Cytometry Part A ( IF 3.7 ) Pub Date : 2020-06-24 , DOI: 10.1002/cyto.a.24166 David R Parks 1
With the introduction of “spectral” flow cytometers and the general trend toward increasing numbers of dyes and fluorescence detectors, researchers need to understand how much improvement in actual measurements can be obtained with such systems compared to the traditional approach with a specific detector for each dye. However, neither Multispectral Flow Cytometry: The Consequences of Increased Light Collection by Feher et al. (1) nor its Corrigendum (2) provides useful guidance in understanding or using multispectral flow cytometry.
The original study summed fluorescence signals from detectors in six wavelength ranges and compared the results to those with a single typically configured detector leading to claims of greatly increased detection efficiency and dramatically lowered fluorescence detection limits (DLs). The fundamental problem with the study is that the signals‐addition model and experimental design used in Feher et al. is invalid in the multispectral realm where detectors receive light from multiple dyes and evaluation of the individual dyes depends on multiplicative spectral unmixing (= compensation for an equal number of dyes and detectors).
After the invalid experiment design and major data evaluation errors in the original paper were pointed out, the authors published a corrigendum (2) in which the main data evaluation errors were corrected and the photoelectron scale estimates were re‐evaluated. The corrigendum makes no attempt to address the fundamental point of the inapplicability of the data model and the experiment design.
In the corrigendum, the authors claim that “The basic conclusion concerning the performance of multispectral versus conventional cytometry remains the same, namely that multispectral detection of fluorescein and phycoerythrin is significantly more efficient than conventional cytometric detection, in terms of detection efficiency (Q) and detection limit (DL).”
The revised data evaluations change the originally claimed ninefold increase in photoelectron detection to 2.5‐fold for FITC and 1.2‐fold for PE. Estimates obtained from the on‐line ThermoFisher Fluorescence SpectraViewer (https://www.thermofisher.com/us/en/home/life‐science/cell‐analysis/labeling‐chemistry/fluorescence‐spectraviewer.html) predict multiple detector sums of 1.77 times the single detector signal for FITC and 1.31 times for PE, so the revised results indicate that nothing interesting or unexpected was detected. The revised DLs with multiple detectors are “1.3 times (FITC) and 1.2 times (PE) lower” compared to a single detector. These 20–30% decreases are not substantial and are qualitatively different from the originally claimed 5.4‐fold and 3.8‐fold decreases in DL.
In summary, the revised results do not support the claim that significant improvements in Q and DL were observed.
What experiments would actually illuminate important issues in spectral flow cytometry? With an appropriate experimental design, an instrument like that used by Feher et al. could provide a valid comparison between conventional and more “spectral” configurations. However, for a comparison of detection sensitivity between spectral and conventional configurations, I would recommend a test with a spectral instrument comparing analysis in full spectral mode with the same data analyzed by aggregating selected detectors (“pseudo‐filters”) to approximate compensation with a conventional filter configuration for the selected dye set. The test samples would be a selection of single dye controls including some with closely spaced spectra (capture beads would be fine). The results of spectral unmixing versus conventional compensation unmixing for each of the samples would be evaluated by comparing the rSDs of the main positive population of the dye X sample as seen in the unmixed dye dimensions Y, Z, and so on, where the interesting Y and Z dyes would be ones with substantial spectral overlap with dye X. If those rSDs are significantly smaller for the spectrally unmixed data than for the compensated data, the benefit of spectral unmixing in improving low signal detection would be confirmed.
The other point of great interest in spectral analysis is the utility of including an autofluorescence (AF) spectrum for unstained cells in the unmixing. This could be tested as an addition to the experiment described above by including samples of low and high AF unstained cells (e.g., lymphocytes and fibroblasts). The data analysis would be to unmix the unstained cell data using (spectral) matrices derived from a well‐spaced subset of the single dye spectra and from a set of spectra with more dense coverage of the spectral range. Comparing unmixed results for the cells with and without inclusion of cell AF spectra in the unmixing would indicate how much improvement (if any) would be observed in detection of low‐level dye on the cells. The relevant statistics would be the rSDs of the unstained cell populations with and without AF in the unmixing for each of the dye dimensions. My expectation would be that including AF will be helpful with high AF cells but not with low AF cells and that the benefit will be clearer for the well‐spaced dye set.
中文翻译:
多光谱流式细胞术:未解决的问题和改进建议。
随着“光谱”流式细胞仪的推出以及染料和荧光检测器数量不断增加的总体趋势,研究人员需要了解与使用每种染料使用特定检测器的传统方法相比,使用这种系统可以在实际测量中获得多少改进。然而,这两种方法都没有:Feher等人的《多光谱流式细胞术:增加光收集的后果》。(1)或其更正(2)为了解或使用多光谱流式细胞术提供了有用的指导。
最初的研究对来自六个波长范围的检测器的荧光信号求和,并将结果与具有单个典型配置的检测器的荧光信号进行比较,从而声称大大提高了检测效率,并大大降低了荧光检测极限(DLs)。这项研究的根本问题是,Feher等人使用的信号加法模型和实验设计。在检测器从多种染料接收光并且对每种染料进行评估取决于倍增光谱解混(=补偿相同数量的染料和检测器)的多光谱领域中,“光谱”无效。
在指出了无效的实验设计和主要数据评估错误后,作者发表了更正(2),其中纠正了主要数据评估错误并重新评估了光电子规模估计。更正未试图解决数据模型和实验设计不适用的根本问题。
在更正中,作者声称“关于多光谱与常规细胞计数法性能的基本结论是相同的,即,在检测效率(Q)和检测效率方面,荧光素和藻红蛋白的多光谱检测显着高于常规细胞计数检测。检出限(DL)。”
修改后的数据评估将最初声称的光电子检测增加了9倍,FITC为2.5倍,PE为1.2倍。从在线ThermoFisher荧光SpectraViewer(https://www.thermofisher.com/us/en/home/life-science/cell-analysis/labeling-chemistry/fluorescence-spectraviewer.html)获得的估计值可预测多个探测器的总和FITC的单个检测器信号为1.77倍,PE为1.31倍,因此修改后的结果表明未检测到任何有趣或意外的情况。与单个检测器相比,具有多个检测器的修订后的DL分别“低1.3倍(FITC)和1.2倍(PE)”。这20%至30%的下降幅度并不大,并且在质量上与DL最初声称的5.4倍和3.8倍下降有所不同。
总而言之,修改后的结果不支持声称观察到Q和DL显着改善的说法。
哪些实验实际上将阐明光谱流式细胞术中的重要问题?通过适当的实验设计,像Feher等人使用的仪器。可以在常规配置和更多“频谱”配置之间提供有效的比较。但是,为了比较光谱配置和常规配置之间的检测灵敏度,我建议使用光谱仪器进行测试,将全光谱模式下的分析与通过汇总所选检测器(“伪滤波器”)分析的相同数据进行比较,以使用所选染料组的常规过滤器配置。测试样品将是单个染料对照的选择,包括一些具有紧密间隔光谱的染料(捕获珠会很好)。
光谱分析中的另一大关注点是实用性,它可以将未染色细胞的自发荧光(AF)光谱包括在解混中。可以通过包括低和高AF未染色细胞(例如,淋巴细胞和成纤维细胞)的样品,对上述实验进行测试。数据分析将使用(光谱)矩阵将未染色的细胞数据进行混合,该矩阵是从单个染料光谱的间隔良好的子集以及光谱范围更密集的一组光谱中得出的。比较在未混合中是否包含细胞AF光谱的细胞未混合结果,将表明在检测细胞上的低含量染料时会观察到多少改进(如果有)。相关的统计数据是每种尺寸的染料在解混时有和没有AF的未染色细胞群的rSD。我的期望是,包括AF对高AF细胞有帮助,而对于低AF细胞则无济于事,并且对于间隔良好的染料集,其好处将更加明显。