Mechanical Memory Impairs Adipose-Derived Stem Cell (ASC) Adipogenic Capacity After Long-Term In Vitro Expansion

Abstract

Introduction

Adipose derived stem cells (ASCs) hold great promise for clinical applications such as soft tissue regeneration and for in vitro tissue models and are notably easy to derive in large numbers. Specifically, ASCs provide an advantage for in vitro models of adipose tissue, where they can be employed as tissue specific cells and for patient specific models. However, ASC in vitro expansion may unintentionally reduce adipogenic capacity due to the stiffness of tissue culture plastic (TCPS).

Methods

Here, we expanded freshly isolated ASCs on soft and stiff substrates for 4 passages before adipogenic differentiation. At the last passage we swapped the substrate from stiff to soft, or soft to stiff to determine if short term exposure to a different substrate altered adipogenic capacity.

Results

Expansion on stiff substrates reduced adipogenic capacity by 50% which was not rescued by swapping to a soft substrate for the last passage. Stiff substrates had greater nuclear area and gene expression of nesprin-2, a protein that mediates the tension of the nuclear envelope by tethering it to the actin cytoskeleton. Upon swapping to a soft substrate, the nuclear area was reduced but nesprin-2 levels did not fully recover, which differentially regulated cell commitment transcriptional factors.

Conclusion

Therefore, in vitro expansion on stiff substrates must be carefully considered when the end-goal of the expansion is for adipose tissue or soft tissue applications.

Introduction

Adipose-derived stem cells (ASCs) are abundant, multipotent, and harvested with limited morbidity,^21,46 making them an ideal stem cell source for cell-based therapeutics and in vitro tissue models. While there are established surgical techniques for fat grafting involving the injection of lipoaspirate from one depot into the graft sites^3,4,6,37,38 in vitro ASC expansion is necessary to generate a high enough cell count for regenerative therapies and for lipoaspirate enrichment. This expansion often occurs on stiff substrates, such as tissue culture polystyrene (TCPS). While the relationship between a cell and its microenvironment is complex and dynamic, modulating substrate stiffness is a well-documented approach to direct stem cell fate towards osteogenic and adipogenic differentiation, for example.^2,16,19 This complicates the use of ASCs in therapeutic applications which require in vitro expansion but also retention of multipotency, such as in wound healing and burn treatments.^26,37 Further, when ASCs are employed for in vitro models of various tissues, including for adipose tissue models^4–7,12 they undergo in vitro expansion on TCPS to produce enough cells for in vitro tissue models. Thus, this complicates the use of ASCs when in vitro models require cells maintain adipogenic capacity.

Previous work has demonstrated that human mesenchymal stem cells (hMSCs) can remember the mechanical environments of previous culture environments, with cells previously grown on stiff surfaces persisting in osteogenic differentiation even after being moved to a soft substrate.⁴⁴ Additionally, the success of muscle stem cell engraftment is directly related to the in vitro culture conditions of the cells prior to implantation.¹⁸ The exposure to stiff environments alters translocation of signaling proteins and induces chromatin remodeling^22,24; the duration of exposure can determine the irreversibility of these changes.^24,44 Furthermore, adhesion to stiff substrates induces increased intracellular tension that can be transferred to nucleus via the LINC-complex, a protein structure that links the nuclear envelope with the actin cytoskeleton.⁸ This stress on the nucleus has the potential to physically rearrange chromatin and, subsequently, change gene expression patterns.⁴³

While these findings have greatly improved the understanding of mechanotransduction and mechanical memory in multipotent stem cell differentiation, it remains unclear how applicable these results are to ASCs. Furthermore, there has been little focus on capturing potential donor variability and uniqueness, making it unclear how mechanical memory will affect potential treatments on a patient-to-patient basis. Finally, while a recent study saw pre-conditioning ASCs on soft substrates before implantation yielded a better outcome in their disease model,¹⁵ the role of mechanical memory in ASC differentiation specifically has been largely unexplored.

To evaluate whether ASCs expanded on different substrate stiffnesses have a varied response for adipogenic capacity, we expanded freshly isolated ASCs from four different human donors on the standard TCPS, a stiff substrate (> 1 GPa range) or on a soft substrate (~ 100 kPa range) for 4 passages prior to adipogenesis. Additionally, for passage 4, the substrate was swapped in some cultures to determine if the ASCs retain a “memory” of their former substrate. As expected, when ASCs were expanded on the soft substrate then differentiated on the soft substrate, their adipogenic capacity increased nearly 2.5-fold. When the substrates were swapped at the last passage prior to differentiation, from soft to stiff, adipogenesis decreased 3-fold, and the extent of maturation decreased 2.5-fold. By contrast, swapping from a stiff to soft substrate, did not notably increase the number of cells that differentiated, but improved their maturation by 2-fold. These changes in substrate at the last passage led to changes in nuclear area and nesprin-2 gene expression which may lead to altered cell fate commitment. These results demonstrate that ASCs response to their substrate alters their adipogenic capacity and a short-term (1 passage) exposure to an alternate substrate is not sufficient to return the ASCs to their control substrate baseline. This outcome may not be obvious when looking at morphological cues alone, as the ASCs rapidly modify their morphology within that short-term exposure.

Materials and Methods

ASC Isolation

Adipose-derived stem cells (ASCs) were isolated from human subcutaneous adipose tissue via an adhesion-based method. Briefly, subcutaneous adipose tissue was obtained from elective abdominoplasties approved under Temple University IRB (Protocol # 24700) from the Temple University Medical School, Department of Surgery, Division of Plastic Surgery. The adipose tissue was separated from the skin via blunt dissection and chopped. The tissue was transferred to 250 mL conical tubes and washed with equal volumes of warm phosphate-buffered saline (PBS, Fisher Scientific) until free of blood. An equal volume of 1 mg/mL collagenase I (Corning) and 1% bovine serum albumin (BSA, Fisher Scientific) in PBS was added to the tissue, which was then placed under gentle agitation for 1 h at 37 °C. The samples were mixed via aggressive shaking for 30 s and then allowed a second 1 h incubation at 37 °C under gently agitation. The tissue solutions were centrifuged at 240×g for 10 min at room temp, mixed via shaking, and then centrifuged again at 240×g for 10 min at room temp. The supernatant containing the tissue and free lipids was aspirated from the pellet. To remove residual collagenase, the pellet was resuspended in PBS and centrifuged under the same conditions as above. The washed pellet was resuspended in ASC growth media consisting of low glucose DMEM/F12 (Corning), 10% fetal bovine serum (FBS, Sigma-Aldrich), and 1% Penicillin/Streptomycin (P/S, HyClone) and plated so that 15 g of initial tissue volume was added per P100 tissue culture dish. These cells were labeled Passage 0 (P0). ASCs were harvested from a total of 4 female donors (age: 42–63).

PDMS Cultureware

To generate PDMS culture substrates, the elastomer was mixed at a 20:1 base: crosslinker ratio and degassed to generate substrates in the 100 kPa range.²⁴ PDMS was added to either P100 dishes (8 mL/dish) or 12-well culture plates (300 µL/well) and cured for 2.5 h at 60 °C. Once cooled, the dishes were subjected to four 10 min washes with isopropanol to remove unreacted crosslinker and base. This was followed by two 10 min washes with ddH₂O, after which the plates were dried overnight in a fume hood.

ASC Culture and Differentiation

Isolated ASCs were grown on polydimethylsiloxane (PDMS, Sylgard) or TCPS substrates that were coated with collagen I (Corning, 50 μg/mL in ddH2O) for 2 h at 37 °C directly before seeding. For PDMS conditions, surfaces were activated via 2 min plasma treatments in a plasma asher (PE25W, PlasmaEtch) prior to collagen coating. ASCs were grown to 70–80% confluency before being detached with trypsin-EDTA (Gibco) and passaged to new dishes. When seeding P4 cultures, a fraction of the ASCs was ‘swapped’ to the alternative substrate (Fig. 1). For example, cells isolated and grown on PDMS were split into 2 fractions at the end of P3; half of the cells continued culture on PDMS while the other half were seeded on TCPS. At passage 4, ASCs were cultured to confluence. To induce differentiation, ASCs were treated with adipogenic induction media (AM) at 2 days post-confluence at P4 and proceeded for 14 days with media changes every other day. AM consisted of low-glucose DMEM/F-12 supplemented with 3% FBS, 1% PS, rosiglitazone (2 µm, Cayman Chemical), isobutylmethylxanthine (IBMX, 500 µM, Sigma-Aldrich), dexamethasone (1 µm, Acros Organics), D-calcium pantothenate (17 µM, TCI America), biotin (33 µM, Alfa Aesar), and insulin (20 nM, Sigma-Aldrich). Cells were maintained at 37 °C and 5% CO₂.

Figure 1.

Experimental layout.

RNA Isolation and qPCR

RNA was extracted via the TRI reagent method. Briefly, adherent cells were washed with PBS and overlayed with TRIzol (Invitrogen). The solution was pipetted up and down to mix and the pipette tip was used to manually disrupt cells on the substrate surface. Two chloroform (Acros Organics) extractions were performed, and the RNA was precipitated in 50% isopropanol (Fisher Chemical), after which it was washed 2 times in 75% ethanol (Decon Labs) and resuspended in DEPC-treated water (Invitrogen). RNA was converted to cDNA via the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative polymerase chain reaction (qPCR) was performed with custom gene-specific primers (Table 1, Integrated DNA Technologies) and PowerUp SYBR Green Master Mix (Applied Biosystems) on a StepOnePlus PCR System (Applied Biosystems). Relative levels of gene expression were calculated using the $2^{-\mathrm{\Delta }\mathrm{\Delta }{C}_{\text{T}}}$ method.

Table 1.

qPCR primers for genes of interest.

Primer target	Primer sequence
SYNE 2 FWD	CTTCGACGAGGTAGACTCGG
SYNE2 REV	ACTGCTCTGAACTGCTTTGC
PPARG FWD	ACCAAAGTGCAATCAAAGTGGA
PPARG REV	ATGAGGGAGTTGGAAGGCTCT
RUNX2 FWD	GCGGTGCAAACTTTCTCCAG
RUNX2 REV	TGCTTGCAGCCTTAAATGACTC

Fluorescent Staining and Confocal Microscopy

ASCs were fixed in 4% paraformaldehyde (Pierce) for 15 min. To visualize lipids, samples were incubated with Bodipy 493/503 (5 µg/mL in PBS, Invitrogen) for 45 min at RT. Nuclei were stained with Hoechst 33342 (1 µg/mL, Thermo Scientific) for 15 min at RT. For characterizing cell morphology, cells were permeabilized in Triton X-100 (0.1% in PBS, Sigma-Aldrich) for 20 min and then incubated with Alexa Fluor 647 phalloidin (26 µM, Invitrogen) for 45 min at RT. Samples were imaged on an Olympus FluoView FV1200 confocal microscope using a 10 × objective.

Image Analysis

All image analysis was done in FIJI.³⁶ To quantify Bodipy area, the Bodipy channel was converted to a binary display and the area of signal in the field of view was calculated using the ‘measure’ tool. The percent of differentiating cells was determined by dividing the number of Bodipy-positive cells by the total number of nuclei in each field of view. To quantify nuclear area, images were converted to a binary display and the ‘analyze particle’ tool was used to determine the area of each nucleus. For qualitative assessment, representative images of nuclei morphologies were converted to heatmaps by applying the ‘fire’ LUT on FIJI. For all image analysis, n = 4 donors and 3 fields of view were used in analysis for each donor. The mean measurement from each field of view was averaged to generate the final value for each donor.

Statistics

All experiments were conducted with ASCs from at least 3 different donors. All graphs are presented as mean ± standard deviation unless otherwise stated. One-analysis of variance (ANOVA) followed by a student’s t-test was performed using GraphPad Prism 8 with p < 0.05 considered statistically significant.

Results

Substrate Stiffness Reduces Adipogenic Capacity in ASCs

To mimic in vitro cell expansion to generate large numbers of cells, we expanded ASCs for 4 passages (P4) on either soft (PDMS) or the standard, stiff substrate (TCPS) before undergoing adipogenic differentiation for 2 weeks (Fig. 2). Adipogenesis was broadly characterized by visualization of the defining feature of adipocytes, lipid droplets (Bodipy staining) (Fig. 2b). Expansion and differentiation on PDMS yielded about 50% differentiation, while only about 20% of ASCs on TCPS differentiated, as noted by cells containing lipid droplets (i.e. Bodipy positive cells) (Fig. 2d). However, although the number of ASCs differentiated increased, the amount of lipid droplet accumulation per adipocyte was not notably affected by the substrate on a per cell basis (Figs. 2b and 2c).

Figure 2.

Substrate stiffness negatively correlates with adipogenesis. (a) Experimental overview. (b) ASCs differentiated at P4 in adipogenic media for 2 weeks on TCPS or PDMS substrates showed differing degrees of lipid accumulation (green). (c) Both the area of lipid staining and (d) the fraction of ASCs undergoing adipogenesis were decreased when cells were seeded on TCPS as opposed to PDMS. Scale bar = 200 μm. n = 4 donors, 3 fields of view were quantified per donor. For Bodipy area, * denotes significance (p = 0.003) compared to TCPS condition, for percent differentiating * denotes p = 0.007 compared to TCPS.

ASC Nuclear Spreading Increased with Increased Culture Time on Stiff Substrates

Nuclear size and spreading were affected by both culture time and substrate (Fig. 3a). Early passages (P0, P1) did not have statistically significant differences in nuclear area. Starting at P2, ASCs on TCPS had significantly greater nuclear area than both ASCs cultured on PDMS and the P0 and P1 ASCs on TCPS (Fig. 3a). Given that cellular and nuclear spreading are known to affect signal transduction, including factors regulating stem cell fate, we quantified nesprin-2 (SYNE2) gene expression by normalizing PDMS to TCPS at each passage (Fig. 3b). Nesprin-2 expression was similar between substrates at P1; however, starting at P2, culture on PDMS decreased nesprin-2 gene expression 2.6-fold and remained decreased to similar levels through P4.

Figure 3.

Stiff substrates induce increased nuclear spreading and tension while altering the gene expression of lineage-specific transcription factors. (a) The cross-sectional area of nuclei from ASCs on TCPS increased significantly as the cells were cultured from P1 to P4 while those from cells on PDMS remained similar sizes. Scale bar = 20 µm, n = 4 donors, 3 fields of view were quantified per donor, * denotes significance between PDMS and TCPS conditions of p = 0.005, p = 0.0007, and p = 0.004 at P2, P3, and P4, respectively. ^ denotes significance compared to TCPS at P1 of p = 0.008, p = 0.047, and p = p = 0.023 for TCPS conditions at P2, P3, and P4 respectively. (b) The rise in nuclei size coincided with an increase in nesprin-2 and PPARG gene expression alongside a decrease in RUNX2. n = 3–4 donors. For SYNE2, * denotes significant differences in fold change of gene expression between P1 and P2, P3, and P4 of p = 0.032, p = 0.034, and p = 0.049 respectively. For RUNX2, * denotes a significant difference in fold change in gene expression between the P1 and P4 groups.

ASCs Maintain Some Memory of Their Initial Culture Period After Swapping Substrates

After initial 3 passages on one substrate, we swapped the substrates for P4 studies in Fig. 4. Here, ASCs were cultured on TCPS (or PDMS) for P0-P3, then swapped to PDMS (or TCPS) for P4. After the swap, ASC nuclear size was matched to its current substrate, as if the ASCs were never on their previous substrate (Fig. 4c, left). As with nuclear size, the overall cell morphology, as visualized by f-actin (phalloidin), matched its current substrate, i.e. ASCs cultured for the final passage (P4) on TCPS despite being expanded on PDMS, resembled ASCs expanded on TCPS for all 4 passages (Fig. 4d). These cells were generally larger and did not cluster together. By contrast, the ASCs swapped onto PDMS at P4 or those expanded throughout on PDMS were generally smaller and grew in clusters (Fig. 4d).

Figure 4.

ASCs partially maintain stiffness-specific phenotypes even after swapping substrates prior to the final (P4) passage. (a) Schematic of experimental groups, with “swapped” groups outlined in dotted lines. (b) Imaging of the nuclei on swapped substrates showed a reversal in (c) nuclear area after only one passage on the alternate substrate; however, the difference in nesprin-2 gene expression between ASCs on TCPS or PDMS was only partially bridged by the swap. n = 3–4 donors, 3 fields of view were quantified per donor to calculate nuclear area. For nuclear area, * denotes a significant difference (p = 0.004) between nuclear areas of cells on TCPS and PDMS control groups, P* denotes p = 0.0005 between PDMS and the PDMS TCPS, T* denotes p = 0.0016 between TCPS and TCPS PDMS, ** denotes p = 0.002 between the PDMS TCPS and TCPS PDMS swapped groups. For SYNE2 expression, * denotes p = 0.001 between TCPS and PDMS control groups, P* denotes p = 0.024 between PDMS and the PDMS TCPS, T* denotes p = 0.031 between TCPS and TCPS PDMS. C.) ASCs on TCPS appeared more spread while those on PDMS tended to cluster together at increased densities. Once swapped, the ASCs took on the morphological characteristics of cells on the substrate they were swapped to. (d) PPARG was unchanged by the substrate swap while RUNX2 gene expression levels were flipped. n = 3–4 donors, * denotes p = 0.0252 between TCPS and PDMS, P* denotes p = 0.042 between PDMS and PDMS TCPS, T* denotes p = 0.0256 between TCPS and TCPS PDMS. Scale bar (a) =20 μm, (c) =100 μm.

ASCs cultured on PDMS at P4, although expanded for 3 passages on TCPS, had the same nuclear area as ASCs cultured on PDMS, for all 4 passages. Similarly, ASCs cultured on TCPS at P4, although expanded for 3 passages on PDMS, had the same nuclear area as ASCs cultured on TCPS, for all 4 passages. However, nesprin-2 (SYNE2) gene expression did not return to its baseline in the same manner (Fig. 4c, right). In the control conditions, where the ASCs were expanded on the same substrate for all 4 passages, Nesprin-2 was increased 2.6-fold in cells on TCPS compared to those on PDMS. Swapping substrates from TCPS to PDMS and PDMS to TCPS only led to a 1.4-fold decrease and 1.9-fold increase, respectively. The master adipogenic transcriptional regulator, Peroxisome Proliferator Activated Receptor Gamma (PPARG) was not affected by the swap in substrate (Fig. 4e, left). However, the master osteogenic transcriptional regulator, Runt-related transcription factor 2 (RUNX2) was affected by the swap in substrate (Fig. 4e, right). ASCs swapped to TCPS from PDMS at P4, had a 6-fold decrease in RUNX2 expression, while ASCs swapped to PDMS from TCPS at P4, had a 9-fold increase in RUNX2 expression.

Swapping Substrates Altered ASC adipogenesis differentially for Each Substrate

To determine if swapping substrates can alter adipogenesis, we expanded ASCs for 3 passages before swapping substrates at P4, then exposed the cells to adipogenic media for 2 weeks. ASCs swapped to PDMS at P4, then differentiated, did not increase their adipogenic capacity, as quantified by the number of lipid droplet positive cells (Figs. 5b to 5d, black lines). However, ASCs swapped to TCPS at P4, then differentiated, decreased in adipogenic capacity by 3.3-fold (Fig. 5d, red lines). Similar trends were seen with lipid droplet area (Fig. 5c).

Figure 5.

Swapping ASCs from PDMS to TCPS reduced adipogenesis to a greater degree than the TCPS to PDMS swap increased it. (a) Experimental overview. (b) ASCs grown on either TCPS or PDMS and swapped to the other substrate for P4 were differentiated for 2 weeks at P4 in adipogenic media. (c and d) Both the area of lipids and fraction of ASCs differentiating were similar in the swapped conditions, yielding a large decrease in the PDMS TCPS and small increase in the TCPS PDMS condition. Scale bar =200um. n = 4 donors, 3 fields of view were quantified per donor. For Bodipy Area, * denotes a significant difference in area (p = 0.003) between PDMS and TCPS control groups, P* denotes p = 0.012 between PDMS and PDMS TCPS. For percent differentiating, * denotes a significant difference (p = 0.007) between TCPS and PDMS control groups, P* denotes p = 0.001 between PDMS and PDMS TCPS, ** denotes p = 0.011 between PDMS TCPS and TCPS PDMS swapped groups.

Discussion

Human ASCs provide several advantages to other human stem cell sources like bone marrow derived mesenchymal stem cells (BM-MSCs), namely they can be isolated in greater numbers with a minimally invasive process (liposuction) or as a byproduct of elective surgeries, such as abdominoplasties. For example, on average, we obtain ~ 5000 g (11 lbs) of adipose tissue, where each 35 g of tissue, yields approximately 10 million P0 ASCs, or about 1.43 billion P0 ASCs in total per 5000 g adipose tissue. This is contrast to BM-MSCs where the process to harvest bone marrow is painful, not the byproduct of elective procedures, and yields 10–25 g of tissue per aspiration procedure. The stem cell yield per volume tissue is similar between ASCs and BM-MSCs. ASCs are preferred for regenerative medicine applications such as fat grafting and skin grafting as they better mimic the source of tissues to be repaired. ASCs can be isolated directly in the operating room, with equipment such as Cytori Therapeutics, PureGraft, and set aside for further expansion or used for immediate clinical use, such as for Cell-Assisted Lipotransfer (CAL). While the number of ASCs for therapeutic needs would vary greatly based on the application, if we consider that CAL like techniques use approximately 2 million ASCs per mL fat, then for larger scale applications like breast reconstruction, where cases range from 100 to 700 mL fat grafting, that could require up to 1.4 billion ASCs.^25,32

Moreover, ASCs provide a unique opportunity for developing in vitro adipose tissue models, where one can create these in vitro models with cells derived from that same tissue and differentiated into mature adipocytes. This overcomes a major hurdle in adipose tissue engineering in harvesting mature adipocytes, where mature adipocytes, once released from the microenvironment are prone to delipidation and rupture. This rupture occurs easily with adipocytes as they are lipid-laden where they are especially sensitive to mechanical disruption. If they remain intact, they tend to float to the surface again due to their lipid-laden nature. Therefore, differentiating stem cells in a controlled environment, where handling is minimized, allows for a more consistent and reproducible in vitro model of adipocytes or adipose tissue. Furthermore, ASCs can be used to develop patient specific in vitro models of adipose tissue. Despite these major advantages, ASCs have not been as well characterized as BM-MSCs for responses to their environment, especially as it pertains to their in vitro expansion. Here, we sought to evaluate how in vitro expansion may alter the ASCs, particularly to determine if their adipogenic capacity is affected by their expansion protocol.

Adipogenesis is known to be, in part, mechanically regulated. For example, seeding ASCs on soft substrates (low kPa range) increases adipogenesis when compared to higher stiffness substrates.^16,45 This can be partly attributed to the lack of cell spreading, where ASCs or BM-MSCs restricted in spreading in 2D or 3D environments tend to undergo adipogenesis, where cells permitted to spread, favor osteogenesis.²³ A lack of cell spreading and thus reduced cytoskeletal tension can be directly connected to adipogenesis via an inverse relationship between spreading and PPARG gene expression.²⁹ One possibility is that the increased cytoskeletal tension commonly associated with cell spreading is being transferred to the nuclear envelope via the LINC-complex, a protein structure that physically tethers cytoskeleton and nucleus. Increased mechanical tension on the nuclear envelope has the potential to change gene expression by altering transcription factor availability and binding, chromatin remodeling, and nuclear architecture.^8,11,35,40 Indeed, gene expression of adipogenic markers are increased when the LINC-complex is rendered non-functional.⁴⁰ As a result, we evaluated if prolonged exposure to a soft vs. stiff substrate, as seen with in vitro expansion protocols, would affect the ASCs capacity for adipogenesis after 4 passages.

Our results are consistent with the existing literature, where ASCs expanded and differentiated on the stiff substrate (TCPS) had a 50% reduced adipogenic capacity. Since ASCs are a heterogenous population, cells differentiate to differing extents. However, in general, ASCs with the least f-actin stress fibers (visualized by phalloidin) or less cytoskeletal tension, tended to be more differentiated (i.e. had lipid droplets present, as seen by Bodipy staining), regardless of substrate (Fig. 2b). Since our goal was to replicate in vitro expansion, we did not seed at low densities, which would have allowed us to capture individual cell areas and quantify the cell spread-ness. If we assume nuclear area is an analog for cell area, then, we can extrapolate that cell spread area is significantly higher on the stiff substrates (Fig. 3a), where we observed about 50% less adipogenic differentiation than the soft substrate (Fig. 2c). These findings are consistent with the studies correlating increased cell spread-ness to reduced adipogenic capacity.

In a 2D environment, the cross-sectional area of the nucleus correlates with intracellular mechanics, i.e. spread cells on stiff surfaces yield nuclei with greater areas.^1,27,28,41 Given the changes in nuclear area over passage number and substrate (Fig. 3a) and the fact that intracellular tension can influence gene expression via the LINC-complex, we assessed the change in expression of both PPARG and RUNX2, the major transcriptional factors for adipogenesis and osteogenesis, respectively. Surprisingly, PPARG expression was slightly lower in ASCs on soft substrates and did not vary much over passages (Fig. 3b). This is in contrast to RUNX2, which increased substantially on soft substrates over passages (Fig. 3b). We also looked at gene expression of nesprin-2 (SYNE2), a major LINC protein that physically tethers actin to the nuclear envelope, allowing the transfer of forces between the cytoskeleton and nucleus. The statistically significant increase in nuclear size of ASCs on TCPS at P2 correlated to an increase in nesprin-2 (SYNE2), which aligns with previous work that has found stiffer mechanical environments and increased cell spreading lead to higher nesprin-2 expression and nuclear tension.⁹

Expanding large numbers of ASCs requires that we maintain high proliferation rates, which is challenging on soft substrates, where proliferation rates are impaired. We observed a decrease in proliferation rate on the soft substrate compared to the stiff substrate. So, expanding on TCPS until 1 passage before use would be ideal to maximize expansion rate. However, we also know that mechanical memory exists in BM-MSCs and several other cell types, so we sought to determine if expanding ASCs on TCPS before swapping to the terminal differentiation substrate (i.e., soft substrate to enhance adipogenesis) would limit adipogenesis as a result of this mechanical memory. We observed that ASCs grown on a stiff substrate and switched to soft at P4 did not yield the ‘soft’ phenotype observed in ASCs on soft substrates for all 4 passages. While the increase in nuclear area observed in ASCs on TCPS was reversed by swapping to PDMS at P4, SYNE2 expression remained significantly elevated (Figs. 4c and 4d). These results indicate that ASCs grown for multiple passages on TCPS will have elevated expression of LINC-complex components that could substantially alter how the nucleus senses its mechanical environment. Additionally, such differences persist despite changes to the microenvironment (i.e. swapping to PDMS) that lead to decreases in tension (i.e. smaller nuclear area). In agreement with other studies on stem cell mechanical memory,^24,44 the stiff substrate phenotype may be irreversible or require multiple passages on soft substrates. Likewise, ASCs expanded on a soft substrate then switched to a stiff substrate increased SYNE2 expression, but not to the same levels as cells on TCPS. The substrate swap at P4 demonstrated that some substrate-dependent differences, such as morphological properties and nuclear area, were reversible; however, other differences, like transcriptional regulation of nesprin-2 (SYNE2) expression, appeared to be, at least partially, irreversible. This suggests that observing only morphological features is not sufficient for determining the effects of long-term substrate dosing. Perhaps a residual effect of the prolonged mechanical dosing not returning to control levels is a result of PPARG not responding to the swapped substrate in either group (Fig. 4e). This was particularly interesting given that PPARG has been shown to be regulated by mechanical effects that involve cytoskeletal tension and specifically the RhoA/ROCK (Ras homolog family member A/Rho-associated protein kinase) pathway. Future studies will explore the role of RhoA/ROCK pathways in regulating PPARG activity in response to these long-term mechanical cues, especially as they relate to the incomplete nesprin-2 (SYNE2) recovery. Similarly, we will explore if chromatin remodeling may account for the incomplete return to baseline levels after prolonged substrate exposure.^22,24 We did not observe any changes to chromatin remodeling gene expression of Histone Deacetylase 1/2 (HDAC1/2), Histone acetyltransferase 1 (HAT1), after differentiation at P4 (data not shown), however, these effects may not be captured at the gene expression level and thus would warrant further characterization.

The primary aim of this study was to investigate how patient-derived ASCs behave during expansion on different in vitro culture substrates and explore how choice of culture substrate impacts their potential in different downstream applications. As such, ECM identity, architecture, and mechanical properties were not extensively explored here. PDMS was chosen as the soft substrate because it is inexpensive, easy to fabricate, and reproducible. We chose 100 kPa for PDMS stiffness, as this was the lowest stiffness we tested where ASCs were able to maintain proliferative capacity enough to reach confluence. This was found to be consistent even with epithelial cells where cells grown on ECM-coated PDMS of 50 kPa did not achieve confluence but would when cultured at 130 kPa or greater. While materials like polyacrylamide offer more tunability, they are too expensive and cumbersome to be feasible for large-scale in vitro expansions. Furthermore, PDMS has been used extensively in cell culture and, while we did not characterize collagen I adsorption in our system, other studies have demonstrated complete and homogenous coatings on PDMS with a variety of ECM proteins.^17,33 Even so, the way in which cells respond to their environment can be influenced by several properties beyond “bulk” stiffness on 2D surfaces. Differences in elasticity, viscoelasticity and/or tethering of matrix adhesion proteins all can affect the cell’s mechanical response to its substrate.^10,13,39 Therefore, these results are limited to these specific interactions and may not be broadly applicable to all substrates with the same bulk stiffness properties. In this study, it was important for us to understand the effect of in vitro exp ansion on ASCs in the conditions in which we would use (TCPS and PDMS).

Preliminary immunophenotype characterization of these populations resulted in the same general findings as the literature. Namely, ASCs are positive (>70%) for mesenchymal stem cell markers (CD73, CD90, CD105) and low (<5%) for hematopoietic/vascular markers (CD31, CD45). We could not perform flow cytometry characterization on later passages as our cell numbers per donor were limited in the PDMS group (where cells cultured on PDMS required a higher cell seeding density than standard TCPS cultures; for this study all initial cell densities were the same). When looking at single cell populations in general, the major subsets of CD markers, like those listed earlier, are shared by most donors, but minor subsets are highly donor dependent.³⁴ We did not look at these same minor, non-canonical subsets in our study but would explore this in future studies where donor variability is present.

These data include 4 racially diverse female subjects (2 African-American, 2 Caucasian), therefore we can say only within our limited cohort that there are consistent biological responses to these substrates. In general, when normalizing to each group’s controls, adipogenic gene expression trends were maintained regardless of donor. This is consistent with other studies which show that mesenchymal stem cells lose the in vivo phenotype after isolation, as the stromal vascular fraction, and in vitro expansion.^30,31 We cannot rule out donor specific differences that may occur in some disease states or due to biological sex. These findings also could potentially explain an inability of some cells to revert to a healthy state after that disease cue, such as increased stiffness due to fibrosis is corrected.

In summary, we must consider the environment in which a cell is expanded and differentiated in striving to optimize our yield and outcomes, but this comes at a balance between efficiency and time. In our studies, where generating adipose tissue models^{4–7,12,14,20,42} with ASCs or using ASCs for fat grafting or soft tissue regeneration is the ultimate goal, expanding on a softer substrate, for multiple passages prior to differentiation enhances adipogenic capacity 2.5-fold compared to on stiff substrates (Fig. 5d). Swapping to a soft substrate for the final passage was not sufficient to improve adipogenic capacity. Beyond our tissue of interest, not accounting for the ASCs previous expansion substrate may result in a more muted response to a soft substrate.