Network modeling links breast cancer susceptibility and centrosome dysfunction

1
Network modeling links breast cancer
susceptibility and centrosome dysfunction

• Pujana et al.
• Nature genetics, 2007
• Presented by Meeyoung Park
• Feb. 29, 2008

2
Outline

• Introduction
• Methods Results
• Discussion
• Conclusion

3
Motivation

• Most genes and their products interact in complex
  cellular networks, the properties which might be
  altered in cancer cells.
• Modeling the functional interrelationships
  between genes and/or proteins may be required for
  a deeper understanding of cancer molecular
  mechanisms.

4
Objective

• Modeling of global macromolecular networks to
  identify cancer genes and their products.

5
Outline

• Introduction
• Methods Results
• Discussion
• Conclusion

6
A network modeling strategy

• Macromolecular networks can be modeled on the
  basis of global correlations observed among
• Transcriptional profiling compendia,
  protein-protein interaction or interactome
  networks, and genomewide phenotypic profiling
  data sets
• Comparisons of interolog data sets from
  different organisms.

• Interolog A pair of molecular interactions X-Y
  and X'- Y.
• ( X-Y and X'-Y' are in two different
  organisms.)
• X is an interolog of X' while Y is an
  interolog of Y'.

7
Modeling macromolecular networks

• Strategy
• Integrates coexpression profiles in human
• Integrates functional associations derived from
  various functional omic data sets obtained in
  humans and model organisms.
• Reference genes/proteins
• BRCA1 and BRCA2
• identified by high-penetrance mutations
• ATM and CHEK2
• identified by low-penetrance mutations

8
Figure 1. Outline for the generation of the
BRCA-centered Network (BCN) model
9
Coexpression profiling

• Data
• 9,214 human genes in 101 samples and three cell
  lines
• Pearson correlation coefficient (PCC)
• Between each of the reference genes and all of
  the genes on the array

10
Functional associations

• Literature interaction (LIT-Int) network
• To determine the likelihood of predicting
  functional associations
• Curated published data from the scientific
  literature on protein interactions
• 103 proteins and 129 functional associations.

11
LIT-Int network for ATM, BRCA1, BRCA2 and CHEK2
12

• PCC 0.4 captures 36 of the LIT-Int functional
  associations.

b) PDF of transcriptional PCC values between gene
pairs for each of the four reference genes
13
Potential functional associations

• Expression intersection (XPRSS-Int)

• Generate the XPRSS-Int of the four coexpression
  sets
• 164 genes (PCC) 0.4
• 15 are present in
• LIT-Int data set

C) XPRSS-Int of the four reference coexpression
sets using PCC 0.4. LIT-Int genes included in
the XPRSS-Int set are shown.
14
Potential functional associations

• Randomly chosen sets do not overlap in
  coexpression levels.
• Results indicates that P

• Evaluate the significance of the XPRSS-Int

d) Distribution of the coexpression intersection
for randomly chosen sets of four genes and
comparison with the XPRSS-Int set using PCC 0.4.
15
XPRSS-Int and reference genes

• Functionally related in shared characteristics
• An enrichment of Gene Ontology (GO) terms
• Evolutionary conservation of coexpression
  patterns (Orthologs of XPRSS-Int and reference
  genes)
• Significant coexpression among 33 XPRSS-Int genes
  was observed when an expression data set used for
  the analysis of breast tumor cell lines.

16
The functional significance of the XPRSS-Int set
Expression changes in breast tumors
17
BRCA-centered network modeling

• Integrated data
• Gene expression profiling similarity above a
  given threshold.
• Saccharomyces cerevisiae and C. elegans
  microarray profiles (6,174 and 18,451 genes,
  respectively)
• Phenotypic similarity for 661 early embryogenesis
  C. elegans genes above a specific threshold.
• Genetic interactions for 1,347 S. cerevisiae
  genes.
• Protein physical interactions
• binary interactions, complex co-memberships and
  biochemical interactions (protein modification))
  for 3,458 S. cerevisiae, 4,588 C. elegans (WI6
  data set), 7,198 Drosophila melanogaster and
  10,305 Homo sapiens proteins.

18
A BRCA-centered network model
19
Evaluation of BCN connectivity
20
(c) GO terms annotations reveal functional
clusters contained in distinct omic data sets
used to generate the BCN. C. elegans tac-1
functional associations of the C. elegans tac-1
gene and TAC-1 protein with connections to BCN
genes/proteins.
21
(d) Five criteria were integrated to estimate the
overall functional significance of XPRSS-Int
genes/proteins relative to breast cancer
reference genes/proteins. XPRSS-Int genes are
clustered according to the number of criteria
they match (from 5 to 0) and then ordered within
clusters according to their average PCC value for
BRCA1 (PCC-BRCA1).
22
Bioinformatic Analysis

• GO annotations from NetAffix(Affymetrix)
• Only grade 3 (poorly differentiated) tumors were
  used to study gene expression in BRCA1 mutation
  tumors
• P values for differential expression were
  determined by two-tailed Student's t-test (P 0.10)
• The BRCA1mut coexpression network was generated
  with the Graphviz graph visualization package
• Orthologs were defined by reciprocal BLASTP best
  hit (P

23
Predictions based on the BCN model

• HMMR
• encodes the hyaluronan-mediated motility receptor
  (HMMR, also known as RHAMM).
• It has the highest PCC coexpression value
  relative to BRCA1 (0.9)
• It is known that HMMR may have a potential role
  in centrosome function in conjunction with BRCA1.

24
Centrosome

• The main microtubule organizing center (MTOC) of
  the animal cell

Centriole
From http//micro.magnet.fsu.edu
25
Role of the centrosome in cell cycle progression
From http//wikipedia.org
26
New BRCA functional associations

• HMMR-centered interactome map
• Yeast two-hybrid screens

27
Experimentally Testing BCN predictions

• Coimmunoprecipitation

(b) Endogenous coimmunoprecipitation of CSPG6 and
BRCA1, and SMC1L1 and BRCA1 in 293 cells.(c)
Endogenous coimmunoprecipitation of HMMR and
BRCA1 in non- synchronized HeLa S3 cells.(d)
Cell cycle synchronization and coimmunoprecipitati
on assays in HeLa S3 cells.
28
Co-Immunoprecipitation

• Traditional IP

• Co- IP

From http//www.piercenet.com/Proteomics/
29
HMMR-BRCA1 and centrosome dysfunction

• Overexpression of HMMR and/or its biochemical
  modification resulting in centrosome
  amplification are early somatic molecular events
  that contribute to breast tumorigenesis.

30
HMMR and breast cancer risk

• Functional association investigation
• Study of HMMR haplotype-tagging SNPs in 923
  individually matched case-control pairs.
• Found a significant association between genetic
  variation in HMMR and early-onset breast cancer.
• HMMR may be a previously unknown susceptibility
  gene for breast cancer within diverse human
  populations.

31
(No Transcript)
32
Outline

• Introduction
• Methods Results
• Discussion
• Conclusion

33
Discussion

• Network modeling points to functional
  associations for genes/proteins.
• The observation supports the idea that the
  HMMR-centered interactome network described has a
  role in genomic stability and breast
  tumorigenesis.
• Genetic analysis supports HMMR as a newly defined
  breast cancer susceptibility gene, thereby
  delineating a genetic link between risk of breast
  cancer and centrosome dysfunction.

34
Conclusion

• The network modeling strategy is applicable to
  other types of cancer.
• It will help to discover more cancer-associated
  genes and to generate a wiring diagram of
  functional interactions between their products.

35
Thank You !