A conserved role for AMP-activated protein kinase in NGLY1 deficiency

Mutations in human N-glycanase 1 (NGLY1) cause the first known congenital disorder of deglycosylation (CDDG). Patients with this rare disease, which is also known as NGLY1 deficiency, exhibit global developmental delay and other phenotypes including neuropathy, movement disorder, and constipation. NGLY1 is known to regulate proteasomal and mitophagy gene expression through activation of a transcription factor called "nuclear factor erythroid 2-like 1" (NFE2L1). Loss of NGLY1 has also been shown to impair energy metabolism, but the molecular basis for this phenotype and its in vivo consequences are not well understood. Using a combination of genetic studies, imaging, and biochemical assays, here we report that loss of NGLY1 in the visceral muscle of the Drosophila larval intestine results in a severe reduction in the level of AMP-activated protein kinase α (AMPKα), leading to energy metabolism defects, impaired gut peristalsis, failure to empty the gut, and animal lethality. Ngly1–/– mouse embryonic fibroblasts and NGLY1 deficiency patient fibroblasts also show reduced AMPKα levels. Moreover, pharmacological activation of AMPK signaling significantly suppressed the energy metabolism defects in these cells. Importantly, the reduced AMPKα level and impaired energy metabolism observed in NGLY1 deficiency models are not caused by the loss of NFE2L1 activity. Taken together, these observations identify reduced AMPK signaling as a conserved mediator of energy metabolism defects in NGLY1 deficiency and suggest AMPK signaling as a therapeutic target in this disease.


Introduction
The cytoplasmic enzyme N-glycanase 1 (NGLY1) catalyzes the removal of N-linked glycans from glycoproteins and is thought to operate as part of the endoplasmic reticulum-associated degradation (ERAD) pathway [1]. Recessive mutations in human NGLY1 result in a genetic disorder with various phenotypes including developmental delay, seizures, hypo-/alacrima, elevated liver enzymes, diminished deep tendon reflexes, muscle weakness, orthopedic manifestations, and chronic constipation [2][3][4][5][6][7][8]. This disease is a congenital disorder of deglycosylation (OMIM # 615273) and is commonly referred to as NGLY1 deficiency. NGLY1 and its homologs in model organisms have been shown to regulate the proteasomal gene expression by deglycosylating a transcription factor called "nuclear factor erythroid 2-like 1" (NFE2L1; also called NRF1; SKN-1 in worms) [9][10][11]. However, the extent to which this proteasomal defect contributes to NGLY1 deficiency phenotypes in human patients and developmental abnormalities observed in Ngly1-mutant animal models remains to be determined. It is worth mentioning that NFE2L1 and its paralog NFE2L2 have also been called NRF1 and NRF2. However, since NRF1 is the official symbol for a distinct protein called nuclear respiratory factor 1 in mammals, we will use NFE2L1 and NFE2L2 in the current work.
Studies in tissue samples from NGLY1 deficiency patients, patient fibroblasts, Ngly1 -/mouse embryonic fibroblasts (MEFs) and C. elegans mutants for the NGLY1 homolog (png-1 -/-) have shown structural and functional abnormalities in mitochondria [7,12]. Moreover, Ngly1 -/-MEFs fail to properly clear damaged mitochondria via mitophagy [13]. Given that a number of NGLY1 deficiency phenotypes like developmental delay, neuropathy, muscle weakness, and seizures are also observed in mitochondrial disorders [14], they are considered to be one of the differential diagnoses in patients suspected of having NGLY1 deficiency [2]. Therefore, although a specific phenotype in human NGLY1 deficiency patients and Ngly1mutant animals is yet to be directly linked to energy homeostasis defects, these studies suggest that mitochondrial abnormalities might contribute to some aspects of NGLY1 deficiency pathophysiology, and that improving energy homeostasis might be beneficial in this patient population.
The Drosophila genome encodes a single NGLY1 homolog called PNGase-like or Pngl [15]. Under regular culture conditions, Pngl-null (Pngl -/-) Drosophila show a significant developmental delay and~99% lethality [15,16]. We have previously reported that 80%-85% of the lethality observed in Pngl -/animals can be rescued by transgenic expression of human NGLY1 in the mesoderm [16]. Around 20-30% of the lethality can be explained by a tissue-specific requirement for Pngl in the visceral mesoderm to promote bone morphogenetic protein (BMP) signaling mediated by the fly BMP protein Decapentaplegic or Dpp [16]. Pngl -/larvae also show a food accumulation phenotype (severe failure in gut clearance) that cannot be explained by the loss of BMP signaling [16]. However, the molecular basis of gut clearance defects and their contribution to the lethality observed in Pngl -/animals remained to be identified. Moreover, it was not clear whether at a mechanistic level the regulation of Drosophila gut clearance by Pngl has any parallels in mammals.
Here, we show that the food accumulation phenotype in Pngl -/larvae is caused by reduced mesodermal expression of AMP-activated protein kinase α subunit (AMPKα), which encodes the catalytic subunit for a major energy sensor in the cells [17]. The midgut in Pngl -/larvae exhibits abnormal mitochondrial cristae, reduced ATP content and increased oxidative stress, all of which can be improved upon restoring AMPKα levels in the mesoderm. Importantly, both Ngly1 -/-MEFs and fibroblasts from NGLY1 deficiency patients show a similar reduction in AMPKα1 and AMPKα2 expression. Moreover, pharmacological enhancement of AMPK signaling rescues the impaired energy homeostasis in both model systems. Importantly, the reduction in AMPKα level cannot be explained by impaired proteasomal gene expression in NGLY1deficient models. Our work identifies reduced AMPKα expression as an evolutionarily conserved mechanism contributing to the energy homeostasis defects in NGLY1-deficient animals and suggests AMPK signaling as a potential therapeutic target in NGLY1 deficiency patients.

Drosophila Pngl is required in the mesoderm but not neurons or endoderm to promote midgut peristalsis
To determine the mechanism for the food accumulation phenotypes observed in Pngl mutant larvae, we first performed transgenic rescue experiments. Overexpression of wild-type Pngl in the mesoderm by using the GAL4/UAS system [18] fully rescued the food accumulation phenotype of Pngl -/larvae, similar to adding one copy a Pngl genomic transgene, Pngl Dp ( Fig 1A  and S1 Fig). Moreover, mesodermal overexpression (Mef2-GAL4) of Pngl with a C303A mutation in its catalytic domain failed to rescue the food accumulation phenotype in Pngl -/larvae ( Fig 1A). Similar to human NGLY1, mesodermal expression of the fly Pngl (but not Pngl C303A ) rescued the lethality in~80% of Pngl -/animals ( Fig 1B). However, none of these phenotypes were rescued by endodermal (NP-3270-GAL4) or neuronal (elav-GAL4) overexpression of Pngl in Pngl -/animals ( Fig 1A and 1B). Together, these observations indicate a critical role for Pngl's enzymatic activity in the mesoderm to ensure gut clearance and animal survival.
Phalloidin staining and transmission electron microscopy (TEM) of the visceral muscles in third instar larvae do not show an obvious morphological defect in Pngl -/animals ( Fig 1C and  1D and S2 Fig). To examine whether the food accumulation phenotype of Pngl -/larvae can be explained by a functional defect in the midgut, we assessed the effect of loss of Pngl on the peristaltic activity in freshly dissected larval midguts. We found a significant decrease in the number of midgut contractions per minute in Pngl -/midguts compared to control midguts ( Fig 1E). This phenotype was rescued by introducing genomic Pngl Dp or by overexpressing Pngl in the mesoderm, but not in neurons or endoderm (Fig 1E). These observations indicate that Pngl is required in the visceral muscle to promote midgut peristalsis and gut clearance.

AMPKα expression is reduced in Pngl -/larval midguts
We previously reported that the food accumulation phenotype observed in Pngl -/larvae can be recapitulated by Pngl knock-down in the mesoderm by how 24B -GAL4, but not by dpp The food accumulation phenotype in Pngl mutants is associated with gut peristalsis defects and is rescued by mesoderm-specific enzymatic activity of Pngl. (A,B) Gut clearance assays in 3rd instar larvae and lethality rescue assays of the indicated genotypes are shown. The x-axis shows hours (h) after egg laying. The food accumulation phenotype (A) and the lethality (B) of Pngl mutants is rescued by a Pngl genomic duplication (Pngl Dp) and by overexpressing wild-type (WT) Pngl by a mesodermal driver (Mef2-GAL4) but not by expressing an enzymatic-deficient Pngl mutant (C303A) in the mesoderm or by expressing Pngl WT using neuronal (elav-GAL4) and endodermal (NP-3270-GAL4) drivers. Data in (A) represent mean ± SD of three independent experiments. Animal numbers in (B) are 200, 240, 160, 130, 193, and 145 from left to right. (C,D) Representative images of midguts of y w control (C) and Pngl -/-(D) larvae stained with phalloidin (red) do not show gross morphological defects in the mutant visceral muscle. The scale bar in D is 50 μm and applies to (C) as well. (E) Midgut contraction frequencies per minute in Pngl -/larvae is significantly reduced compared to control animals and is rescued by Pngl Dp and mesodermal but not neuronal or endodermal overexpression of Pngl WT

PLOS GENETICS
Role of AMPK in NGLY1 deficiency knock-down using the same driver [16]. We recently reported that one copy of the dpp HA-3NQ knock-in allele, in which three out of the four Dpp N-glycosylation sites are mutated, can fully rescue the BMP signaling defects in Pngl -/midguts [19]. However, this allele failed to improve the food accumulation phenotype in Pngl -/larvae, further indicating that the gut clearance phenotype in Pngl mutants is not caused by impaired BMP signaling (S3A Fig). A previous report had shown that AMPKα is required in the visceral muscle to promote peristaltic activity and gut clearance in Drosophila larvae [20]. Accordingly, we examined whether loss of Pngl affects AMPKα level and activation. As shown in Fig 2A, AMPKα mRNA expression level is significantly reduced in larval midguts homozygous for two different Pngl null alleles (Pngl ex14 and Pngl ex18 ; [15,16]) compared to the control y w animals. Western blotting showed a severe decrease in AMPKα and phospho-AMPKα (pAMPKα; active form) protein levels in Pngl ex14/ex14 and Pngl ex18/ex18 larval midguts compared to control midguts without changing the pAMPKα/AMPKα ratio ( Fig 2B). These phenotypes can be fully rescued upon mesodermal overexpression of wild-type Pngl but not Pngl C303A , indicating that the Pngl enzymatic activity is required for the regulation of AMPKα level (S4A and S4B Fig). AMPKα functions as a heterotrimer with AMPKβ and AMPKγ [17]. However, loss of Pngl does not result in a statistically significant alteration in AMPKβ and AMPKγ mRNA levels in larval midguts (S4C Fig), suggesting that the observed effect is specific for the AMPKα subunit. These data suggest that Pngl is required for AMPKα expression in the fly midgut. Given the similar phenotypes observed here and previously [15,16] for these two null Pngl alleles, we have used Pngl ex14/ex14 animals in the rest of the current study and will refer to them as Pngl -/-.

Restoration of AMPKα expression levels in the mesoderm suppresses the food accumulation and gut peristalsis phenotypes and partially rescues the lethality in Pngl mutant
If a decrease in midgut expression of AMPKα contributes to the Pngl -/food accumulation phenotype, increasing AMPKα expression should improve midgut emptying in these animals. To test this notion, we first used the genomic duplication Dp(1;3)DC102, PBac{DC102}VK33 (referred to as AMPKα Dp hereafter), which contains AMPKα and seven neighboring genes (S4D Fig). One copy of this duplication increased AMPKα mRNA and AMPKα and pAMPKα protein levels in Pngl -/midguts (S4E and S4F Fig). This duplication also improved the gut clearance and significantly increased the number of midgut contractions per minute in Pngl -/animals ( Fig 2C and 2D). Remarkably, adding a copy of this duplication increased the survival of Pngl -/animals to adulthood from~1% to around 44% the expected Mendelian ratio ( Fig  2E). As expected, the AMPKα Dp did not rescue the BMP-dependent phenotypes in Pngl -/midguts (S3B Fig). These observations strongly suggest that a reduction in AMPKα expression level underlies the gut clearance phenotypes of Pngl -/larvae and contributes to the lethality of these animals independently of the BMP signaling defects in these animals. The severe reduction in AMPKα level in Pngl -/midguts (Fig 2A and 2B) and the previous report on the role of AMPKα expressed in the visceral muscle in promoting midgut peristalsis [20] together suggest that increasing the level of AMPKα in the visceral muscle by AMPKα Dp is sufficient for the rescue of both Pngl -/gut clearance and lethality. However, since this duplication harbors seven of AMPKα's neighboring genes and because it should increase AMPKα expression in all cell types, we also performed tissue-specific rescue experiments. Overexpression of wild-type AMPKα (AMPKα WT ) in the mesoderm by how 24B -GAL4 restored gut clearance and peristalsis activity in Pngl -/larvae (Fig 2F and 2G). Moreover, almost 46% of Pngl -/animals reached adulthood upon overexpression of AMPKα WT by how 24B -GAL4 (Fig 2H). Of note, overexpression of a constitutively active form of AMPKα (AMPKα T184D ) in the mesoderm resulted in the rescue of Pngl -/gut clearance and lethality phenotypes to an extent similar to that of AMPKα Dp and AMPKα WT overexpression (S5A and S5B Fig). Together, these data indicate that a reduction in AMPKα expression level in mesodermal cells results in the gut clearance phenotypes observed in Pngl -/animals and contributes to their lethality. Furthermore, the similarity between the level of rescue conferred to Pngl -/animals by wild-type and constitutively active forms of AMPKα and the normal pAMPKα/AMPKα ratio in Pngl -/midguts strongly suggest that phosphorylation of AMPKα are not impaired in these animals.

Pngl -/midguts show abnormal mitochondrial morphology and impaired energy metabolism, both of which are rescued by restoring AMPKα expression
Previous work in C. elegans and mammalian cells has suggested that loss of NGLY1 leads to a reduction in mitochondrial oxidative phosphorylation and an increase in oxidative stress [12]. Therefore, we examined whether Pngl -/larvae exhibit any defects in mitochondrial morphology and energy metabolism and if yes, whether these defects can be improved by increasing AMPKα levels. We performed TEM on control and mutant midguts and classified the visceral muscle mitochondrial morphology into three different categories: normal, mildly damaged (disorganized cristae structures), and severely damaged (disrupted cristae structures) (Fig 3A, TEM images). Based on this analysis, around 70% of mitochondria in wild-type larvae looked normal, with the remaining 30% mostly showing mild damage ( Fig 3A). However, less than 20% of mitochondria in Pngl -/larval midgut musculature were categorized as normal, with 60% mildly damaged and~20% severely damaged ( Fig 3A). The overall mitochondrial morphology in Pngl -/larvae is improved by adding a genomic copy of AMPKα and more so upon mesodermal overexpression of AMPKα WT , strongly suggesting that reduced AMPKα level might contribute to this phenotype.
In agreement with mitochondrial defects reported for Ngly1 mutants in other organisms [12,13], Pngl -/midguts exhibited a statistically significant reduction in ATP levels ( Fig 3B). One copy of the AMPKα Dp fully rescued this phenotype, similar to one copy of Pngl Dp ( Fig  3B). Loss of Pngl also induced a high level of oxidative stress in the larval midgut, as evidenced by a dramatic increase in the level of oxidized-dihydroethidium (DHE) fluorescence, which is an indicator of reactive oxygen species (ROS) (Fig 3C and 3D). Again, adding one genomic copy of AMPKα strongly reduced the ROS marker in Pngl -/midguts ( Fig 3E). The AMPKα Dp is likely to increase AMPKα expression in all cell types that normally express this gene. Therefore, the rescue conferred by this reagent did not allow us to distinguish the contribution of AMPKα expressed in midgut visceral musculature versus endodermal cells to the observed phenotypes. To address this issue, we performed cell type-specific rescue experiments. Overexpression of AMPKα WT in Pngl -/larvae using a mesodermal driver fully rescued ATP levels and resulted in a strong reduction in ROS levels (Fig 3F and 3G). However, only a modest rescue in ATP level and a mild reduction in ROS levels were observed upon endodermal overexpression of AMPKα WT (Fig 3F and 3H). These observations suggest that reduced AMPKα Restoration of AMPKα levels suppresses food accumulation phenotype, gut peristalsis defects, and lethality in Pngl mutants. (A) qRT-PCR assays show a significant decrease in the expression level of AMPKα mRNA in the midguts of Pngl mutant larvae (Pngl ex14/ex14 and Pngl ex18/ex18 ). (B) Western blotting and quantification of phospho-AMPKα (pAMPKα) and AMPKα expression in larval midguts show reduced expression in Pngl mutants. Note that the pAMPKα/ AMPKα ratio is comparable in all three genotypes. (C-E) Adding a genomic AMPKα duplication (AMPKα Dp) in Pngl -/animals improved their gut clearance (C) and midgut contraction defect (D), and rescued their lethality by~44% (E). (F-H) AMPKα overexpression in Pngl -/animals using a mesodermal driver (how 24B -GAL4) improved their gut clearance (F), rescued their midgut contraction defect (G) and rescued their lethality by~46% (H). The x-axis in C and F shows hours (h) after egg laying. Data in A-C and F represents mean ± SD of three independent experiments. Each circle in D and G represents a larva. Animal numbers for each genotype from left to right are 200 and 301 in (E), and 200 and 321 in (H). Significance is ascribed as � P<0.05, �� P<0.01 and ��� P<0.001. NS, not significant.
Using a null allele of AMPKα and transgenic overexpression studies, it was previously reported that expression of AMPKα in the visceral mesoderm is required for midgut peristalsis [20]. If a reduction in the mesodermal expression of AMPKα plays a causative role in the impaired energy homeostasis in Pngl -/midguts, we would expect to see similar phenotypes upon AMPKα knock-down in a Pngl +/+ background. To examine this notion, we used a previously verified UAS-AMPKα RNAi strain [21] to decrease its expression in the mesoderm. In agreement with the above-mentioned report [20], AMPKα RNAi in the mesoderm resulted in abnormal mitochondrial morphology ( Fig 3A) and impaired gut clearance ( Fig 3I). Moreover, midguts from animals with mesodermal AMPKα RNAi showed a reduction in ATP level and an increase in ROS comparable to that observed in Pngl -/midguts (Fig 3J-3L). Altogether, these data support the conclusion that reduced AMPKα expression in the visceral mesoderm plays a major role in the impaired energy homeostasis and gut clearance phenotype observed in Pngl -/animals, which contribute to the lethality observed in these animals.

Reduced AMPKα expression and impaired energy homeostasis in Ngly1 -/mouse embryonic fibroblasts
To examine whether AMPKα plays a conserved role downstream of NGLY1, we used Ngly1 -/mouse embryonic fibroblasts (MEFs) established from animals homozygous for the Ngly1 tm1.1Tasuz allele [22]. Mammalian genomes harbor two AMPKα genes: AMPKα1 (Prkaa1) and AMPKα2 (Prkaa2) [23]. qRT-PCR experiments showed a significant reduction in the expression of both genes in Ngly1 -/-MEFs compared to controls (Fig 4A) but not in the mRNA levels of AMPKβ1, AMPKβ2, AMPKγ1 and AMPKγ2 (S6 Fig). Immunoblotting with antibodies that detect both α1 and α2 proteins [24] showed a significant reduction in AMPKα levels and the pAMPKα/AMPKα ratio in Ngly1 -/-MEFs (Fig 4B and 4C). These observations suggest that in addition to reduced AMPKα expression, AMPKα phosphorylation might also be impaired in Ngly1 -/-MEFs. Indeed, the mRNA level of liver kinase B1 (Lkb1), which encodes a major AMPKα kinase [17], shows a modest but statistically significant reduction in Ngly1 -/-MEFs (S6 Fig), suggesting a potential mechanism for reduced AMPKα phosphorylation in these cells.
Similar to midguts from Pngl -/larvae, Ngly1 -/-MEFs showed a significant increase in the level of ROS and a significant reduction in the level of ATP (Fig 4D and 4E). We next examined whether increasing AMPK signaling can improve these phenotypes in Ngly1 -/-MEFs. For these experiments, we used a pharmacological approach and incubated the cells with two  different compounds known to activate AMPK signaling: metformin and PT1 [25][26][27]. Incubation of both wild-type and Ngly1 -/-MEFs with each of these compounds resulted in a significant increase in the pAMPKα levels without changing the AMPKα expression level (Fig 4B  and 4C). PT1 was more effective than metformin in increasing pAMPKα levels in Ngly1 -/-MEFs, although it showed a comparable effect in control MEFs (Fig 4B and 4C). Both compounds reduced oxidative stress in Ngly1 -/-MEFs ( Fig 4D). Moreover, PT1, but not metformin, increased the ATP levels in Ngly1 -/-MEFs (Fig 4E), potentially because PT1 directly activates AMPK signaling but metformin indirectly activates AMPK signaling by interfering with the function of mitochondrial respiratory chain complex I [25,26]. Together, these results suggest that reduced AMPK signaling contributes to impaired energy homeostasis in Ngly1 -/-MEFs.

The impaired energy metabolism and reduced AMPKα levels in Ngly1/Pngl mutant contexts are not due to NFE2L1-related proteasome and mitophagy defects
Previous work in C. elegans and mammalian cell lines has shown that NGLY1 (PNG-1 in worms) is essential for deglycosylation and activation of NFE2L1/SKN-1 [9][10][11]. In response to proteasome dysfunction, NFE2L1/SKN-1 is activated and induces the transcription of proteasomal genes to restore proteasome function, a process called the "proteasome bounce-back response" [10,11,28]. Transcriptome analysis of adult Drosophila upon RNAi-mediated Pngl knock-down also showed proteasomal gene downregulation [29]. In addition, Pngl -/larvae were 25 times more sensitive to the toxic effects of the proteasome inhibitor bortezomib compared to Pngl +/larvae [30]. These observations indicated that similar to its mammalian and worm homologs, Pngl promotes proteasomal gene expression in flies. Given the decreased AMPKα expression and pAMPKα levels in both Pngl -/larval midguts and Ngly1 -/-MEFs observed in our experiments, we asked whether impaired NFE2L1 activation and proteasome activity contribute to these phenotypes.
In agreement with a previous report [28], loss of Nfe2l1 in MEFs impaired the proteasome bounce-back response normally seen upon treating the cells with the proteasome inhibitor MG132 (Fig 5A). This is very similar to the impaired proteasome bounce-back response observed in Ngly1 -/-MEFs upon bortezomib treatment (S7A Fig) and is in agreement with the notion that impaired NFE2L1 activation underlies the proteasome dysfunction observed upon loss of Ngly1 and its worm homolog [9,11]. However, Nfe2l1 -/-MEFs did not show any decrease in AMPKα and pAMPKα levels ( Fig 5B). Moreover, MG132 treatment did not alter the AMPKα and pAMPKα levels in control and Nfe2l1 -/-MEFs (Fig 5B). It has recently been shown that a chemical activator of the NFE2L1 homolog NFE2L2 called sulforaphane (SFN) [31] is able to bypass the loss of NFE2L1 activity in Ngly1 -/-MEFs and restore the proteasome bounce-back response in these cells [13]. As expected, SFN treatment led to a robust increase in proteasomal gene expression in Ngly1 -/and control cells (Fig 5C). However, this was not accompanied by an increase in the mRNA (S7B Fig) and protein levels of AMPKα and pAMPKα in SFN-treated Ngly1 -/cells (Fig 5D). SFN treatment has also been shown to promote mitophagy in Ngly1 -/-MEFs by enhancing the expression of mitophagy-related genes [13]. In agreement with this report, we observed a robust increase in the expression of mitophagy-related genes in Ngly1 -/-MEFs treated with SFN ( Fig 5E). However, neither ATP nor ROS levels were rescued in these cells upon SFN treatment (Fig 5F and 5G). Together, these data suggest that the reduced AMPKα expression and impaired energy metabolism in Ngly1 -/-MEFs are not due to NFE2L1-related defects in the expression of proteasome and mitophagy genes.
We next performed similar experiments in Drosophila. Compared to control larvae, Pngl -/larvae showed a strong reduction in the expression of a number of proteasomal genes even without treatment with a proteasomal inhibitor (Fig 5H). These data are in agreement with RNA-seq results from adult Pngl knock-down animals [29] and indicate a critical role for Pngl in maintaining proteasomal gene expression in flies. Adding SFN to the fly food resulted in a robust increase in proteasomal gene expression in both control and Pngl -/larval midguts and restored the expression of these genes in Pngl -/midguts to levels comparable to those in control animals without SFN (Fig 5H). However, SFN was not able to restore AMPKα gene expression in Pngl -/midguts ( Fig 5H). Moreover, the SFN-fed Pngl -/larvae still showed impaired gut clearance ( Fig 5I). Finally, SFN did not rescue the lethality of Pngl -/animals beyond the~1% Mendelian ratio usually observed in these animals (n = 150). Altogether, the MEF and Drosophila data provide compelling evidence that the reduced AMPKα expression observed upon loss of Ngly1/Pngl and its functional consequences are not due to impaired NFE2L1 activation.

Reduced AMPKα expression and reduced cellular respiration in NGLY1 deficiency patient fibroblasts
To examine whether AMPKα expression and AMPK activation (measured by pAMPKα level) are also affected in NGLY1 deficiency patients, we performed qRT-PCR and immunoblotting on two patients with different NGLY1 allelic combinations and control fibroblasts. Similar to the other models examined in this study, a significant decrease was observed in the expression levels of AMPKα1 and AMPKα2, but not AMPKβ1, AMPKβ2, AMPKγ1 and AMPKγ2, in NGLY1 deficiency patient fibroblasts compared to NGLY1 +/+ fibroblasts isolated from a healthy control and NGLY1 +/fibroblasts isolated from two NGLY1 deficiency parents (Fig 6A  and S8A Fig). Furthermore, both AMPKα and pAMPKα levels and the ratio of pAMPKα/ AMPKα in NGLY1 deficiency patient fibroblasts are significantly decreased compared to parent fibroblasts (Fig 6B and 6C). These data indicate that regulation of AMPKα expression is an evolutionarily conserved function of NGLY1. The data also suggest that similar to Ngly1 -/-MEFs, patient fibroblasts might be defective in AMPKα phosphorylation. Of note, unlike Ngly1 -/-MEFs, the patient fibroblasts do not show reduced LKB1 levels (S8B Fig). To examine whether NGLY1 deficiency patient fibroblasts exhibit defects in mitochondrial respiration, we evaluated the oxygen consumption rate (OCR) in patient and control fibroblasts. The basal OCR was significantly reduced in both patient fibroblasts compared to an NGLY1 +/+ control (Fig 6D). Treating the patient cells with PT1 fully rescued the basal OCR in both cell lines. The patient cells also showed a significant reduction in maximal OCR, which was rescued by PT1 treatment in both cell lines (Fig 6E). Together, these observations suggest that reduced AMPKα expression contributes to the cellular respiration defects in NGLY1 deficiency patients and could be considered as a target to improve this phenotype.

Discussion
The discovery of recessive NGLY1 mutations in patients with a multi-system developmental disorder has prompted a series of studies on the consequences of loss of this enzyme in several cellular and animal model systems. Perhaps the most-studied cellular process downstream of NGLY1 is the NFE2L1-mediated regulation of the proteasomal gene expression, which depends on NGLY1 in both invertebrates and mammals [9][10][11]29,30]. In addition, our work in Drosophila, MEFs and mouse embryos has identified the Drosophila Dpp and its mammalian homolog BMP4 as direct, biologically relevant targets of Pngl/NGLY1 and has shown that impaired BMP signaling in the visceral mesoderm of Pngl -/larvae results in specific developmental abnormalities in their midgut and contributes to their lethality [16,19]. Our current data indicate that loss of N-glycanase 1 also results in a significant reduction in the level of AMPKα mRNA in the fly larval midgut, MEFs, and human patient fibroblasts, accompanied by reduced AMPKα and pAMPKα protein levels in all three models. We had previously reported that in addition to the tissue-specific loss of BMP signaling in the midgut, Pngl -/larvae exhibit a food accumulation phenotype (failure to empty the gut) that cannot be explained by loss of BMP signaling [16]. We now show that this phenotype is associated with impaired energy homeostasis in the midgut, and that restoring AMPKα expression rescues the food accumulation phenotype and allows 40-45% of Pngl -/larvae to reach adulthood, compared tõ 1% without rescue. In addition, AMPKα RNAi in the mesoderm recapitulates the gut clearance and energy homeostasis phenotypes of Pngl -/larvae. Moreover, pharmacological activation of AMPK signaling significantly improves the energy homeostasis defects observed in NGLY1-deficient MEFs and patient fibroblasts. Together, these observations suggest that reduced AMPK signaling underlies some of the phenotypes observed in NGLY1-deficient models. It is worth noting that in Ngly1 -/-MEFs and patient fibroblasts, but not in Drosophila visceral mesoderm, the pAMPKα/AMPKα ratio was also decreased compared to controls. These observations suggest that in addition to reduced AMPKα levels, NGLY1-deficient mammalian cells might also be defective in AMPKα phosphorylation (activation).
Loss of N-glycanase 1 in worms, MEFs, and fibroblasts and muscle biopsies from NGLY1 deficiency patients results in functional abnormalities in mitochondria, including a reduction in oxidative phosphorylation, basal OCR and maximal OCR, and an increase in oxidative stress [7,12,13]. Moreover, Ngly1 -/-MEFs showed mitochondrial fragmentation [13]. In agreement with these observations, we found that mitochondria in the midgut visceral muscles of Pngl -/larvae exhibit abnormal cristae structure, accompanied by reduced ATP levels and increased oxidative stress. Importantly, increasing the AMPKα expression level in Pngl -/animals partially rescued the mitochondrial morphology in the visceral muscle, significantly reduced the level of reactive oxygen species, and fully restored the ATP levels in the midgut. Moreover, pharmacological activation of AMPKα reduced reactive oxygen species and increased ATP levels in MEFs, and restored the basal and maximal OCR in fibroblasts from NGLY1 deficiency patients. Together, these observations indicate that enhancing AMPK signaling improves mitochondrial energy metabolism in several N-glycanase 1-deficient contexts.
Given the impairment of NFE2L1-mediated gene regulation in all N-glycanase 1-deficient animal models tested so far, we hypothesized that the reduction in AMPKα levels observed in our models might result from the loss of NFE2L1's transcriptional activity. However, Nfe2l1 -/-NGLY1 deficiency patient fibroblasts compared to two independent heterozygous fibroblasts from parents. (C) Graphs showing AMPKα/tubulin, pAMPKα/tubulin and pAMPKα/AMPKα ratios based on quantification of Western blots in (B) and two other independent experiments. (D-E) Basal (D) and maximal (E) oxygen consumption rates are reduced in NGLY1 deficiency patient fibroblasts compared to control (NGLY1 +/+ ) fibroblasts and are rescued upon PT1 treatment. Data represent mean ± SD of three independent experiments. Significance is ascribed as � P<0. 05 and wild-type MEFs showed comparable levels of AMPKα and pAMPKα, despite the impaired proteasome bounce-back response in Nfe2l1 -/-MEFs. Moreover, although treatment with the NFE2L2 activator SFN increased the expression of proteasomal genes in the midguts of Pngl -/larvae and in Ngly1 -/-MEFs, it failed to restore normal AMPKα levels in these models. Lastly, treating wild-type MEFs with a proteasome inhibitor did not affect AMPKα and pAMPKα levels. Together, these data indicate that the reduced AMPKα observed in NGLY1-deficient contexts is not caused by loss of NFE2L1 activity or impaired proteasome function. The mitochondrial fragmentation phenotype observed in Ngly1 -/-MEFs was shown to primarily result from impaired mitophagy due to loss of NFE2L1 activity and was significantly improved upon treating the cells with SFN [13]. However, SFN treatment failed to suppress the energy metabolism defects in Ngly1 -/-MEFs, even though it rescued the mitophagy gene expression in these cells. Finally, adding SFN to the fly food did not lead to any rescue of the AMPKα-dependent food accumulation and lethality in Pngl -/larvae. Taken together, our data suggest that the impairment of energy metabolism observed upon loss of NGLY1 is caused by reduced AMPKα activity and is independent of NFE2L1-related defects in proteasome and mitophagy.
Our previous report [16] and current data (S3 Fig) demonstrate that the loss of BMP signaling observed in Pngl -/midguts does not cause the food accumulation phenotype in these animals. Moreover, the data presented in the current study indicate that loss of NFE2L1 activation cannot explain the reduced AMPKα expression in NGLY1-deficient models. These data suggest that NGLY1 regulates AMPKα levels independently of BMP and NFE2L1 pathways. The mechanisms that regulate the level of AMPKα mRNA are not well understood. MicroRNA 148b (miR-148b) and miR-301a are reported to negatively regulate AMPKα levels in pancreatic cancer and osteosarcoma cell lines, respectively [32,33]. A recent study has shown that TP63 is recruited to the AMPKα1 regulatory region and directly activates AMPKα1 transcription in human mammary gland cells [34]. However, to our knowledge, neither TP63 nor the above-mentioned microRNAs have been linked to NGLY1 so far. In fact, transcription factors and other proteins involved in mRNA stability are not N-glycosylated and therefore cannot be direct targets of NGLY1 (with NFE2L1 as a rare exception). How does then loss of NGLY1 lead to reduced AMPKα levels? Ngly1 -/-MEFs have been shown to accumulate cytosolic aggregates of a model ERAD substrate harboring N-linked N-acetylglucosamine monosaccharides (N-GlcNAc) [22], a type of glycan normally not seen in the cytosol. Accumulation of proteins harboring N-GlcNAc can potentially interfere with the function of O-GlcNAc, which is the major type of glycosylation on nucleocytoplasmic proteins and regulates various cellular processes including transcription and mitochondrial activity [35]. Therefore, one possibility is that NGLY1 regulates AMPKα level indirectly through impaired O-GlcNAc signaling. Another possibility is that NGLY1 is involved in the quality control of an N-glycosylated cell surface receptor that is upstream of AMPKα transcription. The molecular mechanism through which NGLY1 regulates AMPKα levels remains to be determined.

Generation of the Pngl genomic transgene (duplication)
To generate a Pngl genomic rescue transgene, we purchased the bacterial artificial chromosome CH322-03O13 [39] from BACPAC Resources (https://bacpacresources.org/). The 21,853-bp insert present in this construct only contains the full coding and flanking sequences of two genes: Pngl and Actin 42A. The construct was integrated into the VK27 docking site on the third chromosome [40]. The injection to generate the transgene was performed by Best-Gene, Inc.

Lethality rescue assays (eclosion tests)
To test the lethality rescue, we scored the number of eclosed progeny and calculated the estimated total number based on Mendelian inheritance for each genotype. The observed/ expected ratio is reported as a percentage.

Gut clearance assay
To examine gut clearance, 3rd instar larvae were transferred to standard food mixed with 1% FD&C No. 1 Blue, which is a synthetic food color. Wandering stage larvae were collected from the side of the vial with a paintbrush, transferred to wet Whatman paper in a petri dish, and were monitored for the presence of the blue color in the gut until puparium formation. About 30 larvae were scored for each genotype.

Midgut contraction assay
To check the frequency of gut contractions, the intact larval midguts from each genotype were dissected in freshly-made PBS at room temperature without disrupting the attached tissues. After dissection, the number of gut contractions per minute was scored under a light microscope by visual inspection. At least 15 animals per genotype were scored.

qRT-PCR assays
Total RNA was extracted from 3 larval midguts with Trizol (Invitrogen) and dissolved in 25 μL of RNase-free water. cDNA was then synthesized from 1 μg total RNA using amfiRivert II cDNA Synthesis Master Mix (R5500, GenDEPOT), and qPCR was carried out using amfiSure qGreen Q-PCR Master Mix, Low ROX (Q5601, GenDEPOT). Expression levels were normalized to internal control: actin for fly midguts and MEFs, and GAPDH for human fibroblasts. The following oligonucleotide sequences were used to assess target genes expression (5' to 3'; f, fly; m, mouse; h, human). The following primer pairs were obtained from previous reports: fly AMPKγ [41], mouse proteasomal and mitophagy genes [13], mouse and human AMPK subunits [42], mouse Lkb1 [43], and human LKB1 [44].

Measurement of ATP levels
ATP levels were measured using luciferase-based ATP determination kit (Molecular Probes; A22066) by following the manufacturer's descriptions. Briefly, larval midgut or MEFs were lysed in homogenizing buffer [6 M guanidine HCL, 100 mM Tris (pH 7.8), 4 mM EDTA] and then boiled for 5 min and centrifuged for 3 minutes at 4˚C. Supernatant was diluted (1:500) and 10 μL of it was added to 100 μL of luciferase-reaction buffer to start the reaction in plate reader. Luminescence was measured by a LUMIstar OPTIMA microplate reader (BMG LAB-TECH). ATP levels were determined by comparing luciferase measurements to ATP standard curve. For relative quantification, ATP levels in different groups were normalized by their respective protein levels.

Assessment of ROS level
As a measure of oxidative imbalance, ROS levels were detected using the cell permeable dye dihydroethidium (DHE) following an online protocol [46] with minor modifications. When oxidized, DHE forms 2-hydroxyethidium, which intercalates within DNA molecules and generates a bright red fluorescent signal in the nucleus. For the in vivo detection of ROS levels by DHE, Drosophila larval midguts were dissected out in Schneider's medium and incubated with 30 μM DHE for 5 minutes in the dark at room temperature. After 1X PBS wash, midguts were mounted and imaged for oxidized DHE under Leica TCS-SP5 microscope. For the detection of ROS levels in cell lines, MEFs were incubated in 15 μM DHE for 15 minutes. After washing in 1X PBS, fluorescence intensity was measured in a FlexStation 3 Microplate reader. Fluorescence intensity was normalized by the number of cells.

Measurement of the oxygen consumption rate
Oxygen consumption rates (OCR) for cells treated with PT1 and control cells were assessed using the Seahorse XF96 analyzer (Agilent Technologies, CA, USA). Wild-type and patient human fibroblasts were seeded into XF 96-well cell culture plates, and incubated overnight to allow attachment. Maximal OCR was measured after FCCP injection. Cells were treated with the AMPK activator PT1 (10 μM) for 48 hours. Vehicle alone (DMSO) control cells were processed in parallel. After 48 hours of incubation, cells were washed in pre-warmed XF assay media (or for OCR measurement, XF assay media supplemented with 10 mM glucose, 1 mM pyruvate, 2 mM L-glutamine and adjusted at 7.4 pH). Cells were then maintained in 175 μL/ well of XF assay media at 37˚C, in a CO 2 incubator for 1 hour. Measurements were normalized by protein content (Bradford assay). The data were analyzed by XF96 software and GraphPad Prism software, using one-way ANOVA and Student's t-test calculations. All experiments were performed in quintuplicate, three times independently.

Assessment of midgut musculature and muscle-mitochondrial morphology
To examine midgut musculature, 3rd instar larval midguts were stained with phalloidin for 1 hour. Confocal images were taken with Leica TCS-SP8 microscope. All images were acquired using Leica LAS-SP software. Amira 5.2.2 and Adobe Photoshop CS6 were used for image processing.
TEM of visceral muscles in larval midguts was performed using a Ted Pella Bio Wave processing microwave with vacuum attachments. Third instar larvae were dissected in 4% paraformaldehyde, 2% glutaraldehyde, 0.1 M sodium cacodylate, and 0.005% CaCl2 (pH 7.2). The samples were incubated in the same fixative for 3 days at 4˚C overnight. On the third day, the samples were processed in fix again, post-fixed in 2% OsO4, dehydrated in ethanol series and propylene oxide, and then embedded in Embed-812 resin (Electron Microscopy Sciences, Hatfield, PA). Individual larval midguts were then sectioned and stained in 1% uranyl acetate and 2.5% lead nitrate. TEM images of cross sections were taken using a JEOL JEM 1400 plus transmission electron microscope with an AMT XR-16 mid-mount 16 megapixel digital camera.

Statistical analysis
Statistical analysis was performed by Student's t-test or by one-way ANOVA with the Dunnett's Post hoc multiple comparisons test using GraphPad Prism 6. Curves from gut clearance data were analyzed using Log-rank (Mantel-Cox) test. Data were plotted as mean ± SD of at least three independent experiments. P values are mentioned in Figure Legends Note that one copy of the dpp HA-3NQ does not improve the food accumulation phenotype in Pngl -/larvae, even though it fully rescues the dpp loss-of-function phenotypes in Pngl -/midguts [19]. (B) Bright images of the proximal midgut region of 3rd instar larvae of the indicated genotypes fed with bromophenol blue (BPB) are shown. Asterisks mark gastric caeca; dashed box marks the acid zone, which turns yellow upon BPB feeding. Note that one copy of AMPKα duplication (Dp) does not rescue the shortened gastric caeca and loss of acid zone in Pngl -/midguts, even though it significantly improves the gut clearance and contraction phenotypes (Fig 2). Scale bar in the top panel is 100 μm and applies to all panels. (TIF)