Identifying OxPhos dependent tumors across different cancer types Daniel Gusenleitner Adviser: Stefano Monti

Identifying OxPhos dependent tumors across different cancer types

Daniel GusenleitnerAdviser: Stefano Monti

Monti, Savage, Kutok, et al. Blood, 2005

DLBCL - Oxphos Subtype

OxPhos samples respond to targeted treatment

PPARγ antagonistsCaro et al., Cancer Cell, 2012

Trainings dataset

CCCbiomarker

Developing a biomarker for CCC

BCR HR OxPhos

signatures

Validation on

Goal

Identification of OxPhos samples

across different cancer types

Outline

• Evidence for OxPhos – NonOxPhos stratification in other tumor types

• Most genes of DLBCL OxPhos signature are also present in solid tumors, but we found additional concordant genes

• No evidence of association with genomic instability (mutations and/or SCNAs)

• Strong candidates of OxPhos/Non-OxPhos cell-lines

OxPhos subtype in Melanoma

OxP

hos

Sig

nat

ure

GSEA resultsP

GC

1a

p-value: 0.003711

OxPhos metagene

Datasets

• TCGA– LUAD: lung adenocarcinoma

– LUSC: lung squamous cell carcinoma

– BRCA: breast invasive carcinoma

– HNSC: head and neck squamous carcinoma

• Breast Curtis

• CCLE

Outline





Clustering is not always the best way to define a subtype

OxP

hos

sign

atur

e

DLBCL samples

Differences in background expression levels make classification infeasible

DLBCL Lung Cancer

Ox

Ph

os

met

a g

ene

ex

pre

ssio

n

Additional DLBCL

ASSIGN is able to score the activation of a signature

.. and to identify the genes that are the main drivers of the signature

ASSIGN is able to adjust for differing backgrounds

In BRCA OxPhos is activated in comparison to BCR …

… but a stratification in a heatmap is hard to quantify

Unlike active signatures, the weights for random signatures are exponentially

distributedsc

ale

d g

en

e w

eig

hts

OxPhos signature

active signature

random signatures

p-value: 3.28E-06 p-value: 0.06195

sca

led

ge

ne

we

igh

ts

OxPhos signature BCR signature

Gene weights for inactive signatures cannot be distinguished from weights of random signatures

p-values based on whether the signature follows an exponential distribution

OxPhos appears to be active in all TCGA datasets

Signature Dataset p-value

OxPhos BRCA 0.00162

OxPhos HNSC 1.30E-05

OxPhos LUAD 3.28E-06

OxPhos LUSC 8.84E-06

BCR BRCA 0.08553

BCR HNSC 0.17410

BCR LUAD 0.06195

BCR LUSC 0.00011

We can find the OxPhos signal in all cancer datasets we looked at

Expansion of the OxPhos signature

CCC OxPhos signature:

Derived from DLBCL samples robust (>>FC)high minimum expressionRepresented on U133A

108

General OxPhos signature

BIOCARTA MITOCHONDRIA PATHWAYKEGG OXIDATIVE PHOSPHORYLATION

REACTOME ELECTRON TRANSPORT CHAINREACTOME RESPIRATORY ELECTRON TRANSPORT

OXIDOREDUCTASE ACTIVITY ACTING ON NADH OR NADPHREACTOME TCA CYCLE AND RESPIRATORY ELECTRON TRANSPORT

232287

Expanded OxPhos signature in LUAD

DLBCL OxPhos signature

General OxPhos genes

Continuous GSEA to find gene-sets that correlate with ASSIGN score

Cor

rela

tion

with

A

SS

IGN

sco

re

ASSIGN score

DLBCL derived OxPhos signature

Expression Set

Microsoft Office Excel 97-2003 Worksheet

Gene SetAvg. Rank

Curtis discov.

DLBCL 2010

TCGA BRCA

TCGA HNSC

TCGA LUAD

TCGA LUSC

KEGG OXIDATIVE PHOSPHORYLATION 5.8 0.005 NA 0.013 0.007 0.002 0.001

REACTOME METABOLISM OF PROTEINS 8.2 0.001 0.122 0.028 0.003 0 0.001

REACTOME TRANSLATION 9.2 0.002 0.1 0.024 0.003 0.002 0.005

SRP DEPENDENT COTRANSLATIONAL PROTEIN TARGETING TO MEMBRANE

9.7 0.002 0.079 0.022 0.004 0.004 0.006

MITOCHONDRIAL PROTEIN IMPORT 14 0.005 0.073 0.028 0.003 0.005 0.008

REACTOME METABOLISM OF MRNA 16 0.001 0.3 0.019 0.002 0.001 0.003

REACTOME METABOLISM OF RNA 19.3 0 0.32 0.024 0.003 0.002 0.007

REACTOME CYCLIN E ASSOCIATED EVENTS DURING G1 S TRANSITION

22.7 0.005 0.057 0.024 0.033 0.007 0.01

REACTOME SCFSKP2 MEDIATED DEGRADATION OF P27 P21

23.2 0.006 0.071 0.025 0.028 0.009 0.01

KEGG RIBOSOME 23.3 0.005 0.073 0.023 0.005 0.011 0.008

ASSIGN score GSEA - up

FDR

Microsoft Office Excel 97-2003 Worksheet

ASSIGN score GSEA - down

Gene SetAvg. Rank

Curtis dis

DLBCL 2010

TCGA BRCA

TCGA HNSC

TCGA LUAD

TCGA LUSC

REACTOME NRAGE SIGNALS DEATH THROUGH JNK 30.5 0.03 0.091 0.174 0.022 0.019 0.048

KEGG PHOSPHATIDYLINOSITOL SIGNALING SYSTEM 30.5 0.067 0.086 0.172 0.04 0.025 0.058

KEGG DORSO VENTRAL AXIS FORMATION 38.3 0.031 0.09 0.181 0.033 0.028 0.035

PID RHOA REG PATHWAY 41 0.029 0.088 0.243 0.024 0.031 0.026PID FAK PATHWAY 41 0.045 0.091 0.163 0.046 0.059 0.046ST INTEGRIN SIGNALING PATHWAY 41.2 0.026 0.097 0.168 0.023 0.05 0.038BIOCARTA PAR1 PATHWAY 41.8 0.066 0.1 0.133 0.026 0.035 0.029REACTOME SIGNALING BY RHO GTPASES 42.2 0.035 0.156 0.186 0.026 0.031 0.041

REACTOME BMAL1 CLOCK NPAS2 ACTIVATES CIRCADIAN EXPRESSION 45.2 0.022 0.096 0.118 0.038 0.027 0.056

PID KITPATHWAY 47.5 0.028 0.195 0.18 0.033 0.072 0.059

FDR

Outline





There is a good concordance between all TCGA sets

Signatures’ overlap

Good concordance between OxPhos signature and TCGA sets

Lymphoma- derived

signature

There are additional genes in the solid tumors

35 nonDLBCL genes

2010

Outline





Looking for link to genomic instability - GISTIC

DEL 1

DEL 2

DEL N

AMP 1

AMP 2

AMP N

…

…

…

…

TCGA samples

CNV profile

ASSIGN score

pVal 1

pVal 2

pVal N

pVal 1

pVal 2

pVal N

…

…

KS - test

Sig

nific

ant

CN

Vs

(>0.

99 c

onfid

ence

)

The only significantly enriched deletion peak is enriched in the intermediate cases

DEL 3

Looking for link to genomic instability - MutSigCV

Gene 1

Gene 2

Gene 3

Gene X

Gene Y

Gene Z

… …

TCGA samples

Mut profile

ASSIGN score

pVal 1

pVal 2

pVal 3

pVal X

pVal Y

pVal Z

…

KS - test

Sig

nific

ant

Mut

atio

ns

(<0.

25 F

DR

)

There is only one enriched mutation across all TCGA datasets

TBL1XR1

Outline





Testing “OxPhosness” in cell-lines

CCLE - pVal Predictions

Lung 5.53E-06

Lung adeno 0.0073

Lung small cell 0.0154

Lung squamous 0.0553

Breast 5.08E-05

HNCC 5.67E-05

Melanoma 0.001

breast

HNCC

Lung

Lung small cell

Lung squamous

Breast

hncc

Lung

Lung adeno

Lung small cell

Lung squamous

Lung adeno

melanoma melanoma

Lun

g A

den

oca

rcinom

a

Cell-lin

es

Cellline CCLE_score CCLE_probNCI-H1355 0.175 1

HCC-2279 0.193 1

NCI-H1651 0.209 1

NCI-H1573 0.216 1

NCI-H2023 0.232 1

SK-LU-1 0.236 1

MOR/CPR 0.262 1

NCI-H1703 0.269 1

NCI-H2087 0.271 1

Calu-3 0 0

HCC-44 0 0

LXF-289 0 0

NCI-H1373 0 0

NCI-H2009 0 0

NCI-H2085 0 0

NCI-H322 0 0

NCI-H3255 0 0

NCI-H854 0 0

RERF-LC-Ad1 0 0

RERF-LC-Ad2 0 0

Bre

ast C

an

cer

Cell-lin

es

Cellline CCLE_score CCLE_probEVSA-T 0.985 1

SK-BR-3 0.885 1

YMB-1 0.879 1

AU565 0.878 1

ZR-75-1 0.837 1

HCC2218 0.83 1

EFM-192A 0.784 1

MDA-MB-453 0.777 1

MCF7 0.773 1

CAL-148 0.758 1

HDQ-P1 0.027 0.198901

HCC1143 0.001 0.013993

Hs 742.T 0.001 0.009495

HCC1806 0 0.003998

Hs 606.T 0 0.001999

Hs-578-T 0 0.0005

Hs 274.T 0 0.0005

Hs 739.T 0 0.0005

Hs 281.T 0 0

Hs 343.T 0 0

HN

CC

Cell-lin

es

Cellline CCLE_score CCLE_probFADU 0.937 1

SNU-1214 0.911 1

SNU-46 0.831 1

HSC-3 0.787 1

PE/CA-PJ34 (clone C12) 0.728 1

SNU-1041 0.715 1

BICR 18 0.708 1

YD-15 0.689 1

CAL-27 0.679 1

BICR 16 0.673 1

SNU-1066 0.165 0.746127

BICR 22 0 0.002499

BHY 0 0

BICR 56 0 0

SCC-9 0 0

SNU-1076 0 0

Future Work

• Functional validation of the top and bottom cell-lines

• Association with TCGA methylation data

Conclusion





Acknowledgements

Stefano Monti

Ying Shen

Liye Zhang

Francesca Mulas

Yuxiang Tan

Evan Johnson

Marc Lenburg

Luis Carvalho

Björn ChapuyNika DanialMargaret Shipp

Documents

Identifying OxPhos dependent tumors across different cancer types Daniel Gusenleitner Adviser: Stefano Monti