A dicer-related helicase opposes the age-related pathology from SKN-1 activation in ASI neurons

Significance

Unlike activation, an understudied fundamental question across biological systems is how to deactivate a pathway, process, or enzyme after it has been turned on. The irony that the activation of a transcription factor that is meant to be protective can diminish health was documented by us at the organismal level over a decade ago, but it has long been appreciated that chronic activation of the human ortholog of SKN-1, NRF2, could lead to chemo- and radiation resistance in cancer cells. A colloquial analogy to this biological idea is a sink faucet that has an open valve without a mechanism to shut it off, which will cause the sink to overflow. Here, we define one such off mechanism for SKN-1 activity.

Abstract

Coordination of cellular responses to stress is essential for health across the lifespan. The transcription factor SKN-1 is an essential homeostat that mediates survival in stress-inducing environments and cellular dysfunction, but constitutive activation of SKN-1 drives premature aging thus revealing the importance of turning off cytoprotective pathways. Here, we identify how SKN-1 activation in two ciliated ASI neurons in Caenorhabditis elegans results in an increase in organismal transcriptional capacity that drives pleiotropic outcomes in peripheral tissues. An increase in the expression of established SKN-1 stress response and lipid metabolism gene classes of RNA in the ASI neurons, in addition to the increased expression of several classes of noncoding RNA, define a molecular signature of animals with constitutive SKN-1 activation and diminished healthspan. We reveal neddylation as a unique regulator of the SKN-1 homeostat that mediates SKN-1 abundance within intestinal cells. Moreover, RNAi-independent activity of the dicer-related DExD/H-box helicase, drh-1, in the intestine, can oppose the effects of aberrant SKN-1 transcriptional activation and delays age-dependent decline in health. Taken together, our results uncover a cell nonautonomous circuit to maintain organism-level homeostasis in response to excessive SKN-1 transcriptional activity in the sensory nervous system.

A central node in the response to xenobiotics and oxidative stress is the cytoprotective transcription factor SKN-1/NRF2 (skinhead-1/NF-E2 related factor 2), which binds to antioxidant response elements in the promoters of genes, including phase II detoxification and antioxidant synthesis enzymes, that are important for survival under a wide range of stress conditions (1–6). SKN-1 also plays critical roles in embryonic development by specifying the early blastomere identity and formation of the pharyngeal and intestinal tissues (7–11). Recently, SKN-1 activity has been demonstrated to coordinate cellular metabolism, particularly lipid metabolism and mobilization (12–18). Taken together, these studies highlight the complex roles that SKN-1 plays in organismal homeostasis across the lifespan.

Despite the essentiality for development and in the response to stress with age, constitutive activation of SKN-1 has been demonstrated to be pleiotropic for health (13, 15, 17–21). Although the molecular basis of this health detriment remains elusive, turning off SKN-1 activity is equally important to turning it on. SKN-1 abundance and turnover are regulated by the ubiquitin–proteasome system, mediated by the CUL-4/WDR-23 (Culin-4/WD repeat protein-23) ubiquitin ligase complex. While the loss of skn-1 is maternal-effect embryonic lethal (7, 22), and whole-life loss of wdr-23 leads to constitutive SKN-1 activation that results in sickness, the adult-specific inactivation of wdr-23 can lead to increased lifespan (23). Relatedly, gain-of-function (gf) mutations in skn-1 lead to enhanced resistance to acute exposure to oxidative stress early in life, but this stress resistance phenotype is lost when animals reach the post-reproductive period (15). Tied to these changes in stress resistance and reproduction, these skn-1gf alleles also result in the loss of somatic lipid reserves and a significant diminishment of adult lifespan (24) that is tied to the activation of pathogen resistance responses (13). Considering these findings, SKN-1 plays critical roles from embryogenesis through death, which requires sophisticated regulatory mechanisms.

Fifteen years ago, SKN-1 was shown to function in ASI neurons to mediate longevity responses to caloric restriction (25), but multicopy transgenic reporters suggest that SKN-1 activity in nonneuronal tissues is activated in response to stress (11). Moreover, use of transgenic reporters designed to measure SKN-1 transcriptional responses are activated in cells beyond ASI neurons that suggest that either multicopy transgenics do not accurately reflect physiological SKN-1 responses or that cell nonautonomous pathways responding to SKN-1 activation exist but have yet to be elucidated.

To address these questions, we employed CRISPR-mediated genome editing to tag the endogenous skn-1 locus with Green Fluorescent Protein in wild-type and skn-1gf mutant animals and additionally, replaced the commonly used multicopy extrachromosomal transgenic arrays with single-copy tissue-specific constructs that more accurately measure physiological responses. We found that SKN-1 expression, even in animals with genetically encoded constitutive SKN-1 activation, remains detectable only in ASI neurons, which is sufficient to coordinate changes in stress and metabolic homeostatic responses. Comparing the transcriptomic landscapes in whole animals, and specifically transcriptional responses to SKN-1 activation specifically in ASI neurons, we further found that the pleiotropic outcomes from constitutive SKN-1 activation can be delayed by the dicer-related helicase that can temporarily mitigate the increased transcriptional response from SKN-1 activation.

Results

SKN-1 Activation Phenotypes from Two ASI Neurons.

Previously, activation of SKN-1 has been demonstrated to drive the age-dependent reallocation of somatic lipids such that somatic lipids are depleted while germline lipids are maintained during reproduction that subsequently alters organismal stress resistance and reproduction (15). In wild-type animals, this phenotype is one of the earliest detectable responses upon exposure to pathogenic bacteria like Pseudomonas aeruginosa (13). Although the environment is a potent driver of this response, genetic mutations that activate SKN-1, including gf mutations in skn-1 (SI Appendix, Fig. S1A) are potent drivers of this change in lipid partitioning (18, 26, 27); the causality of this single nucleotide mutation on lipid partition was confirmed by genome editing (SI Appendix, Fig. S1 B and C).

To understand how constitutive SKN-1 activation drives premature aging and diminished health, we used CRISPR-Cas9 genomic editing to tag the SKN-1 locus in wild-type animals and animals harboring the skn-1gf(lax188) allele; hereafter referred to as skn-1gf. We used this endogenously tagged version of SKN-1 and SKN-1gf to identify direct transcriptional targets by ChIPseq. RNAseq analyses reveal upregulation of several gene classes including phase II detoxification, host defense factors and lipid metabolism in skn-1gf animals. Additionally, significant enrichment of promoters regulating proteostasis factors (e.g., ribosomal subunits, ubiquitin ligases, and chaperones) were recovered by ChIPseq (Fig. 1 A and B). These direct targets included 70 ribosomal subunits (rps-1 through rps-27 and rpl-1 through rpl-43), ubl-1, ubq-1, lgg-2, hsp-1, hsp-3, hsp-6, and ndk-1 and several genes enriched in neurons (Fig. 1B).

Fig. 1.

Cell autonomous activity of SKN-1gf in ASI neurons. (A) Differentially expressed genes between WT and skn-1gf (Blue) were overlapped with SKN-1gf occupied loci (Yellow). ChIP-seq was normalized to a no antibody control IP, while the DE analysis was performed by comparing skn-1gf to WT transcript levels. Genes with altered expression in skn-1gf mutants and identified with occupancy of SKN-1gf on the promoter region were analyzed in WormCat 2.0 (list of genetic loci can be found in Dataset S1) to reveal enriched classes (B). Expression of SKN-1wt-GFP (C) and SKN-1gf-GFP (D) are indistinguishable and restricted to ASI neurons; costained with DiI (red) that marks ciliated neurons; arrows designate GFP in ASI cell bodies (green). (E) WormCat 2.0 analysis of genes up-regulated in FACS-enriched ASI neuron populations from skn-1gf animals.

We investigated whether the low number of recovered targets bound by SKN-1 could be explained by the expression pattern of the SKN-1gf protein. Surprisingly, and unlike animals with activated SKN-1 stemming from loss of the negative regulator wdr-23 (20, 23, 28–34) or animals exposed to exogenous toxicants (33, 35–37) that have SKN-1 stabilized in the intestine, the expression of SKN-1wt-GFP and SKN-1gf-GFP were only detectable in ASI neurons (Fig. 1 C and D); ASI-specific expression was confirmed by coexpression in animals with ASI neuron expression of mCherry from the gpa-4 promoter (25, 38) ( SI Appendix, Fig. S1N).

Previous transcriptional profiling of animals with activated SKN-1 was performed using RNA samples derived from whole animals. Since SKN-1wt and SKN-1gf protein were only detectable in ASI neurons, we enriched ASI neuronal populations and performed RNAseq. Among the gene targets specifically up-regulated in ASI neurons of skn-1gf mutants were canonical glutathione s-transferases (gst-4 and gst-30), pathogen response genes (irg-5, C55B7.3), and several neuron enriched genes (gcy-7, C50C3.19, nspd-10, and zig-2) (Fig. 1E and Dataset S1). Taken together, these data suggest that skn-1gf activity may change steady-state stress homeostats in ASI neurons.

Neddylation Regulates Intestinal SKN-1 Stability.

SKN-1 coordinates stress adaptation in response to a variety of cellular insults (2, 11–13, 15, 17–20, 39–47) and although skn-1gf mutant animals are stress resistant, several details surrounding the molecular basis of this response remain unclear. Traditionally, when a stressor is encountered, SKN-1 protein is stabilized, translocated to the nucleus, and then mediates the transcription of genes that will mitigate the current stress condition (11, 30, 34). To reconcile the difference in the robust pan-tissue transcriptional response measured in the skn-1gf mutant with the inability to detect stabilized SKN-1gf expression outside of the ASI neurons (Fig. 1 C and D), we measured several characteristics of SKN-1 dynamics. First, we tested the stabilization of SKN-1 in response to oxidative stress by acute exposure to hydrogen peroxide (H₂O₂), which resulted in the predicted accumulation of SKN-1wt-GFP and SKN-1gf-GFP in the intestine of treated animals (Fig. 2 A–D), while mock-treated animals did not display intestinal accumulation. Despite robust resistance to hydrogen peroxide exposure, we noted a delayed accumulation of SKN-1gf-GFP in the intestine as compared to SKN-1wt-GFP, suggesting an improved capacity to turn over the SKN-1gf protein outside of the nervous system.

Fig. 2.

Neddylation regulates nuclear SKN-1 stabilization in the intestine. Unlike mock treatment (A and B), acute exposure to hydrogen peroxide (H₂O₂) drives nuclear accumulation outside of the ASI neurons (white arrows) for SKN-1wt-GFP (C) and SKN-1gf-GFP (D) within intestinal nuclei (yellow arrows). As compared to control RNAi treated animals (E and F), RNAi of uba-1 (G and H) and ned-8 (I and J) stabilizes SKN-1gf-GFP but not SKN-1wt-GFP within intestinal nuclei. (K) Cartoon of the role of ubiquitinylation and neddylation on nuclear SKN-1 stability. All RNAi experiments were conducted with n = 50 N = 3, representative worms are shown.

The stabilization of SKN-1 in response to oxidative stress is linked to its turnover by the CUL4-DDB-1-WDR23 E3 ubiquitin ligase that targets SKN-1 to the ubiquitin–proteosome system (UPS) for degradation (20, 34, 42, 48). As such, we next exposed animals expressing SKN-1wt-GFP or SKN-1gf-GFP to wdr-23 RNA interference (RNAi) and observed accumulation of SKN-1wt-GFP and SKN-1gf-GFP in the intestine (SI Appendix, Fig. S2 A–D). This result confirms that the UPS-mediated control of SKN-1 protein is not perturbed in skn-1gf mutants, which supports previous findings that the increase in transcriptional output stemming in skn-1gf mutants is additive with loss of wdr-23 (18).

Although regulation of SKN-1 by the CUL4/DDB1/WDR23 E3 ligase and the ubiquitin–proteasome system is well established, the proteostasis network is tightly regulated by the coordinated actions of several pathways (49), including the post-translation modification of proteins by the ubiquitin-like molecules NED-8/NEDD8 (50) and SMO-1/SUMO (Small Ubiquitin-like Modifier) (51). We used RNAi to inhibit ubiquitinylation, neddylation, and sumoylation components of the proteostasis machinery to examine their effects on SKN-1wt and SKN-1gf mutants (Fig. 2 E–J and SI Appendix, Fig. S2 E–N). Intriguingly, compared to control RNAi (Fig. 2 E and F), RNAi targeting uba-1, the E1 ubiquitin activating enzyme (Fig. 2 G and H), and ned-8 (neural precursor cell expressed, developmentally down-regulated 8), the ubiquitin-like modifier (Fig. 2 I and J), stabilized SKN-1gf-GFP but not SKN-1wt-GFP within the nucleus of intestinal cells. The preferential stabilization of SKN-1gf-GFP was not observed with RNAi targeting the sumoylation pathway (SI Appendix, Fig. S2 I–N), which suggests the observed differential stabilization of SKN-1gf relative to SKN-1wt is not a generalized sensitivity to proteostatic stress but instead a specific response to the ubiquitin and neddylation pathways (Fig. 2K).

SKN-1gf Activity in ASI Neurons Drives Cell Nonautonomous Decline with Age.

Our finding that SKN-1gf protein is only detectable outside of the ASI neurons only when proteostasis is perturbed raised questions as to where SKN-1gf activity is needed to establish early-life stress resistance and eventual health decline if left unchecked. To generate physiologically relevant models for tissue-specific manipulation of SKN-1wt and SKN-1gf activities, we developed several new strains (Fig. 3A) for tissue-specific expression under established promoter elements as well as tissue-specific degradation by tagging the endogenous skn-1 locus with an auxin-inducible degron (AID) tag for auxin-mediated degradation of SKN-1wt or SKN-1gf protein exclusively in tissues expressing TIR1 (52).

Fig. 3.

SKN-1 activity in ASI neurons mediates peripheral stress responses. (A) Cartoon representation of strains for tissue-specific regulation of skn-1gf expression. (B) Expression of skn-1gf from gpa-4p (ASI neurons), but not vha-6p (intestine), or rgef-1p (pan-neuronal) can establish resistance to acute exposure to H₂O₂ n = 150 N = 3 per condition analyzed by one-way ANOVA *(P < 0.05) ****(P < 0.0001). (C) Tissue-specific expression of skn-1gf results in the cell nonautonomous activation of the gst-4p::gfp reporter, Green (bright reporter activation), Yellow (dim reporter activation), Red (No detectable reporter activation). Tissue-specific expression of skn-1gf is not sufficient to drive Asdf (D). Pan-neuronal, intestinal, and ASI neuron-specific degradation of SKN-1gf can partially suppress somatic lipid depletion (Asdf); n = 300; N = 3 per condition analyzed by one-way ANOVA **(P < 0.01) ***(p < 0.001) ****(p < 0.0001) (E).

Considering the well-established role that SKN-1 plays in cytoprotection against oxidative stress (11, 53), we tested whether the expression of skn-1gf in a single tissue was sufficient to establish resistance to hydrogen peroxide as previously documented for the skn-1gf mutant (15). Both skn-1B and skn-1C regulate oxidative stress resistance and longevity (11, 25, 54, 55); however, because the lax188 gf mutation does not alter the SKN-1b polypeptide (SI Appendix, Fig. S1A), we focused on the skn-1C isoform. Although previously thought that only SKN-1b activity in ASI was sufficient to drive oxidative stress resistance (11, 25), restricted expression of skn-1gf in ASI neurons was sufficient to recapitulate the resistance to acute exposure to hydrogen peroxide (Fig. 3B), suggesting that the gf allele is active in ASI and can stimulate organism-level protection from oxidative insult.

The ability of ASI-restricted expression of the skn-1gf c-isoform to mediate an organism-level oxidative stress response suggests a cell nonautonomous action stemming from SKN-1 activation. Although we are only able to detect SKN-1wt-GFP and SKN-1gf-GFP in the ASI neurons, our initial isolation of the skn-1gf mutant was due to the robust activation of the gst-4p::gfp reporter that was induced across multiple tissue types in the animal (18). As such, we next tested whether tissue-specific expression of skn-1gf was sufficient to activate the gst-4p::gfp reporter cell nonautonomously (Fig. 3C and SI Appendix, Fig. S3 A–D). Although muscle-specific expression (myo-3p) of skn-1gf resulted in muscle restricted activation of gst-4p::gfp, ASI specific expression, under the control of the gpa-4 promoter, induced gst-4p::gfp activation beyond the two ASI neurons, including expression in the intestine and body wall muscle. Animals with pan-neuronal expression (rgef-1p) displayed a similar pattern of gst-4p::gfp activation, except that, multiple neurons displayed activation of the reporter. In contrast, intestine-specific expression (vha-6p) resulted in gst-4p::gfp reporter activation only in the intestine and body wall muscle and expression of a mCherry reporter alone in ASI was not sufficient to drive the activation of the gst-4p::gfp reporter (SI Appendix, Fig. S3).

We next examined whether cell type–specific expression of skn-1gf could stimulate the age-dependent somatic depletion of fat (Asdf). Although SKN-1gf activity in ASI neurons was sufficient to recapitulate oxidative stress resistance, it was not sufficient to induce somatic lipid depletion at day 3 of adulthood (Fig. 3D and SI Appendix, Fig. S3 E–G); however, we did document a trend toward increased somatic lipid depletion in the population at day 5 of adulthood (SI Appendix, Fig. S3 H–K).

Although we could not identify sufficiency for any single tissue for somatic lipid depletion, we next examined which tissues, if any, were necessary for somatic lipid depletion by auxin-mediated degradation. Based on previous reports with this system (52, 56), we expected and observed some auxin-independent degradation effects on lipid depletion (SI Appendix, Fig. S3L). However, the addition of auxin greatly enhanced the suppression of lipid depletion in the intestine-specific TIR1 strain by 21%, while pan-neuronal and ASI-specific TIR1 strains resulted in a 40% and 16% suppression of lipid depletion, respectively (Fig. 3E, SI Appendix, Fig. S3 L–S, and Dataset S2). Collectively, these results reveal that skn-1 activity in neurons are required for the phenotypes associated with constitutive SKN-1 activation and can initiate a cell nonautonomous response in peripheral tissues like the intestine, which normally removes activated SKN-1 through ubiquitin and neddylation proteostasis pathways.

DRH-1 Activation Delays Healthspan Decline from SKN-1 Activity.

Our data suggest that activation of skn-1 in the ASI neurons is sufficient to drive systemic changes in oxidative stress resistance throughout the organism and is needed, at least in part, for the somatic lipid depletion that accompanies SKN-1 activation with age. To identify mediators of the peripheral response to skn-1gf activity in ASI neurons, we performed an unbiased genetic screen with ethylmethanesulfonate to recover suppressors of the activation of the gst-4p::gfp reporter outside of the nervous system in skn-1gf mutants (Fig. 4A). To our surprise, we recovered a suppressor mutant in the F1 generation of the screen and confirmed the dominant nature of this allele by subsequent backcrossing into the unmutagenized parental strain. We identified insertion–deletion polymorphisms (57) linked to the dominant suppressor mutation that mapped to the center of chromosome IV (LGIV) (Fig. 4B). Genome-wide sequencing of the suppressor mutant genomic DNA revealed twenty-six missense mutations within this region (SI Appendix, Fig. S4A). Only RNAi targeting the dicer-related helicase gene, drh-1, restored the peripheral gst-4p::gfp expression observed in the parental skn-1gf mutant (Fig. 4 C and D and SI Appendix, Fig. S4 B–G), which also suggests the drh-1gf(lax257) mutation is gf; hereafter referred to as drh-1gf. We confirmed causality and dominance of the drh-1gf allele by transgenesis (Fig. 4 E and F). The drh-1 locus encodes for two predicted isoforms, DRH-1A and DRH-1B that are 1,037 and 779 amino acids in length, respectively. The drh-1gf(lax257) mutation changes glycine 474 in DRH-1A and glycine 216 in DRH-1B to arginine, which are on the surface of the predicted Caenorhabditis elegans DRH-1 protein (Fig. 4G) that resembles the bridging domain found in the mammalian RIG-I that participates in RNA recognition (58, 59).

Fig. 4.

DRH-1 activation delays SKN-1gf-dependent healthspan decline. (A) Cartoon schematic of EMS genetic screen for suppressors of skn-1gf activation of gst-4p::gfp. (B) Genetic linkage maps the lax257 suppressor to LGIV. As compared to control RNAi (C) drh-1 RNAi abolishes the suppression of drh-1gf (D). Ectopic expression of drh-1gf (E) suppresses skn-1gf activation of gst-4p::gfp as compared to nontransgenic siblings (F). (G) Predicted structure and amino acid substitution (wt-cyan; gf-magenta) in DRH-1gf. drh-1gf suppresses the somatic lipid depletion phenotype of skn-1gf mutant at day 3 (H) but not day 5 (I) of adulthood; n = 300; N = 3 per condition comparisons were made by one-way ANOVA ****(P < 0.0001). The resistance to acute H₂O₂ by skn-1gf exposure at day 1 of adulthood (J) and the sensitivity at day 3 of adulthood (K) is reversed by drh-1gf. (L–N) The suppression of the impaired movement phenotype of skn-1gf by drh-1gf at the day 1 stage (L) is progressively abrogated at day 3 (M) and day 5 (N) of adulthood; n = 50; N = 3 per condition. Oxidative stress assay was analyzed by one-way ANOVA **(P < 0.001) ****(P < 0.0001), n = 100; N = 3 per condition.

We next examined the impact of the drh-1gf mutation on the age-related healthspan phenotypes influenced by skn-1gf; age-dependent somatic lipid depletion (Fig. 4 H and I, SI Appendix, Fig. S4 H–Q, and Dataset S2), oxidative stress resistance (Fig. 4 J and K), lifespan (SI Appendix, Fig. S4U), and movement (Fig. 4 L–N and Dataset S3). skn-1gf drh-1gf double mutant animals display a significant reduction of somatic lipid depletion at day 3 of adulthood (Fig. 4H), whereas the phenotype nears complete penetrance in skn-1gf mutants. Remarkably, the suppression of somatic lipid depletion was not maintained at day 5 of adulthood where animals harboring the drh-1gf allele were indistinguishable from age-matched skn-1gf single mutant animals (Fig. 4I); thus, the effect of the drh-1gf mutation is to delay the impact of the skn-1gf allele. In addition to changing age-dependent distribution of lipids, the skn-1gf mutation has a paradoxical effect on oxidative stress resistance where skn-1gf mutant animals are more resistant to acute exposure to oxidative stress at day 1 of adulthood, as compared to WT (Fig. 4J), but at day 3 of adulthood, when somatic lipid depletion is complete, skn-1gf mutant animals are much more sensitive to the same exposure of oxidant than age-matched wild-type animals (Fig. 4K). The drh-1gf mutation partially reversed the effects of the skn-1gf allele back to wild type (Fig. 4 J and K). The drh-1gf mutation did not reverse the shortened lifespan displayed in skn-1gf mutants (SI Appendix, Fig. S4U). Finally, skn-1gf mutant animals display movement defects compared to wild-type animals that are characterized by reduced speed when crawling starting early in life (Fig. 4L and Dataset S3) that persists throughout adulthood (Fig. 4 M and N). Early on, skn-1gf drh-1gf mutant animals show an intermediary movement defect that is indicative of partial suppression of the skn-1gf mutation which becomes progressively worse and more fully resembles the skn-1gf single mutants later in adulthood. Taken together, these results reveal the ability to delay the diminished health stemming from SKN-1 activation with age by an activating mutation in the dicer-related helicase, drh-1.

Intestinal DRH-1 Reduces Transcriptional Response from SKN-1 Activation.

Universally represented across all genetic mutants with enhanced SKN-1 activity is diminished health with age. Although previous work has demonstrated that the reduced healthspan is associated with the transcriptional activity of SKN-1 (13), our understanding of why constitutive transcription is debilitating requires additional detail. We first examined whether the suppression of the gst-4p::gfp transcriptional reporter outside of the nervous system by drh-1gf was not an artifact of this simple transgenic reporter. We measured the expression of endogenous gst-4 transcripts and multiple phase II detoxification genes (60) (Fig. 5 A–G and Dataset S4), that are strongly induced in skn-1gf (13, 18) and regulated by SKN-1 under normal conditions (61, 62) and found that they were significantly reduced in the skn-1gf drh-1gf double mutant and their expression pattern opposes the age-related change in the skn-1gf single mutant animals.

Fig. 5.

Intestinal DRH-1 activation reduces transcriptional load of activated SKN-1. (A–G) drh-1gf (magenta) suppresses the activation of established SKN-1 targets in skn-1gf mutant animals (blue), Blue lines represent DE between WT and skn-1gf, and magenta lines represent DE between skn-1gf drh-1gf and skn-1gf. (H) drh-1gf mutations abolishes the increased sensitivity of skn-1gf mutant animals to RNAi inhibition of transcriptional regulators n = 100 N = 3. (I) WormCat 2.0 analysis of genes activated by skn-1gf and suppressed by drh-1gf. (J–Q) drh-1gf suppresses the increased expression of ncRNA in skn-1gf mutants. The activation of the gst-4p::gfp in skn-1gf animals is suppressed by intestinal expression of drh-1gf (R and S). (T–V) drh-1gf suppresses the increased expression of signaling molecules. (W) Cartoon schematic of cell nonautonomous signaling by skn-1gf in ASI neurons and drh-1gf in the intestine. A simple linear regression model was used to compare each of the lines with P < 0.05 considered significant.

Next, we examined how drh-1gf influences other transcriptional stress responses in the context of constitutive SKN-1gf activity. Previous work identified several RNAi conditions that induce SKN-1 activation (23, 47, 61, 62), including RNAi targeting RNA polymerases themselves and regulators of transcription (61). RNAi targeting several transcriptional regulators (e.g., rpc-1, F30A10.9, and tif-1A) results in hyperactivation of the gst-4p::gfp reporter which does not occur in the context of the drh-1gf mutation (Fig. 5H and SI Appendix, Fig. S5A). As such, drh-1gf attenuates the response to increased transcriptional load resulting from the activation of cytoprotective transcription factors like SKN-1.

DRH-1 is an ortholog of the mammalian retinoic acid-inducible gene I (RIG-I)-like receptors which can detect viral double-stranded RNA (dsRNA) and promote antiviral defense. In C. elegans, drh-1 is required for antiviral RNA interference (63–65) but also plays a separable signaling role in the regulation of transcription immune responses to viral infection (66). We tested RNAi targeting the CEr1 regulator of viral RNA, cerv-1, which strongly induces gst-4p::gfp in the skn-1gf mutant but not in skn-1gf drh-1gf mutants (Fig. 5H and SI Appendix, Fig. S5A). Additionally, we confirmed that the reduced activation of the gst-4p::gfp reporter was not dependent on the activation of the RNAi-related machinery and as such, unlikely a result of RNAi-induced transgene silencing (SI Appendix, Fig. S5B). Taken together, these results support a role for drh-1 in the regulation of SKN-1-dependent transcriptional homeostasis.

In addition to stress response and cellular metabolism genes that are activated in skn-1gf mutants but repressed by drh-1gf (SI Appendix, Fig. S5C), we noted an increase in the expression of multiple transcripts enriched for double-stranded secondary structure (e.g., non coding RNAs (ncRNAs), 21U-RNA, tRNA, snRNA, snoRNA, and pseudogenes) in skn-1gf mutants but are repressed in the skn-1gf drh-1gf double mutant (Fig. 5 I–Q and SI Appendix, Fig. S5C). Taken together, these data reveal a unique model where activated DRH-1 works to oppose the increased transcriptional load stemming from constitutive SKN-1 activation (Fig. 5W and SI Appendix, Fig. S5D). Although the drh-1gf allele can temporarily counteract SKN-1 activity, if SKN-1 is not attenuated, the ability of DRH-1 to counterbalance SKN-1 activation is eventually overcome. We noted that the classes of noncoding RNAs were not significantly enriched in the genes upregulated from the ASI neuron population suggesting that the regulation of the ncRNAs occurs outside of the nervous system. As such, we were curious where skn-1gf activity was needed to suppress the transcriptional load in the skn-1gf mutant background. DRH-1 activity in the intestine has been linked to its RNAi-independent roles in gene expression (66–68). We expressed drh-1gf exclusively in the intestine, which mitigated much of the skn-1gf-induced activation of the gst-4p::gfp reporter (Fig. 5 R and S). We noted a significant change in the expression of multiple genes that mediate signaling across tissues mediated by the drh-1gf allele in the context of the skn-1gf background (Fig. 5 T–V) that might mediate the cell nonautonomous responses to activated SKN-1. Our results demonstrate that the genetic activation of the dicer-related helicase opposes the activity of SKN-1 in ASI neurons to reestablish transcriptional homeostasis at the organismal level that improves age-dependent health status (Fig. 5W and SI Appendix, Fig. S5D).

Discussion

The paradoxical finding that the activation of cytoprotective transcription factors can be the cause of diminished health over the lifespan points to the complexity of the homeostats that have evolved to ensure survival. Our work supports the notion that turning off cytoprotection when it is not needed is perhaps equally as important as our ability to turn it on when it is; analogous to turning a faucet on and off to fill a sink with water. SKN-1 is the C. elegans ortholog of the NRF2/NFE2L2 family of transcription factors in mammals. In cancer patients, activation of NRF2/NFE2L2 in transformed cells results in chemo- and radiation therapy resistance (69–71) because the players in phase II detoxification that are normally induced by NRF2 in normal cells to protect against toxic conditions protect the cancer cells from treatment. Beyond this action of the protein products of SKN-1/NRF2, our study revealed that several noncoding RNA transcripts that are induced when SKN-1 is activated may also drive phenotypic decline with age.

Although first reported a decade ago, the molecular basis of how a single amino acid substitution in SKN-1 renders this cytoprotective transcription factor constitutively active remains elusive. Our inability to detect activated SKN-1 stabilized at steady state in cells beyond the ASI neurons was suggestive that either SKN-1 is not functioning in those cell and tissue types or that the skn-1gf mutation alters the homeostatic turnover of the gf protein. Our finding that mildly impairing the ubiquitin–proteasome system allows for the detection of intestinal nuclear-localized SKN-1gf but not SKN-1wt supports the latter model. Moreover, the differential hypersensitivity to changes in the ubiquitin-like modifiers of the proteostatic network reveals a previously undescribed role for the posttranslational modification of neddylation on SKN-1 activity. The stabilization of SKN-1gf in the nucleus when neddylation or ubiquitinylation is impaired supports recent evidence demonstrating that alternating chains of NEDD8-Ubiquitin can mediate nuclear proteotoxic stress (72). In addition, NEDD8 conjugation to proteins can decrease stability and is an established mechanism to regulate transcription factor function (73) and neddylation can activate cullin ring ligases, like Cul4, through conformational changes which promote cellular proteostasis (74). Collectively, the observation that proteostatic network has differential effects on SKN-1wt and SKN-1gf is intriguing and suggests the regulation of SKN-1 at the protein level has the capacity for additional regulation.

Beyond the intestine, our work also reveals that while activation of SKN-1 in a single tissue is insufficient to recapitulate all the phenotypes associated with skn-1gf, we found that neurons are essential for this response. We describe how the activation in the two ASI sensory neurons is required to initiate a cell nonautonomous program in the periphery that drives age-related pathology. Specifically, the activation of SKN-1 in the ASI neurons results in a change in the proteostatic balance within intestinal cells that drives the rapid depletion of intracellular lipid stores in these cells to fuel reproduction (15), which supports the recent model of Morimoto and colleagues that the proteostasis network is critical for reproductive success (75).

While previous work has used transcriptomics to determine differentially expressed genes between WT and skn-1gf animals, these results were focused on lipid metabolism and pathogen stress response genes (13). Here, we present a comprehensive multiomic analysis of animals harboring the skn-1gf allele. Previous examinations of SKN-1 in the ASI neuron pair had made observations that it only functioned in determining longevity in a caloric-restricted state (25), but our ASI-enriched transcriptomic analysis reveals that the repertoire of SKN-1-regulated transcripts includes a larger battery of stress response genes. This finding, in consideration of the apparent lack of SKN-1gf stabilization in other nuclei when additional stressors are absent, further confirms the complexity of the cell nonautonomous nature of the response in ASI neurons and communication needed to peripheral tissues.

Furthermore, our work here provides insight into the function of SKN-1 isoforms. SKN-1 has four predicted isoforms (11, 54). SKN-1c has previously been demonstrated to accumulate within intestinal nuclei during oxidative, xenobiotic, and pathogenic stress responses (11, 54, 76) while SKN-1a has a transmembrane domain that allows its association with the mitochondrial (11, 18) and ER (11, 46) membranes that may facilitate organelle specific stress responses. SKN-1b was thought to be the sole isoform accumulating within the ASI neurons and modulating longevity in response to caloric deficit (25); however, our findings reveal a gf mutation that only affects SKN-1a and SKN-1c can drive physiological responses from the ASI neurons and activation of SKN-1a or SKN-1c in the intestine is insufficient. Although the differences in dosage and expression from the single-copy-edited genomes used here and multicopy arrays previously used are likely important, taken together, these results suggest that the rigid roles previously assigned to the SKN-1 isoforms are likely more fluid.

We also demonstrate how intestinal activation of the dicer-related helicase, drh-1, a regulator of both RNA interference and transcriptional responses to pathogen attack that was previously not connected to SKN-1 activity, can oppose the phenotypes of SKN-1 activation. Animals with constitutive activation of SKN-1 display increased expression of several RNAs with complex secondary structures and are also sensitive to changes in the expression of regulators of multiple RNA polymerases and the cellular proteostasis machinery (47); a response that is abolished by the drh-1gf allele. Thus, DRH-1 acts as a regulator of this RNA homeostat, acting as the drain valve in our analogy and can prevent the sink from overflowing. Intriguingly, activation of DRH-1 only provides a temporary relief to oppose the pathology stemming from SKN-1 activation, since if SKN-1 activity is not turned off, the negative consequences of transcriptional increase still emerge, albeit delayed. Remarkably, the activation of skn-1gf in two neurons is sufficient to drive this systemic response that drives multisystem functional decline with age.

Several key questions remain but our results promote the practice of matching exogenous manipulations of cellular cytoprotective responses to physiological signals stemming from cell nonautonomous pleiotropic consequences in distal cells. It remains possible that SKN-1 activation in two or more tissues or that, multiple isoforms, which have different subcellular roles, are necessary to recapitulate the age-related decline in health observed in the skn-1gf genetic mutants. The partial mitigation of the skn-1gf response when drh-1gf is expressed only in the intestine further exemplifies the complexity of SKN-1 signaling and reveals a new layer of cell nonautonomous control in maintaining organismal homeostasis. Moreover, our work provides a framework to refine our models and include alternative approaches that can alleviate the consequences of dysregulated transcriptional activities.

Materials and Methods

C. elegans Strains and Maintenance.

C. elegans were raised on 6-cm nematode growth media (NGM) plates supplemented with streptomycin and seeded with OP50. All worm strains were grown at 20 °C and unstarved for at least three generations before being used. Strains used in this study include: WT, N2 Bristol strain; SPC168, dvIs19[gst-4p::gfp]; skn-1gf(lax188); SPC572, dvIs19[gst-4p::gfp]; skn-1gf(lax188) drh-1gf(lax257); SPC2005, skn-1(lax188syb2619); SPC2004, skn-1gf(syb2597); SPC2058, ttTi4348-Pvha-6-skn-1a(lax188); SPC2065, ttTi5605-Prgef-1-skn-1c(lax188); SPC2067, ttTi5605-Pgpa-4-skn-1c(lax188); SPC2047, skn-1gf-AID; DV3803, ges-1p::TIR1; DV3805, rgef-1p::TIR1; SPC2048, gpa-4p::TIR1; SPC597, Ex[vha-6p::drh-1(lax257)]; skn-1gf; gst-4p::gfp, SPC2094, drh-1gf(syb7642), SPC 600, Ex[gpa-4p::mCherry];skn-1:;GFP, SPC601, Ex[gpa-4p::mCherry];gst-4-4p::gfp. Single, double, and triple mutants were obtained by standard genetic crosses. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). Genetic mapping of the drh-1(lax257) mutation was performed as previously described (18, 19, 39, 40, 42, 77) by linkage of the gst-4p::gfp suppression of the skn-1gf(lax188) phenotype by using the InDel mapping primer set (57).

Transgenic and Genome Editing.

CRISPR-Cas9 genome editing was used to revert the skn-1(lax188) mutation to WT in skn-1gf mutant animals and separately to create an independent gf allele in wild type animals. To generate tissue-specific expression of drh-1gf, the coding and 3′UTR of drh-1(lax257) was cloned between the vha-6p 5′ regulatory sequence and an SL2::WrmScarlet marker and injected along with a myo-2p::mCherry marker into skn-1gf; gst-4p::gfp animals. Pharyngeal expression of mCherry was used to identify transgenic and nontransgenic siblings for imaging of the gst-4p::gfp reporter.

RNAi Treatment.

RNAi treatment was performed as previously described (78). Briefly, HT115 bacteria containing specific dsRNA-expression plasmids were seeded on NGM plates containing 5 mM isopropyl-β-D-thiogalactoside and 50 μg mL⁻¹ carbenicillin. RNAi was induced at room temperature for 24 h. Synchronized L1 animals were added to those plates to knockdown indicated genes. RNAi efficiency was determined by qPCR with primers designed for the untranslated region of each mRNA that does not overlap with the dsRNA generated for RNAi; Only RNAi treatments that reduce mRNA levels by 50% were used. Imaging was performed on a Leica Stellaris confocal platform (PMT-based) or a Zeiss Axiocam (with a color camera) (SI Appendix, Fig. S2 A–D).

DiI Staining.

DiI staining was performed as previously described (79). In brief, worms were washed with M9 and then put on a rotator overnight in 10 μg/mL DiI in M9 at 20 °C. Worms were then washed with M9 twice and imaged on an agar pad. All centrifugation was done at 106 rcf for 30 s.

Oil Red O Staining.

Oil Red O fat staining was conducted as previously described (57, 80, 81). In brief, worms were egg prepped and allowed to hatch overnight for a synchronous L1 population. The next day, worms were dropped onto plates seeded with bacteria and raised to 120 h (day 3 adult stage) or 168 h (day 5 adult stage). Worms were washed off plates with PBS+triton, then rocked for 3 min in 40% isopropyl alcohol before being pelleted and treated with ORO in diH2O for 2 h. Worms were pelleted after 2 h and washed in PBS+triton for 30 min before being imaged at 20× magnification with LAS X software and Leica Thunder Imager flexacam C3 color camera.

For tissue-specific degradation experiments, worms were egg prepped and allowed to hatch overnight for a synchronous L1 population. The next day, worms were dropped onto plates seeded with bacteria and raised to 48 h (L4 stage) and then moved to experiment plates with vehicle 4 mM ethanol or 4 mM auxin. Worms were moved to new plates every day until 144 h post-drop (Day 4 Adults). Worms were washed off plates with PBS+triton, then rocked for 3 min in 40% isopropyl alcohol before being pelleted and treated with ORO in diH2O for 2 h. Worms were pelleted after 2 h and washed in PBS+triton for 30 min before being imaged at 20× magnification with LAS X software and Leica Thunder Imager flexacam C3 color camera.

Asdf Quantification.

ORO-stained worms were placed on glass slides and a coverslip was placed over the sample. Worms were scored, as previously described (57, 80, 81). Worms were scored and images were taken with LAS X software and Leica Thunder Imager flexacam C3 color camera. Fat levels of worms were placed into two categories: non-Asdf and Asdf. Non-Asdf worms display no loss of fat and are stained dark red throughout most of the body (somatic and germ cells). Asdf worms had most, if not all, observable somatic fat deposits depleted (germ cells only) or significant fat loss from the somatic tissues with portions of the intestine being clear (somatic < germ).

Nile Red Staining.

Nile Red fat staining was conducted as previously described (57, 80, 81). In brief, worms were egg prepped and allowed to hatch overnight for a synchronous L1 population. The next day, worms were dropped onto plates seeded with bacteria and raised to 48 h (L4 stage). Worms were washed off plates with PBS+triton, rocked for 3 min in 40% isopropyl alcohol before being pelleted, and treated with Nile Red in 40% isopropyl alcohol for 2 h. Worms were pelleted after 2 h and washed in PBS+triton for 30 min before being imaged at 10× magnification with ZEN Software and Zen Axio Imager with the DIC and GFP filter. Fluorescence is measured via corrected total cell fluorescence (CTCF) via ImageJ and Microsoft Excel. CTCF = Worm Integrated Density-(Area of selected cell X Mean fluorescence of background readings) and normalized to the control.

Hydrogen Peroxide Treatment.

Conducted as previously described (13). Briefly, synchronous worm populations at either day 3 adulthood or 80HPF were washed 3× with M9+.01% Triton. Then, 500 μL of 20 mM H₂O₂ was then added to 600 μL of worms+M9+.01% Triton. Worms were then incubated on a rotator at 20 °C for 25 min before being rinsed 3× with M9+.01% Triton. Worms were counted before and after 24 h to determine survival.

RNAseq Analysis.

RNAseq analysis was conducted as outlined (80, 82). Worms were egg prepped and eggs were allowed to hatch overnight for a synchronous L1 population. The next day, L1s were dropped onto seeded NGM plates and allowed to grow 48 h, 72 h, 120 h, or 168 h (L4, day 1 adult, day 3 adult, or day 5 adult, respectively) before collection. Animals were washed 3 times with M9 buffer and frozen in TRI reagent at −80 °C until use. Animals were homogenized and RNA extraction was performed via the Zymo Direct-zol RNA Miniprep kit (Cat. #R2052). Qubit™ RNA BR Assay Kit was used to determine RNA concentration. The RNA samples were sequenced and read counts were reported by Novogene. Read counts were then used for differential expression (DE) analysis using the R package DESeq2 created using R version 3.5.2. Statistically significant genes were chosen based on the adjusted P-values that were calculated with the DESeq2 package. Gene Ontology was analyzed using the most recent version of WormCat 2.0 (83). Simple Linear Regression analysis was done on each time-course slope to determine significant difference between lines, P < 0.05.

Chromatin Immunoprecipitation (ChIP).

Chromatin was prepared as in Nhan et al. (13) and Wormbook (Modern techniques for the analysis of chromatin and nuclear organization in C. elegans). Approximately 1,000,000 L4 synchronized animals were washed in M9 and collected into lysis buffer and flash-frozen in liquid nitrogen. Worm pellets were ground via mortar and pestle and resuspended with 1.1% formaldehyde to crosslink proteins. Chromatin was fragmented via sonication and SKN-1::GFP was pulled down by overnight incubation with GFP affinity beads. Associated immunocomplexes were eluted by heat denaturing and DNA fragments were purified. Purified DNA fragments were then used as input for sequencing library preparation using the Diagenode MicroPlex Library prep kit V2. Bioinformatic analyses were done using the following software for ChIP-seq. Quality trimming and Adapter sequences were trimmed from raw paired end reads using Trim Galore package v 0.6.4. Reads were mapped to the C. elegans reference genome using bwa v 0.7.17. BAM files were sorted with Samtools v 1.10. BAM files were sorted to contain only uniquely mapped reads using Sambamba v 0.6.8. Peak calls were made using MACS2 v2.2.7.1 Peak files were feature annotated using Chipseeker Bioconductor package in R using the annotate Peak function.

Statistical Analysis.

All statistical analysis were performed using GraphPad Prism version 10.0. Information on specific statistical tests can be found within each figure legend.

ASI-Enriched RNAseq.

Performed as previously described (84, 85). In brief, approximately 250,000 L4 synchronized WT and skn-1gf animals with GFP tagged ASI neurons (daf-7p::gfp) were washed with M9 6 times to remove residual bacteria. Animals were then pelleted and Cell Isolation Buffer (20 mM HEPES, 0.25% SDS 200 mM DTT 3% Sucrose pH8) was added to worms. Worms were incubated in Cell Isolation Buffer for 2 min. Initial lysis was quenched by washing with M9 6 times. Worm pellet was resuspended in 20 mg/mL Pronase and digested for 20 min with vigorous pipetting every 5 min through a P1000 tip. Pronase digestion was quenched by resuspending in FBS. Cells were pelleted by centrifuging at 550rcf and resuspended in fresh FBS. Cells were passed through a 40-µ cell strainer. DAPI was added to cells to assess viability. Cells were then sorted on a Bio-Rad S3e FACS (Fluorescence Activated Cell Sorting) system. Neurons were homogenized, and RNA extraction was performed via the Zymo Direct-zol RNA Miniprep kit (Cat. #R2052). Qubit™ RNA BR Assay Kit was used to determine RNA concentration. Low input RNA libraries were prepped using the Ovation SoLo RNA library kit from Tecan Genomics. The RNA libraries were sequenced by Novogene. Raw paired-end reads were quality trimmed and adapter trimmed using timmomatic v0.39. Quality-trimmed reads were aligned to the C. elegans reference genome using STAR 2.7.6a. Mapped reads were counted via HTseq v2.0.2 union mode. Read counts were then used for DE analysis using the R package DESeq2 created using R version 3.5.2. Statistically significant genes were chosen based on the adjusted p-values that were calculated with the DESeq2 package. Gene Ontology was analyzed using the most recent version of WormCat 2.0 (83).

Movement Measurements—Crawling.

Worms were egg prepped, and eggs were allowed to hatch overnight for a synchronous L1 population. The next day, worms were dropped onto plates seeded with OP50. Worms were then allowed to grow until each time point (48 h postdrop for L4s, 72 h post-drop for day 1 adults, 120 h postdrop for day 3 adults, and 168 h postdrop for day 5 adults). Once worms were at the required stage of development, 30 to 50 worms were washed off of a plate in 50 μL of M9 with a M9+triton coated P1000 tip and dropped onto an unseeded NGM plate. The M9 was allowed to dissipate, and worms roamed on the unseeded plate for 1 h before imaging crawling. Crawling was imaged with the MBF Bioscience WormLab microscope and analysis was performed with WormLab version 2022. Worm crawling on the plate was imaged for 1 min for each condition at 7.5 ms. Worm crawling was analyzed with the software and only worms that moved for at least 90% of the time were included in the analysis. Statistical analysis of crawling speed was done via one way ANOVA with multiple t test.

Protein Prediction.

Phyre2 (86) was used to predict the structure of DRH-1wt and DRH-1gf; WT structure prediction file 25950237 and Missence3D (image with cyan and magenta in same structure, predicted changes in protein structure 30995449).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Data, Materials, and Software Availability

Sequencing data have been deposited in GEO (GSE246467) (87). All other data are included in the manuscript and/or supporting information.