ORIGINAL ARTICLE

The Immune Microenvironment, Genome-wide Copy

Number Aberrations, and Survival in Mesothelioma
Bibhusal Thapa, FCPS,a,b Adriana Salcedo, MSc,c Xihui Lin, BSc,c
Marzena Walkiewicz, MSc,b Carmel Murone, PhD,b,d Malaka Ameratunga, M.B.B.S.,e
Khashyar Asadi, FRACPA,d Siddhartha Deb, M.B.B.S.,b
Stephen Arthur Barnett, FRACS,f Simon Knight, FRACS,f Paul Mitchell, FRACP,e
D. Neil Watkins, PhD,g Paul C. Boutros, PhD,c,h,i Thomas John, PhDb,e,j,*
a
Department of Medicine, University of Melbourne, Parkville, Victoria, Australia
b
Olivia Newton-John Cancer Research Institute, Heidelberg, Victoria, Australia
c
Informatics and Biocomputing Program, Ontario Institute for Cancer Research, Toronto, Canada
d
Department of Pathology, Austin Health, Heidelberg, Victoria, Australia
e
Department of Medical Oncology, Austin Health, Olivia-Newton John Cancer and Wellness Centre, Heidelberg, Victoria,
Australia
f
Department of Thoracic Surgery, Austin Health, Heidelberg, Victoria, Australia
g
Garvan Institute of Medical Research, Darlinghurst, New South Wales, Australia
h
Department of Medical Biophysics, University of Toronto, Toronto, Canada
i
Department of Pharmacology and Toxicology, University of Toronto, Toronto, Canada
j
School of Cancer Medicine, La Trobe University, Bundoora, Victoria, Australia

Received 2 October 2016; revised 19 December 2016; accepted 12 February 2017

Available online - 28 February 2017

ABSTRACT differentiation 8–positive, and forkhead box P3–positive

lymphocytes. High PD-L1–positive expression correlated
Introduction: Results of recent clinical studies of immune
with worse prognosis (hazard ratio ¼ 2.37, 95% confidence
checkpoint inhibitors in malignant pleural mesothelioma
interval: 1.57–3.56, p < 0.001) in univariate analysis but not
(MPM) have dampened initial enthusiasm. However, the
in multivariate analysis. Higher percent genome alteration
immune environment and targets of these treatments such
was associated with epithelioid histological subtype and
as programmed cell death protein 1 and its ligand pro-
poorer survival (hazard ratio ¼ 1.59, 95% confidence inter-
grammed death ligand 1 (PD-L1) have not been well char-
val: 1.01–2.5, p ¼ 0.04) but not PD-L1 expression.
acterized in MPM. Using a large cohort of patients, we
investigated PD-L1 expression, immune infiltrates, and Conclusions: PD-L1 expression was associated with non-
genome-wide copy number status and correlated them to epithelioid MPM, poor clinical outcome, and increased
clinicopathological features. immunological infiltrates. Increased genomic instability did
not correlate with PD-L1 expression but was associated
Methods: Tissue microarrays were constructed and stained
with poorer survival.
with PD-L1(clone E1L3N [Cell Signaling Technology, Dan-
vers, MA]), cluster of differentiation 4, cluster of differenti-
2017 International Association for the Study of Lung
ation 8, and forkhead box P3 antibodies. PD-L1 positivity Cancer. Published by Elsevier Inc. All rights reserved.
was defined as at least 5% membranous staining regardless
Keywords: Mesothelioma; Immunohistochemistry; PD-L1;
of intensity, and high PD-L1 positivity was defined as at least
Tumor-infiltrating lymphocytes; Copy number aberrations
50%. Genomic DNA from a representative subset of 113
patients was used for genome-wide copy number analysis.
The percent genome alteration was computed as a proxy for *Corresponding author.
genomic instability, and statistical analyses were used to Disclosure: The authors declare no conflict of interest.
relate copy number aberrations to other variables. Address for correspondence: Thomas John, PhD, Austin Hospital,
Studley Rd., Heidelberg, Victoria, Australia 3084. E-mail: tom.john@
Results: Among 329 patients evaluated, PD-L1 positivity was onjcri.org.au
detected in 130 of 311 (41.7%), but high PD-L1 positivity was ª 2017 International Association for the Study of Lung Cancer.
Published by Elsevier Inc. All rights reserved.
seen in only 30 of 311 (9.6%). PD-L1 positivity correlated
ISSN: 1556-0864
with nonepithelioid histological subtype and increased infil-
http://dx.doi.org/10.1016/j.jtho.2017.02.013
tration with cluster of differentiation 4–positive, cluster of

Journal of Thoracic Oncology Vol. 12 No. 5: 850-859

May 2017 Mesothelioma: Immunological Characterization 851

Introduction tissue arrayer (Tissue Arrayer I, Beecher Instruments

Despite several trials investigating novel therapies in Inc., Sun Prairie, WI) with 1-mm cores in triplicate from
malignant pleural mesothelioma (MPM), there have been no each patient’s sample placed sequentially. Representa-
practice-changing therapies since pemetrexed and cisplatin tive cores were chosen on the basis of tumor cellularity
were found in a phase III trial to improve survival over and the interface with tumor stroma. Four-micrometer
cisplatin alone.1,2 Immune checkpoint inhibition therapy sections of the TMAs were cut and stained with cluster
has shifted treatment paradigms in many types of cancers of differentiation 4 (CD4), cluster of differentiation 8
and some recent data have shown promise for improving (CD8), forkhead box P3 (FOXP3), and PD-L1 antibodies
MPM treatment.3–6 However, response to checkpoint inhi- as previously published.16 In brief, we used the anti-CD4
bition therapy has been neither uniform nor predictable. To rabbit immunoglobulin G (IgG) (Cell Marque, Rocklin,
date, programmed death ligand 1 (PD-L1) expression in CA) at a concentration of 1.0 mg/mL, anti-CD8 mouse
both tumor and immune cells has been the most readily IgG1 clone C8/144B (Dako, Glostrup, Denmark), anti-
assayable marker. Although there are many caveats to its FOXP3 mouse IgG1 (Abcam, Cambridge, MA) at a
expression and predictive power, it remains in wide use for concentration of 17.0 mg/mL, and anti–PD-L1 rabbit IgG
its ability to enrich patients likely to respond and for its ease (E1L3N) (Cell Signaling Technology) at a concentration
of application as an immunohistochemical (IHC) assay. of 7.0 mg/mL. Antigen retrieval for PD-L1 and FOXP3
MPM was thought to be potentially immunogenic, as its was carried out by boiling in a microwave for 15 minutes
pathogenesis involves substantial inflammation, but in target retrieval solution (pH 9) (Dako). For CD4 and
recent studies have tempered the initial enthusiasm for CD8, we used cell conditioning (CC1) buffer (pH 8.5)
immune checkpoint inhibition.7 The large phase IIb (Ventana Medical Systems, Tucson, AZ) at 95 C for 52
DETERMINE study demonstrated no overall survival (OS) minutes. A horseradish peroxidase system was used to
benefit for the anti–cytotoxic T-lymphocyte–associated detect signals. For all T-cell markers tonsil tissue was
protein 4 (CTLA-4) antibody tremelimumab relative to used as a positive control, and for PD-L1 placenta was
placebo.8 Recent data confirm that mutational load in the positive control.
MPM is low,9 which may in part explain why the results of Quantification for CD4-positive, CD8-positive, and
the DETERMINE study were negative: in both melanoma FOXP3-positive lymphocytes was performed by using the
and NSCLC, mutational load is a predictor of benefit from Leica Aperio positive pixel count algorithm, version 9.1
CTLA-4 or programmed cell death protein 1 (PD-1) inhi- (Leica Biosystems, Nussloch, Germany) as previously
bition.10,11 Structural aberrations resulting in gene copy described17 and expressed as the number of cells per
number gains and losses are thought to be common in 1000 tumor cells. A subset was validated by manual count.
the MPM genome, but whether their burden or pattern An average of the counts in all evaluable cores was taken
influences the immunogenicity and tumor immune for the statistical analysis. For PD-L1 staining, we consid-
microenvironment in MPM has not been studied. ered at least 5% membranous staining (regardless of
PD-L1 is one of two ligands for PD-1. Inhibition of intensity) as positive16,18 and at least 50% moderate or
PD-1–PD-L1 interaction leads to continued T-cell acti- intense staining as highly positive.19 Samples with only
vation, thus potentiating antitumor activity.12 Data on cytoplasmic staining were considered negative and we did
PD-L1 expression in MPM are limited, and reported not score the PD-L1 expression on infiltrating cells. The
series have typically included small numbers of patients IHC scoring was conducted independently by investigators
with positivity ranging from 20% to 50%.13–15 B. T., M.A., K.A., and S. D. (authors K. A. and S. D. are
Using a tissue microarray (TMA) containing a large experienced pathologists). For PD-L1 scoring (by B. T. and
cohort of 329 MPM samples, we sought to characterize K. A.), the initial k score was 0.74 and discrepancies
the tumor–immune cell interface and correlate this with were settled by joint discussion. Cores with minimal or no
clinicopathological features and copy number profiles. tumor tissue and those with tumor necrosis were
excluded from all analyses.

Materials and Methods Statistics

Under an ethics committee-approved protocol, clinical OS was calculated from the time of initial diagnosis to
information on patient demographics, comorbidities, treat- death or last follow-up; patients who were lost to follow-up
ments, follow-up, and outcome was retrieved from medical or were alive were censored. Descriptive statistics were
records. used for clinicopathological data. Comparisons of different
parameters between the groups with respect to PD-L1
TMA and IHC Analysis expression were performed by using an analysis of vari-
After histological confirmation of diagnosis (see the ance test for continuous variables and Fisher’s exact test for
Supplementary Methods), TMAs were created by using a categorical variables. Values for neutrophil-to-lymphocyte
852 Thapa et al Journal of Thoracic Oncology Vol. 12 No. 5

ratio (NLR); CD4-positive, CD8-positive, and FOXP3-positive PD-L1–positive (Fig. 1). There was a strong correlation
lymphocyte counts; and CD4-positive–to–CD8-positive, between PD-L1 expression and nonepithelioid histo-
FOXP3-positive–to–CD4-positive, and FOXP3-positive–to– logical subtype (Table 1). When samples were
CD8-positive lymphocyte ratios were dichotomized using dichotomized on the basis of levels of infiltration with
the median. For survival analysis, a Wald-type test from CD4-positive, CD8-positive, and FOXP3-positive cells
Cox’s proportional hazard (PH) model was performed in the using the median, a higher level of infiltration was seen
R statistical environment. in patients who were highly PD-L1–positive (see
Table 1). However, PD-L1–positive tumors also showed
DNA Extraction and Copy Number Analysis significantly increased CD4-positive and CD8-positive
A representative subset of patient tumors (n ¼ 113) cell infiltration but not FOXP3-positive cell infiltration.
were selected for copy number analysis. One-millimeter
cores were taken from tumor blocks corresponding to Relationship between PD-L1 Status and Immune
sections marked by the pathologist (K.A.). Copy number Infiltration with Survival
profiles were studied by using Affymetrix OncoScan Patients with PD-L1–positive tumors had a nonsig-
Express SNP arrays (Affymetrix, Santa Clara, CA) and data nificant trend toward poorer OS (9.8 versus 13.5
were processed with Nexus Express (Biodiscovery, El months) (HR ¼ 1.19, 95% confidence interval [CI]: 0.91–
Segundo, CA). After removing 13 low-quality samples 1.53, Wald-type p ¼ 0.15). However, when the analysis
(samples [EGA Study Accession: EGAS00001002323] that was restricted to those who were strongly positive,
showed median absolute pairwise difference scores higher high PD-L1–positive expression was associated with a
than 0.3 and displayed a high level of noise), we computed significantly worse OS (HR ¼ 2.37, 95% CI: 1.57–3.56,
the percent genome aberration (PGA) for each sample. PGA Wald-type p <0.001). The poorer prognosis associated
was calculated as the total count of base pairs involved in with high PD-L1–positive status was maintained even
copy number gains or losses divided by the total length of when the epithelioid and nonepithelioid subtypes were
the genome in base pairs.20 Mutated genes were defined on analyzed separately (Fig. 2).
the basis of hg19 reference genome and the GENCODE When dichotomized by the median, levels of
reference gene annotation.21 We first identified signifi- CD4-positive, CD8-positive, and FOXP3-positive infiltra-
cantly recurrent copy number aberrations (CNAs) using tion demonstrated no association with survival, although
GISTIC 2.022 with the default settings and also investigated patients with high CD4-positive–to–CD8-positive ratios
the patterns in CNA occurrence among patients. Consensus achieved better survival (HR ¼ 0.71, CI: 0.55–0.90, Wald-
clustering was used to group patients on the basis of ter- type p ¼ 0.005).
narized CNA profiles (neutral, loss, or gain) for genes by Histological subtype, stage, NLR, and CD4-positive–
using hierarchical clustering with Jaccard distance and to–CD8-positive ratio were found to remain significant
1000 repetitions. We considered up to six clusters and on multivariate analysis. However, high PD-L–positive
determined optimal cluster number by maximizing average status failed the Schoenfeld residual test to validate the
item consensus and inspecting the consensus matrices. PH assumptions, suggesting that that its effect changes
with time. When Schoenfeld residuals were plotted, 10
Results months was found to represent a turning point and was
Between 1988 and 2014, 373 patients with MPM therefore chosen as a threshold. When a multivariate Cox
were captured within our database. After patients with model with time-variate HRs was performed, we found
inadequate formalin-fixed paraffin-embedded samples that although these parameters had a significant effect in
were excluded, TMAs were constructed from the tissues the earlier period (within 10 months), their effect faded
of 329 patients with confirmed MPM. thereafter (Table 2). When the cohort was stratified by
Demographic data are summarized in Supplementary histological subtype, we found that the effect of high
Table 1. The median OS was 12.06 months. On univariate PD-L1–positive status was time dependent in the non-
analysis, advanced age, history of significant weight loss, epithelioid histological subtype but stayed constant with
Eastern Cooperative Oncology Group performance status an HR greater than 6 in the epithelioid cohort (see
higher than 2, nonepithelioid histological subtype, higher Table 2).
stage (III and IV), and high NLR (>3.94) were found to
be related to poorer outcomes (Supplementary Table 2). Association of CNAs with PD-L1 Status, Tumor
Histological Subtype, and Survival
Immunological Characterization Our data revealed a median of 147 CNAs (range 3–
Among 311 evaluable samples, 130 (41.8%) were 866) per sample with losses (median 79, range 3–516)
PD-L1–positive but only 30 (9.64%) were highly being more common than gains (median 57, range 0–516).
May 2017 Mesothelioma: Immunological Characterization 853

Figure 1. Representative images of the immunohistochemistry for programmed death ligand 1 (PD-L1), cluster of
differentiation 4 (CD4), cluster of differentiation 8 (CD8), and forkhead box P3 (FOXP3). (A) PD-L1–negative. (B) Weakly
PD-L1–positive. (C) Highly PD-L1–positive. (D) High infiltration with CD4-positive cells. (E) High infiltration with CD8-positive
cells. (F) High infiltration with FOXP3-positive cells. (G) Low infiltration with CD4-positive cells. (H) Low infiltration with
CD8-positive cells. (I) Low infiltration with FOXP3-positive cells.

PGA in individual samples ranged from 0.1% to 55.5% We initially divided the groups into test and validation
(median 15.2). Using consensus clustering to group cohorts with the same result. When counts of individual
patients on the basis of CNA profiles, we identified two CNAs among the PD-L1 groups were compared by using
primary clusters (Fig. 3). Cluster 1 showed a significantly Fisher’s exact test at each gene, none was found to be
higher PGA than cluster 2 (Wilcoxon test p <0.001), significantly associated with PD-L1 status after correction
but cluster membership was not found to be prognostic for multiple testing.
(HR ¼ 0.99, 95% CI: 0.65–1.51, Cox PH p ¼ 0.97) and did
not correlate with PD-L1 status (Fisher’s exact test Discussion
p ¼ 0.76). Interestingly, cluster 2 contained a greater Using a clinically well-annotated TMA, we have
proportion of nonepithelioid MPMs (Fisher exact test demonstrated that PD-L1 is strongly expressed in a
p ¼ 0.03). PGA was significantly higher in epithelioid than subset of mesotheliomas, but most were weakly positive
nonepithelioid tumors (Supplementary Fig. 1) but did not or negative. The strongly positive samples were associ-
correlate with patient sex, asbestos exposure status, stage, ated with immune infiltration of CD4-positive, CD8-
and (importantly) PD-L1 status. Patients with a lower PGA positive, and FOXP3-positive lymphocytes, as well as
had significantly higher infiltration with CD4-positive with poorer survival, in a time-dependent fashion.
(t test p ¼ 0.03) and CD8-positive (t test p ¼ 0.001) Although we found a large number of CNAs, neither the
cells. When patients were dichomatized by the tertiles PGA nor any particular CNA was found to be associated
(lower third versus upper two-thirds), patients with a with PD-L1 expression. Intriguingly, PGA was higher
lower PGA achieved better survival (HR ¼ 2.01, 95% CI: among epithelioid MPMs, and yet higher PGA was asso-
1.24–3.26, Cox PH p ¼ 0.004) after controlling for age ciated with poorer survival.
and tumor stage. This was true when the epithelioid and Using a predefined stringent scoring criteria in
nonepithelioid tumors were analyzed separately (Fig. 4). accordance with previous reports,18,19 we observed a
854 Thapa et al Journal of Thoracic Oncology Vol. 12 No. 5

Table 1. Correlation of PD-L1 with Clinicopathological Parameters and Immune Infiltrates

Characteristic PD-L1–Negative PD-L1–Positive High Positivity for PD-L1 p Value
No. (%) 181 (58.2) 100 (32.1) 30 (9.6)
Age 65.3 67.7 68.4 0.27a
Sex, n (%) 0.03b
Male 155 (85.6) 79 (79) 27 (90)
Female 26 (14.4) 21 (21) 3 (10)
Stage, n (%) 0.39b
I and II 91 (48.6) 49 (50) 10 (37)
III and IV 87 (51.6) 49 (50) 17 (63)
NLR (mean) 5 5.7 4.9 0.42a
Histological subtype 0.009b
Epithelioid 125 (69.8) 64 (57.4) 9 (31)
Nonepithelioid 54 (30.1) 40 (42.5) 20 (69)
Median survival, mo 13.5 11.33 5.33 <0.001a
CD4-positive lymphoctye count, n (%) <0.001b
Lowc 94 (57.6) 39 (45.8) 4 (17.4)
Highd 69 (42.4) 46 (54.2) 23 (82.6)
CD8-positive lymphocyte count, n (%) <0.001b
Low 94 (58.7) 39 (43.8) 3 (10.3)
High 66 (41.2) 50 (56.2) 26 (89.6)
FOXP3-positive lymphocyte count, n (%) <0.001
Low 87 (57.6) 41 (47.3) 2 (9)
High 64 (42.4) 45 (52.3) 20 (91)
a
p Value from analysis of variance test.
b
p Value from Fisher’s exact test.
c
Median value and lower.
d
Above median.
PD-L1, programmed death ligand 1; NLR, neutrophil-lymphocyte ratio; CD4, cluster of differentiation 4; CD8, cluster of differentiation 8; FOXP3, forkhead box P3.

PD-L1 positivity rate of 41.4%. Mansfield et al.14 group with very high expression (50% staining in our
reported a 40% positivity rate in 106 patients when case) may therefore help identify the truly positive cases.
they regarded both cytoplasmic and membranous Also, the location of PD-L1 on tumor cells is thought to
staining with 5% taken as a cutoff. They found a lower be germane to its prognostic role. It is presumed that
rate (24%) of exclusive membranous staining. Cedrés membranous expression is the most relevant form of
et al.13 found 20% positivity in their 77 evaluable PD-L1 because PD-1 mediates downstream signalling
patients despite considering cytoplasmic and membra- cascades only when it has been ligated.24 Cytosolic
nous staining and staining of infiltrating lymphocytes. PD-L1 has not been reported to correlate with patient
Lack of uniformity in staining procedures, use of response to immunotherapy, and as such its importance
different antibodies, differing cutoffs, and also the fact is unclear.25,26
that PD-L1 expression is dynamic and heterogeneous A significant finding of our study was the correlation
within the tumor makes comparison between different of PD-L1 expression with tumor-infiltrating lympho-
series difficult.23 Given that nonepithelioid MPMs are cytes. Increased CD8-positive T-cell counts with high
more likely to be PD-L1–positive, the comparatively PD-L1positivity is in keeping with the current under-
lower positivity rates noted by Cedrés et al.13 may be standing that the PD-1–PD-L1 interaction results in
explained by a larger proportion of epithelioid MPM in clonal proliferation of CD8-positive lymphocytes that are
their series. Mansfield et al.14 used a different antibody functionally impaired.27 A similar correlation was
(clone 5H1-A3). The effect of different antibodies on observed by Cedrés et al.,13 such that tumors with
PD-L1 positivity rates was recently demonstrated by increased CD8-positive and CD4-positive tumor-
Combaz-Lair et al.15 In their 58 patients with MPM, when infiltrating lymphocyte counts were seen in 68% and
1% membranous staining was used as the cutoff, they 59% of PD-L1–positive patients, respectively. The anti-
found 50% (29 of 58) positivity with clone E1L3N (Cell body used for our IHC (E1L3N XP) has been known to
Signaling Technology) and 29% (17 of 58) with clone stain immune cells and could potentially confound these
SP142 from Spring Bioscience (Pleasanton, CA). How- results; however, we have tried to avoid this by scoring
ever, these numbers were limited and full face was used PD-L1 expression exclusively on malignant cells and
for IHC.15 Setting a higher benchmark by defining a discounting the immune infiltrates.
May 2017 Mesothelioma: Immunological Characterization 855

Figure 2. Overall survival and correlation with histological subtype and programmed death ligand 1 (PD-L1) status. (A)
Kaplan-Meier (KM) curve based on the whole cohort. (B) KM curve by histological subtype. (C) KM curve by PD-L1 in the whole
cohort. (D) KM curve by PD-L1 in epithelioid tumors. (E) KM by PD-L1 in nonepithelioid tumors.

This study confirmed the findings of the two smaller subtype was included in the model, suggesting that the
studies13,14 showing that PD-L1 positivity was related to poor outcome seen in PD-L1–positive patients was more
worse outcome. However, the association with survival a function of histological subtype rather than PD-L1
was lost in multivariate analysis when histological expression alone. Analyses of survival restricted to

Table 2. Multivariate Analysis for Overall Survival

Variable Time Period (mo) Hazard Ratio Lower 95% Upper 95% p Value
Histological subtype <0.001
Epithelioid vs. nonepithelioid 0.58 0.41 0.80
Stage <0.001
III and IV vs. I and II 1.73 1.37 2.20
NLR 0.005
>3.94 vs. 3.94 1.38 1.10 1.74
CD4/CD8 ratio 0.001
>0.93 vs. 0.93 0.67 0.53 0.86
High positivity for PD-L1
Yes vs. no
Nonepithelioid 10 2.98 1.69 5.26 <0.001
Nonepithelioid >10 0.61 .22 1.73 0.36
Epithelioid 10 6.03 2.56 14.19 <0.001
Epithelioid >10 7.81 1.04 58.35 0.045
Note: Wald-type p value from Cox model with time variate hazard ratios for high PD-L1–positive expression.
NLR, neutrophil-lymphocyte ratio; CD4, cluster of differentiation 4; CD8, cluster of differentiation 8; PD-L1, programmed death ligand 1.
856 Thapa et al Journal of Thoracic Oncology Vol. 12 No. 5

Figure 3. Consensus clustering based on copy number profile. Each row represents a tumor sample, and genomic regions are
represented in columns. The bar graph displays the percent genome aberration (PGA) in individual tumor samples. C1, cluster
1; C2, cluster 2; CNA, copy number aberration; PD-L1, programmed death ligand 1; CDK6, cyclin-dependent kinase 6 gene;
MET, MNNG HOS Transforming gene; CDKN2A, cyclin-dependent kinase inhibitor 2A gene; HOXB2, homeobox B2 gene; TIMP2,
metallopeptidase inhibitor 2 gene; PGA, per cent genome aberration.

high PD-L1 positivity showed the hazard associated to be Interestingly, given their overall poorer prognosis, such
significant but time dependent. Time variance of risks is patients are rarely eligible for clinical trials and so there
thought to be common in oncology-related studies with are limited data currently available for these subtypes.
long follow-up, although it is only looked for and The concept that effective checkpoint inhibition
reported in a very small proportion of studies.28 Our therapy requires T-cell responses to neoantigens sec-
finding of a constant HR higher than 6 among epithelioid ondary to mutational burden in cancer cells has been
MPMs could suggest that patients with high PD-L1– substantiated in recent studies in melanoma11 and lung
positive status would stand to benefit more from ant– cancer.10 The mutational load of MPM is known to be
PD-1/PD-L1 therapy. However, the proportion of these relatively low,9 and yet, there have been some reports of
patients in our cohort of epithelioid patients (seven of encouraging response to checkpoint inhibition therapy
154 [4%]) was small and therefore this finding needs to in MPM. CNAs have been reported as the predominant
be validated in another cohort. genetic aberrations in MPM31 and are thought to be able
The association between PD-L1 positivity and histo- to produce proteins with new and modified activities.32
logical subtype is an important finding because it has This study did not find a significant association
been consistent in all reported series. Given that PD-L1 between the PGA and frequency of individual CNA with
expression on tumor cells has to date been the PD-L1 status. We did however find that lower PGA was
most promising predictive biomarker,29,30 our data associated with better survival, and intriguingly, despite
suggest that nonepithelioid mesothelioma, which is a having higher rates of PD-L1 positivity, nonepithelioid
disease with few treatment options and universally poor MPM harbored fewer CNAs than epithelioid MPM. These
outcomes, may be best suited to these therapies. results differ from those of recent reports from Bueno
May 2017 Mesothelioma: Immunological Characterization 857

Figure 4. Percent genome aberration (PGA) and its relationship with programmed death ligand 1 (PD-L1) status and overall
survival. (A) Boxplots comparing PGA among groups with differing PD-L1 expression. p Value from Mann-Whitney U test. (B)
Kaplan-Meier curves comparing survival in PGA groups in the whole cohort and the Cox proportional hazard p value controlling
for age and stratified by histological subtype. PGA is divided between the lower and two upper tertiles. (C) Kaplan-Meier
curves comparing survival in PGA groups in epithelioid malignant pleural mesothelioma. Abbreviation: HR, hazard ratio.

et al.,9 who found no differences in somatic mutation Our characterization of the immunological infiltrates,
rates between histotypes. However, CNAs and point PD-L1, and CNAs in MPM provides important data from a
mutations arise from distinct mutational processes, and large patient cohort. Our data suggest that the immune
it is not surprising that patterns observed in one muta- axis is a valid target in MPM; however, single-agent
tion class would not necessarily reflect the other. If MPM immune checkpoint inhibitors are unlikely to be effec-
is a class C tumor driven more strongly by CNAs than tive in most patients with MPM, given the low rates of
point mutations (as suggested by the low mutation rate), strong PD-L1 expression and previous data confirming
our assessment of PGA would then be more likely to low mutational burden. As combination immune check-
detect differences in genomic instability among subtypes point inhibition threatens to shift treatment paradigms
than would comparing point mutation rates. in melanoma and NSCLC, these data provide further
Using a TMA in lieu of full sections is subject to rationale for implementation of such combinations in
sampling error, and this could be considered a limitation this rare but deadly disease.
of this study. To reduce this, we sampled three cores
from separate areas of the tumor. TMAs allow efficient Acknowledgments
assessment of large cohorts and as such have been used This study was supported by the Lyall Watt’s Mesothe-
to assess PD-L1 and immune infiltrates for a multitude of lioma grant awarded by Cancer Council Victoria and
malignancies.33,34 Also, high concordance between TMA Victorian Cancer Agency and in part by research funds
and whole sections in the evaluation of IHC in meso- from Slater and Gordon, Asbestoswise, and the Opera-
thelioma has been previously demonstrated, although tional Infrastructure Support Program provided by the
our study focuses not just on tumor expression but Victorian Government, Australia. Dr. John is the recipient
also on immune cell infiltrates, which may not be as of a National Health and Medical Research Council Early
concordant across a sample.35 Career Fellowship (APP1074035), and Dr. Thapa is
858 Thapa et al Journal of Thoracic Oncology Vol. 12 No. 5

supported by an Australian Postgraduate Award post- 10. Rizvi NA, Hellmann MD, Snyder A, et al. Cancer immu-
graduate scholarship. This study was also supported by nology. Mutational landscape determines sensitivity to
the Ontario Institute for Cancer Research to Dr. Boutrous PD-1 blockade in non-small cell lung cancer. Science.
2015;348:124–128.
through funding provided by the government of Ontario.
11. Snyder A, Makarov V, Merghoub T, et al. Genetic basis for
Dr. Boutros was also supported by a Terry Fox Research clinical response to CTLA-4 blockade in melanoma.
Institute New Investigator Award and a Canadian In- N Engl J Med; 2014:2189–2199. http://dx.doi.org/10.
stitutes of Health Research New Investigator Award. The 1056/NEJMoa1406498.
funding sources had no role in collection, analysis, and 12. Ahmadzadeh M, Johnson LA, Heemskerk B, et al. Tumor
interpretation of data or in the writing of the manuscript. antigen–specific CD8 T cells infiltrating the tumor express
high levels of PD-1 and are functionally impaired. Blood.
2015;114:1537–1545.
Supplementary Data 13. Cedrés S, Ponce-Aix S, Zugazagoitia J, et al. Analysis of
Note: To access the supplementary material accompa- expression of programmed cell death 1 ligand 1 (PD-L1)
nying this article, visit the online version of the Journal of in malignant pleural mesothelioma (MPM). PLoS One.
Thoracic Oncology at www.jto.org and at http://dx.doi. 2015;10:e0121071.
14. Mansfield AS, Roden AC, Peikert T, Sheinin YM. B7-H1
org/10.1016/j.jtho.2017.02.013.
Expression in malignant pleural mesothelioma is associ-
ated with sarcomatoid histology and poor prognosis.
References J Thorac Oncol. 2014;9:1036–1040.
1. Vogelzang NJ, Rusthoven JJ, Symanowski J, et al. Phase 15. Combaz-Lair C, Galateau-Sallé F, McLeer-Florin A,
III study of pemetrexed in combination with cisplatin et al. Immune biomarkers PD-1/PD-L1 and TLR3 in
versus cisplatin alone in patients with malignant pleural malignant pleural mesotheliomas. Hum Pathol.
mesothelioma. J Clin Oncol. 2003;21:2636–2644. 2016;52:9–18.
2. Chia PL, Russell P, Scott A, John T. Targeting the 16. Ameratunga M, Asadi K, Lin X, et al. PD-L1 and tumor
vasculature: anti-angiogenesis agents for malignant infiltrating lymphocytes as prognostic markers in resec-
mesothelioma. Expert Rev Anticancer Ther. ted NSCLC. PLoS One. 2016;11:e0153954.
2016;16:1235–1245. 17. Garcia-Rojo M, Sanchez J, de la Santa E, et al. Auto-
3. Calabrò L, Morra A, Fonsatti E, et al. Tremelimumab for mated image analysis in the study of lymphocyte
patients with chemotherapy-resistant advanced malig- subpopulation in eosinophilic oesophagitis. Diagn Pathol.
nant mesothelioma: an open-label, single-arm, phase 2 2014;9(suppl 1):S7.
trial. Lancet Oncol. 2013;14:1104–1111. 18. Taube JM, Klein A, Brahmer JR, et al. Association of
4. Calabrò L, Morra A, Fonsatti E, et al. Efficacy and PD-1, PD-1 ligands, and other features of the tumor
safety of an intensified schedule of tremelimumab for immune microenvironment with response to anti-PD-1
chemotherapy-resistant malignant mesothelioma: an therapy. Clin Cancer Res. 2014;20:5064–5074.
open-label, single-arm, phase 2 study. Lancet Respir 19. Garon EB, Rizvi NA, Hui R, et al. Pembrolizumab for the
Med. 2015;3:301–309. treatment of non-small-cell lung cancer. N Engl J Med.
5. Hassan R, Thomas A, Patel MR, et al. Avelumab 2015;372:2018–2028.
(MSB0010718C; anti-PD-L1) in patients with advanced 20. Lalonde E, Ishkanian AS, Sykes J, et al. Tumour genomic
unresectable mesothelioma from the JAVELIN solid tumor and microenvironmental heterogeneity for integrated
phase Ib trial: safety, clinical activity, and PD-L1 prediction of 5-year biochemical recurrence of prostate
expression [abstract]. J Clin Oncol. 2016;34(suppl):8503. cancer: a retrospective cohort study. Lancet Oncol.
6. Alley EW, Molife LR, Santoro A, et al. Clinical safety and 2014;15:1521–1532.
efficacy of pembrolizumab (MK-3475) in patients with 21. Harrow J, Frankish A, Gonzalez JM, Frazer KA. GENCODE:
malignant pleural mesothelioma: preliminary results the reference human genome annotation for the ENCODE
from KEYNOTE-028 [abstract]. Proceedings of the 106th project. Genome Res. 2012:1760–1774.
Annual Meeting of the American Association for Cancer 22. Mermel CH, Schumacher SE, Hill B, Meyerson ML,
Research; 2015 Apr 18–22. Philadelphia, PA: AACR; Beroukhim R, Getz G. GISTIC2.0 facilitates sensitive and
2015:CT103. confident localization of the targets of focal somatic
7. Thapa B, Watkins DN, John T. Immunotherapy for copy-number alteration in human cancers. Genome Biol.
mesothelioma: reality check. Expert Rev Anticancer 2011;12:R41.
Ther. 2016:Oct 6:1-10. 23. Momtaz P. Immunologic checkpoints in cancer therapy:
8. Kindler HL, Scherpereel A, Calabrò L, et al. Trem- focus on the programmed death-1 (PD-1) receptor
elimumab as second- or third-line treatment of unre- pathway. Pharmgenomics Pers Med. 2014;7:357–365.
sectable malignant mesothelioma (MM): results from the 24. Hamid O, Carvajal RD. Anti-programmed death-1 and
global, double-blind, placebo-controlled DETERMINE anti-programmed death-ligand 1 antibodies in cancer
study [abstract]. J Clin Oncol. 2016;34(suppl):8502. therapy. Expert Opin Biol Ther. 2013;13:847–861.
9. Bueno R, Stawiski EW, Goldstein LD, et al. Comprehen- 25. Weber JS, Kudchadkar RR, Yu B, et al. Safety, efficacy,
sive genomic analysis of malignant pleural mesothelioma and biomarkers of nivolumab with vaccine in
identifies recurrent mutations, gene fusions and splicing ipilimumab-refractory or -naive melanoma. J Clin Oncol.
alterations. Nat Genet. 2016;48:407–416. 2013;31:4311–4318.
May 2017 Mesothelioma: Immunological Characterization 859

26. Bigelow E, Bever KM, Xu H, et al. Immunohistochemical 31. De Assis LVM, Locatelli J, Isoldi MC. The role of key
staining of B7-H1 (PD-L1) on paraffin-embedded slides genes and pathways involved in the tumorigenesis
of pancreatic adenocarcinoma tissue. J Vis Exp. 2013;1:1–6. of malignant mesothelioma. Biochim Biophys Acta.
27. Freeman BGJ, Long AJ, Iwai Y, et al. Engagement of the 2014;1845:232–247.
PD-1 immunoinhibitory receptor by a novel B7 family 32. Hastings P, Lupski JR, Rosenberg SM, Ira G. Mechanisms
member leads to negative regulation of lymphocyte of change in gene copy number. Nat Rev Genet.
activation. J Exp Med. 2000;192:1027–1034. 2009;10:551–564.
28. Bellera CA, MacGrogan G, Debled M, de Lara CT, 33. Zhang Y, Kang S, Shen J, et al. Prognostic significance of
Brouste V, Mathoulin-Pélissier S. Variables with time- programmed cell death 1 (PD-1) or PD-1 ligand 1 (PD-L1)
varying effects and the Cox model: some statistical expression in epithelial-originated cancer: a meta-anal-
concepts illustrated with a prognostic factor study in ysis. Medicine (Baltimore). 2015;94:e515.
breast cancer. BMC Med Res Methodol. 2010;10:20. 34. Donnem T, Hald SM, Paulsen E-E, et al. Stromal CD8þ
29. Brahmer JR, Tykodi SS, Chow LQ, et al. Safety and T-cell density—a promising supplement to TNM staging
activity of anti–PD-L1 antibody in patients with advanced in non-small cell lung cancer. Clin Cancer Res.
cancer. N Eng J Med. 2012;366:2455–2465. 2015;21:2635–2643.
30. Herbst RS, Soria J-C, Kowanetz M, et al. Predictive cor- 35. Kao SC, Lee K, Armstrong NJ, et al. Validation of tissue
relates of response to the anti-PD-L1 antibody MPDL3280A microarray technology in malignant pleural mesotheli-
in cancer patients. Nature. 2014;515:563–567. oma. Pathology. 2011;43:128–132.