Getting More from Cell Size Distributions: Establishing More Accurate Biovolumes by Estimating Viable Cell Populations Stanislav Sokolenko, Yu-Lei Cheng, and Marc Gordon Aucoin Dept. of Chemical Engineering, University of Waterloo, Waterloo, Ontario, Canada, N2L 3G1 DOI 10.1002/btpr.486 Published online September 23, 2010 in Wiley Online Library (wileyonlinelibrary.com). Current approaches for cell size distribution modeling are attempting to describe the behavior of the entire distribution with respect to time. Although some advances have been made in this area, the modeling process requires a large number of culture-specific parame- ters and an a priori assumption of the distribution nature (Poisson, Gaussian, etc.). In this work, we propose a deconvolution of the distribution into size ranges and an iterative regression process with respect to a single culture variable, such as viability. Following this approach, two example applications are outlined using data collected with a Coulter Coun- ter Multisizer. In the first, traditional biovolume measurements are corrected to account for the noneven distribution of nonviable cells. These corrections amount to an average increase of 7–65% in the calculated biovolume from 24 to 72 h postinfection and are expected to aid in the development of a new basis for nutrient consumption postinfection. In the second example, viability is predicted from the cell size distribution using both linear and exponen- tial regressions. Differences between predicted and measured viabilities were found to be normally distributed with means of 0.4% and 0% as well as standard deviations of 7.6% and 8.1% for linear and exponential regression, respectively. Although only viability rela- tionships were tested, our approach yielded significant results for both applications, allowing the possibility for further development. V V C 2010 American Institute of Chemical Engineers Biotechnol. Prog., 26: 1787–1795, 2010 Keywords: cell size distribution, biovolume, Sf9 insect cell, viability, regression analysis Introduction The popularity of the baculovirus expression vector sys- tem has made the study of Spodoptera frugiperda (Sf9) insect cell culture of prime interest in the last two decades. The importance of the cell size distribution in the characteri- zation of the culture has been reported repeatedly throughout this time. The cell size distributions of healthy and unin- fected insect cells have been shown to remain more or less constant as the cells proliferate and increase in cell density; however, once infected by baculovirus, the cells cease to divide and are arrested in the S and G 2 /M phase of their cell cycle because of the expression of early viral genes. The infected cells then increase in size because of the additional production of viral nucleic acids and proteins, 1 until their viability decreases because of infection-induced cell death. Increases in mean cellular volume (or diameter) have been used to track the progress of infection 2 as well as to estimate protein production 3–5 and baculovirus titer. 6,7 Cell size distri- butions have also been used for off-line biovolume measure- ment either in their own right or for on-line capacitance measurement calibration. 8,9 In many of these cases, the cell size distribution is examined in its entirety, usually through a statistical parameter such as the mean or the mode. This approach may work for cell counters capable of distinguish- ing between viable and nonviable cells, but the case is not so clear for nondiscriminating cell counters such as the Coulter Counter Multisizer. The advantage of the Coulter Counter Multisizer is in the number of events it can capture, making the statistics associated with the measurement robust and reliable. Strong evidence exists for the need to separate the overall distribution according to the underlying cellular populations. Sf9 cell death characterization reveals a strong correlation between the fraction of low volume and nonvi- able cells in the culture. 10 Meneses-Acosta et al. 10 have observed that the distribution of low viability cells is bi- modal in nature and correlated the number of lower volume cells (5–10 lm diameter when compared with 12–18 lm for larger cells) to the number of nonviable cells in uninfected culture. These observations were correlated to apoptotic cells through flow cytometry. 10 As infected cell death has also been reported to be apoptotic in nature, 11 similar trends were expected for infected culture. The presence of a direct rela- tionship between viability and a proportion of the distribu- tion suggests that specific culture parameters could be extracted by examining specific sections of the distribution. The presence of distinct viable and nonviable populations in the overall distribution however goes directly against the general assumption used for biovolume calculation that non- viable cells are evenly distributed throughout the distribu- tion. 9 This suggests that a correction to the biovolume calculation could produce better correlations for on-line measurements that rely on viable cell sizes calculated using Correspondence concerning this article should be addressed to M. G. Aucoin at maucoin@uwaterloo.ca. V V C 2010 American Institute of Chemical Engineers 1787