Casein kinase 1 gamma regulates oxidative stress response via interacting with the NADPH dual oxidase complex

Oxidative stress response is a fundamental biological process mediated by conserved mechanisms. The identities and functions of some key regulators remain unknown. Here, we report a novel role of C. elegans casein kinase 1 gamma CSNK-1 (also known as CK1γ or CSNK1G) in regulating oxidative stress response and ROS levels. csnk-1 interacted with the bli-3/tsp-15/doxa-1 NADPH dual oxidase genes via genetic nonallelic noncomplementation to affect C. elegans survival in oxidative stress. The genetic interaction was supported by specific biochemical interactions between DOXA-1 and CSNK-1 and potentially between their human orthologs DUOXA2 and CSNK1G2. Consistently, CSNK-1 was required for normal ROS levels in C. elegans. CSNK1G2 and DUOXA2 each can promote ROS levels in human cells, effects that were suppressed by a small molecule casein kinase 1 inhibitor. We also detected genetic interactions between csnk-1 and skn-1 Nrf2 in oxidative stress response. Together, we propose that CSNK-1 CSNK1G defines a novel conserved regulatory mechanism for ROS homeostasis.


Introduction
Reactive oxygen species (ROS) are universal life molecules. ROS at physiological levels are essential signaling molecules that regulate metabolism, cell growth, cell survival and cell proliferation [1]. Endogenous ROS are primarily generated by the NADPH oxidases and mitochondria [1,2]. Abnormal cellular processes, xenobiotics, some metals and pathogens can induce excessive ROS generation [3][4][5], which may damage proteins, lipids and nucleic acids and cause oxidative stress [2]. Animals use conserved molecular mechanisms to regulate ROS levels and mitigate oxidative stress [1,2]. For example, the nuclear factor erythroid 2-related factor 2 (Nrf2) pathway plays a major role in the defense response to oxidative stress triggered by endogenous or exogenous stressors [1,2,6].
Iodine is an essential mineral nutrient for the synthesis of thyroid hormones and is routinely added to salt and foods. Surprisingly, some studies found that excess intake of iodide salt was associated with several thyroid diseases with unclear molecular mechanisms [50]. We became interested in this phenomenon and used C. elegans as a model to examine the biological effects of excess iodide intake. We found that C. elegans treated with excess iodide had a significant increase of ROS levels and wildtype animals exhibited larval arrest potentially due to oxidative stress [19,20]. This is likely a conserved effect of iodide because several earlier studies showed that excess iodide can induce ROS generation in mammalian thyroid cells [51][52][53][54]. To investigate whether any genes might be involved in the larva-arresting effect of excess iodide, we screened for mutants that can survive into adults in excess iodide and isolated multiple loss-of-function (lf) mutations in bli-3, tsp-15, doxa-1, wdr-23 and gain-of-function (gf) mutations in skn-1 [19,20]. These findings suggest that new genes involved in oxidative stress response might be identified by examining whether their mutations can promote C. elegans survival in excess iodide [19,20].
In this study, we found that mac397, a previously isolated mutation that promoted C. elegans survival in excess iodide [20], caused a missense mutation in csnk-1. Further analyses found that csnk-1 genetically interacted with the bli-3/tsp-15/doxa-1 genes by nonallelic noncomplementation to affect animal survival in excess iodide. Consistent with this special genetic interaction, we detected specific biochemical interactions between CSNK-1 and DOXA-1 and potentially between their human orthologs. We also found that CSNK-1 was required for maintaining ROS levels in C. elegans and human CSNK1G2 can promote ROS levels in cultured cells. We propose that CSNK-1 CSNK1G is a conserved novel regulator of ROS levels and oxidative stress response by interacting with the NADPH dual oxidase complex.

csnk-1 loss of function promotes C. elegans survival in excess iodide
We previously performed an ethyl methanesulfonate (EMS) screen for F 1 mutants that can grow into adults in 5 mM NaI (the survived-in-sodium-iodide phenotype, abbreviated as Sisi hereafter) and isolated six unidentified mutants [20]. We mapped one of the mutations, mac397, to Chr. I. Within the mapped region of mac397, genomic sequencing detected a point mutation in csnk-1 (exon4:c.G482A:p.R161H, named mac397. Fig 1A and S1 Table), a point mutation in tsp-15 (exon3:c.G331A:p.G111R, named mac499. S1 Table), and nine other coding variants (S1 Table). Deficiencies in csnk-1 or tsp-15 caused by RNAi or genetic mutations led to the Sisi phenotype (S1, S2 and S3 Tables). Transgene rescue and genocopy experiments suggest that tsp-15(mac499) caused a loss of function (S3 Table). The analyses of other isolates are ongoing.
To investigate how csnk-1 transcript levels change in these mutants, we performed RT-qPCR experiments using primer pairs that are supposed to recognize all transcripts (wildtype and mutants) or wildtype-only transcripts (S2A Fig). We found that both mutants exhibited significantly reduced levels of total csnk-1 transcripts at the L4 larval stage (S2B Fig), implying a positive feedback effect of CSNK-1 on the expression of its own transcripts or that the mutant transcripts were degraded via nonsense-mediated decay (NMD). Furthermore, wildtype transcripts were essentially depleted in both mutants at the L4 larval stage (S2C and S2D  The egg-hatching, egg-laying and molting defects of csnk-1(mac495) mutants were obviously more severe than those of csnk-1(mac494) mutants. Since mac494 and mac495 both caused frameshifts close to the start codon and are predicted to encode severely truncated proteins (Figs 1 and S1), it is intriguing that the severity of their phenotypes was quite different.
We have yet to determine whether this difference is caused by a potential hypomorphic nature of mac494 or unknown modifier mutations in them.
To further understand csnk-1 function, we expressed a csnk-1 cDNA transgene driven by an endogenous csnk-1 promoter (3.1 kb upstream of the start codon) in csnk-1(mac494) mutants. The transgene significantly rescued the Sisi phenotype of the mutants (Table 1). Because the majority of csnk-1(mac494) mutants were able to survive into adults probably due to apparently weaker molting defects (S3E Fig), which makes our analyses of the Sisi phenotype significantly easier, we used this mutation as our reference loss-of-function (lf) allele in the following studies, if not specified.
We next examined how mac397 R161H affects CSNK-1 activity. Driven by the endogenous csnk-1 promoter, a csnk-1(R161H) cDNA transgene failed to rescue or only weakly rescued the Sisi phenotype of csnk-1(lf) mutants (Table 1). This transgene also failed to cause Sisi in wildtype animals, suggesting that mac397 was not dominant negative (Table 1). We propose that mac397 caused a loss of function.

CSNK-1 primarily functions in the epidermis to affect the Sisi phenotype and its function is conserved between C. elegans and human
We next examined the expression pattern of csnk-1. A GFP transgene driven by the csnk-1 endogenous promoter was broadly expressed in the wildtype background ( Fig 1C). The expression was obvious in the pharynx, head neurons, epidermis (hypodermis), intestine, ventral cord, dorsal cord, vulval muscles and body-wall muscles (Fig 1D-1H).
To examine whether the function of CSNK1G was conserved, we expressed human CSNK1G1, CSNK1G2 or CSNK1G3 cDNA transgenes in the epidermis of csnk-1(lf) mutants. We found that these transgenes significantly rescued the Sisi phenotype of the mutants and CSNK1G2 transgenes appeared to be more effective (Table 1). Hence, the function of csnk-1 is likely conserved and human CSNK1G1/2/3 exhibited similar activities. CSNK1Gs contain a conserved C-terminal palmitoylation signal (TKCCCFFKR) required for membrane localization and the activities of the kinases [34,55]. A conserved palmitoylation sequence was also found in CSNK-1 (VKCCCFRRR, aa 386-394, S1 Fig). We generated two mutant csnk-1 transgenes disrupting this sequence (p.C388-K407del or p.C388-390S, S1 Fig). These transgenes failed to or barely rescued the Sisi phenotype of csnk-1(lf) mutants (Table 1).

Conserved interaction between CSNK-1 and DOXA-1
The genetic findings imply that CSNK-1 and BLI-3/TSP-15/DOXA-1 might form a protein complex or function in the same pathway. Furthermore, the more special interaction between csnk-1 and doxa-1 prompted us to speculate whether DOXA-1 might physically interact with CSNK-1. To investigate the possibility, we co-expressed a DOXA-1::GFP fusion protein and a CSNK-1::mCherry fusion protein in C. elegans epidermis. Here we detected colocalization of the fluorescent signals in epidermal subcellular structures (Fig 2A-2C). The colocalization is likely caused by specific interaction between DOXA-1 and CSNK-1 because GFP alone did not colocalize with CSNK-1::mCherry, nor did mCherry alone colocalize with DOXA-1::GFP (S7 Fig). We next transiently expressed a FLAG::CSNK-1 fusion protein with HA tag alone or with a DOXA-1::HA fusion protein in HEK293T cells. From the cell extracts, FLAG::CSNK-1 could be coimmunoprecipitated with DOXA-1::HA using an anti-HA antibody, while FLAG:: CSNK-1 co-expressed with HA tag alone could not be coimmunoprecipitated using this antibody ( Fig 2D). We further performed reverse immunoprecipitation using an anti-FLAG antibody. Similarly, DOXA-1::HA could be coimmunoprecipitated with FLAG::CSNK-1 but not with the FLAG tag alone (Fig 2E).
To examine the interaction between CSNK-1 and DOXA-1 by a different approach, we purified a bacterially expressed His::TF::DOXA-1 fusion protein and co-incubated the purified protein with the lysates of HEK293T cells expressing FLAG::CSNK-1. Here FLAG::CSNK-1 could be specifically pulled down by His::TF::DOXA-1 ( Fig 2F). In reverse, a His::TF::CSNK-1 fusion protein purified from bacterial expression could specifically pull down a DOXA-1::HA fusion protein expressed in HEK293T cells ( Fig 2G). Together, these results suggest a specific interaction between CSNK-1 and DOXA-1. Nevertheless, caution should be exercised in interpreting the pulldown results as proteins expressed in E. coli may be incorrectly folded, which might cause non-specific interactions.
DOXA-1 is orthologous to the human DUOX maturation factor 1 and 2 (DUOXA1 and DUOXA2) (S8 Fig), which are required for dual oxidase activities [17,59]. To investigate whether human CSNK1Gs and DUOXAs exhibit similar interactions like that between CSNK- Table 2 Table). csnk-1 (lf)/hT2 was treated as csnk-1(lf)/+. We failed to place tsp-15(lf) and csnk-1(lf) on the same chromosome due to the close linkage of the two genes. For this reason, we only examined male progeny from crosses between tsp- 15 1 and DOXA-1, we co-expressed FLAG::CSNK1G2 and DUOXA2::HA fusion proteins in HEK293T cells. In these cells we detected co-localization of the two proteins on the plasma membrane and subcellular structures (Fig 3A). Similar colocalization was also observed in a transfected HeLa cell (S3I Fig). [CSNK1G2 was chosen because its transgenes showed stronger rescuing effects (Table 1).] We also detected a potentially specific interaction between these proteins using immunoprecipitation (Fig 3B).
To investigate whether human CSNK1 might affect ROS levels, we treated HEK293T cells with D4476, a selective small molecule inhibitor of CSNK1 [61]. Using DCFDA and Amplex Red staining, we detected a dose-dependent inhibition of ROS levels by D4476 (Fig 3C and  3D). Furthermore, overexpression of CSNK1G2 led to increased ROS levels in the cells as measured by DCFDA fluorescence, which were inhibited by D4476 (Fig 3E). The increased ROS levels caused by DUOXA2 overexpression were also inhibited by D4476 (Fig 3E).
The interaction between CSNK-1 and DOXA-1 prompted us to explore whether DOXA-1 activity might be related to any potential phosphorylation sites. We inspected DOXA-1 sequence using Scansite 4 (www.scansite4.mit.edu) and the top three predicted sites were shown in S8 Fig (red boxes and red arrowheads). These sites have the characters of the consensus CSNK-1 phosphorylation sites, D/E/(p)S/T(X) 1-3 S/T [55], in which the phospho-acceptor residue is underlined. We mutated each individual residue to alanine (A) in doxa-1 transgenes. Interestingly, only the T343A mutation significantly weakened the rescuing effects of the transgene (Table 3). These results suggest that T343 is important for DOXA-1 activity. However, it remains unclear whether the importance of this site depends on its phosphorylation.

Discussion
In this study, we identified a novel role of CSNK-1 CSNK1G in oxidative stress response. We uncovered nonallelic noncomplementation interactions between C. elegans csnk-1 and bli-3/ tsp-15/doxa-1 NADPH dual oxidase genes, which led to the findings of a potentially conserved biochemical interaction between CSNK-1 and DOXA-1 and a conserved function of CSNK-1 in promoting ROS levels. We also detected genetic interactions between csnk-1 and skn-1 in oxidative stress response. CSNK-1 is an embryonically essential kinase with novel function in postembryonic oxidative stress response csnk-1 is an essential gene that affects the embryonic asymmetric spindle positioning and oocyte meiosis in C. elegans [44][45][46][47][48][49]. Consistently, we found that eggs laid by csnk-1(lf) mutants failed to hatch. The effect of csnk-1 on the Sisi phenotype is specific, as RNAi knockdown of kin-19, kin-20 and three other kinase genes did not obviously affect the phenotype  Table 5. csnk-1(lf) and skn-1(gf) mutations exhibit synergy on the Sisi phenotype in 50 mM NaI. (S2 Table). The importance of the conserved C-terminal palmitoylation for CSNK-1 activity is supported by the finding that mutations of the site caused a loss of the function. Our transgene rescue results suggest functional similarity among the three human CSNK1Gs in affecting oxidative stress response. This is consistent with a recent finding that human CSNK1Gs exhibited functional redundancy in regulating Wnt signaling [63]. Together with our findings that CSNK-1 and human CSNK1G2 can promote ROS levels, we suggest that CSNK1Gs are important regulators of ROS homeostasis across species.

CSNK1G interacts with the NADPH dual oxidase complex
Mammals have seven membrane-localized NADPH oxidases as major generators of endogenous ROS [2,31], among which the dual oxidase DUOX2 is essential for thyroid hormone synthesis, while DUOX1 is more involved in immune cell functions [31]. It is not well understood how DUOX activities are regulated. DUOXA1 and DUOXA2 are conserved facilitators for the ER-Golgi-plasma membrane translocation of DUOX1 and DUOX2 [59]. In C. elegans, BLI-3 DUOX, DOXA-1 DUOXA and TSP-15 tetraspanin form a dual oxidase complex [16,17]. Recently, the highly conserved MEMO-1 protein was identified as a negative regulator of BLI-3 activity via binding the RHO-1/RhoA/GTPase [15], providing a new regulatory mechanism for dual oxidase activities.
We suggest that CSNK-1 interacts with the dual oxidase complex and might regulate dual oxidase activity. First, the nonallelic noncomplementation interactions between csnk-1 and bli-3/tsp-15/doxa-1 imply physical and/or functional interactions between CSNK-1 and BLI-3/ TSP-15/DOXA-1. Second, CSNK-1 was colocalized with DOXA-1 on C. elegans epithelial subcellular structures. So did human CSNK1G2 and DUOXA2 in HEK293T cells. Third, that CSNK-1 and DOXA-1 were detected in a same protein complex by immunoprecipitation and pull-down experiments suggests a possible direct physical interaction between CSNK-1 and DOXA-1, and human CSNK1G2 and DUOXA2 were detected in a same protein complex by immunoprecipitation. Fourth, CSNK-1 is required for normal ROS levels in C. elegans and CSNK1G2 and DUOXA2 each can promote ROS levels in HEK293T cells.
We found that D4476, a commonly used CSNK1 inhibitor, can decrease ROS levels in HEK293T cells in a dose-dependent manner and suppress the promotion of ROS levels by overexpressed CSNK1G2 and DUOXA2, supporting the involvement of CSNK1 in regulating ROS levels. However, our evidence is insufficient for connecting CSNK1s directly with ROS level regulation before extensive analyses of the expression and function of each CSNK1 in HEK293T are performed.
Mutagenesis analyses suggest that T343, the phospho-acceptor of a potential CSNK1 site in DOXA-1 C-terminal domain (S8 Fig), was required for DOXA-1 activity, implying phosphorylation as a mechanism for regulating DOXA-1 activity. However, the lack of conservation in the aligned DUOXA2 sequence raises the question about the importance of this site in other species. Considering that four potential CSNK1 phosphorylation sites are detected in the Cterminal domain of DUOXA2 (S8 Fig), two of which appear to be conserved in DOXA-1, we suspect that the interaction between CSNK-1 and DOXA-1 and the phosphorylation of DOXA-1, if it exists, might be located at different sites. To answer this question, it is necessary to identify the interacting domains between these proteins, the potential phosphorylation sites on DOXA-1, and the functional importance of these sites in future.

csnk-1 genetically interacts with skn-1
SKN-1 is the C. elegans ortholog of Nrf2 [8,9] and a key regulator of oxidative stress response and other cellular processes [6]. We previously found that skn-1 can genetically interact with bli-3/tsp-15/doxa-1 to affect the Sisi phenotype [20]. Similarly in this study we found that csnk-1 and skn-1 can genetically interact to affect the Sisi phenotype. Specifically, the lack of Sisi phenotype in skn-1(lf) mutants in 5 mM NaI was suppressed by csnk-1(lf), while csnk-1(lf) and skn-1(gf) exhibited synergy in promoting the Sisi phenotype in 50 mM NaI. These interactions imply an opposite effect of csnk-1 on oxidative stress response that is in parallel with skn-1. However, the molecular genetic mechanism underlying such interactions remains to be understood considering that the genetic analyses were performed on mutants with maternal effects and that reducing ROS levels (like in bli-3/tsp-15/doxa-1 loss-of-function mutants) [19,20], increasing antioxidant expression (like in skn-1(gf) mutants) [20] or adding antioxidants such as vitamin C or NAC to NGM plate [20] all can cause the Sisi phenotype.
We found that the expression of the SKN-1 reporter transgene dvIs19 (gst-4p::GFP) was increased in csnk-1(lf) mutants and the increase existed even with skn-1(RNAi) treatment. However, considering that RNAi knockdown of skn-1 could be partial, we have yet to interpret such an effect of csnk-1 as skn-1-independent. To gain a deeper understanding of the interaction between CSNK-1 and SKN-1, it will be important to determine how ROS levels are affected in csnk-1; skn-1 double mutants, whether endogenous gst-4 expression is similarly affected by csnk-1(lf), or whether any other gene expression is affected by csnk-1. We hope to address these questions in future using multidisciplinary approaches such as transcriptome and phosphoproteomic analyses.

Conclusions
Among casein kinase 1 members, CSNK1Gs are understudied and are classified by the NIH program "Illuminating the Druggable Genome" as "dark kinases" calling for extra exploration (https://commonfund.nih.gov/IDG). We uncovered an important role of CSNK-1 CSNK1G in regulating oxidative stress response and ROS levels, which involves potentially conserved interactions between CSNK1G and the NADPH dual oxidase complex. Our findings provide novel insights into redox biology and may facilitate the understanding of diseases caused by mutations in these genes and more broadly by defective oxidative stress response.

Materials and methods
Mapping and cloning of csnk-1 mac397 was previously isolated as a Sisi F 1 progeny of EMS-mutagenized wildtype P 0 animals grown on plates with 5 mM NaI [20]. To map mac397, males of the Hawaiian strain CB4856 were crossed with mac397 hermaphrodites. F 1 male progeny were crossed with CB4856 hermaphrodites on plates with 5 mM NaI to generate Sisi F 2 male progeny. Similar crosses were performed three more rounds to generate Sisi F 5 hermaphrodites, which were picked to individual plates with 5 mM NaI and allowed to propagate. Animals on these plates were examined for single nucleotide polymorphisms (SNPs) [64,65]. We established a linkage on Chr. I to the right of SNP WBVar00240395 (SNP W03D8: 34384 D1 = T, genetic location: -5.68; GenBank accession no. FO081764) and to the left of SNP WBVar00245221 (SNP F58D5: 17029-17032 D4 = ACTA, genetic location: 13.02; GenBank accession no. AL137227).
To determine the genomic sequences of mac397 mutants, progeny of unbackcrossed original isolate were washed and starved for~4 hrs in H 2 O. Genomic DNAs were extracted by proteinase K digestion, followed by RNase A treatment and two rounds of phenol-chloroform extraction. Three genomic DNA libraries (350 bp inserts) were constructed by Annoroad Gene Technology Corporation (Beijing) using Illumina's paired-end protocol. Paired-end sequencing (100-bp reads) was performed on the Illumina HiSeq X Ten. 4G clean bases were mapped to the N2 genome (Wormbase release 220) after removal of duplicated reads.
Genomic sequencing identified 7 homozygous and 4 heterozygous missense mutations between the mapped SNPs on Chr. I, among which were a homozygous mutation in tsp-15 and a heterozygous mutation in csnk-1 (S1 Table). The homozygosity of the tsp-15 mutation might be caused by prolonged propagation of the original isolate in 5 mM NaI before the strain was sequenced (see below).
We initially suspected that the tsp-15 mutation (exon3:c.G331A:p.G111R, named mac499) (S1 and S3 Tables) might be a dominant negative allele, thus causing the Sisi phenotype of the mac397 isolate as a heterozygote. To test this, we generated two knock-in mutations genocopying mac499 (S3 Table, mac500 and mac501). The heterozygous knock-in mutants were not Sisi, while the homozygous mutants were (S3 Table). In addition, a tsp-15 transgene can rescue the Sisi phenotype of tsp-15(mac499) homozygous mutants (S3 Table). Therefore, tsp-15 (mac499) is recessive, which led us to suspect that a linked mutation, either alone as heterozygote or together with tsp-15(mac499)/+, caused the Sisi phenotype of the original isolate.
To identify this gene, we treated wildtype animals with feeding RNAi targeting the genes listed in S1 Table. Besides tsp-15(RNAi), we found that csnk-1(RNAi) also made animals Sisi (S1 and S2 Tables). The corresponding mutation in csnk-1 was name mac397.
After we found that csnk-1(RNAi) can cause Sisi, we re-examined our mac397 frozen stocks but were only able to detect tsp-15(mac499) mutation. mac397, either as heterozygote or homozygote, was not found in different batches of thawed animals.
Considering that mac397 was a loss-of-function mutation (Table 1) and csnk-1(lf) homozygous mutations cause lethality, we postulate that a prolonged propagation of the mutants before freezing probably caused this phenomenon. In the original isolate, csnk-1(mac397) was likely heterozygous (S1 Table, genomic sequencing results in the early propagation phase). A prolonged propagation of the strain in 5 mM NaI, with tsp-15(mac499) csnk-1(mac397)/+ + as the presumptive starting genotype, would generate crossover between tsp-15 and csnk-1 and allow tsp-15(mac499) (viable as heterozygotes or homozygotes) to outcompete csnk-1(mac397) (only viable as heterozygotes) in the population. At the time of freezing, most animals, if not all, might have carried only a tsp-15(mac499) homozygous mutation and lost the csnk-1 (mac397) mutation.

C. elegans survival assay in excess iodide (Sisi phenotype assay)
The survival assay was performed as described [19] with modification.
In general, bleached eggs were grown on OP50-seeded NGM plates with different concentrations of NaI (5 mM or 50 mM). The number of L1 larva was counted and the number of adults was counted 3 days post L1.
For heterozygous males in Table 2, males and hermaphrodites of the desired genotypes were crossed on plates with 5 mM NaI. The cross progeny were allowed to grow in the same plate for 4-6 days and the appearance of young adult males was recorded.

Generation of csnk-1 or tsp-15 mutations using CRISPR/Cas9
We followed the method [66] with modifications. The DNA mixture for injection contained 50 ng/μl Peft-3::Cas9-SV40_NLS, 25 ng/μl PU6::sgRNA (specific for csnk-1 or tsp-15), and 25 ng/μl PU6::sgRNA (specific for dpy-10) as co-injection marker. To generate tsp-15 knockin mutations, we included a 130 nt synthesized oligo (500 nM) containing the desired tsp-15 mutation as the repair template and a repair template for dpy-10 as described [67]. F 1 animals with the Dpy/Rol phenotype were picked to individual plates and their progeny were analyzed for the desired mutations by Sanger sequencing. Target sequences of sgRNAs and the sequence of tsp-15 repair template are shown in S6 Table. mac494 and mac495 homozygous mutants derived from heterozygous parents can grow into adults and were able to lay multiple eggs. However, these eggs would not or rarely hatch. We maintained csnk-1 mutations using the hT2 [bli-4(e937) let-?(q782) qIs48] (I; III) balancer.
For csnk-1 rescue experiments, the transgene mixture containing 0.5 ng/μl or 5 ng/μl of the transgenes of interest with 20 ng/μl pPD95_86 (myo-3p::GFP) (a gift from Andrew Fire) as co-injection marker was injected to wildtype animals to generate at least two stable lines. The transgenes were crossed into csnk-1(mac494lf)/hT2[qIs48] animals and csnk-1(mac494lf) homozygous progeny carrying the transgenes were examined. For the Sisi phenotyping, transgenic adults were bleached, and eggs were placed on regular or 5 mM NaI NGM plates for 12 hrs to hatch.~100 transgenic homozygous csnk-1(mac494lf) L1 larva were picked to a new 5 mM NaI NGM plates. The numbers of adults were counted after 3 days.

Hoechst 33258 staining
Hoechst 33258 staining was performed as described [16,19] with modifications. Young adult animals (24 hrs post mid-L4) were washed off plates and incubated for 15 min in M9 containing 1 μg/ml Hoechst 33258 (Sigma, 861405) at 20˚C with gentle shaking. After staining, animals were washed three times with M9 and observed under a Leica DM5000B fluorescence microscope.

RNA interference
Bleached eggs were transferred to plates seeded with HT115 (DE3) bacteria expressing RNAi plasmids on NGM plates with 1 mM IPTG and 0.1 mg/ml Ampicillin. The progeny were examined under dissecting microscope for the Sisi phenotype. RNAi feeding bacterial strains for lrp-2, rbpl-1, uggt-2, K07A1.1, sys-1 and gsk-3 were picked from a whole-genome RNAi library [69], and the inserts were confirmed by sequencing. RNAi plasmids for other genes were generated in this study (see Plasmids).

Confocal microscopy
Fluorescent images of immunostained cells or transgenic animals expressing GFP and/or mCherry reporters were captured using a Leica TCS SP5 II or Zeiss LSM 880 laser confocal microscope.

Immunoprecipitation, pull-down and western blotting
For immunoprecipitation, HEK293T cells were lysed in ice-cold IP buffer (Beyotime, P0013) with 1% protease inhibitor cocktail and phosphatase inhibitor cocktail (Bimake, B14001). Lysed cells were incubated at 4˚C for 4 hrs and centrifuged at 12000 rpm for 20 min at 4˚C. The clear supernatants were transferred to a new tube and incubated with an anti-HA antibody (1 μl per sample, Cell Signaling Technology, #3724) or anti-FLAG antibody (1 μl per sample, Sigma, F9291) and protein A/G-beads (Bimake, B23201, 10 μl) at 4˚C overnight. The beads were centrifuged and washed 5 times in TBST. Proteins associated with the beads were mixed with 32 μl 2x SDS lysis buffer and 8 μl 5x SDS loading buffer and boiled for 5 min at 95˚C.
For pull-down, BL21 bacteria transformed with pCold TF:: doxa-1_cDNA or pCold TF:: csnk-1_cDNA plasmid were cultured in Luria broth (LB) at 37˚C to OD 0.6, and 1 mM IPTG was added to induce protein expression at 16˚C for 24 hrs. Bacteria were collected and lysed in PBST (1% Triton X-100) with 1 mM lysozyme (Beyotime, ST206) on ice for 30 min, followed by centrifugation at 12000 rpm, 4˚C for 20 min. The clear supernatant was transferred to a new tube and incubated with His-tag resin (Beyotime, P2229S-1) at 4˚C for 2 hrs. The resin was collected after centrifugation at 1200 rpm for 20 sec and used to incubate at 4˚C overnight with the lysate supernatant of HEK293T cells transfected with pCMV-Tag2B (FLAG)::csnk-1_cDNA or pcDNA3.1-doxa-1_cDNA::HA plasmid. On the next day the mixture was washed three times in TBST. Proteins associated with the resin were mixed with 20 μl 2x SDS lysis buffer and 5 μl 5x SDS loading buffer and boiled for 5 min at 95˚C.

ROS measurement using DCFDA
Whole-animal ROS levels were measured using 2,7-dichlorofluorescein diacetate (DCFDA) (Sigma, D6883) following previous methods [70,71] with modifications. Synchronized L4 larval animals were washed with H 2 O for three times. The population density was adjusted with H 2 O to 50 to 200 animals per 100 μl. Two to three 100 μl aliquots were transferred to individual wells of a black-walled 96-well plate. DCFDA dissolved in DMSO was added to a final concentration of 50 μM. Animals were incubated at 20˚C for 30 min and measured for DCFDA fluorescence intensity using a fluorimeter (BioTek, Synergy 2) at the excitation wavelength of 485 nm and the emission wavelength of 528 nm at room temperature. After the measurement, animals in each well were emptied to an agar plate and counted.
For DCFDA signals of individual animals, synchronized L4 animals were washed three times in H 2 O and resuspended in M9 to a final volume of 95 μl. 5 μl 1 mM DCFDA solution was added to a final concentration of 50 μM. Animals were incubated at 20˚C for 30 min, washed in M9 three times, transferred to a glass slide, and covered with a cover glass. Fluorescent pictures of multiple animals were taken immediately with the same exposure intervals. Whole body fluorescence intensities of individuals were measured with ImageJ.
To measure dose-dependent effects of D4476 (Sigma, D4476) on ROS levels, HEK293T cells at 70% confluency in 24-well plates (coated with 0.01 mg/ml poly-D-lysine (Sigma, P6407)) were treated with different concentrations of D4476 in duplicate wells. After 24 hrs, the culture media was removed, and cells were washed once with DMEM. A 500 μl volume of 50 μM DCFDA in DMEM was added to each well, followed by incubation at 37˚C for 20 min. The cells were washed three times with PBS and DCFDA fluorescence intensities were immediately measured using a fluorimeter (BioTek, Synergy 2) at the excitation wavelength of 485 nm and the emission wavelength of 528 nm at room temperature.
For CSNK1G2 or DUOXA2 overexpression, cells at 70% confluency were transfected with 200 ng plasmids per well in a 24-well plate. After 24 hrs, D4476 was added to the media at a final concentration of 30 μM and DCFDA staining was performed after 24 hrs.

ROS measurement using Amplex Red
Whole-animal ROS levels were measured using Amplex Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen, A22182) in a 96-well plate following manufacturer's instructions and previous methods with modifications [72,73]. Amplex Red was added to a final concentration of 50 μM. Synchronized L4 animals were incubated at 20˚C for 30 min and measured for fluorescence intensity using a fluorimeter (BioTek, Synergy 2) at the excitation wavelength of 530 nm and the emission wavelength of 590 nm at room temperature. After the measurement, animals in each well were emptied to an agar plate and counted.
To measure dose-dependent effects of D4476 (Sigma, D4476) on ROS levels, HEK293T cells at 70% confluency in 24-well plates coated with PDL were treated with different concentrations of D4476 in duplicate wells. After 24 hrs, the culture media was removed, and cells were washed once with DMEM. A 300 μl volume of 50 μM Amplex Red was added to each well followed by incubation at 37˚C for 20 min and measuring for fluorescence intensity using a fluorimeter (BioTek, Synergy 2) at the excitation wavelength 530 nm and the emission wavelength of 590 nm at room temperature.

ROS measurement using transgenic HyPer reporter jrIs1
Synchronized L4 animals were picked to triplicate wells of a 96-well plate in 100 μl H 2 O, with 20 animals per well. Fluorescent signals were measured using a fluorimeter (BioTek, Synergy 2) with excitation wavelength of either 490 nm (for oxidized HyPer) or 405 nm (for reduced HyPer) and emission filter of 535 nm. The quantification of the oxidized/reduced HyPer signals was based on the method described by Back et al [60].