Ring canals in the larval adipose of Drosophila buffer stress response

Abstract

Cells in metabolically active tissues with high biosynthetic and secretory demands often use robust stress-responsive mechanisms to maintain homeostasis. Coordinating such robust stress response mechanisms requires intercellular communication and coordination. Such modalities of intercellular communication have been relatively understudied in the context of stress tolerance. Here, we use the Drosophila melanogaster third instar fat body to demonstrate that adipocytes communicate with each other through intercellular bridges called ring canals to buffer endoplasmic reticulum stress. The fat body supports the exponential growth from embryo to late larval stage over a short period of time through its energy storage and secretory functions, enduring a high basal level of stress in the process. We discovered that individual cells in the fat body are paired to one neighboring cell through ring canals. We further demonstrate that ring canals mediate rapid and highly specific intercellular cargo and organellar trafficking, and allow the transport of cytoplasmic, ER-bound and Golgi vesicular proteins. Disrupting fat body ring canals resulted in higher levels of stress response markers, aberrant cell size, as well as increased cell sensitivity and lethality in response to various exogenous stressors. We also find that animals with disrupted fat body ring canals display an overall delay in larval development, likely due to reduced secretion of larval serum proteins from the fat body. In sum, our work reveals a novel feature of intercellular communication in adipose tissue that serves to buffer stress across cells which is required for both homeostatic secretory function and maintaining tissue viability under exogenous stress.

eTOC:

Whether and how individual cells within a tissue cooperate with each other to respond to cellular stress has remained elusive. Nandakumar et al show that Drosophila adipocytes transport select subcellular components across intercellular bridges with their neighbors, and that such transport is required for tissue resilience to various stressors.

Introduction

Metabolically active tissues such as adipocytes and hepatocytes have high biosynthetic and secretory burdens. Due to their secretory demands, they also rely heavily on pathways that maintain endoplasmic reticulum (ER) homeostasis, such as the unfolded protein response (UPR) for their function (Ryoo, 2024; Han and Kaufman, 2016). However, chronic activation of stress response pathways can result in pathologies, metabolic dysfunction, and ultimately, cell death. Thus, for a tissue to function properly under basal stress levels without invoking pathological stress programs, proper coordination and intercellular communication within a tissue is required. Intercellular communication within a tissue can be achieved through a number of avenues including short-range (physical connections such as cell-cell junctions (Cavey and Lecuit, 2009)), and cytonemes (González-Méndez et al., 2019)) and long-range methods (such as secreted molecules and hormones (Nässel and Zandawala, 2020)). How and whether these intercellular communication modalities support tissue-wide stress tolerance is relatively understudied. Here, we use the Drosophila adipose tissue, also called the ‘fat body’, to address this open question.

The larval fat body has several vital secretory, endocrine, innate immune, and lipid storage functions(Tennessen and Thummel, 2011). In addition to these functions, the larval fat body supports the growth of the embryo to pupa over a short developmental period of 120h (Church and Robertson, 1966). To support such exponential growth during development, adipocytes themselves undergo multiple rounds of endocycling which involve alternating growth and DNA synthesis phases to grow in cell size but not in number (Colombani et al., 2003; Smith and Orr-Weaver, 1991; Pierce et al., 2004). By the last larval instar stage, the fat body is organized as multiple lobes across the entire body plan, comprising approximately ~2,000 cells each with an average ploidy of 255C (Nordman et al., 2011). Across all these stages, but particularly in the third instar stage, the fat body secretes vast amounts of proteins and lipids into the hemolymph to support the growth of all the developing tissues, as well as in preparation for pupariation (Powell et al., 1984; Valzania et al., 2024; Heier et al., 2021; Jowett et al., 1986). Metabolic tissues with high secretory activity, such as the Drosophila fat body, often display high basal levels of several stress response markers. We and others have shown that the third instar larval fat body shows constitutive basal activation of ER stress, amino acid deprivation, and oxidative stress amongst others (Huang et al., 2017; Kang et al., 2015, 2017; Sone et al., 2013; Saeedi et al., 2022; Yamada et al., 2023). Whether and how the stress burden is coordinated across this tissue to cope with the constant secretory demand and to maintain metabolic function has not been studied. Here, we report that individual adipocytes in the larval fat body are connected to exactly one neighboring cell through an intercellular bridge (called ‘ring canal’ in Drosophila), and demonstrate that this connection is required for buffering stress and maintaining tissue function.

Intercellular bridges are formed at the end of a cell division cycle when, following an incomplete cytokinesis, the cytokinetic furrow is stabilized to form a stable connection between two daughter cells (Lu et al., 2017; Chambaud et al., 2024). Such structures are well conserved across multicellular organisms, albeit known by different names across popular model organisms; septin rings in yeast (Kukhtevich et al., 2024), ring canals in Drosophila (Robinson et al., 1994; Robinson and Cooley, 1996), intercellular bridges in mammals and other vertebrates (Andreuccetti et al., 1999; Mullins and Biesele, 1973; Fawcett et al., 1959), and as plasmodesmata in plants (Cilia and Jackson, 2004; Zani and Edelman, 2010). The function of intercellular bridges in animals is best studied in the context of the germline (Price et al., 2023; Lee et al., 2013; Airoldi et al., 2011; Roth and Lynch, 2009; Kloc et al., 2004). In the Drosophila female germline, ring canals allow trafficking of maternally loaded mRNAs and proteins synthesized in the nurse cells into the developing oocytes (Shaikh et al., 2024; Bernard et al., 2018; Hsu et al., 2015; McLean and Cooley, 2014; Roth and Lynch, 2009; Petrella et al., 2007). In the male germline, they function to synchronize meiotic division and development of male spermatids (Kaufman et al., 2020; Eikenes et al., 2015; Miyauchi et al., 2013; Montembault et al., 2010). Ring canals have been described in a few other somatic tissues such as the Drosophila wing disc in larvae (McLean and Cooley, 2014), main cells of the male accessory glands (Box et al., 2024) and in follicle cells of adult female ovary (Airoldi et al., 2011; McLean and Cooley, 2013), but their specific function in each of these tissues remains unstudied. Our work is not only the first description of ring canals in adipocytes but also uncovers a novel function for ring canals in a somatic tissue with critical roles in development.

Results

Pairs of Drosophila larval adipocytes are connected via ring canals

When studying the stress tolerance of the fat body using a transgenic reporter that reads out the activity of integrated stress response signaling (ISR) (4E-BP^intron-DsRed, (Kang et al., 2017)), we observed a pattern where neighboring adipocytes appear to display similar levels of reporter gene expression (Fig. S1A), suggesting that the cells may somehow be paired to one another. This pattern resembled the expression pattern of transgenes in other somatic tissues of Drosophila which have been shown to contain ring canals, such as the accessory glands (Box et al., 2024) in adult males and follicle cells in adult female ovaries (Airoldi et al., 2011; McLean and Cooley, 2013). We thus sought to test if the neighboring adipocytes in the fat body were connected via ring canals using the pavarotti-GFP (pav-GFP) transgenic reporter (Airoldi et al., 2011). Pavarotti is a kinesin family protein (with close homology to mammalian KIF23) which is known to localize to ring canals in many Drosophila tissues (Bassi et al., 2013; Carmena et al., 1998; Eikenes et al., 2013). We found pav-GFP puncta throughout the wandering third instar fat body and present exclusively on the cell membrane (Fig. 1A-F). We determined that individual adipocytes only have one ring canal puncta per cell, indicating that pairs of adipocyte nuclei are potentially connected through ring canals. We next examined whether such putative ring canals are present throughout the fat body. Using fluorescence confocal imaging, we determined that adipocytes in all lobes of the larval fat body are paired with exactly one neighbor via a pav-GFP puncta (Fig. 1A-F). While we observed ring canals in both male and female fat bodies, we chose to utilize females for the rest of this study since they express higher basal levels of 4E-BP^intron-based reporters (Grmai et al., 2024).

Figure 1: Super resolution microscopy reveals pav-GFP puncta found throughout the Drosophila larval fat body on adipocyte cell membranes and represent open ring canals.

Schematic (left). Cartoon of the larval fat body (made using Biorender) displaying orientation of the Drosophila larval fat body with anterior-posterior orientation denoted. Dotted and labelled rectangles show the regions of the fat body that were imaged in panels A-F’’’.

A-F. Confocal images of wandering third instar larval fat bodies expressing pav-GFP (green) and counterstained with Phalloidin (magenta) to mark cell membranes. Note that pav-GFP localizes to the nucleus, putative ring canals and midbody remnants. Inset regions marked by white rectangles in panels A’’-F’’ are magnified and shown in A’’’-F’’’. Yellow arrows in A’’’-F’’’ point to minute pav-GFP puncta between two adipocyte nuclei.

G-H. Representative high magnification (93X Objective) confocal image of pav-GFP (green) puncta from Fig. 1A with STED super resolution (G-G’’). Cell membranes are counterstained with phalloidin (magenta). H-H’’’’ show deconvolved image from G-G”.

I-K. STED super resolution and deconvolution confocal images of pav-GFP puncta from fat bodies at different developmental timepoints corresponding to 48h after egg deposition (AED) (I-I’’), 72h AED (J-J’’) and 112h AED (K’-K’’). Yellow dotted lines outline the inner and outer circumference of the ring canal in I”-K”.

L-H. Measurements of the inner and outer circumference of ring canals from super-resolved images in C-E across larval development at indicated timepoints. n=10 ring canals from ten animals

Scale bars: A-F, A’-F’, A’’-F’’ = 50μm; A’’’-F’’’ = 5μm; G-H’’ = 1μm; I-K = 0.25μm ns= not significant; ****=p<0.00001; ***=p<0.0001; *=p<0.01 here and in all other figures.

Ring canals are open structures, and their size remains unchanged throughout larval development

With the goal of ultimately understanding the specificity of ring canal-mediated intercellular communication, we wanted to determine whether the pav-GFP puncta represented open ring-canals or simply midbody remnants. Midbody remnants are vestiges of cytokinesis, and sometimes remain intact in daughter cells after cell division has occurred (Price et al., 2022). Due to their size, we were unable to resolve whether the GFP puncta represent open ring canals or midbody remnants even at the highest magnifications of traditional, high-resolution confocal microscopy (Fig. 1G). To understand the dimensions and structure of the puncta better, we turned to STED super resolution (Vicidomini et al., 2018) followed by Huygens deconvolution, which revealed less pixel dense center of pav-GFP puncta resembling open ring canal structure, on the cell membrane (Fig. 1H). Scaled measurements revealed that the open ring canals are approximately 1.8-3μm in outer circumference with a 0.5-1.2 μm inner circumference (Fig. 1L-M).

Since the larval fat body is post-mitotic and the number of adipocytes is determined at the end of embryogenesis (Smith and Orr-Weaver, 1991; Pierce et al., 2004; Butterworth et al., 1988; Edgar and Orr-Weaver, 2001), we surmised that ring canal formation must have occurred early in fat body development. Consistent with this, we found that ring canals are present in all in larval fat bodies from 48h AED (hours after egg deposition) to previously tested wandering third instar larva at 112h AED (Fig. 1I-K). We next measured the dimensions of ring canals across development from 48h AED to 112h AED which represents the majority of larval growth stages. Surprisingly, despite these stages coinciding with exponential endocycle-mediated growth of individual adipocytes (Pierce et al., 2004; Guarner et al., 2017; Nordman et al., 2011), we found that ring canals don’t change in size between 48h AED and 112hAED (Fig. 1L-M). These results strongly suggest that ring canal formation occurs as a result of an incomplete cytokinesis at the end of the last mitotic cell cycle in late embryogenesis (Smith and Orr-Weaver, 1991; Pierce et al., 2004).

Since our discovery was propelled by visualizing similar levels of fluorescent reporter protein expression between paired cells (Fig. S1A), we asked whether paired cells display similar levels of other proteins. Such “equilibration” in cytoplasmic content has been previously reported in the ovary follicular epithelium where cells are connected by ring canals (McLean and Cooley, 2013). To test this, we used a fat body driver (Dcg-GAL4, (Asha et al., 2003)), to express cytoplasmic GFP (UAS-GFP) in adipocytes (Fig. S1B). We then measured integrated GFP intensity of paired cells and their non-sister neighbor cells in fat bodies from wandering third instar larvae 112h AED (Fig. S1B’-B’’’). We found that pairs of cells connected through ring canals show similar levels of GFP with ~10% or less deviation in GFP intensity (Fig. S1C, D). In contrast, non-sister neighboring cells displayed up to ten-fold deviation (up to >1000%) in GFP intensity (Fig. S1C, D), strongly suggesting that similar GFP levels is a defining feature of paired adipocytes.

Intercellular transport across ring canals shows cargo-bias

The equal levels of GFP across paired cells (Fig. S1B-D) indicate that sister cells may equilibrate cytoplasmic content by transporting cargo across ring canals similar to follicular epithelial cells (McLean and Cooley, 2013). Further, the maintenance of a relatively small ring canal size throughout larval stages of adipocyte growth (Fig. 1L-M) suggests that cargo transport across adipocyte ring canals is size-selective. To test candidate cargos that may be transported across ring canals, we devised a fluorescence recovery after photobleaching (FRAP) experimental paradigm. We also developed a protocol for ex vivo live imaging of fat body from wandering third instar larva (see methods) expressing these fluorescently labeled organelles for FRAP experimentation. Using a fat body driver (Dcg-GAL4, (Asha et al., 2003)), we drove the expression of various compartment-bound fluorescent proteins in adipocytes. To efficiently perform live-imaging on friable adipose tissue ex vivo while limiting non-experimental light exposure, we required a fast, reliable method to identify paired cells. Given the very small dimensions of the ring canals (<1μm), they are difficult to locate in live fat tissue, which can be up to 50μm thick in the Z axis. Thus, we utilized equivalent levels of Dcg>GFP intensity, which we demonstrated to be a feature of paired cells (Fig. S1B-D), to reliably guide our identification of paired cells for these experiments. We first tested whether cytoplasmic proteins could pass through the ring canal using a cytoplasmic GFP. We photo-bleached cytoplasmic GFP in one adipocyte, and measured changes in fluorescence over time in the bleached cell, the neighboring cell paired through a ring canal, and an “unpaired” adjacent neighbor not connected to the cell through a ring canal (Fig. 2A, Video 1). We observed that the bleached cell rapidly recovered GFP fluorescence in 30 seconds, with a concomitant decrease in fluorescence in the paired neighbor and no change in fluorescence in the unpaired neighboring cell (Fig. 2A’, A”). To ensure that such recovery in paired cells accounts only for ring canal mediated transport of cytoplasmic GFP and not de novo fluorescent protein biosynthesis, we photo-bleached both paired neighbors (Fig. 2B, Video 2). Across similar time scales of imaging as Fig. 2A, we found that neither of the bleached pair showed any recovery in fluorescence during the course of imaging (Fig. 2B’, B”). These experiments demonstrate that cytoplasmic content is continually and rapidly exchanged between adipocytes connected by ring canals.

Figure 2: Cytoplasmic, ER and Golgi-bound fluorescent proteins are transported rapidly through rig canals, but not mitochondrial proteins.

A-B. Fluorescent image montages showing time course of GFP signal over 4.5 minutes from Dcg-GAL4>UAS-GFP fat bodies subjected to photobleaching followed by recovery (FRAP) for several minutes. Panel A (from Video 1) shows montage corresponding to one bleached cell and panel B (from Video 2) shows experimental paradigm where both the adipocytes connected through ring canal are subjected to photobleaching (solid orange outlines). GFP fluorescence intensity is measured in the bleached cell (orange outline), paired neighboring cell (dashed orange outline) and a third neighboring cell (green outline) and plotted as recovery following photobleaching normalized to initial values in A’-B’ and as absolute values in A”-B”.

C-E. FRAP analysis on ER-bound KDEL-GFP (C, Video 3), Golgi bound-GFP (D, Video 4) and mitochondria-GFP (E, Video 5) as performed in A-B where GFP recovery is measured in the bleached cell (blue outline). Note that C-D were performed over a time course of six minutes and E was performed over 30 minutes. Quantifications in C’-E’ represent the mean change in GFP recovery post bleaching in the bleached cell. Standard deviation is plotted as the shaded error envelope. n=10 independent FRAP experiments from ten animals over 3 independent crosses. Quantification in C”-E” are absolute GFP intensity values from images in C-E.

Scale bars for all panels = 50μm

Since the fat body is a highly metabolically active and secretory tissue, we next tested whether mitochondria, ER, or Golgi compartments maybe transported across ring canals. To do so, we used Dcg-GAL4 to drive expression of either UAS-KDEL-GFP (marking the ER), UAS-Golgi-GFP (marking Golgi vesicles) or UAS-mito-GFP (marking mitochondria). Similar to the cytoplasmic GFP, we found that both KDEL-GFP and Golgi-GFP are transported rapidly (in under three minutes) across paired neighbors (Fig. 2C-D, Videos 3-4). We ensured that the KDEL-GFP and Golgi-GFP utilized in these experiments faithfully marked the ER and Golgi respectively as reported before using organelle-specific markers (Fig. S2A-B)(Okajima et al., 2005; Cole et al., 1996). In contrast, the mitochondrial GFP did not recover at even 30 minutes after photobleaching (Fig. 2E) suggesting that mitochondria likely do not get transported through ring canals in third instar larval adipocytes. These findings demonstrate that cargo transport across paired adipocytes is selective and permits transfer of secretory pathway components.

Kelch, pnut, and septin1 are required for maintaining fat body ring canal integrity

Previous work from our lab and others has shown that the developing larval fat body endures high levels of intrinsic ER, nutrient deprivation, and oxidative stress, due to the nature of its function (Kang et al., 2015, 2017; Sone et al., 2013; Yamada et al., 2023; Saeedi et al., 2022). Since we found intercellular exchange of secretory pathway components, we hypothesized that ring canals may play a role in regulating ER stress across paired adipocytes. To test this, we sought to disrupt ring canal structure and examine the consequences of such disruption. The composition and size of the intercellular bridges is cell type dependent. In Drosophila, the largest ring canals connect nurse cells in the female germline and contain actin among dozens of other structural proteins (Kelso et al., 2002; Hudson and Cooley, 2002; Warn et al., 1985) but in contrast, the male germline ring canals do not contain actin (Hime et al., 1996; Gerdes et al., 2020; Haglund et al., 2011). Ring canals have been demonstrated to require several protein complexes to both be established, and to maintain their ‘openness’, including actin remodeling complexes, GTP-binding cytoskeletal proteins such as septins, plasma membrane bound regulators of signaling pathways, and proteosome-associated proteins (Robinson et al., 1994; Robinson and Cooley, 1996; Guarnieri et al., 1998; Cooley, 1998; Djagaeva et al., 2005; O’Neill and Clark, 2016; Hime et al., 1996; Greenbaum et al., 2007). Perhaps the best studied of these proteins is the E3 ubiquitin ligase, Kelch, which has been shown to keep the ring canal structure open by targeted substrate degradation to keep the ring canals open (Hudson et al., 2015; Robinson and Cooley, 1997; Kelso et al., 2002; Hudson et al., 2019; Xue and Cooley, 1993).

We first tested whether adipocyte ring canal integrity is regulated by Kelch by performing fat body-specific knockdown of kelch using the Dcg-Gal4 driver (henceforth called Dcg>kelch^RNAi) and comparing these animals to a control RNAi (Dcg>LacZ^RNAi). We validated that knockdown of kelch resulted in reduced kelch mRNA using qPCR analysis (Fig. S2C). Super resolution STED, followed by deconvolution of ring canals from fat bodies from Dcg>kelch^RNAi wandering third instar larva revealed that loss of kelch resulted in decreased inner circumference of ring canals but that the outer circumference of the ring canal was unaltered (Fig. 3A-B, F-F’). We also observed similar reduction in inner circumference of the ring canals in kelch mutant heterozygotes (kelch^DE1/+) (Fig. 3E-F), corroborating our RNAi result (Hudson et al., 2019). These results suggest that Kelch is required in the fat body to maintain ring canals in an ‘open’ state, similar to its role in the germline.

Figure 3: Kelch, pnut and sep1 are required for ring canal structure and function.

A-C. Representative high magnification confocal image acquired as in 1G-H of pav-GFP with STED resolution (A-C) and Hugyens deconvolution (A-C’) from 112h AED adipocytes expressing a control LacZ^RNAi (A), kelch^RNAi (B), pnut^RNAi(C), sep1^RNAi (D), or from kelch heterozygotes (kelch^DE1/+, E). RNAi expression was driven using Dcg-GAL4. A”-E” are images from A’-E’ with inner and outer circumference of the ring canal marked in yellow dotted lines.

F-G. Measurement of average inner (F) and outer circumference (F’) from A-E. n= 7-10 ring canals from at least five different animals per sample. Deviation from circularity (G) is measured as a ratio of the major axis diameter to the perpendicular minor axis diameter.

H-I. Time lapse montage over six minutes of fluorescent images showing GFP expression driven by Dcg-GAL4 in FRAP experiments as in 3A-B from 112h AED adipocytes also expressing LacZ^RNAi (H, Video 5) or kelch^RNAi (I, Video 6). The bleached cell is outlined in blue in H and magenta in I.

J-K Quantification of GFP recovery in bleached cells from H-I. Data represent average of 10 independent photobleaching experiments from ten animals collected across 3 independent crosses, standard deviation is plotted as the shaded error envelope.

Scale bars: A-E = 0.1μm; H-I= 50μm

We next tested whether previously reported structural components of ring canals such as septins are required in fat body ring canals (Hime et al., 1996; Price et al., 2023; Kaufman et al., 2019; Ji et al., 2005; O’Neill and Clark, 2016). Based on publicly available fat body gene expression data in FlyAtlas or Flyatlas2 (Robinson et al., 2013; Leader et al., 2018), we determined that several septin encoding genes are abundantly expressed in the larval fat body including pnut (septin7) and septin1 (sep1). We tested whether depleting pnut or sep1 disrupted fat body ring canal structure by measuring the dimensions of ring canals. Interestingly, the inner circumference of fat body ring canals was unaltered in Dcg>pnut^RNAi and Dcg>septin1^RNAi (Fig. 3C-D, F-F’). However, we found that the ring canals displayed aberrant morphology` and appear more ellipsoid than circular in comparison to control and Dcg>kelch^RNAi (Fig. 3C’-D’, G). These results suggest that pnut and sep1 proteins are important for maintaining the structural integrity of fat body ring canals.

We next tested whether disruption of ring canal integrity using Dcg>kelch^RNAi impacted intercellular cargo transport. We performed FRAP experiments as in Fig. 2A to observe recovery of cytoplasmic GFP after photobleaching in control and Dcg>kelch^RNAi animals. We observed no recovery in GFP fluorescence in six minutes in Dcg>kelch^RNAi adipocytes (Fig. 3I-K, Video 7), whereas control adipocytes showed visible and measurable recovery in fluorescence as before (Fig. 3H, I-K, Video 6). These results establish that disrupting the fat body ring canal impairs the rapid transport of cytoplasmic proteins through ring canals across paired adipocytes.

Ring canal disruption results in increased stress and uncoupling of adipocytes

With a suitable method to disrupt ring canals in the fat body using Dcg>kelch^RNAi, we sought to test our hypothesis that disrupting ring canal structure negatively impacts the stress response in the fat body. To interrogate the effects of kelch knockdown on stress tolerance, we used the ISR reporter, 4EBP^intron-DsRed (Kang et al., 2017). We found that knocking down kelch in 112h AED adipocytes resulted in higher levels of DsRed expression, suggesting that these animals are experiencing higher levels of stress in adipose tissue than control Dcg>LacZ^RNAi animals (Fig. 4A-B, D). Notably, these fat bodies do not show disparities in the expression of a control UAS-GFP also driven by Dcg-GAL4 (Fig. 4A”-C”, E), indicating that the differences in DsRed reporter levels are not due to global protein synthesis defects. We also plotted average DsRed measurements from individual animals to ensure that the difference in stress response reporter activation was not due to a single outlier animal (Fig. S3A).

Figure 4: Disruption of ring canals results in uncoupled stress response between paired cells.

A-C, I-L. Representative confocal images of 112h AED adipocytes expressing 4EBP^intron-DsRed (Magenta) and GFP (green) driven by Dcg-GAL4, which also drives expression of control LacZ^RNAi (A, I), kelch^RNAi (B-C), pnut^RNAi (J), or sep1^RNAi (K). C represents example of very high DsRed expression in kelch^RNAi (see text).

(D-E), (M-N). Quantification of DsRed (D, M) and GFP (E, N) signal from (A-C) and (I-L) respectively. n=50 cells for each genotype, from images taken on fat bodies from at least five different animals across two independent crosses.

(F-H), (O-Q). Measurement from ring-canal paired 112h AED adipocytes in (A-C) and (I-L) respectively, showing differences in absolute values of 4EBP^intron-dsRed intensity (F, N), GFP intensity (G, O) and cell area (H, P). n=25 pairs for each genotype from five animals across two or more independent crosses.

Scale bars: A-C, I-K= 50μM

Interestingly, in some Dcg>kelch^RNAi fat bodies, we also observed cells displaying highly disparate DsRed expression between paired adipocytes (Fig. 4C, C’), which contrasts with our original observation that putatively paired adipocytes show similar levels of DsRed reporter under normal conditions (Fig. S1A, 4A, A’). These results prompted us to ask whether ring canals might mitigate or “buffer” stress across adipocytes. To understand this, we performed pairwise measurements in all fat bodies from Fig. 4A-C to compare differences in control GFP, 4EBP^intron-DsRed, and cell size, between paired adipocytes. We found that while the difference in control GFP intensity did not vary between paired adipocytes in control and kelch^RNAi (Fig. 4F), DsRed fluorescence, as well as the difference in cell area between paired cells showed a significant difference (Fig. 4G-H). We ensured that such disparate ISR reporter expression is not due to differences in expression, stability or maturation of the DsRed protein itself by utilizing another variant of this reporter, 4E-BP^intron-GFP (Grmai et al., 2024), which showed similar results as with 4E-BP^intron-DsRed (Fig. S3B-B’’). We further quantified this by measuring the number of ‘outlier’ singlet cells that display high levels of reporter expression when no other neighboring cells do, and found that Dcg>kelch^RNAi fat bodies had significantly higher number of such outlier singlet cells (Fig. S3B’’). We next tested whether disrupting other fat body ring canal components, pnut and sep1, similarly led to loss of stress buffering between paired adipocytes. As with depletion of kelch, adipocytes from Dcg>pnut^RNAi and Dcg>sep1^RNAi also showed in both an overall increase in 4E-BP^intron-DsRed expression, mismatched DsRed expression and cell size in sister adipocytes with no change in control GFP expression (Fig. 4I-Q). These results showing that disrupting ring canals results in “uncoupling” of stress response levels in paired adipocytes collectively suggest a critical role for ring canals in buffering stress levels across paired cells.

Disrupting ring canal structure results in lowered survival in with exogenous stressors

Since we found that kelch^RNAi expressing fat bodies display higher levels of ISR reporters (Fig. 4A-H, S2B-D), we next investigated a role for ring canals in the response to acute exogenous stress. To assess the stress-responsiveness of fat body to extrinsic stressors in real time, we developed an ex vivo live imaging protocol that allows for long-term monitoring of adipose tissue in the presence of stress-inducing drugs. This strategy allowed us to monitor the effects of tissue-autonomous effects of exogenous ER stress on fat body. We embedded third instar larval fat bodies from Dcg>GFP animals expressing either LacZ^RNAi or kelch^RNAi in low-melting agarose and incubated them in complete Schneider’s insect medium containing either vehicle (Veh = DMSO) or the ER stress-inducing drug, Tunicamycin. This experimental set up allowed for time-lapse imaging over 18 hours to assess adipocyte viability by measuring the irreversible loss of GFP. While Dcg>LacZ^RNAi fat bodies treated with DMSO remained viable for the entire time course (Fig. 5A, E, Video 8), comparably treated Dcg>kelch^RNAi fat bodies rapidly lost GFP expression in cells rapidly, with up to 50% of cells losing GFP expression within 11 hours, and nearly 95% in 18 hours (Fig. 5C, E, Video 10). This loss of viability was even more pronounced in kelch^RNAi fat bodies that were treated with Tunicamycin, where over 50% of cells lost GFP expression within six hours and complete GFP loss at 13 hours (Fig. 5D-E, Video 11). In comparison, the Dcg>LacZ^RNAi fat bodies showed a gradual loss in viability, with only 2% cells losing GFP expression over the time course of 18 hours (Fig. 5B, E, Video 9). We confirmed that the loss of GFP indeed was an indicator of cell death by incubating these tissues with a low concentration of the nuclear marker, DAPI, which selectively stains for dead cells (Wallberg et al., 2016). Indeed, cells that showed loss of GFP expression showed preferential DAPI uptake (Fig. S4), validating that GFP-negative cells were indeed dead. Further, we also observed changes in nuclear morphology in these dying cells including karyorrhexis and pyknosis, which have been reported previously to be a reliable marker of cell death in polyploid cells that do not undergo canonical apoptosis (Mehrotra et al., 2008; Zhang et al., 2014; McCall, 2010; Serizier et al., 2022; Yamada et al., 2023). This underscores the role of ring canals in buffering ER stress across paired adipocytes.

Figure 5: Disruption of ring canals increases susceptibility to exogenous stress.

A-D. Time lapse fluorescence image montages over 18 hours of GFP expression (green) in 112h AED fat bodies treated with either vehicle (A, C; Videos 7, 9) or Tunicamycin (B, D; Videos 8, 10). Dcg-GAL4 drives GFP expression and expression of LacZ^RNAi (A-B) or kelch ^RNAi (C-D). Please note that these images are at lower magnification showing a larger section of the fat body.

E. Quantification of GFP signal from A-D. Error bars represent standard deviation across three independent live imaging experiments from three independent crosses.

F-I. Representative images showing fat bodies treated with either DMSO (F, H) or paraquat (PQ) (G, I) from LacZ^RNAi (F-G) or kelch^RNAi (H-I) for 2h, then fixed and imaged to evaluate change in in gstD-GFP reporter expression.

J. Quantification of mean GFP intensity per cell from each condition in F-I.

K. Number of isolated outlier cells with high GFP expression (presumed uncoupled but paired cells).

Scale bars: A-D = 500μm; F-I = 25μm

We next tested whether other stress modalities are also buffered by ring canals in the fat body. To do so, we utilized a reporter of oxidative stress, gstD-GFP (Sykiotis and Bohmann, 2008). We then treated fat bodies from control and kelch^RNAi animals with the oxidative stress inducing chemical, paraquat. While basal gstD-GFP levels were not altered in kelch^RNAi fat bodies in comparison to control (Fig. 5F, H, J), inducing further oxidative stress by paraquat treatment resulted in higher levels of gstD-GFP induction in kelch^RNAi fat bodies (Fig. 5G, I-J). Further, we observed that fat bodies subjected to such exogenous oxidative stress showed uncoupling of adipocytes as we observed using our ISR reporter (Fig. 5K). Taken together, these findings corroborate our hypothesis that ring canal-mediated transport results is required for buffering several modalities of stress across coupled adipocytes and mediates cell survival under conditions of very high extrinsic stress. Indeed, consistent with our finding that disrupted ring canals result in lowered stress tolerance, we also find that larvae carrying heterozygous mutations in kelch and PERK (encoded by PEK in Drosophila) are inviable (Fig. S5C), suggesting further that ring canal integrity is crucial for animals to cope with intrinsic ER stress occurring during development.

Disrupting fat body ring canals impacts secretory capacity of adipocytes and larval developmental timing

Given that the fat body is a critical tissue for proper development, and that it relies on stress response pathways for its homeostatic function (Sone et al., 2013; Kang et al., 2017; Huang et al., 2017; Saeedi et al., 2022; Yamada et al., 2023), we sought to determine the phenotypic consequences of disrupting ring canals in adipocytes. The primary function of the fat body particularly during the third instar stage is to assimilate dietary nutrients from the hemolymph, which is then used to 1) to synthesize, store, and secrete larval serum proteins (Lsps) back into the hemolymph (Powell et al., 1984; Roberts et al., 1977), and 2) synthesize and store lipid reserves in the form of triacylglycerides in adipocytes (Grönke et al., 2005). These proteins and lipids are sequestered in the fat body and hemolymph until later non-feeding stages of development, when they are broken down to support the energetic needs of metamorphosis (Jowett et al., 1986; Heier and Kühnlein, 2018). We asked whether the high levels of stress, as well as disparate stress reporter between sister cells seen with ring canal disruptions (Fig. 4) impacted protein secretory capacity or lipid storage in Dcg>kelch^RNAi animals.

We first examined protein secretory capacity by western blotting analysis of two larval serum proteins, Lsp1α and Lsp2, in whole larvae extracts from either Dcg>LacZ^RNAi or Dcg>kelch^RNAi animals. We found that the levels of both Lsp1α and Lsp2 were significantly lower in Dcg>kelch^RNAi (Fig. S5A-A’) animals with no change in total protein (see Fig. 6C) suggesting that disrupting ring canals likely impairs the secretory function of adipocytes, which could impact development. Since Lsps are only a small portion of the fat body’s secretory output, we sought to examine total secretory output from the fat body into the hemolymph. To do so, we used an ER-linked biotin ligase (BirA) that biotinylates all proteins that pass through the secretory pathway (Droujinine et al., 2021). We specifically expressed this in fat body using r4-GAL4 since Dcg-GAL4 is also expressed in hemocytes (Asha et al., 2003; Armstrong et al., 2014; Weaver et al., 2020; Winant et al., 2024). Western blotting analysis of hemolymph extracts from control and r4>BirA,kelch^RNAi animals showed that disrupting of ring canals led to reduced secretion from the fat body into the hemolymph when normalized to total hemolymph protein (Fig. 6A-B). Consistent with our analysis using whole larval extracts in Dcg>kelch^RNAi (Fig. S5), we observed a decrease in Lsp1α in hemolymph from r4>BirA,kelch^RNAi animals (Fig. 6B, B’). However, we did not see a significant change in Lsp2 (Fig. 6B, B’’) suggesting that Lsp2 secreted from the fat body is not primarily stored in the hemolymph. Nonetheless, these results together show that disrupting ring canals in the fat body reduces its secretory capacity.

Figure 6: Disruption of ring canals reduces protein secretion from the fat body and results in delayed development.

A-B. Western blot analysis of hemolymph from animals of indicated genotypes also expressing an ER-associated biotin ligase (UAS-BirA-ER-HA) driven by r4-GAL4 which also drives expression of either LacZ^RNAi or kelch^RNAi. Representative blot in (A) shows biotinylated secreted proteins from indicated genotypes, open arrowheads indicate proteins secreted from the fat body that are likely not sensitive to kelch depletion (no change in intensity in comparison to control) and black arrowheads mark proteins which show a notable decrease in secretion with kelch depletion.

Representative blot in B shows expression of Lsp1α and Lsp2. Graphs in A’, B’, B’’ show mean of data across three independent experiments from two independent crosses, error bars represent standard error, *=p<0.05 as calculated by a two-tailed student t-test.

C-D. Relative total protein levels measured by BCA assay levels (C) or triacylglyceride (TAG) levels (D) in whole 112h AED larvae where Dcg-GAL4 drives expression of either LacZ^RNAi or kelch^RNAi. Graph represents the mean of six biological replicates from two independent crosses and error bars indicate standard error.

E-F. Developmental timing of animals from C-D as measured by the percentage of larva either foraging or wandering at 112h AED when raised on protein rich food (‘R’) or basic food (‘B’). n=5 vials per genotype with 25 larvae per vial (see methods).

G. Percentage of animals that survive to adulthood from D-E.

We next tested whether disrupting ring canals in the fat body resulted in lipid storage defects. The majority of lipids are stored in the form of triglycerides in Drosophila. To measure potential defects in lipid storage, we performed triglyceride assays on whole larval extracts from Dcg>LacZ^RNAi or Dcg>kelch^RNAi animals. Interestingly, we observed no difference in the amount of total triglyceride content between the two genotypes (Fig. 6D), suggesting that the ER stress defects observed in Dcg>kelch^RNAi (Fig. 4A-H) impacts fat body protein secretory capacity but not lipid metabolism.

Protein secretion from the fat body is critical for timely metamorphosis as described above and since kelch^RNAi animals showed reduced fat body secretory capacity (Fig. 6A-B, S5A), we examined the developmental timing of these animals. While control larvae reared on a protein-rich diet (‘R’ food) begin wandering around ~110h AED as expected, Dcg>kelch^RNAi larvae displayed a modest but significant delay in developmental timing when reared on a regular diet (‘R’ food) as evidenced by a much lower population of animals wandering at the same time point (Fig. 6E). We observed a similar delay in r4>kelch^RNAi animals as well (Fig. S5B), suggesting that this defect is due to disruption of ring canals in the fat body. Next, since developmental timing is sensitive to protein availability and we found that disrupting fat body ring canals resulted in reduced protein secretion (Fig. 6A-B, S5A), we performed these experiments in animals reared on a reduced protein food (‘B’ food, see Methods). Interestingly, the developmental delay in Dcg>kelch^RNAi animals was even more pronounced on ‘B’ food (Fig. 6F), suggesting an additional role for ring canals in buffering nutrient deprivation stress. Notably, the disruption of ring canal structure did not result in any lethality with nearly all animals surviving to adulthood (Fig. 6G), indicating that ring canals might contribute to the general robustness of larvae but are not necessary for their survival. Together, these phenotypic analyses underscore the physiological importance of stress buffering across neighboring cells by ring canals in fat body function.

Discussion

In this study we describe the first instance of intercellular bridges in adipose tissue in any organism and establish a function for them. We demonstrate a role for specialized sharing of cytoplasmic and secretory pathway contents across adipocytes paired by ring canals in the metabolic and secretory larval fat body, a tissue with several critical functions. Our findings point to a role for rapid intercellular transport of specific organellar cargo in buffering the high levels of intrinsic stress adipocytes endure due to their function and also mediating adipocyte survival under conditions of extrinsic stress. Further, we find that the proper structure and function of ring canals are required for the normal secretory function of the fat body, and that their disruption results in nutrient-dependent developmental defects. This underscores the importance of intercellular communication in normal adipocyte function. Integrating these observations, we postulate that buffering several types of stress across cells is necessary in secretory and metabolic cell types for both normal tissue function and survival under acute stress The broader implications of our findings are discussed below.

The nanoscale dimensions of ring canals confer functional and cargo specificity

Our super-resolution imaging shows that ring canals in the Drosophila larval adipocytes are extremely small but open structures, having an inner diameter of less than 500 nm even at developmental stages when adipocytes span several tens of microns (Fig. 1). This implies that adipose ring canals are up to two orders of magnitude smaller than the cell. Our work also shows that the ring canals adipose do not grow even as adipocyte nuclei undergo rounds of endoreplication to reach a final average ploidy of ~255(Nordman et al., 2011). This is in contrast to other ring canal structures; for example ring canals of nurse cells in the female germline grow in scale with the nuclei reaching up to 10μm in diameter, while the cells themselves have 512-1024 copies of the genome (Cooley, 1998; Dej and Spradling, 1999). Although both adipocytes and nurse cells are highly secretory, this contrast in their respective ring canal sizes can be partly explained by the differences in tissue architecture and the role ring canals play in these respective cell types. In the germline, ring canals permit the transport of mRNAs and proteins synthesized in the nurse cells to be deposited into the developing oocyte in an efficient manner (Shaikh et al., 2024; Bernard et al., 2018; Hsu et al., 2015; McLean and Cooley, 2014; Roth and Lynch, 2009; Petrella et al., 2007). The structure of the ovarian follicle is such that not every nurse cell makes contact with the oocyte, and thus ring canals are the primary mode of intercellular transport from distal nurse cells to the oocyte. However, the larval fat body is a relatively flat monolayer of adipocytes that each make contact with the circulating hemolymph, which is the primary destination for secretory products from this tissue. Adipocytes hence rely on canonical vesicular trafficking via the secretory pathway to deposit proteins and lipids into the hemolymph (Ugrankar-Banerjee et al., 2023; Yang et al., 2021). Thus, unlike in the nurse cells, fat body ring canals appear to be cargo selective, likely restrained by their size-this is supported by our finding that they only permit the transport of some cytoplasmic, ER and Golgi proteins, but not mitochondria (Fig. 2). However, despite differences in their functions, ring canals in both nurse cells and adipocytes appear to share core components for maintaining their structural integrity such as Kelch, pnut ,and sep1 (Fig. 3). The structural components of ring canals in adipocytes likely will inform their cargo selectivity, in addition to specifying their size. Future studies on larval adipocyte ring canal composition and comparative analyses with other ring canal structures may reveal the mechanistic basis for tissue-specific cargo selectivity.

It is also interesting to note that our various genetic manipulations resulted in different effects on the structural integrity of fat body ring canals – kelch^RNAi reduced the opening size of ring canals (Fig. 3F), while pnut^RNAi and sep1^RNAi resulted in altered shape (Fig. 3G). However, all manipulations resulted in a compromised stress buffering capacity (Fig. 4), pointing to the crucial role of structural integrity and function of ring canals in the resilience of the tissue. Future work will investigate both how these manipulations alter the ultrastructure and how dynamically ring canal shape changes in live tissues during active periods of stress.

Tissue architecture and implications for mosaic analysis

All larval adipocytes undergo specification and terminal differentiation during embryogenesis by stage 15, which corresponds to a time point earlier than 13.5 hours AED (Hoshizaki et al., 1994; Klapper et al., 1998). We observe ring canals in fat bodies as early as 48h AED (Fig. 1I), suggesting that these structures are very likely formed at the end of the last mitotic cell cycle during embryogenesis and prior to the onset of adipocyte endoreplication at ~15h AED (Smith and Orr-Weaver, 1991). One practical implication of such early formation of ring canals in development is the timing of clone induction for mosaic analyses such as MARCM or FLP/FRT (Lee and Luo, 1999; del Valle Rodríguez et al., 2011). Mitotic recombination to generate mutant clones in the fat body necessitates inducing clone formation prior to 12h AED, during or before the penultimate mitosis. The early formation of ring canals also has implications for the use of FLP/FRT recombination which generates ‘flip-out’ clones by excising a stop cassette or other transgenic sequences flanked by FRT sites. Any FLP/FRT-mediated recombination ‘flipping out’ event after ring canals have formed might render individual nuclei with different genotypes, but the phenotype of the resultant adipocyte pair may represent the average of the individual nuclei or the more dominant phenotype of the pair. Given that many cargo types can be transported across ring canals, individual FLP/FRT clone cells expressing UAS-driven transgenes will be coupled to a sister cell that may represent a ‘shadow’ clone (Bosch et al., 2016). Conversely, since we find that the fat body ring canal is size-selective for cargo, some protein complexes may be excluded from ring canals, representing two cells within a syncytium having uneven expression profiles, similar to the accessory gland (Box et al., 2024; Church et al., 2025). Thus, future work in the larval fat body should consider the above caveats and carefully examine phenotypes to distinguish cell non-autonomous potential phenotypes from the consequences of cytoplasm sharing between paired cells.

Adipocytes buffer stress through ring canals

The fat body functions to secrete several important proteins into the hemolymph that are required for supplying nutrients to all the developing tissues of Drosophila in larvae (Musselman and Kühnlein, 2018). The fat body also simultaneously functions in the absorption and storage of nutrients as lipids which are required for the animal to successfully undergo metamorphosis. We employed two reporters in this study, 1) 4E-BP^intron-DsRed which measures activation of the nutrient-sensitive ISR kinase GCN2 in the developing fat body, and ER stress response mediated by the ISR kinase, PERK, (Kang et al., 2017) and, 2) gstD-GFP, which responds to oxidative stress downstream of Keap1/Nrf2 signaling ((Sykiotis and Bohmann, 2008)). Consistently, we find that loss of ring canal integrity results in increased susceptibility to nutrient availability (Fig. 6E-F), exogenous ER stress (Fig. 5A-E), and oxidative stress (Fig. 5F-K), suggesting that ring canals buffer multiple flavors of cellular stressors.

We find that compromising ring canals by knocking down kelch, pnut, or sep1 in the fat body results in higher levels of average ISR stress reporter expression, indicative of general cellular malfunction (Fig. 4). Consistent with such a malfunction in a secretory and metabolic tissue, we observe reduced secretory function in the fat body of these animals (Fig. 6A-B, S5). More intriguingly, we also find the disrupting ring canals led to the “uncoupling” of adipocytes such that sister cells now showed significant differences in cell size and stress reporter levels (Fig. 4F-H, N-P, S3B). These results lead us to consider that one of the critical functions of ring canals is to mitigate high levels of stress in individual cells during development by facilitating rapid sharing of cytoplasmic/ER/Golgi contents. This is supported by other work demonstrating that ring canals in somatic ovarian follicle cells serve to equilibrate protein levels across cells connected through ring canals (McLean and Cooley, 2013; Airoldi et al., 2011), though the need for such equilibration was not fully understood.

We propose that ring canal-mediated equilibration of cellular contents in secretory tissues such as the fat body and follicular epithelium serves to coordinate stress levels between connected cells such that no single cell experiences the detrimental effects of chronic ER stress response activation. From a mechanistic standpoint, we posit that hyperactivation of stress response signaling in an individual cell because of inherent or exogenous stress results in increased sharing of specific cargoes with the sister cell that serves to immediately reduce the stress burden on the individual cell, thus promoting its survival. Our data showing that disrupting ring canals in fat bodies subjected to exogenous ER or oxidative stress results in rapid adipocyte death (Fig. 5A-E, S4) strongly supports this hypothesis that ring canals serve to “buffer” stress in sister cells. An open question for future studies is whether such stress buffering is mediated by transfer of specific proteins or molecules. Such potential cargo candidates include signaling entities such as stress response transcription factors themselves or smaller signaling moieties such as calcium; alternatively stress buffering may occur by ‘dilution’ of potentially toxic entities such as misfolded proteins or reactive oxygen species or by actively supplying stress mitigating entities such as proteosome components or chaperones.

Finally, recent work from our lab shows a role for stress response in sexual dimorphic stress tolerance in larvae (Grmai et al., 2024). While all the experiments performed in this study are in female larvae, it would be interesting to determine whether the function of ring canals may contribute differentially to larval adipose sexual dimorphism in future studies(Grmai et al., 2024).

Does multinucleation confer advantages to metabolic tissues?

Ring canal-connected sister cells in the fat body do not undergo complete cytokinesis, and remain connected to one another partially sharing cytoplasm, making each pair a bi-cellular syncytium. Syncytia or syncytial structures form either through cell fusion or through incomplete cell cycles resulting in multiple nuclei sharing cytoplasm (Peterson and Fox, 2021). It is tempting to propose that the extent of intercellular transport and cargo sharing render paired larval adipocytes binucleate. Interestingly, the adult fat body of Drosophila comprises binucleate and tetranucleate cells which form as a result of cell fusion of a different set of progenitor cells during late metamorphosis (Lei et al., 2022). However, we did not find evidence of ring canals in adult adipocytes, at least as marked by Pav-GFP (data not shown). Nonetheless, both the larval and the adult fat body contain syncytial cells - the larval fat body comprising several bi-cellular (or binucleate) paired adipocytes, and the adult fat body containing binucleate and tetranucleate syncytial adipocytes (Lei et al., 2022). Thus, context dependent variation in DNA content and nuclear number appears to be an important facet of Drosophila adipocyte physiology. Multinucleation has also been reported in cultured vertebrate adipocytes (Xu et al., 2017) and in several other metabolically active tissues such as cardiomyocytes (Gan et al., 2020), liver hepatocytes (Fortier et al., 2017), lactating mammary epithelial cells (Rios et al., 2016), and the exocrine pancreas (Oates and Morgan, 1989). Our work adds to the substantial and growing body of literature demonstrating the diversity of variant cell cycle processes resulting in tissues containing cells with polyploid DNA content and binucleate cells (Nandakumar et al., 2021). However, despite the descriptions of multinucleation in various tissues across phyla, our understanding of the functions and consequences of having multiple nuclei in these vastly different tissues remains largely speculative. Emerging work presents tantalizing hints that multinucleate cells, specifically cardiomyocytes and enterocytes, show improved cellular function in comparison to their mononucleate counterparts (Lam et al., 2025; Lessenger et al., 2025). Our findings here put forth Drosophila adipose tissue as an excellent genetically pliable model to investigate the role of both the extent and mode of multinucleation in the functioning of metabolic tissues.

Materials and Methods

Fly Rearing

All flies were raised at 25°C. Crosses were set up with ten female virgins and five males no more than three days post-eclosion at the time of setting up the cross. Unless otherwise mentioned, all crosses and rearing were performed on ‘R’ food formulation from Lab Express (https://www.lab-express.com/DIS58.pdf). Crosses were flipped every day to ensure no overcrowding of vials occurred. All fly stocks used in the study are listed in Table 1. Specific genotypes for each experiment are listed in Table 2. For assessment of genotype viability in Fig. S5C, crosses were set up as described above, and the number of larval progeny wandering were scored by phenotype (i.e presence or absence of balancer marker) and graphed as a percentage of total F1 animals in each cross.

Table 1.

List of fly stocks and corresponding stock numbers used in this study.

Transgene	Source (RRIDs from BDSC)
pav-GFP	BDSC_81650
UAS-GFP	BDSC_1522
UAS-KDEL-GFP	BDSC_9898
UAS-Golgi-GFP	BDSC_30902
UAS-mito-GFP	BDSC_8442
Dcg-GAL4	BDSC_7011
UAS-LacZRNAi	(Kang et al., 2017)
UAS-kelch RNAi	BDSC_31251
kelchDE1	BDSC_4893
UAS-pnutRNAi	BDSC 30485
UAS-septin1RNAi	BDSC 27709
UAS-GFPNLS	BDSC_4776
4EBPintron-DsRed	(Kang et al., 2017)
4EBPintron-GFP	(Grmai et al., 2024)
Pav::mCherry	Cooley Lab (Yale University)
r4-GAL4	BDSC 33832
UAS-BirA-ER-HA	BDSC_93427
gstD-GFP	BDSC_605991
PEK1	(Monteiro et al. 2023)
Canton-S	BDSC_64349 (Originally BDSC 1)

Table 2.

List of genotypes by figure.

Figure	Genotype
S1A	w-;Dcg-GAL4,UAS-GFPNLS/UAS-LacZRNAi;4EBPintron-DsRed/+
S1B	w-;Dcg-GAL4,UAS-GFPNLS/+;pav::mCherry/+
1	pav-GFP
2A-B, Video 1-2	w-;Dcg-GAL4/UAS-GFP;
2C, S2A Video 3	w-;Dcg-GAL4/UAS-KDEL::GFP;
2D, S2B Video 4	w-;Dcg-GAL4/+;UAS-Golgi-GFP/+
2E	w-;Dcg-GAL4;UAS-mito::GFP;
S2C	w-;r4-GAL4/+;UAS-LacZRNAi/+ and w-;r4-GAL4/+;UAS-kelchRNAi/+
3A	w-;Dcg-GAL4,pav-GFP/UAS-LacZRNAi;
3B	w-;Dcg-GAL4,pav-GFP/+;UAS-kelch RNAi/+
3C	w-;Dcg-GAL4,pav-GFP/UAS-pnutRNAi
3D	w-;Dcg-GAL4,pav-GFP/UAS-sep1RNAi
3E	w-;kelchDE1/pav-GFP;
3H, Video 5	w-;Dcg-GAL4/UAS-LacZRNAi;
3I, Video 6	w-;Dcg-GAL4/+;UAS-kelchRNAi/+
4A, I	w-;Dcg-GAL4,UAS-GFPNLS/UAS-LacZRNAi;4EBPintron-DsRed/+
4B-C	w-;Dcg-GAL4,UAS-GFPNLS/+;4EBPintron-DsRed/UAS-kelchRNAi
4J	w-;Dcg-GAL4, UAS-GFPNLS/UAS-pnutRNAi; 4EBPintron-DsRed/+
4K	w-;Dcg-GAL4, UAS-GFPNLS/UAS-sep1RNAi; 4EBPintron-DsRed/+
S3B	w-;Dcg-GAL4,4E-BPintron-GFP/UAS-LacZRNAi; and w-;Dcg-GAL4,4E-BPintron-GFP/+; UAS-kelchRNAi/+
5A-B, S4A-B, Video 7, 8	w-;Dcg-GAL4,UAS-GFPNLS/UAS-LacZRNAi;4EBPintron-DsRed/+
5C-D, S4C-D, Video 9,10	w-;Dcg-GAL4,UAS-GFPNLS/+;4EBPintron-DsRed/UAS-kelchRNAi
5F, G	w-;gstD-GFP/UAS-LacZRNAi;r4-GAL4/+
5H-I	w-;gstD-GFP/+;r4-GAL4/UAS-kelchRNAi
6A	w-; UAS-BirA-ER-HA/UAS-LacZRNAi;r4-GAL4 /+ and w-; UAS-BirA-ER-HA/+;r4-GAL4 /UAS-kelchRNAi
6C-G, S5A	w-;Dcg-GAL4,UAS-GFPNLS/UAS-LacZRNAi;4EBPintron-DsRed/+ and w-;Dcg-GAL4,UAS-GFPNLS/+;4EBPintron-DsRed/ UAS-kelchRNAi
S5B	w-; UAS-LacZRNAi/+;r4-GAL4 / + and w-;+;r4-GAL4 /UAS-kelchRNAi/+
S5C	w−/+;; PEK1/+ w−/+;+/Cyo,Tb w−/+; kelDE1/+ w−; kelDE1/+; PEK1/+ w−;+/Cyo,Tb; PEK1/+

Staining for confocal microscopy

Larval fat bodies were micro dissected in 1X phosphate buffer saline (PBS) and fixed in 4% paraformaldehyde (PFA) made fresh to a final concentration of 1X PBS for 25 minutes on a nutator. Tissues were washed three times each in 0.1% PBS-Tween between different stains. 4′,6-diamidino-2-phenylindole (DAPI, 200 μM final) and optional phalloidin Alexafluor 647 (Thermo Fisher Scientific, 1:1000) staining was performed to counterstain for nuclei and actin/cell membrane respectively for 1h at room temperature. Tissues were mounted on glass slides in a solution of 70% glycerol prior to placing coverslips. Slides were sealed with a thin coating of nail polish to prevent evaporation and stored protected from light until imaging using a Nikon A1 confocal microscope. Slides are stable at room temperature for a month and at 4°C for several weeks with limited reduction in fluorescence of GFP, RFP and common fluorescent dyes.

Super resolution microscopy

Fat bodies mounted in 70% glycerol without DAPI were imaged using a 93x glycerin immersion objective (refractive index-matched) on a Leica-SP8 imaging system. Each ring canal image was captured as a z-stack with 10 steps (0.1-0.2μm step size, system optimized conditions used) 3D STED imaging of pav-GFP marked ring canals was performed using a 592mm wavelength depletion laser. Further deconvolution was performed using the Huygens suite with the same iteration settings across all experimental conditions. Super-resolved images were then used to measure ring canal dimensions by manually tracing inner and outer circumferences, and long and short radii of inner circumference using FIJI/ImageJ.

Fat body mounting for live imaging

Fat bodies dissected in 1X PBS were carefully mounted on MatTek glass bottom culture dishes in a solution of 1% low melting agar in 1X PBS as follows. A 3X agar stock solution was melted at 90°C and diluted with chilled PBS to bring the working agar solution to 25°C. 250μl of agar was carefully pipetted on the glass bottom dish divot in which fat bodies are placed within two minutes. After agar gelling occurs (5 minutes), specimens were submerged in Schneider’s cell culture medium with desired concentration of drug or vehicle and subsequently imaged on a Nikon A1/Ti2 inverted line scanning confocal or epifluorescence microscope, either for a few minutes for FRAP, or over longer periods (up to 18 hours) cell viability time course experiments. For detection of dead cells, we incubated samples prepared as above with 2 μM of DAPI.

Fluorescence recovery after photobleaching (FRAP) experiments

FRAP imaging was performed using a 20X dry objective on a Standard Nikon A1 plus Confocal with the pinhole opened to allow for increased fluorescence detection of 3D mounted tissues at a depth of up to 25μm. Individual cells were traced using the polygon ROI function. Standard Nikon A1 plus stimulation settings were used to deplete fluorescence reporters (100% laser power for depletion, 1/12 scanning speed, four rounds of stimulation to achieve 95% or higher depletion in total fluorescence of ROI of interest). Processing and quantifications of all FRAP experiment images were performed using the NIS-elements advanced microscopy suite. A custom GA3 workflow pipeline was developed to perform timed image quantifications, which were exported to excel spreadsheets and plotted using Graphpad prism.

SDS-PAGE and Western Blotting

Whole larvae (3 per sample) were flash frozen before being crushed using a pestle in 125μL of Radio-immunoprecipitation assay lysis buffer (RIPA, 50 mM Tris HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.5% Sodium deoxylcholate, 0.1% SDS, 1 mM EDTA) supplemented with protease inhibitor (Roche). The lysate was centrifuged at 16000 RCF for 10 minutes, and clear supernatant was transferred to a new tube while avoiding the floating fat layer. For hemolymph western blots, 5 larvae were ‘bled’ by nicking their cuticle in a droplet of 50 μl PBS. Hemolymph was then centrifuged at 16000 RCF for 10 minutes to remove hemocytes. 25μl of lysate prepared as above mixed with 4X Laemmli buffer containing 4% β-mercaptoethanol was analyzed by SDS-page followed by western blotting onto nitrocellulose membranes. Membranes were blocked with 5% milk in 1xPBS containing 0.1% Tween-20 and then probed overnight at 4°C with FITC-conjugated anti-biotin (1:1000, Jackson Immunoresearch labs 50-194-2530), Rat anti-Lsp1α (1:1000) and Lsp2 (1:2500) (Valzania et al., 2024). Proteins were detected using Rat anti-HRP secondary antibody. Total protein was either measured using 5μl lysate in a standard BCA assay in a plate reader or by staining membranes using Ponceau-S and measuring intensity. Blots were quantified in ImageJ by measuring intensity of specific bands and normalizing it to total protein.

Measurement of developmental delay

Crosses with 20 virgins and 10 males were set up in population cages with a plate of grape agar supplemented with fresh yeast paste. Crosses were flipped into new vials after 1h egg lays. 24 hours after egg lays were complete, 25 1st instar larvae of similar size were picked from grape plates and placed in vials containing either ‘R’ (regular/protein rich) or ‘B’ (Bloomington formulation, https://bdsc.indiana.edu/information/recipes/bloomfood.html) food, which contain 4% yeast w/v and 1.5% yeast w/v respectively. To determine percentage of wandering larvae, measurements were made at 112h AED (after egg deposition). Total survival to adulthood was measured by counting animals post eclosion from individual vials.

Triacylglycerides (TAG) assay

Three whole larvae per sample were crushed in a solution of 0.5%SDS-1X PBS, incubated on ice for 15 minutes, and then centrifuged at 12000 RCF at 4°C for 5 minutes. 3μL of the lysate was added to 100μL of Triglyceride reagent solution (Pointe Scientific) in 96 well dish wells, in technical duplicates. After 15 minutes of incubation, the plate was analyzed on an absorbance plate reader at 540nm. Total protein was simultaneously measured using 5μl lysate in a standard BCA assay in a plate reader.

Statistical analyses

The figure legends describe the sample sizes for the corresponding data. Significance in graphs in Figures 1-5 were calculated using unpaired T-test with Welch’s correction for unequal standard deviations. Comparison of multiple groups as in Figure S1C-D was done using two-way ANOVA followed by a Tukey test.

Supplementary Material

11

Figure S1: Paired adipocytes show similar levels of fluorescent protein expression

A. Representative image showing general heterogeneity in 4EBP^intron-DsRed expression across the fat body, with pairs of neighboring cells (yellow arrows) showing similar DsRed levels.

B. Representative image from Dcg>GFP;pav::mCherry fat bodies counterstained for Phalloidin (magenta) displaying sister cells connected to each other through ring canals (marked by pav-mCherry, red) with similar levels of GFP (green). B’ shows magnified image from B with example of measurement performed, and formula used for quantifying differences is shown at the bottom. Subject cell is marked in dotted black outline, the sister cell connected to the subject cell through ring canal (indicated by red arrow) is marked in solid black outline, and the non-sister neighboring cells are marked in solid gray outline. B’’ plots raw integrated GFP density values from B’, B’’’ plots the mathematical difference in GFP intensity between subject cell and its neighbors from B’.

C. Difference in GFP intensity between subject cell and neighbors from images of fat bodies from at least 30 cells collected from five animals.

D. Difference in GFP intensity in C plotted as % of subject cell intensity.

Scale bars: A= 100μm; B-B’ = 50μm

Figure S2. Validation of organellar markers and knockdown efficiency

A-B. Representative images showing fat bodies from 2C (A) and 2D (B) where Dcg-GAL4 drives expression of UAS-KDEL-GFP (A) and UAS-Golgi-GFP (B). Tissues were fixed and immunostained using anti-KDEL to mark the ER (A) and anti-GM130 to mark the Golgi (B). Organelle markers are in magenta, and GFP transgenes are in green.

C. qPCR analysis of isolated fat bodies from indicated genotypes. X-axis indicates the primers utilized. **=p<0.01 as measured by a one sample Wilcoxon test for normalized samples.

Scale bars: A, B= 25μm

Figure S3: Increased ISR reporter activation in kelch^RNAi is not due to outlier animals or fluorescent protein type

A. Reanalysis of data from Fig. 4D showing data points for individual animals. Each symbol corresponds to data collected from the same animal.

B. Representative confocal images of 112h AED adipocytes expressing 4EBP^intron-GFP (Green) and DAPI (magenta) in animals where Dcg-GAL4 drives expression of control LacZ^RNAi or kelch^RNAi. B’ shows quantification of mean GFP intensity per cell in n=75 cells from over 10 animals. B’’ quantifies number of outlier singlet cells observed per frame as a measure of uncoupled sister cells.

Scale bars: B= 25μm

Figure S4: Validating loss of GFP as a marker of cell death in long-term live imaging experiments

Representative time lapse fluorescence images from 6 hours (A, C) and 12 hours (B, D) of GFP expression (green) in 112h AED fat bodies treated with Tunicamycin as in Fig. 5A-D. Dcg-GAL4 drives GFP expression and expression of LacZ^RNAi(A-B) or kelch ^RNAi (C-D). Cells that lost expression of GFP (arrows) also showed DAPI incorporation, which is indicative of dead or dying nuclei. Yellow arrows point to karyorrhectic nuclei and white arrowheads point to pyknotic nuclei, often appearing in pairs (sister cells).

Scale bars: A-D= 75μm

Figure S5: Loss of kelch in the fat body results in reduced secretion and developmental delay

A. Western blot analysis of Lsp1α and Lsp2 protein levels in hemolymph extracts from whole 112h AED larva where Dcg-GAL4 drives either control LacZ^RNAi or kelch ^RNAi. Data quantified in A’ bars represent mean of three independent experiments from two independent crosses, and error bars represent standard error.

B. Developmental timing as measured by the percentage of larva either foraging or wandering at 112h AED when raised on protein rich food (‘R’), where r4-GAL4 drives either control LacZ^RNAi or kelch ^RNAi.

C. Percentage of PEK¹, kel^DE1 heterozygotes, or kel^DE1; PEK¹ double heterozygotes or balancer containing larvae recovered from the crosses of the indicated genotypes (X axis). Progeny genotype shown in legend