Automated Counting of Colony Forming Units (CFU): Towards Standardization of the Measurement of Potency in Cord Blood Cell Therapy Products

Abstract 485

CFU content of a cord blood (CB) graft is a potency indicator that evaluates the number, viability, functionality and repopulation capability of its hematopoietic progenitor cells (HPC). However, the CFU assay has not yet been amenable to standardization necessary for potency comparisons between CB units from different CB banks. As a result, this key graft attribute cannot be used widely in CB unit selection. One of the critical steps of the CFU assay is CFU counting, which, together with dish plating, contributes to the 14–36% of variability in CFU counts, as evaluated in the NCBP laboratory in a set of 64 CB samples that were plated and 240 CFU dishes independently red and counted by three operators.

Our goal is to establish a method to determine HPC potency able to replace the manual CFU counting, that can be standardized, proving accurate high-throughput performance.

Our technology includes traditional, manual CFU assay and High Resolution Digital Imaging of stained colonies (CFU-HRDI) [work in progress presented at ASH 2008 and 2009; abstracts 2306 and 2160]. After 14 days of CB culture in Stem Cell Technologies media, culture dishes contents were stained with MTT (3-[4,5-dimethylthiazol-2yl]-2,5-diphenyltetrazolium bromide). A high resolution image of each dish was captured which allowed a clear view of HPC colonies in a color different for CFU-M/GM than for CFU-E. The image also included the dish's barcoded ID. CFU manual counting on the HRDI images showed good correlation with microscope counts (R² = 0.95; p<0.01; n=151).

To improve consistency, eliminate the variability of manual CFU counting and allow high-throughput performance that can be applied widely with acceptable ruggedness, we developed and validated a system that performs automated counting of CFU [ACC] imaged with HRDI. Algorithms were designed to count objects within a specific region of the dish, defined as a “CFU” based on size, optical density and separateness from other objects and also, to read a barcode ID label on the culture dish. The ACC required 20 seconds per image to complete the capture and counting of CFU. The system produces a batch report with CFU counts/dish including the CB unit ID, date and other desired data such as technologist ID, time or lot numbers. A total of N= 12,263 CB samples were tested in duplicate CFU assays. CFU were counted manually and using HRDI-ACC over a period of 24 months. Thus, N= 24,486 CFU culture dishes were evaluated. Assays and manual counting were performed by 8 operators. ACC and manual CFU counts correlated well (R² = 0.85, slope 1.14, intercept 2.0). An average of n=490 CB samples were evaluated per month. Monthly correlations between manual CFU counts and ACC were consistent (range R² =0.81–0.89; SD=0.024). In addition, CFU counts by operator and ACC correlated, with R²= between 0.8–0.9 (mean=0.84 and slope 1.11; range 0.92–1.21). ACC was highly reproducible: CFU counts of the same dish were within 1–2%, with n=50 culture dishes each read 10 times.

To evaluate the inter-laboratory reproducibility of results and their stability despite operational and environmental variables the ruggedness of the CFU-HRDI-ACC system was evaluated by the NCBP and MDA teams in their respective laboratories. Aliquots of the same unprocessed CB units (n=173) were tested for CFU in duplicate and in parallel by two laboratories following a jointly-developed protocol. Automated CFU counts (ACC) per culture dish were comparable between laboratories. The CB units evaluated had CFU counts in the range of 3 to 93 and 4 to 88 CFU/culture dish in each lab respectively with a highly significant linear correlation (R² = 0.62, slope 0.8, intercept 6). Manual CFU counts of the same culture dishes using HRDI images were also comparable between both labs (R²=0.67, slope 0.8, intercept 3).