RESEARCH ARTICLE

Telomere Dysfunction Drives Chromosomal Instabilityin Human Mammary Epithelial Cells

David Soler, Anna Genesca, Gema Arnedo, Josep Egozcue, and Laura Tusell*

Unitatde Biologia Cellular,Departamentde Biologia Cellular,Fisiologia i Immunologia,Universitat Auto' noma de Barcelona,08193 Bellaterra,Spain

The development of genomic instability is an important step toward generating the multiple genetic changes required for can-

cer. Telomere dysfunction is one of the factors that contribute to tumorigenesis. Telomeres shorten with each cell division in

the absence of telomerase. Human mammary epithelial cells (HMECs) obtained from normal human tissue demonstrate two

growth phases. After an initial phase of active growth, HMECs exhibit a growth plateau termed selection. However, some cells

can emerge from this growth plateau by spontaneously losing expression of the p16INK4a protein. These post-selection HMECs

are capable of undergoing an additional 20–50 population doublings in culture. Continued proliferation of these post-selection

HMECs leads to further telomere erosion, loss of the capping function, and the appearance of end-to-end chromosome

fusions that can enter bridge-fusion-breakage (BFB) cycles, generating massive chromosomal instability before terminating in a

population growth plateau termed agonescence. We have found that the chromosome arms carrying the shortest telomeres

are those involved in telomere–telomere type rearrangements. In addition, we found that the risk of a particular chromosome

being unstable differs between individuals. Most importantly, we identified sister chromatid fusion as a first event in generating

genomic instability in HMECs. During post-selection HMEC growth, double strand breaks are generated by both fused chro-

mosomes as well as individual chromosomes with fused chromatids entering BFB cycles. These broken chromosome extrem-

ities are able to join other broken ends or eroded telomeres, producing massive chromosomal instability at the later passages

of the cell culture. This article contains Supplementary Material available at http://www.interscience.wiley.com/jpages/1045-

2257/suppmat. VVC 2005 Wiley-Liss, Inc.

INTRODUCTION

Telomeres are specialized structures at chromo-

some ends that are essential for maintaining the

stability of eukaryotic genomes. The primary role

of telomeres is to protect chromosome ends from

recombination and to enable the cells to distin-

guish between random DNA breaks and natural

chromosome ends. Telomeric sequences are added

to the ends of chromosomes by the ribonucleopro-

tein enzyme telomerase. Most somatic human tis-

sues and primary cells possess low or undetectable

telomerase activity, and telomeres shorten with

each cell division in vivo and in vitro. When telo-

meres shorten below a critical length, they lose

their capacity to provide an adequate cap to the

chromosome end. In human primary cells, it has

been suggested that short telomeres signal entry

into replicative senescence or Hayflick limit (Har-

ley et al., 1990; Bodnar et al., 1998). Viral oncopro-

tein-induced inactivation of p53 and pRb allows

for extended cell division (reviewed by Wright and

Shay, 1995). Continued proliferation of cells

beyond the Hayflick limit leads to further telomere

erosion, loss of the capping function, and entry into

a phase of rampant chromosomal instability that is

associated with increased numbers of telomere

associations, dicentric chromosomes, ring chromo-

somes, and massive cell death (Counter et al.,

1992; Rogan et al., 1995). The massive genetic

instability associated with this stage may well be

the mechanism by which unusual cells acquire the

constellation of genomic alterations needed for

malignant transformation (de Lange et al., 1990;

Hastie et al., 1990; Counter et al., 1992; Harley

et al., 1992). Although most cells transiting through

crisis die, rare survivor cells emerge that can main-

tain telomere length through the activation of

either telomerase or the alternative lengthening of

telomeres (ALT) mechanisms (Shay and Bacchetti,

1997; Henson et al., 2002), which may pave the

transition to a more stable genome in which more

subtle changes promote tumor progression.

Studies of the mTerc�/� mouse model have cor-

related the genomic instability resulting from telo-

mere dysfunction with the appearance of carcino-

mas (Artandi et al., 2000; Hackett et al., 2001).

*Correspondence to: Laura Tusell. E-mail: laura.tusell@uab.es

Supported by: Research grant sponsors FIGH-CT2002-217,2001SGR-00202, FI6R-CT-2003-508842, and SAF2002-11833-E.

Received 7 March 2005; Accepted 10 June 2005

DOI 10.1002/gcc.20244

Published online 28 July 2005 inWiley InterScience (www.interscience.wiley.com).

VVC 2005 Wiley-Liss, Inc.

GENES, CHROMOSOMES & CANCER 44:339–350 (2005)

Ageing mTerc�/� p53�/� mutant mice exhibited a

pronounced shift in their tumor spectrum to carci-

nomas of the breast, colon, and skin. This result,

together with the fact that telomeres in human can-

cer cells are often significantly shorter than their

normal tissue counterparts (Pathak et al., 1994; de

Lange, 1995; van Heek et al., 2002; Meeker et al.,

2004), has fuelled speculation that telomere erosion

might be a risk factor in the genesis of certain

tumors. A dysfunctional telomere-induced genomic

instability model was recently proposed to explain

the origin of epithelial cancers (O’Hagan et al.,

2002). It has been suggested that end-to-end

fusions originated by telomere shortening could

generate the important chromosome instability that

seems necessary to initiate epithelial carcinogene-

sis (Artandi et al., 2000; Gisselsson et al., 2001;

Hackett et al., 2001). Chromosomal instability is

intimately linked to cancer development, as it is

thought to generate chimeric genes or to deregulate

oncogenes, and induce changes in the gene dosage

needed for epithelial carcinogenesis.

Most human breast cancers originate from mam-

mary epithelial cells (HMECs). Normal HMECs

cultured in serum-free medium exhibit an initial

growth phase, followed by a transient growth plateau

in which most cells show proliferative arrest while a

small number of cells maintain good growth. These

post-selection cells, which present functional TP53

(Romanov et al., 2001) and do not express CDKN2A

mRNA and protein (Brenner et al., 1998), are capa-

ble of undergoing further population doublings

(PDs) in culture. It has been shown that post-selec-

tion HMECs exhibit eroded telomeric sequences

and ultimately enter agonescence, where the types

of chromosomal abnormalities seen in the earliest

lesions of breast cancer are generated (Romanov

et al., 2001). Recently, it has been observed that the

genomic events occurring in cultures of HMECs

before, during, and after ZNF217-mediated immor-

talization were remarkably similar to those occurring

in breast cancer during progression from usual ductal

hyperplasia (UGH) to ductal carcinoma in situ

(DCIS) (Chin et al., 2004). To evaluate fully the

telomere-based crisis of post-selection HMECs, we

have performed an exhaustive cytogenetic analysis

of these cells during their passage in culture. Our

results confirm that the set of chromosomes with the

shortest telomeres are those chromosomes most

prone to fusion events and demonstrate that the

chromosomes with the shortest telomeres are specific

to each individual. Most importantly, our results pro-

vide the first evidence that chromosomal instability

in HMECs is initiated by chromatid-bridge-fusion-

breakage cycles of individual chromosomes that

result in chromosomal amplification and deletion.

Likewise, primary telomere–telomere fusions

between different chromosomes arising through the

culture of cells can also enter BFB cycles. Therefore,

the broken chromosome extremities that are gener-

ated are not only able to join another broken chromo-

some extremity but also to join uncapped chromo-

somes, giving rise to the extensive chromosomal

instability observed at the later passages of HMECs.

MATERIALS ANDMETHODS

Cells and Culture Conditions

Post-selection 219-7 HMECs (BioWhittaker,

Walkersville, MD) and AG11137A HMECs (Coriell

Cell Repositories, Camden, NJ) were derived from

normal breast tissue. Cells were cultured in serum-

free MEGM medium (BioWhittaker) supplemented

with epidermal growth factor, insulin, hydrocorti-

sone, gentamicin/amphotericin-B, and bovine pitui-

tary extract. The cells were counted, plated at 2 3105 cells per 75 cm2 flask, and grown at 378C in a 5%

CO2 atmosphere. The number of accumulated PDs

per passage was determined using the equation PD

¼ PDinitial þ log (n8 viable cells harvested/n8 viablecells plated)/log 2. The finite life span of post-selec-

tion HMECs when cultured in MEGM medium

is about 22 passages, equivalent to �65–75 PDs

(Hammond et al., 1984; Stampfer, 1985).

Obtaining of Metaphase and Binucleated Cells

Exponentially growing HMECs were treated

with Colcemid 0.02 lg/ml for 8 hr, followed by

hypotonic shock and methanol/acetic fixation. Cell

suspensions were dropped onto clean slides, which

were stored at �208C. Before hybridization, the

slides were mounted with DAPI staining. Meta-

phase karyotyping was performed by reverse DAPI

staining, which results in a reproducible G-band-

like pattern that allows accurate individual chro-

mosome identification. Binucleated HMECs were

obtained after adding cytochalasin B (Sigma, St.

Louis, MO) before harvesting the culture, so as to

block cytokinesis (Ponsa et al., 2001).

Fluorescence In Situ Hybridization (FISH)

PNA-FISH

Telomere and centromere PNA-FISH were per-

formed using a Cy3-(CCCTAA)3 PNA-probe for

telomeres and a FITC-AAACACTCTTTTTGT-

AGA probe for centromeres (PE Biosystems, Foster

City, CA) (Lansdorp et al., 1996; Martın et al., 2003).

340 SOLER ETAL.

Subtelomeric 1q-FISH, chromosome painting,

and M-FISH

After PNA-FISH, the same slides were used to per-

form an in situ hybridization with the subtelomeric

1q probe (Vysis, Downers Grove, IL) and the paint-

ing of chromosome 1 (Vysis). A multiplex FISH

(Vysis) was performed at PD59 of the cells because

of highly rearranged karyotypes. All probes were

applied in accordance with the manufacturer’s

instructions. After hybridization and counterstaining,

fluorescence signals were visualized under a micro-

scope equipped with epifluorescent optics, and were

captured and analyzed using Cytovision software

(Applied Imaging, Inc., Santa Clara, CA)

Comparative Genome Hybridization (CGH)

Control and 219-7 HMECs DNA at different

PDs were labeled with Spectrum Red-dUTP and

Spectrum Green-dUTP by nick translation, using a

commercial kit (Vysis). Subsequently, equal

amounts of control and HMEC-labeled probes

(700 ng) were hybridized to normal metaphase

cells, as previously described by Prat et al. (2001).

Fluorescent hybridization signals and DAPI-stain-

ing patterns were captured using MetaSystems Isis

V5.C software (MetaSystems, Altlussheim, Ger-

many). Ratio values obtained from at least 10

metaphase spreads were averaged, and the result-

ing profiles with 95% confidence intervals were

plotted next to the chromosomal ideograms.

Telomerase Activity Assay

Protein extracts from 219-7 HMECs were

obtained using a CHAPS detergent method (Kim

et al., 1994). Telomerase activity was assessed in

HMECs at different PDs, as described by Martın

et al. (2003).

Western Blot Analysis

Protein extracts for Western blot analysis were

obtained as previously explained for the telomer-

ase-activity assay. Protein concentrations were

determined using the Bradford assay (Bio-Rad).

Lysates were boiled for 5–10 min, and 50 lg of pro-

tein was electrophoresed in an SDS-polyacryl-

amide gel and transferred to nitrocellulose mem-

branes by electroblotting. A mouse anti-p16INK4a,

Clone DCS-50.1/A7 (Neomarkers, Fremont, CA)

antibody was used as a probe. Blots were then

probed with alkaline-phosphatase-conjugated goat

anti-mouse antibodies (Bio-Rad, Hercules, CA)

and detection was performed by colorimetric sub-

strates BCIP/NBT.

Immunofluorescence

HMECs and human-skin fibroblasts were grown

on chamber slides, washed twice in PBS, fixed in

methanol, and kept at �208C. After washing with

PBS-1%Triton X-100, cells were incubated with

blocking buffer (PBS-0.1%Triton X-100-2% goat

serum) for 30 min. Incubation was then performed

with mouse anti-cytokeratin 8 (C5301, Sigma) and

mouse anti-vimentin (MU074-UC, Biogenex, San

Ramon, CA) antibodies in blocking buffer for 1 hr at

room temperature. After three washes in PBS-

0.1%Triton X-100, the secondary antibody (rabbit

anti-mouse Alexa 488, Molecular Probes) in blocking

buffer was applied for 1 hr. After three washes, the

slides were dehydrated, left to dry, and mounted on a

Vectashield containing DAPI. Immunofluorescence

signals were visualized under a microscope equipped

with epifluorescence optics and a camera.

Cytogenetic Analysis

Cytogenetic analysis in 219-7 HMECs was per-

formed at PD32, PD37, and PD59 to monitor the

karyotype evolution of these cells in culture,

whereas in AG11137A HMECs, it was performed

only at PD43. Chromosomal aberrations were clas-

sified according to five different categories: (1)

telomere–telomere fusions, when two eroded chro-

mosome extremities fuse; (2) telomere–double-

strand break (DSB) fusions, when an eroded telo-

mere joins a broken chromosome; (3) DSB–DSB

fusions, when two broken extremities join to pro-

duce a conventionally rearranged chromosome; (4)

deleted chromosomes in which no signs of joining

are observed; and (5) others.

Statistical Analysis

A v2 test was applied to determine whether

there was a relationship between the increased fre-

quencies of different types of chromosomal aberra-

tions throughout the culture of HMECs. Expected

frequencies of chromosomal aberrations per PD

were calculated assuming a homogeneous distribu-

tion of aberrations among cells at the different PDs

analyzed. For all types of aberrations (tel–tel; tel–

dsb; dsb–dsb and del), the v2 tests provided evi-

dence of highly significant differences between

expected and observed frequencies of cells with

aberrations at the different PDs analyzed.

RESULTS

Post-selection HMECs were derived from the

mammary reduction of two women. Although fibro-

blasts do not proliferate well when cultured in

341KARYOTYPE EVOLUTION OF HUMAN MAMMARY EPITHELIAL CELLS

serum-free MEGM medium, an immunofluores-

cence study with anti-cytokeratin 8 and anti-vimen-

tin antibodies was performed to confirm the epithe-

lial origin of the cells used in our study. Most of these

were positive for cytokeratin 8, providing evidence of

their epithelial origin, but they also expressed vimen-

tin. In contrast, cultured-skin fibroblasts were only

positive for vimentin antibodies (see supplementary

Fig. 1; Supplementary material for this article can be

found on the Genes, Chromosomes, and Cancer websiteat http://www.interscience. wiley.com/jpages/1045-

2257/suppmat/index.html). It has been reported that

post-selection epithelial cells gradually acquire a

luminal epithelial phenotype with an increasing

expression of cytokeratin 8 (Taylor-Papadimitriou et

al., 1989), and that epithelial cell cultures—in addi-

tion to cytokeratins—coexpress vimentin as strongly

as do cultured fibroblasts (Mork et al., 1990).

HMECs emerging from the first plateau lose

CDKN2A protein expression (Brenner et al., 1998;

Huschtscha et al., 1998). To ascertain whether this

was the case for our cells, CDKN2A was assessed in

serial subcultures of HMECs and, in accordance

with previous reports, no protein expression was

observed (data not shown). The TP53 gene

sequence is wild-type in these cells and still func-

tional (Romanov et al., 2001). These post-selection

HMECs that lack telomerase activity have a divi-

sion potential of 65–75 PDs before they reach a

telomere-based crisis (Hammond et al., 1984;

Stampfer, 1985). To confirm the absence of telomer-

ase activity in our HMECs throughout the post-

selection culture, a Telomere Repeat Amplification

Protocol (TRAP) was applied at different PDs. At

all PDs analyzed, protein extracts of HMECs lacked

telomerase activity (see supplementary Fig. 2).

Figure 1. Distribution of telomere signal-free ends and chromosomal aberrations among chromosomesthroughout the culture of 219-7 HMECs at PD32 (A), PD37 (B), and PD59 (C), respectively.

342 SOLER ETAL.

The karyotype evolution of 219-7 HMECs was

followed by means of an exhaustive cytogenetic

analysis, including conventional cytogenetic analy-

sis by reverse DAPI banding pattern, subtelomeric,

telomeric, and centromeric hybridization, chromo-

some painting, and M-FISH protocols. Cytoge-

netic, immunologic, and molecular analyses of 219-

7 HMECs were performed at three different PDs

before these cells entered a telomere-based crisis.

To find the connection between the progressive

telomere erosion of individual chromosomes and

their fate, we first examined the metaphase cells

for the presence of chromosome ends with unde-

tectable TTAGGG hybridization signals. This

parameter indicates critical telomere shortening

more precisely than do the average telomere

length values (Hemann et al., 2001; Espejel et al.,

2002). To avoid obscuring critically short telo-

meres, we used the counts for telomere signal-free

ends on the homologous chromosome arms in each

analyzed metaphase cell to describe the profile of

critically short telomeres. At PD32, the shortest

telomeres were mainly located on the long arm of

one chromosome 1 and also on the short arm of one

chromosome 22, as these were the chromosome

arms that most frequently lacked labeled telo-

meres in 219-7 HMECs at this passage (Fig. 1A).

Because the corresponding homologues cannot be

formally distinguished, the data for the same arms

of both homologues was pooled for consistency in

presenting whole-genome data. To determine

whether chromosomes with eroded telomeres

retained their subtelomeric region, in situ hybrid-

ization with the subtelomeric 1q probe was per-

formed. Only one of the 73 long arms of chromo-

some 1 lacking visible telomeric signals failed to

label with the subtelomeric probe.

At PD32, an initial passage after selection

�29% of metaphase cells in 219-7 HMECs

showed chromosomal abnormalities (Table 1).

These were mainly primary chromosomal aberra-

tions of the telomere–telomere fusion type origi-

nating from two different chromosomes (0.149 per

metaphase cell, Table 2), giving rise mainly to

dicentric chromosomes. In accordance with their

origin through the erosion of telomeric DNA, the

chromosomes involved in telomere–telomere

fusions were not randomly distributed and

occurred preferentially between the chromosome

arms that most frequently lacked telomeric signals.

A total of 25 dicentric chromosomes were observed

at PD32 219-7 HMECs. Nearly all fusions affected

the long arm of chromosome 1 or the short arm of

chromosome 22 (Fig. 2 and supplementary Table

1). In situ hybridization with the subtelomeric spe-

cific probe for 1q revealed that most fused chromo-

some 1 long arms retained their subtelomeric

region, indicating that there was limited erosion of

Figure 2. Partial metaphase chromosomesshowing telomere–telomere fusions at PD32 219-7HMECs. (A) Reverse DAPI banding. (B) PNAhybridization of telomeres and centromeres. (C)Subtelomeric 1q hybridization.

TABLE 1. Chromosomal Aberrations in 219-7 HMECsat Different PDs

PD

Number ofmetaphase

cells analyzed

Number ofcells with

aberrationsaNumber ofaberrationsb

32 195 56 (29) 65 (0.33)37 91 39 (43) 49 (0.54)59 34 34 (100) 136 (4.00)

aValues in parentheses indicate frequency per 100 cells (in percentage).bValues in parentheses indicate frequency per cell.

TABLE 2. Types and Frequencies of Chromosome Aberrationsin 219-7 HMECs at Different PDs

Type

Frequency per metaphase cell

PD32 PD37 PD59

Telomere–telomere 0.149 0.132 0.735Telomere–DSB 0.046 0.231 2.117DSB–DSB 0.015 0.011 0.764Deleted chromosomes 0.071 0.132 0.176Other 0.046 0.033 0.206

343KARYOTYPE EVOLUTION OF HUMAN MAMMARY EPITHELIAL CELLS

unprotected chromosomes before fusion to another

chromosome end.

In addition to primary telomere–telomere fu-

sions, secondary chromosomal aberrations were

also observed at PD32 219-7 HMECs (Table 2).

These consisted mainly of deleted chromosomes

and rearrangements resulting from the non-homol-

ogous fusion of a broken chromosome with an

intact chromosome with eroded telomeres (telo-

mere–DSB fusion). When analyzing the chromo-

somes that were participating in these rearrange-

ments, we observed that 70% of these involved at

least one of the chromosome arms with critically

short telomeres. It is noteworthy that 1q and 22p

were the chromosome arms most frequently

involved in structural aberrations. Further analysis

showed that chromosome 1 was frequently

involved in aberrations as a broken chromosome

(Fig. 1B). The rearranged chromosome 1 contained

partial 1q amplifications as well as complementary

partial 1q deletions (Fig. 3A). These abnormalities

are secondary aberrations and are thought to origi-

nate from individual chromosomes with fused sis-

ter chromatids (SCFs) entering BFB cycles. SCF

events are usually missed during routine cytoge-

netic analysis, because they look like normal chro-

mosomes except for their rounded ends and the

absence of telomeric labeling. However, they can

cause extensive chromosomal instability if the cells

continue dividing. During mitosis, this particular

chromosome creates an anaphase bridge, because

the sister chromatids are actually one continuous

DNA molecule. Analysis of anaphase bridges in

binucleated 219-7 HMECs at PD32 using chromo-

some 1 painting revealed, as expected, that this

chromosome was present on the bridge in most

cases (Fig. 3A). Resolution of the anaphase bridge

generates one chromatid containing two copies of

specific genes oriented as inverted repeats,

whereas the other chromatid contains neither of

these genes. In accordance, large chromosomal

deletions and amplifications of the long arm of

chromosome 1 were observed. The generation of

these new chromosome 1 configurations with open

ends makes them susceptible to further fusion

events with broken or eroded extremities, creating

a prolonged period of chromosome instability. At

PD32, accordingly, dicentric chromosomes and

non-reciprocal translocations (NRTs) between the

deleted or amplified long arm of chromosome 1

and the uncapped short arm of chromosome 22

were also observed, although to a lesser extent

(Fig. 3B). To ascertain whether the observed

amplifications and deletions of 1q could give rise

to net changes in DNA sequence copy number,

CGH analysis was performed. No net differences

were observed in DNA sequence copy number in

the long arm of chromosome 1, which suggests that

there is compensation between gains and losses at

this early PD.

In 219-7 HMECs at PD37, we observed a slight

increase in the frequency of metaphase cells with

chromosomal aberrations (Table 1). At this stage,

the chromosome arms that most frequently lacked

telomeric signals were still 1q and 22p. However,

other chromosome arms such as 6p, 7p, 9q, and 12p

were devoid of telomeric labeling and emerged to

be involved in chromosomal rearrangements, indi-

cating the onset of their structural instability

(Fig. 1C). When analyzing the types of chromoso-

mal aberrations at this PD, we observed that the

frequency of telomere–telomere-type aberrations

was similar to that found at PD32 (Table 2). How-

ever, at PD37, as there were larger number of dif-

ferent chromosomes without telomeric signals,

there was more diversity in the chromosomes

involved in this primary type of aberration (see sup-

plementary Table 1). Secondary chromosomal rear-

rangements of the telomere–DSB type increased

significantly at this PD. These aberrations were

dicentric chromosomes, usually a chromosome 1

with a partial long-arm amplification or deletion

joined to the uncapped p-arm of chromosome 22,

and NRTs involving the same chromosomes.

In 219-7 HMECs at PD59, all cells analyzed had

chromosomal abnormalities, and the number of

aberrations per cell increased greatly (Table 1,

Fig. 4A). At this stage, telomere data were calcu-

lated according to extremities lacking telomeric

signals, plus those eroded telomeric extremities

fused to other extremities, to avoid obscuring

results. At this PD, most chromosomes with crit-

ically short telomeres were fused to other chromo-

somes, forming structural aberrations. The chromo-

some arms that most frequently lacked telomeric

signals were 1p, 5q, 6p, 7q, 8p and q, 9q, 11q, 12p,

16p and q,17p, and 22p (Fig. 1D). An overall

increase in all types of chromosomal aberrations

was observed (Table 2). Chromosomes involved in

telomere–telomere fusions at PD59 were mainly

those that failed to label telomeres at the earlier

PD analyzed, such as 6p, 7p and q, 8p, 9p and q,

and 12p (see supplementary Table 1). Interest-

ingly, no chromosomal rearrangements involving

the long arm of chromosome 1 were found, sug-

gesting a selective disadvantage of cells carrying

1q fusions or the derivative aberrations generated

through their entrance into BFB cycles. Secondary

344 SOLER ETAL.

Figure 3. Chromosome rearrangements resulting from telomereloss and BFB cycles. (A) When telomere erosion occurs before or dur-ing DNA synthesis and no other free extremity is present in a cell, theresulting sister chromatids undergo fusion after DNA replication. Asthis chromosome is a linear DNA molecule, anaphase bridges areformed during cell division. Hybridization of binucleated HMECs withchromosome 1 painting provided evidence that this chromosome waspresent in the bridge. Following unequal breakage and cytokinesis, onedaughter cell receives a chromosome containing an inverted duplicationon its end, while the other cell receives a chromosome with a terminaldeletion. (B) The rearranged chromosomes produced after SCF: del(1q)and amplified(1q) will once again be without a telomere cap and there-fore can re-enter BFB cycles. During the subsequent cell cycle, the rear-

ranged chromosome extremities are able to fuse with a chromosomewith eroded telomeres generating a telomere–DSB type aberration, orto fuse with a DSB generating a DSB–DSB type aberration. The fusionof the deleted or amplified long arm of chromosome 1 with the erodedtelomeres at 22p, generating a telomere–DSB dicentric chromosome, isshown. (C) When telomere erosion affects more than one chromo-some, telomere–telomere fusions can be produced between differentchromosomes before DNA replication. The resulting dicentric chromo-somes can also create an anaphase bridge if a twist between chromatidstakes place. In this case, it might break and could form new chromoso-mal rearrangements by fusion of broken ends with either a brokenchromosome or an unbroken but eroded chromosome, giving rise tocomplex NRTs or dicentric chromosomes.

345KARYOTYPE EVOLUTION OF HUMAN MAMMARY EPITHELIAL CELLS

chromosomal aberrations of telomere–DSB and of

DSB–DSB types increased dramatically at this PD.

Moreover, when analyzing the individual chromo-

somes involved in aberrations as broken chromo-

somes in this stage, we observed that those partici-

pating in primary aberrations (telomere–telomere

fusions) at PD37 were now involved in secondary

aberrations as broken chromosomes (Fig. 3C). The

overbreakage of chromosome 10 (Fig. 1F) was due

to the clonal expansion of an NRT in which the

long arms of chromosome 10 fused to the eroded

telomeres of 22p. At PD59, HMECs showed

highly rearranged karyotypes with a mean of four

aberrations per cell; however, most cells had differ-

ent chromosomal rearrangements, and there were

few identical rearrangements in more than one

Figure 4. (A) Karyotype of 219-7 HMEC at PD59 showing multiple chromosomal aberrations. (B) CGHprofiles of DNA-sequence copy increases (green) and decreases (red) observed in HMECs at PD59.

346 SOLER ETAL.

cell. When these cells were analyzed by CGH, the

ideograms showed genomic imbalances in five

chromosome regions. CGH profiles showed gains

in 1p, 5q, and 13q and decreases in DNA sequence

copy number at 12q and Xq. Imbalances near the

centromeric or telomeric regions of different chro-

mosomes were not considered, as CGH at these

regions usually shows artifactual results (Fig. 4B).

To ascertain whether the profile of critically

short telomeres differs between donors, we

extended our study to HMECs derived from

another woman. When analyzing the metaphase

cells from AG11137A at PD43 for the presence of

chromosome ends with undetectable TTAGGG

hybridization signals, we found that the particular

chromosomes showing preferentially telomere ero-

sion in those cells did not coincide with the ones

observed in 219-7 HMECs. In the AG11137A,

telomere erosion preferentially affected the chro-

mosome arms 1q, 2q, 11p, and 22p (Table 3). With

regard to the types of chromosomal aberrations in

the HMECs from this additional woman, we found

telomere–telomere fusions between the short arm

of chromosome 11 and the short arm of chromo-

some 22.

DISCUSSION

As an overall result, throughout the culture of

post-selection HMECs, there is an increase in the

frequency of metaphase cells with aberrations

(Table 1). In 219-7 HMECs at PD32, 29% of cells

showed chromosomal abnormalities, whereas at

PD59 all metaphase cells analyzed (100%) showed

chromosomal aberrations. Similarly, the number of

aberrations per cell increased throughout the cell

culture. At early PD, approximately 0.4 aberrations

per metaphase cell were observed, whereas at

PD59 the frequency increased to 4.0 aberrations

per cell. These results are in accordance with the

reported accumulation of abnormal cells when

HMECs are near the second growth plateau

(Romanov et al., 2001).

HMECs are able to emerge from a first growth

plateau by spontaneously losing CDKN2A protein

expression (Brenner et al., 1998; Huschtscha et al.,

1998); however, the TP53 gene sequence is wild-

type and still functional (Romanov et al., 2001).

Ongoing proliferation of post-selection HMECs in

the absence of telomerase expression produces

telomere erosion. Progressive telomere shortening

in those chromosomes with shorter telomeres

would leave them uncapped and, therefore, sus-

ceptible to fusion events through non-homologous

end joining (Smogorzewska et al., 2002). Consis-

tent with this, primary chromosomal abnormalities

at PD32 were mainly telomere–telomere fusions,

and the particular chromosomes involved in these

fusion events were the long arm of chromosome 1

and the short arm of chromosome 22, the chromo-

some arms that preferentially showed eroded telo-

meres in 219-7 HMECs. Therefore, these results

support the idea that chromosome arms carrying

the shortest telomeres are those involved in telo-

mere fusions, as has been observed previously in

transformed epithelial cells before reaching crisis

(Deng et al., 2004; der-Sarkissian et al., 2004).

The Set of Chromosomes with the Shortest

Telomeres is Specific to Each Individual

Studies using quantitative fluorescence in situ

hybridization with telomeric PNA probes (Q-

FISH) have demonstrated that individual telomere

lengths in normal somatic cells are heterogeneous

and vary between donors (Lansdorp et al., 1996;

Martens et al., 1998). Consequently, there must be

some chromosomes with shorter telomeres than

others that will differ in different cells or cell lines.

In our study, the analysis of the specific chromo-

somes with the shortest telomeres in HMECs

derived from two women provided evidence that

the set of chromosome arms showing preferentially

eroded telomeres depends on each individual, as

has been shown recently in transformed human

epithelial cell lines (Deng et al., 2004). Chromo-

somes lacking TTAGGG signals in 219-7 HMECs

were located on the long arm of chromosome 1 and

on the short arm of chromosome 22, and telomere-

to-telomere fusions were mainly between 1q and

22p, while in AG11137 HMECs, telomere erosion

preferentially affected chromosome arms 1q, 2q,

11p, and 22p, and telomere–telomere fusions

between 11p and 22p were observed. Thus, the

risk of a particular chromosome arm becoming

unstable differs among individuals. Therefore, this

TABLE 3. Comparison of the Chromosome Arms withPreferentially Telomere Erosion and the Types ofTelomere–Telomere Fusion in HMECs Derived

from Two Donors

HMECs

PD ofcytogeneticanalysis

Frequencyof telomereerosion permetaphasea

Chromosomearms

involved infusion events

219-7 32 1q (72.67) 1q and 22p22p (55.81)

AG11137A 43 11p (66.10) 11p and 22p22p (38.98)

aValues in parentheses are percentages.

347KARYOTYPE EVOLUTION OF HUMAN MAMMARY EPITHELIAL CELLS

variability may be responsible for the extensive

intratumor heterogeneity in the pattern of struc-

tural chromosomal aberrations found in human

neoplasms (Gisselsson et al., 2002), and may also

explain the observed karyotype differences among

tumors of the same type.

Initiation of Chromosomal Instability by Sister

Chromatid Fusion and End-to-End Fusions

When telomere erosion occurs, chromosomes

remain unstable until they are capped. In most

instances, uncapped chromosomes will join

another broken or eroded chromosome end, pro-

ducing dicentric chromosomes or NRTs. Accord-

ingly, in the 219-7 HMECs analyzed at PD32,

dicentric chromosomes resulting from the fusion of

two different chromosomes with eroded telomeres

were the most frequent type of chromosomal aber-

ration. These dicentric chromosomes can promote

the formation of new chromosomal aberrations by

entering BFB cycles (Fig. 3C). However, a signifi-

cant fraction of secondary aberrations observed at

this early passage did not derive from primary telo-

mere–telomere fusions between two different

chromosomes. At PD32 HMECs, we observed 1q

amplifications (Fig. 3A), as well as rearrangements,

resulting from amplified 1q arms fused to other

chromosomes with eroded telomeres (Fig. 3B).

These secondary aberrations, as well as comple-

mentary deletions, cannot result from unequal

breakage of dicentric chromosomes. Chromosome

amplification can result only from fusion of sister

chromatids entering BFB cycles. Thus, the chro-

mosomal aberrations observed originated from

unprotected chromosome 1 that joined its sister

chromatids before PD32. When only one chromo-

some in a cell is unprotected and no other free end

is available, this uncapped chromosome end has no

chance of joining. Then, if this cell replicates its

DNA, a continuous DNA molecule is formed by

covalent fusion of the unprotected sister chroma-

tids. Resolution of SCF by BFB cycles is thus the

first event in generating chromosomal instability in

219-7 HMECs. Later on, due to progressive telo-

mere shortening as HMECs divide in culture, the

number of chromosome arms lacking telomeric sig-

nals increased (Fig. 1C and E). When this occurs,

uncapped extremities can join together, giving rise

to telomere–telomere fusions between different

chromosomes before DNA synthesis occurs (Fig.

3C). As with SCF, the entrance of these aberrations

into BFB cycles is also responsible for promoting

chromosomal instability in HMECs.

Massive Chromosome Instability by Telomere–DSB

and DSB–DSB Aberrations

In telomerase-deficient mice, it has been shown

that shortened telomeres are able to join DNA

breaks induced by radiation (Latre et al., 2003).

When telomeres shorten below a critical length,

the chromosomes become uncapped. Unprotected

ends are sensed by the DNA repair machinery as if

they were DSB ends that might join DSB to form

telomere–DSB-type rearrangements (d’Adda di

Fagagna et al., 2003). This situation is similar to

what we have observed in HMECs. At early PDs,

when broken ends are still scarce, it is very prob-

able that eroded chromosome ends will join with

another eroded telomere, either in a sister chroma-

tid or in a different chromosome. The entrance of

these telomere–telomere fusions into BFB cycle

generates broken chromosome extremities. Conse-

quently, eroded telomeres might join each other

but they might also join DSBs (Fig. 3B and C). At

PD32, the frequency of telomere–DSB fusions was

only 0.046 per metaphase cell (Table 2). However,

the increasing number of chromosomes with

eroded telomeres together with many DSBs gener-

ated by BFB cycles resulted in an increase in sec-

ondary telomere–DSB-type aberrations throughout

the passages in culture (0.231 and 2,117 per meta-

phase cell at PD37 and PD59, respectively). In

addition, telomere–DSB fusion-type aberrations

were not the only type of secondary rearrange-

ments present in these cells. Because of the

increasing number of breaks at the later PDs, a sig-

nificant increase in DSB–DSB fusions was also

observed (0.015 per metaphase cell at PD32 versus

0.764 at PD59).

All metaphase cells showed chromosomal aber-

rations and presented highly rearranged karyo-

types when nearly at the end of the HMECs cul-

turing process (Fig. 4A). However, chromosome

rearrangements affecting 1q were not found, sug-

gesting a selective pressure against cells with 1q

gains and losses. It has been suggested that mas-

sive genetic instability can generate chimeric

genes, deregulate oncogenes, and induce changes

in the gene dosage needed for cancer develop-

ment. CGH analysis of clonal DNA sequence

copy number changes across the entire genome of

219-7 HMECs at PD32 showed neither gains nor

losses in DNA sequences. In contrast, at PD59,

net changes in sequence copy number were

observed for five chromosome arms that, in some

cases, coincided with the imbalances observed

through the cytogenetic analysis (Fig. 4B).

348 SOLER ETAL.

Although extensive changes in gene dosage were

not detected, the massive chromosomal instability

observed is reminiscent of the abundant and het-

erogeneous chromosomal changes observed in

premalignant and malignant breast cancer (Teix-

eira et al., 2002).

Taken together, our results indicate that chromo-

somal instability in HMECs is initiated by SCF of

chromosomes with eroded telomeres. The chromo-

somes involved in primary aberrations depend on

the initial telomere length of individual chromo-

somes, which is highly heterogeneous and varies

among individuals. However, whereas the initial

telomere length at a given chromosome has some

value in predicting the likelihood of that chromo-

some being involved in rearrangements, other fac-

tors such as selective pressure, architecture of indi-

vidual chromosomes, or recombination between

telomeric repeats (Bailey et al., 2004) limit such

correlations. Once initiated, chromosomal instabil-

ity increases through the entry of chromosomal

rearrangements into BFB cycles, leading to com-

plex chromosomal aberrations.

Recently, silencing of CDKN2A has been

observed as a common event in normal breast

specimens (Holst et al., 2003). These cells were

found in discrete foci in a substantial fraction of

women with no indication of or predisposition for

breast cancer (Holst et al., 2003). In HMECs,

two events provide the initiating and promoting

forces that drive the acquisition of a premalig-

nant program: epigenetic modulation of

CDKN2A and chromosomal instability due to

telomere dysfunction (Crawford et al., 2004). It

has been observed that the genomic events

occurring in cultures of HMECs before, during,

and after ZNF217-mediated immortalization were

remarkably similar to those occurring in breast

cancer during progression from UGH to DCIS

(Chin et al., 2004); therefore, the in vitro cell sys-

tem of HMECs will be highly useful in deter-

mining the initial pathways of carcinogenesis in

human breast cancer.

ACKNOWLEDGMENTS

We greatly appreciate the donations of antibod-

ies from A. Ramırez and the generous advice of

R. Miro and J. del Rey for CGH, and J.A. Perez for

Western analysis. We thank M. Puigcerver for tech-

nical assistance and SiMTRAD (the Translation and

Text Correction Service at the UAB’s School of

Modern Languages) for correcting our English. We

specially thank Joan Aurich and M.A. Blasco for

critically reviewing the manuscript.

REFERENCES

Artandi SE, Chang S, Lee SL, Alson S, Gottlieb GJ, Chin L,DePinho RA. 2000. Telomere dysfunction promotes non-recipro-cal translocations and epithelial cancers in mice. Nature 406:641–645.

Bailey SM, Brenneman MA, Goodwin EH. 2004. Frequent recombi-nation in telomeric DNA may extend the proliferative life of telo-merase-negative cells. Nucleic Acids Res 32:3743–3751.

Bodnar AG, Ouellette M, Frolkis M, Holt SE, Chiu CP, Morin GB,Harley CB, Shay JW, Lichtsteiner S, Wright WE. 1998. Extensionof life-span by introduction of telomerase into normal humancells. Science 279:349–352.

Brenner AJ, Stampfer MR, Aldaz CM. 1998. Increased p16 expres-sion with first senescence arrest in human mammary epithelialcells and extended growth capacity with p16 inactivation. Onco-gene 17:199–205.

Chin K, de Solorzano CO, Knowles D, Jones A, Chou W, RodriguezEG, Kuo WL, Ljung BM, Chew K, Myambo K, Miranda M, KrigS, Garbe J, Stampfer M, Yaswen P, Gray JW, Lockett SJ. 2004.In situ analyses of genome instability in breast cancer. Nat Genet36:984–988.

Counter MC, Avilion AA, LeFeuvre CE, Stewart NG, Greider CW,Harley CB, Bacchetti S. 1992. Telomere shortening associatedwith chromosome instability is arrested in immortal cells whichexpress telomerase activity. EMBO J 11:1921–1929.

Crawford YG, Gauthier ML, Joubel A, Mantei K, Kozakiewicz K,Afshari CA, Tlsty TD. 2004. Histologically normal human mam-mary epithelia with silenced p16INK4a overexpress COX-2, pro-moting a premalignant program. Cancer Cell 5:263–273.

d’Adda di Fagagna F, Reaper PM, Clay-Farrace L, Fiegler H, CarrP, von Zglinicki T, Saretzki G, Carter NP, Jackson SP. 2003. ADNA damage checkpoint response in telomere-initiated senes-cence. Nature 426:194–198.

de Lange T. 1995. Telomere dynamics and genomic instability inhuman cancer. In: Blackburn EH, Greider CW, editors. Telo-meres. New York: Cold Spring Harbor Laboratory Press. pp. 265–293.

de Lange T, Shiue L, Myers RM, Cox DR, Naylor SL, Killery AM,Varmus HE. 1990. Structure and variability of human chromo-some ends. Mol Cell Biol 10:518–527.

Deng W, Tsao WS, Guan X-Y, Lucas JN, Si HX, Leung CS, Mak P,Wang LD, Cheung ALM. 2004. Distinct profiles of critically shorttelomeres are a key determinant of different chromosome aberra-tions in immortalized human cells: whole-genome evidence frommultiple cell lines. Oncogene 23:9090–9101.

der-Sarkissian H, Bacchetti S, Cazes L, Londono-Vallejo JA. 2004.The shortest telomere drive karyotype evolution in transformedcells. Oncogene 23:1221–1228.

Espejel S, Franco S, Rodrıguez-Perales S, Bouffler SD, CigudosaJC, Blasco MA. 2002. Mammalian Ku86 mediates chromosomalfusions and apoptosis caused by critically short telomeres. EMBOJ 21:2207–2219.

Gisselsson D, Jonson T, Petersen A, Stromberck B, Cin PA,Hoglund M, Mitelman F, Mertens F, Mandahl N. 2001. Telomeredysfunction triggers extensive DNA fragmentation and evolutionof complex chromosome abnormalities in human malignanttumors. Proc Natl Acad Sci USA 98:12683–12688.

Gisselsson D, Petterson L, Hoglund M, Heidenblad M, GorunovaL, Wiegant J, Mertens F, Cin PD, Mitelman F, Mandahl N. 2002.Chromosomal breakage-fusion-bridge events cause genetic intra-tumor heterogeneity. Proc Natl Acad Sci USA 97:5357–5362.

Hackett JA, Feldser DM, Greider CW. 2001. Telomere dysfunctionincreases mutation rate and genomic instability. Cell 106:275–286.

Hammond SL, Ham RG, Stampfer MR. 1984. Serum-free growth ofhuman mammary epithelial cells: rapid clonal growth in definedmedium and extended serial passage with pituitary extract. ProcNatl Acad Sci USA 81:5435–5439.

Harley CB, Futcher AB, Greider CW. 1990. Telomeres shorten dur-ing ageing of human fibroblasts. Nature 345:458–460.

Harley CB, Vaziri H, Counter CM, Allsopp RC. 1992. The telomerehypothesis of cellular aging. Exp Gerontol 27:375–382.

Hastie ND, Dempster M, Dunlop MG, Thompson AM, Green DK,Allshire RC. 1990. Telomere reduction in human colorectal carci-noma and with ageing. Nature 346:866–868.

Hemann MT, Strong MA, Hao L-Y, Greider CW. 2001. The shortesttelomere, not average telomere length, is critical for cell viabilityand chromosome stability. Cell 107:67–77.

Henson JD, Neumann AA, Yeager TR, Reddel RR. 2002. Alterna-tive lengthening of telomeres in mammalian cells. Oncogene 21:598–610.

349KARYOTYPE EVOLUTION OF HUMAN MAMMARY EPITHELIAL CELLS

Holst CR, Nuovo GJ, Esteller M, Chew K, Baylin SB, Herman JG,Tlsty TD. 2003. Methylation of p16INK4a promoters occurs in vivoin histologically normal human mammary epithelia. Cancer Res63:1596–1601.

Huschtscha LI, Noble JR, Neumann AA, Moy EL, Barry P, MelkiJR, Clark SJ, Reddel RR. 1998. Loss of p16INK4 expression bymethylation is associated with lifespan extension of human mam-mary epithelial cells. Cancer Res 58:3508–3512.

Kim NW, Piatyszek MA, Prowse KR, Harley CB, West MD, Ho PL,Coviello GM, Wright WE, Weinrich SL, Shay JW. 1994. Specificassociation of human telomerase activity with immortal cells andcancer. Science 266:2011–2015.

Lansdorp PM, Verwoerd NP, van de Rijke FM, Dragowska V, LittleMT, Dirks RW, Raap AK, Tanke HJ. 1996. Heterogeneity intelomere length of human chromosomes. Hum Mol Genet 5:685–691.

Latre L, Tusell L, Martın M, Miro R, Egozcue J, Blasco MA, Gen-esca A. 2003. Shortened telomeres join to DNA breaks interferingwith their correct repair. Exp Cell Res 287:282–288.

Martens UM, Zijlmans JM, Poon SS, Dragowska W, Yui J, ChavezEA, Ward RK, Lansdorp PM. 1998. Short telomeres on humanchromosome 17p. Nat Genet 18:76–80.

Martın M, Genesca A, Latre L, Ribas M, Miro R, Egozcue J, TusellL. 2003. Radiation-induced chromosome breaks in AT cellsremain open. Int J Radiat Biol 79:203–210.

Meeker AK, Hicks JL, Iacobuzio-Donahue CA, Montgomery EA,Westra WH, Chan TY, Ronnett BM, De Marzo AM. 2004. Telo-mere length abnormalities occur early in the initiation of epithe-lial carcinogenesis. Clin Cancer Res 10:3317–3326.

Mork C, van Deurs B, Petersen OW. 1990. Regulation of vimentinexpression in cultured human mammary epithelial cells. Differen-tiation 43:146–156.

O’Hagan RC, Chang S, Maser RS, Mohan R, Artandi SE, Chin L,DePinho RA. 2002. Telomere dysfunction provokes regional ampli-fication and deletion in cancer genomes. Cancer Cell 2:149–155.

Pathak S, Dave BJ, Gagos S. 1994. Chromosome alterations in can-cer development and apoptosis. In Vivo 8:843–850.

Ponsa I, Barquinero JF, Miro R, Egozcue J, Genesca A. 2001. Non-disjunction and chromosome loss in gamma-irradiated humanlymphocytes: a fluorescence in situ hybridization analysis usingcentromere-specific probes. Radiat Res 155:424–431.

Prat E, Bernues M, Caballın MR, Egozcue J, Gelabert A, Miro R.2001. Detection of chromosomal imbalances in papillary bladdertumors by comparative genomic hybridization. Urology 57:986–992.

Rogan EM, Bryan TM, Hukku B, Maclean K, Chang AC, Moy EL,Englezou A, Warneford SG, Dalla-Pozza L, Reddel RR. 1995.Alterations in p53 and p16INK4 expression and telomere lengthduring spontaneous immortalization of Li-Fraumeni syndromefibroblasts. Mol Cell Biol 15:4745–4753.

Romanov SR, Kozakiewicz BK, Holst CR, Stampfer MR, Haupt LM,Tlsty T. 2001. Normal human epithelial cells spontaneously escapesenescence and acquire genomic changes. Nature 409: 633–637.

Shay JW, Bacchetti S. 1997. A survey of telomerase activity inhuman cancer. Eur J Cancer 33:787–791.

Smogorzewska A, Karlseder J, Holtgreve-Grez H, Jauch A, de LangeT. 2002. DNA ligase IV-dependent NHEJ of deprotected mam-malian telomeres in G1 and G2. Curr Biol 12:1635–1644.

Stampfer MR. 1985. Isolation and growth of human mammary epi-thelial cells. J Tissue Cult Methods 9:107–116.

Taylor-Papadimitriou J, Stampfer M, Bartek J, Lewis A, Boshell M,Lane EB, Leight IM. 1989. Keratin expression in human mammaryepithelial cells cultured from normal and malignant tissue: relation toin vivo phenotypes and influence of medium. J Cell Sci 94:403–413.

Teixeira MR, Pandis N, Heim S. 2002. Cytogenetic clues to breastcarcinogenesis. Genes Chromosomes Cancer 33:1–16.

van Heek NT, Meeker AK, Kern SE, Yeo CJ, Lillemoe KD,Cameron JL, Offerhaus GJ, Hicks JL, Wilentz RE, Goggins MG,De Marzo AM, Hruban RH, Maitra A. 2002. Telomere shorteningis nearly universal in pancreatic intraepithelial neoplasia. Am JPathol 161:1541–1547.

Wright WE, Shay JW. 1995. Time, telomeres and tumours: is cellu-lar senescence more than an anticancer mechanism? Trends CellBiol 5:293–297.

350 SOLER ETAL.