Exceptional Response to Pembrolizumab in a Metastatic, Chemotherapy/Radiation-Resistant Ovarian Cancer Patient Harboring a PD-L1-Genetic Rearrangement

Ovarian carcinoma remains the cancer with the highest mortality rate among gynecologic tumors. In 2017, 22,440 new cases of ovarian carcinoma are predicted in the United States, with 14,080 deaths secondary to this disease (1). While the majority of patients with advanced-stage disease (i.e., stages III–IV) initially respond to first-line therapy based on cytoreductive surgery and platinum-based chemotherapy, ovarian cancer will recur in the majority of cases. Overall, the 5-year survival rate for FIGO (International Federation of Gynecologists and Obstetricians) stage III disease is 20% to 25% and for stage IV disease is only 5% (2). These figures illustrate the dire need for the development of novel, more effective approaches for the treatment of chemotherapy-resistant ovarian cancer.

Next-generation sequencing (NGS) studies from The Cancer Genome Atlas (TCGA) network and other research groups have recently demonstrated that high-grade serous ovarian cancer, the most common histological type of ovarian carcinoma, is characterized by TP53 mutations in over 96% of cases, a relative low number of single-nucleotide variants (SNVs) when compared with other gynecologic and non-gynecologic tumors and a widespread copy-number variants (CNVs), a genetic signature correlated with high genomic instability (3, 4).

In the last few years, the successful treatment of a variety of human patients with cancer using antibodies against programmed cell death 1 (PD1) and its ligand PD-L1 (CD274) has demonstrated the intrinsic immunogenicity of at least a subset of human tumors and the critical importance of the PD-1/PD-L1 axis in tumor immune evasion. In some of these trials, a high mutational tumor burden and a high expression of PD-L1 protein on the surface of human tumors have been identified as predictive biomarkers for the identification of patients most likely to respond to the immunotherapy treatment (5). However, the mechanisms controlling response to current immune checkpoint inhibitors remain poorly understood as demonstrated by the fact that some human tumors with low mutation burden and/or lacking expression of PD-L1 may still respond to checkpoint blockade (5, 6).

Several studies have recently investigated whether PD-L1 expression is associated with tumor-infiltrating T cells (TILs) and a favorable prognosis in patients with ovarian cancer (7–11), and multiple antibodies directed against PD1 or PD-L1 are currently being tested in clinical trials (for review see ref. 11). While many of these studies are currently ongoing or waiting for results to mature, few have been published. For example, Hamanischi and colleagues (12) treated 20 patients with platinum-resistant ovarian cancer with an intravenous infusion of nivolumab every 2 weeks at a dose of 1 or 3 mg/kg. The best overall response was 15%, which included two patients who had a durable complete response (in the 3 mg/kg cohort). The disease control rate in all 20 patients was 45%. Of interest, complete responses were seen in patients with ovarian cancer with both serous and clear cell histology (12). Varga and colleagues reported on the activity of pembrolizumab in the ovarian cancer cohort of advanced ovarian epithelial, fallopian tube, or primary peritoneal carcinoma patients with PD-L1⁺ tumors treated within the KEYNOTE-028 (NCT02054806) study. Out of the 26 patients enrolled, the best overall (confirmed) response was 11.5% with one patient achieving a complete response and two patients experiencing partial responses (13). Disis and colleagues recently reported on 124 patients with recurrent/refractory ovarian cancer treated with avelumab (i.e., anti–PD-L1 humanized antibody; NCT01772004; ref. 14). The overall response rate was 9.7% based on 12 partial responses (14). Importantly, in all these trials, single-agent anti-PD1 or anti–PD-L1 antibodies showed an acceptable safety profile in heavily pretreated patients with ovarian cancer (11–14).

Here, we describe a patient with ovarian cancer treated with neoadjuvant chemotherapy, followed by surgery and progressing on radiation and multiple regimens of chemotherapy, experiencing an exceptional clinical response to anti-PD1 therapy (i.e., pembrolizumab). An analysis of the pathological and genomic characteristics of the ovarian tumor demonstrated a low/intermediate number of mutations by WES and a PD-L1 gene structural variation causing aberrant PD-L1 surface expression and heavy infiltration of T lymphocytes in the tumor tissue. These results strongly suggest immune evasion as a key mechanism of tumor growth and survival for this chemotherapy/radiation resistant ovarian cancer.

This study was conducted in accordance with recognized ethical guidelines, after approval by the institutional review board, and after informed written consent was obtained from the patient.

Briefly, for IHC, 4-μm sections were cut from the formalin-fixed paraffin embedded (FFPE) blocks of the cancer patient collected at the time of cytoreductive surgery and 79 additional high-grade serous ovarian carcinoma from a recently described tissue microarray ovarian cancer cohort (15), and stained with the following antibodies according to the manufacturers' instructions: CD8 (clone 144B, ready to use, DAKO), CD4 (clone 1F6, 1:40 dilution, Vector), CD3 (clone 2GV6, Ventana), CD56 (clone 1B6, 1:200 dilution, Vector), CD68 (clone PG-M1, ready to use, DAKO), CD20 (clone L26, 1:200 dilution, DAKO), TIA-1 (clone TIA1, ready to use, Biocare), CK7 (clone OVTL, Dako), and PD-L1 (clone E1L3N, 1:200 Cell Signaling). For antigen retrieval, the sections were pretreated at low pH for PD-L1 and CD8, CD4, CD20, CD56, and CD68. PD-L1 antibody and membranous immunoreactivity was assessed semiquantitatively in tumor cells as follows: <1% staining was considered negative, staining in 1% to 50% of tumor cells was scored as focal, and >50% staining was scored as diffusely positive.

PD-L1 (Spring Bioscience; clone SP142; 1:800), CD68 (DAKO; clone PG-M1; 1:200), and CD8 (DAKO; clone C8/144B; 1:250) were used for immunofluorescence multiplex staining using a previously descried method (16). Simultaneously, cytokeratin and DAPI were stained for detection of tumor and nucleus, respectively. Briefly, the patients' whole tissue section slides and a technical control tumor microarray slides (TMA; ref. 16) were deparaffinized. Antigen retrieval was performed for 20 minutes at 97°C in the PT module from LabVision (Thermo Scientific), in EDTA pH8 buffer. Endogenous peroxidases were blocked for 30 minutes in 2.5% hydrogen peroxide in methanol and subsequently blocked by 0.3% BSA in 0.05% tween with 30-minute incubation. The mixture of primary antibodies (PD-L1, CD68/CD8) was incubated overnight at 40°C in 0.3% BSA in TBST. For the PD-L1/CD68 multiplex panel, primary antibodies incubation was followed by anti-mouse IgG3 secondary antibody (Abcam; ab97260; 1:1000) incubation in 0.3% BSA in TBST for an hour at room temperature to detect CD68. Signal was amplified with Biotinylated Tyramide (Perkin Elmer) at 1:50 dilution and Alexa750–streptavidin (Life Technologies) at 1:100. Following 1-hour incubation, rabbit EnVisionreagent (Dako) was used for detecting PD-L1 and amplified by Cyanine 5(Cy5) tyramide (Perkin Elmer) at 1:50 dilution. Cytokeratin was detected with 1-hour incubation of anti-rabbit cytokeratin (Dako) at 1:100 and goat anti-rabbit secondary antibody (Life Technologies) at 1:100. ProLong mounting medium (Molecular Probes) was the final step to stain nuclei with 4,6-diamidino-2-phenylindole (DAPI). Between each step, slides were washed using TBS/Tween and TBS, each for 2 minutes. For the PD-L1/CD8 multiplex panel, primary antibodies were followed by anti-mouse IgG1 secondary antibody (eBioscience) at 1:100, and signal was amplified with Plus Cyanine 3 (Perkin Elmer) at 1:100. Similar steps were used in this panel for staining PD-L1, cytokeratin, and DAPI.

Briefly, DNA was extracted from the patient's peripheral blood as source of germinal DNA and FFPE tumor samples using a BiOstic FFPE Tissue DNA Isolation Kit (MO BIO Laboratories #12250-50) with a modified protocol. Genomic DNA was captured on the NimbleGen 2.1M human exome array and subjected to 74 base paired-end reads on the Illumina HiSeq 2000 instrument as described (17). Sequence reads were mapped to the reference genome (hg19) using the ELAND program. Reads outside the targeted sequences were discarded, and statistics on coverage were collected from the remaining reads using in-house Perl scripts as previously described (17).

Briefly, genomic DNA derived from the patient's tumor was tested by real-time PCR using a Custom TaqMan Assay (ID APTZ94T, Life Technologies Inc.) designed to specifically amplify the region of interest. The forward primer (CTTCAAGCAGGGATTCTCAACCT) was designed on the CD274 sequence while the reporter sequence (ATGGAACCCTTTTAGAACCC) and the reverse primer (GCCCCGATACCAGTATGACTTG) were designed to include part of the CD274 sequence and part of the PLGRKT insertion. The detection of the genetic rearrangement in the tumor sample was confirmed by Sanger sequencing.

The patient-specific four-digit HLA class I allele genotype was assessed by ATHLATES (18) in silico using whole-exome sequencing (WES) data. All somatic mutations including nonsynonymous mutations, frameshift insertions, and deletions were translated into 17-mer polypeptides flanking the mutant amino acid. The binding affinity of wild-type and mutant nonamers to the patient-specific HLA class I allele was predicted using NetMHCcons (19) algorithms. Nonamers with IC₅₀ below or equal to 500 nmol/L were further evaluated for the recognition by the T-cell receptor using class I immunogenicity (20) resulting in putative neoantigens. The putative neoantigens were then classified as a strong binder (IC₅₀ < 50 nmol/L), intermediate binder (50 nmol/L < IC₅₀ < 150 nmol/L), or weak binder (150 nmol/L < IC₅₀ < 500 nmol/L).

The patient is an 80-year-old Caucasian woman originally diagnosed in June 2014 with an advanced-stage, high-grade, ovarian carcinoma with clear cell features. The initial CAT scan revealed extensive abdominal carcinomatosis, ascites, pelvic lymphadenopathy, and a CA125 tumor marker elevated at 1,810 U/mL (normal value less than 35 U/mL). After a diagnostic biopsy she received six cycles of neoadjuvant chemotherapy (i.e., carboplatin AUC 5 and paclitaxel 175 mg/m²) followed by a radical surgical procedure. Persistent widespread abdominal carcinomatosis was found at the time of the interval debulking surgery with gross residual disease present in the omentum (nodules up to 2.9 cm), ovaries, fallopian tubes, and pelvic lymph nodes. Final pathology was consistent with the prechemotherapy biopsy and demonstrated a high-grade adenocarcinoma of Mullerian origin with clear cell features most likely started in the ovaries and treatment effect. Because of symptomatic neuropathy, postoperatively she received six additional cycles of carboplatin (AUC 5) and Topotecan (2 mg/kg days 1 and 15) in combination with bevacizumab (10 mg/kg, days 1 and 15). The patient remained in remission until March 2015, when an MRI of the pelvic and abdomen followed by a PET/CT imaging revealed the presence of recurrent disease in the pelvis (i.e., a mass of 2.6 cm above the vaginal cuff). She was dispositioned to receive vaginal cuff brachytherapy followed by pelvic radiation in the form of 28 fractions completed on July 7, 2015. Unfortunately, a PET/CT performed on October 2015 demonstrated persistent/recurrent disease with interval increase in size and hypermetabolism of the pelvic lesion involving the superior aspect of the vaginal cuff. She was therefore dispositioned to receive dose-dense paclitaxel protein-bound particles (abraxane), initially at the dose of 100 mg/m² on days 1, 8, and 15 and subsequently dose-reduced to 80 mg/m² of a 28-day cycle in combination with bevacizumab (10 mg/kg, days 1 and 15). She did well until June 2016 when a pelvic exam followed by a CAT scan guided biopsy of the enhancing 3.1- × 4.8-cm mass along the superior/cephalad border of the vaginal cuff confirmed the presence of a metastatic/recurrent adenocarcinoma histological consistent with the original high-grade mullerian tumor. On August 2016, she was enrolled at Yale University within a phase II study of pembrolizumab entitled “Efficacy and Safety Study of Pembrolizumab (MK-3475) in Women With Advanced Recurrent Ovarian Cancer” (MK-3475-100/KEYNOTE-100). Pembrolizumab was administered at a flat dose of 200 mg i.v. on day 1 of each 3-week cycle. CA125 tumor marker at this time was found elevated at 230 U/mL. A CAT scan performed on October 2016 (i.e., after three pembrolizumab infusions) demonstrated stable disease (i.e., larger diameter of the pelvic/vaginal mass stable at 5.4 cm vs. 5.6 cm at baseline; Fig. 1, top). Strikingly, at the time of the second CAT scan imaging obtained 4 months from study initiation (i.e., December 2016, Fig. 1, left bottom), a remarkable complete response by RECIST v1.1 criteria was noted to the immune checkpoint inhibitor. Confirmatory CAT scans obtained 8 weeks later (i.e., February 2017, Fig. 1, right bottom) and in April and June 2017 (data not shown) demonstrated a sustained complete response of the large pelvic/vaginal tumor with no new lesions. The patient continued to improve in her performance status during treatment with the normalization of the CA125 tumor marker. Per protocol, pembrolizumab treatment has been discontinued after the completion of a total of nine cycles of treatment (24 weeks) and two cycles after the confirmatory complete response (i.e., CAT scan imaging). The patient's remarkable clinical response remains unchanged at the time of the writing of this report with no side effects reported to date and the patient experiencing good quality of life.

To gain additional knowledge in the genetic and pathologic characteristics of the patient's tumor demonstrating such exquisite sensitivity to anti-PD1 treatment, we initially performed NGS testing using a commercially available platform (Foundation One, Foundation Medicine (FM) Inc., a test sequencing the coding region of 315 genes plus introns from 28 genes to a median depth of coverage of greater than 500×; ref. 21) followed by WES (performed at the Center for Genome Analysis at Yale University School of Medicine). We found the ovarian carcinoma to harbor a low/intermediate number of mutations (i.e., tumor mutation burden, TMB = 4.31 mut/Mb, total number of mutations = 164; Fig. 2) when compared with 316 high-grade ovarian cancer samples previously reported by the Cancer Genome Atlas Research Network (mean TMB = 1.84 mut/Mb, total number of mutations = 61; ref. 3) and 114 high-grade ovarian cancer samples recently reported by Patch and colleagues (mean TMB = 5.64 mut/Mb, total number of mutations = 271; ref. 4). Using accurate typing of human leukocyte antigen through exome sequencing, we predicted a total of 40 potential neoantigens harbored in the tumor including 11 strong, nine intermediate, and 20 weak immunogenic antigens (Fig. 2). Somatic mutations (SNV), loss of heterozygosity (LOH), and copy-number variations (CNV) in the whole-exome-sequenced tumor are presented in Fig. 3. Nonsynonymous mutations (i.e., SNV) in the TP53 and MLH1 genes (the later not causing microsatellite instability, data not shown) and a stop mutation in TGF-β (a key player cytokine during host–immune system interaction) amongst other mutations were detected (Fig. 3). Gain of function in known oncogenes (i.e., the PIK3CA gene on chromosome 3 and in c-MYC on chromosome 8) and in BCL2L1, a gene endowed with antiapoptotic properties and a crucial role in controlling survival during lymphocyte development and following B- and T-cell activation, was also identified (Fig. 3). Of great interest, both FM and WES testing demonstrated a peculiar PD-L1 gene rearrangement (i.e., chr9 duplication event affecting the 5′ end of PD-L1, breakpoint in exon 7 [3′ untranslated region (3′-UTR); Fig. 2E]. As described in the Materials and Methods section, using primers specific for the predicted fusion, RT-PCR, and Sanger sequencing, we confirmed the detection of the genetic rearrangement in the tumor sample (Supplementary Fig. S1). The event consists of a translocation/insertion of a 32 nucleotide fragment from the exon 5 of the PLGRKT gene (plasminogen receptor, C-terminal lysine transmembrane protein) in chr9:5361091–5361122 of the hg19 assembly (Fig. 2E), into 3′UTR of the PD-L1 gene. PLGRKT, however, remains intact after the tandem duplication event. These genetic characteristics were found to differ from previously demonstrated gain of function of the PD-L1 gene secondary to gene amplification and, importantly, have recently been described in a subset of T-cell leukemia and B-cell lymphoma to lead to a marked elevation of PD-L1 extracellular domain secondary to the stabilization of the PD-L1 transcripts through the truncation of the 3′ -UTR of the PD-L1 gene (22). To confirm the potential high expression of PD-L1 receptor on the surface of the ovarian tumor and thoroughly evaluate the presence or absence of lymphocytic infiltration we next performed PD-L1 IHC testing and quantitative multiplex fluorescence analysis in the tumor tissue. We found strong tumor cell expression for PD-L1 by IHC as well as heavy intratumoral T lymphocytic (i.e., CD8 and CD4) and to a lower extent B-cell (CD20) and histiocytic infiltration (i.e., CD68) by hematoxylin–eosin (H&E) stain and immunostains (Fig. 4). Using an immunofluorescence multiplexing panel (16), we confirmed colocalization of PD-L1 with cytokeratin-positive tumor cells and negligible to no PD-L1 expression in infiltrating CD68⁺ macrophages and CD8⁺ T lymphocytes (Fig. 5). Finally, we compared PD-L1 expression in the patient's tumor with PD-L1 expression in 79 cases of high-grade serous ovarian carcinoma available to our laboratory from a recently characterized tissue microarray (15). As representatively shown in Fig. 6, out of the 79 tumors, only 15 (19%) showed focal PD-L1 expression in tumor cells while the remaining 64 cases (81%) were negative for PD-L1. None of the cases present in the TMA showed diffuse immunoreactivity for PD-L1 as described in Fig. 4 for the patient with ovarian cancer harboring the PD-L1–genetic rearrangement.

Recurrent ovarian carcinoma resistant to salvage chemotherapy treatments remains an incurable disease. In the last few years, multiple preclinical and clinical studies have demonstrated that one mechanism by which multiple human cancers evade host immunity is via upregulation of PD1 ligand (PD-L1) and its interaction with the PD1 ligand on antigen specific T-lymphocytes (5, 6). Consistent with this view, a variety of human tumors have demonstrated sensitivity to immune checkpoint inhibitor antibodies targeting PD1 or PD-L1 in vivo and, accordingly, pembrolizumab (Merck & Co., Inc.), nivolumab (Bristol-Myers Squibb Company), and atezulumab (Genentech/Roche) have recently received approval by the FDA for the treatment of multiple human malignancies. Unfortunately, however, a large number of patients treated do not respond to immune checkpoint treatment and little is currently known about the potential genetic mutations regulating response to the blockade of the PD1/PD-L1 axis.

We describe a heavily pretreated patient with ovarian cancer with recurrent disease experiencing a complete and sustained response to pembrolizumab. Remarkably, this ovarian cancer was found exquisitely sensitive to the immune checkpoint blockade regardless to its high resistance to chemotherapy and radiation as demonstrated by the complete disappearance of a large recurrent tumor mass at 16 weeks from treatment initiation confirmed with multiple serial CAT scans. Of interest, WES results demonstrated the tumor to have a relatively low number of mutations when compared with melanoma, lung cancer, ovarian cancer (4), or gynecologic and non-gynecologic tumors characterized by high microsatellite instability (i.e., MSI-H endometrial and colorectal tumors; ref. 23). These data combined with our results predicting at least 40 potential neoantigens in the WES tumor suggest that a high tumor mutation burden may not represent a mandatory requirement for clinical response to immune checkpoint inhibitors in patients with ovarian cancer.

Of great interest, we found the tumor to harbor a structural variation in the PD-L1 gene recently reported in about 27% of adult T-cell leukemia/lymphoma, 8% of B-cell lymphoma and 2% of patients with stomach cancer to lead to marked elevation of PD-L1 transcripts secondary to the truncation of the 3′-UTR of the PD-L1 gene (22). Importantly, this rare PD-L1 mutation has recently been reported to provide high sensitivity to immune checkpoint inhibitor in preclinical animal models and speculated to potentially confer high susceptibility to blockade of the PD-1/PD-L1 axis in human patients (22). Consistent with these preclinical data in xenografted animals, we found our patient tumor to express high levels of the PD-L1 receptor and to be characterized by a high intratumoral CD8, CD4, and to a lower extent CD20 B lymphocytic and macrophage (i.e., CD68) infiltration by IHC staining. Taken together, the relative low tumor mutation burden, strong PD-L1 surface expression and heavy inflammatory infiltration of this ovarian cancer suggest the PD-L1 structural variation to be the reason of its exquisite sensitivity to immune checkpoint treatment. Mutations in other immunologically relevant genes such as TGF-beta and CD27 were also detected in the tumor. The possible influence of these mutations on the exceptional response of this patient to pembrolizumab are currently unclear.

In conclusion, to our knowledge, this is the first report demonstrating a remarkable clinical activity of an immune checkpoint inhibitor in a patient harboring a rearrangement in the PD-L1 gene. While the PD-L1 structural variation we found in this patient with ovarian cancer seems to be a rare event in human tumors with only 31 cases expressing 3′-UTR-truncated PD-L1 transcript out of 10,210 cancer samples from 33 tumor panel analyzed by TCGA reported to date (22), the exquisite sensitivity of this patient to pembrolizumab suggests that such PD-L1 structural variants may serve as genetic marker for the identification of human cancers selectively upregulating PD-L1 in order to evade antitumor immunity. More importantly, these peculiar genetic characteristics may be able to identify human tumors that regardless to their high resistant to surgery, chemotherapy, and radiation and a low mutation burden may still be highly susceptible to immune checkpoint PD1 blockade.

L. Gay and J.A. Elvin hold ownership interest (including patents) in Foundation Medicine. D.L. Rimm reports receiving commercial research grants from Navigate, other commercial research support from Perkin Elmer, and is a consultant/advisory board member for Cell Signaling Technology. E. Ratner is a consultant/advisory board member for Tesaro. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Liu, P.E. Schwartz

Development of methodology: S. Bellone, Y. Liu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Bellone, N. Buza, Y. Liu, E.S. Ratner, A.D. Santin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Bellone, N. Buza, J. Choi, L. Zammataro, L. Gay, J. Elvin, D.L. Rimm, Y. Liu, P.E. Schwartz, A.D. Santin

Writing, review, and/or revision of the manuscript: S. Bellone, N. Buza, J. Elvin, Y. Liu, P.E. Schwartz, A.D. Santin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Bellone

This work was supported in part by grants from NIH U01 CA176067-01A1, the Deborah Bunn Alley Foundation, the Tina Brozman Foundation, the Discovery to Cure Foundation, and the Guido Berlucchi Foundation to Alessandro D. Santin. This investigation was also supported by NIH research grant CA-16359 from the NCI.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.