Technical note: hESC/hiPSC culture quality assessment based on brightfield images

Updated: Nov 28

Contributor: Dr Mahfuz Chowdhury

Assessment of morphological features of human Embryonic stem or induced Pluripotent Stem Cells (hESC/hiPSC) from brightfield/phase contrast images is a critical, non-invasive and low-cost quality procedure that stem cell researchers routinely use.

Because of various factors e.g., different batches of reagents, operator variability, acquisition of genomic changes by the cells during the culture period, and non-consistent cell culture equipment performance, a stem cell line morphology can gradually change indicating a change in cell quality.

Morphological characteristics of hESC/hiPSC can be assessed in two levels: colony level and single-cell level refs. [1] and [2].

hESC/hiPSC colony level

This assessment provides a quick assessment of the whole stem cell culture, therefore it is used more routinely than single-cell level.

The morphological features that are assessed at this level are:

Round and flat colony
Well-defined and smooth colony edge
cells in the colony are tightly packed

A comparison [ref. 3] of these features between good and bad colonies of a stem cell line (H9p35) is shown below.

Stem cell colony: good vs bad, adapted from ref. [3]

hESC/hiPSC single cell level assessment

The morphological features that are assessed at this level are:

round and small cell
high nucleus: cytoplasm ratio i.e., nucleus occupies most of the cell region
prominent nucleoli/nucleolus

Single-cell morphological features and a schematic illustrating changes in these features, indicative of deviation from an undifferentiated state, are presented below.

Stem cell morphological features at single-cell level, adapted from ref. [2].

schematic showing change in stem cell single-cell morphogical features indicating change in undifferentiated state — Schematic showing change in stem cell single-cell morphological features indicating deviation form undifferentiated state, taken from ref. [3]

Recently, machine-learning approaches is being developed for automated assessment of these morphological features [refs. 3, 4-5], but they are not still widely used.