Digital image analysis of HL-60 proliferation and granulocytic differentiation.

Abstract

Biochemical and computer-assisted digital image analysis techniques were used to study the proliferation and granulocytic differentiation of HL-60 leukemia cells which were induced to differentiate by retinoic acid (RA). HL-60 cells were synchronized by density arrest (DA), double thymidine block (TB), or flow cytometric sort (FCM). Following release from the 3 modes of synchrony, proliferation was evaluated by immunoperoxidase staining for incorporated bromodeoxyuridine (BrdUrd, % S phase cells), mitotic index, cell counts, and digital image analysis of Feulgen stained cells. Data revealed increased synchronization of cells by TB. Digital image analysis of changes in total optical density (TOD) of Feulgen stained cells following release from TB verified the greater synchrony achieved by TB. Fisher linear discriminant analysis using chromatin textural features was used to distinguish G₁ from S and G₂ phase cells. Evaluation of biological descriptors over a 7 day period demonstrated less proliferation and increased differentiation for RA-induced TB cells than for RA-induced DA cells. These descriptors included differential counts of Wright-Giemsa stained cells, nitroblue tetrazolium (NBT) reduction (superoxide production), staining for myeloperoxidase (MPO) and immunofluorescence staining for C3bi receptors using OKM1 monoclonal antibody. Digital image analysis revealed suppression of DNA synthesis by lower Feulgen TOD or RA-treated cells when compared with uninduced controls. The suppression of DNA synthesis was noted as early as one day after induction. The decreasing trend in Feulgen TOD of RA-treated cells correlated with decreased percentages of these cells incorporating BrdUrd. Similarly, differentiation related changes in many image feature values (e.g., decreasing Feulgen TOD, decreasing nuclear area and increasing nuclear convolution) of RA-treated cells, corresponded with decreased c-myc expression, decreased MPO production, increased C3bi receptor expression and increased superoxide production beginning on day 1 after RA-induction. Changes in differentiation-related image feature values preceded decreases in the intensity of bands on silver stained polyacrylamide gels. Additionally, several textural image features demonstrated statistically significant differences from time-matched controls as early as day 1 after RA induction. A combination of 4 image features was used in Bayesian analysis to achieve an approximately 90% correct classification of RA-treated and untreated cells by the 7th day of RA exposure.