Biochimica et Biophysica Acta 1861 (2017) 28992911

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

journal homepage: www.elsevier.com/locate/bbagen

Integrative genomic and network analysis identied novel genes

associated with the development of advanced cervical squamous
cell carcinoma
Anirban Roychowdhury a,1, Sudip Samadder a,1, Pijush Das b, Sapan Mandloi b, Sankar Addya c,
Chandraditya Chakraborty a, Partha Sarathi Basu d, Ranajit Mondal e, Anup Roy f, Saikat Chakrabarti b,
Susanta Roychoudhury g,, Chinmay Kumar Panda a,
a

Department of Oncogene Regulation, Chittaranjan National Cancer Institute, Kolkata, India

Structural Biology and Bioinformatics Division, CSIR-Indian Institute of Chemical Biology, Kolkata, India
Department of Cancer Biology, Sidney Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, PA, USA
d
India Screening Group (SCR), Early Detection and Prevention Section (EDP), International Agency for Research on Cancer (IARC), World Health Organization (WHO), Lyon, France
e
Department of Gynaecology Oncology, Chittaranjan National Cancer Institute, Kolkata, India
f
North Bengal Medical College and Hospital, West Bengal, India
g
Saroj Gupta Cancer Centre & Research Institute, Kolkata, India
b
c

a r t i c l e

i n f o

Article history:
Received 26 March 2016
Received in revised form 17 August 2016
Accepted 12 September 2016
Available online 15 September 2016
Keywords:
Cervical cancer
Copy-number variations
Protein-protein interaction network
PARP1
ATR

a b s t r a c t
Background: CSCC is one of the most common cancer affecting women globally. Though it is caused by the infection of hrHPV but long latency period for malignant outcome in only a subset of hrHPV infected women indicates
involvement of additional alterations, primarily CNVs. Here, we showed how CNVs played a crucial role in development of advanced tumors (stage III/IV) in Indian patients.
Methods: Initially, high-resolution CGH-SNP microarray analysis pointed out frequent CNVs followed by signicantly altered genes. After comparison with TCGA dataset, expressions of the genes were checked in three
CSCC datasets to identify key genes followed by Ingenuity Pathway analysis. Then node effect property analysis
was applied on the constructed PPI network to rank the key proteins. Finally, validations in independent samples
were performed.
Results: For the rst time, frequent chromosomal amplications at 3q13.133q29, 1p36.111p31.1, 1q21.11q44
and 5p15.335p12 followed by common deletions at 11q14.111q25, 2q342q37.3, 4p16.34p12 and 13q13.3
13q14.3 were identied in Indian CSCC patients. Integrative analysis found 78 key genes including several novel
ones, which were mostly associated with Cancer and may regulate DNA repair and metabolic pathways. Analysis
showed PARP1 and ATR were among the top ranking protein interactors.
Conclusions: Frequent amplication and over-expression of ATR and PARP1 were further conrmed in cervical lesions, indicating their association with poor prognosis of advanced CSCC patients.
General signicance: Our novel approach identied precise CNVs along with several novel genes within these loci
and showed that PARP1 and ATR, having biologically signicant interactions, may be involved in development of
advanced CSCC.
2016 Elsevier B.V. All rights reserved.

1. Introduction
Abbreviations: CSCC, cervical squamous cell carcinoma; hrHPV, high-risk human papilloma virus; CNVs, copy number variations; CGH-SNP, comparative genomic
hybridization-single nucleotide polymorphism; TCGA, The Cancer Genome Atlas; PPI,
protein-protein interaction; PARP1, poly-(ADP-ribose) polymerase-1; ATR, ataxia telangiectasia and Rad3-related protein.
Correspondence to: S. Roychoudhury, Saroj Gupta Cancer Centre & Research Institute,
Mahatma Gandhi Road, Thakurpukur, Kolkata 700063, India.
Correspondence to: C.K. Panda, Chittaranjan National Cancer Institute, Department of
Oncogene Regulation 37, S. P. Mukherjee Road, Kolkata 700 026, India.
E-mail addresses: susanta@iicb.res.in, susantarc@gmail.com (S. Roychoudhury),
ckpanda.cnci@gmail.com (C.K. Panda).
1
Authors contributed equally.

http://dx.doi.org/10.1016/j.bbagen.2016.09.014
0304-4165/ 2016 Elsevier B.V. All rights reserved.

Cancer of the uterine cervix (CC) is the fourth most frequent carcinoma among women worldwide. In developing nations like India, both the
incidence (132,000) and mortality (74,000) is high accounting for 1/3rd
of the global burden [1]. This may be due to lack of public awareness and
affordability of vaccines. As a result, Indian women are getting diagnosed at advanced stages of the disease [2].
Among the two types of CC, cervical squamous cell carcinoma
(CSCC) which originates from the ectocervix is more common. It is
known that high risk human papilloma virus (hrHPV) (mostly HPV16

2900

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

and HPV18) infects the basal epithelial cells which in most of the cases
are transient and are taken care by body's immune surveillance. But untreated persistent infection may lead to pre-cancerous lesions called
cervical intraepithelial neoplasia (CIN) and nally to invasive carcinoma
[3]. Interestingly, long latency period of around 1520 years is reported
between HPV infection and development of invasive lesion. This indicates that other than persistent hrHPV infection, additional chromosomal alterations are also required to bring about carcinogenesis [4].
Since the rst report of changes in chromosomal content of CC, [5,6]
considerable efforts have been made towards cataloguing the genetic
alterations of the tumor. Several meta-analyses revealed that chromosomal gain of 3q (3q2429), 1q (1q22q23, 1q25.3q32.1) and 5p
(5p12p13) are recurrent, while loss of 3p (3p1223), 11q
(11q22.325) and 4p (4p16.3p16.1) are frequent in CC. It is also observed that there is an increase in rate of copy number (CN) alterations
(gain/loss) at these distinct chromosomal loci from CIN to invasive CSCC
[7,8]. Such observations suggest that the genes located within these frequently altered chromosomal loci may play a crucial role in development of more aggressive form of the tumor, if left untreated. It is also
indicative that the cancer cells containing these chromosomal aberrations are primarily selected to sustain the changing tumor microenvironment. This is supported by the fact that there is a signicant
decrease in 5-year survival of CC patients with increase in aggressiveness of the tumor [8]. Generally, around 80% patients survive without
disease in case of early stage (I/II) tumor while, approximately only
25% survive if diagnosed at later stages (III/IV). Thus, the failure of conventional treatment in advanced stage (III/IV) CC patients may be associated with the deregulation of the genes due to alterations in their
chromosomal loci. So, the present investigation was carried out to analyze such plausible role of chromosomal copy number variations (CNVs)
in development of advanced (stage III/IV) CSCC. Further, in silico functional analysis and validation in independent patient population were
also performed to analyze how such aberrations may contribute to
poor prognosis in advanced CC patients.
In the present study, at rst, genome-wide frequent CNVs were
identied in advanced stage (III/IV) CSCC using high-resolution CGHSNP analysis. The alteration (amplied/deleted) of genes located within
frequent CNVs was compared with The Cancer Genome Atlas (TCGA)
CN dataset for CSCC to identify genetically altered genes. It was followed
by integration of publicly available expression datasets to nd out differentially expressed genes (DEGs) in CC that were frequently
deregulated due to CNVs. Further, Ingenuity Pathway analysis (IPA)
followed by protein-protein interaction (PPI) network analysis revealed
that these identied genes may modulate key cellular pathways and
crosstalk with other important proteins as well. Finally, alterations
and expression of key genes were validated and found to play a crucial
role in development of the aggressive tumor having poor prognosis.

2. Materials and methods

2.1. Tissue specimen
Primary uterine cervical lesions (n = 55) and corresponding peripheral blood lymphocytes (PBL) were collected from the hospital section
of Chittaranjan National Cancer Institute, Kolkata, India after appropriate approval of the Institutional Ethical Committee and written informed consent from respective patients were taken. According to
FIGO classication, primary tumors that belonged to early clinical
stage I/II (n = 23) and advanced clinical stage III/IV (n = 32) were
taken [9]. Normal ectocervical tissue (n = 3) was obtained from patients who underwent hysterectomy for benign reasons. Freshly collected tissues and corresponding blood were stored at 80 C for DNA
extraction. Some portion of the tumor tissues were collected in
RNAlater RNA Stabilization solution (Ambion) for RNA isolation.
SiHa (HPV-16 positive CSCC cell line) were procured from the National

Centre for Cell Sciences (NCCS), Pune, India. Demography of patients

with survival status upto 36 months was provided (Table S1).
2.2. DNA isolation
Based on microscopic examination of Hematoxylin-Eosin
(HE) stained cryosections (5 m), primary tumor tissues and
normal ectocervix were conrmed by experienced pathologist(s)
as squamous cell carcinoma and absolute normal respectively. If
contaminant normal cells were reported in any tumor tissue(s), mainly
stage I/II, then microdissection was done for further enrichment. Genomic DNA from cryosections (5 m) of tissue specimen and PBL were isolated by standard phenol/chloroform method [10].
2.3. RNA isolation
Total RNA from samples kept in RNAlater was isolated by TRIzol
Reagent according to the manufacturer's protocol (Invitrogen). Reverse
transcription was done with 1 g total RNA using Random hexamer
(Invitrogen) and M-MuLV-Reverse Transcriptase (Promega, USA).
2.4. Typing of hrHPV
Detection of HPV in the cervical lesions was detected by Polymerase
chain reaction (PCR) using primers (MY09 and MY11) designed from
the consensus L1 region. It was followed by typing for high risk HPV
16/18 in the L1 positive samples [11].
2.5. CGH-SNP microarray experiment
High-resolution whole-genome analysis of CNVs (n = 11) was done
using SurePrint G3 CGH + SNP Array 2x400K platform (Agilent Technologies, Santa Clara, CA, USA) according to manufacturer's protocol.
Briey, about 1000 ng of control and test DNA was used for the restriction digestion in the master mix containing AluI and RsaI restriction enzymes. The samples were incubated at 37 C for 2 h followed by heat
inactivation of enzymes at 65 C for 20 min. Labelling of samples was
done by random priming method, in which the random hexamers,
Cy3-dUTP & Cy5-dUTP, dNTP, Buffer and Klenow enzyme was used.
Master mix for Cy3 and Cy5 dNTPs was done separately to ensure
that, the control sample is labelled with Cy3 and test sample with Cy5
respectively. About 19 l of labelling master mix prepared as per manufacturers recommendation was added to the denatured control and test
DNA sample and incubated at 37 C for 2 h followed by enzyme heat inactivation at 65 C for 10 min. Equal amount of labelled Test & Control
DNA sample was added into a fresh tube containing 25 l of Human
Cot-1 DNA (1 mg/ml), 26 l of Agilent 10 blocking agent and 130 l
Agilent 2 hybridization buffer. The above hybridization mix was denatured at 95 C for 3 min and incubated the microfuge tubes at 37 C for
30 min. Following hybridization at 65 C for 40 h in the hybridization
chamber, the slides were washed using aCGH Wash Buffer1 (Agilent
Technologies, Part Number 5188-5221) at room temperature for
5 min and aCGH Wash Buffer 2 (Agilent Technologies, Part Number
5188-5222) at 37 C for 1 min. The slides were then washed with
Acetonitrile for 10 s.
2.6. CGH-SNP microarray data analysis
The microarray slide was scanned using Agilent Scanner (Agilent
Technologies, Part Number G2565CA). Image analysis was performed
using Agilent Feature Extraction software. Feature extracted raw data
was normalized using lowest normalization and analyzed by Agilent
CytoGenomics 2.7 Software. According to manufacturer's recommendation, to genotype our own (Indian) reference, a normal female DNA
(Sample ID: 854) was hybridized with Coriell European female
(NA12878). Other samples were then analyzed using the deduced

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

genotype as reference. Aberration Detection Method II algorithm

(ADM2) was applied to identify signicant regions having amplications and deletions among each of the samples. GC Correction algorithm
was then applied to correct aCGH log ratio data for the presence of
wavy artefacts. The centralization normalization algorithm was used
as a parameter for detecting aberrant regions or regions of constant
copy number using ADM2. The Diploid Peak Centralization algorithm
centres the data so that the log ratios of the probes that are in copy number 2 regions are centred on zero.
For SNP analysis, Allele-specic copy number (ASCN) detection algorithm was applied to distinguish the two alleles of a SNP by whether or
not the SNP site is cleaved by the AluI/RsaI restriction enzyme mixture
that is used during the sample labelling process. ASCN values are used
by the LOH (Loss of heterozygosity) detection algorithm to deduce the
genotypes of SNP sites. The algorithm reports LOH for the regions that
are also detected as deletions in standard CGH analysis. The data was
submitted to the NCBI Gene Expression Omnibus (GEO) repository
and assigned GEO accession number GSE76911.
2.7. Statistical analysis
ADM-2 algorithm uses probe quality information to weight the log
ratios before calculating the score for a particular genomic interval
(CGH Interactive Analysis User Guide). So, along with log-ratios, it also
uses log-ratio error information; hence it generates Quality-Weighted
Interval Score. The following describes how the statistical score was
generated:
The algorithm takes the following vector of pairs as inputs:
(L1, LE1) , (L2, LE2) , . , (Ln, LEn) where Li was the log-ratio signal for the
ith probe and LEi was the log-ratio error for the ith probe ordered.
The dLRsd can be dened as a robust estimate of noise in the presence of highly aberrant samples. So, if dLRsd N LEi then LEi was set to
the dLRsd value. Now, qi was dened as
qi

1
LEi 2

Generally, it was assumed that under the null model, Li ~ N (0, 1/qi )
and the different Li are independent of each other, where N is the number of probes in the genomic interval.
The weighted sum for a genomic interval (I) was considered as:
X

qi Li

iI

Variance of S(I) was computed:

X
 X
 X
varSI var
qi Li
qi
q2i varLi

Therefore, the ADM-2 score for a genomic interval:

X

q i Li
SI q
X
qi

This score reects the deviation of the weighted sum from its expected value (of zero) in units of standard deviation. It searches for intervals in which a statistical score based on the average quality
weighted log-ratio of the sample and reference channels exceeds a
user specied threshold (t). It uses a recursive process which by default
stops when no interval with score exceeding the user-specied threshold is found. Here, the threshold (t) of 6.0 was used. Putative amplied
or deleted genomic intervals containing b 3 probes and/or having minimum average absolute log-ratio of less than 0.25 was excluded. Positive
statistical score meant amplication (Amp) while negative score indicated deletion (Del). It also reports the p-value corresponding to each

2901

score for a genomic interval. The p-value was calculated using the normal probability distribution function and the score of that interval.
2.8. Identication of signicantly altered genomic regions
Using ADM-2 generated statistically signicant interval based amplication and deletion data, penetrance analysis was performed to nd
the percentage of samples that share aberrations in a particular genomic
region among multiple samples (amplication and deletions are considered separately). Common aberrations among the samples were identied (Table S2). A genomic region has been identied as frequently
amplied if it has shown amplication (percent penetrance) in 50%
of samples (6 out of 11 samples) and for loss it was taken as 35% (4
out of 11 samples).
2.9. Identication of signicantly altered genes
Several in-house algorithms were then developed to generate the
amplication and deletion score of the genes (with corresponding pvalue for each sample) located within these frequently altered genomic
regions using the same interval based statistical scores, generated by
ADM-2 algorithm (mentioned above). Adjusted p-values were calculated using Benjamini & Hochberg method (False discovery rate). Genes
located within these frequently altered chromosomal loci that showed
signicant (p-value 0.01) statistical score (Amp/Del score) were compared with TCGA (146 normal and 105 tumor samples) and Scotto
(7 normal and 79 tumor samples) CN datasets for CSCC. The frequently
altered genes were considered as those that showed similar (Amp/Del)
alteration in our CGH-SNP data as well as in the other two CN datasets.
The two datasets were obtained using Oncomine Research Premium
Edition. It is an online application that integrates and unies highthroughput cancer proling data across a large number of cancer
types, so that the status of user-dened genelist in a particular tumor
can be analyzed [12]. It provides Fold-change (positive/negative)
along with p-value (based on t-statistics) for each gene when compared
in normal vs cervical squamous cell carcinoma (DNA) scenario. Positive
fold-change meant amplication whereas, negative fold-change indicated deletion.
2.10. Analysis of expression proles
The RNA expression of the frequently altered genes were also obtained using Oncomine database [12]. Three datasets were used
(Fig. S1): Biewenga (5 normal and 40 tumor samples), Scotto (24 normal and 32 tumor samples) and Zhai (10 normal and 21 tumor samples). Oncomine output contained Fold-change along with p-value
(based on t-statistics) for each gene when compared in normal vs cervical squamous cell carcinoma (RNA) scenario. Positive fold-change
meant over-expression whereas, negative fold-change indicated
under-expression. Genes that showed signicant (p-value 0.01) fold
change 1.5 and also showed similar (over/under) expression in all
three datasets were taken for further analysis.
2.11. Pathway analysis
The identied deregulated genes were analyzed using QIAGEN's
Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City, www.
qiagen.com/ingenuity). The Canonical Pathway analysis and related
Diseases and BioFunctions analysis were performed by the use of
QIAGEN's Ingenuity Pathway Analysis. To nd out a specic function
or pathway to be signicant in relation to the genelist, Ingenuity recommends uncorrected p-value calculation using right-tailed Fisher's
Exact Test. It measures how likely the observed result (the number of
mapped genes) would be if the association was just random. If this is
very unlikely (i.e. the p-value is below the threshold) then the function
is said to be statistically signicant. Thus, the associated p-value is a

2902

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

measure of the likelihood that the association between a set of genes

and a given biological process or pathway is due to random chance. A
p-value of 0.05 (threshold) indicates that the association is signicant
and not random.
2.12. Construction of the human Cervical Cancer Interaction Network
(CCIN)

PPIN of STRING Ensemble Protein IDs (ENSP), which were further converted to HGNC gene symbol for better understanding. Second level interactions (interactor of interactor) of 48 deregulated genes (out of 78
deregulated genes/proteins from CNV and transcriptomic analysis of
cervical cancer patient samples) consisting of 1903 nodes and 4431 interactions form the CCIN.
2.13. Assigning weights to proteins in the CCIN

Human PPI information was downloaded from STRING v9.1 [13] database. STRING contains interaction data based on various data collection principles. However, we have extracted interactions, which are
experimentally veried with high condence score (score 700). Cervical Cancer Interaction Network (CCIN) was created by pooling up the

To identify proteins with important interactions in CCIN, we calculated node (protein) effect property. Node effect is a weighted network
property based on functional knowledge of pathways, expression and
regulatory features such as rate limiting enzymes or cross-talk genes/

Fig. 1. Circos plot representing chromosomal alteration in three (normal vs tumor) CSCC patients. Here, human chromosomes staring from chromosome 1 to chromosome Y are
represented with different colours and are marked in Mbs. Amplication (Log-ratio 0.25) and deletion (Log-ratio 0.25) along with unchanged (Log-ratio = 0.25 to 0.25)
chromosomal regions were shown.

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

Table 1
List of frequently amplied (50%) and deleted (35%) chromosomal loci as evident from
CGH-SNP microarray analysis.
Chr
name

Start (hg38)

Stop (hg38)

Aberration size
(in kb)

Cytogenetic
location

Frequent (50%) gain of chromosomal loci

chr1
26,537,121
80,039,972
chr1
143,960,023
248,903,563
chr3
109,308,013
198,118,383
chr5
26,143
46,279,633
chr8
144,868,300
145,054,180
chr16
70,818,537
71,162,574
chr19
37,887,886
39,204,515
chr19
39,308,261
39,604,381
chrX
13,303
58,517,390
chrX
62,779,286
68,096,929

53,502
104,943
88,810
46,253
185
344
1316
296
58,504
5317

1p36.111p31.1
1q21.11q44
3q13.133q29
5p15.335p12
8q24.3
16q22.2
19q13.1319q13.2
19q13.2
Xp22.33Xp11.21
Xq11.2Xq12

Frequent (35%) loss of chromosomal loci

chr2
208,862,492
242,005,889
chr4
123,924
45,692,717
chr4
86,514,543
86,598,624
chr4
86,615,109
86,745,018
chr8
226,814
893,442
chr8
3,756,723
3,907,556
chr8
39,379,919
39,505,368
chr11
84,319,407
135,034,169
chr13
34,898,407
51,117,548
chr19
281,067
2,515,068
chr19
5,222,171
5,229,153

33,143
45,568
84
129
666
150
125
50,714
16,219
2234
6

2q342q37.3
4p16.34p12
4q21.3
4q21.3
8p23.3
8p23.2
8p11.22
11q14.111q25
13q13.313q14.3
19p13.3
19p13.3

proteins involved in the interaction network. This method considers

heterogeneous datasets into the network of interactions like gene ontology information and topological property (degree) of network for

2903

biological interpretation. To understand complex biological networks

and to identify proteins with important interacting partner's node effect
(effs) property is utilized. Here, we have provided weights to node/proteins in the network based on the differential expression at genomic
level (CNV; Gexp), differential expression at transciptomic level (Texp),
rate limiting enzyme (RLE) and signalling cross-talk proteins (SC), respectively. Effect of neighbours (up to second level) on nodes (s, protein) in CCIN is calculated by
 n
k
eff s ;
j1

i1

wj
wi
; ;
degreei
nj

where, k is the degree of s, nj is the degree of protein j, wj and wi are the

node weights of protein j and i.
2.14. Validation of genomic copy number
Genomic copy number of ATR and PARP1 were validated in normal/
tumor paired samples (n = 13) and SiHa cell line using quantitative PCR
(qPCR) in Roche LightCycler 96 Real-Time PCR System [14]. Primer sequences for ATR: Forward (Fwd): 5-GATCTTGACATCTTATCCAGG-3;
Reverse (Rev): 5-TGAAATCAAGCAACATCACG-3 and for PARP1 were:
Fwd: 5-AGCTAGGATCTGACAGCCTG-3; Rev: 5-GGCACTCTTGGAGACC
ATGTCA-3; the Beta-2-Microglobulin (B2M) gene was used as reference as no alterations of this gene were seen in our CGH-SNP analysis
of CSCC patients. Primer sequences for reference gene B2M were:
Fwd: 5-CACTTGGTGCCTGATATAG-3; Rev: 5-TCAATGTCGGATGGAT
GAAA-3. It is assumed that for a gene a normal sample has two copies.

Fig. 2. The CGH-SNP data of the four most frequent chromosomal alterations. The altered chromosomal segment comprises of several precise coordinates showing different frequency of
aberration. They were represented by different colours and the adjacent pie chart showed the number of samples showing alteration of that precise region. The absolute copy numbers
(CNs) of the informative SNPs were provided for each sample to show concordance of CGH and SNP data. If the CN value ranges from 0 to 1 then the SNP was deleted. While if the CN
value was N2 then the SNP was amplied and 2 signied no change in CN. Amplication of chromosomal region 1p36.111p31.1 and 1q21.11q44 along with amplication of
3q13.133q29 and 5p15.335p12 was shown. Loss of genomic region 11q14.111q25 was seen.

2904

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

But as a primary tumor tissue may still be contaminated by normal cells

(after micro- dissection as well), a copy-number range approach was
used and following was considered: borderline LOH-normal N1.41.6;
normal N 1.62.4; borderline amplication N2.42.6; and amplication
N2.6. For CN calculation in SiHa cell line, average Ct values of reference
gene (B2M) and target genes (ATR and PARP1) for normal samples
were taken.
Further analysis of amplication of ATR and PARP1 was carried out
by differential polymerase chain reaction (DPCR) method using the
same primer sets in independent normal/tumor paired samples [15].
The primers of both ATR and B2M (reference gene) were co-amplied
by multiplex PCR with 50 ng of DNA. In case of PARP1, separate PCR reactions were performed with equal amount of DNA (50 ng) and equal
volume of PCR products were mixed. The products were then visualized
after electrophoresis in 2.5% agarose gel followed by ethidium bromide
staining. The band intensities were measured by Image J software
version-3.0 (Bio-Rad). Amplication status of target gene was calculated as follows: Relative copy number = [Ttarget gene / Treference gene] /
[Ntarget gene / Nreference gene]; where Ttarget gene and Treference gene are the intensities of the bands in tumor tissues and Ntarget gene and Nreference gene
are the intensities of the bands in the corresponding normal PBL DNA.
Target gene was considered as amplied when the ratio was 2.
Deletion analysis of SLIT2 gene in tumor/normal paired samples
(n = 11) was carried out by an intragenic non-polymorphic microsatellite marker, D4S1372 [chr4:20,299,82620,299,990] using multiplex
PCR. D4S1372: Fwd 5-GATTAACTTTACCTTGAATGATGATAT-3 and Rev
5-GCTCCGTAACAGATACTCTTTGC-3. It was performed using [-P32]
end labelled forward primer of the marker and the reference gene

(B2M). PCR products were run in 7% denaturing polyacrylamide sequencing gel containing 8 M urea followed by autoradiography [4].
2.15. Analysis of mRNA expression
Expression of ATR and PARP1 were analyzed in normal ectocervical
tissue (n = 3), CSCC samples (n = 14) and SiHa cell line. The real
time quantitation of RNA expression was done by Power SYBR-green
PCR assay (Applied Biosystems, USA) [4]. Following primer sets were
used: PARP1: Fwd 5-AAGGCGAATGCCAGCATTAC-3; Rev: 5-GGCACT
CTTGGAGACCATGTCA-3; ATR: Fwd 5-CAGTGCCACACCAGAGGAAT-3;
Rev: 5-TGAAATCAAGCAACATCACG-3. B2M was used as the internal
control. Primers for B2M: Fwd 5-GTGCTCGCGCTACTCTCTCT-3; Rev
5-TCAATGTCGGATGGATGAAA-3. The relative level of gene expression
was determined by comparative threshold cycle (Ct) method based
on the average Ct value of the genes in three normal tissues. A
log2^( ddCt) value of 0.1769 (fold change 1.5) was considered as
over-expression.
2.16. Protein expression by immunocytochemical analysis
Cervical squamous cell carcinoma (HPV-16 positive) cell line, SiHa,
were cultured on coverslips until sub-conuency followed by methanol
xation. Further blocking was done with 3% BSA and incubated with the
respective primary antibodies for ATR (sc-1887) and PARP1 (sc-1562)
in dilution of 1:100. After that uorescein isothiocyanate (FITC)conjugated rabbit anti-goat secondary antibody (sc-2777) was used
in 1:500 dilution, then washed and briey incubated with DAPI

Fig. 3. Heatmaps showing number of genes that are frequently altered (amplied = 1224 and deleted = 583) both in A. TCGA dataset (normal samples = 146; tumor samples = 105) and
B. Scotto dataset (normal samples = 7; tumor samples = 79).

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

(sc-3598) and subsequently mounted on slides, viewed and

photographed by uorescence microscope (Leica DM 4000B; Germany).

2905

3. Results

amplications at distinct chromosomal loci were catalogued as follows

(Table 1): 1p36.111p31.1, 1q21.11q44, 3q13.133q29, 5p15.33
5p12, 8q24.3, 16q22.2, 19q13.1319q13.2, Xp22.33Xp11.21 and
Xq11.2Xq12. While common (N35%) loss at chromosomal cytobands:
2q342q37.3, 4p16.34p12, 4q21.3, 8p23.3, 8p23.2, 8p11.22, 11q14.1
11q25, 13q13.313q14.3, and 19p13.3 were also observed (Figs. 2
and S4). Amplication of chromosome 3 at precise chromosomal
coordinates (GRCh38/hg38) of 3q25.23q26.1 (chr3:154,427,429
162,796,745), 3q26.13q26.31 (chr3:162,901,354175,146,595) and
3q26.323q29 (chr3:175,208,719198,094,926) were found in all
patients. Chromosomal loci 11q24.311q25 (chr11:128,926,744
135,034,169) showed highest (72%) loss (Table S3).

3.1. Identication of frequent CNVs in CSCC patients

3.2. Identication of genes having signicant CNVs in CSCC patients

To identify genes associated with the development of advanced

(stage III/IV) CSCC, a unique workow have been used (Fig. S2). At
rst, high resolution CGH-SNP analysis of advanced CSCC revealed
CNVs at different chromosomal arms (Figs. 1 and S3). Then, the percentage of samples that share common aberrations in a particular chromosomal locus were analyzed (Table S2). Finally, frequent (N 50%)

Our further analysis identied the genes (n = 2946) within frequent

CNVs which were signicantly (p-value 0.01) altered (Table S4). It
consists of both protein-coding genes (n = 2874) and miRNA genes
(n = 72). Among the signicantly altered protein-coding genes, 1944
protein-coding genes were from common amplied chromosomal
intervals while 930 protein-coding genes reside within the frequently

2.17. Overall survival (OS) analysis

Overall survival (OS) was measured from the date of surgery to the
date of most recent follow-up or death (up to 36 months). Survival
curves based on 36 months follow-up data (Table S1) were obtained
using KaplanMeier method. The analysis was done using Epi Info
(Centres for Disease Control and Prevention).

Fig. 4. Representative heatmaps showing signicantly (p-value 0.01) altered genelist (n = 78) which are over-expressed (due to amplication) and under-expressed (and deleted) in all
three expression datasets A. Biewenga (normal samples = 5; tumor samples = 40) B. Scotto (normal samples = 24; tumor samples = 32) C. Zhai (normal samples = 10; tumor
samples = 21).

2906

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

deleted chromosomal regions. For further validation, we compared alterations (amplication/deletion score) of the identied proteincoding genes (n = 2874) with that of the fold-change (positive/
negative) of TCGA and Scotto CN datasets obtained from Oncomine
database. This analysis showed that 1807 protein-coding genes (out of
2874) were signicantly (p-value 0.01) altered in both of these two
datasets and showed concordance with CGH-SNP data as well
(Table S5). Thus, these amplied (n = 1224) and deleted genes (n =
583) were found to be signicantly altered in advanced CSCC (Fig. 3).

3.3. Correlation between CNVs and their expression to identify key genes
A gene associated with the development of advanced CSCC must be
deregulated at the transcriptomic level due to its genetic alteration.
Gene that is amplied may be over-expressed and a deleted gene may
be under-expressed. Hence, fold-change (FC) in expression of these
genes (n = 1807) were correlated with their alterations. Among the altered genes, similar differential expressions of only 78 DEGs (p-value
0.01) have been seen in all three CSCC expression datasets (available
in Oncomine database). Under-expressed (FC 1.5) 11 proteincoding genes belong to deleted regions while rest of the 67 amplied
(FC 1.5) protein-coding genes were over-expressed (Table S6 and
Fig. 4). It was evident from the present analysis that the differential expression of these DEGs (n = 78) in CSCC were primarily due to their
CNVs and thus, these identied key genes may deregulate crucial cellular pathways.

3.4. Identied genes may modulate important biological pathways

To realize the role of the identied 78 DEGs in development of CSCC,
QIAGEN's Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City,
www.qiagen.com/ingenuity) tool were used. The identied candidate
genes (30 mapped out of 78) were shown to be associated with 103 signicant canonical pathways. It showed that DNA damage and repair
pathways were mostly enriched. Interestingly, several metabolic pathways showed signicant enrichment as well (Table S7 and Fig. 5A). Disease and BioFunctions association clearly showed that 67 (out of 78)
genes were related to Cancer. Identied genes mostly shown to modulate cell cycle to promote cellular growth and survival (Table S8 and
Fig.5B). It appears that these cancers associated genes may interact to
regulate both the signalling and metabolic pathways.

3.5. Functional in silico analysis by ranking proteins in CCIN

Further attempt has been made to understand how the identied
genes/proteins functionally involved in pathogenesis of CSCC using
human protein-protein interaction (PPI) network (interactome). We
performed weighted network analysis to understand the role of 48
(out of 78) proteins that were found to be deregulated both at the genomic (CNV) and transcriptomic level. We constructed a cervical cancer
related proteins' interaction sub-network (CCIN) (see Materials and
methods for more detail). Number of interactions or degree of a protein
in the interaction network may not always reect the importance of the

Fig. 5. Ingenuity Pathway analysis (IPA) revealed signicant A. Canonical pathways and B. Disease and Biofunctions associated with the identied altered genes. The p-value (Log) was
plotted as blue bars. The ratio (represented as orange squares) is the number of genes from our identied list that belong to the pathway divided by the total number of genes in the same
pathway based on Ingenuity's Knowledge Base. The threshold line means a p-value of 0.05.

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

interaction module. Hence, we determined the degree of each protein

(node) in the CCIN and calculated the node effect (effs; see Materials
and methods for more details) property based on biological information
like expression status and involvement in signalling/metabolic pathways (Table S9). Correlation between degree of a protein with its
node effect (effs) property was found to be low (R2 = 0.248) indicating
that the node effect property is not directly related to number of interactions of a protein. Higher node effect score for a protein indicates existence of other important protein-protein interactors within its local
network. So, node effect score can be used to rank the proteins based

2907

on the importance of local interaction module. Identied genes involved

in CCIN (48 out of 78) showed differential interaction modules
(Table S10). Weighted network of CCIN (Fig. 6) was created and visualized using Cytoscape program [16].
Our network analysis showed that high node effect score proteins
were indeed related to cancer development. Gene set analysis using
GeneAnalytics [17,18] of 48 protein-coding genes showed that
among top 10 diseases 18 (38%), 8 (17%) and 6 (13%) genes/proteins
were related to Cervical Squamous Cell Carcinoma, Lung cancer and
Breast cancer respectively. Around 10 (20%), 4 (8%) and 4 (8%) genes/

Fig. 6. Cervical cancer related protein's interaction sub-network (CCIN) represented 1903 nodes based on weighted network analysis. The size of the nodes/genes increased with the score
(according to node effect property). Triangle shaped nodes (outer most circle) were the top 100 based on the scores. The represented colour (red: up-regulation to green: downregulation) of the nodes was based on the copy number (n = 2) and/or expression (n = 3) values available for all the genes in CSCC datasets used in the present study. The circle next
to it showed the identied 78 genes (deregulated at both genomic and transcriptomic level) and among them PARP1 and ATR (triangle shaped) were among the top 100 gene rank.
Next two circles are for genes deregulated at transcriptomic and CNV (octagon shaped) respectively. The innermost circle represented the genes which had no copy number variation
and/or expression values available in CSCC datasets used in the current investigation.

2908

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

proteins were involved in cell cycle, DNA repair and cancer pathways/
processes respectively (Fig. S5). These observations suggest that high
scored deregulated genes/proteins have functional association with
cancer. Interestingly, PARP1 and ATR, which were found to be highly
deregulated at genomic/CNV and transcriptomic level in cervical squamous cell carcinoma scenario were also found to possess the top node
effect scores (top 1%) via our weighted network analysis.
3.6. Validation of genomic alteration and expression
Network based analysis revealed that PARP1 (1q42.12) and ATR
(3q23) were the two (out of the identied 78 genes) potential candidates that may play an essential role in development of advanced
CSCC by virtue of their important interactions. So, their genomic copynumber (CN) and expression status in CSCC samples (n = 55, including
previous 11 samples) along with hrHPV-16 positive CSCC cell line, SiHa,
were checked.
Our DNA microarray showed amplication (Table S4) in both ATR
(avg. amp score = 0.74) and PARP1 (avg. amp score = 0.42). Likewise,
overall amplication of 60% (33 out of 55 CSCC samples) was observed
in ATR (11 out of 23 in Stage I/II and 22 out of 32 in Stage III/IV samples)
while, in case of PARP1 (8 out of 23 in Stage I/II and 15 out of 32 Stage
III/IV samples) it was 41.82% (23 out of 55 CSCC samples). Moreover,
over-expression of ATR [mean log2^( ddCt) = 1.032] and PARP1
[mean log2^( ddCt) = 0.894] was also observed similar to their expression status in three CSCC datasets (Table S1 and Fig. 7). Moreover,
analysis in 4 CSCC samples and SiHa cell line (where both CN and expression were done) showed signicant positive correlation between
high copy-number and over-expression in ATR (R2 = 0.91, p-value =
0.013) and PARP1 (R2 = 0.73, p-value = 0.067). Interestingly, a gradual
increase in deregulation (CNV/expression) from early to advanced CSCC
was observed in both ATR and PARP1 (Fig. 7). High nuclear expression
of both ATR and PARP1 was concordant with increased copy-number
and over-expression in SiHa (Fig. 7). Moreover, it was evident from
overall survival (OS) amplication of ATR or PARP1 may be associated
with poor prognosis of CSCC patients (Fig. S6). Additionally, deletion
of SLIT2 (avg. del score = 0.54) was also concordant with the DNA
microarray data (Fig. S7).
4. Discussion
The primary aim of the present investigation was to catalogue the
chromosomal alterations associated with advanced CSCC (clinical
stage III/IV) followed by in depth analysis to understand the role of
these aberrations in development and poor prognosis of this cancer.
Chromosomal alterations primarily CNVs are dened as gains or
losses of genomic DNA that are larger than 1 kilobase (kb) in size and
often not visible by standard G-banding karyotyping [19]. For
genome-wide identication of these precise CNVs in advanced CSCC,
CGH-SNP microarray was performed. Detection of CNVs simultaneously
by oligonucleotide (average 60mer; average spacing 7 kb) based CGH
and SNP array enhanced our precision [20,21]. To the best of our knowledge this is for the rst time such high-resolution mapping of chromosomal aberrations have been performed among Indian CSCC patients.

2909

The altered chromosomal regions ranged from 6 kb (19p13.3) upto

105 Mb (1q21.11q44). Our present study showed that gain of chromosomal 3q (3q25.23q26.1; 3q26.13q26.31; 3q26.323q29) was the
most common (100%) chromosomal aberration found in Indian CSCC
cases. Therefore, in Indian patients 3q gain may be used as a potential
biomarker for CC progression [2225]. Additionally, about 90% of the
CSCC patients showed simultaneous gain of precise regions of 3q and
1q in our study. Moreover, loss of chromosomal 11q (11q14.111q25)
was reported to be the most frequent among Indian CC patients. The
identied frequently altered chromosomal regions were also reported
in several studies done previously [7,8].
Next, we identied signicantly (p-value 0.01) altered genes (n =
2874) residing within these chromosomal coordinates in order to realize the importance of these CNVs. To extrapolate our ndings in large
number of CSCC patients, comparison with TCGA and Scotto CN datasets
(obtained from Oncomine database) were done. After the critical comparison, around 63% (1807 out of 2874 genes) showed similar alteration
pattern.
To play a role in tumor development, a gene residing in the amplied
regions often may be over-expressed while deleted genes may get
under-expressed [7,26]. To illustrate such events, the expression status
of these 1224 amplied and 583 deleted genes were screened through
three CSCC expression datasets (available in Oncomine database). As
a result, 67 amplied genes were found to be over-expressed while 11
deleted genes were under-expressed. Therefore, it is evident that
these DEGs were deregulated due to their CNV.
Among the 78 identied DEGs, some genes were novel and may play
a role in CSCC progression. To name a few of them like, YEATS2component of the complex ATAC having histone acetyltransferase activity, NUP155-essential component of nuclear pore complex, GTPBP8GTP-binding protein containing EngB-type G domain, and C3orf52 upregulated by 12-O-tetradecanoylphorbol-13-acetate (TPA). Whereas,
association of the identied genes namely: CDCA8 (1p34.3), DTL
(1q32.3), B4GALT5 (3q13.32), MCM2 (3q21.3), ATP2C1 (3q22.1),
PLOD2 (3q24), TNFSF10 (3q26.31), ECT2 (3q26.31), LAMP3 (3q27.1),
ALG3 (3q27.1), RFC4 (3q27.3), RPL39L (3q27.3), ATP13A3 (3q29),
TFRC (3q29), MUC4 (3q29) and TRIO (5p15.2) were also reported in
previous genome-wide studies in CC patients [4,2738]. On the other
hand, loss of SLIT2 (4p15.31) was reported previously by our research
group and found be associated with the development of CSCC in
Indian patients [4,35]. Some of the identied candidate genes like:
PARP1 (1q42.12), EXO1 (1q43), ATR (3q23), ACTL6A (3q26.33), PAK2
(3q29), ACAT1 (11q22.3), CRYAB (11q23.1) were reported to be associated with other cancers but here we report their association with CSCC
for the rst time [3945]. It appears that the frequent alteration of these
key genes could lead to the deregulation of cellular pathways that may
help the tumor cells to sustain the hostile microenvironment.
To look for the deregulated pathways, QIAGEN's Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City, www.qiagen.com/
ingenuity) was further employed in this study. Most of the genes
seem to regulate Cell cycle and DNA damage and repair pathways
followed by several metabolic pathways like Pyrimidine Ribonucleotides synthesis, Ketolysis and Ketogenesis. Interestingly, this nding showed that signalling and metabolic pathways may crosstalk

Fig. 7. Representative gures showing validation of alteration and expression of PARP1 and ATR in CSCC samples. A. Graphical representation of the copy-number (CN) of PARP1 and ATR
(determined by qPCR) in 13 CSCC samples and SiHa cell line. Below the amplication (Amp) status of PARP1 and ATR was provided, as interpreted from the CN values (N2.6). Amplication
is denoted by + and means no change in copy number. B. Representative agarose gel image showing amplication [(15,550,075 / 4,642,599) / (6,686,154 / 4,602,478) = 2.305] of PARP1
in Sample ID 3380 based on relative band intensity calculation. While, there was no change in relative copy-number [(9,966,949 / 5,982,192) / (9,380,108 / 5,334,452) = 0.947] was
observed in Sample ID 4495. B2M was used as the control gene. C. Amplication [(12,165,898 / 5,654,662) / (3,806,205 / 5,601,860) = 3.167] of ATR was shown in the same sample
through representative agarose gel. In Sample ID 4495, ATR showed no change [(5,866,781 / 3,065,916) / (9,273,509 / 4,647,091) = 0.97] in relative copy-number. B2M was used as
the reference gene. (M = pUC19/MspI marker; NTC = non-template control). D. Relative mRNA expression [log2^(ddCt)] of PARP1 and ATR in 14 CSCC samples and SiHa cell line
was graphically represented. Below the amplication (Amp) status of PARP1 and ATR was given. Amplication is denoted by + and means no change in copy number. E. Overall
percentage of amplication of both the genes was seen with stage-wise progression of CSCC. F. Mean value of log2^(ddCt) of stage I/II (n = 6) and stage III/IV (n = 8) was plotted
showing signicant rise in expression of PARP1 (p-value = 0.002344) and ATR (p-value = 0.02692) from early to late stage. G. Immunocytochemical analysis of PARP1 and ATR in
SiHa cells showed up- regulation of nuclear expression.

2910

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

with each other and play a crucial role in development of advanced

CSCC. Recent reports also shed light on this hallmark of carcinomas
and justied our ndings [46,47]. Moreover, most (67 out of 78) of the
altered genes were associated with Cancer, which further support
our data in CSCC and emphasize on their probable role in maintenance
of solid tumors.
To accomplish a biological event, functional proteins may crosstalk
with each other. Therefore, we created the protein interactome using
the known interactions of the identied functional genes, called the
Cervical Cancer Interaction Network (CCIN), in order to realize the
role of these proteins in this disease condition. In addition, identication
of proteins with having biologically signicant interactions may also aid
the search for new therapeutic options to treat advanced CSCC patients.
Though several previous networks based analysis was done in CC
[24,31,48,49] but here, we applied a unique weighted network analysis
(the node effect property) to the CCIN containing experimentally veried protein-protein interactions (PPIs) of the 78 identied genes. Correlation between degree of a protein and its node effect (effs) property
was found to be low indicating that high degree may not always reect
the important players. Based on Node effect score rank, PARP1
(1q42.12) and ATR (3q23) appears to play a crucial role in CSCC. Therefore, we tried to validate amplication and over-expression of PARP1
and ATR during tumor development (23 FIGO stage I/II and 32 FIGO
stage III/IV). Similar to our integrative analysis results, frequent amplication of their respective loci and over-expression at transcriptome
level was found. Immunocytochemical analysis in CSCC cell line SiHa
provided evidence of ATR and PARP1 to be over-expressed (intense nuclear localization) at the proteome level as well. Overall 74.54% (41 out
of 55) of CSCC patients showed amplication of either ATR and/or
PARP1. Interestingly, more number of CSCC tumors showed amplication of ATR (60%) than PARP1 (41.82%), with a gradual increase in alteration was observed from early (FIGO stage I/II) to advanced (FIGO stage
III/IV) CSCC. Moreover, both these genes are involved in DNA damage
repair and alteration of either ATR and/or PARP1 showed overall poor
prognosis of Indian CSCC patients. It seems that more number of CSCC
cells, harbouring alteration of ATR and/or PARP1, may be getting selected to keep up with the change in microenvironment during tumor progression. This is in order with the thought that inhibitors against (overexpressed) PARP1 and ATR may target defects in DNA repair of cancer
cells leading to enhanced efcacy of conventional DNA damaging chemotherapeutic drugs [5052]. Finally, strengthening our observation,
recently, a clinical trial using ATR inhibitor (AZD6738) in combination
with PARP1 inhibitor, Olaparib had been under way (module 2) to
treat advanced malignancies including advanced cervical cancer
(NCT02264678).
5. Conclusion
Altogether, the present study helped us to identify frequent chromosomal aberrations in advanced CSCC patients of Indian origin. It also
shed light on the biological relevance of these alterations based on our
novel systemic approach. The identied 78 altered genes have been observed to interact with other proteins to regulate crucial signalling and
metabolic pathways. Among these, amplication and over-expression
of ATR (3q23) and PARP1 (1q42.12) was validated in independent
CSCC samples paving the way to study all these newly identied
genes in detail to completely understand their synergistic role in pathogenesis of cervical squamous cell carcinoma leading to better therapeutic intervention and prognosis.
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.bbagen.2016.09.014.
Conict of interest
The authors have no conict of interest to declare.

Transparency Document
The Transparency document associated with this article can be
found, in online version.
Acknowledgement
The authors thank the Director, Chittaranjan National Cancer Institute, Kolkata, India. They acknowledge Genotypic Technology Private
Limited, Bangalore, especially Mr. Mohamed Aiyaz for the microarray
processing and data analysis. They also thank Dr. Partha Sarathi
Dasgupta, Emeritus Scientist, Chittaranjan National Cancer Institute,
for his valuable suggestions during the study. We would also like to
thank Mr. Md. Saimul Islam and Dr. Santu Kumar Saha for their enormous help and valuable suggestions during preparation of manuscript.
This work was supported by UGC-NET/JRF grant [Sr No. 2121130723]
to Mr. S. Samadder from University Grants Commission (UGC), Council
of Scientic and Industrial Research CSIR-JRF/NET grant [File No.09/
030(0059)/2010-EMR-I] to Mr. C. Chakraborty and grants from Department of Science and Technology (DST), Government of India [SR/SO/HS116/2007 of dt 07/09/2011] and CSIR, Government of India [No.
60(0111)/14/EMR-II of dt 03/11/2014] to Dr. C. K. Panda.
References
[1] A. Jemal, F. Bray, M.M. Center, J. Ferlay, E. Ward, D. Forman, Global cancer statistics,
CA Cancer J. Clin. 61 (2011) 6990.
[2] K. Kaarthigeyan, Cervical cancer in India and HPV vaccination, Indian J. Med.
Paediatr. Oncol. 33 (2012) 712.
[3] X. Castellsague, Natural history and epidemiology of HPV infection and cervical
cancer, Gynecol. Oncol. 110 (2008) S4S7.
[4] S. Mitra, D. Mazumder-Indra, R.K. Mondal, P.S. Basu, A. Roy, S. Roychoudhury, C.K.
Panda, Inactivation of SLIT2-ROBO1/2 pathway in premalignant lesions of uterine
cervix: clinical and prognostic signicances, PLoS One 7 (2012), e38342.
[5] R. Richart, A.S. Ludwig Jr., Alterations in chromosomes and DNA content in gynecologic neoplasms, Am. J. Obstet. Gynecol. 104 (1969) 463471.
[6] R. Wakonig-Vaartaja, D.T. Hughes, Chromosomal anomalies in dysplasia, carcinomain-situ, and carcinoma of cervix uteri, Lancet 2 (1965) 756759.
[7] G. Narayan, V.V. Murty, Integrative genomic approaches in cervical cancer: implications for molecular pathogenesis, Future Oncol. 6 (2010) 16431652.
[8] L.K. Thomas, J.L. Bermejo, S. Vinokurova, K. Jensen, M. Bierkens, R. Steenbergen, M.
Bergmann, M. von Knebel Doeberitz, M. Reuschenbach, Chromosomal gains and
losses in human papillomavirus-associated neoplasia of the lower genital tract - a
systematic review and meta-analysis, Eur. J. Cancer 50 (2014) 8598.
[9] S.J. Freeman, A.M. Aly, M.Y. Kataoka, H.C. Addley, C. Reinhold, E. Sala, The revised
FIGO staging system for uterine malignancies: implications for MR imaging, Radiographics 32 (2012) 18051827.
[10] S. Chakrabarti, S. Dasgupta, S. Roy, A. Bhar, A. Sengupta, A. Roy, K. Ray, N.P.
Bhattacharyya, B. Roy, C.K. Panda, S. Roychoudhury, Microsatellite instability in
squamous cell carcinoma of head and neck from the Indian patient population,
Int. J. Cancer 92 (2001) 555561.
[11] D. Mazumder Indra, R.K. Singh, S. Mitra, S. Dutta, C. Chakraborty, P.S. Basu, R.K.
Mondal, S. Roychoudhury, C.K. Panda, Genetic and epigenetic changes of HPV16 in
cervical cancer differentially regulate E6/E7 expression and associate with disease
progression, Gynecol. Oncol. 123 (2011) 597604.
[12] D.R. Rhodes, J. Yu, K. Shanker, N. Deshpande, R. Varambally, D. Ghosh, T. Barrette, A.
Pandey, A.M. Chinnaiyan, ONCOMINE: a cancer microarray database and integrated
data-mining platform, Neoplasia 6 (2004) 16.
[13] A. Franceschini, D. Szklarczyk, S. Frankild, M. Kuhn, M. Simonovic, A. Roth, J. Lin, P.
Minguez, P. Bork, C. von Mering, L.J. Jensen, STRING v9.1: protein-protein interaction
networks, with increased coverage and integration, Nucleic Acids Res. 41 (2013)
D808D815.
[14] C. Lopez-Gines, R. Gil-Benso, R. Ferrer-Luna, R. Benito, E. Serna, J. Gonzalez-Darder,
V. Quilis, D. Monleon, B. Celda, M. Cerda-Nicolas, New pattern of EGFR amplication
in glioblastoma and the relationship of gene copy number with gene expression
prole, Mod. Pathol. 23 (2010) 856865.
[15] G.P. Maiti, P. Mondal, N. Mukherjee, A. Ghosh, S. Ghosh, S. Dey, J. Chakrabarty, A.
Roy, J. Biswas, S. Roychoudhury, C.K. Panda, Overexpression of EGFR in head and
neck squamous cell carcinoma is associated with inactivation of SH3GL2 and
CDC25A genes, PLoS One 8 (2013), e63440.
[16] P. Shannon, A. Markiel, O. Ozier, N.S. Baliga, J.T. Wang, D. Ramage, N. Amin, B.
Schwikowski, T. Ideker, Cytoscape: a software environment for integrated models
of biomolecular interaction networks, Genome Res. 13 (2003) 24982504.
[17] N. Rappaport, M. Twik, N. Nativ, G. Stelzer, I. Bahir, T.I. Stein, M. Safran, D. Lancet,
MalaCards: A Comprehensive Automatically-Mined Database of Human Diseases,
Current protocols in bioinformatics / editoral board, Andreas D. Baxevanis
[et al.], 47 (2014) 1 24 21-21 24 19.
[18] M. Safran, I. Dalah, J. Alexander, N. Rosen, T. Iny Stein, M. Shmoish, N. Nativ, I. Bahir,
T. Doniger, H. Krug, A. Sirota-Madi, T. Olender, Y. Golan, G. Stelzer, A. Harel, D.

A. Roychowdhury et al. / Biochimica et Biophysica Acta 1861 (2017) 28992911

[19]
[20]

[21]

[22]
[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

Lancet, GeneCards Version 3: the human gene integrator, Database 2010 (2010)
(baq020).
A. Shlien, D. Malkin, Copy number variations and cancer, Genome Med. 1 (2009) 62.
M.T. Barrett, A. Scheffer, A. Ben-Dor, N. Sampas, D. Lipson, R. Kincaid, P. Tsang, B.
Curry, K. Baird, P.S. Meltzer, Z. Yakhini, L. Bruhn, S. Laderman, Comparative genomic
hybridization using oligonucleotide microarrays and total genomic DNA, Proc. Natl.
Acad. Sci. U. S. A. 101 (2004) 1776517770.
P. Van Loo, S.H. Nordgard, O.C. Lingjaerde, H.G. Russnes, I.H. Rye, W. Sun, V.J.
Weigman, P. Marynen, A. Zetterberg, B. Naume, C.M. Perou, A.L. Borresen-Dale,
V.N. Kristensen, Allele-specic copy number analysis of tumors, Proc. Natl. Acad.
Sci. U. S. A. 107 (2010) 1691016915.
S.J. Hwang, K.R. Shroyer, Biomarkers of cervical dysplasia and carcinoma, J. Oncol.
2012 (2012) 507286.
A. Rodolakis, I. Biliatis, H. Symiakaki, E. Kershnar, M.W. Kilpatrick, D. Haidopoulos, N.
Thomakos, A. Antsaklis, Role of chromosome 3q26 gain in predicting progression of
cervical dysplasia, Int. J. Gynecol. Cancer 22 (2012) 742747.
V.V. Sahasrabuddhe, P. Luhn, N. Wentzensen, Human papillomavirus and cervical
cancer: biomarkers for improved prevention efforts, Future Microbiol 6 (2011)
10831098.
T.C. Wright, J. Compagno, P. Romano, V. Grazioli, Y. Verma, E. Kershnar, T. Tafas,
M.W. Kilpatrick, Amplication of the 3q chromosomal region as a specic marker
in cervical cancer, Am. J. Obstet. Gynecol. 213 (51) (2015) e51e58.
O. Vazquez-Mena, I. Medina-Martinez, E. Juarez-Torres, V. Barron, A. Espinosa, N.
Villegas-Sepulveda, L. Gomez-Laguna, K. Nieto-Martinez, L. Orozco, E. RomanBasaure, S. Munoz Cortez, M. Borges Ibanez, C. Venegas-Vega, M. GuardadoEstrada, A. Rangel-Lopez, S. Kofman, J. Berumen, Amplied genes may be
overexpressed, unchanged, or downregulated in cervical cancer cell lines, PLoS
One 7 (2012), e32667.
M. Bierkens, O. Krijgsman, S.M. Wilting, L. Bosch, A. Jaspers, G.A. Meijer, C.J. Meijer,
P.J. Snijders, B. Ylstra, R.D. Steenbergen, Focal aberrations indicate EYA2 and hsamiR-375 as oncogene and tumor suppressor in cervical carcinogenesis, Genes
Chromosom. Cancer 52 (2013) 5668.
Y.W. Choi, S.M. Bae, Y.W. Kim, H.N. Lee, Y.W. Kim, T.C. Park, D.Y. Ro, J.C. Shin, S.J.
Shin, J.S. Seo, W.S. Ahn, Gene expression proles in squamous cell cervical carcinoma using array-based comparative genomic hybridization analysis, Int. J. Gynecol.
Cancer 17 (2007) 687696.
J.N. Kloth, J. Oosting, T. van Wezel, K. Szuhai, J. Knijnenburg, A. Gorter, G.G. Kenter,
G.J. Fleuren, E.S. Jordanova, Combined array-comparative genomic hybridization
and single-nucleotide polymorphism-loss of heterozygosity analysis reveals complex genetic alterations in cervical cancer, BMC Genomics 8 (2007) 53.
M. Koch, M. Wiese, Gene expression signatures of angiocidin and darapladib treatment connect to therapy options in cervical cancer, J. Cancer Res. Clin. Oncol. 139
(2013) 259267.
K.L. Mine, N. Shulzhenko, A. Yambartsev, M. Rochman, G.F. Sanson, M. Lando, S.
Varma, J. Skinner, N. Volfovsky, T. Deng, S.M. Brenna, C.R. Carvalho, J.C. Ribalta, M.
Bustin, P. Matzinger, I.D. Silva, H. Lyng, M. Gerbase-DeLima, A. Morgun, Gene network reconstruction reveals cell cycle and antiviral genes as major drivers of cervical cancer, Nat. Commun. 4 (2013) 1806.
G. Narayan, V. Bourdon, S. Chaganti, H. Arias-Pulido, S.V. Nandula, P.H. Rao, L.
Gissmann, M. Durst, A. Schneider, B. Pothuri, M. Mansukhani, K. Basso, R.S.
Chaganti, V.V. Murty, Gene dosage alterations revealed by cDNA microarray analysis
in cervical cancer: identication of candidate amplied and overexpressed genes,
Genes Chromosom. Cancer 46 (2007) 373384.
T. Rajkumar, K. Sabitha, N. Vijayalakshmi, S. Shirley, M.V. Bose, G. Gopal, G.
Selvaluxmy, Identication and validation of genes involved in cervical
tumourigenesis, BMC Cancer 11 (2011) 80.
L. Scotto, G. Narayan, S.V. Nandula, H. Arias-Pulido, S. Subramaniyam, A. Schneider,
A.M. Kaufmann, J.D. Wright, B. Pothuri, M. Mansukhani, V.V. Murty, Identication of

[35]

[36]

[37]

[38]

[39]

[40]
[41]

[42]

[43]

[44]
[45]

[46]
[47]
[48]
[49]

[50]
[51]
[52]

2911

copy number gain and overexpressed genes on chromosome arm 20q by an integrative genomic approach in cervical cancer: potential role in progression, Genes
Chromosom. Cancer 47 (2008) 755765.
R.K. Singh, D. Indra, S. Mitra, R.K. Mondal, P.S. Basu, A. Roy, S. Roychowdhury, C.K.
Panda, Deletions in chromosome 4 differentially associated with the development
of cervical cancer: evidence of slit2 as a candidate tumor suppressor gene, Hum.
Genet. 122 (2007) 7181.
J.Y. Song, J.K. Lee, N.W. Lee, H.H. Jung, S.H. Kim, K.W. Lee, Microarray analysis of normal cervix, carcinoma in situ, and invasive cervical cancer: identication of candidate genes in pathogenesis of invasion in cervical cancer, Int. J. Gynecol. Cancer 18
(2008) 10511059.
S.M. Wilting, J. de Wilde, C.J. Meijer, J. Berkhof, Y. Yi, W.N. van Wieringen, B.J.
Braakhuis, G.A. Meijer, B. Ylstra, P.J. Snijders, R.D. Steenbergen, Integrated genomic
and transcriptional proling identies chromosomal loci with altered gene expression in cervical cancer, Genes Chromosom. Cancer 47 (2008) 890905.
Y.F. Wong, T.H. Cheung, G.S. Tsao, K.W. Lo, S.F. Yim, V.W. Wang, M.M. Heung, S.C.
Chan, L.K. Chan, T.W. Ho, K.W. Wong, C. Li, Y. Guo, T.K. Chung, D.I. Smith,
Genome-wide gene expression proling of cervical cancer in Hong Kong women
by oligonucleotide microarray, Int. J. Cancer 118 (2006) 24612469.
C. Gao, T. Ma, L. Pang, R. Xie, Activation of P21-activated protein kinase 2 is an independent prognostic predictor for patients with gastric cancer, Diagn. Pathol. 9
(2014) 55.
S. Matt, T.G. Hofmann, The DNA Damage-induced Cell Death Response: A Roadmap
to Kill Cancer Cells, Cellular and molecular life sciences: CMLS, 2016.
J. Park, J.M. Kim, J.K. Park, S. Huang, S.Y. Kwak, K.A. Ryu, G. Kong, J. Park, B.S. Koo, Association of p21-activated kinase-1 activity with aggressive tumor behavior and
poor prognosis of head and neck cancer, Head Neck 37 (2015) 953963.
B. Salhia, J. Kiefer, J.T. Ross, R. Metapally, R.A. Martinez, K.N. Johnson, D.M. DiPerna,
K.M. Paquette, S. Jung, S. Nasser, G. Wallstrom, W. Tembe, A. Baker, J. Carpten, J.
Resau, T. Ryken, Z. Sibenaller, E.F. Petricoin, L.A. Liotta, R.K. Ramanathan, M.E.
Berens, N.L. Tran, Integrated genomic and epigenomic analysis of breast cancer
brain metastasis, PLoS One 9 (2014), e85448.
M.J. Schiewer, J.F. Goodwin, S. Han, J.C. Brenner, M.A. Augello, J.L. Dean, F. Liu, J.L.
Planck, P. Ravindranathan, A.M. Chinnaiyan, P. McCue, L.G. Gomella, G.V. Raj, A.P.
Dicker, J.R. Brody, J.M. Pascal, M.M. Centenera, L.M. Butler, W.D. Tilley, F.Y. Feng,
K.E. Knudsen, Dual roles of PARP-1 promote cancer growth and progression, Cancer
Discov. 2 (2012) 11341149.
P.T. Tran, N. Erdeniz, L.S. Symington, R.M. Liskay, EXO1-a multi-tasking eukaryotic
nuclease, DNA Repair 3 (2004) 15491559.
S. Xiao, R.M. Chang, M.Y. Yang, X. Lei, X. Liu, W.B. Gao, J.L. Xiao, L.Y. Yang, Actin-like
6A predicts poor prognosis of hepatocellular carcinoma and promotes metastasis
and epithelial-mesenchymal transition, Hepatology 63 (4) (2016) 12561271.
C.V. Dang, Links between metabolism and cancer, Genes Dev. 26 (2012) 877890.
V. Sherwood, WNT signaling: an emerging mediator of cancer cell metabolism? Mol.
Cell. Biol. 35 (2015) 210.
S. Jalan, K. Kanhaiya, A. Rai, O.R. Bandapalli, A. Yadav, Network topologies decoding
cervical cancer, PLoS One 10 (2015), e0135183.
N. Wang, Z. Xu, K. Wang, M. Zhu, Y. Li, Construction and analysis of regulatory genetic networks in cervical cancer based on involved microRNAs, target genes, transcription factors and host genes, Oncol. Lett. 7 (2014) 12791283.
K.M. Foote, A. Lau, J.W. Nissink, Drugging ATR: progress in the development of specic inhibitors for the treatment of cancer, Future Med. Chem. 7 (2015) 873891.
N.V. Malyuchenko, E.Y. Kotova, O.I. Kulaeva, M.P. Kirpichnikov, V.M. Studitskiy,
PARP1 inhibitors: antitumor drug design, Acta Nat. 7 (2015) 2737.
S.S. Singh, J.A. Sarma, L. Narasu, R. Dayam, S. Xu, N. Neamati, A review on PARP1
inhibitors: pharmacophore modeling, virtual and biological screening studies to
identify novel PARP1 inhibitors, Curr. Top. Med. Chem. 14 (2014) 20202030.