Immunotherapy and Cancer: Harnessing the Power of Diagnostic Assays - 18

Immunotherapy and Cancer | Harnessing the Power of Diagnostic Assays

POSTERS

Click for Full Screen
Immunotherapy biomarker assessment in FFPE sampleƐ from solid tumors using IHC, gene expression profiling and mutation burden assessment
Figure 1. Expression of PD-L1 and infiltrating immune cells.

Figure 4. Heatmap of 25-gene IFNG signature in urothelial cancer samples.

Table2. Evaluation of tumor mutation burden and mutational spectrum using CCP409 and OCAv3.
OCAv3

CCP409

Oncomine Variants

TMB Total
(No.Var No. SBS Missense Synonymous Indel
/Mbp) Var

Checkpoint inhibitors have been approved for treatment of solid tumors and hematological malignancies. PD-L1
immunohistochemistry (IHC) assays are currently being used as companion or complementary diagnostics to select patients for
treatment. However, patient responses are variable and there is a need to identify additional predictive biomarkers beyond PDL1 levels as measured by IHC.

Sample

Cancer Type

Tumor
Content
(%)

The assessment of tumor infiltrating lymphocytes (TILs) in addition to PD-L1 staining using IHC has also been proposed for
patient stratification (Ref.1). However, IHC scoring and classification can be complicated by the definition of positivity and intratumoral heterogeneity.

RCC-102

RCC

None

1.6

RCC-107

RCC

BRCA2 K3326X 65%

1.6

RCC101

RCC

None

2.4

RCC-103

RCC

None

2.4

RCC-099

RCC

None

RCC-104

RCC

SETD2 G988X 29%

4.8

RCC-100

RCC

SETD2 S203fs 18%

5.6

MSH6 K1358fs 43%; EGFR L858R 6%

1.6

KRAS G12F 19%; ATM G1672X 29%;
NOTCH2 Q508X 4%

5.6

A more comprehensive assessment of the tumor and its microenvironment using WES for tumor mutation burden analysis
(TMB) and/or WTS for gene expression profiling (GExP) has also been shown to be informative for predicting patient
responses. However, comprehensive analysis using these methods can be challenging with poor quality and limited amounts of
FFPE DNA and RNA from clinical samples. Here we demonstrate the feasibility of performing TMB analysis and GExP in clinical
FFPE samples using targeted NGS panels. The results from these studies highlight the value of using IHC in combination with
targeted NGS methods to comprehensively assess the tumor and its microenvironment.

Materials and Methods

NSCLC
(Papillary adenocarcinoma)
NSCLC
(Adenocarcinoma)

NSCLC-065

Immunohistochemistry
5-micron sections were stained using mAbs, α-CD3 (clone LN10, Leica), α-CD163 (clone 10D6, Leica), α-CD8 (clone 4B11,
Leica) and α-PD-L1 (clone E1L3N, Cell Signaling) on the Leica BOND III platform.

NSCLC-077

Detection of the target protein was visualized using the Bond Polymer Refine Detection (Leica).
IHC images were captured via an Aperio AT2 Slide Scanner (Leica).

RNA QC Analysis

Targeted RNA Seq

FFPE RNA samples with a delta-Ct (CtSample- CtControl) <9.7 were analyzed by the OIRRA .

128 FFPE specimens were analyzed, including renal cell carcinoma (RCC), colorectal cancer (CRC), non-small cell lung
cancer (NSCLC), urothelial cancer (UC) and other solid tumor types.

RNA samples with a lower delta-Ct value generated better sequencing QC metrics, i.e., more valid reads (Figure 2A)
and longer read length (Figure 2B), demonstrating the utility of our custom in-house RNA QC assay.

The quality of the RNA samples was assessed using a custom, multiplexed RT-qPCR assay. The expression level (Ct) of 5
housekeeping genes in each FFPE sample was compared to the levels in an RNA control (the Human Lung Total RNA,
Thermo Fisher).

Duplicate analysis of each sample allowed for the elimination of deamination errors in FFPE (C>T), especially in two poor-quality
samples (NSCLC-066 and NSCLC-077). It also allowed for reliable detection of low-level mutations (3-10%) (Figure 5 B and C).

The Oncomine Immune Response Research Assay (OIRRA) from Thermo Fisher was used for GExP of 391 genes involved in
tumor-immune cell interactions.

The TMB range was lowest in RCC samples (1.6-5.6/Mb). TMB in the NSCLC and CRC samples ranged from 1.6-16.8 to 4-28/Mb,
respectively (Table 2). These results are consistent with what has been previously reported for these tumor types (Ref.7).

10 ng RNA input per sample was used for library preparation and 32 barcoded libraries were sequenced on a 540 chip
using the Ion Chef-S5 XL system.

Delta-Ct

The Ion AmpliSeq™ Comprehensive Cancer Panel (CCP409) from Thermo Fisher was used for TMB analysis. CCP409
provides full coding exon coverage of 409 cancer related genes using 4 multiplexed primer pools.

Immune response signatures
Three different IFNG immune response signatures were assessed using OIRRA data. The signatures ranged in
content from 4 to 25 genes (Table 1). Higher expression of each of these gene signatures was predictive of better
response to the relevant immunotherapy (Refs.4-6). Mean expression (log2(RPM+1)) of the component genes in
respective IFNG signatures was used to score the IFNG signatures.

The Oncomine Comprehensive Assay v3 (OCAv3) from Thermo Fisher was used to detect clinically relevant SBS, INDEL,
CNV, and fusions.
40 ng DNA input per sample was used for CCP409; 20 ng DNA and 20 ng RNA per sample was used for OCAv3.
8 barcoded libraries were sequenced on a 540 chip using Ion Chef-S5 XL system.

34 NSCLC and 48 UC samples were analyzed.

Oncomine® variant analysis: alignment was performed by the Ion Torrent 5.4 software, variant calling and annotation by
the Ion Reporter™ Software (IR 5.4). The driver gain-of-function/loss-of-function variants were identified by the
Oncomine® Variant Annotator plugin (v2.0), with the Oncomine® Knowledgebase.

The mean expression scores for the three IFNG signatures correlated to each other in urothelial cancer (Figure 3,
comparing the trend of green, red and blue lines) and NSCLC (data not shown).

*The genes corresponding to the pink-highlighted cells are included in the respective immune response signatures.

CRC-061

CRC-113

CRC-058

CRC-124

NSCLC-070

NSCLC-066

NSCLC-077

NSCLC-065

RCC-100

RCC-104

RCC-099

RCC-103

RCC-101

CD274

C>G

IFNG 25 genes

No Variants

CD8A

C>T

CRC-113

CRC

PIK3CA E542K 22%; BRAF V600E 22%;

20.8

CRC-2-4-061

CRC

BRAF V600E 23%; NF1 Q1086X 21%,
H1494fs 26%; EGFR G465R 4%; PIK3CB
E1051K 9%

The study also suggested that tumor mutation burden in clinical FFPE tumor samples could be analyzed without the inclusion of the matched
normal.
The use of GExP, TMB analysis and PD-L1/TIL IHC allows for a more comprehensive assessment of the tumor and its microenvironment. A multiassay approach may improve patient selection and stratification criteria and lead to the identification of more cancer patients who may benefit
from immunotherapy.

Future Directions

80
60
40
20
0

Comparison of GExP, TMB, mutation profiles and MSI status.
Optimization of TMB assay workflow and tumor-specific cut-offs.
Evaluation of SETD2 mutations, TMB and therapy response in a RCC patient cohort.
Evaluation of NF1 mutations, TMB and therapy response in a CRC patient cohort.

References

>=50%
20-50%
10-20%
5-10%
3-5%

3. Khagi Y, et al., (2017) Hypermutated Circulating Tumor DNA: Correlation with Response to Checkpoint Inhibitor-Based Immunotherapy. Clin. Cancer Res., 23(19):5729
4. Sharma P, et al., (2017) Nivolumab in metastatic urothelial carcinoma after platinum therapy (CheckMate 275): a multicentre, single-arm, phase 2 trial. Lancet Oncol., 18(3):312
5. Streicher K, et al., (2017) Gene expression analysis of tumor biopsies from a trial of durvalumab to identify subsets of NSCLC with shared immune pathways. 2017 ASCO Annual meeting, Chicago, IL
6. Ayers M, et al., IFN-gamma gene signature biomarkers of tumor response to pd-1 antagonists. WIPO Patent WO2015094992 A1
7. Ludmil B, et al., (2013) Signatures of mutational processes in human cancer. Nature, 500: 415

25-gene
4-gene
10-gene
housekeeping genes

8. Hirotsu Y, et al., (2015) Multigene panel analysis identified germline mutations of DNA repair genes in breast and ovarian cancer. Mol Genet Genomic Med.,3(5):459
9. The Cancer Genome Atlas Network (2015) Genomic Classification of Cutaneous Melanoma. Cell, 161:1681

CRC-061

CRC-113

CRC-058

CRC-124

RCC-100

NSCLC-070

NSCLC-066

NSCLC-077

Urothelial Cancer Samples (N=48)

NSCLC-065

RCC-104

RCC-099

0
RCC-103

2. Chalmers ZR, et al., (2017) Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med., 9(1):34

8
4

1. Teng MW, et al., (2015) Classifying Cancers Based on T-cell Infiltration and PD-L1. Cancer Res., 75(11):2139

RCC-101

Mean expression level
(Log2(RPM+1))

CD4

T>A

9.6

The criteria by which TMB is considered as high or low may vary in different tumor types.

Mutation Frequency

RCC-107

CDK12 R1356X 12%; TP53 R306X 10%;
TP53 R248Q 12%; SMAD4 R361H 11%;
PIK3CA R88Q 12%; BRAF V600E 9%

Assessment of the mutation profile in addition to TMB can identify specific variants that may aid in the design of an individualized therapeutic
strategy.

C>A

RCC-102

Figure 3. Comparison of immune response signature scores

Treg-mean

Number of Variants

10-gene

TBX21

INDEL
T>G
T>C

RCC-107

CCR5
IFNG
CXCL9

4-gene

Immunohistochemistry
Expression of PD-L1 and presence of total T cells (CD3), M2 macrophages (CD163), and cytotoxic T cells (CD8) in tumor
and its microenvironment varied dramatically among the FFPE tumor samples examined (Figure 1).
RCC-099 and CRC-061 showed prominent PD-L1 staining on tumor cells, whereas cases RCC-107 and CRC-124 showed only
immune cell staining.
All 4 cases showed a range of tumor heterogeneity, with "hot spots" of immune infiltrate and regions of necrosis.

PIK3CA K111E 53%

TMB generated from targeted NGS panels such as CCP409 and OCAv3 are comparable to those obtained via whole exome sequencing.

25-gene

Results

CRC

Gene expression profiling with the OIRRA allows for flexibility in classification of tumor samples using different immune response and immune cell
subset signatures.

40
30
20
10
0
10
20
30
40
50
60
70

TBX21
PTPRC
PDCD1
IL2RG
IL2RB
HLA-DRA
GZMB
CXCR6
CXCL13
CD8A
CD74
CD4
CD3D
CD274
CD27
CD2
CCL5
STAT1
PRF1
LAG3
IDO1
GZMA
CXCL11
CXCL10

Signatures

CRC

The high TMB sample CRC-061 contains three mutations in NF1, BRAF and PIK3CB, which have been shown to be associated
with high mutation burden (Ref.2). Moreover, loss-of-function NF1 mutations and high TMB has been observed in melanoma
(Ref.9).

Figure 5. Spectrum of tumor mutation burden and its association with expression of immune response signature genes.

Table1. Gene content of 3 published immune response signatures

NSCLC (Large cell)

CRC-124

NSCLC-065 has a low TMB while harboring a MSH6 frameshift mutation (Table 2). This germline mutation was predicted to be
nonpathogenic which is consistent with the low TMB score (Ref.8).

In CRC samples, there was a positive correlation between the 25-gene IFNG GEx signature score, GEx signatures for T cell
subsets (Th1 and Treg), PD-L1 expression and TMB (Table 2, bottom four samples and Figure 5A, right four samples).

UC samples can be further clustered into subgroups based on low to high expression of the 25 gene IFNG signature
(Figure 4). Similar results were also observed in NSCLC (data not shown).

9.6
16.8

1
1

The custom multiplexed RNA-QC assay is valuable in predicting the quality of RNA samples to be analyzed by RNA-Seq assay.

Constant expression of housekeeping genes (Figure 3, purple line) suggested an accurate measurement of IFNG
signatures.

TMB analysis: each sample was analyzed in duplicates. Synonymous and non-synonymous SBS and small INDEL in coding
regions were included in the total count. Variants were then removed from the total count based on the following filters:
1) those not reproducibly detected in the sample replicates 2) those detected in FFPE GM12878 3) those found in dbSNP
(germline mutations) and the COSMIC database 4) those with mutant allele coverage below 20 read depth and 5) those
that failed in IGV inspection. Removal of variants that were not found in sample replicates reduced artifacts caused by FFPE
deamination. Removal of the COSMIC variants reduced the bias toward the cancer-related genes by the CCP409 panel
(Refs 2 and 3). TMB was calculated as the rate of the true variants per million bases of CCP409 ROI. The ROI is calculated
at 1.25 million bases with an R-script to eliminate the overlapped amplicon design among four (4) primer pools.

NOTCH1 S2467X 14%
TP53 R158P 70%; STK11 K81X 70%

2
1

This study demonstrated the feasibility of employing targeted RNA-Seq and targeted DNA-Seq in the analysis of immune response signatures and
tumor mutation burden in clinical FFPE tumor samples.

NSCLC-070 has a high TMB (Table 2). The high proportion of C>A change is consistent with a mutation signature caused by
tobacco (Figure 4A) (Ref.7).

Delta-Ct

1
1

Conclusions

3 RCC samples harbored mutations in genes involved in DNA repair (Table 2). Samples with somatic SETD2 mutations had higher
TMB scores than the sample with a BRCA2 germline mutation.

Data analysis was performed using Torrent Suite 5.4, EXCEL, dCHIP and pheatmap in R package.
7 RCC, 4 CRC and 4 NSCLC were analyzed with FFPE processed GM12878 as the negative control.

NSCLC
(Squamous)

2
1

CRC-058

The majority of mutations (88%) detected were at >10% frequency (Figure 5B), suggesting that a 10% cut-off may be sufficient
for TMB analysis.

Targeted DNA Seq

NSCLC-066
NSCLC-070

Tumor mutation burden

Figure 2. Correlation of RNA QC Assay with RNA-Seq performance

RCC-102

Immunotherapy biomarker assessment
in FFPE samples from solid tumors
using IHC, gene expression profiling and
mutation burden assessment

Peng Fang, Zhenyu Yan, Xiaodong Wang, Wes Chang, Chad Galderisi, Cindy Spittle and Jin Li, MolecularMD Corp., Portland OR and Cambridge MA
Introduction

For Further Information

Please contact BD@molecularmd.com or visit
www.molecularmd.com.

Evaluation of a commercial targeted NGS panel for tumor mutation burden assessment in FFPE tissue
Peng Fang, Zhenyu Yan, Quyen Vu, Dave Smith, Chad Galderisi, Cindy Spittle and Jin Li, MolecularMD Corp., Cambridge, MA and Portland, OR

Targeted NGS Analysis
A preliminary assessment of the performance of a targeted panel compared to Whole
Exome Sequencing (WES) for TMB assessment was conducted using the Ion
Comprehensive Cancer Panel. This panel targets 409 genes and is similar in content to
the Oncomine™ Tumor Mutation Load Research Assay (see below). When performing
TMB analysis using the CCP409 Panel synonymous and non-synonymous single
nucleotide variants (SNVs) in coding regions were included in the total count. Variants
were then removed from the total count based on the following filters: 1) variant
frequency below 10% 2) those found in dbSNP (germline mutations) and the COSMIC
database 3) those with mutant allele coverage below 20 read depth and 4) those that
failed in inspection using the Integrative Genomics Viewer (IGV) .
The Thermo Fisher Oncomine™ Tumor Mutation Load Assay (TML) was used to assess
the TMB level in the tumor samples selected (Table1). The TML Assay evaluates TMB
(mutations/Mb) by interrogating 409 cancer-related genes, spanning ~1.7 megabases
of the genome. 20ng dsDNA, measured by Qubit, was used for library preparation as
per the manufacturer's instruction.
Subsequently, the TMB in each sample was assessed with two workflows, namely, the
manufacturer's (THF) workflow and MolecularMD's (MMD) workflow (Figure 1). In
both workflows, alignment was performed by the Ion Torrent 5.6 software, and
variant calling and annotation by the Ion Reporter™ Software (IR 5.6).
In the THF workflow, a single library was prepared for each sample. Eight (8) barcoded
libraries from eight (8) unique samples were sequenced on one (1) 540 chip using the
Ion Chef-S5 XL system. TMB was measured by counting the somatic single-base
substitutions per Mb at ≥5% allele frequency.
In the MMD workflow, 2 libraries were independently prepared for each sample.
Sixteen barcoded libraries representing duplicates of eight (8) unique tumor samples
were sequenced on one (1) 540 chip using the Ion Chef-S5 XL system (TS5.6 and
IR5.6). The TMB was determined by counting the somatic SNVs per Mb that 1) were
reproducibly detected in each duplicate sample; 2) were exonic; 3) had a mutant
allele frequency above 10% in at least one duplicate; and 4) were not found in dbSNP
(germline mutations) or the COSMIC databases.
Figure 1: TMB analysis workflows

THF

MMD

WORKFLOW

A Clinical OMICs eBook

8 Samples/Chip

300 - 500x
8 Duplicated
Samples/Chip

250 - 300x

Counts of variants:
 Intronic  Synonymous
 Exonic
 Missense

 Nonsense

Counts of REPRODUCED variants:
 Exonic
 Non-dbSNP
 Missense
 Synonymous  Non-COSMIC  Nonsense

R² = 0.9484

CCP409-MMD

TMB by the THF workflow
TMB Analysis - THF workflow
When assessed with the THF workflow (see Figure 1), eight (8) libraries were
sequenced on one (1) 540 chip. The percentage reads on target ranged from 82% to
99%, uniformity from 87% to 97%, and mean depth from 312X to 897X (Figure 3).
TMB results ranged from 10-2779 variants/MB. The majority of samples were found
to harbor <125 variants/MB. Two samples were notable outliers (>400 variants/MB).
The TMB determined for each sample using the THF workflow showed a certain
degree of positive correlation with the MSI status measured in these samples (Figure
4), especially within related cancer types (e.g., gastrointestinal tract or the
reproductive system (comparing group "All", vs group "GI" or group "Cervix_uterus",
in Figure 4).
The analysis of 4 samples was repeated using the THF workflow, including the 2
samples with TMB >400 and 2 samples with TMB <100 variants/MB. The variants
reproducibly detected between the two runs ranged from 2% to 68% out of those
initially detected ("Reproduced/THF-mean" in Table 2). Previous data generated with
these samples suggested that poor DNA quality/deamination could lead to the
detection of false positive SNVs due to FFPE deamination errors.

300
Read on target
Uniformity
mean depth

20%

All-MSS
(n=10)

All-MSI
(n=12)

GI-MSS
(n=7)

GI-MSI
(n=10)

Cervix_Uterus-MSS
(n=3)

Sample ID

THF-1

THF-2

MMD-2-2-066
MMD-2-2-077
MMD-2-4-061
MMD-2-4-113

539
2697
108
42

586
2779
126
40

Reproduced MMD-pipeline*
19
56
31
28

7
25
13
7

Sample ID

Analysis of duplicate libraries allows removal of the false-positive SNV calls
(predominantly G>A or C>T errors caused by deamination). These results suggest
that 35% to 95% of the initial SNV calls can be due to errors (see numbers listed in
Column "Reproduced/mean-MMD" in Table 4). These results demonstrate the
advantage of using the duplicate library analysis approach, especially when sample
quality is poor. Note that the two (2) potentially poor-quality DNA samples had only
4 to 5% reproduced SNVs between duplicate libraries (highlighted in yellow in Table
4).

MMD-1 MMD-2 Reproduced MMD-pipeline

MMD-0262-129
MMD-0262-001
MMD-0262-024
MMD-0262-036
MMD-0262-045
MMD-0262-047
MMD-0262-052
MMD-0262-051
MMD-0262-D2
MMD-0262-D3
MMD-0262-D7
MMD-0262-D8
MMD-0262-157
MMD-0262-061
MMD-0262-010
MMD-0262-115
MMD-0262-012
MMD-0262-017
MMD-0262-015
MMD-0262-016

We hypothesized that duplicate analysis could eliminate false positive results due to
poor DNA quality. Therefore the MMD workflow was designed to perform duplicate
library preparations while minimizing the additional cost. To demonstrate the
feasibility of this approach, an in-silico analysis was first performed where 50% of the
aligned reads in the bam files generated using the THF workflow were randomly
removed to mimic the 50% drop in mean depth when using the MMD workflow.
These "half" bam files were then re-analyzed in IR5.6 using the same setting in the
THF "Oncomine Mutation Load" workflow, except that the "min SNP cov" for calling
a SNV was reduced from 60 to 30. The results indicate that TMB result from "full"
bam files and "half" bam files is comparable (Table 3).
Table 3. In-silico comparison of the THF workflow and the MMD workflow

When assessed with the MMD workflow (Figure 1), 16 libraries generated from 8
unique samples were sequenced on one 540 chip. The percentage reads on target
and uniformity from the MMD workflow were comparable to those from the THF
workflow, but as expected, the mean depth dropped about 50% (compare Figures 3
and 5).

MMD-0262-016

MMD-0262-015

MMD-0262-017

MMD-0262-012

MMD-0262-115

MMD-0262-010

MMD-0262-061

MMD-0262-D8

MMD-0262-D7

MMD-0262-D3

MMD-0262-D2

Table 4. TMB determined by the MMD workflow
Reproduced
/THF-mean
3.4%
2.1%
26.4%
68.8%

TMB Analysis - MMD workflow

MMD-0262-157

MMD-0262-051

MMD-0262-052

MMD-0262-047

MMD-0262-045

Sample ID

Cervix_Uterus-MSI
(n=2)

Sample ID Mean depth-THF Mean depth-MMD TMB-THF TMB-MMD
S1
451
226
21
17
S2
383
191
16
12
S3
409
204
22
18
S4
385
192
20
14
S5
405
202
20
18
S6
424
212
14
16
S7
297
148
45
25
S8
277
138
539
301
S9
403
201
97
57
S10
262
131
2697
1597
S11
433
216
49
44
S12
379
189
109
100
S13
496
248
42
41
S14
495
248
21
19
S15
577
289
197
202
S16
526
263
17
15

100
0

200
0

200

Average Mena Depth (X)

400

60%
40%

MMD-0262-036

Average Read on Target
/Uniformity (%)

500

80%

MMD-0262-024

Mena Depth (X)

MMD-2-4-113

MMD-2-4-061

MMD-2-2-077

MMD-2-2-066

MMD-0262-016

MMD-0262-D8

MMD-0262-015

MMD-0262-017

MMD-0262-D7

MMD-0262-D3

MMD-0262-012

MMD-0262-115

MMD-0262-D2

MMD-0262-010

MMD-0262-061

MMD-0262-157

MMD-0262-051

MMD-0262-052

MMD-0262-047

MMD-0262-045

Read on Target/Uniformity (%)

MMD-0262-036

MMD-0262-024

MMD-0262-001

600

100%

100

*: the variant filtering method in the MMD workflow (see Materials and Methods).

Figure 2. Correlation of the TMB determined by MMD workflow and the WES

400

120%

300

In silico TMB analysis - MMD workflow

In a separate pilot study, a set of sixteen FFPE samples was evaluated and TMB
assigned using WES and the CCP409 targeted gene panel. The CCP409 panel data
was analyzed using a MMD filtering method (See Materials and Methods). A good
correlation was observed (Figure 2), demonstrating the feasibility of using a targeted
gene panel for TMB analysis. Use of a targeted gene panel provides advantages over
WES when performing clinical sample analysis including lower DNA input
requirements, lower cost and faster TAT.

Figure 5. QC Metrics of the Sequencing Runs using MMD workflow

500

Cancer type and MS status

WES vs Targeted NGS for TMB Analysis

Sample ID

The TMB determined in each sample using the MMD workflow correlated well with the MSI status for those samples
(Figure 7).

Table 2. Reproducibility of TMB results using the THF workflow

Results

45
40
35
30
25
20
15
10
5
0

The MMD workflow TMB results correlated well with the duplicate THF workflow TMB results (Figure 6A, R2 >0.92 ).
The correlation between the standard (singlet) THF workflow TMB results and the MMD workflow TMB results was low
(Figure 6B, R2 <0.06). In addition, a poor correlation was also observed between the duplicate THF workflow TMB (i.e.,
no MMD-filtering) and standard (singlet) THF workflow analysis (Figure 6C, R2 <0.08).

MMD-0262-001

0
0
0
0
0
0
0
1
5
5
5
5
5
5
4
5
5
0
0
0
3
4
NA
NA

40%
20%

MMD-0262-129

MSS
MSS
MSS
MSS
MSS
MSS
MSS
MSI-L
MSI-H
MSI-H
MSI-H
MSI-H
MSI-H
MSI-H
MSI-H
MSI-H
MSI-H
MSS
MSS
MSS
MSI-H
MSI-H
ND
ND

Read on target
Uniformity
mean depth

Figure 4. Correlation of the TMB and MSI status using the THF workflow

Tumor Content (%) MSI-Status Altered Alleles
60
75
80
53
68
80
48
50
50
50
NA
80
80
N/A
N/A
N/A
N/A
45
100
25
40
100
70
70

60%

39
52
68
38
233
58
50
481
47
63
79
74
46
32
24
39
19
18
29
19

21
44
106
33
234
34
31
504
38
63
60
58
53
29
17
23
14
16
20
20

8
8.0
10.1
12.3
12.5
8.6
4.9
21.4
23
45.7
19
43.8
9.2
18.0
6
11.0
4.9
8.6
12.9
11.7

1
1.2
1.3
1.2
1.9
0.6
0.6
7.1
12
21.6
7
22.2
2.5
6.2
2
3.7
1.2
0.6
3.1
1.8

Reproduced/
MMD-mean
27%
17%
12%
35%
5%
19%
12%
4%
54%
73%
27%
66%
19%
59%
27%
36%
30%
51%
53%
60%

Figure 6. Correlation of TMB results using different
workflows

Figure 7. Correlation of the TMB and MSI status using the MMD workflow
25
20
15

A
TMB-Reproduced

A set of 24 FFPE tumor samples from cancers that occur in the gastrointestinal tract
(esophagus, stomach, large intestine), the reproductive system (cervix and uterus)
and lung were analyzed (Table 1). DNA was extracted using either the RecoverAll Total
Nucleic Acid Isolation Kit, or Promega Maxwell CSC DNA FFPE kit on the Maxwell CSC
instrument. DNA was quantified using NanoDrop and Qubit.

Tissue
Stomach
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Esophagus
Stomach
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Large Intestine
Cervix
Uterus
Uterus
Uterus
Uterus
NSCLC
NSCLC

80%

B
TMB-THF

Samples

Sample ID
MMD-0262-129
MMD-0262-001
MMD-0262-024
MMD-0262-036
MMD-0262-045
MMD-0262-047
MMD-0262-052
MMD-0262-157
MMD-0262-010
MMD-0262-051
MMD-0262-061
MMD-2-4-061
MMD-2-4-113
MMD-0262-D2
MMD-0262-D3
MMD-0262-D7
MMD-0262-D8
MMD-0262-115
MMD-0262-012
MMD-0262-017
MMD-0262-015
MMD-0262-016
MMD-2-2-066
MMD-2-2-077

1000
900
800
700
600
500
400
300
200
100
0

100%

R² = 0.9292

40
30
20
10
0

500
400
300
200
R² = 0.0508

100
0

TMB-MMD
50
40
30

R² = 0.0706

20
10
0

100

200

300

400

500

TMB-THF

5
0

TMB-MMD

TMB-Reproduced

Materials and Methods

TMB Analysis - MMD workflow

120%

TMB (No.Var/Mb)

In order to fully determine the value of TMB as a predictive biomarker for
immunotherapy, a standardized panel, workflow and data analysis pipeline for TMB
assessment are needed. In this study we evaluated the performance of a
commercially available targeted NGS panel and workflow for TMB analysis. The
correlation between the MSI status and the TMB level in each sample was also
evaluated.

Results continued

Figure 3. QC Metrics of the Sequencing Runs using THF workflow

MMD-0262-129

Increased TMB in cancer cells could be caused by abnormal activity in several cellular
pathways, including DNA damage repair and DNA replication. Microsatellite instability
(MSI) is a molecular marker of a deficient mismatch repair (MMR) pathway. The
positive correlation between MSI and TMB has been observed in certain cancer
types.

MSI Analysis
The MSI status in each sample was determined using the Promega MSI Analysis
System v1.2 on ABI 3500DX Genetic Analyzer. The system is a fluorescent PCRbased assay to detect the presence of five (5) mononucleotide repeat markers
(BAT-25, BAT-26, NR-21, NR-24 and MONO-27) and two pentanucleotide repeat
markers (Penta C and Penta D). Two (2) ng DNA from each sample, measured by
Qubit, was used as input into the assay. The MSI status was determined based on
comparing allelic profiles generated from tumor vs normal DNA. The presence of
alleles with altered length in the tumor sample is interpreted as exhibiting
microsatellite instability. Tumor samples with no altered markers are classified as
microsatellite stable (MSS), with ≥2 out of 5 altered marker as high MSI (MSI-H), or
with 1 altered marker as low MSI (MSI-L) (Table 1).
Table 1. FFPE clinical samples evaluated and their MSI status

TMB (No.Var/Mb)

Tumor mutation burden (TMB) has been shown to correlate with response to
checkpoint inhibitors and is emerging as a potentially important predictive biomarker.
So far, the methods used to assess tumor mutation burden have included exome
sequencing and multiple laboratory-developed targeted NGS panels (Refs.1 and 2).

WORKFLOW

Results continued

Materials and Methods continued

Checkpoint inhibitors have been approved for the treatment of solid tumor and
hematological malignancies. PD-L1 immunohistochemistry (IHC) assays are currently
being used as companion or complementary diagnostics to select patients for
treatment. While significant responses have been observed in a subset of patients,
outcomes are variable and there is a need to identify additional predictive biomarkers
beyond PD-L1 levels as measured by IHC.

WES

Evaluation of a commercial targeted
NGS panel for tumor mutation burden
assessment in FFPE tissue

Introduction

All-MSS
(n=10)

All-MSI
(n=12)

GI-MSS
(n=7)

GI-MSI
(n=10)

Cervix_Uterus-MSS Cervix_Uterus-MSI
(n=3)
(n=2)

Cancer type and MS status

Conclusions
* Commercially available targeted NGS panels such as the Oncomine™ TML Assay can provide advantages over WES
for TMB analysis such as lower FFPE DNA input requirements, lower cost and faster TAT.
* Poor quality FFPE DNA samples can generate false positive SNV calls and falsely elevated TMB scores.
* The MolecularMD workflow, including duplicate analysis and variant filtering, reduces the false positive variant calls
in the TMB calculation allowing for higher accuracy in the TMB determination.
* Additional studies are on-going to determine the clinical utility of TMB analysis for predicting patient response to
checkpoint inhibitors.

References
1.

Chalmers ZR, et al., (2017) Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med., 9(1):34.

Rizvi H, et al (2018) Molecular Determinants of Response to Anti-Programmed Cell Death (PD)-1 and Anti-Programmed Death-Ligand 1 (PD-L1) Blockade in
Patients With Non-Small-Cell Lung Cancer Profiled With Targeted Next-Generation Sequencing. J Clin Oncol., 36(7):633-641.

For Further Information

Please contact BD@molecularmd.com or visit www.molecularmd.com.

www.clinicalomics.com

http://www.clinicalomics.com

Table of Contents for the Digital Edition of Immunotherapy and Cancer: Harnessing the Power of Diagnostic Assays