评估多重数字PCR中的光谱重叠与串扰，以有效监测癌症临床试验中的下一代CAR T细胞

Introduction: As new chimeric antigen receptor (CAR) T cells emerge as secondary therapies in patients already treated with an approved CAR, clinical trial investigation of next generation constructs requires monitoring for the presence of residual first generation CAR-T therapies, often from limited patient samples. Multiplex digital PCR (dPCR) enabling absolute quantification without standard curves and supporting detection of as few as 10 copies, is an ideal approach to monitor multiple approved CAR-Ts in a limited clinical trial specimen (Morley, 2014; QIAGEN, n.d.). While multiplexing improves efficiency, it introduces challenges such as spectral overlap, potentially causing crosstalk and misinterpretation. Herein, we investigate how bleed-through affects signal interpretation in multiplex dPCR and evaluate the effectiveness of threshold adjustments and software updates in mitigating crosstalk and preserving monitoring accuracy of multiple CAR-Ts on the QIAcuity platform. Method: A multiplexed dPCR assay targeting three CAR domains, a viral safety marker, and one reference gene was tested using probes labeled with FAM, TxRED, VIC, TAMRA, and Cy5 across five optical channels. Two formats were run on QIAcuity: singleplex reactions to confirm specificity and baseline crosstalk, and multiplexed reactions to assess cumulative interference. Analysis was performed using Software Suite (SS) versions 2.5 and 3.1. Fluorescence plots assessed signal clarity, crosstalk, and the impact of optical and threshold settings on droplet classification and quantification.  Result : Singleplex reactions showed no bleed-through in FAM, VIC, TAMRA, or Cy5 channels. TxRed bled into Cy5. TxRED bled into Cy5, and TAMRA into VIC, confirmed by 2D plots showing double-positive patterns and unexpected VIC-positive droplets in TAMRA-only samples. VIC-to-TAMRA crosstalk was concentration-dependent, decreasing at lower input levels, while TxRED-to-Cy5 remained constant. Signal clarity for TAMRA improved with optical or threshold adjustments. Despite visible bleed-through, multiplex quantification remained accurate. Cy5-labed targets were correctly identified, with software ignoring TxRED bleed-through. No differences in crosstalk or quantification accuracy were observed between SS versions 2.5 and 3.1, confirming consistency across updates. Conclusion:  Spectral bleed-through, especially from TxRED into Cy5, was a key source of signal misclassification. Manual thresholding and optical adjustments in both software versions effectively reduced crosstalk and improved quantification of all CAR domains. These findings highlight the importance of optimized assay design and analysis tools for reliable quantification of residual CAR-T therapies, which are critical for safe advancement of next generation CAR-T therapies.