Deep abscopal response to radiotherapy and anti-PD-1 in an oligometastatic melanoma patient with unfavorable pretreatment immune signature

Case 1 In January 2012, a 69-year-old woman was diagnosed with BRAF-wild-type melanoma on the upper left leg. The primary tumor and the sentinel inguinal lymph nodes were resected. The disease was staged pT2a pN1a cM0. From 5/2012 to 12/2012, the patient received low-dose adjuvant IFNα therapy. In May 2015, she presented with macroscopic lymph node metastases in the left groin, which were R1-resected (Fig. 1a). In June 2015, three liver metastases, a muscle lesion in the upper left leg, and a lesion in the left groin were detected by FDG-PET/CT. Since only three body regions were affected, with a limited number of macroscopic metastases, the patient was treated with potentially curative SBRT to the three liver lesions (Fig. 1b, c, e). The SBRT (3 fractions of 15 Gy on consecutive days) was delivered in September 2015, 4 weeks after initiation of nivolumab (between the 3rd and 4th nivolumab doses) (Fig. 1a). Due to non-symptomatic grade 3 elevation of liver enzymes, anti-PD-1 was discontinued after the 6th nivolumab infusion, and systemic corticosteroids were initiated. The first response evaluation (12 weeks after the nivolumab initiation/8 weeks after SBRT) showed a partial response of all irradiated and non-irradiated lesions. At 6 months after the nivolumab initiation/5 months after SBRT, FDG-PET/CT showed CR, despite early nivolumab discontinuation and continued corticosteroids (Fig. 1d). In March 2020, more than 4.5 years after the concurrent SBRT and nivolumab, there was still no evidence of disease.

Case 2 In September 2013, a 77-year-old man was diagnosed with BRAF-wild-type melanoma in the right breast region; the primary tumor and the sentinel lymph nodes in the axilla were resected. The disease was staged pT3b pN1a cM0. The patient refused adjuvant IFNα therapy. In February 2015, a solitary liver metastasis was detected, which was surgically resected. In July 2015, CT revealed three new liver metastases. Since only one organ showed a limited number of macroscopic metastases, potentially curative SBRT (3 fractions of 15 Gy every other day) was delivered to the two largest lesions located close together in a non-central liver region. The SBRT was delivered in September 2015, 4 weeks after the pembrolizumab initiation (between the 2nd and 3rd pembrolizumab doses) (Fig. 2). At the first response evaluation (10 weeks after the pembrolizumab initiation/6 weeks after SBRT), the irradiated lesions had regressed and the non-irradiated lesion was stable. At 5 months after the pembrolizumab initiation/4 months after SBRT, FDG-PET/CT revealed progression, with a new liver lesion. Three months later (April 2016), a new lung lesion and another new liver lesion appeared; in addition, a large abdominal lymph node metastasis was detected (Fig. 2). The tumor board decided to again try to induce an abscopal effect by delivering a second SBRT (8 fractions of 7.5 Gy over 2.5 weeks) to the large lymph node metastasis and one liver metastasis; the three other detectable lesions (two liver lesions and the lung lesion) were not irradiated. The 2nd SBRT was delivered in June 2016, after the 14th cycle of pembrolizumab (Fig. 2). FDG-PET/CT in July 2016 revealed a CR with radiologic complete regression of all irradiated and non-irradiated metastases, i.e., a strong RT-induced abscopal effect (Fig. 2).

In September 2016, a pruriginous exanthema appeared which initially responded to topical steroids. In December 2016, the skin symptoms exacerbated and histological, clinical, and blood (BP180-ELISA) examinations led to the diagnosis of autoimmune bullous pemphigoid, likely related to pembrolizumab. Systemic corticosteroids were then administered for approx. 12 months, which improved the skin condition. Pembrolizumab was nevertheless discontinued in March 2017. In May 2017, FDG-PET/CT revealed a new lung lesion, which was treated by SBRT (Fig. 2). Afterward, all lesions were controlled at least until December 2018. However, FDG-PET/CT in July 2019 revealed progression with a new vertebral lesion (Fig. 2), several lymph node metastases in the cervical region, and a suspicious soft-tissue lesion in the left thigh.

After multifocal progression was diagnosed in July 2019, the patient at first refused further therapy apart from RT of the symptomatic vertebral bone metastasis. According to the patient’s will, RT of the 3rd lumbar vertebra metastasis was performed with 5 fractions of 3 Gy per week for 2 weeks (cumulative dose 30 Gy), which led to local tumor control as documented by CT scans in December 2019. But CT scans in December 2019 revealed a further progression with peritoneal tumor spread, and pembrolizumab therapy was resumed in January 2020. Pembrolizumab therapy led to a decrease in the tumor marker S100 and the patient’s general condition is now stable without symptoms of metastatic disease. The bullous pemphigoid has not yet recurred since the reinitiation of pembrolizumab.

Pretreatment tumor material obtained through surgical resection a few weeks before the start of ICB was analyzed by bulk whole-exome sequencing (Fig. 3a, b, Supplementary Fig. 1) and IHC (Fig. 3c, Supplementary Fig. 2). Both patients’ tumors showed a similar mutational burden (Fig. 3a) in the upper-intermediate range for melanoma [17]. However, there were strong differences in the immune status of the resected tumor material. High levels of CD8+ T cells were found in the resected lymph node metastasis of patient 1; the resected liver metastasis of patient 2 showed a 20 times lower CD8+ T cell density both by RNAseq and by IHC (Fig. 3b, c, Supplementary Fig. 2).

Moreover, the lymph node metastasis of patient 1 revealed an exhaustion signature (Fig. 3b) [18]. Besides PD-1 and PD-L1, the exhaustion markers TIM-3 and LAG-3 and the transcription factor TOX, which is required for the formation of exhausted T cells [19], were expressed. Exhausted T cells secrete high levels of cytotoxic molecules (perforin, granzymes) and the T cell effector cytokine IFNγ, but not IL-2. TNFα expression is also usually reduced. Another typical feature is high expression of IFNγ-inducible T cell-recruiting chemokines (CXCL9, CXCL10), which were found, in addition to the T cell-recruiting chemokine CCL5, and CXCL13 recently reported to be expressed in tertiary lymphoid tissue [20]. Expression of indoleamine 2,3-dioxygenase (IDO1), also typically induced by IFNγ, was also found (Fig. 3b). Remarkably, there was only little evidence for the presence of immunosuppressive leukocytes such as CD4+ Tregs; immunosuppressive arginase typically expressed by immunosuppressive myeloid cells was not found at all.

In contrast, the resected liver metastasis of patient 2 showed no expression of PD-1, TIM-3, or LAG-3. However, some expression of the PD-1 ligand, PD-L1, which can occur dependently or independently of IFNγ [21], was detected (Fig. 3b, Supplementary Fig. 2). Perforin, IFNγ, IL-2, TNFα, and granzymes were virtually undetectable (Fig. 3b). In accordance with the low IFNγ expression, expression of IDO1, CXCL9, and CXCL10 was also low. However, the Treg-specific transcription factor FoxP3 was expressed at higher levels than in the lymph node metastasis of patient 1, and expression of immunosuppressive arginase was also relatively high (Fig. 3b). Expression of TOX was low and CXCL13 was undetectable. Together, these findings did not show evidence for pretreatment tumor infiltration by exhausted T cells that could respond to anti-PD-1.

Both patients’ tumors showed high expression of β2-microglobulin (β2M) as well as expression of MHCI. Expression of MHC class II, typically found on professional antigen presenting cells, was higher in the lymph node metastasis of patient 1 (Fig. 3b).

ICB can induce the proliferation of PD-1+ T cells [22, 23]; it is unclear whether conventional treatments (including different modes of RT) also increase it. We analyzed PBMCs before ICB and 7–10 days after the first SBRT. PBMCs of patient 2 were also analyzed before and after the second SBRT. After initiation of ICB and the subsequent first SBRT, patient 1, who rapidly developed a CR, showed a stronger increase in Ki67+ CD8+ T cells compared to the pretreatment levels (approx. 3.5-fold; post-treatment: 20%) than patient 2 (approx. twofold; post-treatment: 12%) (Fig. 4a). The increase in Ki67 was mainly seen in PD-1+ CD8+ T cells (patient 1: post-treatment 54%; patient 2: 24%) (Fig. 4b). The frequencies of Ki67+ cells indeed seemed to correlate with the treatment response. After the second SBRT in patient 2, which provoked the strong abscopal response despite resistance to long-term anti-PD-1 therapy, the Ki67+ CD8+ T cells increased to 19%, and the PD-1+ CD8+ T cells expressing Ki67, to 37%, i.e., to a similar level as seen in patient 1 after successful combined SBRT/ICB. When patient 2 clearly progressed (August 2019), the CD8+ T cells expressing Ki67 dropped to 9%.

We then investigated whether tumor-specific CD8+ T cells were detectable in the blood of patient 2 during progression between the first and second SBRT. Known HLA-A2-restricted epitopes from melanoma differentiation antigens, which were highly expressed in the pretreatment melanoma tissue, and potential neoepitopes, predicted based on whole-exome DNA and RNA sequencing, were selected (Fig. 4c left). Only a few of these potential neoepitope peptides bound with high affinity to HLA-A2.1 (Fig. 4c right). By using peptide pools, we detected reactivity to differentiation antigens and to neoepitopes (Fig. 4d). The response of differentiation antigen-specific CD8+ T cells, which mainly recognized the MART-1_26–35 epitope, was particularly high (Fig. 4e). Tumor-specific T cells could be detected despite a pronounced lymphopenia, which deepened with each round of SBRT (Fig. 4f left, Supplementary Fig. 3). Of note, both patients also had an unusually low proportion of CD8+ T cells among the CD3+ cells in the blood (normally approx. 30%) (Fig. 4f right).

The ratio of Ki67+ CD8+ T cells to the tumor burden might be a better predictor for ICB response than Ki67+ T cells alone [23]. In our study, a large proportion of the visible tumor mass was irradiated in the two patients who were “oligometastatic” [24]. In the successful second SBRT in patient 2, more than 90% of the tumor mass was irradiated, thus leaving only a very small non-irradiated tumor mass (approx. 10 cm³) (Table 1).