Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=kepi20 Epigenetics ISSN: 1559-2294 (Print) 1559-2308 (Online) Journal homepage: https://www.tandfonline.com/loi/kepi20 Human rDNA copy number is unstable in metastatic breast cancers Virginia Valori, Katalin Tus, Christina Laukaitis, David T. Harris, Lauren LeBeau & Keith A. Maggert To cite this article: Virginia Valori, Katalin Tus, Christina Laukaitis, David T. Harris, Lauren LeBeau & Keith A. Maggert (2019): Human rDNA copy number is unstable in metastatic breast cancers, Epigenetics, DOI: 10.1080/15592294.2019.1649930 To link to this article: https://doi.org/10.1080/15592294.2019.1649930 Accepted author version posted online: 28 Jul 2019. Published online: 12 Aug 2019. Submit your article to this journal Article views: 32 View Crossmark data
Transcript
Full Terms & Conditions of access and use can be found athttps://www.tandfonline.com/action/journalInformation?journalCode=kepi20
Human rDNA copy number is unstable inmetastatic breast cancers
Virginia Valori, Katalin Tus, Christina Laukaitis, David T. Harris, LaurenLeBeau & Keith A. Maggert
To cite this article: Virginia Valori, Katalin Tus, Christina Laukaitis, David T. Harris, Lauren LeBeau& Keith A. Maggert (2019): Human rDNA copy number is unstable in metastatic breast cancers,Epigenetics, DOI: 10.1080/15592294.2019.1649930
To link to this article: https://doi.org/10.1080/15592294.2019.1649930
Accepted author version posted online: 28Jul 2019.Published online: 12 Aug 2019.
Human rDNA copy number is unstable in metastatic breast cancersVirginia Valori a, Katalin Tusb*, Christina Laukaitisc,d, David T. Harrise,f, Lauren LeBeaub, and Keith A. Maggert d,g
aDepartment of Applied Biosciences, University of Arizona, College of Medicine, Tucson, AZ, USA; bDepartment of Pathology, University ofArizona, College of Medicine, Tucson, AZ, USA; cDepartment of Medicine, University of Arizona, College of Medicine, Tucson, AZ, USA;dUniversity of Arizona Cancer Center, University of Arizona, College of Medicine, Tucson, AZ, USA; eDepartment of Immunobiology,University of Arizona, College of Medicine, Tucson, AZ, USA; fArizona Health Sciences Center Biorepository, University of Arizona, College ofMedicine, Tucson, AZ, USA; gDepartment of Cellular and Molecular Medicine, University of Arizona, College of Medicine, Tucson, AZ, USA
ABSTRACTChromatin-mediated silencing, including the formation of heterochromatin, silent chromosometerritories, and repressed gene promoters, acts to stabilize patterns of gene regulation and thephysical structure of the genome. Reduction of chromatin-mediated silencing can result ingenome rearrangements, particularly at intrinsically unstable regions of the genome such astransposons, satellite repeats, and repetitive gene clusters including the rRNA gene clusters(rDNA). It is thus expected that mutational or environmental conditions that compromise hetero-chromatin function might cause genome instability, and diseases associated with decreasedepigenetic stability might exhibit genome changes as part of their aetiology. We find the supportof this hypothesis in invasive ductal breast carcinoma, in which reduced epigenetic silencing hasbeen previously described, by using a facile method to quantify rDNA copy number in biopsiedbreast tumours and pair-matched healthy tissue. We found that rDNA and satellite DNA sequenceshad significant copy number variation – both losses and gains of copies – compared to healthytissue, arguing that these genome rearrangements are common in developing breast cancer.Thus, any proposed aetiology onset or progression of breast cancer should consider alterations tothe epigenome, but must also accommodate concomitant changes to genome sequence atheterochromatic loci.
ARTICLE HISTORYReceived 16 May 2019Revised 7 July 2019Accepted 22 July 2019
KEYWORDSRibosomal DNA (rDNA);invasive breast carcinoma;heterochromatin; repeat;qPCR; copy numberpolymorphism
Introduction
The human genome contains significant amounts ofrepetitive sequences. Many of the most repetitious –the alphoid satellite repeats, Satellites-I, -II, and -III,and the telomeres – may consist of kilobases up tomegabases of nearly identical repeats that, if damaged,may repair using sister chromatids, homologous chro-mosomes, non-homologues, or even repeats in cis asrepair templates [1–8]. Such events may generateinterchromatid crossovers, translocations, acentric/dicentric chromosomes, repeat expansions and con-tractions, and/or extrachromosomal circles.Normally,these genome-damaging repair events are disfavouredby the packaging of repetitive sequences as
constitutive heterochromatin, usually chemically hall-marked by histone H3 lysine 9 methylation,association with the heterochromatin protein HP1a,histone hypoacetylation, and expansive DNA methy-lation [9–11]. Heterochromatin potentiates pairing,regulates repair, and inhibits recombination. Forexample, in Drosophila special repair processes disfa-vour non-allelic crossovers during repair ofdouble-strand breaks in heterochromatin [12,13].Mutations in Drosophila that compromise hetero-chromatin formation allow recombination withinconstitutive centric heterochromatin and at telomeres[14,15], deregulates telomere length [16,17], destabi-lizes repeat gene clusters [18], and derepresses
CONTACT Keith A. Maggert [email protected] University of Arizona Cancer Center, University of Arizona, College of Medicine, Tucson,AZ, 85724, USA
*Present Address: Carolinas Pathology Group, Spartanburg Medical Center, Spartanburg, South Carolina.*Present Address:Spartanburg Regional Healthcare System, SC, USAAuthors’ statementOne of the common hallmarks of cancer is genome instability, including hypermutation and changes to chromosome structure. Using tumourtissues obtained from women with invasive ductal carcinoma, we find that a sensitive area of the genome – the ribosomal DNA gene repeatcluster – shows hypervariability in copy number. The patterns we observe as not consistent with an adaptive loss leading to increased tumourgrowth, but rather we conclude that copy number variation at repeat DNA is a general consequence of reduced heterochromatin function incancer progression.
expression of genes transposed into heterochromaticrepeats [9]. Heterochromatin is considered a primeexample of epigenetic gene regulation because silen-cing and genome stability are stochastic at these loci,variations in the degree of silencing does not corre-spond to genome changes, and the silencing and sta-bilizing effects of heterochromatin are transmittedthrough S-phase [11,19]. Whether such processesexist in other organisms is not yet know, but muta-tions in the DNA methyltransferases of mouse andhumans compromise heterochromatin formation,leading to hypervariability – predominantly loss – ofsatellite copy number, and derepression of transposa-ble elements [20].
In Drosophila, many mutational, developmental,and environmental factors affect heterochromatinstability [21]. Reduction through any of thesemeans may lead to instability of the repeat sequencesreplete in the Drosophila genome [22]. Someexpressed genes are organized as tandem repeatsand, for unknown reasons, are generally subject toepigenetic regulation such that the arrays consist ofinterspersed expressed and non-expressed copiesdespite identical sequence [23–25]. This is perhapsa strategy to maximize both expression and stability,but it renders such genes particularly sensitive to lossand gain in conditions that reduce heterochromatinformation or function as these gene arrays are onlypartially packaged as silenced heterochromatin. Weand others have observed that the 18S/5.8S/28S ribo-somal RNA gene cluster (henceforth referred to asthe ‘45S rDNA,’ from the human nomenclature) isa sensitive locus of copy number changes inDrosophila induced by mutation or by ecologicalconditions [22,26,27]. This is expected to be broadlypleiotropic to a cell since the rDNA not only controlsrRNA production and translational capacity, but therDNA also mediates other processes regulated by thenucleolus [28–31]. These processes are not fullyinvestigated, nor is the roster of roles in regulatingcell-biological responses mediated by the nucleolusfully enumerated. Recently, hints at function inradiation sensitivity and DNA repair, stressresponse, metabolic rate, and developmental deci-sions, have become more concrete. Even vaguernotions of roles, such as in ‘stability’ or ‘heterochro-matin formation’ have been confirmed andexpanded, suggesting that the pleiotropy of rDNAcopy number may expand well-beyond the expected
impacts on protein synthesis and includemanymoreaspects of nuclear function such as telomere main-tenance, transposable element silencing, satelliteDNA stability, and others [32–34].
Heterochromatin formation and regulation isnot well understood in humans, especially duringdisease onset and progression, although there isevidence that some diseases may have heterochro-matin loss and repeat instability as part of theircomplex aetiology [35–40]. Breast cancers, alongwith leukaemias and lymphomas, colorectal can-cers, and others, are known to involve expansivegenome instability, including chromosome aneu-ploidies and rearrangements [41–44]. These can-cers, and others, are expected to show defects inrDNA regulation because of the cytological pre-sence of Argyrophilic Nucleolar Organizers(AgNORs), silver-stained multiple or aberrantnucleoli. It is not known whether AgNORs arethe manifestation of extrachromosomal circlesresulting from damage and incorrect repair of therDNA, as they appear to be in Drosophila. IfAgNORs, rDNA loss, and nucleolar defects sharecommon features, then one may reasonably expectthat AgNORs and disease-related changes torDNA copy number may portend changes in stressresponse, differentiation program, DNA damageresponse, metabolism, chromosome structural sta-bility, epigenetic instabilities, or other nucleolus-related processes.
In breast cancer, de novo mutations and copynumber variations are known to exist but havebeen difficult to quantify or monitor because ofthe heterogeneity of typical tumours in situ, thusmuch of the mutation and copy number muta-tional analyses have been investigated using cellcultures which can be made clonal and grown tolarge numbers [31]. Studies investigating copynumber changes in cancer have so-far analyzedsuch changes in the context of adaptive advantageby the cancer phenotype [1,45], but find them tobe small and variable in scope, and without anyphenotypic consequence. Some cell lines mayshow distinct interline differences in rDNA copynumber, but it is not known whether these existedprior to, or as a consequence of, culturing [46,47].The large variation in natural and presumablyhealthy human rDNA copy number [8] also raisesthe possibility that the ‘aberrant’ copy numbers in
2 V. VALORI ET AL.
these cell lines are merely captured isolates fromwithin natural human population variation [47].Although copy number variation analysis is pro-gressing with the advent of new low-copy-numbersequencing technologies [31], repetitious regionsof the genome remain difficult to analyze.Repeated DNAs including the rDNA continue tobe under-reported in databases or under-investigated in the literature, either because theydo not have a tradition of being considered asmutagenic or capable or regulatory function, orbecause they are refractory to sequencing technol-ogies and assemblies. It is routine to cull thesesequences from databases prior to curation oranalysis, and even those few reported analyseshave not been confirmed as accurate using othermethods [27].
The role of rDNA copy number variation inpopulations or single cells, or any changes insomatic tissues linked to disease risk, onset, orseverity, will remain hypothetical until we havean easy way to ascertain rDNA copy numberfrom small samples. It is with this intention thatthese studies were undertaken. Herein, we describea method to rapidly quantify rDNA copy numberfrom fresh or stored human samples and show it isrobust and sensitive, exceeding other techniquesthat rely on bioinformatic analyses. We employedthis method to characterize rDNA copy number inbreast tumour samples and neighbouring non-tumour cells, discovering that rDNA copy num-bers are discordant in the two samples. Notably, insome tumours the copy number of rDNAdecreases, and in others, it increases. These find-ings challenge the previous common notions thatrDNA copy numbers increase in cancers (toaccommodate increased cell division rates) orthat they decrease (as a consequence of genomeinstability). Rather, we propose that rDNA copynumbers generally become unstable as the stabiliz-ing effects of heterochromatin are compromisedby tumour progression.
Results and discussion
Real-time (quantitative) PCR (qPCR) has been usedeffectively for rDNA copy number determination inDrosophila [26,48]. The benefit of this approach isthat it provides robust and sensitive copy number
determination with small amounts of genomic DNAextracted from fresh or fixed tissue, and at low cost.The preparation of tissue is technically simple andrapidly done. We applied this general approach tohuman tissues, where the sensitivity and low costallowed investigation of rDNA copy number changesin small samples, and from many individuals acrosslarge populations.
Design and validation of real-time quantitativePCR primers for copy number determination
Based on validated primer design for quantifyingrDNA copy number in Drosophila [26], we designedsix primers to amplify three regions of the human45S pre-rRNA transcription unit, two sets directed atsequences in the 18S region (h18S.1 paired withh18S.2, and h18S.3 paired with h18S.4), and one setdirected at the 28S region (h28S.1 and h28S.2)(Figure 1(a)). Primer sets amplified approximately100 base pair regions of the rRNA genes with com-parable annealing temperatures (Table 1). Primersets were verified by amplifying rDNA segmentsfrom whole genomic DNA isolated from blooddrawn from an apparently healthy 46-y-old whitemale with European ancestry and no known familyhistory of cancer. This has been the lab standard fornormalization between samples, experiments, ortechnical replicates.
Target rDNA sequences are under strong selec-tive pressure and are not known to vary betweenindividuals or between repeat units within indivi-duals [49–54]. Without known exception amongeukaryotes, the 18S and 28S rRNA subunits of theribosome are transcribed as a single pre-rRNAtranscript from the 45S rDNA gene, then post-processed in the nucleolus into independent struc-tural ribosomal RNAs (rRNAs). As such, weexpected a strong correlation between the two18S targets’ and between the 18S and 28S targets’copy numbers, as we are aware of no mechanismor rearrangement in any organism that breaksthe fundamental correlation of these two co-transcribed subunit sequences. We validatedrDNA copy number determination across a 100-fold dilution range for both 18S and the 28S targetsequences, centred on 10 nanograms per reaction,which was empirically determined to be the mostrobust in studies of Drosophila rDNA copy
EPIGENETICS 3
a
b
c
d e f
Figure 1. Schematic map of the ribosomal DNA (rDNA) Repeat and validation of the Real-Time/quantitative PCR (qPCR) approachtaken in this study. (a) The 45S rDNA repeat showing how the structure of the 45S primary rRNA transcription unit corresponds to thepost-processed 18S, 5.8S, and 28S rRNA subunits. Locations of the primer sets are indicated. NTS = Non-transcribed Spacer, IGS =Intergenic Spacer, ETS = External transcribed spacer/sequence, ITS = Internal transcribed spacers/sequences. (b) Responsiveness ofqPCR crossing thresholds (Ct) to DNA concentration for three different primer sets in (A) (see also Table 1), including regressioncoefficient m (of y = mx+b) and coefficient of determination (R2) of the lines. Graphs that do not have numerical ordinal values sharean ordinate with other graphs in the same row. (c) Correlations between qPCR amplifications of the subunits of the 45S unit at fourdifferent DNA concentrations (from left to right, 0.1 ng, 1 ng, 10 ng, and 100 ng). (d) Determination of human 18S rDNA copynumber (using 18S.12 primer set) from 10 ng template DNA, by normalization to human tRNAMet gene copy number (see Table 1), inthe presence of different concentrations of competing Drosophila DNA. (e) Determination of human rDNA copy number from variedamounts of template DNA, by normalization to human tRNAMet gene copy number, in the presence of competing Drosophilagenomic DNA such that the total concentration of template DNA was kept constant at 16 ng. Data from (D) are included in thisgraph as the green, yellow, and orange data points. (f) Correlation between copy number determination of enhanced GFP (eGFP)and human Bloom Syndrome helicase (BLM) from cell lines bearing stable integrations of varied copy numbers of a BLM::eGFP fusiontransgene. Throughout this figure, error bars indicate standard error of the mean (S.E.M.) for triplicate or quadruplicate technicalreactions.
4 V. VALORI ET AL.
number [48]. Coefficients of determination of0.94, 0.97, and 0.97, respectively (Figure 1(b)),and coefficients of determination of 1.0, 0.98, and0.99 for pairwise comparisons (Figure 1(c)) con-firmed the robustness of the primer sequences inquantifying rDNA copy number, even from lowDNA concentration samples (1 ng/reaction).These regressions are much stronger than thosereported in other studies using other techniques[27], which is an indicator that qPCR may yieldmore accurate data than high-throughput sequen-cing in determining copy number. The ability toperform multiple reactions also allows us to eval-uate accuracy in copy number determination.Recalculation of regression excluding the 10 nano-gram data points did the least to alter these values,suggesting this concentration is most robust, as inDrosophila. Henceforth, in this study, unlessotherwise indicated, rDNA copy number determi-nations were done with 10 nanograms of genomicDNA per individual reaction, performed at least intriplicate. This amount of template DNA is wellwithin an acceptable range of DNA yield fromdried blood cards (average yield > 500 ng/1 cm2),fresh unspun blood (average yield > 100 µg/200 µLwhole blood), or FFPE sections (average yield ~500 ng/five 10 µm sections). Although we did nottest multiple DNA sources from any single indivi-dual, the consistency of low- and single-copy num-ber genes (see below) across all samples at all
concentrations suggests the source/storage ofDNA does not detectably affect copy numberdetermination.
To compare relative rDNA copy numbersbetween different samples, we designed primersto three tRNA genes. Each is multicopy, but theirdistributed location throughout the genome makesthem relatively stable in copy number provided thegenomes in question are free from overlappingsegmental duplications or deficiencies (i.e., copynumber variations), or from chromosomal aneu-ploidies. To assure that these would be minorsources of error, we also designed primers to mul-tiple single-copy genes amplified in normal cells(Materials and Methods). DNA extracted from thesame male peripheral blood and from unrelatedprimary male human foreskin fibroblast cultureswere tested for copy numbers of each tRNA andeach single-copy gene.
Specificity of amplified and quantified sequenceswere confirmed by analysis of the post-hoc melt-curves for each of the rRNA and tRNA sequences,each primer set producing a single peak in a graph ofthe first-order derivative of fluorescence changewith respect to temperature. In each case, high-concentration (1.7%) agarose gel electrophoresiswas performed and a band of the expected size wasthe only visible PCR product. Sanger sequencing offive individual reactions further confirmed that thepopulation of amplified DNA was homogenous and
Table 1. Characteristics of real-time PCR primers in this study.Primer Name Locationa Sequence (5ʹ-3ʹ) TMelt (°C)
aLocation refers to the beginning of the alignment of the first nucleotide of the primer with the primary pre-rRNA transcripts (35S/45S or 5S). For thetRNA gene, the location is relative to the mature tRNA.
bPrefix ‘h’ indicates a primer specific to human rDNA, ‘d’ indicates Drosophila.crc(N) indicates that the first nucleotide of the primer aligns with the indicated nucleotide (N) of the reverse-complement of the primary transcript.dThis primer set detects the initiator Met isotype (tRNA-iMet-CAT-1–1 through tRNA-iMet-CAT-1–8, and likely also tRNA-iMet-CAT-2–1).
EPIGENETICS 5
corresponded to the desired target sequence. In ourhands, the primer sets directed at tRNAMet and18S.12 performed the best, so unless otherwise indi-cated these are used for the remainder of the work.
We challenged quantitative amplification ofhuman rDNA (and tRNA gene) in two ways. First,we held constant the concentration of human DNAand challenged it with no, 10 nanograms (equivalentmass of DNA), and 100 nanograms (10-fold excess) ofDrosophila genomicDNA.Given the approximate 10-fold larger genome size of humans than ofDrosophila,the last condition represents an approximate 100gene-molar excess of the single-copy genes.Although rDNA copy numbers vary within bothhumans and Drosophila, they can be consideredapproximately equal in copy number in broad popu-lations [8]. Comparable sequences between these twospecies are less than perfectly complementary (95%for the 18S.2 primer, 91% and 64% for the 18S.34 set,and 89% for the 28S.1 primer, Table 1), despite beingin conserved regions of the rRNAs. Even with molarexcess of Drosophila genomic rDNA target, there wasno detectable difference in human rDNA quantifica-tion (Figure 1(d)); the coefficient of determinationwas very high, and the regression coefficient was near-zero (m = −0.003), suggesting that competingDrosophilaDNA had nomeasurable impact on quan-tification of human rDNA copy number.
Second, we challenged human rDNA amplificationby titrating the human DNA to correspond withincreased titer of Drosophila DNA, keeping the totalamount of genomic DNA constant at 16 nanograms.In this configuration, the molar-competition spansfrom 0 to 150-fold gene-molar excess but at a 10-foldlower DNA concentration than the previous experi-ment, and again we detected no difference in rDNAcopy number relative to tRNA normalization (Figure1(e), R2 = 0.98). Based on these two competitiveexperiments, we conclude that the quantification ofrDNA copy number in humans is remarkably robustto DNA concentration, even when in competitionwith vast excess of a homologous ‘contaminating’animal DNA. We expect that contamination byorganisms more-diverged than animals (e.g., bacteria,fungi) would be of negligible concern in routinelaboratory applications.
A similar analysis was performed using rDNAand tRNA gene copy number with three single-copy genes. Without exception, copy numbers of
rDNA and tRNA relative to these ‘denominator’copy numbers were consistent across the range ofconcentrations used above (R2 for Rnmt2 (Gene ID:1787) was 0.98 over the same concentration rangeshown in Figure 1(b), R2 for Snail2 (Gene ID: 6591)was 0.99, R2 for Bloom Helicase (Gene ID: 641) was0.98). Finally, we confirmed the sensitivity androbustness of relative copy number determinationof the last gene by using a series of establishedBloom Syndrome cell line clones (all derived froma precursor GM08505 strain), each containing stablytransformed (i.e., integrated) BLM-GFP fusiongenes. Eight sub-lines have different numbers andloci of integrations, but BLM and GFP sequencesalways co-varied (R2 = 0.97, Figure 1(f)).
It is critical to emphasize that absolute copy num-ber comparisons of rDNA, tRNA gene, and uniquegenes are not possible. This is the unavoidable expec-tation, as copy number is derived from qPCR reac-tion crossing thresholds (Cts), which are affected byprimer sequence, annealing kinetics and tempera-ture, the length and sequence composition of theinter-primer sequence, subtle sequence biases inSYBR binding, etc. For this reason, while compari-sons of relative copy number between samples orconditions are valid, absolute determination ofcopy number of one gene relative to another in anyone sample is certainly not. This is clear from thedata of Figure 1(f), where the Y-intercept of a known1-to-1 correlation is not zero (b = 1.01).
Relative amplification of single-copy and tRNAgenes were compared between male blood andgenomic DNA extracted from tissue sectionsfrom ovaries of three healthy human women.tRNA/single-copy genes did not vary, althoughrDNA, telomeric repeats, and satellites did(Figure 2(a)). This situation is expected given theknown variation in these repetitive sequencesbetween individuals in the population. Threegenomic DNA preparations from three sequentialsections of the same ovarian tissue showed rDNArepeat determinations to be robust across technicalreplicates (Figure 2(b)).
Epigenetic instabilities in breast cancer –AgNOR to rDNA loss?
Breast cancer is among one of the cancers with clearcytological manifestations of nucleolar instability
6 V. VALORI ET AL.
[55–58]. The presence of supernumerary argyrophi-lic nucleolar organizing regions (AgNORs) is inter-preted as either fragmentation of nucleoli or thederepression of inactive rDNA arrays (or both). InDrosophila, presence of supernumerary nucleoli cor-relates with rDNA excision to create extrachromoso-mal circles of rDNA genes [5], likely a result ofderepression, damage, and repair from templatecopies in cis [22,59]. Further, the appearance ofsupernumerary or ‘fragmented’ nucleoli correlateswith the severity of rDNA loss, as is expected froma simple excision of acentric rDNA extrachromoso-mal circles followed by a cell division [33]. WhetherAgNORs present in breast cancer tumours corre-spond to rDNA copy number loss in those tumoursis not known.
We obtained 29 samples of breast cancer tumours(Table 2). Tumours were from invasive ductal
carcinomas with evidence for lymph node metastases,and were of various genetic subtypes – EstrogenReceptor positive or negative, Progesterone Receptorpositive or negative, HER2 over-expression or not(typed on the 0–3+ scale, where 1–2 were taken asnegative), and of varied Ki-67 scores; two were ‘triple-negative’ and had Ki-67 fractions of 90%. In each case,formalin-fixed paraffin-embedded tumour tissueswere obtained and two sequential 10 µm slicesmade. The first was haemotoxilin-eosin stained todefine tumour tissue and healthy marginal tissuefrom the same patient sample. We attempted to nor-malize areas on the slides with similar cell numbers,avoiding connective tissue and adipose when possible.Tumour and non-tumour areas weremarked, and thecorresponding areas on the unstained second sectionwere scraped and DNA isolated using filter-bindingafter xylenes extraction. For each paired sample, we
a
b
Figure 2. Variation in repeat DNA copy number in different individuals. (a) Variation in copy numbers of single-copy genes (BLM andSNA), three rDNA targets (18S, 28S, and 5S), telomeric repeats, and satellite-III, relative to tRNAMet. Blue data are from the laboratorystandard DNA, and are by definition set to unity. Green, yellow, and red datasets are from three different individual ovaries obtainedfrom the Univeristy of Arizona Tissue Acquisition and Cellular/Molecular Analysis Shared Resource without any identifying informa-tion. (b) Data obtained from three consecutive sections of one individual ovary. ‘Std gDNA’ is the laboratory standard humangenomic DNA from peripheral blood. Throughout this figure, error bars indicate standard error of the mean (S.E.M.) for triplicate orquadruplicate technical reactions. S.E.M. from the laboratory standard is pooled into the data from the other individuals’ values.
EPIGENETICS 7
determined tRNAM-AUG gene, 18S, 28S, and 5S rDNAcopy numbers. Throughout our samples, whentumour and non-tumour samples are independentlyanalyzed, the 18S and 28S remained correlated,further indicating bona fide changes in 45S rDNAcopy number and not artefactual vagaries of DNAextraction or the qPCR reactions.
By analyzing 28S/tRNA as a function of 18S/tRNAseparately for both tumour (Figure 3(a), red) andnon-tumour (blue) samples, we could conclude thatthe expected 28S-18S correlation exists and is of equalslope (regression coefficients of 1.83 for tumour and1.66 for non-tumour, both statistically indistinguish-able from the slope of 1.5 representing linearity, as inFigure 1(c)) and correlation (R2 = 0.74 for tumourand 0.53 for non-tumour) for both types of tissues.The difference in coefficients of determination islikely due to a larger experimental error in thetumour-derived samples. Thus, we conclude that the
overall rDNA array structure remains unaffected intumours, suggesting that any changes in copy numberwould involve addition or subtraction of whole 45SrRNA genes, rather than amplification/loss of the 18Sor 28S independently. We also detected that the rela-tive rDNA copy numbers span the same ranges, withno strong evidence for any bias in the spread of valuesin tumour or non-tumour samples (Figure 3(b)). Butit is likely that the high variation in copy numbersbetween individuals might obscure any differences inrDNA copy numbers within individuals as a result ofthe disease.
To more sensitively detect changes in rDNA copynumber, we analyzed the data from Figure 3(a) aspaired (tumour and non-tumour from the same indi-vidual biopsy) samples.We considered 5S rDNA copynumber changes by plotting the difference betweentumour 5S/tRNA and non-tumour 5S/tRNA (5ST/tRNAT – 5SnT/tRNAnT) as a function of the
Table 2. Demographics of breast samples.Sample IDa Ageb Raceb Stage EST-Rc PROG-R3 Her2c Ki-67c
1 42 White invasive ductal carcinoma + 1 Lymph. metastasis NEG NEG 1+ 0.92 66 Asian invasive ductal carcinoma + 1 Lymph. metastasis3 79 White Multiple foci of invasive ductal carcinoma + 6/19 Lymph. metastasis POS NEG 1+ 0.024 47 White invasive ductal carcinoma + 1/23 Lymph. metastasis POS POS 0+ 0.255 57 White invasive ductal carcinoma + 2/15 Lymph. metastasis POS POS 0+ 0.26 58 White invasive ductal carcinoma + 5/19 Lymph. metastasis POS POS 0+ 0.37 59 White invasive ductal carcinoma + 2/10 Lymph. metastasis POS POS8 62 Other infiltrative ductal carcinoma + 2/10 Lymph. metastasis POS POS 1+ 0.19 62 Other infiltrative ductal carcinoma + 2/10 Lymph. metastasis POS POS 1+ 0.210 56 White invasive ductal carcinoma + 1 Lymph. metastasis POS POS 0+ 0.111 85 White invasive ductal carcinoma12 50 White invasive ductal carcinoma + 4/22 Lymph. metastasis POS POS 0+ 0.1513 62 White invasive ductal and lobular carcinoma + no Lymph. metastasis POS NEG 1+ 0.714 64 White invasive ductal carcinoma + 1 Lymph. metastasis POS POS 2+ 0.515 52 American
a No identifying information is presented here; Sample ID refers to a serial nomenclature for this study alone, and connection of these numbers tode-identified University of Arizona College of Medicine sample numbers are curated in the laboratory of Keith A. Maggert.
b Self-identified.c EST-R = Estrogen Receptor; PROG-R = Progesterone Receptor; Her-2 = Her-2, using the 0–3+ scale, with 0–1 taken as negative and 2–3 taken aspositive; Ki67 = fraction of cells positive for Ki-67.
8 V. VALORI ET AL.
differences between 18S rDNA copy number changes(18ST/tRNAT – 18SnT/tRNAnT) (Figure 3(c)) or 28SrDNA copy number changes (Figure 3(d)). In both
cases, we could clearly detect individuals with alteredrDNA copy numbers (both 18S/28S and 5S) intumours compared to non-tumours as those data
a b
c d
e
Figure 3. Comparisons of rDNA copy number changes in breast cancer tumours. (a) Correlation of 18S and 28S rDNA copy numbers isretained in both tumour (red) and non-tumour (blue) samples. (b) Data from (A) projected into one dimension to visualize thespread of individual data points. Black lines show the means. ‘nT’ = non-tumour, ‘T’ = tumour. The ordinal scale is shared with (a). (c)Plot of differences in 18S and in 5S rDNA copy numbers between tumour and non-tumour samples from the same individual. Blackheavy lines highlight 0,0 origin, indicative of no change between 5S and/or 18S copy number between tumour and non-tumour;deviation from the origin is indicative of differences in one or both copy numbers. (d) As in (c), but comparing differences in 28S and5S rDNA copy numbers. (e) Replotting of the absolute value of the data from 3C and 3D to highlight the discordance in the extent ofchanges to the 5S and 45S rDNA copy numbers. Throughout this figure, error bars indicate standard error of the mean (S.E.M.) fortriplicate or quadruplicate technical reactions.
EPIGENETICS 9
that deviated from the 0,0 origin. In about half, a cleardifference in rDNA copy number was detectablebetween tumour and non-tumour tissue from thesame patient. For example, in 10 (of 29) samples, thecopy number of 28S rDNA was larger in the tumourrelative to the non-tumour, with an average gain of26% and a population deviation of 15%, and in 6 (of29) it was smaller, with an average loss of 22% ± 11%.For the 5S, 10 samples showed an increase in rDNAcopy number (21%± 9%), and four showed a decrease(16% ± 9%). As expected from Figure 3(a), differencesin 18S and 28S copy number retained their linearrelationships with R2 = 0.83 and a regression coeffi-cient of 1.84.
We plotted the absolute value of differences in5S as a function of differences in 18S or 28S(Figure 3(e)) to demonstrate that changes in thecopy numbers of rRNA genes in these two clustersare not themselves correlated positively or nega-tively (R2 of 0.02 and 0.04, respectively). Thus, it isequally likely in any given tumour sample with anincrease in 45S rDNA copy number to have anincrease or a decrease in 5S copy number, anddecreases in 45S are not enriched for eitherincreases or decreases in 5S. These results suggestthat the changes to the 45S and 5S rDNA clustersare independent, and the degrees of changes areuncorrelated. Our finding adds illuminating detailto a previous report [45], which showed increasesin 5S and decreases in 45S copy numbers in multi-ple cancers (including breast cancer) but did notanalyze co-relation in the same individual. Wedetect hyper-variability rather than uniformincreases or decreases, indicating that the reportedcorrelation between 45S and 5S rDNA copy num-bers [27] is regulated in a way that is ineffective inbreast cancer tissues. This possibility would bea striking departure from the expected biology ofthe 45S and 5S rDNA concerted copy numbermaintenance, and might serve as a powerful diag-nostic for breast cancer onset or progression.
That some samples had higher rDNA copy num-bers, and others had lower, suggests that rDNA arraysare subjected to general instability with losses andgains both occurring. This argues against develop-mental differences, since it seems likely that develop-mentally programmed changes to rDNA copynumber would be uniform in direction, if not indirection and degree. Similarly, it seems unlikely that
selective pressures would enrich for tumours withboth increases and decreases in copy number ifthere is a growth (or cancer) advantage to either lossesor gains in rDNA copy number. Instead, we find thesedata most easy to reconcile as a result of destabiliza-tion of repeat copy number in general. This assertion,and our data, are consistent with those recently pub-lished by Xu and colleagues [31], and by Wang andLemos [45], although both of those studies analyzetheir data in the context of a cancer adaptive pheno-type for rDNA copy number changes. These research-ers showed that in tumours or samples derived fromleukaemias and lymphomas, medulloblastomas,osteosarcomas, and oesophagal adenocarcinomas,although the average rDNA copy number wasreduced, there were clear cases of individuals withincreased copy numbers.
Those studies reported much larger changes incopy number than did we, but some of the data inthose studies indicate copy numbers that weresurprisingly low (in some cases, fewer than fivecopies of the 5S or widely discordant 18S, 5.8S,and 28S copy numbers), indicating their approachmay not have generated interpretable absolutecopy numbers (see below).
Despite the heterogeneity in genetic subtypes(Table 2), we could detect no correlation betweenrDNA variability and any of the genetic markers ascategorical variables, or with Ki-67 as a continuousvariable (Figure 4). We conclude from this that it isunlikely that these factors, or the known genomeinstability in triple-negative breast cancers, are driv-ing rDNA instability. Rather, we favour the interpre-tation that the universal loss of heterochromatinfunction in breast cancers is the cause.
Variations in telomeres and satellite repeats
The rDNA repeats are regulated and stabilized by theformation of heterochromatin [25]. We expectedthat if the underpinning defect in cancer cells is toheterochromatin function, rather than specific regu-lation of the rDNA, then other repeat-sequenceswould also be affected. To test this, we analyzed‘first-order’ (repeat-to-tRNA) changes in other repeatDNA sequences: the telomeric repeats and satellite-III sequences. The former are simple and non-transcribed, and altered heterochromatin structuresat chromosome ends contributes to telomere length
10 V. VALORI ET AL.
misregulation and allows alternative lengthening oftelomeres [60,61] or de novo capping by heterolo-gous sequences [15]. The latter are stress-responsivetranscribed pericentric repeats [62–64].
We analyzed the telomeric repeats using anapproach modified [26] from Richard Cawthon[65,66]. Telomere repeat copy number in tumourDNA was plotted as a function of copy number innon-Tumour DNA from the same biopsy (Figure 5(a)). It is clear that deviation from the diagonal
(dashed line) is statistically distinguishable from theregression line, but is moderate-to-non-existent inextent (uniformly less than 2%, dotted lines). Thecoefficient of determination is very low (R2 = 0.33),and even these very-small effects are attributable inpart to individual variation, rather than the tumourphenotype. Correspondingly, the coefficients of varia-tions (C.V.s) do not differ, C.V. is 0.8% for the non-tumour samples, and 0.6% for the tumour samples.These results are consistent with the relatively small
a b
c d
e
Figure 4. 18S and 5S rDNA copy number changes as a function of breast cancer genetic subtype. (a) Differences in 18S rDNA copynumbers between paired tumour and non-tumour samples as a function of categorical grouping of Estrogen Receptor Negative (ER-)and Positive (ER+), and Progesterone Receptor (PR). Data from Figure 3(c). (b) As in (a), but with 5S rDNA. (c) As in (a), butcategorically grouped by Her2 expression phenotype. Grades 0–1+ were called ‘negative’ and 2–3+ were called ‘positive.’ (d) as in(e), but with 5S rDNA. (e) Differences in 18S (blue) and 5S (green) rDNA copy numbers between paired tumour and non-tumoursamples as a function of Ki-67 expression, R2 (18S) = 0.003, R2 (5S) = 0.001. Data and S.E.M. are from Figure 3(c).
EPIGENETICS 11
changes in telomere length in breast cancers, despitethe common reactivation of telomerase in this cancertype [67–70].
For the satellite-III DNA, we used techniquesdesigned for Drosophila satellite sequences [26].Three trends were evident upon analysis (Figure 5(b)). First, tumours exhibit increased variationwithin the population of 29 individuals analyzedhere. Specifically, the coefficient of variation nearlydoubles in tumours (C.V. is 4.3% for non-tumours and 8.3% for tumours). Second, in general,tumours possess more Sat-III copies than do non-tumour tissues from the same individual (range of1.04 to 1.24 for non-tumours, and range of 1.00 to1.41 for tumours). Third, the average Sat-III copynumbers were unchanged in tumours compared tonon-tumours when taken as populations (mean was1.14 for non-Tumour samples and 1.17 for tumoursamples). These latter two points seem contradictory,highlighting one of the enduring problems of analyz-ing repeat copy number changes in large populations,namely that copy number is so variable between indi-viduals that even significant changes within indivi-duals are often shrouded when (even small)populations are analyzed (e.g., [45]). But it is clearfrom analyzing data from individuals that there arejust as many decreases in Sat-III number as increases,and that the increases are larger in scale (+7% ± 7%,
range of 1% to 25%) than are the decreases (−4%±4%,range of −1% to −13%).
We interpret these data as further support of ourinterpretation that the proximate cause of genomeinstability at repeats is a reduction in epigeneticstability of repeat sequences in general. And, aswith the rDNA, there was no clear trend towardgains or losses of repeat number, merely the occur-rence of increased variability. While this may ulti-mately derive from loss of the rDNA itself [71], thealternative that a proximate cause of epigeneticinstability affecting all repeats lies elsewhere cannotbe ruled out. These results indicate that repeat DNAin general, including rDNA specifically, are unstablein progressing breast cancer cells.
Assessing whether 45S-5S is a risk factor forbreast cancer
Blood is of mesodermal origin, and most of the cellsin the breast cancer tumours are of endodermalorigin, so these two cell types last shareda common cellular ancestor prior to gastrulation.This allowed us to probe whether we could detectrDNA copy number polymorphisms that prefigurelater epigenetic instability, as ametric for early risk oflater development of breast cancer. This hypothesisderives credibility from the demonstration that
a b
Figure 5. Comparison of telomere and satellite repeats in paired tumours and non-tumour tissues. (a) Very weak correlationbetween telomeric repeat copy number in tumours versus non-tumours from the same individuals. Mean copy numbers areindicated (mean for non-tumours, vertical blue line, is 0.007; mean for tumours, horizontal red line, is 0.008). X- and Y-axis values areper cent difference from the lab standard; dashed diagonal is no difference between tumour and non-tumour (y = x), and dotteddiagonals demarcate a 2% difference (y = 1x ± 0.02). (b) No correlation between Satellite-III in tumours versus non-tumours from thesame individuals. Throughout this figure, error bars indicate standard error of the mean (S.E.M.) for triplicate or quadruplicatetechnical reactions.
12 V. VALORI ET AL.
naturally and experimentally low rDNA copy num-ber at fertilization result in adult epigenetic instabil-ities in model systems, and the finding that loss ofrDNA leads to cell-autonomous defects in epigeneticstability [33]. Blood from the patients from whichbreast cancer tumours were derived was not avail-able, but the commonality of response (differencesbetween tumour and non-tumour) suggested that ifrDNA copy number detects prefigured breast cancerin individuals, it might be detected in blood. Wetherefore obtained genomic DNA from 51 patientsinvolved in a clinical study designed to ascertainbreast cancer risk alleles; about half were diagnosedwith breast cancer, and none were from the sameimmediate families. We screened these samples forrDNA copy number in the attempt to find if uni-formly low, uniformly high, or notably altered 45S-to-5S ratio of rDNA copy number was correlated withcancer diagnosis.
The known broad variance in rDNA copy num-ber, both 18S and 28S (Figure 6(a)), argued againstthis simple hypothesis, and against the possibilitythat rDNA copy number could be used as a pre-clinical screen for cancer risk. However, we diddiscover that the 5S copy number in all of thestudy participants was low and did not varymuch between patients (Figure 6(b), note themuch smaller range of the abscissa compared tothe ordinate), and showed no detectable correla-tion with 45S copy number (R2 is 0.08 for patientswith a cancer diagnosis, and is 0.03 for non-diagnosed patients). It is of some debate whether45S and 5S copy numbers are equally variable.Analysis of few family lineages has demonstratedthat the copy number of 5S appears more stablethan the 45S [3], however the published correla-tion between 45S and 5S copy number in healthypeople is strong and robust even between differentracial groups worldwide, and under experimentalperturbation [27], suggesting that both copy num-bers must be variable. The latter studies reportedan approximately 10-fold variance in 45S and 40-fold variance in 5S rDNA copy numbers inEuropean men and women (135 of 201 of oursamples are non-hispanic white women), whilewe found only a 1.6-fold variance in the medial90% of samples.
It is formally possible that the lack of 45S-5Scorrelation in our samples is itself a pre-breast
cancer indicator, and reflects an instability andloss of 5S sequences in women at risk for breastcancer development. However, our results andthose of Gibbons and colleagues [27] were derivedfrom two different methodologies. Ours was rDNAcopy number ascertainment from qPCR (relativeto tRNA gene copy number), while the correlationreported by Gibbons and colleagues was derivedfrom bioinformatic analysis of high throughputDNA sequencing, the real-time PCR validationdata for which were not published. Therefore, thedifference between 45S-5S correlation in healthyand breast cancer patients could be a bona fidebiomarker, or it could be an artefact of eithermethod of determination.
Quantification of rDNA copy number in normalhuman blood
To address the correlation between 45S and 5SrDNA copy numbers in individuals not selected tohave a history of breast cancer, we obtained 201blood samples from the Arizona Health SciencesCenter Biorepository at the University of Arizona(Table 3). Too few healthy people donated blood forour needs, so we instead obtained blood from peo-ple of a broad range of age and both sexes whoseblood was drawn as part of diagnosis of non-cancerafflictions. The blood was prepared as before, and45S and 5S were determined relative to tRNAMet. Inthose samples, we observed the same lack of correla-tion between 45S and 5S copy number (Figure 6(c)).This adequately refutes our hypothesis that rDNAcopy number differences could prefigure breast can-cer development late in life, and further refutes thepossibility that breast cancer development disruptsa natural coupling of 45S and 5S copy number.However, it does not identify the origin of the dis-parity between the data of Gibbons and colleaguesand our own: it could be that informatic calculationof rDNA copy number is error-prone, or it could bethat qPCR determination is insensitive, particularlywith respect to the 5S. We disfavour the latter pos-sibility as we see no evidence of such when quanti-fying 5S copy number over a range of genomic DNAconcentrations (Figure 6(d)).
To resolve this conflict, we selected blood sampleswith large and small 5S rDNA counts and subjectedthem to Southern dot-blot analysis. We observed
EPIGENETICS 13
a very strong correlation between Southern-basedcopy number determination and the results fromqPCR across a broad range (Figure 6(e)). This is in
good agreement with similar experiments inDrosophila, where very small differences in 18SrDNA copy number by qPCR could be robustly
a b
c
de
Figure 6. Comparisons of 18S rDNA and 5S rDNA in blood. (a) The correlation between 28S and 18S copy number is retained in blood samplestaken from both women diagnosed with (red) or not diagnosed with breast cancer (blue). (b) The correlation between 5S and 18S rDNA copynumbers reported in [27] is not found in our dataset (no diagnosis, blue, R2 = 0.03; positive diagnosis, red, R2 = 0.08). (a) and (b) share anordinate, error bars are S.E.M., and data are copy numbers relative to tRNAMet. (c) Extremely weak correlation between 18S rDNA copy numberand 5S copy number in whole blood taken from people with no indication of any cancer diagnosis. Data are presented without error bars andwith censoring of the highest values for clarity, but all data are present in the inset graph. (d) Template dose response for the 5S rDNA primers,as in Figure 1B. (e) Comparison of 5S rDNA copy number determined by qPCR and Southern blot quantification.
14 V. VALORI ET AL.
Table 3. Demographics of blood samples.Sample IDa Ageb Sex Ethnicityb
B1-2 62 F Not Hispanic or LatinoB1-4 34 F UnknownB1-9 74 F Not Hispanic or LatinoB2-1 62 F Not Hispanic or LatinoB2-2 78 F Not Hispanic or LatinoB2-3 47 F Hispanic or LatinoB2-4 64 F Not Hispanic or LatinoB2-5 50 F Hispanic or LatinoB2-6 57 F Hispanic or LatinoB2-7 48 F Not Hispanic or LatinoB2-8 48 F Not Hispanic or LatinoB2-9 70 F Not Hispanic or LatinoB2-10 48 F Not Hispanic or LatinoB2-11 65 F Not Hispanic or LatinoB2-14 59 F Hispanic or LatinoB2-15 62 F Not Hispanic or LatinoB2-16 62 F Not Hispanic or LatinoB2-17 85 F Not Hispanic or LatinoB2-18 80 F Not Hispanic or LatinoB2-20 62 F Hispanic or LatinoB3-3 75 F Hispanic or LatinoB3-4 84 F Not Hispanic or LatinoB3-5 69 F Not Hispanic or LatinoB3-6 67 F Hispanic or LatinoB3-7 80 F Not Hispanic or LatinoB3-8 44 F Not Hispanic or LatinoB3-9 68 F Hispanic or LatinoB3-13 61 F Not Hispanic or LatinoB3-14 55 F Hispanic or LatinoB3-16 48 F Hispanic or LatinoB3-17 62 F UnknownB3-18 68 F Not Hispanic or LatinoB3-19 48 F Hispanic or LatinoB4-1 70 F Not Hispanic or LatinoB4-2 77 F Not Hispanic or LatinoB4-3 54 F Not Hispanic or LatinoB4-5 77 F Not Hispanic or LatinoB4-9 68 F Not Hispanic or LatinoB4-10 68 F Not Hispanic or LatinoB5-1 68 F Not Hispanic or LatinoB5-2 68 F Not Hispanic or LatinoB5-3 68 F Not Hispanic or LatinoB5-4 77 F Not Hispanic or LatinoB5-6 66 F Not Hispanic or LatinoB5-7 59 F Hispanic or LatinoB5-8 77 F Not Hispanic or LatinoB5-9 51 F Not Hispanic or LatinoB5-10 51 F Not Hispanic or LatinoB5-11 51 F Not Hispanic or LatinoB5-12 73 F Not Hispanic or LatinoB5-13 73 F Not Hispanic or LatinoB5-15 73 F Not Hispanic or LatinoB5-16 50 F Not Hispanic or LatinoB5-17 73 F Not Hispanic or LatinoB5-20 65 F Not Hispanic or LatinoB6-1 56 F Not Hispanic or LatinoB6-4 71 F Not Hispanic or LatinoB6-5 62 F Not Hispanic or LatinoB6-6 76 F Not Hispanic or LatinoB6-9 68 F UnknownB6-11 68 F Not Hispanic or LatinoB6-12 65 F Not Hispanic or LatinoB6-13 69 F Not Hispanic or LatinoB6-16 57 F Hispanic or Latino
(Continued )
Table 3. (Continued).
Sample IDa Ageb Sex Ethnicityb
B7-3 53 F Not Hispanic or LatinoB7-4 53 F Not Hispanic or LatinoB7-5 53 F Not Hispanic or LatinoB8-1 88 F Hispanic or LatinoB8-2 85 F Hispanic or LatinoB8-4 43 F Hispanic or LatinoB8-5 43 F Hispanic or LatinoB8-7 55 F Hispanic or LatinoB8-8 48 F Hispanic or LatinoB8-9 48 F Hispanic or LatinoB8-11 44 F Hispanic or LatinoB8-13 44 F Hispanic or LatinoB8-15 61 F Not Hispanic or LatinoB8-16 72 F Not Hispanic or LatinoB8-18 66 F Not Hispanic or LatinoB9-1 78 F Not Hispanic or LatinoB9-2 43 F Not Hispanic or LatinoB9-9 85 F Not Hispanic or LatinoB9-12 81 F Not Hispanic or LatinoB9-13 45 F Not Hispanic or LatinoB9-14 82 F Hispanic or LatinoB9-16 71 F Not Hispanic or LatinoB9-18 61 F Hispanic or LatinoB10-4 58 F Not Hispanic or LatinoB10-5 58 F Not Hispanic or LatinoB11-2 61 F Hispanic or LatinoB11-3 86 F Not Hispanic or LatinoB11-4 80 F Not Hispanic or LatinoB11-9 67 F Not Hispanic or LatinoB11-12 32 F Not Hispanic or LatinoB11-17 53 F Not Hispanic or LatinoB11-18 70 F UnknownB11-20 77 F Not Hispanic or LatinoB12-11 41 F Not Hispanic or LatinoB12-12 44 F Not Hispanic or LatinoB12-13 93 F Not Hispanic or LatinoB12-14 47 F Not Hispanic or LatinoB12-15 49 F Hispanic or LatinoB12-16 34 F Not Hispanic or LatinoB12-19 68 F Not Hispanic or LatinoB12-20 31 F Not Hispanic or LatinoB1-1 73 M UnknownB1-3 58 M Hispanic or LatinoB1-5 46 M Not Hispanic or LatinoB1-6 69 M Not Hispanic or LatinoB1-7 64 M Not Hispanic or LatinoB1-8 56 M Not Hispanic or LatinoB1-10 83 M Not Hispanic or LatinoB2-12 50 M Not Hispanic or LatinoB2-13 50 M Not Hispanic or LatinoB2-19 70 M Not Hispanic or LatinoB3-1 29 M Hispanic or LatinoB3-2 72 M Not Hispanic or LatinoB3-10 78 M Not Hispanic or LatinoB3-11 55 M Not Hispanic or LatinoB3-12 55 M Not Hispanic or LatinoB3-15 60 M Not Hispanic or LatinoB3-20 60 M Not Hispanic or LatinoB4-4 34 M Not Hispanic or LatinoB4-6 72 M Not Hispanic or LatinoB4-7 72 M Not Hispanic or LatinoB4-8 66 M Not Hispanic or LatinoB5-5 77 M Not Hispanic or Latino
(Continued )
EPIGENETICS 15
quantified and then independently validated bygenetic means [48]. Thus, our findings question thevalue of rDNA copy number determination usingDNA deposited in high-throughput sequencing data-bases, consistent with those reported rDNA countsbeing so at-odds with other methodologies [8]. Wecannot speculate whether any potential error that mayexist in databases is a result of biases in sequencing,quality control, the extraction and analysis of copynumber data, or some other step in the informatics‘pipeline.’ It seems prima facie erroneous that somesamples had fewer than 10 copies of either 45S or 5SrDNAs [45], which is incompatible with life in anyknown eukaryote [8,51]. Instead, our data support ourassertion that the existing databases are not reliablesources of rDNA copynumber data, and it is likely thatif rDNA copy number proves to be a risk factor ora diagnostic for human disease, then a simple andreliable protocol, such as ours, will be necessary tovalidate data in databases or from patients.
rDNA copy numbers do not correlate with age,sex, or ethnicity
Our collection of blood from breast cancer patients,and from patients with non-cancer diagnoses, allowedus to ascertain deviations in average rDNA copy num-ber as a consequence of age, sex, and ethnicity. Thesedata could prove useful in evaluating hypotheses link-ing rDNA copy number to age [72,73], disease status,or involvement in lifestyle choices that lead to greaterdisease risk (e.g., alcohol consumption, cigarettesmoking, overeating [74]). We re-analyzed the rDNAcopy number from the non-cancer-diagnosis
Table 3. (Continued).
Sample IDa Ageb Sex Ethnicityb
B5-14 86 M Not Hispanic or LatinoB5-18 53 M Hispanic or LatinoB5-19 71 M Not Hispanic or LatinoB6-2 70 M Not Hispanic or LatinoB6-3 75 M Not Hispanic or LatinoB6-7 28 M Not Hispanic or LatinoB6-8 57 M Not Hispanic or LatinoB6-10 79 M Not Hispanic or LatinoB6-14 45 M Not Hispanic or LatinoB6-15 54 M Not Hispanic or LatinoB6-17 72 M Not Hispanic or LatinoB6-18 72 M Not Hispanic or LatinoB6-19 72 M Not Hispanic or LatinoB6-20 38 M Not Hispanic or LatinoB7-1 68 M Hispanic or LatinoB7-2 68 M Hispanic or LatinoB7-6 48 M Hispanic or LatinoB7-7 48 M Hispanic or LatinoB7-8 48 M Hispanic or LatinoB7-9 45 M Hispanic or LatinoB7-10 45 M Hispanic or LatinoB8-3 54 M Not Hispanic or LatinoB8-6 55 M Not Hispanic or LatinoB8-10 60 M Not Hispanic or LatinoB8-12 73 M Not Hispanic or LatinoB8-14 68 M Not Hispanic or LatinoB8-17 74 M Not Hispanic or LatinoB8-19 68 M UnknownB8-20 68 M Not Hispanic or LatinoB9-3 81 M UnknownB9-4 83 M UnknownB9-5 81 M UnknownB9-6 74 M UnknownB9-7 69 M Not Hispanic or LatinoB9-8 84 M Not Hispanic or LatinoB9-10 82 M Not Hispanic or LatinoB9-11 78 M Not Hispanic or LatinoB9-15 76 M Not Hispanic or LatinoB9-17 74 M Not Hispanic or LatinoB9-19 77 M Not Hispanic or LatinoB9-20 62 M Not Hispanic or LatinoB10-1 35 M UnknownB10-2 35 M UnknownB10-3 38 M Not Hispanic or LatinoB10-6 37 M Not Hispanic or LatinoB10-7 37 M Not Hispanic or LatinoB10-8 37 M Not Hispanic or LatinoB10-9 45 M Hispanic or LatinoB10-10 45 M Hispanic or LatinoB11-1 61 M Not Hispanic or LatinoB11-5 51 M Hispanic or LatinoB11-6 77 M Not Hispanic or LatinoB11-7 56 M Not Hispanic or LatinoB11-8 56 M Not Hispanic or LatinoB11-10 67 M Hispanic or LatinoB11-11 72 M Not Hispanic or LatinoB11-13 61 M UnknownB11-14 61 M UnknownB11-15 57 M Not Hispanic or LatinoB11-16 57 M Not Hispanic or LatinoB11-19 46 M Hispanic or LatinoB12-1 52 M Hispanic or LatinoB12-2 70 M Hispanic or Latino
(Continued )
Table 3. (Continued).
Sample IDa Ageb Sex Ethnicityb
B12-3 36 M Hispanic or LatinoB12-4 36 M Hispanic or LatinoB12-5 70 M Hispanic or LatinoB12-6 59 M Hispanic or LatinoB12-7 52 M Hispanic or LatinoB12-8 36 M Hispanic or LatinoB12-9 70 M Hispanic or LatinoB12-10 59 M Hispanic or LatinoB12-17 83 M Not Hispanic or LatinoB12-18 64 M Not Hispanic or Latino
aNo identifying information is presented here; Sample ID refers toa serial nomenclature for this study alone, and correlation of thesenumbers to de-identified University of Arizona College of Medicinesample numbers are curated in the laboratory of Keith A. Maggert.
bSelf-identified.
16 V. VALORI ET AL.
individuals from the Tucson area and this time lookedfor correlations between average rDNA copy numberand sex and ethnicity as categorical conditions (Figure7(a-b)). No condition grouped rDNA copy numbersignificantly away from the others. Thus, we concludethat in our samples, there is no evidence to supportthe hypothesis that rDNA copy number differs in theblood of different sexes or ethnicities.
Ageing has been linked to rDNA loss in yeast, andthe application of that phenomenon to humans hasbeen proposed and, in some cases, supported [73]. Incontrast to those studies, we could detect no
correlation between individual rDNA copy numberor C.V. with age at time of blood collection (Figure 7(c)). This lack of apparent correlation is no-doubtaffected by the widely variant copy number at birthof these patients, and it is still possible that indivi-duals will exhibit rDNA loss as a function of ageing.At present, sufficiently broad sampling of blood overthe lifetimes of individuals is not available, so whilethe issue of individual rDNA copy number loss is stilloutstanding, it can be said that ‘snapshot’ rDNAcopy number is not predictive of biological age.This alone makes rDNA copy number a poor metric
a b
c
Figure 7. Comparisons between rDNA copy numbers and population variables. (a) Consistent lack of correlation between differentrDNA cluster copy numbers (18S and 5S) in males (blue, y = 1.61x + 0.83, R2 = 0.24) and females (pink, y = 0.78x + 1.06, R2 = 0.19).(b) Consistent weak correlation between different rDNA cluster copy numbers (18S and 5S) in hispanic (green, y = 1.20x + 0.72, R2 =0.37) and non-hispanic white (orange, y = 1.75x + 0.87, R2 = 0.30) individuals. (c) Consistent lack of correlation between 18S and 5SrDNA copy numbers as a function of age of blood donor. All values in this graph are represented as rDNA copy number relative totRNAMet, and normalized to the lab standard.
EPIGENETICS 17
for age or sex, and extreme caution must beemployed before considering such metrics in diseaserisk, disease onset, or forensic identification.
Concluding remarks
We have presented a method for quickly, inexpen-sively, sensitively, accurately, and reproduciblydetermining the copy number of repeated DNAs,in particular, the ribosomal RNA genes (45S and5S rDNAs), from patient samples. This approachwas pioneered in the model system Drosophila, butits application here has allowed us to directlytest multiple hypotheses concerning repeat DNAstability in both healthy individuals and breasttumours.
We find, first, that the rDNA show signs of gen-eral instability, consistent with previous work show-ing derepression of heterochromatin-mediatedstability in cancer. This finding is supported by ourdemonstration that the satellite repeats vary in timewith (although not in the same direction as)increased rDNA variation. As all of the repeatsequences are stabilized by heterochromatin, weenvision that an early step in cancer progression isthe loss of heterochromatin function. Changes incopy numbers of rDNA, satellites, telomeres, andperhaps other repetitive sequences are then second-ary and random. Second, we find that the reportedcovariation between 45S and 5S gene copy numbersis not evident using techniques other than high-throughput sequencing. Our data demonstratingthe sensitivity, responsiveness, and confirmation bySouthern blot analyses suggests that great cautionmust be exercised when deriving repeat copy num-bers from curated sequencing data. Third, our find-ings refute hypotheses suggesting that increases ordecreases in rDNA copy number are adaptive fordisease. Rather than being selected for by increaseddemands for protein synthesis, or selected against bytrimming genomes of superfluous DNA, we con-clude that hyper-variability is a general outcome ofdisease onset. The salient phenotype is increasedvariability, not an increased or decreased rDNAtranscription or rRNA output. It is likely that theobserved preferential losses in cultured cancer celllines are an artefact of growth in vitro, and that lossesmay merely be more stable than gains in culture.Fourth, we have demonstrated a facile, rapid,
inexpensive, precise, and accurate real-time PCR-based strategy for rDNA copy number quantificationusing very small amounts of fresh or fixed tissue.
Materials and methods
DNA from tumour and tissue samples
Five consecutive 1 µm slices were cut from forma-lin-fixed-paraffin-embedded tissue blocks andDNA purification was obtained by usingQIAamp DNA FFPE Tissue Kit (Qiagen).Basically, 1 ml of xylenes was added to each tube,vortexed for 1 min, then centrifuged at 14,000 rpmfor 2 min and the supernatant removed throughtwo ethanol washes. For each, ATL buffer andproteinase K were added, the samples groundwith a pestle, and incubated for 1 h at 56°C and1 h at 90°C. RNAse A was added to the samplesand left at room temperature for 2 min, then thesamples were transferred to columns and werewashed with buffer AL, buffer AW1, and finallybuffer AW2. Samples were eluted with 100 µLbuffer ATE and quantified using a Synergy H1Microplate Reader (BioTek).
DNA from blood cards
DNA purification was performed using GenSolveDNA Recovery Kit (Gentegra), the QIAshredderKit (Qiagen), and the QIAamp Blood Mini Kit(Qiagen). Basically, half of each blood card wascut and placed in individual microcentrifugetubes with Recovery Solution A and the bloodallowed to resuspend overnight at room tempera-ture. Tubes were then incubated, with rotation, for1 h at 56°C. Recovery Solution B was added alongwith the samples to QIAshredder columns andcentrifuged at 13,300 rpm for 2 min. The columnswere discarded and ethanol added, the samplesvortexed and briefly centrifuged, then transferredto QIAamp columns and centrifuged at 8000 rpmfor 1 min. Columns were washed with buffer AW1then with buffer AW2. Finally, samples wereeluted in two steps of buffer AE, in a total of100 μL of DNA solution, then quantified usinga Synergy H1 Microplate Reader (BioTek).
18 V. VALORI ET AL.
DNA from drosophila and human blood
Tissue was ground with a mini-pestle ina microcentrifuge tube in a solution containing100 mM Tris pH 8.0, 50 mM ethylenediaminete-traacetic acid, 1% Sodium Dodecylsulfate, and 1 µgProteinase K. The slurry was digested for an hourat 65°C, then extracted through a series of phenol,phenol-chloroform, chloroform, and ether. TheDNA was ethanol-precipitated and resuspendedin TE (10 mM Tris pH 8.0, 1 mM ethylenediami-netetraacetic acid) containing 0.1 µg RNAseA.DNA concentration was determined usinga Synergy H1 Microplate Reader (BioTek), andDNA samples were stored at high concentrationat −80°C until use.
Real-time PCR reactions
Real-time PCR reactions were done as describedpreviously [26,48], including controls on an ABIStep-One or Stop-One-Plus using SYBR Greenchemistry, full-length (2 h) reaction cycle, andobligate post-hoc melt curve analysis. Reactionswere adjusted to be 12 µL total reaction volume.Reactions were performed in triplicate or more,and data were accepted as valid if the standarderror of the mean of the replicates was less than0.1. Primers for non-rDNA targets are: BloomHelicase (GGCTGCTGTTCCTCAAAATAATCTACAG and ATTATTAAGTGTTCTGGCTGAGTGACG), Snail2 (CCCGTATCTCTATGAGAGTTACTCC and GTATGCTCCTGAGCTGAGGATCTC), RNMT2 (CTTTGATGGCAGCATACAGTGTTCTGG and CCTGTGAATTTCTTCTGCAGTTTCAAGC), eGFP (GAGGGTGAAGGTGATGCAACATACGG and GCCATGGAACAGGTAGCTTCCCAG).
Numerical analysis
Analyses were performed on an Apple MacBook Prousing Numbers version 6.0 (build 6194). Descriptiveand frequentist statistical analyses (e.g., regression,slope/intercept) were from embedded functions.
For simple comparisons between copy numbersof two genes in the same sample (e.g., Figure 1(b,c)),Crossing thresholds (Cts) could be directly
compared. Ct is calculated by determining the PCRcycle at which the signal first crosses the average plus10 standard deviations of all preceding cycles. Errorsare generally presented as Standard Errors of theMean, derived from the pooled errors of the copynumber of the gene in question and the tRNA nor-malizer, run in triplicate or quadruplicate. Standarderror of the mean is justified as these data are fromtechnical (assay) replicates of DNA extracted fromsingle individuals. Errors were pooled using the stan-dard summation of errors (SEM-pooled = √((SEM-target-1)2) + (SEM-target-2)2).
For copy number determination between differentsamples (e.g., Figure 1(d), 5(a-c)) crossing thresholdswere converted to relative amounts by first subtractingtRNACt from the target gene Ct, then using that as anexponent (2(Ct-target – Ct-tRNA); this is commonlyreferred to as the ‘ΔCt’ method. Where appropriate(e.g., Figure 6(c), 7), the ‘ΔΔCt’ method was used,which calculates ΔCts for the sample and fora standard (in this case a human DNA sample that isa study-wide standard in the laboratory); these dataindicate the proportion of rDNA-to-tRNA in the sam-ple relative to the rDNA-to-tRNA of the standard,allowing us to compare values between qPCRreactions.
Correlations were considered weak if R2 wasbelow 0.65, and non-existent if below 0.20. A prioricriteria for accepting Type-I errors (alpha) was set at0.01 at the beginning of the study.
Human data
Breast cancer tumour and adjacent samples from 29individuals were obtained from Dr. L. LeBeau. It wasdetermined by the University of Arizona InstitutionalReview Board that the use of the tissues did notrequire board oversight for this study (protocol#15–0477-0333) as the work did not meet the defini-tion of ‘human subjects’ by U.S. Department ofHealth and Human Services which state that ‘humansubject means a living individual about whom aninvestigator (whether professional or student) con-ducting research obtains data through interventionor interaction with the individual, or identifiable pri-vate information.’ Samples were de-identified, andshared without personal information for the purposeof this study.
EPIGENETICS 19
DNA samples from 51 individuals taking part ina trial to identify cancer risk alleles were collectedunder protocol #12–0138 (to C. Laukaitis), whichwas approved by the University of ArizonaInstitutional Review Board. Samples were de-identified, and shared without personal informationfor the purpose of this study.
Blood samples from 200 individual patientsobtained from the Arizona Health SciencesCenter Biorepository (from D. Harris) were de-identified, and shared without personal informa-tion for the purpose of this study.
Samples were further de-identified by assigningnew serial numbers to all samples; the key linkingthe original de-identified number from the source(i.e., LeBeau, Laukaitis, Harris) to the Table 1n thisstudy is safeguarded by our laboratory.
Acknowledgments
GFP-BLM cell lines were obtained from Dr. Mary Yagle andDr. Nathan Ellis. The University of Arizona Cancer Centerprovided core and facilities support, funded through NationalInstitutes of Health Support Grant P30CA023074. Blood DNA(Figure 6(a-b,e)) or blood samples (Figures 6(c) and 7) wereobtained from the AHSC Biorepository at The University ofArizona College of Medicine. Ovary samples (Figure 2) wereobtained from the Univeristy of Arizona Tissue Acquisition andCellular/Molecular Analysis Shared Resource (TACMASR).TACMASR also performed the tissue sections of the breasttumour samples. Support also came from the Department ofCellular and Molecular Medicine, the Univeristy of ArizonaCollege of Medicine, the Arizona State Museum, andTransformative Research Award (R01) GM123640, granted toKeith A. Maggert. Finally, Dr. Nicholas Ratterman’s excellentsociolinguistic parsimony and keen advice was invaluable, aswere the delicate urgings of Dr. Diana Darnell.
Disclosure statement
No potential conflict of interest was reported by the authors.
Funding
This work was supported by the NIH [GM123640].
ORCID
Virginia Valori http://orcid.org/0000-0003-4963-4285Keith A. Maggert http://orcid.org/0000-0002-9291-9481
References
[1] Stults DM, Killen MW, Williamson EP, et al. HumanrRNA gene clusters are recombinational hotspots incancer. Cancer Res. 2009;69:9096–9104.
[2] Gonzalez IL, Gorski JL, Campen TJ, et al. Variationamong human 28S ribosomal RNA genes. Proc NatlAcad Sci U S A. 1985;82:7666–7670.
[3] Stults DM, Killen MW, Pierce HH, et al. Genomicarchitecture and inheritance of human ribosomalRNA gene clusters. Genome Res. 2008;18:13–18.
[4] Krystal M, D’Eustachio P, Ruddle FH, et al. Humannucleolus organizers on nonhomologous chromosomescan share the same ribosomal gene variants. Proc NatlAcad Sci U S A. 1981;78:5744–5748.
[5] Peng JC, Karpen GH. H3K9 methylation and RNAinterference regulate nucleolar organization andrepeated DNA stability. Nat Cell Biol. 2007;9:25–35.
[6] Ranzani GN, Bernini LF, Crippa M. Inheritance ofrDNA spacer length variants in man. Mol Gen Genet.1984;196:141–145.
[7] Cohen S, Yacobi K, Segal D. Extrachromosomal circu-lar DNA of tandemly repeated genomic sequences inDrosophila. Genome Res. 2003;13:1133–1145.
[8] Long EO, Dawid IB. Repeated genes in eukaryotes.Annu Rev Biochem. 1980;49:727–764.
[9] Elgin SC, Reuter G. Position-effect variegation, hetero-chromatin formation, and gene silencing inDrosophila. Cold Spring Harb Perspect Biol. 2013;5:a017780.
[10] Vogel MJ, Guelen L, de Wit E, et al. Human hetero-chromatin proteins form large domains containingKRAB-ZNF genes. Genome Res. 2006;16:1493–1504.
[11] Bintu L, Yong J, Antebi YE, et al. Dynamics of epige-netic regulation at the single-cell level. Science.2016;351:720–724.
[12] Ryu T, Bonner MR, Chiolo I. Cervantes and Quijoteprotect heterochromatin from aberrant recombinationand lead the way to the nuclear periphery. Nucleus.2016;7:485–497.
[13] Chiolo I, Minoda A, Colmenares SU, et al. Double-strand breaks in heterochromatin move outside ofa dynamic HP1a domain to complete recombinationalrepair. Cell. 2011;144:732–744.
[14] Weiler KS, Wakimoto BT. Heterochromatin and geneexpression in Drosophila. Annu Rev Genet.1995;29:577–605.
[15] Perrini B, Piacentini L, Fanti L, et al. HP1 controlstelomere capping, telomere elongation, and telomeresilencing by two different mechanisms in Drosophila.Mol Cell. 2004;15:467–476.
[16] Garcia-Cao M, O’Sullivan R, Peters AH, et al.Epigenetic regulation of telomere length in mammaliancells by the Suv39h1 and Suv39h2 histonemethyltransferases. Nat Genet. 2004;36:94–99.
[17] Savitsky M, Kravchuk O, Melnikova L, et al.Heterochromatin protein 1 is involved in control of
20 V. VALORI ET AL.
telomere elongation in Drosophila melanogaster. MolCell Biol. 2002;22:3204–3218.
[19] Hathaway NA, Bell O, Hodges C, et al. Dynamics andmemory of heterochromatin in living cells. Cell.2012;149:1447–1460.
[20] Damelin M, Bestor TH. Biological functions of DNAmethyltransferase 1 require its methyltransferaseactivity. Mol Cell Biol. 2007;27:3891–3899.
[21] Spofford JB. Position-effect variegation in Drosophila. In:M. Ashburner, E. Novitski, editors. The genetics and biol-ogy of drosophila. Academic Press, London; 1976. p.955–1019.
[22] Aldrich JC, Maggert KA. Transgenerational inheritanceof diet-induced genome rearrangements in Drosophila.PLoS Genet. 2015;11:e1005148.
[23] Miller OL Jr., Beatty BR. Visualization of nucleolargenes. Science. 1969;164:955–957.
[24] McStay B, Grummt I. The epigenetics of rRNA genes:from molecular to chromosome biology. Annu RevCell Dev Biol. 2008;24:131–157.
[25] Guetg C, Lienemann P, Sirri V, et al. The NoRC com-plex mediates the heterochromatin formation and sta-bility of silent rRNA genes and centromeric repeats.EMBO J. 2010;29:2135–2146.
[26] Aldrich JC, Maggert KA. Simple quantitative PCRapproach to reveal naturally occurring andmutation-induced repetitive sequence variation on theDrosophila Y chromosome. PLoS One. 2014;9:e109906.
[27] Gibbons JG, Branco AT, Godinho SA, et al. Concertedcopy number variation balances ribosomal DNAdosage in human and mouse genomes. Proc NatlAcad Sci U S A. 2015;112:2485–2490.
[28] Kobayashi T. A new role of the rDNA and nucleolus inthe nucleus–rDNA instability maintains genome integ-rity. Bioessays. 2008;30:267–272.
[30] Bughio F, Maggert KA. The peculiar genetics of theribosomal DNA blurs the boundaries of transgenera-tional epigenetic inheritance. Chromosome Res.2019;27:19–30.
[31] Xu B, Li H, Perry JM, et al. Ribosomal DNA copynumber loss and sequence variation in cancer. PLoSGenet. 2017;13:e1006771.
[32] Zhou J, Sackton TB, Martinsen L, et al. Y chromosomemediates ribosomal DNA silencing and modulates thechromatin state in Drosophila. Proc Natl Acad SciU S A. 2012;109:9941–9946.
[33] Paredes S, Maggert KA. Ribosomal DNA contributes toglobal chromatin regulation. Proc Natl Acad Sci U S A.2009b;106:17829–17834.
[34] Larson K, Yan SJ, Tsurumi A, et al. Heterochromatinformation promotes longevity and represses ribosomalRNA synthesis. PLoS Genet. 2012;8:e1002473.
[35] Slee RB, Steiner CM, Herbert BS, et al. Cancer-associated alteration of pericentromeric heterochroma-tin may contribute to chromosome instability.Oncogene. 2012;31:3244–3253.
[36] Dialynas GK, Vitalini MW, Wallrath LL. LinkingHeterochromatin Protein 1 (HP1) to cancerprogression. Mutat Res. 2008;647:13–20.
[37] Atkin NB, Brito-Babapulle V. Heterochromatin poly-morphism and human cancer. Cancer GenetCytogenet. 1981;3:261–272.
[38] Ci X, Hao J, Dong X, et al. Heterochromatin protein1alpha mediates development and aggressiveness ofneuroendocrine prostate cancer. Cancer Res.2018;78:2691–2704.
[39] Zhang R, Adams PD. Heterochromatin and its rela-tionship to cell senescence and cancer therapy. CellCycle. 2007;6:784–789.
[40] Pezer Z, Ugarkovic D. Role of non-coding RNA andheterochromatin in aneuploidy and cancer. SeminCancer Biol. 2008;18:123–130.
[41] Storchova Z, Pellman D. From polyploidy to aneu-ploidy, genome instability and cancer. Nat Rev MolCell Biol. 2004;5:45–54.
[42] Chin K, de Solorzano CO, Knowles D, et al. In situanalyses of genome instability in breast cancer. NatGenet. 2004;36:984–988.
[43] de Vargas Wolfgramm E, Alves LN, Stur E, et al.Analysis of genome instability in breast cancer. MolBiol Rep. 2013;40:2139–2144.
[44] Richard SM, Bailliet G, Paez GL, et al. Nuclear andmitochondrial genome instability in human breastcancer. Cancer Res. 2000;60:4231–4237.
[45] Wang M, Lemos B. Ribosomal DNA copy numberamplification and loss in human cancers is linked totumor genetic context, nucleolus activity, andproliferation. PLoS Genet. 2017;13:e1006994.
[46] Schawalder J, Paric E, Neff NF. Telomere and riboso-mal DNA repeats are chromosomal targets of thebloom syndrome DNA helicase. BMC Cell Biol.2003;4:15.
[47] Killen MW, Stults DM, Adachi N, et al. Loss ofBloom syndrome protein destabilizes human genecluster architecture. Hum Mol Genet.2009;18:3417–3428.
[48] Paredes S, Maggert KA. Expression of I-CreI endonu-clease generates deletions within the rDNA ofdrosophila. Genetics. 2009a;181:1661–1671.
[49] Erickson JM, Rushford CL, Dorney DJ, et al.Structure and variation of human ribosomal DNA:molecular analysis of cloned fragments. Gene.1981;16:1–9.
[50] Wilson GN, Szura LL, Rushford C, et al. Structure andvariation of human ribosomal DNA: the external tran-scribed spacer and adjacent regions. Am J Hum Genet.1982;34:32–49.
EPIGENETICS 21
[51] Prokopowich CD, Gregory TR, Crease TJ. The correla-tion between rDNA copy number and genome size ineukaryotes. Genome. 2003;46:48–50.
[52] Richard GF, Kerrest A, Dujon B. Comparative geno-mics and molecular dynamics of DNA repeats ineukaryotes. Microbiol Mol Biol Rev. 2008;72:686–727.
[53] Hugerth LW,Muller EE,HuYO, et al. Systematic design of18S rRNA gene primers for determining eukaryotic diver-sity in microbial consortia. PLoS One. 2014;9:e95567.
[54] Wellauer PK, Dawid IB. The structural organization ofribosomal DNA in Drosophila melanogaster. Cell.1977;10:193–212.
[55] Bankfalvi A, Ofner D, Schmid KW, et al. Standardizedin situ AgNOR analysis in breast pathology: diagnosticand cell kinetic implications. Pathol Res Pract.1999;195:219–229.
[56] Lesty C, Chleq C, Contesso G, et al. Nucleoli and AgNORproteins in 32 cases of primary breast carcinoma. Spatialpattern of interactions between 50 clinical and histometriccriteria. Anal Quant Cytol Histol. 1992;14:175–186.
[57] Barwijuk-Machala M, Musiatowicz B, Cylwik J,et al. AgNOR, Ki-67 and PCNA expression infibroepithelial tumours of the breast in correlationwith morphological features. Folia Morphol(Warsz). 2004;63:133–135.
[67] Bednarek AK, Sahin A, Brenner AJ, et al. Analysis oftelomerase activity levels in breast cancer: positive detec-tion at the in situ breast carcinoma stage. ClinCancer Res.1997;3:11–16.
[58] Derenzini M, Ceccarelli C, Santini D, et al. The prog-nostic value of the AgNOR parameter in human breastcancer depends on the pRb and p53 status. J ClinPathol. 2004;57:755–761.
[59] Guerrero PA, Maggert KA. The CCCTC-binding factor(CTCF) of Drosophila contributes to the regulation ofthe ribosomal DNA and nucleolar stability. PLoS One.2011;6:e16401.
[74] Holland ML, Lowe R, Caton PW, et al. Early-lifenutrition modulates the epigenetic state of specific
rDNA genetic variants in mice. Science.2016;353:495–498.
[61] Jiang WQ, Nguyen A, Cao Y, et al. HP1-mediated for-mation of alternative lengthening of telomeres-associatedPML bodies requires HIRA but not ASF1a. PLoS One.2011;6:e17036.
[60] Jiang WQ, Zhong ZH, Nguyen A, et al. Induction ofalternative lengthening of telomeres-associated PMLbodies by p53/p21 requires HP1 proteins. J Cell Biol.2009;185:797–810.
[64] Jolly C, Lakhotia SC. Human sat III and Drosophila hsromega transcripts: a common paradigm for regulationof nuclear RNA processing in stressed cells. NucleicAcids Res. 2006;34:5508–5514.
[63] Jolly C, Metz A, Govin J, et al. Stress-induced tran-scription of satellite III repeats. J Cell Biol.2004;164:25–33.
[71] Kobayashi T. Regulation of ribosomal RNA gene copynumber and its role in modulating genome integrityand evolutionary adaptability in yeast. Cell Mol LifeSci. 2011;68:1395–1403.
[73] Malinovskaya EM, Ershova ES, Golimbet VE, et al.Copy number of human ribosomal genes withaging: unchanged Mean, but narrowed range anddecreased variance in elderly group. Front Genet.2018;9:306.
[70] Nawaz S, Hashizumi TL, Markham NE, et al.Telomerase expression in human breast cancer withand without lymph node metastases. Am J ClinPathol. 1997;107:542–547.
[68] Rha SY, Jeung HC, Park KH, et al. Changes of telo-merase activity by alternative splicing of full-length andbeta variants of hTERT in breast cancer patients. OncolRes. 2009;18:213–220.
[69] Umbricht CB, Sherman ME, Dome J, et al. Telomeraseactivity in ductal carcinoma in situ and invasive breastcancer. Oncogene. 1999;18:3407–3414.
[62] Valgardsdottir R, Chiodi I, Giordano M, et al.Structural and functional characterization of non-coding repetitive RNAs transcribed in stressedhuman cells. Mol Biol Cell. 2005;16:2597–2604.
[72] Wang M, Lemos B. Ribosomal DNA harbors an evo-lutionarily conserved clock of biological aging.Genome Res. 2019;29:325–333.
