Rescuing mesenchymal stem cell regenerative properties on hydrogel substrates post serial expansion

Abstract The use of human mesenchymal stem/stromal cells (hMSCs) in most clinical trials requires millions of cells/kg, necessitating ex vivo expansion typically on stiff substrates (tissue culture polystyrene [TCPS]), which induces osteogenesis and replicative senescence. Here, we quantified how serial expansion on TCPS influences proliferation, expression of hMSC‐specific surface markers, mechanosensing, and secretome. Results show decreased proliferation and surface marker expression after five passages (P5) and decreased mechanosensing ability and cytokine production at later passages (P11‐P12). Next, we investigated the capacity of poly(ethylene glycol) hydrogel matrices (E ~ 1 kPa) to rescue hMSC regenerative properties. Hydrogels reversed the reduction in cell surface marker expression observed at P5 on TCPS and increased secretion of cytokines for P11 hMSCs. Collectively, these results show that TCPS expansion significantly changes functional properties of hMSCs. However, some changes can be rescued by using hydrogels, suggesting that tailoring material properties could improve in vitro expansion methods.


| INTRODUCTION
Human mesenchymal stem/stromal cells (hMSCs) are multipotent cells capable of differentiating into cell types found in tissues of the mesoderm (bone, cartilage, and fat), ectoderm (epithelium and neural), and endoderm (muscle, gut, and lung). 1 hMSCs are characterized by a cell surface maker profile, which is constituted by the positive expression of CD105, CD90, and CD73 and negative expression of CD34, CD45, and CD14. 1 hMSCs are also capable of secreting various cytokines and chemokines to modulate immune responses and promote wound healing. As a result of their myriad capabilities in regenerative therapies, hMSCs are one of the most widely used stem cells in clinical trials with over 800 trials registered worldwide. 2 hMSCs are being tested as cell-based therapies for the treatment of graft versus host disease, myocardial infraction, various neurological diseases, and bone and cartilage regeneration. Although the number of trials using hMSCs has increased three-fold over the past decade, the percentage of these trials that have advanced to Phase III/IV has stagnated around 2%-7% for multiple years. 3,4 While the lack of late phase trials is the result of many compounding problems, one contributing factor is a lack of robust, scalable, and reproducible methods that allow efficient Thus, the clinical use of hMSCs is contingent upon their successful ex vivo expansion. 8,9 While regenerative medicine applications exploit the multipotency and differentiation of hMSCs, they are also known to secrete many trophic factors that impact therapeutic outcomes. hMSCs secrete various cytokines and chemokines involved in immunodulation, especially those related to inflammation signaling, cell trafficking, and lymphocyte differentiation and proliferation. [10][11][12][13] Although the precise mechanisms involved in hMSC immunomodulation are largely unknown, several molecules, such as TNF-α and IL-6, 14,15 have been cited as potent regulators of initial inflammatory responses, while others, such as VEGF or HGF, 10,16,17 can aid in angiogenesis and wound healing. Additionally, hMSCs can suppress immune responses and promote tissue homeostasis by secreting PGE2 or IDO to promote M2 macrophage polarization. 15 hMSCs have the ability to sense the mechanics of their environment through integrins that translate extracellular mechanical cues into intracellular biochemical signaling. One output of this mechanotransduction is the nuclear shuttling of yes-associated protein (YAP) on culture substrates with high elastic moduli. 18 Many studies have used biomaterials with tunable elastic moduli and viscoelasticity to investigate the influence of these mechanical properties on the differentiation of hMSCs. For example, hMSCs have been shown to commit to a single cell lineage when cultured on substrates with moduli corresponding to tissue-specific matrix properties (e.g., E~0.1 kPa for neurogenesis, E~10 kPa for myogenesis, and E > 25 kPa for osteogenesis). 19 Following up on this work, Yang et al. 20 found that mechanosensing ability of hMSCs, and ultimately their multipotency, depended on the time of exposure to stiff matrix environments. Uninterrupted culture on substrates with stiff moduli (E~40 kPa) for 10 days caused irreversible YAP nuclear localization, even when the substrate was in situ softened (E~2 kPa). However, the effects were reversible when the exposure to the stiff microenvironment was shorter (<7 days). 20 Additionally, cells exposed to longer stiff mechanical doses were biased towards osteogenesis, losing their multipotency.
The time course of matrix stiffness has also been shown to influence angiogenesis. 17 hMSCs cultured on 4 kPa hydrogels showed increased mRNA levels of genes associated with new blood vessel formation compared to mRNA levels of hMSCs cultured on hydrogels of lower modulus. 21 hMSCs primed on soft hydrogels (~2 to 5 kPa) show reduced α-smooth muscle actin expression, a marker for a pro-fibrotic response, even after transfer to stiff hydrogels (100 kPa), indicating soft mechanical memory. 22 Previous studies have indicated that prolonged culture on traditional tissue culture plates and the use of enzymatic passaging methods can bias hMSCs toward an osteogenic fate, 23 cause loss of chondrogenic and adipogenic differentiation ability, 23,24 cause loss of DNA repair ability, 25 induce replicative senescence, and decrease cell surface markers essential to hMSC function. 23,26,27 Less is known about the effect that prolonged expansion on stiff surfaces may have on hMSC mechanosensing and secretory properties. Motivated by the growing body of evidence that hMSCs respond to both the magnitude and dose of their substrate modulus, we sought to further characterize the temporal changes that occur in hMSC properties both under typical expansion conditions and when transferred to hydrogel substrates. Together, we hypothesized that exposure to soft matrix cues after serial passaging on TCPS could recover or maintain the regenerative and multipotency properties of hMSCs lost during expansion.

| hMSC isolation and expansion
Fresh human bone marrow aspirate was purchased from Lonza (donor 18-year-old black female) and the hMSCs were isolated based on preferential adhesion to TCPS plates, using previous published protocols. 20,28 Freshly isolated hMSCs (P0) were detached with 0.05% trypsin-EDTA (Sigma) and subsequently centrifuged, counted, and frozen down in 80% fetal bovine serum (FBS; Invitrogen) and 20% dimethylsulphoxide and stored in liquid nitrogen. For passaging, hMSCs were cultured for 3 days on TCPS at an initial density of 4,000 cells/cm 2 in expansion media, detached with 0.05% trypsin-EDTA, centrifuged, and replated at the same density. Expansion media consisted of low glucose (1 ng/mL glucose) Dulbecco's Modified Eagle Medium (ThermoFisher) supplemented with 10% FBS (ThermoFisher), 1 ng/ml fibroblast growth factor basic (Life Technologies), 50 U/ml penicillin (ThermoFisher), 50 μg/ml streptomycin (ThermoFisher), 0.5μg/ml of Amphotericin B (ThermoFisher). This method was repeated to generate desired passage numbers. For subsequent analyses, cells at desired passage numbers (P2 for early, P5-P7 for middle, and P11-P12 for late) were frozen in cell freezing medium (Sigma) and stored in liquid nitrogen.

| Hydrogel fabrication
Hydrogels were polymerized as described previously described. 30 Briefly, polymer precursor solution was prepared by mixing 2wt/v% 40 kDa PEG-8NB (synthesized), 20 kDa PEG-8SH (Jenkem), 2 mM photoinitiator lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP), and 2 mM CRGDS adhesive peptide (Bachem) in PBS at a thiol:ene ratio of 1. The photoinitiator LAP has been used extensively in our group 31,32 and others 33,34 and has been shown to be cytocompatible. After vortexing, 12 or 50 μl of the solution was pipetted onto a hydrophobic Sigmacote (Sigma) treated slide. Sigmacote-treated slides were made by flaming glass microscope slides (VWR, 3 00 × 1 00 × 1 mm), soaking in Sigmacote for 30 min, washing thoroughly with DI water, and air drying. A 12 mm or 25 mm thiolated coverslip was placed on top of the droplet, and it was allowed to spread fully. Glass coverslips (VWR) were thiolated by vapor deposition of (3-mercatopropyl) triethoxy-silane performed overnight at 80 C. The polymer precursor solution was photopolymerized in between a sigmacote-treated glass slide and a thiolated coverslip with exposure to 365 nm UV light at 10 mW/cm 2 for 3 min to form hydrogels with a diameter of 12 or 25 mm and thickness of 100 μm. Hydrogels were equilibrium swollen in sterile PBS overnight before use.

| Rheological characterization
All rheological measurements were performed using a DHR3 rheometer (TA instruments) fitted with a UV light guide accessory with an 8 mm parallel plate tool. Optically thin hydrogels with a thicknesses of 250 μm were formed in situ by irradiating with 365 nm light (I 0 = 10 mW/cm 2 , Omnicure 1,000, Lumen Dynamics) for 30 s. The shear storage modulus (G 0 ) was characterized at constant strain (1%) and angular frequency (1 rad/s). The Young's modulus, E, was calculated using the following relationship.

| Proliferation
A Click-iT EdU Imaging Kit (ThermoFisher) was used to characterize proliferating cells at pre-selected passage conditions (P1, P7, and P12) and the manufacturer's protocol followed. In brief, hMSCs were seeded on either hydrogels or coverslips and treated with 10 μM EdU RT. Afterward, immunostaining was continued as described before.

| YAP nuclear localization and proliferation quantification
Using the Harmony software (Perkin Elmer), DAPI and Rhodamine Phallodin channels were used to identify the nuclear and cytoplasmic region of each cell in a single imaging plane. Using the YAP fluorescent channel, the average YAP intensity in the nuclear and cytoplasmic areas was calculated for each cell. Next, the YAP nuclear to cytoplasmic ratio was calculated as the average YAP intensity in the nucleus divided by the average YAP intensity in the cytoplasm. For proliferation, the total number of nuclei was calculated using DAPI staining.
Numbers of proliferating cells were quantified by the nuclei stained EdU+. Percent of proliferating cells was calculated for each field of view analyzed.

| Cytokine secretion analysis
Secretory profiles were assessed for early (P1) and late (P11) passage cells on TCPS and hydrogels using a Human Cytokine Array C5 (RayBiotech) and the manufacturer's protocol was followed. Hydrogels were pooled to ensure sufficient cell numbers for cytokine detection (>200,000 cells

| Statistics
All experiments were performed with at least three replicates per con- 3 | RESULTS

| Hydrogel characterization
Peptide-functionalized PEG hydrogels were synthesized via a thiol-ene photoclick reaction. 29 Eight-arm PEG-thiols were co-polymerized with eight-arm PEG-norbornene (equal stoichiometry) to form predominantly elastic hydrogels (E = 1-20 kPa), where the final modulus was controlled by the concentration of PEG macromolecules in the initial solution (Figure 1a,b). The fibronectin-derived integrin binding motif, CRGDS, was incorporated into the hydrogels at a 2 mM concentration to promote hMSC attachment. 20 While the hydrogels biochemical and biomechanical properties can be further tuned by selection of the initial formulation, all future studies used the 2 w/v% hydrogels (E~1 kPa). This modulus was selected as prior literature has reported the modulus of bone marrow to be~300 Pa. 36,37 The gel formulation also provided structural integrity during fabrication, culturing and transferring of cells. Furthermore, hMSCs cultured on hydrogel substrates with this elastic modulus had largely cytoplasmic YAP and remained proliferative. Thus, we aimed to compare the differences between hMSC properties when expanded on TCPS, where YAP nuclear localization predominates and the modulus is 6 orders of magnitude higher than these PEG hydrogel microenvironments. We were particularly interested in whether or not transfer of hMSCs to the PEG hydrogels post TCPS culture would allow them to recover their initial phenotype that might drift with TPCS expansion (Figure 1c).

| hMSC mechanosensing ability is lost with TCPS expansion
In addition to decreased proliferative capacity, hMSC mechanosensing ability was assessed at early (P1), middle (P7), and late passages (P12).  (d) Tukey plot reporting the YAP nuclear to cytoplastic ratios for each cell cultured on either TCPS or the hydrogel conditions. Statistics were performed on the mean YAP nuclear to cytoplasmic ratio for each condition (n = 3; #-late TCPS relative to early TCPS, ****p < 0.0001, # p < 0.0001, n.s., nonsignificant) hydrogels. Initially, 88 AE 1% of early passage hMSCs on TCPS were CD90+CD105+CD73+ (Supporting Information Figure S1). This population of cells decreased significantly for middle (56 AE 4%) and late passages (52 AE 13%); expanded on TCPS (Figure 5b). When early passage hMSCs were transferred to hydrogels post TCPS expansion, results showed that they maintained their cell surface marker expression over the entire 9 day experimental time course (Figure 5c). Strikingly, middle passage hMSCs on TCPS (56 AE 4%) were able to recover their immunophenotype when transferred to soft gels, with 86 AE 3% of the population expressing CD90, CD105, and CD73 after just 3 days on the hydrogels (Figure 5d). This~50% increase compared to TCPS controls was maintained over the course of 9 days on the hydrogels. In contrast, late passage hMSCs showed no recovery of their immunophenotype when transferred to the hydrogels, suggesting an irreversible change in the hMSC population after extended culture times on TCPS (Figure 5e).

| Enhanced hMSC secretome on hydrogels
Finally, the influence of transferring hMSCs to soft hydrogels on their secretory properties was measured using a cytokine array. Compared to their TCPS controls, secretion of most cytokines increased when either early or late passage hMSCs were cultured on hydrogels  Log fold change (Late TCPS relative to Early TCPS) FIGURE 4 Expansion of hMSCs on TCPS decreases their cytokine secretion. Log fold change in cytokine secretion of late (P11) relative to early (P1) passage hMSCs expanded on TCPS. Cytokines with no significant changes in their secretion are not reported hMSCs, such as graft versus host disease. Further, only hMSC populations positive for CD90, CD105 and CD73 over a specific threshold, usually 90-95%, are currently being administered to patients. 6,42 In this work, cell surface marker expression was increased for middle passage hMSCs by transferring them to hydrogels post TCPS expansion.
The percent of triple CD90+CD105+CD73+ cells increased from 56% to over 80% after 3 days on hydrogels, compared to their TCPS control ( Figure 5d). Cell surface marker recovery strategies like this could be of use in hMSC manufacturing, allowing for decreased expansion times while still achieving high cell numbers.
With respect to mechanosensing, significant differences were observed between middle and late passage hMSCs. As indicated by the higher mean YAP nuc/cyt ratios (Figure 3b), middle passage hMSCs remain sensitive to the culture substrate stiffness. Thus, culture on a substrate with an elasticity similar to their in vivo niche may have prompted the cells to begin restoring their cell surface markers when transferred to hydrogels. In addition, about half of the middle passage cells were still able to proliferate on hydrogels, increasing turnover ( Figure 2c). In contrast, the late passage hMSCs lose their responsiveness to the mechanical properties of their microenvironment, and with their low proliferation rates on hydrogels, are unable to recover their immunophenotype. However, both differences in the stiffness and biochemical surface properties of the hydrogels and TCPS are substantially different and could influence cell-matrix interactions. The thiolene PEG hydrogel system was formulated to present a single integrinbinding RGDS epitope, while TCPS is a surface that is highly modified , and late (e) passage hMSCs cultured on soft hydrogels and TCPS for 1, 3, 9 days (**p < 0.01, ***p < 0.001, **** p < 0.0001; n.s., nonsignificant) FIGURE 6 Secretory properties of hMSCs is enhanced of soft hydrogels. (a) Heatmap of cytokine secreted by hMSCs on early TCPS, late TCPS each relative to TCPS control, and early hydrogel, and late hydrogel each relative to hydrogel control. (b) Relative secretion of cytokines related to cell growth, chemoattractant, pro-inflammatory, anti-inflammatory, or lymphocyte response functions with adsorbed serum proteins. As a result, hMSC-material interactions and the strength of adhesion vary between the two systems. Increased cytokine secretion on hydrogels for both early and late passage hMSCs, the first being able to sense stiffness and another unable, could indicate that the change in surface chemistry from TCPS to hydrogel is involved in promoting hMSC secretory abilities. Additionally, this increase could indicate a connection to other mechanical sensing pathways, independent of YAP, that could still be active at late passages. Overall these results indicate that both expansion time and soft gel culture have an effect on the hMSC cytokine secretion. To further increase cytokine production, preconditioning strategies with proinflammatory molecules have been employed by other groups. 14,43,44 As cells were not pretreated in any way, further experiments help elucidate the effect of IFN-γ, IL-1β, or TNF-α simulation on secretion properties for early or late passage cells. As their immunomodulary and inflammatory response is better understood and defined, hMSC culturing conditions should be tailored to ensure maximum therapeutic potency.
Ultimately, it is important to recognize that each component of the hMSC phenotype is linked to the performance of another. For example, mechanical stiffness of microenvironment, sensed through integrins on the cell surface, can direct differentiation. Additionally, the cell surface marker CD73 has been shown to enhance immunosuppression by reducing inflammatory molecules in both B-cells and hMSCs, useful in treating autoimmune disorders like rheumatoid arthritis. 45,46 In umbilical cord derived hMSCs, loss of CD105 expression has been linked to decreased ability to inhibit Th1 lymphocyte proliferation in co-culture. 47 Decreased hMSC secretory potency and chemokine receptor expression can reduce homing ability to injured tissues. 48 The results of this study indicate that loss of properties is also linked. Loss in hMSC properties in vitro can have detrimental effects during in vivo transplantation. The success of stem cell therapies is contingent on the design of biomaterial systems to expand multipotent and regenerative hMSCs. By recovering immunophenotype and improving cytokine secretion during expansion, in vivo therapeutics of hMSCs could be improved.

| CONCLUSION
The goal of this study was two-fold: quantify the phenotypic drift of hMSCs during expansion on TCPS and then assess whether transfer of hMSCs to soft hydrogel matrices could restore lost phenotype.
Expansion solely on TCPS decreased hMSC proliferation rates, mechanosensing ability, cell surface marker expression, and secretory profile. Transfer of middle passage hMSCs to PEG hydrogels formed via a thiol-ene bioclick reaction (E~1 kPa) was able to restore expression of CD90, CD105, and CD73, cell surface markers crucial to hMSC definition and function. In contrast, late passage hMSC (P12) lost their YAP-associated mechanosensing and had low proliferation rates, and transfer to hydrogels was unable to recover their immunophenotype.
In addition, culture of hMSCs on hydrogels promoted cytokine and chemokine secretion from both early and late passage hMSC populations. The simultaneous quantification of changes in multiple cell properties with exposure to TCPS and soft hydrogel culture can inform a more optimal expansion time course designed to preserve desired hMSC properties.