A Two-tiered compensatory response to loss of DNA repair modulates aging and stress response pathways

Oxidative stress response and reduced insulin/IGF-1 signaling in nth-1(ok724)

Since DNA damage responses often show small changes on the transcriptional level [24], we analysed the gene expression signatures using a fold-change cut-off criterion ≥1.8. We found a high number of differentially expressed transcripts between the N2 reference strain and the nth-1 mutant considering the unstressed conditions of the animals: 2074 probe sets were differentially expressed ≥1.8 fold (Supplementary Table 1). The low number of transcripts regulated ≥4-fold (185 probe sets) suggests that there is a focused transcriptomic response to loss of the NTH-1 enzyme.

Gene ontology (GO) enrichment analysis revealed that genes involved in determining adult lifespan (p< 0.007) were enriched among the regulated genes in the nth-1 mutant (Figure 1). Of these, 17 are known to act through the insulin/IGF-1 signaling (ILS) pathway. Reduced signaling through the canonical ILS pathway leads to nuclear localisation of the FOXO transcription factor DAF-16 [25]. A total of 84 genes previously identified as downstream targets of DAF-16 (dod) [26,27], were differentially regulated in nth-1, of which 67 were not assigned to the aging cluster based on present GO annotation. However, some confirmed targets of DAF-16 (e.g. hsf-1, hsp-90, hsp-70) were not differentially regulated, and there was no significant overlap between our dataset and the previously reported daf-16 dataset [27] (data not shown). Moreover dao-6, which is positively regulated by DAF-16 and negatively regulated by DAF-2, was downregulated by 7.7-fold. Thus, the transcriptional changes in the nth-1 mutant appear not to be dominated by DAF-16. The downregulation of ins-1 and ins-7 (2.17 and 3-fold, respectively), two DAF-2 agonists whose expression are repressed by DAF-16, likely reflects negative feedback inhibition of ILS rather than sensory neuronal input to the ILS pathway.

Previous genetic and genomic studies have demonstrated that there is a close interconnection between the ILS and stress-response pathways in C. elegans [8,28]. This is reflected in thenth-1 dataset: The genes in the aging cluster, as well as individual genes regulated more than 4-fold (Supplementary Table 1), indicate that oxidative stress responses are activated. SOD-3 is a mitochondrial Mn-containing superoxide dismutase [29] and increased expression of sod-3 has been reported in response to oxidative stress [30]. sod-3 is a target of DAF-16, and the well established inverse regulation between ins-7 and sod-3 [31] is observed in nth-1 (-3 and 1.84-fold, respectively). Activation of an oxidative stress response in nth-1 is further suggested by the upregulation of gst-4 (2.31-fold), a regulator of SKN-1 which is a transcription factor mediating transcriptional responses to oxidative stress [32]. Regulation of steroid signaling and stress responses are also reflected in the second GO enriched cluster, proteolysis (p < 0.01) (Figure 1).

A search for functional interactions among upregulated genes in the two clusters using the online functional interaction browser FunCoup revealed a close interconnection between the two enriched clusters involving 31 of the 58 regulated genes (12 and 19 from cluster I and II, respectively) (Figure 2A). The expression of the CeTOR (let-363) kinase is upregulated in nth-1 (2.33-fold), possibly indicating activation of a survival response to stress. The TOR pathway controls protein homeostasis and contributes to longevity, and the network analysis indicates that TOR might connect the two clusters via the AAA+ ATPase homolog RUVB-1, a component of the TOR pathway [33]. Direct protein-protein interactions involving RUVB-1 have been demonstrated with several upregulated genes in both enriched GO clusters.

The protein-interaction network (Figure 2A) suggests that the transcriptional changes may also involve regulation of the redundant activities of the conserved p38 and JNK stress-activated protein kinase pathways: MFB-1, for example, directly interacts with SEK-1, a MAPK kinase required for germline stress-induced cell death independent of the CEP-1 (C. elegans p53) DNA damage response [34]. SEK-1 is also required for nuclear localisation of DAF-16 in response to oxidative stress [35]. It was suggested that oxidative stress mediates regulation of DAF-16 through activating the p38 signal transduction pathway upstream of DAF-16. Therefore, regulation of DAF-16 target genes in nth-1 is consistent with activation of an oxidative stress response. Alternatively, the regulation of DAF-16 targets could be secondary to aqp-1 upregulation. Aquaporin-1, a glycerol channel protein, was recently demonstrated to modulate expression of DAF-16-regulated genes and suggested to act as a feedback regulator in the ILS pathway [36]. There is a strong upregulation of aqp-1 in nth-1 (31-fold). Moreover, 6 out of 7 genes negatively regulated by AQP-1 are repressed in nth-1 (Supplementary Table 2).

Network analysis of the 45 downregulated genes resulted in a network involving 71 genes from both clusters (Figure 2B). The pronounced downregulation of genes specifically responding to exogenous oxidative and heat stress (such as the hsp-16 family, ftn-1 and gst-10,lys-7, mtl-1) and anti-microbial immunity (several c-lectins, cpr-2, ilys-3, abf-2, cnc-7) in the nth-1 mutant suggests that a specific response to endogenous stressors is triggered. Hence, loss of BER in C. elegans appears to induce transcriptional responses involving similar pathways as those regulated in mammalian NER mutants [19].

Shared transcriptional responses in nth-1 and xpa-1 mutants

To experimentally validate whether there is similarity between the transcriptional programs associated with loss of BER and NER capacity in C. elegans, we collected the expression profile of the xpa-1(ok698) mutant. In xpa-1, we identified 2815 differentially expressed transcripts having a fold-change of ≥1.8 (Supplementary Table 3).

GO enrichment analysis revealed four significantly regulated clusters in xpa-1 (Figure 3). The GO process determination of adult lifespan (p < 0.0007) was shared with nth-1 (Figure 1), and 67% (28 out of 42) of the individual genes in this cluster in nth-1 were shared with xpa-1. In xpa-1, genes that respond to oxidative stress and redox homeostasis are not only represented in the "aging" cluster, but are also found in the clusters containing genes involved in the ER unfolded protein response (p < 0.05) and regulation of carboxylic acid metabolism (p < 0.03). Network analysis of the 57 downregulated genes within the enriched clusters resulted in a network resembling that of nth-1 (Supplementary Figure 1). Thus, qualitatively similar responses were activated to compensate for the loss of NTH-1 or XPA-1. Regression analysis using the 1.8-fold-change data confirmed the similarity of the nth-1 and expression profiles (R₂ = 0.96). However, there is stronger modulation of gene expression in xpa-1 compared to nth-1, with an increased number of transcripts with higher fold-change; e.g. expression of ins-7 (8.8-and 3-fold), aqp-1 (34.8- and 31-fold, respectively), and hsp-16.49 (-15.99 and -5.58- fold) in xpa-1 and nth-1, respectively (Supplementary Table 1 and 3).

Somatic preservation in nth-1;xpa-1

Genes regulating adult lifespan were not among the four enriched GO processes identified from the 2787 regulated probe sets with fold-change ≥1.8 (1225 up and 1562 down) in nth-1;xpa-1 (Figure 4 and Supplementary Table 4). Instead, the transcriptional response was dominated by genes involved in cell-cycle regulation (clusters I-III) and DNA repair (cluster IV). Cluster I (p< 0.003) and II (p < 0.01) contain genes that function in mitosis-related processes such as chromosome segregation, mitotic spindle assembly and stability, and replication licensing. Only 2 out of 36 genes in cluster I are upregulated. In contrast, 15 of the 64 genes present in cluster II are upregulated and most encode histone genes. Cluster III (p < 0.00001) share many genes with cluster I and II but reflect regulation of progression through meiosis.

Genes involved in DNA repair and DNA damage checkpoint pathways are enriched in Cluster IV (p <0.02). Naively, it could be expected that the double mutant would compensate for loss of integrity of two DNA repair pathways by upregulating alternative DNA repair modes. However, the opposite seems to be the case. Several mismatch repair, homologous recombination (HR), non-homologous end-joining (NHEJ), and DNA damage checkpoint genes, such as the C. elegans homolog of BRCA1 (brc-1) and its associated proteins, brd-1 and dog-1, are downregulated (Table 1). Many uncharacterized genes that have previously been identified in screens for genes that result in mutator phenotypes when depleted by RNAi [37,38] were also suppressed in the nth-1;xpa-1 mutant. Network analyses illustrate close interrelation of clusters I through IV also on protein level returning protein-protein interactions between 157 of the 212 genes in all clusters (data not shown).

Suppression of the Aurora-B kinase and Polo-like kinase 1 regulatory network in nth-1;xpa-1

The GO analysis suggests that the double mutant differs from either single mutant. Linear regression analysis comparing the overlapping transcripts in the ≥1.8-fold-change lists from nth-1;xpa-1 and xpa-1 confirmed this difference (R2 = 0.12) whereas the single mutants show significantly stronger correlation (R₂ = 0.94). Principal Component Analysis (PCA) on the entire dataset (Figure 5A) confirms that the overall expression profiles of the single mutants cluster together and therefore resemble each other but, although nth-1;xpa-1 clusters separately from the wild type, it seems to be in closer proximity to it than to either single mutant. Hierarchical clustering confirmed the closer relationship between nth-1;xpa-1 and the wild type (Supplementary Figure 2). Hierarchical clustering of the mutants only revealed even more clearly that the single mutants are more similar to each other than either are to the double mutant. Several transcripts have opposite regulation, most notably in xpa-1 and nth-1;xpa-1 (Figure 5B). DNA repair and DNA damage response genes are prominent among the genes regulated in an opposite direction (a selection is presented in Table 1).

Polo-like kinase 1 (PLK-1), which is upregulated in xpa-1 (1.97-fold) but repressed in nth-1;xpa-1 (-2.11-fold), has emerged as an important modulator of DNA damage checkpoints [39,40]. Moreover, Aurora B kinase (air-2) is downregulated (-1.83-fold), and an inhibitor of AIR-2 activation, gsp-2, is one of the few upregulated genes in nth-1;xpa-1. Several other components of AIR-2 and PLK-1 networks are represented in the enriched GO clusters in the double mutant (Figure 4). Moreover, the transcriptional changes observed in nth-1;xpa-1 involved several genes that are validated interactors of AIR-2 and PLK-1. The direction of the expression changes suggests that there is a concerted response that suppresses AIR-2 and PLK-1 signaling networks in the double mutant (Figure 6) that are consistent with published literature evidence: Plk-1 stimulates the activation of Cdk-1, several cyclin B proteins and a G2/M specific cyclin A through Cdc-25.1 [40]. The inner centromere protein (INCENP), ICP-1, coordinates cytokinesis and mitotic processes in the cell and integrates the PLK-1 and AIR-2 signaling at kinetochores. AIR-2 and PLK-1 regulate mitosis and cytokinesis through CYK-4 and ZEN-4 [41]. Downregulation of MCM2-7 could prevent firing of dormant replication origins which are often used when the transcriptional machinery is blocked or otherwise impaired [42]. Hence, the suppression of DNA metabolism suggested by the transcriptional signature reflects a concerted response. In summary, there seems to be a two-tiered compensatory response to loss of DNA repair in C. elegans: While lack of either BER or NER results in activation of genes responding to endogenous stressors and suppression of ILS, lack of both BER and NER shifts the transcriptional response to reduction of proliferation and somatic preservation through modulation of AIR-2 and PLK-1 signaling networks.

Depletion of NTH-1 and XPA-1 induces oxidative stress responses

Transcriptomic profiling strongly indicates that the nth-1 and xpa-1 mutants experience oxidative stress. To experimentally validate whether loss of NTH-1 and XPA-1 induces oxidative stress, we took advantage of the established reporter strain CL2166, which expresses green fluorescent protein (GFP) under the control of the glutathione-S-transferase GST-4 promoter [32]. gst-4 expression is upregulated in both nth-1 and in xpa-1 (2.31, and 2.01-fold respectively). Whereas GFP is normally expressed in hypodermal muscle, GFP-fluorescence increases in the body wall muscles and translocates to the intestinal nuclei upon oxidative stress As expected, paraquat, which generates superoxide in vivo, increases the average number of GFP positive intestinal nuclei up to 46 compared to 15 in untreated animals (p < 0.001). Depletion of NTH-1 or XPA-1 by RNAi significantly increased the number of foci to 26 and 25, respectively (p < 0.001) (Figure 7), thus demonstrating that even transient depletion of NTH-1 or XPA-1 induces oxidative stress responses. Codepletion of NTH-1 and XPA-1 did not increase the number of intestinal GFP-positive foci. The gst-4::GFP reporter assay therefore experimentally validated the high throughput genomic results and confirmed that loss of NTH-1 or XPA-1, but not both, leads to oxidative stress and activation of oxidative stress responses.

The transcriptional changes do not protect against exogenous acute stress

The expression profiling indicated that the transcriptional responses in the single-mutants are aimed at compensating for oxidative stress resulting from DNA-repair deficiency. The responses appear to be selectively tuned to compensate for endogenous stress. The down-regulation of other stress induced factors, such as the hsp-16 family, may serve to prevent unsolicited activation of a full-blown stress response. Thus, we would not expect the DNA repair mutants to show resistance to oxidizing agents which is correlated with reduced ILS in C. elegans [8,30]. In agreement with previous reports, neither xpa-1 [43], nth-1 [7] nor nth-1;xpa-1 were hypersensitive to paraquat (data not shown). However, all mutants showed mild sensitivity to an acute exposure to another superoxide generating agent, juglone (Figure 8A) and mild heat-shock (data not shown). Hence, the upregulation of oxidative stress responses do not confer resistance to acute exogenous stress. These phenotypes are consistent with downregulation of genes responding to exogenous stressors as observed.

Loss of NTH-1 restores normal lifespan in xpa-1

Reduced ILS induces longevity in C. elegans [44], but reduced ILS is also seen in segmental progeroid NER defective mice [21]. This apparent paradox can be interpreted as the reduced ILS in the DNA repair defective mice is part of a compensatory attempt to extend lifespan in organisms suffering from DNA damage associated stress. Thus, we were interested to test whether the reduced ILS in nth-1 and xpa-1 observed here was accompanied by reduced lifespan - or whether the compensatory response was sufficient to sustain normal lifespan. The lifespan of nth-1 was indistinguishable from the wild type, as was recently shown [7], whereas the xpa-1 mutant displayed reduced lifespan compared to the wild type (mean survival of 14.5 and 17.3 days, respectively) (Figure 8B). In C. elegans therefore, as in mice, the challenges that loss of NER poses to the organism is more severe than loss of a single DNA-glycosylase. Our results demonstrate that this difference in challenge can be read out as a stronger activation of the antioxidant defense and reduction in ILS.

The nth-1;xpa-1 double mutant has a more profound DNA repair defect and is expected to be unable to repair a much wider spectrum of DNA lesions. If the accumulation of DNA damage itself is the bigger lifespan reducing challenge in xpa-1, we would expect nth-1;xpa-1to be more severely affected. Interestingly, normal lifespan was restored in nth-1;xpa-1 with a mean survival of 17.4 days. One possible interpretation of these results is that the oxidative lesions most relevant for aging are those that are normally repaired by NER, but are attempted processed by BER in the absence of the preferred repair pathway.

Oxidative stress induced in DNA repair mutants

Here we present the first comprehensive report describing compensatory transcriptional responses to loss of base excision repair genes in animals. Using a well established transgenic reporter assay, we show that transient depletion of NTH-1 and XPA-1 by RNAi induces oxidative stress, thus it seems likely that this initiates transcriptome-modulation in the mutants. We show that lack of the NTH-1 and XPA-1 enzymes are accompanied by upregulation of oxidative stress responses tuned towards endogenous stressors. This is in agreement with identification of a focused compensatory response to BER intermediates (AP-sites and strand breaks) previously shown in S.cerevisiae, where the transcriptional responses differed from the common environmental stress response or the DNA damage signature [45]. A DNA-damage dependent ROS response to unrepaired oxidative DNA damage was previously demonstrated in S. cerevisiae BER and NER mutants [46]. Interestingly, no indication of oxidative stress or increased expression of oxidative stress response genes was observed in a mutant lacking both NTH-1 and XPA-1. This transcriptomic shift argues against a DNA-base damage dependent activation of oxidative stress responses, but instead indicates that the DNA repair enzymes mediate signaling to activate stress response pathways. Although the upstream signaling events in the nth-1;xpa-1 double mutant remain to be elucidated, the modulation of AIR-2 and PLK-1 interaction networks may be a consequence of absence of DNA repair enzyme-mediated signaling of transcription blocking lesions. Alternatively, the extensive new synthesis of histone genes suggests that signaling involves chromatin dynamics in the absence of the global genome damage binding proteins, NTH-1 and XPA-1.

The biological significance of transcriptome modulation seen here is confirmed by the DNA repair mutants showing a mild sensitivity to oxidizing agents. It seems likely that the downregulation of the hsp-16 family and ftn-1 contributes to the higher sensitivity to juglone in xpa-1 and nth-1, particularly as it is unlikely that a short acute exposure to oxidizing agents (or heat-stress) may lead to DNA-damage mediated toxicity on organismal level.

Conserved compensatory responses to BER and NER deficiency

Few systematic studies have been performed to look at transcriptomic changes in BER mutant animals and none, to the best of our knowledge, have compared mutants in BER and NER.

A study on gene expression profiling in BER- or NER-defective mutant S. cerevisiae showed transcriptional changes in mutants defective in both pathways after treatment with hydrogen peroxide [22], but not in the double mutant. Instead, transcriptome changes in the BER/NER defective strain were already elicited from unrepaired spontaneous DNA damage [23]. To test the generality of our finding, we re-analyzed the baseline data sets from S. cerevisiae BER (Ntg1, Ntg2, and Apn1 deficient), NER (Rad1 deficient) and BER/NER defective mutants. We performed GO enrichment analysis on expressed transcripts in the individual strains (Supplementary Table 5). This analysis showed that BER-defective cells had few expressed transcripts and only one enriched GO process, DNA replication (p< 0.05). Informative enriched GO processes found only in the NER defective strain include transcription regulation, ubiquitin dependent protein degradation, sister chromatid segregation, and cell communication (p < 0.01). The BER/NER and NER defective cells share many enriched GO clusters and show 38% overlap of individual expressed transcripts. Enriched GO processes found only in the BER/NER mutant include DNA repair, DNA packaging, response to DNA damage stimulus, and cell-cycle checkpoint. Regulation of RNA polymerase II transcription was not enriched in the BER/NER mutant. Therefore, the main conclusions drawn here from BER, NER and BER/NER deficient C. elegans, resemble those previously seen in S. cerevisiae [22,23,45]: i) BER mutants show transcriptomic changes. ii) Loss of NER induces more substantial transcriptional responses than BER involving modulation of RNA metabolism and regulation of transcription iii) Loss of BER in the NER mutant shifts the response to regulate processes that maintain DNA integrity.

Reduced mean lifespan in xpa-1(ok698)

The first xpa-1 mutant identified,rad-3(mn159), was reported to have a near-normal lifespan [15] but there are conflicting reports on the lifespan of xpa-1(ok 698) allele ranging from normal [43] to a maximum lifespan of 15 days compared to 25 days in the wild type [17]. Here, we show a moderate reduction of lifespan in xpa-1. We did therefore not anticipate that the XPA-1 mutant would display a transcriptional profile resembling that of the segmental progeroid NER defective mice [19]. Nevertheless, the reduced lifespan is entirely consistent with the transcriptional changes observed. A recent transcriptomic signature of Xpa^-/- mouse dermal fibroblasts shows that suppression of ILS and activation of oxidative stress responses is also seen in DNA repair mutants that do not exhibit accelerated aging [47]. In support of this, GO enrichment analysis, performed as part of the present study, on the differentially expressed genes in Xpa^-/- mice [21] showed enrichment of genes that regulate lifespan. Hence, the transcriptomic changes in NER mutants are conserved.

However, the consequences of loss of XPA-1 appear more severe in C. elegans compared to mice, both with respect to transcriptional regulation and lifespan. This might indicate that C. elegans XPA-1 contributes to repair of spontaneous DNA damage. The 13-fold elevated mutation accumulation rate in xpa-1 compared to 7-fold in nth-1 [9], supports this possibility.