Flavio Licciulli – Ricerca bioinformatica e sue applicazioni per l’identifica...
BiPday 2014 --Creanza Teresa
1. Teresa M.CREANZA1,2, A. Piepoli3, A. Andriulli3, N. Ancona1
1 Institute of Intelligent Systems for Automation, National Research Council, Bari, Italy,
2Center for Complex Systems in Molecular Biology and Medicine, University of Torino, Italy,
3Department of Medical Sciences, Division and Laboratory of Gastroenterology,
IRCCS Casa Sollievo della Sofferenza, San Giovanni Rotondo, Italy
19/12/2014
Differential miRNA-mRNA co-expression networks
in colorectal cancer
2. Outline
I. Changes in connectivity of gene co-expression networks in cancer
II. Re-wiring of miRNA-mRNA co-expression networks in colorectal cancer
Motivations and Methodologies
Why should one use differential miRNA-mRNA co-expression measures?
Strategies to score miRNAs and biological processes for differential
co-expression in tumor tissues
Results in terms of
miRNA-mediated molecular mechanisms underlying the pathogenesis and
new potential biomarkers
III. The differential approaches based on conditional dependencies to reveal
alterations of mRNA modulatory activities on the interactions between miRNAs
and known cancer genes.
IV. Conclusions and Future perspectives
219/12/2014
3. 19/12/2014
Research direction
Background. “Several recent interaction mapping studies have demonstrated
the power of differential analysis for elucidating fundamental biological responses,
revealing that the architecture of an interactome can be massively re-wired during
a cellular or adaptive response.”
Ideker et al., Molecular Systems Biology 2012.
The goal is to develop computational approaches to investigate the re-wiring of
gene networks depending on the disease state. This requires the inferring of gene
networks from experimental RNAseq or microarray expression data in normal and
tumor phenotypes.
Changes in connectivity of gene co-expression networks in cancer
4. Re-wiring of gene cancer networks.
19/12/2014
A Kolmogorov-Smirnov test showed that all
cancer networks are characterized by a gene
degree which is stochastically decreased
with respect to the normal graphs
(P < 10−100).
Loss of Connectivity in Cancer Co-Expression Networks
Analysis of differential gene connectivity to capture alterations in gene co-expression
networks as a consequence of genetic alterations and post-translational modifications
that critically modify the activity of the coded protein.
Its potential to reveal
1. general traits of cancer networks
2. new potential biomarkers.
5. Differentially connectivity highlights known cancer genes.
The findings of our study show a correspondence between known cancer
biomarkers and differentially connected (DC) genes.
For colorectal cancers (CRC), DC gene list (P<0.005) resulted enriched in
tumor-suppressor genes and oncogenes commonly associated with
colorectal cancer:
• Cancer Gene Census (P=0.0067, Fisher’s exact test),
• KEGG Disease H00020 Colorectal cancer gene list (P=0.063)
• genes mutated in CRC as reported in Wood et al. (P=0.05).
Hence they yield the encouraging evidence that the biological meaning of
co-expression changes can be interpreted in terms of modifications of
cancer genome landscape.
Re-wiring of gene cancer networks.
Futreal PA, Coin L, Marshall M, Down T, Hubbard T, et al. (2004) A census of human cancer genes. Nat Rev Cancer
4: 177–183.
Wood LD, Parsons DW, Jones S, Lin J, Sjo ̈blom T, et al. (2007) The genomic
landscapes of human breast and colorectal cancers. Science 318: 1108–1113
19/12/2014
Anglani R, Creanza TM, Liuzzi VC, Piepoli A, Panza A, et al. (2014) Loss of Connectivity in
Cancer Co-Expression Networks. PLoS ONE 9(1): e87075.
6. Differential miRNA-mRNA co-expression networks in colorectal cancer.
The problem.
Colorectal cancer (CRC) is one of the most common neoplasms in the world and its molecular biology is one
of the most intensively and successfully studied.
MicroRNAs (miRNAs) are small, non-coding RNAs that function as post-transcriptional regulators of mRNA
expression. A single miRNA can target hundreds of mRNAs and a single mRNA can contain functional binding
sites for several miRNAs, building miRNA-mRNA networks that regulate the expression of a large portion of
the human transcriptome.
19/12/2014
miRNA
miRNA
miRNA
mRNA
mRNA
mRNA
Calin GA, Croce CM., MicroRNA signatures in human cancers. Nat Rev Cancer (2006);6:857–66.
Volinia S, et al., Reprogramming of miRNA networks in cancer and leukemia. Genome Res. 2010;20:589–599.
Cell differentiation
Cell proliferation
Apoptosis
Dysregulation of miRNA networks has been shown to be a key feature of many human
cancers
7. In cancer, genetic variants in miRNA genes and mRNA targets can alter miRNA-mediated gene
regulation. Moreover, mRNAs or non-coding RNAs that regulate miR-activities on their target
RNAs may have an altered modulatory acitvity.
Our work aims to to investigate alterations in miRNA-mRNA interactions resulting from the
aforementioned modifications that critically influence miRNA activity on gene transcription.
A system-level analysis: miRNA and mRNA expression arrays in cancer and normal tissues.
A data driven approach that does not consider any a priori information
• a complete coverage of the human genes on the chip,
• little bias due to the knowledge obtained from the published literature,
• the ability to infer condition specific relationships.
19/12/2014
Ryan BM, et al., Genetic variation in microRNA networks: the implications for cancer research. Nat. Rev. Cancer (2010); 389-402.
Differential miRNA-mRNA co-expression networks in colorectal cancer.
Our approach.
8. 19/12/2014
Our work aims to capture alterations in miRNA-mRNA co-expression
networks resulting from the aforementioned modifications that critically
influence miRNA activity on gene transcription.
Interaction term
Differential miRNA-mRNA co-expression networks in colorectal cancer.
Our approach.
9. miRNA
miRNA
miRNA
mRNA
mRNA
mRNA
mRNA-microRNA
Spearman correlations
on normal tissue data
miRNA
miRNA
miRNA
mRNA
mRNA
mRNA
mRNA-microRNA
Spearman correlations
On tumor tissue data
ri,j
hc
ri,j
d
Predicted miRNA-target interactions overlapped with CLIP-Seq
data were collected from StarBase v2.0
APC
let-7a
miR-155
KRAS
Differential miRNA-mRNA co-expression networks in colorectal cancer.
The analysis.
Li, J H, et al. starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction
networks from large-scale CLIP-Seq data. Nucleic Acids Res 42, D92–D97 (2014).
We analyzed paired expression levels of human mRNAs and miRNAs on 14 cancer and 14 matched normal
colorectal tissues measured on Affymetrix Human Exon 1.0 ST Array platform (17400 mRNAs) and on
[miRNA-1_0] Affymetrix miRNA Array platform (847 miRNAs), respectively.
19/12/2014
Kolmogorov-Smirnov test
P < 10−100
10. Differential miRNA-mRNA co-expression networks in colorectal cancer.
Our approach.
Analysis of changes in miRNA-
mRNA CRC interactions in terms
of differential co-expressions
relative to normal condition.
Differential co-expression (DC)
P-values are computed by label
permutation tests.
Pi,j
= # |Di,j
*
|>|Di,j
|{ }/ s
Di,j
= ri,j
d
-ri,j
hc
Anglani R, Creanza TM, et al. (2014) Loss of Connectivity in Cancer Co-Expression
Networks, PLoS ONE 9(1): e87075
Good P (2000) Permutation Tests: A Practical Guide to Resampling Methods for Testing
Hypotheses (Springer Series in Statistics). Springer.
19/12/2014
Fisher transformation of Spearman correlation
between the i-th transcript and the j-th mRNA.
s number of permutations
ri, j
11. Jiang Q et al. miR2Disease: a manually curated database for microRNA deregulation in human disease.
Nucleic Acids Res. 2009; 37: D98–D104
Rossi S et al. microRNAs in colon cancer: a roadmap for discovery. FEBS Lett. 2012; 586(19):3000-7.
miRNAs causal in CRC
miRNAs causal in CRC are characterized by DC P-values
which are stochastically decreased with respect to the
random miRNAs (P< 10−100).
19/12/2014
Differential miRNA-mRNA co-expression networks in colorectal cancer.
g
j
jiiZ
1
,
g = number of genes on the chip
P-values by permutation tests
12. An enrichment pathway analysis for differential co-expression.
Newton MA, et al, Random-set methods identify distinct aspects of the enrichment signal in gene-set analysis.
Ann. Appl. Stat.1(1): 85-106 (2007)
METHOD
g = number of genes in the pathway
M = number of miRNAs on chip
Enrichment P-values.
Random set procedure:
• label permutation and
• re-standardization
Z = Di,j
j=1
M
å
i=1
g
å
19/12/2014
We assessed the evidence of association of a pathway with the phenotype supported by
changes in the interactions between the pathway and the entire miRNome.
13. Our integrative analysis suggested an alteration in CRC tissues in the interplay
between miRNAs and the eukaryotic translation initiation factor 3 (eIF3) which has a
central role in recruiting both mRNAs and the cellular translation machinery to form
translation initiation complexes.
Differential miRNA-mRNA co-expression pathway analysis.
The top ranked canonical pathway C2Cp (MsigDB)
MIPS EIF3 COMPLEX P-value=0.002
GeneOntology
GeneOntology Terrms pvalue fdr
EUKARYOTIC_TRANSLATION_INITIATION_FACTOR_3_COMPLEX 0.000000 0.000000
CENTROSOME 0.001000 0.275000
TRANSLATIONAL_INITIATION 0.001000 0.275000
ESTABLISHMENT_OF_LOCALIZATION 0.003000 0.315000
PHOTORECEPTOR_CELL_MAINTENANCE 0.003000 0.315000
TRANSPORT 0.003000 0.315000
KINESIN_COMPLEX 0.004000 0.315000
MICROTUBULE_ORGANIZING_CENTER 0.004000 0.315000
REGULATION_OF_TRANSLATIONAL_INITIATION 0.004000 0.315000
19/12/2014
Jackson RJ, Hellen CU, Pestova TV (2010) The mechanism of eukaryotic translation
initiation and principles of its regulation. Nat Rev Mol Cell Biol 11:113–127.
14. 19/12/2014
Polymorphism chromosome gene
Rs16892766 8q23.3 EIF3H
Pittman et al. (2010) Allelic variation at the 8q23.3 colorectal cancer risk locus
functions as a cis-acting regulator of EIF3H. Plos Genet. 6 e1001126
15. All miRNAs were ranked on the basis of the number of
mRNAs differentially co-expressed (P<=0.001)with
them.
miR-23b
miR-132
miR-152
miR-27b
miR-99b-star
miR-574-3p
miR-648
miR-634
miR-521
miR-138
miR-141
miR-214 miR-23b
EUKARYOTIC TRANSLATION
INITIATION FACTOR 3 COMPLEX
BIOPOLYMER
CATABOLIC PROCESS
ORGANELLE MEMBRANE
TRANSLATIONAL INITIATION
RIBONUCLEOPROTEIN COMPLEX
BIOGENESIS AND ASSEMBLY
PROTEIN RNA COMPLEX
ASSEMBLY
DOUBLE STRAND
BREAK REPAIR
CELLULAR MACROMOLECULE
CATABOLIC PROCESS
PROTEIN AMINO ACID LIPIDATION
LIPOPROTEIN METABOLIC PROCESS
Zhang H, et al. Genome—wide functional screening of miR-23b as a pleiotropic modulator suppressing cancer
metastasis. Nat. Commun. 2011;2:554.
To the best of our best knowledge, this is the first evidence in which so many critical pathways are
interconnected in the regulation of metastasis under the control of a single miRNA miR-23b
19/12/2014
16. Dysregulation of physiologic microRNA (miR) activity has been shown to play an
important role in tumor initiation and progression. Therefore, molecular species
that can regulate miR activity on their target RNAs without affecting the
expression of relevant mature miRs may play equally relevant roles in cancer.
The role of modulators of miRNA functions in tumor onset.
We developed a strategy to evaluate the role of modulators (M) of miRNA
activities in controlling normal cell physiology and pathogenesis. We evaluated if
changes of their modulatory activity are significantly associated to tumor
progression.
Sumazin, Pavel, et al. "An extensive microRNA-mediated network of RNA-RNA interactions regulates
established oncogenic pathways in glioblastoma." Cell 147.2 (2011): 370-381.
19/12/2014
17. mRNA
miRNA
r mRNA,miRNA( )¹ 0 r mRNA,miRNA | M( )= 0AND
mRNA M
miRNA
The role of modulators of miRNA functions in tumor onset.
We examined changes in the modulatory activity of M that lead to the differential
miRNA-mRNA co-expression.
19/12/2014
18. competing endogenous RNA (ceRNA) pairs
Mediators of miRNA activity on cancer driver genes
APC , FBXW7,KRAS, PTEN: CRC cancer genes from Census and
differential co-expressed with a significant number of transcripts.
The respective ceRNAs: Predicted ceRNA pairs by integrating the
interactions from potential microRNA targets (miRanda/mirSVR)
overlapping with CLIP-Seq data were collected from StarBase v2.0.
19/12/2014
RNAs modulate each other through their common miR regulatory program (sponge effect).
19. For APC, we considered all possible triplets including APC, its ceRNAs and all measured
miRNAs.
r APC,miRNA( )¹ 0
r APC,miRNA|ceRNA( )= 0
AND
P<0.05
APC
M1 M2 M3
miRNA1 miRNA2 miRNA3 …
APC-miRNA networks in tumor and in normal tissues.
Normal tissues
APC
M1 M2 M3
miRNA1 miRNA2 miRNA3 …
Cancer tissues
miRNA-modulator pairs relative to APC in each condition.
19/12/2014
20. Anglani R, Creanza TM, et al. (2014) Loss of Connectivity in Cancer
Co-Expression Networks, PLoS ONE 9(1): e87075
APC KRAS FBXW7 PTEN
P = 10−13
Paired t-test P-values
P = 10−7
P = 10−9 P = 10−14
mRNAs Mediators of miRNA activity on cancer driver genes
The average number of modulators per miRNA significantly decreases in tumor tissues.
19/12/2014
21. APC KRAS FBXW7 PTEN
P = 10−38P = 10−51 P < 10−100
The average number of miRNAs per modulator significantly decreases in tumor tissues.
Mediators of miRNA activity on cancer driver genes
Paired t-test P-values
P = 10−9
CRC tumors are characterized by a lower modulatory activity on the interactions
between CRC genes and miRNAs.
19/12/2014
22. mRNAs Mediators of miRNA activity on cancer driver genes
All measured miRNAs were ranked on the basis of the loss of the number of mRNAs
mediators in tumor tissues.
19/12/2014
APC KRAS FBXW7 PTEN
hsa-miR-1297_st hsa-miR-1273_st hsa-miR-1273_st hsa-miR-106a-star_st
hsa-miR-520f_st hsa-miR-16-1-star_st hsa-miR-19b-2-
star_st
hsa-miR-1273_st
hsa-miR-1825_st hsa-miR-521_st hsa-miR-16-1-star_st hsa-miR-1293_st
hsa-miR-521_st hsa-miR-500-star_st hsa-miR-574-3p_st hsa-miR-19b-2-star_st
hsa-miR-1273_st hsa-miR-574-3p_st hsa-miR-33b-star_st hsa-miR-520f_st
hsa-miR-518d-3p_st hsa-miR-19b-2-
star_st
hsa-miR-517b_st hsa-miR-517b_st
hsa-miR-16-1-star_st hsa-miR-106a-star_st hsa-miR-521_st hsa-miR-448_st
hsa-miR-19b-2-
star_st
hsa-miR-517b_st hsa-miR-507_st hsa-miR-1282_st
hsa-miR-527_st hsa-miR-527_st hsa-miR-335-star_st hsa-miR-578_st
hsa-miR-1293_st hsa-miR-561_st hsa-miR-527_st hsa-miR-640_st
hsa-miR-609_st hsa-miR-744_st hsa-miR-500-star_st hsa-miR-505-star_st
hsa-miR-937_st hsa-miR-23b_st hsa-miR-23b_st hsa-miR-518d-3p_st
hsa-miR-448_st hsa-let-7d_st hsa-miR-744_st hsa-miR-609_st
hsa-miR-561_st hsa-miR-520f_st hsa-miR-106a-star_st hsa-miR-125b-2-star_st
hsa-miR-33b-star_st hsa-miR-1297_st hsa-miR-367-star_st hsa-miR-23b_st
hsa-miR-15a_st hsa-miR-648_st hsa-miR-505-star_st hsa-miR-1825_st
Our analysis suggests that there are miRNAs whose mRNA-mediated action on the
colorectal cancer genes is critically altered in the tumor tissues.
23. Differential miRNA-mRNA co-expression networks in colorectal cancer.
Other References.
Conclusions
I presented a systems-level analysis of miRNA expression in CRC, by analysing miRNA levels and
integrating them with matched mRNA expression measured in both normal and tumor tissues.
Compared with other types of miRNA-mRNA interaction scores, using co-expression coefficients with
any a priori information has several advantages:
• a complete coverage of the human genes on the chip,
• little bias due to the knowledge obtained from the published literature,
• the ability to infer condition specific relationships.
Unveiling differential miRNA-mRNA co-expression properties allows
1. to gain insights into miRNA-mediated molecular mechanisms underlying the pathogenesis of the
disease
2. and may suggest novel miRNA drug targets to be validated.
Finally, considering differential approches based on conditional dependencies allows to reveal
alterations on modulatory activity on the interactions between miRNAs and known cancer genes.
Future works
Investigating the pathologic role of long non coding RNAs by differential correlation approaches.
19/12/2014