Steering cell behavior through mechanobiology in 3D: A regenerative medicine perspective

Abstract

Mechanobiology, translating mechanical signals into biological ones, greatly affects cellular behavior. Steering cellular behavior for cell-based regenerative medicine approaches requires a thorough understanding of the orchestrating molecular mechanisms, among which mechanotransducive ones are being more and more elucidated. Because of their wide use and highly mechanotransduction dependent differentiation, this review focuses on mesenchymal stromal cells (MSCs), while also briefly relating the discussed results to other cell types. While the mechanotransduction pathways are relatively well-studied in 2D, much remains unknown of the role and regulation of these pathways in 3D. Ultimately, cells need to be cultured in a 3D environment to create functional de novo tissue. In this review, we explore the literature on the roles of different material properties on cellular behavior and mechanobiology in 2D and 3D. For example, while stiffness plays a dominant role in 2D MSCs differentiation, it seems to be of subordinate importance in 3D MSCs differentiation, where matrix remodeling seems to be key. Also, the role and regulation of some of the main mechanotransduction players are discussed, focusing on MSCs. We have only just begun to fundamentally understand MSCs and other stem cells behavior in 3D and more fundamental research is required to advance biomaterials able to replicate the stem cell niche and control cell activity. This better understanding will contribute to smarter tissue engineering scaffold design and the advancement of regenerative medicine.

1. Introduction

Mechanobiology is an emerging field investigating the translation of physical forces into molecular biological signals. These forces can greatly impact cell behavior, countering or synergizing with other cellular signals from soluble factors. A fundamental understanding of mechanobiology is of importance for many different research areas. In the fields of tissue engineering and regenerative medicine, this fundamental knowledge is being translated to practical applications, but many challenges still remain. One of the hurdles is that the basic mechaniobiological machinery is fairly well established in 2D, but research in 3D is still in its early stages. In order to create functional tissues, cells need to be in a 3D environment. Recent studies illustrate the differences in cell behavior and mechanobiology in 2D vs 3D, which will be discussed in detail in this review. To design smart materials for tissue engineering, a thorough understand of mechanobiology in 3D is critical. In this review, we will briefly go over the current understanding of mechanobiology in 2D and go deeper into how this translates to a 3D environment.

We will focus on the role of mechanobiology in 3D tissue engineering and regenerative medicine. Mesenchymal stromal cells (MSCs) are among the most widely used cell-type for tissue engineering, and they are also a good cell-type model for mechanobiological studies because their multi-potent differentiation potential is highly dependent on mechanobiology. For these reasons, MSCs will be taken at the center of the stage in this review. While well-established in 2D, we will discuss how the extra cellular properties affect cell behavior in 3D and how this is different compared to 2D.

Much research has been done on MSC differentiation into other lineages. MSCs have been proposed to differentiate into lineages of all three germ layers, such as: muscle, heart, liver, and neuronal, to name a few examples (the reader is referred to a number of critical reviews on this topic: [1–3]). This potential pluripotent differentiation has been proposed to be mechanosensitive [4] and is being investigated in different 3D scaffolds [5,6]. However, no functional in vitro or in vivo differentiation of MSCs into lineage other than osteo-, chondro- or adipogenic, has yet been shown. All experiments so far reported are performed by assessing cell morphology and the expression of only a few proteins, thus keeping the possible pluripotent character of MSCs contended. For these reasons, these studies were not considered in this review. If researchers are to further investigate the effect of materials, scaffolds and mechanobiology on this potential pluripotent differentiation of MSCs, we argue for the inclusion of functional tests, before strong conclusions are drawn. Therefore, in this review we will focus mainly on adipogenic/osteogenic lineage commitment of MSCs, while chondrogenic differentiation, proliferation and other cell behaviors are discussed to a lesser extent.

Cell morphology is another well-studied variable in 2D that can greatly influence MSC differentiation. Here, we will explore the influence of cell morphology and cell volume on MSC differentiation in 2D and 3D. Then, we will travel through the cell, discussing integrins, actinmyosin, Rho and ROCK, mechano-sensitive transcription factors and the nuclear lamina, focusing on their role in steering cell behavior in 3D. First, a brief general introduction to mechanobiology will be given, after which we will explore how this is similar or different in 3D environments. The reader is referred to excellent in-depth reviews on these topics that are mentioned throughout the text, for more focused reading on specific sub-sections of this review.

While we are beginning to get an understanding of mechanobiology in 3D, much is still unknown, as will be highlighted throughout this review. With more research, the better fundamental understanding of MSCs and other stem cells behavior in 3D will greatly aid the smarter design of tissue engineering scaffolds.

2. General introduction to mechanobiology

MSCs can differentiate to bone, fat and cartilage and are therefore widely used in tissue engineering applications (for specific MSC reviews the reader is directed to [7,8]). In 2D, MSCs are well studied and their lineage commitment and efficiency of differentiation is highly dependent on mechanobiology. Initial studies showed that MSCs on stiff substrates differentiated preferentially to bone, while soft substrates aided adipogenic and chondrogenic differentiation [9,10]. Since then, a number of material properties besides stiffness have been identified in 2D as potent guiders of cell behavior. These include, among others, material properties such as chemistry, degradability, stress-relaxation, nano- and microtopography, ligand type and density and externally applied force [11–13]. Besides MSCs, these material properties also influence the proliferation and differentiation of other cell types, such as pluripotent stem cells (induced (iPS), or embryonic (ES)), muscle stem cells, fibroblasts, adipocytes, osteoblasts, chondrocytes, among others [11,14].

These effects of extra cellular properties are orchestrated by a machinery of mechanosensitive proteins [15]. A cell adheres to the extra cellular matrix using integrins. Different combinations of α and β integrins can recognize different adhesion sites on extra cellular matrix (ECM) proteins, such as the widely used RGD-motif on fibronectin, or other motifs on other ECM proteins (for a specific integrin review the reader is directed to [16]). When the integrin dimer binds to a ligand and force is applied, early focal adhesion proteins such as focal adhesion kinase (FAK), paxillin and talin, among others, bind to the intracellular part of the integrin subunits. This starts a cascade that will cluster integrins and recruit other proteins to the focal adhesions, including zyxin, vinculin and actin filaments. When enough force is applied, these focal adhesions will mature into large protein complexes, creating a trans-membrane connection from the ECM, through multiple integrin dimers, to the actin cytoskeleton (for focused focal adhesion reviews, the reader is referred to [16–18]). Force on integrins is applied extracellularly, and/or intra-cellularly. An important force-generating element in cells are actin filaments with incorporated non-muscle myosin II (NM-II). Actin-myosin filaments join together to form large contractile bundles that exert force on the focal adhesions, creating tension in the cell (for a thorough actin-myosin review, we suggest [19]). The actin-myosin filaments can be connected to two focal adhesions, or a focal adhesion and the nucleus. The actin-myosin filaments connect to the nucleus through the linker of nucleoskeleton and cytoskeleton (LINC) complex [20]. The LINC complex bridges both nuclear membranes and connects to the nuclear lamina. The nuclear lamina is a meshwork of proteins just under the inner nuclear membrane and gives structural integrity to the nucleus [21,22]. The cellular tension leads to changes in gene expression through a number of different pathways. Mechanosensitive transcription factors such as Yes-associated protein (YAP), transcriptional coactivator with PDZ-binding motif (TAZ), myocardin related transcription factor (MRTF), and serum response factor (SRF), directly affect gene expression when translocated to the nucleus due to mechanical tension in the cell (the reader is referred to detailed reviews on YAP/TAZ [23] and MRTF/SRF [24,25]). The nuclear lamina can also change gene expression upon force on the nucleus, affecting among others chromosomal organization and histone remodeling. As referred throughout this section, there are plenty of reviews on mechanobiology and the individual proteins and pathways. However, these reviews are mainly based on data from 2D studies. While the behavior of these proteins in different environments and their effect on MSC differentiation and proliferation is fairly well established in 2D, their role and regulation are still vague in 3D. Here, we focus on the differences between 2D and 3D in response to extracellular properties and cell shape and how important mechanotransduction proteins have a different role and regulation in 3D than in 2D.

3. Extracellular material properties influence cellular behavior differently in 2D vs 3D

3.1. Substrate stiffness in 2D

Stiffness is arguably the most widely researched material property to influence cell proliferation and differentiation. In 2D, stiff substrates allow cells to build up high cellular tension and spread, with pronounced focal adhesions, actin stress fibers and YAP nuclear localization [4, 26–32]. On soft substrates, cells maintain a less spread morphology and display small focal adhesions, few actin stress fibers and cytoplasmic YAP [4,26,27,29–33]. This is true for many different cell-types, but in this review we will focus mainly on MSCs, because of their wide use in tissue engineering. MSCs on stiff substrates have increased osteogenic potential, while softer substrates increase adipogenic potential [26,28, 30–32]. In 2D, MSCs on substrates stiffer than 30–70 kPa undergo efficient osteogenic differentiation, while softer substrates inhibit osteogenesis [4,26–28,30–32,34]. Adipogenesis is optimal around 0.3–3 kPa [26–32]. Although generally inefficient in 2D, chondrogenic differentiation of hMSCs is also influenced by substrate stiffness. Softer substrates with reduced cellular tension increased chondrogenesis at 1 kPa [35], in line with results for chondrocytes [36]. Differentiation of other cell types is also dependent on substrate stiffness in 2D, in a similar stiffness range, e.g. muscle stem cells [37–40], induced pluripotent stem cells [41,42], embryonic stem cells [43], cardiac muscle cells [44] and fibroblasts [45]. Besides differentiation, proliferation is also controlled by substrate stiffness and cellular tension in 2D, in MSCs [34,46,47] and other cell types [36,40,48–50], where increased stiffness and higher cellular tension results in higher proliferation rates than on softer substrates. Altogether, the role of substrate stiffness in steering cell behavior in 2D is well established, particularly MSC differentiation and lineage commitment in 2D. However, to form functional tissues, cell behavior needs to be influenced in 3D. In the next section, we will go into the material properties that influence cell behavior in 3D and see how this is different from 2D.

3.2. Substrate stiffness in 3D

Similarly to the role of stiffness in 2D, Huebsch et al. showed that for MSCs encapsulated in 3D in RGD-modified alginate, agarose, and PEG hydrogels increasing stiffness led to increased osteogenesis and reduced adipogenesis [51]. The optimal stiffness for osteogenesis was around 20 kPa, whereas ~2.5 kPa for adipogenesis, similar to what was found in 2D. Others have shown a similar effect of matrix stiffness on osteogenesis in 3D gelatin gels [52]. Interestingly, however, stiffnesses of around 100 kPa inhibited osteogenesis in the alginate, agarose and PEG hydrogels [51]. In 2D, there does not seem to be such an upper limit to substrate stiffness, demonstrated by efficient osteogenic differentiation on extremely stiff polystyrene (such as conventional tissue culture plastic dishes) and glass (both in the GPa range) substrates. This upper limit to substrate stiffness in 3D likely arises from a reduced ability for cells to remodel the matrix and gather adhesive ligands when encapsulated in hydrogels. The importance of matrix remodeling is illustrated in a paper by Khetan et al. [53]. In contrast to Huebsch et al., Khetan et al. showed that irrespective of hydrogel stiffness (~4–91 kPa), only adipogenesis and no osteogenesis occurred in covalently crosslinked hyaluronic acid hydrogels [53]. Khetan et al. elegantly demonstrated that these differences arise from a difference in remodeling of the matrix. Alginate, agarose and PEG hydrogels are relatively dynamic due to non-covalent crosslinks, whereas the covalently crosslinked hyaluronic acid gels are static. Indeed, when 20 kPa alginate (inducing osteogenesis when non-covalently crosslinked) was covalently crosslinked, only adipogenic differentiation occurred [53]. This highlights that matrix stiffness on its own is not enough to guide MSC differentiation in 3D, as it is in 2D. In 3D, it seems that cells need to be able to remodel the matrix sufficiently in order to build up the required cellular tension to undergo efficient osteogenic differentiation. This is likely also the explanation for the upper limit of substrate stiffness in the 3D hydrogels of Huebsch et al.; even though the hydrogels are non-covalently crosslinked, at higher stiffnesses the gels become less dynamic. One strategy to improve the ability of a hydrogel to be remodeled by cells is to make the gels degradable. The next section will discuss the influence of degradability and show how this can be a key aspect in controlling MSC differentiation and proliferation.

3.3. Substrate degradability in 3D

While in 2D an increase in stiffness leads to increased cellular tension, Caliari et al. showed that the opposite is true in 3D covalently crosslinked gels [54]. The increased extracellular stiffness in 3D covalently crosslinked gels prevents cell spreading, similar to the results of Huebsch et al. [51]. Others have found a similar confining effect of matrix stiffness in 3D [29]. Interestingly, when Caliari et al. added MMP degradable peptide motifs to the covalently crosslinked matrix, spreading and YAP nuclear localization increased, indicating increased cellular tension. In line with this, when Khetan et al. added MMP-degradable peptides to covalently crosslinked hydrogels, so that the MSCs could remodel the matrix, almost all cells preferentially differentiated to the osteogenic lineage in 4.4 kPa gels [53]. Ferreira et al. also found that inhibiting hyaluronidase in 1 kPa degradable hyaluronic acid hydrogels inhibited osteogenic differentiation [55]. In type-I collagen gels, inhibiting membrane bound MT1-MMP (a collagen degrading enzyme) in hMSCs crippled degradation and osteogenic differentiation in vitro [56] and in vivo [57], while enhancing adipogenic differentiation. Lastly, bone formation was enhanced in vivo in MSC-laden degradable alginate hydrogels, as opposed to more slowly degradable gels [58]. On top of 2D collagen gels, however, the lack of MT1-MMP and degradation does not affect adipo- or osteogenic differentiation [57]. While stiffness properties of the degradable hydrogels are usually well documented, a quantification of degradability is not present in the highlighted papers mentioned in this section. Minimally, the final concentration of degradable peptides in the hydrogels should be measured in future studies, which could help the comparisons between studies and provide valuable starting points for new studies. Also, quantification of degradability by specific enzymes, ideally accompanied by quantitative information of how much the cells degrade the matrix would greatly aid comparability and a deeper understanding of the 3D cell-matrix interactions. Even though not quantified, these papers demonstrate the importance of matrix degradation in 3D hydrogels to build up cellular tension and undergo osteogenic differentiation. In 2D, this does not seem to be of importance. Also, stiffness seems to be of subordinate importance in 3D, highlighted by efficient osteogenic differentiation in 1 and 4.4 kPa hydrogels if the gels can be sufficiently degraded [53,55] (Table 1). Another method to increase the remodeling of hydrogels is by increasing stress-relaxation properties. In the next section, we will discuss how this effects MSC differentiation and how it relates to substrate stiffness. In the section after that, we will go deeper into why this matrix-remodeling is of such important in 3D, and not in 2D.

Table 1.

Overview of literature investigating the effect of stiffness on MSC differentiation in 2D and 3D, showing the importance of matrix modeling, more than substrate stiffness, in 3D environments.

Differentiation	Paper	Stiffness	Matrix type	Remodelability
Osteogenesis
2D	[4,26–28, 30–32,34]	>30 kPa	Flat 2D hydrogels	Non-remodelable 2D substrates
	[66]	1.1 kPa	2D collagen hydrogels	Remodelable, long, soft collagen fibers
3D	Huebsch et al. [51]	12–25 kPa	Alginate-RGD, agarose-RGD or PEG-RGD	Remodelable, non-covalent matrix bounds
	Khetan et al. [53]	4.4 kPa	MMP-degrable hyaluronic acid hydrogels	Remodelable, MMP degradable motifs
	Ferreira et al. [55]	1 kPa	Hyaluronic acid hydrogels	Remodelable, degradability
	Chaudhuri et al. [67]	17 kPa	Alginate-RGD hydrogels	Remodelable, fast stress relaxation
	Darnell et al. [61]	18 kPa	Alginate-RGD hydrogels	Remodelable, fast stress relaxation
	Nam et al. [62]	15 kPa	Alginate-RGD hydrogels	Remodelable, fast stress relaxation
	Lee et al. [60]	20 kPa	Alginate-RGD hydrogels	Remodelable, fast stress relaxation
No/Little osteogenesis
2D	[4,26–28, 30–32,34]	<30 kPa	Flat 2D hydrogels	Non-remodelable 2D substrates
	[66]	9.3 kPa	2D collagen hydrogels	Non-Remodelable, short, stiff collagen fibers
3D	Huebsch et al. [51]	>100 kPa	Alginate-RGD, agarose-RGD or PEG-RGD	Non-remodelable, too high stiffness
	Huebsch et al. [51]	2.5–5 kPa	Alginate-RGD, agarose-RGD or PEG-RGD	Remodelable, non-covalent matrix bounds
	Khetan et al. [53]	4–91 kPa	Crosslinked hyaluronic acid hydrogels	Non-remodelable, covalent matrix bounds
	Khetan et al. [53]	20 kPa	Crosslinked alginate-RGD	Non-remodelable, covalent matrix bounds
	Chaudhuri et al. [67]	17 kPa	Alginate-RGD hydrogels	Non-remodelable, slow stress relaxation
	Chaudhuri et al. [67]	9 kPa	Alginate-RGD hydrogels	Remodelable, fast stress relaxation
	Darnell et al. [61]	18 kPa	Alginate-RGD hydrogels	Non-remodelable, slow stress relaxation
	Nam et al. [62]	15 kPa	Alginate-RGD hydrogels	Non-remodelable, slow stress relaxation
	Lee et al. [60]	20 kPa	Alginate-RGD hydrogels	Non-remodelable, slow stress relaxation
Adipogenesis
2D	[26–32]	0.3–3 kPa	Flat 2D hydrogels	Non-remodelable 2D substrates
	[66]	9.3 kPa	2D collagen hydrogels	Non-Remodelable, short, stiff collagen fibers
3D	Huebsch et al. [51]	2.5–5 kPa	Alginate-RGD, agarose-RGD or PEG-RGD	Remodelable, non-covalent matrix bounds
	Khetan et al. [53]	4–91 kPa	Crosslinked hyaluronic acid hydrogels	Non-remodelable, covalent matrix bounds
	Khetan et al. [53]	20 kPa	Crosslinked alginate-RGD	Non-remodelable, covalent matrix bounds
	Chaudhuri et al. [67]	9 kPa	Alginate-RGD hydrogels	Remodelable, fast stress relaxation

3.4. Stress-relaxation properties in 3D

The studies on matrix degradability highlight the importance of matrix remodeling in 3D. In line with this, Chaudhuri et al. found that with faster stress relaxation in 3D RGD-alginate gels, osteogenesis greatly increased, while adipogenesis decreased [59]. Often used synthetic hydrogels are mostly elastic, whereas natural ECMs derived hydrogels are viscoelastic [59]. By increasing viscoelastic (stress relaxation) properties of the alginate hydrogels, the MSCs are able to harness more RGD than in slower relaxing gels. Indeed, increasing RGD density by 10-fold increased osteogenic potential in the slow and fast relaxing gels. However, MSCs in low-RGD fast relaxing gels still exhibited much greater osteogenic potential than in high-RGD slow relaxing gels [59]. Others have found similar importance of stress relaxation for osteogenesis, where faster stress relaxation greatly enhances osteogenic differentiation of MSCs in vitro [60–62]. When implanted in vivo, faster stress relaxing hydrogels increased bone formation [61]. In accordance with the differentiation results, proliferation also increased in degradable or fast-relaxing gels for MSCs [60] and other cell types [59,62–64]. To quantify stress relaxation properties, t_1/2 is extracted from stress relaxation tests, which is the time it takes for the stress to relax to half of its original value. Chauduri et al. found better osteogenesis in t_1/2 = 300 s than 2300 s, but no difference between 60 s, 140 s and 300 s [59]. Adipogenesis was decreased in 60 s vs 140 s, 300 s and 2300 s. Lee et al. demonstrated better osteogenesis in t_1/2 100 s than 500 s, and no difference between 100 s and 25 s [60]. Nam et al. obtained better osteogenesis in t_1/2 20 s hydrogels than in 20,000 s hydrogels [62]. Lastly, Darnell at all showed better in vitro and in vivo osteogenesis in t_1/2 50 s than 800 s [61]. Together, this data seems to indicate that stress relaxation of t_1/2 < ~500 s is good for osteogenesis of MSCs, below which no difference in osteogenesis has yet been demonstrated. Only one study investigated adipogenesis and indicated a lower threshold of t_1/2 = 140 s or more [59]. This information is, however, not readily available in the discussed papers and (except from Chauduri et al.) had to be estimated from the graphs. The mentioned t_1/2 should therefore not be read as precise measurements, but rather estimates. Future studies should mention precise t_1/2 and could investigate a wide range of stress relaxation properties in different materials, stiffnesses and concentrations of adhesive ligands, to get a better idea of the ideal stress relaxation range for each of the MSC tri-lineage differentiations.

Altogether, the fast-relaxing and degradable hydrogels demonstrate that for a cell to build cellular tension, it needs to harvest adhesive ligands. While pulling on the ligands, the matrix deforms, and its deformability determines the cellular tension. However, when a cell can harvest enough adhesive ligands, it can build up high (enough) cellular tension even in soft materials. Indeed, it has been shown that the faster stress relaxation in hydrogels allows the cells to gather more adhesive ligands [65]. This brings us to the next section where we discuss the importance of gathering adhesive ligands 3D, and in 2D.

3.5. Adhesive ligand gathering in 2D and 3D

In 2D, adhesive ligands are plentiful and in close proximity. This allows easy integrin clustering and large focal adhesion formation, which allows pulling on many ligands and the build-up of cellular tension. If the matrix is soft, the force that can be built-up is low, if the matrix is stiffer, higher cellular tension can be built-up. In 3D, however, adhesive ligands are distributed in three dimensions, and thus the matrix needs to be remodeled in order for the cell to gather many adhesive ligands in close proximity. Research with nanopatterned surfaces in 2D has revealed that to form focal adhesions and spread on thin nanometer adhesive lines, the lines have to be closer than 110 nm apart, or cross with each other [68]. This resembles the situation of the hydrogels, consisting of nanometer sized fibers, and demonstrates the need for close proximity of the ligands. In 2D, a cell can move freely over the surface where adhesive ligands are presented. When adhesion sites are in close enough proximity, integrin clusters can form to create focal adhesions and build up cellular tension, without the need to remodel the substrate. In 3D, however, adhesive ligands are embedded in a matrix and need to be discovered and gathered before integrin clusters can form to build up cellular tension. To gather multiple adhesive ligands, the matrix needs to be remodeled. This could explain the importance of remodeling the matrix in 3D, which is illustrated elegantly in the degradation and stress-relaxation studies discussed above.

The question then arises whether matrix remodeling is only important in 3D, or also in 2D. When using the right conditions, 2D matrix deformation also plays a role in cellular tension. The effect of stiffness on cellular tension and MSC differentiation in 2D is mostly tested on slowrelaxing or covalently crosslinked gels. Here, significant adhesive ligand clustering is difficult for cells [65], so cellular tension is mostly determined by matrix stiffness and ligand proximity [68]. However, osteosarcoma cells increased spreading on soft substrates (~1 kPa) with fast stress-relaxation, while cells remained round on soft slow-relaxing substrates [67]. This shows that adhesion ligand gathering can allow cells to build up cellular tension in 2D, like in 3D. The importance for ligand gathering in 2D is also nicely illustrated in a set of studies that use micrometer sized fibers, rather than the nanometer sized fibers of fibrillary hydrogels. These micrometer sized fibers also partially recapitulate the fibrillary ECM in vivo, making it an interesting substrate type to study [69–71]. On top of a thin layer of stiff, unmovable electrospun micrometer sized fibers, MSCs and fibroblasts displayed little cellular tension and proliferation [72]. When fibers were softer and could be displaced more easily, the cells built up more cellular tension and significantly increased proliferation. Similarly, MSCs increased osteogenic differentiation on long, movable collagen fibers (1.1 kPa), while adipogenic differentiation was enhanced on smaller, stiffer, less movable fibers (9.3 kPa) [66]. Bulk stiffness was very low in both conditions (16.4 Pa). On soft, movable fibers, MSCs spread and formed focal adhesions, while no focal adhesions were observed on stiffer, less movable fibers [73]. Similar results have been shown for other cells [74], and this has been further investigated on a variety of different materials. In these studies, thin microfibrillar meshes were used, so cells grow on top and were in a 2D environment. The studies on microfibrillar substrates are interesting because they highlight the importance of ligand gathering and matrix remodeling in a 2D environment and argue against the direct influence of material stiffness. The 2D microfibrillar studies show a similar importance for ligand gathering to build up cellular tension in 2D, similar to the 3D hydrogel studies. The 3D stiffness, degradation and stress-relaxation studies, together with the 2D stress-relaxation and microfibrillar substrate studies, illustrate the importance for adhesive ligand gathering and the subordinate role for stiffness in both 2D and 3D.

In summary, in most 2D culture substrates, stiffness is very important for cellular tension because adhesive ligands are in close proximity and easily accessible. In 3D, however, adhesive ligands need to be gathered, so the matrix must be remodeled. When adhesive ligands can be gathered sufficiently, cells can build up high cellular tension in 3D even in soft materials (Fig. 1). This is demonstrated by the fact that osteogenesis can occur at much lower stiffnesses (1–4 kPa) in 3D when cells are able to remodel the matrix sufficiently, than in 2D [53,55] (Table 1).

Fig. 1.

Schematic visualization of differences in adhesive ligand gathering between 2D and 3D. When a 3D matrix is remodelable, due to (a combination of) degradability, fast stress relaxation, non-covalent matrix bounds and a low-enough stiffness, cells can harvest multiple adhesive elements to build up high cellular tension. In a static 3D matrix, which is non-degradable, has slow stress relaxation, is covalently bound and/or too stiff, cells cannot gather enough adhesive ligands to build up high cellular tension. In a classic flat 2D culture setting, the substrate is also static, but there are enough adhesive ligands so that the matrix does not need to be remodeled. In the presence of enough adhesive ligands in 2D, the stiffness of the matrix determines the cellular tension. Black lines represent ECM with adhesive ligands. Outlook on extracellular material properties in 2D vs 3D.

Now that several individual matrix properties that influence MSC differentiation in 3D have been identified, investigating a combination of these factors could result in superior scaffolds for tissue engineering. For example, as discussed throughout this section, matrix deformability seems to be more important than stiffness in 3D, but stiffness still plays an important role. It is likely that there is an optimal combination of the two to guide MSC differentiation. Combinations of faster stress-relaxation and degradation of a 3D matrix could also result in even further enhancement of cellular tension, especially when done in a gel of optimal stiffness for a targeted cell differentiation lineage.

How cells are cultured before such experiments is also important, as MSCs have been shown to have mechanical memory. When cultured on soft surfaces after pre-culture on stiff surfaces, MSCs preferentially differentiated to the osteogenic lineage. Without such pre-culture on stiff surfaces, MSCs efficiently underwent adipogenesis [75]. This mechanical memory is likely regulated (partly) by epigenetic changes [76]. The mechanical memory of MSCs has until now only been studied in 2D, but it would be important to study in 3D as well.

Whereas adipogenic and osteogenic differentiation are highly influenced by cellular tension, chondrogenesis is also highly dependent on specific protein interactions, such as cadherins [77]. Most MSC mechanobiological studies, therefore, focus on adipogenesis and osteogenesis. However, material properties also influence chrondrogenesis in 3D. Faster stress relaxation hydrogels, for example, allow for the deposition of ECM further away from chondrocytes during chondrogenesis, creating a more interconnected matrix [78]. A deeper investigation of the influence of material properties on chondrogenesis in 3D could advance cartilage tissue engineering approaches. In addition, most of the research on matrix properties and mechanobiology in 3D is done with hydrogels. However, many tissue engineering approaches utilize other scaffold types, such as electrospun- or additive manufactured scaffolds, for their superior structural mechanical properties. Very few studies investigate how MSC differentiation is affected directly by these environments. We have recently shown that cellular tension of MSCs is reduced in 3D electrospun and additive manufactured scaffolds, made up of stiff materials, when compared to the same stiff materials in 2D [79]. Others have also hinted at a decrease in cellular tension in epithelial cells on electrospun scaffolds, showing a decrease in focal adhesions in 3D electrospun scaffolds compared to 2D [80]. This shows that even when using stiff materials, cellular tension can be low and is affected by the properties of the 3D environment. Effects of material properties in 2D cannot be translated directly to 3D, as we have discussed in this section. More research into how different material properties affect MSC proliferation and differentiation is required for more effective design of biomaterials. For example, properties of electrospun scaffolds such as optimal fiber stiffness, diameter and inter-fiber linkage, are yet to be investigated in a comparative manner for their effect on mechanobiology of MSCs. Individual parameters of additive manufactured scaffolds would also be interesting to further investigate and optimize, such as fiber diameter and curvature, pore size and shape and material stiffness. In addition, excreted proteins during culture time has also been shown to influence cellular tension and differentiation in 3D [81]. A deeper investigation into how material properties influence this ECM and remodeling protein secretion and how these in turn influence cell behavior would aid the understanding of these complex interactions. A better understanding of how these individual parameters influence MSC differentiation, and which of these are dominant factors, could greatly aid the design of smart scaffolds for tissue engineering and regenerative medicine strategies. In addition, fundamental knowledge of how MSCs gather adhesive ligands and build up cellular tension in 3D is still lacking. In the sections below, we will go into more detail of the molecular mechanisms that orchestrate cellular tension and MSC differentiation in 3D. First, we will look at the literature of cell morphology and cell volume, and how it influences MSC differentiation in 2D and 3D.

4. Cell morphology and volume in 2D and 3D

4.1. Cell morphology

In 2D, MSC osteogenic differentiation is greatly enhanced by cellular tension, marked by actin stress fibers with non-muscle myosin-II, large focal adhesions, a stiff nucleus (high lamin A/C expression) and YAP nuclear translocation [4,26–32,34]. The opposite is true for adipogenic differentiation, which is aided by low cellular tension [26–32]. Generally, in 2D stiff substrates such as standard tissue culture plastic, MSCs are allowed to spread and build up high cellular tension. This spread morphology is a good indicator for cellular tension in 2D, and it can also directly influence cellular tension [82], due to an increase in adhesion area [83]. Restriction of MSC spreading by confinement demonstrates that small cells display little cellular tension and are biased towards adipogenesis, while large cells exhibit increased cellular tension and improved osteogenesis [84]. Also, when MSCs are forced in a specific cell shape, cellular tension and differentiation is influenced. More circular shapes allow less cellular tension and bias MSCs towards adipogenesis, while rectangular shapes increase cellular tension and drive MSCs towards osteogenesis [85]. Cell shape also influences behavior in other cell types in 2D, such as proliferation [86,87], migration [88,89], apoptosis [86] and tumorigenicity [90].

Based on these studies, one would expect that cell spreading and morphology in 3D settings also greatly influences differentiation. However, differentiation of MSCs has been fully decoupled from cell morphology in 3D. In hydrogels, both small, circular cells and larger, spread cells have been shown to undergo efficient osteogenesis [51,53, 55,59,60,91]. Also, both circular and spread cells have been shown to differentiate to adipocytes [51,53,59,60,91]. In 3D additive manufactured scaffolds, less spread cells more efficiently underwent osteogenic differentiation than more spread cells, contrary to 2D results [92]. This highlights once again that studies performed in 2D cannot be expected to directly translate to a 3D setting. Cell morphology does not seem to have the same predictive power in 3D as it does in 2D. However, recent studies have demonstrated a role for cell volume in 3D differentiation. In the next section, we will discuss the role of cell volume in 2D and 3D differentiation.

4.2. Cell volume in 3D

Rather than cell morphology, a few studies have shown the importance of cell volume as a regulator of cellular tension and differentiation in 3D. Bao et al. investigated cellular tension and differentiation in cells that were confined to different morphologies and volumes in 3D, in specifically shaped wells with lid in hydrogels [93]. They found more actin stress fibers, focal adhesion and YAP nuclear translocation in smaller cells, regardless of cell geometry, up to a certain lower limit. Too little cell volume also decreased cellular tension. Indeed, cells with optimal volume for cellular tension (~3600 μm³) differentiated preferentially towards osteoblasts, while cells with a bigger volume (~6000 or ~9000 μm³) decreased osteogenic differentiation and increased differentiation towards adipocytes. Cells with smaller volumes (~2200 μm³) slightly decreased osteogenesis and showed no difference in adipogenesis. In the cells with optimal volume for osteogenic differentiation, cell morphology influenced differentiation, but to a much smaller extend than cell volume. Seemingly contradictory, Lee et al. later reported that increasing cell volume (by changing stress-relaxation properties or osmotic pressure) increased osteogenesis [60]. However, the investigated cell volumes were in a different range. Cells with a volume of ~1000 μm³ displayed little osteogenic differentiation, while ~2000 μm³ displayed efficient osteogenesis, in accordance with Bao et al. Perhaps a further increase in osteogenesis would be observed if cell volume was increased to 3000–4000 μm³ (the optimal volume for osteogenesis in the study of Bao et al. [93]). The increase in osteogenesis in response to volume increase was regulated through TRPV4, actin-myosin and RUNX2. Because faster stress-relaxation increased cell volume and osteogenic potential, it is likely that other 3D environments that allow matrix remodeling (such as a degradable hydrogel) also increase volume and osteogenesis. This has, however, not yet been investigated. The two studies investigating cellular volume in 3D show the importance of cell volume for MSC differentiation in 3D. Is this a unique attribute of MSC differentiation in 3D, or is this also important in 2D?

4.3. Cell volume in 2D

As cells spread and increase their surface area, cell volume decreases [94,95]. The change in cell volume upon cell spreading is true for a range of cell types, including MSCs. When cell spreading was inhibited, by physical confinement or a decrease in substrate stiffness, cell volume remained larger [94,95]. When the volume of MSCs cultured on a stiff surface (spread morphology, reduced cell volume) was increased with osmotic pressure to the volume of MSCs on soft surfaces (round morphology, more cell volume), adipogenic differentiation of MSCs was enhanced [95]. Vise-versa, when the volume of MSCs cultured on soft surfaces was decreased with osmotic pressure to the volume of MSCs on stiff surfaces, osteogenic differentiation was improved. The cell volume of MSCs on spread surfaces, where osteogenic differentiation is most efficient, was around 2000 μm³. Round MSCs on a soft surface had a volume of around 2500 μm³. The optimal volume for osteogenic differentiation seems to be different between 2D and 3D (~2000 μm³ in 2D vs ~3600 μm³ in 3D) (Fig. 2). This study shows that cell volume can also directly influence differentiation of MSCs in 2D, like in 3D.

Fig. 2.

Influence of cell volume, rather than cell morphology, on MSC differentiation. In 2D, MSCs cultured on a stiff substrate have a large surface area, high cellular tension, and a volume of ~2000 μm³. On soft substrate, MSCs spread less and have lower cellular tension with a volume of ~2500 μm³. By changing the volume with different substrates or osmotic pressure, a dominant role for cell volume on MSC differentiation was found [95]. In 3D, MSCs that are able to remodel the matrix sufficiently can build up higher cellar tension, which influences cellular volume. Based on two papers investigating 3D cellular volume, an optimal volume of ~2000–36000 μm³ to induce osteogenesis was found, while larger or smaller volumes drove cells more towards adipogenesis [60,93]. Black lines represent ECM with adhesive ligands. Outlook on cell morphology and cell volume in 2D and 3D.

The 2D and 3D cell volume studies discussed above question the direct influence of cell morphology on MSC fate. In 2D, when cell morphology was spread, but volume increased, MSCs underwent more efficient adipogenic differentiation, while round cells with decreased volume increased osteogenic potential. As cell volume changes with cell morphology, this questions whether the studies on cell morphology in 2D are not indirectly looking at the effect of cell volume, rather than cell morphology. In 3D, cell morphology has also been decoupled from cell volume, and a dominant role for cell volume over morphology was found. Only one study directly investigates cell volume vs cell morphology in MSC differentiation in 2D, and only two studies in 3D. From these studies, many questions arise. The optimal cell volume is still to be determined in both 2D and 3D. Ideally, a comparative study between 2D and 3D could be done, using the same materials and investigating a large range of cell volumes. Chondrogenesis in 3D has also been correlated to a change in cell volume, but direct causality or the underlying mechanisms have not yet been explored [78]. The molecular mechanisms underlying the role of cell volume in MSC differentiation to any of the three lineages is still unclear. Does cell volume induce a change in cellular tension and affect MSC differentiation through that, or is there an independent pathway? Confirming the dominant role of cell volume over cell morphology, determining the optimal cell volume for MSC differentiation, and investigating the molecular pathways could greatly improve understanding and aid design of tissue engineering constructs. In the next section, we will go over some of the main players in mechanosensing and how their role or regulation can be different in 2D vs 3D.

5. Molecular mechanisms of mechanosensing in 2D and 3D

In this section, we will explore the role and regulation of some key actors in the mechanotransduction pathway. We will focus on their role in 3D mechanobiology and guiding MSC differentiation, and what the differences and similarities are compared to 2D culture systems. Many individual proteins have been shown to play a role in mechanotransduction and MSC differentiation and proliferation. Here, we will focus on the most widely studied proteins that play critical roles in MSC mechanotransduction (Fig. 3). Starting from the cell membrane and traveling inwards, we will start with integrins, then go to focal adhesions, actin-myosin, RhoA and ROCK, YAP/TAZ, MRTF/SRF and the nuclear lamina. The findings of the sections below are summarized in Table 2.

Fig. 3.

Overview of mechanotransduction pathways highlighted in this review. The interactions between these proteins have mainly been discovered in 2D and is now being investigated in 3D. Each of the highlighted proteins, when active, has a positive influence on osteogenesis of MSCs in 2D and a negative influence on adipogenesis. In 3D, this distinction is not so clear and is still subject to investigation. In short: integrin binds the extracellular matrix and senses mechanical forces. Large protein complexes called focal adhesions form on the intracellular part of integrin clusters and attach to actin-myosin polymers. Actin-myosin chains attach to other focal adhesions or to the LINC complex on the nucleus. RhoA and ROCK play an important, but not exclusive, role in myosin phosphorylation and actin-myosin force generation. The LINC complex passes both nuclear membranes and attaches to the nuclear lamina, consisting of Lamin A, C, B1, B2 and emerin, among others. YAP and TAZ enter the nucleus upon sufficient cellular tension where they serve as co-transcription factors. MRTFs bind to globular-actin monomers that prevent nuclear entry. When concentrations of G-actin are sufficiently low, MRTF can enter into the nucleus and bind to SRF to initiate target-gene transcription.

Table 2. Summary of the influence of mechanosensitive and mechanotransducive proteins on MSC differentiation in 2D and 3D.

	Reference	Action	Result
Integrin
2D	[96–99]	Inhibition	Decreased osteogenesis, increased adipogenesis
3D	[51,55,57,81,100–102]	Inhibition	Decreased osteogenesis, increased adipogenesis
Focal adhesions
2D	[28]	Vinculin knock down	Decreased osteogenesis, increased adipogenesis
	[102–105]	FAK knock down/inhibition	Decreased osteogenesis
	[106]	FAK knock down	Decreased adipogenesis
3D	Focal adhesions remain unstudied in 3D MSC differentiation
Actin-Myosin
2D	[4,29,107–109]	Blebbistatin (NM-II inh.)	Decreased osteogenesis, increased adipogenesis
	[29,110]	Latrunculin A (actin pol. inh.	Decreased osteogenesis
3D	[53,59,60]	Blebbistatin, Latrunculin A, ML-7 (MLC inh.)	Decreased osteogenesis, increased adipogenesis
	[101]	Blebbistatin	No effect on osteogenesis
	[101]	Latrunculin A	Increased osteogenesis
RhoA/ROCK
2D	[84,97,107,108, 111–114]	RhoA or ROCK inh.	Decreased osteogenesis, increased adipogenesis
	[114]	Constitutively active ROCK	Increased osteogenesis
3D	[53]	ROCK inh.	Decreased osteogenesis, increased adipogenesis
	[53]	Constitutively active ROCK	Increased osteogenesis, decreased adipogenesis
	[59]	ROCK inh.	Increased osteogenesis
	[101]	ROCK inh.	No effect on osteogenesis
	RhoA remains unstudied in 3D MSC differentiation
YAP/TAZ
2D	[29,115–117]	Constitutively active YAP or TAZ	Increased osteogenesis, decreased adipogenesis
	[115–120]	YAP and/or TAZ knock down	Decreased osteogenesis, increased adipogenesis
3D	[29]	Constitutively active YAP	Reduced adipogenesis, increased osteogenesis
	[57]	Constitutively active YAP or TAZ	Increased osteogenesis in vivo
	[116]	Activated TAZ	Increased osteogenesis in vivo
MRTF/SRF
2D	[121]	MRTF-A knock out	Decreased osteogenesis, increased adipogenesis
	[121,122]	MRTF-A or SRF overexpression	Increased osteogenesis, decreased adipogenesis
	[122]	MRTF/SRF inh.	Increased adipogenesis
3D	The MRTF/SRF pathway remains unstudied in 3D MSC differentiation
Nuclear lamina
2D	[26,123]	Lamin A/C knock down	Decreased osteogenesis, increased adipogenesis
	[26,124]	Lamin A/C overexpression	Increased osteogenesis
	[125]	MAN1 knock down	Increased osteogenesis, decreased adipogenesis
	[123]	LINC complex inh.	Increased adipogenesis
3D	The nuclear lamina remains unexplored in 3D MSC differentiation

5.1. Integrins

Different combinations of α and β integrin subunits form dimers to bind different ECM proteins and form a mechanosensitive physical link between the ECM, through the cell membrane to the cell interior [126]. Clustering of different integrin dimers allows cells to build up cellular tension, driving MSC differentiation towards osteogenesis in 2D. Similarly, in 3D osteogenic differentiation has also been shown to be dependent on integrin clustering [51]. Indeed, inhibiting integrins in 2D inhibits osteogenesis and favors adipogenic differentiation of MSCs [96–99]. Like in 2D, inhibiting integrins in MSCs in 3D also leads to decreased osteogenesis [51,55,57,81,100,101] and increased adipogenesis [51,55,81,102]. The role of integrins in 3D differentiation of MSCs has been well studied in different 3D environments. As opposed to other proteins discussed later in this review, the role of integrins seems to be very similar in 2D as in 3D.

In 2D, when ligand density is sufficiently high, cellular tension through integrins is mainly dependent on stiffness [127]. In 3D, however, integrin clustering is regulated by local reorganization of the matrix, to gather sufficient adhesive ligands. This is dependent on stress relaxation and degradation in hydrogels [57,62,67], and fiber stiffness and movability in fibrous meshes [65,66]. Increasing the stiffness in 3D reduced integrin bonds, arguably due to a decrease in cells’ ability to remodel the matrix [51]. We refer the reader to the ‘adhesive ligand gathering in 2D and 3D’ section above for a more elaborate discussion on adhesive ligand gathering in 3D. Next, we will discuss the role of focal adhesions in 3D mechanosensing.

5.2. Focal adhesions

Focal adhesions are complex structures and consist of more than 60 different proteins that link the ECM-binding integrins to the actin cytoskeleton, of which some are mechanosensitive, and others are mechanotransducers [128]. Focal adhesions play an important role in transducing mechanical signals from the outside of the cell to the inside. On 2D stiff materials, number and size of focal adhesions vary per cell type, but all adherent cells form many relatively large focal adhesions [128,129]. On 2D soft materials, fewer and smaller focal adhesions form [129]. Important focal adhesions proteins, such as vinculin [130], focal adhesion kinase (FAK) [131], paxillin [132], talin [133] and zyxin [134], among others, all contribute to mechanotransduction. This contribution is both mechanical, creating the bridge between ECM and actin cytoskeleton, and by signaling, activating other effector proteins.

Focal adhesions increased during osteogenesis of MSCs, while they decreased during adipogenesis [135]. However, surprisingly little is known about the effects of these proteins on lineage commitment of MSCs. Vinculin knockdown decreased osteogenesis and increased adipogenesis [28]. In line with this, inhibition of FAK inhibited osteogenesis in MSCs, osteoblasts and fibroblasts in 2D [102–105,111]. Also, proliferation of MSCs [136] and other cells [131] was reduced by FAK inhibition. Interestingly, however, FAK was also required for adipogenesis in MSCs and adipocytes and fibroblasts [106,137,138], and for the formation of fat tissue in vivo in mice [139]. Adipogenesis was inhibited by high cellular tension, but few and small focal adhesions were still present [4,26,27,29–33]. This suggests that FAK has a signaling role for FAK in adipogenesis, rather than aiding in the build-up of cellular tension. Other proteins such as paxillin, zyxin and talin remain unexplored in their role in MSC differentiation in both 2D and 3D, even though they are likely to play an important role. More research is required to investigate the specific role of the different focal adhesion proteins in MSC differentiation. Understanding how MSCs convey mechanical signals from outside the cell to the inside, and specifically which proteins are required for the right cellular response, is critical for the smart design of tissue engineering constructs.

In 2D there is a strong correlation between many large focal adhesions and osteogenesis, and few small focal adhesions and adipogenesis [4,26,27,29–33]. In 3D, however, there are often very few and small focal adhesions in MSCs and other cell types [140–142], but osteogenesis can still efficiently occur [53,81,93]. Fraley et al. [140] have shown that this reduction in focal adhesions is a direct response to the 3D environment. On top of 2D gels, many large focal adhesions formed, but when a gel was dispensed on top of MSCs attached to 2D gels, or when cells were embedded in 3D hydrogels, only few and small focal adhesions were observed. We have recently shown a similar decrease of focal adhesions in stiff 3D electrospun and additive manufactured scaffolds, compared to 2D films of the same stiff material [143,144]. Rather than forming large protein complexes that tightly anchor the cell to the 2D substrate, the focal adhesion proteins are distributed more diffusely through the cell and regulate protrusion activity and matrix deformation [140,145]. Creating protrusions and deforming the matrix is required to gather adhesive ligands in 3D, while it does not play a great role on stiff 2D substrates with plenty closely spaced adhesive ligands. This suggests a critical difference between the role of focal adhesions in 2D and in 3D. This difference is likely to translate more broadly to the build-up of cellular tension, and thus to proliferation and differentiation. Further exploring the role of individual focal adhesion proteins, and the focal adhesion protein complex as a whole in a 3D setting, could greatly aid both fundamental understanding and tissue engineering. This could be investigated in hydrogels, allowing for relatively easy control of properties such as stiffness, stress relaxation and ligand density, among others. However, as most scaffolds exploited for tissue engineering are made up of stiff materials, researching the role of focal adhesion proteins in these settings will also greatly improve understanding of cellular behavior in tissue engineering scaffolds.

5.3. Actin-myosin

The actin cytoskeleton is a critical mechanosensitive component in the mechanotransduction machinery. Actin monomers polymerize in filaments and multiple filaments come together to form large actin fibers. Non-muscle myosin II (NM-II), consisting of two heavy chains, two essential light chains and two regulatory light chains, incorporates within the actin fibers to create contractile actin-myosin fibers, called stress-fibers. When the actin-myosin elements contract, tension is created between its two attachment points [19]. These two attachment points can be two focal adhesions, or a focal adhesion and the nucleus. On 2D stiff materials, large actin stress fibers are observed in many different cell types. On soft materials, a more diffuse actin network was observed, with fewer stress fibers, thinner actin fibers and more actin around the cell periphery [4,26,27,30,31]. Actin-myosin contractility greatly influenced MSC differentiation, as inhibiting NM-II activation by blebbistatin on 2D stiff materials greatly reduced osteogenesis and enhanced adipogenesis [4,29,107–109]. Actin polymerization inhibitors, such as latrunculin A, also inhibited osteogenesis in MSCs [29, 110]. In 3D, however, pronounced stress fibers as in 2D were often not observed, regardless of stiffness, degradation or stress-relaxation properties of the matrix [47,51–53,57,59,60,81,146]. However, both osteogenesis and adipogenesis can still occur in cells without pronounced stress fibers in 3D [47,51–53,55,59,60,143,146]. Osteogenic differentiation of MSCs without pronounced stress fibers has been extensively shown in hydrogels and we have recently also shown this for MSCs in 3D electrospun and additive manufactured scaffolds [143]. This demonstrates that the actin cytoskeleton does not have the same predictive power in 3D as it has in 2D. Yet, consistent with 2D studies, inhibiting myosin activation or actin polymerization in 3D hydrogels results in reduced osteogenesis and increased adipogenesis [53,59,60]. This shows that actin-myosin tension is still very important in 3D MSC differentiation, but that thick actin-stress fibers are not required for sufficient cellular tension to induce osteogenesis. One study, however, has shown no difference in osteogenic differentiation in the presence of myosin II inhibitor blebbistatin [101]. On top of this, surprisingly, the actin polymerization inhibitor latrunculin A increased osteogenic potential. Interestingly, inhibition of integrins did inhibit osteogenesis, showing that the induced osteogenesis was reliant on cell adhesion to the matrix, but not reliant on actin-myosin contraction. This is contradictory to the above-mentioned studies that found decreased osteogenesis upon blebbistatin inhibition. The concentrations of blebbistatin are the same in the papers (50 μM), but one key difference is the use of osteo-inductive medium supplements. Whereas Khetan et al., Lee et al. and De La Cruz et al. used osteo-inductive factors and show a dependence on actin-myosin for osteogenic differentiation, the hydrogel used by Parekh et al. induced osteogenesis without osteogenic medium. Lee et al. showed that TRPV4 activation can induce osteogenesis even in the presence of blebbistatin by directly regulating cell volume [60]. This shows that there are (at least one) NM-II independent pathways that can induce osteogenesis. Perhaps the hydrogel used by Parekh et al. activates such a pathway. Together, these studies show that actin-myosin tension is required for medium-induced osteogenesis, but that there might be alternative pathways that can bypass the requirement of actin-myosin.

Besides NM-II, smooth muscle myosin has also been suggested to be involved in osteogenic differentiation of MSCs [147], and smooth muscle actin is also important for MSC osteogenesis [27]. Little research has been done to further explore the role of smooth muscle actin and smooth muscle myosin, particularly in 3D. Further research is required to understand the role of other myosins and other actin-modulating proteins in MSC lineage commitment. Also, proteins such as TRPV4, that seem to influence differentiation independent of actin-myosin tension, would also be interesting to further investigate. This could help to better understand results obtained from specific inhibitors, such as blebbistatin. Also, as stress fibers don not seem to have the same predictive power in 3D as they do in 2D, having a profile of proteins that are important for MSC differentiation could greatly aid tissue engineering scaffold optimizations.

5.4. RhoA/ROCK

Rho GTPases and ROCK are mechanotransducers and play important and diverse roles in cell adhesion and cellular tension. RhoA, one of the Rho GTPases, stimulates stress fiber formation and cellular tension mainly, but not exclusively, through ROCK. The reader is referred to an excellent review on Rho GTPases regulation [148]. ROCK phosphorylates myosin regulatory light chain (MLC), which activates the ATPase activity of NM-II, resulting in actin-myosin contraction. In addition, ROCK inhibits MLC phosphatases, and affects a wide range of actin modifying proteins, resulting in the stabilization and formation of actin stress fibers [149]. Other well studied GTPases such as Rac and CDC42 have different effects on actin and are more involved in the formation of protrusions [150]. In MSCs in 2D, the role of RhoA and ROCK in adipo- and osteogenic differentiation is well studied. In human MSCs, inhibition of either RhoA or ROCK reduced cellular tension [107], promoted adipogenic differentiation and reduced osteogenesis [84,97,107,108, 111–114]. In murine MSCs, ROCK inhibition also increased adipogenesis and decreased osteogenesis [109,110,122,151]. Commonly used ROCK inhibitors such as fasudil and Y27632 have been shown to also inhibit a wide range of other protein kinases to varying degrees [152, 153]. It is important to understand that the effects ascribed to ROCK inhibition in the above-mentioned papers could also be partially ascribed to other inhibited kinases. Introduced expression of active proteins is a more direct way of measuring the effect of a protein. Supporting these inhibitor studies, constitutive expression of RhoA increased stress fibers [154] and constitutive activity of ROCK increased osteogenic differentiation [114] in human MSCs. Together, these works demonstrate an important role for RhoA and ROCK in MSC fate-commitment, osteogenic differentiation and mineralization. Other Rho GTPases remain understudied, however, and further research is required to better understand the interplay between the different GTPases, kinases and differentiation of MSCs.

The role of RhoA and ROCK is far less clear in osteoblast differentiation. In rat or murine primary calvaria osteoblasts, inhibition of ROCK increased differentiation and mineralization [155–157], opposite to the effect on MSC osteogenesis. In the murine pre-osteoblast cell line MC3T3-E1, inhibition of ROCK with fasudil led to increased mineralization [158]. In a different paper, but the same cell line, ROCK inhibition with Y27632 decreased mineralization [159]. This suggests a different effect of fasudil and Y27632 on differentiation of this cell line, potentially due to effects on proteins other than ROCK. In murine primary calvaria osteoblasts, however, both Y27632 and fasudil increased mineralization [156]. In the human osteoblast cell line MG63, ROCK inhibition with Y27632 decreased differentiation [160]. Also for RhoA, mixed results have been found. Inhibition of RhoA with C3 toxin reduced mineralization in the MC3T3-E1 cell line [159], but increased mineralization in murine neonatal calvaria osteoblasts [156]. These conflicting results show an opposite effect of ROCK or RhoA inhibition depending on the specific cell-type and inhibitor that is used. Also, the increased mineralization after ROCK or RhoA inhibition in some of the osteoblast studies is contrary to the results in MSCs. A potential explanation for this is the state of differentiation. ROCK and RhoA may be important for initial differentiation of MSCs towards osteoblasts, but inhibit further differentiation or mineralization at a more mature osteoblastic stage. Further research into these differences could greatly aid tissue engineering. First, a better understanding of the role of ROCK, RhoA and other GTPases in osteogenic differentiation is required. Then, the ideal timing of activation of these proteins should be studied. If no inhibitors are used, one could think of materials that change properties over time, for example, to inhibit RhoA or ROCK activation after an initial differentiation stage.

In 3D, human MSCs display fewer stress fibers upon RhoA inhibiton [93], illustrating a similar role of actin modulation in 3D as in 2D. In line with this, constitutively active ROCK in human MSCs induced osteogenesis in covalently crosslinked HA gels that normally induce adipogenesis [53]. In degradable HA hydrogels, inhibition of ROCK decreased cellular tension on the matrix and drove cells from osteogenesis to adipogenesis [53]. These results are in line with the results for MSCs in 2D and show a similar role of RhoA and ROCK in driving MSCs towards osteogenesis. However, one study found no effect of ROCK inhibition on osteogenesis of human MSCs in PEG hydrogels [101]. The hydrogels used in this study induced osteogenesis without the addition of osteogenic factors. Osteogenic differentiation in these hydrogels was independent of actin-myosin tension and ROCK. This study suggests an alternative pathway of osteogenic differentiation, independent of actin-myosin and ROCK, as already discussed above in the actin-myosin section. In mouse MSCs, ROCK inhibition leads to increased ALP activity in slow-stress relaxing 3D alginate gels. In fast-stress relaxing gels, where cellular tension was highest, inhibition of ROCK had no effect [59]. These results are contradictory to the results of human MSCs, where ROCK inhibition decreases osteogenesis.

These conflicting results highlight the need for further research. A better understanding of these important proteins and in what situations they are required for osteogenesis and when they are not is needed for a more intelligent design of tissue engineering constructs. Analyzing the effect of ROCK or RhoA inhibition in MSCs in hydrogels with different properties (e.g. stress relaxation, degradability, stiffness) could aid this research. Also, the differences between mouse and human MSCs could be directly compared to exclude a species-specific effect.

The roles of CDC42, Rac1 or other Rho GTPases in MSC differentiation are mostly unstudied in 3D, nor in 2D. CDC42 and Rac1 are more involved in protrusion activity, rather than cell contractility like RhoA and ROCK [150]. One study inhibited Rac1 in 3D hydrogels and found no difference in osteogenesis [67]. This suggests no role for Rac1 in osteogenesis, although it could be worth investigating in different types of hydrogels or other 3D scaffolds. It could be that Rho GTPases other than RhoA are not important for 2D differentiation of MSCs, but are important for 3D osteogenesis. As discussed earlier, MSCs in 3D gather and cluster adhesive ligands by reorganizing the ECM to build up cellular tension. Protrusion activity could play an important role in a cells’ ability to sense the surrounding matrix. Studies inhibiting and overexpressing the different Rho GTPases, preferably with knock down/out or specific inhibitors, in different 3D environments would advance fundamental knowledge on 3D MSC differentiation.

5.5. YAP/TAZ

YAP and TAZ are a well-studied pair of mechanotransducive co-transcription factors that transfer to the nucleus upon high cellular tension in 2D. This nuclear translocation is regulated by two different pathways. One is phosphorylation, which increases nuclear export and degradation by the proteasome. The other is by direct force on the nucleus, opening nuclear pores and allowing YAP to move inside the nucleus by active transport. The reader is referred to an excellent review on YAP regulation [161]. On 2D stiff substrates, where MSCs favor osteogenic differentiation, YAP is located in the nucleus. On soft substrates, YAP is predominantly cytoplasmic and MSCs favor adipogenesis. YAP has been directly involved in the differentiation of MSCs. Expression of non-degradable YAP or activation of TAZ inhibits adipogenesis and increases osteogenesis [29,115–117], while knockdown of YAP, TAZ or YAP and TAZ has the opposite effect [115–120]. Also, bone specific YAP/TAZ knockout mice have severely impacted bone formation [162]. Together, these papers clearly show an important role for both YAP and TAZ in differentiation of MSCs.

In 2D YAP and TAZ are regulated by integrin, RhoA, ROCK and actin-myosin tension [29,115,161]. In 3D, YAP and TAZ have also been shown to be dependent on integrin [57,81], ROCK and actin-myosin [57]. However, in 3D, as opposed to 2D, nuclear entry of YAP does not always correlate with adipo- or osteogenic differentiation of MSCs [59,60,143], while in other studies it does correlate [29,57,93]. In 3D, osteogenic differentiation can still occur without pronounced nuclear translocation of YAP [67,143], adipogenic differentiation with mainly nuclear YAP [67], and different efficiencies of differentiations do not correlate with different levels of YAP [60,67]. Similar to 2D, non-degradable YAP reduced adipogenesis in 3D [29]. Also, MSCs expressing non-degradable YAP or TAZ exhibit increased osteogenesis in vivo [57]. Another study also found increased osteogenic differentiation and in vivo bone formation by pharmacologically activating TAZ in MSCs [116]. In summary, YAP and TAZ are not yet well studied in 3D tissue engineering, but the first results seem to reveal a similar function as in 2D. The studies where YAP and MSC lineage commitment did not correlate, found YAP nuclear localization in gels where MSCs preferred adipogenesis, and more cytoplasmic YAP in gels where MSCs preferred osteogenesis. This shows that YAP alone is not sufficient to guide MSC cell behavior and that other factors also play an important, and sometimes dominant, role. It could be that the low levels of YAP are still sufficient to allow for osteogenesis, and that high levels of YAP alone are not enough to inhibit adipogenesis. Further research into the function of YAP and TAZ in 3D is required to better understand their role in 3D differentiation. Also, in 2D nuclear entry of YAP and TAZ is often used as a read-out of cellular tension (and extrapolated to MSC lineage commitment). As discussed above, this correlation does not always seem to hold true in 3D. More in-depth investigations into the regulation of YAP and TAZ in different 3D environments could shed light on this and help to better interpret YAP and TAZ results in 3D.

5.6. MRTF/SRF

The MRTF family of mechanotransducive co-transcription factors (Myocardin, MRTF-A and MRTF-B) are regulated by actin. Myocardin is mostly expressed in cardiac and smooth muscle cells, while MRTF-A and B are more widely expressed. MRTFs bind to G-actin in the cytoplasm and are then prevented from entering the nucleus. When G-actin polymerizes into F-actin, MRTFs release from G-actin and enter the nucleus. Here, they bind to SRF to initiate transcription of a variety of genes [163]. Most, but not all, SRF target genes are MRTF controlled [164]. The MRTF/SRF pathway has been shown to be important in 2D MSC adipogenic and osteogenic differentiation. MRTF-A knock-out murine MSCs exhibit increased adipogenesis and reduced osteogenesis [121]. In line with this, overexpression of either MRTF-A or SRF leads to increased osteogenesis and decreased adipogenesis [121,122]. Inhibiting the MRTF/SRF pathway with CCG1423 increased adipogenesis in MSCs [122]. Pre-adipocytes also increased adipogenesis after knockdown of either MRTF-A or -B [165], or SRF [166]. SRF knock-out osteoblasts had reduced osteogenic differentiation capacity [167]. In vivo, MRTF-A or SRF knock-out mice displayed decreased bone-mass [121,167], illustrating the importance of MRTF/SRF also in 3D settings. However, the MRTF/SRF pathway has not yet been investigated in 3D MSC differentiation in tissue-engineering constructs. We have recently shown that the MRTF/SRF pathway influenced proliferation of MSCs in 3D electrospun scaffolds through the regulation of FGFR1 [144]. Given the profound effects of MRTF/SRF on 2D MSC differentiation, investigating their role in a 3D setting would be highly interesting. MRTFs are regulated by G-actin and enter the nucleus when actin polymerizes. In 3D, however, large actin (stress) fibers are mostly absent (see ‘actin-myosin’ section). Investigating the regulation of MRTF nuclear entry in 3D could shed light on actin dynamics and the role of MRTF in steering MSC behavior. Also, YAP/TAZ nuclear entry does not always correlate with MSC osteogenesis in 3D, as opposed to 2D (see ‘YAP/TAZ’ section). Understanding a similar correlation for MRTF could help to better unravel the role of the MRTF/SRF pathway in 3D differentiation, and aid in the optimization of biomaterials. Also, we have shown that SRF expression was decreased in 3D electrospun scaffolds, compared to 2D controls. This is another potential pathway of regulating MRTF/SRF activity and would be interesting to investigate in different 3D biomaterials.

5.7. The nuclear lamina

Part of the extra- and intracellular forces are transmitted directly to the nucleus through the cytoskeleton. Forces on the nucleus can directly regulate gene expression through histone modifications and chromatin remodeling [168–171]. These force-induced changes in gene expression can be very rapid and can precede the effects of nuclear translocation of transcription factors such as MRTF-A [171]. The actin cytoskeleton connects to the nuclear lamina through the LINC complex. The LINC complex spans through the inner- and outer nuclear membrane and connects to the nuclear lamina [172,173]. The nuclear lamina is a meshwork of mechanosensitive proteins just under the inner-nuclear membrane that gives structural integrity to the nucleus. It consists of a number of proteins, among which lamin A, -B1, –C and emerin are arguably the most well-studied. Lamin A and C, more than the other nuclear lamina proteins, give the nucleus its mechanical properties [174–176]. In response to force, lamin A and C stiffen the nucleus and emerin plays an important role in regulating this stiffness increase [177]. Besides the mechanical functions, the nuclear lamina plays important roles in gene regulation. The proteins in the nuclear lamina can change the spatial distribution of DNA, impact histone modifications or directly bind to promotor regions [170,178–181]. In addition, lamin A, –C, emerin and the LINC complex also impact nuclear localization of mechanosensitive transcription factors such as MRTF-A and YAP [26, 119,123,182,183].

In 2D, MSCs expressed more lamin A and C, but not B1, on stiff vs soft surfaces [31]. Also, upon adipogenic differentiation, MSCs [26] and adipocytes [184] reduced lamin A and C expression. During osteogenic differentiation, MSCs increase lamin A and C expression [169,185]. This correlation is also functional, as lack of lamin A, C or emerin restricted the transmission of forces to the nucleus [174,186,187]. After lamin A and C knock down, MSCs exhibit reduced osteogenic and increased adipogenic potential [26,123]. In addition, lamin A/C overexpression increased osteogenesis of MSCs [26,124] and lamin A and C knock-out mice have reduced bone formation and turn-over [188]. Knockdown of MAN1, another nuclear lamina protein, increased osteogenesis and reduced adipogenesis in MSCs in 2D [125]. Lastly, disruption of the LINC complex increased adipogenesis of MSCs [123]. Besides MSCs, differentiation of other cell types is also affected by the nuclear lamina. Lamin A and C or emerin knock down decreased differentiation of, for example, muscle stem cells [173] and osteoblasts [123]. Together, these studies highlight the importance of the nuclear lamina and the LINC complex in transducing forces to the nucleus to guide differentiation of MSCs and other cell-types in 2D.

In 3D, however, very little research has been performed to elucidate the functions of the nuclear lamina in MSC differentiation. We have recently shown that the 3D environment greatly reduced lamin A and C expression, even in stiff materials [143]. We hypothesized that this was due to a difference in force distribution on the nucleus. In 2D, the actin fibers going over the nucleus created indentation sites on the nucleus [189,190]. Lamin A, –C and LINC proteins accumulate at these indentation sites. At these indentation sites, a change in chromatin [189] and histone acetylation [190] has been observed, showing that these indentation sites can directly affect gene expression. In 2D, actin fibers going over the nucleus push down on the nucleus, creating the indentation sites. On concave surfaces, where the downward force on the nucleus is further increased, lamin A and C expression was higher [191]. In 3D, however, cells adhere in a volumetric manner, with adhesions distributed above and below the nucleus, rather than in a single 2D plane. Indeed, on convex surfaces, where downward force on the nucleus is reduced due to cellular adhesions above the nucleus, lamin A and C expression was lower. This illustrates the importance of force distribution within the cell in regulating lamin A and C expression, greatly influenced by the adhesion sites in all three dimensions (Fig. 4). The impact of the change on lamin A and C expression in 3D has not yet been investigated. Actin-myosin has been shown to influence histone acetylation [93] and -methylation [192] in 3D, which it also does in 2D [193]. It has not yet been investigated whether this is coordinated by the nuclear lamina, or through other pathways. Investigating the effect of overexpression or knock-down of nuclear lamina proteins on MSC lineage commitment in different 3D settings could shed light on its role in 3D. In addition, mapping the interaction with- and re-organization of DNA by the nuclear lamina in 3D could elucidate more specific functions in gene regulation. Specifically, a better understanding is required of how forces on the nucleus regulate the expression of individual or groups of genes, and what forces nuclei experience in 3D.

Fig. 4. Forces on the nucleus are dependent on the three-dimensional distribution of adhesions.

On flat 2D surfaces, adhesions form below the nucleus. Actin fibers attach to these adhesions and attach to- or go over the nucleus, creating downward force on the nucleus. On concave surfaces, this effect is enhanced because many adhesions are even further below the nucleus. On convex surfaces, many adhesions are above the nucleus, therefore partly pulling the nucleus upwards and reducing the downward force on the nucleus. In 3D environments, adhesions are distributed equally in all directions, above and below the nucleus, greatly reducing the force on the nucleus. Blue arrows indicate the forces between focal adhesions and/or focal adhesions and the nucleus. Dark grey arrows indicate the resulting force on the nucleus. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

5.8. Outlook on molecular mechanisms of mechanobiology in 2D and 3D

Altogether, mechanobiology in 3D remains largely underexplored. The current studies indicate roles of integrins, actin-myosin, ROCK and YAP in 3D MSC differentiation like in 2D. However, there are also conflicting reports for these proteins, hinting at alternative and independent pathways that remain to be described. RhoA and actin-myosin have also been shown to influence chondrogenesis in 2D [194], but remain poorly explored in 3D. Also, some important players in 2D MSC differentiation, such as lamin A and C and MRTF/SRF, remain almost completely unstudied in 3D MSC differentiation. On a more fundamental level, studies using atomic force microscopy and magnetic/optical tweezers to exert forces on isolated nuclei [177] could investigate the precise reaction of chromatin remodeling, histone modification, and other gene regulatory processes, to these forces. With a combination of analysis of what forces nuclei experience in 3D cell cultures, such experiments could help to better understand how the 3D environment can directly regulate gene expression.

A better understanding of the role and regulation of the proteins discussed in this section and how they work together in 3D to guide MSC differentiation could greatly aid tissue engineering. Expanding the research to include chondrogenesis as well as adipo- and osteogenesis, would aid in understanding the trilineage potential of MSCs and how to guide MSCs into a specific lineage. In the next section, we will speculate how a better understanding of cellular response to 3D material properties and the underlying molecular mechanisms could help tissue engineering and where future research could be aimed towards.

5.9. Future outlook

On flat 2D substrates, a spread morphology, actin stress fiber, nuclear YAP, among others are fairly good predictors of MSC lineage commitment and osteo- or adipogenic differentiation efficiency. In 3D, however, all of these variables do not have the same predictive power of MSC differentiation as in 2D. Having such a profile of variables that predict differentiation of MSCs in 3D could greatly aid optimization of tissue engineering scaffolds. Differentiation experiments take long (typically at least 21 days), making it difficult to test many different variables and fully explore individual material properties. A profile of variables that accurately predicts MSC differentiation in 3D could greatly reduce the time and costs of experiments. This would allow for the screening of more variables and ease material optimization, before the final scaffold is properly tested for long-term MSC differentiation.

With a better understanding of the fundamental process underlying MSC mechanosensing and differentiation, smarter tissue engineering constructs could be developed. As discussed in the RhoA/ROCK section, materials that change properties over time could help to activate specific pathways at specific stages of differentiation. Other proteins, such as lamin A/C or LINC complex proteins because of their key role in gene regulation, could also have an optimal window of activation. Before such materials can sensibly be developed, a thorough understand of the timing and levels of activation is required. Lastly, most 3D mechanobiological studies of MSCs are done in hydrogels. It would be beneficial to the tissue engineering field to extend this research to other 3D tissue engineering constructs, such as electrospun, additive manufactured, or other types of scaffolds. Individual parameters are more difficult to control when creating more complex tissue engineering scaffolds with stiffer materials, making the investigation of single variables trickier. However, most parameters in these scaffolds (such as fiber size, surface roughness, pore size and shape, fiber interconnectivity, among many others) have been modified. Even though the modification of one variable is likely to change others too, tools are also available to map the different parameters. With in-depth material characterization, smart material design and the development of new materials and techniques, mechanobiology in the more complex 3D materials can surely be properly investigated.

Stiffness, stress-relaxation and degradation have been investigated in 3D hydrogels. Even though some ranges of individual parameters have been tested and optimized, in vivo osteogenesis is still poor [61]. In our opinion, before this can be improved, the fundamental understanding of 3D MSC differentiation needs to be advanced. Quantification and testing of wide ranges of combinations of these properties could greatly aid smarter tissue engineering design. On top of this, both cell volume and mechanotransduction pathways have not been widely investigated in 3D and many conflicting results still remain unresolved. A thorough investigation in the role and the predictive power of all these different proteins should be performed. The study of different combinations of scaffold properties could be combined with measurements of and interference with cell volume and mechanotransduction pathways to truly get an in depth understanding of 3D mechanobiology and differentiation of MSCs. This fundamental research should also be extended to other scaffold types such as electrospun- or additive manufactured scaffolds.

Altogether, many recent developments have shed light on mechanobiology in 3D, but many questions still remain. Even though molecular biological studies are easiest in 2D, we argue for a greater and quicker move towards 3D. Even though more challenging, all the tools and methods are available to study 3D mechanobiology. Even though the focus of many translational tissue engineering and regenerative medicine studies is on the final application, we argue for more research on the fundamental processes that orchestrate the cellular behavior in 3D. A greater fundamental understanding could speed up translational research in the long term and will be crucial in the creation of functional tissue-engineering tissues.

Data availability

Data analysis and original figures of this review article are available upon request.