Loss of circRNAs from the crh‐1 gene extends the mean lifespan in Caenorhabditis elegans

Abstract Accumulation of circular RNAs (circRNAs) during aging occurs on a genome‐wide level for multiple organisms, but its significance is unknown. Generating circRNA loss‐of‐function mutants is difficult because the vast majority of these RNAs are comprised of exons shared with protein‐coding mRNAs. In Caenorhabditis elegans, most circRNAs were previously found to accumulate during aging. Two of the most abundant, age‐accumulating circRNAs are generated from exon 4 of the crh‐1 gene (circ‐crh‐1). Here, we found that the biogenesis of circ‐crh‐1 was regulated by the double‐stranded RNA‐binding protein ADR‐1. We identified Reverse Complementary Match (RCM) sequences in introns flanking circ‐crh‐1. Using CRISPR‐Cas9, we deleted the downstream RCM and found that this completely eliminated expression of the circRNA without affecting linear mRNA expression from the crh‐1 gene. Remarkably, worms lacking circ‐crh‐1 exhibited a significantly longer mean lifespan. Lifespan was partially restored to wild type by expression of circ‐crh‐1 in neural tissues. Widespread transcriptome alterations in circ‐crh‐1 mutants were identified using RNA‐Seq. Moving forward, intronic RCM deletion using CRISPR should be a widely applicable method to identify lifespan‐regulating circRNAs in C. elegans.

also accumulate during aging and that several age-accumulated cir-cRNAs are generated from genes with known roles in lifespan regulation (Cortes-Lopez et al., 2018). One of these genes, crh-1, encodes an ortholog of the cAMP Response Element-Binding Protein (CREB) which plays a role in longevity (Chen et al., 2016;Lakhina et al., 2015). We previously reported that the two abundant circRNAs generated from the crh-1 gene greatly increase in abundance during aging (i.e., cel_circ_0000438 and cel_circ_0000439) (Cortes-Lopez et al., 2018). These circRNAs differ only in six nucleotides as a result of an alternative splice acceptor site and are collectively referred to from here on as circ-crh-1 (Figure 1a).
For certain circRNAs, RCMs within flanking introns facilitate backsplicing ostensibly by bringing the splice donor and acceptor sites into closer proximity (Jeck et al., 2013). The two crh-1 circRNAs generated from exon 4 of the crh-1 gene are flanked by long introns that contain sequences complementary to one another (RCM-L and RCM-R) (Figure 1a, Figure S1a). ADAR is a double-stranded RNAbinding protein that when knocked down increases the expression of some circRNAs in mammalian cells Rybak-Wolf et al., 2015). We investigated crh-1 expression in two independent ADAR null mutant alleles, adr-1(gv6) and adr-1(tm668) (Hundley et al., 2008;Tonkin et al., 2002). RT-qPCR analysis of whole worms showed that circ-crh-1 expression was significantly increased in both adr-1 mutants, whereas linear crh-1 mRNA was unchanged ( Figure 1b). These data demonstrate that ADR-1 negatively regulates circ-crh-1, most likely through interacting with intronic RCMs. F I G U R E 1 circ-crh-1 regulation by ADR-1 and generation of CRISPR/Cas9 deletion alleles. (a) Schematic showing exon 3 to exon 5 of crh-1 gene (chrIII:11685086-11691812). Two circRNAs are generated by backsplicing of exon 4, using two alternative splice acceptors (SA) and one shared splice donor (SD). Reverse complementary matches (RCM-L and RCM-R) predicted to facilitate backsplicing of crh-1 circRNAs are shown as blue rectangles. (b) Mutations in adr-1 result in increased expression of circ-crh-1 but not linear crh-1 in day-1 adult worms, as determined by RT-qPCR. Linear adr-1 transcripts are not expressed in adr-1(gv6) or adr-1(tm668) mutant alleles as expected. n = 2 independent biological samples. (c) Schematic of crh-1 exon 4 and flanking intronic sequences from either wild-type (WT) or crh-1 circRNA mutant genotypes. Intronic deletions targeting the downstream RCM-R region were introduced by CRISPR/Cas9 and are presented as red rectangles. (d) Northern analysis of day-1 adult whole worms using dsDNA probe complementary to crh-1 exon 4. Signal from crh-1 circRNAs is absent in crh-1(syb385) mutant worms compared to wild type. (e) RT-qPCR expression analysis of linear and circular crh-1 transcripts in day-7 adult crh-1 circRNA mutants compared to wild-type worms. Both circRNAs are significantly reduced compared to wild type, whereas the crh-1 linear RNA is unchanged. n = 3 independent biological samples. For RT-qPCR expression analyses, data were normalized to cdc42 mRNA and are represented as mean ± SEM; n.s., not significant; *p < 0.05; **p < 0.01; ***p < 0.001. Strains used in this study are found in Table S1. RT-qPCR and northern blot primers can be found in Table S2. Also see Figure S1  Generating circRNA loss-of-function organisms is challenging because circRNAs are derived from protein-coding genes, and thus attempts to disrupt circRNA expression can interfere with the biogenesis of protein-coding transcripts. Recently, CRISPR/Cas9 was used to delete RCMs in the introns flanking mouse circKcnt2 which abolished the circRNA and did not affect the linear RNA (Liu et al., 2020). We similarly used CRISPR/Cas9 to generate a 377 bp deletion overlapping the RCM-R region to generate the crh-1(syb385) mutant strain (Figure 1c). To confirm the loss of circ-crh-1, we performed northern blot analysis and found circ-crh-1 to be undetectable, whereas linear crh-1 expression persisted ( Figure 1d). By RT-qPCR analysis, circ-crh-1 expression was barely detectable in syb385 mutants in both 1-and 7-day-old adult worms ( Figure 1e and Figure S1b). Importantly, we found that expression of linear crh-1 was not significantly affected ( Figure 1e and Figure S1b).
In addition, we generated adr-1(gv6); crh-1(syb385) double mutants and examined whether suppression of circ-crh-1 by ADR-1 depends on RCM base-pairing. We found that in contrast to adr-1(gv6) mutants, circ-crh-1 expression in these double mutants remained unchanged compared to crh-1(syb385) (Figure 1b and Figure S1b). Together, this suggests that RCMs are likely required for ADR-1 regulation of circ-crh-1 biogenesis. Next, we generated a second independent allele, crh-1(syb2657), which was designed to delete a slightly larger portion of the intronic region surrounding the RCM-R sequence (Figure 1c), and confirmed by RT-qPCR that only the circle and not mRNA from the crh-1 gene was altered ( Figure 1e and Figure S1c). These two alleles, thus, represent genuine circRNA-specific loss-of-function mutants.
To directly test whether circ-crh-1 plays a role in aging, we per- Remarkably, the removal of just circ-crh-1 caused a significant extension of mean lifespan compared to wild-type controls for both alleles (syb385, 11.5% extension, p < 0.0001; syb2657, 9.8% extension, Table S3). This suggests that circ-crh-1 might contribute to age-related decline. Expression of circ-crh-1 under control of the germline promoter pie-1 or the pan-neural promoter rab-3 restored circ-crh-1 expression levels in syb385 mutants ( Figure   S2a,b,d). Expression of circ-crh-1 driven by rab-3, but not pie-1, was able to partially rescue the lifespan phenotype observed in syb385 mutants (Figure 2b, Figure S2c and Table S3). These results indicate that circ-crh-1 expression in neurons is an important determinant of C. elegans lifespan.
To demonstrate that lifespan extension observed in crh-1 cir-cRNA mutants was independent of CRH-1 protein expression, we measured phosphorylated CRH-1 protein (p-CREB) via western blot.
As expected, p-CREB was undetectable in null crh-1(n3315) and crh-1(tz2) mRNA mutants (Figure 2c). p-CREB expression was not altered in circ-crh-1 mutants compared to wild type (Figure 2c). Similarly, we observed that the expression of circ-crh-1 was unchanged in the p-CREB protein null condition ( Figure S2e). Together, these data suggest that p-CREB is not regulated by circ-crh-1 and likewise, CREB does not regulate circ-crh-1 biogenesis. However, we cannot eliminate the possibility that circ-crh-1 and CREB might regulate one another at particular timepoints or in specific tissues not examined here.
As previously reported (Chen et al., 2016;Templeman et al., 2020), p-CREB mRNA mutants (n3315 and tz2) are short-lived, whereas we observed a longer mean lifespan for circ-crh-1 mutants ( Figure 2a). In order to identify transcriptomic differences that might contribute to the opposing lifespan phenotypes in p-CREB mRNA mutants versus circ-crh-1 mutants, we turned to RNA-Seq analysis.
We sequenced 1-day-old adult worms from crh-1(syb385) and crh-1(n3315) mutant backgrounds (Table S4) and identified hundreds of differentially expressed genes (DEGs) when compared to wild-type controls (Figure 2d and Table S5). To our surprise, we discovered a significant overlap (p < 0.0001; Fisher's exact test) for both upregu-  Table S6). Upregulation of collagens is a shared feature found in several different long-lived mutants (Ewald et al., 2015). While this finding is supportive of the increased lifespan observed in crh-1(syb385) mutants described here, crh-1(n3315) mutants have a reduced lifespan. Hence, other differentially expressed genes unique to each mutant genotype (e.g., non-overlapping DEGs) are likely responsible for determining C. elegans overall lifespan.
CircRNAs generally increase in expression during aging, which has been attributed to their high degree of stability, especially in post-mitotic tissues such as neurons (Knupp & Miura, 2018). We have speculated previously that the age accumulation of circRNAs could be detrimental to cellular function due to the progressive nature of the age accumulation (Knupp & Miura, 2018). Given the large number of circRNAs increased with aging, perhaps such a detrimental effect is independent of the specific identity of the individual circRNAs. Despite our hypothesis that circRNA accumulation could decrease lifespan, it nonetheless was quite surprising to find that loss of a single circRNA can increase the mean lifespan. Which of the other hundreds of age-accumulated circRNAs might impact lifespan?
Given the enrichment of intronic RCMs near C. elegans circRNA loci (Cortes-Lopez et al., 2018), the relative ease of CRISPR gene editing, and the general utility of C. elegans in aging research, screening additional age-associated circRNAs appears to be a logical next step. Some age-accumulated circRNAs might have beneficial roles in aging cells. Recently, circSfl transgenic overexpression was found to extend lifespan in Drosophila (Weigelt et al., 2020). However, siRNA knockdown of circSfl was unsuccessful, and genetic manipulations performed to reduce the circRNA also impacted linear mRNA isoforms. This highlights the difficulties in generating circRNA lossof-function mutants to study their impact on aging. In vivo siRNA knockdown has been successfully implemented for certain circRNAs in Drosophila (Pamudurti et al., 2020), but some circRNAs simply cannot be specifically targeted due to sequence limitations of the back-spliced junction region. C. elegans have generally shorter introns than Drosophila and mammalian model systems. This might make them more amenable to efficient CRISPR manipulation of RCMs; however, it remains to be seen whether many C. elegans cir-cRNAs can be specifically reduced or abolished using the RCM deletion methodology.

| ME THODS
A full description of methods can be found in the Supporting Information. Briefly, C. elegans were cultivated on the surface of NGM agar seeded with the Escherichia coli strain OP50 and grown in 20°C incubators using standard protocols. CRISPR deletion mutants were generated by Suny Biotech. Lifespan analysis was carried out with synchronized adult N2 or outcrossed mutant worms with or without transgenes at 20°C. Lifespan curves were analyzed using OASIS2 (Han et al., 2016). Raw RNA-Seq reads are deposited at GEO (GSE190124). RNA-Seq alignment was performed using STAR v2.7.5a (Dobin et al., 2013) and differential expression was performed using DESeq2 (Love et al., 2014).

ACK N OWLED G M ENTS
We thank members of the van der Linden and Miura labs for use-

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests. AVDL and PM supervised the study.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the