Fluorescence-based actin turnover dynamics of stem cells as a profiling meth od for stem cell functional evolution, heterogeneity and phenotypic lineage parsing

Abstract

Stem cells are widely explored in regenerative medicine as a source to produce diverse cell types. Despite the wide usage of stem cells like mesenchymal stem cells (MSCs) and induced pluripotent stem cells (iPSCs), there is a lack of robust methods to rapidly discern the phenotypic and functional heterogeneity of stem cells. The organization of actin cytoskeleton has been previously used to discern divergent stem cell differentiation pathways. In this paper, we highlight the versatility of a cell profiling method for actin turnover dynamics. Actin filaments in live stem cells are labeled using SiR-actin, a cell permeable fluorogenic probe, to determine the endogenous actin turnover. Live MSC imaging after days of induction successfully demonstrated lineage specific change in actin turnover. Next, we highlighted the differences in the cellular heterogeneity of actin dynamics during adipogenic or osteogenic MSC differentiation. Next, we applied the method to differentiating iPSCs in culture, and detected a progressive slowdown in actin turnover during differentiation upon stimulation with neural or cardiac media. Finally, as a proof of concept, the actin dynamic profiling was used to isolate MSCs via flow cytometry prior to sorting into three distinct sub-populations with low, intermediate or high actin dynamics. A greater fraction of MSCs with more rapid actin dynamics demonstrated increased inclination for adipogenesis, whereas, slower actin dynamics correlated with increased osteogenesis. Together, these results show that actin turnover can serve as a versatile biomarker to not only track cellular phenotypic heterogeneity but also harvest live cells with potential for differential phenotypic fates.

Graphical Abstract

SiR-actin (SA), a cell-permeable fluorogenic probe (white stars), markedly increases its intensity only when bound (red stars) to actin filaments (F-actin). Following a wash after prolonged staining, F-actin turnover results in net detachment of the probe due to disassembly and/or reorganization of dynamic actin filaments. Temporal declining intensity of the probe can be harnessed for analysis of actin turnover dynamics of single cells in diverse ways, as shown on the right.

1. Introduction

Mesenchymal stem cells (MSCs) have been widely explored for cell based regenerative applications. Knowhow and control of differentiated MSCs are integral to the cell therapeutic efficacy following cell transplantation but the clinical outcomes can be challenged by the innate heterogeneity and lineage propensity of MSCs. Typically, the phenotypic classification of MSCs is defined by their ability to demonstrate plastic adherence, self-renewal and tri-lineage differentiation potential, along with expression of CD105, CD90, CD73 and lack of CD45 [1]. However, several studies have recommended the consideration of novel surface markers such as CD146, CD271 given that these biomarkers correlated with increased differentiation or proliferation capacity [2, 3]. Yet, the expression of these newly reported markers varies depending on the donor or the tissue of origin, and the relevance of these markers during dynamic cycles of lineage differentiation and dedifferentiation remains to be fully validated. Thus, there is a need to identify a robust and universal phenotypic determinant to simplify the processability of MSC characterization.

The cytoskeleton has been shown to possess the ability to both influence and correlate with MSC differentiation [4]. Naïve MSCs have a fibroblastic spindle shape morphology with actin cytoskeleton organized as thin, parallel microfilaments spread across the cytoplasm [5]. When naïve MSCs differentiate, they undergo distinct cytoskeletal changes depending on the final cell type [5–7]. Osteogenic differentiation results in increased actin polymerization leading to an intertwined patterning of actin filaments with thickened stress fibers [5, 6]. In contrast, adipogenic differentiation results in a rounded cell morphology, with reduced focal adhesions and a disrupted actin network [8, 9]. Differentiation induced actin reorganization takes place over the course of several days to attain a distinct lineage specific cell shape. Interestingly, much before the appearance of gross morphological changes in the differentiated cells, higher order variations in cell shape and cytoskeletal organization have been shown to have the ability to forecast the lineage fates in MSCs within few hours of induction [7, 10]. In addition to being a morphological marker of differentiation, the actin cytoskeleton actively regulates differentiation. In their seminal study, Mcbeath et al. demonstrated that manipulation of cytoskeletal morphology conveyed mechanical cues that influenced lineage commitment of MSCs. Furthermore, several studies have also shown that disruption of actin cytoskeleton can modify cell differentiation[5, 11–13]. For instance, modulation of actin network using cytoskeletal drugs results in downregulation of osteogenesis [11], while the inhibition of actin polymerization results in promotion of adipogenesis [11].

Another major stem cell source for cell-based therapies are human pluripotent stem cells. These include but are not limited to human embryonic stem cells (hESC) and induced pluripotent stem cells (iPSC) [14–19]. Many protocols have been advanced to optimize the differentiation process to generate cells belonging to the three possible lineages: endoderm, ectoderm, and mesoderm (reviewed in selected publications: [20–29]). However, a major challenge is the ability to establish a highly differentiated population of cells without significant phenotypic heterogeneity. Cytoskeletal markers, such as alpha and beta tubulins, are among the earliest proteins that emerge during iPSCs differentiation to become cardiomyocytes or neurons [30]. Similarly, proteins including alpha- and beta- MHC appear when IPSCs adopt cardiac phenotype [31]. Several retrospective studies have more broadly investigated cytoskeletal changes [30–36]. The role of the immediate morphological changes and the relation to differentiation has also been examined with greater focus [34]. While these studies will help understand how spontaneous differentiation, or particular signals alter the cytoskeleton and associated pathways, they do not afford the ability to address the heterogeneity within a cell population during differentiation.

The actin cytoskeleton has been largely studied as a static marker in the context of stem cell differentiation, as exemplified by the gold standard F-actin probe, phalloidin, which requires cell fixation. For visualization of actin in live cells, some of the widely used methods involve administration of fluorescent actin-binding proteins (e.g. LifeAct), actin-directed nanobodies (e.g. Actin-Chromobody) or fluorescent-labeled actin (e.g. GFP-actin derivatives) in live cells [37]. Recently, we showed a new approach to benchmark actin turnover using SiR-actin (SA), a fluorogenic live-cell F-actin specific probe [38]. Unlike the aforementioned live F-actin reporters, SA offers a unique feature since its fluorescent labeling correlates with the endogenous filament actin turnover in live cells. Therefore, cells with high levels of actin polymerization or pronounced stress fibers demonstrate brighter staining with SA, whereas cells with low actin polymerization level or high actin turnover display dimmer probe labeling. Quantitative image analysis of the changing SA fluorescence intensity as a result of actin reorganization could be used as a metric of the real-time actin turnover[38].

In this study, we highlight the use of SA intensity based live stem cell profiling method in terms of harnessing actin turnover in multiple ways, from assessing stem cells during differentiation to isolating MSC sub-populations to achieve enhanced differentiation. By using SA imaging, we found a correlation between actin turnover and differentiation in a heterogenous cell population. Similarly, SA labeling was found to correlate with the expression of lineage specific markers when differentiating to cardiomyocyte or neuronal lineages. Lastly, we employed cell sorting to isolate and characterize MSC subpopulations for their inclination to differentiate. Given the universal nature of actin cytoskeleton, we suggest that our method could serve as a proxy marker for probing live cell differentiation or for selective enhancement of stem cell differentiation.

2. Materials and methods

2.1. MSC culture

Human bone-marrow derived MSCs were provided by Dr. Rick Cohen (Rutgers University). Cells were maintained in VWR™ T-75 or Corning® T-175 tissue culture flasks. After the initial expansion of the cells in Peptrotech® MSC media, cells were cryopreserved at passage 3 (P3). Upon thawing, cells were maintained in basal growth media (BA) prepared with Gibco Minimum essential medium α (MEMα) supplemented with 10% Fetal Bovine Serum-premium select (Atlanta Biologicals™) and 0.1%v/v penicillin-streptomycin (Lonza). The growth media was changed every third day and passaged when 70–80% confluence was reached. For passaging, MSCs were dissociated with TrypLE™ Express (Gibco) and a seeding density of 3000–4000 cells/cm² was used. Flow cytometric isolation of cells based on actin turnover was conducted at P6 or P7 upon reaching cell count of at least 5 million cells. Post-sort cells were allowed to attach and grow. Subsequent proliferation and differentiation studies were done for 2 more passages.

Adipogenic media was prepared with BA media supplemented with 1 μM dexamethasone, 10 μg/mL insulin, 500 μM isobutyl-1-methyl-xanthine and 200 μM indomethacine. Osteogenic media was prepared by adding 500 μM L-ascorbic acid-2phosphate, 1 μM dexamethasone, and 10 nM b-glycerophosphate to the BA media. Chondrogenic media bullet kit was obtained from Lonza. Cell differentiation experiments were conducted on 96-well multi-well dish at 10,000 cells/cm² seeding density.

2.2. iPSC culture

Human foreskin Fibroblasts (HFFs) were derived from discarded tissue (CHTN) using the protocol similar to Bryne and coworkers using optimized media conditions to support the expansion of Skin Derived Precursors [39]. All cell culture is carried out in 5% O2, 5%CO2 atmosphere. The HFF cultures were electroporated (NEON, ThermoFisher) with a single EBNA1/Ori plasmid containing a polycistronic vector with Oct4, Sox2, KLF4, LMyc and a fusion of mRFP and Blasticidin S Deaminase using conditions similar to Okita and coworkers [40]. iPSC colonies appeared within 30 days and were subcultured using standard enzyme free techniques onto vitronectin (PeproTech) coated plasticwear with albumin free low protein culture media (PeproTech). Under these conditions, the IPS cells were Oct4+/SSEA4+/SSEA1−, and Nanog+/Lin28+/Tra-1–60+, and found to reliably form both neurons [41] and cardiomyocytes (TnT1+, beating clusters) using standard protocols [42, 43]. A more detailed protocol is included in the supplementary material.

2.3. Immunostaining pre-SA labeled MSCs to determine actin turnover correlation with differentiating MSCs

MSCs were stained with 100 nM SA overnight in BA followed by 7 day culturing in BA, AD or OS media. Then cells were fixed in 4% PFA and permeabilized with 0.1% triton-X100. PPARG and RUNX2 specific antibodies (Cell Signaling Technology, Inc.) were used as differentiation specific markers for AD and OS were labeled respectively [44–46]. To evaluate differentiation specific dynamics of actin turnover, cells in AD and OS were compared separately against cells in BA. For image analysis, Hoescht 33342, Invitrogen™ (nuclear stain) was used to mark single cells. For BA vs AD, PPARG intensity was plotted against SA after normalization using the highest value of SA and PPARG among both media. Similar approach was followed for BA vs OS, but with RUNX2 quantification instead of PPARG. Adipogenesis and osteogenesis were assessed by plotting PPARg or RUNX2 intensities respectively against SA after normalization using their highest intensity values and compared to BA media.

2.4. Flow cytometry for isolation of MSC sub-populations

MSCs were labeled with 100nM SA overnight, then transferred to the sorting buffer (2% fetal bovine serum, 0.5mM EDTA, 20 μg/mL DNaseI (Worthington Biochemical, Lakewood, NJ) in Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12. Prior to sorting, cells were filtered through a 40um nylon mesh to eliminate cell aggregates. BD Biosciences Influx High Speed Cell Sorter equipped with 100μm nozzle was used for aseptic isolation of MSC sub-populations. 640nm solid-state laser was used to sort MSCs based on SA probe intensity. DAPI staining was done to determine the gating for excluding dead cells. The sorted cells were collected based on SA intensity gradation into 3 sub-populations: dim, intermediate and bright.

2.5. SA quantification based on live iPSC imaging

Live iPSCs were labeled with 100nM SA overnight in iPSC growth media (IGM) following the methodology described by 2019 Mishra et al[38]. The SA staining media was aspirated, and a quick media change was done to remove residual SA. The cell culture dish was placed on Zeiss LSM780 laser scanning confocal microscope equipped with a stage-top incubator. The reference intensity images were captured for all test conditions. Time-lapse imaging was performed following induction with IGM, cardiac differentiation media (CDM) or neural differentiation media (NDM) for 4 hours. Imaging was done with the 10× objective in triplicates for all test conditions and 4 distinct fields were imaged in each replicate. Subsequent image analysis was done using ImageJ to measure the changing SA intensity of the cell populations compared to the reference images. The mean gray values of SA were plotted against time to quantify changing actin turnover.

2.6. Flow cytometry analysis to determine the correlation between SA and iPSC pluripotency/differentiation markers

iPSCs were labeled with 100nM SA overnight, then the cells were cultured in label free IGM, CDM or NDM for up to 3 days. At day 0, 2 and 3, cell monolayer was dissociated with TrypLE™ Express (Gibco) and cells were fixed using 2% paraformaldehyde in PBS. Subsequently, cells were permeabilized and immunostained with primary labeled SOX-1 (early neuronal or ectodermal marker) [47–49], SSEA-1 (general early differentiation marker) [50–53], OCT3/4 (stemness marker) [54] and BRACHYURY (early cardio or mesodermal marker) [55–57] antibodies (BioGems). The correlation between SA and iPSC markers was assessed by FACS analysis of the immunolabeled cells using BD Biosciences Influx High Speed Cell Sorter.

2.7. Differentiation assay for MSCs

MSCs were seeded at the density of 10,000 cells/cm² and allowed to attach overnight in 96-well dish. MSC differentiation were induced with adipogenic or osteogenic media for 7 or 14 days. The cells were fixed with 4% paraformaldehyde and stained with fast blue RR (Sigma) and AdipoRed (Lonza) reagents to stain for alkaline phosphatase (osteoblast) and intracellular triglycerides (adipocyte) respectively. For cellular enumeration, Hoechst dye (33342, Invitrogen) was used to label nuclei. Subsequently, to scan the entire wells, automated confocal images were taken of 3 wells per condition, and fast blue RR (Supplementary Fig. 1) and AdipoRed fluorescence intensity signals were normalized to the number of cells in each well.

3. Results

3.1. Method for tracking actin dynamics during stem cell differentiation induction

Actin filaments (F-actin) are dynamic structures that rapidly polymerize and depolymerize, with the addition and detachment of actin monomers driven by ATP hydrolysis and several cellular factors[58]. SA binds F-actin at sites with 3 contiguous monomeric units and increases its brightness by 100 fold[59]. When the F-actin undergoes depolymerization, the bound probe comes off resulting in loss of fluorescence[60]. Our approach to assessment of actin turnover involves temporal analysis of changing SA labeling. A workflow for the method is illustrated in Figure 1. Briefly, the F-actin in live cells is probed with SA in basal growth media overnight to allow maximal SA labeling in all cells. Subsequent removal of staining media followed by addition of a cytoskeletal perturbation results in varying loss of SA from cells based on their actin turnover status. In this paper, the cytoskeletal perturbation was induced with differentiation media, but it is possible to influence cell dynamics in other ways as well. For instance, cytoskeleton perturbing drugs that arrest actin dynamics (e.g. cytochalasin D) result in prolonged retention of SA probes on actin filaments compared to untreated cells[38]. High SA label indicates a slow actin turnover and vice versa. SA labeled cells can be subsequently evaluated by various methods for qualitative or quantitative assessment of actin turnover in cell populations.

Figure 1. Actin turnover imaging methodology for stem cell phenotypic dynamics.

Workflow for SiR-actin (SA) labeling protocol enables live cell actin turnover analysis and cellular profiling: 1) Cell seeding and adhesion in basal media, followed by SA labeling in basal media; 2) Repeated washes with basal media followed by introduction of cytoskeletal perturbation cues; 3) With time, cells with high actin turnover display lower SA fluorescence (a, c) compared to the cells with low actin turnover (b,d); 4) The variable SA labeling can be assessed qualitatively or quantitatively, and used as a basis for further cell analysis or cell sorting.

3.2. Live tracking of actin dynamics during MSC differentiation

In order to monitor how differentiating cells modulate their actin turnover, cells were stained with SA in basal media (BA) followed by stimulation with adipogenic (AD), osteogenic (OS) and chondrogenic (CH) media (Fig. 2). In BA, the cells demonstrated elongated spindle morphology with less defined stress fibers, indicating a dynamic actin cytoskeleton as this group demonstrated lowest SA intensity at all time-points. Adipogenic induction begins with dramatic reduction in actin turnover at day 2 as evident by the SA brightness. The cells in AD continue to show slow actin turnover like the chondrogenic induction till day 7. Similar to AD, CH showed high retention of SA during 1 week of induction with relatively brighter staining of stress fibers indicating reduced actin turnover i.e. less dynamic cytoskeleton. OS induction demonstrated higher SA labeling of actin stress fibers compared to BA at day 2, but the subsequent loss of SA staining indicates a higher level of cytoskeletal reorganization unlike AD and CH.

Figure 2. Multiday live cells imaging of MSCs showed lineage specific changes in SA labeling.

Representative images showing changing SA staining during induction for 7 days with basal (BA), adipogenic (AD), chondrogenic (CH) or osteogenic (OS) media. The imaging settings were kept consistent Day 2 onwards (calibration bar unit: mean gray value). Higher SA brightness indicates a slowdown in actin turnover.

3.3. Correlating changing actin turnover with differentiation specific markers to highlight heterogeneity of MSC states

After discerning an early slowdown in actin turnover in MSCs upon differentiation induction, next we sought to understand how actin turnover dynamics correlate with MSC differentiation. MSC differentiation was induced after initial overnight SA labeling in BA media, followed by cell culture in BA, AD or OS media for 7 days. After 7 days, the differentiated cells were fixed and immunostained for lineage specific reporters, PPARG (master regulator of adipogenesis [45]) or RUNX2 (master regulator of osteogenesis [46]). Cells in BA showed minimal expression of PPARG or RUNX2 (Fig. 3 A-D). To determine the correlation between SA labeling due to MSC differentiation markers, SA intensity was plotted against PPARG (AD marker) or RUNX2 (OS marker). The intensity of fluorophores in single cells were normalized with the highest mean gray value within BA vs AD or BA vs OS groups. An ellipse was drawn encompassing all the datapoints in BA to define the basal level of reporter intensities for the undifferentiated cells (Fig. 3 E, H). In AD media, the average SA intensity was higher compared to BA (Fig 3 A, E, F, G). PPARG^high cells exhibited low levels of SA labeling (Fig. 3 F-ii). A possible explanation could be that adipogenic induction results in slowdown in actin-turnover (Fig. 2) but as the adipocytes mature, SA is eliminated due to the loss of actin filaments [9]. Cells in OS also showed elevated levels of SA labeling compared to BA (Fig. 3 I, J). But in contrast to adipogenic differentiation, higher osteogenic differentiation (RUNX2^high) resulted in a higher relative SA retention compared with cells in BA. A possible explanation for retention of SA during transition to OS lineage could be due to preservation of F-actin regions during differentiation. Interestingly, cells committed to either AD or OS lineages demonstrated a more uniform range of SA expression, while the undifferentiated cells had a more variable SA labeling (Fig. 3 F, I). Taken together, the SA dynamics reveals actin turnover heterogeneity during MSC differentiation. Thus, this fluorescent labeling and imaging methodology has the potential to reflect intrapopulation heterogeneity during lineage commitment based on actin dynamics.

Figure 3. Immunofluorescence based correlation of differentiation and actin turnover in MSCs:

After SA labeling, MSCs were cultured for 7 days in BA, AD or OS media. Subsequently, cells were fixed and immunostained. The representative images show differences in expression of SA (actin turnover marker), Hoechst (nuclei), PPARG [adipogenic marker (A) BA media, (B) AD media) or RUNX2 (osteogenic marker (C) BA media, (D) OS media]. The correlation plots for PPARG (E, F ) and RUNX2 (H, I) intensities are shown along with best-fit line. In addition, an elliptical region was marked for AD (E, F) and OS (H, I) based on corresponding BA plots to mark the undifferentiated cells. The single cell scatter plots for SA and PPARG (G) or RUNX2(J) are also represented along with means to summarize differences among cell populations during AD and OS differentiation respectively compared to BA. Two-tailed Pearson correlation test for E and H showed p<0.001, but n.s. for F and I. [n=315(E), 285 (F), 252 (H) and 350 (I)]. Comparison of means of BA vs AD or BA vs. OS of RUNX2, PPARG, and SA was done using unpaired student t-test method (p<0.001 ***, p<0.01 **).

3.4. Early cardiomyocyte or neuronal differentiation of iPSCs involve reduction in actin turnover

iPSC maturation towards distinct lineages involves extensive cytoskeletal reorganization [30, 31] but the early changes during differentiation are not reported. Therefore, the SA-labeling based approach was extended to assess the early changes in actin turnover during iPSC differentiation. The naïve cells were SA stained in iPSC growth media (IGM), followed by stimulation with either neural differentiation (NDM) or cardiac differentiation media (CDM). Differentiation induction resulted in elevated retention of SA staining in both NDM and CDM compared to the basal growth media (Fig. 4A). After 2 days, cells cultured in CDM showed highest probe retention with well-defined actin filaments marking the periphery of cells but with diffused cortical actin. Cells cultured in NDM showed a more dispersed SA staining but at an elevated level compared to cells cultured in growth media (Fig. 4B).

Figure 4. Early changes in actin dynamics during iPSC differentiation.

A. Intensity quantification of SA labeling during Cardiac and neural differentiation B. Representative images to show lineage specific SA retention after 2 days of stimulation following removal of SA staining media. C, D. FACS analysis to show expression of stemness and differentiation markers for neural (C) and cardiac (D) lineages [x-axis: SiR-actin, y-axis: iPSC markers]

Next, FACS analysis was performed at different timepoints to assess the correlation between SA staining (actin dynamics) and early differentiation markers. Cells were cultured in NDM or CDM and fixed and immunostained at Day 0, 2, and 3. Given the lack of evidence in literature for early expression of the markers (Brachyury or Sox-1), we designed a pilot experiment to gain insights past the day 1 timepoint. Thus, we focused on conditions where we could pool together adequately large cell numbers to enable a comparison of all controls within a single stained sample that would allow for expression of the aforementioned marker [25, 55, 61, 62]. As early as 2 days of stimulation, cells in NDM resolved into two populations and high SA expressing cells were found to be positive for Sox-1 (neuroprogenitor marker) and SSEA-1 (general differentiation marker). Oct-4 (pluripotency marker) expression declined after 2 days of stimulation compared to naïve cells and resolved into two distinct cell populations similar to SOX-1 and SSEA-1 (Fig. 4C). A possible explanation for 2 sub-populations could be that the initiation of neural differentiation induced a slowdown of actin turnover resulting in elevated SA levels while naïve cells lost more SA due to relatively higher actin dynamics. Upon stimulation with CDM, cells showed higher levels of SA compared to basal media or NDM indicating slower actin turnover. Unlike NDM, CDM did not resolve the iPSCs in 2 populations. However, at both timepoints (Day 2, 3), SA^high cells were found to have a higher expression of Brachyury (mesoderm marker, early cardiomyocyte) and SSEA-1 compared to IGM (Fig. 4D). Interestingly, the expression of OCT-4 did not decline from day 2 to day 3. Therefore, SA labeling was found to correlate with the expression of early neural or cardiac markers indicating that iPSC differentiation initiates with slowing down of the actin turnover.

3.5. Actin turnover based sorting of MSCs results in subpopulations with distinct proclivities for differentiation

SA labeling is indicative of the actin cytoskeleton dynamics in live cells. Cells with more dynamic actin structures are poorly labeled while the cells with long-lived, less dynamic actin cytoskeleton demonstrate weak labeling [59, 60]. We hypothesized that the intensity of SA probe could be used as a predictive marker to isolate cells based on the dynamic status of the actin cytoskeleton. MSCs were stained with SA in basal growth media. Cell sorting was done based on SA intensity into three sub-populations, designated for simplicity as: Dim, Intermediate, and Bright cells (i.e. high, intermediate and low actin turnover of single cells, respectively) (Supplementary Fig. 2). Post-sorting, cells were cultured in basal media. Next day after sorting, MSCs were evaluated for their cytoskeletal features. “Dim” cells displayed a spindle shaped cell body with smaller area. While, the “Bright” cells had larger surface area, and more stress fibers. “Intermediate” cells were closer to “Dim” cells morphologically, but the SA intensity was much higher (Supplementary Fig. 3). Therefore, SA labeling could be used to resolve MSCs into distinct sub-populations based on their actin dynamics.

Next, we investigated whether SA sorted naïve MSC populations demonstrate a higher intrinsic proclivity for cell differentiation. The sorted cells were expanded, and differentiation studies were conducted for 2 consecutive passages. “Dim” cells with faster actin dynamics showed highest adipogenesis among all groups (Fig. 5D). A possible explanation for this observation could be that adipogenesis involves gradual degeneration of the actin cytoskeleton and the “Dim” cells with a less complex cytoskeleton and fewer stress fibers might have a higher proclivity to make adipocytes. On the contrary, “Bright” cells with slower actin dynamics showed highest osteogenesis among all groups (Fig. 5B). During osteogenesis, actin cytoskeleton leads to increase in actin stress fibers and actin cytoskeleton [9]. Since, the “Bright” cells have an inherently actin filament rich cytoskeleton, they might be inclined to become osteoblastic. Next, we allowed the cells to grow for another passage to test if the differentiation proclivity maintains in the subpopulations. Upon OS induction, “Bright” cells showed highest differentiation among all groups (Fig. 5C). While, after AD induction, both “Dim” and “Bright” groups showed higher differentiation compared to unsorted groups (Fig. 5D). Therefore, actin turnover has the potential to be harnessed as a single marker for not just isolation of cells based on their inherent actin dynamics but also to isolate sub-populations with a higher proclivity for distinct lineages.

Figure 5. SiR actin intensity-based cell sorting of MSCs into three potential sub-populations for enhanced MSC differentiation potential.

A) Cell sorting strategy. SA intensity based sorted cells showed distinct proclivities after cell culture in OS (B, C) or AD (D, E) media for 2 weeks. For image analysis, fast blue (FB) and adipored (AR) staining were normalized with hoescht (n+1: after one passage, n+2: after two passages following cell sorting). The mean values were compared using one-way analysis of variance (ANOVA) *p<0.05, **p<0.01 vs Unsorted cells.

4. Discussion

The actin cytoskeleton plays a vital role in diverse cellular functions and has been extensively correlated with stem cell differentiation for regenerative medicine [5, 7, 11, 63]. In this paper, we highlight a live stem cell tracking methodology focused on harnessing the dynamics of the actin reorganization in the context of stem cell differentiation. We have described multiple ways in which SA labeling based actin turnover tracking can be used for long-term evaluation of actin dynamics during stem cell differentiation and as a dynamic marker for isolation of MSCs using flow-cytometry.

It is well recognized that cells undergo rapid change in cell shape during differentiation, regulated via the cytoskeleton, resulting in divergent phenotypic fates [5–7]. Most of the prior work focused on cytoskeleton-based cell profiling metrics was based on fixed cells; therefore, only considered the cytoskeleton as a static snap-shot of the cellular phenotype [5–7, 9, 12] Given the highly dynamic nature of actin cytoskeleton, the cytoskeletal dynamics itself can offer deeper insights about the temporal trajectory of cell populations that could be used to sensitively parse the cellular lineage potential. Previously, we reported a new approach involving SA, which has the unique feature of labeling actin structures based on their dynamic status to illustrate early changes in the actin turnover of differentiating MSCs [38]. Yet, the prior study was conducted only for few hours after stimulation with differentiation media, which restricted its utility to early divergence of cellular phenotypes.

Using SA labeling, we are able to gain new insights about the status of actin turnover during the course of MSC differentiation for several days (Fig. 2). The onset of AD and CH differentiation involves a decline in RhoA/ROCK signaling resulting in reduced actin polymerization and loss of F-actin [63]. But we observed retention of SA after 7 days of culture in AD and CH media, instead of losing the probe due to loss of F-actin. This could be because of diminished actin polymerization, suggesting the actin filaments were preserved or remained static until transition to a functional specific lineage cell type. On the other hand, OS differentiation is supposed to increase actin polymerization and stress-fibers that should provide more binding sites for SA binding [6]. At Day 2, OS showed higher SA labeling compared to BA but we observed a small fraction of SA labeling compared to AD or CH. A possible explanation for lower SA labeling could be the OS induced actin reorganization that resulted in rapid loss of SA from the actin cytoskeleton. After 7 days of induction, SA labeling declined sharply in OS condition (Fig 2). In order to more carefully evaluate the late stage dynamics of actin turnover, we re-labeled the differentiating cells described in Fig. 2 overnight with SA after 6 days of induction, removed the staining media the following morning and imaged further until the 10^th day of initial differentiation stimulation i.e. 3^rd day after relabeling [Supplementary Fig. 4]. Both AD and CH showed higher SA labeling like Fig. 2, but cells in CH had more prominent labeling of stress fibers (Supplementary Fig. 4). Therefore, SA labeling offers a new way to probe actin turnover in live cells for several days. Live cell change in SA intensity conveys that upon actin cytoskeletal reorganization, a highly dynamic process leads to decline in observed probe intensity as SA comes off from its binding site on F-actin [60]. The non-dynamic changes in SA labeling are not clearly described in the literature, but SA labeling is likely to be influenced by intracellular concentration, availability of binding sites and several other stochastic events that involve actin reorganization (e.g. cell cycle). There is a need for further in-vitro characterization of SA labeling, but regardless, the method can be extended iteratively as cells can be relabeled with SA to extend the duration of actin dynamics tracking.

Following SA labeling, immunostaining of differentiated cells showed intra-population heterogeneity of actin dynamics. A quantitative correlation plot of SA against differentiation specific marker helps to understand transition of actin dynamics from naïve to differentiated cells (Fig. 3). Especially during AD induction, it is possible to observe distinct sub-populations after 7 days of induction. Previously, we reported immediate slowdown in actin turnover after AD induction [38]. Therefore, it is possible that all cells start as SA^high PPARG^low during AD induction. After 7 days, the observation of undifferentiated SA^high cells could be due to senescence or spontaneous differentiation towards a different lineage that enabled SA retention (Fig. 3D). Subsequent transition to PPARG^high happened after gradual decline in SA. On the other hand, OS differentiation induction showed that cells with intermediate SA labeling changed to OS phenotype without much change in SA labeling. Given the highly dynamic nature of cells in OS [Fig 2], SA labeling after 7 days of induction suggests that that certain regions on F-actin are preserved during OS differentiation, where SA remains even when cells started expressing RUNX2. In future, we plan to perform this study with longer differentiation stimulation period (2–4 weeks) with the hope of obtaining improved dataset. However, it will be important to validate SA expression at measurable levels, especially in basal condition, where low SA expression was observed after a week of staining (Fig. 2). Therefore, SA labeling in conjunction with immunolabeling highlights the heterogeneity of actin turnover in single cells during MSC differentiation.

Early dynamics of actin cytoskeleton during iPSC differentiation have not been explored in the literature to date[31, 33, 34, 64, 65]. Similar to MSCs, a differentiation induced immediate slowdown in actin turnover was observed during stimulation with both CDM and NDM as early as 3.5 hours of stimulation. Next, we explored if the higher retention of SA probe during short term live-imaging experiment could be due to lineage progression. Immunostaining pre-SA labeled cells after differentiation induction as early as day 2 showed that cells positive for differentiation markers showed a higher retention of SA. This phenomenon could be due to a slowdown in actin dynamics of the differentiating cells. After 2 days of induction, NDM demonstrated two distinct subpopulations where SOX1^high cells correlated with brighter SA. On the other hand, SA^low cells showed low SA intensity. Perhaps, SA^low cells represent the naïve cells that remained undifferentiated and had a more dynamic cytoskeleton, similar to cells cultured in basal media as shown in Fig 4A, B. CDM stimulation resulted in prolonged retention of the probe in all cells but did not resolve into distinct sub-populations (Fig. 4B). This could be due to two factors, first SA retention in CDM was higher compared to NDM i.e. undifferentiated cells could be showing similar levels of SA labeling to the differentiated cells. Second, longer CDM induction might help to separate the differentiating cells from the naive cells. Still, BRACHYURY^high cells showed a positive correlation with SA expression. Our findings suggest that actin dynamics could be used as an early marker to parse differentiating iPSCs from naïve cells. In future, the genetic and epigenetic profiling of the SA^low vs SA^high sub-populations would further reveal the role of key factors associated with cytoskeletal dynamics in the context of stem cell differentiation.

We have also introduced a new method where actin turnover can be used as a dynamic marker for selection of subsets of MSC populations (Fig. 5). This promises to be the first such methodological approach in the literature to segregate cells based on the native actin dynamics, which could correlate with lineage-related enrichment of differentiation. Next, we examined the effect of sub-culturing on the differentiation proclivity of the sorted cells. A decline in the purity of the sorted cells would indicate a reduction in AD and OS differentiation to the levels of unsorted cells. In contrast, however, we found a higher differentiation tendency in the sorted cells compared to the unsorted cells in terms of both AD and OS induction. This suggests that cells exhibiting equivalent levels of actin dynamics can be enriched during cell culture which resulted in increased differentiation proclivity. The observed propensity towards enhanced differentiation warrants further studies to investigate the interplay between actin dynamics and stem cell differentiation and elucidate via epigenetic or transcriptional profiling of the sorted MSC sub-populations the mechanistic and molecular basis for the variation among the groups that underlies enhanced differentiation. Another application of the method could be used to isolate and enrich young MSCs from senescent cells, given the correlation between slower actin dynamics and cellular aging [38, 66].

We have described four different methods to use SA based labeling in the context of following actin dynamics in MSCs and iPSCs. Namely, live cell tracking (Fig. 2), immunostaining (Fig. 3), flow cytometry (Fig. 4) and cell sorting (Fig. 5). These methods might be helpful to other researchers who are interested in observing actin dynamics with other cells as well, but it is important to conduct preliminary experiments to determine the appropriate staining protocol. We used SA at 100nM concentration for all experiments, as this dosage has minimal influence on the actin dynamics, cytotoxicity and differentiation [38, 59]. A more detailed description of the influence of SA labeling and dosage with multiple in vitro assays and cell types can be found in the first SA paper by Lukinavičius et al [59].

Taken together, we have introduced a novel dynamic marker-based cell sorting method that has the potential to discern cell phenotypic populations based on actin turnover. Given our observation of a prolonged retention of SA probe during both iPSC and MSC differentiation, we propose that cells might have an inherent inclination towards a certain lineage based on the endogenous actin turnover of each cell. To explore the efficacy of our approach we conducted a proof of concept study of SA-labeling and imaging followed by sorting. We found increased osteogenesis and adipogenesis with low and fast actin turnover displaying cells respectively (Fig. 5). Interestingly, the cells retained this increased inclination towards a specific lineage even after cell passaging. Moreover, the differentiation was enhanced further which could be due to increase in proportion of the sorted cells during tissue culture maintenance. Therefore, our approach has the potential to be used for isolation of cells based on actin turnover for robust promotion of MSC differentiation even after cell passaging.

Conclusion

Actin turnover has the potential to be used as a dynamic marker of stem cell differentiation. Fluorescent labeling of live MSCs revealed cell fate-specific changes in the actin turnover. Additionally, SA labeling was found to correlate with differentiation markers in both MSCs and iPSCs. Therefore, we have introduced a new versatile platform that not only enables FACS-based high throughput probing of heterogeneity among differentiating cells but also offers a more informative dataset in conjunction with immunolabeling based approach to illustrate changing cell states during cell differentiation. As demonstrated by our pilot study, SA can be used to isolate cells based on their actin turnover to study how cellular functions are influenced by the dynamic status of actin turnover of the cells.

Highlights.

• Fluorescence-based actin turnover is a biomarker for stem cell differentiation

• Actin turnover reveals lineage heterogeneity at the single stem cell level.

• Actin turnover of mesenchymal stem cells forecasts divergent cell functions.

• Actin turnover discerns differentiation of live induced pluripotent stem cells

• (iPSCs)

• Actin turnover dynamics identifies iPSCs positive for early cardiomyocyte or neuronal markers.

• Live tracking of actin turnover dynamics can be coupled with flow cytometry based isolation to enrich cells with a desired functionality.