In preparation for this study, we sought to determine PD-L1 expression in archived TET samples by IHC. Resection or biopsy specimens were obtained from 54 patients with the following clinical characteristics: median age 47 years (range, 17–77); 35 males/19 females; World Health Organization histology: Thymoma = 19 (AB 1, B1 1, B2 8, B3 8, unclassified thymoma 1), Thymic carcinoma = 35; Masaoka stage: IIB 1, III 1, IVA 16, IVB 36; history of prior systemic therapy in 45 (83%) cases.

PD-L1 expression was observed in 16 (84%) thymomas and 17 (49%) thymic carcinomas, whereas it was focal in 14 (40%) thymic carcinomas, and absent in three (16%) thymomas and four (11%) thymic carcinomas. Representative images of PD-L1 expression are shown in Additional file 2: Figure S1; online only.

Seven patients with recurrent thymoma and one patient with recurrent thymic carcinoma were enrolled and received avelumab at doses of 10 mg/kg to 20 mg/kg intravenously every 2 weeks. Patient characteristics are presented in Table 1.Tab1

Patient Characteristics

Overall, an objective response was observed in four of seven (57%) patients with thymoma, and the response could be confirmed with repeat imaging in two of seven (29%) patients. Details and duration of response are summarized in Table 2.Tab2

Clinical Activity of Avelumab in Thymic Epithelial Tumors

Three patients were treated with avelumab at the 20 mg/kg dose level. One patient had a confirmed partial response (WHO B3 thymoma, maximum tumor change from baseline: 48% after one dose of avelumab), one patient had an unconfirmed partial response (WHO B2 thymoma, maximum tumor change: 30% after one dose of avelumab) and one patient had stable disease (WHO B3 thymoma, maximum tumor change: 8% reduction after three doses of avelumab).

Four patients with thymoma (WHO B1 = 1; WHO B2 = 2; WHO B3 = 1) and one patient with thymic carcinoma received avelumab at a dose of 10 mg/kg. The patient with B1 thymoma had a confirmed partial response and one patient with a B2 thymoma had an unconfirmed partial response with maximum tumor shrinkage of 37 and 31%, respectively. One patient with B2 thymoma and the patient with thymic carcinoma had stable disease with no tumor shrinkage and the patient with B3 thymoma had disease progression.

Response to therapy and duration of response are illustrated in Fig. 1. Figure 2a illustrates changes in selected target lesions in patients responding to treatment.Fig1Fig. 1

Response to therapy and duration of response. a Waterfall plot of best response to treatment. Four patients with thymoma achieved a partial response to treatment, including three patients who received only one dose of avelumab (*). Patients with any tumor shrinkage also developed irAE. b Duration of response. Change in the size of target lesions over time during treatment and after discontinuation of therapy (until the last follow-up time point) is illustrated. Three patients (1, 3, and 6) received one dose of avelumab

Treatment-related AEs are summarized in Table 3. Grade 3 and 4 AEs were observed in 3 (38%) patients each. Most AEs were mild (grade 1 or 2) and consistent with previously observed toxicities associated with ICIs. However, a disproportionately large number of patients developed signs and symptoms suggestive of autoimmunity that were included under the umbrella term “autoimmune disorder,” which accounted for all the grade 3 and 4 AEs observed in this series (with the exception of one case of grade 4 hyperkalemia in a patient with grade 3 diarrhea due to autoimmune enteritis). These AEs included muscle weakness, myalgia, myositis, respiratory muscle insufficiency, hoarseness, paresthesia, dysphagia, dyspnea, diarrhea and elevated creatine phosphokinase (CPK). Details of irAEs are presented below and in Additional file 1: Table S2; online only. Neuromuscular AEs observed in our series have been reported separately [21].Tab3

Adverse Events, at Least Possibly Related to Treatment with Avelumab

Patient 1 (stage IVA, WHO subtype B3 thymoma) developed grade 3 CPK elevation and grade 1 transaminitis 2 weeks after administration of avelumab (Fig. 2b). Oral steroids were started on day 18 and tapered over a 6-week period with resolution of laboratory abnormalities. A liver biopsy performed on day 43 showed no evidence of drug-induced liver injury despite elevated transaminases (Fig. 2c). The patient was not re-challenged with avelumab.

Patient 2 (stage IVB, WHO B3 thymoma) received three doses of avelumab uneventfully before an elevation in CPK and liver transaminases was noted (Fig. 2b). The patient developed bulbar weakness with mild sensory loss in feet, facial diplegia, weakness of the tongue and hypophonia. Avelumab was discontinued; oral steroids were started on day 49 and discontinued on day 171 with partial resolution of symptoms.

Patient 3 (stage IVB, WHO subtype B2 thymoma) experienced an increase in CPK 1 week after starting avelumab and it peaked 23 days later. Magnetic resonance imaging (MRI) of bilateral thighs showed changes consistent with myositis. Changes in laboratory parameters are shown in Fig. 2b. Oral steroids were initiated on day 23 and laboratory abnormalities resolved gradually. Steroids were discontinued on day 87. The patient was not re-treated with avelumab.

Patient 6 (stage IVA, WHO B2 thymoma) developed grade 2 dysphagia and generalized muscle weakness 2 days after starting avelumab. CPK, aspartate transaminase (AST) and alanine transaminase (ALT) elevation were observed 8 days after treatment and oral prednisone was started at a dose of 60 mg per day (Fig. 2b). Due to worsening dyspnea and dysphagia and a vital capacity of 790 ml, the patient was admitted to the intensive care unit 13 days after treatment and underwent elective intubation and mechanical ventilation. After three doses of intravenous methylprednisolone and five doses of intravenous immunoglobulin, a transient decrease in CPK, AST and ALT was observed. Two more doses of methylprednisolone were administered on days 18 and 19, resulting in normalization of CPK. However, due to persistently elevated transaminases and the need for continued ventilatory support, intravenous cyclosporine A was started on day 23, resulting in gradual resolution of transaminitis within 3 weeks. Recovery from respiratory failure was partial and prolonged.

Patient 8 (stage IVB, WHO B1 thymoma) received 11 doses of avelumab before developing grade 3 diarrhea. Colonoscopy and small bowel biopsy revealed ileitis with villous blunting and active inflammation (not shown). Diarrhea subsided after treatment with oral prednisone. Despite the presence of autoimmune enteritis, no abnormalities of CPK or liver transaminases were observed.

Six of eight patients had received sunitinib previously, including all four patients with an objective response to avelumab. These four responding patients also developed irAEs as described below. Of the remaining two patients previously treated with sunitinib who had stable disease with avelumab, one developed irAEs (only the patient with thymic carcinoma received prior sunitinib and did not develop irAEs). In two out of eight patients (both with B2 thymoma) who had not received prior sunitinib, one had stable disease and one had progressive disease with avelumab; neither developed irAEs. Details of a possible association of response and irAEs with prior sunitinib are presented in Additional file 1: Table S3; online only.

Interestingly, development of a response was accompanied by irAEs (myositis in three cases and enteritis in one case). Patients with clinical evidence of myositis developed sustained elevations of CPK, AST and ALT as illustrated in Fig. 2b. Post-treatment tissue inflammation was also demonstrated by imaging studies (MRI of the thighs showed myositis, and cardiac MRI showed myocarditis), histopathological analysis (small bowel biopsy in the patient developing enteritis) and electrophysiological studies (myopathic findings on an electromyogram in two of three patients with myositis).

The temporal association between onset of steroid therapy and evidence of tumor shrinkage is depicted in Additional file 1: Table S4; online only. Tumor shrinkage was observed either before development of irAEs and use of steroids (in one case) or shortly after onset of steroid therapy (9 and 20 days later in two cases). This observation suggests that tumor response was related to avelumab rather than steroids used to treat irAEs.

Paired tumor biopsies were analyzed for PD-L1 expression in three patients. In one of three cases the post-treatment biopsy showed necrotic tissue with no viable tumor and was not suitable for immunohistochemical analysis. In the other two cases, diffuse membranous staining pattern in the epithelial component was seen in both pre- and post-treatment biopsies (Additional file 3: Figure S2; online only). Additionally, scattered PD-1 positive lymphocytes were seen before and after treatment (Additional file 3: Figure S2; online only). Staining of normal thymus for control purposes also showed scattered PD-1-positive lymphocytes, predominantly in the medulla (not shown).

Intratumoral immune infiltrates before and after treatment were also evaluated in the two cases described above (Additional file 3: Figure S2; online only). The immune infiltrate in pre-treatment tumor samples of both patients was composed of immature T cells expressing TdT, CD1a and CD5, CD4 and CD8. However, the lymphoid infiltrate in the post-treatment biopsy in one case did not express TdT or CD1a and showed primarily lymphocytes with a mature CD8 positive T-cell phenotype. This patient had a confirmed partial response to treatment. In contrast, the immune infiltrate in the post-treatment biopsy for the other patient showed a phenotype consistent with immature T cells (thymocytes) expressing TdT, CD1a and both CD4 and CD8. This patient had stable disease in response to treatment.

PBMC were monitored at various times pre– and post–anti-PD-L1 therapy for 123 immune cell subsets. The most profound differences were observed prior to therapy between patients who did or did not develop a clinical response to avelumab. Compared to non-responders, responders had a higher absolute lymphocyte count, and lower frequencies of B cells, Tregs, conventional dendritic cells (cDCs), and NK cells before treatment (Fig. 3d). While some of these differences were statistically significant, they should only be considered trends, due to the small number of patients analyzed. Patient 2, a non-responder who developed irAEs like the responders, had an immune profile more similar at baseline to responders than non-responders, including very low levels of Tregs and B cells. Decreases in Tregs and increases in myeloid derived suppressor cells (MDSC) were also noted following steroids in clinical responders who developed irAEs (Additional file 4: Figure S3; online only).

Unsupervised hierarchical clustering of major PBMC subsets prior to therapy was employed to determine if any signature emerged that would distinguish clinical responders (R) vs. non-responders (NR). As seen in Fig. 4a, the five patients who developed irAEs (four of whom also had a radiological response) clustered separately, thus confirming and expanding on the results shown in Fig. 3d, which was performed employing retrospective clinical data.Fig4Fig. 4

Immune phenotype associated with development of clinical responses and autoimmunity, and the effect of steroids. a Unsupervised hierarchical clustering of indicated immune populations in PBMC prior to treatment with avelumab. Higher levels of expression are indicated in red and lower levels of expression are indicated in blue. Patient response (R, responders; NR, non-responders) and development of immune-related adverse event (irAE) are indicated. b Diversity of TCR repertoire in PBMC of patients prior to therapy with avelumab. c Diversity of the TCR repertoire in PBMC of patients pre- and post-steroid treatment. TCR diversity was measured by the metric of repertoire size; values in panels B and C indicate the number of individual clonotypes comprising the top 25th percentile by ranked molecule count after sorting by abundance. The day (D) PBMC were assessed for TCR diversity pre- and post-steroids is indicated