Smith David, Glen Katie, Thomas Robert
Centre for Biological Engineering, Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough, LE11 3TU, U.K.
Biotechnol Prog. 2016 Jan-Feb;32(1):215-23. doi: 10.1002/btpr.2199. Epub 2015 Nov 28.
The translation of laboratory processes into scaled production systems suitable for manufacture is a significant challenge for cell based therapies; in particular there is a lack of analytical methods that are informative and efficient for process control. Here the potential of image analysis as one part of the solution to this issue is explored, using pluripotent stem cell colonies as a valuable and challenging exemplar. The Cell-IQ live cell imaging platform was used to build image libraries of morphological culture attributes such as colony "edge," "core periphery" or "core" cells. Conventional biomarkers, such as Oct3/4, Nanog, and Sox-2, were shown to correspond to specific morphologies using immunostaining and flow cytometry techniques. Quantitative monitoring of these morphological attributes in-process using the reference image libraries showed rapid sensitivity to changes induced by different media exchange regimes or the addition of mesoderm lineage inducing cytokine BMP4. The imaging sample size to precision relationship was defined for each morphological attribute to show that this sensitivity could be achieved with a relatively low imaging sample. Further, the morphological state of single colonies could be correlated to individual colony outcomes; smaller colonies were identified as optimum for homogenous early mesoderm differentiation, while larger colonies maintained a morphologically pluripotent core. Finally, we show the potential of the same image libraries to assess cell number in culture with accuracy comparable to sacrificial digestion and counting. The data supports a potentially powerful role for quantitative image analysis in the setting of in-process specifications, and also for screening the effects of process actions during development, which is highly complementary to current analysis in optimization and manufacture.
将实验室流程转化为适用于制造的规模化生产系统，对于基于细胞的疗法来说是一项重大挑战；尤其是缺乏对过程控制信息丰富且高效的分析方法。本文以多能干细胞集落作为一个有价值且具挑战性的范例，探讨了图像分析作为解决此问题方案一部分的潜力。使用Cell-IQ活细胞成像平台构建了形态学培养属性的图像库，如集落的“边缘”、“核心周边”或“核心”细胞。通过免疫染色和流式细胞术技术表明，传统生物标志物，如Oct3/4、Nanog和Sox-2，与特定形态相对应。使用参考图像库对这些形态学属性进行过程中的定量监测，结果显示对不同培养基更换方案或添加中胚层谱系诱导细胞因子BMP4所诱导的变化具有快速敏感性。为每个形态学属性定义了成像样本大小与精度的关系，以表明使用相对较少的成像样本就能实现这种敏感性。此外，单个集落的形态状态可与单个集落的结果相关联；较小的集落被确定为均匀早期中胚层分化的最佳选择，而较大的集落则保持形态学上的多能核心。最后，我们展示了相同图像库在评估培养物中细胞数量方面的潜力，其准确性与牺牲性消化和计数相当。这些数据支持了定量图像分析在制定过程规范以及筛选开发过程中工艺操作效果方面可能发挥的强大作用，这与当前优化和制造中的分析高度互补。
