Gain-of-Function Alleles in Caenorhabditis elegans Nuclear Hormone Receptor nhr-49 Are Functionally Distinct

Nuclear hormone receptors (NHRs) are transcription factors that regulate numerous physiological and developmental processes and represent important drug targets. NHR-49, an ortholog of Hepatocyte Nuclear Factor 4 (HNF4), has emerged as a key regulator of lipid metabolism and life span in the nematode worm Caenorhabditis elegans. However, many aspects of NHR-49 function remain poorly understood, including whether and how it regulates individual sets of target genes and whether its activity is modulated by a ligand. A recent study identified three gain-of-function (gof) missense mutations in nhr-49 (nhr-49(et7), nhr-49(et8), and nhr-49(et13), respectively). These substitutions all affect the ligand-binding domain (LBD), which is critical for ligand binding and protein interactions. Thus, these alleles provide an opportunity to test how three specific residues contribute to NHR-49 dependent gene regulation. We used computational and molecular methods to delineate how these mutations alter NHR-49 activity. We find that despite originating from a screen favoring the activation of specific NHR-49 targets, all three gof alleles cause broad upregulation of NHR-49 regulated genes. Interestingly, nhr-49(et7) and nhr-49(et8) exclusively affect nhr-49 dependent activation, whereas the nhr-49(et13) surprisingly affects both nhr-49 mediated activation and repression, implicating the affected residue as dually important. We also observed phenotypic non-equivalence of these alleles, as they unexpectedly caused a long, short, and normal life span, respectively. Mechanistically, the gof substitutions altered neither protein interactions with the repressive partner NHR-66 and the coactivator MDT-15 nor the subcellular localization or expression of NHR-49. However, in silico structural modeling revealed that NHR-49 likely interacts with small molecule ligands and that the missense mutations might alter ligand binding, providing a possible explanation for increased NHR-49 activity. In sum, our findings indicate that the three nhr-49 gof alleles are non-equivalent, and highlight the conserved V411 residue affected by et13 as critical for gene activation and repression alike.


Introduction
Nuclear hormone receptors (NHRs) are transcription factors that modulate gene expression in response to extrinsic and intrinsic cues, and they are essential regulators of many developmental and physiological processes [1]. Mammalian Hepatocyte Nuclear Factor 4 alpha (HNF4α) is an illustrative example of an NHR with diverse functions. Genetic analyses in mice have shown that HNF4α is required for normal gastrulation via a functional requirement in the visceral endoderm, for proper terminal hepatocyte differentiation, and for normal function of the adult liver, including key roles in lipid, bile acid, and xenobiotic metabolism. In the pancreatic β-cells, HNF4α is required for normal glucose homeostasis and for β-cell mass expansion during pregnancy [2][3][4][5]. In line with HNF4α's important metabolic regulatory roles, loss-of-function mutations in the human HNF4A gene have been linked to maturity onset diabetes of the young (MODY) and to type 2 diabetes [3,6,7].
C. elegans NHR-49 controls multiple aspects of fatty acid metabolism and is linked to various physiological and molecular phenotypes. nhr-49 is required to express genes involved in fatty acid β-oxidation, especially upon starvation, and also promotes the expression of the two fatty acid desaturases fat-5 and fat-7 [14,15]. By regulating these key enzymes, NHR-49 promotes metabolic reprogramming that allows adaptation to starvation and is also essential for the extended life span of various mutant and transgenic C. elegans strains [18,19,23,24]. nhr-49 dependent fatty acid desaturation also appears to contribute to low-temperature adaptation by altering membrane lipid composition and thus membrane fluidity [25]. Additionally, nhr-49 is required to repress several additional lipid metabolic genes such as sphingolipid breakdown enzymes and lipases, and also regulates various non-lipid metabolism genes [13]. Whether and how these NHR-49 regulated genes contribute to the organismal functions and phenotypes of nhr-49 mutant worms has not yet been elucidated.
NHRs contain two evolutionarily conserved signature domains: an N-terminal zinc-finger DNA binding domain (DBD) that enables interaction with genomic hormone response elements; and a C-terminal ligand binding domain (LBD) that mediates reversible binding to ligands and also enables NHR dimerization and other protein-protein interactions [1]. Ligand binding induces structural changes that enable NHRs to differentially interact with transcriptional coregulators including coactivators and corepressors. In turn, this allows NHRs to implement specific gene programs upon ligand binding or dissociation. Changes in NHR dimer and NHR:coregulator interactions are thus essential to adapt genome expression in response to altered ligand availability [26][27][28]. For NHR-49, several functional partner NHRs and one coregulator have been identified to date. Specifically, NHR-13 and NHR-80 appear to cooperate with NHR-49 in the activation of fatty acid desaturase genes, whereas NHR-66 is thought to cooperate with NHR-49 to repress sphingolipid breakdown and lipase genes [13,17,19]. No NHR- 49 corepressor has yet been identified, but the Mediator subunit MDT-15 physically binds to NHR-49 and the two proteins share numerous downstream targets, including fatty acid desaturase genes and starvation induced β-oxidation genes [29]. Thus, MDT-15 likely serves as a coactivator for NHR-49. The physical interaction between NHR-49 and these interacting proteins is mediated by the LBD [13,29], but has not yet been characterized in greater detail.
Recently, three dominant nhr-49 gain-of-function (gof) alleles have been identified that overcome the cold-sensitive phenotype of a paqr-2/adiponectin receptor mutant [25]. One of these gof alleles promotes the expression of fat-7, which resulted in increased levels of monounsaturated fatty acids that were essential for cold resistance [25]. We reasoned that further study might reveal whether these gof mutations selectively activate fatty acid desaturase genes or whether they indiscriminately influence all nhr-49 regulated genes, i.e. whether they represent potential specificity determinants within NHR-49. Additionally, as all three alleles are point mutations near NHR-49's LBD [25], we reasoned that they might alter protein-protein interactions with partner NHRs or MDT-15 to e.g. induce fat-7. We thus set out to characterize the gene expression, protein interaction, and in vivo phenotypes of these alleles.

Results
nhr-49 gof mutations affect variably conserved residues and locate close to predicted functional motifs Svensk et al. described three gain-of-function (gof) mutations in the nhr-49 gene, et7(P479L), et8(S432F), and et13(V411E) (amino acid numbers refer to the largest NHR-49 isoform, NHR-49C) [25]. All three alleles act in dominant fashion to suppress the cold-sensitivity and a morphological phenotype of the paqr-2/adiponectin receptor mutant. For et8(S432F), the mode of paqr-2 suppression has been described: this mutation causes the upregulation of the known NHR-49 regulated gene fat-7 and a concomitant increase in unsaturated fatty acid levels [25]. However, other consequences of the nhr-49 gof alleles have not yet been described and their molecular characterization remains incomplete.
To gain better insight into the residues affected by the gof mutations, we aligned the sequence of C. elegans NHR-49 to the predicted NHR-49 orthologs of other Caenorhabditis species (Fig 1A), a group of parasitic nematodes (Fig 1B), and to the HNF4α and HNF4γ orthologs of the arthropod Drosophila melanogaster, the vertebrate Danio rerio, the mammal Mus musculus, and of H. sapiens (Fig 1C and 1D; see S1 Table for identity/similarity of NHR-49 to H. sapiens NHRs). As noted previously [25], all three mutations affect residues in or near the ligand-binding domain (LBD) of C. elegans NHR-49. Our alignments revealed that the three gof alleles arose in residues that exhibit distinct patterns of evolutionary conservation. P479 is identical in the five compared Caenorhabditis species, but is missing in all other investigated sequences; S432 is identical in all studied Caenorhabditis species and in four of eight queried non-Caenorhabditis nematodes (the other four nematodes encode an L at the position), but absent in the higher metazoans analyzed; and V411 is conserved but not identical in all sequences inspected (a V in all tested nematodes, an I in D. melanogaster, D. rerio, mouse, and human HNF4). It is therefore likely that the three gof mutations affect distinct properties of the C. elegans NHR-49 protein; et13(V411E) appears particularly interesting as it probably affects an evolutionarily conserved molecular function of NHR-49/HNF4.
The et13(V411E) mutation affects a residue predicted to be involved in the activation/repression switch and in ligand binding alleles all lie within or near the LBD, which is involved in transcriptional activation and repression, we assessed whether they are situated near transcriptional activation sequences, specifically the Nine Amino Acids Transactivation Domain (9aaTAD) [30,31]. The 9aaTAD is a structurally and experimentally defined motif that occurs in transcription factors that bind KIX-domain containing coregulators such as p300, CREB binding protein (CBP), and MED15, the ortholog of the NHR-49 coregulator MDT-15 [29,32]. We found that V411 (mutated in ; residues corresponding to NHR-49C V411, S432, and P479 are indicated by arrows. (E) Predicted 9aaTAD motifs in NHR-49. Note that V411 is adjacent to two overlapping high-confidence 9aaTAD motifs, and P479 is adjacent to an imperfect 9aaTAD motif. The table below the sequence indicates each predicted high stringency 9aaTAD and whether or not it conforms to each of 12 specific refinement criteria of the 9aaTAD prediction algorithm (C1-C12; see [30,31] for details on algorithm and refinement criteria).
As et13(V411E) lies within the LBD, an evolutionarily conserved and structurally defined domain, we performed structural homology modeling of NHR-49 to gain further insight into the putative effects of the V411E substitution; we also modeled the effects of an E327A mutation that is predicted to abrogate NHR-49 dimerization [33]. Although no experimentally derived structure is currently available for NHR-49, the structures of HNF4α and HNF4γ have been solved [34][35][36] and allow in silico modeling of NHR-49. As expected, all five NHR-49 isoforms could be modeled onto existing HNF4 structures with high confidence; these models include the DBD and the LBD (Fig 2A and 2B). For all five isoforms (NHR-49A-E) the Caenorhabditis specific C-terminal extension of the LBD, which harbors the residues S432 and P479, were also resolved fully. The inherent amino acid sequence differences across isoforms result in distinct quaternary structures in the C-termini of the respective isoforms with similar architectures for the DBD and LBD domains.
NHR-49 shares some functional similarity with mammalian PPARα, and thus it is sometimes considered a functional ortholog of that NHR [14,15,18]. Therefore, we modeled NHR-49C onto existing, experimentally derived PPARα and PPARγ structures ( Fig 2C). Although we again obtained high confidence models, these were inferior to the one generated using HNF4α as template, as expected given the higher primary sequence similarity between NHR-49 and HNF4α (S1 Table).
The V411E substitution occurring in nhr-49(et13) affects an evolutionarily conserved amino acid. Thus, we studied the biophysical consequences of et13(V411E) on all NHR-49 isoforms to assess the structural, and consequently predicted functional, disruption, as measured by energetic analysis. Interestingly, the V411E mutation modifies the structure of the loop between two helices and causes an enthalpy change, which is isoform dependent. Specifically, the mutation is predicted to have neutral to moderately stabilizing effects on isoforms B, D, and E (Fig 3A and 3B; S2 Table). Importantly, regardless of impact on protein structure in the static conformation, these two helices represent helix 10 and helix 12 of the conserved NHR structure, which is a dynamic and flexible molecular switch known to change conformation upon NHR transition from transcriptional repressor to activator [37]. As such, the et13 (V411E) mutation (as well as the et8(S432F) and et7(P479L) mutations) is anticipated to perturb NHR-49 dependent activation and/or repression by interfering with NHR-49's capacity to adopt an activating or repressing conformation and, possibly, interaction with MDT-15 and/or other interaction partners ( Fig 3B).

nhr-49 gof mutations broadly activate NHR-49 target genes
Having established that the three missense mutations in the NHR-49-LBD likely affect biophysical properties, we next wished to evaluate their impact on transcription. All three gof alleles originate from the same cold sensitivity suppressor screen and presumably compensate for the paqr-2 mutation by the same mechanism, i.e. by increasing membrane fluidity through the upregulation of unsaturated fatty acid levels via transcriptional activation of the fatty acid desaturase fat-7 [25]. However, this model has only been ascertained for et8(S432F), and only in the paqr-2/adiponectin receptor mutant background, which itself might cause the deregulation of some lipid metabolism genes [25,38]. To gain better insight into the gene expression nhr-49 gof Mutants Are Distinct changes caused solely by the gof mutations, we outcrossed them to wild-type worms, removing the paqr-2 mutation. In the resulting gof single mutant strains, we performed real-time quantitative PCR (qPCR) analysis of NHR-49 activated genes. We used nhr-49(nr2041) null mutants as a positive control. We also assessed gene expression in nhr-66(ok940) null mutants, as NHR-66 is an NHR-49 dimerization partner that appears to exclusively affect NHR-49 repressed genes [13]. As expected, all tested NHR-49 activated genes were downregulated in the nhr-49 (nr2041) null mutants but largely unaffected in nhr-66(ok940) null mutants (Fig 4).
First, we studied the three fatty acid desaturase genes and NHR-49 targets fat-5, -6, and -7 [15] (Fig 4A). In line with previously published data [25], nhr-49(et8) mutants showed a significant upregulation of fat-7, and also induced fat-5 expression. nhr-49(et13) worms showed similar changes, whereas the effects were less pronounced in nhr-49(et7) worms (significant only for fat-5, but not for fat-7). In contrast, we observed no significant upregulation of fat-6, in agreement with published findings that fat-6 dependence on nhr-49 is much weaker than that of fat-5 and -7 [14,15,29]. Collectively, these data are consistent with the model that the nhr-49 gof alleles increase the levels of unsaturated fatty acids by upregulating fatty acid desaturase genes.
Besides fatty acid desaturases, NHR-49 also affects fatty acid β-oxidation enzymes and nonfatty acid metabolism genes [13][14][15]. To test whether the effects of nhr-49 gof alleles were restricted to fatty acid desaturase genes or broadly affect NHR-49 targets, we next assessed the expression of β-oxidation enzyme genes acs-2 and cpt-5, and non-fatty acid metabolism genes sodh-1, ddo-2, dhs-18, and icl-1 (Fig 4B and 4C). Most genes were induced or trended towards induction in all three gof strains, with nhr-49(et13) generally showing the strongest activations and nhr-49(et7) displaying the weakest inductions. Lastly, we also found that all three gof strains induced several stress response genes [39,40] (Fig 4D). We conclude that, although identified in a screen for suppressors of cold sensitivity, the nhr-49 gof mutations do not only affect the fatty acid desaturase genes known to impact membrane fluidity and low temperature adaptation, but also broadly promote the expression of NHR-49 activated genes.
To corroborate the qPCR data, we generated a transgenic C. elegans strain carrying a transcriptional gfp reporter for lips-6. We used the 2kb region upstream of the predicted transcriptional structure: black; LBD: cyan). Both structures successfully model the NHR-49C-LBD (182 aa homology in both models), but neither has the robustness of the HNF4α-derived NHR-49C model. The PPARα-derived model had a 1.6535 RMSD and the PPARγ-derived model a 1.7144 RMSD from HNF4α-derived NHR-49C, indicating a worse overall fit. start site that likely reflects the putative lips-6 promoter (see Materials and Methods). At the L4 stage, expression of this reporter was very faint or completely absent in wild-type worms (Fig 5B). In contrast, although there was some variation of fluorescence signal between individual worms, plips-6::gfp expression was evident when the transgene was crossed into the nhr-66(ok940) null or Comparison of the biochemical differences between the wild-type V411 (grey) and mutant E411 (red) amino acid residue. The gof mutation has a substantially longer, negatively charged (acidic) side chain, which energetically destabilizes the protein structure.   Fig 5B). Notably, we observed expression in the intestine, a tissue known to express nhr-49 and many metabolic genes [15,18]. Taken together, based on the upregulation of NHR-49 repressed genes including lips-6 in the nhr-49(et13) strain, we conclude that the et13(V411E) mutation likely causes a dual loss-and gain-of-function.

The gof mutations do not alter NHR-49 expression or subcellular localization
To determine possible mechanisms leading to increased transcriptional activity, we next tested whether the gof mutations affect NHR-49 protein levels and/or localization. To date, no nuclear localization signal or nuclear export signal has been defined in NHR-49, and computational NLS prediction revealed only a low-confidence NLS in the N-terminus of NHR-49. To test whether the gof mutants affect NHR-49 protein levels or subcellular distribution, we generated transgenic strains expressing each individual gof mutation within a translational NHR-49::GFP fusion protein and compared them to wild-type NHR-49::GFP [18]. We found that NHR-49(et7)::GFP, NHR-49(et8)::GFP, and NHR-49(et13)::GFP showed strong similarity to wild-type NHR-49::GFP: they were expressed at similar overall levels, and in similar patterns in nuclei and cytoplasm of intestinal cells, seam cells, and the hypodermis (Fig 6; S1 Fig). Thus, the gof alleles are unlikely to achieve their effects through altered NHR-49 expression and/or localization.

Differential effects of the gof mutations on ligand-protein interactions in silico
Besides affecting NHR-49 level or localization, the gof alleles might alter ligand and/or protein binding, as all three gof mutations afflict residues situated in or near NHR-49's LBD. Notably, the HNF4 residue orthologous to NHR-49 V411, which is mutated in nhr-49(et13), helps coordinate the binding of a fatty acid ligand [34,36]. This suggests that et13(V411E) may alter ligand binding in NHR-49. No ligand has yet been identified for NHR-49. Thus, we employed unbiased in silico screening of a library comprised of 6,313 lipid molecules (<C19) to assess the functional impact of gof substitutions on ligand binding relative to the wild-type protein.
Initial ligand docking experiments in the NHR-49C homology model yielded >450,000 poses for 5,654 lipids docked in the LBD binding site (see Materials and Methods for details). Lib-Dock scores ranged from 29.065 (C 3 H 6 O short chain fatty acyl) to 133.845 (C 17 H 33 O 7 P glycerophosphates). For example, oleoylethanolamide, which binds NHR-80, an NHR-49 dimerization partner also related to HNF4 [19], had a LibDock score of 119.228 (Fig 7A). These data suggest that more complex, longer chain lipid molecules (C14-C21) may be the preferential ligands for NHR-49, with fatty acyls and polyketides having the highest ranked scores. Using the top-ranked lipid ligands (LibDock score >115) from the WT model, a subset of ligands (n = 165) was compared across the four mutant models, E327A, V411E, S432F, and P479L ( Fig 7A; see S3 Table for docking scores of the top 165 ligands). The structural modifications resulting from the mutations caused a number of lipid ligands to fail docking, with a maximum of 163 ligands docked for E327A and a minimum of 132 lipids docked for S432F ( Fig  7B). Intriguingly, V411E and S432F mutations failed to interact with 26 ligands, 19 of which also failed to dock in the P479L model. Moreover, even with successful docking, the energetics of the lipid:protein interaction, nature and number of bonds formed, and the resulting LibDock score varied widely for a given lipid. Notably, proteins encoding a gof allele generally trended towards reduced ligand binding capacity (Fig 7C). Specific classes of lipids showed substantial differences in their predicted ability to interact with WT or NHR-49(gof) proteins. In particular, saccharolipids completely failed to dock in any of the NHR-49(gof) mutant proteins. Similarly, fewer polyketides and glycerophospholipids were able to dock the LBDs of NHR-49(gof) proteins than that of WT NHR-49. In contrast, the modeled interactions of WT NHR-49 with target ligands were largely maintained in the NHR-49-E327A mutant model (n = 165 vs. 163). This demonstrates that loss-of-function mutations in the dimerization region have limited influence on the LBD's ligand binding capacity, as expected.

Differential effects of the gof mutations on protein-protein interactions
Besides ligand binding, the LBD also serves as a docking site for NHR coregulators and as a surface enabling NHR dimerization [9,10]. Indeed, the NHR-49 LBD is sufficient and necessary for interaction with the coactivator MDT-15, for homodimerization, and for heterodimerization with other NHRs, including NHR-66 [13,29,40]. We hypothesized that the gof mutations might potentially increase interaction with the coactivator MDT-15 and, for et13 (V411E), possibly decrease binding to the repressive dimerization partner NHR-66. To test these hypotheses we used the yeast-two-hybrid (Y2H) system. To study the effects of the gof point mutations, we introduced each mutation individually by site-directed mutagenesis into the NHR-49-LBD bait, and assessed the effect on MDT-15 and NHR-66 prey binding. As expected [29], the wild-type NHR-49-LBD bait (aa 130-501) strongly bound the MDT-15 prey (Fig 8A; note that NHR-49-LBD baits bearing gof substitutions were expressed at levels similar to wild-type NHR-49-LBD, Fig 8B). Compared to this interaction, the et8(S432F) substitution caused a significant, but modest increase in MDT-15 binding, whereas the et13(V411E) nhr-49 gof Mutants Are Distinct  (Fig 8A). Additionally, the et7(P479L) and et13(V411E) substitutions weakly but significantly increased NHR-66 binding (Fig 8C). Thus, et8(S432F) may cause gof by increasing NHR-49 binding to MDT-15, whereas et13 (V411E) substitution apparently achieves gof and lof effects despite decreased MDT-15 binding and increased NHR-66 binding.
As et13(V411E) causes severely reduced MDT-15 binding, we hypothesized that it might have lost mdt-15 dependence, with the gof effect arising from neofunctionalization, i.e. the acquisition of a molecular function not present in wild-type NHR-49. To test this hypothesis, we performed mdt-15 depletion by RNAi interference in wild-type N2 and nhr-49(et13) gof worms. The expression of NHR-49 target genes activated in nhr-49(et13) remained dependent on mdt-15 ( Fig 8D). We conclude that although MDT-15:NHR-49 binding is reduced by the et13(V411E) allele, MDT-15 remains a critical coactivator in the nhr-49(et13) strain.
et7(P479L) resides in the C-terminal extension that is only conserved in NHR-49 orthologs from Caenorhabditis species, and is adjacent to a predicted 9aaTAD motif (Fig 1). We speculated that the C-terminal extension of NHR-49's LBD might be required for MDT-15 binding. Indeed, C-terminal truncation of the LBD (NHR-49dCT, aa 130-434 instead of aa 130-501) completely abrogated MDT-15 binding (Fig 8C), whereas NHR-49-LBD bait expression was not substantially affected ( Fig 8B). Thus, the C-terminal extension harboring et7(P479L) is required for MDT-15 binding, at least in the Y2H system.

nhr-49 gof mutations differentially affect life span
In the paqr-2 mutant background, strains carrying nhr-49 gof mutations display increased viability and brood size at 16°C as well as amelioration of a morphological phenotype [25]. We wondered whether the gof mutations would affect any other phenotype linked to nhr-49. Specifically, because nhr-49 null mutation or depletion results in a shortened lifespan [15,18,29], we tested whether strains carrying nhr-49 gof mutations display altered population survival, expecting an increase in average lifespan. Interestingly, we found that the three strains displayed distinct phenotypes: nhr-49(et7) was long-lived, nhr-49(et8) was short-lived, and nhr-49 (et13) displayed a wild-type life span (Fig 9). Thus, the three nhr-49 gof mutations evoke distinct, non-equivalent effects on animal life span.

Discussion
C. elegans NHR-49 is a sequence ortholog of HNF4 and an important regulator of lipid metabolism and life span [13][14][15]18,19,23,40]. Svensk et al. previously reported three nhr-49 gof mutations that suppress the cold sensitivity and a morphological phenotype of the paqr-2/adiponectin receptor mutant [25]. The fact that three distinct nhr-49 mutations were isolated provided an opportunity to learn about how individual residues determine NHR-49's transcriptional activity. Using molecular and computational methods, we found that all three gof alleles likely achieve suppression of cold sensitivity through the same mechanism, namely induction of fatty acids desaturation. In contrast, the three alleles are non-equivalent in regards to gene expression and phenotypes not related to this condition. Most notably, nhr-49 (et13) is qualitatively different from nhr-49(et7) and nhr- 49(et8). Mechanistically, the protein LBDs compared with the WT or the E327A dimerization mutant. (C) Comparison of the LibDock scores quantifying the energetics and interactions of the lipid ligand (a higher score indicates more favorable binding) within the mutant protein structures by class (FA = fatty acyls; GP = glycerophospholipids; PK = polyketides; PR = prenol lipids; Sx = sterol lipids (ST), sphingolipids (SP), saccharolipids (SL)). Note that ligands that cannot dock within mutant LBDs are not represented as no LibDock score is generated.
doi:10.1371/journal.pone.0162708.g007 interaction and subcellular localization studies did not identify clear molecular mechanisms of action for the three mutations. However, our computational modeling suggests that the gof missense mutations may alter ligand-binding abilities compared to WT, providing a possible mechanism of action.
In the present study we set out to better characterize three nhr-49 gof mutations. We especially attempted to determine whether the residues they alter represent putative specificity determinants, i.e. whether they lie within surfaces or functional regions that allow NHR-49 to specifically regulate certain downstream processes (e.g. fatty acid desaturation vs. fatty acid βoxidation). As expected due to the origin of these mutations from a suppressor screen for coldsensitivity [25], we observed induction of fat-5 and/or fat-7 by all three gof alleles. Other NHR-49 activated genes were not as consistently activated, with e.g. et7(P479L) failing to induce ddo-2 and dhs-18, and et13(V411E) consistently causing stronger activation than the other two alleles. Collectively, these data suggest that the three nhr-49 gof alleles differ at the very least in how strongly they induce NHR-49 activated genes. Additionally, we observed a striking difference in nhr-49-dependent repression, which was largely unaltered by et7(P479L) and et8 loading control; full immunoblots are shown in S2 Fig. (D) nhr-49(et13) requires mdt-15 to express nhr-49 activated genes. Bars represent average mRNA levels of nhr-49 activated genes in synchronized, L4 stage wild-type N2 and nhr-49(et13) gof worms grown on control RNAi (empty vector) or on mdt-15 RNAi (n = 3). RNA levels were normalized to act-1, tba-1, and ubc-2. Errors bars show SEM. ** represents p<0.01 (unpaired t-test).
doi:10.1371/journal.pone.0162708.g008  by et13(V411E). Although identified as a gof allele, nhr-49(et13) thus also displays some loss-of-function characteristics, rendering this allele qualitatively distinct from the other two. From these studies we conclude that V411 is essential for both NHR-49 driven activation and repression whereas P479 and S432F represent residues that are specific to gene activation. Unbiased expression profiling by mRNA sequencing will substantiate this notion and should also reveal whether any of the gof mutations affect the expression of genes not regulated by wild-type NHR-49.
In line with the quantitative and qualitative differences in gene expression, we found that the three mutant worms strains were non-equivalent with regards to a phenotype investigated, namely life span. Specifically, nhr-49(et7) worms are long-lived, nhr-49(et8) worms are shortlived, and nhr-49(et13) mutants display a normal life span. The differential effects of the gof mutations on a phenotype are distinct from the effects observed previously: Svensk et al. showed that, in the paqr-2 mutant background, all three nhr-49 gof alleles are similarly capable of ameliorating the cold sensitivity and withered tail phenotype [25]. At this time we do not understand what underlies the diversity of longevity phenotypes caused by the three alleles. Perhaps, the comparably mild gene activation in nhr-49(et7) mutants resembles that seen in wild-type worms overexpression NHR-49, a genetic change sufficient to induce a long life span [18]. In contrast, the more substantive and/or qualitatively different changes caused by et8 (S432F) and et13(V411E) might counteract any pro-longevity changes in gene expression, e.g. that of increased fatty acid desaturase expression [13,41,42], to the extent that the ultimate effect is neutral (for et13(V411E)) or even negative (for et8(S432F)).
We attempted to delineate how, molecularly, the three substitutions affect NHR-49 activity. All three affected residues are situated within or near NHR-49's LBD, a domain critical for transcriptional regulation Thus, we hypothesized that interactions with molecular partners such as the coactivator MDT-15 [29] or the dimerization partner NHR-66, involved in NHR-49 dependent repression [13], might be altered. Surprisingly, NHR-66 interaction was only modestly affected and MDT-15, although bound less strongly by NHR-49(et13), remained critical for gene activation in worms expressing this nhr-49 mutant protein. We also failed to detect major changes in NHR-49 protein expression and localization, as determined by translational reporter analysis. It is possible that putative effects of the substitutions on the protein are masked in these experiments relying on extrachromosomal arrays, which result in NHR-49 overexpression; thus, endogenous NHR-49, which is not studied by this approach, may be expressed and/or localized differently in the gof strains. Nevertheless, based on our studies we consider it likely that the three gof alleles appear to increase NHR-49 activity through mechanisms other than changing protein abundance, nuclear accumulation, or interaction with MDT-15. Future experiments such as Y2H screens using point mutant NHR-49-LBDs as bait may help identify mechanisms underlying the increased transcriptional activity.
The LBD is also the binding site for ligand. Because no ligand is yet known for NHR-49, we attempted to delineate the potential influence of the gof missense mutations by in silico ligand docking. To our knowledge this is the first attempt of computational modeling and ligand docking of a C. elegans HNF4α-like NHR, and this methodology strongly suggests that NHR-49 adopts a structure resembling HNF4α, as expected based on the sequence homology [9]. This approach did not identify any particular class of molecules that could explain the allelespecific phenotypes or the et13(V411E)-specificgene repression defects. However, our modeling suggested that the proteins produced by the gof alleles lack the ability to bind saccharolipids, polyketides, and glycerophospholipids. Either of these changes could promote gene activation by NHR-49 missense mutants; in particular, polyketides are of interest as they bind and/or regulate NHRs such as human PPARs and PXRs [43][44][45]. Thus, it is possible that reduced binding of such molecules causes the increased gene activation observed in strains bearing nhr-49 gof mutations. Moreover, it is possible that such molecules might represent ligands for other, NHR-49-related NHRs of C. elegans.
The most striking observation we made was the surprising finding that NHR-49 repressed genes were unexpectedly upregulated in the presumed gof mutant nhr- 49(et13). Our structural homology modeling predicts that the V411E mutation substantially alters the biophysical properties of this residue. Notably, V411 is part of helix 12, and thus constitutes an important part of the repressor-to-activator switch in the classical NHR LBD. The mixed gene expression phenotypes of nhr-49(et13), with both strong activation and de-repression, suggest that V411 is primarily required for repression. In contrast, activation can still be successfully achieved in this mutant. It is possible that this is due to multi-surface contact with coactivators such as MDT-15 through the C-terminal extension including the 9aaTAD motif near P479. No corepressor has yet been identified for NHR-49, and only two corepressor have been described for C. elegans NHRs to date, DIN-1 and GEI-8 [46,47], none of which appear to influence nhr-49 repressed genes (KL and ST, unpublished). The transcriptional lips-6 reporter described in this study should be a useful tool to identify candidate NHR-49 (and NHR-66) corepressors in the future using forward or reverse genetic screens.
In sum, our analysis shows that, although they were identified in the same screen and similarly affected the phenotype used as a readout in said screen, the three nhr-49 gof mutants display qualitative and quantitative differences in gene expression and phenotypes, likely by influencing distinct molecular properties of NHR-49.

In silico structural modeling and docking of NHR-49
All in silico analysis was performed using Accelrys DiscoveryStudio v4.1 (Biovia). The protein sequences for the five NHR-49 isoforms were imported from NCBI (https://www.ncbi.nlm.nih. gov/protein) and sequence aligned. Using the built in BLAST search algorithm (http://blast. ncbi.nlm.nih.gov/Blast.cgi),three high confidence crystal structures for HNF4 from RCSB (http://www.rcsb.org/) were identified (PDBs 4IQR; 1LV2; 1M7W) and the individual isoform amino acid sequences were threaded onto the HNF4 sequence using a Blosum62 scoring matrix. For PPARα and PPARγ, we identified one high confidence crystal structure each (PDB 2REW and PDB 3E00, respectively) that were used. The MODELLER algorithm was used to generate homology models (>10,000 permutations) using the Discrete Optimized Protein Energy (DOPE) method to refine loops. Isoform models were then refined, subjected to energy minimization (CHARMm), and side-chain rotamers were verified. In silico mutagenesis was performed and Gibb's free energy (kcal/mol) was calculated for individual mutant models. The lipid ligand structures were obtained from http://www.lipidmaps.org/data/structure/ using < = C18 as the ontology search and downloaded as.sdf files and prepared for docking at physiological pH. The ligand-binding site was defined for each prepared homology model using nhr-49 gof Mutants Are Distinct coordinates from the HNF4 crystal structure template (PDB 4IQR). High throughput screening of the wild-type and mutant protein models was performed using LibDock [49,50]. Lib-Dock is a rigid docking algorithm in which the protein template is held in constant conformation while the molecular ligand is allowed limited flexibility to generate novel conformations based on atomic, polar, and apolar interactions within the protein cleft. After docking, a final optimization step identifies steric clashes followed by an arbitrarily scored ranking of every potential conformation (aka poses) within the protein target using pairwise atomic clustering and comparative scoring. This involves bond interaction analysis, which prioritizes and refines hydrogen bonding. Similar to a piecewise linear potential the docking score is generated by summation of potential energetics of interacting atoms in the protein-ligand complex relative to bond distance. The atoms are divided into four atom types-apolar, acceptor, donor, and donor/acceptor-and the score between interacting atoms is scored using either a hydrogen binding potential or a steric potential. More favorable bond interactions and bond distances result in a higher LibDock score. For reference, linoleic acid, considered an endogenous ligand for HNF4α [51], generated a LibDock score of 113.42 when docked onto the HNF4α structure. The subset of NHR-49 ligands analyzed in Fig 7B and  C. elegans growth and strains C. elegans strains were cultured at 20°C, as described [52]. We used nematode growth medium (NGM)-lite (0.2% NaCl, 0.4% tryptone, 0.3% KH 2 PO 4 , 0.05% K 2 HPO 4 ) agar plates seeded with Escherichia coli strain OP50 except for feeding RNAi, which was performed using HT115 bacteria. Feeding RNAi was performed as described [40]. The empty vector (negative control) and the mdt-15 RNAi clone (plate 74, well C09) are from the Ahringer library [53] and were sequenced prior to use.
To generate synchronized worm populations, embryos were isolated by sodium hypochlorite treatment and allowed to hatched overnight on unseeded NGM-lite plates; arrested, synchronized L1 larvae were then grown to the desired stage, as indicated, and growth times were adapted to ensure developmental synchronicity of slow-growing mutants.

RNA isolation and quantitative real-time PCR
Total RNA was extracted from developmentally synchronized mid-L4 stage worms as assessed by vulval morphology. RNA isolation and qPCR were performed as described [40]. We used ttests (two-tailed, unequal variance) to calculate statistical significance of gene expression changes between different strains (Gaussian distribution). qPCR primers were designed with Primer3web (bioinfo.ut.ee/primer3/) and tested on serial cDNA dilutions to ensure PCR efficiency. Primer sequences are listed in S5 Table. Yeast-two-hybrid assays and immunoblots Yeast-two-hybrid assays to study protein-protein interactions and immunoblots to ascertain protein expression were performed as described [29,40].

DIC and fluorescence microscopy
For microscopy, worms were transferred onto 2% (w/v) agarose pads, which contained NaN 3 . We captured images on a CoolSnap HQ camera (Photometrics) attached to a Zeiss Axioplan 2 compound microscope, and used MetaMorph Imaging Software with Autoquant 3D digital deconvolution for image acquisition.

Lifespan assays
Lifespan assays were conducted as described [29,40]. Day 2 of adulthood was used as t = 0 for lifespan analysis. GraphPad Prism 6 software was used for statistical analysis. P values were calculated using the Log-rank (Mantel-Cox) method. Statistics for all lifespan assays are listed in S4 Table. Supporting Information