Introduction
Deficient DNA mismatch repair (MMR) results in a hypermutated phenotype characterized
by microsatellite instability (MSI) and a higher burden of mutation-associated neoantigens
that are targeted by the immune system. Deficient MMR and high frequency MSI (MSI-H)
have been shown to be predictive biomarkers for immune checkpoint inhibitor drugs
which block the programmed death protein-1/programmed death ligand-1 (PD-1/PD-L1)
interaction between tumor cells and activated T cells. These agents include antibodies
against PD-1 and PD-L1 that have significantly altered the treatment algorithm for
several solid tumors. Presence and level of PD-L1 tumor expression are associated
with responsiveness to these antibodies in certain malignancies that include advanced
melanoma, non-small cell lung cancer, renal cell carcinoma, urothelial cancer and
Hodgkin’s lymphoma.
In 2017, the US Food and Drug Administration (FDA) granted accelerated approval to
pembrolizumab (KEYTRUDA) for patients with unresectable or metastatic d-MMR solid
tumors based upon an assessment of 15 cancer types, among which, 149 MSI-H patients
were enrolled across 5 clinical trials [1]. A complete or partial response was experienced by 39.6 % of patients, with responses
lasting ≥ 6 months in 78 %. Based upon these data, the FDA approved a cancer treatment
for the first time based upon tumor biomarker status, rather than a specific disease-based
approach.
Vanderwalde and colleagues recently published their experience evaluating a broad
spectrum of 11,348 solid tumors noting the frequency of dMMR based on having MSI-H:
PD-L1 expression in endometrial cancer (17 %: 16.2 %), gastric adenocarcinoma (8.7 %:
18.5 %), colorectal adenocarcinoma (5.7 %: 7.2 %), cholangiocarcinoma (2.3 %: 18.6 %),
pancreatic ductal adenocarcinoma (PDAC) (1.2 %: 21.6 %), renal cell carcinoma (0.6 %:
29.7 %) and melanoma (0 %: 42.3 %) [2]. The fact that dMMR is rarely present among PDAC patients was further demonstrated
in a study of 833 surgically resected PDAC tumors revealing a frequency of 0.8 %,
all of whom were patients found to have Lynch syndrome [3].
In an era of evolving tumor agnostic immunotherapy, we questioned if pancreatic endoscopic
ultrasound-guided fine-needle biopsy (EUS FNB) provides sufficient material for dual
immunohistochemistry (IHC) for dMMR and PD-L1 status evaluations. We sought to determine
the prevalence of MSI-H status and quantification of PD-L1 expression to determine
their utility in guiding disease-agnostic immunotherapy based upon biomarker status.
Patients and methods
IHC was performed on consecutive archived treatment-naïve formalin-fixed paraffin-embedded
EUS pancreas core biopsy specimens (2001 – 2017 IRB # 17-006528). Briefly, 4-μm-thick
tissue sections were stained using the Ventana BenchMark XT automated slide-staining
system using the following antibodies: Anti-PD-L1 (clone SP263, VENTANA, Tucson, AZ),
MLH1 (clone G168-728, Cell Marque, Rocklin, California, United States), MSH2 (clone
FE11, Biocare Medical, Concord, Massachusetts, United States), MSH6 (clone BC/44,
Biocare Medical, Concord, Massachusetts, United States), and PMS2 (clone A16-4, Biocare
Medical, Concord, M Massachusetts, United States). Antigen-antibody reactions were
visualized using UltraView detection with diaminobenzidine as the chromogen.
A minimum of 100 viable tumor cells were required on one stained slide for the specimen
to be considered adequate for evaluation. Positive PD-L1 expression was defined as
membranous staining. The approximate percentage of positive tumor cells versus all
tumor cells provided the Tumor Proportion Score (TPS). The specimen was considered
to have PD-L1 expression if PD-L1 was expressed in ≥ 1 % of tumor cells and a high
level of expression if ≥ 50 %. Tumors were classified as dMMR if they exhibited absent
nuclear staining of DNA mismatch repair proteins (MLH1, MSH2, MSH6, or PMS2).
Results
Clinical demographics
Thirty-nine treatment-naïve patients with either a primary or secondary pancreas malignancy
who underwent EUS with FNB were assessed for histologic specimen adequacy for MMR
and PD-L1 expression status. Age of the overall cohort was 71.2 ± 10.2 years, 61.5 %
were male, and the overall mortality rate was 25.6 % at 7 months (4.9 – 15.7) following
EUS diagnosis. The study cohort was composed of 21 patients with PDAC who had previously
reported PD-L1 expression analysis and with patient details as follows: 72.9 ± 8.9
years; 57 % male; CA19-9 level = 143 U/mL (18 – 998); 57 % 8th AJCC Stage ≥ III) [4]. The study also included 18 patients (69.3 ± 11.6 years; 66 % male) with either
a renal cell cancer (RCC) (n = 16) or melanoma that had metastasized to the pancreas
(n = 2). Using either a 19G Trucut biopsy needle (n = 9 patients) (Quick-Core Biopsy
Needle, Cook Medical, Winston-Salem, North Carolina, United States) or a 22G Sharkcore
FNB needle (n = 30 patients) (SharkCore FNB needle, Medtronic, Dublin, Ireland), tumor
tissue was sampled with an overall median tissue aggregate that was 19 mm (12 – 20)
in length following 2 (3 – 5) EUS-FNB passes.
DNA mismatch repair status
Among the 39 patients, 36 had specimen blocks available for retrieval from the tissue
archive and 35 (97.2 %) patients (FNB passes 2 [3]
[4]
[5]) had histologic specimen adequacy to assess dMMR by IHC. One patient (2.9 %) displayed
loss of MLH1-PMS2 ([Fig. 1]). The patient was a 64-year-old male with decompensated cirrhosis secondary to nonalcoholic
steatohepatitis, a splenic artery pseudoaneurysm resulting in hemosuccus pancreaticus
and previously resected colon cancer. He presented with cholestatic jaundice and an
elevated CA19-9 (1,184 U/mL) as a consequence of a 4-cm EUS T4N0M0 PDAC head mass.
The additional immunostains performed on the tumor showed that the tumor cells were
positive for CK7 and focally positive for CDX2 and CK20. This immunophenotype supports
the diagnosis of a primary PDAC. He was started on FOLFIRINOX. He subsequently underwent
germline testing which detected Lynch syndrome and received pembrolizumab. He had
stable disease on therapy, but succumbed to vascular complications, presumably from
his cirrhosis, 7.7 months following EUS.
Fig. 1a – e Loss of MLH1-PMS2 protein expression in a PDAC EUS-FNB specimen.
MMR status and PD-L1 expression
Thirty-seven patients (94.9 %) had specimens that were satisfactory for PD-L1 expression
assessment ([Fig. 2]). Expression thresholds of ≥ 1 %, ≥ 5 %, ≥ 10 %, ≥ 25 % and ≥ 50 % expression in
tumor cells were identified in 16 (43.2 %), 13 (35.1 %), 7 (18.9 %), 5 (13.5 %) and
3 (8.1 %) patients, respectively ([Fig. 3]). Tumor tissue from 33 patients underwent dual MSI and PD-L1 immunostaining ([Fig. 4]). PD-L1 expression (> 1 %) with proficient (p) MMR or dMMR status was identified
in 13 (39.4 %) and 1 (3 %) patient, respectively. The tumor from the patient with
Lynch syndrome had PD-L1 expression in 10 % of tumor cells.
Fig. 2 a EUS-FNB malignant melanoma metastasis to the pancreas with b PD-L1 expression > 90 %.
Fig. 3 Patient prevalence and quantification of PD-L1 expression positivity for EUS-FNB
specimens of treatment-naïve PDAC, clear cell renal cell carcinoma and malignant melanoma.
Fig. 4 Dual MSI and PD-L1 IHC assessment yield.
Biomarker-based immunotherapy eligibility
Based upon the prevalence of dMMR status, 3 % of our patients would be eligible to
receive pembrolizumab in the setting of a FDA-approved disease-agnostic immunotherapy
approach. PD-L1 expression was high (TPS > 50 %) in 8.1 % of patients, suggesting
that expression of this biomarker by their tumors may indicate potential sensitivity
to immune checkpoint inhibitor therapy.
Discussion
Among a unique patient cohort, we determined that a paired evaluation of MMR status
and PD-L1 expression by IHC is possible using specimens obtained with EUS-FNB. Our
findings demonstrate the capability of using tissues obtained with EUS-FNB for molecular
classification of treatment-naïve solid pancreas masses by identifying biomarkers
that may guide use of checkpoint inhibitor therapy. The 3 % prevalence of dMMR in
our cohort indicates that patients with these tumors are candidates for treatment
with pembrolizumab. Furthermore, 8.1 % of evaluated tumors had a high level of expression
of PD-L1 (TPS > 50 %), which in certain tumor types is associated with responsiveness
to checkpoint inhibitor therapy.
Following much debate, it has been established that endoscopic ultrasound-guided fine-need
aspiration and EUS-FNB needles yield a similar satisfactory specimen for diagnostic
purposes, with the caveat that fewer FNB passes are required to achieve such a yield
[5]
[6]
[7]
[8]
[9]
[10]. However, when tissue for companion diagnostic or prognostic tumor information is
required for integration of molecular analyses into clinical decision-making, the
optimal needle landscape changes depending upon the specific objective. Options for
use of tissue specimens include IHC, genotyping using next-generation sequencing (NGS),
or establishing tumor organoids in a research setting. PDACs demonstrate only 5 %
to 20 % neoplastic cellularity and are characterized by a prominent desmoplastic reaction
considered to be a hostile tumor microenvironment (TME) for subsequent therapy [11]
[12]. Cytology smears are thought to be superior for multiplex gene panel NGS evaluations,
as their overall cellularity, tumor fraction, and sequencing metrics are considered
to be superior to that of FNB, yet, successful preclinical disease models to include
organoid development require that EUS-FNB deliver tumor cells and the corresponding
TME [13]
[14]
[15].
As endosonographers, while we try to navigate the provision of satisfactory material
for ancillary molecular analysis to improve cytologic adequacy and identify predictors
of prognosis and chemosensitivity, we are now poised to assist in patient selection
for precision immuno-oncology. There are a number of barriers to the utility of immunotherapy
in PDAC as the mutational burden is low; the TME is characterized by dense stroma
with very few T cells, collectively creating a “cold” or non-immunogenic environment
and potentially making such tumors less responsive to checkpoint inhibitors. Yet,
there is much need for enhanced tissue access and sampling for analysis to utilize
biomarkers in patients with solid tumors, including PDAC, which can determine eligibility
for checkpoint inhibitor therapy.
To date, the MMR/MSI status of pancreatic lesions has only been evaluated in surgically
resected specimens, which represent a small proportion of patients that have earlier,
less advanced disease, as patients with more advanced disease are no longer surgical
candidates. MMR and PD-L1 status may be influenced by prior neoadjuvant or adjuvant
therapy as has been observed in both breast and colon cancer analyses [16]
[17]. Therefore, for primary pancreatic masses to include PDAC, it may be more desirable
to perform such analysis on treatment-naïve specimens for which EUS-FNB is ideally
suited to establish the molecular signature. For secondary pancreatic metastasis,
we evaluated RCC and melanoma treatment-naïve lesions, which is the ideal scenario.
However, from a practical perspective, patients with these lesions have usually undergone
prior treatment that requires careful consideration of the molecular testing.
Conclusion
In conclusion, we highlight that our early experience with dual MMR IHC and PD-L1
expression analysis shows that it is technically feasible to assess biomarker-based
immunotherapy eligibility with EUS-FNB specimens from solid primary and secondary
pancreatic masses. Based upon the prevalence of dMMR and high PD-L1 expression, 3 %
of the treatment-naïve cohort is a candidate for checkpoint inhibitor therapy and
high PD-L1 expression may provide additional information regarding sensitivity to
immunotherapy depending upon tumor type.
