A high-throughput functional 3D droplet microfluidic platform for function-to-transcriptome profiling identifies candidate lncRNAs associated with engineered NK cell function

Background Natural killer (NK) cell-based immunotherapies show considerable promise for cancer treatment; however, their efficacy against solid tumors remains limited by functional heterogeneity and the restrictive tumor microenvironment. Engineered NK cells, including mesothelin-directed chimeric antigen receptor cytokine-induced memory-like (MSLN-CAR CIML) NK cells, exhibit enhanced cytotoxic potential, yet the transcriptional programs underlying effective tumor killing remain poorly defined. Although protein-coding regulators of NK cell function have been extensively studied, the contribution of long non-coding RNAs (lncRNAs) to cytotoxic NK cell states remains largely unexplored, particularly in engineered NK cell therapies. Methods To address this gap, we developed and applied a three-dimensional (3D) Matrigel-based droplet microfluidic screening platform that enables single-cell resolution analysis of NK-tumor interactions. Unlike conventional bulk transcriptomic approaches, this function-to-transcriptome workflow enables functional isolation of killer and non-killer NK cell populations prior to transcriptomic profiling enabling the direct comparison of transcriptional programs associated with cytotoxic function while minimizing confounding effects of bulk population analyses. Using this approach, we profiled donor-matched MSLN-CAR CIML NK cells and CIML NK controls from three independent donors. Results Comparative transcriptomic analysis revealed widespread differences in lncRNA expression between killer and non-killer NK cell populations, characterized predominantly by broad lncRNA downregulation that was consistently observed across donors and more pronounced in MSLN-CAR CIML NK cells than in CIML controls. Integrated lncRNA-mRNA analyses prioritized candidate lncRNA-gene pairs based on genomic proximity and coordinated co-expression patterns, identifying several lncRNAs associated with cytotoxic gene programs. These transcriptional signatures were further associated with enrichment of cytotoxic and inflammatory pathways characteristic of highly functional NK cell states. Conclusions Together, these findings demonstrate the utility of our functional droplet microfluidic screening platform in prioritizing candidate lncRNAs associated with cytotoxic NK cell states within matrix-restricted tumor environments. By integrating functional phenotyping with transcriptomic profiling, this platform provides a framework for identifying candidate lncRNA and other potential biomarkers associated with cytotoxic NK states for future mechanistic validation and the rational optimization of NK cell-based immunotherapies for solid tumors.