Overall survival in metastatic melanoma correlates with pembrolizumab exposure and T cell exhaustion markers

Abstract Trial data support an absence of an exposure–survival relationship for pembrolizumab. As these relationships remain unexamined in a real‐world setting, we determined them in metastatic melanoma prospectively in an observational study. Translational objectives included identifying biomarkers of progressive disease (PD). Checkpoint blockade naïve patients receiving 2 mg/kg Q3W pembrolizumab had pharmacokinetic and clinical outcome data collected. Trough, a valid surrogate for drug exposure, was assessed using ELISA. T‐cell exhaustion and chemokine markers were determined using flow cytometry. Geometric means of exposures and biomarkers were tested against objective response groups using one‐way ANOVA. The cohort was split by the median into high versus low pembrolizumab exposure groups. Kaplan–Meier progression‐free survival (PFS) and overall survival (OS) curves were estimated for high versus low exposure, compared using the log rank test. The high pembrolizumab exposure group (n = 14) experienced substantially longer median OS (not reached vs. 48 months, p = .014), than the low exposure group (n = 14). A similar positive exposure PFS relationship was found (median not reached vs. 48 months, p = .045). The frequency of TIM‐3 expression on CD4+ T cells was significantly higher in PD (mean 27.8%) than complete response (CR) (13.38%, p = .01) and partial response (12.4%, p = .05). There was a near doubling of CXCR6 and TIM‐3 co‐expression on CD4+ T cells in PD (mean 23.3%) versus CR (mean 11.4, p = .003) and partial response (9.8%, p = .0001). We describe positive exposure‐PFS and exposure‐OS relationships for pembrolizumab in metastatic melanoma. TIM‐3, alongside co‐expression of CXCR6 and TIM‐3 on circulating CD4+ T cells are potential bio markers of treatment failure.


| INTRODUC TI ON
Targeting the inhibitory interaction between T cell checkpoint, programmed cell death 1 (PD-1), and its tumoral and stromal ligands, PD-L1/PD-L2 has transformed outcomes across oncological indications. 1 The immunosuppressive tumor microenvironment potentiates T cell exhaustion and restrains the anti-tumoral immune response. 2 Restoration of deficient anti-tumoral immunity by pembrolizumab, a humanized anti-PD-1 IgG4 monoclonal antibody (mAb), is an established standard of care that has led to durable responses in metastatic melanoma. 3 However, there is a need to understand the factors driving failure of immune checkpoint blockade (ICB) in order to help the majority of patients that do not respond to these agents. 4 Pembrolizumab has pharmacokinetic (PK) similarities with other large molecular weight ICB mAbs; a low central volume of distribution, linear PKs at clinically relevant doses, confinement primarily to the vascular compartment, and a prolonged half-life. 5 The seamless trial design of multiple expansion cohorts when promising early efficacy was first noted, 6 together with lack of dose-limiting toxicities, meant that the traditional approach of obtaining a maximum tolerated dose to guide pivotal registrational trials was not undertaken. 7 Subsequently, in silico PK and pharmacodynamic (PD) studies were key for regimen selection. 8 Modeling pharmacodynamic data found that peripheral target saturation for pembrolizumab begins at 1 mg/kg Q3W with a steady-state dose of 2 mg/kg Q3W reaching a 90% probability of 95% target engagement, 9 suggesting a flat dose-response relationship in the clinic. Maximal lymphocyte stimulation was seen around 1 mg/kg. 10 Trial data supported an absence of a dose or exposureresponse relationship at clinically relevant doses 11 and suggested that the classic clinicopathologic features known to influence mAB did not affect pembrolizumab PKs in a clinically meaningful manner.
This inferred limitations to inter-patient variability. 12 However, prospective real-world data with another anti-PD1 mAb, nivolumab, in metastatic non-small cell lung cancer (NSCLC) and melanoma found gender, baseline albumin, and body surface area affected PKs to a clinically meaningful extent, throwing previous assumptions into doubt. 13 Real-world data regarding exposure-response relationships with ICB conflict with the trial evidence. A cohort of pre-treated metastatic NSCLC patients given 3 mg/kg Q2W nivolumab found patients with higher exposures of drug defined by trough measurements had notably improved best overall response (BOR) p = .002 and overall survival (OS) p = .001. 14 Trough concentrations are a regulatory body approved surrogate for drug exposure. 15 There are no clinically validated biomarkers to identify early resistance or predict lack of response to pembrolizumab in metastatic melanoma. TIM-3 is another immune checkpoint that marks the most terminally exhausted subset of CD8 + tumor infiltrating lymphocytes. 16 The exact mechanism by which TIM-3 contributes to T cell dysfunction remains to be defined, but may involve antagonism of the T cell stem-like state and decreased CD8 + differentiation. 17 TIM-3 expression has been associated with rapid tumor progression, and our previous work highlighted that high TIM-3 expression on CD8 + T cells was associated with poor treatment response to ICB. 18 Chemokines are chemotactic cytokines that regulate leukocyte trafficking. Chemokine CXCL16 and its T cell ligand CXCR6 are also key propagators of melanoma. 19 CXCL16 can be expressed by malignant cells as a transmembrane molecule and mediate effect via autocrine binding to CXCR6. 20 The CXCR6/CXCL16 axis is proinflammatory, 21 CXCR6 is expressed by a self-renewing subset of melanoma stem cells, 22 and melanoma secretes CXCL16, contributing to CXCR6-mediated leukocyte recruitment. 19 We previously identified patients with disease progression on pembrolizumab had consistently higher proportions of CD4 + and CD8 + T cells-expressing CXCR6. 18 Given the number of confounders at play, both patient (interpatient variability in plasma exposures and clearance) and malignancy (histopathology, tumor burden, immunogenicity) related, it is challenging to unravel whether lower plasma exposures of ICB are cause or effect of a lack of response. PK and pharmacodynamic relationships have been primarily studied in the peripheral circulation which has questionable relevance to the tumor microenvironment. 23 We aimed to determine the relationship between pembrolizumab drug exposure and clinical outcomes such as BOR, progression-free survival (PFS), and OS in patients with metastatic melanoma in a real-world setting. We also sought to identify whether circulating T cell exhaustion markers and specific chemokines may help to identify patients with progressive disease (PD).

| Participants and treatment
This study was approved by local institutional review boards.
Patients gave informed written consent. Individuals with metastatic melanoma receiving 2 mg/kg Q3W pembrolizumab had serial PK trough blood draws ≤48 h prior to their next scheduled dose, up to a maximum of 22 cycles. Peripheral blood mononuclear cells (PBMCs) and plasma were harvested from the same draws and stored at −80°C.
All patients were treated with pembrolizumab administered over 30 min intravenously. Treatment was continued until the physician assessed disease progression (clinically or radiologically), patient decision to cease treatment or unacceptable toxicity. Treatment discontinuation in the context of sustained complete response (CR) was allowed at the discretion of the treating physician.

| Study endpoints
Pharmacokinetic data, PBMCs, patient baseline characteristics, clinical outcome data and BOR, PFS, and OS were collected prospectively. OS was defined as time from first pembrolizumab initiation until death from any cause. PFS was defined as the time from pembrolizumab initiation until documented progression (clinical or radiological) or death from any cause.
Imaging assessment was undertaken according to immune response evaluation criteria in solid tumors (iRECIST) 24 by 2 unblinded investigators (VN and AvW) and confirmed by a separate blinded investigator (HM). BOR groups were defined as CR, partial response (PR), stable disease (SD), and progressive disease (PD). Contrastenhanced computerized tomography scanning was used for the imaging assessments.

| Plasma concentrations
Plasma trough pembrolizumab concentrations were determined using the Abcam ® pembrolizumab ELISA kit as per manufacturer instructions. The lower limit of detection was 10 ng/ml. 25 The mean of duplicate biological plasma samples was used for each timepoint.
Geometric mean trough concentrations were calculated for this continuous variable.

| T-cell marker immune subsets
The immune subsets based on T-cell markers were determined using Flow Cytometry on the BD Science (Becton, Dickinson and Company) Fortessa ×20 as previously described. 18  The samples were gated for lymphocytes, single cells, and then live cells. The CD3 subset was gated as the fluorophore (BUV737) versus side scatter (SSE). To separate into CD4 + and CD8 + subsets, all positive CD3 + cells were then further divided into CD4 + and CD8 + by gating CD3 versus CD4 + (BUV496) or CD8 + (APC-H7 SK1).

| Statistical analysis
Descriptive statistics included mean, range, and standard deviation of the continuous baseline patient characteristics. Non-parametric correlation analysis between clinical characteristics was performed using Spearman's rho. Statistically significant differences in clinical characteristics, plasma pembrolizumab concentrations, and T-cell markers between the BOR groups were tested for using one-way ANOVA with Bonferroni correction.
The cohort was split into high versus low pembrolizumab exposure groups, divided by the median trough. Kaplan-Meier survival analysis for PFS and OS was undertaken with the logrank test used to compare survival between pembrolizumab exposure groups. Due to the signal seeking nature of this early observational work, no formal pre-specified statistical power calculations were undertaken to compare results between exposure groups. Statistical analysis was performed using IBM SPSS Statistics V27. A two-sided p-value < .05 was considered statistically significant.

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to the corresponding entries in http://www.guide topha rmaco logy.org, the common portal for data from the IUPHAR/BPS guide to pharmacology, 26 and are permanently archived in the concise guide to pharmacology 2019/20. 27

| RE SULTS
Clinical characteristics and demographics are summarized in Table 1 (range) [standard deviation] 28 patients participated, 5 patients Age at diagnosis of primary melanoma, age of diagnosis of metastatic disease and baseline LDH were not different between the high and low pembrolizumab exposure groups (Table 1).

| Plasma trough concentrations and BOR
The geometric mean pembrolizumab plasma trough concentrations across all timepoints were not a statistically significant difference between the BOR groups. The number of participants with SD (n = 2) was too small for statistical analysis. Trends observed included CR (n = 11) with 34.5% higher geometric mean pembrolizumab trough concentrations (90.8 mcg/ml) than PR (n = 10) (67.5 mcg/ml, p = ns). CR had 27.8% higher trough concentrations than PD (n = 5) (71.5 mcg/ml, p = ns). SD (n = 2) had mean trough pembrolizumab concentrations of 106.4 mcg/ ml. The median pembrolizumab plasma concentrations for each BOR followed the same trend of being higher in the CR (91.8 mcg/ml) and PR (81.4 mcg/ml) groups compared with PD (64.7 mcg/ml), with no statistically significant difference, but the trough concentration variability in exposures was high within the groups (Figure 1).

| Pembrolizumab trough concentrations, T-cell exhaustion, and chemokine markers
There were no statistically significant or clinically meaningful associations between pembrolizumab exposure groups and upregulation of T-cell exhaustion or chemokine markers over time. 15.87%-28.90%) (p = .14) ( Figure 4B).

| CXCR6
There was a higher frequency of CXCR6 expression on CD4 + T cells  Figure 4C). There was no significant difference between expression of CXCR6 on the surface of CD8 + T cells in CR compared with PR or PD ( Figure 4D).

| DISCUSS ION
Evasion of immune surveillance is an emerging hallmark of cancer. 28 Metastatic melanoma is the archetypal immunosensitive malignancy. The high tumoral mutational burden induced by ultraviolet light, brisk stimulation of an innate response by malignant melanocytes, and dense lymphocytic infiltration throughout the tumor microenvironment are well established. 29 Pembrolizumab can maintain durable responses in advanced melanoma, 30 with 5 year survival rates approaching 40%. However predictive biomarkers of response remain elusive, which is key to in- in either clearance 11 or population PK model. 38 Given that clearance and clinical status at baseline and throughout treatment is associated with confounding BSLD, sufficient drug concentrations may not be reached at higher tumor burdens due to an intra-tumoral sink effect on PK with increasingly advanced disease. 41 Some patients, as identified by those with <median drug exposure and dramatically inferior OS in our work, may not reach effective exposure of pembrolizumab, leading to inferior outcomes.
Moving to potential future biomarkers, increased frequency of expression of TIM-3 and CXCR6 + on CD4 + T cells in metastatic melanoma patients with PD was confirmed ( Figure 5A). This correlates Clinically impactful inter-patient PK variability is seen in a real-world setting, and important limitations to the presumed effect of baseline advanced disease state and cachexia have been outlined. Modern frameworks for assessing the exposure-efficacy relationship within ICB monoclonal antibodies have moved from a simplistic one-way correlation between independent and dependent variables, respectively, to a complex multi-faceted set of interactions that likely are influenced multi-directionally and by baseline prognostic characteristics. 43 Given the large inter-individual exposure variability, mooted exposure-survival relationship and availability of a reproducible assay, further prospective therapeutic drug monitoring studies are planned in order to establish whether low pembrolizumab drug exposure is a modifiable variable that may lead to improvement in OS.
Upregulation of T-cell exhaustion and leukocyte trafficking markers on CD4 + cells was associated with PD. This is the first work outlining these relationships with pembrolizumab.

D I SCLOS U R E
AvdW and NAB receive funding for investigator-initiated clinical trials from Merck KGaA & Bristol Myers Squibb that is not related to this study. All other authors have no relevant conflicts of interest.

AUTH O R CO NTR I B UTI O N
VN collected and interpreted data and wrote the manuscript; MCG conceived the project, collected and analyzed data and contributed to the manuscript, GCM collected clinical data, HM collected clinical data and reviewing imaging findings, AvdW collected and interpreted data and contributed to the manuscript, NAB supervised the collection of data, analyzed and interpreted data and wrote the manuscript.

E TH I C S A PPROVA L A N D CO N S E NT TO PA RTI CI PATE
This study was carried out in accordance with the recommendations

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.