Response to Immune Checkpoint Inhibition in Two Patients with Alveolar Soft-Part Sarcoma

Alveolar soft-part sarcoma (ASPS) is a rare mesenchymal tumor that typically arises in adolescent and young adults (1). Tumors usually present as a slow-growing mass but metastatic disease is frequently identified at diagnosis (2). Morphologically, ASPS is characterized by a stereotypic pattern of nested cells separated by fibrous septae, which are often associated with a pseudoalveolar architecture. Molecularly, tumors are characterized by an unbalanced translocation between Xp11 and 17q25, leading to the formation of an ASPSCR1–TFE3 chimeric transcription factor (3). Despite slow tumor growth, the 5-year overall survival is poor at 20% (4, 5).

Given the rarity of ASPS, the literature is limited to small case series. Aside from surgery, there are few therapeutic options, with data suggesting limited value of systemic therapies (6, 7). Antiangiogenic agents have been investigated (8, 9) but the benefits are typically not durable. Given the need for effective therapy, we enrolled 2 patients with ASPS in early-phase trials utilizing mono or combination therapy with durvalumab, a monoclonal antibody (mAb) delivered intravenously that blocks the interaction of programmed death ligand 1 (PD-L1) and PD-1. In parallel, we conducted a multipronged preclinical institutional investigation into ASPS, including the samples from the above 2 patients.

Tumors classified as ASPS were identified from an institutional database (identified since 1990) following REB approval. A pathology re-review was conducted by a dedicated sarcoma pathologist (BCD) based on the criteria detailed in the WHO Classification of Tumors of Soft Tissue and Bone.

IHC was performed using standard methods. The Dako Autostainer Link 48 was used for CD3 (clone: F7.2.38; Dako), CD4 (clone: SP35; Roche), CD8 (clone: C8/144B; Dako), CD20 (clone: L26; Dako), MLH1 (clone: ESO5; Novocastra), MSH2 (clone: FE11; Dako), MSH6 (clone: SP93; Cell Marque), and PMS2 (clone: MRQ-28; Cell Marque). The Ventana Benchmark ULTRA was used for PD-1 (clone: NAT105; Roche) and PD-L1 (clone: SP263; Roche).

All slides were quantified by a subinvestigator (B.Y. Lau) with 10% selected for independent secondary quantification (BCD). Tumor-infiltrating lymphocytes (TIL) were assessed using a semiquantitative 4-tiered scale: 0 (no lymphocytes); 1 (1–10/HPF); 2 (11–50/HPF); 3 (51–100/HPF); 4 (>100/HPF) (10). In addition, TIL staining with PD-L1 and PD-1 was also scored, whereby the overall percentage of positive cells on the entire slide was quantified. This latter method was also used to assess PD-L1 tumor staining.

Tumor DNA extracted from paraffin-embedded samples was prepared using Agilent's exome enrichment kit (Sure Select V5). DNA samples were sequenced on an Illumina HiSeq2500 for whole-exome sequencing and an Illumina HiSeq X for whole-genome sequencing. Base calls and intensities from the Illumina HiSeq 2500 were processed into FASTQ files using CASAVA and/or HAS. The paired-end FASTQ files were aligned to UCSC's hg19/GRCh37 with BWA-mem (v.0.7.8). Picard MarkDuplicates (v.1.108) was used to mark PCR duplicates. Aligned reads were realigned for known insertion/deletion events using SRMA and/or GATK. Base quality scores were recalibrated using the Genome Analysis Toolkit (v.2.8.1).

Somatic mutations were identified using Mutect and filtered against a panel of healthy control samples. The genome sample was filtered against a panel of noncancer genomes. Mutations were further filtered against common SNPs found in dbSNP (v132), the 1000 Genomes Project (February 2012), a 69-sample Complete Genomics data set (11), and the Exome Sequencing Project (v6500).

Signature analysis was performed using the DeconstructSigs R package.

MSI analysis was carried out using the Promega Panel, consisting of five specific microsatellite loci (Bat-25, Bat-26, NR-21, NR-24, and Mono-27). These five loci are PCR-amplified from DNA of patient tumor and nonmalignant tissue using fluorescently labeled primers. The lengths of the amplified fragments were measured by fluorescent capillary electrophoresis, which produces an electrophoretogram with peaks representing each loci. Comparing the loci lengths of patient tumor and nonmalignant tissue using Peak Scanner Software, a length change of 3 base pairs (bp) or more (which presents as a separate peak for heterozygous length changes) in the tumor relative to the nonmalignant tissue is considered a significant change. A length alteration in one of the five loci of 3 bp or more indicates that the tumor has a low level of MSI (MSI-low). Length alterations in two or more of the five loci indicate that the tumor has a high amount of MSI (MSI-high). If a significant length change is not seen in any of the five loci, the tumor is designated as microsatellite stable (MSS). In addition, immunohistochemical assessment for loss of protein expression for MLH1, PMS2, MSH2, and MSH6 was conducted.

Interobserver variability was calculated by Cohen kappa analysis. In the event of discordance, a consensus was reached at reevaluation.

Patient 1 was a 22-year-old female who in 2014 identified a growing lesion on her dorsal left hand. Imaging revealed a destructive lesion in the 4th/5th metacarpals and the presence of multifocal pulmonary metastases. Biopsy demonstrated a neoplasm composed of polygonal cells with a nested architecture. IHC was positive for TFE3, and molecular testing confirmed the presence of the ASPSCR1–TFE3 fusion product, consistent with ASPS. The patient was treated with a multikinase inhibitor against Aurora A, VEGFRs, and FGFRs (April 2014–July 2014) with progression at the primary site. Subsequently, preoperative radiation was delivered followed by palliative ray amputation. Upon pulmonary progression, between December 2014 and May 2015, she was treated sequentially without benefit on trials with an OX40 agonist and then a NOTCH inhibitor. After further progression, the patient was treated within the D4190C00010 trial, which investigates the combination of durvalumab and tremelimumab (NCT02261220). Treatment was started on June 18, 2015. On October 5, 2015, CT images demonstrated 30% reduction in the sum of target lesions as per Response Evaluation Criteria in Solid Tumors (RECIST; Fig. 1). After 4 cycles of combination treatment, the patient was admitted with an immune-related (irAE) grade 4 diarrhea requiring treatment with intravenous methylprednisolone. A colonoscopy demonstrated features of colitis requiring a 4-month prednisolone taper and drug hold from August 2015 until March 2016. Despite being off therapy for 8 months, the patient continued to have ongoing response in her target lesions with tumor reduction of 53% on March 10, 2016. Given the absence of residual irAE, the patient was retreated with durvalumab from March 22, 2016, until May 31, 2016, completing the treatment as per protocol. On the last follow-up in January 2018, continual response was demonstrated with 73% tumor reduction despite being off therapy for >18 months.

Patient 2 was a 41-year-old female who had undergone resection and postoperative radiation of ASPS in 1998. In June 2012, she presented with a solitary brain metastasis treated with surgical resection. Histologic assessment of the tumor was consistent with her previous ASPS. Postoperatively the patient received whole brain irradiation, and restaging scans confirmed the presence of multifocal pulmonary metastases. After a period of surveillance, the patient was enrolled in a clinical trial investigating a multikinase inhibitor against Aurora A, VEGFRs, and FGFRs (March 2014 to August 2014). In September 2014, due to disease progression, she was enrolled in MEDI 4736-1108, which is a study investigating durvalumab in subjects with advanced solid tumors (NCT01693562). The patient received durvalumab from October 1, 2014, and showed reduction of target lesions of 58% (Fig. 2). Treatment continued until October 14, 2015 (12 months), whereupon CT scans confirmed pulmonary and nodal progression. The patient was subsequently enrolled in additional clinical trials without response and unfortunately succumbed to progressive disease in April 2016.

Simultaneously, a multipronged investigation study into this sarcoma subtype was conducted. Pretreatment immunoprofiling of both patients 1 and 2 showed scant immune infiltration with tumoral PD-L1 positivity of 2% and 0%, respectively, and immune cell PD-1 positivity of 0% (Fig. 3). We sequenced the whole exomes of both patients' tumors (>80× coverage). Exomic characterization demonstrated no excess mutational burden, with mutational levels in keeping with other fusion-driven sarcomas (Supplementary Fig. S1). We searched for signatures of somatic mutations, using an established nomenclature (12). We identified signatures associated with defective mismatch-repair (MMR) pathways in both patients [patient 1: signature (S) 6 (6%), S15 (9%); patient 2: S26 (35%); ref. Fig. 3]. Point mutation and indel analysis did not identify aberrations in the MMR or polymerase genes. IHC for MLH1, MSH2, MSH6, and PMS2 showed intact IHC expression (Supplementary Figs. S2 and S3).

In order to investigate whether the preclinical features were generalizable, we reviewed a total of 18 patients (including the 2 case reports described here) with ASPS (identified since 1990; Supplementary Table S1).

Paraffin-embedded tumor was available in 12 cases. PD-1 and PD-L1 expression (threshold set at ≥ 1% positive cells) was seen in 17% and 50% of the cases, respectively. CD3⁺, CD4⁺, and CD8⁺ T cell infiltration (threshold set ≥ 11 cells/HPF) was seen in 42%, 25%, and 33% of the cases, respectively (Table 1). Clinical responses to PD-1 blockade (n = 2) did not appear to correlate with PD-1 or PD-L1 expression or with lymphocyte subtypes. IHC for MLH1, MSH2, MSH6, and PMS2 on patients who underwent exomic characterization (patients 1–7) showed intact IHC expression (Table 1).

To determine if the sustained response of ASPS patients to immunotherapy could be attributed to a high mutational load, exomic mutation burden was calculated. As expected for tumors lacking a matched–normal control, approximately 60 germline variants per exome were detected. Tumor mutation burden was calculated from all exonic mutations with a variant allele fraction >0.1 called from a minimum coverage of 30× in the tumor. In order to control for the lack of a matched–normal, we performed the same tumor-only analysis on the fusion-driven Ewing sarcoma. We additionally performed the analysis with a matched–normal for Ewing sarcoma and synovial sarcoma (defined by SS18–SSX fusion). In cases where whole-exome sequencing was conducted, no excess mutation burden in ASPS was identified when compared with tumor-only Ewing sarcoma. When analyzed with a matched–normal, both Ewing and synovial sarcoma had low mutation burdens (Supplementary Fig. S1). These data suggest that ASPS mutation burden is similar to other fusion-driven sarcomas.

Mutational signatures via COSMIC were identified in the MMR deficiency pathway in 5 of 7 cases (patient 1: S6, S15; patient 2: S26; patient 3: S6, S26; patient 4: S6, S26; patient 5: S6, S26), of which 2 of had partial responses to immunotherapy (patients 1 and 2; Fig. 3). The other patients were not treated with immunotherapy. Patient 1 also underwent whole-genome sequencing, which confirmed a COSMIC signature in the MMR deficiency pathway.

Indel analysis did not confirm aberrations in standard MMR or polymerase genes in patients 1 and 2 (Supplementary Table S2; length values for microsatellite loci). Patient 1′s tumor was considered MSS (Supplementary Fig. S4A). Patient 2 would technically have a classification of MSI-L (Supplementary Fig. S4B) from the Promega Panel of five microsatellite sites (13); however, assessing the fluorescence peaks from the MSI electrophoretograms suggested the DNA amplification was suboptimal, with the likely cause being low input DNA concentration.

We developed an informatics pipeline to detect point mutations and indels in single-sample mode using GATK v.3.8 Mutect2 tool and custom, in-house filters. To compensate for the lack of a matched l healthy control, we performed standard point mutation and indel filtering, then filtered against variants common in the population, as well as against a panel of healthy specimens from blood-derived DNA. As a negative control, we performed variant calling and filtering on 145 Ewing sarcoma (also in single-sample mode). On average, the point mutation to indel ratio for ASPS was 5.6:1 (n = 6), whereas the point mutation-to-indel ratio for Ewing sarcoma was 10.2 (n = 145; Supplementary Fig. S5). In general, Ewing sarcoma tumors had higher ratios; however, most ASPS tumors were within the lower range of Ewing sarcoma tumors, suggesting that ASPS tumors are generally MSS.

With respect to the frameshift frequency, most Ewing sarcoma tumors harbored 10 frameshift mutations/tumor. In contrast, ASPS tumors harbored a wide range of frameshift mutations, including three with frameshift frequencies (5–8 frameshift indels) similar to that seen in Ewing sarcoma range, and three having >30 frameshift indels. Patient 2 had the lowest substitution:indel ratio and a high frequency of frameshift indels. However, patient 1 responded to immunotherapy despite having <10 frameshift indels. Taken together, neither the overall mutation burden nor the ratios of substitutions to indels or frameshift variants provide a genetic explanation for the responses seen.

ASPS is a rare sarcoma with limited options in the metastatic setting (8, 9). Both patients 1 and 2 failed to respond to antiangiogenic therapy and after disease progression were placed on either single-agent or combination immune checkpoint inhibitors. Both patients showed substantial and sustained partial responses. Pathologic assessment of IHC and TILs in these patients indicated that the responses to immunotherapy appeared unrelated to baseline numbers of immune cell infiltrates. Exomic characterization demonstrated an MMR deficiency signature, which may, in part, explain these responses; however, conventional MMR aberrations were not identified. Although puzzling, a similar situation has been reported in breast cancer where BRCA mutational deficiency signatures can be identified in patients with no identifiable somatic BRCA mutations (14). The reason for the mutational signature is unclear although was identified in 3 of 5 additional cases where whole-exome sequencing was undertaken. There are several hypotheses to explain why MMR-deficient immune signatures were identified in patients who are MMR proficient with low mutational burden tumors. For example, it is possible that the genetic signatures of MMR are caused by noncanonical members of the MMR pathway (rather than the core members that are assayed by IHC). In addition, it is possible that MMR status is not the same across the entirety of the tumor, reflecting genetic heterogeneity in the tumor. Alternatively, nongenetic causes (e.g., exogenous or mutagenic exposures) may contribute.

The low mutational load in ASPS cases appears to contradict the tenet that only highly mutated tumors respond to immune checkpoint inhibition (15). The mechanism of response in ASPS is thus dissimilar to response to immunotherapy in tumors with high rates of mutation in microsatellites, such as those showing biallelic MMR deficiency, which is typically driven by hypermutation profiles (16). The FDA has approved antibodies to PD1 for treatment of non-colorectal MMR-deficient tumors as assessed by either PCR or IHC, based on the study data reporting 53% of those treated had a radiographic responses and 21% had complete responses (17). These clinical data may also be relevant to the translocation-associated variant of renal cell cancer that harbors the ASPL–TFE3 translocation on chromosome Xp11.2 (18).

Although checkpoint inhibitors have revolutionized care for some types of solid tumors, their value for treatment of advanced soft-tissue sarcoma (STS) is uncertain. In the SARC028 trial, pembrolizumab was investigated in selected histologic subtypes of STS (19, 20). In the 40 evaluable patients, the overall response rate was 18%, with responses occurring in subtypes with mutational heterogeneity. Related to ASPS, a few studies report response to checkpoint blockade using immunotherapy (21, 22). Single-agent checkpoint inhibitor studies such as SARC028 did not included ASPS subtypes as it was assumed there would be low efficacy given it is a fusion-driven sarcoma. It remains to be seen whether our findings will be confirmed in immunotherapy studies that include ASPS subsets (NCT02609984, NCT02815995, and NCT02636725).

In STS, it is unknown whether PD-L1 expression may be used as a biomarker (23). Pretreatment tumoral PD-L1 expression was low in both patients 1 and 2; however, this was in keeping with studies where pretreatment PD-L1 expression does not predict response in other tumor types (24). In the biomarker substudy of SARC028, PD-L1 expression was low at baseline and inconsistently associated with clinical response to immunotherapy (20). Similarly, the response to durvalumab in patients 1 and 2 appeared unrelated to baseline immune infiltrate; thus, it is likely that other mechanisms are driving ASPS immunogenicity. For example, TFE3 may modulate the immune response through upregulation of the TGF-B pathway/Foxp3 (25) and activation of CD40 ligand (26).

We acknowledge several limitations in our report. First, this is a small retrospective study of a rare sarcoma subset. Second, there was insufficient sample material to measure elevated MSI at selected tetranucleotide repeats for assessment of MSI. Finally, the lack of matched control subjects made assessment of mutation burden difficult, although this problem was mitigated by cross referencing these samples' genomic information with databases of germline variants. Nevertheless, the observation in these 2 patients, in combination with preclinical work, supports the investigation of immune checkpoint blockade as a therapeutic strategy for ASPS.

M.O. Butler is a consultant/advisory board member for Merck, Bristol-Myers Squibb, Immunocore, EMD Serono, and Novartis. A.R.A. Razak reports receiving commercial research funding from Medimmune. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Lewin, S. Davidson, S. Salah, M.O. Butler, A.R.A. Razak

Development of methodology: J. Lewin, S. Davidson, N.D. Anderson, U. Tabori, M.O. Butler, A. Shlien, A.R.A. Razak

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Lewin, S. Davidson, B.Y. Lau, J. Kelly, S. Salah, M.O. Butler, K.L. Aung, A. Shlien, A.R.A. Razak

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Lewin, S. Davidson, N.D. Anderson, J. Kelly, U. Tabori, M.O. Butler, K.L. Aung, A. Shlien, B.C. Dickson, A.R.A. Razak

Writing, review, and/or revision of the manuscript: J. Lewin, S. Davidson, J. Kelly, U. Tabori, S. Salah, M.O. Butler, K.L. Aung, A. Shlien, B.C. Dickson, A.R.A. Razak

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Lewin, S. Davidson, A.R.A. Razak

Study supervision: A.R.A. Razak

We acknowledge the support of MedImmune (Gaithersburg, MD) in allowing the publication of these two patients on NCT02261220 and NCT01693562. The authors declare that no funding was received from MedImmune for development or research associated with this manuscript. Funding for this work was directly received from the philanthropic efforts of the Sarcoma Cancer Foundation of Canada and The Princess Margaret Cancer Foundation/Nicol Family Foundation.