|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: AG ARM GC MBJ TK M-WT JDL. Performed the experiments: ARM GC MBJ TK LMK VC AG. Analyzed the data: SH MBJ ARM GC AG. Contributed reagents/materials/analysis tools: MBJ JDL. Wrote the paper: AG.
Insulin signaling has a profound effect on longevity and the oxidative stress resistance of animals. Inhibition of insulin signaling results in the activation of DAF-16/FOXO and SKN-1/Nrf transcription factors and increased animal fitness. By studying the biological functions of the endogenous RNA interference factor RDE-4 and conserved PHD zinc finger protein ZFP-1 (AF10), which regulate overlapping sets of genes in Caenorhabditis elegans, we identified an important role for these factors in the negative modulation of transcription of the insulin/PI3 signaling-dependent kinase PDK-1. Consistently, increased expression of pdk-1 in zfp-1 and rde-4 mutants contributed to their reduced lifespan and sensitivity to oxidative stress and pathogens due to the reduction in the expression of DAF-16 and SKN-1 targets. We found that the function of ZFP-1 in modulating pdk-1 transcription was important for the extended lifespan of the age-1(hx546) reduction-of-function PI3 kinase mutant, since the lifespan of the age-1; zfp-1 double mutant strain was significantly shorter compared to age-1(hx546). We further demonstrate that overexpression of ZFP-1 caused an increased resistance to oxidative stress in a DAF-16–dependent manner. Our findings suggest that epigenetic regulation of key upstream signaling components in signal transduction pathways through chromatin and RNAi may have a large impact on the outcome of signaling and expression of numerous downstream genes.
Reduced activity of the insulin-signaling pathway genes has been associated with a longer lifespan and increased resistance to oxidative stress in animals due to the activation of important transcription factors, which act as master regulators and affect large networks of genes. The ability to manipulate insulin signaling and reduce its activity may allow activation of oxidative-stress response programs in pathological conditions, such as neuronal degeneration, where oxidative stress plays a significant role. Here, we describe a new way of inhibiting insulin signaling that exists in the nematode Caenorhabditis elegans. We find that transcription of one of the insulin-signaling genes is inhibited by mechanisms involving chromatin and RNA interference, a silencing process that depends on short RNAs. We demonstrate that mutants deficient in RNA interference are more susceptible to stress due to increased insulin signaling and that increased dosage of a chromatin-binding protein repressing insulin signaling and promoting RNA interference leads to better survival of nematodes grown under oxidative stress conditions. Since there is a clear homolog of this chromatin-binding protein in mammals, it may also act to promote resistance to oxidative stress in human cells such as neurons.
The role of RNA interference (RNAi) in the silencing of transposons and other repetitive elements is well documented , , while the knowledge of its impact on endogenous genes and signaling pathways is limited. In this article we investigate the remarkable and similar effects of the Caenorhabditis elegans RNAi-promoting factors RNAi DEficient 4 (RDE-4)  and Zinc Finger Protein 1 (ZFP-1) on the expression of stress-related genes. We focus on the key gene regulated by RDE-4 and ZFP-1, pdk-1, which encodes 3-phosphoinositide-dependent kinase-1 (PDK-1) , a component of a conserved insulin-signaling pathway. We describe a functional connection between zfp-1, rde-4 and insulin signaling in genetic epistasis experiments and demonstrate the significance of pdk-1 regulation by zfp-1 and rde-4 for C. elegans fitness.
ZFP-1, a Plant Homeo Domain (PHD) zinc finger protein, was first identified as a factor promoting RNAi interference in C. elegans –. It is a homolog of mammalian AF10 (Acute Lymphoblastic Leukemia 1-Fused gene from chromosome 10)  and plays a key role in leukemias caused by Mixed Lineage Leukemia MLL-AF10 fusion due to the recruitment of histone methyltransferase Dot1 by the AF10 portion of the fusion protein . The developmental and physiological roles of AF10 are largely unknown. RDE-4 is a double-stranded RNA (dsRNA)-binding protein and a component of the Dicer complex responsible for the production of short interfering RNAs (siRNAs) from exogenous dsRNA . The rde-4(ne299) null mutation was discovered in a screen for RNAi resistant mutants . rde-4(ne299) does not have obvious developmental abnormalities, but shows synthetic phenotypes when combined with the null mutant in C. elegans Retinoblastoma gene lin-35  and appears to be less healthy at elevated temperatures . Also, rde-4 mutants were reported to have a slightly reduced lifespan . The effects of rde-4 loss-of-function are likely to be related to recently identified endogenous siRNAs (endo-siRNAs), which perfectly match thousands of genes in C. elegans either in sense or antisense orientation –. Indeed, the expression of some endo-siRNAs is diminished in the absence of rde-4 , .
Our recent genome-wide mRNA expression study has revealed that ZFP-1 and RDE-4 affect the transcript levels of close to 250 overlapping genes . Furthermore, putative target genes of endo-siRNAs  showed a significant enrichment among genes upregulated in the rde-4(ne299) null  and zfp-1(ok554)  loss-of-function mutant larvae . We proposed that ZFP-1 and endo-siRNAs produced in an rde-4-dependent manner cooperate in the repression of target genes in the nucleus. Here, we confirm a direct repressive effect of ZFP-1 on endo-siRNA targets by comparing gene expression changes in zfp-1(ok554) and rde-4(ne299) with genome-wide localization of ZFP-1. Moreover, using functional analysis of misregulated genes we predict a role for RDE-4 and ZFP-1 in modulating insulin signaling and further demonstrate that regulation of pdk-1 transcription by ZFP-1 and endogenous RNAi underlies the oxidative stress sensitivity and short lifespan of zfp-1(ok554) and rde-4(ne299) mutants.
In order to elucidate the common biological roles of ZFP-1 and endogenous RNAi we analyzed gene sets misregulated in zfp-1(ok554) and rde-4(ne299) mutants . We found that genes with lowered expression in the mutants compared to the wild type were enriched in metabolic, oxidative stress-related and anti-pathogenic factors present in the intestine (Table S1). Since insulin signaling mutations lead to increased expression of factors important for defense against oxidative stress and pathogens –, we decided to compare the lists of genes downregulated in zfp-1(ok554) and rde-4(ne299) with longevity-promoting “Class 1” genes found upregulated in the daf-2 mutant in a daf-16-dependent manner .
Insulin-like signaling in C. elegans via the DAF-2 insulin receptor and phosphatidylinositol 3-kinase (PI3K) negatively regulates the DAF-16/FOXO ,  and SKN-1/Nrf  transcription factors. When insulin signaling is reduced, the enhanced DAF-16 and SKN-1 activities contribute to longer lifespan and stress resistance in worms due to concerted regulation of many of their targets –, , . DAF-16 and SKN-1 are negatively regulated in part at the level of their nuclear localization; therefore, mutants in this pathway are long-lived due to a higher level of the active nuclear DAF-16 and SKN-1 and appropriate transcriptional activation or repression of their direct targets. Our analyses revealed that genes downregulated in the zfp-1 and rde-4 mutants significantly overlapped with “Class 1” longevity promoting genes upregulated in the daf-2 mutant (a condition when DAF-16 and SKN-1 are activated)  (Figure 1, Table 1, Table S1). Examples of genes whose expression is negatively regulated by daf-2 and positively regulated by zfp-1 and/or rde-4 include glutathione transferases gst-4 and gst-38, and aquaporin (aqp-1) (Table 1, Figure 2A). Since RNAi is a gene-silencing phenomenon and gene sets expressed lower in zfp-1(ok554) and rde-4(ne299) are not enriched in endo-siRNA targets , we predict that “Class 1” longevity-promoting genes are regulated by ZFP-1 and RDE-4 indirectly. Consistently, genome-wide localization data showed no enrichment of ZFP-1 at longevity-promoting genes (Figure 1).
We considered the possibility that a direct target gene negatively regulated by rde-4 and zfp-1 would be de-repressed in the mutants to account for the reduced expression of the secondary targets, which may therefore be regulated by these factors indirectly. Indeed, a component of the insulin-signaling pathway, the kinase PDK-1, was among the most upregulated genes in zfp-1 and rde-4  (Figure 2A). Although our microarray study was performed on zfp-1 and rde-4 mutant larvae (L1–L2), we found that pdk-1 expression was increased in these mutants at other developmental stages as well (Figure 2A).
The zfp-1 gene was shown to be a direct target of DAF-16 by chromatin immunoprecipitation (ChIP) combined with sequencing  and, more recently, using chromatin profiling by DNA adenine methyltransferase identification (DamID) . However, it was not clear whether DAF-16 had a significant role in the regulation of zfp-1. We found that zfp-1 mRNA expression in the daf-2 mutant background was influenced by daf-16 and was 2-fold lower in the daf-2; daf-16 double mutant strain (Figure 2B). Therefore, DAF-16 appears to enhance transcription of zfp-1, although not nearly to the same extent as other prominent DAF-16 targets, such as sod-3 (Figure 2B).
The analyses of gene expression described above suggest a model where ZFP-1 and RDE-4 modulate the insulin-signaling pathway by repressing pdk-1 and that a DAF-16-dependent enhancement of zfp-1 expression under conditions of low insulin signaling may contribute to a positive-feedback loop enhancing the effect of DAF-16 on other targets (Figure 2C).
Next, we determined a molecular lesion in the weak loss-of-function pdk-1 allele sa709  and tested whether the pdk-1(sa709) mutant mRNA was still regulated by ZFP-1and RDE-4. We found that sa709 affects pdk-1 mRNA splicing and leads to the incorporation of intron three into the mature pdk-1 mRNA with a very low expression level of the correctly spliced mRNA in the mutant (Figure 3A, 3B). We combined pdk-1(sa709) with zfp-1(ok554) and found the level of mutant pdk-1 mRNA expression to be elevated in the double mutant compared to pdk-1(sa709) alone (Figure 3C). The pdk-1(sa709) mRNA expression was also elevated in rde-4(ne299); pdk-1(sa709) (Figure 3C). Therefore, regulation of pdk-1(sa709) mRNA expression by ZFP-1 and RDE-4 was similar to that of wild type pdk-1 mRNA.
Since loss-of-function mutations in insulin-signaling components lead to increased nuclear localization of DAF-16::GFP , we tested the pdk-1(sa709) allele in this assay and found that DAF-16::GFP had more prominent nuclear localization in pdk-1(sa709), while it was mostly cytoplasmic in wild type, zfp-1(ok554) and rde-4(ne299) worms (Figure 3D, 3E). Nuclear localization of DAF-16::GFP persisted in pdk-1; zfp-1 and pdk-1; rde-4 double mutant animals (Figure 3D, 3E). These results demonstrate that pdk-1(sa709) is epistatic to zfp-1(ok554) and rde-4(ne299) and support a model where ZFP-1 and RDE-4 affect expression of DAF-16 targets through regulation of pdk-1.
Since longevity-promoting genes have lower expression in the zfp-1 and rde-4 mutants, it is expected that they may live shorter than wild type worms. Indeed, a decrease in lifespan of zfp-1(ok554)  and rde-4(ne299)  has been reported, with the zfp-1 mutant exhibiting a stronger phenotype than rde-4. In order to test whether upregulation of PDK-1 and therefore increased insulin signaling may contribute to the short lifespan of zfp-1, we conducted epistasis experiments with a reduction-of-function mutation in the PI3 kinase AGE-1, age-1(hx546) , . We found that the short lifespan phenotype of zfp-1(ok554) was suppressed by age-1(hx546) (Figure 4A), i.e. the reduction in lifespan of the mutant was dependent on the active insulin signaling. Also, the extended lifespan of age-1(hx546) was dependent on ZFP-1 function, consistent with the possibility that enhanced PDK-1 dosage may suppress the defect in signaling conferred by the non-null age-1 mutation (Figure 4A). Indeed, increased pdk-1 dosage suppresses the constitutive dauer phenotype of age-1(mg44) .
In order to show that high levels of pdk-1 expression contributed to the short lifespan of zfp-1(ok554) we attempted to combine zfp-1(ok554) with a strong loss-of-function mutation pdk-1(sa680)  for genetic suppression analyses. We were not able to recover zfp-1(ok554); pdk-1(sa680) and assume that this double mutant is not viable. Therefore, all epistasis analyses described below were performed with the sa709 allele.
zfp-1 and rde-4 affect expression of multiple target genes, and some phenotypes of zfp-1(ok554), such as dauer promotion , are similar rather than opposite to the phenotypes of insulin-signaling mutants. We have found that zfp-1; pdk-1 and rde-4; pdk-1 double mutants display some egg-laying deficiency, which complicates the longevity assays that we conduct in the absence of drugs inducing sterility. However, although zfp-1; pdk-1 and rde-4; pdk-1 worms were undoubtedly sicker than zfp-1 or rde-4 single mutants, we found that the reduction of pdk-1 function significantly suppressed the decreased lifespans of zfp-1(ok554) and rde-4(ne299) (Figure 4B, 4C). These results further support the idea that ZFP-1 and RDE-4 affect insulin signaling through the negative regulation of pdk-1.
The gene expression signatures of zfp-1 and rde-4 mutants suggested that they could be deficient in oxidative stress response. We induced oxidative stress by soaking L4 animals in 100mM paraquat and found that the zfp-1(ok554) mutant strain was much more sensitive to this treatment compared to the wild type (Figure 4D), similarly to daf-16(mu86) (Figure 4D), while rde-4(ne299) showed moderate sensitivity (Figure 4D, Figure S1A), and age-1(hx546) and pdk-1(sa709) were more resistant than wild type (Figure 4D). We found that zfp-1; age-1, zfp-1; pdk-1 and rde-4; age-1, rde-4; pdk-1 double mutants were less sensitive to oxidative stress than zfp-1 and rde-4, respectively (Figure 4D), indicating that the stress sensitivity of zfp-1(ok554) and rde-4(ne299) was due to active insulin/PI3K signaling.
In order to determine whether increased pdk-1 expression may be sufficient to cause a stress sensitivity phenotype, we tested the SP940 strain, which contains the free duplication mnDp (II;X;f) that includes the pdk-1 locus. We found that pdk-1 mRNA levels are increased about 2.5-fold in this strain (Figure S1C), close to that observed in rde-4(ne299), and it shows a comparable sensitivity to paraquat (Figure S1A, S1B). These data are consistent with the idea that regulating pdk-1 dosage is important for animal fitness.
We generated transgenic lines expressing ZFP-1::GFP and ZFP-1::FLAG fusion proteins by introducing tags into the C-terminal region of ZFP-1 through fosmid recombineering in bacteria . The resulting genes are expressed from the 30kb fosmid and are subject to the same regulatory inputs as the endogenous zfp-1 locus; the transgenes fully rescued the stress sensitivity and reduced lifespan phenotypes of the zfp-1 mutant (Figure 5A, 5B). zfp-1 mRNA expression was about two-fold greater in ZFP-1 transgenic lines compared to the control (Figure S2). Moreover, we found that these ZFP-1 overexpressing lines were more resistant to oxidative stress compared to the control line generated by a similar technique of unc-119 mutant rescue but not containing the ZFP-1 fosmid (Figure 5A). The stress resistance of ZFP-1 overexpressing lines was dependent on DAF-16 function (Figure 5A). This is consistent with the repression of insulin signaling and therefore indirect activation of DAF-16 by ZFP-1. We have not observed lifespan extension in the ZFP-1 overexpressing lines (Figure 5B), which indicates that a higher level of ZFP-1 may be advantageous only in acute stress situations.
An example of an acute stress response is the response of animals to pathogens. The human pathogenic bacterium Pseudomonas aeruginosa (PA14) inhibits DAF-16 nuclear localization and therefore downregulates the expression of defense factors that are dependent on DAF-16 . We tested the effect of the loss of ZFP-1 function on innate immunity by assaying the survival of zfp-1(ok554) animals. Upon exposure to PA14 under the standard infection assay conditions , we observed that the zfp-1(ok554) mutants were significantly more susceptible to P. aeruginosa infection- mediated killing (Figure 6A, 6B). The pathogen sensitivity seen in zfp-1(ok554) mutants was due to loss of ZFP-1 function as was confirmed using a ZFP-1::GFP transgene that rescued the mutant phenotype (Figure 6B).
Next, we tested whether the increased susceptibility of zfp-1(ok554) to PA14 was dependent on insulin signaling. We confirmed that age-1(hx546) was more resistant to the infection (Figure 6C) and tested age-1; zfp-1 double mutants. The results were similar to those found in the longevity assays: age-1 and zfp-1 suppressed each other's phenotypes (Figure 6C). The survival of the double mutant was closer to that of zfp-1(ok554) than age-1(hx546), although age-1 significantly suppressed the sensitivity of zfp-1 to PA14 killing. We conclude that PI3K signaling contributes to the pathogen-sensitivity of zfp-1(ok554).
Consistent with our expression and genetic epistasis data suggesting a direct role of ZFP-1 in repressing pdk-1 transcription, a strong peak of ZFP-1 localization was found at the pdk-1 promoter in ChIP/chip experiments conducted by the modENCODE (model organism ENCyclopedia Of DNA Elements) project (Figure 7A). We confirmed ZFP-1 localization to the pdk-1 promoter by ChIP/PCR with antibodies specific to endogenous ZFP-1 (Figure 7B) as well as with anti-FLAG antibodies in experiments with ZFP-1::FLAG transgenic lines (Figure 8A).
ZFP-1 was not localized to the promoters of other genes of the insulin signaling pathway (daf-2, age-1, akt-1, sgk-1); it was also not present at DAF-16 target genes that have reduced expression in zfp-1(ok554) and appear to be positively regulated by this factor, as discussed earlier (Figure 1, Figure 7B, Figure 8A, and Table S1). There was no enrichment in direct ZFP-1 targets among the longevity-promoting genes (P-value 0.83) (Figure 1). Therefore, ZFP-1 is likely to target directly only genes whose transcription it inhibits (Table S1).
We reported previously a very significant overlap between genes negatively regulated by zfp-1 and rde-4 and endogenous siRNA target genes . Consistently, we find that direct ZFP-1 target genes are overrepresented among genes expressed higher in the rde-4 mutant (Figure 7C). pdk-1 is repressed by ZFP-1 and is also negatively regulated by RDE-4, which is a dsRNA-binding protein required for the biogenesis of siRNAs in the exogenous RNAi pathway  and contributing to the biogenesis of some endo-siRNAs , . Knowing this, we searched available deep sequencing data – for endo-siRNAs mapping to the pdk-1 locus. There were few endo-siRNAs corresponding to the coding region of pdk-1, and more siRNAs mapped to the promoter region of the gene (5kb upstream of the transcription start site), including a predicted open reading frame, H42K12.2. (Figure 7A). However, for this open reading frame, no transcriptional evidence exists, neither from EST collections nor from deep sequencing runs undertaken in the context of the modENCODE project, and it therefore appears to be a mis-annotated gene , . We were able to detect ~100–250 nt transcripts at the pdk-1 promoter produced from both the plus and minus DNA strands, consistent with the possibility of dsRNA production and processing by RDE-4 and Dicer (Figure 7D and Text S1). Moreover, we detected an elevated level of this RNA in the rde-4 mutant (Figure 7E), further supporting the possible involvement of RDE-4 in the dsRNA processing. Unfortunately, pdk-1 promoter-specific endo-siRNAs are expressed at a very low level, and we were not able to reliably detect them by RT-qPCR. Nevertheless, additional evidence for pdk-1 regulation by endogenous RNAi comes from the observation that pdk-1 mRNA levels are increased in drh-3(ne4253), a loss-of-function mutant in dicer-related helicase 3 , (Figure 8B). DRH-3 is thought to participate in multiple branches of endogenous RNAi in C. elegans .
We have shown that both zfp-1 and rde-4 affect the longevity of C. elegans and its ability to resist oxidative stress and that pdk-1 mRNA levels are elevated in zfp-1(ok554) and rde-4(ne299) ( and Figure 1 and Figure 3). Furthermore, we have demonstrated that ZFP-1 binds the pdk-1 promoter and that endogenous siRNAs also have a potential to regulate pdk-1 directly. Next, we analyzed RNA polymerase II (Pol II) occupancy at the pdk-1 promoter and coding region by ChIP in wild type, zfp-1(ok554) and rde-4(ne299) L3-L4 animals and found it to be significantly increased in both mutants (Figure 8C, 8D). Consistent with transcriptional regulation, pdk-1 pre-mRNA levels were elevated in both mutants as well (Figure 8E). RDE-4, and therefore rde-4-dependent endo-siRNA production, did not affect ZFP-1 localization to the pdk-1 promoter (Figure 8A). It is possible that ZFP-1 and the RNAi machinery are independently recruited to the same targets and cooperate in inhibiting their transcription. Alternatively, ZFP-1 may help stabilize downstream RNAi factors at the endo-siRNA target genes.
Pol II levels increased only at the promoters, but not at the coding regions of indirect target genes expressed lower in zfp-1(ok554) and rde-4(me299) (Figure 8C, 8D), a signature consistent with a slower rate of transition from transcriptional initiation to elongation . This finding reflects the lower expression of these genes in the mutants, although they are not regulated directly by ZFP-1 and do not belong to the group of prevalent endo-siRNA targets ( and Table S1).
We have previously described a very significant overlap between genes misregulated in zfp-1(ok554) and genes misregulated in rde-4(ne299) and noted that the level of expression of zfp-1 mRNA did not change in rde-4(ne299) and vice versa . Since the rde-4 mutation has milder effects on gene expression than zfp-1(ok554), they could potentially be due to zfp-1 misregulation. Therefore, we further confirmed that protein levels of ZFP-1 are not decreased in rde-4(ne299) (Figure S3).
ZFP-1 localizes to the pdk-1 promoter and both the pdk-1 mRNA level and Pol II occupancy at the pdk-1 gene are increased in zfp-1(ok554). These findings strongly suggest that transcription of pdk-1 is directly and negatively modulated by ZFP-1. Our genetic and molecular data also clearly demonstrate that rde-4 has a role in the transcriptional regulation of pdk-1. Several lines of evidence provide correlative support for a possible direct role of endo-siRNAs in pdk-1 regulation: endo-siRNAs match the pdk-1 promoter in a region also targeted by ZFP-1, dsRNA production is detected at the promoter and is increased in rde-4(ne299), and pdk-1 mRNA levels are elevated in at least two RNAi pathway mutants. However, since the endo-siRNAs targeting pdk-1 are not very abundant, we were not able to determine whether they change in rde-4(ne299), and there is a possibility that rde-4 affects pdk-1 transcription indirectly. In either case, RDE-4 is most likely involved in gene regulation through endo-siRNA production since this is the only known molecular function of this protein. The relatively more abundant endo-siRNAs matching the pdk-1 promoter (Figure 7A) are not unique and correspond to the repeat sequences. The Argonaute proteins that bind endo-siRNAs and work downstream in the RNAi pathways have been described and include at least two separate branches: the CSR-1 branch  and the WAGO branch . Although CSR-1-bound endo-siRNAs are enriched in sequences antisense to protein-coding genes, they also include endo-siRNAs matching repeats , while the WAGO system appears to preferentially target repeats and pseudogenes . Both the WAGO and CSR-1 systems have been shown to have some connection to the RDE-4-regulated genes , , and ZFP-1 ChIP/chip targets are enriched in both WAGO and CSR-1-dependent endo-siRNA target gene sets (G. Cecere, M. Jensen, et al., manuscript in preparation). Therefore, we think that regulation of some endogenous genes, such as pdk-1, which contain simple repeats in their promoters, may have evolved to depend on the RNAi surveillance system, either WAGO or CSR-1-based.
This work has revealed a direct repression of pdk-1 transcription by C. elegans AF10 homolog ZFP-1 and the significance of this transcriptional regulation in modulating insulin signaling. We have demonstrated that overexpression of ZFP-1 leads to enhanced resistance to oxidative stress in nematodes in a DAF-16-dependent manner. The role of DAF-16/FOXO in longevity and stress response is conserved in animals , and it would be interesting to see whether AF10 has a role in promoting stress resistance through the activation of FOXO. FOXO proteins have been shown to cause a neuroprotective effect in C. elegans, Drosophila and mammalian models of neurodegeneration . Another transcription factor involved in the antioxidant response, Nrf2 – a homolog of C. elegans SKN-1 - has been implicated in the neuroprotection of motor neurons in a mouse model of ALS , while SKN-1 was shown to be important for protection of dopamine neurons against methylmercury-induced degeneration in C. elegans . Since both DAF-16 and SKN-1 are negatively regulated by insulin/PI3K signaling in C. elegans  (Figure 2C), perhaps inhibition of this signaling pathway in mammalian neurons will lead to activation of both FOXO3a and Nrf2. Our work suggests that the homolog of ZFP-1, AF10, may have a neuroprotective effect by indirectly activating FOXO3a and Nrf2 if the regulation of pdk-1 by ZFP-1/AF10 is conserved in animals.
RNAi was discovered in C. elegans as a response to exogenously introduced dsRNA ,  and was considered to be primarily an anti-viral mechanism also directed against repetitive elements , especially since the first RNAi-resistant mutants did not have obvious developmental phenotypes . The discovery of mutants in RNA-dependent RNA polymerase (RdRP) genes that displayed developmental phenotypes ,  and were either RNAi-resistant  or more sensitive to exogenous RNAi , highlighted the possibility that RNAi may be used for regulating endogenous genes. Indeed, endogenous siRNAs antisense to protein-coding genes and similar to those produced during exogenous RNAi were discovered first in the worm  and then in other animals . It became apparent that in mutants lacking specific endo-siRNAs, corresponding mRNAs become upregulated , , , and microarray and deep sequencing approaches have been used for identifying genes that change expression in the RNAi mutants , , , –, , –. However, the significance of misregulation of specific genes for the biology of the worm has not been clearly demonstrated and phenotypes described for RNAi-related mutants , , , – were not connected to specific targets by functional epistasis experiments. This study interprets the microarray signature of zfp-1 and rde-4 mutants, demonstrates short lifespan and stress sensitivity phenotypes consistent with the signature, and provides functional evidence that pdk-1 is a major target responsible for these phenotypes through genetic epistasis, RNA expression and ChIP analyses.
RNAi in C. elegans has the potential to cause both post-transcriptional  and transcriptional ,  gene silencing. It is possible that endogenous RNAi utilizes multiple mechanisms and that some genes are subject mostly to post-transcriptional regulation while others are regulated at the transcriptional level; the latter are likely to have fewer matching endo-siRNAs to the coding region and relatively more promoter-specific endo-siRNAs, like pdk-1. We surveyed the genes upregulated in rde-4(ne299) for an endo-siRNA signature similar to that of the pdk-1 locus and found a number of examples (Figures S4 and S5). Interestingly, most of these types of genes, including pdk-1, have repetitive elements at the promoters and endo-siRNAs matching them. It appears that a modulating effect of RDE-4 on the transcription of some endogenous genes is linked to the control of repetitive elements.
RNAi-dependent silencing of long terminal repeats (LTR) and non-coding RNA genes located in euchromatic regions that functions with trace amounts siRNAs has been described recently in S. pombe . The lack of abundant siRNA species was remarkable, considering that Dicer and RdRP interacted physically with the loci and that LTR transcript levels were significantly elevated in the dcr1, ago1 and rdp1 mutants. This type of RNAi-based regulation appears to be very similar to that operating on the pdk-1 gene in C. elegans that we describe here.
Examples of genes regulated by RNAi through repetitive elements in promoters already exist in Arabidopsis and include the FWA gene, which affects flowering time ,  and, more recently, an extracellular peroxidase Ep5C gene . High levels of Ep5C promote susceptibility to Pseudomonas syringae and mutation in the Argonaute 4 gene was recovered in an unbiased screen for increased susceptibility to infection . It is interesting that both in plants and animals regulation of endogenous genes by RNAi has evolved to promote fitness.
Strains were maintained at 20°C unless otherwise noted, using standard methods . The following mutants were used: LGI: daf-16(mu86), daf-16(mgDf50), LGII: age-1(hx546), LGIII: daf-2(e1370), rde-4(ne299), zfp-1(ok554), LGX: pdk-1(sa709).
Compound mutant strains and transgenes used are as follows:
CF1595: daf-16(mu86)I; daf-2(e1370)III, AGK138: zfp-1(ok554)III; pdk-1(sa709)X, AGK241: rde-4(ne299)III; pdk-1(sa709)X, AGK25: age-1(hx546)II; zfp-1(ok554)III, AGK264: age-1(hx546)II; rde-4(ne299)III, AGK72: daf-16(mgDf50)I; armEx5, TJ356: zIs356 IV, AGK30: zfp-1(ok554)III; zIs356 IV, AGK262: zfp-1(ok554)III; zIs356 IV; pdk-1(sa709)X, AGK377: rde-4(ne299)III; zIs356 IV, AGK 265: rde-4(ne299)III; zIs356 IV; pdk-1(sa709)X, AGK267: zfp-1(ok554) unc-119(ed3)III; armIs5, AGK248: rde-4(ne299) zfp-1(ok554) unc-119(ed3)III; armIs5, AGK260: zIs356 IV; pdk-1(sa709)X, SP940: unc-52(e444)II; unc-1(e538)X; mnDp11(II;X;f).
Transgenic worms were created by microparticle bombardment using a PDS-1000 Hepta Apparatus (Bio-Rad) . All strains were made by co-bombardment of both a fosmid of interest and plasmid pMM016b (AddGene) for unc-119(ed3)III rescue. Strains created are as follows: AGK29: armIs2 Is[unc-119+] – control strain, AGK128: armIs5 Is[ZFP-1::FLAG,unc-119+], AGK26: armEx5 Ex[ZFP-1::GFP,unc-119+].
The WRM0629bD09 fosmid containing the ZFP-1 locus was obtained from the C. elegans fosmid library generated by C. elegans Reverse Genetics Core Facility, Vancouver, B.C., Canada.
We generated derivative fosmid constructs to express recombinant ZFP-1 protein tagged with GFP or FLAG at the C-terminal portion of the protein by a fosmid recombineering method as described by .
Paraquat sensitivity assays were done essentially as described by . L4 animals were transferred from NGM agar plates into 24-well plates (10 per well) containing 300 µL of 100 mM paraquat dissolved in M9. Worms were then incubated at 20°C and scored for survival after 20 hours. Dead animals were scored by their continuous absence of swimming movements and pharyngeal pumping. A t-test between two means was used to calculate statistical significance.
Assays were performed as described by . Worms were kept at 20°C on NGM plates (10 animals per plate). Day of hatching was used as the first time point. Dead animals were scored as dead when they refused to move after repeated prodding with a pick. Animals that crawled away from the plate, exploded, or contained internally hatched worms were excluded from the analysis. Life spans were determined in parallel for all strains shown together on graphs. Statistical significance was determined by a log-rank analysis using Prizm software.
C. elegans survival assays were performed as described earlier . To avoid the confounding effects of varying brood sizes, egg laying rates and progeny hatching within the infected worms on worm mortality, we used worms rendered sterile by RNAi of pos-1, loss of which results in inviable embryos , . Worms that died due to desiccation on the walls of the Petri dish or due to bursting vulva were censored from further analysis. Statistical analysis was performed using Kaplan-Meier non-parametric survival analysis using the software Statview (Version 5.0.1 SAS Institute Inc.). P<0.001 was considered significantly different than wild type.
Since the addition of the DAF-16::GFP transgene to the zfp-1(ok554); pdk-1(sa709) double mutant strain led to a penetrant dauer phenotype at 20°C, all DAF-16::GFP strains were maintained at 16°C. L4 and adult stage worms were used for scoring nuclear localization. Worms were mounted on agarose pads and DAF-16::GFP localization was assessed in 10–20 worms at a time using 200X magnification on a Zeiss AxioImager Z1 immediately, higher magnification images of DAF-16::GFP localization in intestinal cells were done at 630X.
Synchronous populations of animals were grown at 20°C on NGM plates seeded with OP50 E. coli at a density of approximately 100,000 animals per 15 cm Petri dish and harvested at specific stages of development. The harvested animals were washed three times with M9 buffer and the pellet was frozen in dry ice with TRI Reagent (MRC, Inc.). After five times of freeze and thaw, total RNA was isolated according to the TRI Reagent protocol. Ten micrograms of the total RNA sample was digested with 2U of Turbo DNase (Ambion) at 37°C for 1hr followed by phenol-extraction and ethanol-precipitation. cDNA was generated from 2 µg of total RNA, using oligo-dT primer and RevertAid Reverse Transcriptase (Fermentas). Quantitative PCR was performed on the Mastercycler ep realplex (Eppendorf) using the QuantiFast SYBR Green PCR Kit (Qiagen). Thermocycling was done for 40 cycles in a two-step cycling, according to the manufacturer's instructions, with 25 µl of reaction containing 12.5 µl SYBR master mix, 0.15 µl of 100 µM primers, 5 µl of diluted cDNA, and 7.2 µl dH2O. Each PCR reaction was performed in triplicate. We used the ΔΔCt method to quantify the change in mRNA expression in the mutant samples compared to wild type and act-3 mRNA was used as a reference gene. The primers used were as follows: Forward CACGAGACTTCTTACAACTCC and Reverse GCATACGATCAGCAATTCCT for act-3 mRNA detection, Forward AGCCATCAACACCGTCTAAC and Reverse CGAATTGGCGCGTGGTGC for pdk-1 mRNA detection, Forward GCTAGGATGTCAGGTGGTC and Reverse CCAAGAGAAGCCACGAAAGC for aqp-1 mRNA detection, Forward ATGCTCGTGCTCTTGCTGAG and Reverse GACTGACCGAATTGTTCTCCAT for gst-4 mRNA detection, Forward TACCGATGAGGAGTGGGAGA and Reverse CGAATTCCCGAGCAAGATAA for gst-38 mRNA detection, Forward TTTCAGAATCACAGAGCAACAC and Reverse TGCGATACATGTTCAGAAGAG for zfp-1 mRNA detection, Forward ACACTATTAAGCGCGACTTCG and Reverse AGTTGGCAATCTTCCAAATAGC for sod-3 mRNA detection, Forward pdk-1 ex2-ex3 junction CCTACAGCCAGGTATTCCG and Reverse pdk-1 intron 3 ACAAGTGGATTTTGATGGGTTC for detecting the mutant sa709 pdk-1 mRNA and pre-mRNA and Reverse pdk-1 ex3-4 junction GATCACGAAATAAATTCTAGCCTGG-for detecting the wild-type pdk-1 mRNA.
For detection of bi-directional transcription at the pdk-1 promoter the primers used were as follows. Region 1 RT primers: detecting (−) strand transcript CCGAGGTTATAATTTTGGCTAAACTT; detecting (+) strand transcript ATCAAGAGATACAGCGGGAG. Region 1 PCR primers:
Region 2 RT primers: detecting (−) strand transcript CTCCCGCTGTATCTCTTGAT
detecting (+) strand transcript GTACGGTTGTTATCGCTTTCAGG.
Region 2 PCR primers: forward - GAATGTTCAAAGCCTTAAAGC
reverse – AGGGATAATTGGAGTGACATGG.
Chromatin immunoprecipitation was performed following the modENCODE Protocol from the Lieb Lab with the following modification: 2.5–3mg of cross-linked extract from L3 or adult worms was incubated for 1h at 4°C with the specific antibody and the immune complexes were then incubated with 60 µl IgG Dynabeads (Invitrogen) for 1h at 4°C. DNA was cleaned up with the Qiagen PCR purification kit. For the FLAG ChIP, we incubated the cross-linked extract with ANTI-FLAG M2 Affinity Gel (Sigma) for 2h at 4°C and, after the washing steps, eluted with 300 µg/ml of FLAG peptide (Sigma) for 30min at 4°C. The other antibodies used were anti-ZFP-1 (generated by the Lieb Lab) and anti-Pol II 8WG16 (Covance).
The immunoprecipitated DNA was quantified by qPCR using the ΔΔCt method to calculate the percentage of immunoprecipitation relative to the input. We used the following specific primers: Forward AAACAACACATAGACTTGTGCC and Reverse GTACGGTTGTTATCGCTTTCAG to amplify the promoter region of the pdk-1 gene; Forward pdk-1 ex2 GCAAGTGAATCGGAGAACAG and Reverse pdk-1 ex2 TGAAGAAACATGAAGTGCTTGG to amplify the coding region of the pdk-1 gene; Forward TTTCAGAACTATCATGCCACG and Reverse TCTCTGAGCACACTTTGAGG to amplify the promoter region of the aqp-1 gene; Forward aqp-1 ex5 TTGCCAGTTATCCATCTCCA and Reverse aqp-1 ex5 CTCTCATCAATAACAACGCAG to amplify the coding region of the aqp-1 gene; Forward TTAGATAGAGAATTGGCGAGAG and Reverse CAAGTAGCAAAGCGATAAACC to amplify the promoter region of the gst-4 gene; Forward gst-4 ex4 TGAAGTTGTTGAACCAGCC and Reverse gst-4 ex4 CCCAAGTCAATGAGTCTCCA to amplify the coding region of the gst-4 gene.
To investigate the function of ZFP-1 with ChIP we first developed an antibody (termed JL00006_ZFP1) to the C-terminal portion of the protein. Alternative transcription start sites give rise to two ZFP-1 protein isoforms with identical C-terminal domains. As expected, both isoforms are recognized by the JL00006_ZFP1 antibody. The protocols used for generating ZFP-1 ChIP/chip data are described at http://www.modencode.org/Lieb.shtml.
C. elegans genes (refSeq id) from genome build CE4 (ws170) were extracted from the UCSC genome browser's refGene table. A gene was called bound by ZFP-1 if the center base pair of a ZFP-1 peak overlapped the ORF or the 1,500 bp upstream region. Overlap calls were done using the Galaxy web tool. Of the total 24,901 genes, 3,598 were bound by ZFP-1. Genome-wide ZFP-1 localization data are available at modENCODE: http://intermine.modencode.org/.
Increase in pdk-1 expression correlates with susceptibility to oxidative stress. (A, B) Survival of L4 larva (n=90) from indicated strains after 20 hour incubation period in 100mM paraquat; ** indicates significance of P<0.01 and * - P<0.05 compared to wild type. (C) RT-qPCR detecting an increase in pdk-1 expression in SP940, results of three biological replicas are shown, error bars represent standard deviation.
RT-qPCR confirming enhanced expression of zfp-1 mRNA in ZFP-1::FLAG and ZFP-1::GFP transgenic strains.
ZFP-1 levels do not change in rde-4(ne299). Western blot analysis with anti-ZFP-1 C-terminal antibody; actin levels and Ponceau S staining of the membrane are shown as loading controls.
An example of a gene with promoter siRNAs matching repetitive elements and expressed higher in rde-4(ne299) according to . A screen shot from the UCSC browser and PolII ChIP results showing enhanced occupancy in rde-4(ne299).
Examples of genes with promoter siRNAs matching repetitive elements and expressed higher in rde-4(ne299) according to . Screen shots from the UCSC browser.
Dataset table allowing identification of overlaps between the following datasets: microarray data listing genes misregulated in rde-4(ne299) and zfp-1(ok554) from , microarray data of Class 1 and Class 2 genes acting downstream of daf-16 from , rde-4-dependent siRNA target genes identified by , and ZFP-1 target genes where the ZFP-1 peak was found in the 1,500bp promoter window identified by ChIP/chip; functional mount and category groups from .
We thank I. Greenwald for critical reading of the manuscript; S. Palani and D. Avgousti for helpful discussions; S. Palani, W. Gu, J. Kim, and N. Welker for help with endo-siRNA bioinformatics; B. Tursun and L. Cochella for technical advice on fosmid recombineering; C. Benard for advice on longevity assays; R. Ruiz for technical assistance; D. Conte and C. Mello for providing mutant strains. The zfp-1(ok554) strain was provided by the C. elegans Gene Knockout Project at OMRF, which is part of the International C. elegans Gene Knockout Consortium.
The authors have declared that no competing interests exist.
This work was supported by 3260-07 Special Fellow Award from The Leukemia and Lymphoma Society, Arnold and Mabel Beckman Foundation Young Investigator Award, and NIH Director's New Innovator Award (1 DP2 OD006412-01) to AG; National Institutes of Health grant GM66269 to M-WT; and modENCODE grant U01 HG004270 to JDL. LMK was supported by NIH training grant 5T32 GM07088-34. Some of the strains used in this study were provided by the Caenorhabditis Genetics Center, which is supported by the National Institutes of Health-National Center for Research Resources. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.