Mesenchymal stromal or stem cells (MSCs) exhibit the capacity to self-renew and, through in vitro induction, to differentiate into multiple tissue cell lineages [1] including bone, cartilage, and fat. Such cells derived from human bone marrow can be isolated and expanded in vitro, and are therefore feasible for clinical applications such as tissue regeneration [1]. However, the potential consequences of current in vitro isolation approaches and expansion conditions for these cells [2], as well as the exact function of such MSCs in vivo after re-administration to the body [3], are not well understood. Indeed, the exquisite sensitivity of stem cells to changes in the extracellular environment [4] renders MSC identity and behavior challenging to characterize and manipulate in vitro.

For stem, progenitor, and differentiated tissue cell types, cell-material interactions are critical to the adhesion, proliferation, and migration of cells in both physiological and pathological states [3], though the mechanisms driving these interactions remain elusive [5], [6], [7]. Historically, the materials comprising the in vitro extracellular environment reflects a compromise between the critical in vivo physical and chemical properties that facilitate cell proliferation and the simplified conditions amenable to large-scale repetition and facile observation of cell response. This tension is particularly apparent in the solid materials with which cells interact: MSCs and other adherent cell types are routinely isolated and expanded in number on optically transparent glass or polystyrene surfaces, which are topographically flat and also orders of magnitude stiffer than extracellular matrices in vivo. However, isolation, expansion, and phenotypic differentiation of MSCs in particular have been correlated with mechanical cues (e.g., stiffness [8] or topography of an underlying substratum [9] or three-dimensional scaffold [10], or interstitial flow [11]), as well as more well-studied biochemical cues (e.g., soluble growth factors and tethered ligand density [7]). Thus, most recent efforts to study and engineer the function of MSCs and other animal cells in vitro have focused on the solid materials to which these cells adhere. In contrast, the macromolecular nature of the fluid component of in vitro microenvironments has been much less scrutinized. We show here that the simplifications adopted in the vast majority of in vitro studies result in interesting consequences for the organization of and reciprocity between cells and their microenvironments.

Indeed, it has been largely overlooked that the in vivo soluble environment exhibits a much higher overall concentration of several macromolecular species than the in vitro counterpart [12]. Again, for reasons of practical convenience and well-controlled aqueous environments, the fluid component of in vitro cell studies comprises a so-called essential medium defined by concentrations of specific ions, amino acids, and growth-factor proteins; this medium is supplemented by an ill-defined but vital mixture of animal serum proteins [13]. The serum admixture, or protein additions that replace specific serum components, contribute the sole macromolecular element of in vitro culture (typically 1–20 vol%). However, the high dilution of serum proteins in the final culture medium results in much lower concentration (1–10 g/L or lower [12]) than in interstitial fluids (30–70 g/L [14], [15]), blood plasma (80 g/L [16]), or the cell interior (200–350 g/L [12]). The concept of macromolecular crowding effects is well described in material physics and is attributed to an excluded volume effect by which even low concentrations of small macromolecules can induce entropic segregation of other molecules [17]. Generally, the addition of as little as 1 vol% macromolecules to the in vitro environment sufficiently crowds the functional proteins of interest to accelerate biochemical reactions and assembly – including enzymatic and polymerization reaction rates, binding and folding kinetics, and gelation and protein fibril formation [18], [19] – to approach in vivo levels. As a result, many theoretical and experimental studies of macromolecular crowding in protein solutions have posited that consequences may exist for cell-material interactions. However, to our knowledge, the effects of macromolecular crowding on the cell and matrix materials have not yet been demonstrated directly. To better understand the role of extracellular macromolecules, we investigated the impact of macromolecular crowding on organization of the intra- and extracellular protein networks, as well as on specific behaviors of non-differentiating MSCs. We posit a reciprocity between the cells and the matrix material, whereby crowding directly affects matrix architecture and thus indirectly alters intracellular organization, matrix protein production, and cell behavior.

We increased the degree of extracellular crowding by adding Ficoll®, a specific macromolecular crowding agent, directly to basal cell-culture media. Ficoll is a non-interacting, neutral polymer comprising a copolymerization of sucrose and epichlorhydrine; this epoxide-linked polysaccharide exhibits a mean hydrodynamic radius of ∼4 nm (Fc70) or ∼8 nm (Fc400) [20], and is commonly utilized as a vitrification agent and in gradient centrifugation of cells. It is non-cytotoxic and does not significantly impact solution viscosity at relevant concentrations [21]. Further, unlike other macromolecules that have been considered crowding agents (e.g., dextran sulfate), Ficoll is net charge-neutral and thus does not conflate the excluded volume effects of a crowding agent [22] with the charge-based aggregation of proteins. Lareu et al. have demonstrated previously that addition of this macromolecule to the culture media of fibroblasts accelerated the enzymatic conversion of pro-collagen to collagen [23], [24], a crucial prerequisite for collagen matrix formation. However, the impact of this change on extracellular matrix (ECM) structure or on cell structure or function was not considered. In the present study, adult human MSCs were seeded at low density (∼7500 cells/mL or ∼2100 cells/cm²) in complete basal culture media; media was replaced 24 h after seeding, with or without addition of 62.5 mg/mL Ficoll mixture (see Methods) that is referred to hereafter as MMCs (macromolecular crowders). This crowding concentration represents the same order of magnitude of the total protein concentration found in blood plasma (80 mg/mL) [25]. More relevant to considerations of excluded volume effects, the corresponding fractional volume occupancy, defined as the fraction of total volume occupied by crowder macromolecules, is 17 vol%; this degree of crowding is representative of the perfused bone marrow compartment [20], [26].

Figure 1 shows the organization of the extracellular matrix when crowding is induced within the culture media. First, to characterize the matrix independently of the cells, MSCs were removed at day 7 via deoxycholate lysis buffer, retaining the ECM structure to enable imaging via atomic force microscopy (AFM, fully hydrated in aqueous solution). Organization of the extracellular matrix networks was quantified as the average angular standard deviation <α_sd> among filaments in such images. This approach captures the distribution width of fibril alignment in networks [27], where lower <α_sd> indicates increased degree of alignment, independent of relative fibril orientation (see Methods and Figure S1). As demonstrated by Figures 1A–C, the ECM deposited by cells cultured in crowded media conditions (+MMC) exhibited an increased degree of alignment of ECM fibers (p = 0.1363). However, ECM deposited by MSCs cultured in standard media conditions (−MMC) also resulted in poorly adsorbed ECM after lysing, ostensibly due to the proposed enhancement of surface attraction under crowded conditions [28]. Second, immunochemical staining of the intact ECM showed significantly increased alignment of both fibronectin (Figures 1D–F, p<0.001) and collagen IV (Figures 1G–I, p<0.03) under induced macromolecular crowding. Here, the intact ECM was primarily co-located with and surrounding cells; image locations for alignment quantification were thus selected randomly via 4′,6-diamidino-2-phenylindole (DAPI) nuclear DNA staining. Similar differences in matrix alignment were observed regardless of whether +MMC media was introduced 24 h after cell seeding (as in Figure 1) or at the initial seeding (data not shown). In summary, significant increases in ECM alignment are observed in crowded media conditions.

Enhanced physisorption of macromolecules including matrix proteins from solution to surfaces is a general hallmark of macromolecular crowding [22]. To determine whether MSCs produced and deposited more matrix proteins in the presence of MMCs, we compared amounts of collagen deposited with amounts of collagen secreted into the media. Figure 1J shows via Western blot that more collagen type-I was deposited by MSCs under +MMC conditions within 48 h, in agreement with earlier findings [24]. Further, normalized densitometry (Figure 1K) demonstrates a shift in collagen distribution from the media to the matrix under +MMC conditions. Specifically, the relative fraction of crosslinked collagen (β-bands) increases with respect to the uncrosslinked collagen (α-bands) in the ECM deposited by MSCs under crowded conditions. This is consistent with previous findings that crowding accelerates procollagen conversion and lysyl oxidase-mediated crosslinking [29]. Similar results were also observed for fibronectin (Figure S2), and have been reported for other crowding agents and cell types [30]. These data show that the amount and crosslinked fraction of a specific ECM protein are increased in crowded media conditions.

Figure 1 illustrates the effect of MMCs on organization of extracellular protein networks. To investigate the influence of MMCs on the organization of intracellular networks, we quantified the angular distribution of the cytoskeletal F-actin via immunocytochemical staining (Figures 2A–B). As in the above characterization of the ECM, image locations were selected randomly via DAPI nuclear DNA staining, to avoid biased analysis of cells exhibiting any particular morphology or orientation. Figure 2C shows that <α_sd> of the actin cytoskeleton decreased significantly for MSCs under +MMC conditions (p<0.02). This increased alignment of the cytoskeleton was thus observed concomitantly with increased alignment of the extracellular matrix. Further, Figure 2D indicates that the stiffness of the MSC cortical cytoskeleton, as quantified via AFM-enabled nanoindentation of these cells at 37°C, indented to a maximum depth of 20 nm, also increased significantly with macromolecular crowding (p<0.0001). This local stiffening is consistent with the increased alignment of the actin cytoskeleton in MSCs and other adherent cell types [2].

Although Ficoll is considered a neutral, non-interacting macromolecule, we explored whether it would reach the cell interior via endocytosis, thereby inducing crowding within the cell and altering cytoskeletal alignment independently of matrix alignment. We observed only minor uptake of both molecular weights of Ficoll by MSCs over 24 h, and confirmed via calcein labeling that these punctate regions did not release MMCs into the cytoplasm. Instead, MMCs remained confined within vesicles of mm-scale diameter (Figure S3). Further, even in dilute media conditions, the cell interior is already a highly crowded environment of total macromolecular concentration >200 mg/mL [31]. It is therefore unlikely that minor uptake of this crowder within vesicles significantly increased the degree of intracellular crowding.

It has been shown previously that dilute polymer solutions can undergo a transition with increasing concentrations of polymer, from disordered crosslinked networks to entropically demixed or more ordered phases similar to the nematic phase of liquid crystals [32]. To consider whether in vitro crowding of ECM proteins could directly produce aligned networks in the absence of adherent cells, we generated fluorescently labeled collagen matrices from pure collagen solutions, deposited on plasma-treated glass coverslips in the absence or presence of this same concentration and composition of the MMCs. We observed that addition of the MMCs to 0.5 mg/mL collagen resulted in a significant change in network orientation (p<0.0001), with increased collagen alignment and a correspondingly decreased <α_sd> (Figures 2E–G). When MSCs were seeded onto such paired unlabeled collagen matrices in the absence of MMCs in the culture media – i.e., seeded in uncrowded media onto a collagen matrix assembled under crowded conditions – cells exhibited a significantly increased actin cytoskeleton alignment (Figures 2H–J, p<0.0001) that was comparable to that exhibited by cells exposed only to crowded media (Figures 2A–B). These findings are consistent with the concept that macromolecular crowding promotes alignment of the extracellular matrix material at relatively low matrix protein concentrations, which in turn directs the organization of cell-matrix adhesion and cytoskeletal alignment. Further, we note that although other macromolecular crowding agents can also alter the organization of these collagen networks at similar fractional volume occupancy (Figure S4), these changes are typically conflated by additional factors such as charge-associated or hydrogen bonding-associated interactions with proteins [33].

Finally, to consider whether induction of crowding and the resulting increase in ECM alignment would affect cell behavior as well as cytoskeletal organization, we quantified MSC adhesion, proliferation, and migration in these in vitro environments. Cell adhesion and proliferation were quantified by seeding MSCs onto tissue culture-treated polystyrene at low density (∼2100 cells/cm²), with macromolecular crowders added during, or 24 hours after, initial seeding; whole-well cell counts were obtained via DAPI nuclear staining at 1, 3, and 7 days (see Methods). Figure 3A shows that MSC proliferation increased significantly (p<0.0001) by day 3 when crowding was induced 24 h post-seeding, and even more so when crowding was induced at initial cell seeding. This increased proliferation rate cannot be attributed to crowder-induced acceleration of DNA synthesis by cells, in that EdU assays of cells entering S-phase within 24 h was not statistically different in the presence of MMCs (data not shown). However, Figure 3B shows that MMCs also promoted initial adhesion of cells, with a significant increase in the number of cells adhered when seeded directly into crowded media. This finding is consistent with the increased production of matrix proteins by cells under crowded conditions and enhanced alignment of such proteins upon deposition under crowded conditions (Figure 1), as well as enhanced macromolecular adsorption to surfaces under crowded conditions [22]. Thus, enhanced proliferation when crowder was introduced at initial seeding, as compared to 24 hr later (Figure 3A), can be attributed to both the longer duration of crowding-facilitated matrix deposition and alignment at day 3, as well as the increased number of adhered cells at day 0. (It is established that MSC proliferation rates depend on initial cell seeding density [34].) This interpretation is further supported by Figure 3C, which shows that MSCs exhibited significantly increased proliferation (p<0.0001) when seeded onto matrices comprising collagen that was deposited with or without macromolecular crowders (as in Figure 2, but with unlabeled collagen). Here, no crowders were added to the cell culture media, and there was no significant difference in initial adhesion of MSCs to such matrices (p>0.4). Figures 3D–F shows that MSCs also exhibited significantly decreased migration efficiency (p<0.0001) in a classical in vitro wound healing assay [35] in which both the cell and matrix layers were mechanically removed from the surface, over 24 h in +MMC media. The reduced capacity of MSCs to migrate into a disrupted cell-matrix region is consistent with increased adhesion between the cells and the ECM that these cells deposited over ∼7 days in vitro prior to the wound (see Methods). This finding is further supported by oriented focal adhesions observed in Figures 2A and 2B, as well as a significant increase in effective Young's elastic modulus of the cortical actin cytoskeleton under macromolecularly crowded conditions, as demonstrated in Figure 2D. This finding is similar to the generally observed reduction in cell migration speed at sufficiently high concentration of adsorbed matrix proteins in vitro [36]. The present study was intentionally constrained to non-differentiating conditions of MSCs (see Figure S5), as differentiation is concomitant with complex matrix remodeling processes. However, as will be outlined separately [37], we have observed that the consequences of macromolecular crowding also extend to the in vitro differentiation potential of these cells.

Our understanding of stem cell characteristics and function, and therefore our progress in designing relevant in vivo applications, is derived primarily from in vitro cell culture models. Indeed, this intended correlation between in vitro material microenvironments and in vivo niches extends to nearly all cell types and cell responses quantified in laboratory settings. However, one important difference that we consider here is the orders-of-magnitude lower concentration of macromolecules in vitro [12], even when other microenvironment characteristics such as matrix stiffness, fluid flow rates, and specific protein concentrations are increasingly engineered to approach in vivo conditions. Macromolecular thermodynamics and kinetics in solution are significantly affected by excluded volume effects [38], [39], [40], [41], [42], and the cell exterior and interior are naturally crowded in vivo and subject to such excluded volume effects [43], [44], [45]. Thus, Ellis et al. have long argued that macromolecular crowding should be maintained explicitly when studying biological systems under physiologically relevant conditions [12], [46]. Indeed, it is well established that crowding via addition of more serum proteins is not a viable option, commonly resulting in cell overproliferation and other dysfunctions associated with superphysiological concentrations of growth factors and other proteins [47]. Therefore, synthetic polymeric crowding agents such as the MMCs considered here are desirable to minimize unintended biochemical interactions while inducing a level of macromolecular crowding commensurate with in vivo levels [48].

By inducing physiological levels of macromolecular crowding in the culture media of adult human bone marrow-derived MSCs, we identified clear correlations between the increased alignment and deposition of the extracellular matrix fibrillar networks (Figure 1), and the intracellular actin cytoskeleton (Figures 2A–D). These correlations in the presence of macromolecular crowding could be rationalized either by the cell guiding organization of the matrix material, or by the matrix guiding organization of the cytoskeleton. In other words, did MMCs primarily induce outside-in signaling from the material to the cell, or inside-out signaling from the cell to the matrix material [49]? We found that induction of crowding during deposition of collagen resulted in enhanced matrix alignment, even in the absence of cells (Figures 2E–G); and that cells that were subsequently deposited on such aligned collagen exhibited increased actin cytoskeleton alignment, even in the absence of MMCs in the media (Figures 2H–J). These findings are consistent with the concept that physiological levels of macromolecular crowding enhance matrix deposition and alignment, which in turn guides cytoskeletal actin alignment. This reciprocity is then manifested in associated increases in the cells' effective stiffness and matrix protein secretion, and potentially even the forces generated by the cell against the matrix material that may further promote ECM alignment.

Figure 4 summarizes this cell-matrix reciprocity under conditions of extracellular macromolecular crowding. Such chemomechanical coupling at the cell-material interface directly affects adhesion, morphology, and motility of adult human MSCs in the basal or undifferentiated state (Figure 3), and future studies can now consider whether in vitro differentiation potential into mesenchymal tissue lineage cells is also impacted. Note that this perspective does not obviate intracellular molecular signaling cascades as mediators of this cell-matrix reciprocity, but rather shows that physical interactions among extracellular macromolecules can significantly and independently modulate the local material environment of the cell [20], [26]. In fact, it is quite plausible that this increased matrix deposition and alignment are also highly correlated with local mechanical stiffness, densities of key tethered proteins, and other cues that have demonstrated correlations with cell morphology or function via increasingly detailed signal-response pathways [50]. However, these results demonstrate directly that induction of physiological crowding levels affects not only the long-noted increase in reaction rates among macromolecules in solution, but also the architecture and functionality of the cell-matrix environment. The induction of macromolecular crowding for in vitro systems enables one of the several critical components required to simulate in vivo niches with high fidelity, particularly via the extracellular material that is ultimately generated and maintained by the cells. Thus, increased consideration of such excluded volume effects in the laboratory can close the performance gap between engineered environments and biological niches, for applications ranging from basic cell science to stem cell delivery to tissue-based cancer therapies [26], [51], [52].