group A and B converge on -linked resolution of non-B DNA

Morten Scheibye-Knudsena,b,1, Anne Tsengb, Martin Borch Jensenc, Karsten Scheibye-Alsinga, Evandro Fei Fangb, Teruaki Iyamab, Sanjay Kumar Bhartib, Krisztina Marosid, Lynn Froetscherb, Henok Kassahune, David Mark Eckleyg, Robert W. Maulh, Paul Bastiang, Supriyo Deg, Soumita Ghoshi, Hilde Nilsene,f, Ilya G. Goldbergg, Mark P. Mattsond, David M. Wilson IIIb, Robert M. Brosh Jr.b, Myriam Gorospeg, and Vilhelm A. Bohrb,1

aCenter for Healthy Aging, Department of Cellular and Molecular Medicine, University of Copenhagen, 2300 Copenhagen, Denmark; bLaboratory of Molecular Gerontology, National Institute on Aging, National Institutes of Health, Baltimore, MD 21224; cBuck Institute for Research on Aging, Novato, CA 94945; dLaboratory of Neuroscience, National Institute on Aging, National Institutes of Health, Baltimore, MD 21224; eInstitute of Clinical Medicine, 0167 Lorenskog, Norway; fUniversity of Oslo and Akershus University Hospital, 0167 Lorenskog, Norway; gLaboratory of Genetics and Genomics, National Institute on Aging, National Institutes of Health, Baltimore, MD 21224; hLaboratory of Molecular Biology and Immunology, National Institute on Aging, National Institutes of Health, Baltimore, MD 21224; and iLaboratory of Clinical Investigation, National Institute on Aging, National Institutes of Health, Baltimore, MD 21224

Edited by Philip C. Hanawalt, Stanford University, Stanford, CA, and approved September 19, 2016 (received for review June 23, 2016) Cockayne syndrome is a neurodegenerative accelerated aging disor- to melt G4 structures in a process that involves annealing the two der caused by in the CSA or CSB . Although the DNA strands. Furthermore, stabilization of G4 structures leads to pathogenesis of Cockayne syndrome has remained elusive, recent PARP1 activation and accelerated aging in Caenorhabditis elegans, + work implicates mitochondrial dysfunction in the disease progression. which can be rescued by replenishment of NAD . Combined, our Here, we present evidence that loss of CSA or CSB in a neuroblastoma findings suggest a role for CS proteins in transcription and may explain cell line converges on mitochondrial dysfunction caused by defects why PARP1 might be activated in this premature aging disorder. in ribosomal DNA transcription and activation of the DNA damage sensor poly-ADP ribose polymerase 1 (PARP1). Indeed, inhibition of Loss of CSA or CSB Leads to Mitochondrial Dysfunction ribosomal DNA transcription leads to mitochondrial dysfunction in a To understand why a similar pathology is present in both com- number of cell lines. Furthermore, machine-learning algorithms pre- plementation groups of CS, CSA and CSB, we generated stable dict that diseases with defects in ribosomal DNA (rDNA) transcription knockdown of CSA or CSB in the neuroblastoma cell line SH- have mitochondrial dysfunction, and, accordingly, this is found when SY5Y using lentivirus (Fig. S1 A and B). In accordance with our factors involved in rDNA transcription are knocked down. Mechanis- findings for CSB (8), loss of CSA or CSB led to activation of tically, loss of CSA or CSB leads to polymerase stalling at non-B DNA PARP1 and mitochondrial changes consisting of increased mito- in a neuroblastoma cell line, in particular at G-quadruplex structures, chondrial membrane potential, superoxide production, and in- and recombinant CSB can melt G-quadruplex structures. Indeed, creased oxygen consumption rates (Fig. 1 A–D and Fig. S1C). stabilization of G-quadruplex structures activates PARP1 and leads Increased oxygen consumption rates were also found in C. elegans to accelerated aging in Caenorhabditis elegans. In conclusion, this worms defective in csa-1 or csb-1 (Fig. 1E). Notably, microarray work supports a role for impaired ribosomal DNA transcription in analysis showed very similar expression changes when either Cockayne syndrome and suggests that transcription-coupled resolu- CSA or CSB were deficient (Fig. 1F and Dataset S1). Parametric tion of secondary structures may be a mechanism to repress spurious Analysis of Gene Set Enrichment showed significant up-regulation activation of a DNA damage response. of (GO) terms describing genes involved in ribo- some biogenesis and translation as well as in mitochondrial ATP Cockayne syndrome | aging | polymerase I transcription | nucleolus | CSA | production, likely reflecting the observed functional mitochondrial CSB changes (Fig. 1F). Down-regulated ontology terms included ubiq- uitin pathways and transcriptional regulation (Fig. 1F). ockayne syndrome (CS) is an early onset accelerated aging Cdisorder characterized by growth retardation, progressive Significance neurodegeneration, and typically death in the second decade of life (1, 2). The disease is caused by mutations in either the ERCC8 In this paper we describe a possible pathogenesis for the (CSA) or ERCC6 (CSB) genes, which encode proteins that are accelerated aging disease Cockayne syndrome that entails thought to be involved in DNA repair, transcription, and chromatin defective transcription through DNA secondary structures remodeling (3–5). Recent work has demonstrated mitochondrial leadingtoactivationoftheDNA damage response enzyme dysfunction in cell and animal models deficient in CSB (6–10). The poly-ADP-ribose polymerase 1 and downstream mitochondrial disease mechanism may involve persistent activation of a DNA derangement. These findings are important because they signify damage response orchestrated by the enzyme poly-ADP ribose- a possible new role of transcription in the resolution of DNA + polymerase 1 (PARP1) leading to loss of NAD and increased structures that form spontaneously and suggest a possible lactate production (8, 11), features that are also observed in normal pathogenesis for this accelerated aging disease. aging (12–14). However, it is unclear why PARP1 is activated upon loss of CSB or in normal aging and whether persistent PARP1 Author contributions: M.S.-K., M.B.J., H.N., M.P.M., D.M.W., R.M.B., M.G., and V.A.B. designed activation also occurs in CSA deficiency. Our results support the research; M.S.-K., A.T., M.B.J., K.S.-A., E.F.F., T.I., S.K.B., K.M., L.F., H.K., D.M.E., R.W.M., P.B., notion that the mitochondrial changes observed in CSA or CSB S.D., S.G., and I.G.G. performed research; M.S.-K., T.I., and V.A.B. contributed new reagents/ analytic tools; M.S.-K., A.T., K.S.-A., E.F.F., T.I., S.K.B., K.M., L.F., H.K., D.M.E., R.W.M., P.B., S.D., deficiency are caused by stalled ribosomal DNA (rDNA) tran- S.G., H.N., I.G.G., R.M.B., and M.G. analyzed data; and M.S.-K., L.F., M.P.M., D.M.W., and V.A.B. scription and subsequent PARP1 activation. Accordingly, we find wrote the paper. that diseases with defects in rDNA transcription have dysfunctional The authors declare no conflict of interest. mitochondria whereas classic ribosomopathies do not appear to This article is a PNAS Direct Submission. have this defect. rDNA is prone to form secondary structures, such 1To whom correspondence may be addressed. Email: mscheibye@sund..dk or vbohr@ as G-quadruplexes (G4), and loss of CSA or CSB leads to tran- nih.gov. scriptional pausing at these DNA conformations. G4-induced This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. transcriptional stalling is also found in vivo in humans. CSB is able 1073/pnas.1610198113/-/DCSupplemental.

12502–12507 | PNAS | November 1, 2016 | vol. 113 | no. 44 www.pnas.org/cgi/doi/10.1073/pnas.1610198113 Downloaded by guest on September 30, 2021 these data suggest that CSA and CSB deficiency leads to decreased rDNA transcription. To address this hypothesis, rDNA transcription was measured and found to be lower in short hairpin (sh)CSA and shCSB cell lines, as demonstrated by the loss of 47S pre-rRNA by RT-qPCR as well as by loss of nucleolar ethynyl uridine (EU) incorporation (Fig. 1 H and I and Fig. S1D). Loss of rDNA transcription was also found in fibroblasts from CSA- and from CSB-deficient patients compared with isogenic controls transfected with wild-type (WT) CSA or CSB (Fig. S1 E–G). Because microarray analysis showed up-regulation of genes involved in translation and mitochondrial pathways, we postulate that this outcome reflects compensation due to loss of rDNA transcription. Indeed, inhibition of rDNA transcription using CX5461 led to a significant increase in mito- chondrial membrane potential, content (by mitotracker green staining), and superoxide production (Fig. 1J and Fig. S1 H and I). Furthermore, there was no additive effect between loss of CSA or CSB and the CX5461 treatment, suggesting that the mitochondrial signaling comes from the same pathway (Fig. 1J and Fig. S1 H and I). In summary, loss of rDNA transcription may be driving mitochondrial changes in CS. Loss of rDNA Transcription Leads to Mitochondrial Dysfunction We next asked whether inhibition of transcription produced a mitochondrial response across different cell lines. We treated SH- SY5Y, HeLa (WT and PARP1 null), HEK293T, WI38, primary MEFs, U2OS, and HCT116 (WT and p53 null) with different concentrations of inhibitors of RNA polymerase I (CX-5461 and Fig. 1. Loss of rDNA transcription leads to mitochondrial dysfunction in triptolide), II (α-amanitin and triptolide), or III (ML-60218) and models of Cockayne syndrome. (A) Western blots of SH-SY5Y cells subjected measured mitochondrial changes. Interestingly, RNA polymerase I to stable knockdown of CSA (shCSA) and CSB (shCSB). (Magnification: 60×.) inhibition led to a mitochondrial phenotype similar to what is seen (B) Mitochondrial membrane potential in the SH-SY5Y cells measured using in CS, with significantly increased mitochondrial membrane po- tetramethylrhodamine methyl ester (TMRM) (mean ± SEM, n = 5). (C)Mi- tential, content, and mitochondrial superoxide production in all tochondrial superoxide production (mitosox fluorescence) in the SH-SY5Y cell lines tested (Fig. 2A;seeFig. S2A for raw values of control cell cells (mean ± SEM, n = 5). (D) OCR in SH-SY5Y cells (mean ± SEM, n = 3). (E) OCR in WT (N2), csa-1–, and csb-1–deficient worms (mean ± SEM, n = 3). lines). Notably, the effect of RNA polymerase I inhibition on mi- (F and G) Altered GO terms, hierarchical clustering, and heat map of altered tochondrial function did not appear to be dependent on p53 be- GO terms in SH-SY5Y cells after knockdown of CSA or CSB or indicated cause HCT116 WT cells overall showed less of a response to the −/− treatmets. (H) Quantitative PCR (qPCR) of 47S rRNA levels in SH-SY5Y cells. inhibitors than the p53 cells. Immortalization did not appear to (I) Confocal microscope images of the SH-SY5Y cells after short-term treat- affect the response either. However, HeLa cells where PARP1 was − − ment with EU and staining with antibody against nucleolin. (J) Mitochon- deleted (PARP1 / ) showed a significantly attenuated response to drial superoxide production in the SH-SY5Y cells treated with 1 μM CX5461 RNA polymerase I inhibition compared with the parental HeLa or vehicle (veh) (mean ± SEM, n = 3–5). cell line (Fig. 2A). To further test whether these changes depended on PARP1 activation, we treated HeLa cells with RNA polymerase I inhibitor alone or in combination with the PARP1 inhibitor PJ34. Given that CSA and CSB have been implicated in transcription As shown, the mitochondrial changes that are caused by inhibition driven by RNA polymerases I (15, 16), II (17), and III (18), we of RNA polymerase I were completely reversed by treatment with measured changes in unmodified SH-SY5Y cells PJ34 (Fig. 2B and Fig. S2 B–D). To test the effect of the inhibitors after treatment with specific transcriptional inhibitors compared at an organismal level independent of CS, C. elegans were treated with controls (RNA polymerase I: CX5461; RNA polymerase I/II: with RNA polymerase I, II, or III inhibitors and examined for

α GENETICS triptolide; RNA polymerase II: -amanitin; RNA polymerase I/II/ mitochondrial changes. In agreement with the cellular data, RNA III: actinomycin D; and RNA polymerase III: ML60218). We polymerase inhibitors increased oxygen consumption rates, with treated the cells with the inhibitors at several concentrations to RNA polymerase I inhibition having the greatest effect (Fig. 2C). avoid data bias that might occur when choosing only one con- Thus, there appears to be a direct link between transcriptional centration. Additionally, we treated the knockdown cells with the efficiency, particularly of rDNA, and mitochondrial function. PARP inhibitor PJ34. To validate the results, we included gene expression array data from the cerebellum of human CS patients Ribosomal Defects Do Not Cause Mitochondrial Dysfunction and their controls from a recently published study (19). Notably, A number of diseases have been described with defects in rDNA hierarchical clustering showed an association between the CS pa- transcription [spinocerebellar ataxia 17 (SCA17) and hypomyeli- tients and the transcription inhibitor treatments, despite batch and nating leukodystrophy 7 (HLD7)], pre-rRNA processing [North tissue differences. Intriguingly, clustering revealed close associa- American Indian Childhood Cirrhosis (NAIC) and Treacher tion between the loss of CSA or CSB and the inhibition of rDNA Colins syndrome (TCOF)], and ribosome assembly [Diamond- transcription, and these changes were completely rescued by Blackfan anemia (DBA)] (20–25).Tounderstandthemecha- PARP inhibition (Fig. 1G). As a positive control, very low con- nism of how transcriptional defects generate mitochondrial centrations of ML60218 (1 μM) or CX5461 (0.001 μM) led to very changes, we used our recently developed in silico algorithms to few significantly changed GO terms (Fig. 1G). Indeed, there was predict mitochondrial involvement in the above-listed disorders no association between these two treatments and CSA or CSB based on clinical features (11, 26). Considering the clinical spectrum knockdown in the hierarchical clustering (Fig. 1G). In summary, of these diseases, hierarchical clustering showed an association

Scheibye-Knudsen et al. PNAS | November 1, 2016 | vol. 113 | no. 44 | 12503 Downloaded by guest on September 30, 2021 cycloheximide did not result in a mitochondrial phenotype (Fig. S2 E and F). Likewise, only modest changes in translation rates were seen by measuring incorporation of 35S-labeled methio- nine and cysteine and by assessing polysome profiles in CSA- or CSB-deficient SH-SY5Y cells as well as in SH-SY5Y cells treated with transcriptional inhibitors (Fig. S2 G and H). In summary, loss of rDNA transcription, but not translation, leads to mitochondrial dysfunction in Cockayne syndrome, and a mitochondrial etiology mediated by PARP1 may underlie the pathology observed in SCA17 and HLD7. Loss of CSA or CSB Leads to Transcriptional Stalling at Non-B DNA Because inhibition or loss of PARP1 attenuated the mitochondrial phenotype caused by defective rDNA transcription, we asked if PARP1 might be activated in the nucleolus of cells with reduced CSA or CSB. Using immunocytochemistry, we found increased levels of poly-ADP ribose (PAR) in the nucleolus (marked by nucleolin) of cells deficient in CSA or CSB (Fig. 3A). Notably, nucleolin levels appeared to be increased in the CSA- or CSB- deficient cells (Fig. 3A). Nucleolin binds DNA secondary structures such as G4 structures (28), and these structures have been pro- posed to activate PARP1 (29, 30). We therefore asked if CSA- or

Fig. 2. Loss of rDNA transcription and not ribosome dysfunction leads to mitochondrial dysfunction. (A) Heatmap of changes in mitochondrial membrane potential (TMRM), mitochondrial content (mitotracker green), mitochondrial superoxide production (mitosox), and whole-cell superoxide production [dihydroethedium (DHE)] (each square represents the mean of three independent experiments normalized to vehicle). (B) Mitochondrial membrane potential in HeLa cells treated for 24 h with vehicle (veh), 1 μM CX5461, 10 μM PJ34, or 1 μM CX5461 and 10 μM PJ34 (mean ± SEM, n = 3–6). (C) OCR in worms treated with CX5461, ML60218, and α-amanitin for 24 h (mean ± SEM, n = 3). (D) Hierarchical clustering of diseases based on their clinical parameters (red: mitochondrial diseases; green: nonmitochondrial diseases; blue: diseases with transcription defects; purple: classic ribosomo- pathies). (E and F) Mitochondrial and support vector machine scoring for likely mitochondrial involvement. (G) Mitochondrial membrane potential in WT or PARP1−/− HeLa cells subjected to knockdown of the indicated genes.

between defects in rDNA transcription (SCA17 and HLD7) and mitochondrial disorders, whereas NAIC, TCOF, and DBA did not show this relationship (Fig. 2D). A network algorithm corrobo- rated these findings where close associations were observed between SCA17 and HLD7 and mitochondrial diseases (see interactive figure on www.mitodb.com/network.html or Fig. S3). Furthermore, a machine-learning algorithm and a mitochondrial- scoring algorithm predicted mitochondrial involvement in SCA17 and HLD7, but not in NAIC, TCOF, or DBA (Fig. 2 E and F). Indeed, siRNA knockdown of TBP (mutated in SCA17) or POLR3A (mutated in HLD7) led to changes of mitochondrial membrane potential similar to those observed in CSA and CSB Fig. 3. CSA and CSB mediate transcription through secondary DNA struc- tures. (A) Representative immunocytochemistry micrograph of cells stained knockdown cells, whereas knockdown of CIRH1A (mutated in with antibody against poly-ADP ribose or nucleolin and quantification of the NAIC), TCOF1 (mutated in TCOF), or Rps19 (mutated in DBA) data (mean ± SEM, n = 8–15). (B) Change in coverage in the vicinity of a G4 did not result in mitochondrial membrane potential alterations structure. Data represent the mean of ∼10,000 transcripts. (C) Representa- (Fig. 2G). Indeed, TCOF1 has recently been shown to be down- tive Sybr gold-stained gel of recombinant CSB incubated with G4 DNA and stream of rDNA transcription (24) as well as of PARP1 re- complementary oligo (rc) with a 3′ prime or 5′ prime overhang. (Magnifi- × ± = cruitment to rDNA (27). Notably, the mitochondrial membrane cation: 60 .) (D) Quantification of C (mean SEM, n 3). (E) Quantified flow potential changes associated with TBP or POLR3A knockdown cytometry data of CS3BE cells stably expressing GFP-tagged CSA, CSB, nu- cleolar targeted CSB (NOLS-CSB), or GFP and stained with mitosox (mean ± were not observed in PARP1-null cells (Fig. 2G). In support of the SEM, n = 3). (F) OCR in CS3BE cells stably expressing GFP, GFP-CSA. GFP-CSB, observation that general translation defects did not lead to mito- or GFP-tagged nucleolar targeted CSB (mean ± SEM, n = 3–5). (G) Clonogenic chondrial changes, inhibition of translation by treatment with UV-survival of cells described in E (mean ± SD, n = 3).

12504 | www.pnas.org/cgi/doi/10.1073/pnas.1610198113 Scheibye-Knudsen et al. Downloaded by guest on September 30, 2021 CSB-deficient SH-SY5Y cells might show defects in transcription by CSB in the presence of cDNA, leading to annealing of the through G4 DNA sequences or other DNA secondary structures. two oligonucleotides (Fig. 3 C and D and Fig. S7D). This ac- Toward this aim, we performed RNA sequencing (RNA-seq) to tivity was independent of ATP hydrolysis because the reaction determine whether certain DNA regions might influence the occurred equally efficiently with or without ATP. The anneal- transcription efficiency in CSA- or CSB-deficient cells. Aligned ing activity of CSB was contingent upon either a 3′ or a 5′ tail reads were normalized to the control to minimize possible se- because CSB was unable to anneal the G4 structure with a quencing artifacts. After alignment, we developed a custom script peptide nucleic acid complementary to only the G4 sequence. to look at the normalized RNA-seq coverage in the 400 nucleotides FANCJ, a known G4 DNA , however, was able to re- preceding and following predicted DNA secondary structures in solve this structure (Fig. S7E). Because CSB appeared to be the genome. Inverted-, direct-, or mirror repeats, z-DNA se- independently able to melt G4 structures, we asked if we could quences, or short tandem repeats only led to minor pausing of the rescue mitochondrial features in CSA-deficient CS3BE cells RNA polymerase in the CSA- or CSB-deficient cells, and there by overexpressing CSB targeted to the nucleolus (Fig. S8A). was, as expected, no strand bias, whereas A-phased repeats in the Indeed, that appeared to be the case, as assessed by mito- nontranscribed strand led to moderate stalling (Fig. S4). However, chondrial membrane potential, ROS production, and oxygen the strongest stalling of the polymerase was seen with G4-forming consumption rates (Fig. 3 E and F and Fig. S8B). Interestingly, sequences as evidenced by increased coverage approaching the G4 only CSA overexpression and not CSB was able to rescue UV structure. Loss of coverage was observed in the nucleotides around sensitivity in CS3BE cells, suggesting a decoupling between UV the G4 structure, likely representing alignment artifacts due to the sensitivity and mitochondrial phenotypes as recently also sug- G-rich sequence. Interestingly, polymerase stalling displayed strand gested from clinical data (31) (Fig. 3G). bias in the CSA- or CSB-deficient cells, with particularly prominent stalling occurring when the G4 is in the transcribed strand (Fig. Stabilization of G4 Structures Leads to Accelerated Aging 3B). Furthermore, the amount of stalling depended on the amount We next asked if rDNA G4 structures could activate PARP1. In- of guanines in the sequence and thus the likelihood of the for- deed, recombinant PARP1 was activated by single-stranded rDNA mation of a G4 (Fig. 3B). Notably, there was no loss of signal and rRNA that contained G4-forming sequences whereas the downstream of the structures, suggesting that the polymerase complementary controls did not activate PARP1 (Fig. S9A). To eventually makes it through. avoid the confounding effects of G4 stabilization on replication, we Obstructive DNA structures might lead to less transcription in performed a number of experiments on primary, fully differenti- thedistalcomparedwithproximalpartofgenes.Totestthis,we ated, nondividing rat cortical neurons. Treatment of these cells performed nuclear run-on experiments on a few genes that showed with the G4-stabilizing drug pyridostatin or the RNA polymerase I altered expression in the microarray experiment. Although loss of inhibitor CX5461 led to PARP1 activation (Fig. 4A). Notably, transcription elongation was found in CSA- or CSB-deficient SH- treatment with these drugs did not lead to histone H2AX phos- SY5Y cells as previously reported (4, 17), there was no overall loss phorylation, suggesting that strand breaks are not driving the ac- of transcription when we compared the signal from the 5′ end with tivation of PARP1 (Fig. 4A). + the 3′ end of the genes (Fig. S5 A and B). Furthermore, the RNA- PARP1 activation may lead to loss of NAD and increased seq data did not show loss of transcription in CSA- or CSB-deficient nicotinamide (NAM). We therefore measured the levels of these cells relative to the control when investigating loss of coverage as a metabolites in the neurons treated with CX5461 or pyridostatin and + function of gene size across the entire transcriptome (Fig. S5C). In found a significant decrease in NAD /NADH and a trend to in- + addition, we did not find a generalized change in mitochondrial creased NAM (Fig. 4 B and C). Loss of NAD will lead to increased transcription rates in CSA- or CSB-deficient cells that could explain shunting of pyruvate to lactate. Consistently, by investigating oxygen alterations in mitochondrial function (Fig. S5D). consumption (OCR) and extracellular acidification (ECAR), we We next studied whether changes in transcription through G4 found loss of OCR/ECAR ratios in primary neurons treated with quadruplex structures are involved in normal aging by analyzing a CX5461, pyridostatin, or another G4 stabilizer, TMPYP4, and these published RNA-seq dataset of cells harvested from the iliac crest of changes were rescued by treatment with the PARP1 inhibitor PJ34 young and elderly women (Gene Expression Omnibus accession (Fig. 4D and raw data in Fig. S9 B and C). In addition, treatment of no. GSE72815). Although we found strand bias and transcriptional neurons with pyridostatin led to a dose-dependent increase in mi- pausing at G4 sequences, we did not find increased pausing with tochondrial membrane potential (Fig. S9D). Furthermore, treat- aging (Fig. S6A). However, we did find increased variability in ment of HeLa cells with pyridostatin or CX5461 led to a similar transcriptional rates with age around G4 regions, perhaps sug- mitochondrial profile as observed in CSA- or CSB-deficient cells, gesting alterations in the way G4 structures are handled (Fig. S6A). and this effect was nonadditive (Fig. 4E and Fig. S9 E–G). In ad- If G4 structures have to be resolved for transcription to occur, one dition, pyridostatin treatment led to loss of rDNA transcription as would expect that regions that show less transcription would have evidenced by decreased levels of 47S rRNA (Fig. 4F). GENETICS more stalling compared with highly transcribed regions. Indeed, To further understand if defects in transcription through DNA that appears to be the case in vivo when we measured the amount secondary structures could lead to accelerated aging, we treated of stalling as a function of coverage in the G4-containing region. C. elegans (the somatic cells of which are all postmitotic) with Here we found that regions with low coverage have greater stalling pyridostatin as well as with the rDNA transcriptional inhibitor than regions with high coverage, and this effect is only seen when CX-5461. These treatments led to decreased pharyngeal pumping, the G4 is on the transcribed strand (Fig. S6B). In summary, tran- loss of mobility, and shortened life span, all hallmarks of scriptional stalling occurs at DNA secondary structures, and there accelerated aging (Fig. 4 G–I). Accelerated aging was also found appears to be transcription-coupled resolution of DNA secondary when using a machine classifier to quantify morphological changes structures involving CSA and CSB. that occur with aging in the head of nematodes (32–34) (Fig. 4J). Strikingly, the accelerated aging changes could be rescued by + CSB Is Able to Resolve G4 Structures treatment with the NAD -precursor nicotinamide riboside (NR), Because CSB contains a helicase domain and is able to anneal although at higher pyridostatin or CX5461 concentrations life- cDNA strands, we asked whether recombinant CSB can resolve G4 span shortening could not be rescued (Fig. 4 G–J and Fig. S9H). structures and thereby facilitate transcription through them. We In summary, our results suggest that loss of transcription-coupled found that CSB is unable to unwind bimolecular and tetramo- resolution of DNA secondary structures leads to accelerated aging + lecular G4s (Fig. S7 A–C). However, unimolecular G4s that con- through G4 formation, PARP1 activation, NAD depletion, and tained either a 3′ or a 5′ oligonucleotide tail were efficiently melted mitochondrial dysfunction.

Scheibye-Knudsen et al. PNAS | November 1, 2016 | vol. 113 | no. 44 | 12505 Downloaded by guest on September 30, 2021 activate PARP1 in vitro and in vivo. Recombinant CSB is able to resolve G4-forming DNA structures through an ATP-independent DNA-annealing event, supporting the hypothesis that an accu- mulation of irresolvable G4 structures in Cockayne syndrome un- derlies the PARP1 activation in this disease. Finally, accelerated aging caused by G4 stabilization in nematodes can be rescued through replenishment of the central bioenergetics molecule + NAD . Although we believe that the most likely explanation for PARP1 activation is through direct activation by G4 structures, alternatively, PARP1 activation could occur through signaling from posttranslational modification of the stalled RNA polymerase or from other chromatin-associated proteins. Combined, our findings suggest an explanation for the PARP1 activation and mitochon- drial alterations observed in Cockayne syndrome (Fig. S9I). Previous findings implicate both CSA and CSB in rDNA transcription. Our data support this hypothesis and further underscore the role of rDNA in the aging process (15, 16, 18). Indeed, an expansion of extrachromosomal rDNA circles (ERCs) were among the first proposed causes of aging in eu- karyotes (42). Although ERC accumulation appears to be an age-associated phenomenon particular to yeast, recent data have shown increased rDNA content in the brains of patients suffering from of Alzheimer’s disease or Lewy body dementia (43, 44). Notably, PARP1 is activated in these diseases (45, 46). + Perhaps age-associated PARP1 activation and NAD depletion could be caused by rDNA accumulation leading to loss of activity + of NAD -dependent enzymes, such as sirtuins, and consequent mitochondrial dysfunction. G4 structures have been implicated in both replication and Fig. 4. G4 structure stabilization impairs neuronal bioenergetics and acceler- transcription regulation (47–50). Here we provide in vivo evidence ates aging. (A) Western blot of primary rat cortical neurons treated with 5 μg/mL for transcriptional pausing at DNA sequences that can form G4 α-amanitin, 1 μM CX5461, 1 μM pyridostatin, 10 μM PJ34, or 20 μMmenadione + structures. In Cockayne syndrome cells, this pausing is exacerbated, for 2 h. (Magnification: 60×.) (B) NAD /NADH in primary rat cortical neurons μ ± = leading to mitochondrial dysfunction. Although our study shows treated with 1 M pyridostatin or CX5461 for 2 h (mean SEM, n 3). (C) NAM that CSB can biochemically resolve G4 structures, the role of CSA levels in primary rat cortical neurons treated with 1 μM pyridostatin or CX5461 for 2 h (mean ± SEM, n = 2–3). (D) OCR/ECAR in primary neurons treated with G4 in this process is unknown. CSA is part of the DDB1-Cul4-Rbx stabilizers, pyridostatin and TMPYP4, and RNA polymerase I inhibitor CX5461 complex that is involved in ubiquitination of various substrates. It in combination with the PARP1 inhibitor PJ34 (mean ± SEM, n = 3). (E)Mito- is possible that CSA ubiquitinates proteins involved in the sta- chondrial membrane potential in HeLa cells treated with 1 μM pyridostatin, 1 μM bilization or resolution of G4 structures. Loss of CSA could thus CX5461, or 1 μM pyridostain and 1 μM CX5461 for 24 h (mean ± SEM, n = 3–6). lead to accumulation of G4 structures and subsequent PARP1 (F) qPCR of 47S rRNA in SH-SY5Y cells treated with 1 μM pyridostatin for 24 h activation. ± = (mean SEM, n 3). (G) Pharyngeal pumpting rates of nematodes treated with G4 structures occur at a high frequency in G-rich regions pyridostatin, CX5461, and/or NR. (H) Swimming ability of nematodes treated as such as the rDNA loci; however, this particular DNA structure in f. (I) Life-span measurements of nematodes treated as in F.(J) Age-state score for pharynx terminal bulb of nematodes treated as in F. is also observed in a number of other regions in the genome. It is possible that the effect that we see when inhibiting RNA polymerase I is due to the large amount of G4 structures in this Discussion region, and it is also possible that there are other properties PARP1 activity increases with normal aging and in accelerated inherently different about the rDNA. Indeed, inhibition of aging diseases, such as xeroderma pigmentosum group A, ATM, RNA polymerase II also leads to changes in mitochondrial and Cockayne syndrome (8, 11). We present evidence that PARP1 function although these changes are much smaller than ob- activation in Cockayne syndrome is caused by stalled transcription served after RNA polymerase I inhibition. Notably, although at DNA secondary structures that are enriched in rDNA. Using in we find transcriptional pausing at G4 structures in cells from silico and in vivo techniques, we find that transcription per se normal humans, aging does not exacerbate this pausing. This rather than downstream translational defects lead to mitochondrial observation suggests that a possible PARP1 activation in nor- pathology, a conclusion partly based on the discovery of mito- mal aging does not depend on decreased transcription-coupled chondrial dysfunction in the diseases hypomyelinating leukodys- resolution of DNA secondary structures. trophy 7 and spinocerebellar ataxia 17. Interestingly, the observed In conclusion, this work represents a mechanistic explanation mitochondrial changes with increased mitochondrial membrane for the pathogenesis of Cockayne syndrome through a con- potential and oxygen consumption suggest that the changes are not vergence of the CSA and CSB pathways and describes a role of caused by primary mitochondrial dysfunction but rather by a sec- transcription-coupled resolution of secondary DNA structures. ondary compensatory response due to nuclear changes, as pro- Materials and Methods posed previously (35, 36). This is in agreement with a number of previous studies showing similar mitochondrial changes as a re- All cell culture and nematode work was performed according to standard procedures. Western blot, RT-PCR, flow cytometry, Seahorse XF analyses, and sponse to nuclear DNA damage whereas studies in nonisogenic nuclear run-ons were performed according to standard procedures. In silico cell lines in the context of Cockayne syndrome have been in- – analyses of diseases was done as previously described (26) using the prevalence conclusive (7, 10, 37 41). Notably, G4-forming DNA sequences, of clinical features in each disease. RNA-seq was performed by the National which are abundant in rDNA, appear to be potent obstacles to Cancer Institute sequencing core. DNA helicase and DNA-annealing assays transcription in CSA- or CSB-deficient cells, and these structures were done with gel-purified oligonucleotides purchased from Lofstrand Labs.

12506 | www.pnas.org/cgi/doi/10.1073/pnas.1610198113 Scheibye-Knudsen et al. Downloaded by guest on September 30, 2021 One-way ANOVA with Tukey’s posttest was used to determine significant ACKNOWLEDGMENTS. We thank Jennifer L. Martindale and Dr. Ruin Moaddel difference across multiple samples. Two-tailed t tests were used to compare for technical assistance and Elayne Fivenson and Mustafa Okur for feedback on single groups. Statistical analyses were done with GraphPad Prism (GraphPad the manuscript. H.N. is supported by funding from the South East Regional Software, Inc.) or R. For detailed materials and methods, see SI Materials Health Authority. This research was supported by the Intramural Research Pro- and Methods). gram of the NIH, National Institute on Aging.

1. Wilson BT, et al. (2015) The Cockayne Syndrome Natural History (CoSyNH) study: Clinical 25. McCann KL, Baserga SJ (2013) Genetics. Mysterious ribosomopathies. Science 341(6148): findings in 102 individuals and recommendations for care. Genet Med 18(5):483–493. 849–850. 2. Nance MA, Berry SA (1992) Cockayne syndrome: Review of 140 cases. Am J Med Genet 26. Scheibye-Knudsen M, Scheibye-Alsing K, Canugovi C, Croteau DL, Bohr VA (2013) A 42(1):68–84. novel diagnostic tool reveals mitochondrial pathology in human diseases and aging. 3. Lake RJ, et al. (2014) The sequence-specific transcription factor c-Jun targets Cockayne Aging (Albany, NY) 5(3):192–208. syndrome B to regulate transcription and chromatin structure. PLoS Genet 27. Larsen DH, et al. (2014) The NBS1-Treacle complex controls ribosomal RNA tran- 10(4):e1004284. scription in response to DNA damage. Nat Cell Biol 16(8):792–803. 4. van Gool AJ, et al. (1997) The Cockayne syndrome B protein, involved in transcription- 28. Hanakahi LA, Sun H, Maizels N (1999) High affinity interactions of nucleolin with G-G- coupled DNA repair, resides in an RNA polymerase II-containing complex. EMBO J paired rDNA. J Biol Chem 274(22):15908–15912. 16(19):5955–5965. 29. Soldatenkov VA, Vetcher AA, Duka T, Ladame S (2008) First evidence of a functional 5. Troelstra C, et al. (1992) ERCC6, a member of a subfamily of putative , is interaction between DNA quadruplexes and poly(ADP-ribose) polymerase-1. ACS ’ involved in Cockayne s syndrome and preferential repair of active genes. Cell 71(6): Chem Biol 3(4):214–219. – 939 953. 30. Salvati E, et al. (2010) PARP1 is activated at telomeres upon G4 stabilization: Possible 6. Scheibye-Knudsen M, et al. (2012) Cockayne syndrome group B protein prevents the target for telomere-based therapy. Oncogene 29(47):6280–6293. accumulation of damaged mitochondria by promoting mitochondrial autophagy. 31. Calmels N, et al. (2016) Uncommon nucleotide excision repair phenotypes revealed by – J Exp Med 209(4):855 869. targeted high-throughput sequencing. Orphanet J Rare Dis 11:26. 7. Pascucci B, et al. (2012) An altered redox balance mediates the hypersensitivity of 32. Orlov N, et al. (2008) WND-CHARM: Multi-purpose image classification using com- – Cockayne syndrome primary fibroblasts to oxidative stress. Aging Cell 11(3):520 529. pound image transforms. Pattern Recognit Lett 29(11):1684–1693. + 8. Scheibye-Knudsen M, et al. (2014) A high-fat diet and NAD( ) activate Sirt1 to rescue 33. Eckley DM, et al. (2013) Molecular characterization of the transition to mid-life in – premature aging in Cockayne syndrome. Cell Metab 20(5):840 855. Caenorhabditis elegans. Age (Dordr) 35(3):689–703. 9. Cleaver JE, et al. (2014) Mitochondrial reactive oxygen species are scavenged by 34. Johnston J, Iser WB, Chow DK, Goldberg IG, Wolkow CA (2008) Quantitative image Cockayne syndrome B protein in human fibroblasts without nuclear DNA damage. analysis reveals distinct structural transitions during aging in Caenorhabditis elegans Proc Natl Acad Sci USA 111(37):13487–13492. tissues. PLoS One 3(7):e2821. 10. Chatre L, Biard DSF, Sarasin A, Ricchetti M (2015) Reversal of mitochondrial defects 35. Scheibye-Knudsen M, Croteau DL, Bohr VA (2013) Mitochondrial deficiency in Cock- with CSB-dependent serine protease inhibitors in patient cells of the progeroid ayne syndrome. Mech Ageing Dev 134(5-6):275–283. Cockayne syndrome. Proc Natl Acad Sci USA 112(22):E2910–E2919. 36. Fang EF, et al. (2016) Nuclear DNA damage signalling to mitochondria in ageing. Nat 11. Fang EF, et al. (2014) Defective mitophagy in XPA via PARP-1 hyperactivation and Rev Mol Cell Biol 17(5):308–321. NAD(+)/SIRT1 reduction. Cell 157(4):882–896. 37. Mizutani H, et al. (2002) Mechanism of apoptosis induced by a new topoisomerase in- 12. Gomes AP, et al. (2013) Declining NAD(+) induces a pseudohypoxic state disrupting hibitor through the generation of hydrogen peroxide. JBiolChem277(34):30684–30689. nuclear-mitochondrial communication during aging. Cell 155(7):1624–1638. 38. Facompré M, Wattez N, Kluza J, Lansiaux A, Bailly C (2000) Relationship between cell 13. Mouchiroud L, et al. (2013) The NAD(+)/Sirtuin pathway modulates longevity through cycle changes and variations of the mitochondrial membrane potential induced by activation of mitochondrial UPR and FOXO signaling. Cell 154(2):430–441. etoposide. Mol Cell Biol Res Commun 4(1):37–42. 14. Leen WG, Willemsen MA, Wevers RA, Verbeek MM (2012) Cerebrospinal fluid glucose 39. Fu X, Wan S, Lyu YL, Liu LF, Qi H (2008) Etoposide induces ATM-dependent mito- and lactate: Age-specific reference values and implications for clinical practice. PLoS chondrial biogenesis through AMPK activation. PLoS One 3(4):e2009. One 7(8):e42745. 40. Kluza J, et al. (2004) Mitochondrial proliferation during apoptosis induced by anti- 15. Bradsher J, et al. (2002) CSB is a component of RNA pol I transcription. Mol Cell 10(4): 819–829. cancer agents: Effects of doxorubicin and mitoxantrone on cancer and cardiac cells. – 16. Koch S, et al. (2014) Cockayne syndrome protein A is a transcription factor of RNA po- Oncogene 23(42):7018 7030. lymerase I and stimulates ribosomal biogenesis and growth. Cell Cycle 13(13):2029–2037. 41. Reipert S, Berry J, Hughes MF, Hickman JA, Allen TD (1995) Changes of mitochondrial 17. Le May N, et al. (2010) NER factors are recruited to active promoters and facilitate mass in the hemopoietic stem cell line FDCP-mix after treatment with etoposide: A chromatin modification for transcription in the absence of exogenous genotoxic at- correlative study by multiparameter flow cytometry and confocal and electron mi- – tack. Mol Cell 38(1):54–66. croscopy. Exp Cell Res 221(2):281 288. 18. Yu A, Fan HY, Liao D, Bailey AD, Weiner AM (2000) Activation of p53 or loss of the 42. Sinclair DA, Guarente L (1997) Extrachromosomal rDNA circles: A cause of aging in – Cockayne syndrome group B repair protein causes metaphase fragility of human U1, yeast. Cell 91(7):1033 1042. U2, and 5S genes. Mol Cell 5(5):801–810. 43. Hallgren J, Pietrzak M, Rempala G, Nelson PT, Hetman M (2014) Neurodegeneration- 19. Wang Y, et al. (2014) Dysregulation of gene expression as a cause of Cockayne syn- associated instability of ribosomal DNA. Biochim Biophys Acta 1842(6):860–868. drome neurological disease. Proc Natl Acad Sci USA 111(40):14454–14459. 44. Pietrzak M, Rempala G, Nelson PT, Zheng J-J, Hetman M (2011) Epigenetic silencing of 20. Martianov I, Viville S, Davidson I (2002) RNA polymerase II transcription in murine cells nucleolar rRNA genes in Alzheimer’s disease. PLoS One 6(7):e22585. lacking the TATA binding protein. Science 298(5595):1036–1039. 45. Love S (2001) Damage to nuclear DNA in Lewy body disease. Neuroreport 12(12): 21. Nakamura K, et al. (2001) SCA17, a novel autosomal dominant cerebellar ataxia 2725–2729. caused by an expanded polyglutamine in TATA-binding protein. Hum Mol Genet 46. Strosznajder JB, Czapski GA, Adamczyk A, Strosznajder RP (2012) Poly(ADP-ribose) po- 10(14):1441–1448. lymerase-1 in amyloid beta toxicity and Alzheimer’s disease. Mol Neurobiol 46(1):78–84. 22. Saitsu H, et al. (2011) Mutations in POLR3A and POLR3B encoding RNA Polymerase III 47. Bochman ML, Paeschke K, Zakian VA (2012) DNA secondary structures: Stability and subunits cause an autosomal-recessive hypomyelinating leukoencephalopathy. Am J function of G-quadruplex structures. Nat Rev Genet 13(11):770–780. Hum Genet 89(5):644–651. 48. Belotserkovskii BP, Mirkin SM, Hanawalt PC (2013) DNA sequences that interfere with 23. Freed EF, Baserga SJ (2010) The C-terminus of Utp4, mutated in childhood cirrhosis, is transcription: Implications for genome function and stability. Chem Rev 113(11):8620–8637. GENETICS essential for ribosome biogenesis. Nucleic Acids Res 38(14):4798–4806. 49. Maizels N, Gray LT (2013) The G4 genome. PLoS Genet 9(4):e1003468. 24. Werner A, et al. (2015) Cell-fate determination by ubiquitin-dependent regulation of 50. Murat P, Balasubramanian S (2014) Existence and consequences of G-quadruplex translation. Nature 525(7570):523–527. structures in DNA. Curr Opin Genet Dev 25:22–29.

Scheibye-Knudsen et al. PNAS | November 1, 2016 | vol. 113 | no. 44 | 12507 Downloaded by guest on September 30, 2021