Introduction

The last decade has demonstrated that immunotherapies can be robust and effective treatments for several previously untreatable cancers. Adoptive cell therapy (ACT) of T cells engineered to express a tumor antigen-specific T-cell receptor (TCR) allows for the generation of large numbers of T cells that target a uniform antigen and has shown clear evidence of antitumor efficacy in patients with metastatic solid tumors.1 2 Initial work used this approach against melanoma-specific antigens like MART-1 and gp100 to induce tumor regression,1–3 though more recently, the cancer testis antigen NY-ESO-1 has been shown to be an effective target for a variety of solid tumors.3–7

Given the dismal overall survival in patients with advanced and metastatic sarcomas in response to conventional therapies, a number of recent studies have attempted to use immunotherapies to improve treatment responses.6–9 ACT targeting of NY-ESO-1 has been one of the most promising avenues as multiple studies have shown that it is possible to induce tumor regression in a large fraction of patients.6–8 However, the antitumor activity induced tends to be transient, with most patients progressing within months of initiation of treatment. This underscores the need to better understand the mechanisms driving treatment resistance to inform the development of treatment approaches that can lead to more durable responses.

Here we show for the first-time disease relapse in the setting of loss of NY-ESO-1 expression in sarcoma. The patient described was treated with NY-ESO-1-targeted transgenic ACT with dendritic cell (DC) vaccination and nivolumab. The addition of antigen in the form of a vaccine has been shown in animal models to stimulate T-cell expansion in vivo, in turn, enhancing the effectiveness of ACT.10 11 Indeed, our group has shown that, for both MART-1 and NY-ESO-1, addition of a peptide-pulsed DC vaccine to ACT is both safe and feasible in humans.3 6 PD-1 blockade with nivolumab was added to our dual-cell therapy clinical protocol6 in an attempt to increase duration of the antitumor response, as preclinical models have shown PD-1 blockade enhances antitumor efficacy of transgenic ACT.12 13 The patient had an initial response to therapy but later developed disease progression. This tumor growth coincided with loss of NY-ESO-1 expression in tumor tissue obtained at time of relapse. This novel mechanism of immune escape in sarcoma demonstrates a novel vulnerability in cellular therapy approaches.

MethodsStudy overview

The schedule of events for the patient receiving the NY-ESO-1-specific TCR transgenic lymphocytes with NY-ESO-1 peptide-pulsed DC vaccination and concurrent PD-1 blockade is outlined in figure 1A. Briefly, a pretreatment biopsy was obtained, and on day 0, the patient underwent ACT of 109 transgenic TCR lymphocytes as an intravenous infusion of freshly prepared cells. The patient received intravenous nivolumab (3 mg/kg) over 60 min on day 1 with treatment repeating every 2 weeks, along with intradermal administration of NY-ESO-1157–165 peptide-pulsed DCs and low-dose interleukin (IL)-2 therapy (500 000 IU/m2, subcutaneous) two times a day for up to 14 doses. A repeat biopsy was scheduled for ~3 months following treatment initiation. Standard supportive care included antibiotics and filgrastim, and blood product transfusions were administered as needed and adverse events were analyzed following NCI CTCAE V.3.0.

Figure 1

ACT of NY-ESO-1 transgenic T cells and given with DC vaccination and concurrent anti-PD-1 therapy resulted in transient antitumor activity. (A) Overview of the clinical and sample collection timeline for the analysis performed in this study. Dosing schedule for nivolumab was every 2 weeks and is represented by red arrows. (B) Pretreatment and (C) post-treatment PET-CT images from the patient showing evidence of initial antitumor activity (decreased 18F-FDG uptake). Arrows point to the pulmonary nodule. (D–G) CT scans of the inguinal mass (arrows point to the mass) are shown at (D) baseline before ACT, (E,F) during tumor regression on days +22 and +61 of ACT, and (G) during tumor progression on day +112 of ACT. ACT, adoptive cell therapy; Cy, cytarabine; DC, dendritic cell; Flu, fludarabine; IL, interleukin; PET, positron emission tomography; TCR, T-cell receptor; 18F-fluorodeoxyglucose, 18F-FDG.

MHC dextramer immunological monitoring

Detection of NY-ESO-1 TCR expression using fluorescent major histocompatibility complex (MHC) dextramer analysis for NY-ESO-1 (Immudex; HLA-A-*0201, SLLMWITQV) was performed on cryopreserved peripheral blood mononuclear cells (PBMCs) collected at different time points, as described previously.3 6

Flow cytometry phenotyping analysis

PBMCs were centrifuged (500×g for 5 min), resuspended in 100 µL of adult bovine serum (Omega Scientific, Tarzana, California, USA) and stained with preconjugated fluorescent antibodies for flow cytometry and acquired on an Attune flow cytometer (Thermo Fisher). Detailed description of the antibodies and staining is described in our previously published work14 and can be found in the online supplemental methods. Gating strategy can be found in online supplemental figure 1A.

Deep sequencing of TCRβ alleles

Genomic DNA for TCRβ sequence analysis was isolated, and productive TCRβ sequences were identified from formalin-fixed, paraffin-embedded tumor biopsies, patient-matched infusion products, and postinfusion PBMCs as previously described.6

RNA sequencing and analysis

mRNA libraries were generated using the KAPA Stranded mRNA Kit and were sequenced on the Illumina HiSeq 3000 platform (2×150 bp). Sequencing data were aligned to the human reference genome (GRCh38) using HISAT2, and gene expression quantification was performed using Stringtie and the Ensembl reference transcriptome V.84. Differential gene expression was performed with GFold, and TRUST4 was employed for immune repertoire analysis.

Immunofluorescence experiments

Multiplex immunofluorescence analysis was conducted on baseline and on-treatment biopsies. Full details of antibodies used, antigen retrieval techniques, antibody dilutions and incubation times can be found in online supplemental table 1 and in the online supplemental methods.

Promoter methylation

Post-treatment tumor DNA (350 ng) was used for bisulfite conversion (Zymo D5001). Bisulfite converted DNA was used as template for PCR amplification of the NY-ESO-1 promoter region (for primers, see online supplemental methods). The PCR product was cloned into a pCR4-TOPO linearized vector via the TOPO TA Cloning kit (Thermo K4575J10) and transformed into TOP1 chemically competent bacteria. Plasmid DNA was isolated from 20 colonies, 19 of which had an identity of >25% to the target sequence and were thus used for analysis.

ResultsClinical course

The patient was in their 40s and diagnosed with a high-grade undifferentiated pleomorphic sarcoma (HGUPS) of the right thigh with pulmonary metastases. The sarcoma was refractory to standard-of-care chemotherapy (ifosfamide and doxorubicin), radiation therapy as well as radical resection of the right thigh mass. The patient was HLA-A*0201-positive, and the tumor was NY-ESO-1-positive (figure 1A). At the time of enrollment in the clinical trial, the baseline positron emission tomography (PET)-CT showed disease progression in the lungs and thigh. Adoptive cell transfer of NY-ESO-1 TCR-transduced T cells occurred on April 11, 2018. Following ACT, the patient received a combination of low-dose IL-2, nivolumab, and DC vaccine loaded with NY ESO-1 peptide per study protocol (figure 1A). IL-2 administration was held two times for sinus tachycardia, but treatment was otherwise well tolerated. PET-CT at day 22 showed partial response by RECIST V.1.1. Notably, there was uniform improvement in pulmonary disease burden (figure 1B,C) and significant improvement in the size of the right lateral thigh and inguinal mass (figure 1D,E). Repeat CT scan in June showed continued response at the right proximal thigh/inguinal mass (figure 1F), and the patient was able to return to work. A post-treatment biopsy was performed on day +76. Unfortunately, the patient developed new right thigh pain near the end of July, at which point a CT scan showed progressive disease (figure 1G). The patient continued nivolumab through August, though they were removed from the trial and started on pazopanib following continued progression. Pazopanib failed to stop further disease progression, and the patient passed away in early 2019.

Immunophenotyping of the peripheral T cells shows predominantly effector memory (EM) phenotype over time

Consistent with previous work from our laboratory for patients on different ACT protocols,3 6 TCR transgenic cell frequency was determined to peak within 2 weeks of ACT, after which the percentage and absolute number of TCR transgenic cells in peripheral blood decreased (figure 2A). Less differentiated memory T cells are the preferred population, given the fact they have superior in vivo expansion, persistence, and antitumor activity.15 16 The infusion products for the patient contained large proportions of effector memory (EM; CCR7−/CD45RA−) and effector memory re-expressing CD45RA (EMRA; CCR7−/CD45RA+) (figure 2B) cells. However, in contrast to prior work from our laboratory,6 the proportion of these less differentiated phenotypes persisted throughout treatment for this patient instead of progressing to a more terminally differentiated effector phenotype (figure 2B). This persistence of less differentiated phenotypes was also noted when analyzing the total CD8+ T-cell population (online supplemental figure 2A).

Figure 2

Immunophenotyping of peripheral T cells shows a predominantly EM phenotype over time and confirms nivolumab-mediated PD-1 blockade. (A) Postinfusion peripheral blood levels of NY-ESO-1 TCR transgenic CD3+ cells over time. (B,C) Immunophenotyping of adoptively transferred TCRs overtime the CD8+ population. T-cell compartments analyzed include naïve, CM, EM, and EMRA. (C) analysis of T cell exhaustion among total CD8+ and NY-ESO-1 TCR transgenic CD8+ populations. exhaustion determined by CD39/PD1 positivity. (D) PD-1-receptor occupancy analysis. IgG4 positivity serves as a measure of nivolumab binding. PD1 positivity represents residual PD1 not bound by nivolumab. (E) same as in (D) excluding PD1−/IgG4− population. CM, central memory; EM, effector memory; EMRA, effector memory cells re-expressing CD45RA; TCR, T-cell receptor.

Peripheral T cells showed no evidence of exhaustion and confirmed nivolumab-mediated PD-1 blockade

A challenge associated with ACT involves exhaustion of the adoptively transferred T-cell population due to chronic antigen stimulation. The adoptively transferred T cells in this work showed no evidence of significant exhaustion throughout treatment (figure 2C). To confirm PD-1 blockade on adoptively transferred lymphocytes, PD-1-receptor occupancy was estimated by analyzing the percentage of PD-1/IgG4 on CD8+ T cells (figure 2D) by means of flow cytometry. Nivolumab is a human IgG4 monoclonal antibody, and as such, surface IgG4 binding can be used as a measure of nivolumab saturating the PD-1 receptor.17 By day 4 of treatment, a significant percentage (~20%) of NY-ESO-1 CD8+ T cells (figure 2D,E) and the overall CD8+ T-cell population (online supplemental figure 2B,C) had evidence of nivolumab binding as measured by IgG4 positivity (figure 2D,E). The PD-1−/IgG4+ population was significantly greater than the PD-1+/IgG4+ population throughout the course of treatment. Additionally, the PD-1−/IgG4+ population was significantly greater than the PD-1+/IgG4- population at all time points while the patient was enrolled on the trial except for days 1 and 161 (figure 2D,E). The patient received the first nivolumab dose on day 1, and thus, at this time, there were presumably low levels of circulating antibody. By day 161, the patient had been removed from the trial and had not received a nivolumab dose in over 3 weeks. As such, our data indicate limited exhaustion and effective and efficient binding of PD-1 by nivolumab on adoptively transferred lymphocytes.

Nivolumab and NY-ESO-1 transgenic TCR track to the site of the tumor

We next used immunohistochemistry (IHC) to confirm the presence of tumor-infiltrating CD8+ T cells at the tumor site in the post-treatment biopsy. We found that CD8+ T cells were present at the site of the tumor, and further, IgG4 partially colocalized with CD8 staining, indicating nivolumab binds CD8 T cells at the tumor (figure 3A). We then stained for PD-L1 with TLE1 serving as our tumor marker to confirm strong expression of PD-L1 by tumor cells (figure 3B). Finally, we stained for PD-1, which was negative, allowing us to conclude effective and efficient binding of PD-1 by nivolumab at the tumor site (figure 3C). To confirm the presence of the transgenic TCR at the tumor site, we performed TCRseq of the on-treatment biopsy sample. TCRseq revealed that the major clone in the on-treatment biopsy is the NY-ESO-1 transgenic TCR (figure 3D). Analysis of clinical transgenic ACT products in vivo has shown that while the transgenic TCR may exist at the DNA level, epigenic suppression can lead to profound decreases in the levels of TCR RNA and protein expression.18 To confirm RNA expression of the NY-ESO-1 TCR, we analyzed on-treatment bulk RNA sequencing of the tumor and employed the TRUST4 algorithm19 to reconstruct the immune repertoire from bulk RNA-seq samples. Both the α-chain and β-chain of the NY-ESO-1 transgenic TCR were the most abundant in the sample, making up nearly half of the TCR repertoire (figure 3E,F). Of note, no NY-ESO-1 TCR was detected in the pretreatment biopsy (online supplemental figure 3AB). Further, repertoire clonality dramatically increased (online supplemental figure 3CD) and repertoire entropy significantly decreased in the post-treatment sample (online supplemental figure 3EF). Thus, we conclude that adoptively transferred NY-ESO-1 T cells localized to the tumor and continued to express their transgenic TCR.

Figure 3

Nivolumab and NY-ESO-1 transgenic TCR track to the site of the tumor. (A) IHC of the patient’s tumor at the time of biopsy for CD8, IgG4, and 4′,6-diamidino-2-phenylindole (DAPI). (B) Same as in (A) but for sarcoma marker TLE1, PD-L1, and DAPI. (C) Same as in (A) but for sarcoma marker PD-1 alone, a combination of CD8, IgG4, TLE1, PD-L1, PD-1, or the same combination with DAPI. (D) Post-therapy TCRβ repertoire as determined by deep sequencing of TCRβ alleles using genomic DNA from post-therapy tumor biopsy. Blue represents the frequency of NY-ESO-1 TCRβ alleles; beige represents the frequency of all other sequenced TCRβ alleles. (E) Post-therapy TCRα repertoire as determined by repertoire reconstruction using bulk RNA-sequencing from post-therapy tumor biopsy. Blue represents the frequency of NY-ESO-1 TCRα alleles; beige represents the frequency of all other sequenced TCRα alleles. (F) Same as in (E) except for TCRβ repertoire. IHC, immunohistochemistry; TCR, T-cell receptor.

NY-ESO-1 antigen expression is lost following ACT treatment in the setting of promoter methylation

The patient’s tumor was noted to be NY-ESO-1-positive on enrollment for the trial, and indeed, our pretreatment RNA-sequencing sample confirmed NY-ESO-1 was significantly expressed (figure 4A). However, gene expression analysis of the post-treatment biopsy showed a transcript per million value less than 1, indicating a complete loss of NY ESO-1 expression (figure 4B,C and online supplemental figure S4A). In fact, NY-ESO-1 was one of the most downregulated genes when comparing expression both pretreatment and post treatment (online supplemental figure 4B). This loss of NY-ESO-1 was confirmed at the protein level as IHC was positive in the tissue from the pretreatment biopsy (figure 4D) and negative in the tissue from the post-treatment biopsy (figure 4E). Given that DNA demethylation is a primary mechanism underlying aberrant re-expression of cancer–testis antigens like NY-ESO-1,20 we hypothesized that promoter methylation might play a role in the post-treatment loss of NY-ESO-1 expression. To probe for this, we performed bisulfite sequencing on a region of the NY-ESO-1 promoter (figure 4F). We found that the NY-ESO-1 promoter was extensively methylated in the post-therapy sample (figure 4G). This strongly implies that disease relapse occurred in the setting of loss of NY-ESO-1 expression driven by extensive methylation of its promoter region during therapy.

Figure 4

NY-ESO-1 antigen expression is lost following ACT treatment. (A) Histogram of mapped reads corresponding to NY-ESO-1 expression in the pretreatment biopsy sample. (B) Same as in (A) but for two post-treatment biological replicates. Scale is normalized to pretreatment maximum. (C) NY-ESO-1 expression (TPM) from the pretreatment biopsy and both post-treatment biological replicates. (D) IHC for NY-ESO-1 using pretreatment tumor biopsy. (E) Same as in (D) but for post-treatment tissue. (F) Genomic structure of the NY-ESO-1 promoter region and transcriptional start site. Bent arrow represents the +1 TSS; blue arrow indicates CDS; blue arrowheads represent bisulfite primer sites; upright lines represent CpG dinucleotides. (G) Analysis of methylation status at CpG dinucleotides in the NY-ESO-1 promoter region. CpG sites assayed for methylation are depicted (top) and numbered relative to the TSS. Methylation level for the aggregate experiments quantified for each site (middle). Schematic depiction of the aggregate methylation level depicted (bottom). ACT, adoptive cell therapy; CDS, coding sequence; CpG, cytosine-phosphate-guanine; IHC, immunohistochemistry; TPM, transcript per million; TSS, transcription start site.

Discussion

While transgenic ACT has been demonstrated to be a potent form of cancer immunotherapy, clinical responses are often transient with patients frequently developing progressive disease despite an initial clinical response to treatment. Here, we describe novel mechanism of immune escape in sarcoma whereby NY-ESO-1 expression was lost in response to TCR transgenic ACT with DC vaccination and PD-1 blockade.

With respect to this case, the patient with HGUPS had an initial partial response to therapy with uniform improvement in pulmonary disease burden and significant improvement in size of the right lateral thigh and inguinal mass. The biopsy showing NY-ESO-1 loss was performed at the end of June, a few weeks prior to the development of significant disease progression. The timing of disease progression in relation to the biopsy suggests loss of NY-ESO-1 expression as the primary mechanism for resistance to therapy. While loss of expression has been documented in NY-ESO-1-targeted melanoma trials,21 this is a heretofore unexplored mechanism of resistance in sarcoma. It is possible that this loss of expression is driven by evolution of a clone that initially expressed NY-ESO-1. However, it is also possible that the adoptively transferred lymphocytes were very effective at targeting immunogenic, NY-ESO-1-expressing clones, leaving behind a population of non-immunogenic clones lacking NY-ESO-1 expression. As to why this mechanism of resistance has not previously been seen in sarcoma, it is important to note that prior trials have targeted synovial sarcoma (SS).6–8 It might be that the characteristic SS18/SSX fusion mediating the epigenetic dysregulation in SS drives more durable expression of NY-ESO-1.22 Alternatively, the oncogenic program in SS has been shown to actively repress immunogenicity,23 and thus, it is possible adoptively transferred lymphocytes are less effective at completely eliminating NY-ESO-1-expressing populations in SS.

The finding that antigen loss can lead to disease progression in sarcoma highlights the importance of developing new technologies that prevent this mechanism of therapy resistance. One possible approach would be combinatorial targeting of antigens, an approach most notably being performed with chimeric antigen receptor T-cell (CAR-T) therapies in leukemia.24 Another approach involves inducing durable expression of NY-ESO-1. As expression of NY-ESO-1 is known to be regulated by methylation,20 the addition of a hypomethylating agent like the azanucleosides azacytidine and decitabine could induce durable expression of NY-ESO-1 in tumors exposed to NY-ESO-1-targeted therapies and, as such, could prevent or delay disease progression. Similarly, this approach could be pursued by other therapies which rely on a single molecular target to maintain tumor homogeneity.

With respect to the adoptively transferred lymphocytes, less differentiated memory T cells are the preferred population due to the fact they have been shown to have superior in vivo expansion, persistence, and antitumor activity when compared with the more terminally differentiated effector T-cell subsets.15 16 In this patient, we found they had a predominantly EM phenotype over time. This contrasts with our previous experiences with transgenic ACT in combination with DC vaccination where adoptively transferred cells displayed a shift toward more terminally differentiated phenotypes.3 6 It is possible that integration sites within the host cell’s genomes might have affected the differentiation of adoptively transferred lymphocytes, a phenomenon previously reported in the generation of CAR-T cells.25 However, it has previously been shown that there is a marked accumulation of CD8+ EM cells in lymphoid organs and tissues of PD-1-deficient mice,26 and further, patients who responded to PD-1 blockade had tumor-infiltrating lymphocytes (TILs) that displayed a predominantly EM phenotype.27 Given the addition of nivolumab to the regimen for the patient in this work, the sustained predominately EM phenotype is consistent with the anticipated mechanism of action of therapy.

In conclusion, ACT of NY-ESO-1 TCR transgenic T cells combined with DC vaccination and anti-PD-1 therapy resulted in transient antitumor activity. Adoptively transferred lymphocytes displayed a predominantly EM phenotype without progression to exhaustion/terminal differentiation. On-treatment biopsy showed complete loss of NY-ESO-1 expression with extensive methylation of the promoter sequence. This work highlights a new mechanism of treatment resistance in sarcoma that must be considered in efforts for ACT to have a durable antitumor response.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study involves human participants who gave informed consent to participate in the study before taking part. The patient was non-randomly enrolled in the trial after signing a written informed consent approved by the UCLA Institutional Review Board (#15–0 01 433) under an investigational new drug (IND#15167) for the NY-ESO-1 T-cell receptor.

Acknowledgments

The authors extend a thank you to the patient presented in this study. Biopsy tissue processing was performed at the University of California, Los Angeles (UCLA) Translational Pathology Core Laboratories, and sequencing data were generated in the UCLA Technology Center for Genomics and Bioinformatics.