Telomere Shortening Relaxes X Chromosome Inactivation and Forces Global Transcriptome Alterations

Telomere shortening relaxes X chromosome inactivation and forces global transcriptome alterations

Stefan Schoeftnera,1, Raquel Blancoa,1, Isabel Lopez de Silanesa, Puriﬁcacio´ n Mun˜ oza,2, Gonzalo Go´ mez-Lo´ pezb, Juana M. Floresc, and Maria A. Blascoa,3

aTelomeres and Telomerase Group, Molecular Oncology Program, Spanish National Cancer Centre (CNIO), Madrid 28029, Spain; bBioinformatics Unit, Structural Biology and Biocomputing Program, Spanish National Cancer Centre (CNIO), Madrid 28029, Spain; and cAnimal Surgery and Medicine Department, Facultad de Veterinaria, Universidad Complutense de Madrid, Madrid 28040, Spain

Edited by Jasper Rine, University of California, Berkeley, CA, and approved September 16, 2009 (received for review August 18, 2009) Telomeres are heterochromatic structures at chromosome ends es- system, we provide evidence that progressive telomere shortening in sential for chromosomal stability. Telomere shortening and the ac- stratified epithelia, such as the skin, is linked to global deregulation of cumulation of dysfunctional telomeres are associated with organis- the mammalian transcriptome and loss of maintenance of epigenetic mal aging. Using telomerase-deficient TRF2-overexpressing mice silencing mechanisms, exemplified by the re-expression of an Xi-linked K5TRF2/Terc؊/؊) as a model for accelerated aging, we show that transgene. Indicative of the induction of a stress response, we find a) telomere shortening is paralleled by a gradual deregulation of the down-regulation of genes promoting cell cycle progression and mammalian transcriptome leading to cumulative changes in a defined up-regulation of the mTOR and Akt survival pathways. In addition, set of genes, including up-regulation of the mTOR and Akt survival cells with critically short telomeres show down-regulation of various pathways and down-regulation of cell cycle and DNA repair path- DNA repair pathways. These findings suggest that progressive ways. Increased DNA damage from dysfunctional telomeres leads to telomere shortening and the accumulation of dysfunctional telo- reduced deposition of H3K27me3 onto the inactive X chromosome meres with age may constitute a unique source of DNA damage, (Xi), impaired association of the Xi with telomeric transcript accumu- sufficient to induce global alterations in genome regulation. CELL BIOLOGY lations (Tacs), and reactivation of an X chromosome-linked K5TRF2 Results transgene that is subjected to X-chromosome inactivation in female We previously generated mice overexpressing the telomere-binding mice with sufficiently long telomeres. Exogenously induced DNA Ј damage also disrupts Xi-Tacs, suggesting DNA damage at the origin protein TRF2 under the control of the 5 regulatory region of the of these alterations. Collectively, these findings suggest that critically keratin 5 gene (PM K5TRF2 trangenic line) (23). TRF2 is a key player short telomeres activate a persistent DNA damage response that in the regulation of telomere length and telomere protection (23–25). In accordance with this, K5TRF2 mice showed severe telomere short- alters gene expression programs in a nonstochastic manner toward ening, increased sensitivity to UV radiation, premature skin aging (hair cell cycle arrest and activation of survival pathways, as well as impacts loss, skin hyperpigmentation, skin dryness), and increased skin cancer the maintenance of epigenetic memory and nuclear organization, (23, 26). In this transgenic line, skin phenotypes and embryonic lethality thereby contributing to organismal aging. were restricted to male mice, whereas female littermates remained phenotypically normal (23). We show here that PM K5TRF2 females ͉ ͉ ͉ ͉ aging chromosome X inactivation DNA damage epigenetics display TRF2 protein levels only slightly above wild-type levels, com- telomeres pared with robust TRF2 overexpression in littermate transgenic males (Fig. 1A and B). These findings suggest that the K5TRF2 transgene is ysfunctional, critically short telomeres elicit a DNA damage located at the X chromosome and specifically silenced in females. To Dresponse (DDR) that triggers senescence or apoptosis in address this, we performed DNA FISH on male PM K5TRF2 kera- mammalian cells, two processes that are associated with organismal tinocytes and mapped the integration site for the PM K5TRF2 trans- aging (1–9). Mice with a targeted deletion of the RNA component gene to the X chromosome (supporting information (SI) Fig. S1A). of telomerase (TercϪ/Ϫ) display accelerated telomere shortening, These findings suggest that the K5TRF2 transgene is silenced by a non- premature loss of tissue renewal, and decreased longevity (3, 7–9). random X inactivation event in female K5TRF2 mice, thereby prevent- DNA damage signals originating from critically short telomeres in ing TRF2 overexpression and the onset of severe skin pathologies. these mice is in line with current models proposing a causative role Next, we crossed PM K5TRF2 mice into a telomerase-deficient Ϫ Ϫ for DNA damage in organismal aging (10–13). Interestingly, epi- (Terc / ) background to address the impact of telomere shortening genetic alterations at heterochromatic regions are proposed to lead on global epigenetic alterations, including chromosome X inacti- to changes in gene expression associated with aging (14–16). In S. vation. To this end, we generated increasing generations (G1–G3) cerevisiae, induction of DNA double-strand breaks (DSBs) or of female PM K5TRF2 transgenic mice in a telomerase-deficient Ϫ/Ϫ cellular stress causes a dramatic redistribution of telomeric silent background (K5TRF2/G1–G3 Terc ; see Methods). Progressive information regulator (Sir) proteins and yKU proteins (17–19), thus linking changes in telomere chromatin to global epigenetic alter- Author contributions: S.S. and M.A.B. designed research; S.S., R.B., I.L.d.S., and J.M.F. per- ations. Sir complex relocalization is known to alter the expression formed research; P.M. contributed new reagents/analytic tools; S.S., R.B., I.L.d.S., G.G.-L., of stress response genes, survival factors, and ribosomal biogenesis (20, J.M.F., and M.A.B. analyzed data; and S.S. and M.A.B. wrote the paper. The authors declare no 21). In functional analogy to yeast, mammalian SIRT1 is redistributed conflict of interest. upon induction of DNA damage, causing broad alterations in global This article is a PNAS Direct Submission. gene expression (22). Collectively, these findings suggest that aging- Freely available online through the PNAS open access option. related DNA damage drives gene expression alterations that could 1S.S. and R.B contributed equally to this work. promote the development of aging pathologies. 2Present address: Programa de Epigene´tica y Biología del Cańcer, Institut Catala` An important question to determine is how the various types of DNA d’Oncología (ICO), Gran Via s/n, 08907 LЈHospitalet de Llobregat, Barcelona, Spain. damage impact gene expression changes associated with organ- 3To whom correspondence should be addressed. E-mail: [email protected]. ismal aging. In this study, we focused on the isolated effect of dysfunc- This article contains supporting information online at www.pnas.org/cgi/content/full/ tional telomeres on global genome regulation. Using a mouse model 0909265106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0909265106 PNAS Early Edition ͉ 1of6 Downloaded by guest on September 27, 2021 p=0.035 p<0.0001 p<0.0001 -/- AB -/- -/- p<0.0001 p<0.0001 p=0.9 25 n=4 AB p=0.2 p=0.01 p=0.02 20 -/- Wild type Wild

K5TRF2 K5TRF2 1250 -/- p<0.0001 p<0.0001 p<0.0001 PM PM PM 15 K5TRF2 K5TRF2 G1 Terc 1000 K5TRF2/G3 Terc Wild type K5TRF2 K5TRF2/G1 Terc K5TRF2/G2 Terc 10 kDa G3Terc kDa TRF2 5 n=4 70 TRF2 70 n=4 750 relative TRF2 protein relative 0 50 actin levels (Wild settype ”1”) 50 actin 35 500

30 p=0.014 n=4 25 p=0.056 250 20 (a.f.u) fluorescence telomere p=0.006 n=2 K5TRF2 n=3 K5TRF2 CD type Wild 15 0 -/- -/- n=3 PM 10 PM -/- -/- -/- -/- -/- -/- PM n=4 n=2 5 n=4 n=2 n=4 (wild type set ”1”) Terc Terc TRF2 protein levels 0 -/- -/- -/- -/-

75 -/- -/- K5TRF2/ G3 K5TRF2/ G3 K5TRF2 K5TRF2 K5TRF2 K5TRF2 G1 Terc G2 Terc G3 Terc Wild type Wild 70 K5TRF2 K5TRF2 G3Terc Wild type G2 Terc 10 G1 Terc K5TRF2/G1 Terc K5TRF2/G3 Terc 5 K5TRF2/G2 Terc K5TRF2/G2 Terc K5TRF2/G3 Terc K5TRF2/G1 Terc skin phenotype (%) phenotype skin frequency of severe of frequency 0 -/- -/- -/- -/- -/- -/-

K5TRF2/ p=0.35 K5TRF2 K5TRF2 -/- p=0.88 Wild typeWild -/- D

G2 Terc C

G3 Terc G3Terc G1 Terc G3 Terc K5TRF2 25/435 10 p=0.01 n=2

8 p=0.005 78/1401 K5TRF2/G2 Terc K5TRF2/G1 Terc K5TRF2/G3 Terc n=4 E DAPI 6 16/373 n=2 n.s 200200 16/660 *** 4 n=2 150150 15.4 42.9 *** ** 2 2/698 100100 53.8 28.6 -/- (%) cells positive H2AX n=2 6.3 PM K5TRF2/G3 Terc H2AX *** F 0 -/- -/-

5050 -/-

28.1 69.2 85.7 SCC -/- % lesions in non- 6.8 glandular stomach glandular 2.7 5.4 2.9 4.4 (cumulative numbers) 0 0 31.3 36.4 0 4.2 0 14.7 8.7 7.7 dysplasia K5TRF2

n.s. G3Terc G2 Terc 150150 hyperplasia ** stomach merge non-glandular SCC 57.1 ulcers 1001001 0 53.8 K5TRF2/G2 Terc 0 n.s. K5TRF2/G3 Terc 50 ** 50 * 85.7 5.9 21.9 61.5 K5TRF2/ 4.4 K5TRF2/2.3 2.3 K5TRF2/ K5TRF2/ 4.1 2.9 G1Terc-/- G2Terc-/- G3Terc-/- 4.2G3Terc-/- 0 21.9(n=32) 8.7 (n=44)13.6 0 (n=13) (n=7) 0 esophagus -/- -/- -/- -/- -/- -/- (cumulative numbers) % lesions in esophagus Fig. 2. Loss of silencing of the X-linked PM K5TRF2 transgene leads to telomere Ϫ/Ϫ K5TRF2 K5TRF2 G2Terc G3Terc G1Terc Wild type

Wild type dysfunction in K5TRF2/Terc females. (A)(Top) TRF2 protein levels in back skin keratinocytes; (Bottom) quantification of TRF2 levels after normalizing against ␤-ac- K5TRF2/G3Terc K5TRF2/G1Terc K5TRF2/G2Terc tin. n, 2–4 experiments. (B) Telomere Q-FISH in tail skin. Red bars, average telomere fluorescence intensity. A Student’s t test was used to calculate statistical significance. a.f.u, arbitrary fluorescence units. (C) ␥H2AX-positive cells (white arrowheads) in the Fig. 1. An X-linked transgene is re-expressed upon telomere shortening. (A) Male indicated skin sections. (Scale bar: 10 ␮m.) (D) Quantification of ␥H2AX immunos- PM K5TRF2 mice display elevated TRF2 protein levels compared with female litter- tainings. n, mice analyzed; ␥H2AX-positive nuclei and total number of cells analyzed mates. Actin, loading control. (B) Quantification of Western blots; n, number of are indicated. A Fisher’s exact test was used to calculate statistical significance. keratinocyte preparations; standard error is indicated. A Student’s t test was used to calculate statistical significance. (C) Skin phenotypes in PM K5TRF2/TercϪ/Ϫ females. (D) Quantification of skin disorders. (E) Quantification of abnormalities in stratified epithelia. Female mice: wild type, n ϭ 68; K5TRF2, n ϭ 74; G1 TercϪ/Ϫ, n ϭ 34; TRF2 overexpression leads to dramatic telomere shortening in K5TRF2/G1 TercϪ/Ϫ,nϭ32; G2 TercϪ/Ϫ,nϭ23; K5TRF2/G2 TercϪ/Ϫ,nϭ44; G3 TercϪ/Ϫ, PM K5TRF2 males (23). To investigate whether reexpression of the n ϭ 4; K5TRF2/G3 TercϪ/Ϫ, n ϭ 13. Male mice: wild type, n ϭ 24; K5TRF2, n ϭ 7. A silenced PM K5TRF2 transgene induces further telomere shorten- Ϫ Ϫ Fisher’s exact test was used to calculate statistical significance. (F) Histopathological ing in K5TRF2/G1–G3 Terc / females compared with their G1–G3 Ϫ Ϫ findings in stratified epithelia. Black arrowheads, displastic nuclei; white double- Terc / controls, we performed quantitative telomere Q-FISH on headed arrows, hyperplastic areas. skin sections (see Methods). Telomerase deficiency in G1–G3 TercϪ/Ϫ mice results in slow but continuous telomere shortening telomere shortening in K5TRF2/G1–G3 TercϪ/Ϫ females resulted in with increasing mouse generations (Fig. 2B). Consistent with our gradual appearance of K5TRF2-associated skin phenotypes. previous findings (23), the skin of PM K5TRF2 males displays a K5TRF2/G2–G3 TercϪ/Ϫ females display hair loss, skin hyperpig- dramatic telomere shortening (Fig. 2B). Slightly elevated TRF2 mentation, and increased skin lesions, reaching a severity in levels in PM K5TRF2 females cause only mild telomere shortening K5TRF2/G3 TercϪ/Ϫ females that is similar to that in K5TRF2 males (Fig. 2 A and B). Importantly, loss of silencing of the K5TRF2 (Fig. 1 C and D and Fig. S1B). Histopathological analyses further transgene in K5TRF2/G1 TercϪ/Ϫ females induces accelerated Ϫ Ϫ showed that K5TRF2/G2–G3 Terc / females also develop preneo- telomere shortening compared with their G1 TercϪ/Ϫ controls (Fig. plastic (dysplasia, hyperplasia) and neoplastic (squamous cell car- 2B), and this effect is exacerbated in K5TRF2/G2–G3 TercϪ/Ϫ cinoma, SCC) lesions in other stratified epithelia with reported K5 females (Fig. 2B). We confirmed these results by Southern blot- promoter activity, such as nonglandular stomach and esophagus based telomere restriction fragment (TRF) analysis (Fig. S1C). (Fig. 1 E and F). Again, the penetrance of these epithelial lesions Critically short telomeres elicit a DDR provoking entry into cell in PM K5TRF2/G3 TercϪ/Ϫ females was comparable with that of Ϫ/Ϫ cycle arrest/senescence or apoptosis (1–9). To test whether reacti- male PM K5TRF2 mice (Fig. 1 E and F). Littermate G3 Terc Ϫ Ϫ vation of the K5TRF2 transgene in K5TRF2/Terc / females re- females did not develop any of these pathologies (Fig. 1 C–F). Ϫ/Ϫ sulted in increased DNA damage signaling at dysfunctional telo- Interestingly, telomere shortening in K5TRF2/G1–G3 Terc ␥ females coincided with a gradual increase in TRF2 protein levels, meres, we quantified the abundance of H2AX foci in the skin of reaching the highest levels in K5TRF2/G3 TercϪ/Ϫ females and in experimental mice. Consistent with their shorter telomeres (Fig. 2B), we observed a significant increase of ␥H2AX-positive nuclei PM K5TRF2 males (Fig. 2A). These findings suggest that progres- Ϫ Ϫ Ϫ Ϫ sive telomere shortening in K5TRF2/G1–G3 TercϪ/Ϫ females drives in K5TRF2/G2 Terc / females compared with their G2 Terc / a gradual loss of silencing of the Xi-linked K5TRF2 transgene and female littermates (Fig. 2 C and D). This effect was further increased expression of TRF2, which in turn triggers epithelial increased in females lacking telomerase activity for 3 generations, pathologies in PM K5TRF2/TercϪ/Ϫ females that recapitulate those K5TRF2/G3 TercϪ/Ϫ mice, reaching similar DNA damage levels to of PM K5TRF2 transgenic males. those of K5TRF2 males (Fig. 2 C and D).

2of6 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0909265106 Schoeftner et al. Downloaded by guest on September 27, 2021 DAPI H3K27me3 RNA Pol II merge p<0.0001 C p=0.32 ABp=0.09 p=0.37 p=0.23 p=0.28 N=3 N=2 N=2 100 N=3 n=267 N=3 Wild type 85 n=274 N=2 n=274 n=164 n=331 n=164 N=2 90 75 n=267 80 N=3 65 n=331 70 55 60 50 Fig. 3. Critical telomere shortening 45 K5TRF2/ 40 (H3K27me3) results in impaired H3K27me3 of the -/- 35 30 % of% Xi cells with G2 Terc (H3K27me3)at Xi 20 Xi in vivo. (A) Combined immuno-

25 Pol% RNA IIexclusion -/- -/- -/- -/- staining for H3K27me3 and RNA polymerase II in E12 embryo skin sec- K5TRF2 K5TRF2 K5TRF2/ Wild type Wild G3 Terc G2 Terc K5TRF2/ Wild type Wild G2 Terc

G3 Terc tions. (B) Reduced deposition of H3K27me3 on the Xi in K5TRF2/G2 Ϫ/Ϫ K5TRF2/ p=0.0002 Terc skin sections. (C) Quantiﬁca- D Wild type K5TRF2 G3 Terc-/- G3 Terc-/- E p=0.0008 p=0.694 tion of RNA polymerase II exclusion N=4 N=3 at the Xi (decorated with 70 n=711 n=519 60 N=4 H3K27me3). (D) Immunostaining for DAPI n=620 N=3 50 n=523 H3K27me3 in adult tail-skin sections. 40 (E) Reduced frequency of Xi-coupled 30 20 H3K27me3 in adult skin sections of (H3K27me3) Ϫ/Ϫ

% of% cells with Xi 10 K5TRF2/G2 Terc females. N, num- H3K27me3 0 ber of mice; n, nuclei analyzed. Error -/- -/- bars, standard error. An unpaired Student’s t test was used to calculate K5TRF2 G3Terc Wild type Wild K5TRF2/ G3 Terc statistical signiﬁcance.

Mammalian X chromosome inactivation (XCI) ensures dosage- K5TRF2/G3 TercϪ/Ϫ (culture D) cells, which was confirmed in vivo in compensated expression of X-linked genes in male and female cells. the epidermis of K5TRF2/G2 TercϪ/Ϫ E12 embryos (Fig. 4 G–J). Tac-Xi

XCI is initiated during female embryonic development by the association was normal in cells and embryos with sufficiently long telo- CELL BIOLOGY up-regulation and spreading of the noncoding Xist RNA along the meres (Fig. 4 G–J). Of notice, the frequency of Tacs was decreased in future inactive X chromosome (Xi), triggering long-range transcrip- skin sections from E12 K5TRF2/G2 TercϪ/Ϫ embryos (Fig. 4K), sug- tional silencing and the deposition of repressive chromatin marks gesting that TelRNA/TERRA expression levels influence Tac forma- such as histone H3 lysine 27 trimethylation (H3K27me3) in cis (27, tion. To test whether a DNA damage signal elicited from dysfunctional 28). Although decoration of the Xi with Xist is not significantly telomeres could directly contribute to Tac-Xi dissociation, we exposed affected upon telomere shortening (Fig. S2 A and B), we observed primary MEF (pMEF) to ionizing radiation. Thirty minutes after irra- a significant decrease of Xi-associated H3K27me3 in skin sections diation, cells exhibited an efficient DDR activation, as indicated by the Ϫ/Ϫ of K5TRF2/G2 Terc E12 embryos as well as in adult skin from appearance of abundant ␥H2AX foci, which decreased in the following Ϫ/Ϫ Ϫ/Ϫ G3 Terc and K5TRF2/G3 Terc mice, as detected by confocal hours postirradiation (Fig. S3A). Treatment with ionizing radiation immunofluorescence (Fig. 3 A–E). Exclusion of RNA polymerase caused only minimal alterations in TelRNA/TERRA expression levels Ϫ/Ϫ II from the H3K27me3 mark at the Xi in K5TRF2/G2 Terc and Tac frequency in the experimental time window used here, as embryos indicates that the establishment of Xist-mediated silencing determined both by Northern blotting and RNA FISH (Fig. S2 C–G). is normal in the epidermis of E12 mice cells (Fig. 3 A and C). These Importantly, we observed a severe reduction in Tac-Xi associations 3 h findings suggest that increased DNA damage originated from short postirradiation, suggesting that these associations are disrupted as a telomeres leads to a defective maintenance of X inactivation, consequence of DNA damage (Fig. S3B). These findings suggest that marked by a reduced frequency of H3K27me3 at the Xi and the DNA damage signaling originating from dysfunctional telomeres alters reexpression of the Xi-linked K5TRF2 transgene. the maintenance of epigenetic gene silencing on the Xi but also affects Vertebrate telomeres are transcribed by RNA polymerase II, giving nuclear organization of chromatin territories, as exemplified by Tac-Xi rise to UUAGGG repeat-containing, noncoding RNAs (TelRNA/ dissociation in cells with critically short telomeres. These findings also TERRA) that localize to mammalian telomeres but also form accu- raise the possibility that defective Xi maintenance may be part of the mulations (Tacs, TelRNA accumulations) in the vicinity of the inactive X chromosome (Xi) in female cells (29, 30). To test a possible link with global chromatin defects triggered by telomere shortening. XCI, we determined telomere length, TelRNA/TERRA expression, A recent report indicated that the accumulation of DNA damage TRF2 protein levels, and Tac-Xi localization pattern in primary kera- during mammalian aging results in altered gene expression patterns due tinocytes derived from female wild-type, K5TRF2, and K5TRF2/G2–G3 to the redistribution of the histone deacetylase SIRT on chromatin (22). TercϪ/Ϫ newborn mice. G2 TercϪ/Ϫ cells displayed telomere shortening To investigate whether progressive telomere shortening selectively that was further augmented in K5TRF2/G2 TercϪ/Ϫ keratinocytes (Fig. affects transcriptional memory on the Xi or is associated with a global 4 A and B) (26). Selection for cells that rescue telomere length by deregulation of gene expression, we performed comparative transcrip- recombination-based mechanisms explains the heterogeneity of telo- tome analyses using primary skin keratinocytes from wild-type, Ϫ/Ϫ Ϫ/Ϫ Ϫ/Ϫ mere lengths in K5TRF2/G3 TercϪ/Ϫ keratinocytes (see cultures F and K5TRF2, G2 Terc , K5TRF2/G2 Terc , and K5TRF2/G3 Terc 4 in Fig. 4 A and B) (26). Telomere attrition in K5TRF2/G2 TercϪ/Ϫ PM females. To control for sex-specific gene expression differences, we (cultures A and B) and K5TRF2/G3 TercϪ/Ϫ (culture D) cells is accom- included wild-type and K5TRF2 PM males in the analysis (23). Differ- panied by loss of silencing of the X-linked K5TRF2 transgene, resulting ences in transcript levels of individual genes obtained in pairwise tran- in increased TRF2 protein levels (see cultures A, B, and D in Fig. 4 scriptome comparisons were blotted along cytogenetic maps of all A–D), further confirming our in vivo data (Fig. 2A). Previous reports mouse chromosomes and revealed that progressive telomere shorten- suggested a positive correlation between telomere length and TelRNA/ ing is associated with increasing transcriptome alterations, affecting all TERRA levels (29, 30). In agreement with this, we observed reduced mouse chromosomes to a similar extent (Fig. S5 A–J and Table S1). TelRNA/TERRA levels in keratinocyte cultures with short telomeres Genome-wide alterations reached high statistical significance in female (Fig. 4 E and F). Furthermore, Xist–TelRNA/TERRA double-RNA K5TRF2/G3 TercϪ/Ϫ keratinocytes (P Ͻ 0.001) when compared with FISH on female keratinocytes revealed an impaired association of Tacs other genotypes with longer telomeres (Fig. S4A). We conclude that with the inactive X in female K5TRF2/G2 TercϪ/Ϫ (culture B) and PM telomere attrition leads to a global impairment in the accomplishment

Schoeftner et al. PNAS Early Edition ͉ 3of6 Downloaded by guest on September 27, 2021 A 4000 B

3000 1000

2000 800 729 670 619 600 483 units (T.F.U.)

intesity units 1000 400 349 304 fluorescence intensity 222 217 242 200 Telomere fluoresence 0

Average telomere 0 fluorescence units

-/- AF4DB -/- AF4DB K5TRF2 K5TRF2 K5TRF2 K5TRF2 G2Terc-/- G3 Terc-/- K5TRF2 K5TRF2 -/- -/- K5TRF2 K5TRF2 G2Terc G3 Terc Wild type Wild type G2 Terc G2 Terc

C -/- K5TRF2 K5TRF2 D G2Terc-/- G3 Terc-/- 6 K5TRF2 Wild type K5TRF2 G2 Terc A B D4F 4

TRF2 2 actin 0 Relative TRF2protein

levels; Wildset “1”type Fig. 4. Critical telomere shortening re- -/- AF4DB -/- -/- sults in a loss of Xist-Tacs association. (A) K5TRF2 K5TRF2 -/- -/- -/- Distributions of individual telomere

K5TRF2 K5TRF2 G2Terc G3 Terc K5TRF2 K5TRF2 Wild type Wild G2 Terc G3Terc G2 Terc lengths in metaphases from primary kera- G2 Terc K5TRF2 Wild type B FA 4 D E F 150 tinocytes. a.f.u., arbitrary ﬂuorescence n=2 kb 125 units; red bars, medium telomere ﬂuores- 100 n=3 cence. (B) Average telomere length per ge- 6 4 75 TelRNA/ notype. Average values are shown on top 3 TERRA 50 of the bars. (C) TRF2 protein expression. (D) 2 25

Wild type set “100” Quantiﬁcation of TRF2 levels normalized 0 Relative TelRNA levels; against ␤-actin. (E) TelRNA/TERRA levels as

-/- AF4DB K5TRF2 K5TRF2 detected by Northern blotting. (F) Quanti- gapdh G2Terc-/- G3 Terc-/- ﬁcation of TelRNA/TERRA levels normal- K5TRF2 G2 Terc Wild type Wild ized against gapdh. n, cell lines analyzed; 100 G TelRNA/ H n=7 n=4 standard error is indicated. (G) Combined Xist DAPI TERRA merge 80 Xist-TelRNA/TERRA RNA FISH to detect Xist- 60 Tac association. Cells shown are polyploid Wild type 40 and present 2 Xist signals. DNA was stained 20 no Xi with DAPI. (H) Xist-TelRNA association is

Tac-Xi association (%) 0 lost in most cultures with short telomeres

-/- A B F4 D and high TRF2. n, cell lines analyzed; stan- K5TRF2 K5TRF2 K5TRF2 G3 Terc-/- G2Terc-/- G3 Terc-/- dard error is indicated. (I) Xist-TelRNA RNA K5TRF2 K5TRF2 Wild type G2 Terc FISH in E12 embryonic skin sections. Green,

p=0.002 Xist RNA; red, TelRNA/TERRA; blue, DAPI. p=0.781 p<0.001 IJp=0.229 Kp=0.002 Gray arrowheads, telomere-associated Tel- p=0.329 N=2 N=3 RNA/TERRA; white arrowheads, TelRNA ac- n=356 p=0.207 K5TRF2/ 24 p=0.004 n=437 -/- -/- N=2 cumulations (Tacs) in the vicinity of the Xist- Wild type K5TRF2 G3 Terc G2 Terc N=3 80 N=3 n=488 20 N=2 n=400 n=400 N=2 n=488 n=356 coated inactive X chromosome. (J) 16 60 DAPI K5TRF2/G2 TercϪ/Ϫ E13 embryos display re- Xist 12 N=3 40 n=437 duced Xi-Tac association. (K) Quantiﬁca- TelRNA/ 8 TERRA association 20 tion of Tac frequency in E12 embryos. N, 4

% of cells without Tac number of mice; n, nuclei analyzed. Error % of cells with Tac-Xi 0 0 -/- -/- -/- -/- bars, standard error. An unpaired Student’s t test was used to calculate statistical K5TRF2 K5TRF2 Wild type Wild type G3 Terc G3 Terc K5TRF2/ G2 Terc K5TRF2/ G2 Terc signiﬁcance.

of gene expression programs. Focusing our analyses on individual of genes up-regulated and 20% of genes down-regulated in male genes, we found 213 affected genes in female G2 TercϪ/Ϫ keratinocytes K5TRF2 keratinocytes (FDR Ͻ0.05) showed the same regulation in [false detection rate (FDR) Ͻ0.05], and this number was further female K5TRF2/G3 TercϪ/Ϫ samples. This indicates that telomere increased in K5TRF2/TercϪ/Ϫ cells reaching a maximum of 2,757 gene attrition affects the expression of an overlapping set of genes in male expression changes in female K5TRF2/G3 TercϪ/Ϫ keratinocytes (Fig. and female cells (Fig. 5A). Focusing on individual probe sets displaying S4B). We note that highly significant gene expression changes (FDR most robust alterations, we confirmed that robust gene expression Ͻ0.001) were detected only in K5TRF2/G3 TercϪ/Ϫ keratinocytes (Fig. changes in K5TRF2/G3 TercϪ/Ϫ cells (FDR Ͻ1.00E-07) are gradually S4B). Progressive telomere shortening results in cumulative transcrip- diminished in K5TRF2/G2 TercϪ/Ϫ and G2 TercϪ/Ϫ cells (Fig. 5B and tome changes affecting 15.1% of all genes (6.7% up-regulated; 8.4% Tables S2–S4). down-regulated) in K5TRF2/G3 TercϪ/Ϫ mice (Table S1). Importantly, Importantly, despite showing slightly elevated TRF2 expression, transcriptome alterations were reproduced in male K5TRF2 keratino- no significant alterations in gene expression were detected in female cytes, ruling out a sex-specific effect on transcriptome alterations K5TRF2 keratinocytes (Fig. S4 and Table S1). In addition, high associated with critically short telomeres (Fig. S4B, Fig. S5 A–J, and TRF2 levels in female K5TRF2/G1–G2 TercϪ/Ϫ keratinocytes did Table S1). We next addressed whether genes altered (FDR Ͻ0.05) in not significantly alter the global transcriptome (Fig. S4 and Table female K5TRF2/G3 TercϪ/Ϫ keratinocytes were also affected in cells S1). These findings indicate that short telomeres and not increased with less severe telomere shortening. Venn diagrams demonstrated a TRF2 expression are responsible for the transcriptome alterations highly overlapping pattern of affected genes (FDR Ͻ0.05) between described here. We conclude that progressive telomere shortening K5TRF2/G3 TercϪ/Ϫ and K5TRF2/G2 TercϪ/Ϫ keratinocytes when is linked with a continuous loss of stringency of global transcrip- compared with wild-type or G2 TercϪ/Ϫ transcriptomes (Fig. 5A). An tional control that affects the expression of a defined set of genes independent set of gene expression-profiling analyses revealed that 24% in a nonstochastic manner.

4of6 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0909265106 Schoeftner et al. Downloaded by guest on September 27, 2021 A K5TRF2/ G3 Terc-/- K5TRF2/G2 Terc-/- -/- -/- -/- vs. K5TRF2/ G3 Terc K5TRF2/ G3 Terc vs. K5TRF2/G3 K5TRF2/G3 Terc vs. m K5TRF2 vs. f K5TRF2/G3 Terc-/- Wild type -/- -/- -/- 296 vs. vs. K5TRF2/G2 Terc Terc vs. G2 Terc m Wild type vs. Wild type Wild type Wild type 504 f Wild type 627 821 926 952 1247 genes 743 1238 genes 426 679 genes 1247 genes 620 860 genes 718 genes up regulated 1222 genes 1248 genes up-regulated up-regulated up-regulated up-regulated up-regulated 117 up-regulated up-regulated 25 98 179 3 398 505 1086 1331 genes 1005 1223 genes 1112 711 1332 424 502 genes down-regulated down-regulated down-regulated 1510 genes 1510 genes 1510 genes 890 genes 1511 genes 78 down-regulated down-regulated 218 down-regualted 219 down-regulated down-regulated

) 7 2 B 6

(log 5 4 K5TRF2/G2 Terc-/- . levels 3 vs Wild type 2 -/- 1 K5TRF2/G3 Terc vs. Fcgr3 4732466D17Rik 1810011O10Rik Fstl1 2810055F11Rik S100a4 A730090H04Rik Slc27a3 Scd2 Aldh18a1 Parp1 Edg8 Rerg Slc16a9 BC029169 Npl Tnc Galnt14 Dmpk Wild type ratio RNA 0 0 NA NA

App -/-

Lnx1 K5TRF2/G3 Terc . Lnx1 -1 vs Blcap Ifngr1 Fkbp5 Fkbp5 Fkbp5 Slc6a6 -/- Ctdsp2 Gpsm1

Sdcbp2 G2 Terc Slc38a2 -2 -/- BC011467 K5TRF2/G3 Terc vs -/- -3 K5TRF2/G2 Terc 2810457I06Rik 2310002A05Rik 2810003C17Rik 5430432H19Rik 9630055N22Rik -4 ratio RNA levelsratio RNA (log)

Fig. 5. Deregulation of the mouse transcriptome is linked to progressive telomere shortening. (A) Venn diagrams showing overlapping gene expression patterns between the indicated pairwise comparisons of transcriptomes. Telomere shortening affects the expression of a similar group of genes between different pairwise comparisons. Red circles, down-regulated genes; green circles, up-regulated genes. (B) Transcript levels ratios of 20 probe sets robustly up-regulated or down-regulated in K5TRF2/G3 TercϪ/Ϫ versus wild-type comparisons. Probe sets are ranked according to the false detection rate (FDR Ͻ1.00E-07; also see Tables S2 and S3). Changes in transcript levels are tendentially reduced when the difference in telomere length between experimental samples is reduced.

K5TRF2/G2 TercϪ/Ϫ show only limited transcriptome changes when compared with wild-type keratinocytes. CELL BIOLOGY

Next, we investigated whether transcriptome alterations were is associated with the development of aging pathologies (6–9). Telo- biased toward particular pathways, which could provide an expla- mere shortening is paralleled by activation of a DDR (4, 5). Recent nation for the premature aging phenotypes in K5TRF2/G2–G3 studies show that oxidative stress as well as mutations in DNA repair TercϪ/Ϫ females and K5TRF2 males. Pathways analysis revealed an pathways provoke epigenetic alterations and transcriptional changes overrepresentation (FDR Ͻ0.25) of genes involved in the mTOR, paralleling those observed in aged animals (13, 22, 32). Here we show Akt, eIF4 and ErbB pathways in K5TRF2/G3 TercϪ/Ϫ females and that accumulation of dysfunctional telomeres, associated with broad to a lesser extent in male K5TRF2 keratinocytes, presumably alterations of the mouse transcriptome and impaired maintenance of reflecting the activation of a cell survival program to escape cell epigenetic silencing during X inactivation. cycle arrest/cellular senescence induced by critically short telomeres Gene expression changes accumulate with progressive telomere (Table S5 A–D). In this regard, we found an underrepresentation shortening and are detected along all chromosomes and are unbiased of genes involved in cell cycle regulation (FDR Ͻ0.25) in female toward gene location, excluding a telomere position effect as a causative K5TRF2/G3 TercϪ/Ϫ and male K5TRF2 keratinocytes (Table S5 event for the observed transcriptome alterations (33, 34). Comparative A–D). Interestingly, despite the presence of an increased DNA transcriptome analyses further revealed that progressive telomere shor- damage load, genes involved in DNA damage repair pathways, such tening gradually affects a defined set of genes driving the up-regulation as base excision, nucleotide excision, and mismatch repair, were of the mTOR, ErbB, and Akt survival pathways and down-regulation underrepresented in female K5TRF2/G3 TercϪ/Ϫ and male K5TRF2 of cell cycle-promoting genes and DNA damage pathways. The ob- keratinocytes (Table S5 A–D). Mutations in components of nucle- served induction of a cellular stress response promoting growth repres- otide excision repair are responsible for the development of pre- sion and cell survival is in line with previous findings in yeast (17–20). mature aging syndromes such as xeroderma pigmentosa, Cockayne In particular, Sir relocation causes gene expression changes involving syndrome, or trichothiodystrophy (31). These findings suggest that the up-regulation of survival genes and stress response factors and the the decreased ability of male K5TRF2 and female K5TRF2/G2–G3 down-regulation of ribosomal biogenesis, thus limiting proliferation and TercϪ/Ϫ keratinocytes to up-regulate DNA damage-response path- promoting cell survival (20, 21). Interestingly, yeast mTOR antagonizes ways in a context of severe telomere dysfunction may contribute to Sir complex relocation upon the induction of cellular stress (20). To- the known hypersensitivity to UV irradiation and premature ap- gether with our current data, these findings support the view of mTOR pearance of skin-aging phenotypes in these mice (23). as a major integrator of stress signals, including severe telomere dys- Interestingly, we found that Ϸ10% of the gene expression changes function. Finally, the activation of survival mechanisms via the mTOR detected in mouse models for aging, such as nucleotide excision and Akt pathways is in line with previous findings from our group repair-deficient mice (Ercc1Ϫ/Ϫ mice vs. wild-type mice) and very old showing that mouse spermatocytes with critically short telomeres up- mice (130 vs. 8 weeks old) (13), overlapped with alterations in regulate the PI3-kinase survival pathway (35). Interestingly, we also K5TRF2/G3 TercϪ/Ϫ females (Fig. S6 and Tables S6–S8). However, this detected an underrepresentation of genes involved in DNA damage overlap was less evident when comparing K5TRF2/G3 TercϪ/Ϫ female repair pathways in cells with critically short telomeres. Conceivably, this keratinocytes with ES cells treated with oxidating agents (22) or with could lead to an impaired response to DNA damage, which is in line the the aging mouse neocortex (22) (Fig. S6 and Tables S6–S8). reported hypersensitivity of these mice to UV irradiation, and could contribute to genomic instability and premature tissue deterioration Discussion (23, 26). Identifying sources of DNA damage associated with aging and analyz- Alterations in epigenetic regulators have been previously re- ing their effect on cellular metabolism is essential to understanding the ported to cause gene expression patterns related to aging (14–16, aging process. Accumulation of dysfunctional, critically short telomeres 22). We show here that DNA damage originating from critically

Schoeftner et al. PNAS Early Edition ͉ 5of6 Downloaded by guest on September 27, 2021 short telomeres causes the re-expression of an X-linked K5TRF2 Telomere Length Analyses on Skin Sections and Skin Keratinocyte. Q-FISH on transgene that is subjected to X chromosome inactivation in cells keratinocytes and tissue sections was carried out as described (23). Nuclei or ϫ with sufficient telomere length. Importantly, loss of silencing is metaphases were captured at 100 magnification using a Leica CTR MIC microscope and a Cohu High-Performance CCD camera. Telomere fluorescence was accompanied by a reduced deposition of H3K27me3 at the Xi and determined as described (40). by a dissociation of TelRNA accumulations (Tacs) from the Xi. Interestingly, Tac-Xi dissociation is also induced by exogenously TRF-Based Telomere Length Analysis of Adult Keratinocytes. Adult epidermal induced DNA damage. Together, this indicates an impact of keratinocytes were isolated as described (23) and included in agarose plugs telomere-originated DNA damage on the organization of chroma- following instructions provided by the manufacturer (Bio-Rad). TRF analysis was tin territories, such as that of the Xi. The failure of transgene performed as described (3). silencing in cells with short telomeres provides further evidence for Isolation and Culture of Keratinocytes from Newborn Mouse Skin. Newborn skin the hypothesis of an age-related decrease in the stability of the keratinocytes were isolated as described (23). X-inactivation mechanism (36, 37). Genome deregulation caused by telomere attrition and telomere ␥ Irradiation of Primary Mouse Embryonic Fibroblast (pMEFs). pMEFs were exposed dysfunction is in line with a current model proposing that the accumu- to3Gyof␥ irradiation, and 30 min and 3 h postirradiation induction of DNA lation of DNA damage causes the deregulation of epigenetic control damage was monitored using an anti-phospho-histone H2AX antibody (1:500; and altered gene expression during organismal aging (13, 22). Consid- Upstate Biotechnology). In parallel, Xist and TelRNA/TERRA RNA FISH stainings were performed. ering that progressive telomere shortening is observed in elderly mice and humans (38, 39), our study indicates that DNA damage checkpoint Western Blotting. Western blots of primary keratinocytes were carried out as activation by dysfunctional telomeres is sufficient to induce epigenetic described (23). Twenty-five micrograms to 70 ␮g were used per condition. A alterations and gene expression changes that antagonize proliferation polyclonal antibody to TRF2 (1:1,000; SF08, provided by E. Gilson, Lyon, France) and promote the activation of survival pathways. These findings support and a monoclonal antibody to ␤-actin (1:10,000; Sigma) were used. the notion that dysfunctional telomeres represent a unique and cumu- RNA Analysis. Northern blot analysis for TelRNA/TERRA was performed as de- lative source of DNA damage, which can promote organismal aging by scribed (30). impairing tissue regeneration. RNA Fluorescence in Situ Hybridizations (RNA FISH). RNA FISH analysis was carried Materials and Methods out as previously described (30). RNA FISH analysis of embryos was performed on Mice. PM K5TRF2 and K5TRF2/G1–G3 TercϪ/Ϫ mice were generated and main- frozen OCT sections (5 ␮m) (30). tained as previously described (23, 26). Microarray Analyses. Total RNA was prepared from adult female tail-skin kera- Histopathology and Immunohistochemistry. Immunohistochemistry was per- tinocytes (23, 35) using the RNAeasy kit (Invitrogen): PM wild-type (n ϭ 4), PM Ϫ/Ϫ Ϫ/Ϫ formed on deparaffinated adult skin sections or frozen OCT sections (E12) using K5TRF2 (n ϭ 4), G2 Terc (n ϭ 2), PM K5TRF2/G2 Terc (n ϭ 3), and PM Ϫ/Ϫ ϭ a mouse monoclonal anti-phospho-histone H2AX antibody (1:500; Upstate Bio- K5TRF2/G3 Terc (n 3). RNA quality was tested using the Agilent 2100 technology), a polyclonal rabbit anti-H3K27me3 antibody (07–449; Upstate Bio- Bioanalyzer (Agilent Technologies). RNA was labeled in a one-color format and hybridized to 44K Whole Mouse Genome Oligo microarrays (G4122F; Agilent technology), and a monoclonal anti-RNA polymerase II antibody (ab5408; Ab- Technologies). Details on gene expression profiling and analysis are in SI Text. cam). Xi-linked H3K27me3 and exclusion of RNA polymerase II from the inactive X (marked by focal H3K27me3) was analyzed by visual appearance using confocal ACKNOWLEDGMENTS. M.A.B.’s laboratory is funded by the Spanish Ministry of microscopy. Images were obtained using a confocal ultra-spectral microscope Science and Technology, the European Union, European Research Council Ad- (TCS-SP2-A-OBS-UV; Leica) or captured at 100ϫ magnification using a Leica CTR vanced Grants, the Spanish Association Against Cancer, and the Ko¨rber European MIC microscope and a Cohu High-Performance CCD camera (Cohu, Inc.). Science Award 2008.

1. Chan SW, Blackburn EH (2002) New ways not to make ends meet: Telomerase, DNA 22. Oberdoerffer P, et al. (2008) SIRT1 redistribution on chromatin promotes genomic stability damage proteins and heterochromatin. Oncogene 21:553–563. but alters gene expression during aging. Cell 135:907–918. 2. de Lange T (2002) Protection of mammalian telomeres. Oncogene 21:532–540. 23. Munoz P, Blanco R, Flores JM, Blasco MA (2005) XPF nuclease-dependent telomere loss and 3. Blasco MA, et al. (1997) Telomere shortening and tumor formation by mouse cells lacking increased DNA damage in mice overexpressing TRF2 result in premature aging and cancer. telomerase RNA. Cell 91:25–34. Nat Genet 37:1063–1071. 4. d’Adda di Fagagna F, et al. (2003) A DNA damage checkpoint response in telomere- 24. Celli GB, de Lange T (2005) DNA processing is not required for ATM-mediated telomere initiated senescence. Nature 426:194–198. damage response after TRF2 deletion. Nat Cell Biol 7:712–718. 5. de Lange T (2005) Shelterin: The protein complex that shapes and safeguards human 25. Smogorzewska A, et al. (2000) Control of human telomere length by TRF1 and TRF2. Mol telomeres. Genes Dev 19:2100–2110. Cell Biol 20:1659–1668. 6. Goytisolo FA, Blasco MA (2002) Many ways to telomere dysfunction: In vivo studies using 26. Blanco R, et al. (2007) Telomerase abrogation dramatically accelerates TRF2-induced mouse models. Oncogene 21:584–591. epithelial carcinogenesis. Genes Dev 21:206–220. 7. Herrera E, et al. (1999) Disease states associated with telomerase deficiency appear earlier 27. Heard E (2005) Delving into the diversity of facultative heterochromatin: The epigenetics in mice with short telomeres. EMBO J 18:2950–2960. of the inactive X chromosome. Curr Opin Genet Dev 15:482–489. 8. Lee HW, et al. (1998) Essential role of mouse telomerase in highly proliferative organs. 28. Payer B, Lee JT (2008) X chromosome dosage compensation: How mammals keep the Nature 392:569–574. balance. Annu Rev Genet 42:733–772. 9. Rudolph KL, et al. (1999) Longevity, stress response, and cancer in aging telomerase- 29. Azzalin CM, et al. (2007) Telomeric repeat containing RNA and RNA surveillance factors at deficient mice. Cell 96:701–712. mammalian chromosome ends. Science 318:798–801. 10. de Boer J, et al. (2002) Premature aging in mice deficient in DNA repair and transcription. 30. Schoeftner S, Blasco MA (2008) Developmentally regulated transcription of mammalian Science 296:1276–1279. telomeres by DNA-dependent RNA polymerase II. Nat Cell Biol 10:228–236. 11. Martin GM (2005) Genetic modulation of senescent phenotypes in Homo sapiens. Cell 31. Garinis GA, van der Horst GT, Vijg J, Hoeijmakers JH (2008) DNA damage and ageing: 120:523–532. New-age ideas for an age-old problem. Nat Cell Biol 10:1241–1247. 12. Martin GM, Oshima J (2000) Lessons from human progeroid syndromes. Nature 408:263–266. 32. Bahar R, et al. (2006) Increased cell-to-cell variation in gene expression in ageing mouse 13. Niedernhofer LJ, et al. (2006) A new progeroid syndrome reveals that genotoxic stress heart. Nature 441:1011–1014. suppresses the somatotroph axis. Nature 444:1038–1043. 33. Baur JA, Zou Y, Shay JW, Wright WE (2001) Telomere position effect in human cells. Science 14. Imai S, Kitano H (1998) Heterochromatin islands and their dynamic reorganization: A 292:2075–2077. hypothesis for three distinctive features of cellular aging. Exp Gerontol 33:555–570. 34. Koering CE, et al. (2002) Human telomeric position effect is determined by chromosomal 15. Vijg J (2004) Impact of genome instability on transcription regulation of aging and context and telomeric chromatin integrity. EMBO Rep 3:1055–1061. senescence. Mech Ageing Dev 125:747–753. 35. Franco S, Canela A, Klatt P, Blasco MA (2005) Effectors of mammalian telomere dysfunction: 16. Villeponteau B (1997) The heterochromatin loss model of aging. Exp Gerontol 32:383–394. A comparative transcriptome analysis using mouse models. Carcinogenesis 26:1613–1626. 17. Martin SG, et al. (1999) Relocalization of telomeric Ku and SIR proteins in response to DNA 36. Bennett-Baker PE, Wilkowski J, Burke DT (2003) Age-associated activation of epigeneti- strand breaks in yeast. Cell 97:621–633. cally repressed genes in the mouse. Genetics 165:2055–2062. 18. McAinsh AD, Scott-Drew S, Murray JA, Jackson SP (1999) DNA damage triggers disruption 37. Wareham KA, Lyon MF, Glenister PH, Williams ED (1987) Age related reactivation of an of telomeric silencing and Mec1p-dependent relocation of Sir3p. Curr Biol 9:963–966. X-linked gene. Nature 327:725–727. 19. Mills KD, Sinclair DA, Guarente L (1999) MEC1-dependent redistribution of the Sir3 38. Flores I, et al. (2008) The longest telomeres: A general signature of adult stem cell silencing protein from telomeres to DNA double-strand breaks. Cell 97:609–620. compartments. Genes Dev 22:654–667. 20. Ai W, et al. (2002) Regulation of subtelomeric silencing during stress response. Mol Cell 39. Harley CB, Futcher AB, Greider CW (1990) Telomeres shorten during ageing of human 10:1295–1305. fibroblasts. Nature 345:458–460. 21. Taddei A, et al. (2009) The functional importance of telomere clustering: Global changes 40. Samper E, et al. (2000) Mammalian Ku86 protein prevents telomeric fusions independently in gene expression result from SIR factor dispersion. Genome Res 19:611–625. of the length of TTAGGG repeats and the G-strand overhang. EMBO Rep 1:244–252.

6of6 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0909265106 Schoeftner et al. Downloaded by guest on September 27, 2021