Patients and baseline characteristics
Patients with metastatic melanoma progressing on any PD-1 ICI were eligible for this study. Potentially eligible patients provided informed consent prior to a baseline screening blood draw. Patients were screened from December 2020 to February 2023 with an accrual goal of between 17 and 20 evaluable patients.
Thirty-four (34) patients were screened for eligibility (see CONSORT diagram in Supplemental Fig 2). Sixteen (16) patients were excluded due to low sPD-L1 (n=8), poor venous access (n=2), unwillingness to undergo treatment or follow-up (n=2), contraindications to ICI (n=2), or death before sPD-L1 level could be assessed (n=2).
Eighteen (18) patients were eligible (Table 1). Mean age of patients in the study was 63 years (SD 13), and 7 (39%) were female. Mean baseline sPD-L1 measured by ELISA was 26.31 ng/mL. All patients had previously received at least one PD-1 inhibitor that had failed, and fifteen (83%) had previously received at least one ICI doublet (e.g., ipilimumab and nivolumab) that had failed. Half of patients had received two prior lines of therapy, and an additional one third of patients had received four or more lines of therapy before enrollment in the study.
All eligible patients (n=18) underwent treatment (Supplemental Fig 1 and 2). Treatment comprised 1-5 days of stereotactic body radiotherapy (SBRT) to at least one—but not all—sites of disease seen on imaging. This was followed by three consecutive days of one session each of TPE, followed by re-initiation of ICI therapy. ICI therapy was the same as previously received in 16 patients (89%). One patient was unable to complete the third TPE treatments due to a line infection. All patients were included in the intent to treat analysis regardless of treatment completion.
SBRT, TPE, and ICI re-challenge is safe and well-tolerated.
The primary safety endpoint of this trial was to assess the adverse events (AE) observed during SBRT, TPE, and ICI re-challenge in patients with melanoma receiving PD-1 immunotherapy (Table 2). AEs were collected and reported according to CTCAE V.5.0. In a pre-planned stopping rule, if 3 of the first 10 patients were to experience grade 4 or greater adverse events attributable to the intervention, the study would be halted.
In total, one patient experienced AE grade 3 and 4 thromboembolic event and sepsis, respectively, secondary to an infection of a vascular access device. This patient required hospitalization with line removal and intravenous antibiotics. Other AEs were grade 1-2 and included fatigue and arthralgias/myalgias. No deaths were attributable to therapy.
Therapeutic plasma exchange removes soluble PD-L1 with varying reconstitution rates.
The primary efficacy endpoint of the study was the reduction of sPD-L1 levels by TPE (Fig 1A, Supplemental Figure 3A). Relative and absolute levels of sPD-L1 were significantly reduced by TPE (mean 80.2% reduction, p=6.1*10-5). Mean recovery rate of sPD-L1 at the second cycle of ICI re-challenge was 21.4% of pre-TPE levels.
SBRT, TPE, and ICI re-challenge resensitized melanomas to immunotherapy.
A secondary efficacy endpoint of the study was the proportion of patients achieving CR (defined as absence of radiologically apparent disease), PR (defined as greater than 30% radiographic reduction), or SD (defined as less than 30% radiographic reduction) as a best response per Response Evaluation Criteria in Solid Tumors (RECIST) guidelines version 1.1 in unirradiated lesions assessed radiographically.47 Response was assessed by a trained radiologist. Two patients (11.1%) experienced CR, one (5.6%) PR, three (16.7%) SD, and 12 (66.7%) progressive disease (PD) (Table 3). Notably, there was a deeper absolute sPD-L1 reduction with TPE in patients who experienced CR or PR (p=0.02, Supplemental Fig 3B).
A further secondary endpoint of the study was overall survival (OS). Median overall survival for the cohort by Kaplan-Meier estimate was 18.9 months (95% CI 15.1-NR) (Supplemental Fig 4). CR, PR, or SD response to treatment predicted superior overall survival (OS) (p=0.002) (Fig 1B, Supplemental Fig 5). Median OS for patients experiencing PD was 15 months. Median OS for responders was not reached.
Representative irradiated and unirradiated lesions were selected and measured by a blinded radiologist across time in approximately 3-month intervals and assessed for standardized uptake value (SUV) and tumor size. The majority of measured irradiated and unirradiated lesions showed a greater than 50% reduction in SUV (Fig 1C). Approximately one third of measured irradiated and unirradiated lesions showed a greater than 50% reduction in size (Fig 1D). An example FDG PET scan image of irradiated and non-irradiated lesions in a patient experiencing CR is shown in Fig 1E. A swimmer plot of all patients is shown in Fig 1F.
Tumor-reactive effector T cell changes predict outcomes to immunotherapy re-challenge after radiotherapy and TPE.
Peripheral blood mononuclear cell (PBMC) subpopulations act as liquid biomarkers and as critical substrates of ICI immunotherapy. Changes in these subpopulations reflect the effects of therapy on anti-tumor immunity.
We isolated PBMCs and measured subpopulations by flow cytometry according to a broad panel (Supplemental Tables 1-2, Supplemental Fig 6-7). We compared subpopulations from pre-treatment to the beginning of the second cycle of ICI re-challenge. A heatmap of changes is shown (Fig 2A). We further evaluated how these changes correlated with overall survival in our cohort by Cox proportional hazards regression (Fig 2B, Supplemental Tables 3-4).
Our group previously identified tumor-reactive CD8+ T cells (TTR) with high expression of CD11a, Granzyme B, and CX3CR1 (GZMB+/CX3CR1+/CD11ahigh or TTR).44–46 In a prespecified analysis, peripheral blood TTR changes from baseline to the next cycle of ICI re-challenge predicted overall survival (OS) with a Cox proportional hazard ratio (HR) of 0.16 (95% confidence interval [CI] 95% CI 0.03-0.8, p=0.026). We plotted survival by TTR change using Kaplan-Meier estimates (Fig 2C, log rank p=0.01). When TTR levels were analyzed dichotomously as increasing versus decreasing (cutoff ratio=1), the results were identical.
Regulatory T cells marked by CD4, CD25, and FOXP3 expression (Treg) are well-established predictors of poor response to immunotherapy. In a prespecified analysis, peripheral blood Treg changes from baseline to the next cycle of ICI re-challenge predicted OS (HR 7.87 [95% CI 1.6-38.7], p=0.01). We plotted survival by Treg change using Kaplan-Meier estimates (Fig 2D, log rank p=0.004). When Treg levels were analyzed dichotomously as increasing versus decreasing, these cells did not predict a significant difference in OS (HR 2.89 [95% CI 0.74-11.3], p=0.13).
We previously showed that NKG7 marks resilient T cells that are critical to immunotherapy response.44,48 In an exploratory analysis, changes in peripheral blood CD11ahigh CD8+ T cells with NKG7 expression from baseline to the next cycle of ICI re-challenge predicted superior OS (HR 0.19 [95% CI 0.04-0.88], p=0.03). We plotted survival by NKG7-positive T cell change using Kaplan-Meier estimates (Fig 2E, log rank p=0.02).
In addition, we previously reported that Bim (BCL-2-interacting mediator of cell death) is upregulated in T cells by PD-L1/PD-1 engagement and that T cell Bim expression drives poor anti-tumor immunity.44,49 Persistent CD8+ T cell Bim expression despite ICI predicts poor responses to immunotherapy.50 In an exploratory analysis, changes in peripheral blood CD11ahigh CD8+ T cells with Bim expression from baseline to the next cycle of ICI re-challenge predicted inferior OS (HR 6.8 [95% CI 1.58-29.32], p=0.01). We plotted survival by Bim+ T cell change using Kaplan-Meier estimates (Fig 2F, log rank p=0.003).
Soluble mediators of immunosuppression are elevated in patients with ICI-refractory melanoma, are reduced by TPE, and rapidly reaccumulate.
As noted above, soluble factors beyond sPD-L1 contribute to ICI resistance.27–42 We next sought to determine the effect of TPE on these soluble factors in systemic circulation. We measured a large library of soluble factors in sufficient plasma samples from patients on the study and compared these to matched healthy controls using a high-throughput O-link multiplex assay. A variety of immunosuppressive soluble factors were elevated at baseline in the blood of patients with ICI-refractory melanoma versus healthy controls (Fig 3A, Data File 1), including sCTLA-4, sLAG3, sPD-1, and sTIM-3.
We next measured reduction and rebound of these factors in the plasma of patients from SBRT to TPE, pre- to post-TPE, and at the second cycle of ICI re-challenge (see Data File 1). TPE significantly reduced exemplary immunosuppressive factors sPD-1, sLAG3, and sTIM3 (HAVCR2) (Fig 3B-D). Notably, sCTLA4, which was only moderately elevated in this cohort versus healthy controls, was unchanged (Fig 3E). Conversely, Prostaglandin E synthase 2 (PTGES2), a marker of systemic inflammatory response, increased significantly after TPE (p=0.0004) and recovered by cycle 2 of ICI re-challenge (p=0.016) (Fig 3F).
Given the direct connection between soluble immunosuppressive factors and ICI resistance, we sought to determine whether the level of these factors at the time of ICI re-challenge (i.e., post-TPE) could predict overall survival. Level of sPD-L1 (log rank p=0.003) and sTIM3 (log rank p=0.007) after TPE predicted OS (Fig 4A-B). Conversely, OS was not predicted by sPD-1, sLAG3, or sCTLA-4 in this cohort.
As outlined above, we observed rapid reconstitution of most factors prior to second cycle ICI re-challenge. We next sought to determine whether levels of these factors at the second cycle of ICI re-challenge could predict outcomes. These factors did not predict OS in our cohort (Fig 4C-D).
Exosomal PD-L1 was not significantly changed on treatment.
We and others have previously showed that malignant cells release exosomes or extracellular vesicles bearing PD-L1 that can drive immunosuppression and can release sPD-L1 by proteolytic cleavage.23–26 Radiation can cause tumor exosome shedding by multiple mechanisms, including cell death.26 However, exosomal PD-L1 (exPD-L1) was previously reduced by TPE in a pilot study.43 Thus, it is unclear whether exPD-L1 would be expected to be reduced or increased after both SBRT and TPE. We measured exPD-L1 by flow cytometry before and after TPE in the present cohort (Supplemental Fig 8A). Circulating exPD-L1 was not significantly changed. However, change in exPD-L1 after TPE predicted improved OS (HR 0.28 [95% CI 0.08-0.98], p=0.047) in the cohort (Supplemental Fig 8B).