Gradual conversion of cellular stress patterns into pre-stressed matrix architecture during in vitro tissue growth

Abstract

The complex arrangement of the extracellular matrix (ECM) produced by cells during tissue growth, healing and remodelling is fundamental to tissue function. In connective tissues, it is still unclear how both cells and the ECM become and remain organized over length scales much larger than the distance between neighbouring cells. While cytoskeletal forces are essential for assembly and organization of the early ECM, how these processes lead to a highly organized ECM in tissues such as osteoid is not clear. To clarify the role of cellular tension for the development of these ordered fibril architectures, we used an in vitro model system, where pre-osteoblastic cells produced ECM-rich tissue inside channels with millimetre-sized triangular cross sections in ceramic scaffolds. Our results suggest a mechanical handshake between actively contracting cells and ECM fibrils: the build-up of a long-range organization of cells and the ECM enables a gradual conversion of cell-generated tension to pre-straining the ECM fibrils, which reduces the work cells have to generate to keep mature tissue under tension.

1. Introduction

Connective tissues mainly consist of cells, collagen and other extracellular proteins such as fibronectin and proteoglycans (PGs), sometimes reinforced by mineral particles such as in bone [1–4]. The arrangement and mechanical state of these fibrous components and especially of the collagens, determine the shape, the mechanical properties and thus the mechanical function of the tissues they constitute [5–7]. Because tissues are assembled during morphogenesis, remodelled and healed throughout the life of organisms, fundamental understanding of tissue organization during growth both inspires and profits from the research on de novo tissue formation performed in the context of tissue engineering [8–10]. Despite their importance, it is still unclear how the extracellular components of a tissue become and remain organized on scales larger than individual cells.

Cells are tightly aligned with respect to their extracellular matrix (ECM) both in vivo and in vitro, suggesting that they actively control ECM organization and fibre alignment [11–15]. Mechanical forces that are generated by the contractile cytoskeletal actomyosin fibres and transmitted to the ECM through focal adhesion complexes, play an important role in this process [16–23] as well as in morphogenesis at larger length scales [24–26]. On the one hand, it is known that cellular forces are required for fibre assembly [27,28], and that these forces influence the organization, mechanical state and conformation of the ECM [21,29–32]. On the other hand, the spatial arrangement and mechanical state of the ECM influences the ability of cells to spread, migrate, proliferate and generate forces [33–36], resulting in continuous feedback between cells and the surrounding ECM [11,21,23,37]. Although it has been shown that forces can also be powerful signals for cells to communicate and synchronize over large distances [22,38,39], such long-range interactions are as yet poorly understood, being so far studied mostly in reconstituted or decellularized ECM on relatively short timescales [39–42]. The question remains therefore how the transient forces generated by cells and transmitted via the ECM act at longer time and length scales during growth and remodelling, and orchestrate the spatial organization of collagen fibres, for example in bone.

To study how transient cellular stress patterns regulate ECM architectures during the maturation of growing tissues, we used a previously described scaffold model system with millimetre-sized pores of controlled geometries [43] in which pre-osteoblasts deposited a thick ECM-rich tissue over the duration of a few weeks [43–45]. In this system, the local rates of tissue growth were observed to be proportional to the local curvature of the tissue interface. Overarching actin rings were also seen along the interfaces between the tissue and the medium [43–45] and these were found to co-localize with myosin [46], implying an important role of mechanical signalling on growth. The growth kinetics and tissue structure are severely affected in these systems when cell contractility is impaired by driving mesenchymal stem cells towards adipogenic instead of osteogenic differentiation [47]; however, a detailed understanding of the role of mechanical forces on three-dimensional tissue growth has not yet been achieved.

Because fibronectin is one of the first provisional matrix components assembled by cells, while collagen forms the most prominent ECM fibres in mature tissue, we asked here how cells exploit tension to transform provisional ECM into mature tissue with well-defined collagen architecture. The spatio-temporal sequence of events during growth was investigated by tracking the incorporation of labelled fibronectin (Fn), whereas cells and collagen-I were imaged using confocal and second harmonic microscopy. We then used laser cutting and biochemical treatments to show that cell-generated tensile stresses guide the growth process and are transferred into persistent tissue tension and ECM architecture. Finally, we assessed the role of mechanical tissue integrity for balancing cellular tension during growth by impairing collagen ECM assembly and stability using ascorbic acid (ASC) deprivation and collagenase treatment, respectively.

2. Results

2.1. Cell morphology and extracellular matrix organization

To investigate the structure of ECM-rich tissue, MC3T3-E1 pre-osteoblasts were cultured on three-dimensional hydroxyapatite (HA) scaffolds containing channel-like pores with triangular cross section of about 1 mm edge length, as described in [43] (figure 1a). The resulting tissue samples were fixed at different time points, and various visualization techniques were used to investigate the internal organization of the osteoid-like tissue consisting of osteoblast cells embedded in an ECM rich in fibronectin and collagen fibres.

Figure 1.

Cell organization during tissue growth in triangular millimetre-sized pores. (a) ECM-rich tissue was grown in triangular pores of HA scaffolds incubated in culture medium containing MC3T3-E1 cells. After fixation, the tissue was stained for actin (green) and nuclei (red) for fluorescent confocal imaging. Samples were fixed after 2 days (b) to reveal the elongated shape of the cells, which occasionally pulled out of the surface of the scaffold (white dashes) by the associated forces (c). Scaffolds fixed after 35 days of culture reveal the organization of the cells in the tissue at a later stage of growth and the apparition of an actin ring at the tissue–medium interface (pink dashes) (d). Throughout the culture period, cells at the tissue–medium interface have an elongated morphology (e), whereas cells embedded in the bulk spread in three dimensions (d, arrow). This transition in cell morphology as they become embedded in a three-dimensional environment also appears at the centre of a pore filled with tissue after 35 days of growth (f). Scale bar, 50 µm.

2.1.1. Cells have an elongated morphology at the tissue–medium interface

Tissue containing scaffolds fixed after 2 or 35 days of cell culture were stained for actin (green) and nuclei (red) and imaged with a fluorescent confocal microscope. After the first days, a few cells had settled on the surface and their actin cytoskeleton adopted an elongated morphology bridging across the scaffold corners (figure 1b,c). At later stages of growth, cells close to the substrate were still aligned along the scaffold-tissue interface, whereas cells on the tissue–medium interface aligned along the surface of the tissue to the medium and formed an overarching actin ring (figure 1d). As highlighted in figure 1e, the actin pattern along the tissue–medium interface was highly oriented, which suggests that the forces generated and experienced by the cells are highly anisotropic. In contrast, cells found deep in the corner of the pore (figure 1d) or in the pore after closure (figure 1f) were more randomly oriented with respect to each other and do not show long-range order.

2.1.2. Spatio-temporal formation of fibronectin fibres follows actin organization

The spatio-temporal organization of fibronectin within the de novo assembled ECM was assessed by adding trace amounts of fibronectin labelled with different fluorophores (either yellow or red) at two time points for 1 day each using well-established protocols [12,48–50], and by observing how cells incorporate it into their own ECM (see Material and methods). Figure 2 shows the evolution of the actin and the fibronectin networks observed in tissues fixed after different durations of culture. It was verified that the presence of labelled fibronectin did not significantly affect the kinetics of tissue deposition (electronic supplementary material, figure S1). First, newly growing tissue was exposed to yellow-labelled fibronectin between days 4 and 5 and directly fixed for actin immunostaining at the end of exposure on day 5. The actin as well as the Fn-546 signal intensity was highest at the tissue–medium interface (figure 2a). Although the fibronectin fibres were not as highly organized as the actin stress-fibres, there is a general coalignment.

Figure 2.

Actin and fibronectin organization during tissue growth. Fluorescent-labelled fibronectin was added to the culture medium 4 and 11 days after seeding the cells (yellow or red, respectively). Tissue was then cultured further for different periods of time, stained for actin (green) and imaged by fluorescent confocal microscopy (a). The labelled fibronectin was taken up by the cells and incorporated into the fibrous ECM they produced, the overall organization of which is similar to the actin pattern. (b) Intensity profiles averaged over seven lines per corner for three corners (one per pore) were derived from the centre of the pore to the edge of the scaffold for each channel and each time point. The evolution of the position of the peaks of intensity shows that fibronectin is integrated locally in the tissue (scale bar, 100 µm).

Second, to explore how deposited fibronectin matrix evolves as the tissue is growing, the cells were then cultured for six additional days in standard culture medium without labelled fibronectin in solution. When fixed at day 10, the cells had completely overgrown the labelled fibronectin matrix integrated between days 4 and 6. As a consequence, very little signal could be observed close to the tissue–medium interface, and the brighter yellow fibronectin signal no longer follows the actin ring.

Third, to investigate the deposition of new relative to existing matrix, some samples received additional red-labelled fibronectin after 11 days of regular culture. In the scaffold fixed 1 day later (on day 12), the labelled fibronectin has been incorporated into fibres throughout the tissue, although the fluorescent signal was slightly higher along the tissue–medium interface. Further samples were fixed and imaged after days 17, 28 and 35 (figure 2a). In all of these pores, the two fibronectin layers could still be seen distinctly, although the strong actin signal indicating the tissue–medium interface moved closer to the centre of the pore with growth.

To analyse the spatio-temporal evolution of the actin and fibronectin networks, we measured intensity profiles from the centre of the pore to the edge of the scaffold for each channel and each time point (figure 2b). In every sample, the maximum intensity of the actin signal was closest to the centre of the pore and corresponds to the actin ring along the tissue–medium interface. The signal of the first fibronectin added to the medium (yellow) at day 4 initially coincided with the actin peak on the first day after addition, and had moved only slightly towards the centre of the pore at day 11, but not as far as the actin ring. This indicates that the ECM deposited at an earlier time point does not move together with the tissue–medium interface. The signal from the second labelled fibronectin added at day 11 (red) followed the same trend, but is shifted in time and space owing to its later incorporation. Together, these findings show that new tissue is deposited layer-by-layer along the tissue–medium interface as cells continue to assemble new fibronectin fibrils on top of the existing tissue.

To measure the degree of fibre alignment between the incorporated fibronectin and the actin, we performed a fast Fourier transform (FFT) analysis of the different confocal images. The degree of fibre alignment was then measured by the local difference in orientation angles (Δϑ) between the two images. Figure 3 shows two example orientation analyses for tissues fixed at day 12 (figure 3a–e) and day 28 (figure 3f–j). Figure 3a,f shows the confocal images of tissues stained for actin. The differences in matrix fibre orientation with respect to the local actin orientation are shown for Fn546 (b,g), Fn633 (c,h) and collagen (see §2.4) (d,i) as well as their associated histograms (e,j). For a full set of histograms of all samples measured in figure 2, see electronic supplementary material, figure S4. An average degree of orientation can be calculated by simply taking the average cosine of the local difference in orientation angles (Δϑ). Perfect alignment between F-actin and ECM fibres is represented by a value of 1, randomly aligned fibres will have a value of 0.626. Electronic supplementary material, table S1 shows orientation parameters for all samples, and highlights that ECM components become aligned with the local orientation of the actin stress fibres, although for later time points, the data indicate a lower actin–collagen alignment. This might be partially owing to the lower actin signal obtained from imaging thick tissues.

Figure 3.

Degree of fibre alignment. The local orientations of actin, Fn546, Fn633 and collagen (SHG) were measured using FFT analysis of the confocal images. Shown in (a) and (f) are confocal images of samples stained for actin at two different time points (days 12 and 28) (scale bar, 100 µm). The degree of fibre alignment is measured by differences in orientation angles (Δϑ) for actin–Fn546 (b,g), actin–Fn633 (c,h) and actin–collagen (d,i) as well as their associated histograms (e,j). (For histograms of the other experimental images, see electronic supplementary material, figure S4 and table S1.)

2.1.3. Spatio-temporal assembly of collagen fibrils follows cell and fibronectin deposition with a slight delay

The ECM produced by MC3T3-E1 pre-osteoblasts not only contains fibronectin, but also collagen, whose thick fibrils can be imaged with second harmonic generation (SHG) microscopy [51], as tissues grown for 14 days contained collagen fibres that are ‘mature’ enough for SHG (figure 4a). SHG is specific for fibrillar collagen made of non-centrosymmetric units forming triple helical superstructures [52,53]. Such fibres first appeared deep in the tissue, and their orientation follows the actin and fibronectin patterns along the interfaces (figure 3) but appears to be less aligned deeper down in the bulk of the tissue. After 35 days of growth (figure 4f), the collagen fibres were thicker and essentially located in ‘mature’ tissue, i.e. towards the corners, whereas the region along the tissue–medium interface was empty of visible collagen fibrils. The bright spots lining the edge of the scaffold are artefacts from the HA. Note that while cell and collagen organization highly depends on the geometry of the pore in which the tissue has been produced, both mostly co-align with each other (figure 3 and electronic supplementary material, figure S2).

Figure 4.

Perturbations of tissue mechanics. Control: ECM-rich tissues grown in the control conditions were fixed after 14 days (a) and 35 days (f) of culture, stained for actin (green) and nuclei (red) and imaged by fluorescent confocal microscopy using SHG signal to visualize collagen fibrils (white). Cell contractility: samples were treated with blebbistatin, either temporarily with 20 µM for 24 h after 14 days of culture (red band, m) or continuously with 2 µM over the whole culture time. In both cases, impairing cell contractility both affected the organization of the tissue (b,c,g,h) and the kinetics of tissue growth (l,m). Growth kinetics of the samples continuously treated with blebbistatin were significantly different to the control (two-way ANOVA, p < 0.001), and significantly different to the control (two-way ANOVA, p < 0.05) for day 7 and from day 18 till day 25 for the pulse treatment of blebbistatin. When treated for 24 h, cells temporarily lost their elongated shape and sharp organization (b,c) and the formation of thick collagen fibres seemed to be impaired after 35 days of growth (h). When treated continuously with a low dose of blebbistatin, global tissue production and organization was impaired compared with the control conditions (g). Extracellular matrix: samples were either treated temporarily with 0.1% collagenase for 2 h after 14 days of culture (red line, n) or continuously deprived with ascorbic acid (ASC) over the whole culture time. In both cases, impairing the stability or formation of collagen prevented the tissue from being tightly anchored to the scaffold as shown by the large holes without cells (d,e). After 35 days of growth, the formation of thick collagen fibres is rare in the treated samples compared with the control condition and a hole in the collagen network suggests that the ECM produced in such conditions cannot bear the tension imposed by the surrounding cells (i,j, arrow). The collagenase treatment suddenly increased the projected tissue area (n), whereas the deprivation of ASC did not significantly (two-way ANOVA) alter tissue growth (o). (Scale bar, 100 µm.) Error bars in (l–o) indicate the standard deviation, (n = 9) for each dataset.

2.2. Tensile stresses in the tissue

Because the presence of an actin band lining the tissue–medium interface suggests an important role of cell contractility in tissue formation, the following experiments were designed to study the role of mechanics in the control of tissue patterning and stability. For each treatment, the effect was assessed qualitatively by observing structural changes with immunofluorescent techniques and quantitatively by comparing the evolution of the overall growth rate of the projected tissue area (PTA) with control kinetics curves (figure 4).

2.2.1. Tensile stress is released by laser cutting

The presence of internal mechanical stress in the tissue was first directly visualized by performing laser dissection of the tissue–medium interface and by observing the immediate relaxation of stored mechanical stress in the cells and ECM. Figure 4k (electronic supplementary material, figure S2) shows a laser cut performed on living tissue grown in a small triangular pore. Although the section was linear and perpendicular to the tissue–medium interface, the resulting ‘wound’ presents a large opening angle, which appeared already after the first layers of cells were damaged. As suggested by the actin patterns, higher forces are generated at the tissue–medium interface compared to the interior, because cutting deeper into the tissue did not further enlarge the opening of the actin ring (electronic supplementary material, figure S3).

2.2.2. Inhibiting cell contractility impairs both cell growth and extracellular matrix formation

We next asked if inhibition of cell contractility impacts growth kinetics and tissue structure. For this purpose, a set of samples grown for 14 days was temporarily incubated in medium containing 20 µM blebbistatin to inhibit myosin activity for 24 h. Another set of scaffolds was continuously treated throughout the tissue culture with medium containing a lower concentration of blebbistatin (2 µM), starting from day 4. The lower concentration was chosen such that cells generate less contractile tension, but still continue to proliferate and deposit tissue.

These experiments revealed that impairing cell contractility during tissue culture affects the overall kinetics of growth (figure 4l,m). Indeed, tissue deposition was reduced during the temporary blebbistatin treatment with the PTA dropping from 0.130 ± 0.017 to 0.119 ± 0.017 mm² 2 h after starting the pulse of blebbistatin. This indicates that while the tissue relaxes, it moves away from the centre of the pore thus causing a reduction in tissue volume. Growth resumed when the tissue was put back in fresh regular medium, returning to a rate of 0.012 ± 0.0033 mm² day⁻¹ after 7 days (figure 4m) which is similar to the growth rate of 0.011 ± 0.0025 mm² day⁻¹ before the inhibition pulse. During continuous blebbistatin treatment, growth persisted with a slower initial rate (0.0047 ± 0.004 versus 0.020 ± 0.014 mm² day⁻¹ for the control after 7 days), and similar rates at later time points compared with the control when it reached a plateau (figure 4l).

Samples fixed right after the temporary blebbistatin treatment indicated reduced force generation characterized by a blurred actin pattern and a roughened tissue–medium interface when compared with the control conditions (figure 4b). After reverting back to regular culture medium for 24 h, cells recovered their spindle-like shape, and a smooth tissue–medium interface formed again on which tissue growth resumed (figure 4c). At the end of the experiment (D35), no gaps could be observed, and the actin and collagen fibres were still organized as in the control situation (figure 4h). However, a weaker SHG signal suggests that the blebbistatin treatment also delayed the maturation of collagen fibres.

Although the growth was initially slower, tissue formed under continuous blebbistatin treatment and showed a similar organization as in the control conditions. The smooth tissue–medium interface suggests that cells still formed a continuous contractile structure, but the weak SHG signal observed after 35 days of growth revealed that fewer collagen fibres were produced compared with normal conditions, suggesting that tissue growth and maturation requires proper cell contractility (figure 4g).

2.3. Impairing mechanical stability of the extracellular matrix network disturbs tissue growth and organization

2.3.1. Collagenase treatment leads to tissue reorganization by the contractile cells

To study how the cells in matured tissue respond to a selective disruption of the ECM, we enzymatically degraded the collagen ECM with collagenase. A set of samples treated at day 14 with 0.1% collagenase in PBS revealed slight changes in tissue properties (figure 4d,i). Samples fixed and stained for actin directly after the treatment did not show organizational alterations of the actin network. However, in tissue fixed after 4 days of recovery in fresh regular medium, the tissue–medium interface was no longer smooth, cells were no longer densely packed, and actin was less organized (figure 4d). Three weeks after the treatment, the tissue recovered a structure similar to control samples, but the collagen signal was found to be weaker (figure 4i).

Tissue growth graphs showed a sudden increase of the PTA right after the collagenase treatment, increasing by 0.002 mm² after 30 min and 0.02 mm² after 2 h (figure 4n). The short timescale involved is characteristic of tissue disruption induced by the contractile actin ring pulling on a weakened ECM rather than by tissue formation. Four days after the collagenase treatment, PTA was still higher than in the control sample. A control experiment in pure PBS ensured that the changes in tissue and cell behaviour were not an effect of culture medium deprivation during the treatment (data not shown). Altogether, these results clearly show the importance of an intact collagen ECM to counterbalance the cell-generated tension and to stabilize the growing tissue.

2.3.2. Ascorbic acid deprivation leads to formation of unstable tissue

To further investigate the importance of collagen integrity in tissue organization, tissue growth experiments were performed entirely without additional ASC in the medium, as ASC is known to be necessary for cells to assemble collagen into fibrils [54]. Immunofluorescent images show that cells deprived of ASC organized as in the control conditions, but the tissue they produced contained gap regions devoid of cells along the scaffold figure 4e. After 35 days of culture, cells deprived of ASC produced less collagen compared with the control, but the few visible thin fibres were well organized and showed an orientation similar to the actin fibres (figure 4j). Note that a larger gap also appeared in the collagen pattern (figure 4j arrow). The PTA measured in samples grown without ASC was slightly less than in the control at the beginning of the experiment although this was not significant, but increased faster and became almost equal to the control after two weeks of culture (figure 4o). These results indicate that if the ability of cells to assemble collagen fibres is impaired, tissue still forms but is less dense and contains defects owing to a lack of a mechanically stable collagen network.

3. Discussion

The goal of this study was to understand the role of mechanics in the long-range organization of cells and ECM during the formation of three-dimensional tissues. Specifically, we were interested in how collective cellular tension develops during tissue formation and is transferred into a permanent matrix with pre-stressed collagen fibres. Our in vitro system to model bone-like tissue growth using pre-osteoblastic cells in scaffolds with pores of defined geometry revealed the following as summarized in figure 4: (i) the cells have an elongated morphology at the tissue–medium interface (figure 1), (ii) spatio-temporal formation of fibronectin fibres follows actin organization (figure 2), (iii) spatio-temporal assembly of collagen fibrils follows cell and fibronectin deposition (figure 4a,f), (iv) inhibiting contractility impairs both cell growth and ECM formation (figure 4b,c,g,h,l,m), (v) impairing the mechanical stability of the ECM network disturbs tissue growth and organization (figure 4d,e,i,j,n,o) and (vi) tensile stress is partially released by laser cutting (figure 4k).

Figure 5a highlights the sequence of cell and ECM deposition during tissue growth, schematically summarized in figure 5b. The highest concentration of actin stress fibres (green) appears in the youngest tissue at the tissue–medium interface (see also figure 2). Cells adopt an elongated chord-like appearance, as we reported earlier [44], owing to the balance of contractile actomyosin forces and substrate geometry [34,55]. Moreover, the change of cellular behaviour around pore closure compares with that described in a study showing the relevance of the geometrical model in smaller pores [56] (figure 1f). Also, the continuous actin band forming parallel to the tissue–medium interface is reminiscent of the actin ring observed in other contexts of tissue growth, such as wound healing [57].

Figure 5.

Structural and mechanical transfer from the active cells to the passive ECM. (a) Tissue grown for 35 days with addition of yellow- and red-labelled fibronectin at days 4 and 10, respectively, was fixed, stained for actin (green) and imaged with fluorescent and SHG microscopy to visualize actin, fibronectin and collagen fibres simultaneously. (b) The internal structure of the tissue is well described by a geometrical model based on the sequential assembly of fibrous cellular elements, the organization of which is progressively transferred to fibrous extracellular elements synthetized by the cells. (c) Similarly, we propose a model of mechanical handshake between the cells and their ECM that enables the progressive transfer of the tension actively generated by the contractile actin network into tension born in the passive ECM as the tissue matures (scale bar, 100 µm).

In figure 5a, the clear concentric yellow and red bands indicate the locations at which the yellow- and red-labelled fibronectins introduced into the culture medium at days 4 and 11, respectively, were assembled into fibres, whereas the green actin band indicates the presence of a contractile layer of cells on top of the pre-existing ECM. Small amounts of labelled fibronectin can also be observed in the older tissue (figure 2a). This incorporation can be both owing to active remodelling of the existing matrix and to passive adsorption of fibronectin onto collagen fibres [27]. Nevertheless, the intensity profiles (figure 2b) confirm that labelled fibronectin is mostly incorporated in the vicinity of the actin ring, along the tissue–medium interface at the time of addition. This suggests that adherent cells rapidly start to assemble extracellular fibronectin, stabilize their focal adhesions, generate traction forces [58], and then spin an extensible fibronectin network [59] that follows the alignment of the actin network using their contractile cytoskeleton as evident in figure 2a [60–62].

During later phases of tissue development, mature collagen fibres appeared initially co-aligned with the cells and the fibronectin ECM parallel to the surface, but showed a more radial orientation deeper in the tissue. The orientation of the collagen fibrils follows the actin pattern, regardless of the initial geometry of the substrate (electronic supplementary material, figure S2). SHG images suggest that the first mature collagen fibrils form during the second week of growth (figure 4a). Although SHG cannot detect early collagen fibrils, our data suggest that the assembly and maturation of collagen into fibres does not occur as quickly as fibronectin, but rather at later time points following cell proliferation, adhesion and contraction [63–65]. The schematic summary of the sequential deposition of tissue components during growth given in figure 5b, is reminiscent of the geometrical model proposed to describe the organization of tensile cells on a curved surface [44], and indicates a transfer of this same organization to the fibrous components of the ECM as previously suggested [45]. It has been shown that collagen can self-organize into highly aligned, chiral and cholesteric liquid crystal-like structures in a completely cell-free system [66–69], while it can only be assembled in cell culture in the presence of fibronectin [65]. Our results on the co-alignment of collagen with previously deposited cells and fibronectin, and on the importance of cellular tension for ECM organization, emphasize the importance of active cellular control over such self-organizing processes.

These results suggest a mechanical model for the process by which active tension generated by actin stress fibres is being transferred to passive tension within the ECM, as outlined in figure 5c. The youngest tissue along the tissue–medium interface consists of very little ECM, and contains the largest amount of actin stress fibres indicating circumferential tension in the tissue (green arrow in figure 5c). This is confirmed by the significant triangular opening of a laser cut applied to this region of the tissue (figure 4k), which raises the question whether the entire circumferential tensile load is carried by the contracting cells, or whether part of the load is taken up by the ECM. To address this, we applied a blebbistatin pulse to the cell culture at day 14 (figure 4b,c,h). Figure 4m shows that the application of the blebbistatin pulse (red vertical line) only leads to a minimal relaxation of the tissue indicating that a substantial fraction of the circumferential tension is carried by the ECM (orange arrow in figure 5d). Further tissue growth is somewhat retarded owing to the temporary inhibition of the actin–myosin complex. When tissue treated with blebbistatin is cut with a UV laser, a smaller but still appreciable opening is observed confirming the circumferential tension in the collagen matrix (electronic supplementary material, figure S3b). A treatment with collagenase after 14 days of culture leads to a partial closure of the pore in the tissue, sometimes accompanied by a detachment from the HA channel walls. This shows that ECM is not only carrying circumferential loads but is also stabilizing the central pore by radial tensions connecting it to the walls of the pore (grey arrow in figure 5c).

Our model of transfer from cellular to ECM tension during tissue growth is in agreement with cells stretching fibronectin molecules during fibre assembly [12,48–50], transient cellular contractility stored in soft and highly compliant fibronectin meshwork [12,48–50], and mechanical stress transferred from the fibronectin matrix to the more rigid collagen network appearing after 2–3 days of culture [65]. Moreover, our observation of the sequential occurrence of cellular tension, fibronectin meshwork and finally collagen fibres supports the interpretation that the fibronectin ECM mediates the transfer of transient cytoskeletal tension into permanent ECM pre-stress in the mature collagen network. An understanding of how cells pre-stress the ECM, is fundamental not only for understanding processes of morphogenesis [24], but is important in understanding diseases such as cancer and heart disease that may be promoted by compromised tensional homeostasis [23]. Furthermore, our in vitro observations may also help shed light on the enigmatic behaviour of myofibroblasts in wound healing, that are also strongly influenced by tension in the tissue [70]. Although a different type of cell to those studied here, myofibroblasts also create contractile actin–myosin structures that apply tension to the wound boundary, produce and align collagen to fill the wound and then transfer tension to the aligned ECM to allow for regeneration of tissue function [71].

Taken together, our findings suggest that cell contractility first determines both the internal organization and the mechanical tension of the growing tissue, which are then stored as pre-strain in the ECM fibrils behind the growth front. Recent studies discovered how stresses up to several megapascals in collagen can develop purely owing to changes in osmotic pressure in ranges that can occur in the presence of PGs in the ECM [72]. Although we have not yet investigated the role of PGs on the development of ECM tension in our in vitro system, it presents a plausible mechanism by which ECM internal stresses can be pre-programmed by cells. By controlling first the spatial arrangement and pre-tensioning of ECM fibrils followed by local chemical changes that induce osmotic tension in the fibrils, cells are able to modulate ECM stresses in mature tissues. Through this mechanical handshake between actively contracting cells and pre-stressed ECM fibrils taking up passive load, internal stresses may be inscribed permanently into a tissue without the requirement of energy supply in the form of ATP, necessary for actin contraction.

4. Material and methods

4.1. Growing tissue in control conditions

The tissue investigated in this study was deposited by bone cells seeded on 2 mm thick HA scaffolds containing straight-sided pores of controlled geometries. The substrates were produced by casting HA slurry in wax moulds made by rapid prototyping. More details about scaffold fabrication can be found in previous works [43,44,73]. Murine pre-osteoblastic cells MC3T3-E1 (provided by the Ludwig Boltzmann Institute of Osteology, Vienna, Austria) were seeded with a density of 10⁵ cells cm⁻² on the top of the HA scaffolds and cultured in alpha-MEM (Sigma-Aldrich) supplemented with 10% fetal calf serum (PAA Laboratories), 0.1% ASC (Sigma-Aldrich) and 0.1% gentamicin (Sigma-Aldrich). Cells in culture were incubated at 37°C in a humidified atmosphere with 5% CO₂, and the culture medium was changed every 3–4 days over the time of the experiment.

4.2. Visualizing tissue organization

4.2.1. Cells

After the culture time of interest, the scaffolds were washed in phosphate-buffered saline (PBS). The tissue was then fixed in 4% paraformaldehyde for 10 min at room temperature, permeabilized in 1% Triton X100 (Sigma-Aldrich) at 4°C overnight and washed again in PBS. The cytoskeletal actin stress fibres were stained in a solution of Alexa-phalloidin fluorescein isothiocyanate (488 nm; Invitrogen, Molecular Probes) at a concentration of 3 × 10⁻⁷ M for 90 min. The samples were then incubated for 5 min in TO-PRO 3 692-661 (Invitrogen, Molecular Probes) at 3 × 10⁻⁶ M for staining cell nuclei. After a last wash in PBS, the tissue was imaged with a fluorescent confocal microscope (Leica TCS SP5), using 488 and 633 nm lasers and a 25× water immersion objective to visualize actin and nuclei, respectively.

4.2.2. Fibronectin

Because stains for fibronectin could not diffuse homogeneously throughout fixed thick tissues, cells were provided fluorescent-labelled fibronectin in the culture medium to be incorporated into the fibronectin fibres they synthesize and assemble [49]. The fibronectin used in this experiment was isolated from human plasma (Swiss Red Cross) using gelatin–sepharose chromatography. For fluorescent labelling on free amines, fibronectin was incubated with a 60-fold molar excess of Alexa 546 or 633 succinimidyl ester (Molecular Probes) for 1 h, and dialysed overnight in PBS to remove free dye, following published protocols [49]. The final protein concentration and labelling ratio were quantified in a Nanodrop UV–Vis spectrophotometer (Thermo Scientific).

In the regular supplemented medium described above, fibronectin (labelled and unlabelled) was added at a total concentration of 25 µg ml⁻¹ (20 µg ml⁻¹ unlabelled and 5 µg ml⁻¹ labelled) with 75% excess of unlabelled fibronectin. Different scaffolds received this particular medium for 3 days after various times of culture: Alexa-546-labelled fibronectin was added at day 4, and Alexa-633-labelled fibronectin was added at day 11.

In those particular samples, nuclei were not stained. Images were taken using a confocal laser scanning microscope (Leica TCS SP5) with a 25× water immersion objective (NA = 0.95). Excitation was performed via an Ar laser at 488 nm for actin and via HeNe lasers at 546/633 nm for fibronectin. Emitted light from the samples was detected using internal photomultiplier detectors (PMT) set for actin and fibronectin signal according to the emission spectra. Laser power and NDD parameters were optimized for three-dimensional image stack recording by z-value-dependent linear compensation for imaging each of the different pores.

4.2.3. Collagen

Collagen fibrils were imaged by SHG microscopy. Second-harmonic was generated using a Spectra Physics Ti : sapphire laser (Mai Tai HP) with 100 fs pulse width at 80 MHz and wavelength of 910 nm and signal emitted from collagen fibrils was detected around 455 nm (two times the energy of excitation).

4.3. Measurement of extracellular matrix fibre alignment

An FFT analysis of confocal images was used to quantify fibre alignment between ECM components and actin. For a given time point, the projected z-stack from each channel (i.e. actin, yellow and red fibronectin, and collagen) was subdivided into 32 pixel × 32 pixel = 1024 pixel² equally sized blocks. In each of these blocks, an FFT algorithm is applied, and an ellipse is fitted to the real part of the Fourier transform with the orientation of the major axis being the anisotropic direction in real space +90°. The degree of fibre alignment is then measured by difference in orientation angle (Δϑ) between the ECM angles (fibronectin and collagen) and the angles measured for actin in each of the corresponding blocks.

4.4. Intensity profiles

To quantify the relative deposition of actin and yellow 546 and red 633 fibronectin, intensity profiles were derived from a series of images. For each corner imaged, the actin image was used to manually define a region of interest, within which seven lines going from the centre of the pore to the edge of the scaffold were drawn following consistent rules. For each channel, intensity profiles were computed along the seven lines, the distance to the interface was normalized by the length of the line and the intensity by the maximum intensity, and averaged using a binning method (bin = 0.01). For each time point, the intensity profiles of three independent corners were then averaged.

4.5. Following growth kinetics

Projection images of the tissue grown in the pores were taken twice a week by phase contrast microscopy (Nikon Digital sight DS 2Mv). Because the amount of tissue produced in the three-dimensional pores could not be determined on such images, the evolution of the PTA, i.e. the difference between the initial and the current pore areas (PA), was used to estimate tissue growth kinetics.

4.1

PAs were measured manually by using the program NIS-Elements v. 3.10 (Nikon). For each condition of culture, kinetics was followed in three scaffolds containing three pores, i.e. n = 9. Statistics of the PTA curves were analysed using SigmaPlot v. 12.3.

4.6. Perturbing tissue mechanics

In order to investigate the role of mechanics in the stability of the tissue at the macroscopic level, the following treatments were performed on living tissues cultured for 35 days.

4.6.1. Cell contractility

At day 14, samples were incubated 24 h in supplemented culture medium containing the myosin II inhibitor blebbistatin at 20 µM. At that point, some of them were fixed, whereas others were put back in regular supplemented medium for an additional 24 h before fixation. The effect of a continuous partial inhibition of cell contractility was investigated with another set of samples cultured in medium containing additional blebbistatin at 2 µM from day 4 up till the end of the experiment.

4.6.2. Extracellular matrix stability

At day 14, samples were incubated for 30 min in 0.1% collagenase (Sigma-Aldrich) diluted in PBS to disrupt the ECM network. Some of them were then fixed, whereas the others were put back in regular culture conditions for 24 h before fixation. The quality of the collagen produced by the cells has also been impaired continuously by growing cells in the absence of ASC and scaffolds were fixed at different time points for similar structural investigations. After each treatment, actin, nuclei and collagen were observed following the protocol described above.

4.6.3. Tension release

Tissue organization was also disturbed mechanically by micro-sectioning the living tissue under a bright-field microscope equipped with a laser (P.A.L.M. Microbeam, Zeiss). Linear sections were cut perpendicular to the tissue–medium interface and with a length depending on the local tissue thickness. The changes in tissue shape were simultaneously imaged under the microscope (Hitachi HV D30).

Authors' contributions

C.M.B., P.K. and J.W.C.D. designed the study, C.M.B., S.E., and V.G. performed cell culture experiments, and P.K. prepared the labelled fibronectin. C.M.B., V.G., S.E., P.J. and A.P., imaged the tissues, C.M.B. and S.E. analysed the images and C.M.B., P.K., P.F., V.V. and J.W.C.D. interpreted the data. C.M.B., P.K. and J.W.C.D. wrote the manuscript, and all authors edited the manuscript and gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

We acknowledge funding from the Leibniz prize of PF running under DFG contract number FR2190/4-1. C.M.B. and P.J. were members of the Berlin-Brandenburg School for Regenerative Therapies (GSC 203). Parts of the research leading to these results have received funding by the EU Seventh Framework Programme (FP7/2007–2013) under grant agreement no. 327065 (P.K.) and from the Swiss National Science Foundation (SNF Nr. CR32I3-156931/1) to V.V.