A set of candidate genes with potential roles in RQ-dependent metabolism

We previously developed an image-based assay that allows us to test whether any particular gene is required for RQ-dependent metabolism in C. elegans [34,37]. If so, it is a candidate target for anthelmintics since inhibiting its function would kill the STH. In brief outline, we expose C. elegans L1 larvae to potassium cyanide (KCN) for 15 hours—KCN prevents oxygen binding to Complex IV of the mitochondrial electron transport chain (ETC) and thus blocks oxidative phosphorylation. As we previously showed [34], this forces C. elegans to rely on anaerobic RQ-dependent metabolism and wild-type worms readily survive 15 hours of KCN treatment and rapidly recover full movement after removal of KCN. Worms that cannot make RQ, such as animals with mutations in the gene kynu-1, cannot survive and thus show no recovery of movement once KCN is removed [34]. We note that this assay works not just on L1 worms but can also be used to drive C. elegans into using RQ-dependent metabolism at other developmental stages (S1 Fig), suggesting our findings should likely be applicable to other life stages. This assay is powerful but does not have sufficient throughput to carry out genome-scale screens and thus we were forced to take a candidate gene approach. Fortunately, much is known about the pathways that may regulate or be required for RQ-dependent metabolism and thus we manually assembled a set of candidate genes to test for their involvement in RQ-dependent metabolism. These genes are all potential targets for anthelmintics and are predicted from previous literature to act in one of 4 different aspects of RQ-dependent metabolism (shown schematically in Fig 1B; genes are listed in Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Candidate genes with potential roles in RQ-dependent metabolism.

https://doi.org/10.1371/journal.pntd.0009991.t001

First, helminths must sense a requirement to switch from UQ to RQ-dependent metabolism [39] such as a change in environment from an aerobic niche to an anaerobic one. Interfering with this sensor could thus block the ability of parasites to activate RQ-dependent metabolism. Second, in response to the change in environment, parasites alter their quinone synthesis pathways to make more RQ and less UQ. For example, ~14% of the quinone in free-living miracidia-stage Fasciola hepatica is RQ, but ~93% is RQ in the adult stage in the host [40]. Blocking RQ synthesis will block RQ-dependent metabolism, and this is a key potential target for anthelmintics. Third, once RQ has been synthesised, it affects metabolism principally by directly docking to the quinone-coupled dehydrogenases (QDHs) that are linked to the mitochondrial ETC [41,42]. If a compound blocks the docking of RQ to the critical QDHs, this should block RQ-dependent metabolism. There are multiple QDHs that link to the ETC and we must test which are required for RQ-dependent metabolism and thus which are the potential targets. Finally, multiple other metabolic enzymes are required during RQ-dependent metabolism downstream of the interaction between RQ and QDHs [24]. By screening 25 candidate genes in our RQ-dependent metabolism assay, we aim to find a set of genes that are essential for C. elegans to carry out anaerobic RQ-dependent metabolism and suggest that their orthologues will act similarly in parasites. We describe the data from these assays on each of the four classes of genes below.

The switch to RQ-dependent metabolism is not a classic HIF-1-dependent hypoxia response

Parasitic helminths switch from aerobic UQ-dependent metabolism to anaerobic RQ-dependent metabolism when they are exposed to prolonged hypoxia. They must sense this change in environment and blocking the hypoxia response might prevent helminths from switching into anaerobic metabolism. The hypoxia-sensing machinery is thus an excellent potential target for anthelmintics.

The machinery for sensing hypoxia has been well-described in C. elegans [43–48] and requires the core hypoxia-regulated transcription factor HIF-1 [43,48] along with its negative regulators EGL-9 [44,49] and VHL-1 [47,50], and the atypical splicing regulator IRE-1 [51,52] and its target XBP-1 [52], which respond to an increase in unfolded proteins in the ER due to altered redox levels (Fig 2A). We used our assay to examine whether loss-of-function mutations in this core hypoxia-sensing pathway affected the ability of C. elegans to carry out RQ-dependent metabolism. If the hypoxia-sensing machinery is required for the switch from UQ-dependent aerobic metabolism to RQ-dependent anaerobic metabolism, mutants lacking the ability to sense hypoxia should be unable to survive 15 hours of KCN exposure.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The AMPK/mTOR pathway and not the hypoxia pathway is required for RQ-dependent metabolism.

(a) Model of the hypoxia response pathway. (b) Wild-type (N2) and mutant L1 worms for components of the hypoxia pathway were exposed to 200 μM KCN for 15 h. KCN was then diluted 6-fold and worm movement was measured over 3 h to track recovery from KCN exposure (see Materials and methods for details). kynu-1(e1003), a mutant that lacks RQ and cannot survive prolonged KCN treatment was used as a positive control. Data points show biological replicates of movement scores of ~150 worms per well after 3 h recovery, with 2–3 technical replicates for each experiment. (c) Activation of the nhr-57::GFP reporter in response to KCN is dose-dependent. L1 worms were treated with 50 or 200 μM KCN and imaged after 1.5 and 4.5 hours. Worms incubated in M9 buffer alone were used as a negative control. Images are representative of 10–15 worms imaged. (d-e) Wild-type N2 (black curve) and hif-1(ia4) (red curve) L1 worms were exposed to 50 (d) and 200 (e) μM KCN and their movement measured over 3 hours. Curves show the mean of 3 biological replicates. (f) aak-2 has a 50–60% reduction in recovery (1.00e-04 < p < = 1.00e-03) and rict-1 has a 75% reduction (1.00e-04 < p < = 1.00e-03). Wild-type (N2) and mutant L1 for components of the anoxia pathway were exposed to 200 μM KCN for 15 h. KCN was then diluted 6-fold and worm movement of ~150 worms per well was measured after 3 h to track recovery from KCN exposure (see Materials and methods for details). aak-2(ok524) is a loss-of-function mutant with a 408-bp deletion [32,33,60], while aak-2(rr48) is a semidominant-negative mutant with a substitution mutation [61]. Both rict-1(ft7) and rict-1(mg360) are loss-of-function mutants. rict-1(ft7) has a nonsense allele [62] while rict-1(mg360) has a missense allele [63].

https://doi.org/10.1371/journal.pntd.0009991.g002

We examined a set of 4 strains (see Table 1) containing mutations in the hypoxia sensing machinery and found no requirement for the core hypoxia-sensing pathway for C. elegans to survive long-term exposure to KCN (Fig 2B and S2 Fig). Worms homozygous for loss-of-function mutations in hif-1 can survive 15 hours of KCN treatment as well as wild-type animals, as can strains with loss of function mutations in ire-1 or xbp-1. Although xbp-1 is significantly different (1.00e-03 < p < = 1.00e-02) than N2, there is only a ~25% reduction in recovery which is not sufficient for a potential drug target. This was a surprising result to us as we had expected that KCN treatment would activate a hif-1-dependent hypoxia response. To test whether KCN addition caused activation of HIF-1 target genes, and thus invoked a hif-1-dependent hypoxia response, we examined induction of expression of nhr-57, a well characterized transcriptional target of HIF-1 [53]. We find that doses of KCN up to ~50 μM result in potent induction of nhr-57 expression as measured using a fluorescent transcriptional reporter (Fig 2C). Wild-type animals initially slow their movement when exposed to these lower doses of KCN, but subsequently recover full movement within ~3 hours; this acute recovery is hif-1-dependent (Fig 2D). However, at higher KCN doses (200 μM), we see no induction of nhr-57 expression and no difference in movement between wild-type and hif-1(ia4) mutant animals—both strains slow their movement and become immobile after ~90 mins (Fig 2E). We also see no increase in nhr-57 expression levels at higher KCN doses from RNA-seq data, suggesting the lack of GFP expression is unlikely to be due to quenching by KCN (S3 Fig). It thus appears that at lower KCN doses, worms invoke a hif-1 transcriptional response that allows them to recover from acute KCN treatment, and this recovery is not RQ-dependent. At higher doses of KCN, worms do not activate hif-1 but instead enter an immobile state where they rely on RQ-dependent metabolism. We suggest that the higher doses of KCN that are needed to drive worms to use RQ-dependent metabolism are mimicking anoxia rather than hypoxia. Consistent with this, C. elegans is known to be able to survive long periods of anoxia [54–58] and hif-1 is not required for this survival [59].

Our data show that the switch to RQ-dependent metabolism in C. elegans does not require the core hypoxia-sensing pathway. Compounds that block the core hypoxia-sensing pathway are thus unlikely to affect the ability of STHs to activate RQ-dependent metabolism and this is a useful negative result since the hypoxia sensing machinery appeared an obvious candidate target for this. Biologically, it is also important: if C. elegans do not use the core hypoxia pathway to induce the switch to RQ-dependent metabolism in hypoxic conditions, what is the switch? We address this in the next section.

RQ-dependent metabolism is activated in response to metabolic consequences of anoxia

We found that the core hypoxia-sensing machinery is not involved in the switch to RQ-dependent metabolism. We reasoned that if C. elegans is not sensing levels of oxygen directly it might sense the consequences of reduced oxygen instead. Hypoxia has diverse effects on the organism including changes in metabolism [64]. One immediate consequence of the inability of C. elegans to use the aerobic mitochondrial ETC is a reduction in efficiency of ATP generation and thus a likely drop in ATP levels and growth rate. Reduced ATP levels are sensed by AMP-activated kinase [65], a heterotrimeric kinase which in C. elegans requires the aak-1 and aak-2 encoded catalytic subunits [61,66,67] and we also note that AMPK activity is required for C. elegans to survive anoxia [54]. We find that C. elegans requires AMPK activity to efficiently activate RQ-dependent metabolism (Fig 2F and S2 Fig) suggesting that the key switch is not a lack of oxygen per se but the metabolic consequences of reduced oxygen levels. We also examined whether the mTOR pathway is required for the shift to RQ-dependent metabolism. The mTOR complex is required for sensing whether there is sufficient nutrient availability to drive growth and is regulated by AMPK phosphorylation [68–70]. rict-1 encodes the C. elegans orthologue of RICTOR [62,63], a core component of the mTOR complex and loss-of-function mutations in rict-1 (Fig 2F and S2 Fig) prevent RQ-dependent recovery from KCN. We thus find that C. elegans also requires the mTOR complex for efficient activation of RQ-dependent metabolism. We conclude that C. elegans activates RQ-dependent metabolism via both AMPK and mTOR in response to dropping ATP levels in anoxic conditions.

Finally, we examined whether sgk-1 is required for the switch to RQ-dependent metabolism in C. elegans. sgk-1 encodes the serum and glucocorticoid induced kinase [63,71] and is involved in maintaining mitochondrial homeostasis [72], in mitochondrial stress responses [73], and it affects sensitivity to paraquat which generates Reactive Oxygen Species [74,75]. We found that worms homozygous for the sgk-1(ok538) loss-of-function mutation show a reduced (~50%) but variable (1.00e-02 < p < = 5.00e-02) recovery after 15 hours of KCN exposure which suggests it is also partially required for C. elegans to use RQ-dependent metabolism (Fig 2F and S2 Fig).

Together, our findings here indicate that the switch from UQ-dependent aerobic metabolism to RQ-dependent metabolism is not regulated by the core hypoxia-sensing machinery but instead requires the AMPK/mTOR machinery that senses nutrient levels including ATP levels and the SGK-1 kinase that responds to a variety of cellular stresses. AAK-2, RICT-1 and SGK-1 are all required for C. elegans to enter RQ-dependent metabolism and thus these three proteins are potential targets. We next examined each amino acid sequence to identify whether there were specific residues that are identical across the helminths studied (Fig 1A) and that differ from those in a selection of their mammalian hosts (Fig 1A). Helminth species were selected so that 4/5 clades were represented and so that platyhelminthes were also included in the form of Fasciola hepatica (Fig 1A). We also included two molluscs and an annelid as their own group since they are also known to produce RQ [reviewed in 24]. Residues were considered of interest if they were conserved amongst the helminth (nematode+platyhelminth), C. elegans and mollusc+annelid group and were differentially conserved in mammalian hosts. We also checked whether any of those ‘helminth-specific’ residues reside at or near the active site.

AAK-1 and AAK-2 belong to a complex gene family whose evolutionary relationships were too complex for the method used here. In contrast, both RICT-1 and SGK-1 have clear evolutionary trees and conserved divergent residues between hosts and parasites. RICT-1 shows divergence between hosts and STHs in 22 residues. However, these are scattered throughout the protein and are not close to any binding interface or active site (S4 Fig). SGK-1 has 7 residues that differ between STHs and hosts and three that are near the active site. Two of these (residues 130 and 140 in C. elegans) are close to the active site while Phe136 is right at the active site [76,77]. This suggests it might be possible to generate an inhibitor that would block the STH SGK-1, disrupting the switch to RQ-dependent metabolism, without affecting the host orthologue. We note however that this active site change is not RQ-specific since it is present in species that do not make or use RQ (S5 Fig). These residues changes may or may not be sufficient to enable a helminth-specific SGK-1 drug. Taken together, while AMPK, mTOR, and SGK-1 all appear required for the efficient activation of RQ-dependent metabolism, only SGK-1 has STH-specific residues near the active site and these are not uniquely associated with RQ metabolism. We thus conclude that there is no high confidence target for anthelmintics within our candidate ‘sensor’ genes regulating the switch from UQ to RQ-dependent metabolism, though SGK-1 is perhaps still tractable.

Multiple enzymes required for RQ synthesis have helminth-specific active sites

We and others previously showed that UQ and RQ synthesis share a similar pathway [34,35,78]. The critical difference in UQ and RQ synthesis is the precursor—UQ synthesis in animals uses 4-hydroxybenzoate (4HB) whereas RQ synthesis starts with 3-hydroxyanthranilate (3HA) as the precursor [34,78]. A crucial early step in synthesis of these distinct quinones is the prenylation of either of these two precursors by the polyprenyltransferase COQ-2—the substrate choice by COQ-2 is the key event that dictates whether UQ or RQ will be synthesized [35]. We recently found that the choice of substrate for COQ-2 is determined by a single alternative splicing event which changes the enzyme core from a conserved 4HB-utilizing enzyme to a helminth-specific 3HA-preferring enzyme [35]. The splice variant that uses 3HA as its preferred substrate has two helminth-specific residues deep in its active site making it an attractive drug target since it should be possible to identify inhibitors that specifically inhibit this RQ-synthesising form.

3HA is generated by metabolism of tryptophan via the kynurenine pathway (Fig 3A) [79,80]. In C. elegans, the kynurenine pathway requires the genes tdo-2, kmo-1, amfd-1 and kynu-1 and we previously showed that kmo-1, amfd-1 and kynu-1 are required for RQ synthesis [34]. These 3 enzymes are thus potential targets for anthelmintics—if we could identify a STH-specific inhibitor of any of these, it should block 3-HA synthesis and thus block RQ synthesis. Comparative genomics and structural modeling showed that the sole potential candidate is KMO-1 where there is a helminth-specific change in a critical region required for binding of kynurenine. This Tyr334Phe is not a major change so we do not believe it to be a high priority target (S6 Fig). Our structural analyses did not find STH-specific residues at the active site of KYNU-1 (S7 Fig) while AMFD-1 protein sequences were too divergent from each other for our analysis pipeline. This suggests that it will be hard to identify STH-specific inhibitors to these 3 enzymes. We therefore turned our attention to Tryptophan Dioxygenase (TDO; TDO-2 in C. elegans) which catalyses the initial step in the kynurenine pathway [81–83], the oxidation of L-tryptophan to N-formyl-L-kynurenine.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. TDO-2 is required for RQ-dependent metabolism.

(a) Schematic of kynurenine pathway. 3-hydroxyanthranilate (3HA), a precursor for RQ synthesis, is derived from tryptophan. Tryptophan 2,3-Dioxygenase (TDO; TDO-2 in C. elegans) is required for the oxidation of L-tryptophan to N-formyl-L-kynurenine while AMFD-1, KMO-1, and KYNU-1 function downstream and were previously shown to be required for RQ synthesis. (b) Wild-type and tdo-2 loss-of-function mutants were treated with 200 μM KCN for 15 h followed by a 6-fold dilution with M9. Curves show the recovery from KCN treatment and represent the mean of 4 biological replicates and error bars are standard error. (c) Conserved divergent residues from host (blue), C. elegans (yellow), parasite (red) and mollusk/annelids (green) were identified using S3Det [38] with a cutoff of 2 (see Materials and methods). Residues are numbered according to C. elegans. One of the residues (arrow) is part of the PLD loop, a previously identified and evolutionarily conserved structural loop required for enzymatic activity [84]. (d) Crystal model of human TDO (PDB: 5TI9) with the single monomer highlighted in teal as visualized in Chimera [85]. (e) Single monomer of human TDO (PDB: 5TI9) in complex with Tryptophan and dioxygen with C. elegans sequence threaded over by homology (pink residues, circled in red). PLD loop is highlighted in yellow and the A vs P residue is circled in green and is close to tryptophan (orange) and to heme complexed with iron (red).

https://doi.org/10.1371/journal.pntd.0009991.g003

We found that like kmo-1, amfd-1 and kynu-1, tdo-2 is also required for survival in conditions where RQ-dependent metabolism is required, confirming that it is also a potential target for anthelmintics (Fig 3B). We therefore examined TDO sequence and structure in STHs and their mammalian hosts. We identified 4 residues that consistently differ between STHs and their mammalian hosts (Fig 3C). One is of particular interest since it sits at the active site close to the haem and tryptophan binding site in a small loop (Fig 3D and 3E). This is the same region where characterized TDO inhibitors are known to bind [84,86,87] and which is required for the enzymatic activity of TDO-2 in C. elegans [84]. STHs and mammalian hosts thus differ in a single critical residue at the active site of TDO and we suggest that this single amino acid difference may make it possible to develop TDO inhibitors that will be helminth specific and that could block the generation of RQ precursors and thus block the synthesis of RQ itself.

Finally, we also examined COQ-3 and COQ-5, two genes downstream of COQ-2 in the pathway for RQ synthesis. Both these genes were previously shown to be required for RQ synthesis and thus are potential targets [78]. However, RNAi knockdown of coq-3 and coq-5 did not greatly affect the ability of the worm to carry out RQ-dependent metabolism (S8 Fig). COQ-5 has no obvious helminth-specific residues that differ consistently between mammalian host and helminth. COQ-3 has a single such residue but this is far from the active site (S9 Fig) and thus is unlikely to result in a helminth-specific inhibitor. We conclude that these are not good anthelmintic targets.

We have thus identified two excellent targets for potential anthelmintics in the RQ synthesis pathway: COQ-2 from our previous study [35] and TDO-2 as we describe here. Both are required for RQ synthesis in C. elegans and thus for RQ-dependent metabolism and both have helminth-specific residues at the active site.

Two Quinone-coupled dehydrogenases are required for RQ-dependent metabolism

The direct consequence to energy metabolism of switching the main electron-carrying quinone from UQ to RQ is at the level of the quinone-coupled dehydrogenases (QDHs) that are connected to the mitochondrial electron transport chain (ETC) [24]. In most animals, these QDHs allow electrons to enter the ETC downstream from Complex I and onto the UQ pool to participate in oxidative phosphorylation. In helminths, however, these QDHs can act as either entry points or exit points for electrons. In aerobic conditions, they act as dehydrogenases where they shuttle electrons into the ETC by transferring electrons from substrates onto UQ, just as in other animals; however, in anaerobic conditions, they reverse direction and act as reductases where they transfer electrons from RQ and out of the ETC [23,26,28,88] (Fig 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. ETFDH is required for RQ-dependent metabolism.

(a) Schematic showing role of QDHs in the electron transport chain in UQ and RQ-dependent metabolism. (b) Requirement for QDHs in RQ-dependent metabolism. RNAi knockdown of QDH genes was performed by feeding dsRNA-expressing bacterial strains to L1 worms until they developed to fertile adults. Their L1 progeny were subjected to 15 h of 200 μM KCN treatment followed by a 3 h recovery (see Materials and methods). An RNAi clone targeting GFP was used as a negative control. Data show movement scores after 3 h recovery and represent biological replicates, with 2–3 technical replicates in each. (c) Conserved divergent residues from host (blue), C. elegans (yellow), parasite (red) and mollusk/annelids (green) were identified using S3Det [38] with a cutoff of 2 (see Materials and methods). Residues are numbered according to C. elegans. Phe437 (arrow) is conserved in RQ-containing species but is not present in hosts. (d) Crystal structure of porcine ETFDH (PDB: 2GMH) in complex with UQ5 (orange) as visualized in Chimera [85] with C. elegans sequence (pink residues) and H. sapiens sequence (cyan residues) threaded over by homology (residues from (c) circled in red). The Phe437Cys residue (green) is located near the quinone binding site [89] where both RQ and UQ must be able to bind in C. elegans and parasites.

https://doi.org/10.1371/journal.pntd.0009991.g004

QDHs are excellent candidate targets for anthelmintics since they are the direct site of action of RQ. If we could selectively inhibit the QDHs that are required for RQ-dependent metabolism, we should kill the helminth parasite. STH QDHs have different quinone-binding requirements to those of their mammalian hosts: the only quinone that interacts with the host QDHs is UQ but QDHs in STHs and other RQ-using animals must permit both UQ and RQ to dock at their quinone-binding pockets [26]. In Complex II, this difference in quinone binding requirement between host and STH is known to have driven the divergence of the quinone-binding pocket and this has allowed the identification of Complex II inhibitors [31,90] that are highly selective for STH Complex II. These inhibitors include flutolanil [89] and wact-11 [31] and these inhibitors bind directly at the quinone-binding pocket [31,90]. We reasoned that if it is possible to generate inhibitors that selectively bind to the quinone-binding pocket of Complex II, it may be possible to find inhibitors that selectively bind to the quinone binding pocket of other QDHs that use both UQ and RQ in helminths.

There are 5 QDHs in animals that are linked to the mitochondrial ETC. These comprise Succinate dehydrogenase (also known as Complex II) [26,91–93]; Electron-Transferring Flavoprotein (ETF) dehydrogenase [94–96]; Proline dehydrogenase [97–99]; Glycerol 3-phosphate dehydrogenase [100–102]; and Dihydroorotic acid dehydrogenase [103]. We note that there are other quinone-coupled enzymes in animals including sulphide:quinone oxidoreductase (SQRD-1 in C. elegans [104]) as well as orthologues of D- and L-hydroxyglutarate dehydrogenase [105–107] (D2HGDH and L2HGDH; F45D5.12 and Y45G12B.3 respectively). These are not thought to have any role in energy metabolism, our key focus here, and thus we have not pursued these in this study. We previously showed that Complex II is required for RQ-dependent metabolism in C. elegans and showed that helminth-specific residues in the quinone-binding pocket of Complex II in particular are important for RQ utilization[34]. However, we do not know which of the other QDHs may also be required and thus we examined the requirement for each QDH in our RQ-dependent metabolism assay.

We used RNAi to examine which of the QDHs are required for RQ-dependent metabolism using our RQ assay in C. elegans. We found that in addition to Complex II, ETFDH is also required for survival for 15 hours in KCN where worms rely on RQ-dependent metabolism (Fig 4B and S10 Fig). Complex II acts as a fumarate reductase in these conditions, transferring electrons from the RQ pool and generating succinate as a product. ETFDH links BCAA and fatty acid metabolism to the ETC during UQ-dependent aerobic metabolism [94,96,108,109]—when acting in reverse during RQ-dependent anaerobic metabolism, it transfers electrons from the RQ pool and generates BCAAs and short chain fatty acids including 2-methyl valerate and 2-methyl butanoate [95,110]. Since Complex II has been extensively studied as an anthelmintic target, we focused our attention on ETFDH.

We used sequence alignment to identify amino acids that consistently differ in ETFDH between hosts and helminths (Fig 4C) and find 5 such residues i.e. all helminths have one sequence, all mammlian hosts have a different sequence. We modelled C. elegans ETFDH (LET-721) on the Sus scrofa ETFDH (PDB:2GMH) crystal structure [89] to see where these helminth-specific amino acids sit. While some are dispersed around the structure, one of these host-helminth changes is located in the quinone-binding pocket of ETFDH (Fig 4D). It is a significant change: while mammalian hosts have a redox-sensitive cysteine at this position, helminths have the aromatic amino acid phenylalanine that may aid in charge delocalisation. This change right in the quinone-binding pocket suggests that just as it has been possible to identify potent helminth-specific succinate dehydrogenase inhibitors that bind to the quinone-binding pocket of Complex II, it should be possible to identify helminth-specific ETFDH inhibitors that bind to the quinone-binding pocket of ETFDH. We thus consider ETFDH to be another high confidence molecular target for anthelmintics.

Metabolic enzymes that change directionality in RQ-dependent metabolism

One of the key features of the switch from UQ-dependent aerobic metabolism to RQ-dependent metabolism is the rewiring of a subset of core metabolic reactions [24,111,112]. This often takes the form of a reverse in the direction of a reaction. The canonical example is that of Complex II which uses succinate as a substrate and generates fumarate as a product during aerobic conditions; in anaerobic conditions, it uses fumarate as a substrate and generates succinate as a product [92,113]. There is a similar reverse in reaction direction for ETFDH and as we showed in this paper, ETFDH is also critical for RQ-dependent metabolism. This switch in the direction of core reactions is not unique to quinone-coupled dehydrogenases and other helminth enzymes also switch their direction during RQ-dependent metabolism. The best characterized are those that involve the TCA cycle and specifically the reactions that lead to the generation of malate and fumarate and this is where we focus.

Malate metabolism is a key decision point between the aerobic and anaerobic TCA cycles in helminths. In aerobic conditions, malate is largely converted to oxaloacetate (OAA) by mitochondrial malate dehydrogenase (MDH-2 in C. elegans) as part of a normal animal TCA cycle [114]. In anaerobic conditions, however, malate is converted to fumarate by the enzyme fumarase (FUM-1 in C. elegans) acting in the reverse direction to the standard aerobic TCA cycle. Generation of fumarate is crucial since it acts as the substrate for RQ-coupled Complex II which is essential for RQ-dependent metabolism. Malate can be generated from a variety of substrates including from fumarate by fumarase (FUM-1), from OAA by reversal of malate dehydrogenases, from pyruvate either directly by the action of malic enzyme (MEN-1 in C. elegans [115]) or via OAA through the action of pyruvate decarboxylase (PYC-1 in C. elegans [115,116]). We used a combination of RNAi and classical genetics to test whether fum-1, mdh-1, mdh-2, pyc-1 and men-1 are required for RQ-dependent metabolism. In addition, we tested two genes, F36A2.3 and VF13D12L.3 that encode members of a distinct subfamily of malate dehydrogenases [117,118] that are typically found in archaeal genomes. Importantly for this study, these genes have no host homologues but are present in all STHs making them an attractive target.

As shown in Fig 5A, we find that none of the genes tested appear to be required for RQ-dependent metabolism in C. elegans. There are a number of caveats to this. First, fum-1 and mdh-2 are lethal by RNAi so we could not assess their involvement. This is not unexpected—they are required for the TCA cycle which is essential in both aerobic and anaerobic conditions. Second, F36A2.3 and VF13D12L.3 are closely related and the lack of a strong effect of these genes on RQ-dependent metabolism might be due to paralogue redundancy. We therefore used CRISPR genome engineering to make a presumptive null mutation in VF13D12L.3 and carried out RNAi against F36A2.3 in a strain homozygous for the VF13D12L.3 null allele to reduce activity of both paralogues at once—we saw no effect on RQ-dependent metabolism (S11 Fig). This could reflect the lack of requirement for these genes, but could also be the result of poor knockdown of F36A2.3 by RNAi. We suspect this to be the case since we used CRISPR engineering to make a presumptive null mutation in F36A2.3 and found that this is lethal and we were unable to generate a strain without balancers. The fact that the mutant phenotype is lethal but the RNAi phenotype is not suggests that the F36A2.3 RNAi is only giving a weak knockdown. It is intriguing to note that F36A2.3 does not show any reproducible strong effect on RQ-dependent metabolism by RNAi but has a variable phenotype (1.00e-2 < p < = 5.00e-2)—some biological repeats show strong effects, others very little. This is the sole gene for which we saw such variability and variable RNAi phenotypes can often result from weak or incomplete RNAi knockdown. We would thus not feel confident either excluding it as a possible target or including it as a high priority target based on this weak data. Overall, we conclude that we did not find that any of these metabolic enzymes to be definitely required for RQ-dependent metabolism in C. elegans.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Requirement for various metabolic enzymes for RQ-dependent metabolism.

(a) The involvement of various metabolic enzymes in RQ-dependent metabolism was tested using either loss-of-function mutants or RNAi knockdown. Wild-type, mutants, and RNAi-treated L1 worms were exposed to 200 μM KCN for 15 h followed by a 6-fold dilution, and worm movement was measured after 3 h recovery. For loss-of-function mutants, movement scores were compared to wild-type N2 worms, while for RNAi-treated worms, scores were compared to RNAi against a GFP control. Data show movement scores after 3 h recovery and represent biological replicates, with 2–3 technical replicates in each. (b) Evolutionary tree of the three types of malate dehydrogenase found in helminths as calculated by BEAST [119] (see Materials and methods) and visualized with FigTree. The colour of the circles reflects their posterior probability with the more blue being the most likely. Unlike hosts, helminths possess a third malate dehydrogenase which is more closely related to archaean MDHs (black).

https://doi.org/10.1371/journal.pntd.0009991.g005

More positively, the finding that F36A2.3 is an essential gene in C. elegans suggests that this malate dehydrogenase is a good target for anthelmintics. All animals have both cytoplasmic and mitochondrial malate dehydrogenases (mdh-1 and mdh-2 in C. elegans). However, a small number of animal species have an additional family of malate dehydrogenases that is closely related to those found in archaea (Fig 5B). Intriguingly, no mammalian STH hosts have this third class of malate dehydrogenase but all species that make RQ (nematodes, STHs, annelids, and molluscs) have these. Crucially, since there is no similar gene in any STH mammalian host and since F36A2.3 is an essential gene in C. elegans, it may be an additional target for new anthelmintics.

In addition to these changes in reaction direction, helminths can also truncate their TCA cycles to avoid the two steps where carbon atoms are lost in the form of CO₂. This may be beneficial during times of limited carbon source availability [120]. These steps occur in the serial conversion of isocitrate to alpha-ketoglutarate (α-KG) and α-KG to succinyl-CoA. The truncated TCA cycle that avoids these steps requires the isocitrate lyase ICL-1 to drive the glyoxylate shunt [121,122]. Plants, fungi and helminths can also carry out the glyoxylate shunt [123–125] but hosts have no ICL-1 homologues and thus cannot do this. ICL-1 has thus previously been proposed as a potential anthelmintic target [111,126]. We examined whether ICL-1 was required for RQ-dependent metabolism in C. elegans. As shown in Fig 5A, worms homozygous for a loss of function mutation in icl-1 were no different to wild-type animals in their ability to carry out RQ-dependent metabolism and we conclude ICL-1 is not required at least within the assays used here.

Together, we conclude that while there are extensive metabolic changes downstream of quinone-coupled dehydrogenases that occur during RQ-dependent metabolism, none of the enzymes involved appear to be individually required for efficient RQ-dependent metabolism. They are thus not worthwhile targets for drugs designed to block RQ-dependent metabolism in STHs. We suggest that this may be because there are many parallel routes for the generation of the key metabolite malate including through multiple different malate dehydrogenases and malic enzyme and that inhibiting any single route has little effect. The sole target we identified here is F36A2.3 which appears to be an essential gene rather than having a dedicated role in RQ-dependent metabolism. Since there is no mammalian homologue for this unusual malate dehydrogenase, it is potentially an excellent target for anthelmintics.

Functional significance of STH-specific residues

We identified 4 enzymes that are required for RQ-dependent metabolism and that also have helminth-specific residues at their active sites: Complex II, ETFDH, COQ-2e and TDO-2. These are all high confidence targets for anthelmintics since it should be possible to identify compounds that inhibit the helminth and not the host enzyme and thus block RQ-dependent metabolism without affecting the mammalian host. We previously showed that mutations that affect the helminth-specific binding pocket of Complex II can block RQ-dependent metabolism and wanted to specifically test the functional significance of the helminth-specific residues in the other three targets we identified. We did this in two ways.

First, we looked at the phylogenetic distribution of the amino acids that differ between STHs and their mammalian hosts. These differences at the active sites could have arisen through random genetic drift and have no functional significance. Alternatively, the amino acid changes seen at the active sites of the STH enzymes may be important for their function in RQ-dependent metabolism—if so we would expect to see the same residues in the other animal lineages that use RQ, molluscs and annelids. If they are simply the result of genetic drift, we would not expect this distribution. We see identical amino acids in the active sites of COQ-2e, TDO-2, and ETFDH in STHs, annelids, and molluscs but not in any of the mammalian hosts of STHs. This suggests that these residues may be functionally important for the ability to do RQ-dependent metabolism.

Second, we used genetics to directly test whether the identified residues are critical for the ability to carry out RQ-dependent metabolism. For COQ-2, we had previously shown that all RQ-using animals make an unusual splice form (COQ-2e in C. elegans) that includes the STH-specific residues [34], but did not confirm the functional importance of these two amino acids for RQ-synthesis. Here, we used CRISPR genome engineering to change these two residues in the COQ-2a isoform. COQ-2a has the pan-eukaryotic active site and only participates in UQ synthesis; we previously showed that animals that can only make the COQ-2a isoform cannot make RQ and thus cannot survive in conditions where RQ-dependent metabolism is required. However, here we generated a C. elegans strain in which animals cannot make the RQ-synthesising COQ-2e isoform but can only make a CRISPR-edited version of COQ-2a where two amino acids in COQ-2a are changed from the pan-eukaryotic consensus to the STH-specific residues in the COQ-2e splice form (Fig 6A). If these two residues are indeed sufficient for RQ synthesis, this strain that expresses only the edited COQ-2a form should now be able to make RQ and thus to survive in conditions where RQ is required. We find that mutant worms with this edited COQ-2a form can survive where RQ is required (Fig 6B), suggesting that this mutant can now make RQ. This demonstrates that these two STH-specific residues are indeed critical for RQ synthesis. In the case of TDO-2, we took advantage of an existing mutant, tdo-2(zg218) [84]. The critical STH-specific amino acid in the active site, P133 in C. elegans TDO-2, is in a short 3 amino acid loop (PLD, residues 133–135 in C. elegans). Crucially, the mutant strain tdo-2(zg218) has been engineered to delete only that PLD loop [84]. We find that tdo-2(zg218) animals with the deleted PLD loop cannot carry out RQ-dependent metabolism confirming the functional importance of this STH-specific region (Fig 6C). Finally for ETFDH, we also used CRISPR genome engineering to switch the single STH-specific amino acid in the quinone binding pocket to the mammalian host consensus. We find that this single change reduces the ability of C. elegans to carry out RQ-dependent metabolism, confirming its functional importance, though we note the effect is weaker (Fig 6D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Mutations in STH-specific residues affect ability to survive extended KCN treatment.

(a) Schematic of CRISPR-mediated engineering of a COQ-2 mutant. coq-2 exons 6a and 6e are alternatively spliced in a mutually exclusive manner in wild-type worms. Mutant worms with exon 6e deleted (coq-2Δ6e) are unable to synthesise RQ. Exon 6e contains 2 helminth-specific residues and mutations were generated in the coq-2Δ6e mutant, such that two residues in exon 6a are replaced with their helminth-specific counterparts (F204L, A243S). (b) Helminth-specific residues in COQ-2 are required for survival in KCN. L1 worms were treated with 200 μM KCN for 15 h followed by a 6-fold dilution, and worm movement was tracked for 3 h. The RQ-deficient coq-2Δ6e mutant cannot survive extended KCN treatment while the coq-2Δ6e F204L;A243S mutant (coq-2a(syb2949);coq-2e(syb1721)) can now recover from KCN treatment, suggesting restored ability to synthesise RQ. (c) The TDO-2 PLD loop is required for survival in KCN. Wild-type (N2, black) or tdo-2(zg218) mutant worms (blue) L1 worms were treated with 200 μM KCN for 15 h followed by a 6-fold dilution, and worm movement was tracked for 3 h (d) LET-721 helmint-specific quinone binding pocket is required for normal survival in KCN. Wild-type (N2, black) or let-721 mutant L1 worms were treated with 150–250 μM KCN for 15 h followed by a 6-fold dilution, and worm movement was tracked for 3 h. All curves show the mean of 4 biological replicates and error bars represent standard error.

https://doi.org/10.1371/journal.pntd.0009991.g006

Taken together, both the phylogenetic distribution of the active site amino acid changes and the direct functional testing show that not only do STHs have changes at the active sites of COQ-2e, TDO-2, and ETFDH, compared to their hosts but these active site changes are critical for their ability to use RQ-dependent metabolism. These three enzymes are thus high confidence targets for new anthelmintics.

Caenorhabditis elegans strains and maintenance

The F36A2.3 (syb769)/hT2, VF13D12L.3.1(syb243), coq-2a(syb2949); coq-2e(syb1721) and let-721(syb2817) mutant strains were generated by SunyBiotech (Fuzhou City, China) using the CRISPR-Cas9 system. The specific mutations were confirmed by DNA sequencing and details of each of these strains are provided in S1 Table. The coq-2e(syb1721) strain used was previously described [35] and was a kind gift from Prof. Gustavo Salinas. The tdo-2(zg216), tdo-2(zg217) and tdo-2(zg218) strains were a kind gift from Prof. Ellen Nollen [84]. The wild-type N2 strain and all other strains listed in Table 1 were obtained from the Caenorhabditis Genetics Center (University of Minnesota, USA), which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). All worm stocks were maintained at 20°C on NGM agar plates seeded with the E. coli strain OP50 as described elsewhere [146].

Image-based KCN assays

The 15 h KCN recovery assay was performed as previously described [34,37]. In outline, L1 worms were washed from NGM agar plates and isolated by filtration through an 11 μm nylon mesh filter (Millipore: S5EJ008M04). Approximately 150 L1 worms in M9 were dispensed to each well of a 96 well plate and an equal volume of potassium cyanide (KCN) (Sigma-Aldrich St. Louis, MO) solution was then added to a final concentration of 150–250 μM KCN. Upon KCN addition, plates were immediately sealed using an aluminium seal and incubated at room temperature for 15 h on a rocking platform. After 15 h, the KCN was diluted 6-fold by addition of M9 buffer. Plates were then imaged on a Nikon Ti Eclipse microscope every 10 min for 3 h.

After imaging, fractional mobility scores (FMS) were calculated using a custom image analysis pipeline [37]. For each strain, FMS scores for the KCN-treated wells were normalized to the M9-only control wells for the WT strain at the first timepoint. FMS scores of the non-WT wells were then normalized for each timepoint to the average of N2 (mutant) or HT115 treated with GFP RNAi (RNAi) wells that were also treated with 200μM KCN. For endpoint data, an average of the normalized FMS scores at 160, 170 and 180 min was taken for each biological replicate and 2–3 technical replicates were carried out for each experiment. Welsh’s t-test was used to test for the statistical significance between WT/GFP RNAi and the mutants/RNAi.

For the acute 3 h KCN assay, L1s were isolated and treated with KCN as described above. After KCN addition, the plates were immediately imaged every 5 min for 3 h. FMS scores of KCN-treated wells were then normalized to the M9-only control wells at each timepoint.

Analysis of the nhr-57::GFP reporter

The nhr-57::GFP reporter [53] was used to assay HIF-1 activity after KCN treatment. L1 worms were purified by filtration and treated with 40 μl KCN of various concentrations. Worms incubated in M9 alone were used as the control. After 1.5 h and 4.5 h of KCN treatment, worms were imaged using a Leica DMRE fluorescence compound microscope, with levamisole used as an anesthetic. All images were taken using the 40x objective.

RNAi experiments

RNAi-by-feeding was performed as previously described [146]. RNAi clones were grown overnight at 37°C in LB media with 1 mM carbenicillin and seeded onto NGM plates supplemented with 1 mM IPTG and 1 mM carbenicillin. For each RNAi knockdown, approximately 350 wild-type L1 worms were grown on NGM plates seeded with dsRNA-expressing bacteria. GFP-targeting dsRNA-expressing bacteria was used as a negative control. After one generation, worms were washed with M9 buffer and filtered to collect L1 progeny. These worms were then transferred to 96-well plates to be used in the KCN recovery assay as described above.

Sequence identification

To identify host, parasite, mollusk/annelid and invertebrate sequences, four search strategies were made using BLASTP (protein databases). The search strategies are detailed below. For each search, the first 100 sequences which met at minimum 30% query cover were selected for further processing. Sequences were retrieved from NCBI protein and nucleotide databases and from RCSB Protein Data Bank (PDB).

For parasites, each search was made in the non-redundant protein sequences (nr) database against Fasciola hepatica (taxid:6192), Ancylostoma ceylanicum (taxid:53326), Strongyloides ratti (taxid:34506), Brugia malayi (taxid:6279), Trichuris suis (taxid:68888), Haemonchus contortus (taxid:6289) and Ascaris suum (taxid:6253). For hosts as well as C. elegans, each search was made in the reference proteins (refseq_protein) database against Caenorhabditis elegans (taxid:6239), Ovis aries (taxid:9940), Bos taurus (taxid:9913), Mus musculus (taxid:10090), Sus scrofa (taxid:9823) and Homo sapiens (taxid:9606). For mollusks and annelids, each search was made in the nr database against Magallana gigas (taxid:29159), Biomphalaria glabrata (taxid:6526) and Capitella telata (taxid:283909). In order to capture a broad select of other invertebrates, the search strategy used the refseq_protein database included Insecta (taxid:50557), Arachnida (taxid:6854), Mollusca (taxid:6447), Crustacea (taxid:6657), Anthozoa (taxid:6101), Onychophora (taxid:27563), Xiphosura (taxid:6845), Cnidaria (taxid:6073), Echinodermata (taxid:7586) and Porifera (taxid:6040) while excluding Vertebrata (taxid:7742), Crassostrea (taxid:6564), Biomphalaria (taxid:6525), Capitella (taxid:51293) and Drosophilidae (taxid:7214).

The crystal structures were found on PDB by searching for the name of the enzyme and finding the best match, preferentially in a host or parasite and/or one bound with the ligand(s) or an inhibitor.

Sequence analysis

Sequences were aligned using Clustal Omega [147] with {msa} [148] then ModelTest-NG [149] was used to identify the substitution model and its parameters for BEAST [119]. Speciation was based on the Yule Process [150,151] and a random local clock was used. BEAST was run using the CIPRES Science Gateway [152]. Tracer was used to confirm that an effective sample size (ESS) had been reached [153] and the trees were summarized using TreeAnnotator with a burnin percentage of 10% and a posterior probability limit of 0.5. Trees were visualized using FigTree.

Based on the distribution of sequences in the tree, sequences were selected for each species. Two sets of sequences were selected for each protein; one included only host, mollusk/annelid, C. elegans and parasite sequences in addition to any further crystal structure sequences while the second also included the expanded invertebrate sequences. If a sequence was clustered with other sequences from the same species, only one sequence was selected with a preference for first PDB sequences and then XP sequences. If there were multiple clear gene families across species, one sequence per species was selected based on gene families. Gene families which included sequences from other proteins outside of the protein of interest were not included.

Selected sequences were realigned using Clustal Omega [147] and {msa} [148] then visualized in JDet [154]. S3det [38] was used with a cut-off of 2 with categories based on host, mollusk/annelid, C. elegans and parasite. In cases where there had been a gene family expansion, additional groups were used. These residues were also checked against the expanded invertebrate list in order to determine whether they were RQ-specific.

Protein analysis

Crystal structures were visualized using Chimera [85] and the C. elegans and H. sapiens (where needed) sequences were threaded by homology using Modeller [155,156]. Active sites were determined based on the original published crystal structures. Residues of interest were visualized relative to ligands and known active sites.