High‐throughput mechano‐cytometry as a method to detect apoptosis, necroptosis, and ferroptosis

Abstract In recent years, the importance of the investigation of regulated cell death (RCD) has significantly increased and different methods are proposed for the detection of RCD including biochemical as well as fluorescence assays. Researchers have shown that early stages of cell death could be detected by using AFM. Although AFM offers a high single‐cell resolution and sensitivity, the throughput (<100 cells/h) limits a broad range of biomedical applications of this technique. Here, a microfluidics‐based mechanobiology technique, named shear flow deformability cytometry (sDC), is used to investigate and distinguish dying cells from viable cells purely based on their mechanical properties. Three different RCD modalities (i.e., apoptosis, necroptosis, and ferroptosis) are induced in L929sAhFas cells and analysed using sDC. Using machine learning on the extracted parameters, it was possible to predict the dead or viable state with 92% validation accuracy. A significant decrease in elasticity can be noticed for each of these RCD modalities by analysing the deformation of the dying cells. Analysis of morphological characteristics such as cell size and membrane irregularities also indicated significant differences in the RCD induced cells versus control cells. These results highlight the importance of mechanical properties during RCD and the significance of label‐free techniques, such as sDC, which can be used to detect regulated cell death and can be further linked with sorting of live and dead cells.


Regulated cell death (RCD) is defined by the Nomenclature
Committee on Cell Death (NCCD) as 'any form of cell death that results from the activation of one or more signal transduction modules, and hence can be pharmacologically or genetically modulated'. 1 The first identified, and consequentially most thoroughly studied, form of RCD is apoptosis which is introduced in 1972 by Kerr et al. 2 However, in recent years an increasing number of RCD modalities have been discovered among which, but not limited to necroptosis (a RCD which mainly is regulated by the kinase proteins, RIPK1 and RIPK3, and dependant on their substrate MLKL for the execution) and ferroptosis (an iron-dependant RCD that results from the lethal accumulation of peroxidised lipids). 3 Most of these RCD modalities have shown to play an important role in several diseases, [4][5][6] which has led to an enormous increase of interest in the cell death field. One of the best-known examples is the evasion of apoptosis in tumour development, which has been defined as one of the hallmarks of cancer, 7 other important examples are, but not limited to, the role of different RCD modalities in heart diseases, 8 neurological disorders 9 and the role of different RCD modalities in the development of immune responses. 10,11 When researching the potential of drugs or other compounds to induce cell death, a common strategy is to use fluorescent dyes which mark the occurrence of an essential step in the cell death process to indicate cellular membrane permeabilization (e.g., propidium iodide or Sytox green). However, these markers also entail some limitations, among which, prolonged excitation of fluorescent dyes which can lead to phototoxicity, 12 markers often only indicate later stages of cell death and fluorescent labels, in some cases, do not allow for discrimination in between different modalities. 13 To overcome these limitations label-free microscopy techniques have gained increasing interest.
In the past decades, a range of different strategies to analyse regulated cell death in a label-free manner have been emerging among which are Raman spectroscopy, 14 digital holography microscopy (DHM) 15 and mechanobiology analysis. 16 A range of biological states and processes have previously been associated with changes in mechanical properties. 17 Atomic force microscopy (AFM), a nanoindenter tool initially developed for material science applications, can be used to analyse the mechanical properties of soft samples. An example of this is found in a paper analysing the mechanical properties of polymeric capsules, 18 this data could then be used to derive the forces that cells exert on such capsules. 19 Recent developments in AFM techniques have greatly increased the use of this technique in imaging and mechanical measurements of biological samples. [20][21][22][23] It has been reported that a distinction between different modes of RCD could be made by comparing dynamics in elasticity and microrheology, using AFM ( Figure 1A). 16 However, the limitation inherently connected to the AFM technique is a low throughput. To improve this, techniques are being developed that allow to analyse mechanical properties of cells at a higher throughput. Often a microfluidics approach is used which limits free movement of the cell in order to analyse deformation. 24 This technique is the mechanical simile to flowcytometry, a technique well engrained in cell death analysis. 25 Some examples of this approach include constriction-based deformability cytometry (cDC), 26 shear flow deformability cytometry (sDC) 27 and extensional flow deformability cytometry (xDC) 28 (schematic representations of these techniques are shown in Figure 1B). Although all these techniques show promising results and allow throughputs up to 200 cells per second, they have not been applied to identify induction of RCD.
Even though the implementations of these high throughput methods are quite recent, there is a surge studies that are looking into possible applications of these methods. A first possible type of application is the analysis of the effect of drugs on mechanobiology. Examples of this are shown in research on the effect of cytoskeletal drugs in a human leukaemia cell line, 29 the effect of a V-ATPase inhibitor on the mechanical properties of cancer cells with the outlook of developing new treatment strategies 30 and treatment of Choreaacanthocytosis with Dasatinib or Lithium. 31 Furthermore, deformability cytometry can be used to characterize changes of cellular states such as maturation of dendritic cells 32 and differentiation of pluripotent stem cells 33 or changes in mechanical properties due to certain pathologies as was done for Plasmodium falciparum infection. 34 Besides its use for fundamental research purposes, several papers   have also been already exploring more applied uses for this technique, examples are the quality assessment of cellular blood products 35 and the differentiation between healthy and tumorous tissue in biopsy samples using a machine learning based analysis. 36 It was reported that sDC and cDC work at similar low strain rated while xDC applies higher strain rates, therefore cDC or sDC were posed as most suited to measure changes in the actin cytoskeleton, 24 since at higher strain rates the actin cytoskeleton can fluidize. Considering the knowledge gained from previous research on the changes in mechanical properties of cells due to breakdown dynamics these two techniques are most promising. 16 Of note that during induction of regulated cell death, cells detach from the substrate and tend to agglomerate, which would lead to an increased risk of channel clogging in case of cDC compared to sDC. Therefore, here, a sDC setup is used to analyse the mechanical changes during RCD. For sDC two syringe pumps, one flowing sheath fluid while another has cells suspended in sheath fluid, are connected to a specialized chip that contains a narrow channel on where cells are focused through ( Figure 1C). At the end of that channel, a transmission image is acquired. By performing automated object detection on these images, the cell can be distinguished from the background and a mask can be created. This mask is then used to extract a multitude of parameters from the cell ( Figure 1D). It is hypothesized that by inducing RCD in cancer cells, a significant drop in cellular elasticity will occur, which will be measurable by analysing the change in deformability of the cells flowing through the narrow channel allowing to distinguish these cells from control viable cells. Furthermore, cells undergoing RCD show characteristic changes in morphology which can further help distinguish the dying cells from the control cells.

| Induction of RCD and time kinetics analysis
With the aim of investigating the kinetics of the RCD modalities, an overtime induction experiment was performed. Cells were seeded in a 96-well plate at a concentration of 10,000 cells per well 24 h prior to the experiment. Cell death induction was performed using the mentioned above concentrations. These cells were imaged every hour using fluorescence microscopy (Ti-e, Nikon). Two fluorescent labels were added to the cells to allow analysis of the cell state.

| Analysis of mechanical properties by AFM
To analyse the mechanical properties of the cells, AFM was used. The cells were trypsinized right before the experiment and seeded in a confocal imaging dish (VWR) to a density of 500,000 cells/mL. To prevent any changes in mechanical properties due to formation of stress fibres (involved in cell attachment), the mechanical properties were measured immediately after adding the cells to the dish (while still in suspension).
The measurements were performed in a liquid environment using an in-house-made colloidal probe with the radius of 5 μm. To extract the Young's modulus, the obtained force curves were processed in the JPK DP software using an adjusted Hertz model for spherical probes.

| Deformability cytometry
The cells were prepared for the sDC as follows. First, the cells are a Nikon TE2000 setup. Images were acquired using a highspeed CMOS camera (QVIT, AOS technologies) at speeds higher than 3000 frames per second Post-processing of the captured images was performed using a personal script in MATLAB (v. 2022; Supporting Information S1). During the post-processing, a size filter is applied to objects with an area between 78 μm 2 which is the reported upper size limit of apoptotic bodies 38 and 284 μm 2 which is the equivalent area of a circle with a diameter of 40 μm 2 , the mesh size used in the previous filtration.
The following six parameters are isolated for cells in these experiments: area, perimeter, area ratio, deformability, eccentricity, and aspect ratio ( Figure 1D, Table S1). Area, perimeter, and area ratio are related to the morphological state of the cells. The area ratio offers information on the roughness of a cell. Higher values indicate a bigger difference between the actual cell area and the convex hull around the cell, to prevent that doublets of cells or objects that are too irregular are included in the dataset a threshold value is set at 1.2, these values are based on previously reported data. 39 Detected objects with an area ratio higher than this value are excluded from the further analysis.
Deformability, eccentricity, and aspect ratio, on the other hand, offer information on the mechanical state of the cells. Deformability, defined as 1-circularity (a measure of how much an object deviates from a circle, with a circle being the extreme case: 0) is an indicator for cellular elasticity, 40 higher deformability indicates a softer cell and vice versa.
Eccentricity is another measure for the elongation of an object, this value is calculated from an ellipse with the same second moments as the cell mask. Eccentricity offers information on the elongation and has two degenerate cases: 0 for a circle and 1 for a line. The aspect ratio is a less accurate measurement of elongation which takes the ratio of the x-axis over the y-axis of the bounding box around the detected cell (a more detailed calculation of these parameters is shown in Table S1).
The aforementioned parameters are not completely independent of each other. An important consideration that needs to be made is that objects with a bigger area, but the same elasticity will show a larger deformability when flown through the same channel diameter.
Another parameter that should be interpreted carefully is area ratio.
Normally control cells are smooth which results in an area ratio close to 1. Irregularities on cells, such as blebs, will increase this value. In relation to that, deformability is artificially increased by these irregularities (since it is calculated from both perimeter and area). Eccentricity, on the other hand, is less dependent on these irregularities. Taking this into account, it has previously been suggested that at the threshold of an area ratio higher than 1.05, alternative parameters describing elongation should be preferred over deformability for the interpretation of mechanical properties of the cell, 41 in this work, eccentricity is preferred when this threshold has been crossed.
2.5 | Live/dead cell pre-filtering using support vector machine classification model

| Changes in mechanical properties of suspension cells after induction of RCD using AFM
The difference in elasticity between viable and dead cells was assessed using a well-established method namely AFM. In this work, a specific adherent cell line namely murine fibrosarcoma L929sAhFas is used which allows to induce three different RCD modalities (i.e., apoptosis, necroptosis and ferroptosis) in a similar timeframe by adding different inducers. 15,16,42 Since the sDC setup requires cells to be in suspension, the L929sAhFas cells were brought into a suspen- The AFM analysis showed that the Young's modulus of adherent control L929sAhFas is approximately 700 Pa, whereas the L929sAh-Fas control cells in the suspension have a significantly lower average value of the Young's modulus (336.7 Pa). It has been reported previously that cellular elasticity decreases due to cellular detachment. 43 The focal adhesion points, and more specifically the stress fibres that originate in these points, are disturbed during cellular detachment (e.g., by enzymatic detachment). Given that detaching cells from the surface already leads to a decrease in elasticity, here, still, a significant decrease can be observed when RCD is induced in the cells. On average, the Young's modulus of the control viable cells is halved (from ±300 Pa to ±150 Pa upon RCD induction). Importantly, this difference in elasticity is essential for the next steps in the analysis of this work.
Since it has been established that the stress fibres and the amount of them can strongly affect the cellular elasticity, 44 it is to be expected that detached cells will still show a decreased elasticity after induction of RCD. In between the different cell death modalities, no significant differences in Young's modulus are reported. This is to be expected since the previously reported differences between these modalities occur at earlier stages during the cell death process than are investigated in this work. 16 3.2 | Live/dead cell filtering using support vector model Due to an inherent presence of heterogeneity during RCD induction, not all cells will enter the cell death process at the same time. These variations can depend on many parameters, among which, the cellular micro-environment, cell density or the stage of cells in the cellcycle. 45,46 To get an idea how this is present in the L929sAhFas cell line, a fluorescence microscopy kinetics experiment was performed ( Figure S1). For this experiment, a fluorescent label is used that robustly indicates end-stage cell death for apoptosis, necroptosis and ferroptosis, which is propidium iodide. High percentages of end-stage cell death are only observed after 24 h (95%, 88%, and 98% for apoptosis, necroptosis, and ferroptosis, respectively) which are not a sudden switch but rather a gradual increase. The question posed here is if sDC can be used to detect the onset of cell death by analysing the mechanical properties. We have found a significant decrease in elasticity after 5 h for apoptosis, necroptosis, and ferroptosis. Importantly properties of cells undergoing cell death. 47 For control cells, an average estimated 14% of cells were classified as dead. This value is higher compared to that obtained from the fluorescence microscopy experiments (3%). However, these are some factors that potentially affect this number. First, L929sAhFas cells are inherently adhesive cells, so in order to detach them from the substrate these need to be treated with trypsin. Previous research on the damaging effect of different cell harvesting methods reported an average cell death of 10% after trypsin treatment, which is in line with the results presented here. 48

| Comparison of the effects caused by apoptosis, necroptosis, and ferroptosis in L929sAhFas cells using deformability cytometry
The data of the extracted parameters, after filtering the data for all the RCD modalities, is presented in Table 1 (area, perimeter, deformability, area ratio, eccentricity, aspect ratio; a detailed explanation of these parameters see in Section 2 while statistical information is shown in Table S2). To give an indication of how a representative cell looks during the different RCD modalities, three cells are represented in Figure 3D with the corresponding values of area, area ratio and eccentricity.
Looking first at the area ( Figure 3A) and perimeter, it can be observed that for all RCD modalities a significant increase is observed compared to the control (p < 0.001; Table 2). Further intercomparing the different RCD modalities between each other also shows significant differences in the area and perimeter. The smallest increase of area and perimeter is noticed for apoptotic cells (233 μm 2 ) which are significantly smaller than necroptotic cells (252 μm 2 ) which on their turn are significantly smaller than ferroptotic cells (281 μm 2 ; p < 0.001, Table S2). While no clear reports exist of cells swelling during apoptosis, it is possible that this small area increase is caused by the presence of membrane blebs on the cellular surface, most likely caused by membrane destabilization. 49 Another characteristic that can be noticed about the cell area for apoptotic cells is that a small subpopulation exists with a much lower area than those at all other conditions ( Figure 3E). A potential explanation for this sub-population could be that these are apoptotic bodies originating from the apoptotic cells or debris released from apoptotic cells. During necroptosis, the pore-forming molecule, mixed-lineage kinase domain-like protein (MLKL) is incorporated into the cell membrane, which leads to an influx of extracellular and water molecules resulting in cell swelling and leading to eventual cell lysis. 50 Researchers have also previously reported that cell rupture occurs during ferroptosis which eventually leads to cellular swelling similarly as in necroptosis. 51 This is in line with what is observed in our data. Also, the presence of ferroptotic blebs can increase this value. 52 For area ratio again a significant increase can be reported for all RCD modalities compared to control cells (p < 0.001) indicating that cells undergoing cell death exhibit a rougher morphology. The biggest increases are noticed here for apoptosis and ferroptosis which both have a significantly higher area ratio compared to the necroptotic cells. For apoptotic cells, this is most likely caused by the process of apoptotic blebbing. 49 Recently, it was reported that a type of macroblebbing can also occur during the process of ferroptosis. 52 In some of the captured images of the ferroptotic cells, these structures are visible ( Figure S3). This phenomenon explains the similar increase in the area ratio in ferroptotic cells to apoptotic cells. In the necroptotic cells, a smaller increase is observed. During necroptosis, it is known that the plasma membrane is permeabilized which can lead to irregular shape changes in cells resulting in a higher area ratio. 50 When looking at the parameters related to the deformation of the cells namely deformability, eccentricity, and aspect ratio, all RCD modalities show a significant increase compared to the control (p < 0.001). Care should be taken however since a high area ratio could potentially artificially increase the deformability values (in other research, as a safety an upper limit of 1.05 is set). 41 For apoptosis, T A B L E 1 A summary of area, perimeter, deformability, area ratio, eccentricity, and aspect ratio per condition which were analysed in the L929sAhFas cells (± indicates standard deviation).

Control
Apoptosis ( necroptosis and ferroptosis, the area ratio exceeds 1.05. Therefore, to interpret the mechanical changes, the eccentricity has been analysed. The highest increase in eccentricity is noticed in the necroptotic cells, which is significantly higher ( p < 0.001) than the eccentricity of both apoptotic and ferroptotic cells.
The significant increase in eccentricity compared to the control cells suggests that the cells undergoing RCD are softer and thus deformed more (Figure 2A). A significant increase in eccentricity of necroptotic cells in comparison with apoptotic cells can be noticed, which can be explained by the increase in cellular area of necroptotic cells compared to apoptotic cells. For ferroptosis, a significantly lower eccentricity is measured compared to necroptosis. However, the ferroptotic cells have a significantly higher area compared to the necroptotic cells, thus it would be expected that ferroptotic cells have a higher eccentricity. This can be explained by the aforementioned ferroptotic blebs, 52 which form irregularities on the cell surface and therefore increase the cellular area but do not contribute to the deformability, since these do not contain intact cytoskeletal protein fibres as was previously shown. 16 To better understand which parameters are the major contributors to the variability in the data, a principal component analysis was performed (PCA). The first three principal components (PC) explain a cumulative 94% of variance (Table S3). A loadings plot, containing all the data points is presented in Figure 3E (the eigenvalues table is presented in Table S4). In this plot, the first PC is mainly dominated by three parameters namely deformability, eccentricity, and aspect ratio.  Table 2 (results of the statistical tests are represented in Table S5).
Examples of representative cells in these experiments are shown in Figure 4A. When analysing the cell area ( Figure 4B) and perimeter, a significant increase can be noticed after 5 h of apoptosis induction (p < 0.001). However, at 24 h, a decrease compared to control is observed for both (p < 0.001). A potential explanation for this could be the phenomenon of secondary necrosis, which occurs when apoptotic cells are not cleared by immunological cells. 53 During secondary necrosis the cell membrane is permeabilized and the cellular components start to spill out of the cell. Which eventually leads to an empty hull of a cell membrane left over. Losing all these cellular components also logically leads to decrease in cell area.

At 5 and 24 h after induction, a significant increase in deformabil-
ity and eccentricity is observed ( Figure 4C) can be noted ( p < 0.001), which further increased at 24 h to an average of 1.11 ( p < 0.001).
As mentioned above with a high area ratio (>1.05) eccentricity should be preferred over deformability, since eccentricity proves to be a more robust indicator of elasticity for irregularly shaped cells.  Figure 4E). The average cell area of cells at 24 h is decreased. Therefore, the cells encounter less shear force if the channel width remains constant. Therefore, it is likely that cells with equal elasticity, but lower cell area will have a decreased eccentricity (which supports our corresponding AFM data; Figure 4E).
It can be noticed on the scatterplot ( Figure 4D) that the spread of the population at 24 h is large. Taking into account that we mentioned above that due to secondary necrosis; apoptotic cells lose their contents and only an empty shell remains one should be critical when analysing these results. This deformability cytometry technique was developed for intact cells, and it makes some assumptions in developing this technique namely that the object is an isotropic, linearly elastic sphere, which is not the case anymore in our situation. 40   shown to be possible AFM. 16 The duration of deformation in the sDC is significantly shorter than that of the AFM (in the range of milliseconds for sDC and seconds for AFM) which could lead to discrepancies in measured (AFM) and the observed deformation (with the sDC). 40 This is an essential remark since it was mentioned previously that cells are not purely elastic but rather viscoelastic. 58 Since it has been shown cells behave more like a viscous fluid as they progress further into the process of RCD, 16 it will take even longer to observe the resulting total deformation, that the shear force of the walls causes on the cells.
Knowing this it is possible that the real difference in deformability between control cells and dead cells is still larger than measured in the results. Thus, essential information on the viscoelastic state of cell in this technique is missing. However, new avenues are being explored that would enable this in the future. 59 Recently, advancements have led to development of a sorting system linked to deformability cytometer with the possibility of measuring fluorescence signal, enabling a user to sort samples based on their mechanical properties and fluorescent labelling(with a throughput up to 200 cells/second). 39 Combining this sorting technique with the data presented in this work provides strong outlook for future applications.
Dead cells tend to have a higher unspecific uptake of probes, altered antigen expression 60 or increased autofluorescence. 61 The ability to remove dead cells from a sample in a label-free manner (based on their mechanical properties) before staining can greatly increase the quality of the results. Inversely, for other applications, the early-stage dead cells are of significant interest. An example of this can be found in testing the immunogenicity of early-stage ferroptotic cells. 57 The data shown in this work show the perspective to split such a sample into two categories: early-stage dead cells and viable cells allowing to exclude more variability in the data. Since these systems work with immune responses, any foreign components (such as fluorescent labels) should be avoided, thus showing the need for effective label-free techniques.