Cancer is a genetic disease, characterized by the activation of oncogenes and inactivation of tumor suppressors. These changes can be driven by a variety of genomic mechanisms, including protein-altering point mutations, copy number alterations, and promoter hypermethylation. In prostate cancer (PCA), structural genomic rearrangements represent the most common mechanisms that effect the genetic changes leading to oncogenic events. Genomic rearrangements result in cardinal genetic alterations in PCA such as amplification of oncogenes like MYC, androgen-receptor (AR), and PIK3CA, and deletion of tumor suppressors such as PTEN, TP53, and NKX3-1. Oncogenic gene fusions also result from genomic rearrangements in PCA; fusion of the androgen-driven TMPRSS2 gene with the oncogenic transcription factor ERG occurs in nearly 50% of PCA [1,2]. The fact that these events represent the most common genomic abnormalities in PCA reinforces the central role of structural genomic rearrangements in the inactivation of tumor suppressors and activation of oncogenic signaling pathways and their critical importance to the pathogenesis of PCA.
In this review, we examine recently reported data regarding genomic rearrangements in PCA, associations with underlying molecular features, and implications for patent prognosis and clinical decision-making.
Discovery of recurrent ETS gene fusions: a paradigm shift in cancer research
The report of recurrent gene fusions in PCA in 2005 represents a paradigm shift regarding genomic rearrangements in cancer . Prior to this discovery, genomic rearrangements, and particularly translocations, were thought to be largely restricted to hematologic malignancies and sarcomas, playing a minimal role in solid epithelial tumors. We now know that gene fusions between androgen-regulated genes and ETS genes are common, recurrent, and act as drivers of prostate tumorigenesis. In addition, rare fusion events, such as recurrent rearrangements of BRAF [3▪,4] and KRAS , can also act as oncogenic drivers in PCA.
ETS family gene fusions
ETS family fusions are present in about 50% of PCA. In general, these rearrangement events are characterized by the presence of an androgen-regulated gene as the 5′ fusion partner, with an oncogenic transcription factor of the ETS family as the 3′ partner, thus driving overexpression of an oncogenic factor with a tissue-specific promoter. The archetype of this class of genomic rearrangements is the TMPRSS2-ERG fusion, the most common genomic abnormality in PCA. As shown in a recent meta-analysis of over 10 000 men, TMPRSS2-ERG is present in 47% of PCA , and resulting aberrant ERG expression acts as a driver of prostate tumorigenesis in a number of model systems [7,8,9▪,10,11].
Since the initial description of fusion genes involving ETS family transcription factors in PCA, studies have revealed additional fusion events following this paradigm . Androgen-driven 5′ partners include SLC45A3 [13,14], HERPUD1 , and NDRG1 , whereas additional ETS family members acting as downstream 3′ partners include ETV1, ETV4, ETV5, ELK4, and FLI1 [2,17–19]. This repertoire of fusions is analogous to hematopoietic malignancies and sarcomas, in which typically there is one highly recurrent translocation (e.g., TMPRSS2-ERG) and a large number of rare variants of the theme, thus confirming the biologically significant role of ETS transcription factors in PCA carcinogenesis.
Other fusion events
Non-ETS fusion events are also observed in PCA. Some, like rearrangements of genes in the RAF pathway [3▪,4] and the oncogene KRAS , are classic drivers of carcinogenesis, although rearrangements in this category are rare (1 to 2% of cancers). Interestingly, recurrent alterations in the Ras/Raf pathways appear to be restricted to ETS-negative cancers, possibly suggesting an overlap in oncogenic mechanisms. Others, such as rare rearrangements involving CDKN1A, CD9, PIGU, RSRC2, and IKBKB, have less known about their functional importance in PCA, although these genes exhibit biological roles in other cell types . The recurrent nature of these alterations is unknown.
Types of genomic rearrangements in prostate cancer
Structural alterations to the genome can occur at several different scales with a variety of subtypes. Large-scale alterations at the level of entire chromosomes or chromosome arms (often referred to as chromosomal instability) can result in aneuploidy and gross karyotypic abnormalities. Smaller scale events result from imperfect repair of double strand DNA breaks and can manifest as interchromosomal rearrangements, in which different chromosomes exchange material with minimal loss of genetic material, or intrachromosomal rearrangements, occurring within a single chromosome but often resulting in DNA deletions (Fig. 1a–c). Broadly speaking, these general processes result in the translocations, genomic deletions, and amplifications that activate oncogenes and inactivate tumor suppressors in cancer.
In addition to these basic components, complex patterns of genomic rearrangements have been described and observed specifically in PCA. Chromoplexy, composed of complex, coordinated chains of balanced, copy number neutral rearrangements, was described by Baca et al. [21▪▪] in 2013 (Fig. 1d). Although evidence of these chained rearrangements can be found in other tumor types, it appears that this phenomenon is highly prominent in PCA, possibly indicating a tissue-specific mechanism for formation of these events. Furthermore, chained rearrangements are particularly associated with ERG fusion-positive PCAs, again implicating prostate-specific mechanisms. Finally, other complex genomic rearrangements in cancer are associated with a phenomenon known as chromothripsis, a striking pattern of hundreds of clustered rearrangements thought to arise through the shattering and inaccurate reassembly of a single chromosome . Chromothripsis-like patterns have been reported in PCA, particularly in ETS-negative classes [21▪▪,23]. However, true chromothripsis generates clonal alterations from a single catastrophic event . It remains to be established whether the chromothripsis-like complex rearrangements observed in PCA occur as a single event or through a progressive, stepwise mechanism.
Factors predisposed to rearrangements
A variety of mechanisms have been proposed to associate with genomic rearrangements and possibly have a causative relationship. Several studies have noted a relationship of the position of genomic rearrangements with a pattern of localized hypermutation involving highly localized substitution mutations. This hypermutation phenomenon, known as kataegis (from the Greek for ‘thunder’), is characterized primarily by C>T and C>G point mutations, and it has been proposed to be associated with activity of the APOBEC family of cytidine deaminases and subsequent base excision repair activity. Although multiple reports have shown the spatial relationship between regions of kataegis and genomic rearrangements [25–28], the biological underpinnings of this phenomenon, and whether the point mutations have a causative relationship with observed genomic rearrangements, remain unclear.
Telomere dysfunction and genomic instability
Telomeres, and their associated proteins, are specialized structures located at the ends of chromosomes, and they serve to distinguish chromosome ends from DNA breaks, preventing inappropriate activation of the DNA damage response. Genomic instability, through fused chromosome ends, results when the protective function of telomeres is compromised (reviewed in ). In the absence of appropriate cell cycle control, fused chromosomes break randomly during cell division, resulting in unequal distribution of genetic material. This phenomenon, described as ‘fusion–breakage–bridge cycles,’ will continue to propagate with further cell division and will result in multiple genomic aberrations.
In PCA, telomere shortening has been correlated with some of the most common genomic rearrangements in the disease, specifically, abnormalities on chromosome 8, including deletions of 8p and amplification of 8q . In addition, reactivation of telomerase following telomere dysfunction was shown to promote tumor metastasis in a nonmetastatic mouse model of PCA . Notably, telomere length has been proposed as a potential marker associated with metastasis and death from PCA . Many questions regarding the relationship between telomere dysfunction, genomic rearrangements, and PCA outcomes remain to be answered.
Transcriptional activity and genomic rearrangements
It is well established that highly accurate DNA replication and repair processes are critical for genome maintenance. However, transcriptional activity has also emerged as a potential predisposing factor for genomic rearrangements, particularly in PCA. Transcribing DNA to RNA is a highly regulated and complex process involving the coordination of multiple levels of cellular events, including transcription factors and binding sites on the DNA template, alteration of the DNA template, remodeling of the DNA template with topoisomerases, local chromatin remodeling, and processing of the RNA product. Furthermore, the transcriptional and replication machineries are required to share the same DNA template, requiring coordination between these processes to avoid collisions of RNA and DNA polymerases. With errors in these events, transcription therefore has the potential to generate diverse types of genomic rearrangements.
In PCA, accumulating evidence supports transcription as a source of genomic instability predisposed to structural rearrangements. The closed chain pattern of rearrangements frequently observed in whole-genome sequencing of PCA is consistent with the idea that the genomic loci involved in these rearrangements are spatially colocalized prior to the actual translocation events. ‘Transcription factories’ refer to discrete sites in the nucleus in which active RNA polymerases colocalize to coordinate gene expression from multiple loci [33,34]. Consistent with this concept, rearrangement break points from PCA whole-genome sequencing data spatially correlated with chromatin marks of active transcription – interestingly, however, this phenomenon may be restricted to ERG fusion-positive tumors . Furthermore, interchromosomal, chained rearrangements are significantly more pronounced in tumors harboring ETS fusions, and ERG fusions frequently arise in this context [21▪▪]. Finally, using a genome-wide chromosome conformation capture approach, ERG itself has been shown to affect global changes in the topology of chromosome organization, inducing new patterns of spatial associations between genomic loci . Therefore, the temporal dependency of transcriptional activity, ETS fusions, and genomic rearrangements is still unclear. There is evidence for both the hypothesis that ETS fusions cause genomic rearrangements and the hypothesis that they represent consequences of transcriptional-induced rearrangements. The true mechanisms underlying these phenomena may represent a complex combination of these concepts.
Androgen-stimulated transcription has received particular attention as a potential mechanism for the induction of genomic rearrangements in PCA. Androgen signaling induces proximity of the TMPRSS2 and ERG genomic loci (as well as other AR targets), predisposing to the formation of the TMPRSS2-ERG gene fusion [37–39]. In addition, early-onset PCAs (<age 50), which are enriched for ETS fusions [40▪▪,41,42] have been reported to exhibit rearrangement break points spatially localized near AR-binding sites [40▪▪]. These findings suggest that androgen-driven transcription preferentially induces specific types of interchromosomal rearrangements, leading to a cascade of oncogenic molecular events such as recurrent ETS rearrangements.
Taken together, these concepts paint a mechanistic picture of the critical nature of androgen-stimulated transcription in the formation of genomic rearrangements in PCA. Potentially, AR and its cofactors promote formation of highly active transcriptional centers, bringing together ‘transcriptional hubs’, composed of genomic loci widely separated in linear chromosomal space and across different chromosomes, and in doing so form a conducive environment for structural rearrangements.
In contrast ETS-negative PCAs have their own, distinct patterns of genomic rearrangements. In general, these prostate tumors are more heavily enriched in intrachromosomal rearrangements, leading to copy number losses and have specific patterns of SCNAs [21▪▪,43,44]. The importance of these rearrangement patterns, the mechanisms responsible for their differences among different molecular classes of PCA, and how they reflect the underlying biology of these tumors are all areas of active investigation.
RELATIONSHIP OF REARRANGEMENTS TO PROGNOSIS
Defining the clinical impact of genomic alterations in prostate cancer has been complicated by the long natural history of the disease, and the need for large cohorts with long follow-up and extensive molecular annotation. We summarize current knowledge below.
TMPRSS-ERG and association with clinical outcome
Numerous studies have examined the effect of TMPRSS2-ERG on clinical outcome in PCA. No definitive conclusion has been forthcoming because of conflicting data. ERG rearrangement is reported as associated with both more aggressive and more indolent PCA, which may represent differences in patient cohorts and management strategies, the impact of sampling, PCA multifocality molecular heterogeneity, and differences in specific choices of outcomes. This has been reviewed in detail by others . In brief, population-based studies focused on patients diagnosed by TURP (non-PSA screened) and managed with watchful waiting have shown a significant association between ERG status and adverse pathology, metastases, or death from PCA [46,47]. TMPRSS2-ERG is reported to be associated with adverse pathologic features (tumor volume and Gleason grade) in an active surveillance population . The impact of ETS rearrangement on adverse pathologic features or clinical outcome following radical prostatectomy is less clear with several studies reporting association between ETS status and features of aggressive PCA (including increased Gleason grade, stage, or BCR), whereas others report no associations or even association with indolent features (lower Gleason grade or improved recurrence-free survival). Finally, a recent study of patients undergoing treatment with brachytherapy showed that concurrent ERG rearrangement and PTEN deletion, rather than ERG fusion alone, was predictive of poorer recurrence-free survival [49▪].
In summary, population-based studies of watchful waiting cohorts have shown ETS fusions associated with poorer outcomes, whereas retrospective radical prostatectomy series have conflicting results regarding prognosis. A unifying conclusion across studies is made difficult by the variation in techniques to detect ERG rearrangement, variation in patient populations, and lack of PSA screening in some cohorts.
Total burden of genomic rearrangements and clinical outcome
Several studies have reported the ability to classify PCA on the basis of patterns of SCNAs and have attempted to correlate these findings with patient outcome [44,50,51▪▪]. Most recently, the total burden of SCNAs – defined as the percentage of the genome altered by copy number changes, rather than specific deletions or amplifications – was reported to be associated with clinical outcomes. In over 250 clinically localized PCA cases, Hieronymus et al. [51▪▪] showed that increasing numbers of SCNAs was an independent predictor of biochemical recurrence and metastasis after radical prostatectomy, adding predictive information to standard clinicopathologic parameters, including Gleason grade. This finding implies that the total number of genomic rearrangements, and by inference the degree of genomic instability, may be a factor highly predictive of PCA aggressiveness. This concept has been established in other cancer types (breast, ovarian) in which the degree of genomic instability, and particularly chromosomal instability, correlates with aggressive features.
Structural genomic rearrangements play a critical role in the pathogenesis of PCA, resulting in activation of oncogenes, inactivation of tumor suppressors, and production of oncogenic gene fusions. The mechanisms responsible for the predilection of PCA toward these genome changes and the biology underlying the difference in structural rearrangements between molecular classes of PCA represent exciting areas of active investigation. Establishing the clinical significance of these events will require large cohorts, well annotated for both clinical and molecular characteristics, with long-term follow-up.
We thank Erick Herrscher for editorial assistance and evaluation of the manuscript and many investigators whose long-time collaborations have made this work possible.
Financial support and sponsorship
This work was supported by the National Cancer Institute (2R01CA125612-05A1, M.A.R.; 1K08CA187417-01, C.E.B.), a Prostate Cancer Foundation Young Investigator Award (C.E.B.), Urology Care Foundation Research Scholar Award (C.E.B.), and the Frederick J. and Theresa Dow Wallace Fund of the New York Community Trust (C.E.B.).
Conflicts of interest
There are no conflicts of interest.
REFERENCES AND RECOMMENDED READING
Papers of particular interest, published within the annual period of review, have been highlighted as:
- ▪ of special interest
- ▪▪ of outstanding interest
1. Perner S, Demichelis F, Beroukhim R, et al. TMPRSS2:ERG fusion
-associated deletions provide insight into the heterogeneity of prostate cancer. Cancer Res 2006; 66:8337–8341.
2. Tomlins SA, Rhodes DR, Perner S, et al. Recurrent fusion
of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 2005; 310:644–648.
3▪. Beltran H, Yelensky R, Frampton GM, et al. Targeted next-generation sequencing of advanced prostate cancer identifies potential therapeutic targets and disease heterogeneity. Eur Urol 2013; 63:920–926.
Established frequencies of many alterations in CRPC, while highlighting heterogeneity.
4. Palanisamy N, Ateeq B, Kalyana-Sundaram S, et al. Rearrangements of the RAF kinase pathway in prostate cancer, gastric cancer and melanoma. Nat Med 2010; 16:793–798.
5. Wang XS, Shankar S, Dhanasekaran SM, et al. Characterization of KRAS rearrangements in metastatic prostate cancer. Cancer Discov 2011; 1:35–43.
6. Pettersson A, Graff RE, Bauer SR, et al. The TMPRSS2:ERG rearrangement
, ERG expression, and prostate cancer outcomes: a cohort study and meta-analysis. Cancer Epidemiol Biomarkers Prev 2012; 21:1497–1509.
7. King JC, Xu J, Wongvipat J, et al. Cooperativity of TMPRSS2-ERG with PI3-kinase pathway activation in prostate oncogenesis. Nat Genet 2009; 41:524–526.
8. Tomlins SA, Laxman B, Varambally S, et al. Role of the TMPRSS2-ERG gene fusion
in prostate cancer. Neoplasia 2008; 10:177–188.
9▪. Chen Y, Chi P, Rockowitz S, et al. ETS factors reprogram the androgen receptor cistrome and prime prostate tumorigenesis in response to PTEN loss. Nat Med 2013; 19:1023–1029.
Defined ERG as acting cooperatively with AR to coordinately regulate target genes in PCA pathogenesis.
10. Klezovitch O, Risk M, Coleman I, et al. A causal role for ERG in neoplastic transformation of prostate epithelium. Proc Natl Acad Sci U S A 2008; 105:2105–2110.
11. Carver BS, Tran J, Gopalan A, et al. Aberrant ERG expression cooperates with loss of PTEN to promote cancer progression in the prostate. Nat Genet 2009; 41:619–624.
12. Rubin MA, Maher CA, Chinnaiyan AM. Common gene rearrangements in prostate cancer. J Clin Oncol 2011; 29:3659–3668.
13. Esgueva R, Perner S, LaFargue CJ, et al. Prevalence of TMPRSS2–ERG and SLC45A3–ERG gene fusions in a large prostatectomy cohort. Mod Pathol 2010; 23:539–546.
14. Rickman DS, Pflueger D, Moss B, et al. SLC45A3-ELK4 is a novel and frequent erythroblast transformation-specific fusion
transcript in prostate cancer. Cancer Res 2009; 69:2734–2738.
15. Maher CA, Palanisamy N, Brenner JC, et al. Chimeric transcript discovery by paired-end transcriptome sequencing. Proc Natl Acad Sci U S A 2009; 106:12353–12358.
16. Pflueger D, Rickman DS, Sboner A, et al. N-myc downstream regulated gene 1 (NDRG1) is fused to ERG in prostate cancer. Neoplasia 2009; 11:804–811.
17. Paulo P, Barros-Silva JD, Ribeiro FR, et al. FLI1 is a novel ETS transcription factor involved in gene fusions in prostate cancer. Genes Chromosomes Cancer 2012; 51:240–249.
18. Tomlins SA, Mehra R, Rhodes DR, et al. TMPRSS2:ETV4 gene fusions define a third molecular subtype of prostate cancer. Cancer Res 2006; 66:3396–3400.
19. Helgeson BE, Tomlins SA, Shah N, et al. Characterization of TMPRSS2:ETV5 and SLC45A3:ETV5 gene fusions in prostate cancer. Cancer Res 2008; 68:73–80.
20. Pflueger D, Terry S, Sboner A, et al. Discovery of non-ETS gene fusions in human prostate cancer using next-generation RNA sequencing. Genome Res 2011; 21:56–67.
21▪▪. Baca SC, Prandi D, Lawrence MS, et al. Punctuated evolution of prostate cancer genomes. Cell 2013; 153:666–677.
Reported whole-genome sequencing of 57 PCAs, the largest dataset currently. Defined the phenomenon of ‘chromoplexy’ – complex, balanced, chained rerarrangements – as highly prevalent in PCA and associated with specific subtypes.
22. Forment JV, Kaidi A, Jackson SP. Chromothripsis and cancer: causes and consequences of chromosome shattering. Nat Rev Cancer 2012; 12:663–670.
23. Wu C, Wyatt AW, McPherson A, et al. Poly-gene fusion
transcripts and chromothripsis in prostate cancer. Genes Chromosomes Cancer 2012; 51:1144–1153.
24. Korbel JO, Campbell PJ. Criteria for inference of chromothripsis in cancer genomes. Cell 2013; 152:1226–1236.
25. Alexandrov LB, Nik-Zainal S, Wedge DC, et al. Signatures of mutational processes in human cancer. Nature 2013; 500:415–421.
26. Nik-Zainal S, Alexandrov LB, Wedge DC, et al. Mutational processes molding the genomes of 21 breast cancers. Cell 2012; 149:979–993.
27. Davis CF, Ricketts CJ, Wang M, et al. The somatic genomic landscape of chromophobe renal cell carcinoma. Cancer Cell 2014; 26:319–330.
28. Drier Y, Lawrence MS, Carter SL, et al. Somatic rearrangements across cancer reveal classes of samples with distinct patterns of DNA breakage and rearrangement
-induced hypermutability. Genome Res 2013; 23:228–235.
29. O'Sullivan RJ, Karlseder J. Telomeres: protecting chromosomes against genome instability. Nat Rev Mol Cell Biol 2010; 11:171–181.
30. Joshua AM, Shen E, Yoshimoto M, et al. Topographical analysis of telomere length and correlation with genomic instability in whole mount prostatectomies. Prostate 2011; 71:778–790.
31. Ding Z, Wu CJ, Jaskelioff M, et al. Telomerase reactivation following telomere dysfunction yields murine prostate tumors with bone metastases. Cell 2012; 148:896–907.
32. Heaphy CM, Yoon GS, Peskoe SB, et al. Prostate cancer cell telomere length variability and stromal cell telomere length as prognostic markers for metastasis and death. Cancer Discov 2013; 3:1130–1141.
33. Sutherland H, Bickmore WA. Transcription factories: gene expression in unions? Nat Rev Genet 2009; 10:457–466.
34. Osborne CS, Chakalova L, Brown KE, et al. Active genes dynamically colocalize to shared sites of ongoing transcription. Nat Genet 2004; 36:1065–1071.
35. Berger MF, Lawrence MS, Demichelis F, et al. The genomic complexity of primary human prostate cancer. Nature 2011; 470:214–220.
36. Rickman DS, Soong TD, Moss B, et al. Oncogene-mediated alterations in chromatin conformation. Proc Natl Acad Sci U S A 2012; 109:9083–9088.
37. Mani RS, Tomlins SA, Callahan K, et al. Induced chromosomal proximity and gene fusions in prostate cancer. Science 2009; 326:1230.
38. Haffner MC, Aryee MJ, Toubaji A, et al. Androgen-induced TOP2B-mediated double-strand breaks and prostate cancer gene rearrangements. Nat Genet 2010; 42:668–675.
39. Lin C, Yang L, Tanasa B, et al. Nuclear receptor-induced chromosomal proximity and DNA breaks underlie specific translocations in cancer. Cell 2009; 139:1069–1083.
40▪▪. Weischenfeldt J, Simon R, Feuerbach L, et al. Integrative genomic analyses reveal an androgen-driven somatic alteration landscape in early-onset prostate cancer. Cancer Cell 2013; 23:159–170.
Reported whole-genome sequencing of early-onset PCAs, and implicated AR-driven transcription as the critical mechanism for formation of genomic rearrangements, ETS fusions, and pathogenesis of these cancers.
41. Steurer S, Mayer PS, Adam M, et al. TMPRSS2-ERG Fusions Are Strongly Linked to Young Patient Age in Low-grade Prostate Cancer. Eur Urol 2014; [Epub ahead of print].
42. Schaefer G, Mosquera JM, Ramoner R, et al. Distinct ERG rearrangement
prevalence in prostate cancer: higher frequency in young age and in low PSA prostate cancer. Prostate Cancer Prostatic Dis 2013; 16:132–138.
43. Barbieri CE, Baca SC, Lawrence MS, et al. Exome sequencing identifies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nat Genet 2012; 44:685–689.
44. Taylor BS, Schultz N, Hieronymus H, et al. Integrative genomic profiling of human prostate cancer. Cancer Cell 2010; 18:11–22.
45. Tomlins SA, Bjartell A, Chinnaiyan AM, et al. ETS gene fusions in prostate cancer: from discovery to daily clinical practice. Eur Urol 2009; 56:275–286.
46. Attard G, de Bono JS, Clark J, Cooper CS. Studies of TMPRSS2-ERG gene fusions in diagnostic trans-rectal prostate biopsies [author reply]. Clin Cancer Res 2010; 16:1340.
47. Demichelis F, Fall K, Perner S, et al. TMPRSS2:ERG gene fusion
associated with lethal prostate cancer in a watchful waiting cohort. Oncogene 2007; 26:4596–4599.
48. Lin DW, Newcomb LF, Brown EC, et al. Urinary TMPRSS2:ERG and PCA3 in an active surveillance cohort: results from a baseline analysis in the Canary Prostate Active Surveillance Study. Clin Cancer Res 2013; 19:2442–2450.
49▪. Fontugne J, Lee D, Cantaloni C, et al. Recurrent prostate cancer genomic alterations predict response to brachytherapy treatment. Cancer Epidemiol Biomarkers Prev 2014; 23:594–600.
Study higlighting the potential clinical utility of genomic aberrations as predictive markers for specific therapies.
50. Lapointe J, Li C, Giacomini CP, et al. Genomic profiling reveals alternative genetic pathways of prostate tumorigenesis. Cancer Res 2007; 67:8504–8510.
51▪▪. Hieronymus H, Schultz N, Gopalan A, et al. Copy number alteration burden predicts prostate cancer relapse. Proc Natl Acad Sci U S A 2014; 111:11139–11144.