Case series

We identified seven patients with angiosarcoma treated with ICIs on clinical trials or off-label since 2016 (Table 1). Most patients had cutaneous angiosarcoma (5/7), one had primary breast angiosarcoma and one radiation-associated breast angiosarcoma. Among the patients with cutaneous angiosarcoma (n = 5), three had scalp and two facial involvement. The median age was 68 years and 5/7 patients were females. Patients had either metastatic or locally advanced disease with multiple recurrences. All patients had received prior systemic therapies (range 1–6, mean 3), and received 4–14 doses of ICIs (median 5).

Response to therapy was evaluated every 2 to 3 months by radiographic imaging as well as by physical examination. Patients with measurable disease by RECIST 1.1 criteria were assessed for response by comparing baseline and on-treatment imaging, with partial response (PR) requiring greater than 30% decrease in the sum of greatest diameter of target tumor lesions. Patients with cutaneous involvement not assessible by radiographic imaging were monitored as non-target lesions by RECIST 1.1 guidelines.

Five out of seven patients achieved partial or complete response as best overall response. Patient 3 met RECIST 1.1 criteria for complete response (CR) and patient 2 for PR. Patients 5 and 6 had PR based on evaluation of cutaneous, non-measurable lesions. Patient 4 achieved PR based on RECIST 1.1 criteria as well as based on cutaneous lesion measurements (Table 1).

Three patients (patient 2, 4 and 6) are presently on ICI with continuing response (Fig. 1c, Table 1). Patient 5 (cutaneous angiosarcoma) discontinued Pembrolizumab due to personal preference. Patient 3 (cutaneous angiosarcoma) received 14 doses of AGEN1884, an anti-CTLA-4 antibody at a low dose of 0.1 mg/kg, and achieved CR by imaging, physical exam and biopsy. Patient remains in CR more than 1 year off therapy.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1

a. Clinical photographs of patient 4. b. PET imaging showing the site of progression of patient 4 before and after switching therapy to Nivolumab/Pembrolizumab. c. Clinical photographs of patient 6

By 12 weeks after initiating ICIs, 2/7 (28%) patients met radiographic criteria for progressive disease; patient 1 with primary breast angiosarcoma and patient 7 with cutaneous multifocal angiosarcoma. Patient 1 was switched to another therapy shortly after progressing to Pembrolizumab but eventually succumbed to her disease 31 weeks after initiation of ICI. For patient 7, therapy was continued for two additional doses past radiographic progression per protocol, and was changed to Pazopanib when confirmed progressive disease. One patient with metastatic cutaneous angiosarcoma (patient 4), developed a mixed response at 14 weeks on pembrolizumab, with some lesions improving whereas others were worsening (Fig. 1a), but achieved partial response after changing therapy to Ipilimumab/Nivolumab (Fig. 1b). The median duration of response to ICI for the three patients who progressed was 3.4 months and not reached for the rest. No patient developed any grade 3 or 4 immune related adverse effects (irAE).

Correlative studies

Based on the remarkable activity of ICIs in angiosarcoma patients, we performed exploratory analysis of immune and genetic features of patient 3 who achieved CR on a Phase I clinical trial of anti-CTLA4 antibody AGEN1884 (NCT02694822) (Fig. 2). AGEN1884 is a fully human monoclonal immunoglobulin G1 κ subclass (IgG1κ) antibody that specifically recognizes CTLA-4 and mediates strong inhibition of CTLA-4:CD80/CD86 axis [8]. While of the same IgG1 class as ipilimumab, preclinical data suggests this molecule may have enhanced activity against T regulatory cells. We obtained a core needle tumor biopsy 12 days after the first dose of AGEN1884, and isolated tumor infiltrating lymphocytes (TILs) by flow cytometry, with attention to CD8⁺, CD4⁺, and T regulatory cells (Fig. 2e). CD4⁺ T cells consisted of central memory T cells (79%) and effector memory T cells (17%), whereas CD8⁺ T cells consisted of central memory T cells (78%) and less of effector memory T cells (5.5%). Both CD4⁺ and CD8⁺ T cells expressed PD-L1 as 17 and 31%, respectively. Tregs had a predominantly central memory phenotype, with an almost absent population of CD49b⁺Lag3⁺ (TR1) cells which typically represent an exhausted T cell phenotype. Expression of other checkpoint receptors is noted in Fig. 2e.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2

(Patient 3) a. Clinical photographs of cutaneous angiosarcoma lesion before and after treatment with AGEN1884, a monoclonal antibody to immune checkpoint CTLA-4. b. Magnetic resonance imaging before and after treatment with AGEN1884. c. Immunohistologic appearance of angiosarcoma showing malignant cells that line poorly-formed vascular lumens and infiltrate the dermis. d. Relative proportions of circulating immune cells within the peripheral blood at baseline and with subsequent treatments with AGEN1884. e. Immune phenotyping by multiparameter flow cytometry of tumor-infiltrating lymphocytes isolated from angiosarcoma tissue biopsy 12 days after the first dose of AGEN1884

Additionally, mononuclear cells from peripheral blood of patient 3 were isolated on the first day of the first four cycles (Fig. 2d). Two populations decreased from baseline: the natural killer cells (NK) and CD4⁺ T cells. However, both CD4⁺ T cells and NK cells were above 50% of the peripheral blood immune cells at baseline. Peripheral B cells and CD8⁺ T cells increased above baseline with treatment, from 35.8 to 47.5% and from 37.8 to 43.8%, respectively.

Further, we measured numerous cytokines on peripheral blood of Patient 3 at different time points during the first two cycles. Measurable cytokines were IL-6, IL-8, TNF and IFN-γ. Interestingly, IFN-γ and IL-6 decreased relative to baseline before the second cycle and remained as such 7 days after but no particular response pattern was observed in the other measurable cytokines.

Lastly, next-generation whole exome (WES) and RNA sequencing (RNA-Seq) were performed on archival tumor tissue from this patient’s original surgery (rhinectomy), and WES on blood-extracted DNA. Comparison of tumor and normal DNA identified 287 tumor mutations, with only 6 missense mutations. Missense mutations occurred in the genes NBPF10, NBPF15, ZNF678, VPS8, PCLO and ABCB1. The function as well as the exact clinical relevance of these genes are not known in sarcomas. Mutations in the PLCO gene have been found in poorly differentiated hepatocellular carcinoma [9] as well as in hematological malignancies [10], however its role in the pathogenesis of either is not known. The NBPF genes encode proteins whose function are still not fully understood, but have been shown to be highly expressed in sarcoma with unknown clinical or prognostic relevance [11, 12]. Interestingly, the ABCB1 gene has been implicated in the export of taxanes and other cytotoxic agents, and gene polymorphisms have been shown to have both predictive value in ovarian cancer [13]. The overall tumor mutation burden (TMB) was low at only 0.09 mutations/mb. Multiple putative fusion transcripts were identified, including 31 fusions predicted to generate novel protein sequences. Additionally, 20 of 246 genes associated with cancer testis antigens [14] were expressed at over 1 FPKM (fragments per Kilobase per million) in the tumor RNA-Seq data.

Sufficient archived tumor tissue was unavailable for the other patients treated in the study for correlative analysis. However, patients 4 and 5, both of whom achieved partial response, had previously undergone tumor comprehensive genomic profiling (CGP) performed by FoundationOne™. Both patients had intermediate tumor mutation burden based on pre-established histology nonspecific cutoffs per FoundationOne™. Remaining genomic findings from these two patients are listed in Table 2.

Characterization of TILs

biopsy specimens were processed in real time and a single cell suspension was made by digesting tissue with dispase/collagenase (Roche Liberase DH) in media with addition of DNAse at 37 degrees, and single cells isolated using a Ficoll gradient. Cells were stained with antibodies, followed by FACS analysis for profiling. The antibodies used for FACS: CD3 Alexa, CD4 APC cy7, CD25 BV650, CD27 PE-CF594, CD45RA BV421, CTLA4(CD152) BV786 and PD-1 BB515 were purchased from BD Biosciences; LIVE/Aqua BV510 from Invitrogen; CD8 BV570 from Biolegend; CD127 PE from Beckman Coulter; Lag3 PE-Cy7 and TIM-3 APC from eBioscience.

Analysis of circulating peripheral blood mononuclear cells

samples were processed by Ficoll gradient to isolate serum, peripheral mononuclear cells (PBMC), and plasma. PBMCs were stained with antibodies followed by FACS analysis.

Measurement of cytokines

Luminex Multiple assays were used.

Genetic profiling of tumor

WES of tumor and matched patient PBMCs were conducted by Personalis Inc. using the Personalis ACE Exome™ Assay (ACE v3). Personalis Cancer DNA Pipeline was used to identify tumor somatic variants, short insertions and deletions. Similarly, RNA-Seq was performed using Personalis ACE Transcriptome™ Assay (ACE v3), and the Personalis Cancer RNA Pipeline was used for gene expression analysis. All sequencing reads were aligned to hs37d5 reference genome build. The analysis pipeline performs alignment, duplicate removal, and base quality score recalibration using best practice guidelines recommended by the Broad Institute.

Tumor mutation burden

Calculated as the number of non-synonymous somatic mutations per DNA megabase, as derived from the WES-based mutation discovery.

Gene fusion detection

After filtering the RNA-Seq reads for quality and removal of bacterial and viral sequences, bioinformatic tools were used to identify gene fusions [29–31].

*Performed by Agenus.