Introduction

Cutaneous squamous cell carcinoma (cSCC) represents a significant burden in dermatological malignancies, being the second most common skin tumor arising from the malignant progression of keratinocytes.1 It accounts for approximately 20% of all deaths from skin cancer globally, with a notable rise in its incidence worldwide. Despite its prevalence, most patients present with a localized disease that can be managed with surgical excision. However, a small yet critical subset of patients faces unresectable, locally advanced, or metastatic disease, for whom therapeutic options are limited.

Immune checkpoint inhibitors (ICIs) have revolutionized the treatment landscape of various malignancies, including cSCC. In 2018, the Food and Drug Administration approved cemiplimab, a high-affinity, highly potent human monoclonal antibody targeting programmed death 1 (PD-1).2 Later, in 2020, another compound was approved, pembrolizumab, following the results of the KEYNOTE-629 (NCT03284424), a multicenter, multicohort, non-randomized, open-label trial.3 Anti-PD1s have emerged as a favorable therapeutic option for patients with advanced cSCC who are not suitable candidates for curative surgery or radiation therapy.4 Clinical data supporting the efficacy of cemiplimab and pembrolizumab in advanced cSCC are robust, with notable response rates ranging from 50% to 60% and a favorable safety profile.2 5 Despite the remarkable clinical outcomes associated with anti-PD1 therapies, several critical clinical challenges persist within the field of cSCC management, increasing the need for reliable biomarkers that predict response to ICIs in patients with cSCC. The heterogeneous nature of cSCC underscores the complexity of its molecular landscape, necessitating the identification and validation of predictive and prognostic biomarkers to guide treatment decisions and optimize patient outcomes.

Here, we describe the immunological changes in the tumor microenvironment (TME) and peripheral blood of patients with cSCC undergoing treatment with cemiplimab. Transcriptional analysis of tumor specimens identified CD8+T cells as the most activated immune cell in the tumors, and interleukin (IL)8 and IL1β as the most downmodulated ILs in responder patients after cemiplimab treatment. Blood-based investigations confirmed the putative role of IL8 and highlighted regulatory T cell (Treg) PD1+ as early markers of response. IL8 peripheral levels decreased after one cycle of treatment only in responder patients and Treg PD1+decreased after 3 weeks of treatment in responders, while they showed a rebound enrichment in non-responders after three cycles of cemiplimab. In conclusion, our study showed that peripheral lymphocytes and cytokines, regulated by immune checkpoints, may reflect disease progression or regression and could serve as a useful non-invasive approach to enhance precision immune-oncology efforts.

MethodsStudy design and patients

This single-arm open-label translational prospective study was conducted in the Department of Clinical Medicine and Surgery of the University of Naples “Federico II” (Italy) and was reviewed and approved by the ethics committee. Written informed consent was obtained from all the participants enrolled in this study. Eligible patients were 18 years of age or older, with an Eastern Cooperative Oncology Group (ECOG) performance status ≤2, with a histologically confirmed diagnosis of cSCC, and planned to be treated with cemiplimab as first-line systemic therapy. For safety reasons, patients with conditions that could potentially elevate the risk of adverse events were excluded from the study. None of the patients included in this study had any known factors impacting immunotherapy activity, including recent or concurrent treatment with immunosuppressive drugs (chemotherapies for other tumors, corticosteroids, etc), previous transplant, or diagnosis of Chronic Lymphocytic Leukemia (CLL). Physical examination and imaging were used to evaluate the disease extent at baseline, and reassessments were performed every 2–3 months, or earlier if the patients showed signs and/or symptoms suspicious of disease progression. Objective responses were evaluated using Response Evaluation Criteria in Solid Tumors (RECIST V.1.1) criteria.6 Progression-free survival (PFS) and overall survival (OS) were estimated using the Kaplan-Meier method. PFS was evaluated from the start of treatment to disease progression or death from any cause, and OS from the start of treatment to death from any cause, with patients censored at the date of last follow-up if no event occurred. Tissue biopsies were performed at baseline (T0) and after 3 weeks of treatment (T1). For peripheral blood mononuclear cell (PBMC) and serum studies, blood samples were collected at baseline (T0), after 3 weeks (T1), after 9 weeks (T3) (figure 1). Tissue sampling and blood collection were performed at baseline (T0) for 21 and 44 patients, respectively. A new biopsy and blood samples were collected 3 weeks after the start of anti-PD1 treatment (T1 tissue n=12 matched, T1 blood n=42). Additional blood samples were collected after 9 weeks from the start of anti-PD1 treatment (T3 blood n=14) and at every following clinical/imaging disease evaluation (figure 1). Peripheral blood samples were analyzed for patients with all the available time points (T0, T1, and T3). All patients enrolled in this study received at least one administration of cemiplimab and were evaluated for objective response rates. In this cohort, 44 male patients (80%) and 11 female patients (20%) were enrolled.

Figure 1

Study workflow. Schematic representation of the translational study. Patients diagnosed with cutaneous squamous cell carcinoma (cSCC) and candidates for systemic cemiplimab treatment were enrolled in this study. Using the RECIST V.1.1 criteria, we categorized patients into responders and non-responders for our analysis. Responders achieved an objective response (complete response (CR) or partial response (PR) at the first assessment and did not progress for at least 6 months. Non-responders exhibited either stable or progressive disease at the first assessment. Tissue samples were collected at baseline (T0) and after 3 weeks of treatment (T1). Blood samples were collected at T0, T1, after 9 weeks (T3), and at each subsequent assessment (every 2–3 months, as detailed in the Methods section). RNA was extracted from tissue biopsies and analyzed using the NanoString nCounter PanCancer Immune Profile. Peripheral blood mononuclear cells (PBMCs) isolated from blood samples were used to study lymphocyte populations, whereas sera were analyzed for circulating cytokines. PD1, programmed cell death 1. Created by BioRender.com

The whole cohort of patients was divided for further analysis into: Responder patients, if the patient reported an early objective response (complete or partial) that persisted or improved in the following 6 months; and non-responder patients, if they reported stable disease (SD) or progressive disease (PD) at T3.

RNA extraction and quantification

RNA extraction was performed manually using the Maxwell RSC RNA formalin-fixed paraffin-embedded (FFPE) kit. Extractions using the Maxwell kit were automated using Promega’s Robotics platform. The FFPE tissues were scraped off with a clean razor blade and then placed in 1.5 mL tubes. After brief centrifugation, mineral oil was added, and the mixture was vortexed. The samples were heated at 80°C for 10 min and then allowed to cool to room temperature. A solution (master mix) containing Lysis Buffer, Proteinase K solution, and Blue Dye was prepared according to the protocol. The master mix was added to the sample tubes and vortexed. The tubes were incubated at 56°C for 15 min, and then at 80°C for 1 hour. Subsequently, DNase cocktail was prepared and added to the tubes. After further incubation and centrifugation, the aqueous phase was transferred to a Maxwell FFPE cartridge. Immediately after extraction, the total RNA concentration and A260:A280 ratio of each sample were assessed using a NanoDrop 2000 (Thermo Scientific).

nCounter gene-expression analysis

Samples were prepared according to the manufacturer’s guidelines and, for the most concentrated samples, diluted to ensure that each well’s final volume of 5 µl contained 150–200 ng of total RNA. Gene expression in immune cells was assessed using NanoString nCounter technology, using the nCounter PanCancer Immune Profile kit.

Two pairs of probes were used: reporter probes, which were labeled at the 5' end with a sequence of six fluorophores, the combination of which was specific to each gene, and capture probes, which were biotinylated at the 3' end. Both the probe pairs were hybridized to the target RNA. The samples were hybridized for 22 hours at+65°C. Excess probes were removed using the nCounter Prep Station, which employs magnetic beads. The next step involved immobilizing the target-probe complexes onto a streptavidin-coated cartridge, and aligning the reporter probes. Finally, the cartridge was transferred to the nCounter Digital Analyzer, where it was scanned, and each barcode was counted.

The differential expression analysis data model selected the most suitable statistical method for each gene based on its variable distribution, prioritizing the following model’s sequence: (1) mixture negative binomial model, (2) simplified negative binomial model, and (3) log-linear model. Adjustment of false discovery rate (FDR) p values was performed using the Benjamini-Hochberg method. All results were normalized using the geometric mean of the housekeeping genes.

PBMC and sera collection

Peripheral blood was collected in EDTA tubes and subjected to centrifugation at 250×g for 10 min at 4°C. Collected blood samples were diluted with an equal volume of phosphate-buffered saline (PBS) in a sterile tube. Diluted blood was carefully layered onto an equal volume of Ficoll-Paque in a fresh sterile tube, maintaining a 1:1 ratio. The tubes were centrifuged at room temperature at 400×g for 30 min without applying a brake. Following centrifugation, distinct layers were formed: PBMCs at the interface between the plasma and Ficoll-Paque, and erythrocytes and granulocytes at the bottom. The PBMC layers were gently collected using a sterile transfer pipette and transferred to sterile tubes. The collected PBMCs were washed two times with PBS by centrifugation at 250×g for 10 min each. After the final wash, the PBMC pellets were resuspended in fetal bovine serum and 10% Dimethylsulfoxide and stored at −80°C.

For sera extraction, peripheral blood was collected in serum separation tubes. The tubes were then centrifuged at 2000×g for 10 min at room temperature. Following centrifugation, a clear yellowish liquid (serum) formed on top of the clot. The serum was carefully collected using a sterile transfer pipette and transferred to a new sterile tube. Serum was stored at −80°C.

Flow cytometry analysis

Human CD4 T cell subsets were identified using the following antibodies: PerCP/Cyanine5.5 anti-human CD3 (SP34-2, BD Biosciences, 1:50), FITC anti-human CD4 (OKT4, BioLegend, 1:50), APC-Cy7 anti-human CD45RO (UCHL1, BioLegend, 1:50), PE anti-human CRTH2 (CD294) (BM16, BioLegend, 1:50), PE-Cy7 anti-human CXCR3 (1C6, BioLegend, 2:50), and APC anti-human CD196 (CCR6) (G034E3, BioLegend, 2:50). Human Treg were identified using the following antibodies: PE-Cy7 anti-human CD127 (IL-7Rα) (A019D5, BioLegend, 1:100), PE anti-human CD25 (B261355, BioLegend, 1:50), APC anti-human CD278 (ICOS) (C398.4A, BioLegend, 1:100), and APC-Cy7 anti-human CD279 (PD-1) (EH12.2H7, BioLegend, 1:100). Dead cells were excluded using the viability dye Zombie Yellow or Zombie Green (BioLegend). Cells were gated on live CD3+CD4+CD45RO+ cells and the Th subsets were identified as follows: Th1, CXCR3+CRTH2−CCR6−, Th2 as CRTH2+ and Th17 as CXCR3−CRTH2−CCR6+ and Treg as CD25+CD127low. Human innate lymphoid cells (ILC) were identified as lineage (Lin) negative and CD127 positive cells. Lin markers, all FITC-conjugated, include: anti-human CD3 (SP34-2, BD Biosciences, 1:50), anti-human CD4 (OKT4, BioLegend, 1:50), anti-human CD8 (MEM- 31, ImmunoTools, 1:50), anti-human CD14 (RMO52, BC, 2:50), anti-human CD15 (80H5, BC, 2:50), anti-human CD16 (3G8, BC, 1:100), anti-human CD19 (J3-119, BC, 1:100), anti-human CD20 (2H7, BioLegend, 1:100), anti-human CD33 (HIM3-4, BioLegend, 1:50), anti-human CD34 (561, BioLegend, 1:100), anti-human CD203c (E-NPP3, 1:50) (NP4D6, BioLegend, 2:50), anti-human FcεRIα (AER-37, BioLegend 2:50), anti-human CD56 (NCAM16.2, BD Biosciences, 1:50). Additional markers used include APC-Cy7 anti-human CD127 (IL-7Rα) (A019D5, BioLegend, 1:100), APC anti-human CD117 (cKit) (B290144, BioLegend, 1:50), and PE anti-human CRTH2 (CD294) (BM16, BioLegend, 1:50). Samples were acquired on a BriCyte E6 flow cytometer (Mindray Medical Italy Srl, Milan, Italy) or Sony MA900 flow cytometer (Sony Biotechnologies, San Jose, California, USA) and data were analyzed using FlowJo software (Tree Star V.10).

Multiplex cytokine assay

The concentrations of various cytokines in patients’ sera were determined using the multi-LEGENDplex analyte flow assay kit (Human Cytokine Panel 2 and Human Th Cytokine Panel, BioLegend) according to the manufacturer’s instructions. Samples were acquired at a BriCyte E6 flow cytometer. The results were analyzed using LEGENDplex software Qognit.

ResultsTreatment with ICI cemiplimab shows excellent response in patients with cSCC

Between December 2020 and December 2023, 55 patients diagnosed with advanced cSCC who were not considered candidates for curative treatment were enrolled in the study (figure 1). Patients’ baseline characteristics are shown in table 1. The median age was 83 years (IQR 62–93). The cohort included 45 (81.8%) patients with locally advanced cSCC, 6 (10.9%) with recurrent/refractory cSCC, and 4 (7.3%) with metastatic disease. All patients received cemiplimab as first-line systemic treatment, administered intravenously at a dose of 350 mg every 3 weeks. At the data cut-off (March 1, 2024), the median duration of follow-up was 14.5 months, and the median duration of treatment was 26 weeks (IQR 12–49 weeks). Median PFS was 18.6 months for the entire cohort, while median OS was not reached (figure 2a and b).

Table 1

Patients’ characteristics at baseline

Figure 2

Cemiplimab treatment delivers excellent outcomes in∼70% of patients. (a–b) Kaplan-Meier curves for progression-free survival (a) and overall survival (b) in responders (green), non-responders (blue), and the overall population (black). (c) Swimmer plot illustrating clinical responses at T3, including, where relevant, treatment interruption, disease progression, and death. (d) Diagram showing the availability of tissue and blood samples for each patient in the cohort, along with their best-reported response and responder/non-responder status. (e–f) Representative examples of patients responding (e) and not responding (f) to cemiplimab. CR, complete response; PBMC, peripheral blood mononuclear cells; PD, progressive disease; PR, partial response; SD, stable disease.

In our cohort, we observed 38 confirmed objective responses by RECIST V.1.1 (figure 2c), including 26 patients with complete response (CR, 47.3%) and 12 with partial response (PR, 21.8%). SD and PD were the best responses in 17 patients (30.9%).

For the purposes of this translational study, we analyzed the available samples from responder and non-responder patients. In particular, we selected among patients achieving an objective response (CR+PR) within three cycles of immunotherapy (T3), those who did not progress for an additional 6 months (responder patients). Conversely, the patients who reported SD or PD at T3 were considered non-responder patients (figure 2d). Median PFS was 4.87 months in the non-responder group and not reached in responders (figure 2a, HR 0.18, log-rank p<0.0001). Median OS was not reached in either group (figure 2b). Examples of responder versus non-responder cases are shown in figure 2e and f, respectively.

PD-1 blockade enhances early immune cell activation in patients’ tumor biopsies

Before delving into the modulations induced by PD-1 blockade, we sought to understand the landscape of the pretreatment TME and whether this might correlate with response to therapy. To address this, we profiled baseline tissue samples of 21 patients, using the nCounter NanoString immune-related panel, which encompasses genes related to innate and adaptive immune cell populations and their activation. Transcriptional analysis revealed that non-responders exhibited upregulation of the IL-2/STAT5 pathway and downregulation of interferon (IFN)-alpha and IFN-gamma pathways compared with responders’ tissues (figure 3a). Then, to investigate the early immune-related transcriptional changes in cSCC cells on PD-1 blockade, we collected matched pretreatment (diagnostic biopsy) and post-treatment (3 weeks after the first treatment) tumor specimens from 12 patients exhibiting different cemiplimab responses. Unsupervised hierarchical clustering revealed that transcriptional profiles were segregated primarily by treatment status (pretreatment or post-treatment) rather than by patient sample (figure 3b, online supplemental table 1). Comprehensive analysis of enriched pathways in the top differentially expressed genes (DEGs) revealed positive regulation of immune response and leukocyte activation/interactions as the most significant activated signaling pathways post-treatment (figure 3c). Gene Set Enrichment Analysis (GSEA) demonstrated significant enrichment of genes regulated by PD-1 blockade (revealing the bona fide effect of the treatment; Normalized Enrichment Score=2.03) and genes involved in the immune receptor activity (Normalized Enrichment Score=1.61) (figure 3d). Indeed, hierarchical clustering revealed that anti-PD1 treatment boosted the activation of dendritic cells, macrophages, and T cells in most samples, whereas no changes were observed in Treg cells (figure 3e, online supplemental table 2). Focusing on lymphocyte populations, we determined that responder patients displayed decreased T helper and CD4+T cells in tumor specimens after anti-PD1 therapy, whereas CD8+T cells and B cells increased (figure 3f). Interestingly, non-responder tissues showed an opposite trend on cemiplimab treatment (figure 3f). IHC assays were subsequently performed to validate the findings obtained through deconvolution analysis. IHC confirmed differential recruitment of CD4+T cells, CD8+T cells, and B cells in tumor tissues after 3 weeks of treatment in responders and non-responders (online supplemental figure 1a, b). Notably, the abundance of tumor-infiltrating lymphocytes (TILs) in pretreatment tumors was not significantly associated with OS and PFS (online supplemental figreu 2a–b), further emphasizing the importance of evaluating changes on cemiplimab treatment—more than the baseline immunity—as determinants of response.

Figure 3

Cemiplimab modulates the immune cell landscape in cutaneous squamous cell carcinomas. (a) Gene Set Enrichment Analysis of differentially expressed genes in non-responders versus responders pretreatment biopsies. (b) Unsupervised hierarchical clustering of 12 matched biopsy samples at baseline (T0) and after 3 weeks of treatment (T1). (c) Gene Ontology and (d) Gene Set Enrichment Analysis of the top 119 differentially expressed genes in (b). (e) Unsupervised hierarchical clustering of immune cell-related gene signatures. (f) Stacked barplot representation of CIBERSORT deconvoluted gene expression profiles of responder patients (R) and non-responders (NR) at T0 and T1. Cell-type Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORT);DC, dendritic cell; NES, Normalized Enrichment Score; Treg, regulatory T cells.

IL1β and IL8 down-regulation in tissues’ specimens is associated with response to cemiplimab treatment

GSEA analyses were performed on the DEGs separately for responders and non-responders, comparing T1 versus T0 samples. Notably, only in responder patients, the treatment with cemiplimab was associated with significant downregulation of the “TNF-α signaling via NF-κB” Hallmark signature (Normalized Enrichment Score=−2.3, FDR q-value=0.0, figure 4a). In contrast, no such downregulation was observed in non-responders. This pathway is implicated in fostering an immunosuppressive TME, thereby facilitating immune escape and resistance to immunotherapy, as well as angiogenesis, invasion and metastatic processes.7 Next, genes exhibiting the most robust and statistically significant downregulation following cemiplimab treatment were subjected to Gene Ontology analysis. We observed a significant enrichment of genes associated with chemokine and cytokine signaling pathways, particularly the IL-1 receptor pathway (figure 4b–c). To further investigate this observation, cytokine and chemokine expression profiles were assessed in responders and non-responders. Notably, two key protumorigenic ILs, IL1β and IL8, displayed significantly decreased expression exclusively in responders (figure 4d) with logFC values of approximately −2.5 and −4, respectively, in post-treatment biopsies compared with pre-cemiplimab specimens (figure 4e). These findings suggest that cemiplimab-mediated modulation of these IL signaling pathways potentially contributes to its therapeutic efficacy in cSCC. To ascertain whether these genes had prognostic/predictive value, we queried publicly available clinical and gene expression data via the “Kaplan-Meier Plotter” database (www.kmplot.com). Specifically, we examined the association between the expression of each IL across all tumor types and clinical outcomes, in patients treated with pembrolizumab (anti-PD1). Intriguingly, our analysis revealed that patients receiving anti-PD1 treatment with elevated levels of IL1β and IL8 (also referred to as CXCL8) had a higher risk of progression compared with patients with low IL1β and IL8 levels (HR = 4.37 and 6.09, respectively; figure 4f and g). Further, to evaluate these cytokines as potential predictive biomarkers, receiver operating characteristic curves were generated for CXCL8 and IL1β, including all tumor types treated with immunotherapies. We found that CXCL8 gene expression levels, but not IL1β, during treatment with ICIs had a statistically significant area under the curve of 0.663 and 0.746 for any ICIs and anti-PD1, respectively (online supplemental figure 3).

Figure 4

Cemiplimab treatment downregulates IL1β and IL8 in responsive cSCC tissues. (a) Gene Set Enrichment Analysis of differentially expressed genes in T1 versus T0 tumor biopsies in responder patients. (b) String analysis of the most significantly downregulated genes in the samples at T1. (c) Gene Ontology analysis of (b). (d) Volcano plot of differentially expressed genes in T1 compared with T0 in responders. (e) Violin plots of gene expression of IL1β (on the left) and IL8 (on the right) in responder patients, comparing T0 and T1. (**p value<0.01; statistical differences were assessed using one-way ANOVA test with Dunnett’s correction for multiple testing). (f–g) Kaplan-Meier curves of progression-free survival (PFS) in patients with high versus low expression of IL1β (e) or IL8 (f),during the treatment with the anti-PD1 pembrolizumab. ANOVA, analysis of variance; cSCC, cutaneous squamous cell carcinoma; FDR, false discovery rate; IL, interleukin; PD1, programmed cell death 1.

Peripheral lymphocytes abundance predicts anti-PD1 treatment efficacy

In light of the observed modulation of lymphocyte activity and tissue-specific cytokine and chemokine profiles, we sought to elucidate the corresponding changes in the peripheral blood. Thus, to evaluate the potential for minimally-invasive prediction of early treatment response, we first assessed changes in lymphocyte abundance in peripheral blood within a subset of our cohort (figure 2d). Specifically, we compared lymphocyte counts after one (T1) and three cycles (T3) of cemiplimab treatment with pretreatment (T0) levels in both responder and non-responder patients. Given that RNA profiling of tumor biopsies revealed T cells as the most activated lymphocytes after cemiplimab treatment, we investigated whether this modulation was also reflected in the peripheral blood of patients. Among all the T cell subtypes, only T helper cell types changed in abundance in the peripheral blood. Notably, Th1 cells exhibited minimal changes after cemiplimab treatment in the entire cohort. In contrast, Th2 and Th17 cell counts decreased following two treatment cycles (figure 5a). However, unlike the analysis of tissue biopsies, the modulation of T helper cells did not distinguish between responder and non-responder patients. Intriguingly, we observed parallel modulation in ILC subsets defined as the innate counterparts of T helper cells.8 In particular, type-1 ILCs (ILC1) abundance decreased in both responder and non-responder groups, whereas ILC2 and ILCp populations increased (figure 5b). Moreover, consistent with our previous transcriptional profiling results, Treg populations within the CD3+/CD4+subset did not exhibit significant changes following cemiplimab treatment across the entire cohort (figure 5c–d). This common trend in both responder and non-responder groups confirmed the limited role of Tregs as determinants of response. Nevertheless, by focusing on the expression of PD1, we identified a subpopulation of Treg PD1+ cells with a differential trend between responders and non-responders (figure 5e). Indeed, the divergent patterns observed in responders versus non-responders revealed their differential treatment outcomes (figure 5f, left and right panels). In line with the action of cemiplimab, which specifically targets PD1+ cells, responder patients had decreased PD1+Treg cells as early as after the first cycle of therapy. Conversely, in non-responder patients, after an initial decrease, the PD1+Treg population increased at T3, representing an early manifestation of resistance to treatment. This finding suggests the potential for monitoring peripheral blood PD1+Tregs to evaluate anti-PD1 treatment efficacy, in addition to tumor specimens monitoring.

Figure 5

Peripheral blood PD1+Treg abundance may predict response to cemiplimab therapy in patients with cSCC. (a–b) Bar plots of the percentage of T helper cells (a) and innate lymphoid cells (b) in the overall cohort at T0 (circle), T1 (square) and T3 (triangle). (c–d) Bar plots of CD3+CD4+ (c) and Treg (d) cells abundance in the overall cohort at T0, T1 and T3. (e) Representative examples of flow cytometry analysis of PD1+Treg cells in a responder (R) and a non-responder (NR) peripheral blood. (f) Bar plots of (e). cSCC, cutaneous squamous cell carcinoma; FSC, forward scatter; ILC, innate lymphoid cell; PD1, programmed cell death 1; TH, T helper; Treg, regulatory T cells.

Predictive significance of IL8 concentrations in the peripheral blood of patients with cSCC

To better understand the impact of immunotherapy on activated immune cells and to determine potential correlations between lymphocyte population changes and cytokine/chemokine levels in peripheral blood, we analyzed a panel of 25 molecules using a multiplex array. First, we assessed whether the cytokine/chemokine profile in patients differed between responders and non-responders at baseline, and we found that none of them showed a statistically significant difference in concentration between the two groups (online supplemental figure 3a). Instead, focusing on the modulation following cemiplimab treatment, we observed alterations in various cytokines and chemokines in the peripheral blood of patients (online supplemental figure 4a). We specifically aimed to identify molecules exhibiting differential patterns between responders and non-responders to pinpoint potential markers of response (figure 6a–b). While responders exhibited minimal changes or a slight decrease in the levels of TNF-α, IL5, IL6, IL9, IL27, and IL33, the non-responder group showed an increase in these cytokines (online supplemental figure 5a–f). This suggests a potential role for these cytokines in mediating treatment resistance. Conversely, IL11 and IL13 levels exhibited opposing trends, with a decrease in responders and an increase in non-responders post-treatment, although without reaching statistical significance (online supplemental figure 5g–h). Next, we compared our peripheral blood cytokine results with those obtained from tissue biopsy analysis. In peripheral blood, IL1β did not exhibit a differential pattern between responders and non-responders but decreased in both groups (figure 6c). In contrast, IL8 maintained its ability to discriminate treatment response, with circulating levels significantly decreased in responder patients after one cycle of treatment, whereas non-responders displayed an increase (figure 6d). Thus, our findings provide compelling evidence supporting IL8 as a promising biomarker for predicting response to cemiplimab treatment in patients with cSCC.

Figure 6

IL8 in peripheral blood is a potential marker of response to anti-PD1 therapy in cSCC. (a–b) Heatmaps showing cytokine levels in the sera of responders (a) and non-responders (b). The x-axis represents different cytokines, the y-axis represents time points (T0, T1, T3), and the color scale indicates the concentration of cytokines in pg/ml. (c) Bar plots of IL1β data in responders (green) and non-responders (blue) from (a). (d) Bar plots of IL8 data in responders (green) and non-responders (blue) from (a). (*p<0.05; statistical differences were assessed using t-test). cSCC, cutaneous squamous cell carcinoma; IL, interleukin; PD1, programmed cell death 1.

Discussion

In this study, we demonstrate that cemiplimab induces significant transcriptional changes in the TME of cSCC, particularly by enhancing T cell activation and modulating cytokine signaling. At baseline, patients who did not respond to treatment showed modulation of pathways sustaining an immunosuppressive milieu (high IL2/STAT5 and low IFN-alfa and IFN-gamma Hallmark signatures). Interestingly, patients who achieved an early and sustained response to cemiplimab (responders) exhibited increased infiltration of B cells and CD8+T cells in the TME three weeks after the initial treatment. Additionally, we observed downregulation of the cytokines IL1β and IL8 in tissue biopsies of responders, and of IL8 also in peripheral blood responder samples. Conversely, non-responder patients exhibited opposite trends in these circulating cytokines, especially IL8, which may serve as a potential biomarker for predicting treatment response. Furthermore, we identified a differential modulation of PD1+ regulatory T cells (Tregs) between responder versus non-responder patients, suggesting its potential role in longitudinal patient monitoring.

The introduction of cemiplimab for the treatment of advanced cSCC in 2018 replaced platinum-based therapies, which have been used as a standard treatment for various cancers, including advanced cSCC, for decades. Platinum therapy poses challenges due to the advanced age and comorbidities, particularly in terms of renal function, of the patient population. In this setting, cemiplimab has been a revolution, although in the registrative trials the benefit was limited to ∼50% of patients. In real-world settings, objective response rates between 50% and 77% have been reported,9 which is similar to our cohort results (∼70%). Nevertheless, there is a pressing need to identify biomarkers to predict treatment response, given the significant impact of cSCC on patients’ quality of life and prognosis. Moreover, a comprehensive understanding of the influence of cemiplimab on antitumor immunity, both locally—within the TME—and systemically, is still lacking.

Our study elucidates the effects of ICIs in the tumor specimens of patients with cSCC. Of note, our study focuses exclusively on cemiplimab as the anti-PD1 therapy under investigation, since this agent is currently the only anti-PD1 therapy approved in Italy for patients with locally advanced or metastatic cSCC. Nonetheless, we propose that the findings presented here may have applicability to patients treated with other anti-PD1 agents, offering insights relevant to the broader landscape of PD-1 inhibition in cSCC management. Across the entire cohort, cemiplimab treatment led to the upregulation of immune-related gene signatures. However, we found a distinct modulation of immune cell subtypes in responder and non-responder patients, at least in part due to a different landscape even before treatment initiation. In non-responders baseline tissues, the activation of IL-2/STAT5 signaling suggests a preferential differentiation of infiltrating T cells toward a Treg-dominated environment, potentially contributing to immune evasion.10 Moreover, the suppression of IFN pathways in non-responders could impair the recruitment of cytotoxic immune cells, fostering an immunosuppressive microenvironment. Consistent with this hypothesis, using Cell-type Identification by Estimating Relative Subsets of RNA Transcripts (CIBERSORT) deconvolution analysis, we found that non-responder tissues exhibited a decrease in B cells and CD8+T cells and an increase in T helper subtypes after cemiplimab treatment. Conversely, responders showed enrichment of B cells and CD8+T cells following treatment, reflecting a potential cytotoxic immune environment. Peripheral blood analysis revealed overall stability in Th1 cells, and reduction in Th2 and Th17 cells, indicating that the results from the TME were not systemically mirrored. Although different studies have highlighted pretreatment TILs as a crucial marker of response in solid tumors (eg, melanoma,11–13), our results emphasize the importance of verifying immune modulation on pharmacological treatment.

The inflammatory cascade both locally—in the TME—and systemically is driven by cytokine production and regulation.9 We identified IL8 and IL1β as the most downregulated cytokines in the tumor biopsies of responder patients, aligning with their reported protumorigenic roles in invasion, metastasis, and chemotherapy resistance.14 15 High levels of these cytokines are correlated with poorer prognosis in different cancer types, and a higher risk of progression following anti-PD1 therapy. Moreover, targeting IL1β and IL8 enhances antitumor immune responses, promoting TME remodeling, and increasing T-cell infiltration, thus determining tumor regression as a result.16 17 IL1β is a well-characterized activator of Th17 cells that subsequently promotes IL8 production in neighboring cells.18 19 Interestingly, IL1β has been described as a critical regulator of ILCs function and plasticity, demonstrating therapeutic benefit in chronic inflammatory diseases.20 Moreover, a population of CD56+ ILC1-like cells has been demonstrated to secrete IL8 under proper stimulation in acute myeloid leukemia setting.15 Our findings suggest that this IL1β/IL8 axis could be involved in the modulation of both the adaptive and innate arms, supporting the protumorous microenvironment and therapy resistance. In our cohort, cemiplimab treatment appears to disrupt this inflammatory loop in responsive patients, leading to a marked and statistically significant reduction in IL8 levels, probably also by interrupting the angiogenic axes TNF-α/NF-kB.21 Intriguingly, changes in IL8 are also observable in the peripheral blood of responder patients, before evident tumor response (T1). Early downregulation of IL8 levels may serve as a predictor of favorable clinical outcomes, aligning with previous reports implicating IL8 as a potential biomarker for anti-PD-1 therapy response.22 Further prospective studies are warranted to validate the clinical utility of IL8 as a biomarker and to explore its potential role in guiding personalized treatment decisions for patients with cSCC receiving cemiplimab immunotherapy.

Recent studies have associated increased serum levels of IL6 after cemiplimab treatment in patients with cSCC with poorer responses.23 Consistently, non-responder patients exhibited an increase in IL6 after just one treatment cycle, unlike responders whose levels remained unchanged. Similarly, IL9, IL27, and IL33 with known protumorous roles and involvement in regulating the immune response against the tumor, remained unchanged after cemiplimab treatment in responders, whereas their levels seemed to increase in non-responders, suggesting a potential role as biomarkers for poor prognosis after cemiplimab treatment. Moreover, responder patients displayed a slight decrease in IL13, IL11, and TNF-α levels, whereas non-responders showed increasing levels, in line with their role in promoting immune evasion and tumor progression.

Our study also provides the first evidence that the modulation of circulating PD-1+Tregs could indicate the efficacy of cemiplimab treatment. In the TME, PD-1 positive Tregs may be converted into immunosuppressive effector Tregs upon PD-1 blockade, promoting tumor progression.24 We observed a decrease in PD-1+Tregs in responder patients, and importantly, this decrease could be monitored in peripheral blood, offering a less invasive alternative to tissue biopsies.

In conclusion, our study identifies unique transcriptomic signatures and reliable biomarkers for predicting response to ICIs. To our knowledge, this is the first study to offer intriguing insights into the potential of peripheral blood as a surrogate marker for TME dynamics and treatment response in cSCC. By examining immune modulation induced by ICIs in tumor samples and peripheral blood, we provide insights into key events associated with positive outcomes in patients with cSCC treated with ICIs and pave the way for novel combination therapies for resistant patients.