Multi-Analyte Profiling Reveals Relationships Among Circulating Biomarkers in Colorectal Cancer Daniel Delubac, Eric Ariazi, Jonathan Berliner, Adam Drake, John Dulin, Riley Ennis, Erik Gafni, Kate Niehaus, Gabriel Otte, Jennifer Pecson, Girish Putcha,

Corey Schaninger, Aarushi Sharma, Mike Singer, Abraham Tzou, Jill Waters, David Weinberg, Brandon White, Imran S. HaqueFreenome Inc, South San Francisco, CA

INTRODUCTION• Blood-based tests hold great promise as cancer diagnostics, but most current tests are restricted to the analysis of a

single class of molecules (e.g., circulating tumor DNA, circulating mRNA, circulating proteins)1-4

• The ability to analyze multiple analytes simultaneously from the same biological sample may increase the sensitivityand specificity of such tests by exploiting independent information among signals5

OBJECTIVE• Develop and implement an experimental and analytical system for the integrated analysis of multiple analytes from

a single blood sample

METHODSMulti-Analyte Approach (Figure 1)• De-identified blood samples were obtained from healthy individuals and individuals with non-advanced adenomas

(non-AAs), advanced adenomas (AAs), and stage I-IV colorectal cancer (CRC)

• After plasma separation, multiple analytes were assayed as follows: – Cell-free DNA (cfDNA) content was assessed by low-coverage whole-genome sequencing (lcWGS) andwhole-genome bisulfite sequencing (WGBS)

– Cell-free microRNA (cf-miRNA) was assessed by small-RNA sequencing – Levels of circulating proteins were measured by quantitative immunoassay

FIGURE 1. Overview of Multi-Analyte Approach for ‘Liquid Biopsy’

Plasma

Multi-Analyte Assays

Low-coverage Whole-Genome SequencingCNV callingTumor fraction (TF) estimation

Whole-Genome Bisulfite SequencingLINE-1 CpG methylation56 genes CpG methylation

Circulating Proteins Immunoassay

cf-miRNA Sequencing

Whole blood was collected in K3-EDTA tubes, and double-spun to isolate plasma. Plasma was split into aliquots for cfDNA lcWGS, WGBS, cf-miRNA sequencing, and quantitative immunoassays. CNV, copy number variation; LINE-1, long interspersed nuclear element 1.

Analysis• Sequenced cfDNA, WGBS, and cf-miRNA reads were analyzed as follows:

– cfDNA (lcWGS): Aligned to Genome Reference Consortium Human Build 38 (GRCh38) using Burrows-WheelerAligner (BWA) software. Fragments that aligned within annotated genomic regions were counted and normalizedfor depth of sequencing to produce a 30,000-dimensional vector per sample• Samples with high TF were identified via manual inspection of large-scale CNV

– WGBS: Aligned to GRCh38 using BWA-meth software. Percentage of methylation was calculated per sampleacross LINE-1 CpGs and CpG sites in targeted genes (56 genes)

– cf-miRNA: Aligned to miRbase 21 using Bowtie 2. Fragments that aligned to annotated miRNA genomic regionswere counted and normalized for depth of sequencing to produce a 1700-dimensional vector per sample

• Absolute protein abundances were determined using standard curves

• Principal component analysis (PCA) was performed per analyte

RESULTScfDNA: lcWGS• lcWGS of plasma cfDNA was able to identify CRC samples with high TF on the basis of large-scale CNV across the

genome (Figure 2)

• High TFs, while more frequent in late-stage CRC samples, were observed in some stage I and II samples (Figure 2).High TFs were not observed in samples from healthy individuals or those with non-AAs or AAs

FIGURE 2. Distribution of High TF Samples, as Inferred by CNV, Across Clinical Stage

Stage I, T2N0M0, Grade 2 Stage IV, T3N1M1, Grade 1

Stage II, T3N0M0, Grade 1 Stage IV, T3N1M1, Grade 2

Stage II, T3N0M0, Grade 2 Stage IV, T3N1M1, Grade 2

Stage III, T3N2M0, Grade 2

Healthy Non-AA AA Stage I CRC Stage II CRC Stage III CRC Stage IV CRC

0N with high TF

Total N

0 0 1 2 1 4

26 13 10 3 7 4 5

Stage IV, T3NxM1

A.

B.

40000

30000

20000

10000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr10

chr11

chr12

chr13

chr14

chr15

chr16

chr17

chr18

chr19

chr2

0ch

r21

chr2

2ch

rX

chrY

50000

3000040000

2000010000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr1

0

chr1

1

chr1

2

chr1

3ch

r14

chr1

5ch

r16

chr1

7ch

r18

chr1

9ch

r20

chr2

1ch

r22

chrX

chrY

3000040000

2000010000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr1

0

chr1

1

chr1

2

chr1

3ch

r14

chr1

5ch

r16

chr1

7ch

r18

chr1

9ch

r20

chr2

1ch

r22

chrX

chrY

30000

2000025000

100005000

15000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr1

0

chr1

1

chr1

2

chr1

3ch

r14

chr1

5ch

r16

chr1

7ch

r18

chr1

9ch

r20

chr2

1ch

r22

chrX

chrY

3500030000

2000025000

100005000

15000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr10

chr11

chr12

chr13

chr14

chr15

chr16

chr17

chr18

chr19

chr2

0ch

r21

chr2

2ch

rX

chrY

50000

3000040000

2000010000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr1

0

chr1

1

chr1

2

chr1

3ch

r14

chr1

5ch

r16

chr1

7ch

r18

chr1

9ch

r20

chr2

1ch

r22

chrX

chrY

300002500020000

1000015000

50000

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr1

0

chr1

1

chr1

2

chr1

3ch

r14

chr1

5ch

r16

chr1

7ch

r18

chr1

9ch

r20

chr2

1ch

r22

chrX

chrY

2000025000

100005000

15000

0

chr1

chr2

chr3

chr4

chr5

chr6

chr7

chr8

chr9

chr1

0

chr1

1

chr1

2

chr1

3ch

r14

chr1

5ch

r16

chr1

7ch

r18

chr1

9ch

r20

chr2

1ch

r22

chrX

chrY

(A) CNV plots for individuals with high TF based on cfDNA-seq data. (B) Distribution of high TF cfDNA samples across clinical stage. High TF samples do not necessarily correspond to samplesclinically classified as late stage. Chr, chromosome.

Methylation: WGBS• Genome-wide hypomethylation at LINE-1 CpG loci was only observed in individuals with CRC (Figure 3)

• Hypomethylation was not observed in samples from healthy individuals or those with non-AAs or AAs

FIGURE 3. CpG Methylation Analysis at LINE-1 Sites

Normal Non-AA AA CRC (stages I-IV)50

60

70

80

90

%5m

C in

LIN

E-1 C

pG s

ites

1-way ANOVA P=0.0088Post-test:Normal vs CRC adj P=0.004

Significance was assessed by 1-way analysis of variance (ANOVA) followed by Sidak’s multiple comparison test. Only significant adjusted P values are shown. CpG hypomethylation of LINE-1 was only observed in CRC cases. 5mC, 5-methylcytosine.

miRNA• miRNAs suggested as potential CRC biomarkers in the literature tended to be present in higher abundance relative

to other miRNAs (Figure 4)

FIGURE 4. cf-miRNA Sequencing Analysis

1

10

100

1,000

10,000

100,000

1,000,000

10,000,000

1 301 601 901 1201 1501 1801

Tota

l num

ber o

f rea

ds

miRNA expression rank

Aggregate read counts for detected miRNAs

All others Literature support

Shown are the number of reads mapping to each miRNA after pooling reads from all samples. miRNAs indicated in blue have been suggested as potential CRC biomarkers in the literature. Adapter-trimmed reads were mapped to mature human microRNA sequences (miRBase 21) using bowtie2. More than 1800 miRNAs were detected in plasma samples with at least 1 read, while 375 miRNAs were present at higher abundance (detected with an average of ≥10 reads per sample).

Proteins• In CRC samples, circulating levels of alpha-1-antichymotrypsin (AACT), C-reactive protein (CRP), and serum amyloid

A1 (SAA1) proteins were elevated, while urokinase-type plasminogen activator (u-PA) levels were lower compared withhealthy controls (Figure 5)

• In AA samples, circulating levels of fibroblast activation protein (FAP) and Flt3 receptor-interacting lectin precursor(FRIL) proteins were elevated, while CRP levels were lower compared with CRC samples (Figure 5)

FIGURE 5. Circulating Protein Biomarker Distribution

2

0

–2

–4

Classification

Normal/Healthy

Non-AA

AA

CRC–6

A1G

1

C3a

des

Arg

CA7

2-4

Cat

D

CEA

AAC

T

CRP

CTA

CK

CYF

RA21

-1

FAP

FRIL

MM

P9

SAA1

u-PA

Prot

ein

leve

l rel

ativ

eto

nor

mal

(z-s

core

s)

Protein

Normal

Non-AA AACRC

0

20

40

60

80

100

AACT

ng/m

l

Normal

Non-AA AACRC

0

50

100

150

200

CRP

pg/m

l

Normal

Non-AA AACRC

0

100

200

300

FAP

pg/m

l

Normal

Non-AA AACRC

0

100

200

300

pg/m

l

FRIL

Normal

Non-AA AACRC

0

5

10

15

20

SAA1

ng/m

l

Normal

Non-AA AACRC

0

10

20

30

40

u-PA

pg/m

l

1-way ANOVA P=0.024

Normal vs. CRC adj. P=0.034

1-way ANOVA P<0.0001

Normal vs. CRC adj. P<0.0001AA vs. CRC adj. P=0.006

1-way ANOVA P=0.014

AA vs. CRC adj. P=0.020

1-way ANOVA P=0.049

AA vs. CRC adj. P=0.030

1-way ANOVA P=0.0035

Normal vs. CRC adj. P=0.001

1-way ANOVA P=0.050

AA vs. CRC adj. P=0.029Post-test: Post-tests: Post-test: Post-test: Post-test: Post-test:

A.

B.

(A) Levels of all circulating proteins assayed. (B) Proteins which show significantly different levels across clinicopathologic stages according to 1-way ANOVA followed by Sidak’s multiple comparison test. Only significant adjusted P values are shown. Proteins measured using SIMOA (Quanterix): ATP-binding cassette transporter A1/G1 (A1G1), acylation stimulating protein (C3a des Arg), cancer antigen 72-4 (CA72-4), carcinoembryonic antigen (CEA), cytokeratin fragment 21-1 (CYFRA21-1), FRIL u-PA. Proteins measured by ELISA (Abcam): AACT, cathepsin D (CATD), CRP, cutaneous T-cell-attracting chemokine (CTACK), FAP, matrix metalloproteinase-9 (MMP9), SAA1.

Multi-analyte• Strikingly, aberrant profiles across analytes were indicative of high TF (as estimated from cfDNA CNV), rather than

cancer stage (Figure 6)

• All individual analyte profiles were discordant between CRC samples and normal controls

FIGURE 6. PCA of cfDNA, CpG Methylation, cf-miRNA and Protein

B. miRNA

PCA comp. 1, 47.6% var0 2 4

3.0

PCA

com

p. 2

, 15.

4% v

ar 2.5

2.0

1.5

1.0

0.5

0.0

–0.5

–1.0

Methylation

PCA comp. 1, 20.1% var0 1 2 3

1.25

PCA

com

p. 2

, 13.

1% v

ar 1.00

0.75

0.50

0.25

0.00

–0.25

–0.50

–0.75

PCA

com

p. 2

, 14.

4% v

ar

cfDNA

PCA comp. 1, 30.4% var0 1 2 3 4

2.5

2.0

1.5

1.0

0.5

0.0

–0.5

–1.0

–1.5

Proteins

PCA comp. 1, 31.2% var–5.0 –2.5 0.0 2.5 5.0

PCA

com

p. 2

, 16.

9% v

ar

6

4

2

0

–2

–4

Normal/HealthyNon-AAAACRC

(A) PCA as a function of TF. (B) PCA as a function of tissue class. High TF samples have consistently aberrant behavior across all 4 analytes investigated.

CONCLUSIONS• These data suggest that TF is correlated with cancer stage, but has a large potential range, even in early stage

samples

• We found that aberrant profiles among cfDNA CpG methylation, cf-miRNA, and circulating protein levels were morestrongly associated with high TF than with late stage. This may explain the variability reported in cancer detectionrates for other screening tests6-8

• These findings suggest that some positive “early stage” detection results may in fact be “high TF” detection results7,8

• The presence of low TF CRC outlier data in non-cfDNA analytes, including miRNA and protein, suggests that assayingmultiple analytes from a single sample may enable the development of classifiers that are reliable at low TF

ACKNOWLEDGMENTSMedical writing and editorial support, which was in accordance with Good Publication Practice (GPP3) guidelines, was provided by Amy Agbonbhase, PhD of The Lockwood Group (Stamford, CT), and funded by Freenome Inc.

REFERENCES1. Bettegowda C, et al. Sci Transl Med. 2014;6(224):224ra24.2. Phallen J, et al. Sci Transl Med. 2017 Aug 16;9(403).3. Lim SY, et al. Mol Cancer. 2018;17:8.4. Best MG, et al. Cancer Cell. 2015;28(5):666-676.5. Haque IS, et al. AACR 2018;Poster 2225.6. Merker JD, et al. J Clin Oncol. 2018 Mar 5:JCO2017768671 [Epub ahead of print].7. Cohen JD, et al. Science. 2018;359(6378):926-930.8. Razavi P, et al. J Clin Oncol. 2017;35(suppl):abstr LBA11516.

2227

Presented at the American Association for Cancer Research (AACR) Annual Meeting, April 14-18, 2018; Chicago, Illinois, USA.

miRNAMethylationA. cfDNA Proteins

0 2 4PCA comp. 1, 47.6% var

3.0

PCA

com

p. 2

, 15.

4% v

ar 2.5

2.0

1.5

1.0

0.5

0.0

–0.5

–1.0

PCA comp. 1, 20.1% var0 1 2 3

1.25

PCA

com

p. 2

, 13.

1% v

ar 1.00

0.75

0.50

0.25

0.00

–0.25

–0.50

–0.75

2.5

PCA

com

p. 2

, 14.

4% v

ar

1 2 3PCA comp. 1, 30.4% var

0 4

2.0

1.5

1.0

0.5

0.0

–0.5

–1.0

–1.5

PCA comp. 1, 31.2% var

Low TFHigh TF

–5.0 –2.5 0.0 2.5 5.0

PCA

com

p. 2

, 16.

9% v

ar

6

4

2

0

–2

–4