Robust measurement of telomere length in single cellsFang Wanga,b, Xinghua Panc,1, Keri Kalmbacha, Michelle L. Seth-Smitha, Xiaoying Yeb, Danielle M. F. Antumesa,d,e,Yu Yinb, Lin Liub,1, David L. Keefea,1, and Sherman M. Weissmanc,1

aDepartment of Obstetrics and Gynecology, New York University Langone Medical Center, NY 10016; bState Key Laboratory of Medicinal Chemical Biology,College of Life Sciences, Nankai University, Tianjin 300071, China; cDepartment of Genetics, Yale University School of Medicine, New Haven, CT 06520-8005;dDepartment of Pathology, Fluminense Federal University, 24033-900, Rio de Janeiro, Brazil; and eCoordenacao de Aperfeicoamento de Pessoal de NivelSuperior (CAPES) Foundation, Ministry of Education of Brazil, 70040-020, Brasilia, Brazil

Contributed by Sherman M. Weissman, April 9, 2013 (sent for review February 5, 2013)

Measurement of telomere length currently requires a large popula-tion of cells, which masks telomere length heterogeneity in singlecells, or requires FISH in metaphase arrested cells, posing technicalchallenges. A practical method for measuring telomere length insingle cells has been lacking. We established a simple and robustapproach for single-cell telomere length measurement (SCT-pqPCR).We first optimized a multiplex preamplification specific for telo-meres and reference genes from individual cells, such that the ampli-con provides a consistent ratio (T/R) of telomeres (T) to the referencegenes (R) by quantitative PCR (qPCR). The average T/R ratio of multi-ple single cells corresponded closely to that of a given cell populationmeasured by regular qPCR, and correlated with those of telomererestriction fragments (TRF) and quantitative FISHmeasurements. Fur-thermore, SCT-pqPCRdetected the telomere length forquiescent cellsthat are inaccessible byquantitative FISH. The reliability of SCT-pqPCRalsowas confirmedusing sister cells from two cell embryos. Telomerelength heterogeneity was identified by SCT-pqPCR among cells ofvarious human and mouse cell types. We found that the T/R valuesof human fibroblasts at later passages and from old donors werelower and more heterogeneous than those of early passages andfrom young donors, that cancer cell lines show heterogeneous telo-mere lengths, that human oocytes and polar bodies have nearly iden-tical telomere lengths, and that the telomere lengths progressivelyincrease from the zygote, two-cell to four-cell embryo. This methodwill facilitate understandingof telomere heterogeneity and its role intumorigenesis, aging, and associated diseases.

single cell analysis | stem cell | cell senescence

Telomeres are the ribonucleoprotein structures that cap andprotect linear chromosome ends from genomic instability and

tumorigenesis (1, 2). Intriguingly, telomere shortening protectsagainst tumorigenesis by limiting cell growth (3, 4), but also canimpair tissue regenerative capability and cell viability (5, 6).Thus far, most assays of telomere length measure average

telomere length from aggregates of many cells derived fromdissected tissues, cultured cells, or blood (7). Telomere restrictionfragment (TRF) determination (1, 8), a Southern blot-based tech-nique, remains the “gold standard” for determining absolute telo-mere length, but requires a large amount of starting material (0.5–5μg DNA) and several days for processing. Moreover, the require-ments for gel electrophoresis and hybridization limit the scalabilityof this assay. Recently, a quantitative PCR (qPCR)-based methodfor telomere length measurement was developed, providing theconvenience and scalability of PCR (9). Although the DNA re-quirement (35 ng) for qPCR is significantly less than TRF, it stillrelies on populations of cells to derive sufficient amount of DNA.Quantitative FISH (Q-FISH) allows sensitive visualization of

relative telomere length from individual cells and individual telo-meres, but this method requires many cells or metaphase arrestedcells, which precludes its application to many sample types, in-cluding postmitotic cells, senescent cells, and other nondividingcells, and when only one actual cell is required to test. In addition,preparing chromosome spreads requires significant technical skill,and only proliferating cells within a population reach metaphasestage, so this analysis potentially biases the estimates of telomere

length for a given cell population (10–12). High-throughputQ-FISH, flowFISH, and single telomere length analysis can be usedfor telomere measurement of dividing, nondividing, and senescentcells, but these methods also require large cell populations (13–15).The ability to measure telomere length in single cells rather

than relying upon average telomere length in cell populations orthe entire tissue enables the study of biological heterogeneity ona cell-by-cell basis, an issue of fundamental importance for studiesof aging, development, carcinogenesis, and many other diseases.Here, we demonstrate an accurate determination of telomerelength in individual cells, with the resolution and scalability of theqPCR telomere length assay.The basis of qPCR is that within a given cell, the ratio of the copy

number of telomere repeats to the copy number of a multicopyreference gene is fixed (3), and this method, because of its sim-plicity, has been widely used to investigate a variety of telomereshortening-associated diseases (7), even sensitive enough to iden-tify mild telomere dysfunction resulting from chronological lifestress (16, 17). We adapted qPCR to measure telomere length inindividual cells by using a preamplification step that specificallytargets both the telomere andmulticopy genes, followed by a qPCRassay to obtain telomere to reference gene (T/R) ratio. A single-celltelomere (SCT) length measurement method (SCT-pqPCR) runsrobustly, and shows an identical T/R ratio for two sister blastomeresfrom two-cell–stagemouse embryos. The average result from SCT-qPCR with multiple single cells is linearly correlated to Q-FISH,TRF, and conventional qPCR assays designed for a large numberof cells. The heterogeneity of telomere length among severalpopulations of cells by SCT-pqPCR run on multiple single cells isconsistent with—and sometimes superior to—results obtained byQ-FISH. Application of SCT-pqPCR to study telomere lengthduring early embryo development, aging, and cancer demonstratethe value of this single-cell telomere length assay method.

Significance

Telomeres are the structures at the ends of chromosomes thatprotect these ends from degradation or joining to one another.Telomeres consist of repeat DNA sequences and the length isgradually eroded as the cell ages. The ability to measure telo-mere length in individual cells would be important for studiesof cell senescence, malignancy, stem cell renewal, and humanfertility. We have developed a robust and practical method forestimating the telomere length of single cells, and used thismethod to demonstrate the heterogeneity or changes oftelomere length in several systems.

Author contributions: X.P., L.L., D.L.K., and S.M.W. designed research; F.W. and K.K.performed research; M.L.S.-S., X.Y., and Y.Y. contributed new reagents/analytic tools;F.W., X.P., D.M.F.A., L.L., D.L.K., and S.M.W. analyzed data; and F.W., X.P., K.K., L.L.,D.L.K., and S.M.W. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence may be addressed. E-mail: xinghua.pan@gmail.com, liutelom@yahoo.com, David.Keefe@nyumc.org, or sherman.weissman@yale.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1306639110/-/DCSupplemental.

E1906–E1912 | PNAS | Published online May 9, 2013 www.pnas.org/cgi/doi/10.1073/pnas.1306639110

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

ResultsEstablishment of Single-Cell qPCR for Telomere Length Measurement.A single cell possesses ∼6–7 pg of genomic DNA (18), but theoriginal qPCR assay requires nanogram amounts of DNA tomeasure telomere length (9). To analyze telomere length in singlecells, we developed a strategy based on qPCR of the telomereversus reference gene after parallel preamplification of these tar-get sequences. The preamplification step consists of a multiplexPCR for a limited number of cycles in a single tube containinga single cell without physical gDNA extraction, to generate suffi-cient DNA for downstream qPCR, yet retain the natural ratio ofthe targeted sequences (Fig. 1A).Robust and faithful preamplification of the telomere se-

quence and reference genes from single cells, without distor-tion of their native ratio, is critical for this assay. To avoidDNA loss during physical purification, we used a lysis buffercontaining EDTA and Nonidet P-40 to gently release DNA,retain DNA integrity, and remain compatible with the sub-sequent PCR. In addition, we tested various reference genes,from a single-copy gene to several multicopy genes, to nor-malize the telomere units. The single-copy gene, 36B4, serves asa reference gene in the conventional qPCR for telomere mea-surement (9). However, 36B4 did not produce robust results forsingle-cell amplification, and the melting curve of 36B4 ampli-cons (mouse and human) frequently showed multiple peaks, incontrast to very robust telomere DNA amplification (see for exam-ple, Fig. S2A), suggesting that the single-copy gene from a single cellmay be unavailable for amplification in many conditions. Multicopy

genes, some with thousands of copies of sequences throughput thegenome, similar to telomeres, could provide more dependable am-plification.We tested threemulticopy genes (Alu,B1, and 18S rRNA)as reference genes and found that the relative telomere lengthsmeasured with T/R for populations of human cells using multicopygenes Alu or 18S rRNA by qPCR were consistent with the resultsusing the single-copy gene (36B4) (Fig. S1A). The correlation be-tween repeat Alu and 36B4 (R2 = 0.9836) was slightly better thanbetween 18S rRNA and 36B4 (R2 = 0.9527) (Fig. 1B).We also measured the absolute telomere length of human

cell lines with the TRF method (Fig. S1B). The telomere lengthof human cell lines by TRF was proportional to the T/R ratio,and again the multicopy gene Alu (R2 = 0.8963) better corre-lated with TRF than did 18S rRNA (R2 = 0.7064) (Fig. 1C). Theslight distortion of 18S rRNA may aise from an outlier in themeasurement. We chose the multicopy gene Alu for human cellsor the B1 sequence for mouse cell as our reference gene insingle-cell telomere analysis, and suggest taking 18S rRNA as analternative when required.With single-cellDNAor amounts up to 10 ngDNA fromHeLaS3

cells, the PCR reached a plateau when the cycle number was morethan 20, regardless of the primers used. Themouse tail-tip fibroblast(TTF) showed a similar result (Fig. S2B). We preamplified single-cell DNA from human lung fibroblasts using telomere and Alu pri-mers simultaneously for 20, 18, 16, 14, or 12 cycles and found the Ctvalue proportionally increased with decreasing cycle number from18 to 14 (Fig. S1E). To determine the optimal pre-PCR cyclenumber, we prepared standard curves for telomere and reference

Fig. 1. Design of SCT-pqPCR. (A) Schematic diagram of SCT-pqPCR. Green and red dots represent telomere sequence and reference gene (Alu) sequence inthe chromosome, respectively. Green and red lines (thin) represent PCR productions by Tel and Alu primers, respectively, after preamplification. (B) Linearcorrelation analysis of relative telomere length of a population of cells by regular qPCR between the single-copy gene and multicopy genes as the referencegene. (C) Linear correlation analysis of telomere length measurement of a population of cells by TRF and regular qPCR with multicopy genes.

Wang et al. PNAS | Published online May 9, 2013 | E1907

GEN

ETICS

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

genes by amplification of a series of dilutions of genomic DNA. Thestandard curves were quite linear within DNA concentrations from1 ng/μL to 16 ng/μL for human cells (Fig. S2 C andD) and 0.375 ng/μL to 6 ng/μL for mouse cells (Fig. S2C). A comparison of the Ctvalue of standard DNA (Fig. S2D) and the Ct value after differentnumbers of pre-PCR cycles (Fig. S2E) indicated that pre-PCR cyclenumbers from 16 to 18 were optimal; therefore, we chose the pre-amplification cycle number 17 for human cells and 16 formouse cells,because the telomere length ofmouse generally is longer than that ofhuman and, expectedly, requires fewer cycles for amplification. Weshould point out that the T/R ratio obtained for mouse telomerescannot be directly comparedwithT/R ratio for human telomeres.Wesuggest that a correlation of T/R ratiomeasured by SCT-pqPCRwithTRF (kb) by Southern blot should be established initially using a cellpopulation for each species and for each laboratory.

Correlation of Average Telomere Length Estimated by SCT-pqPCR andTelomere Length of Cell Populations by Conventional Methods. Al-though the results above demonstrated that the telomere lengthof single cells can be measured by our assay, it was necessary tocompare average telomere length calculated by averaging theresults of single-cell measurements by SCT-pqPCR to establishedmethods, including qPCR, Q-FISH, and TRF, which measureaverage telomere length of cell populations. Based on the abovedata, we compared telomere length of a population of cells mea-sured by conventional qPCR (Fig. S1A) and Q-FISH (Fig. 3 andFig. S3) to average telomere length of single cells measured bySCT-pqPCR in human cell lines F171, F200, HeLa S3, RuES2,U2OS, and SaOS2 (Table S1). As expected, the data showed that

single-cell telomere length by SCT-pqPCR significantly correlatedwith average telomere length of cell populations measured byconventional qPCR and Q-FISH; the Pearson test produced Pvalues of 0.001 and 0.006, respectively (Fig. 2 A and B). We alsoanalyzed the correlation between SCT-pqPCR and TRF (Fig.S1B) in human cell lines (F171, F200, HeLa S3, RuES2, andSaOS2). The Pearson test again showed that the average of single-cell telomere lengths measured by SCT-pqPCR was highly corre-lated with absolute telomere length by TRF, with a P value of 0.015(Fig. 2C). Interestingly, we found lower correlation betweenQ-FISH and TRF in these five cell lines (Fig. 2D). Presumably,the heterogeneity of single-cell telomere lengths in cancer celllines exceeds that of normal cell lines. In addition, variationsin telomere lengths between dividing and nondividing cells maybias Q-FISH results, because Q-FISH measures telomere lengthsonly of dividing cells capable of arrest at metaphase, but the TRFand qPCR methods measure average telomere length of all cells.This finding further supports the value of the SCT-pqPCR methodfor telomere measurement of any individual cells independent oftheir replication rate or potential.To further validate our assay, we measured telomere lengths in

dilutions of preamplified DNA of F171 and F200 after pre-amplifying DNA with target primers (telomere and referencegene Alu). When the input DNA was 2 ng, the relative telomerelength in F171 was longer than that of F200 (Fig. S4), but whenthe DNA quantity decreased to 0.4 ng, there was no significantdifference in telomere lengths between fibroblast F171 andF200, P > 0.05 (Fig. S4). Therefore, when the prepurified DNAdrops below threshold value, one aliquot of the diluted DNAdoes not fairly represent the entire genome. The approximately0.5-ng to 1-ng threshold for purified human genome DNA wasobserved in a whole genome amplification effort (19, 20). The locusrepresentation was significantly distorted when the input gDNAaliquoted from a large DNA pool is <0.5–1 ng. On the other hand,an intact single cell, although it contains only about 6–7 pg DNA,contains an entire set of genomic sequences including all telomeres.

Validation of Single-Cell Telomere Length Measurements by SCT-pqPCR Using Various Assays. To validate single-cell telomere lengthmeasurements using our method, we chose two human cell typeswith different telomere lengths: HeLa S3 and 1301 human celllines with average telomere lengths of 5 kb (15) and 70 kb, re-spectively. We also studied two mouse cell lines with differenttelomere lengths: embryonic stem cell (ESC) and TTF (21). Thetelomere length for each single cell in the same population variedby SCT-pqPCR analysis, and these results were consistent with theQ-FISH telomere lengths (Fig. 3 A, C,D, and F). Not unexpectedly,the calculated average telomere length of multiple single cells wassignificantly longer in mESCs than TTF in mice, and longer in 1301human cells than HeLa S3 by SCT-pqPCR. The single-cell telo-mere length variations were found by SCT-pqPCR, and the varia-tion also was identified by Q-FISH, with an independent sampledt test. The average T/R ratio of single cells measured by SCT-pqPCR was consistent with that of a cell population measured bySCT-pqPCR (T/R) or by conventional qPCR (T/S) (S, single-copygene) (Fig. 3 B and E).The limitation for the validation step above was that SCT-pqPCR

and Q-FISH could not be performed for the same individual cells,and no somatic cells with identical telomere lengths could be reliablyidentified for this validation. To further validate the SCT-pqPCRassay, we measured telomere length in pairs of sister cells derivedfrom two-cellmouse embryos and oocytes. Two-cell mouse embryosexhibit identical telomere lengths between sister blastomeres (22),and human polar bodies (PBs) show telomere lengths nearly iden-tical to their matched MII oocytes by Q-FISH analysis (23). Telo-mere lengths were remarkably similar between sister blastomeres bySCT-pqPCR,P>> 0.1 (Fig. 4A andTable S2), although the one-wayANOVA (Tukey test) indicated that differences existed between

Fig. 2. Linear correlation of relative telomere length between single cellsand their populations with various human cell lines. (A) Average telomerelength in single cells as T/R ratio by SCT-pqPCR is significantly correlated withthat of a cell population shown as the T/S ratio by regular qPCR. The Pearsontest is applied. (B) Average telomere length in single cells by SCT-pqPCR issignificantly correlated with quantitative telomere length by Q-FISH. (C) Av-erage telomere length of single human cells by SCT-pqPCR is highly correlatedwith absolute telomere length of population cells by TRF. (D) Quantitativetelomere length of metaphase cells by Q-FISH is correlated with the absolutetelomere length of cell populations by TRF. The trapezoid represents HeLa S3cancer cell, the hexagon represents human lung fibroblast from a 71-y-olddonor (F200), the pentagon represents human lung fibroblast from a 14-wkgestation (F171), the triangle represents SaOS2 cells, the square representshuman ESC (RuES2), and the circle represent U2OS cell.

E1908 | www.pnas.org/cgi/doi/10.1073/pnas.1306639110 Wang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

pairs of sister cells from different embryos. Correlation analysisshowed the telomere lengths between sister cells were proportionalby the Pearson test (P=0.005) (Fig. 4B). Telomere lengths betweenhuman oocytes (O) and their PBs demonstrated no significant dif-ferences by the paired samples t test, t = 0.603, P = 0.569 (Fig. 4Cand Table S2). Occasional (e.g., PB1/O1, PB6/O6) PB and oocytesexhibited different telomere lengths, which could represent de-generation in telomere DNA or biological differences. Curiouslytelomere lengths of oocytes and PBs in patients 6 and 7 were re-markably longer than other patients by one-way ANOVA test, P <0.001 (Fig. 4C). The telomere lengths of human oocytes correlatedhighly with the paired PB telomere lengths by liner correlationanalysis (Fig. 4D). Measurements of telomere length in pairs ofhuman oocytes and PBs by SCT-pqPCR agreed with our previousstudies (23). Unlike sister cells of mouse two-cell embryos, telomerelengths between daughter cells of HeLa S3 frequently showed dif-ferences, although the correlation between telomere lengths indaughter cells was still high (Fig. 4 E and F, and Table S2).

SCT-pqPCR Identifies Telomere Length and Its Heterogeneity in SingleCells from Various Cell Types. We used SCT-pqPCR to measuretelomeres of human lung fibroblast cell lines from different ageddonors and passage numbers. When comparing telomere lengthin cell populations, the average telomere length of human fi-broblast F171 passage number (P)16 (from 14-wk gestation) and

F204 P14 (from a 35-y-old donor) was longer (P < 0.05) than thatof human fibroblast F200 P7 (from a 71-y-old donor) by Q-FISHand conventional qPCR, but the average telomere length did notdiffer (P > 0.05) between F171 P16 and F204 P14 (Fig. S5). Wethen analyzed the single-cell telomere lengths between F171 P16and F200 P7 by SCT-pqPCR. Remarkably, the telomere lengthsof single cells differed in the same population of both F171 P16and F200 P7 cells. Indeed, some single cells from F200 P7 hadlonger telomeres than F171 P16, as measured by SCT-pqPCR,a finding confirmed by Q-FISH analysis (Fig. 5 A and B and Fig.S3). The coefficient of variation (CV) showed single-cell telo-mere length in F200 P7 to be more heterogeneous than F171 P16(Table 1). When human fibroblasts were continuously cultured(F171 P16 to P31 and F200 from P7 to P12), metaphase chro-mosome spreads were rarely available for analysis of cells at laterpassage, presumably because these cells had undergone senes-cence and failed to divide. Interestingly, SCT-pqPCR demon-strated increased variation in telomere length among cells atlater passage compared with early passages (CV 0.486 in F171P31 and 0.398 in F200 P12 vs. 0.169 in F171 P16 and 0.233 inF200 P7, respectively) (Fig. 5A and Table 1).The above data show that telomere length varies among in-

dividual cells in a given population. We also analyzed telomerelength of single cells in human ESC cultures (RuES2, telomeraseactivity-positive), cultured cancer cells (HeLa S3, telomerase

Fig. 3. Measurement of relative telomere length by SCT-pqPCR. (A and D) Analysis of single-cell telomere length by SCT-pqPCR (A and D) and QFISH (A′ andD′) in mouse (TTF and mES) and human cell lines (HeLa S3 and 1301). T/R ratio is against the multicopy gene B1 and Alu for mouse and human cells, re-spectively. The cycle number of pre-PCR is 16 and 17 for mouse and human cells, respectively. n = 6. Fluorescence intensity represents the telomere lengthsignal by the Q-FISH method. (B and E) Validation of average telomere length of single cells by comparing to the average telomere length of cell populationsin mouse and human cells. (B and E) Average telomere length of single cells as mean of T/R ratio by SCT-pqPCR. n = 10. (B′ and E′) Average telomere length ofcell populations as T/S ratio by regular qPCR. n = 6. Bar is ± SD. (C and F) Telomere length distribution in metaphase chromosomes of mouse and human cellpopulations by Q-FISH. Average telomere length as telomere fluorescence unit (TFU) is indicated as mean ± SD.

Wang et al. PNAS | Published online May 9, 2013 | E1909

GEN

ETICS

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

activity-positive), and cultured osteosarcoma cells (U2OS andSaOS2, telomerase activity-negative) by SCT-pqPCR. As expec-ted, average telomere length and variability among cells of a givenpopulation differed among these various cell types (Fig. 5A, andFig. S3 A and C). The telomere lengths of HeLa S3 cells variedmore than other cell types, with a CV of 0.294 (Table 1). Telo-mere lengths of single cells at metaphase in these cells (HeLa S3,Rues2, SaOS2, and U2OS) also showed extensive variations whenmeasured by Q-FISH, confirming that single-cell telomere lengthvaries within the same cell population (Fig. 5B, and Fig. S3 B andD). From the CV of different human cell lines as estimated bySCT-pqPCR and Q-FISH, the single-cell telomere lengths incancer cell populations vary more than in early passage humanfibroblast cell lines (Table 1). Similarly, fibroblasts from olderpeople show more heterogeneity in telomere length than thoseof young people, and cells at later passages show telomerelength variations more than those at earlier passages.In addition, as a very delicate case, we also successfully mea-

sured the telomere lengths of individual cells from mouse zygoteto four-cell embryos (Fig. 5C), and the average telomere lengthincreased along with development of mouse embryos by SCT-pqPCR (Fig. 5D). Consistently, telomere length in developingmouse embryos was found to lengthen significantly from zygotes tothe four-cell embryo stage (22).

DiscussionHere we report the development and validation of single-celltelomere assay by qPCR (SCT-pqPCR), which provides sim-ple, robust and high-quality analysis of telomere length in in-dividual cells. SCT-pqPCR can measure the telomere lengthwith an actual single cell, for example, each sister cell of a two-cell mouse embryo, which are confirmed identical in T/R ratio. Alinear relationship was demonstrated between average T/R valuesgenerated with SCT-pqPCR and absolute telomere length gener-

ated by TRF, Q-FISH, and conventional qPCR. SCT-pqPCR showsthat the average telomere length and its variation within multiplesingle cells for human lung fibroblasts, embryos, and cancer cells,correlate with data obtained by conventional qPCR and Q-FISHwith the corresponding populations of cells. Using this method, wefound that the T/R value of early passage human fibroblasts to berelatively uniform, but later passages of human fibroblasts andcancer cells become increasingly variable, that human oocytes andPBs share nearly identical telomere length (T/R), and that thetelomere length (T/R) is significantly increased from zygote to two-cell and to four-cell embryos. SCT-pqPCR not only allows mea-surement of telomere length in single cells, independent of theirability to divide, but also provides a tool to estimate the hetero-geneity of telomere length in cell populations.One key to the successful establishment of our assay was iden-

tifying reliablemulticopy genes,Alu andB1, to normalize telomeremeasurements. It is possible that aneuploidy may exist in cancercells and embryos, and single loci may drop out during lysis andpre-PCR. The use of a multicopy gene also may help avoid am-plification bias from single cells because it targets an abundantsequence, similar to the abundant telomere units, within the singlecells. Because of the large number of Alu (500,000 copies) or B1(150,000 members per genome) family units and their interspersionthroughout the genome, small copy number variations, such asa deletion of a fewmegabases or even a few hundred new insertions,would not substantially affect the denominator. When desired, tworeference genes [i.e., Alu (or B1) and 18S rRNA] may be used incombination in one assay. If the copy number of the reference gene(Alu or B1) were to change extremely significant among the testsamples, 18S rRNAmay be used alone to replace Alu (or B1) as thereference gene. Another key is the single-tube preamplification ofthe reference gene in parallel with the telomere sequences fora limited number of cycles by multiplex PCR, which avoids sampleloss, and retains a faithful ratio of products, yet significantly expands

Fig. 4. Measurements of telomere length in mouse two-cell stage embryos and pairs of human oocytes and PBs by SCT-pqPCR. (A) Measurement of single celltelomere length by SCT-pqPCR between two sister cells from a mouse two-cell (2c) embryo. n = 6. (B) Telomere length is highly correlated between two sistercells of a two-cell embryo. (C) Comparison of telomere length in pairs of human MII oocytes and PBs by SCT-pqPCR. n = 6. (D) The telomere length of humanoocyte is proportional to that of the paired PB. (E) Measurement of single-cell telomere length by SCT-pqPCR between two daughter cells from the HeLa S3.H represents HeLa S3. n = 6. Bar is ± SD. (F) The telomere length correlation of two daughter cells for 10 pairs of cells of HeLa S3.

E1910 | www.pnas.org/cgi/doi/10.1073/pnas.1306639110 Wang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

material for measurements by qPCR. We found that SaOS2 cellsproliferated more slowly. In addition, many cells undergo senes-cence and fail to reach metaphase stage, and thus cannot be mea-sured by Q-FISH. These data demonstrate that SCT-pqPCR has anadvantage over Q-FISH for telomere length at single-cell levelsbecause it is not dependent on cell division and does not bias resultswhen testing a heterogeneous population of cells consisting of bothdividing and quiescent cells.SCT-pqPCR is a simple, efficient, and accurate method for

measuring cell-to-cell differences and heterogeneity in telomerelength at the molecular level, comparable to RNA-seq, CpG

methylation, and other functional genomics technologies. Thesignificance of measuring telomere length in individual cells is ev-ident in several respects. This process will facilitate the identifica-tion and functional studies of cancer stem cells and pluripotent stemcells, and distinguish them from other cells in mixed populations(24), and advance our understanding of the molecular mechanismsunderlying variation of individual cells in development, disease,senescence, and tumorigenesis (25). Single-cell telomere estimationalso may be particularly useful to predict the viability and chro-mosomal stability of oocytes and embryos for women undergoingassisted reproductive technologies.

MethodsSingle-Cell Isolation and DNA Extraction. Human and mouse cell lines werecultured in regular media. Mouse embryos and human germinal vesicleoocytes were cultured in K -modified simplex optimized medium and humantubal fluid, respectively. The single cells were isolated and lysed in a PCRtube without physical purification before the first PCR. Genomic DNA of cellpopulations was extracted by the tissue and blood DNA extraction kit(Qiagen) following the manufacturer’s instructions. The details are shown inSI Methods. See Fig. S6 for derivation of daughter cells and Table S3 for thesequence of primers used by qPCR.

Single-Cell Genomic DNA Amplification by PCR (pre-PCR) with Primers ofTelomere and Reference Gene, and Product Purification. Pre-PCR was per-formed using DNA Polymerase Hot Start Version (TAKARA). The reactionswere set up by aliquoting 38 μL of master mix into the 0.2 mL PCR tubes eachwith 2 μL single-cell genomic DNA. Each reaction was set up with by 4 μL 10×PCR buffer, 4 μL 2.5 mM dNTP, 0.25 μL DNA polymerase, 1 μL each of telo-mere forward and reverse primer (10 μM), and 1 μL each of multicopy geneforward and reverse primer (10 μM) or single gene primer 36B4 (10 μM) andenough water to make up a 40-μL final volume. Thermal cycler reactionconditions were set at 94 °C for 5 min followed by different cycles of 94 °Cfor 15 s, 60 °C annealing for 30 s and extension at 72 °C for 30 s, with a finalextension for 10 min at 72 °C. PCR products were purified following theprotocol of the purification Kit (DNA clean and concentrator-5; Zymo Re-search). Finally the purified PCR products were eluted in 64 μL of doubledistilled water.

Q-PCR Assay for Average Telomere Measurement of a Cell Population. Averagetelomere length was measured from total genomic DNA of human cell linesby using the qPCR method previously described (9). First we followed thesame protocol and chose the single-copy gene (36B4) as the reference. Inaddition, we chose multicopy genes (Alu and 18S rRNA) as reference genesand measured the average telomere length of different human cell lines.Each reaction included 10 μL 2× SYBR Green mix (Bio-Rad), 0.5 μL each of10 μM forward and reverse primers, 4 μL molecular-filter water and 5 μLgenomic DNA (7 ng/μL) to yield a 20-μL reaction. DNA samples were placedin adjacent three wells of a 96-well plate for telomere primers and ref-erence gene primers, respectively. A Bio-Rad thermocycler (CFX systemtest) was used with reaction conditions of 95 °C for 10 min followed by 40cycles of data collection at 95 °C for 15 s, 60 °C anneal for 30 s and 72 °Cextend for 30 s along with 80 cycles of melting curve from 60 °C to 95 °C.To serve as a reference for standard curve calculation, mouse TTF and humanfibroblast were serially diluted and qPCR performed as described above.

After thermal cycling was completed, the CFX manager software was usedto generate standard curves and Ct values for telomere signals and referencegene signals. Here we used different reference genes for telomere lengthmeasurement, and each sample of DNA had one telomere signal and ref-erence gene signal. The average telomere length was termed the T/R ratioinstead of the T/S ratio.

Q-PCR Assay for Single-Cell Telomere Measurement. Telomere length of singlecells was measured by qPCR after pre-PCR, purification and aliquoting. Thereaction was set up with SYBR Green I in 96-wells plates. The purifiedproducts of Pre-PCR for each single cell were aliquoted with 5 μL into eachwell of a 96-well plate. Three repeat reactions were performed for eachsample plus each pair of primers. The final master mix of each well in theqPCR was 10 μL 2× SYBR Green mix, 0.5 μL each of forward and reverseprimer (10 μM for Tel, Alu, B1, and 36B4) and 4 μL of molecular-filteredwater. The qPCR conditions were the same as above. The results wereanalysis by CFX manager software and the relative telomere length of singlecells was calculated by the T/R ratio.

Fig. 5. Variations of telomere length in single cells. (A and B) A summary oftelomere length for multiple individual cells measured by SCT-pqPCR (A) andQ-FISH (B), shown as box-plot. The small circle represents an outlier; theasterisk represents extreme value. The result for each individual cell is shownin Fig. S3. F171 and F200 are human lung fibroblast from a 14-wk gestationand a 71-y-old donor, respectively. P16, P31, P7, and P12 refer to passagenumber. HeLa S3: human cancer cell. RuES2: human embryonic stem cell.U2OS and SaOS2: human osteosarcoma. Single-cell telomere length of hu-man fibroblasts shortened with the age and passage number, and the het-erogeneity of F171 and F200 increased with the passage. The later passageF171 P31 and F200 P12 were unavailable for metaphase telomere Q-FISHbecause of cell senescence and compromised division. (C) Telomere length-ening of single cells derived from mouse zygotes to four-cell embryos bySCT-pqPCR. Z, zygote; 2c, two-cell embryo; 4c, four-cell embryo. n = 6. (D)Analysis of average telomere length in different stage mouse embryos byStudent Newman–Keuls test shows that telomere length elongates fromzygote to four-cell embryo. n = 4. Bar is ± SD.

Table 1. Variation of single-cell telomere length by SCT-pqPCRand Q-FISH in different cell lines

Cell line

T/R ratioQ-FISH (fluorescence

intensity)

Mean SD CV Mean SD CV

F171 P16 0.845 0.143 0.169 422.652 77.626 0.184F171 P31 0.749 0.364 0.486 — — —

F200 P7 0.710 0.165 0.233 249.855 67.251 0.269F200 P12 0.724 0.289 0.398 — — —

RuES2 1.274 0.299 0.234 578.253 138.486 0.239HeLa S3 0.680 0.200 0.294 339.530 113.267 0.334SaOS2 1.202 0.291 0.242 625.677 240.469 0.384U2OS 1.872 0.419 0.224 1550.925 468.434 0.302

The cycle number of pre-PCR is 17. n = 6.

Wang et al. PNAS | Published online May 9, 2013 | E1911

GEN

ETICS

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021

Telomere Q-FISH. Telomere Q-FISH and quantification were performed asdescribed previously (11, 22).

Southern Blot Analysis of TRF. Telomere restriction fragment analysis wasperformed as described by using the TeloTAGGG Telomere Length Assay Kit(Roche) according to the protocol provided by the manufacturer. The meantelomere length was calculated using the following (26): TRF = ∑(ODi)/∑(ODi/Li), where ODi is the chemiluminescent signal and ODi/Li is the lengthof the TRF at position.

Data Statistics. All of the data statistics were obtained by SPSS 13.0 software.The P value for comparison of two groups was derived from the in-dependent-samples t test. The multiple groups’ statistical data were ana-

lyzed with one-way ANOVA. The frequency of telomere distribution wascompared by a nonparametric test. The correlation analysis was performedby the Pearson correlation test; α is set at 0.05 for all tests.

ACKNOWLEDGMENTS. We thank Susan Smith (New York University LangoneMedical Center) for providing the U2OS and Hela S3 cancer cell line, andBrigitte L. Arduini (Laboratory of Molecular Vertebrate Embryology, HumanPluripotent Stem Cell Core Facility (The Rockefeller University) for providingthe human embryonic stem cell RuES2. This work was supported by fundingfrom the New York University Department of Obstetrics and Gynecology andthe Clinical and Translational Science Institute at New York University(NIH1UL1RR029893); the Most National Major Basic Research Program(2009CB941000, 2011CBA01002); and National Institute of Health Grants1P01GM099130-01 and 1R21HD066457-01.

1. Blackburn EH (2000) Telomere states and cell fates. Nature 408(6808):53–56.2. Blackburn EH (2001) Switching and signaling at the telomere. Cell 106(6):661–673.3. Wright WE, Shay JW (2001) Cellular senescence as a tumor-protection mechanism: The

essential role of counting. Curr Opin Genet Dev 11(1):98–103.4. Gomes NM, et al. (2011) Comparative biology of mammalian telomeres: Hypotheses

on ancestral states and the roles of telomeres in longevity determination. Aging Cell10(5):761–768.

5. Marión RM, Blasco MA (2010) Telomeres and telomerase in adult stem cells andpluripotent embryonic stem cells. Adv Exp Med Biol 695:118–131.

6. Lee HW, et al. (1998) Essential role of mouse telomerase in highly proliferative organs.Nature 392(6676):569–574.

7. Aubert G, Hills M, Lansdorp PM (2012) Telomere length measurement-caveats anda critical assessment of the available technologies and tools. Mutat Res 730(1-2):59–67.

8. Allshire RC, Dempster M, Hastie ND (1989) Human telomeres contain at least threetypes of G-rich repeat distributed non-randomly. Nucleic Acids Res 17(12):4611–4627.

9. Cawthon RM (2002) Telomere measurement by quantitative PCR. Nucleic Acids Res30(10):e47.

10. Lansdorp PM, et al. (1996) Heterogeneity in telomere length of human chromosomes.Hum Mol Genet 5(5):685–691.

11. Poon SS, Martens UM, Ward RK, Lansdorp PM (1999) Telomere length measurementsusing digital fluorescence microscopy. Cytometry 36(4):267–278.

12. Zijlmans JM, et al. (1997) Telomeres in the mouse have large inter-chromosomalvariations in the number of T2AG3 repeats. Proc Natl Acad Sci USA 94(14):7423–7428.

13. Baerlocher GM, Lansdorp PM (2003) Telomere length measurements in leukocytesubsets by automated multicolor flow-FISH. Cytometry A 55(1):1–6.

14. Baird DM, Rowson J, Wynford-Thomas D, Kipling D (2003) Extensive allelic variationand ultrashort telomeres in senescent human cells. Nat Genet 33(2):203–207.

15. Canela A, Vera E, Klatt P, Blasco MA (2007) High-throughput telomere lengthquantification by FISH and its application to human population studies. Proc Natl

Acad Sci USA 104(13):5300–5305.16. Entringer S, et al. (2011) Stress exposure in intrauterine life is associated with shorter

telomere length in young adulthood. Proc Natl Acad Sci USA 108(33):E513–E518.17. Epel ES, et al. (2004) Accelerated telomere shortening in response to life stress. Proc

Natl Acad Sci USA 101(49):17312–17315.18. Dolezel J, Bartos J, Voglmayr H, Greilhuber J (2003) Nuclear DNA content and genome

size of trout and human. Cytometry A 51(2):127–128, author reply 129.19. Dean FB, et al. (2002) Comprehensive human genome amplification using multiple

displacement amplification. Proc Natl Acad Sci USA 99(8):5261–5266.20. Pan X, et al. (2008) A procedure for highly specific, sensitive, and unbiased whole-

genome amplification. Proc Natl Acad Sci USA 105(40):15499–15504.21. Wang F, et al. (2012) Molecular insights into the heterogeneity of telomere

reprogramming in induced pluripotent stem cells. Cell Res 22(4):757–768.22. Liu L, et al. (2007) Telomere lengthening early in development. Nat Cell Biol 9(12):

1436–1441.23. Keefe DL, Liu L, Marquard K (2007) Telomeres and aging-related meiotic dysfunction

in women. Cell Mol Life Sci 64(2):139–143.24. Amit M, et al. (2000) Clonally derived human embryonic stem cell lines maintain

pluripotency and proliferative potential for prolonged periods of culture. Dev Biol

227(2):271–278.25. Ma H, et al. (2011) Shortened telomere length is associated with increased risk of

cancer: A meta-analysis. PLoS ONE 6(6):e20466.26. Kimura M, et al. (2010) Measurement of telomere length by the Southern blot

analysis of terminal restriction fragment lengths. Nat Protoc 5(9):1596–1607.

E1912 | www.pnas.org/cgi/doi/10.1073/pnas.1306639110 Wang et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

021