In recruited subjects, the accuracy of PBLs or lymphocyte subset (CD8 lymphocytes) isolations from fresh blood was compared between the Ficoll method (N>35 subjects) (A,B) and the manual magnetic method (N>335 subjects) (C,D). Accuracy of the isolates was analyzed by viability (A,C), purity (A,C), yield (A,C), and cellular composition by analysis of FSC (forward)/SSC (side) scatter (B,D) from flow cytometry. The Ficoll method (A) was inferior to the magnetic method on viability and purity (C). The magnetic method (C) also generated higher yields but the variability was high, reflecting inherent variability found across healthy subjects. In addition, the cellular composition, as determined by scatter plots (B,D), revealed that the magnetic method isolated the cells of interest, CD8 T cells, whereas Ficoll produced a mix of PBLs sufficiently contaminated with dead cells and mononuclear cells that it would require further purification to detect autoreactive CD8 T cells.

In order to detect rare lymphocyte populations, it is critical that the separation method yields healthy cells of sufficiently high purity. (A) Schematic depiction of magnetic and gradient methods. The magnetic method is gentle, highly selective and retains the lymphocytes of interest (blue cells), whereas the gradient method has remaining red blood cells (red dots) and low viability (dark cells), while the cells of interest are largely lost due to cell death. (B) Comparison of performance parameters of magnetic and Ficoll gradient separation methods. Autoreactive CD8 T cells were isolated from 56 blood samples from human type I diabetics with both Ficoll and magnetic methods. Using the insensitive WST biomarker assay of rare autoreactive T cells, only magnetically separated cells showed the presence of the cells (Pass); Ficoll isolated cells uniformly lacked autoreactive T cells. The accuracy of the cell separation method was established by determining viability, purity, and yield. The results show that extremely efficient cell isolation procedures with high yield and high viabilities are required to preserve the cells of interest. (C) Flow cytometry with tetramers specific for autoreactive CD8 T cells shows that these cells are present and detectable after separation from patient blood samples using our magnetic method.

We compared two different plate designs for the robotic platform (A,B). For each test system, Nunc vials were loaded with fresh human blood and paramagnetic beads tagged with the antibody to the whole blood cells of interest was added to the sterile vials. This blood and antibody mixture typically was tested for its ability to separate lymphocyte subsets, such as CD8 or CD4 white blood cells. The most robust platform design gradually culminated in two different plate designs with varying configurations of the external magnets. On the left (A) is a plate design with a post magnet on the outside of the Nunc vial. On the right (B) is a plate design with a ring magnet design at the base of the Nunc vial. The resulting separated lymphocytes were evaluated for the highest yield, viability and purity. It was first thought that a full-length magnet along the side of the tube would facilitate the cellular separations. However, the disadvantage of this plate design was that every magnet experienced interference from the surrounding magnets, thus decreasing the overall density of Nunc vials for maximal efficiency and/or decreasing the magnetic force. As a result, this method produced lymphocyte preparations contaminated with RBCs and magnetic debris, as shown in the left scatter plot of the yielded cells. In contrast, the ring magnets in (B) produced internally consistent magnetic fields that did not interfere with adjacent ring magnets since the pulling force was exclusively directed to the interior of the vial. Also, the ring magnet configuration maximizes the plate capacity for the maximal number of Nunc vials per plate, and allowed the isolation of CD8 lymphocytes from whole blood with much higher purity (right scatter plot).

Based on our workflow, one research associate works 4.5 hours to process blood samples from one diabetic patient and one control; or a total of 18 hrs for 8 patients. In our current automated setup one associate can initiate the processing of 8 samples in parallel and finish in 3 hrs while being able to perform other activities once setup is completed. Such dramatic savings in amount of direct labor will translate in large cost savings and increased throughput.