Recruitment of Peroxin 14 to lipid droplets affects lipid storage in Drosophila

Peroxisomes are responsible for catabolism of very-long-chain fatty acids (VLCFAs) and branched chain fatty acids, biosynthesis of ether lipids and regulation of reactive oxygen (He et al., 2021). They are composed of a membrane bilayer with peroxisome membrane proteins (PMPs), surrounding a dense core of enzymes (Fujiki et al., 2020). Peroxisome number, size and enzyme composition change dynamically in response to demand (Honsho et al., 2016). New peroxisomes arise via fission of existing organelles but can be assembled de novo. Either process requires new membrane supplied as pre-peroxisomal vesicles (PPVs) into which PMPs are inserted (Fujiki et al., 2020). The peroxisome assembly process requires the coordinated activity of ten uniquely numbered Peroxin (Pex) proteins that are highly conserved in eukaryotes, as well as Pex11, which directs fission of existing peroxisomes (Fujiki et al., 2020). In animal cells, PPV budding from the endoplasmic reticulum (ER) requires Pex3 and Pex16 (Fakieh et al., 2013; Geuze et al., 2003; van der Zand et al., 2010; van der Zand and Tabak, 2013). Mitochondrial-derived PPVs are also thought to form in a Pex3-dependent manner (Kim, 2017; Rucktaschel et al., 2010; Sugiura et al., 2017). Pex19 recruits PMPs from the cytosol, directing them to membranes in a Pex3 (Type I)- or Pex3/Pex16 (Type II)-dependent process (Fujiki et al., 2020). Together, PMPs Pex13 and Pex14 form a transmembrane (TM) pore (docking complex) through which enzymes with a peroxisome-targeting sequence 1 (PTS1, SKL) motif are recruited from the cytoplasm Pex5 binding Pex14 in the docking complex. Pex5 is subsequently cycled out of the peroxisome by the Pex2, Pex10 and Pex12 E3 ligase complex and AAA-ATPase proteins Pex1 and Pex6 (Fujiki et al., 2020). Defects in any step of this process due to Pex gene mutations cause peroxisome biogenesis disorders like the Zellweger spectrum disorders (Fujiki et al., 2020).

Lipid droplets (LDs) are the primary cellular storage reservoir for fatty acids and cholesterol. LDs are large and stable in adipocytes, but smaller and transient LDs are present in most cell types (Fujimoto and Parton, 2011). LDs have a single phospholipid layer surrounding a neutral lipid core, primarily composed of triglycerides (TGs) and cholesterol esters (Olzmann and Carvalho, 2019). LDs form when individual fatty acids are enzymatically fused to a glycerol backbone to form neutral lipids (e.g. TG), which are inserted between membrane leaflets at specialized regions of the ER. In animal cells, large LDs bud from the ER and can remain for extended periods of time (Walther et al., 2017). Recruitment of specific lipases to the LD surface reverses this process, freeing lipids and individual fatty acids for use by the cell. LD lipases include adipose triglyceride lipase (ATGL), which generally cleaves the initial fatty acyl chain from TG leaving diacylglycerol (DG) (Zimmermann et al., 2004), and Hormone-sensitive lipase (Hsl), which cleaves a fatty acid from DG leaving monoacylglycerol (MG), and MG lipase, which cleaves the remaining fatty acid, freeing the glycerol backbone (Lass et al., 2011). Thus, LD-stored TG can supply individual fatty acids, but LDs are also a source for TG, DG and MG needed for cellular functions. The primary regulators of lipases at the LD surface are the Perilipin (PLIN) proteins (Jackson, 2019). PLINs can suppress or enhance recruitment of lipases (Ducharme and Bickel, 2008; Itabe et al., 2017).

Organelles can coordinate their activities to facilitate overall cellular homeostasis. This includes peroxisomes and LDs, which both play crucial roles in regulating cellular lipids (Lodhi and Semenkovich, 2014; Thiam and Dugail, 2019). The metabolic functions of peroxisomes and LDs overlap via common metabolites or their intermediates (Choudhary and Schneiter, 2021; Lodhi and Semenkovich, 2014; Thiam and Dugail, 2019). Both organelles also require membrane contribution from the ER via specialized domains (Joshi and Cohen, 2019). In yeast cells grown in excess lipid, peroxisomes associate with newly forming LDs that remain at the ER (Binns et al., 2006). In COS7 cells, clusters of mature peroxisomes were observed near the LD surface (Schrader, 2001). LD-resident spastin M1 protein interacts with peroxisome ATP-binding cassette subfamily D member 1 (ABCD1) to promote interaction of peroxisomes with LDs (Chang et al., 2019). In Caenorhabditis elegans, PEX-5 mediates ATGL translocation to LDs, facilitating fasting-induced TG lipolysis (Kong et al., 2020). Finally, PEX2 was identified as part of a regulatory loop that couples ATGL activity to peroxisomal β-oxidation and reactive oxygen production (Ding et al., 2021).

The Drosophila homologs of the human PEX and yeast Pexp proteins are highly conserved (Anderson-Baron and Simmonds, 2019; Pridie et al., 2020; Mast et al., 2011), except that Drosophila peroxisome import does not employ an alternative pathway mediated by the PTS2 motif, recognized by Pex7. Notably, Pex7 requires docking with Pex5 to complete the process of protein import into the peroxisome (Kunze, 2020). Functionally, Drosophila peroxisomes are required for both VLCFA metabolism as well as managing reactive oxygen species (ROS) (Di Cara et al., 2017; Faust et al., 2014; Nakayama et al., 2011). Similarly, enzymes and regulatory proteins involved in LD formation and lipolysis are also highly conserved in flies (Heier and Kühnlein, 2018), as is the process of LD formation (Kuhnlein, 2012). Although protein families are conserved, the total number of LD regulatory proteins in Drosophila is less than that in mammals. Flies have only two PLINs: Lsd-1 and Lsd-2 (Beller et al., 2006; Bi et al., 2012; Guo et al., 2008). Upon phosphorylation, Lsd-1 facilitates lipid mobilization by recruiting Drosophila Hsl to the LD surface, facilitating lipolysis (Bi et al., 2012). Lsd-2 serves to protect the surface of LDs from lipases, such as the Drosophila ATGL homolog Brummer (Bmm) (Grönke et al., 2005). Embryonically derived Drosophila Schneider 2 (S2) cells have been used extensively to study LDs as they rapidly form multiple LDs when cultured in oleic acid (Beller et al., 2010; Guo et al., 2008; Kory et al., 2015; Krahmer et al., 2011; Sui et al., 2018; Wang et al., 2016; Wilfling et al., 2014, 2013). However, despite clear requirements for LDs during Drosophila development (Heier and Kühnlein, 2018), and conservation of Pex proteins and peroxisome metabolism (Anderson-Baron and Simmonds, 2019), little is known about how these organelles coordinate their activities.

Drosophila Pex gene mutations lead to altered levels of circulating fatty acids in larvae, causing developmental defects. These may be related to stored lipid availability, as altering the fatty acid in the larval diet can suppress some phenotypes associated with Pex gene mutations (Bülow et al., 2018; Chen et al., 2010; Di Cara et al., 2019; Liu et al., 2021; Sellin et al., 2018). The wide variety of lipids needed for cell structure, metabolism and energy storage during Drosophila development can be synthesized de novo or obtained by feeding (Heier and Kühnlein, 2018). The fat body constitutes a significant proportion of the larval body and is the major storage location for lipids (Church and Robertson, 1966; Musselman and Kuhnlein, 2018). Drosophila larvae feed almost constantly, storing TGs in the fat body (Heier and Kühnlein, 2018). Subsequent pupal development depends on fat body TG stores until the adult emerges from the pupa and begins to feed again (Musselman and Kuhnlein, 2018). The proportion of the larval fat body within a larva can be rapidly assayed by a simple buoyancy assay (Reis et al., 2010). To probe the requirements for peroxisomes during this stage of Drosophila development, we performed targeted RNA interference (RNAi) knockdown of Pex genes (Pexi) in the larval fat body. Although Pexi had the expected effect of inhibiting peroxisome proliferation, we noted that a few Pex genes had much stronger effects on lipid storage than others. To understand the underlying mechanism for how these few Pex proteins affected lipid storage, we compared mRNA expression in Drosophila S2 cells that were serum starved, cultured with oleic acid (+Oleate) or serum starved in the presence of a chemical inhibitor of Catalase, the primary ROS metabolism enzyme in peroxisomes. Of the 11 Peroxins needed for peroxisome assembly, only Pex2, Pex13 and Pex14 showed significant changes in mRNA levels when these various conditions were compared pairwise. To probe the mechanisms underlying this more direct linkage of a subset of Pex proteins to lipid storage, we examined LD formation in S2 cells cultured with excess oleic acid. There was a novel activity associated with only three Pex proteins – Pex3, Pex13 and Pex14 – in that they surrounded LDs independently of markers that would indicate they were incorporated into peroxisomes. This association was enhanced by mutations or Pexi suppressing Pex16 or Pex19, the key proteins that mediate insertion of Pex13 and Pex14 into the peroxisome bilayer membrane. Given that Pex14 was altered in differential RNA sequencing (RNASeq) comparison of serum-starved cells to those fed excess oleic acid. Pex14 knockdown had the strongest effect in the fat body; we focused on Pex14 in terms of LD function. We found that recruitment of Pex14 to LDs promotes TG storage and suppresses lipolysis, especially as cells transition from storing to releasing LD-stored fatty acids via lipolysis.

To characterize requirements for peroxisome activity in the fat body, we performed a systematic Pexi screen, targeting each Pex gene via fat body-targeted transgenes (Fig. 1A,B; Fig. S1). Fat body-targeted Pex14i, and Pex16i, had much larger effects on larval buoyancy; Pex19i had almost no effect (Fig. 1C). Fat body Pex14i had the strongest effect on buoyancy and was coupled with a significant reduction in TG, as measured by a glycerol assay, compared to control (P<0.01) (Fig. 1D). Most (>80%) larvae with targeted Pexi in the fat body survived to pupal stage on standard cornmeal media (Fig. 1E), suggesting that extensive peroxisome biogenesis is not prevalent in larval fat body cells. To examine the role of dietary lipids on the fat body-targeted Pexi phenotypes, second-instar, feeding-stage larvae were transferred to a lipid-reduced diet (Piper et al., 2014). Fat body Pexi larvae raised on a lipid-reduced (holidic) medium survived less well than control larvae, with the strongest effects seen with Pex3i (50% survival), Pex14i (80% survival) or Pex19i (60–90% survival) (Fig. 1E). Larvae will consume a lipid-enriched diet containing lard (Woodcock et al., 2015). When larvae were fed holidic food plus lard, the survival rate of Pex3i larvae was also lower than that of control larvae (30%; Fig. 1E). Fat body-targeted Pex13i (15–20%, P<0.01) or Pex14i (15–20%, P<0.001) significantly reduced larval survival on lard food (Fig. 1E). Neutral lipid staining showed LDs filling each cell, with peroxisomes distributed throughout the cytoplasm (Fig. 1F, LipidTOX). The Pex14 protein had two patterns of cellular localization; the first was punctate sites enriched in proteins with the SKL (PTS1) motif (Fig. 1F, SKL). Pex14 was also observed surrounding the LDs in a pattern that did not overlap with SKL (Fig. 1F, Pex14). In Pex14i larvae, Pex14 signal and mature peroxisomes (punctate SKL) were largely absent (Fig. 1G). The LDs in the Pex14i fat body appeared smaller and were fewer in number than those in control fat body (Fig. 1G). Pex19i fat body LDs were larger, and peroxisomes were absent (Fig. 1H). In Pex19i fat body cells, Pex14 surrounded very-large LDs (Fig. 1H). Enlargement of LDs and reduced peroxisome number were seen in Pex16i fat body (Fig. 1I), although to a lesser extent than in Pex19i fat body. Pex13i and Pex5i fat body had reduced peroxisome abundance and increased Pex14 signal adjacent to LDs (Fig. 1J,K). Pex11i fat body cells had a pattern of LDs and peroxisomes like that of control fat body cells (Fig. 1L). The fewer number of peroxisomes (punctate SKL) would be expected in Pex11i cells as the primary role of Pex11 is regulating fission of existing peroxisomes (Schrader et al., 2016). Some of the Pex14 signal surrounding LDs in fat body cells colocalized with Calnexin 99A (Cnx99A), an ER marker, but considerable ER-independent Pex14 signal surrounding LDs was also present (Fig. 1M).

Drosophila S2 cells form LDs consistent in both number and volume (Guo et al., 2008) when the culture medium is supplemented with oleic acid, an 18-carbon monounsaturated fatty acid (Darfler, 1990). Omitting serum from standard Schneider's medium (serum starved) induced a starvation response and increased the level of mRNAs encoding proteins involved in ROS and reactive nitrogen species (RNS) production (Fig. 2A; Table S1). Lsd-1 was significantly different [P<0.05, adjusted P (Padj)<0.1], and Pex2 and bmm were also different but at lower confidence (Padj>0.1). However, comparison of serum-starved to +Oleate cultured cells showed the greatest change in mRNAs of pathways affecting fatty acid metabolism and peroxisomal protein import, including Pex14 (Fig. 2B; Table S2). Other comparisons of serum-starved, standard and oleic acid culture showed changes in multiple genes previously associated with LDs (Figs S2 and S3). Notably, of all Pex genes, only Pex14 levels were significantly differently (Padj<0.1) enriched in S2 cells cultured in +Oleate condition compared to serum starvation condition. To address the potential effect of peroxisome ROS metabolism in the observed effects, RNASeq was also performed on serum-starved cells treated with 3-amino-1,2,4-triazole (3AT), which suppresses Catalase (Samis et al., 1972). Notably, very few mRNAs had different levels in serum-starved versus serum-starved +3AT cells (Table S4), suggesting that ROS metabolism is not linked to response to starvation or oleic acid in S2 cells (Table S1). Comparing serum-starved +3AT cells with those in standard culture showed (Table S5) differences in Lsd-1 and Pex13 at high confidence and Pex11ab at lower confidence (Padj>0.1), while comparison with +Oleate cells (Table S6) showed Pex14 changes at high confidence and Lsd-1 at lower confidence (Padj>0.1). For all comparisons, high-confidence changes in mRNAs encoding genes known to encode LD-resident or regulatory proteins (Beller et al., 2006; Cermelli et al., 2006; Guo et al., 2008) were present (Tables S1–S6).

The observed changes in mRNA levels of a subset of peroxisome protein import genes did not correlate with a corresponding increase in other Pex genes required for peroxisome proliferation. To confirm whether these Pex genes were linked to conditions that would alter the formation/lipolysis of LDs, we compared expression of each when +Oleate cells were subsequently transferred to serum-starved culture for 24 h, which would induce lipolysis of LD-stored TGs (lipolytic). The increased level of Pex14 in +Oleate cells was not maintained when cells were induced to metabolize LD-stored neutral lipids by transfer to serum-starved conditions (Lipolytic, Fig. 2C). The relative increase in Pex14 mRNA levels in +Oleate cultured cells was coupled with an increase in Pex14 protein (Fig. 2D). The relative change in Pex14 protein level was less than that seen at the mRNA level by quantitative reverse transcription PCR (qRTPCR), but this may be due to post-transcriptional regulation of the Pex14 mRNA (Dahan et al., 2022).

When S2 cells were cultured in standard conditions, mature peroxisomes (punctate SKL) and Pex14 largely overlapped (Fig. 2E). In +Oleate cultured S2 cells, Pex14 could be observed surrounding LDs. This signal did not overlap peroxisomes (punctate SKL, Fig. 2F). When S2 cells were subsequently placed in lipolytic conditions, Pex14 surrounded larger LDs (Fig. 2G). Quantification of three-dimensional colocalization showed that peroxisome numbers were relatively similar in cells cultured under standard, +Oleate and lipolytic conditions (Fig. 2H). In both fat body and S2 cells, the Pex14 surrounding LDs was independent of Abcd3, a non-Peroxin PMP inserted into the peroxisome membrane (Fig. S4A–D), and the punctate SKL that marks proteins concentrated inside mature peroxisomes (Fig. S4E–I). Pex14 protein was present in isolated LD fractions from Pex19 knockout (KO) S2R+ cells lacking peroxisomes (Fig. 2I). Notably, LD-associated Pex14 was resistant to alkaline carbonate treatment (Fig. 2I,J), indicating that Pex14 has a stable association with the LD surface.

In Pex14i S2 cells, PTS1-mediated (SKL) peroxisomal import was reduced (Baron et al., 2016). Pex14i affected peroxisome volume and number in +Oleate and lipolytic cultured S2 cells as well as influenced LD volume and number (Fig. 3A–H). Peroxisome volume was significantly reduced (P<0.01) in +Oleate Pex14i cells but had little effect in cells in lipolytic conditions (Fig. 3E). Conversely, Pex14i had little effect on peroxisome number in +Oleate cultured cells but a significant effect on peroxisome number in cells in lipolytic conditions (P<0.01, Fig. 3F). Changes in LD number and volume have been shown previously to correlate strongly with changes in TG storage in S2 cells (Guo et al., 2008). Pex14i had relatively small effects on LD volume or number in +Oleate cultured cells but caused LD fragmentation (increased number and smaller volume) when cells were transferred to lipolytic conditions (Fig. 3G,H). TG lipolysis is also reflected in changes in free glycerol levels in cell lysates (Tennessen et al., 2014). Pex14i led to an increase in lipolysis in lipolytic culture conditions (Fig. 3I), coupled with reduced peroxisome proliferation (Fig. 3F). Pex14i efficacy of >90% was confirmed by qRTPCR (Fig. 3J).

Peroxisome formation in Drosophila requires the collective activity of 11 conserved Pex proteins (Anderson-Baron and Simmonds, 2019; Baron et al., 2016; Faust et al., 2012). Peroxisomes have a half-life of ∼2 days (Nordgren et al., 2013), and it is unlikely Pex14 is released once incorporated into the peroxisomal membrane (Natsuyama et al., 2013). Myc-tagged Pex14 surrounds peroxisomes in standard culture (Fig. 3K) and surrounds newly formed LDs after +Oleate culture (Fig. 3L). To determine whether the Pex14 surrounding newly formed LDs was newly translated from the increased Pex14 mRNA, we performed pulse-chase protein labeling. Myc-Pex14 transfected cells were incubated with ³⁵S-Met immediately after transfection (0–24 h) or 24 h after transfection (24–72 h). Cells that were labeled immediately after transfection were washed after 24 h to remove excess ³⁵S-Met. Autoradiography of total labeled protein from the LD fraction of control and +Oleate cultured cells pulse labeled at 0 h showed that only a few newly synthesized proteins were recruited to LDs in cells grown in +Oleate culture. This increased over time in +Oleate culture with a prominent band at 45 kD (Fig. 3M). In addition, 6xMyc-Pex14 immunoprecipitation/autoradiography showed that newly translated Pex14 was recruited to LDs (Fig. 3N). Pex14 surrounded large LDs when cells were transferred from +Oleate to lipolytic conditions (Fig. 3O).

Pex19 mediates Pex13 and Pex14 insertion into bilayer membranes that contain Pex3 (Itoh and Fujiki, 2006). Pex14 can associate with mitochondrial membranes when Pex19 is absent (Sacksteder et al., 2000). In +Oleate cultured, Pex19KO, S2R+ cells, the majority of Pex14 LDs was associated with LDs (Fig. 4A), similar to Pex19i fat body (Fig. 1H). Confocal imaging showed that Pex14 was near ER surrounding LDs (Fig. 4A; Fig. S5A). However, Pex14 surrounding LDs did not appreciably colocalize with Cytochrome-C (CytC), which marks mitochondria (Fig. 4B; Fig. S5B). Pex19KO S2R+ cells co-expressed mNeonGreen-SKL, which was dispersed throughout the cell, indicating that the Pex13/Pex14-mediated peroxisome matrix protein import was absent. Using super-resolution (30 nm) stimulated emission depletion (STED) imaging, we found that Pex14 surrounding LDs was largely overlapping with a protein marker for the LD surface (Fig. 4C, LiveDrop) (Wang et al., 2016). The relative spatial relationship of Pex14 and Cnx99A (ER) at the LD surface (LiveDrop-GFP) at 30 nm resolution was examined in S2 cells stably expressing LiveDrop-GFP, which suggested that Pex14 was more closely associated with the surface of LDs than the surrounding ER (Fig. 4D).

mRNAs encoding LD-associated proteins like Hsl and Lsd-2 were significantly altered (P<0.05, Padj/FDR<0.1) in S2 cells cultured in +Oleate compared to serum-starved conditions, as were those encoding Pex3 and Pex13 (Tables S2 and S6). Overexpression of Bmm in S2 cells caused rapid LD fragmentation, indicating increased lipolysis (Fig. 5A). When Pex14 was co-overexpressed in these cells, LD fragmentation also occurred (Fig. 5B). Overexpression of Hsl similarly caused LD fragmentation, but this was suppressed when Pex14 was overexpressed in the same cells (Fig. 5C,D). Co-overexpression of Pex13 and Hsl also suppressed LD fragmentation to a lesser extent (Fig. S5C,D). Three-dimensional quantification of relative protein localization in these cells showed that Pex14 was concentrated in a region immediately adjacent to LDs, where more than 50% of the signal corresponding to overexpressed Bmm and Hsl lipase was concentrated (Fig. 5B,D,E). However, in lipolytic conditions, although Pex14 and Bmm remained concentrated adjacent to LDs, there was a reduction in Hsl overlap when Pex14 was at high levels (Fig. 5D,F). Pex14-induced repression of Hsl surrounding LDs was linked to an increase in LD volume and decrease in number (Fig. 5C,D,G,H), suggesting suppression of Hsl-mediated lipolysis (Marcinkiewicz et al., 2006).

Although there is some overlap in Drosophila PLIN activities, it is thought that Lsd-1 facilitates LD lipid mobilization by Hsl while Lsd-2 suppresses Bmm-mediated lipolysis at LDs (Beller et al., 2010; Marcinkiewicz et al., 2006). Simultaneously overexpressing Pex14 and Lsd-2 resulted in both being largely adjacent to LDs (Fig. 5I). Conversely, co-overexpression of Pex14 and Lsd-1 resulted in suppression of Pex14 recruitment to LDs (Fig. 5J). Although overexpressed Lsd-1 or Lsd-2 are both strongly recruited to LDs (Fig. 5K), Lsd-1 significantly repressed Pex14 co-recruitment (P<0.01, Fig. 5L). Notably, whereas Pex3 localization to LDs was unaffected by Lsd-1 overexpression in +Oleate conditions, Pex13 was strongly affected (Fig. 5M,N).

When inserted into the peroxisome bilayer membrane, PEX14 is proposed to have an N-in C-out topology (Barros-Barbosa et al., 2019; Reuter et al., 2021). Several N- or C-terminal truncations of Pex14 were tested for recruitment to LDs or peroxisomes in +Oleate cells (Fig. 6A). mRFP-tagged full-length Pex14^aa1–280 produces a punctate pattern indicating peroxisomes as well as strongly overlaps with LiveDrop-GFP in live-imaged S2 cells (Fig. 6B). The N-terminal 1–87 amino acids (aa) of Pex14 contain a region homologous to the ‘PEX14’ domain that mediates interaction with PEX5 and PEX13. This region was insufficient to mediate LD or peroxisome association and was relatively homogeneous within the cell (Fig. 6C). A longer truncated form of Pex14^aa1–117 including the N-terminal Pex14, but not the TM, domains was cytosolic (Fig. 6D). Inclusion of the predicted TM domain (aa119–141) of Pex14 retained LD association (Fig. 6E). Truncations beginning with the N-terminal end found that a region surrounding the TM domain (aa76–147) that did not include the Pex14 domain (aa34–75) or the C-terminal portion of the protein (aa117–280) was sufficient to mediate recruitment to LDs (Fig. 6F–H). The C-terminal of Pex14 missing the TM domain (aa141–280) does not surround LDs (Fig. 6I). In all cases, if the TM domain was not present, Pex14 truncations were not seen surrounding LDs (Fig. 6D,I). Only Pex14 transgenes that contained the TM domain were recruited to LDs (Fig. 6J). In S2 cells overexpressing Hsl, full-length Pex14 or truncations that contained both the TM domain and the C-terminal end suppressed Hsl-mediated LD fragmentation (Fig. 6K–M).

To probe the linkage between the canonical role for Pex14 in peroxisome assembly and the newly identified role at LDs, we tested the effect of knockdown of Pex genes that affected different aspects of PMP insertion and enzyme import in S2 cells stably transformed with LiveDrop-GFP. In control cells cultured in +Oleate conditions, Pex14 localized to both punctate spots indicating peroxisomes and colocalizes with LiveDrop (Fig. 7A; Fig. S6A). Pex16 initially recruits Pex3 to the ER or to existing peroxisomes (Aranovich et al., 2014; Matsuzaki and Fujiki, 2008). Pex3 then serves as the docking module for the cytosolic chaperone, Pex19, to import PMPs including Pex14 (Fang et al., 2004). In Pex3i cells, fewer Pex14 punctate spots were observed, as PMP insertion would be suppressed, and the colocalization of Pex14 with LiveDrop was enhanced (Fig. 7B; Fig. S6B). Similarly, in Pex16i or Pex19i cells, colocalization of Pex14 with LiveDrop was enhanced, and the number of punctate Pex14 spots in the peroxisome characteristic pattern was suppressed (Fig. 7C,D; Fig. S6C,D). Pex13 is thought to interact directly with Pex14 to help its trafficking to peroxisomes and to form the docking complex for importing proteins such as peroxisomal enzymes (Itoh and Fujiki, 2006; Matsuzaki and Fujiki, 2008). Reduced punctate Pex14 spots were observed in Pex13i cells, but colocalization of Pex14 with LiveDrop was enhanced (Fig. 7E; Fig. S6E). In Pex1i cells, a similar pattern of punctate spots was observed relative to control, but they were smaller in size (Fig. 7F; Fig. S6F), as Pex1 functions after PMP insertion to recycle Pex5 and promote enzyme import through the Pex14 docking complex (Fujiki et al., 2020). There was no noticeable enhancement of Pex14 localization to the LD surface in Pex1i cells (Fig. 7F; Fig. S6F). Pex5 interacts directly with Pex14 to import peroxisome enzymes through the docking complex (Gaussmann et al., 2021). In Pex5i cells, fewer punctate Pex14 spots were present but there was no enhancement of Pex14 colocalization with LiveDrop (Fig. 7G; Fig. S6G). In Pex14i cells, few Pex14 punctate spots were observed and little colocalization with LiveDrop was observed (Fig. 7H; Fig. S6H). Only Pex3i, Pex13i, Pex16i and Pex19i caused significant increases in the proportion of Pex14 surrounding LDs (Fig. 7I). This is notable as Pex3, Pex16 and Pex19 are the Pex proteins needed to insert PMPs like Pex14, and likely Pex13, into the peroxisome membrane (Fujiki et al., 2020).

A subset of Pexi was tested in Pex19KO cells under lipolytic conditions to examine peroxisome-independent functions of each Pex gene in lipid storage. Pex1i and Pex3i had little effect on the volume or number of LDs in cells cultured in lipolytic conditions (Fig. 7J–O; Fig. S6I). Pex5i reduced the volume of LDs but increased their average number, suggesting increased fragmentation (Fig. S6I). Pex13i or Pex14i caused significantly reduced total LD volume in lipolytic cells (P<0.001, Fig. 7P; Fig. S6I).

Very few peroxisomes (punctate SKL) overlap LDs (LiveDrop) in S2 cells cultured in +Oleate conditions (Fig. 8A). Transient transfection of transgenes expressing HA-tagged Pex proteins was used to determine whether Pex proteins other than Pex14 were also associated with LDs. HA-Pex1 was largely cytoplasmic, but a proportion of HA-Pex1 was concentrated at peroxisomes (Fig. 8B). Overexpressed HA-Pex3 suppressed peroxisome formation. HA-Pex3 formed a reticular pattern surrounding the nucleus as well as surrounding, but not overlapping, LiveDrop-GFP (asterisks) and was also concentrated at the few remaining peroxisomes (Fig. 8C). HA-Pex5 was also homogenous throughout the cytoplasm as expected by its function, but again a significant proportion was concentrated at peroxisomes. However, no colocalization with LiveDrop was ever observed (Fig. 8D). HA-Pex10, which together with Pex2 and Pex12 forms a E3 ligase complex needed to traffic Pex5 back out of the peroxisome (Fujiki et al., 2020), strongly overlapped with peroxisomes but did not appreciably colocalize with LiveDrop (Fig. 8E). Two distinct phenotypes were observed with overexpression of HA-Pex13. In cells in which transgene expression was relatively lower (Fig. 8F, right-most cell), HA-Pex13 overlapped with peroxisomes. However, when HA-Pex13 levels were relatively higher, fewer peroxisomes were observed and the majority of HA-Pex13 overlapped the LiveDrop signal (Fig. 8F, second left, asterisk). Finally, HA-Pex16 cells had a reduced number of peroxisomes compared to control cells (Fig. 8A) and the HA-Pex16 signal did not overlap with LiveDrop (Fig. 8G). HA-Pex19 cells also had a similar number of peroxisomes to control cells, but they were often larger (Fig. 8H). HA-Pex19 also did not show observable colocalization with LiveDrop (Fig. 8H).

The recruitment of Pex14 to LDs independently of mature peroxisomes was conserved in mammalian cells. When NRK cells were cultured in standard conditions, the PEX14 signal overlapped peroxisomes (Fig. 8I). When cultured in +Oleate conditions (1 mM oleic acid for 48 h), a portion of the PEX14 distinct from ABCD3 surrounded LDs (Fig. 8J). The proportion of PEX14 adjacent to LDs increased significantly from 4.7% in standard conditions to 32.8% in +Oleate conditions (Fig. 8K). However, the proportion of mature peroxisomes adjacent to LDs was largely the same in standard or +Oleate conditions (Fig. 8L). Like what was observed in S2 cells, no change in NRK cell peroxisome volume was observed in +Oleate conditions (Fig. 8M), but peroxisome number increased significantly (P<0.01, Fig. 8N). Corresponding to what was seen in +Oleate cultured S2 cells (Fig. 2D), PEX14 levels were relatively higher in NRK cells cultured in +Oleate at 24 h and remained higher even after 48 h (Fig. 8O).

Here, in Drosophila, we identified lipid-responsive LD association of a subset of Pex proteins (Pex3, Pex13 and Pex14) that were not part of mature peroxisomes. When associated with LDs, Pex14 affected recruitment of Hsl and influenced TG storage and mobilization of TGs (Fig. 5C,D). Although Pex13 overexpression affected Hsl recruitment to LDs, the effect was more muted (Fig. S5C,D). Thus, our data favor a model whereby Pex14 plays a more active role in LD regulation that is functionally and/or temporally distinct from previously characterized roles for Pex14 in peroxisome biogenesis or movement (Bharti et al., 2011; Gaussmann et al., 2021; Reuter et al., 2021). When Pex14i +Oleate cultured cells were transferred to lipolytic conditions, LD fragmentation occurred (Fig. 3A–D,G,H). Further supporting a role for Pex14 in promoting TG storage, the LDs in cells overexpressing both Pex14 and Hsl were significantly larger and less numerous than those in cells overexpressing Hsl alone (Fig. 5G,H). The Drosophila DGAT homolog, Diacylglycerol O-acyltransferase 2 (Dgat2), acts in opposition to LD lipolysis (Wilfling et al., 2013). However, RNASeq indicated that Dgat2 mRNA was absent from S2 cells in all conditions. The relative increase in TG lipolysis (free glycerol) levels in Pex14i lipolytic cells compared to control (Fig. 3I) supports a model in which Pex14 also suppresses lipase activity at LDs. This model is also supported by observations that LD volume in the fat body and overall TG content is significantly reduced by Pex14i in the larval fat body (Fig. 1E).

The mechanism underlying non-peroxisomal retention of Pex14 at LDs is suggested by Pex19KO cells (Fig. 7P). Pex19KO or Pex19i cells had a marked increase in Pex14 surrounding LDs in +Oleate cultured cells (Fig. 4A,B and Fig. 7D). Drosophila Pex19 mutants have elevated activity of cytoplasmic (non-LD-associated) lipases, causing mitochondrial dysfunction and lipotoxicity (Bülow et al., 2018). Given that Pex19 directs Pex14 to the peroxisome membrane, loss of Pex19 would promote association of Pex14 with LDs, which could also contribute to disruption of lipid metabolism. This could also explain enhancement of Pex14 LD recruitment by Pex3i, Pex13i and Pex16i (Fig. 7B,C,E). The targeting of Pex14 to a peroxisomal membrane requires Pex3, Pex16 and Pex19 (Fujiki et al., 2020). Pex13 also assists insertion of Pex14 into the peroxisome membrane (Itoh and Fujiki, 2006). Enhanced localization of Pex14 to the LD surface in Pex3i, Pex13i, Pex16i or Pex19i cells strongly suggests that Pex14 association with LDs is not part of canonical peroxisomal protein import (Fig. 7A–E; Fig. S6A–E). Notably, Pex14 trafficking to the LDs was not affected by Pexi of genes needed for later stages of peroxisome biogenesis like Pex1 or Pex5 (Fig. 7F,G; Fig. S6F,G). In Pex3-deficient yeast, Pex14 is trafficked to the ER via the Sec61 translocon, which was also shown recently to assist in translocation of LD proteins to the ER in vitro (Leznicki et al., 2022; van der Zand et al., 2010).

The targeting of Pex14 to peroxisomes relies on successful binding to the cytosolic chaperone, Pex19, via its N-terminal Pex14 domain (Itoh and Fujiki, 2006; Neufeld et al., 2009). In Pex19-deficient cells, Pex14 accumulates at the ER and at mitochondria in yeast and mammals, respectively (Sacksteder et al., 2000; van der Zand et al., 2010). It has been proposed that Pex14 can be inserted into mitochondrial membranes (Sugiura et al., 2017). In Pex19KO cells, we found that all LD-associated Pex14 signal was distinct from mitochondria (Fig. 4B), while some signals were in close proximity to the ER (Fig. 4A). Super-resolution imaging showed that Pex14 surrounding LDs is not entirely associated with ER (Fig. 4D). This is different from other proteins identified previously to mediate peroxisome LD-association, such as FAF2 (also known as UBXD8) (Schrul and Kopito, 2016).

Blocking peroxisome Class I and II PMP bilayer membrane insertion pathways via Pex19KO, Pex3i, Pex16i or Pex19i increased the relative Pex14 association with LDs (Fig. 4A,B and Fig. 7B–D), suggesting that Pex14 at LDs is not part of a PPV or peroxisome bilayer membrane. Mutant forms of PEX14 in which the TM domain was non-functional did not localize to any membranes in Chinese hamster ovary cells (Itoh and Fujiki, 2006). Similarly, although the TM domain is sufficient to associate Pex14 with the LD surface (Fig. 6), the Pex14 domain that mediates formation of the docking complex is not, consistent with enhanced Pex14 localization observed in Pex19KO/Pex19i cells (Fig. 4A-C and Fig. 7D).

Insertion of Pex14 into the peroxisome membrane to form the docking complex creates large pores through which folded proteins can be imported by Pex5 (Azevedo and Schliebs, 2006; Will et al., 1999). This likely occurs in partnership with Pex13 (Otera et al., 2002). Thus, the TM domain of Pex14 faces hydrophilic and hydrophobic environments and has an amphipathic helix conformation. This same amphipathic property of the Pex14 TM domain (Fig. S7) might insert into the LD monolayer membrane. This would account for the alkaline carbonate-resistant retention of Pex14 in fractionated LDs (Fig. 2I). Co-overexpression of Pex14 and Hsl shows that the C-terminal region of Pex14 can inhibit Hsl-mediated LD fragmentation (Fig. 6K–M). However, the C-terminal half of Pex14 is poorly characterized. There is a coiled-coil domain (CCD) near the TM domain sufficient for homodimerization/oligomerization and subsequent interaction with Pex13 (Itoh and Fujiki, 2006). The CCD is predicted within the C-terminal end (aa159–209) of Pex14 (Ludwiczak et al., 2019). The role for this CCD or dimerization/oligomerization of Drosophila Pex14, or interaction of Pex14 with Pex13 at LDs, remains to be determined. PEX14–PEX5 interaction is required to regulate peroxisome–LD interactions and ATGL recruitment during fasting-induced lipolysis at LDs (Kong et al., 2020). Retention of Pex14 at the LD surface may play a role in priming such events.

Whether Pex14 requires the canonical partner Pex13 remains to be investigated. Pex3 is likely to have a separate role at the LDs from Pex13 and Pex14 because FAF2, trafficked to LDs via the action of Pex3 and Pex19, inhibits the action of ATGL (Olzmann and Carvalho, 2019; Schrul and Kopito, 2016). This provides explanation for the sensitivity of Pex3i larvae in lipid-reduced condition (Fig. 1E) and less pronounced LD fragmentation in Pex19KO Pex3i cells under lipolytic condition (Fig. 7P; Fig. S6I). Both Pex13i larvae (Fig. 1E) and Pex19KO Pex13i cells (Fig. 7P) suggest that Pex13 confers a similar, but less pronounced, function to Pex14 at the LD surface. Co-overexpression of Hsl and Pex13 in lipolytic condition suppressed LD fragmentation like Pex14 (Fig. 5C,D; Fig. S5C,D). Therefore, Pex13 may assist Pex14 in mitigating Hsl-mediated lipolysis without contributing to its trafficking to the LD surface (Fig. 7E). Supporting the model that Pex14 and Pex13 work together at the LD surface is the observation that changes in the amount of PLIN affects LD localization of Pex14 and Pex13 (Fig. 5J,N). Overexpression of Lsd-1 suppressed the localization of Pex13 and Pex14 to the LD surface (Fig. 5J,N), but elevated Lsd-2 levels did not affect Pex14 association with LDs (Fig. 5I).

A major mechanism regulating protein association with the LD surface is inter-molecular competition for limited space as the LD grows and shrinks (Kory et al., 2015). As the LD surface shrinks during TG lipolysis, proteins lose association from the LD surface, creating space for lipases (Kory et al., 2015). Exclusion of Pex14 from LDs when Lsd-1 levels are elevated (Fig. 5L) suggests that this is occurring. As overexpression of Lsd-2 does not affect Pex14-LD localization, the effect of Lsd-1 overexpression on Pex14 is likely a specific, rather than general, effect. The major circulating neutral lipid in Drosophila is DG, which is accumulated in the fat body for energy storage (Heier and Kühnlein, 2018). Although Pex14 levels at the LD have little effect on the TG lipase Bmm, Pex14 antagonizes Hsl at the LD (Fig. 5A–H). This supports a model whereby recruitment of Pex14 to the LD surface perturbs the interaction between Hsl and Lsd-1, blocking the recruitment of Hsl to the LD, with the antagonistic effects of overexpression of Lsd-1 on Pex14-LD localization (Fig. 5J). Further studies are required to determine the molecular mechanism by which Pex14 suppresses Hsl recruitment to LDs.

Given the absence of peroxisome proliferation observed in S2 cells transferred to +Oleate conditions, it is likely that newly synthesized Pex13 and Pex14 directly associate with LDs. A local pool of proteins associated with the LD surface needed to form the peroxisome docking complex could couple local peroxisome assembly to TG lipolysis. However, this model, including any Pex14-dependent or -independent role for Pex3 and Pex13 recruitment to LDs, needs to be addressed experimentally.

S2 and S2R+ Pex19KO cells were obtained from the Drosophila Genomics Resource Centre. Cell identity was validated by RNASeq and lack of contamination tested by visual observation or qRTPCR as needed. ‘Standard’ conditions for culturing S2 and S2R+ cells were Schneider's medium (Sigma-Aldrich S0146) containing 10% fetal bovine serum (FBS; Thermo Fisher 12483-012) at 25°C. The standard culture conditions for NRK cells were Dulbecco's modified Eagle medium (Sigma-Aldrich D5796) containing 10% FBS at 37°C and 5% CO₂. The +Oleate culture conditions used in this study are the same as used previously to induce LDs in S2 cells (Guo et al., 2008; Kramer et al., 2018), where standard medium was supplemented with 1 mM oleic acid (+Oleate, Sigma-Aldrich O1008) bound to fatty acid-free bovine serum albumin (BSA; Sigma-Aldrich A8806). Cells were maintained in +Oleate conditions for 24 h to 48 h. To induce LD lipolysis, S2 cells cultured in +Oleate conditions for 24 h cells were washed 1× in fresh Schneider's medium and transferred to Schneider's medium without FBS or oleic acid supplementation for 24 h (Guo et al., 2008). All culture media were supplemented with 100 U penicillin/ml and 100 μg streptomycin/ml (Thermo Fisher 15140-122). S2, S2R+ and NRK cells were passaged at log phase before they reached confluency. Cultures were not used beyond passage 25.

The full-length Pex14 open reading frame (Baron et al., 2016) cloned into PENTR-D (Thermo Fisher) was transferred to pDEST-17 (Thermo Fisher) using LR ClonaseII (Thermo Fisher 11791-020) and expressed in BL21-AI Escherichia coli (Thermo Fisher C6070-03) grown to an OD₆₀₀ of 0.4 at 37°C with shaking by addition of 0.2% L-arabinose and then cultured at 25°C for 3 h. Bacterial cell pellets were lysed by incubation in 8 M urea, and cleared by centrifugation at 20,000 g for 30 min at 25°C. The cleared lysate was applied to a 1 ml HisTrap column (Cytavia 17524701), using the Akta-Start His-tagged purification protocol (Cytavia). Purified protein was eluted using a stepwise imidazole gradient. Fractions containing purified Pex14 were combined, transferred to dialysis tubing (8000 MWCO, Spectrum 132660) and desalted by buffer exchange in 5 l 1× PBS overnight. Purified protein was concentrated in an Amicon Ultra 15 centrifugal filter (Millipore-Sigma UFC900308) to 1 mg/ml and injected into guinea pigs (Pocono Rabbit Farms and Laboratories). The partially purified serum was validated for antigen specificity by western blotting against purified bacterially expressed Pex14 as well as lysates from S2 cells, S2 cells expressing Pex14-GFP fusions and Pex14i S2 cells, confirming the presence or absence of a single band of expected size.

The w¹¹¹⁸ strain was obtained from the Bloomington Drosophila Stock Center (BDSC). All crosses were performed at 25°C. TRiP UAS-RNAi transgenes against Pex genes were obtained from the BDSC. These included the following: Pex1i BDSC:28979 (y¹ v¹; P{y^+t7.7 v^+t1.8=TRiP.HM05190}attP2); Pex1i BDSC:51497 (y¹ v¹; P{y^+t7.7 v^+t1.8=TRiP.HMC03252}attP2); Pex5i BDSC:55322 (y¹ v¹; P{y^+t7.7 v^+t1.8=TRiP.HMC04009}attP40); Pex14i BDSC:77180 (y¹ v¹; P{y^+t7.7 v^+t1.8=TRiP.HMC06491}attP40); and Pex16i BDSC:57495 (y¹ sc* v¹sev²¹; P{TRiP.HMC04810}attP2). Additional UAS-RNAi transgenic stocks targeting Pex genes were obtained from the Vienna Drosophila Resource Center (VDRC). These included the following: Pex2i VDRC:108578 KK; Pex3i VDRC:11017 GD; Pex10 VDRC:110405 KK; Pex11abi VDRC:105654 KK; Pex12i VDRC:34671 GD; Pex13i VDRC:39544 GD; Pex14i VDRC:42590 GD; Pex16i VDRC:34296 GD; Pex19i VDRC:100746 KK; and Pex19i VDRC:22064 GD. Driver lines used were as follows: y¹, w*, UAS-Dicer; P{w^+mC=r4-GAL4}3; y¹, w*, UAS-Dicer; P{TubP-GAL4}LL7/TM3(Ubi-GFP), Sb¹ (Kyoto Drosophila Stock Center 108069). Fly stocks were maintained on the BDSC standard cornmeal food recipe and passaged once a week to prevent overcrowding. To obtain second-instar larvae for feeding assays, flies of each genotype were transferred to bottles with food for 24 h. After 3 days, the food was scraped from the bottle and rinsed with distilled water to expose larvae. The developmental stage was confirmed by the characteristic shape of the larval mouth hooks.

Myc-, HA- or FLAG-tagged Pex vectors were described previously (Baron et al., 2016). Pex14 truncations were generated by PCR into the pENTR/D Gateway entry vector by TOPO cloning (Thermo Fisher K240020) and recombined into pARW or pAWR using LR ClonaseII. To generate LiveDrop (Wang et al., 2016), DNA encoding aa160–216 of Drosophila Glycerol-3-phosphate acyltransferase 4 (Gpat4) was cloned via PCR into pENTR/D and recombined into pAWR or pAWG vectors. Plasmids were transfected into Drosophila S2 cells using Effectene transfection reagent (Qiagen 301425). S2 cells were passaged 24 h before transfection. Approximately 5.0×10⁵ cells were transfected with 150 ng plasmid DNA. Transfected S2 cells were incubated at 25°C for 48–72 h before fixation for imaging. Clonal populations of stably transfected LiveDrop GFP S2 cells were selected by fluorescence-activated cell sorting (Faculty of Medicine and Dentistry Flow Cytometry Facility, University of Alberta) followed by limiting serial dilution in conditioned Schneider's medium in 96-well plates (Luhur et al., 2019).

To label newly synthesized protein in S2 cells, Schneider's medium (Schneider, 1972) without L-methionine, L-Cysteine or yeast extract was supplemented with 10% dialyzed FBS (Thermo Fisher A33820) and 100 μl Easy Tag Express ³⁵S Methionine/Cysteine (Perkin Elmer NEG 772002MC) at either 0 h or 24 h after transformation with a 6xMyc-Pex14 (pAMW-Pex14) as described above. Twenty-four hours after transformation, 1 mM oleic acid was added. Cells in which ³⁵S was added at 0 h were washed in complete Schneider's medium (Schneider, 1972) at 24 h. Cells in which ³⁵S Methionine/Cysteine mix was added at 24 h were washed in Schneider's medium at 72 h. The cells were pelleted at 72 h and rinsed with PBS containing cOmplete Protease Inhibitor (Millipore 04693159001) and the LDs selectively fractionated as described below. The presence of radiolabeled 6xMycPex14 was analyzed by immunoprecipitation from each fraction and detected by autoradiography.

Fixed S2 cells were observed using a Zeiss 63× oil immersion objective (1.4 NA) on a Zeiss Axio Observer M1 microscope with an ERS spinning disk confocal and a C9100 EMCCD camera (Hamamatsu) using Volocity imaging software (PerkinElmer) or a Zeiss LSM700 confocal and Zen 2009 software (Zeiss). Live S2 cells suspended in Schneider's medium were observed using a C-Apochromat 63×/1.2 W Corr (0.14–0.19 mm) objective. Super-resolution imaging was performed using a Leica Plan Apochromat 100× oil immersion (1.4 NA) objective on a Falcon SP8 STED. The depletion lasers were set to provide 30 nm lateral resolution. Image stacks were captured at 130 μm vertical (z) spacing (ERS) or 10 to 25 nm (LSM700).

Cells were grown in six-well dishes containing a #1.5 cover glass coated with concanavalin A (Millipore-Sigma; Rogers et al., 2002). Cells were fixed for 10–20 min in 4% paraformaldehyde in PBS (Baron et al., 2016). Fat bodies were hand dissected from third-instar larvae and fixed for 60 min in 4% paraformaldehyde dissolved in PBS. Following fixation, cells or tissues were incubated in blocking solution (4% BSA in PBS) and then for 1 h or overnight in primary antibody in blocking solution. For Pex14, CytC and Cnx99A primary antibodies, 0.1% Tween-20 was used instead of 4% BSA in the blocking solution. Primary antibodies included monoclonal mouse anti-FLAG M2 (1:200, Sigma-Aldrich F3165), rabbit anti-Myc (1:200, Sigma-Aldrich SAB4301136), monoclonal mouse anti-Myc (1:250, 9B11, Cell Signaling Technology 2276S), monoclonal rat anti-HA (1:2000, Millipore-Sigma 3F10), guinea pig anti-Pex14 (1:1000), rabbit anti-Abcd3 primary antibody (1:500) (Huang et al., 2020), mouse anti-Cytochrome C (1:500, BD Pharmingen 7H8.2C12), mouse anti-Cnx99A (6-2-1, S. Munro, Developmental Studies Hybridoma Bank) and rabbit anti-SKL antibody (1:200, R. Rachubinski, University of Alberta). Primary incubation was followed by washing 3×20 min in block solution followed by incubation with donkey anti-mouse AlexaFluor 568 (Life Technologies A10037), donkey anti-mouse AlexaFluor 647 (Invitrogen A31571), donkey anti-guinea pig Cy3 (Jackson ImmunoResearch 706-165-148), donkey anti-guinea pig 647 (Jackson ImmunoResearch 706-606-148), donkey anti-rat A568 (Abcam ab175476), donkey anti-rabbit AlexaFluor 488 (Invitrogen A21206), or donkey anti-rabbit 647 (Abcam ab150067), all at 1:2000 dilution. To detect LDs, fixed cells or tissues were incubated with LipidTOX Deep Red (647 nm, Thermo Fisher) or AUTODOT (405 nm Monodansylpentane, Abcepta) at 1:500 dilution for 1 h after secondary antibody incubation, while live cells were incubated in AUTODOT for 30 min prior to imaging. Cells and tissue were mounted in ProLong Gold mounting medium (Thermo Fisher P36930).

NRK cells were fixed in 4% paraformaldehyde and blocked in 3% BSA for 1 h, incubated in rabbit anti-PEX14 (Thermo Fisher PA5-78103) and mouse anti-ABCD3 (R. Rachubinski, University of Alberta) at 1:200 dilution, for 1 h. Primary antibody incubation was followed by 3× washes in blocking solution followed by incubation with goat AlexaFluor 568 anti-mouse or AlexaFluor 488 anti-rabbit (1:2500, Abcam ab175473 and ab150077, respectively).

Image stacks of individual images sampled at sub-Nyquist Z-resolution comprising the entire cell volume were deconvolved by Classical Maximum Likelihood Estimation (confocal) algorithm in Huygens Professional Software (Scientific Volume Imaging) using an experimentally determined point spread function constructed from multiple images of 0.1 μm Tetraspeck beads (Thermo Fisher T7279). Three-dimensional colocalization analysis was performed using Huygens Professional Software (Scientific Volume Imaging) using Pearson's coefficient (Adler and Parmryd, 2010). Average peroxisome or LD volume and number per cell were calculated using IMARIS software v9 (Oxford Instruments). To validate our estimation of background fluorescence colocalization, measurements were also calculated on images in which one channel was shifted 90°, relative to the other (Dunn et al., 2011). In all cases, measured background colocalization was less than 10%. The proportion of signal surrounding LipidTOX or AUTODOT marked LDs (% Adjacent) was performed using a modified method used previously for signal adjacent to peroxisomes (Dahan et al., 2022). Briefly, a surface corresponding to each LD signal above the local background was generated encompassing the LD volume using the ‘Surfaces’ function. This volume was expanded by 6 voxels, and the signal within inner shell subtracted from the outer shell and reported as a percentage of total signal within the cell. All quantifications were based on at least three independent biological replicates with five or more cells in each replicate. Unless stated otherwise, colocalization significance and organelle volume/number data were assayed using an unpaired two-tailed Student's t-test.

A prevalidated template library (Foley and O'Farrell, 2004) was used to generate dsRNA using T7 RNA Polymerase (Thermo Fisher EP0111) or a HiScribe T7 High Yield RNA Synthesis Kit (New England BioLabs E2040S). S2 cells were passaged 24 h prior to dsRNA treatments. Effectene Transfection Reagent (Qiagen 301425) was used to enhance dsRNA uptake. Cells were incubated with ∼15 µg dsRNA for 4 days (reapplied after 48 h) and subsequently incubated in 1 mM oleic acid containing media free of dsRNA for 24 h. A scrambled dsRNA amplicon was used as a control (forward primer sequence, 5′-GTGAAGAGGTCAGAGGCCTG-3′; reverse primer sequence, 5′-ACAGTCTAGCGTTCCTTGAGG-3′).

RNA was isolated from S2 cells or from larvae using the RNeasy Plus Mini Kit (Qiagen 74134) and reverse transcribed using the Maxima H minus system (Thermo Fisher K1681). Quantification was performed using Perfecta SYBR Green FastMix (QuantaBio 95118) and an Eppendorf MasterCycler RealPlex2. All samples were measured in triplicate and calculations were made relative to Ribosomal protein L30 (RpL30) expression. Primers used for each of the target genes were experimentally validated pairs reported in FlyPrimerBank (Hu et al., 2013). For all qRTPCR experiments, values reported are averages based on three biological replicates. Statistical significance was determined by unpaired Student's t-test or one-way ANOVA.

Three independent replicates of S2 cells cultured in Schneider's medium (starved), Schneider's+FBS (standard) or Schneider's plus oleic acid (+Oleate) culture conditions were compared. A replicate of starved cells was also treated with 3AT (Samis et al., 1972). Total RNA was isolated from each using the RNeasy Plus Mini Kit. RNA integrity was verified using an Agilent RNA Nano assay (Agilent Genomics 5067-1511). Ribosomal RNA was subtracted from samples using a Ribo-Zero Gold rRNA Removal Kit (Illumina 20040526). Libraries were prepared using a NEBNext Ultra RNA Library Prep Kit and NEBNext Multiplex Oligos (New England Biolabs E7530L and E7335L). Library quality and size distribution were confirmed by running an Agilent High Sensitivity DNA assay (Agilent Genomics 5067-4626), and the average size of library inserts was verified to be 290–300 bp. Then, 10 pM of each of the libraries was loaded onto an Illumina MiSeq v2 300 cycle kit (2×150 cycles, paired-end reads, MS-102-2003). Paired-end reads were aligned to the Drosophila melanogaster genome (6.28 release) using HiSat2 (Kim et al., 2015). Individual read counts were mapped to specific genes using HTSeq (Anders et al., 2015). Reads with less than one count per million in at least three samples were filtered. Differential analysis was performed using the SARTools pipeline (version 1.74) pipeline (Varet et al., 2016) that evaluates differential mRNA levels using both EdgeR (Robinson et al., 2010) and DESeq2 (Love et al., 2014) algorithms. Transcripts with significantly differential expression (P<0.05, Padj<0.1) were considered for subsequent analysis. Gene set enrichment analysis (GSEA) was performed on both the EdgeR and DESeq datasets using EasyGSEA, which is part of the easy Visualization and Inference Toolbox for Transcriptome Analysis (eVITTA) toolbox (version 1.31) (Cheng et al., 2021). RNASeq datasets were deposited in the NCBI BioProject database (PRJNA807290).

Ten y¹, w*, UAS-Dcr; r4-GAL4 females were crossed to ten UAS-Pexi males and placed in a vial with fresh BDSC food and raised at constant 25°C. After 3–4 days, late third-instar larvae were isolated, rejecting larvae that had begun to pupariate (everted spiracles). Larvae were transferred to a 1 ml clear plastic cuvette containing 1 ml solution of 16% sucrose (w/v) dissolved in PBS (Reis et al., 2010). Images were captured of each replicate compared to a corresponding y¹, w*, UAS-Dcr; r4-GAL4 x w¹¹¹⁸ control. Using ImageJ, images of each replicate were analyzed by drawing a line from the base of the cuvette to the center of the liquid meniscus. The positions of larvae relative to the cuvette base were then measured and their vertical position expressed as a fraction relative to the distance to the meniscus. Combined measurements for three biological replicates of each genotype were analyzed using SuperPlotsofData (Goedhart, 2021). Statistical significance relative to the control for each genotype was measured by an unpaired Welch's t-test.

Third-instar larvae were rinsed in sterile PBS and homogenized in 5% NP-40 in distilled, deionized water. Samples were heated at 80°C for 5 min, cooled to room temperature and centrifuged to remove insoluble material. TG measurements were made using a Triglyceride Assay Kit (Abcam ab65336), measured on a BioTek Synergy 4 plate reader with Gen 5 software. TG measurements were normalized to the protein concentration of each sample, measured using a Pierce BCA Protein Assay Kit (Thermo Fisher 23225). Ten larvae for each trial were tested and reported as averages from three biological replicates. Statistical significance was measured by unpaired Student's t-test.

TG lipolysis was quantified using a Glycerol Assay Kit (Sigma-Aldrich MAK117). S2 cells were pelleted by centrifugation and the pellet diluted 1:1000 in water. End-point fluorescence was measured at 587 nm in a BioTek Synergy 4 plate reader with Gen 5 software. Glycerol measurements equalized relative to the protein content measured using a Pierce BCA Protein Assay Kit (Thermo Fisher 23225). For protein measurements, cells were lysed in mild lysis buffer (20 mM HEPES pH 7.0, 50 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM DTT, 1.0% Triton X-100, protease inhibitors), and protein measurements were taken. Colorimetric absorption was measured at 562 nm using a BioTek Synergy 4 plate reader with Gen 5 software. Statistical significance was determined using unpaired Student's t-test.

Second-instar larvae were collected from standard cornmeal food and transferred to lipid reduced holidic food made using the HUNTaa formulation with reduced levels (50%) of choline chloride, myo-inositol, cholesterol and sucrose (Piper et al., 2014), or holidic food supplemented with lard at 22.2 g/l (Woodcock et al., 2015). For each trial, 50 second-instar larvae from each genetic cross were transferred to holidic food or lard food. The values shown are averages from three independent genetic crosses.

LDs were isolated from transfected S2 cells, as described (Brasaemle and Wolins, 2016; Krahmer et al., 2011, 2013) with the following modifications. Parallel fractions were treated with 100 mN sodium carbonate, pH 11.5 (Ding et al., 2012). Cells were lysed using a 10 μm ball bearing (Isobiotec) cell homogenizer. The LD fraction was isolated, and the proteins precipitated by methanol: chloroform extraction (Wessel and Flügge, 1984). The dried protein pellet was resuspended in 30 μl gel sample buffer, boiled and size separated by SDS-PAGE. LD fractionation quality was monitored by western blotting with markers for mitochondria (CytC), ER (Cnx99A) and LDs/PLINs (Lsd-2) as well as the proportion of each protein in the LD fraction versus the pellet fraction.

Protein samples were separated by SDS-PAGE, transferred to nitrocellulose membrane (Bio-Rad 1620112), blocked in Odyssey Blocking Buffer (LI-COR) and incubated with primary antibody diluted in Blocking Buffer. Primary antibodies included guinea pig anti-Pex14 (1:2000), rabbit anti-Abcd3 primary antibody (1:1000) (Huang et al., 2020), rabbit anti Lsd-2 (1:200) (Welte et al., 2005), mouse anti-Cytochrome C (1:500, BD Pharmingen 7H8.2C12), mouse anti-Cnx99A (6-2-1, S. Munro, Developmental Studies Hybridoma Bank) and mouse anti-β-tubulin (E7, M. Klymkowsky, Developmental Studies Hybridoma Bank). For NRK cell lysates, membranes were probed with rabbit anti-PEX14 primary antibody (Thermo Fisher). Secondary antibodies used were AlexaFluor anti-rabbit A680 secondary antibody, AlexaFluor anti-mouse A790 secondary antibody, AlexaFluor anti-mouse A680 and AlexaFluor anti-guinea pig A790 (Abcam ab175773, ab186698 and Jackson ImmunoResearch 706-655-148, respectively). Membranes were visualized using an Odyssey Infrared Imaging System (LI-COR), and band intensity (relative to tubulin) was quantified using Odyssey software (LI-COR). Western blots were representative of three independent biological replicates.

The region corresponding to the ‘PEX14’ domain within Drosophila Pex14 was predicted using Interpro annotation database (Apweiler et al., 2000), the CCD with DeepCoil2 (Ludwiczak et al., 2019) and the TM domain with TMHMM 2.0 (Krogh et al., 2001). The helical wheel projection in Fig. S7 was generated using HeliQuest (Gautier et al., 2008).

Experiments were performed with assistance from the University of Alberta Faculty of Medicine and Dentistry Cell Imaging Core, Applied Genomics Core and Flow Cytometry Facility (RRID:SCR_019195, with financial support from the Faculty of Medicine and Dentistry, University Hospital Foundation and Canada Foundation for Innovation).

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259092.