Objective: Shape-sensing robotic-assisted bronchoscopy is an emerging technology for the sampling of pulmonary lesions. We seek to characterize the shape-sensing robotic-assisted bronchoscopy learning curve at an academic center. Methods: Shape-sensing robotic-assisted bronchoscopy procedures performed by 9 proceduralists at a single institution were analyzed. Cumulative sum analyses were performed to examine diagnostic sampling and procedure time over each operator's first 50 cases, with the acceptable yield threshold set to 73%. Results: During the study period, 442 patients underwent sampling of 551 lesions. Each operator sampled 61 lesions (interquartile range, 60-63 lesions). Lesion size was 1.90 cm (interquartile range, 1.33-2.80 cm). The median procedure time for single-target cases decreased from 62 minutes during the first 10 cases to 39 minutes after case 40 (P < .001). The overall diagnostic yield was 72% (range, 58%-83%). Six of 9 operators achieved proficiency over the study period. An aggregated cumulative sum analysis of those who achieved competency demonstrated a steep improvement between lesions 1 and 21 and crossing of the competency threshold by lesion 25. Temporal analysis of yield-related lesion characteristics demonstrated that at approximately lesion 20, more challenging lesions were increasingly targeted, as evidenced by smaller target size, higher rates of unfavorable radial endobronchial ultrasound views, and a negative bronchus sign. Conclusions: Skills acquisition in shape-sensing robotic-assisted bronchoscopy is variable. Approximately half of proceduralists become facile with the technology within 25 lesions. After the initial learning phase, operators increasingly target lesions with more challenging features. Overall, these findings can inform certification and competency standards and provide new users with expectations related to performance over time.

Supplemental Figure S2. Relative expression on a per cell basis in KitV558Δ/+ tumors of (A) Trdv4 and (B) multiple genes known to be members of the SCART1+ γδ T cell phenotype. Dot plots represent scRNAseq data of γδ T cells from 6 mice as detailed in the methods section.

Viability of HG129 cells or T1 GIST cells after 24 hours of (A, B) human IL17a or (C, D) anti-human IL17Ra. Representative of 2 separate experiments with experimental groups in triplicate. Data represent mean ± SEM; p values were calculated using a two-sided Student’s t test; *, P < 0.05.

Abstract γδ T cells are a rare but potent subset of T cells with pleiotropic functions. They commonly reside within tumors but the response of γδ T cells to tyrosine kinase inhibition is unknown. To address this, we studied a genetically engineered mouse model of gastrointestinal stromal tumor (GIST) driven by oncogenic Kit signaling that responds to the Kit inhibitor imatinib. At baseline, γδ T cells were antitumoral, as blockade of either γδ T-cell receptor or IL17A increased tumor weight and decreased antitumor immunity. However, imatinib therapy further stimulated intratumoral γδ T cells, as determined by flow cytometry and single-cell RNA sequencing (scRNA-seq). Imatinib expanded a highly activated γδ T-cell subset with increased IL17A production and higher expression of immune checkpoints and cytolytic effector molecules. Consistent with the mouse model, γδ T cells produced IL17A in fresh human GIST specimens, and imatinib treatment increased γδ T-cell gene signatures, as measured by bulk tumor RNA-seq. Furthermore, tumor γδ T cells correlated with survival in patients with GIST. Our findings highlight the interplay between tumor cell oncogene signaling and antitumor immune responses and identify γδ T cells as targets for immunotherapy in GIST.

(A) Prevalence of IL17A+ γδ T cells in the spleens of KitV558Δ/+ mice after vehicle or imatinib treatment for 1 week. (B) Relative expression of Il6 by PCR of bulk KitV558Δ/+ tumors after 1 week of vehicle or imatinib. (C) Tumor weight in KitV558Δ/+ mice with 2 weeks of anti-IL23 therapy or isotype. Data are representative of 2 experiments each, 4 mice/experiment. Data represent mean ± SEM; p values were calculated using Student’s t test.

Supplemental Figure S1. (A) Alternate gating strategy for identifying γδ T cells in KitV558Δ/+ tumors by flow cytometry (B) Prevalence of γδ T cells as a percentage of T cells in KitV558Δ/+ mice across multiple anatomic compartments. (C) Percent of cells positive for IFNγ in KitV558Δ/+ tumor by flow cytometry after intracellular cytokine stimulation. (D) Histograms of flow cytometry staining for the transcription factors T-bet and EOMES in T cells from KitV558Δ/+ tumors. (E) Vγ2 expression in tumor and spleen γδ T cells from KitV558Δ/+ mice. Representative of 4 mice. Data represent mean ± SEM; p values were calculated using a two-sided Student’s t test; *, P < 0.05.

Clinicopathological features of the patients with GIST in this study.

Supplementary Figure 3: the interaction between endogenous β-catenin and COP1 in human GIST specimens.

HIF1a mRNA expression in KitV558del/+ mice and 72h cell viability assay of GIST-T1 cells treated with ferric iron chloride + VLX600

Primary untreated KIT+ vs. Metastatic resistant KIT+, pathways with P< 0.05.

Supplementary Figure 5: the effect of imatinib on Wnt/β-catenin signaling in Kit^V558Î"/+ tumors

Serum metabolic chemistries and splenic immune cell frequencies in VLX600-treated mice

72h cell viability assays of GIST cell lines treated with VLX600

ROS production in 2 week imatinib-treated and vehicle-treated KitV558del/+ murine CD45+ immune cells

Supplementary Table from Tyrosine Kinase Inhibition Alters Intratumoral CD8+ T-cell Subtype Composition and Activity

Supplementary Tables 1-4 and Supplementary Figure Legends

Real-time changes in OCR and ECAR in GIST-T1 cells treated with varying doses of VLX600 + imatinib