Eligible patients had newly diagnosed and histology-confirmed OSCC, with clinical T1 or T2 (American Joint Committee on Cancer, 8^th Ed [AJCC 8]) without nodal involvement (N0) and distant metastases (M0) assessed by imaging studies (TNM Stages I and II), and Eastern Cooperative Oncology Group (ECOG) Performance Status ≤2. The patients were treatment-naïve and planned for surgical resection with curative intent.

This was a single-arm, open-label pilot trial to assess the feasibility, efficacy, and safety of utilizing imiquimod 5% cream applied topically in a neoadjuvant setting for 28 days immediately followed by surgical resection of the tumor. The trial was conducted following the protocol approval from the Institutional Review Boards. Patients were enrolled at Columbia University Irving Medical Center (CUIMC) in New York, NY from September 2021 to February 2022, and at Medical University of South Carolina (MUSC) in Charleston, SC from May 2022 to September 2023.

Topical imiquimod 5% cream was self-administered starting day 1 of a 28-day cycle. Patients applied the cream directly onto the tumor and the surrounding area at bedtime, left it on for 20 minutes, and rinsed thoroughly. Self-application of imiquimod was performed daily for 28 days, and the date and time of the application were documented in the Daily Diary form. Clinical tumor size was measured at the baseline, midpoint (14 days into the trial), and post-therapy by measuring the longest perpendicular bidirectional size of the clinically visible lesion. The adverse events (AEs) were recorded and graded for safety evaluation according to the NCI Common Terminology Criteria for Adverse Events (CTCAE version 5.0) throughout the treatment period.

The primary endpoint was a minimum of 50% reduction in tumor cell count assessed by quantitative multiplex immunofluorescence (qmIF) within the tumor bed of the surgical tissue (post-treatment) compared to the biopsy tissue (pre-treatment). In addition, the major pathologic response was assessed using Immune-Related Pathologic Response Criteria (irPCR) in the tumor bed of the post-treatment surgical tissue. The secondary endpoint was treatment-related toxicity graded by CTCAE, defined as safe if no life-threatening (Grade 4) is reported. The correlative endpoint was a shift in antitumor immune profile in post-treatment tissue compared to pre-treatment tissue measured by qmIF. PD-L1 expression was evaluated on tumor and tumor-infiltrating immune cells to assess the combined positivity score (CPS). The electronic chart review was conducted every 12 weeks to determine recurrence and survival status. The one-year recurrence-free survival (RFS) was measured from the time of surgery to the time of biopsy-confirmed recurrent OSCC.

The study was approved by the Institutional Review Board and conducted in accordance with institutional and federal guidelines for human investigation in accordance with the Declaration of Helsinki. All participants were informed of the investigational nature of the study and provided written informed consent prior to enrollment. Ongoing safety oversight was conducted by the Institutional Review Board and Safety Monitoring Committee.

Immune-related Pathologic Response Criteria were used to assess the pathologic response to immunotherapy (19–21). The H&E stained slides were prepared from formalin-fixed, paraffin-embedded surgical specimens. Two board-certified pathologists blinded to patient outcome scored the slides by identifying the areas of tumor bed and residual viable tumor. The tumor bed was defined as the sum of 1) residual viable tumor (RVT), 2) necrosis, and 3) regression bed characterized by the presence of dense tumor-infiltrating lymphocytes (TIL) infiltrates, tertiary lymphoid structure (TLS), features of cell death (keratinous debris and giant cells), and tissue repair (neovascularization, fibrosis). The %irRVT was calculated as RVT area/tumor bed area x 100. If the estimate of %irRVT differed by >10% between two pathologists, a third pathologist served as an adjudicator to finalize the score. pCR (complete response) was defined as 0% irRVT, irMPR (major pathologic response) as <10% irRVT, PR (partial response) as >10 to 90% irRVT, and irNR (no response) as >90% RVT.

Tissues were macrodissected to separate the tumor bed from the overlying surface epithelium and surrounding stroma in the biopsy and surgical tissues by a pathologist blinded to the pre- and post-treatment samples. Tumor and immune cells in the TME were quantified using qmIF. The slides were stained with pan-cytokeratin using Lunaphore Comet automated stainer (Fortis Life Sciences, TX) for quantification of pan-cytokeratin (pan-CK) positive tumor cells (22). In addition, tumor cells were assessed for proliferating cell nuclear antigen (PCNA) expression. Similarly, the immune profile between the pre- and post-treatment TME was compared using qmIF. To characterize immune cells, the markers for CD3 (T-cells), FoxP3 (Regulatory T-cells), CD4 (Helper T-cells), CD8 (Cytotoxic T-cells), PD-L1 (Immune response co-inhibitory factors on tumor and macrophages), CD45 (Lymphocytes), CD45RO (memory T-cells), Granzyme B (T-cell activation), CD68 (Macrophages) were used.

The digital image analysis and quantification, segmentation, and classification of immunofluorescence-positive cells were meticulously overseen by a pathologist, adhering to the methodology previously outlined (23). Initially, images were imported into the QuPath (RRID: SCR_018257) version 0.3.2. Cellular segmentation was executed using the StarDist algorithm. To ensure accuracy, marker-specific thresholds were established using a composite image that encapsulated representative samples from the entire image batch. These individual thresholds were amalgamated to form a composite classifier, enabling the categorization of each cell into distinct classes based on marker positivity. Within 6-8 regions of interest (ROI), quantitative analyses were conducted to enumerate the cells corresponding to each marker or combination thereof. This count was then standardized against the total area of the tissue region under examination, thereby deriving cell-type density metrics and expressed as cells/mm².

Spatial Score was calculated to generate mean spatial ratios per sample between permutations of sets of 3 cells (26, 27). The X/Y coordinates for each cell type were determined during cellular segmentation. The minimal distance between each cell type and its nearest other cell types and the averages of these minimal distances per tissue spot were calculated in a python script adapted from the original R script (github.com/nolanlab/SpatialScore). To assess the relationship of cell distances between three cell types [Tumor cell (C1), T-cell (C2), and Macrophage (C3)], the minimal distances between CT1-CT2 (left distance) versus CT2-CT3 (right distance) were calculated. This distance ratio (Spatial Score) was determined as left distance/right distance. The median ratio from all samples was compared between pre- and post-treatment.

The tumor proportion score (TPS), which measures the proportion of residual viable tumor cells expressing PD-L1, and the combined positive score (CPS), which measures the proportion of PD-L1 on both tumor cells and macrophages, were assessed using qmIF (24, 25). CPS was defined as the proportion of PD-L1+ tumor cells divided by the sum of PD-L1+ tumor cells and PD-L1+ immune cells. The TPS cutoff values were 1% and 50%. The threshold for CPS was <1%, 1-9%, 10-19%, and >20%.

Based on the sample size calculation, with 15 subjects and an estimated response rate of 0.5, we would achieve the 95% confidence interval (CI) of 0.239 to 0.761, using a two-sided exact binomial test with a type I error of 0.05. Baseline characteristics were summarized using median and range for continuous variables and frequency and percent for categorical variables. For tumor size, we reported median percent change from pre- to post-therapy and tested the significance of the change from baseline using the Wilcoxon signed rank test with continuity correction. The mean pre- and post-treatment tumor cell count/mm² is reported with standard deviation, and the significance of differences in cell counts was assessed using the Wilcoxon signed rank exact test and Wilcoxon signed rank test with continuity correction. The proportion of patients experiencing a reduction in tumor cell count/mm² from pre- to post-therapy of at least 50% was summarized as a percentage and an exact binomial 95% confidence interval. The percentage of patients with irMPR, PR, and irNR in surgical specimens was reported with the range. The significance of the change in cell counts for immune markers comparing post-therapy to pre-therapy was performed using Wilcoxon signed rank test with continuity correction as needed. The PD-L1 expression assessed in terms of TPS and CPS, and the spatial ratio between macrophages, T-cells, and tumor cells are reported in descriptive summaries. Recurrence-free survival (RFS) was defined as the time from surgery to disease recurrence/death due to any cause or date of last follow-up, whichever occurred first. RFS was censored for patients free of recurrence at the last date of follow-up. Only patients with a recurrence or minimum follow-up of 1 year were evaluated and included in the one-year RFS analysis. Due to only one recurrence among 14 RFS-evaluable patients, analysis was limited to descriptive summaries. The median follow-up time for the evaluation of RFS was calculated using the reverse Kaplan-Meier (KM) method. All analyses were performed using R version 4.3.0. All hypothesis tests were two-sided, and p-values < 0.05 were considered statistically significant.

A total of sixteen patients with new primary early-stage oral cancer were enrolled from 2021 to 2023, and fifteen patients received trial therapy and were evaluated for safety and anti-tumoral response. One patient with a history of myocardial infarction had sudden cardiac death less than two weeks into the trial and was subsequently excluded from analyses. All 15 patients underwent surgical resection within two weeks of imiquimod therapy completion. The patient characteristics of fifteen evaluable patients are in Table 1 .

Adverse events (AEs), at least possibly related to the study treatment, occurred in 14 out of 15 patients. Table 2 summarizes the number of patients with treatment-related toxicities by NCI CTCAE grade. The most common therapy-related AEs were application site mucositis and discomfort. Short dose interruption (<10% of total treatment duration) occurred in two patients (13%) due to toxicity interfering with oral intake. Two patients (13%) reported fatigue during the first two weeks following the start of the study, with subsequent resolutions. No life-threatening (grade 4) AEs were reported. All patients completed the imiquimod therapy and received scheduled surgery. AEs reported by individual participants are shown in Supplementary Table 1 .

A median of 45% reduction in clinical size (IQR = -55, -28; p=0.001) was noted. The clinical tumor size was measured by examining the longest perpendicular bidirectional dimensions ( Figure 1A ). Fourteen (93%) patients experienced a regression in clinical tumor size after the treatment. One patient had no change in the overall size. However, the rough and irregular surface of the exophytic tumor at baseline regressed to a nearly flat lesion by the endpoint ( Supplementary Figures 1A-F ).

The tumor cell numbers were quantified using quantitative multiplex immunofluorescence (qmIF) from the tumor bed macrodissected from the biopsy and surgical tissues ( Figure 1B ). A significant reduction in tumor cell count was observed post-treatment compared to pre-treatment (p=0.007) ( Figure 1C ). Nine (60%) patients experienced a >50% reduction in tumor cell count after the treatment (95% CI = 32% to 84%). The baseline expression of PCNA on tumor cells was low. An insignificant decrease in PCNA expression was observed after the treatment.

The pathologic response in terms of irRVT ranged from 0 to 65%, with a median of 10% irRVT. The tumor regression bed was characterized by inflammatory infiltrates, keratin debris surrounded by histiocytes, neovascularization, and fibrosis. None of the cases exhibited necrosis. Based on the analysis of surgical specimen using irPRC, nine patients (60%) had irMPR ( Figure 2A ), two (13%) of whom had a complete pathologic response with no residual tumor ( Figure 2B ). Six patients (40%) had PR ( Figure 2C ). All patients showed pathologic response to immunotherapy (no irNR), as illustrated in Figure 2D . The change in clinical tumor size, tumor cell count, %irRVT, and pathologic response in terms of pCR, irMPR, PR, and irNR for individual patients is summarized in Supplementary Table 2 .

To assess the antitumor immune activity, immune cell composition in the tumor bed from paired pre- and post-treatment tissues was evaluated in all fifteen patients by qmIF. As shown in Figure 3 , the most striking findings were a 709% increase in cytotoxic T-cells (p=0.01) and a 593% increase in helper T-cells (p=0.003). 503% and 461% increase in memory cytotoxic and helper T-cells, respectively, were also noted. Although not significant, an increase in regulatory T-cells by 380% was detected. Spearman’s rank correlation assessed the contribution of changes in various tumor-infiltrating lymphocytes (TILs) to the reduction in tumor cell count ( Table 3 ). Both activated cytotoxic and helper T-cells significantly correlated with the reduction in tumor cell count (R=0.55; p=0.036 and R=0.54; p=0.039, respectively), but not the memory T-cells.

The spatial topography of helper and cytotoxic T-cells to tumor cells relative to the proximity of these T-cells to macrophages was measured to explore their spatial relationships. In the pre-treatment tissue, a lower ratio (helper and cytotoxic T-cells in close proximity to tumor than to macrophage) was observed. After the treatment, the helper and cytotoxic T-cells were in closer proximity to macrophage than tumors resulting in a higher spatial ratio as shown in Figure 4 . The helper and cytotoxic T-cells and all T-cells were closer in distance to tumor cells before the therapy and shifted closer to macrophages post-therapy.

PD-L1 expression analysis was performed using the quantified data obtained from qmIF. Supplementary Figure 2 shows a heatmap of the total tumor cell count, total number of tumor and immune cells expressing PD-L1, and PD-L1+ tumor and macrophage count within the tumor bed. As shown in Table 4 , there was no change in the median PD-L1+ tumor cell numbers post-treatment. However, there was a slight reduction in PD-L1+ macrophage cell counts (p=0.041), while the median percent change showed an increase (158%, p=0.007).

TPS accounting for proportions of PD-L1+ viable tumor cells and CPS inclusive of PD-L1+ tumor and immune cells correlated with irRVT and cancer recurrence is shown in Table 5 . TPS positivity (≥1%) was found in 7 patients pre-treatment and 7 patients post-treatment. At baseline, CPS <1% was observed in 7 patients, 1-9% in 5 patients, 10-19% in 0 patients, and >20% in 3 patients. Following imiquimod therapy, CPS <1% was observed in 2 patients, 1-9% in 7 patients, 10-19% in 0 patients, and >20% in 6 patients. Pre-treatment TPS and CPS were <5% (range 0-3%) for all nine patients with irMPR. Interestingly, one patient who had recurrent cancer exhibited the highest pre-treatment CPS of 102%.

Fourteen patients had sufficient follow-up and were evaluable for one-year RFS via electronic chart review. Median follow-up was 17 months (95% CI = 16 months - not reached) based on the reverse KM approach. Following neoadjuvant imiquimod therapy and surgical resection, 13 (93%) of 14 patients had recurrence-free 1-year survival. One patient developed a recurrent OSCC immediately adjacent to the previous tumor site approximately six months after the surgery despite negative surgical margins. The lesion was subsequently resected, and the patient had no further evidence of disease. The duration of follow-up and the one-year recurrence-free survival status of individual patients are shown in Table 6 .