the whole transcriptional landscape of circulating tumor ... · -4-3-2-1 0 1 2 3 4 log fold change...

-4

-3

-2

-1

0

1

2

3

4

log fo

ld ch

ange

The whole transcriptional landscape of circulating tumor cells compared to metastases in stage IV breast cancerAlexander Ring, Tania B. Porras, Daniel Campo, Pushpinder K. Bains, Victoria Forte, Debu Tripathy, Janice Lu, Gabriel Zada, Naveed Wagle, Julie E. Lang

Background

Methods

Conclusion

−60

−40

−20

0

20

40

−50 0 50 100PC1: 46% variance

PC2:

7%

var

ianc

e

groupCTC

Met

PB

Figure 1: Principal component (PC) analysis of CTCs, Metsand PB. The results show separation of the majority of CTCsversus mets and PB in PC1, and separation of CTCs and metsfrom PB in PC2.

Table 1: SNVs common to all CTC/met pairs in ESR1and ERBB2: Using IGV genome viewer, six SNVs inESR1 (n=4 patients) were identified that are sharedbetween CTCs and distant metastases. Five are listed indbSNP. From those listed, rs3798577 hasbeen associated with increased and decreased breastcancer risk in White and Asian women, respectively.rs2228480 is linked to decreased BC risk in whitewomen. rs2747648 occurs in the miR-453 bindingsite resulting in higher ESR1 protein expression. Knownand novel SNVs were identified with RNA Seq.

Patient ID SNVs in ESR1 Rs number in dbSNP

78536_CTCT>C chr6:152,097,179 not found

78536_MET79412_CTC

T>C chr6:152,099,995 dbSNP: rs379857779412_MET

101738_CTC_FOLLOWUP G>T chr6:152,101,052C>T chr6:152,101,200

dbSNP: rs72993667dbSNP: rs2747648101738_MET

36541_CTC C>A chr6:152,098,960C>T chr6:152,101,200

dbSNP: rs2228480dbSNP: rs274764836541_MET_BREAST

PIK3CARPTORNF1AKT1AKT3FBXW7

EGFRGRB7MAP3K1MAP2K1BRAFKRAS

CDKN2ACCND1CCND2CCND3CCNE1CDK4CDK6MYBL2PTTG1

CXXC5SFRP1CXCR4FGF1WNT1NOTCH1NOTCH4ALDH1A1TBX3MET

TCGB1IL4IL6IL12IL15IL23AGATA3CXCL9CSCL13CD3D

BRCA1BRCA2ATMPALB2BARD1

ESR1ESR2PGRHER2HER3

4

2

0

-2

-4

3

1

-3

-1fold

cha

nge

p=0.94 p=0.41

CTCs metsp=0.80p=0.29 p=0.33p=0.80 p=0.66

EGFR

GRB7

MAP3K1

MAP2K1

BRAF

KRAS

-20

-15

-10

-5

0

5

10

15

20

25

19065

28089

36541

68185

78536

78908

79388

79412

79555

79644

80541

81103

101738

112165

112370

113059

113166

113457

113488

C M

C M C M

C M

C MC M

fold

cha

nge

Figure 5: Analysis of potentially clinically actionable genes in breastcancer. 5A: Comparison of overall gene expression in different druggablepathways between CTCs and mets. We queried for n=66 potentially clinicallyactionable genes and on paired T tests for n=9 (7 shown here) pathways therewas no significant difference in mean gene expression between CTCs andmetastases. 5B: Expression of clinically actionable target genes in CTCs andmets per patient for the EGFR/RAF/MEK pathway (C – CTCs, M – mets).

(E) Clinically actionable genes/ signaling pathways

carboplatin/paclitaxel

6/2014

doxorubicin10/2016

2/7/2017

disease progression

CTCs 1st collection12/13/2016

fulvestrant/palbociclib

8/2015

Met collection10/24/2016

gemcitabine1/2017

gemcitabineexemastane/everolimus

2/2017

Sample collectionCTCs

2nd collection3/16/2017

3/23/20178/6/2015

Figure 6: Intra-patient (n=1) two time-point comparison: 6A: Clinical data (includingtreatment and imaging studies) and sample collection are shown. 6B: differentiallyexpressed breast cancer genes (KEGG pathway) in met (ascites), 1st CTC and 2nd CTCsample were analyzed using IPA pathway analysis tool, demonstrating differential geneexpression and pathway activation.

(F) Intra-patient analysis

Met (ascites) vs. PB 1st time point CTCs vs. PB 2nd time point CTCs vs. PB

Results (A) Whole transcriptome RNA Seq gene expression - group analysis

CTCs from 21 MBC patients were enumerated and captured from 10mL peripheral blood(PB) via the ANGLE Parsortix system. RNA Seq was performed on fresh metastatic tumorbiopsies (mets), CTCs and peripheral blood (PB) from all patients. Biopsy sites included:skin (n=1), lung (n=1), pleural effusion (n=5), pericardial effusion (n=1), breast (n=3),lymph node (n=2), brain (n=4), liver (n=1), ascites (n=3), cerebrospinal fluid (n=2), bonetissue (n=1). 19/21 patients were included in subsequent data analysis. (A) Groupcomparison of biologically relevant gene expression patterns in CTCs, mets and PB wasperformed. (B) Differential expression of genes of interest (oncogenes, breastcancer related genes, mesenchymal and cancer stem cell (CSC) genes) betweenCTCs, mets and PB was investigated. (C) Survival analysis based on gene expressionin CTCs and mets compared to PB was performed using data from The Cancer GenomeAtlas (TCGA). (D) Single nucleotide variants (SNV) analysis using IGV was performedin corresponding CTCs/mets pairs. (E) Clinically actionable gene (n=66) expressionand molecular signaling pathways (n=7) for each patient were explored. (F) Intra-patient serial time points were analyzed, and detailed clinical-pathological andtreatment data was evaluated.

(D) Single nucleotide variants (SNV) analysis

Metastasis is responsible for the vast majority of breast cancer related deaths. Metastaticbreast cancer (MBC) is inherently different than primary breast cancer (BC), evolvingunder selection pressure at different organ sites or during systemic therapy. The currentASCO guidelines call for biopsy of a metastatic site to guide decision making for systemictherapy. Yet, biopsies of macro metastasis are oftentimes not feasible in the clinicalsetting. Circulating tumor cells (CTCs) have been shown to be prognostic in MBC, buttheir use as clinical biomarker beyond CTC enumeration has been limited. A betterunderstanding of CTC-biology compared to metastasis may shed light on treatmentopportunities and help advance the application of CTCs as liquid biopsies in clinicalpractice. The ANGLE Parsortix system is a microfluidics device that separates CTCsbased on size and deformability, without the need for cell-surface marker selection. Ourlab has previously demonstrated the feasibility of gene expression profiling of rare CTCs.Here, we evaluated whether whole transcriptome sequencing (RNA Seq) geneexpression profiling of ANGLE Parsortix isolated CTCs may serve as a surrogatefor biopsies of macro metastases.

CTC

s vs

PB

Met

s vs

PB

CTC

s+M

ets

vs P

B

CTC

s vs

Met

s

AKT1CCND1CCNE1CCNE2

PLK1BIRC5

CENPFCENPAAURKAAURKB

CDC25ACDC25C

JUNMYCNMTORABL1ABL2BCR

EGFRFGFR2FGFR1MYBL2

BCL6ELK4

HMGA1PIK3CA

-1 0 1 2 3 4

CTC

s vs

PB

Met

s vs

PB

CTC

s+M

ets

vs P

B

CTC

s vs

Met

s

SHHGLI1GLI2GLI3SMO

RARGSOX3SOX7SOX8

SOX11SOX12SOX13SOX18SOX30

WNT10AWNT3AWNT5A

WNT10BCD133OCT4

KITABCG2

MDR1NESID2

ALDH1A2ALDH1A3

CD44

-1 0 1 2 3 4

CTC

s vs

PB

Met

s vs

PB

CTC

s+M

ets

vs P

B

CTC

s vs

Met

s

SCGB2A2

KRT7

KRT6B

KRT6C

KRT8

KRT3

KRT4

KRT9

KRT13

KRT16

ESR1

PGR

CDH1

KRT14

KRT18

KRT19

-3 -2 -1 0 1 2 3 4

CTC

s vs

PB

Met

s vs

PB

CTC

s+M

ets

vs P

B

CTC

s vs

Met

s

SNAI1TWIST1

FN1CDH2

-2 -1 0 1 2 3 4Figure 3: Expression of genes of interestin CTCs or mets compared to PB: CTCsas a group showed much stronger geneexpression of oncogenes, stem cell genes,keratins and mesenchymal markers than didmets from the same patients.

(B) Differential expression of genes of interest

Oncogenes Mesenchymal genesCSC genes Breast epithelial genes

0 20 40 600%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Months

Perc

ent s

urvi

val

50_gene_coding_CTCsANDMetsGENESinBOTH_vs_PB

Signature expressed

Signature not expressedp = 0.0099

CTCs/Mets common vs. PB

0 20 40 600%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Months

Per

cent

sur

viva

l

50_gene_coding_Cancer_vs_PB

Alteration

No alteration p = 0.0099

CTCs or Mets combined vs. PB

0 20 40 600%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Months

Per

cent

sur

viva

l

50_gene_coding_CTCs_vs_PB

Alteration

No alteration p = 0.8749

0 20 40 600%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Months

Per

cent

sur

viva

l

50_gene_coding_Mets_vs_PB

Alteration

No alteration p = 0.0476

Figure 4: TCGA BC (n=817) overall survival (OS) based on 50-geneexpression signatures. The top 50 highest expressed genes in fourcomparison conditions are shown. The 50-gene signature expressedCTCs and/or mets was superior to all other comparisons in predictingpoor OS (p=0.01).

CTCs vs. PB Mets vs. PB

Signature expressedSignature not expressed

(C) Survival analysis

p=0.84

p=0.87 p=0.048

p=0.01

*

**

Log

(RPK

M+1

)

fold

cha

nge

Figure 2: Differential gene expression analysis of CTCs, mets and PB.Two dimensional hierarchical clustering of all samples based on a 253gene signaturRe that distinguishes CTCs (blue), mets (grey) and PB (red)(FDR adjusted p<0.05).

We present the whole transcriptomic landscape of CTCs with comparison to metastasesand peripheral blood all acquired prior to treatment of Stage IV breast cancer. Multiplegenes, including oncogenes, breast epithelial, mesenchymal genes and CSC genes, werefound with higher expression in CTCs versus metastases. When focusing on 66 knownpotentially clinically actionable genes in breast cancer, represented by 7 molecularsignaling pathways, CTCs did not show significantly different patterns of expressionversus mets compared to PB. Longitudinal analysis of 4 patients over time who had serialCTC assessments showed changing biological characteristics of CTCs isolated atdifferent time points during treatment and disease progression. RNA Seq of CTCs may beutilized to identify molecular alterations in MBC patients that are potentially clinicallyactionable.

Metastatic site profiled: acsites, ER/PR+,HER2-. Other sites ofmetastatic disease: liver, lung, pleural effusion and bone.Molecular profiling with RNA Seq was done to evaluate for potentialtreatment targets in CTCs as a liquid biopsy as well as metastaticdisease.