication between the follicle cells and the germline. It is, therefore, possible that variation in the size or position of Atg1 mutant follicle cell clones in our experiment could elicit a maternal effect, although to our knowledge a maternal effect associated with mosaicism of the follicular epithelium has never been described. Interestingly, whereas our Atg1 M/Z mutants were embryonic lethal, the same study reported that Atg1 M/Z mutants produced by pole cell transplantation displayed lethality in late larval stages.

The process of AS epithelial extrusion has been reported in several studies but there has, in general, been little research conducted on the topic of epithelial extrusion. The measurement of AS apoptosis rates during dorsal closure has been found to directly correlate to the rate of dorsal closure; in addition, delamination of apoptotic AS cells was suggested to contribute toward the generation of a mechanical force that facilitates extrusion while promoting dorsal closure.¹¹ Our data are consistent with this study in that both show a complete absence of extrusions in the context of caspase inhibition and elevated extrusions in the context of pro-apoptotic gene expression. Our data are also consistent with respect to the observation of increased extrusions in the anterior AS. There is, however, a discrepancy in control extrusion rates, which differ approximately 2-fold and where our data show lower rates. It is possible that this discrepancy is associated with the different GFP constructs used to facilitate live-imaging confocal microscopy (GFP-moesin which localizes to F-actin verses cadherin-GFP which localizes to the adherens junction). It is also possible that cadherin-GFP is, itself, associated with reduced extrusion. Our imaging experiments, however, were all performed on embryos carrying a single copy of cadherin-GFP and our measurements, while internally consistent, are likely not directly comparable to the GFP-moesin data.

A curious observation of this study was the finding that expression of dINR^ACT resulted in a striking decrease in both autophagy and caspase activation. The effect of dINR^ACT on autophagy was expected, given that INR/PtdIns3K activity is well established as a mechanism leading to the inhibitory phosphorylation of the key autophagy regulator Atg1 kinase by TOR kinase.²² The incongruity of caspase activation in the AS of embryos expressing dINR^ACT compared with Atg1 M/Z mutants, both of which downregulate autophagy, strongly suggests that INR is capable of impeding the activation of caspases through a pathway that is parallel or upstream of Atg1. One possibility is that dINR^ACT expression activates RAS, leading to an inhibitory phosphorylation of HID by MAPK. RAS activation by INR is well established in mammalian systems. Interestingly, interaction of INR with components of the RAS/MAPK pathway has been reported in the context of photoreceptor differentiation and the timing of neuronal differentiation during Drosophila retinal development, but here a direct relationship has not been established.²⁸ The possibility of a direct interaction between INR and RAS in the context of the autophagic AS warrants further investigation.

Our analysis of autophagy in the AS of the Drosophila embryo also revealed that most autophagic structures observed by TEM are associated with rER. This was further supported by live imaging analysis where the RFP moiety of the caspase sensor Apoliner, which contains a transmembrane domain and serves to label membranes—including the ER compartment—showed dramatic differences between embryos associated with reduced and enhanced autophagy. The term ER-phagy describes the specific degradation of this cellular compartment and has been linked to ER stress and the unfolded protein response (UPR) pathways.²⁹ Most interpretations of UPR and autophagy, however, place the activation of autophagy via the TOR/Atg1 pathway downstream of ER stress (reviewed by Yorimitsu T, Klionsky DJ³⁰). On the other hand, our observations suggest that ER-phagy is downstream of Atg1 kinase in the context of the AS. Clearly, further investigation is required to address the nature of ER-associated autophagic structures observed in the degenerating AS.

Materials and Methods

Genetic stocks and crosses.

All stocks were kept on standard Drosophila medium. Crosses and embryo collections were performed at 25°C. Flies carrying the X-chromosome markers yw¹¹¹⁸ were used as wild-type controls. With the exceptions of UAS-Apoliner, UAS-GFP-mCherry-DrAtg8a, basigin-GFP, and YFP-ER, all genetic stocks and associated insertions used in this study used are described in detail in Mohseni et al.¹⁰ The PTT basigin-GFP stock, also known as G289, is described in detail in Reed et al.³¹ The AS-specific LP1-GAL4 driver promotes expression almost exclusively in the AS from the onset of germ band retraction through dorsal closure and is also known to report for expression of the gene calderón (LP1-GAL4 has also been called cald-GAL4).³² The GAL4 driver NP3312 was obtained from the Kyoto Drosophila Resource Center and drives GAL4 expression under control of the gene hindsight. The YFP-ER marker P{w[+mC] = sqh-EYFP-ER}3 was obtained from the Bloomington Drosophila Stock Center. The design and rationale of the caspase sensor known as Apoliner is reported in Bardet et al.¹³; Atg1 M/Z experiments used a second chromosome insertion (P{UAS-Apoliner}5), other experiments used a third chromosome insertion (P{UAS-Apoliner}8) which was recombined onto the LP1-GAL4 driver chromosome. The UASp-GFP-mCherry-DrAtg8a line (abbreviated in this paper to DT-Atg8a) is described in Nezis et al.¹⁵

Recovery of Atg1 M/Z mutants.

The following chromosomes, listed with complete genetic nomenclature followed by abbreviations used in this paper (in bold) and their source, were used in the synthesis and analysis of Atg1 M/Z mutants:

• X chromosomes:

• P{ry[+t7.2] = hsFLP}1, y¹ w¹¹¹⁸ (= hsFLP) from Bloomington Stock Center;

• 2^nd chromosomes:

• P{Ubi-DEcadherin-GFP}#5 (= Ubi-DEcadherin-GFP) from H. Oda;

• P{UAS-Apoliner}5 (= UAS-Apoliner-5) from J.P. Vincent;

• P{UASp-GFP-mCherry-DrAtg8a} (= UAS-DT-Atg8a) from H. Stenmark;

• UAS-GFP-Ref(2)P^2L (= UAS-GFP-Ref(2)P) from T. Neufeld; 3^rd chromosomes:

• Atg1^Δ3D from T. Neufeld;

• Atg1^Δ3D P{w[+mW.hs] = FRT(w[hs]) }2A (= Atg1^Δ3D FRT-2A) from S. Gorski;

• P{w[+mC] = ovoD1-18}3L P{w[+mW.hs] = FRT(w[hs])}2A (=ovoD-3L FRT-2A) from Bloomington Stock Center;

• Atg1^Δ3D LP1-GAL4 (Atg1^Δ3D LP1) recombinant chromosome made in Reed lab;

• TM3, P{w[+mC] = GAL4-twi.G}2.3, P{UAS-2xEGFP}AH2.3, Sb Ser (= TM3, twiG)a “GFP balancer” from Bloomington Stock Center.

Descriptions of standard balancer chromosomes and genetic markers not listed above are found on flybase or Bloomington Stock Center websites (see http://flybase.org or http://flystocks.bio.indiana.edu/Browse/balancers/balancers.htm).

In the following section the crossing scheme described uses the abbreviations as defined above; boldface font is also used to help distinguish chromosome descriptions from normal text. To recover Atg1^Δ3D germline clones hsFLP; ; Dr/ TM3, Sb females were crossed to ovoD-3L FRT-2A/ TM3, twiG males. Male progeny of this cross of the genotype hsFLP/Y; ovoD-3L FRT-2A/ Dr were then crossed to Atg1^Δ3D FRT-2A/ TM3, Sb virgin females. The mated adults were transferred onto fresh food every two days. When progeny of the mated adults were at the third larval stage or early pupal stage they were heat shocked by immersing the culture vials in a 37°C water bath for 30 min on two consecutive days. Adult virgin females that of the genotype hsFLP/+ ; ovoD-3L FRT-2A/ Atg1^Δ3D FRT-2A were recognized by the absence of both the Dr and Sb dominant markers. Such females were mated to males of the following genotypes for the following purposes: Ubi-DEcadherin-GFP; Atg1^Δ3D / TM3, twiG (for imaging epithelial cells and extrusion); UAS-Apoliner-5/ +; Atg1^Δ3D LP1/ + (for imaging caspase activity); UAS-DT-Atg8a / +; Atg1^Δ3D LP1/ + (for imaging autophagy); UAS-GFP-Ref(2)P / +; Atg1^Δ3D LP1/ + (for imaging autophagy). Note that the latter three genotypes are not balanced genetic stocks; males were recovered by crossing appropriate parental stocks. Since there is no meiotic recombination in Drosophila males, the expression of the reporter from the LP1-GAL4 driver also served to identify the zygotic mutants, which must also carry the Atg1^Δ3D allele of the recombinant chromosome. Zygotic mutants of the first cross were recognized by absence of the GFP balancer TM3, twiG. Using the dominant female sterile ovoD technique, all eggs from the heat shocked hsFLP/+ ; ovoD-3L FRT-2A/ Atg1^Δ3D FRT-2A females were, by necessity, derived from germline clone homozygous for Atg1^Δ3D and therefore maternal mutants. Control crosses, set up using virgin females from cultures that were not heat shocked, were uniformly sterile.

Preparation of embryos for live imaging and confocal microscopy.

Some embryos were prepared for live imaging as previously described (Fig. 1A and B; Movies 2 and 3).^10,31 For all other experiments embryos were prepared using a hanging drop protocol, the details of which are described in detail elsewhere.³³ We developed the hanging drop method in part to facilitate the analysis of epithelial extrusion, as we were concerned that any compression of the mounted embryo might affect extrusion rates and represent an undesirable source of variability in our experiments.

Some of the time-lapse confocal imaging presented was performed as previously described (Fig. 1A and B; Movies 2 and 3).^10,31 All other timelapse confocal microscopy was performed using a Nikon Eclipse 90i microscope fitted with a Nikon D-eclipse C1 scan head. 20X Plan Apo VC (NA 0.75) or 40X Plan Fluor oil (NA 1.30) objectives were used. Images were collected and saved as animated projections using the Nikon EZ-C1 3.70 software. Z-stack acquisitions for most embryos consisted of 6–8 steps, covering a range of 25–40 µm, with steps sizes usually in the range of 3–5 µm. Further processing and compression of movies was performed using ImageJ software (public domain, NIH).

Calculation of extrusion rates.

Three minute time-lapse intervals were found to be optimal for observing extrusion events while maintaining embryo viability during imaging sessions of 3–4 h. For calculating extrusion rates, movies of embryos undergoing dorsal closure were examined frame by frame. Cells observed to undergo a conspicuous reduction in apical surface area relative to its neighbors were scored as extrusion events. Cumulative totals of extrusion events observed in control embryos (progeny of cadherin-GFP; LP1-GAL4 females crossed to yw¹¹¹⁸ males) were plotted vs. elapsed time. A statistical analysis of extrusion data for control embryos showed that the relationship between extrusion events and elapsed time is sufficiently described by a linear model for all embryos examined (range of adjusted R-square values: 0.869 to 0.961; F-test indicated significance at p < 0.05). Extrusion rates were subsequently calculated by dividing the total number of extrusions by the elapsed time in minutes. In calculating extrusion rates the time of the first extrusion was used as the zero time point and a minimum total elapsed time of 100 min was scored for extrusion events.

Transmission Electron Microscopy (TEM).

All TEM, including embryo collection, fixation and embedding, was performed as previously decribed.¹⁰

Abbreviations

AS

amnioserosa

eGFP

enhanced green fluorescent protein

GFP

green fluorescent protein

INR

insulin receptor

mRFP

monomeric red fluorescent protein

NLS

nuclear localization signal

PCD

programmed cell death

PtdIns3K

phosphoinositide-3-kinase

rER

rough endoplasmic reticulum

RFP

red fluorescent protein

TEM

transmission electron microscopy

YFP

yellow fluorescent protein

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.