Flexible origin of hydrocarbon/pheromone precursors in Drosophila melanogaster.

In terrestrial insects, cuticular hydrocarbons (CHCs) provide protection from desiccation. Specific CHCs can also act as pheromones, which are important for successful mating. Oenocytes are abdominal cells thought to act as specialized units for CHC biogenesis that consists of long-chain fatty acid (LCFA) synthesis, optional desaturation(s), elongation to very long-chain fatty acids (VLCFAs), and removal of the carboxyl group. By investigating CHC biogenesis in Drosophila melanogaster, we showed that VLCFA synthesis takes place only within the oenocytes. Conversely, several pathways, which may compensate for one another, can feed the oenocyte pool of LCFAs, suggesting that this step is a critical node for regulating CHC synthesis. Importantly, flies deficient in LCFA synthesis sacrificed their triacylglycerol stores while maintaining some CHC production. Moreover, pheromone production was lower in adult flies that emerged from larvae that were fed excess dietary lipids, and their mating success was lower. Further, we showed that pheromone production in the oenocytes depends on lipid metabolism in the fat tissue and that fatty acid transport protein, a bipartite acyl-CoA synthase (ACS)/FA transporter, likely acts through its ACS domain in the oenocyte pathway of CHC biogenesis. Our study highlights the importance of environmental and physiological inputs in regulating LCFA synthesis to eventually control sexual communication in a polyphagous animal.

(VLCFA) synthesis requires four distinct enzymes to further elongate a fatty acyl-CoA primer ( 11 ). A member of the elongase family, whose various gene products differ from one another in their tissue-specifi c expression and substrate specifi city, catalyzes the process of incorporating malonyl-CoA ( 12 ). The subsequent steps are successively catalyzed by a 3-keto-acyl-CoA-reductase (KAR), a 3-hydroxy-acyl-CoAdehydratase (HADC), and a trans -enoyl-CoA-reductase (TER). Desaturating enzymes can operate between elongation steps, leading to unsaturated FAs ( 11 ).
Genetic studies using the D. melanogaster model have identifi ed several enzymes that are required for CHC biogenesis in specifi c abdominal cells called oenocytes ( 8,(13)(14)(15)(16)(17)(18). More recently, it has been shown that FASN CG3524 , one of the three Drosophila FASN orthologs, is specifi cally expressed in the oenocytes for the synthesis of methylated/ branched (mb) FAs, which are precursors of 2-methylalkanes (mbCHCs) ( 19 ). This study also revealed that FASN CG17374 is expressed in the oenocytes, whereas FASN CG3523 , which was expected to be ubiquitously expressed ( 20 ), is excluded from the oenocytes but is present in the fat body (FB; the organ in charge of hepatic and adipose functions) ( 19 ). Together, these studies support the notion that the entire metabolic pathway sustaining CHC production takes place within the oenocytes.
To comprehensively address this issue, we used D. melanogaster genetics and performed systematic knockdown in the oenocytes of several enzymes that cover the entire process of FA synthesis. We showed that LCFA elongation to VLCFA takes place within the oenocytes. In contrast, LCFA synthesis-defi cient fl ies still produce CHCs, though in reduced amounts, indicating that dietary lipids may partly compensate for LCFA defi ciency. Further, we observed that providing fat-enriched food during larval development impedes pheromone biosynthesis in adult fl ies and decreases their mating success. Finally, we showed that FA homeostasis may infl uence CHC biogenesis and identifi ed a putative acyl-CoA synthase required for CHC synthesis within the oenocytes.
supplementary Table 2C, D), possibly because of variable effi cacy in RNAi knockdown. RNAi to HADC CG6746 led to a strong decrease in CHC levels ( Fig. 1G ; supplementary  Table 2E, F), whereas RNAi to HADC CG9267 led to a moderate decrease in CHC levels ( Fig. 1H ; supplementary Table  2E, F). However, coexpression of both RNAis decreased CHC levels even more ( Fig. 1I ; supplementary Table 2E, F), suggesting that both HADCs contribute to this enzymatic activity. Together, these fi ndings indicate that VLCFA synthesis of CHC precursors takes place exclusively within the oenocytes.

Functional roles of the oenocyte-specifi c FASN genes
To evaluate the role of LCFA synthesis in the oenocytes, we induced RNAi to FASN CG3524 and FASN CG17374 , the two FASN genes reported to be expressed in the oenocytes ( 19 ). Consistent with the study of Chung and colleagues ( 19 ), we observed a dramatic reduction in mbCHCs in Oe > FASN CG3524 -RNAi fl ies, although the total amount of CHCs did not change ( Fig. 2A  FASN CG3524 has been proposed to act in maintaining ecological isolation between two Drosophila species through desiccation resistance and effects on mating choice ( 19 ). Therefore, we investigated these functions in D. melanogaster . Oe>ACC-RNAi fl ies were extremely sensitive to desiccation ( Fig. 2G ), a phenotype suppressed when coexpressing the svp-gal80 transgene (supplementary Fig. 1A). Further, Oe>FASN CG3524 -RNAi fl ies were not sensitive to desiccation, whereas Oe>FASN CG17374 -RNAi fl ies were moderately sensitive ( Fig. 2G ; supplementary Table 4). The desiccation sensitivity of Oe>FASN CG17374 -RNAi fl ies suggests that FASN CG17374 catalyzes the synthesis of precursors for other FA derivatives, potentially waxes or TAGs ( 5 ), that are required to secure cuticular watertightness. The desiccation sensitivity of Oe>ACC-RNAi fl ies may result from a default in CHCs and/or in these non-CHC lipid derivatives. Nonetheless, oenocyte expression of an RNAi to Cyp4g1 , which specifi cally catalyzes decarbonylation to CHC ( 8 ), resulted in desiccation sensitivity similar to that of Oe>ACC-RNAi fl ies (supplementary Fig. 1B, B'). These fi ndings indicate that, although non-CHC lipid derivatives appear to be required to produce an effi cient cuticle, CHC depletion is suffi cient to fully impair cuticular watertightness.
Next, we investigated whether FASN CG3524 or FASN CG17374 knockdown affects mating choice in D. melanogaster . Single wild-type (Canton-S) females did not exhibit any preference when given a choice between Canton-S males and one of either genotype Oe>FASN CG3524 -RNAi or Oe>FASN CG17374 -RNAi ( Fig. 2H ). Reciprocally, single Canton-S males did not exhibit any preference when given a choice between a Canton-S female and a female of either genotype Oe>FASN CG3524 -RNAi or Oe>FASN CG17374 -RNAi ( Fig. 2H ). Although we cannot anticipate the consequence of total mbCHC depletion, our fi ndings indicate that a 50% reduction in mbCHCs does not affect desiccation resistance or sexual recognition in D. melanogaster .
In their progeny, RNAi-expressing fl ies contain the gal4 driver, whereas control fl ies contain the balancer chromosome.

Biochemical analysis
The C23-C29 CHCs, which are synthesized after adult eclosion ( 7 ), were extracted from 4-to 5-day-old fl ies and analyzed by gas chromatography as previously described ( 16 ). Control and test fl ies were issued either from the sibling progeny of the same crosses (RNAi lines and LpR mutants) or from fl ies reared at the same time in the same conditions (nutrition test on control and FASN mutants). At least 10 fl ies were analyzed for each genotype. TAG measurements and quantitative RT-PCR were performed from four replicates of three adult males, as previously described ( 21 ).

Histochemistry
The dorsal parts of abdominal cuticles were dissected from 4to 5-day-old fl ies in PBS, fi xed for 20 min at room temperature in 4% paraformaldehyde, and then washed three times in PBS. Lipid (Nile Red) and nuclear (TO-PRO-3-iodide) staining was performed as previously described ( 23 ). Cuticles were mounted in DABCO and examined using a Nikon (TE-2000-U) confocal microscope.

Fly behavior
Desiccation tests and mate choice tests were performed as previously described ( 8,28 ).

Elongation of LCFA to VLCFA takes place within the oenocytes
To investigate the CHC biosynthetic pathway in D. melanogaster , we fi rst evaluated the ACC requirement. The 1407-gal4 driver that expresses Gal4 in the oenocytes from the late third larval stage ( Fig. 1B ) was used to direct an RNAi to the unique ACC Drosophila gene (hereafter called Oe>ACC-RNAi ). In these fl ies, CHCs were almost fully depleted ( Fig. 1C ; supplementary Table 2A, B). This defect was a direct consequence of ACC knockdown in the oenocytes: expression of svp-gal80 ( 25 ), which specifi cally blocks Gal4 activity in the oenocytes ( Fig. 1B, B' ), completely rescued CHC production ( Fig. 1D ; supplementary Table 2A, B).
Next, we evaluated the requirement for enzymes that catalyze VLCFA synthesis. The D. melanogaster genome encodes 20 elongase members ( 29 ). Stringent in silico analyses indicate that KAR is encoded by seven putative genes, TER by a single gene ( TER CG10849 ), and HADC by two putative genes ( HADC CG6746 and HADC CG9267 ) (supplementary Table 1). We focused on the strongest KAR homolog ( KAR CG1444 ) because we previously reported that it was required early in development for an essential function of larval oenocytes ( 23 ). In Oe>KAR CG1444 -RNAi animals, CHCs were almost completely eliminated ( Fig. 1E ; supplementary Table 2C, D), indicating that knockdown of only the KAR CG1444 ortholog is suffi cient to largely eliminate CHC synthesis in D. melanogaster oenocytes. RNAi to TER CG10849 led to a partial decrease in CHC levels ( Fig. 1F ; that FASN CG17374 can also contribute to the pool of LCFAs used for CHC biogenesis. To evaluate how these FASN mutants consume lipid stores, we analyzed TAG levels either in 0-to 1-day-old or in 4-to 5-day-old adult fl ies. In contrast to CHCs ( Fig. 3A, A' ), TAG levels in 0-to 1-day-old fl ies were slightly higher in control animals fed a lipid-supplemented diet ( Fig. 3B ). Consistent with the higher fasting resistance of newly emerged versus 3-day-old adult fl ies ( 30 ), TAG stores were at very high levels at the day of eclosion and then decreased in 4-to 5-day-old males, irrespective of larval lipid supplementation ( Fig. 3B ). Conversely, in FASN ⌬ 24-23 and FASN ⌬ 24-23;CG17374i mutants, TAG stores were dramatically lower at eclosion and almost eliminated 4 days later ( Fig. 3B ).
Finally, we analyzed FASN expression in fl ies that emerged from larvae fed a lipid-supplemented diet. Quantitative RT-PCR analysis indicated that feeding control larvae with the lipid-supplemented media led to a signifi cant decrease in expression of the three FASN genes in 1-day-old adult males ( Fig. 3C-E ). Moreover, we observed that, in the presence of single Canton-S females, males that emerged from larvae fed a lipid-supplemented media had lower mating success than those raised on standard media ( Fig. 3F ). Furthermore, females that emerged from larvae fed a lipidsupplemented media were less attractive than those raised

FASN -defi cient fl ies produce few CHCs
To get further insights into the requirement of FA synthesis in CHC biogenesis, we took advantage of a mutant ( FASN ⌬ 24-23 ) that removes both FASN CG3524 and FASN CG3523 genes. This mutant is lethal at the L1 stage but can be rescued by an appropriate lipid-supplemented diet ( 21 ). We also induced FASN CG17374 -RNAi with the 1407-gal4 driver in FASN ⌬ 24-23 mutants ( FASN ⌬ 24-23;CG17374i ). We focused on mutant males because they survived better than females after eclosion. Importantly, both control and mutant animals were raised on the lipid-supplemented media during larval development. However, newly emerged fl ies were transferred to standard food for 4 days because adult fl ies tend to stick to the lipid-supplemented media, leading to a high rate of lethality. Surprisingly, control males emerging from larvae fed a lipid-supplemented diet contained roughly half the amount of all CHCs, including mbCHCs, compared with control males raised on standard media ( Fig. 3A , A' ; supplementary Table 5A). Nonetheless, when emerged from larvae fed a lipid-supplemented diet, FASN ⌬ 24-23 mutant males contained a similar amount of CHCs compared with control males raised in the same conditions ( Fig. 3A, A' ; supplementary Table 5B). Further, CHCs were strongly reduced, but not completely eliminated, in FASN ⌬ 24-23;CG17374i males ( Fig. 3A, A' ; Table 6G, H), suggesting that FA synthesis within the FB may participate in feeding the pool of LCFAs used for CHC biogenesis in the oenocytes.

A bipartite FA transporter/acyl-CoA synthase is required in the oenocytes for CHC biogenesis
Based on our previous study, which showed that a default in VLCFA synthesis in larval oenocytes results in tracheal defects ( 23 ), we screened RNAi to 140 genes encoding FA metabolic effectors (supplementary Table 7), using the BO-gal4 driver that is active in embryonic and larval oenocytes ( 25 ). In this way, we found that fatty acid transport protein (FATP) was essential in larval oenocytes (supplementary Table 7). Further, CHCs were almost fully depleted in Oe>fatp-RNAi fl ies ( Fig. 5A ; supplementary  Table 8A, B) indicating that FATP is also required in adult oenocytes.
In addition to its FA-transporter domain, FATP also contains an acyl-CoA synthase (ACS) domain. However, FATP is unlikely to work through its FA-transporter domain, at least for the production of mbCHCs that are depleted in Oe>fatp-RNAi (supplementary Table 8A, B); some of their precursors (mbFAs) are synthesized within the oenocytes and therefore do not require transporter-mediated uptake. Finally, we investigated potential cytological defects in the oenocytes of CHC-depleted fl ies. In the abdomen of adult fl ies, oenocytes and the FB are organized as tightly associated rows of cells that can be easily distinguished ( Fig. 5B-D ). Analysis of lipid content revealed that the FB has a large capacity to store lipid droplets (LDs) ( Fig. 5B ), whereas no LDs could be detected in the oenocytes ( Fig. 5C, D ). Conversely, oenocytes of Oe>fatp-RNAi fl ies contained more LDs than did control oenocytes on standard media ( Fig. 3F ). Consistently, several femalespecifi c pheromones were reduced in the females fed on lipid-supplemented media, while total CHCs were not signifi cantly affected (supplementary Table 5C; Fig. 3G ). Considering that CHC biogenesis occurs after eclosion ( 7 ), these fi ndings indicate that an excess of dietary lipids during the juvenile period restrains pheromone biogenesis in adults and reduces their mating success.

FA metabolism in the FB affects CHC biogenesis
The observations described previously suggest that the LCFAs used for CHC synthesis may originate somewhere outside the oenocytes and, thus, must be taken up into the oenocytes before CHC biogenesis. The lipoprotein receptors LpR1 and LpR2 are expressed in the oenocytes ( 25 ). We previously showed that LpR2 was required for lipid uptake into ACC-defi cient oenocytes in larvae ( 23 ). Therefore, here we monitored CHC amounts in mutants in which LpR1 , LpR2 , or both genes together had been deleted ( Fig. 4A ). CHC amounts were not affected in mutants with either LpR1 or LpR2 deletions ( Fig. 4B, C ; supplementary Table 6A, B) but were severely decreased in the mutants with a double LpR1/LpR2 deletion ( Fig. 4D ; supplementary Table 6A, B). Consistent with this, fl ies expressing the ubiquitous da-gal4 driver to direct LpR1-RNAi and LpR2-RNAi together had far fewer CHCs ( Fig. 4E ; supplementary Table 6C, E). However, CHCs remained unaffected when both RNAi's were directed by either an oenocyte-or a gut-specifi c driver ( Fig. 4F, G ; supplementary Table 6C-F). Conversely, we observed a signifi cant reduction in CHCs when directing LpR1-RNAi and LpR2-RNAi together with C564-gal4 , an FB-specifi c driver ( Fig. 4H ; supplementary Table 6D, F), indicating that alteration of lipid metabolism in the FB affects oenocyte activity. Therefore, to determine whether FA synthesis in the FB plays a role in CHC biogenesis, we used the C564-gal4 driver to knockdown lipid-enriched food to the larvae of the cactophilic Drosophila mojavensis decreases CHC production in adults ( 31 ). Here, we show that this is also the case in the polyphagous D. melanogaster and that the decrease in CHCs impacts sexual communication in adults. Because CHCs are synthesized after adult eclosion ( 7 ), it is tempting to speculate that a larval nutritional signal modulates the competence for CHC biogenesis in adult oenocytes. Alternatively, reduced CHCs induced when dietary lipids are provided during the larval stages may directly depend on repression from circulating lipids that could remain higher in the resulting adult fl ies. This repression operates at least in part on FASN gene transcription, including FASN CG3523 that is not active in the oenocytes. However, further experiments will be required to determine whether this repression results from a developmental event that is induced early or from a direct effect due to higher lipid content.
In an attempt to identify novel genes critical for oenocyte function, we have identifi ed fatp as an essential gene in larval oenocytes and further showed that fatp is required for CHC synthesis in adult oenocytes. Previous Drosophila studies on fatp mutants reported metabolic and eye-specifi c phenotypes ( 32,33 ) but did not discriminate between FAtransporter or ACS activities. Importantly, our observations favor the notion that in the process of CHC synthesis, FATP acts instead through its ACS domain, because fatp knockdown impedes mbCHC synthesis in the oenocytes, and mbCHC precursors are mainly synthetized within the ( Fig. 5E ). Further, we observed a dramatic accumulation of LDs in ACC -defi cient oenocytes ( Fig. 5F ) and to a lesser extent in KAR CG1444 -defi cient oenocytes ( Fig. 5G ). Importantly, the dramatic accumulation of LDs observed in oenocytes of Oe>ACC-RNAi fl ies was still visible when coexpressing fatp-RNAi ( Fig. 5H ). Accumulation of LDs may be due to either an increase in lipid uptake or a decrease in CHC synthesis and a subsequent accumulation of precursors. Nonetheless, in ACC -defi cient oenocytes, there is no malonyl-CoA and therefore no FA synthesis. Thus, the accumulation of LDs must be due to an increase in lipid uptake that does not depend on FATP.

DISCUSSION
Previous studies have suggested that the entire CHC/ pheromone biogenesis pathway takes place within the oenocytes, because targeted knockdown of enzymes acting either in the early or fi nal steps of this metabolic pathway led to depleted CHCs ( 8,13,19 ). Here we provide evidence that VLCFA synthesis in D. melanogaster happens exclusively within the oenocytes, while there is fl exibility in where the LCFAs used to feed this metabolic pathway originate ( Fig. 6 ).
Lipid homeostasis appears to infl uence CHC production because FB disruption of genes encoding lipid metabolic effectors (FASN, LpR1, and LpR2) decreases the amount of CHCs. It has been previously reported that providing Further, in the mouse, skin-targeted knockout of FATP4 provokes watertightness defects that can be rescued by transgenic overexpression of wild-type FATP4, but not of a variant containing two point mutations in its ACS domain ( 35 ). Therefore, because any FA modifi cation fi rst requires CoA esterifi cation ( 8 ), the phenotype induced by fatp knockdown suggests that FATP is required within the oenocytes through its ACS domain, rather than its FA-transporter domain ( Fig. 6 ). In summary, our fi ndings support the notion that lipid uptake into the oenocytes does not involve LpR1, LpR2, or FATP. Therefore, identifi cation of other candidates must be undertaken to determine whether oenocyte lipid uptake proceeds through lipoprotein particles or free FAs.
Our fi ndings highlight the existence of compensatory processes that regulate the production of CHCs. In Drosophila serrata , the synthesis of mbLCFAs used as mbCHC precursors takes place solely within the oenocytes ( 19 ). Here, we show that in D. melanogaster , oenocyte knockdown of FASN CG3524 also results in a strong reduction of mbCHCs. However, FASN ⌬ 24-23 mutants that do not have the FASN CG3524 gene still produce mbCHCs, demonstrating that the mbLCFAs used for mbCHC synthesis can be either synthesized within the oenocytes or provided by dietary lipids. Compensatory processes are also highlighted by oenocyte disruption of the FASN CG17374 gene, which does not affect CHC amounts unless it is produced in an FASN ⌬ 24-23 mutant. This suggests either that oenocyte knockdown of FASN CG17374 is compensated by increased lipid uptake into the oenocytes or that FASN CG17374 is recruited within the oenocytes only when FA synthesis is repressed in the entire animal. In summary, our fi ndings reveal that CHC production integrates various physiological inputs and suggest that regulation of CHC biogenesis operates at the level of LCFA rather than VLCFA synthesis. This regulatory loop integrates nutrition and FB metabolism at the organismal level to guarantee the production of a functional integument. In this process, the FB, which oenocytes and therefore do not require FA-mediated intake. Nonetheless, we cannot exclude that the FA-transporter domain of FATP is required for CHC biogenesis. However, the increase in LD content induced by ACC knockdown in the oenocytes must be due to lipid intake that does not require FATP. Moreover, as for defaults in VLCFA synthesis, oenocyte knockdown of fatp results in almost total CHC depletion, a phenotype that is never observed when LCFA synthesis is impaired, even in a severe FASN mutant combination. Consistent with this, biochemical analysis on the strongest mammalian homolog, FATP4, revealed that it acts as an ACS on either LCFA or VLCFA substrates ( 34 ).   6. CHC biogenesis integrates physiological inputs. Bold arrows are enzymatic processes already known or strongly suspected to sustain CHC synthesis in oenocytes. Dotted arrows are enzymatic or regulatory processes highlighted in our study: synthesis of non-CHC compounds required for cuticle watertightness; malonyl-CoA inhibition of lipid uptake; FB lipid metabolism modulating CHC synthesis; role of dietary lipids in feeding the pool of CHC precursors and repressing FASN gene expression; and FATP potentially acting as an ACS. Note that lipid intake within the oenocytes may proceed through lipoprotein particles or free FAs. has been previously shown to control body homeostasis ( 25,36 ), may provide LCFAs to the oenocytes and/or control oenocyte activity.