PE homeostasis rebalanced through mitochondria-ER lipid exchange prevents retinal degeneration in Drosophila

The major glycerophospholipid phosphatidylethanolamine (PE) in the nervous system is essential for neural development and function. There are two major PE synthesis pathways, the CDP-ethanolamine pathway in the endoplasmic reticulum (ER) and the phosphatidylserine decarboxylase (PSD) pathway in mitochondria. However, the role played by mitochondrial PE synthesis in maintaining cellular PE homeostasis is unknown. Here, we show that Drosophila pect (phosphoethanolamine cytidylyltransferase) mutants lacking the CDP-ethanolamine pathway, exhibited alterations in phospholipid composition, defective phototransduction, and retinal degeneration. Induction of the PSD pathway fully restored levels and composition of cellular PE, thus rescued the retinal degeneration and defective visual responses in pect mutants. Disrupting lipid exchange between mitochondria and ER blocked the ability of PSD to rescue pect mutant phenotypes. These findings provide direct evidence that the synthesis of PE in mitochondria contributes to cellular PE homeostasis, and suggest the induction of mitochondrial PE synthesis as a promising therapeutic approach for disorders associated with PE deficiency.

a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 cellular PE levels is critical for neuronal function and integrity, suggests that PE synthesized in the mitochondria can be exported to the ER to compensate for deficiencies in cellular PE in neurons, and highlights a fundamental function of mitochondria in cellular phospholipid homeostasis.

A forward genetic screen reveals that pect is required for photoreceptor survival
To identify genes required for photoreceptor survival, we performed EMS mutagenesis and screened chromosomes 2 and 3 using the "Rh1::GFP ey-flp/hid" system, which generates homozygous mutant eyes and labels photoreceptors with GFP-tagged Rh1 to track cellular integrity [27,28]. Among the~50 mutants that exhibited low GFP fluorescence on day 5 but not on day 1, two were lethal mutations in the gene pect, which encodes phosphoethanolamine cytidylyltransferase (PECT), a rate-limiting enzyme in the CDP-ethanolamine pathway ( Fig  1A) [29,30]. Both pect 29 and pect 102 alleles contained a single nucleotide change that affected conserved enzymatic domains (Fig 1B and S1A Fig). Levels of PECT protein were also greatly reduced in pect 29 flies (Fig 1C).
We next determined the subcellular localization of PECT by isolating fly heads and separating the cytoplasmic and membrane fractions via subcellular fractionation. PECT immunoreactivity was detected in the cytosol (S2A Fig). We next attempted to use immunohistochemistry to visualize the spatial distribution of PECT within photoreceptors, but the PECT antibody did not work in immunostaining. We therefore expressed Flag-tagged PECT in photoreceptor cells using a promoter associated with the major rhodopsin, ninaE (neither inactivation nor afterpotential E). PECT protein is largely colocalized with the ER marker calnexin (CNX) [31] (S2B Fig). These data indicate that PECT is a cytosolic protein that localizes to the ER. We then purified PECT recombinant protein and measured its enzymatic activity in vitro. CDPethanolamine was detected by the mass spectrum after incubating phosphoethanolamine and CTP with PECT, confirming that Drosophila PECT is a bona fide phosphoethanolamine cytidylyltransferase (S2C and S2D Fig).

PECT is required for maintaining the visual response
We next asked whether major proteins involved in the phototransduction process were correctly localized in pect mutants. Rh1, the major rhodopsin in Drosophila, and TRP, the major downstream Ca 2+ /cation channel, localize exclusively to the rhabdomere, tightly-packed microvilli that are critical for phototransduction. Localization patterns for these proteins were the same in wild-type and pect 29 flies, indicating that loss of PECT did not affect the formation of the rhabdomere (Fig 1D).
To analyze the visual response of pect 29 flies, we performed electroretinogram (ERG) recordings, which measure the summed response of all retinal cells. There are two primary features of the ERG, a sustained corneal negative response, as well as on-and off-transients, which reflect postsynaptic activity in the lamina. In wild-type flies, light induces a rapid corneal negative potential, which quickly returns to baseline after cessation of the 20-s light stimulation. However, light-evoked photoreceptor potential persisted after the 20-s light stimulation, and the ERG transients were diminished in pect 29 flies. Expressing wild-type PECT in the compound eyes of pect 29 flies via the GMR (glass multiple response) promoter completely rescued both these ERG phenotypes. Moreover, expressing the mammalian homolog of PECT, PCYT2, in pect 29 mutant flies also rescued ERG defects (Fig 1E and 1F). These data indicate that PECT activity is required for termination of the visual response.

Mutations in pect lead to light-independent retinal degeneration
We next measured the integrity of photoreceptor cells in pect 29 flies using transmission electron microscopy (TEM). In wild-type flies, seven rhabdomeres were detected regardless of fly  29 and pect 102 mutations via a forward genetic screen. Rh1-GFP fluorescence was detected in the deep pseudopupil (upper panels) and by cornea optical neutralization (lower panel). Images from 1-day-old and 5-day-old wt (wild type, ey-flp rh1-GFP;FRT40A/GMR-hid CL FRT40A), pect 29 (ey-flp rh1-GFP; pect 29 FRT40A/GMR-hid CL FRT40A) and pect 102 (ey-flp rh1-GFP;pect 102 FRT40A/GMR-hid CL FRT40A) flies are shown. Scale bars are 100 μm or 10 μm in upper and lower panels, respectively. (B) The pect locus and mutations associated with the pect 29 and pect 102 alleles. (C) Western blot analysis of retinas dissected from 1-d-old wild-type, pect 29 , and GMR-pect flies that were labeled with antibodies against INAD and PECT. (D) Tangential resin-embedded retina sections of compound eyes from~1-day-old wt, pect 29 flies were labeled using antibodies against Rh1 (green) and TRP (red). Scale bar is 2 μm. (E) ERG recordings from 1-day-old wt, pect 29 , pect 29 ;GMR-pect (ey-flp rh1-GFP;pect 29 FRT40A/GMR-hid CL FRT40A;GMR-pect/+), and pect 29 ;GMR-PCYT2 (ey-flp rh1-GFP;pect 29 FRT40A/GMR-hid CL FRT40A;GMR-PCYT2/+) flies. Flies were dark adapted for 2 min and subsequently exposed to a 1-s pulse followed by a 20-s then a 1-s age and light condition (Fig 2A). However, pect 29 flies displayed a gradual loss of photoreceptor cells and rhabdomeres. At 10 days of age, pect 29 photoreceptor cells severely degenerated and there was a complete loss of rhabdomeres in R1-R6 cells. This was seen when flies were reared in a constant dark, or a light/dark cycle (Fig 2B). This cell death phenotype was rescued by expressing pect, or its mammalian homolog PCYT2, via the GMR or ninaE promoters ( Fig  2C and 2D). This also indicates that mutations in pect do not cause developmental defects as ninaE promoter begins to drive gene expression at the late pupal stage. These results demonstrate that PECT activity is required for photoreceptor cell survival.

Defects in visual response and loss of photoreceptors seen in pect mutants do not result from alternations in DAG and PI metabolism
It is well established that hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) to generate inositol-1,4,5-triphosphate (IP3) and diacylglycerol (DAG) by PLC, which is encoded by the norpA (no receptor potential A) gene, activates the light-sensitive TRP and TRPL channels in Drosophila [24]. Further, DAG or its metabolites may play an excitatory role in phototransduction [32][33][34]. Therefore, DAG must be rapidly degraded to maintain the resting potential of photoreceptor neurons (Fig 3A). DAG levels were increased by 14% in pect 29 mutant retinas, leading us to hypothesize that this excess DAG may sustain active phototransduction in pect mutant animals ( Fig 3B and S1 Data). We tested this idea by genetically reducing DAG levels in pect 29 flies.
We first overexpressed the DAG kinase, RDGA (Retinal Degeneration A) to promote phosphorylation of DAG to phosphatidic acid (PA) [35]. This did not suppress the ERG deficits or photoreceptor neurodegeneration seen in pect 29 mutants (Fig 3C and 3D and S3A Fig). We next reduced the generation of DAG from PA by introducing mutations in the gene lazaro (laza), which encodes a photoreceptor PA phosphatase [36,37]. ERG responses and retinal degeneration phenotypes were indistinguishable between pect 29 ;laza 1 and pect 29 flies (Fig 3C  and 3D and S3A Fig). Finally, we examined norpA P24 ;pect 29 double mutants, and found the same level of retinal degeneration seen in pect 29 mutants ( Fig 3E).
Total PI levels were increased by 41% in pect 29 retina compared with wild-type (Fig 3B and S1 Data). Since PIP 2 must be rapidly replenished to maintain a sustained light response, we tested whether reducing PI level could suppress the degeneration phenotype of pect 29 flies. PI synthase (PIS) catalyzes the final step of the PI regeneration pathway. Knocking down PIS levels via RNAi (pis RNAi ) did not affect the severity of retinal degeneration in pect 29 mutants ( Fig  3F and S8A Fig). We also generated pect 29 ;trp 343 and inaF;pect 29 double mutants, thereby inactivating the major light-sensitive TRP channels in photoreceptors. Both of these double mutants exhibited levels of retinal degeneration seen in pect 29 flies, suggesting that TRP channel activity does not cause retinal degeneration in pect mutants (Fig 3G). In summary, these results demonstrate that the accumulation of DAG and PI did not cause ERG deficits or retinal degeneration in pect mutant flies.

Increasing PECT-independent PE synthesis suppresses retinal degeneration in pect 29 flies
Two major pathways synthesize phosphatidylethanolamine (PE) in the cell, the PECT-dependent CDP-ethanolamine pathway, or through phosphatidylserine (PS) decarboxylation pulse of orange light. On-and off-transients are indicated. (F) ERG amplitudes at the second 1-s pulse are normalized to ERG amplitudes at the first 1-s pulse. Error bars represent SD, and significant differences were determined using the unpaired Student's t-test (n = 6, ��� p < 0.001).

PLOS GENETICS
We next genetically manipulated PC and PE levels and investigated the contribution of lipid composition to the mutant phenotypes. We used a heterozygous null allele of Pcyt1 179 to decrease PC synthesis in pect 29 flies, but this did not suppress the retinal degeneration (S3B and S8B Figs). To examine whether overexpressing PC synthesis enzymes could enhance retinal degeneration, we expressed two enzymes, bbc (bb in a boxcar) and Pcyt1 (Phosphocholine cytidylyltransferase 1), via the GMR promoter. Overexpressing PC synthesis enzymes did not further enhance retinal degeneration in pect 29 flies (S3C Fig).
Strikingly, when we expressed PSD to generate PE from PS in mitochondria, the retinal degeneration seen in pect 29 flies was suppressed (Fig 4C and 4D). In contrast, knocking down Psd in photoreceptor cells via three different Psd RNAi lines enhanced the pect 29 degeneration phenotype. Expressing these Psd RNAi lines alone did not lead to photoreceptor cell death ( Fig  4E, S4A and S8C Figs). Moreover, eye-specific knockdown of PS synthase (pss) in pect 29 ; GMR-Psd flies led to levels of retinal degeneration that were even more severe than those seen in pect 29 flies. Retinal morphology was normal in pss RNAi flies ( Fig 4F and S8D Fig). Since PSS is the only PS synthase in Drosophila, it is speculated that both PC and PE are the substrates of PSS. According to our results, PC may be the main substrates of PSS at least in the absence of PECT. Thus pss knockdown would reduce PS levels, and subsequently reduce the PE levels generated through the PSD pathway. We next asked whether PSD expression could restore visual responses to normal in pect 29 flies. Expressing PSD had little effect on ERG responses in 1-day-old pect 29 flies. In 9-day-old pect 29 ;GMR-Psd flies, however, the activation and prolonged after-potential ERG responses seen in pect 29 flies were rescued; ERG transients were partially rescued (Fig 4G and 4H and S4B Fig). This delayed effect on ERG response may be explained by the low levels of PE synthesized by the PSD pathway, as the exchange of lipids between the ER and mitochondria is a slow process. Finally, the lipidomic analysis revealed that expressing PSD fully restored PE levels in pect 29 cells, both on days 1 and 9 ( Fig 4I and S2 Data). In contrast, PSD induction further decreased PS levels by 56% in pect 29 ;GMR-Psd flies compared with wild type (Fig 4J and S2 Data). We next examined relative levels of specific PE species and found that levels of PE38:1, PE36:4, and PE36:5 were decreased by 54%, 49%, and 39%, respectively. These levels were fully reversed by PSD overexpression (S5A- S5C Fig and  S2 Data). These data indicate that PSD overexpression can restore levels of both total PE and specific PE species in pect mutant cells, and further demonstrate that reduced levels of PE caused retinal degeneration and ERG defects in pect 29 flies.

Mitochondria distribution and activity are normal in pect 29 retinas
We next sought to determine if PSD localizes exclusively to mitochondria. In S2 cells, PSD-GFP colocalized with TOM20 ( Fig 5A). We next generated a transgenic fly that expressed PSD-GFP via the ninaE promoter and found that PSD-GFP colocalized with the mitochondrial marker COX4, but not with the ER marker calnexin (CNX) (Fig 5B). Moreover, western bot analysis of ER and mitochondria isolated using a sucrose gradient revealed that PSD expressed in the eye localized exclusively to mitochondria ( Fig 5C). As a mitochondrial protein, PSD may directly modulate mitochondrial PE concentrations. Thus, mitochondrial integrity, which is important for cell survival and for maintaining visual responses, may be disrupted in pect 29 mutants [38]. To visualize mitochondria we expressed mitoGFP via the GMR promoter in pect 29 eyes. The pattern of mitochondrial localization and GFP signal was similar between pect 29 and control flies ( Fig 5D). We next examined the morphology of the lamina cartridge, the organized synaptic modules where R1-R6 photoreceptors project axons to, in pect 29 flies. The structure of the lamina cartridge was intact and mitochondria were normal in pect 29 flies, although synaptic vesicles were disrupted ( Fig 5E) [25]. We compared ATP levels between wild-type and pect 29 retinas to assess mitochondrial function and found similar ATP levels in pect 29 mutants and controls. In contrast, ATP levels in pink1 B9 mutant, a previously characterized mutant which induces mitochondrial impairment [39], were significantly reduced ( Fig 5F). We also assessed levels of reactive oxygen species (ROS) by measuring hydrogen peroxide levels but found no differences between pect 29 flies and wild type. In contrast, hydrogen peroxide levels were dramatically increased in pink1 B9 flies ( Fig  5G). Furthermore, we measured mitochondrial transmembrane potential by fluorescent dye TMRM and found no differences between wild-type and pect 29 mutant cells (S6A and S6B Fig). Finally, we performed a phospholipid mass spectrometric analysis of isolated mitochondria. Although levels of PE in mitochondria were reduced in pect 29 mutant retina, expressing PSD did not restore mitochondrial PE levels in pect 29 flies. This was true on day 1 and day 9 ( Fig 5H and S3 Data). This may be because PE export from mitochondria is stimulated when substrates for PE production via the CDP-ethanolamine pathway are lacking [14]. Indeed, PSD expression further decreased PS levels. Together, loss of PECT did not affect mitochondrial function, and the disruption of lipid homeostasis in mitochondria did not play a physiological role in pect 29 mutant photoreceptor cells.
https://doi.org/10.1371/journal.pgen.1009070.g004 Cryostat sections from ninaE-Psd-GFP flies were labeled with antibodies against Calnexin (blue), COX4 (red), and GFP (green). Scale bar specific knockdown of proteins important for maintaining ERMCS would disrupt the trafficking of PE from mitochondria to the ER. To examine ERMCS in vivo, we labeled ER and mitochondria by expressing KDEL-GFP and Tom70-RFP via the photoreceptor cell-specific ninaE and trp promoters, respectively. We then knocked down mfn (mitofusin), serca (sarco/endoplasmic reticulum Ca 2+ -ATPase), miro (mitochondrial Rho), porin, and ip3r (inositol 1,4,5-triphosphate receptor), which regulate ERMCS dynamics [41][42][43][44], and quantified ERMCS in photoreceptor cells. There were fewer ERMCS in mfn RNAi and serca RNAi ommatidia, whereas miro RNAi , porin RNAi , and ip3r RNAi were not affected. This suggests that the mitochondrial protein MFN, and the ER protein SERCA, are key regulators of ER-mitochondria contacts in photoreceptor cells (Fig 6A and 6B, S7A and S8E-S8I Figs). We next expressed mfn RNAi , serca RNAi , miro RNAi , porin RNAi , and ip3r RNAi via GMR-Gal4 in pect 29 ;GMR-Psd photoreceptor cells. In agreement with the ERMCS results, both mfn RNAi and serca RNAi prevented PSD overexpression from rescuing the retinal degeneration seen in pect mutants, miro RNAi , porin RNAi , and ip3r RNAi had no effect (Fig 6C and 6D). In contrast, knocking down mfn, serca, miro, porin, or ip3r alone did not affect photoreceptor cells. SERCA is an intracellular calcium pump that transports Ca 2+ ions from the cytoplasm to the ER lumen and plays an important role in maintaining ER calcium stores. Reduced luminal Ca 2+ concentration is known to trigger the unfolded protein response (UPR) [45]. To investigate the specific role of SERCA in PE exchange between ER and mitochondria, we performed phospholipid mass spectrum analysis. PSD overexpression increased PE levels and reduced PC levels in pect 29 flies. Expressing serca RNAi significantly decreased PE levels and increased PC levels of pect 29 ;GMR-Psd flies, while the PE and PC levels are not changed upon expression of serca RNAi alone (S7B and S7C Fig and S4 Data). These results suggest that ER-mitochondria contacts are required for the export of PE from mitochondria to ER to maintain cellular PE levels.

Discussion
In a forward genetic screen to identify genes necessary for photoreceptor cell survival, we isolated mutations in the gene pect, which encodes CTP:phosphoethanolamine cytidylyltransferase. In these mutants, light-evoked photoreceptor potentials persisted after 20-s light stimulation, indicating prolonged activation of the visual response. These mutants also exhibited light-independent degeneration of photoreceptor neurons, and lipidomic analysis revealed alterations in major phospholipid composition. We manipulated phospholipid composition via comprehensive genetic interactions and concluded that pect mutant phenotypes resulted from PE deficiency. Strikingly, increasing PE synthesis through the PSD pathway effectively suppressed retinal degeneration in pect mutants. Finally, the Mitochondria Associated Membrane (MAM)-enriched proteins MFN and SERCA were required for PSD to rescue pect phenotypes. We therefore proposed a model in which PE synthesized in the mitochondria through PSD is transported back to the ER through ERMCS when cellular PE is deficient (Fig 6E). is 10 μm. (C) Western blot analysis of fractionated ER and mitochondria from ninaE-Psd-GFP flies that were labeled with antibodies against GFP, ATP5A, Porin, COX4, and SERCA. (D) Cryostat sections from GMR-mitoGFP and pect 29 ;GMR-mitoGFP (ey-flp rh1-GFP; pect 29 FRT40A/GMR-hid CL FRT40A;GMR-mitoGFP/+) flies were labeled with antibodies against LOVIT (red) and GFP (green). The longitudinal views of the visual system were shown in upper panels and the higher magnification of lamina regions were shown in lower panels. Scale bars are 50 μm or 10 μm in upper and lower panels, respectively. (E) TEM micrographs of cartridges containing wild type, pect 29, and pect 29 ;GMR-Psd terminals, respectively. Scale bar is 1 μm. (F) Rates of ATP synthesis from the dissected fly retinas of indicated genotypes. Data are presented as mean ± SD, ns, not significant (Student's unpaired t-test). n = 3. (G) The hydrogen peroxide levels detected in the dissected fly retinas from different genotypes. Data are presented as mean ± SD, ns, not significant (Student's unpaired ttest). n = 3. (H) Lipidomic analysis of mitochondria isolated from dissected retinas of wt, pect 29 , and pect 29 ;GMR-Psd flies at indicated time points. PE levels expressed in molar fractions are normalized to total phospholipids. Data are presented as mean ± SD, � p < 0.05, ��� p < 0.001 (Student's unpaired t-test). n = 4. https://doi.org/10.1371/journal.pgen.1009070.g005

PE homeostasis is crucial for neuronal survival and activity
Membrane phospholipids, in particular PE, play key roles in regulating neuronal activity and integrity. Drosophila photoreceptor neurons utilize the fastest phospholipid signaling cascade and exhibit high rates of membrane trafficking. Therefore, it is particularly important for neurons to maintain phospholipid pools [23,46]. As phototransduction is completely mediated by phospholipase C (PLC), maintaining levels of PIP 2 and its product DAG are critical for visual responses [24,47]. Mutations in the gene pect prevent the synthesis of PE from DAG, resulting in increased levels of PI and DAG, and prolonged visual responses. Reducing DAG levels by either introducing mutations in the Lazaro enzyme, which converts PA to DAG, or overexpressing the DAG kinase rdgA [36,37], did not suppress the prolonged afterpotentials in pect mutants. Moreover, inhibiting PI synthesis failed to suppress the retinal degeneration and defective ERG responses [48]. In addition, TRP channels were constitutively active in rdgA mutant photoreceptors, and the over-activation and retinal degeneration of rdgA mutants were rescued in rdgA;trp double mutants [34]. In the case of pect, loss of TRP channels did not affect the severity of neurodegeneration, indicating that RDGA and PECT function in different pathways. In conclusion, PECT regulates photoreceptor function and morphology independent of PI and DAG metabolism. Recent evidence suggested TRPs are mechanosensitive channels, and membrane physical properties are involved in channel activation [49][50][51]. Indeed, the rhabdomere size was reduced in the 1-day-old pect mutants, indicating changes in the physical properties of the lipid bilayer. We speculate that alterations in phospholipid composition in pect mutants may change the compact structure and membrane fluidity of rhabdomere, which leads to prolonged activation of TRP channels.
We further provide direct evidence that maintaining cellular PE levels is critical for neuronal function and integrity. First, disruption of the PSD pathway by knocking down PSD enhanced retinal degeneration in pect mutants. More importantly, overexpression of PSD restored cellular PE levels, and thus greatly suppressed both the prolonged afterpotentials in response to light and retinal degeneration of pect mutants. Moreover, this suppression was fully reversed by down-regulating PSS. It has been reported that knockdown of Psd resulted in light-dependent retinal degeneration by preventing autophagy-dependent rhodopsin degradation since the abundance of PE could positively regulate autophagy [26,52]. However, for pect mutants, rhodopsin turn-over is normal and degeneration is independent of light and TRP channel activity, suggesting that rhodopsin homeostasis does not cause pect-induced cell death. Furthermore, the light-independent retinal degeneration in pect mutants suggested that cellular PE homeostasis independently contributes to maintaining neuronal activity and integrity. least 20 ommatidia from each eye were quantified for each genotype. Data are presented as mean ± SD, ns, not significant, ��� p < 0.001 (Student's unpaired t-test). (C) TEM sections were obtained from pect 29 ;GMR-Psd GMR-Gal4 (ey-flp rh1-GFP;pect 29 FRT40A/GMR-hid CL FRT40A;GMR-Psd longGMR-Gal4/+), pect 29  and GMR>ip3r RNAi (longGMR-Gal4/UAS-ip3r RNAi ). All flies were raised for 10 days under 12h-light/12h-dark cycles. Scale bar is 2 μm. (D) Quantification of rhabdomeres per ommatidium in genotypes indicated. At least 10 ommatidia from each section of three different eyes were quantified for each genotype. Data are presented as mean ± SD, ns, not significant, �� p < 0.01, ��� p < 0.001 (Student's unpaired t-test). The knock-down efficiencies of mfn RNAi , serca RNAi , miro RNAi , porin RNAi , ip3r RNAi were 54%, 54%, 56%, 48% and 24%, respectively. (E) Model of lipid exchange between ER and mitochondria.PS is generated from PC or PE through PSS which is localized in the MAMs. PS is then transported to the mitochondria inner membrane where PS is converted to PE through PSD. PE synthesized in the mitochondria is transported back to the ER upon cellular PE deficiency. https://doi.org/10.1371/journal.pgen.1009070.g006

Complementary roles of CDP-ethanolamine and the PSD pathway in PE synthesis
The majority of mitochondrial PE is synthesized in situ in mitochondria via PSD. In contrast, only a small fraction of mitochondrial PE is made in the ER by the CDP-ethanolamine pathway [14,15,53,54]. Maintaining mitochondrial PE levels is critical for mitochondrial respiratory capacity, morphology, and distribution, thus deletion of PSD impairs mitochondrial function, resulting in lethality [6,[55][56][57]. The pect mutant photoreceptor cells are dysfunctional, and both cellular and mitochondrial levels of PE are significantly reduced. Although it has been reported that < 30% depletion of mitochondrial PE by RNAi silencing of PSD alters mitochondrial morphology and function in mammalian cells [5],~40% reduction in mitochondrial PE levels in pect 29 flies did not affect mitochondrial activity, morphology, or axonal localization. Overexpression of PSD, which completely rescued photoreceptor function and integrity in pect mutants, restored total PE levels but not levels of mitochondrial PE. Therefore, disruption of the CDP-ethanolamine PE synthesis pathway did not impair mitochondrial function. Disrupting PE synthesis through the CDP-ethanolamine pathway does not result in a mitochondrial phenotype because mitochondrial PE is synthesized locally through PSD, although the CDP-ethanolamine pathway does contribute to total mitochondrial PE levels.
A recent study reported that yeast Psd1 localizes to both mitochondria and the ER through its transmembrane region (TMR) [58]. This forced us to consider that PSD in the ER may have restored cellular PE homeostasis in pect mutants. However, Drosophila PSD lacks the TMR sequence required for Psd1 to localize to the ER, and PSD remains in mitochondria when overexpression. ER-mitochondria connections are necessary for the efficient exchange of phospholipids between organelles [40,59]. Several proteins, including the mitofusin MFN2 and the ER Ca 2+ ATPase SERCA, have been implicated in maintaining ERMCS, thus facilitating phospholipid exchange. Consistent with previous reports, we found that the fly proteins, MFN and SERCA, stabilize ER-mitochondrial contact sites [60]. Importantly, disrupting ERmitochondria contacts through mfn RNAi or serca RNAi completely blocked the ability of PSD to rescue pect mutant phenotypes. This is also consistent with a recent study that described an unexpected role of MFN2 in PS transfer between the ER and mitochondria [61]. However, as loss of SERCA induces UPR reaction, we cannot rule out the possible role of inducing UPR in disruption of PE homeostasis. Taken together, these analyses prove that cellular PE homeostasis is maintained, in part, by synthesizing PE in mitochondria and then exporting this PE to cellular pools. Thus, mitochondria play a critical role in cellular phospholipid homeostasis.

Similar PE species are generated by the CDP-ethanolamine and PSD pathways
PE species generated from the CDP-ethanolamine and PSD pathways are different, especially fatty acids in the sn-2 position, where PE from the CDP-ethanolamine and PSD pathways prefer mono-/di-unsaturated and polyunsaturated fatty acids, respectively [18]. It has been suggested that individual molecular species of PE may play specialized roles in cellular signaling, which explained why deletion of either Pisd or Pcyt2 causes embryonic lethality in mice [6,12,62]. Here we saw that as total PE levels decreased, the proportion of most PE species were unaffected. In contrast, the PE species PE38:1, PE36:4, and PE36:5 were down-regulated, and PSD overexpression fully restored levels of these three PE species, as well as total PE levels. This suggests that the same PE species are generated by the CDP-ethanolamine and PSD pathway in vivo, although they might prefer different substrates in vitro. These data further showed that PE generated in the mitochondria can compensate for cellular PE deficiency in the Drosophila visual system. A recent study also identified recessive lethal mutations in pect and demonstrated that the biosynthesis of specific phospholipids is linked to neurodegeneration and synaptic vesicle loss in adult Drosophila photoreceptors [25]. In pect omb593 mutants levels of PE 34:1 and PE 36:2 were reduced, but the overall proportions of PE species were not significantly changed, whereas we detected a significant reduction in total cellular PE levels. The different results of phospholipid composition in pect mutants may come from the different alleles analyzed. In contrast to the nonsense pect 29 mutation, the pect omb593 contains a hypomorphic mutation with a single amino acid change (H55Y). Thus PECT H55Y might reduce PE levels to a much lesser extent than complete loss of pect, and this cellular PE reduction could be compensated by the alternative PSD pathway. Moreover, knocking down srebp (regulatory element binding protein) reversed the loss of synaptic vesicles, but failed to suppress axonal or retinal degeneration in pect omb593 mutants [25]. Therefore, SREBP is a downstream factor of PE deficiency and specifically functions in maintaining the synaptic vesicle pools. Here, we demonstrated that increasing PE levels through induction of the PSD pathway restored the cellular PE levels and effectively suppressed retinal degeneration and defective phototransduction in pect mutants, supporting disruption of PE homeostasis is the major cause of photoreceptor cell degeneration and loss of synaptic transmission.

Physiological functions of both PE synthesis pathways
Phospholipid composition is critical for cellular homeostasis, and alterations in the composition of major phospholipid PE are implicated in multiple diseases [61,63]. Cellular PC/PE molar ratios can influence energy metabolism in numerous organelles and thus lead to disease conditions such as steatohepatitis, obesity, and muscular dystrophy [63][64][65]. Cellular PC levels were elevated by 57% in pect 29 retinas compared with wild type, but inhibiting PC synthesis did not suppress the retinal degeneration, suggesting that PE levels, but not the PC/PE ratio, are crucial for neuronal function.
Genetic mutations that affect PE synthesis have been identified in several human autosomal-recessive disorders. In particular, mutations in PISD, the human counterpart of fly PSD, cause Liberfarb syndrome, which is a multisystem disorder affecting the eyes, ears, bone, and brain [22]. Studies in patient-derived fibroblasts revealed impaired phospholipid metabolism, altered mitochondrial respiration, and fragmentation of the mitochondrial network [20,21]. Recently, mutations in PCYT2, the human counterpart of fly PECT, have been associated with hereditary spastic paraplegia. Lipidomic analysis of patient fibroblasts revealed profound lipid abnormalities impacting both neutral etherlipid and etherphospholipid metabolism [7]. As mutations in fly pect affect phospholipid composition, resulting in defective photoresponse and severe retinal degeneration, this system represents a conserved model for studying diseases associated with PE deficiency. Moreover, our data provide direct genetic evidence that induction of mitochondrial PE synthesis can compensate for deficiencies in cellular PE, and suggest that mitochondrial phospholipid synthesis and trafficking represent a promising therapeutic target for treating disorders associated with defective phospholipid composition.

Fly stocks
The following stocks were obtained from the Bloomington Stock Center:

Generation of transgenic flies
The pect, rdgA, Pcyt1, and Psd cDNAs were amplified from the cDNA clones GH23180, RH08828, LD46058, and LD21713, respectively, which were obtained from the Drosophila Genomic Resource Center. The bbc cDNA was amplified from the w 1118 cDNAs. The PCYT2 cDNA was amplified from the cDNA clone IOH3644, which was obtained from the Ultimate ORF Clones (Invitrogen). To express pect or PCYT2 cDNAs under control of the ninaE or GMR promoter [27], the cDNAs were subcloned into the pninaE-attB or pGMR-attB vector between the EcoRI and NotI sites. To express rdgA, Pcyt1, bbc, or Psd cDNAs under control of the GMR promoter, the cDNAs were subcloned into the pGMR-attB vector between the NotI and XbaI sites. The pect cDNA was subcloned into the pIB-c-3xflag vector between the EcoRI and NotI sites. The fragment pect-3xflag was subcloned into the pninaE-attB vector between the EcoRI and XbaI sites. The Psd cDNA was subcloned into the pIB-c-GFP vector between the Acc65I and NotI sites. The fragment Psd-GFP was subcloned into the pninaE-attB vector between the Acc65I and XbaI sites. These constructs were injected into M(vas-int.Dm)ZH-2A; M(3xP3-RFP.attP)ZH-86Fb embryos and transformants were identified based on eye color. The 3xP3-RFP marker was eliminated by crossing to a Cre-expressing line.

Generation of antibodies directed against PECT
The pect cDNA was subcloned into the pRSETB vector between the BglII and EcoRI sites, and the plasmid was transformed into BL21 (DE3) competent cells. Protein expression was induced with 0.5 mM IPTG, and cells were harvested and resuspended in phosphate-buffered saline (PBS, pH 7.4) supplemented with complete EDTA-free protease inhibitor cocktail (Sigma). Cells were lysed by freeze-thaw followed by sonication, and inclusion bodies were collected by centrifugation for 10 min, 13000 rpm at 4˚C. The pellet was washed with washing buffer 1 (PBS, 1 mM EDTA, 1% Triton X-100), followed by washing buffer 2 (PBS, 1 mM EDTA, 2 M urea). The final pellet was homogenized in SDS sample buffer and fractionated by SDS-PAGE. The band corresponding to the size of PECT was sliced and further sent to immunize a rabbit.

Generation of antibodies directed against CNX and Porin
Polyclonal antibodies against Drosophila Calnexin 99A (CNX) were generated by immunizing a rabbit with a synthetic peptide (CESREPAQTEESNTKTRKRQARKEK) conjugated to KLH. Polyclonal antibodies against Drosophila Porin were generated by immunizing a rat with a synthetic peptide (CGYQTAFDTQQSKLTTNNF) conjugated to KLH.

Subcellular fractionation
To separate the cytoplasmic and membranal fractions, subcellular fractionation was performed, as described [67]. Twenty fly heads were collected and homogenized in lysis buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5 mM EDTA) containing complete EDTA-free protease inhibitor cocktail (Sigma). Homogenates were centrifuged at 950 g at 4˚C for 10 min to remove cell debris. The supernatant was centrifuged at 21,000 g at 4˚C for 40 min. The supernatant and pellet fractions were solubilized in SDS sample buffer to equal volumes for western blot analysis. For characterization of subcellular fractions, the following antibodies were used: Tubulin (mouse, 1:2000, Developmental Studies Hybridoma Bank), Rh1 (mouse, 1:2000, Developmental Studies Hybridoma Bank), and PECT (rabbit, 1:2000).
ER and mitochondria fractions were isolated from fly heads, as described [68]. Briefly, 1,000 fly heads were homogenized with MTE (270 mM mannitol, 10 mM Tris, 0.1 mM EDTA, pH 7.4) containing complete EDTA-free protease inhibitor cocktail (Sigma). The homogenate was centrifuged twice at 600 g for 5 min to remove large debris and nuclei. The supernatant was centrifuged for 10 min at 15,000 g at 4˚C to separate crude ER and crude mitochondria. The supernatant containing crude ER was loaded on 1.3 M/1.5 M/2.0 M step sucrose density gradients and centrifuged at 152,000 g for 70 min. The large band at the interface of the 1.3M sucrose gradient layer was collected with a syringe, diluted with MTE, and centrifuged at 126,000 g for 45 min. The final pellet was resuspended in PBS and stored at -20˚C. The mitochondria pellet was washed with MTE three times and resuspended in MTE buffer. The crude mitochondria were loaded on 1.7 M/1.0 M step sucrose density gradients and centrifuged at 40,000 g for 22 min. The band at the interface of the 1.7 M and 1.0 M sucrose layers was extracted with a syringe, diluted with MTE, and centrifuged at 15,000 g for 10 min. The final pellet was resuspended in PBS and stored at -20˚C. For characterization of subcellular fractions, the following antibodies were used: GFP (rabbit, 1:2000, Torrey Pines Biolabs), ATP5A (mouse, 1:2000, #ab14748, Abcam), Porin (rat, 1:500), COX4 (rabbit, 1:2000) [69], and SERCA (rabbit, 1:1000, Santa Cruz).

PECT enzyme activity assays
To purify PECT recombinant protein, the pect cDNA was subcloned into the pMAL vector between the EcoRI and SalI sites. The plasmid was transformed into BL21 (DE3) competent cells and grown at 37˚C in LB medium containing Ampicillin and 0.2% glucose to an OD 600 of around 0.8. Protein expression was induced with 0.1 mM IPTG and culture incubated for 24 h at 13˚C, 220 rpm. Cells were harvested and resuspended in column Buffer (20 mM Tris-HCl, 200 mM NaCl, 1 mM EDTA, pH 7.4) supplemented with complete EDTA-free protease inhibitor cocktail (Sigma). Cells were lysed by freeze-thaw followed by sonication. The supernatant was collected by centrifugation for 30 min, 20,000 g at 4˚C. Then it was diluted 1:6 with Column Buffer and loaded onto an amylose column (New England Biolabs). The fusion protein was eluted with column buffer supplemented with 10 mM maltose.
The purified MBP-PECT or MBP proteins were used for enzyme activity assay, as described [70]. The reaction mixture contained the following components in 25 μl: 20 mM Tris-HCl (pH 7.7), 10 mM MgCl 2 , 2 mM CTP, 1 mM phosphoethanolamine, 1 mM DTT, and 1 μM MBP-PECT or MBP. All reactions were incubated at 37˚C for 15 min followed by boiling for 3 min to inactivate the enzyme. The product CDP-ethanolamine was analyzed with liquid chromatography-mass spectrometry (LC-MS). The Dionex Ultimate 3000 UPLC system was coupled to a TSQ Quantum Ultra triple-quadrupole mass spectrometer (ThermoFisher, Waltham, MA), equipped with a heated electrospray ionization (HESI) probe operated in negative ion mode. The reaction mixture was separated by an Atlantis Silica column (2.1 × 100 mm, 3 μm, Waters, Milford, MA). Separations were performed using binary gradient mobile phases, consisting of water with 0.1% formic acid (A) and ACN with 0.1% formic acid (B). Data were acquired in selected reaction monitoring (SRM) for CDP-ethanolamine with a transition of 447.100/324.148. Data analysis and quantification were performed using software Xcalibur 3.0.63 (ThermoFisher, Waltham, MA).

Electroretinogram recordings
ERG recordings were performed as described [71]. Two glass microelectrodes filled with Ringer's solution were inserted into small drops of electrode cream (PARKER LABORATORIES) that were placed onto the surface of the compound eye and the thorax. A Newport light projector (model 765) was used for stimulation. ERG signals were amplified with a Warner electrometer IE-210 and recorded with a MacLab/4s analog-to-digital converter and the Clampex 10.2 program (Molecular Devices, San Jose, CA). All recordings were carried out at room temperature.
To investigate the localization pattern of major proteins involved in the phototransduction process, resin-embedded sections were labeled as described [72]. Hemisected fly heads were fixed with 4% PFA in PBS and embedded in LR White resin. Cross sections (0.5 μm) of compound eyes were cut through the distal region of the retina, which included the R7 cells. Sections were incubated with primary antibodies against Rh1 (mouse monoclonal 4C5, 1:200, Developmental Studies Hybridoma Bank) and TRP (rabbit, 1:200) [71] at room temperature for 1 h. Secondary antibodies against mouse Alexa Fluor 488 and rabbit Alexa Fluor 568 were used (1:400, Invitrogen).

Cell transfection
S2 cells were transfected by adding 2.5 μg plasmid mixed with 1 μl VigoFect (Vigorous Biotechnology) into the cell media. Twenty-four hours following transfection, cells were suspended and adhered to a poly-lysine coated coverslip for 1 h and then fixed with 4% PFA in PBS for 30 min at room temperature. Cells were incubated with primary antibodies against Tom20 (rat, 1:100) [74] and GFP (rabbit, 1:200, Invitrogen). Secondary antibodies against rat Alexa Fluor 568 and rabbit Alexa Fluor 488 were used (1:400, Invitrogen). Images were captured with a Nikon A1-R confocal microscope (Nikon, Tokyo, Japan). Acquired images were processed using Photoshop CC 2017 and NIS-Elements AR Analysis 5.20.00.

Mitochondria activity assays
The ATP assay kit was obtained from Beyotime and the assay was performed according to the manufacturer's instruction. To measure total ATP level, 10 dissected retinas were homogenized in extraction buffer. After centrifugation to remove cell debris, the supernatant was added to the substrate solution. The luminescence was recorded in an Illuminometer (EnSpire Multimode Plate Reader; PerkinElmer, Waltham, MA).
The intracellular hydrogen peroxide assay kit was obtained from Sigma and the assay was performed according to the manufacturer's instruction. Briefly, 10 dissected retinas were homogenized in assay buffer. The homogenized supernatant was mixed with a fluorescent peroxide sensor and incubated at room temperature for 15 min. The fluorescence intensity at (λex = 490/λem = 525 nm) was measured in a fluorescence plate reader (EnSpire Multimode Plate Reader; PerkinElmer, Waltham, MA).
To check mitochondrial membrane potential, pect 29 FRT40A/CyO flies were crossed with hs-flp;FRT40A ubi-GFP flies, and heat shocked at the 1 st instar larva stage at 37˚C for 1 hour to induce mosaic clone. Pupa eyes were then dissected at 36~40 h APF (after pupa formation), and stained with 100 nM TMRM (tetramethylrhodamine methyl ester, ThermoFisher). GFP positive cells are quantified as wild-type cells, and GFP negative cells are pect 29 mutant cells.

Single ommatidium observation
Ommatidia from 1-day-old flies were dissected in Schneider's Drosophila medium (Invitrogen) as describe [75]. Images were taken within 30 min under a Nikon A1-R confocal microscope (Nikon, Tokyo, Japan). The Pearson's Coefficient, as a measure of colocalization between KDEL-GFP and Tom70-RFP, was calculated using NIS-Elements Advanced Research Analysis 5.20.00 (Nikon, Tokyo, Japan).

Transmission electron microscopy
To visualize Drosophila retina ultrastructure, adult fly heads were dissected, fixed, dehydrated, and embedded in LR White resin (Electron Microscopy Sciences) as described [67]. Thin sections (80 nm) prepared at a depth of 30-40 μm were stained with uranyl acetate and lead citrate (Ted Pella) and examined using a JEM-1400 transmission electron microscope (JEOL, Tokyo, Japan) equipped with a Gatan CCD (4k × 3.7k pixels, USA).
TEM of photoreceptor terminals was performed as described [73]. Adult fly heads were dissected in 4% PFA and the retinas were removed. The dissected lamina was fixed in a solution with 4% PFA and 2.5% glutaraldehyde for 2 h on ice, followed by fixation in 1% osmium tetroxide for 1.5 h at 4˚C. Tissues were then dehydrated in a series of ethanol dilutions at 4˚C (10-min wash in 10,25,40,55,70,85,95, and 30-min wash in 100% ethanol for 5 times). Samples were gradually infiltrated with 2 ratios of ethanol and Eponate 12 (Ted Pella), finally going into 3 changes of pure resin. Samples were allowed to infiltrate in pure resin overnight on a rotator and embedded in Eponate 12 resin (Ted Pella). Thin sections (80 nm) were stained with uranyl acetate and lead-citrate (Ted Pella) and examined using a JEM-1400 transmission electron microscope (JEOL, Tokyo, Japan) equipped with a Gatan CCD (4k × 3.7k pixels, USA).

Lipid extraction and mass spectrometric analysis
Lipids were extracted from retinas or mitochondria as described [76]. Mitochondria were extracted from dissected Drosophila retinas using a mitochondria isolation kit (Abcam) designated for tissue extraction according to the manufacturer's protocol, following tissue homogenization using a Dounce homogenizer (KONTES). Polar lipids were analyzed on an Exion UPLC system coupled with a triple quadrupole/ion trap mass spectrometer (QTRAP 6500 PLUS, Sciex). PC-d 31  . Glycerol lipids triacylglycerides (TAG) and diacyglycerides (DAG) were analyzed using a modified protocol of reverse-phase HPLC/ ESI/MS described previously [77]. DAG species were quantified using 4ME 16:0 Diether DG as an internal standard (Avanti Polar Lipids, Alabaster, AL, USA). Separation of individual lipid classes of polar lipids by NP-HPLC was carried out using a Phenomenex Luna 3 μ-silica column (i.d. 150 × 2.0 mm) with the following conditions: mobile phase A (chloroform: methanol: ammonium hydroxide, 89.5:10:0.5) and mobile phase B (chloroform: methanol: ammonium hydroxide: water, 55:39:0.5:5.5). The initial mobile phase proportion was 95% (A) / 5% (B), which was held for 5 min. A was then linearly decreased to 60% in 7 min, which was held for 4 min. A was further decreased to 30%, which was held for 15 min. Finally, the initial mobile phase conditions were restored and held for 5 min. Data were acquired in multiple reaction monitoring (MRM) in a combined workflow for polar lipids analysis. The source parameters are as follows: curtain gas: 20, ion spray voltage: 5500 V, temperature: 400˚C, ion source gas 1: 35, ion source gas 2: 35. MS profiles were recorded under both positive and negative modes in separate runs (resolution 60,000), and mass accuracy of less than 2 ppm was obtained throughout the analytical runs. The mole fraction of each lipid was normalized to the mole fraction of total polar lipids.

Statistical analysis
All data in bar and line graphs are expressed as Means ± SDs. Significant differences between different groups were determined using Student's unpaired t-tests in Graphpad Prism 6 (Graphpad Software Inc.). ( ��� p < 0.001; �� p < 0.01; � p < 0.05; ns, not significant).  29 ;laza 1 , laza 1 , pect 29 ;GMR-rdgA, GMR-rdgA flies. Flies were dark-adapted for 2 min and subsequently exposed to a 1-s pulse followed by a 20-s then a 1-s pulse of orange light. (B) Reducing PC level did not suppress retinal degeneration in pect 29 flies. TEM sections were obtained from wt, pect 29