*3.3. Autofluorescence-Based Optical Sensors*

Various biomaterials derived from cellular organisms have autofluorescence properties, which allows them the potential to be used in biosensors for invasive and label-free sensing to obtain information about cells and tissues [140]. Furthermore, autofluorescence techniques do not require treatment or fixing of specimens and can be performed in real-time. In addition, autofluorescence-based sensors can be applied to monitor stem cell differentiation with invasive optical sensing methods, such as Raman spectroscopy and NIR, because autofluorescence can indicate a specific cellular component [141,142]. Raman-based sensing applied with autofluorescence is especially advantageous to analyse information about intracellular dynamics, which can be utilised to investigate the intracellular changes during stem cell differentiation [103].

A label-free autofluorescence-based imaging system combining optical metabolic modelling with quantitative image analysis was developed by Qian et al. [143] for monitoring human PSC (hPSC) differentiation into cardiomyocytes (Figure 9). This study was based on the fact that hPSC-derived cardiomyocytes undergo significant metabolic changes during differentiation. Specifically, the amount or ratio of oxidised flavin adenine dinucleotide (FAD) and reduced nicotinamide adenine dinucleotide (phosphate) (NAD(P)H), both autofluorescence metabolic materials, is influenced by the cellular conditions and differentiation and can be imaged to collect metabolic information at the single cell level. Furthermore, the ratio of NAD(P)H to FAD provides information about the relative oxidative state of the cells. Therefore, cardiomyocytes differentiated from different hPSC lines were visualised with NAD(P)H and FAD autofluorescence probes. According to the autofluorescence imaging on the eighth day, the intensity of NAD(P)H autofluorescence differed depending on the differentiation efficiencies. In addition, the cardiomyocytes and noncardiomyocytes showed different autofluorescence intensities. For instance, NAD(P)H and FAD fluorescence after 8 days was exhibited in 84.1% of the differentiated cells, compared with 0.3% in undifferentiated cells.

**Figure 9.** An autofluorescence imaging for monitoring of cardiomyocyte differentiation. Singlecell quantitative analysis of NAD(P)H during 8 days of the differentiation period. Reprinted with permission from [143]. Copyright 2019, Wiley Online Library. EGFP, enhanced green fluorescent protein; NAD(P)H, reduced nicotinamide adenine dinucleotide (phosphate); NKX2.5-EGFP, homeobox protein NKX2.5-EGFP.\*\*\*\* *p* < 0.0001.

In another study, Suhito et al. [144] reported on an autofluorescence-integrated Raman mapping analysis for label-free monitoring of adipogenic differentiation. Raman mapping analysis has the critical issue of long detection time, which results in cell apoptosis. To address this issue, these researchers developed a novel optical sensing method that enabled the rapid and non-destructive analysis of adipogenesis. The authors confirmed that the lipid droplets present in adipocytes were identified with the developed autofluorescenceintegrated Raman sensing method; the Raman scattering of lipid droplets was aroused at 2850–2855 cm<sup>−</sup>1. In addition, this method was utilised in the large-scale sensing analysis of multiple cells in culture plates by obtaining Raman mapping images at low magnification. Moreover, the analysis required a very short time (<20 min) and could scan 440 × 330 μm area per mapping image. Furthermore, the authors analysed in-batch and batch-to-batch variations of adipogenic differentiation throughout the autofluorescence-Raman imaging.

Similarly, Li et al. showed a label-free autofluorescence sensing system capable of monitoring of neurogenesis [145]. The optical sensor was based on tetrapod-shaped ZnO (t-ZnO) microparticles capable of label-free monitoring of neuronal differentiation. Specifically, this sensor formed 3D scaffolds that analysed DA released from neurons embedded on the surface using autofluorescence imaging. Interestingly, t-ZnO nanoparticles with four hexagonal arms were biocompatible and autofluorescence materials that fluoresced under UV light because they contained anion vacancies. The nanoparticles' autofluorescence was demonstrated to be very sensitive to hole scavengers, which was used for quantitative DA analysis. Furthermore, nanoparticles' autofluorescence acted as a quencher for the autofluorescence of the t-ZnO nanoparticles. Due to its 3D structures with high surface area and autofluorescence, the t-ZnO nanoparticles-based sensor showed a high sensing performance toward DA (LOD: 0.137 μM). Furthermore, DA was selectively datable in the presence of interfering molecules, including citric acid, glutamine, ascorbic acid, glucose, KCl and calcium chloride.

Autofluorescence-based optical sensing techniques were reported as suitable for noninvasive, non-destructive, and label-free monitoring of stem cell differentiation. Chronic lymphocytic leukaemia-derived autofluorescence could be utilised to analyse the biological changes in stem cell differentiation. Furthermore, most autofluorescence is biocompatible, which allows stem cells to be stably cultured and differentiated on the sensor for long time periods.
