news and views

Ribonucleotides in DNA: hidden in plain sight Sue Jinks-Robertson & Hannah L Klein

npg

© 2015 Nature America, Inc. All rights reserved.

Mapping of ribonucleotides to single-nucleotide resolution in yeast genomes provides new insight into the enzymology of DNA replication. In the current view of DNA replication, polymerase (Pol) ε is the major leading-strand DNA polymerase (DNAP) and Polδ the major lagging-strand DNAP, with Polα-primase initiating the ends of new DNA fragments1 (Fig. 1). It is critical that DNAPs select the base that pairs properly with the template but also that the selected base be attached to the correct sugar2. It has only recently been appreciated that incorporation of ribonucleoside monophosphates (rNMPs) in place of the corresponding deoxyribonucleoside monophosphates is widespread and associated with unwanted genetic consequences3–6. How, when, where and why does rNMP incorporation occur during replication? Four papers independently address the ‘where’ by mapping rNMP locations genome wide in budding and fission yeasts: Clausen et al.7 and Daigaku et al.8 in this issue of Nature Structural & Molecular Biology, Reijns et al.9 in Nature and Koh et al.10 in Nature Methods. Each study is a technical tour de force, and each provides new and important insight into the complex enzymology of eukaryotic DNA replication. The ‘how’ of rNMP incorporation by DNAPs reflects flexibility in the sugar-discrimination pocket of the replicative DNAPs2 as well as poor recognition and removal of rNMPs by associated nuclease-editing activity11. The use of mutant DNAPs that are either more or less permissive in their use of rNTPs as substrates has demonstrated the ‘when’: rNMP incorporation occurs primarily during replication4. Sue Jinks-Robertson is at the Department of Molecular Genetics and Microbiology, Duke University Medical Center, Durham, North Carolina, USA. Hannah L. Klein is at the Department of Biochemistry and Molecular Pharmacology, New York University School of Medicine, New York, New York, USA. e-mail: [email protected]

176

The consequences of rNMP insertion become particularly evident in the absence of RNase H2, the enzyme that initiates rNMP removal12 via the ribonucleotide excision repair pathway13. Importantly, the DNA strand synthesized by a given rNMP-permissive polymerase becomes densely studded with rNMPs, thus providing a unique footprint of its activity. Each group exploited rNMP-permissive forms of Polα, Polδ or Polε in RNase H2–defective strains and used deep sequencing to map rNMPs to single-nucleotide resolution. The challenge in mapping rNMPs resides in efficiently tagging their positions within duplex DNA, and the studies used four unique approaches (Fig. 2). Three relied on alkaline hydrolysis of the DNA backbone on the 3′ side of the rNMP, which generates a nick bordered by a 2′,3′-cyclic phosphate and a 5′-OH. Two techniques, hydrolytic end-sequencing (HydEn-seq)7 and polymerase usage sequencing (PU-seq)8, used different approaches to tag the 5′-OH thus generated, placing the rNMP at the –1 position when reads were aligned to the reference sequence. The third technique, ribose-seq10, maintained the rNMP in sequenced products and relied on tRNA ligase activity, which can ligate a 2′,3′-cyclic phosphate to a 5′-phosphate14,15. The fourth approach, embedded ribose–sequencing (emRibo-seq)16, used RNase H2 to nick 5′ of an rNMP in vitro, thus generating a 3′-OH that was then ligated to a tag. In this case, the rNMP was at the +1 position relative to the tagged base. Though the techniques differ, we suggest that all subsequently be referred to as ribose-seq, the most intuitive of the names. Together, these studies stunningly confirmed the division of labor for Polε and Polδ as the major leading- and lagging-strand DNAPs, respectively, and additionally revealed that more extensive rNMP incorporation occurred during leading-strand synthesis. There were sharp,

reciprocal changes in rNMP incorporation by Polδ and Polε, and these transitions corresponded to origins of replication (Fig. 3a). Although most switches reflected previously mapped replication origins, new origins were also identified, and their relative timing and strength were determined8. Termination zones between converging replication forks were also evident as broad regions where polymerase usage was reversed, thus suggesting an absence of discrete, programmed termination sites. Finally, at least some DNA synthesized by Polα survived Okazaki fragment (OF) maturation in the budding-yeast genome, and, as expected, such Polα-incorporated rNMPs tracked with those of Polδ (refs. 7,9). As an added bonus, rNMPs were mapped in mitochondrial as well as nuclear DNA in two of the studies7,10. In addition to confirming expected functions, several unanticipated findings further illustrate the power of a genome-wide approach. First, the proportions of distinct rNMPs in genomic DNA did not necessarily reflect the relative dNTP/rNTP ratios in the nucleotide pool, a result indicating that something other than simple mass action must contribute to rNMP Polε

Polαprimase

Okazaki-fragment maturation

Polδ

Figure 1 Division of labor at the replication fork. Polε (green) and Polδ (blue) are the major leading- and lagging-strand DNAPs, respectively. Wavy lines represent RNA primers synthesized by primase; these are initially extended by Polα (magenta) before the switching to Polδ during discontinuous OF synthesis. Segments synthesized by Polα-primase are displaced by Polδ during OF maturation. Nascent DNA strands are color coded to reflect the corresponding DNAP.

volume 22 number 3 march 2015 nature structural & molecular biology

news and views Sonicate

R

Alkaline hydrolysis

Restrict Ligate adapter

5′-OH Random priming incorporates dUTP

R

Ligate adapter RNase H2 Denature 3′-OH

R Kinase 5′ end Ligate adapter 1

Ligate adapter

Alkaline hydrolysis

5′-OH UU Degrade ssDNA Ligate adapters

Anneal adapter 2 2nd-strand synthesis PCR

Self-ligation with tRNA ligase R

Ion torrent

UU Nick at uracil Amplify intact strand

PCR

Illumina

PCR

EmRibo-seq (rNMP at +1)

R HydEn-seq (rNMP at –1)

PU-seq (rNMP at –1)

Illumina Ribose-seq (rNMP at end)

Figure 2 Summary of rNMP-mapping methods. In PU-seq and HydEn-seq, the 5′-OH generated by alkaline hydrolysis is at the end of tagged fragments that are sequenced, thus placing the corresponding rNMP at the –1 position. In ribose-seq, alkaline hydrolysis generates a cyclic phosphate that is intramolecularly ligated to a 5′-phosphate by a tRNA ligase. In emRibo-seq, RNase H2 produces a 3′-OH located at the end of sequenced fragments, with the corresponding rNMP at the +1 position. R, rNMP; filled triangle, cyclic phosphate end; U, uracil; ssDNA, single-stranded DNA; red and blue boxes, adapters that mark yeast DNA ends.

incorporation7,10. Second, because end-read counts at specific positions varied over 100-fold, the distribution of rNMPs was not random; alignment of sequences immediately adjacent to rNMPs also revealed possible context effects7,10. Third, ‘excursions’ from the typical reciprocal pattern of DNAP usage were evident, the significance of which remains to be elucidated7. Fourth, there appeared to be biased use of Polδ near replication origins in fission yeast, thus suggesting that this enzyme, rather than Polε, may initiate some leading-strand synthesis8. This is particularly intriguing because prior studies demonstrated that the catalytic activity of Polε is not necessary for yeast survival17,18. Finally, the observation that Polα contributes to ~1.5% of the mature budding-yeast genome is inconsistent with the currently widespread assumption that Polδ removes Polα-synthesized DNA by strand displacement during OF maturation9. Previous work showed that OF junctions are associated with nucleosome dyads and transcription factor–binding sites19, and Reijns et al.9 now demonstrate that basesubstitution rates in yeast and humans follow a similar pattern. The emerging model suggests that strand displacement during OF maturation is blocked by proteins that bind DNA, thus leading to the persistence of Polα errors and elevated base substitutions (Fig. 3b). It is

possible that protein binding reduces editing of Polα errors by Polδ (ref. 20) or that it affects subsequent mismatch-repair efficiency21. The bottom line is that base-substitution profiles correlate with transcription factor–binding sites, suggesting coevolution.

a

Origin

Origin

b Substitution rate

npg

© 2015 Nature America, Inc. All rights reserved.

Illumina

Though rNMPs are typically viewed as pathological, they have also been reported to have physiological roles. In fission yeast, for example, two consecutively incorporated rNMPs mark the nascent DNA strand and initiate programmed mating-type switching22. In budding yeast, RNase H2–dependent nicks in the leading strand of replication can provide a weak strand-discrimination signal during mismatch repair23,24. Now that the ‘where’ has been firmly established in yeast, the ‘why’ is ripe for further investigation, and progress is expected to be rapid. Additional questions to be addressed include where topoisomerasedependent deletions occur in the genome after rNMP incorporation and persistence, whether there is sequence specificity of RNase H2, whether specialized nonreplicative DNAPs produce distinctive patterns of rNMP incorporation and how altering nucleotide pools by hydroxyurea treatment affects rNMP incorporation. Because hydroxyurea-induced stress is a common tool used in replication studies, the implications of rNMP misincorporation are huge. Finally, it may be possible to apply a similar framework to the genomewide mapping of distinct types of DNA damage in the presence and/or absence of relevant repair pathways. It remains to be seen whether the rNMPsequencing approaches presented in these papers can be applied to organisms whose RNase H2 levels and DNAPs are not readily amenable to manipulation. The ability to edit genes is increasing so rapidly, however, that it is likely that similar experiments will soon be possible in mammalian cells, if they are not

OF junction rate

R

Elevated substitutions Termination zone Figure 3 Schematic of key results from genome-wide rNMP mapping. (a) Sharp and broad transitions of DNAP usage that reflect replication origins (filled diamonds) and termination zones, respectively. (b) Top, correlation between nucleosome positions, OF junctions and base substitution rates. Bottom, model in which DNA-bound proteins block complete removal of Polα errors during OF processing and lead to elevated mutation rates. DNA synthesized by Polε, Polδ or Polα is indicated in green, blue or magenta, respectively.

nature structural & molecular biology volume 22 number 3 march 2015

177

news and views already in the works. Ribose-seq opens the potential to map initiation zones, to monitor changes in initiation zones in different cell types and to correlate initiation zones with the three-dimensional organization of the nucleus. Last, mutation of RNase H2 is one cause of the rare autoimmune disease Aicardi-Goutiéres syndrome25, and recent studies have implicated faulty RNase H2 in systemic lupus erythematosus26. Mapping where rNMPs are located may provide new insights into the causes of these and other devastating autoimmune diseases and suggest new therapeutic approaches for their management.

npg

© 2015 Nature America, Inc. All rights reserved.

COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests.

1. Kunkel, T.A. & Burgers, P.M. Nat. Struct. Mol. Biol. 21, 649–651 (2014). 2. Joyce, C.M. Proc. Natl. Acad. Sci. USA 94, 1619–1622 (1997). 3. Kim, N. et al. Science 332, 1561–1564 (2011). 4. Nick McElhinny, S.A. et al. Nat. Chem. Biol. 6, 774–781 (2010). 5. Nick McElhinny, S.A. et al. Proc. Natl. Acad. Sci. USA 107, 4949–4954 (2010). 6. Potenski, C.J., Niu, H., Sung, P. & Klein, H.L. Nature 511, 251–254 (2014). 7. Clausen, A.R. et al. Nat. Struct. Mol. Biol. 22, 185–191 (2015). 8. Daigaku, Y. et al. Nat. Struct. Mol. Biol. 22, 192–198 (2015). 9. Reijns, M.A. et al. Nature, doi:10.1038/nature14183 (26 January 2015). 10. Koh, K.D., Balachander, S., Hesselberth, J.R. & Storici, F. Nat. Methods 12, 251–257 (2015). 11. Williams, J.S. & Kunkel, T.A. DNA Repair (Amst.) 19, 27–37 (2014). 12. Cerritelli, S.M. & Crouch, R.J. FEBS J. 276, 1494–1505 (2009).

13. Sparks, J.L. et al. Mol. Cell 47, 980–986 (2012). 14. Remus, B.S., Jacewicz, A. & Shuman, S. RNA 20, 1697–1705 (2014). 15. Schutz, K., Hesselberth, J.R. & Fields, S. RNA 16, 621–631 (2010). 16. Reijns, M.A. et al. Cell 149, 1008–1022 (2012). 17. Feng, W. & D’Urso, G. Mol. Cell. Biol. 21, 4495–4504 (2001). 18. Kesti, T., Flick, K., Keranen, S., Syvaoja, J.E. & Wittenberg, C. Mol. Cell 3, 679–685 (1999). 19. Smith, D.J. & Whitehouse, I. Nature 483, 434–438 (2012). 20. Pavlov, Y.I. et al. Curr. Biol. 16, 202–207 (2006). 21. House, N.C., Koch, M.R. & Freudenreich, C.H. Front. Genet. 5, 296 (2014). 22. Sayrac, S., Vengrova, S., Godfrey, E.L. & Dalgaard, J.Z. PLoS Genet. 7, e1001328 (2011). 23. Ghodgaonkar, M.M. et al. Mol. Cell 50, 323–332 (2013). 24. Lujan, S.A., Williams, J.S., Clausen, A.R., Clark, A.B. & Kunkel, T.A. Mol. Cell 50, 437–443 (2013). 25. Crow, Y.J. et al. Nat. Genet. 38, 910–916 (2006). 26. Gunther, C. et al. J. Clin. Invest. 125, 413–424 (2015).

Subversive bacteria reveal new tricks in their cytoskeleton-hijacking arsenal Roberto Dominguez During infection, pathogenic Yersinia species secrete the antiphagocytic factor YopO (or YpkA), which contains a kinase domain and a Rho GTPase guanine nucleotide–dissociation inhibitor (GDI) domain. The structure of YopO in complex with actin, along with biochemical analyses, reveals the mechanism by which YopO uses actin to activate its kinase domain and recruit, phosphorylate and deactivate actin-assembly factors implicated in phagocytic clearance of the bacterium. Many bacterial pathogens1,2 and some viruses3 use an ever-expanding arsenal of mechanisms to exploit the actin cytoskeleton of host eukaryotic cells for invasion, motility, replication and avoidance of the innate immune response. Several pathogens, including Clostridium botulinum, Salmonella enterica, Photorhabdus luminescens and Vibrio cholerae, secrete toxins that covalently modify actin to either block or promote polymerization4. In addition, pathogens such as Listeria monocytogenes, S. enterica and Shigella flexneri, use actin assembly to promote their uptake into eukaryotic cells5. Finally, numerous pathogens, including vaccinia virus, L. monocytogenes and Rickettsia spp. have evolved mechanisms to harness the forces of actin polymerization for motility1. Elegant work in this issue of Nature Structural & Molecular Biology by Lee et al.6 sheds light on the mechanism by which pathogenic Yersinia use the antiphagocytic factor YopO (also known as YpkA) to recruit and deactivate actin-assembly factors through phosphorylation. Roberto Dominguez is at the Department of Physiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, USA. e-mail: [email protected]

178

Pathogenic Yersinia (Y. enterocolitica, Y. pseudotuberculosis and Y. pestis) use a syringe-like type III secretion system (T3SS) to inject six Yersinia outer proteins (YopE, YopT, YopH, YopO, YopM and YopJ) into the mammalian host-cell cytosol7,8 (Fig. 1). The Yops act in concert to neutralize the host innate immune response by inhibiting phagocytosis by macrophages and neutrophils and downregulating the inflammatory response7,8. At least five of the Yops contain eukaryoticlike protein domains, and four (YopE, YopT, YopH and YopO) target the actin cytoskeleton, resulting in rapid actin depolymerization and enhanced virulence of the pathogen. YopE and YopT cause actin depolymerization through effects on Rho family GTPases, which are master regulators of the actin cytoskeleton9, whereas the tyrosine phosphatase YopH dephosphorylates focal adhesion kinase (Fak), paxillin and p130cas, among other targets7. YopO, the largest of the Yops (729 amino acids in Y. enterocolitica, the species studied by Lee et al.6), features an N-terminal membrane-targeting domain (residues 1–89), a serine/threonine kinase domain (residues 108– 434) and a C-terminal GDI domain (residues 435–729) (Fig. 1). Inside the host, YopO is targeted

to the plasma membrane via the N-terminal domain10. Formation of a 1:1 complex with monomeric actin leads to the autophosphorylation of YopO at Ser90 and Ser95 in the loop between the membrane-targeting and kinase domains, a necessary step in the activation of the kinase11,12. Because actin is found only in the host, the kinase domain remains inactive inside the pathogen. For its part, the GDI domain inhibits nucleotide exchange on RhoA and Rac1 (but not Cdc42), and disrupting these interactions through mutagenesis results in impaired YopO-induced cytoskeletal effects and in attenuated virulence in vivo13. Exogenous expression of YopO in eukaryotic cells produces a major disruption of the actin cytoskeleton11–15. This appears to result from the synergistic action of all the YopO domains, because mutations disabling membrane targeting, Rho GTPase binding or kinase activity result in reduced cytoskeleton disassembly11,13,14,16. Although studies using mouse infection models have sometimes produced conflicting results8, they have generally supported the notion that YopO has a substantial contribution towards the virulence of Yersinia13,14,17. The kinase activity of YopO was discovered more than 20 years ago17, yet the mechanism by which it contributes to the neutralization

volume 22 number 3 march 2015 nature structural & molecular biology