Genome-wide DNA profiling better defines the prognosis of
chronic lymphocytic leukaemia
Chronic lymphocytic leukaemia (CLL) is the most common
form of adult leukaemia in the Western world (Dighiero &
Hamblin, 2008; Hallek et al, 2008). The disease has a very
heterogenous clinical course and survival after diagnosis can
range from months to decades. For many years the disease was
considered as a chronic non-curable malignancy that required
only a few diagnostic tools and palliative chemotherapy was
limited to chlorambucil with or without corticosteroids.
However, fast and significant therapeutic progress has been
achieved in the past decade by combining different agents;
indeed fludarabine, bendamustine and three monoclonal
antibodies, alemtuzumab, rituximab and ofatumumab have
Andrea Rinaldi,1* Michael Mian,1,2*
Ivo Kwee,1,3* Davide Rossi,4 Clara
Deambrogi,4 Afua A. Mensah,1 Francesco
Forconi,5 Valeria Spina,4 Emanuele
Cencini,5 Daniela Drandi,6 Marco
Ladetto,6 Rita Santachiara,7 Roberto
Marasca,7 Valter Gattei,8 Franco Cavalli,1
Emanuele Zucca,1 Gianluca Gaidano4
and Francesco Bertoni1
1Oncology Institute of Southern Switzerland
(IOSI), Bellinzona, Switzerland, 2Division of
Haematology, Azienda Ospedaliera S. Maurizio,
Bolzano/Bozen, Italy, 3Dalle Molle Institute for
Artificial Intelligence (IDSIA), Manno,
Switzerland, 4Division of Haematology,
Department of Clinical and Experimental
Medicine and IRCAD, Amedeo Avogadro
University of Eastern Piedmont, Novara,
5Division of Haematology, Department of Clinical
Medicine and Immunological Sciences, University
of Siena and AOUS, Siena, 6Haematology,
University of Turin, Turin, 7Division of
Haematology, Department of Oncology and
Haematology, University of Modena and Reggio
Emilia, Modena, and 8Clinical and Experimental
Onco-Haematology Unit, Centro di Riferimento
Oncologico, I.R.C.C.S., Aviano (PN), Italy
Received 30 March 2011; accepted for
publication 31 May 2011
Correspondence: Francesco Bertoni, MD,
Laboratory of Experimental Oncology,
Oncology Institute of Southern Switzerland
(IOSI), via Vincenzo Vela 6, 6500 Bellinzona,
Switzerland.
E-mail: frbertoni@mac.com
*These authors contributed equally to this study.
Summary
The integration of molecular and clinical information to tailor treatments
remains an important research challenge in chronic lymphocytic leukaemia
(CLL). This study aimed to identify genomic lesions associated with a poor
outcome and a higher risk of histological transformation. A mono-
institutional cohort of 147 cases was used as the test series, and a multi-
institutional cohort of 176 cases as a validation series. Genomic profiles were
obtained using Affymetrix SNP 6.0. The impact of the recurrent minimal
common regions (MCRs) on overall survival was evaluated by univariate
analysis followed by multiple-test correction. The independent prognostic
significance was assessed by multivariate analysis. Eight MCRs showed a
prognostic impact: gains at 2p25.3-p22.3 (MYCN), 2p22.3, 2p16.2-p14
(REL), 8q23.3-q24.3 (MYC), losses at 8p23.1-p21.2, 8p21.2, and of the TP53
locus. Gains at 2p and 8q and TP53 inactivation maintained prognostic
significance in multivariate analysis and a hierarchical model confirmed their
relevance. Gains at 2p also determined a higher risk of Richter syndrome
transformation. The prediction of outcome for CLL patients might be
improved by evaluating the presence of gains at 2p and 8q as novel genomic
regions besides those included in the ‘standard’ fluorescence in situ
hybridization panel.
Keywords: leukaemia, prognostic factors, Richter’s syndrome, transforma-
tion, Affymetrix.
research paper
First published online 12 July 2011
doi:10.1111/j.1365-2141.2011.08789.x ª 2011 Blackwell Publishing Ltd, British Journal of Haematology, 154, 590–599
recently been approved by American and/or European regu-
latory agencies for the treatment of CLL (Hallek, 2009). New
insights into CLL molecular genetics have identified several
features that may predict prognosis (Krober et al, 2002;
Stevenson & Caligaris-Cappio, 2004; Hallek, 2008; Hallek
et al, 2008; Dal-Bo et al, 2009; Malcikova et al, 2009; Rossi
et al, 2009a,b; Zenz et al, 2010; Chiorazzi & Ferrarini, 2011).
Nevertheless, major efforts are still required for the identifi-
cation of the biological and genetic features that may allow a
rational application of the new therapies aimed at increased
curability for poor risk patients and that spare toxicity for the
good risk cases. Here, we present a study of two independent
large series of CLL samples analysed with microarray-based
comparative genomic hybridization (arrayCGH), which iden-
tified novel genomic lesions associated with a poor outcome
and transformation to Richter syndrome, possibly allowing a
more precise definition of the prognostic groups in CLL.
Materials and methods
Tumour panel
Chronic lymphocytic leukaemia samples from two large
consecutive series were investigated. A first mono-institutional
cohort consisted of samples with more than 60% of CD19+
cells and was used as the test series. A second multi-
institutional cohort of unselected consecutive CLL cases was
used as the validation series. All samples were collected at
diagnosis. The study was approved by the Bellinzona ethical
committee. CLL diagnosis and management were based on the
National Cancer Institute (NCI) Working group criteria
(Hallek et al, 2008). Cases of monoclonal B-cell lymphocytosis
were discarded.
Baseline clinical and molecular characteristics were collected
as previously described (Rossi et al, 2009a,b). TP53 inactiva-
tion was defined as the presence of TP53 somatic mutations
and/or del(17p13) by fluorescence in situ hybridization (FISH)
or del(17p13) by arrayCGH. FISH was defined as favourable in
the presence of a normal FISH result, or del(13q14.3) or +12,
whilst unfavourable FISH was defined as the presence of
del(11q22-q23) or del(17p13) (Rossi et al, 2009b).
ArrayCGH
DNA samples were obtained as previously reported (Rossi
et al, 2009a; Rinaldi et al, 2011). DNA integrity was verified by
electrophoresis of 50 ng on a 1% agarose gel (Rinaldi et al,
2011). Genomic profiles were obtained using Affymetrix
Human Mapping GeneChip 6.0 arrays (Affymetrix, Santa
Clara, CA, USA) and data mining was performed as previously
described (Mian et al, 2011). Initial array intensity data was
obtained using Affymetrix GeneChip Command Console
Software (AGCC) with 270 Caucasian HapMap samples as a
reference dataset for normalization (Mian et al, 2011).
Affymetrix Power Tools (APT 1.10.2) were used to obtain
genotype data and raw copy number (CN) and segmentation
was performed with the circular binary segmentation algo-
rithm using default settings (Olshen et al, 2004) on
log 2-transformed data (Mian et al, 2011). After normalization
of each profile to a median log 2-ratio of zero, thresholds for
loss and gain were defined as six times the median absolute
deviation symmetrically around zero with an associated
P < 0Æ001 after Bonferroni multiple test correction (Mian
et al, 2011; Rinaldi et al, 2011). Thresholds for homozygous
deletions and amplifications were established heuristically
because the low frequency of these events impeded the
statistical computation of boundaries (Mian et al, 2011).
Profiles were considered as poor quality and discarded from
further analyses if they showed severe over-segmentation
as evaluated by two independent investigators. Segments
containing <10 probes and/or with more than 80% of overlap
with known copy number variations (CNV, Database of
Genomic Variants, http://projects.tcag.ca/variation) were
discarded. Minimal common regions (MCR) were calculated
in the test set alone according to the algorithm by Lenz et al
(2008). Four distinct types of MCR were defined: short
recurrent abnormality (SRA), long recurrent abnormality
(LRA), abnormal chromosome arm (ACA) and abnormal
whole chromosome (AWC). A maximum segment size of
25 Mb and gap size of 500 kb for SRA analysis, and a minimal
segment size of 15 Mb and gap size of 10 Mb for LRA analysis
were applied (Lenz et al, 2008). MCR frequencies were
calculated as the fraction of samples bearing the same
aberration contained in the MCR without regard to its MCR
type. MCRs containing genes encoding the immunoglobulin
heavy chain (IGHV) genes and the kappa and lambda light
chains were discarded since CN changes in these regions
probably represent physiological rearrangements occurring in
B-cells.
Validation of arrayCGH data
DNA losses at 8p, and DNA gains at 2p and 8q were validated
by FISH and by genomic real-time polymerase chain reaction
(PCR). FISH analysis was performed by staining the nuclei of
peripheral white blood cells used for CLL diagnosis. Nuclei
were counterstained with 4¢,6¢-diamidino-2-phenylindole
(DAPI) and antifade, and signals were visualized using an
Olympus BX51 microscope (Olympus Italia, Milan, Italy). For
each probe, at least 400 interphase cells with well-delineated
fluorescent spots were examined. Validation by FISH of gains
at 2p was performed using an LSI MYCN probe (Abbott,
Rome, Italy), as previously described (Deambrogi et al, 2010).
Validation by FISH of gains at 8q was performed using LSI
IGHV/MYC/CEP8 and MYC break-apart probes (Abbott).
Depending on the extent of the deletions and on the
availability of appropriate biological material, validation of
losses affecting 8p was done by FISH using the LSI LPL probe
(Abbott) (8p21.3) or by real-time PCR on genomic DNA using
the TaqMan Copy Number Assay Hs03275475_cn (Applied
Genomic Lesions and Outcome in CLL
ª 2011 Blackwell Publishing Ltd, British Journal of Haematology, 154, 590–599 591
Biosystems, Rotkreuz, Switzerland) (8p22). Real-time PCR
reactions were run in quadruplicate in 96-well plates with 5 ng
of DNA per reaction in a StepOnePlus instrument and
analysed with CopyCaller software v.1.0 (Applied Biosystems,
Carlsbad, CA, USA).
Statistical analyses
Overall survival (OS) was measured from the date of diagnosis
to the last follow-up or death by any cause. Univariate analysis
was performed with the log-rank test in order to evaluate the
impact of the genetic aberrations on OS. The Benjamini-
Hochberg-method (q-value) was used to control false positive
rate. MCR showing an unadjusted P < 0Æ05 and a q-value
< 0Æ10 and occurring in at least five cases in the test series
underwent univariate analysis in an independent validation
series. For MCR residing in a larger type of MCR (e.g., a SRA
inside a LRA), with both MCR satisfying the significance
criteria in the test series, only the smaller type of MCR was
evaluated in the validation series, and patients carrying a lesion
contained in both a SRA and in a LRA were counted for both
types of MCR (Scandurra et al, 2010a; Rinaldi et al, 2011).
MCRs with a P < 0Æ05 and a q-value < 0Æ1 in the validation
series were considered as statistically significant. As the samples
of the validation set had not been used to calculate MCR, each
sample was classified as bearing or not bearing the MCR
selected in the test series: an overlap of >85% of the probes
reporting the specific aberration was used for SRA and LRA
types, an overlap of >60% for ACA and AWC types.
To identify MCR with an independent impact on survival, a
Cox proportional hazard regression was applied with backward
stepwise selection algorithm. The multivariate analysis was
performed on the whole population of patients to maximize
the statistical power (Janoueix-Lerosey et al, 2009).
A hierarchical model for OS was built based on genetic
features. Patients were classified into different genetic groups
sorted by hazard ratios as obtained with a Cox regression
analysis including the standard karyotype groups (Dohner
et al, 2000) plus the presence of the novel parameters
significant by multivariate analysis. Associations in two way
tables were tested for statistical significance using either the v2
test or Fisher exact test (two-tailed), as appropriate. Statistical
analyses were performed with Statistical Package for the Social
Sciences software v.17.0.2 (SPSS, Chicago, IL, USA) and R
statistical package (version 2.11.1) (R Development Core Team
2009). P values < 0Æ05 were considered statistically significant.
Results
DNA copy number profiles of two independent sets of CLL
identifies genomic aberrations with an impact on outcome
To identify DNA CN changes affecting the outcome of newly
diagnosed CLL patients, we followed a test-validation
approach (Fig S1). Recurrent genomic lesions were first
identified across a test series of 158 CLL patients from a
single institution and tested for their prognostic significance
on the 147 cases with follow-up data. The statistically
significant lesions were then evaluated in an independent
validation group of 176 consecutive CLL patients, collected
from three additional institutions. The most important
clinical characteristics at the time of diagnosis were compa-
rable in both series and reflected the general CLL population
(Table I). Also, no differences could be observed in terms of
the percentage of patients needing treatment (81/147, 55%,
vs. 100/176, 57%, P = n.s.) or for patients receiving a
fludarabine-containing regimen (51/147, 35%, vs. 51/176,
29%, P = n.s.).
Three hundred and sixty-seven MCRs were evaluated for
their prognostic impact on OS in the test series (Fig S2; Table
SI). Univariate analysis followed by multiple test correction
identified 14 MCRs with a statistically significant impact on
OS. Eight MCRs confirmed the prognostic impact on OS in the
validation group (Table II): 2p25.3-p22.3 gains (MYCN),
2p22.3 gains, and 2p16.2-p14 (REL) gains, del(8p23.1-p21.2),
8p21.2 loss, 8q23.3-q24.3 gains (MYC), and, as expected (Rossi
et al, 2009a; Zenz et al, 2010), losses of the TP53 region
[del(17p13.3-p11.2), del(17p12-p11.2)].
Lesions detected by arrayCGH at 2p, 8q were confirmed by
FISH in 7/7 cases tested: gains affecting 2p and 8q were
validated using locus-specific probes for MYCN in five cases
and for MYC in two cases, respectively. Deletions affecting 8p
were confirmed in 12/12 cases tested: by FISH using a
commercially available probe targeting 8p21.3 in four cases
and by real-time PCR using primers for 8p22 in eight cases. In
one of the patients bearing these lesions, it was possible to
show that gains affecting 2p were of somatic origin, not being
detected in the matched normal DNA analysed by arrayCGH.
Table I. Biological and clinical characteristics of the test CLL series
and of the validation CLL series.
Test
series
Validation
series P-value
Median follow-up 60 months 61 months n.s.
Median age (range), years 71 (43–92) 62 (21–94) <0Æ001
No. (%) No. (%)
Age >60 years 125/147 (85) 112/176 (64) <0Æ001
Gender, male 83/147 (56) 110/176 (62) n.s.
Binet, stage A 107/147 (73) 126/175 (72) n.s.
Mutated IGHV genes 91/147 (62) 111/176 (63) n.s.
Favourable FISH panel
[del(13q14Æ3)/normal/+12]
112/142 (79) 140/169 (82) n.s.
Unfavourable FISH panel
[del(11q22-q23)/del(17p13)]
30/142 (21) 29/169 (17) n.s.
Mutated TP53 16/142 (11) 5/32 (16) n.s.
CD38, >30% 44/147 (30) 69/168 (41) 0Æ04
ZAP70, > 20% 36/129 (28) 67/146 (46) 0Æ02
ns, not significant; FISH, fluorescence in situ hybridization.
A. Rinaldi et al
592 ª 2011 Blackwell Publishing Ltd, British Journal of Haematology, 154, 590–599
Gains at chromosome 2p and 8q are independent
prognostic markers
Due to the relatively low frequency of the lesions on 2p and on
chromosome 8, and because both clinical series were compa-
rable regarding their clinical characteristics, we pooled the two
series together to increase statistical power. The lesions
showing the highest level of statistical significance at univariate
analysis, namely del(8p23.1-p21.2), gains at 2p16.2-p14 and
8q23.3-q24.3 (Fig 1, Table II) were evaluated for their inde-
pendent prognostic significance in a Cox regression model
adjusted for age, gender, Binet stage, IGHV mutational status,
and TP53 inactivation (Table III). Gains occurring at 2p16.2-
p14 and 8q23.3-q24.3 maintained their independent prognos-
tic significance on OS. The two lesions also remained
significant in a model including age, gender, Binet stage,
CD38 + >30%, ZAP70 > 20%, del(8p23.1-p21.2), TP53 inac-
tivation, and the presence, as detected by FISH, of
del(13q14.3), del(11q22-q23), and trisomy 12.
Gains at 2p, gains at 8q and losses at 8p are associated
with different clinical and biological features
Patients with 2p gains presented significantly more often with
unmutated IGHV genes (85% vs. 35%; P < 0Æ001), increased
serum lactate dehydrogenase (LDH; 50% vs. 21%; P = 0Æ030)
and a higher rate of transformation to Richter syndrome (22%
vs. 5%; P = 0Æ021). The presence of 2p gains was associated
with concomitant del(13q14.3) (MIR15A/MIR16-1) (77% vs.
48%; P = 0Æ027), del(11q23) (ATM) loss (21% vs. 6%;
P = 0Æ027), and TP53 inactivation (31% vs. 11%; P = 0Æ032).
None of the cases with gains at 2p had concomitant trisomy 12
(P > 0Æ05).
The presence of 8q23.3-q24.3 gains was significantly more
frequent in patients with del(17p13) (TP53) (44% vs. 5%;
P = 0Æ001), del(8p) (44% vs. 2%; P < 0Æ001), gains of the long
arm of chromosome 3 (44% vs. 3%; P < 0Æ001) and inversely
associated with ZAP70 positivity (0% vs. 38%; P = 0Æ038).
Del(13q14.3), del(11q23) and trisomy 12 were equally distrib-
uted in patients with and without 8q gains.
Deletions at 8p were associated with 8q23.3-q24.3 gains
(MYC) (40% vs. 1%; P < 0Æ001) and del(17p13) (TP53) (80%
vs. 10%; P < 0Æ001). Trisomy 12 and del(13q14.3) (MIR15A/
MIR16-1) occurred at the same frequency in patients with and
without 8p deletions, whilst del(11q23) (ATM) was never
observed in patients with del(8p21).
Gains at 2p, gains at 8q and losses at 8p better define CLL
prognostic groups
The impact of the detection of lesions on chromosomes 2 and
8 was evaluated within the CLL prognostic groups defined by
IGHV mutational status, Binet stage, age and FISH. Table IV
shows the prognostic groups that were better defined by the
demonstration of additional genomic lesions.
Gains at 8q23.3-q24.3 identified patients with a different
outcome amongst those older than 60 years, in Binet stage A,
with unmutated or mutated IGHV genes, or with both
favourable and unfavourable FISH (Fig S3).
Similarly, the presence of 8p deletions was a prognosticator
in CLL with Binet stage A, age older than 60, unmutated or
mutated IGHV genes, and an unfavourable FISH (Fig S4).
The presence of gains affecting 2p identified groups with
poorer outcome amongst patients younger or older than 60,
with Binet stage A, or mutated IGHV genes, or with both
favourable and unfavourable FISH (Fig S5).
Resistance to alkylating agents was observed in 3/3 patients
with gains at 8q23.3-q24.3 (two with concomitant TP53
inactivation), in 2/2 with 8p deletions (one with concomitant
TP53 inactivation) and in 6/8 with gains affecting 2p (two with
concomitant TP53 inactivation). Fludarabine-resistance was
observed in 3/4 patients with gains at 8q23.3-q24.3 (two with
concomitant TP53 inactivation), in 4/6 with 8p deletions (all
with concomitant TP53 inactivation) and in 3/4 with gains
affecting 2p (one with concomitant TP53 inactivation).
A hierarchical model (Fig 2) was built including gains at 2p
or 8q and the genetic subgroups according to Dohner et al
(2000), assigning patients to the worst prognostic group on the
basis of the Cox regression analysis (Table SII).
Table II. MCR with a significant impact on OS in the test CLL series, the validation CLL series and in the pooled CLL population.
Lesion Start Size (bp) %
Test series (n = 147)
Validation series
(n = 176) All (n