Introduction

The fluidity of cellular membranes is heavily influenced by the saturation level of the phospholipids composing the membrane. Phospholipids containing saturated fatty acid (SFA) tails are more tightly packed and therefore form more rigid membranes, while phospholipids with an abundance of unsaturated fatty acids (UFAs) are more loosely packed and result in fluid membranes [1,2]. Polyunsaturated fatty acids (PUFAs) themselves can also affect many different cellular processes and are precursors to both anti- and pro-inflammatory PUFA-derived signaling molecules called eicosanoids [36]. Imbalances between SFAs and PUFAs are associated with chronic diseases including coronary heart disease, diabetes, hypertension, and renal disease [4].

C. elegans PAQR-2, and its mammalian ortholog AdipoR2, promote the production and incorporation of PUFAs into phospholipids to restore membrane homeostasis. Whether this is the primary function of PUFAs in cells or organismal physiology is still not resolved. In particular, no unbiased forward genetic screens for suppressors of PUFA deficiency have been reported. Mammals are not able to endogenously synthesize PUFAs and must obtain omega-3 and omega-6 PUFAs from the diet, a fact known since 1930 [7,8]: linoleic acid (LA, 18:2n6) and alpha-linolenic acid (ALA, 18:3n-3) must be dietarily supplied and can be further desaturated and elongated into ≥20-carbon PUFAs used structurally or as precursors of signaling molecules [9]. In contrast, C. elegans expresses many desaturases and elongases that can convert dietary or de novo synthesized SFAs into a wide range of PUFAs: Δ9 desaturases are responsible for converting SFAs into monounsaturated fatty acids (MUFAs) by adding a first double bond, a Δ12 desaturase adds an additional double bond to transform MUFAs into LA, a PUFA with two double bonds, and Δ5 and Δ6 desaturases introduce additional double bonds to produced PUFAs with three, four or five double bonds [1012]. To better understand the essential roles of PUFAs in cells and whole organisms, we leveraged a mutant allele of the C. elegans Δ12 desaturase FAT-2, whose function is to convert oleic acid (OA, 18:1n9) into linoleic acid (LA, 18:2n6) [13] (Fig 1A). Because this is a critical step for PUFA production, worms devoid of FAT-2 activity (i.e. fat-2(-) null mutants) are not able to synthesize any PUFAs and are not viable. In contrast, the fat-2(wa17) allele produces a partially functional protein bearing a S101F substitution (Fig 1B-C): mutants homozygous for this allele produce <10% of the normal levels of PUFAs and are extremely slow growing and sickly but, crucially, are viable [9].

C. elegans fatty acid synthesis pathway and FAT-2 desaturase.

(A) Simplified pathway of fatty acid synthesis and desaturation in C. elegans. Boxes indicate the name of the enzymes, FAT-2 desaturase is indicated in a red box. Fatty acids of which the synthesis is dependent on FAT-2 are indicated in red. Fatty acid abbreviations are as follow: palmitic acid (PA), palmitoleic acid (POA), vaccenic acid (VA), stearic acid (SA), oleic acid (OA), linoleic acid (LA), alpha-linolenic acid (ALA), gamma-linolenic acid (GLA), stearidonic acid (STA), dihomo-gamma-linolenic acid (DGLA), eicosatetraenoic acid (ETA), arachidonic acid (AA), and eicosapentaenoic acid (EPA). (B) AlphaFold2 predicted FAT-2 structure with the serine at position 101 indicated with a red arrow. (C) Same structure as in B, zoomed in and angled to show that the S101 position that is mutated to phenylalanine in the fat-2(wa17) allele lies in a loop connecting two alpha helices.

Although there is no mammalian homolog of FAT-2 [14], the fat-2(wa17) mutant can still serve as a useful genetic model to reveal evolutionarily conserved roles of PUFAs in organisms, and to help identify mechanisms that can compensate for PUFA deficiency. Here, we began by characterizing the fat-2(wa17) mutant in terms of its ability to be rescued with dietary PUFAs or with mutations previously identified as suppressors of paqr-2(tm3410) mutant phenotypes that are attributed to membrane rigidity. We then performed an exhaustive forward genetic screen for fat-2(wa17) suppressors in the hope that the critical functions of PUFAs could be discovered and to identify mechanisms that allows cells/organisms to cope with PUFA deficiencies. This screen yielded ten fat-2(wa17) suppressor alleles that fell into two groups: mutations within fat-2 itself, and mutations in the hif-1 pathway that converge on inhibition of ftn-2 expression.

Results

Characterization of the fat-2(wa17) mutant

The severe growth defect of fat-2(wa17) mutants can be suppressed by the wild-type fat-2(+) allele carried on an extrachromosomal array, confirming that this growth defect is due to reduced fat-2 activity (Fig 2A). As expected, given their low amounts of PUFAs, fluorescence recovery after photobleaching (FRAP) shows that the membranes of intestinal cells in the fat-2(wa17) mutant are excessively rigid and indeed appeared as rigid as those of the paqr-2(tm3410) mutant characterized by an excess of SFAs in its phospholipids [15,16] (Fig 2B-C).

Characterization and rescue of fat-2(wa17).

(A) Introduction of the wild-type fat-2(+) allele on an extrachromosomal array rescues the fat-2(wa17) growth defect. (B-C) FRAP curve and Thalf value show that fat-2(wa17) has rigid membranes similar to paqr-2(tm3410) control. (D-L) The lengths of fat-2(wa17) worms grown from L1 stage for 72h in the indicated conditions; horizontal dashed lines indicate the approximate lengths of the synchronized L1s at the start of the experiments. n=20 for each genotype/condition. Error bars show the standard error of the mean. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Also as expected, low doses of dietary linoleic acid (LA, 18:2n6) fully rescued the fat-2(wa17) growth defect (Fig 2D). This rescue by LA is transient and does not last when the following generation is transferred back to NGM plates (Fig 2E), which is consistent with the rapid turnover of fatty acids in C. elegans [17]. Eicosapentaenoic acid (EPA, 20:5n3) is the longest and most unsaturated PUFA produced in C. elegans (Fig 1A) and is also able to rescue the fat-2(wa17) mutant, albeit requiring higher concentrations than LA (Fig 2F). Surprisingly, docosahexaenoic acid (DHA, 22:6n3), which is not produced by C. elegans, is also able to rescue the fat-2(wa17) mutant (Fig 2G). Temperature has a direct effect on membrane fluidity: given a constant phospholipid composition, lower temperatures cause rigidification while higher temperatures promote fluidity [18]. We found that the fat-2(wa17) mutant is growth-arrested when cultivated at membrane-rigidifying 15°C, and conversely shows improved growth at 25°C, which again is similar to earlier findings with the paqr-2(tm3410) mutant [19] (Fig 2H). However, cultivating the fat-2(wa17) mutant in the presence of two types of fluidizing agents that improve the growth of paqr-2(tm3410), namely the non-ionic detergent NP-40 or the MUFA oleic acid (OA, 18:1) [15], did not suppress the poor growth phenotype of the fat-2(wa17) mutant (Fig 2I-J). Conversely, membrane-rigidifying glucose (which causes a high SFA/UFA ratio in the dietary E. coli [20]) or the SFA palmitic acid (PA, 16:0) did not exacerbate the growth defect of fat-2(wa17) (Fig 2K-L). The observations that NP40 and OA do not rescue fat-2(wa17), and that membrane rigidifying conditions (dietary glucose or PA) were not detrimental to fat-2(wa17), suggest that membrane rigidification may not be the main cause of the fat-2(wa17) growth defects.

Lipidomic analysis of phosphatidylcholines (PCs) and phosphatidylethanolamines (PEs) in the fat-2(wa17) mutant confirmed its reduced levels of PUFAs relative to wild-type worms (Fig 3A; S1A Fig), with the largest loss observed in longer PUFAs, namely dihomo-γ-linolenic acid (DGLA; C20:3), arachidonic acid or eicosatetraenoic acid (AA or ETA, C20:4) and EPA (C20:5) (Fig 3B-C; S1B-C Fig). As expected, the substrate of FAT-2, i.e. OA (18:1), was greatly increased in the fat-2(wa17) mutant while exogenous addition of LA to these worms resulted in an increase of PUFAs. Transferring fat-2(wa17) from LA to NGM 6 hours prior to harvesting did not lessen this increase, suggesting minimal LA depletion during this time (Fig 3A-C; S1A-C Fig). Cultivation at 25°C did not result in any significant changes in the fat-2(wa17) mutant, indicating that the growth rescue seen at 25°C is not due to increased PUFA levels in the lipidome (Fig 3A-C; S1A-C Fig).

Lipidomic analysis of fat-2(wa17) mutant.

(A) SFA, MUFA, and PUFA levels in phosphatidylcholines (PCs) of fat-2(wa17) grown in various conditions. Note that cultivation on 2mM LA boosts PUFA levels. LA to NGM worms were grown on 2 mM LA before being transferred to NGM 6 h prior to harvesting. (B) Heatmap of PC species in fat-2(wa17) in all conditions. (C) Levels of individual FA species in PCs for all conditions. Inset shows levels of 20:5 FA are increased by providing fat-2(wa17) with linoleic acid. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Testing the effect of paqr-2(tm3410) suppressors

The previously characterized paqr-2(tm3410) null mutant has excess SFAs within phospholipids and many of the resulting defects, including membrane rigidity, can be suppressed by mutations that activate fatty acid desaturases or promote the incorporation of UFAs into phospholipids [15,2123]. Given the phenotypic similarities between paqr-2(tm3410) and fat-2(wa17), such as cold intolerance and rigid membranes, we hypothesized that previously characterized paqr-2(tm3410) suppressors may be able to suppress fat-2(wa17) as well. However, the paqr-2(tm3410) suppressors tested (mdt-15(et14), nhr-49(et8), fld-1(et46), paqr-1(et52), acs-13(et54); [15,2123]) resulted in either no or only slight rescue of fat-2(wa17) growth (Fig 4A-D). Additionally, Oil Red O staining of fat-2(wa17) mutants suggests that they have an excessive lipid content, and this too was not normalized by the tested paqr-2(tm3410) suppressors (Fig 4E-G). These results suggest that membrane rigidity is at most only a minor cause of the fat-2(wa17) defects since fluidizing treatments (NP40 or OA) or mutations (the tested paqr-2 suppressors) provide only minimal or no suppression.

Membrane fluidizing mutations partially rescue fat-2(wa17).

(A-D) Fluidizing paqr-2(tm3410) suppressor mutations only slightly rescue fat-2(wa17) growth. Dashed horizontal lines indicate approximate length of L1s at the start of the experiments; length was measured 72 h post-synchronization. Error bars show the standard error of the mean. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control. (E-G) Oil Red O staining of day 1 adults shows that the high lipid abundance in fat-2(wa17) is not suppressed by paqr-2(tm3410) fluidizing mutations.

A forward genetics screen for fat-2(wa17) suppressors

With the aim of identifying essential roles of PUFAs and molecular mechanisms that can compensate for PUFA deficiency, we performed a forward-genetic screen for fat-2(wa17) suppressors that allow growth to adulthood within 72 hours, as opposed to the ∼120 hours needed for the parental strain (Fig 5A). Approximately 40,000 EMS-mutagenized haploid genomes were screened leading to the isolation of ten fat-2(wa17) suppressors which fell into two groups: mutations within the fat-2 locus itself and mutations within genes of the HIF-1 pathway (Fig 5B). The fat-2(wa17) suppressors all reached adulthood within 72 hours (criteria for the screen) and improved the growth of the mutant when assessed by measuring worm length at 72 hours (Fig 5C-D). The hif-1(et69) mutation was recreated by CRISPR-Cas9 within the fat-2(wa17) background to confirm its fat-2(wa17) suppressor activity (S2A Fig); multiple independent isolations of essentially the same alelles within egl-9 (alelles et60-et62), fat-2 (alleles et63-et66), and ftn-2(et67-68) also serves as confirmation for these loci.

A forward genetic screen reveals fat-2(wa17) is suppressed by mutations in the HIF-1 pathway.

(A) Overview of the forward genetics screen strategy to isolate fat-2(wa17) suppressors. (B) Identity and position of the fat-2(wa17) suppressors as well as the positions of functional domains. Novel mutations are marked by a black triangle with corresponding allele name and mutation effect; the red triangle in FAT-2 indicates the original wa17 allele. Gene names in red represent loss-or reduction-of-function mutations; gene names in green represent gain-of-function mutations. (C) Length of all fat-2(wa17) suppressors measured 72 h after L1 stage. (D) Representative images of fat-2(wa17) suppressors after 72 h of growth. (E-G) Null alleles of egl-9 and hif-1 do not rescue fat-2(wa17), but the null allele of ftn-2 does, confirming that ftn-2(et67) and ftn-2(et68) are loss-of-function alleles. Lengths measured 72 h after L1 synchronization. (H) ftn-2(et68) rescue of fat-2(RNAi) worms confirming that the suppressors are not wa17 specific. The horizontal dashed line indicates the approximate length of L1s at the start of each experiment. Error bars show the standard error of the mean. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Three of the four fat-2 intragenic alleles (et64-et66) carried a substitution of serine to leucine at position 99 (S99L), only two amino acids away from the S101F mutation in fat-2(wa17); the fourth, et63, is a missense mutation substituting valine with methionine at position 25 (V25M) (Fig 5B). These four internal fat-2 alleles likely compensate structurally for the S101F mutation in fat-2(wa17) and thus improve its activity.

Three independent fat-2(wa17) suppressor mutations were found to affect the same arginine at position 557 of EGL-9 (et60 and et61 resulted in a R557H missense while et62 caused a R557C missense; Fig 5B). EGL-9 is a proline hydroxylase (PHD) that interacts via its R557 with Fe2+/2-oxoglutarate, a required co-factor for its oxygenase activity [24]. EGL-9 regulates the response to iron depletion and hypoxia: in the presence of sufficient Fe2+ and oxygen, EGL-9 can hydroxylate HIF-1 (hypoxia-inducible factor 1), leading to its ubiquitination and degradation. When either Fe2+ or oxygen are unavailable, HIF-1 is stable and can bind DNA to regulate adaptive transcriptional responses [2426]. It is surprising that all three fat-2(wa17) suppressor alleles affected precisely the same amino acid within EGL-9, and we surmise that a special property is conferred to EGL-9 by this specific mutation. For example, this mutant version of EGL-9 may be unable to inactivate HIF-1 by hydroxylation but still retain other important functions. In agreement with this interpretation, we found that the egl-9(sa307) null mutant cannot act as a fat-2(wa17) suppressor (Fig 5E).

One of the fat-2(wa17) suppressors corresponds to a splice acceptor mutation in the 5th intron of HIF-1, which would result in a frameshift after the first 413 amino acids if splicing instead occurs with the following 6th intron splice acceptor site (Fig 5B). This hif-1(et69) allele is dominant: heterozygosity for hif-1(et69)/+ provides better fat-2(wa17) suppression than hif-1(et69)/hif-1(et69) homozygosity (S2B Fig). This suggests that the hif-1(et69) allele is a gain-of-function allele, which may be because the frameshift occurs just after the first of potentially two prolines that are hydroxylated by EGL-9 when oxygen and Fe2+ levels are sufficient [24]. This is also consistent with the observation that the hif-1(ok2654) null allele is not a fat-2(wa17) suppressor (Fig 5F). Usually, hydroxylation of the prolines P400 and P621 causes recruitment of a ubiquitin ligase leading to HIF-1 degradation [24]. In the case of the hif-1(et69) allele, such regulation is likely impossible, and a constitutive HIF-1 may act as a fat-2(wa17) suppressor in several ways: promote overexpression of lipid metabolism genes including fat-2 [27], inhibit fatty acid beta-oxidation [28,29], which may help PUFAs to reach adequate levels even in the fat-2(wa17) mutant, or suppress the expression of the ferritin-encoding ftn-2, thus increasing the levels of ferrous ions required for desaturase activity [30,31].

Most informatively, the last two fat-2(wa17) suppressor mutations introduced premature STOP codons within the ftn-2 gene (alleles et67 and et68; Fig 5B). Additionally, the ftn-2(ok404) null allele also acted as a potent fat-2(wa17) suppressor (Fig 5G) which is consistent with inhibition of ftn-2 being the key outome from HIF-1 pathway activation. C. elegans ftn-2 encodes a ferritin that is expressed in intestine, muscle and several neurons [32]. The FTN-2 protein is constitutive and 10X faster as a ferroxidase (oxidising the reactive ferrous Fe2+ to the harmless ferric Fe3+) than FTN-1, which is an inducible intestine-specific ferritin in C. elegans [30,3336]. Additionally, FTN-2 is the major binder of iron in worms, and ftn-2 mutants therefore contain much less iron than wild-type worms, though the Fe2+/Fe3+ ratio is increased among the remaining iron [37]. This is likely the mechanism by which the ftn-2(et67) and ftn-2(et68) alleles act as fat-2(wa17) suppressors: increasing the availability of ferrous ions is a potent way to activate desaturases [31] and thus likely increases the activity of the near null fat-2(wa17) allele leading to the production of more/sufficient PUFAs. Importantly, the ftn-2(et68) allele was also able to suppress the growth defect resulting from fat-2 knockdown (using RNAi; Fig 5H); this shows that ferritin mutations compensate for reduced fat-2 activity generally rather than suppressing specifically only the fat-2(wa17) allele. Additionally, the ftn-2(et68) allele was not able to rescue the fat-2(syb7458) null allele (S2C Fig) suggesting that some fat-2 activity must exist for ftn-2(et68) to act upon. Lastly, ftn-2(et68) is still a potent fat-2(wa17) suppressor when hif-1 is knocked out (S2D Fig), and hif-1(et69) is similarly able to suppress fat-2(wa17) when ftn-2 is knocked out (S2E Fig). Altogether the genetic interaction studies suggests that the suppressor mutations in ftn-2 and hif-1 are acting via the same mechanism to rescue fat-2(wa17) and that ftn-2 is downstream of hif-1 in the fat-2 suppression pathway.

Effect of fat-2(wa17) suppressors on HIF-1 and PUFA levels

The results of the fat-2(wa17) suppressor screen support a model where the egl-9 R557 substitution alleles have an impaired ability to suppress HIF-1, while the gain-of-function hif-1(et69) allele constitutively suppresses ftn-2 expression and the ftn-2 null alleles are unable to sequester ferrous ions of which elevated levels increase fat-2 activity (Fig 6A). A 3xFLAG-tagged version of the endogenous HIF-1 allowed us to monitor HIF-1 levels in different conditions using Western blots (Fig 6B-C). As expected, hypoxia caused elevated levels of HIF-1 in wild-type worms. HIF-1 levels are abnormally low in fat-2(wa17) during normoxia but restored to normal levels by the internal fat-2(et65) mutation, suggesting that low PUFA levels cause HIF-1 downregulation. Nevertheless, HIF-1 levels are also increased by hypoxia in the fat-2(wa17) mutant indicating that the hif-1 locus is still responsive to oxygen levels in the fat-2(wa17) mutant. As expected, egl-9(et60) drastically increases HIF-1 expression in fat-2(wa17), which is consistent with the R557 substitution impairing the ability of EGL-9 to inhibit HIF-1. Finally, the null ftn-2(et68) allele caused near-loss (a faint HIF-1 band is occasionally seen) of detectable HIF-1 in fat-2(wa17), suggesting feedback regulation between ftn-2 and hif-1 (Fig 6B-C).

fat-2(wa17) suppressors belong in HIF-1 pathway and influence HIF-1 levels.

(A) Proposed pathway of fat-2(wa17) suppression by mutations in the HIF-1 pathway. Reduction of EGL-9 constitutively activates HIF-1, HIF-1 activation inhibits FTN-2. The loss of FTN-2 increases the levels of Fe2+ outside of iron pools, thus boosting FAT-2 desaturase activity. Gain-of-function mutations are labeled in green, loss-or reduction-of-function mutations are labeled in red. (B) Western blot confirming that hif-1::3xFLAG levels in fat-2(wa17) is increased by egl-9(et60), but not by ftn-2(et68). Hypoxia treatment increases HIF-1 levels in WT and fat-2(wa17), confirming successful protein tagging. (C) Quantification of Western blot in B showing normalized relative intensity of the HIF-1 signal to that of tubulin.

As already mentioned, ferrous ions (Fe2+) are potent activators of desaturases [31]. Given that each of the fat-2(wa17) suppressor mutants within the HIF-1 pathway are predicted to ultimately inhibit ftn-2, thus increasing the ferrous ion pool, we hypothesized that PUFA levels should be at least partially normalized in the fat-2(wa17) suppressors. This was confirmed by lipidomic analysis of phosphatidylcholines (Fig 7A) and phosphatidylethanolamines (S3A Fig). In particular, while levels of 18:2 (LA) were not significantly increased in the suppressor strains, the levels of 20:5 (EPA) were significantly increased (Fig 7B-C, S3B-C Fig), likely because the suppressor mutations allow fat-2(wa17) to produce more LA that is converted by other elongases and desaturases into EPA, the end product.

Lipidomic analysis of fat-2(wa17) suppressors reveals that PUFA levels are increased.

(A) Levels of SFAs, MUFAs, and PUFAs in PCs measured in fat-2(wa17) suppressors confirming that the suppressors increase PUFA levels in fat-2(wa17). Worms were homozygous for all indicated genotypes but note that the hif-1(et69) allele suppresses fat-2(wa17) best in a heterozygous state. (B) Heat map analysis of PC species in suppressor mutants. (C) Levels of individual FA species in PCs in fat-2(wa17) suppressors, insert shows that levels of C20:5 are significantly increased in all double mutant strains. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control. Note that the N2 and fat-2(wa17) samples are the same as in Fig 3.

Multiple stress response pathways are active in fat-2(wa17) and suppressed by ftn-2(et68)

The increase in EPA, and PUFA levels in general, likely explains the improved development and growth of fat-2(wa17). We examined other traits that may be rescued by the fat-2 suppressors, using the ftn-2(et68) mutant as representative because the egl-9 and hif-1 alleles converge on it. We found that the membrane fluidity defects in fat-2(wa17) were suppressed by ftn-2(et68) (Fig 8A-C). Additionally, several stress response pathways that are constitutively activated in the fat-2(wa17) mutant were also rescued by ftn-2(et68). The mitochondrial UPR (visualized with a hsp-60::GFP reporter [38]) is activated in fat-2(wa17) at a level similar to that in afts-1(et15), a known activator of mitochondrial stress [39], and this is suppressed by ftn-2(et68) (Fig 8D-E). Similarly, the metabolic stress reporter DAF-16::GFP [40] is constitutively nuclear-localized in fat-2(wa17) and this is also suppressed by ftn-2(et68) (Fig 8F-G). Using a hsp-4::GFP reporter [41], we found that the ER UPR is only slightly activated in fat-2(wa17) relative to WT (especially in spermatheca), and that this stress response too is partially suppressed by ftn-2(et68) (Fig 8H-I). Altogether, these results show that the PUFA-deficient fat-2(wa17) mutant engages multiple stress response pathways and that these are abated by ftn-2(et68).

ftn-2(et68) rescues fat-2(wa17)’s stress responses.

(A) Representative image of a FRAP experiment, showing pGLO-1::GFP-CAAX-positive intestinal membranes. The rectangle indicates the bleached area. (B-C) Thalf values and FRAP curves showing that ftn-2(et68);fat-2(wa17) has less rigid membranes than fat-2(wa17). (D-E) Representative images and quantification of ftn-2(et68) rescue of fat-2(wa17) mitochondrial stress with a hsp-60::gfp reporter. atfs-1(et15) serves as a control for high mitochondrial UPR activation. (F-G) Representative images and quantification of DAF-16::GFP localization showing that the DAF-16 stress response is constitutively active in the fat-2(wa17) mutant but normalized by ftn-2(et68). Chi-squared test shows that fat-2(wa17) is significantly different from WT. (H-I) Representative images and quantification of mild ER stress in fat-2(wa17) that is slightly rescued by ftn-2(et68) using a hsp-4::gfp reporter. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Mimicking fat-2(wa17) suppressors using hypoxia or iron supplements

Attempts to mimic the effects of the fat-2(wa17) suppressor mutations by hypoxia or supplement treatments were only partially successful. Providing fat-2(wa17) with ferric ammonium citrate (FAC), which increases the levels of ferric ions that can be converted into ferrous ions as well as overall iron levels in worms [42], provided only a slight rescue of the fat-2(wa17) mutant (Fig 9A). Additionally, providing fat-2(wa17) with ferrous ions in the form of ferrous chloride did not provide any rescue (S4A Fig). Reducing the levels of ferrous ions with the iron chelator deferoxamine, which we hypothesized would further hinder fat-2(wa17) growth, had no effect (S4B Fig); however, given that the fat-2(syb7458) null mutant grows at the same rate as fat-2(wa17) in 72 hours (S4C Fig) but never develops into an adult, we theorize that 72 hours may be too short to see a negative effect from deferoxamine on fat-2(wa17). HIF-1-activating paraquat (PQ; [43]) likewise conferred only a small rescue (Fig 9B), while the combination of FAC and PQ did not provide any additional growth rescue, both at 20°C and 25°C (Fig 9C; S4D Fig). The HIF-1 activator hydrogen peroxide [44] also only mildly rescued fat-2(wa17) (Fig 9D), while two separate hypoxia mimetics (cobalt chloride [45] and sodium sulfite [46]) did not suppress the poor growth of fat-2(wa17) (S4E-F Fig). Finally, exposing fat-2(wa17) to multiple short hypoxia treatments slightly increased growth, but longer hypoxia treatments had no effect (Fig 9E). Taken together, these results suggest that increasing iron and activating HIF-1 are beneficial to fat-2(wa17), but that achieving physiologically optimal dosing via experimental treatments is difficult.

Exogenous treatments that mimic fat-2(wa17) suppressors partially rescue fat-2(wa17).

(A-E) Length assay of fat-2(wa17) cultivated with different treatments for 72 h after L1 stage synchronization. The horizontal dashed line represents the approximate length of L1 worms at start of each experiment. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Discussion

That dietary PUFAs are essential for mammalian health, with LA and ALA acting as precursors for the synthesis of other PUFAs, is known since the 1930s [8]. PUFAs have been linked to several important cellular and physiological processes (reviewed in [4749]), including cell membrane properties and organelle dynamics (1,50), autophagy [50], mitochondria function [51], ferroptosis [5254], regulation of the daf-2/insulin, mTOR and p38-MAPK pathways [5557], SREBP stability and signaling [58,59], lipid droplet fusion [60], neuronal signaling and neurotransmission [6163], TRPV-dependent sensory signaling [61], oocyte development [64], and telomere length [65]. Which of these, if any, is the specific essential role of PUFAs in animal physiology? And are there molecular mechanisms that can compensate for PUFA deficiency? In the present study we approached these questions using forward genetics in C. elegans. While C. elegans can de novo synthesize PUFAs, mutations that impair the production of certain PUFAs can lead to developmental defects or lethality [9,17,66], offering opportunities for suppressor screens. Here, we showed that defects in the fat-2(wa17) mutant, which has limited Δ12 desaturase activity and only produces trace amounts of PUFAs, are suppressed by either compensatory intragenic mutations within fat-2 itself or by mutations within the HIF-1 pathway. The fact that screening approximately 40,000 haploid genomes for fat-2(wa17) suppressors and finding mutations only within fat-2 itself or within the HIF-1 pathway suggests that the screen has reached near-saturation and that we may have identified most, if not all, possible genetic ways to compensate for the fat-2(wa17) mutation. Importantly, none of the fat-2(wa17) suppressor mutations that we identified compensate for the PUFA shortage itself. Instead, the fat-2(wa17) suppressors act by boosting desaturase activity to allow the fat-2(wa17) mutant to synthesize more PUFAs; the fat-2(wa17) suppressors therefore cannot suppress the defects of the fat-2 null mutant, as we specifically showed for ftn-2(et68). We draw the important conclusion that PUFAs are not only essential but also that their essential functions cannot be genetically replaced.

The fat-2(wa17) suppressor mutations within the HIF-1 pathway converge on the inhibition of ftn-2. The primary function of ferritin is to provide a harmless storage of iron within cells: ferritin promotes the oxidation of ferrous ions and stores the resulting ferric ions in a mineralized form [67]. Thus ftn-2 inhibition results in reduced total cellular iron but increased levels of ferrous ions, i.e. Fe2+ [68,69]. Importantly, ferrous ions are required for desaturase reactions and increasing ferrous ion concentration is a potent way to increase activity because it accelerates the rate at which the desaturase cycles from the inactive post-reaction Fe3+-bound state to the active Fe2+-bound state [31,70]. Because eukaryotic desaturases are all evolutionarily closely related [71] and act in essentially the same way, ferrous ions must also be potent FAT-2 activators and thus boost the output from the mutant FAT-2(S101F) protein produced by the fat-2(wa17) allele or from the reduced FAT-2 protein levels in fat-2 RNAi-treated worms. Our findings suggest an elegant explanation for the observation that HIF-1 inhibits ftn-2 expression in C. elegans [30]: this is likely an adaptive response to boost desaturase activity when oxygen or iron is limiting, insuring a maximum output under adverse conditions. Fe2+ and HIF may contribute to desaturase boost also in human since CytB5 (which supplies Fe2+ to desaturases) promotes SFA tolerance while VHL (which causes HIF degradation) prevents SFA tolerance in cultured cells [72].

Lipidomic analysis showed that among all PUFAs, it was the EPA levels that were best restored by the fat-2(wa17) suppressors. It is likely that any LA molecule produced in the mutants is quickly acted upon by downstream desaturases and elongases, leading to increased levels of the end product, namely EPA. EPA may be a sufficient or particularly important PUFA for sustaining C. elegans health given that the fat-2(wa17) mutant is well rescued by EPA supplements. Indeed, DHA, which is not produced by C. elegans, is also able to rescue the fat-2(wa17) mutant. Others have shown that supplementing nearly completely EPA-deficient fat-3 C. elegans mutants with DHA significantly restored their EPA levels, suggesting that DHA supplements reduce EPA turnover [73]. EPA and DHA, being long chain PUFAs should have similar fluidizing effects on membrane properties (though in vitro experiments challenge this view [74]), and both can serve as precursors of eicosanoids, particularly inflammatory ones [75]. Abundant literature indicates that EPA is a particularly important PUFA in C. elegans. Phosphatidylcholines containing two attached EPA molecules are very abundant in C. elegans membranes, and their abundance increases the most in response to a temperature shift from 25°C to 15°C, suggesting an important role in fluidity homeostasis [76]. Long chain PUFAs such as EPA are required for efficient neurotransmission in C. elegans: mutants unable to produce them have depleted levels of synaptic vesicles accompanied by poor motility and these defects are rescued by exogenous PUFAs, including DHA [73]. C. elegans can also convert EPA to eicosanoids in a cytochrome P450-dependent manner [77,78]; inhibiting this process results in reduced pharyngeal pumping rate suggesting that regulation of muscular contraction by eicosanoids is conserved from nematodes to mammals [79]. EPA-derived eicosanoids are also required for guiding some cell migrations in C. elegans, including that of sperm [80]. EPA, and other PUFAs, can inhibit the nuclear localization of DAF-16 in fat-2-RNAi-treated worms, suggesting that they mediate signaling via this insulin receptor homolog and thus generally promote growth in C. elegans rather than stress resistance and fat storage [55]. In conclusion, EPA is clearly an important PUFA in C. elegans and our work suggests that its multifaceted functions cannot be replaced by mutations in any one gene.

Finally, the case of the three novel egl-9 alleles isolated in our screen deserves special attention. All three alleles specifically affect the arginine at position 557 of the EGL-9 protein (it is converted to a histidine in two of the alleles, and to a cysteine in the third). This Arg557 in EGL-9 is specifically required for its ability to hydroxylate HIF-1 thus marking it for ubiquitin-dependent degradation [24]. Null alleles of egl-9 were not picked in our screen and directly testing such a null allele revealed it to be ineffective as a fat-2(wa17) suppressor. We conclude that the EGL-9 proteins bearing a mutation at position Arg557 retain important functions while being unable to hydroxylate HIF-1. Others have previously demonstrated that EGL-9 could inhibit HIF-1 even when unable to hydroxylate it [81]. Clearly there is more to EGL-9 than its function as a HIF-1 hydroxylase and it would be interesting in the future to detail this further.

We conclude that PUFA-deficient fat-2(wa17) mutants benefit only slightly from membrane-fluidizing treatments, that there is no genetic way to compensate for PUFA deficiency. fat-2(wa17) mutants can only be rescued by boosting the activity of its defective desaturase, and restoring EPA levels are likely sufficient to suppress most fat-2(wa17) phenotypes suggesting a particularly important role for this PUFA in C. elegans. In the future it will be interesting to determine if boosting desaturase activity by inhibition of ferritin expression via HIF-1 is also a beneficial response to hypoxia in worms and human.

Materials and methods

C. elegans strains and cultivation

The wild-type C. elegans reference strain N2, fat-2(wa17), nhr-49(et8), mdt-15(et14), paqr-1(et52), paqr-2(3410), acs-13(et54), fld-1(et46), hif-1(ok2564), ftn-2(ok404), ftn-1(ok3625), egl-9(sa307), atfs-1(et15), zcIs4 [hsp-4::GFP], zcIs9 [hsp-60::GFP + lin-15(+)] and zIs356 [daf-16p::daf-16a/b::GFP + rol-6(su1006)] are available from the C. elegans Genetics Center (CGC; USA). The PHX7548 (fat-2(syb7458)/nT1[qIs51](IV;V)) strain was created by Suny Biotech Co using CRISPR/Cas9 and carries a deletion of 1387bp between flanking sequences 5’-aaacttggcccccgacgaagatg-3’ and 5’-gtgataatgacgagaataagtcct-3’. fat-2(syb7458) worms were maintained in an unbalanced state on non-peptone plates containing OP50 grown overnight in LB containing 2 mM linoleic acid.

Unless otherwise stated, experiments were performed at 20°C, using the E. coli strain OP50 as a food source, which was re-streaked every 6-8 weeks and maintained on LB plates at 4°. Single colonies were cultivated overnight at 37°C in LB medium before being used to seed NGM plates. Stock solutions of supplements were filter-sterilized and added to cooled NGM after autoclaving to produce supplement plates.

Construction of fat-2(+)

The pfat-2(+) construct was generated with the NEB PCR Cloning Kit for amplification of fat-2(+) with the following primers: 5’-gagctcaagaagcgtttcca-3’ and 5’-gggcaagaatttgtagtgtca-3’ using N2 genomic DNA. Plasmids were prepared with a GeneJet Plasmid Miniprep Kit and injected at the following concentrations: pfat-2(+) of 20 µg/µl, pRF4(rol-6) of 40 µg/µl, and pBSKS of 35 µg/µl into fat-2(wa17) and N2 worms.

Pre-loading of E. coli with fatty acids

Stock solutions of fatty acids dissolved in ethanol (EPA, DHA, OA) or DMSO (LA) were diluted in LB media to the appropriate final concentration, inoculated with OP50 bacteria, and shaken overnight at 37°C. The bacteria was washed twice in M9 to remove fatty acids, concentrated 10X by centrifugation, dissolved in LB and seeded onto NGM plates lacking peptone.

Growth assays

For length measurement studies, synchronized L1s were plated onto test plates seeded with OP50 and worms were mounted and photographed 72 h later. Experiments performed at 15°C were photographed after 144 h. The length of 20 worms was measured using ImageJ.

For hydrogen peroxide treatment, synchronized worms were incubated in 2 mM hydrogen peroxide for 2 h at L1 stage before being plated on NGM plates for 72 h. For hypoxia treatment, synchronized L1s were incubated for 2-6 h in a hypoxia chamber, returned to normoxia for 24h, and hypoxia exposure was repeated as stated in the figure.

Oil Red O staining

Synchronized day 1 adults were washed three times with PBST and fixed for 3 minutes in 60% isopropanol. Worms were then rotated for 2 h in filtered 60% Oil Red O staining solution. The stained worms were washed three times in PBST before being mounted on agarose pads and imaged with a Zeiss Axioscope microscope.

Mutagenesis and screen for fat-2(wa17) suppressors

fat-2(wa17) worms were mutagenized for 4 hours by incubation in the presence of 0.05 M ethyl methane sulfonate (EMS) according to the standard protocol [82]. The worms were then washed and spotted onto an NGM plate. After 2 h, L4 hermaphrodite animals were transferred to new plates. 8-10 days later, F1 progeny were bleached, washed, and their eggs allowed to hatch overnight in M9. The resulting L1 larvae were spotted onto new plates, cultivated at 20°C, then screened after 72 h for gravid F2 worms, which were then picked to new plates for further analysis. In total, approximately 40 000 independently mutagenized haploid genomes were screened. The isolated suppressor mutants were outcrossed 4 to 6 times prior to whole genome sequencing, and 8 to 10 times prior to characterization. Outcrossing was done by mating N2 males to a suppressor, then crossing male progeny to fat-2(wa17) mutant worms. Progeny from this cross were picked to individual plates and kept at 20°C then screened for fat-2(wa17) homozygosity using PCR, followed by testing the F2 progeny for ability to grow to adults in 72 h. Genotyping primers for the suppressor mutants are included in S1 Table.

Whole genome sequencing

The genomes of the ten suppressor mutants that had been outcrossed 4 or 6 times were sequenced by Eurofins (Constance, Germany) with a mean coverage varying from 40.68X to 63.05X and their genomes assembled using the C. elegans genome version cel235 from Ensembl (REF: PMID: 37953337). Eurofins applied customised filters to the variants to filter false positives using GATK’s Variant Filtration module [83,84]. Variants detected were annotated based on their gene context using snpEff [85]. For each suppressor mutant, one or two hot spots, i.e. small genomic area containing several mutations, were identified and candidate mutations tested experimentally as described in the text.

CRISPR-Cas9 genome editing

The recreation of the candidate suppressor mutations and insertion of the 3xFLAG tag into the hif-1 gene was performed using CRISPR-Cas9 gene editing as previously described [86,87]. The insertion of the ssDNA oligos was performed utilizing the homology-direct repair (HDR) mechanisms. The protospacer-adjacent motif (PAM) site of the ssDNA oligo template was flanked by 40 bp homology arms. Design and synthesis of the ssDNA and CRISPR RNA (crRNA) was performed using the Alt-R HDR Design Tool from IDT (Integrated DNA Technologies, Inc.; Coralville, IA, USA), including proprietary modifications that improve oligo stability. To recreate the hif-1(et69) allele, we used the crRNA sequence 5’-UUUCUUAACGUGUGUAUUUCGUUUUAGAGCUAUGCU-3’ and the DNA oligo donor sequence 5’-AGTTCCATACATTTAGCAAGTGATTTCTTAACGTGTGTATTTCAAGAGCACGTAAGAACA GCTACGATGACGTTTTGCAATGGCT-3. To introduce the 3xFLAG at the N-terminus coding end of hif-1 we used the crRNA sequence 5’-GAAAAUAAUCAAGAGAGCAUGUUUUAGAGCUAUGCU-3’ and the DNA oligo donor sequence 5’-AAATGAACAACAGCCTAGTTCTTATTCCCCATTTCCAATGCTCTCTGACTACAAGGACCA CGACGGCGATTATAAGGATCACGACATCGACTACAAAGACGACGATGACAAGTGATTAT TTTCTACCCCCTCTCAAACTGTTCATTGTTTTG-3’. The injection mixes were prepared using 10 μg/μl of the Cas9 enzyme (IDT), 0.4 μg/μl tracrRNA (IDT), 2.8 0.4 μg/μl crRNA (IDT), 1 μg/μl of ssDNA (IDT), and 40 ng/μl of PRF4(rol-6) or Pmyo-2(GFP) plasmid. The mixture was microinjected into the posterior gonad of the worm and the F1 generation was screened for animals expressing the reporter plasmid. Genotypes were tested by PCR and successfully edited genes were confirmed by Sanger sequencing (Eurofins).

Fluorescence recovery after photobleaching (FRAP)

FRAP experiments were carried out using a membrane-associated prenylated GFP reporter expressed in intestinal cells as previous described [88], using a Zeiss LSM700inv laser scanning confocal microscope with a 40X water immersion objective. Briefly, the GFP positive membranes were photobleached over a rectangular area (30 x 4-pixels) using 30 iterations of the 488 nm laser with 50% laser power transmission. Images were collected at a 12-bit intensity resolution over 256 x 256 pixels (digital zoom 4X) using a pixel dwell time of 1.58 µs, and were all acquired under identical settings. The recovery of fluorescence was traced for 25 seconds. Fluorescence recovery and Thalf were calculated as previously described [16].

Stress Response Assay

Worms were imaged with a Zeiss Axioscope and fluorescence intensity was quantified with ImageJ (n≥20 for all experiments). Worm strains carrying hsp-60::GFP were imaged as day 1 adults, and the fluorescence values were taken from a 39 μm circumference circle in the brightest part of the anterior part of the worm. Worm strains carrying DAF-16::GFP were imaged as L4s and the percentage of worms with cytoplasmic or nuclear localization was quantified. Worm strains carrying hsp-4::GFP were imaged as day 1 adults and the fluorescence of the whole worm was quantified.

Lipidomics

Samples were composed of synchronized L4 larvae (one 9 cm diameter plate/sample; each treatment/genotype was prepared in four independently grown replicates) grown on NGM or non-peptone plates seeded with linoleic acid. In the case of LA to NGM samples, worms were grown until late L3/early L4 stage on linoleic acid seeded non-peptone plates before being transferred to NGM plates for 6 h before collection. Worms were washed 3 times in M9, pelleted and stored at −80°C until analysis. For lipid extraction, the pellet was sonicated for 10 minutes in methanol;butanol [1:3] and then extracted according to published methods [89]. Lipid extracts were evaporated and reconstituted in chloroform:methanol [1:2] with 5 mM ammonium acetate. This solution was infused directly (shotgun approach) into a QTRAP 5500 mass spectrometer (Sciex) equipped with a TriVersa NanoMate (Advion Bioscience) as described previously [90]. Phospholipids were measured using precursor ion scanning in negative mode using the fatty acids as fragments [91,92]. To generate the phospholipid composition (as mol%) the signals from individual phospholipids (area under the m/z peak in the spectra) were divided by the signal from all detected phospholipids of the same class. The data were evaluated using the LipidView software (Sciex). The data were further analyzed using Qlucore Omics Explorer n.n (Qlucore AB) for analysis. The data were normalized for the purpose of the heat map visualization (mean = 0; variance = 1).

Protein extraction and western blots

Worms were lysed using lysis buffer containing 25 mM Tris (pH 7.5), 300 mM NaCl, 0.1% NP40, and 1X protease inhibitor on ice with a motorized pestle. Samples were centrifuged at 20000g for 15 min at 4°C and protein sample concentration was quantified using BCA protein assay kit. 15 µg of protein were mixed with Laemmli sample loading buffer contained ý-mercaptoethanol, boiled for 10 min, and loaded in 4% to 20% gradient precast SDS gel. After electrophoresis, the proteins were transferred to nitrocellulose membranes using Trans-Blot Turbo Transfer Packs and a Trans-Blot Turbo apparatus/predefined mixed-MW program. Blots were blocked in 5% nonfat dry milk in PBST for 1 h at room temperature. Blots were incubated with primary antibodies overnight at 4°C (mouse monoclonal anti-FLAG antibody (M2, Sigma Aldrich) 1:5000 dilution) or 1 h at room temperature (mouse monoclonal anti-alpha-Tubulin (B512, Sigma Aldrich) 1:5000 dilution). Blots were then washed with PBST and incubated with swine-anti rabbit HRP 1:3000 dilution or goat anti-mouse HRP 1:3000 dilution for 1 h at room temperature and washed again with PBST. Detection of the hybridized antibody was performed using and ECL detection kit (Immobilon Western, Millipore), and the signal was visualized with a digital camera (VersaDoc, Bio-Rad).

Statistics

Error bars for the worm length measurements show the standard error of the mean, and one-way ANOVA tests were used to identify significant differences from fat-2(wa17) control unless otherwise stated. All experiments were independently repeated at least twice with similar results, and the statistics shown apply to the presented experimental results.

Lipidomics analysis of PEs in fat-2(wa17) in various cultivation conditions.

(A) SFA, MUFA, and PUFA levels in phosphatidylethanolamine (PEs) of fat-2(wa17) grown in various conditions. Cultivation on 2 mM LA boosts PUFA levels. LA to NGM worms were grown on 2 mM LA before being transferred to NGM 6 h prior to harvesting. (B) Heatmap of PE species in fat-2(wa17) in all conditions. (C) Levels of individual FA species in PEs for all conditions. Inset shows that the levels of C20:5 are increased by providing fat-2(wa17) with LA. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

fat-2(wa17) and fat-2(syb7458) with suppressors.

(A) Confirmation of hif-1(et69) suppression of fat-2(wa17) by CRISPR-Cas9. (B) The hif-1(et69) allele acts best as a fat-2(wa17) suppressor when in a heterozygous state. (C) ftn-2(et68) does not act as a suppressor for the fat-2(syb7458) allele. (D) ftn-2(et68) still suppresses fat-2(wa17) in a hif-1 null background. (E) hif-1(et69) suppresses fat-2(wa17) in a ftn-2 null background. (A-E) All length measurements were taken 72 h after L1 synchronization. Horizontal dashed line represents the approximate size of L1s at the start of each experiment. Error bars show the standard error of the mean. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Lipidomics of PEs in fat-2(wa17) suppressors.

(A) Levels of SFAs, MUFAs, and PUFAs in PEs measured in fat-2(wa17) suppressors confirming that the suppressors increase PUFA levels in fat-2(wa17). (B) Heat map analysis of PE species in suppressor mutants. (C) Levels of individual FA species in PEs in fat-2(wa17) suppressors, insert shows levels of 20:5 are increased in all double mutant strains. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

Exogenous treatment of fat-2(wa17) to mimic suppressors.

(A-E) Length of fat-2(wa17) treated with various diets for 72 h after L1 synchronization attempting to mimic effects of egl-9, ftn-2, and hif-1 suppressors. Horizontal dashed line represents approximate size of worms at the start of each experiment. *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences compared to the fat-2(wa17) control.

fat-2(wa17) suppressor genotyping primers.