WDR-23 and SKN-1/Nrf2 Coordinate with the BLI-3 Dual Oxidase in Response to Iodide-Triggered Oxidative Stress

Animals utilize conserved mechanisms to regulate oxidative stress. The C. elegans SKN-1 protein is homologous to the vertebrate Nrf (NF-E2-related factor) family of cap ’n’ collar (CnC) transcription factors and functions as a core regulator of xenobiotic and oxidative stress responses. The WD40 repeat-containing protein WDR-23 is a key negative regulator of SKN-1 activity. We previously found that the oxidative stress induced by excess iodide can be relieved by loss of function in the BLI-3/TSP-15/DOXA-1 dual oxidase complex. To further understand the molecular mechanism of this process, we screened for new mutants that can survive in excess iodide and identified gain-of-function mutations in skn-1 and loss-of-function mutations in wdr-23. The SKN-1C isoform functions in the hypodermis to affect animal’s response to excess iodide, while the SKN-1A isoform appears to play a minor role. wdr-23(lf) can interact with bli-3 mutations in a manner different from skn-1(gf). Transcriptome studies suggest that excess iodide causes developmental arrest largely independent of changes in gene expression, and wdr-23(lf) could affect the expression of a subset of genes by a mechanism different from SKN-1 activation. We propose that WDR-23 and SKN-1 coordinate with the BLI-3/TSP-15/DOXA-1 dual oxidase complex in response to iodide-triggered oxidative stress.

of a SKN-1::GFP fusion protein in intestinal nuclei and enhance the expression of the phase II gene gcs-1 (An and Blackwell 2003).
SKN-1 is a key regulator of the homeostasis of multiple cellular processes. It is required for lipid homeostasis , the expression of extracellular collagens for lifespan extension as a consequence of reduced Insulin/IGF-1 INS-18 signaling  and the stress response to cuticle damage (Dodd et al. 2018). Mitochondrial proline catabolism can activate SKN-1 to affect lifespan (Zarse et al. 2012) or innate immunity (Tang and Pang 2016), and a mitochondria-associated gain-of-function SKN-1 could mediate a conserved starvation response even with ad lib access to food (Paek et al. 2012). SKN-1 is also involved in mitophagy (Palikaras et al. 2015) and can be activated by the IRE protein sulfenylated by ER-or mitochondria-derived ROS (Hourihan et al. 2016). The crosstalk between SKN-1 and mitochondria appears to be conserved across species (Itoh et al. 2015). SKN-1 activity is regulated by multiple signals Inoue et al. 2005;Kell et al. 2007;Tullet et al. 2008;Wang et al. 2010;Li et al. 2011;Robida-Stubbs et al. 2012;Glover-Cutter et al. 2013;Ruf et al. 2013;Ewald et al. 2015).
WDR-23 is a conserved WD40 repeat-containing protein that interacts with the CUL4-DDB1 ubiquitin ligase to promote ubiquitin proteasome system-mediated degradation of SKN-1 in C. elegans (Choe et al. 2009). wdr-23 loss-of-function mutations can lead to constitutive expression of phase II genes, which is similar to the effect of SKN-1 activation (Hasegawa and Miwa 2010). In mammals, a similar WDR23-DDB1-CUL4-dependent mechanism can repress Nrf2 activity independent of the canonical KEAP1-CUL3 pathway, suggesting that WDR-23-dependent regulation of SKN-1 is conserved (Lo et al. 2017).
Several lines of evidence suggest that SKN-1 and the C. elegans NADPH dual oxidase BLI-3 DUOX1 might act together in response to stress. Manganese (Mn)-induced toxicity requires the activity of BLI-3, while SKN-1 can protect against Mn toxicity (Benedetto et al. 2010). Bacterial or fungal pathogens can trigger BLI-3-dependent ROS generation (Chavez et al. 2009;Zou et al. 2013), which can activate SKN-1 target gene expression (Hoeven et al. 2011;Papp et al. 2012;Van Der Hoeven et al. 2012). Loss of the mammalian mediator of ErbB2-driven cell motility, MEMO-1, could lead to enhanced production of ROS by BLI-3, which stimulates SKN-1 to promote stress resistance and longevity (Ewald et al. 2017). Similarly, a redox co-factor, pyrroloquinoline quinone could activate BLI-3 to produce H 2 O 2 at plasma membrane, the effect of which is transduced by SKN-1, JUN-1 and DAF-16 for lifespan extension (Sasakura et al. 2017). These findings suggest that the BLI-3 dual oxidase activity and the SKN-1 activity are probably coordinated to respond to oxidative stress and maintain ROS homeostasis.
Iodine is a micronutrient essential for life and a key ingredient for the synthesis of thyroid hormones. Insufficient intake of iodide can lead to thyroid hormone deficiency and cause severe hypothyroidism and mental retardation (Nussey and Whitehead 2001). However excess iodide intake has been implicated in autoimmune thyroiditis (Bagchi et al. 1985;Rose et al. 1997;Rose et al. 1999;Teng et al. 2006), hyperthyroidism (Nussey and Whitehead 2001), hypothyroidism (Rose et al. 1999;Teng et al. 2006) and thyroid cancers (Lind et al. 1998;Guan et al. 2009;Blomberg et al. 2012;Dong et al. 2013). The molecular mechanism underlying the pathogenic effects of excess iodide is unclear.
We recently used C. elegans as a model to analyze the xenobiotic effect of excess iodide. We found that excess iodide could cause larval arrest, cuticle shedding defects and premature intestinal autofluorescence, phenotypes that can be reversed when animals were moved to normal growth media (Xu et al. 2014). A screen for mutants that can survive in excess iodide isolated loss-of-function (lf) mutations in bli-3 and tsp-15 (Xu et al. 2014), in which tsp-15 encodes a tetraspanin protein required for BLI-3 activity (Moribe et al. 2004;Moribe et al. 2012). We found that the BLI-3/TSP-15/DOXA-1 dual oxidase complex is required for the xenobiotic effects of excess iodide (Xu et al. 2014), which might involve iodide-induced excessive generation of ROS (Xu et al. 2014).
In this study, we report the identification of novel gain-of-function mutations in skn-1 and loss-of-function mutations in wdr-23 and bli-3/ tsp-15/doxa-1 complex and how these genes interact to affect C. elegans cuticle integrity and survival in excess iodide. Besides verifying the known interaction between WDR-23 and SKN-1 in stress responses, we found that wdr-23 can interact with bli-3 and affect gene expression by a mechanism different from SKN-1 activation.

Strains
See Supplementary Materials and Methods.

C. elegans survival in excess iodide
The survival assay was performed as described (Xu et al. 2014). In short, five young adults were grown on E. coli OP50-seeded NGM plates with different concentrations of NaI (5 mM, 10 mM or 50 mM). F 1 progeny were observed for growth and survival until day 8. For transient transgenic experiments, P 0 adult animals injected with transgenes were transferred to OP50-seeded NGM plates with 5 mM NaI and transgene-positive F 1 progeny were examined for growth to adults.

Genetic screens and mapping of mutations See Supplementary Materials and Methods.
Hoechst 33258 staining Hoechst 33258 staining was performed as described (Moribe et al. 2004;Xu et al. 2014) with minor modifications. Synchronized animals (24 hr post mid-L4) were washed off plates and incubated at 20°with gentle shaking for 15 min with 1 mg/ml Hoechst 33258 (Sigma) diluted in M9. After staining, animals were washed three times with M9 and observed under a Leica DM5000B fluorescence microscope.
RNA interference L4 animals were fed HT115 (DE3) bacteria expressing dsRNAs on NGM plates with 1 mM IPTG, 0.1 mg/ml Ampicillin (Timmons et al. 2001) with or without 5 mM NaI for 8 days. The progeny were examined under dissecting microscope for survival. The RNAi feeding bacterial strains for wdr-23 and skn-1 were obtained from a whole-genome RNAi library (Kamath et al. 2003), and the inserts were verified by sequencing. The doxa-1 RNAi feeding bacterial strain was described previously (Xu et al. 2014).

Transgene experiments See Supplementary Materials and Methods.
qRT-PCR Synchronized animals at the L1 larval stage were allowed to recover on OP50-seeded NGM plates with or without 5 mM NaI for 8 hr and subsequently washed three times with H 2 O. RNA was extracted using TRIzol (Invitrogen) and chloroform-isopropanol purification and treated with DNase I (NEB). RNA concentration and quality were measured with a NanoDrop 1000 spectrophotometer (Thermo Fisher). cDNAs were prepared using the Maxima First Strand cDNA Synthesis Kit for qRT-PCR (Thermo Fisher). mRNA levels were quantified from three biological replicates using Maxima SYBR Green (Thermo Fisher) fluorescence on a LightCycler 96 Instrument (Roche). After a pre-incubation step (95°for 10 min), two-step amplification was performed using 40 cycles of denaturation (95°for 15 s) and annealing (60°for 45 s). Target gene expression levels were normalized to that of the reference gene tba-1. Primers for constructs and qRT-PCR experiments are listed in Table S7.

Transcriptome analyses See Supplementary Materials and Methods.
Statistics P values were determined by two-tailed unpaired Student's t-test for comparisons between two samples and Bonferroni test with one-way ANOVA for comparisons of more than two samples. Ã : P , 0.05; ÃÃ : P , 0.01; ÃÃÃ : P , 0.001.

RESULTS
Loss-of-function mutations in wdr-23 and gain-offunction mutations in skn-1 can promote C. elegans survival in excess iodide In a previous screen for mutants that can survive in excess iodide (5 mM NaI) (Xu et al. 2014) (Table S1, Screen 1), we isolated four lf mutations in bli-3 (mac37, mac38, mac40, mac41) and one lf mutation in tsp-15 (mac33) ( Table S1, Table S2 and Figure 1). The lf nature of the tsp-15 (mac33) mutation was further confirmed by transgene rescue experiments in this study (Table S3), showing that wild-type tsp-15 transgenes nearly abolished the survival of tsp-15(mac33lf) mutants in excess iodide.
We mapped the last mutation (mac39) in Screen 1 to Chr. IV within a region containing skn-1. Knowing that WDR-23 is a negative regulator of SKN-1 (Choe et al. 2009), we took skn-1 as a candidate and indeed identified a missense mutation that causes an R43C amino acid change (Table S1, Table S2 and Figure 1) on the SKN-1C isoform in mac39 mutants.
Both mac39 heterozygous and homozygous animals can survive in excess iodide, while the skn-1(zu135lf) (Bowerman et al. 1992) homozygous mutants and animals fed RNAi targeting skn-1 failed to (Table  1). In addition, both heterozygous and homozygous animals of the previously identified skn-1(lax120gf) mutation (Paek et al. 2012) can survive in excess iodide (Table 1). Based on these findings, we propose that mac39 causes a gain of function (gf) in skn-1.
Additional screens isolated novel lf mutations in bli-3, doxa-1 and wdr-23 and gf mutations in skn-1 To identify more genes and mutations involved in animal's response to excess iodide, we performed additional screens (Table S1) for mutants that can survive in 5 mM NaI potentially as homozygotes (Screen 2 for F 2 mutants) or heterozygotes (Screen 3 for F 1 mutants).
All skn-1 mutants can survive in excess iodide as heterozygotes or homozygotes. Using skn-1c as the reference isoform, the amino acid change (G39D) in skn-1(mac53) (Table S2 and Figure 1) is one amino acid away from the R41C change caused by skn-1(k1023gf) (Tang and Choe 2015) and three amino acids away from the R43C change caused by skn-1(mac39gf), suggesting functional importance of a potential domain in SKN-1C that contains these amino acid residues. The skn-1(mac415) mutation causes the same R41C change in SKN-1C as that by skn-1(k1023gf) (Tang and Choe 2015). The mac411, mac413, mac416 mutations cause an E147K amino acid change identical to the previously described skn-1(lax188gf) mutation (Paek et al. 2012). Therefore, we isolated new skn-1 gf mutations as well as mutations that were previously described.
SKN-1C functions in the hypodermis (epidermis) to promote animal survival in excess iodide skn-1 is predicted to express four isoforms, skn-1a, skn-1b, skn-1c and skn-1d , among which skn-1c has been extensively studied. In larvae and adults, skn-1c is apparently expressed in ASI neurons and weakly in intestine (An and Blackwell 2003). skn-1c functions in intestine to regulate a variety of biological processes . SKN-1C expression has also been observed in hypodermis using skn-1c transgenes (Wu et al. 2016) and in hypodermis, pharynx and body-wall muscles based on the expression of SKN-1C target genes (Hasegawa et al. 2008;Paek et al. 2012).
We next examined whether the skn-1c(mac53gf) cDNA transgene could promote the survival in excess iodide. We established stable skn-1c(mac53gf) transgenic lines using the two intestine-specific promoters (nhx-2p and ges-1p) and the body-wall muscle-specific myo-3 promoter. However, these transgenic animals could not survive in excess iodide (Table S4), suggesting that skn-1(gf) might not function in intestine or muscle to promote the survival.
Surprisingly, we failed to establish stable skn-1c(mac53gf) lines using the skn-1c promoter or the hypodermis-specific dpy-7 promoter (Table S4), probably due to the toxicity of hypodermis-specific skn-1c(gf) overexpression. Such toxicity is only obvious in the F 2 generation as we could generate abundant viable skn-1(gf) transgene-positive F 1 animals ( Table 2).
To overcome the toxicity of stable skn-1c(gf) transgenes under control of the skn-1c or the dpy-7 promoter, we examined the survival of skn-1c(mac53gf) transgene-positive (based on co-injection marker expression) F 1 progeny in excess iodide. Here, we found that numerous transgene-positive F 1 animals could grow into adults in excess iodide (Table 2) and the dpy-7 promoter appears to be more robust than the skn-1c promoter. Therefore, skn-1c(gf) can function in the hypodermis to promote the survival.
Since we isolated lf mutations in the bli-3/tsp-15/doxa-1 complex in the same screens, we tested whether these genes function in the hypodermis as well. Indeed, stable tsp-15 cDNA transgenes under control of the dpy-7 promoter could strongly rescue the survival-promoting effect of the tsp-15(mac33lf) mutation (Table S3), suggesting that the bli-3/ tsp-15/doxa-1 complex also functions in the hypodermis to affect the survival in excess iodide.
skn-1c(mac53gf) transgenes can activate the expression of SKN-1C target gene gst-4 To test whether the failure of intestine-specific skn-1c(gf) transgene expression in promoting the survival might be caused by a lack of activated SKN-1C target gene expression, we introduced skn-1c(mac53gf) transgenes to the dvIs19 transgenic animals (Link and Johnson 2002). The dvIs19 transgene expresses GFP under control of the gst-4 promoter and is used as a reliable reporter for SKN-1C activation.
We found that skn-1c(mac53gf) transgenes, under control of either skn-1c promoter or the intestine-specific nhx-2 promoter, can significantly increase GFP expression in the intestines of transgene-positive F 1 progeny ( Figure 2B and 2C, right panels), while skn-1c(wt) transgenes have no obvious effect ( Figure 2B and 2C, left panels). The nhx-2 promoter appears to cause a more robust GFP expression than the skn-1c promoter does.
Under control of the dpy-7 promoter, the skn-1c(mac53gf) transgene resulted in two distinct groups of F 1 transgenic progeny. In most cases, the transgenic animals have normal size with normal intestinal GFP expression and can grow into adults in excess iodide ( Figure 2D, right panel, animals on the left under Tg). Occasionally, we found transgenic animals with a strong Dpy phenotype and an apparent increase in intestinal GFP expression that failed to grow in excess iodide ( Figure 2D, right panel, animals on the right under Tg).
n Table 1 The survival of different mutants and wild-type animals treated with RNAi in excess iodide

Genotype
Survival in 5 mM NaI The mechanism underlying these distinct phenotypes might be related to the difference in levels, temporal stages or leakiness of the transgene expression. These results together suggest that intestinal activation of SKN-1C is not sufficient for animal survival in excess iodide.
To examine whether skn-1a plays a role in animal's response to excess iodide, we generated transgenic lines (Table S4) with skn-1a(wt) or skn-1a(mac53gf) cDNA under control of the skn-1a (Staab et al. 2014) or skn-1c promoter (An and Blackwell 2003) (Fig. S1A). All transgenic lines failed to survive in excess iodide (Table S4). However, we consistently found escapers in the skn-1a(mac53gf) lines controlled by the skn-1a promoter (Table S4). To verify this finding, we examined the survival of F 1 skn-1a transgenic animals in excess iodide (Table S5). The results suggest that both skn-1a(wt) and skn-1a(gf) transgenes under control of either skn-1c or skn-1a promoter could weakly promote the survival.
To examine whether skn-1a transgenes might affect skn-1c target gene expression, we introduced these transgenes to the dvIs19 animals. It appears that the skn-1a(wt) or skn-1a(gf) transgenes under control of the skn-1a promoter could weakly activate the GFP expression (Fig. S1B, two left panels), while these transgenes under control of the skn-1c promoter failed to do so (Fig. S1B, two right panels). These results suggest that both skn-1a(wt) and skn-1a(gf) are capable of weakly activating SKN-1C target gene expression. Furthermore, skn-1a(wt) might carry an activity similar to skn-1a(gf) in promoting animal survival in excess iodide. The underlying mechanism remains to be understood.
skn-1 is required for the survival of bli-3, tsp-15, doxa-1 and wdr-23 lf mutants in excess iodide Since skn-1(zu135lf) and skn-1(RNAi) animals could not survive to adults in excess iodide (Table 1), we examined whether skn-1 is epistatic to bli-3, tsp-15, doxa-1 or wdr-23. We generated double mutants carrying the skn-1(zu135lf) mutation and one of two or more independently isolated mutations in these other genes. Except for bli-3(e767lf); skn-1(zu135lf) double mutants, which were too sick for survival test, all other double mutants grew similarly as skn-1(zu135lf) single mutants Occasionally the transgene caused a strong Dpy phenotype with increased reporter expression in the intestine (animals on the right).
n under normal condition but failed to survive to adults in excess iodide (Table 3). Therefore, skn-1 is required for mutations in the other four genes to promote animal survival in excess iodide.

wdr-23(lf) and skn-1(gf) interact with bli-3(lf) differentially to affect animal survival in high concentration of NaI
To further understand the interactions of bli-3 with skn-1 and wdr-23, we chose two independent alleles of each of the three genes and generated double mutants. We examined the survival of single or double mutants in 10 mM or 50 mM NaI. Higher concentrations of iodide might cause more severe oxidative stress, which can be used for detecting additive or synergistic genetic interactions (Table 4). We found that all single and double mutants can survive in 10 mM NaI ( Table 4), suggesting that at this concentration iodide does not generate a lethal oxidative stress. We next tested 50 mM NaI, in which all single mutants failed to survive (Table 4). Interestingly, bli-3(lf); skn-1(gf) double mutants exhibited split phenotypes: the two bli-3(e767lf); skn-1(gf) double mutants failed to survive in 50 mM NaI, while the two bli-3(mac40lf); skn-1(gf) double mutants could survive. Different from bli-3(lf); skn-1(gf), all four bli-3 (lf) wdr-23(lf) double mutants we initially generated failed to survive (Table 4, First group). Therefore, skn-1(gf) and wdr-23(lf) can interact with bli-3(mac40lf) differentially.
In all double mutants, only bli-3(mac40lf) wdr-23(mac42lf) could not survive in 10 mM NaI. We speculate that an unknown defect(s) in the double mutants that is not derived from oxidative stress contributes to the inviability, since single mutants of either mutation could survive in 10 mM NaI.
skn-1(gf) and wdr-23(lf) interact with bli-3 differentially to affect the cuticle integrity A critical function of the BLI-3 dual oxidase is to catalyze the crosslinking of tyrosyl residues of the cuticle collagens. The process involves the BLI-3 NADPH oxidase domain, the BLI-3 peroxidase domain and the peroxidase MLT-7 (Edens et al. 2001;Meitzler and Ortiz De Montellano 2009;Thein et al. 2009;Meitzler et al. 2010;Moribe and Mekada 2013). To examine whether skn-1 or wdr-23 interacts with bli-3 to affect cuticle formation, we tested the cuticle integrity of mutants by staining with the fluorescent nuclear dye Hoechst 33258 (Thein et al. 2009;Xu et al. 2014).
An examination of bli-3 single mutants suggests that the peroxidase domain and the oxidase domain affect cuticle integrity differentially: the peroxidase domain mutations (mac52 and e767) resulted in cuticles more defective than the oxidase domain mutations did (mac68, mac66, mac40) ( Figure 3A). None of the skn-1(gf) or wdr-23 (lf) single mutants has apparently defective cuticles ( Figure 3A), suggesting that these two genes are not directly involved in cuticle formation.
An examination of double mutants suggests that skn-1(gf) mutations do not apparently alter the cuticle defects caused by mutations affecting either the peroxidase domain or oxidase domain of BLI-3 ( Figure 3B). Surprisingly, wdr-23(lf) mutations affect the cuticles in a bli-3 allele-specific manner: they uniformly and strongly suppress the cuticle defects of the bli-3(mac40lf) (oxidase domain) mutants but not that of bli-3(e767lf) (peroxidase domain) mutants ( Figure 3B).

skn-1(gf) and wdr-23(lf) similarly affect the expression of most, but not all target genes
To understand what downstream genes of skn-1 and wdr-23 might be involved in promoting animal survival in excess iodide, we performed RNA-Seq on synchronized wild-type, skn-1(mac53gf) and wdr-23 (mac35lf) larvae grown with or without excess iodide and analyzed their transcriptomes.
We found that animals of the same genotype exhibit largely similar gene expression profiles with or without excess iodide, as shown in the gene expression heat map ( Figure 4A) and the numbers of differentially expressed genes (DEGs) ( Figure 4B). The gene expression profiles of skn-1(mac53gf) and wdr-23(mac35lf) mutants are apparently different from that of wild type, while a subtle but visible difference is also seen between them ( Figure 4A). Fewer genes are altered in skn-1(mac53gf) mutants compared to wdr-23(mac35lf) mutants ( Figure 4B). Direct comparison of skn-1(mac53gf) and wdr-23(mac35lf) identified 16 (without iodide) and 22 (with iodide) DEGs ( Figure 4B).
Gene ontology (GO) analyses revealed that many of the DEGs in skn-1 and/or wdr-23 mutants belong to signaling pathways that regulate metabolism and defense response, with glutathione metabolic process being the top pathway in each mutant ( Fig. S2 and S3). Cellular process, metabolic process and single-organism process are the top three GO subterms that contain the highest percentage of DEGs in each mutant (Fig. S4).
To determine whether the differential effects on gene expression by skn-1(mac53gf) and wdr-23(mac35lf) are allele-specific, we examined the expression of a subset of these genes in skn-1(lax120gf) and wdr-23 (mac32lf) mutants ( Figure 5C). The expression of these genes exhibits similar patterns in the two independent mutants of skn-1 or wdr-23 ( Figure 5C), suggesting that skn-1(gf) and wdr-23(lf) can affect the expression of a subset of genes, e.g., F56A4.3 and F28B4.3, by a mechanism different from the one used by their shared gst and ugt target genes.

DISCUSSION
In this study, we isolated multiple mutations affecting C. elegans skn-1, wdr-23 and the bli-3/tsp-15/doxa-1 complex by screening for surviving mutants in excess iodide. We suggest that WDR-23 and SKN-1 interact with the BLI-3/TSP-15/DOXA-1 complex to regulate animal's response to oxidative stress. We also suggest that WDR-23 loss of function can affect BLI-3 activity and some gene expression independent of SKN-1 activation.
n Table 4 The survival of single and double mutants in 10 or 50 mM NaI

Genotype
Alleles Survival in 10 mM NaI Survival in 50 mM NaI  SKN-1/WDR-23 and BLI-3/TSP-15/DOXA-1 affect C. elegans survival in excess iodide by a conserved mechanism SKN-1 and BLI-3 are involved in C. elegans response to pathogens (Hoeven et al. 2011;Tang and Pang 2016), oxidative stress (Ewald et al. 2017), manganese toxicity (Benedetto et al. 2010) and ROS-related lifespan extension (Sasakura et al. 2017). The BLI-3/TSP-15/DOXA-1 complex is also required for the formation of C. elegans cuticles by generating H 2 O 2 , which is utilized by the BLI-3 peroxidase domain (Edens et al. 2001) and the peroxidase MLT-7 for crosslinking cuticle proteins (Thein et al. 2009). The concurrent involvement of SKN-1 and BLI-3 in multiple cellular processes suggests functional crosstalk between these two molecules and/or the pathways. It is unclear whether and how the oxidase and peroxidase activities of BLI-3, the activation of SKN-1, and the ROS production are coordinated in vivo.
Our screening for recessive and dominant mutations that can promote animal survival in excess iodide identified lf mutations in the bli-3/ tsp-15/doxa-1 complex and wdr-23 and gf mutations in skn-1. It is plausible that the reduced ROS generation in bli-3/tsp-15/doxa-1 lf mutants and the activation of antioxidant gene expression in wdr-23 (lf) or skn-1(gf) mutants would attenuate the oxidative stress caused by excess iodide, a strong inducer of ROS in C. elegans and mammals (Many et al. 1992;Golstein and Dumont 1996;Corvilain et al. 2000;Vitale et al. 2000;Yao et al. 2012;Serrano-Nascimento et al. 2014;Xu et al. 2014). Consistent with this hypothesis, we found that the antioxidants ascorbic acid (vitamin C) and N-acetylcysteine (NAC) can antagonize the toxic effect of excess iodide (Table S6).
Recent studies found that excess iodide could increase Nrf2 expression in rat thyroid  and activate the Nrf2 pathway in human skin cells (Ben-Yehuda Greenwald et al. 2017). Therefore, it is a conserved mechanism that the BLI-3/TSP-15/DOXA-1 dual oxidase complex and the Nrf2/SKN-1 pathway are both involved in the response to oxidative stress induced by excess iodide.
How C. elegans takes in iodide is unknown. In mammals, iodide uptake is mediated by Na(+)/I(-) symporter (NIS), an integral plasma membrane glycoprotein expressed in multiple tissues including thyroid, the lacrimal sac and nasolacrimal duct, salivary glands, choroid plexus, stomach, intestine, lactating breast, kidney, placenta and ovary (Ravera et al. 2017). We previously found that RNAi targeting two C. elegans genes similar to NIS did not apparently affect animal survival in excess iodide (Xu et al. 2014). It is possible that iodide is absorbed by the intestine and then transported to the hypodermis in C. elegans. Alternatively, iodide might gain access to the hypodermis directly via microscopic openings on the cuticle. The detailed mechanism remains to be understood.

SKN-1C is the primary SKN-1 isoform responsible for promoting animal survival in excess iodide
The SKN-1C isoform normally resides in the cytoplasm and enters the nucleus in response to stress signals (An and Blackwell 2003;Blackwell et al. 2015). The SKN-1A isoform might be associated with mitochondria (Paek et al. 2012) to mediate starvation response. It is also associated with ER (Lehrbach and Ruvkun 2016) to respond to proteasome dysfunction signals. The skn-1 gf mutations we identified affect both A and C isoforms (Fig. S1). Our transgene experiments suggest that SKN-1C is the major isoform that promotes animal survival, while SKN-1A might play a minor role. Hypodermic overexpression of SKN-1C(gf) might be highly toxic, which explains why we failed to obtain any stable transgenic lines using the skn-1c promoter or the hypodermis-specific dpy-7 promoter and suggests that a highly regulated SKN-1C activity in hypodermis is essential for development and survival.
In WormBase (www.wormbase.org), five WDR-23 isoforms are annotated. The mutations we isolated affect all wdr-23 isoforms ( Figure  1D and Fig. S6). Previous studies found that the WDR-23A isoform is associated with outer mitochondrial membranes, while the WDR-23B isoform is localized exclusively in the nucleus (Staab et al. 2013;Staab et al. 2014). We found that wdr-23a and wdr-23b transgenes can strongly rescue the phenotype of wdr-23(lf) mutants with a similar efficiency (Table S3), which is consistent with the previous finding that a functional difference of the two WDR-23 isoforms was not detected in transgenic experiments (Staab et al. 2013).
The peroxidase domain and oxidase domain of BLI-3 are functionally divergent BLI-3 is the only functional dual oxidase in C. elegans that contains an N-terminal peroxidase domain and a C-terminal oxidase domain (Edens et al. 2001;Donko et al. 2005;Bedard et al. 2007). It is also the only NADPH oxidase in C. elegans (Bedard et al. 2007). Studies of missense mutations affecting the peroxidase domain or the oxidase domain of BLI-3 found that peroxidase mutations do not or only weakly affect infection-induced H 2 O 2 production, while an oxidase mutation (Moribe et al. 2012) has an apparently stronger effect (Chavez et al. 2009;Van Der Hoeven et al. 2015). The oxidase domain mutation could also reduce the lifespan and make C. elegans more susceptible to pathogens, while the peroxidase mutations do not or only weakly do so (Chavez et al. 2009;Van Der Hoeven et al. 2015).
In our study, BLI-3 peroxidase domain mutations impair the cuticle integrity more severely than oxidase mutations do ( Figure 3A), suggesting a functional difference of these two domains that is consistent with previous findings (Chavez et al. 2009;Van Der Hoeven et al. 2015). We previously found that ROS overproduction caused by excess iodide in C. elegans can be partially suppressed by both the bli-3(e767lf) (peroxidase) and bli-3(mac40lf) (oxidase) mutations (Xu et al. 2014), suggesting that BLI-3 peroxidase domain mutations also impact the oxidase domain. Therefore, the peroxidase domain [that consumes ROS to Figure 5 qRT-PCR verification of altered gene expression. (A) Top four pathways and seven candidate genes that are altered in both wdr-23(lf) and skn-1(gf) mutants based on KEGG analysis. Among the candidate genes, gst-30,F56A4.3 and ugt-38 are also differentially expressed between skn-1(mac53gf) and wdr-23(mac35lf) based on RNA-Seq. (B) qRT-PCR examination of the eight skn-1/wdr-23 target genes in different genotypes treated with or without excess iodide. (C) qRT-PCR examination of the expression of five skn-1/wdr-23 target genes in independent mutants of skn-1 or wdr-23. Results are from three biological replicates. Reference gene: tba-1. Comparisons were made with wild type or between genotypes. Error bars: Mean 6 SE. Statistics: two-tailed unpaired Student's t-test. Ã : P , 0.05; ÃÃ : P , 0.01; ÃÃÃ : P , 0.001. crosslink tyrosyl residues of collagens] and the oxidase domain [that generates ROS] likely interact and also function differentially to affect cuticle formation, ROS generation and the response to oxidative stress or pathogens.
Our transcriptome analyses suggest that excess iodide does not apparently alter gene expression in wild-type or mutant C. elegans (Figure 4). The survival-promoting effect of skn-1(gf) and wdr-23(lf) mutations likely involve the activation of antioxidant gene expression (Fig. S2, S3, S4 and Result S1), similar to what previous studies have found . However, skn-1(gf) and wdr-23(lf) can differentially affect the expression of a subset of genes. For example, wdr-23(lf) causes higher expression of four genes that are up-regulated in both wdr-23(lf) and skn-1(gf) mutants ( Figure 5B and 5C , sodh-2, ugt-11, gst-30, ugt-38). The expression of F56A4.3 is abolished in wdr-23(lf) mutants but significantly increased in skn-1(gf) mutants ( Figure 5B and 5C), while the expression of F28B4.3 is significantly increased in wdr-23(lf) mutants but unaltered in skn-1(gf) mutants ( Figure 5B and 5C). These differences are not allele-specific for wdr-23 or skn-1, suggesting a potentially novel mechanism underlying these effects of WDR-23 loss of function and SKN-1 gain of function. The biological significance of this differentiation remains unclear.
We generated a working model to describe the known and potential interactions among SKN-1, WDR-23 and BLI-3 ( Figure 6). In this model, the H 2 O 2 generated by BLI-3 for crosslinking cuticle collagens also contributes to the oxidative stress caused by xenobiotic stressors such as iodide. Besides its canonical role as a SKN-1 negative regulator, WDR-23 might also suppress BLI-3 activity, a possibility supported by the genetic interaction between wdr-23(lf) and bli-3(lf) (Figure 3 and Table 4).
In summary, we found that the oxidative stress triggered by excess iodide can be suppressed by defects in the BLI-3/TSP-15/DOXA-1 complex or the activation of SKN-1 either by skn-1(gf) or wdr-23(lf) mutations. We provide further genetic and molecular evidence supporting the consensus that WDR-23 can function as a negative regulator of SKN-1 in activating antioxidant gene expression and also suggest that WDR-23 could interact with BLI-3 and affect some gene expression by a mechanism(s) different from the prior one. Our findings should facilitate the understanding of animal's response to oxidative stress. Future studies are warranted for elucidating the underlying molecular mechanism.
Figure 6 A working model describing the known and potential interactions among SKN-1, WDR-23 and BLI-3. In this model, BLI-3 contributes to oxidative stress by generating H 2 O 2 used for crosslinking cuticle collagens. Besides inhibiting the activation of SKN-1, WDR-23 might also suppress BLI-3 by an unknown mechanism.