COLLAGEN I MATRIX CONTRIBUTES TO DETERMINATION OF ADULT HUMAN STEM CELL LINEAGE VIA DIFFERENTIAL, STRUCTURAL CONFORMATION-SPECIFIC ELICITATION OF CELLULAR STRESS RESPONSE.

Abstract

Previously, we reported that the conformational transition of collagen I matrix plays, along with differentiation stimuli, a regulatory role in determination of differentiation lineage of bone marrow stromal sells via distinct signaling pathways specific for the structural state of the matrix. The present study addresses mechanisms underlying differential structural conformation-specific effects of collagen matrices on differentiation into diverse lineages. The results obtained suggest that the pivotal player in the observed matrix conformation-mediated regulation is a differential cellular stress response elicited by the exposure to native but not to denatured collagen I matrix. The stress causing such a response appears to be generated by matrix contraction and mediated by α₂β₁ integrins engaged on native but not on denatured collagen I matrix. The principal facet of the observed phenomenon is not the nature of a stress but general stress response: when cells on denatured collagen I matrix are subjected to thermal stress, osteogenic pathway shifts to that seen on native collagen I matrix. Importantly, cellular stress response might be commonly involved in determination of differentiation lineage. Indeed, distinct components of cellular stress response machinery appear to regulate differentiation into diverse lineages. Thus, augmentation of Hsp90 levels enables the operation of efficient α₁β₁/α₂β₁ integrin-driven ERK activation pathways hence facilitating osteogenesis and suppressing adipogenesis, whereas myogenesis of satellite stem cells appears to be promoted by native collagen I matrix-elicited activation and nuclear translocation of another stress response component, β-catenin, shown to be essential for skeletal myogenesis, and chondrogenesis may involve stress-mediated elevation of yet another stress response constituent, Hsp70, shown to be an interactive partner of the chondrogenic transcription factor SOX9. The proposed concept of the integral role of cellular stress response in tissue generation and maintenance suggests new therapeutic approaches and indicates novel tissue engineering strategies.

INTRODUCTION.

Previously, we reported that the progression of human bone marrow stromal cells into osteogenic and adipogenic lineages is differentially regulated by the structural conformation of collagen I matrix through distinct signaling pathways specific for each structural state of the matrix (Mauney et al., 2009). Thus, on native collagen I matrix adipogenic differentiation proceeds very inefficiently and is p38-independent, whereas on its denatured counterpart, an efficient adipogenesis is primarily regulated by p38 kinase (Mauney et al., 2009). Inversely, osteogenic differentiation occurs efficiently on native, but not on denatured collagen I matrix (Mauney et al., 2009). Osteogenesis of bone marrow stromal cells on collagen I matrices in both structural conformations is fully dependent on ERK activity (Mauney et al., 2009). However, whereas on native collagen I matrix osteogenic differentiation is Hsp90-dependent, on denatured collagen I matrix it occurs, despite the potential availability of Hsp90-dependent pathway, only in an Hsp90-independent manner (Mauney et al., 2009).

Our previous study (Mauney et al., 2009) suggested that the involvement of Hsp90 occurs at the level of Raf-1, an important and essential link in several ERK-activating cascades wherein Hsp90 is crucially required for Raf-1 activation (Cutforth et al., 1994; van der Straten et al., 1997). On native collagen I matrix, ERK activation is driven by the engagement of triple helix-specific α₁β₁ and α₂β₁ integrins with corresponding binding sites on the matrix and can occur only in a Raf-1, and Hsp90, -dependent manner (Xu et al., 2000; Egan et al., 1993; Schlaepfer et al., 1996; Takeuchi et al., 1997; Wary et al., 1996;1998; Gullberg 2003). In contrast, on denatured collagen I matrix, ERK activation is driven by the engagement of α_Vβ₃ integrins with “cryptic” binding sites which are obscured within triple helical structure of native collagen I, but exposed upon its denaturation (Davis, 1992, Wary et al., 1996; 1998; Blanco-Aparichio et al., 1999; Kaneki et al., 1999; Saxena et al., 1999; Franklin et al., 2000; Short et al., 2000; Hagemann et al., 2001; Gomez et al., 2002; Salasznyk et al., 2004; Mittlestadt et al., 2005; Noon et al., 2005; Rucci et al., 2005; Tapinos et al., 2005; Goessler et al., 2006; Wen-Sheng et al., 2006). α_Vβ₃ integrin-initiated ERK activation can potentially proceed through both Raf- and Hsp90-dependent and –independent pathways, but only the latter was observed (Mauney et al., 2009).

Our earlier study (Mauney et al., 2009) suggested a possible explanation for the differential involvement of Hsp90 in ERK activation and osteogenesis of bone marrow stromal cells on native and denatured collagen I matrices, namely that Hsp90 dependency or -independency reflects differential levels of Raf-1 available in cells on native and denatured collagen I matrices. On native collagen I matrix, the engagement of α₂β₁ integrin leads to activation of protein phosphatase pp2A (Yamagishi et al., 2004; Chetoui et al., 2005) which facilitates the release of Raf-1 sequestered by 14-3-3 proteins and makes it available for interaction with and activation by Ras (Sanders et al., 2004; Abraham et al., 2000). This mechanism is absent in cells on denatured collagen I matrix, therefore, according to this explanation, Raf-1 levels could be insufficient to support ERK activation, and, as a result, it would proceed in a Raf-independent and, consequently, a Hsp90-independent manner. However, there is an alternative, more easily testable, interpretation of Hsp90-independency on denatured collagen I matrix. It presumes that in cells on denatured collagen I matrix there are sufficient levels of Raf-1 to operate a Raf-dependent pathway of ERK activation, but, in contrast to cells on native collagen I matrix, levels of Hsp90 are low and insufficient for Raf activation. Consequently, of potentially available Raf-dependent and Raf-independent pathways of ERK activation only the latter is utilized. The central premise of this interpretation is that that Hsp90 is differentially expressed in cells on native and denatured collagen I matrices. The present study was designed to evaluate this possibility.

RESULTS.

Exposure of bone marrow stromal cells to native collagen I matrix elicits cellular stress response.

As described in “Experimental Procedures” section, the present study employed bone marrow stromal cells (BMSCs) obtained from bone marrow aspirates. Differentiation potentials of these cell populations for both osteogenic and adipogenic lineages have been verified both in vitro and in vivo (Mauney et al., 2006; 2007). For terminological consistency with the majority of publications on adult stem cells it should be emphasized that these differentiation-competent cells are usually referred to as adult mesenmchymal stem cells, commonly defined as adult bone marrow-derived nonhemopoietic stromal cells capable of differentiation into various lineages. However, as a reflection of cells’ origin, we would refer to them as BMSCs throughout this study. As was mentioned above, the observations that osteogenic differentiation of bone marrow stromal cells proceeds via Hsp90-dependent pathways on native collagen I matrix, but in a Hsp90-independent mode on denatured collagen I matrix could be explained by a possibility that Hsp90 is expressed in cells in a differential matrix conformation-specific manner and is present at inconsequentially low, in terms of ERK activation and osteogenesis, levels in cells on collagen I matrix in a denatured conformation but at higher, potentially regulatory levels on its native counterpart. Hsp90 is one of the major stress response proteins and elevation of its levels can be regarded as a manifestation of cellular stress response. To assess a possibility of the involvement of stress response mechanisms in the observed phenomenon of differential, matrix conformation-specific, dependency of osteogenesis on Hsp90, we analyzed levels of Hsp90, as well as of other major stress response markers, in bone marrow stromal cells maintained on either native or denatured collagen I matrices.

As shown in Figure 1, levels of all measured major heat shock proteins, namely Hsp27, Hsp70, Hsp90 and Hsp110 are significantly elevated in bone marrow stromal cells maintained on native collagen I matrix in comparison with the levels of the same proteins seen in cells maintained on denatured collagen I matrix. These observations, therefore, are consistent with the notion that the exposure of cells to native collagen I matrix constitutes stress and elicits a cellular stress response.

Figure 1. Exposure of bone marrow stromal cells to native collagen I matrix substantially increases levels of major stress response proteins.

Hsp: heat shock protein; DC: denatured collagen I matrix; NC: native collagen I matrix. Expression levels on native collagen I matrices designated as “100”. Levels of each particular Hsp on denatured collagen I matrix are shown in relation to its expression on native collagen I matrix. [*= p<0.05], significantly different in comparison to levels of the same Hsp on DC.

Decline of osteogenic potential of bone marrow stromal cells during the expansion on native collagen I matrix is a function of the exposure to native collagen I and coincides with the attenuation of Hsp90 levels.

A concept that the exposure of cells to native collagen I matrix constitutes a stress and elicits cellular stress response is consistent with and provides a plausible explanation for a set of observations made during the expansion of bone marrow stromal cells on native collagen I matrix. We have previously reported that the osteogenic (and adipogenic) potential of mesenchymal stem cells is retained during expansion on denatured collagen I while rapidly declining on tissue culture plastic (Mauney et al., 2004; 2005; 2006). In the view of our current understanding that different pathways of differentiation into osteogenic lineage are utilized on collagen I matrices in different structural states (Mauney et al., 2009), it was of interest to analyze osteogenic potential of bone marrow stromal cells during their expansion on native collagen I matrix. Early passage bone marrow stromal cells and late passage cells expanded on native or denatured collagen I matrices were maintained as untreated controls or induced to undergo osteogenesis on either native or denatured collagen I matrices as described in Experimental Procedures section. Late passage cells were differentiated on matrices in the same structural conformation as were used for their expansion. Osteogenic responses were assessed by monitoring expression of bone sialoprotein, BSP, matrix GLA protein, MGP, and the osteogenic transcription factor Runx2 by real-time RT-PCR after 14 days. The results of this experiment are shown in Figure 2. While osteogenic potential is largely retained by late passage mesenchymal stem cells expanded on denatured collagen I, it dramatically declines in late passage cells expanded on native collagen I matrix, to the levels comparable with those seen on its denatured counterpart.

Figure 2. During the expansion, osteogenic potential of bone marrow stromal cells is retained on denatured collagen I matrix but declines substantially on its native counterpart.

DC: denatured collagen I matrix; NC: native collagen I matrix; OS: osteogenic stimulation. BSP: bone sialoprotein; MGP: matrix GLA protein; Runx2, runt-related transcription factor 2. EP: early passage cells; LP: late passage cells expanded and differentiated on DC or NC. [*= p<0.05]: significantly different in comparison to all other experimental conditions. [Δ= p<0.05]: significantly different in comparison to (EP+OS) cultured on NC. [α= p>0.05]: statistically similar in comparison to (EP+OS) cultured on DC.

The observed decline of osteogenic potential of bone marrow stromal cells expanded on native collagen I matrix may reflect cellular aging or it may be a function of the exposure to the matrix. To distinguish between these two possibilities, early passage cells and late passage cells expanded on denatured collagen I matrix or on tissue culture plastic were maintained as untreated controls or induced to undergo osteogenesis on either native or denatured collagen I matrices by the addition of osteogenic stimulants. Osteogenic responses were assessed by monitoring expression of bone sialoprotein, BSP, by real-time RT-PCR after 14 days. As shown in Figure 3 (middle panel: cells expanded on DC I and differentiated on NC I), no loss of native matrix - characteristic osteogenic potential occurred during the expansion of cells on denatured collagen I matrix. On the basis of our previous results (Mauney et al., 2004; 2005; 2006), it could be argued that general retention of cells’ differentiation potentials occurs during expansion on denatured collagen I matrix and it includes the native collagen I matrix-characteristic osteogenic potential. This, however, is not the case for cells expanded on tissue culture plastic. During such an expansion, the potential to undergo osteogenesis on both tissue culture plastic and denatured collagen I matrix is greatly diminished (Mauney et al., 2004), but native collagen I matrix-characteristic osteogenic potential is fully retained (right panel: cells expanded on TCP and differentiated on NC I). Therefore, we conclude that decline of the potential to differentiate into osteogenic lineage seen in late passage cells expanded on native collagen I matrix is, in fact, a function of the exposure to this matrix. The rational for this conclusion is the following. The data presented in Figure 3 clearly shows that the loss of osteogenic potential is a function of prolonged cultivation on the NC I matrix since the parallel exposure to the DC I matrix or to tissue culture plastic does not generate the same effect. If this phenomenon is underlied by cellular aging (defined in terms of the loss of differentiation potential), it should be argued that, judging by retention of osteogenic potential, cells age slower (or not at all) on DC I than on NC I. By the same criteria of retention, cells certainly age on TCP but fully retain their potential to undergo osteogenesis on NC I (late passage cells expanded on TCP, i.e. not preexposed to NC I, and differentiated on NC I show the same extent of osteogenesis as early passage cells [also not preexposed to NC I] differentiated on NC I). Therefore we conclude that loss of osteogenic potential by late passage cells expanded and differentiated on NC I is due not to aging (as defined above) but to some other process triggered by the exposure to NC I. In other words, late passage cells expanded and differentiated on NC I show the reduction in osteogenic potential not because of being “late passage cells”, but because of their exposure to collagen I matrix in the native structural conformation prior to differentiation.

Figure 3. Decline of osteogenic potential of bone marrow stromal cells during the expansion on native collagen I matrix is the function of the exposure to native collagen I.

DC: denatured collagen I matrix; NC: native collagen I matrix; OS: osteogenic stimulation; BSP: bone sialoprotein. EP: early passage cells; LP: late passage cells; expanded on DC or tissue culture plastic (TCP) and differentiated on DC or NC. DC or NC = differentiation matrix. (DC) or (TCP) = ex vivo expansion matrix. All experimental groups treated with OS were significantly different in comparison to respective untreated controls; p<0.05. [θ= p<0.05]: significantly different in comparison to EP on (DC+OS). [γ= p>0.05]: statistically similar in comparison to EP on (DC+OS). [Δ= p<0.05]: significantly different in comparison to EP on (DC+OS). [*= p>0.05]: statistically similar to EP on (NC+OS).

In terms of a notion that the exposure of cells to native collagen I matrix constitutes a stress and elicits cellular stress response required for osteogenic differentiation, the observed decline of osteogenic potential during the expansion on native collagen I matrix could be interpreted as the attenuation of cellular stress response, a well known phenomenon whereby the extent of stress response diminishes during a prolonged exposure to stress stimulus (Gabai et al., 1998; Volloch et al., 1998; Jeon et al., 2004). If such a phenomenon were to occur during the expansion of cells on native collagen I matrix, it would result in the decline of levels of stress response proteins, including Hsp90. An alternative interpretation is discussed in the Discussion section below. Importantly, as shown in Figure 4, levels of Hsp90 do in fact decline in cells expanded on native collagen I matrix, reaching those seen on the denatured collagen I matrix. Levels of Hsp110, a functional partner of Hsp90 in ERK activation pathway, behave similarly. If Hsp90 were indeed instrumental in supporting a high, native collagen I matrix-characteristic rate of osteogenesis, its decline might explain the observed decrease in osteogenic potential of late passage cells expanded on native collagen I matrix.

Figure 4. Levels of stress response proteins decline during the expansion of bone marrow stromal cells on native collagen I matrix.

EP: early passage cells; LP: late passage cells; Hsp: heat shock protein; DC: denatured collagen I matrix; NC: native collagen I matrix. Expression levels in EP on native collagen I matrices designated as “100”. Levels of each particular Hsp in EP on denatured collagen I matrix and in LP on both NC and DC are shown in relation to its expression in EP on native collagen I matrix. [*= p<0.05]: significantly different in comparison to levels of the same Hsp on DC. [Δ= p<0.05]: significantly different in comparison to EP on NC; [α= p>0.05]: statistically similar to cells on DC for respective Hsps.

Stress-mediated elevation of Hsp90 levels in bone marrow stromal cells shifts osteogenic differentiation on denatured collagen I matrix from an Hsp90-independent to a Hsp90-dependent pathway; these two pathways are mutually exclusive.

Previously, we demonstrated that osteogenic differentiation of bone marrow stromal cells occurs via an Hsp90-dependent pathway on native collagen I matrix, but in an Hsp90-independent manner on its denatured counterpart (Mauney and Volloch, 2009). Above, we suggested that such an Hsp90 independency may be due to insufficient amount of Hsp90, with levels of this chaperone on denatured collagen I matrix representing base levels normally sequestered for housekeeping duties, and therefore unavailable and inconsequential in terms of osteogenic differentiation. If this assumption is correct, and if cellular levels of Hsp90 can be elevated by external manipulations, such as thermal stress, we can expect that (a) if Hsp90 is a limiting factor, the extent of osteogenesis will increase in cells on native collagen I matrix and (b) since elevated levels of Hsp90 will enable Hsp90-dependent osteogenic differentiation pathway in addition to an Hsp90-independent pathway already operating in cells on denatured collagen I matrix under normal conditions, the extent of osteogenesis will also increase in mesenchymal stem cells on denatured collagen I matrix.

These predictions were tested in bone marrow stromal cells with elevated levels of Hsp90. Hsp90 is a stress response protein, therefore its levels can be raised by an external stress. To do so, in our experiments we used continual mild thermal stress. As shown in Figure 5 inset, in early passage cells maintained on denatured collagen I matrix at an elevated temperature, levels of all measured Hsps, including Hsp90, were increased and comparable to those seen in cells on native collagen I matrix at normal temperature. Therefore, this model system was used to test the predictions made above. Cells were maintained as untreated controls or induced to undergo osteogenesis on native or denatured collagen I matrices at either normal or elevated temperatures by the addition of osteogenic stimulants in the presence or in the absence of 17-AAG, a derivative of geldanamycin and an inhibitor of Hsp90. 17-AAG proved to be toxic when used for the full duration of the experiment (14 days). Therefore, in the present study it was applied only during days 7–14, with no toxic effects. This time window was chosen because of the observation (Jaiswal et al., 2000) that during in vitro osteogenesis of mesenchymal stem cells the largest increase in values of ERK occurs in days 7 through 13. After fourteen days the extent of osteogenic differentiation was evaluated by measuring the expression of matrix GLA protein. As shown in Figure 5, in this model system, the extent of osteogenesis on native collagen I matrix was significantly increased at elevated temperature and strongly suppressed by treatment with 17-AAG, an inhibitor of Hsp90, indicative that elevation of Hsp90 was primarily responsible for the observed effect and consistent with a notion that cellular level of Hsp90 is a limiting factor in osteogenic differentiation of bone marrow stromal cells on native collagen I matrix.

Figure 5. Continual mild thermal stress shifts osteogenesis of bone marrow stromal cells on denatured collagen I matrix from Hsp90-independent to Hsp90-dependent pathway; these two pathways are mutually exclusive.

C: untreated controls; OS: osteogenic stimuli; 17-AAG: 17-Allylamino-17-demethoxygeldanamycin, inhibitor of Hsp90; NC: native collagen I matrix; DC: denatured collagen I matrix; Hsp: heat shock protein; 37°C or 39°C: incubation temperature. Insert: relative expression of Hsps on DC at 37°C versus 39°C; [π= p<0.05]: compared to cells at 37°C. All +OS alone compared to respective untreated controls, p<0.05. [*= p<0.05]: compared to (DC+OS). [δ= p<0.05]: compared to (NC+OS) at 37°C; [α= p>0.05]: compared to untreated controls. [Δ= p>0.05]: compared to DC+OS at 37°C. [γ= p<0.05]: compared to (NC+OS) at 37°C. [μ= p<0.05]: compared to (DC+OS) at 39°C; [β= p>0.05]: compared to untreated controls.

The results obtained on denatured collagen I matrix were unexpected. The extent of cells’ osteogenic differentiation did not increase but actually slightly decreased at the elevated temperature. Moreover, whereas at normal temperature osteogenesis was unaffected by treatment with 17-AAG, consistent with its Hsp90-independent nature, it was suppressed to control levels by such a treatment at elevated temperature, indicating a shift from fully Hsp90-independent to fully Hsp90-dependent pathway of osteogenic differentiation. These results indicate that Hsp90-dependent osteogenic differentiation pathway is available in cells on denatured collagen I matrix, but cellular levels of Hsp90 are insufficient to support it. The results also indicate that in this system Hsp90-dependent and –independent osteogenic pathways are mutually exclusive, with the former being a dominant one. A possible reason for Hsp90-dependent osteogenic differentiation on denatured collagen I matrix being not as efficient as that on its native counterpart is discussed below.

Thermal stress drastically reduces the extent of adipogenic differentiation of bone marrow stromal cells on denatured collagen I matrix.

Previously, we demonstrated that adipogenesis on native collagen I matrix occurs inefficiently and is p38-independent, and that on denatured collagen matrix an efficient adipogenic differentiation is predominantly regulated by p38 (Mauney et al., 2009). Our previous study indicated that p38 can potentially, through inhibition of ERK, be activated to similar levels in cells on native and denatured collagen matrices. However, whereas on native collagen I matrix relatively high levels of ERK suppress levels of active p38 below the threshold required to mediate p38-dependent adipogenesis, on denatured collagen I matrix, p38-dependent adipogenesis appears to be enabled by relatively low levels of ERK (Mauney et al., 2009). When considering consequences of thermal stress-mediated Hsp90 elevation in bone marrow styromal cells on their adipogenic differentiation, it could have been predicted (prediction 1) that on native collagen I matrix it would be unaffected due to the fact that because adipogenesis on native collagen I matrix is p38 independent, an increase in ERK mediated by Hsp90 elevation would not have an effect. As for adipogenesis on denatured collagen I matrix, we have demonstrated above (Figure 5) that upon Hsp90 elevation by thermal stress, Hsp90-independent pathway of ERK activation is replaced by an Hsp90-dependent one, however this replacement results in similar levels of active ERK, as reflected in the extents of osteogenesis (Figure 5). Therefore, because levels of ERK on denatured collagen I matrix don’t change significantly upon Hsp90 elevation, it could have been predicted (prediction 2) that levels of adipogenesis on DC I would also be unaffected by thermal stress. To test these two predictions, early passage cells were maintained as untreated controls or induced to undergo adipogenesis on native or denatured collagen I matrices at either normal or elevated temperatures by the addition of adipogenic stimulants (Mauney et al., 2005), and adipogenic responses were assessed by monitoring the frequency of lipid-accumulated Oil Red-O positive cells after 14 days. As shown in Figure 6, when cells are undergoing adipogenic differentiation on native collagen I matrix, no differences can be seen in extent of differentiation at normal and at elevated temperatures, indicating that, as predicted, the p38-independent mode of adipogenesis is not affected by stress. However on denatured collagen I matrix, contrary to expectations, the extent of adipogenic differentiation is drastically reduced under conditions of thermal stress in comparison with that seen at normal temperature. Such a decrease in the extent of adipogenesis indicates a stress-induced reduction in levels of p38 activity. Above, we demonstrated that on denatured collagen I matrix under stressful conditions, an Hsp90-independent mechanism is being excluded and replaced by an Hsp90-dependent one but levels of ERK remain unchanged and thus cannot account for the decrease in p38 and adipogenesis. Therefore, a possible explanation for the thermal stress-induced suppression of adipogenesis on denatured collagen I is that the excluded Hsp90-independent pathway mediates activation of both ERK and p38 whereas its Hsp90-dependent replacement regulates only ERK but not p38. Thus, a stress-induced switch from Hsp90-independent to Hsp90-dependent pathway would eliminate a source of p38 activity and result in suppression of adipogenesis.

Figure 6. The extent of bone marrow stromal cells’ adipogenesis is substantially reduced on denatured collagen I matrix under continual mild thermal stress.

C: untreated controls; AD: adipogenic stimuli; NC: native collagen I matrix; DC: denatured collagen I matrix; 37°C or 39°C: incubation temperature. All experimental groups significantly different from untreated controls, p<0.05. [α= p<0.05]: significantly different in comparison to (NC+AD) at 37°C. [Ω= p<0.05]: significantly different in comparison to (DC+AD) at 37°C. Photomicrographs inset: scale bar = 500μm.

Elicitation of stress response in bone marrow stromal cells on native collagen I matrix requires the engagement of α₂β₁ integrins.

Above, we demonstrated that the exposure to native collagen I matrix elicits cellular stress response. What triggers such a structural conformation-specific reaction? A possible answer is suggested by two types of reported observations. The first one, obtained with cell cultivated on micropatterned surfaces (Chen et al., 1997; Ranucci et al., 2001; Ingber et al., 2005; Nelson et al., 2005), implicates a geometry of the matrix in engendering a mechanical stress, possibly by forcing cell contortion; on native collagen I such a geometry may be generated by ridges between collagen fibrils, which are absent on collagen I in denatured state. A more plausible possibility is that stress is generated through the force exerted by contracted native collagen I matrix. Cell-mediated contraction of collagen matrix, which takes place both in vitro and in vivo (Grinnell et al., 1984; Schiro et al., 1991; Tiollier et al., 1990), was shown to occur on native, but not on denatured collagen I (Tiollier et al., 1990; Yoshizato et al., 1999). This is consistent with the differential matrix structural state specificity in stress generation observed in our study. Cell-mediated contraction of collagen matrix was also shown to require the engagement of α₂β₁ integrins with binding sites on the matrix (Schiro et al., 1991). Bone marrow stromal cells are engaged with native collagen I matrix through six triple helix-specific α₂β₁ binding sites per molecule, which are not functional on non-helical denatured collagen I (Davis, 1992). These features are consistent with the requirement of α₂β₁ engagement for matrix contraction. Moreover, contraction was shown not to occur on the native collagen IV matrix (Tiollier et al., 1990) and to occur only marginally on the native collagen III matrix (Tiollier et al., 1990), consistent with the relative amount of α₂β₁ binding sites per molecule, namely two on native collagen IV, possibly below the required threshold, and three on native collagen III (Kim et al., 2005), versus six on native collagen I and none on denatured collagen matrices, and/or reflecting the structure of collagen fibrils. To assess the possible involvement of α₂β₁ integrins in eliciting stress response in cells on native collagen I matrix, cells were treated with α₂β₁- and α₁β_1-specific antibodies prior to seeding on either native or denatured collagen I matrices and cultiured in the presence of antibodies for four days. As shown in Fig. 7, even a relatively short four-day treatment with α₂β₁-specific antibodies resulted in retardation of Hsp90 levels on native collagen I matrix to the levels seen on denatured collagen I matrix, whereas no effect was seen with α₁β₁-specific antibodies. Thus, the results obtained are consistent with the notion that elicitation of cellular stress response on native collagen I matrix is mediated by α₂β₁ integrins, plausibly through their role in matrix contraction.

Figure 7. Elicitation of stress response in bone marrow stromal cells on native collagen I matrix requires the engagement of α₂β₁ integrins.

NC: native collagen I matrix; DC: denatured collagen I matrix; α₁β₁AB and α₂β₁AB: antibodies to integrins α₁β₁ and α₂β₁. Hsp90 expression: [*= p<0.05], significantly different in comparison to cells on DC; [α= p<0.05], significantly different in comparison to cells on NC without antibodies; [Δ= p>0.05], statistically similar to cells on NC without antibodies; [Δ= p>0.05], statistically similar to cells on DC without antibodies.

Exposure to native collagen IV matrix does not elicit stress response in and does not support osteogenesis of bone marrow stromal cells.

In the preceding segment, it was mentioned that cell-mediated matrix contraction is not observed on native collagen IV. If the cellular stress response seen in bone marrow stromal cells on native collagen I is elicited by matrix contraction, it could be predicted that levels of Hsp90 will not be elevated on native collagen IV matrix. As shown in Figure 8A, in cells on native collagen IV matrix, levels of Hsp90 are indeed much lower than on native collagen I and very similar to those seen on denatured collagen I matrices. In light of the results described above, it could be expected that osteogenic differentiation of mesenchymal stem cells on native collagen IV matrix would occur at a relatively low, denatured collagen I-characteristic level. To test this prediction, mesenchymal stem cells were induced to undergo osteogenic differentiation either on native or denatured collagen I or on native collagen IV matrices by the addition of osteogenic stimulants. After fourteen days the extent of osteogenic differentiation was evaluated by measuring the expression of bone sialoprotein. As shown in Figure 8B, the extent of osteogenesis on native collagen IV matrix is well below that seen on native collagen I matrix, in fact even significantly lower than on denatured collagen I matrix.

Figure 8. Native collagen IV matrix doesn’t elicit stress response in and doesn’t support osteogenesis of bone marrow stromal cells.

NCI: native collagen I matrix; DCI: denatured collagen I matrix; NCIV: native collagen IV matrix; OS: osteogenic stimuli. [Δ = p<0.05]: significantly different in comparison to all other respective experimental groups. [γ= p>0.05]: statistically similar to DCI. [*= p<0.05]: significantly different in comparison to (DCI+OS). Photomicrographs inset: scale bar = 500μm.

Mixed matrices containing both native and denatured collagen I elicit more robust stress response in bone marrow stromal cells and support higher extent of osteogenesis than homogeneous collagen I matrices in either conformation.

In musculoskeletal tissues, where collagen I is the major matrix component, mesenchymal stem cells are often exposed to mixed matrices containing collagen I in both structural conformations. For example, in a model of bone remodeling, recruited mesenchymal stem cells arrive, due to osteoclasts activity, to a bed of denatured collagen I matrix. Because of a high osteogenic differentiation commencement threshold, osteogenesis program (Mauney et al., 2009), which would be inefficient on denatured collagen I matrix, is not activated. Cells rather proliferate, while retaining their differentiation potential, and deposit a de novo native collagen I matrix which lowers the threshold and allows cells to commence osteogenic differentiation (Mauney et al., 2009). Even in this relatively simple model, matrix evolves through stages when it is structurally mixed and contains both native and denatured collagen I. To evaluate the behavior of mesenchymal stem cells during such a transition, we measured osteogenesis on mixed matrices containing both native and denatured collagen I in various defined proportions. Bone marrow stromal cells were maintained as untreated controls or induced to undergo osteogenesis on native or denatured collagen I matrices as well as on mixed matrices containing native and denatured collagen I in proportions 75%:25%, 50%:50% and 25%:75% by the addition of osteogenic stimulants in the presence or in the absence of 17-AAG, an inhibitor of Hsp90. After fourteen days the extent of osteogenic differentiation was evaluated by measuring the expression of matrix GLA protein. The results of this experiment are shown in Figure 9. Contrary to our anticipation of a gradual increase of osteogenesis from the level seen on denatured collagen I to that occurring on its native counterpart, on mixed matrices the extent of osteogenesis significantly exceeded that seen on native collagen I matrix (Figure 9A). Moreover, as shown in Figure 9A, osteogenesis on mixed matrices occurs in Hsp90-dependent manner, seen only on native, but not on denatured, collagen I matrix. These results are reminiscent of the observation described above with cells on native collagen I matrix subjected to external stress, which suggested that that Hsp90 might be a limiting factor in osteogenesis. Consistent with this notion, as shown in Figure 9B, levels of Hsp90 are indeed significantly elevated in cells on mixed matrices, suggesting that these matrices elicit a more robust cellular stress response than homogeneous native collagen I matrix.

Figure 9. Effect of mixed matrices containing both native and denatured collagen I on stress response and osteogenesis of bone marrow stromal cells.

C: untreated controls; OS: osteogenic stimuli; 17-AAG: 17-Allylamino-17-demethoxygeldanamycin, inhibitor of Hsp90; NC: native collagen I matrix; DC: denatured collagen I matrix. [θ = p<0.05]: significantly different in comparison to all other groups in panel A. [*= p<0.05]: significantly different in comparison to all other groups in panel B.

Exposure of satellite stem cells to native collagen I matrix elevates cellular levels of β-catenin.

The results described above suggest that the exposure to native collagen I matrix elicits cellular stress response which, through elevation of Hsp90 levels, plays a regulatory role in selection of pathways for osteogenic and adipogenic differentiation. What is the scope of the involvement of matrix-mediated stress response in regulation of tissue generation, could we find another example of this phenomenon? To answer this question, we turned to skeletal myogenesis. Homeostasis of skeletal muscle tissue is supported by satellite stem cells, which in addition to myogenic potential, also have adipogenic capability (Csete et al., 2001). Both types of differentiation are positively regulated by p38 but differentiation into muscle tissue also requires β-catenin activity, which was shown to be essential and sufficient for skeletal myogenesis (Petropoulos etal., 2002). β-catenin activity is known to be induced by various stresses (Norvell et al., 2004; Douglas et al., 2006), including mechanical stress (Hens et al., 2005). Moreover, a stress was shown capable of inducing translocation of beta-catenin to the nucleus (Norvell et al., 2004). The major matrix component in skeletal muscle tissue is collagen I (Kjaer, 2004). Since the exposure to native collagen I matrix was shown to elicit cellular stress response and since β-catenin levels react to stresses, there is reasonable probability that interaction of satellite stem cells with native collagen I matrix might lead to the induction of the nuclear β-catenin activity.

To test the plausibility of this concept, human satellite stem cells were cultivated on either native or denatured collagen I matrices, and the nuclear content of active β-catenin was analyzed after eight days by immunoblotting. As shown in Fig. 10, nuclear levels of active unphosphorylated β-catenin are indeed significantly increased in cells maintained on native collagen I matrix, consistent with a notion that collagen I matrix can play a regulatory role in myogenesis of satellite stem cells through elicitation of stress response in its native (versus denatured) structural conformation and that differential matrix-mediated stress response might be, through its varied components, involved in regulation of differentiation of stem and progenitor cells into diverse lineages.

Figure 10. Effect of structural conformation of collagen I matrix on nuclear levels of active unphosphorylated β-catenin in muscle satellite stem cells.

Satellite stem cells were cultured on either denatured (DC) or native (NC) collagen I matrices. Nuclear proteins were resolved by PAGE-SDS and immunobloted using antibody for nonphosphorylated (active) β-catenin. Arrow denotes 97 kDa β-catenin band.

DISCUSSION.

In the present study, a number of conventional markers were used separately or in combination to assess the extent of bone marrow stromal cells’ differentiation. For adipogenesis, a direct in situ detection of lipids was employed. For osteogenesis, the following markers were utilized: BSP, bone sialoprotein, MGP, matrix GLA protein, and Runx2, runt-related transcription factor 2. Of those BSP is a conventional marker specific for osteogenic differentiation whereas Runx2 is a confirmatory marker characteristic, albeit not specific, for osteogenesis. The utilization in the present study of MGP as an early osteogenic marker is discussed in our earlier study (Mauney et al., 2009). In a previous study (Mauney et al., 2009) we demonstrated that osteogenic differentiation of human bone marrow stromal cells on collagen I matrix occurs in a manner dependent on the structural conformation of the matrix via distinct matrix structure-specific pathways. On native collagen I matrix an efficient osteogenesis occurs via an Hsp90-dependent pathway whereas on denatured collagen I matrix it proceeds rather inefficiently and in an Hsp90-independent manner. The results of the present study show that the exposure of cells to native collagen I matrix elicits cellular stress response which appears to be mediated by α₂β₁ integrins engaged with corresponding binding sites, six of which per molecule are present on collagen I matrix in native structural conformation but none are available on its denatured counterpart (Davis, 1992). The stress causing such a response appears to be cell-mediated matrix contraction; this process, in turn, appears to be a function of the amount of α₂β₁ sites per molecule of matrix. Indeed, contraction (and stress response) does not occur on native collagen IV matrix (Tiollier et al., 1990) which contains only two sites per molecule and occurs only marginally on native collagen III matrix which contains three sites per molecule (Tiollier et al., 1990; Kim et al., 2005).

The results obtained suggest that high levels of Hsp90 in cells on native collagen I matrix, a manifestation of cellular stress response, enable Hsp90-dependent osteogenic differentiation on this matrix. They also indicate that Hsp90-dependent osteogenic differentiation pathway is available in mesenchymal stem cells on denatured collagen I matrix but cellular levels of Hsp90 are insufficient to support it. In view of such a drastically different behavior of mesenchymal stem cells on collagen I matrices in native and denatured structural conformations it is important to consider a possibility that the observed phenomenon may be due not to a stress response elicited by exposure to native (but not to denatured) collagen I matrix, but to a contaminating factor present in native collagen I preparation and inactivated upon its denaturation. The results of the present study, however, provide evidence that this is not the case. Indeed, when cells on denatured collagen I matrix are subjected to an external treatment (thermal stress) resulting in the elevation of Hsp90 to levels comparable with those seen on native collagen I matrix, osteogenic differentiation is shifted from Hsp90-independent mechanism, specific for denatured matrix, to Hsp90-dependent one, specific for collagen I matrix in native conformation. This observation leaves the remote possibility that a hypothetical contaminating factor elicits a cellular stress response; in this case thermal stress used in our experiments would mimic the effect of such a contamination. However, this possibility is ruled out by the results obtained with mixed matrices comprised of collagen I in both native and denatured conformations; these mixtures elicit a more robust stress response and promote more efficient osteogenesis than homogeneous native collagen I matrices.

The results of the present study strongly suggest that Hsp90 plays a pivotal role in determining the mechanism and, consequently, the efficiency of bone marrow stromal cell differentiation into the osteogenic lineage. On the other hand, previously (Mauney et al., 2009), we demonstrated that osteogenic differentiation of human bone marrow stromal cells is regulated by ERK activity on both, native and denatured, collagen I matrices. Therefore, the plausible role of Hsp90 in regulation of osteogenesis should be considered in terms of its involvement in regulation of ERK activity.

As was mentioned above, interaction of mesenchymal stem cells with native collagen I matrix occurs through the engagement of triple helix-specific integrins α₁β₁ and α₂β₁ with corresponding binding sites on the matrix, six per molecule for each type of integrin. These engagements initiate one shared ERK activation pathway; in addition, both engaged α₁β₁ and α₂β₁ integrins may trigger specific contributions to signal transduction pathway leading to ERK activation. Shared ERK activation pathway initiated by engaged α₁β₁ and α₂β₁ proceeds as follows: α₁β₁, α₂β₁ – FAK -…- Raf – Mek – ERK (Egan et al., 1993; Schlaepfer et al., 1996; Takeuchi et al., 1997). Pathways specific for α₁β₁ proceed either through α₁β₁ – Cav1-…- Raf – Mek – ERK (Wary et al., 1996; 1998) or α₁β₁ – cSrc-…- Raf – Mek – ERK (Schlaepfer et al., 1996); both joining the common pathway (Wary et al., 1996; 1998). A specific contribution of α₂β₁ into ERK activation consists of triggering the activation of PP2A (Yamagishi et al.,2004; Chetoui et al.,2005). This phosphatase acts on Ser259 of Raf-1 (Sanders et al.,2004) facilitating its release from 14-3-3 proteins (Abraham et al., 2000) and substantially increasing levels of Raf-1 available for recruitment to plasma membrane and interaction with and subsequent activation by Ras.

The involvement of Hsp90 in these pathways occurs at the level of activation of Raf-1, a client protein of the Hsp90 chaperone. Indeed, all pathways mentioned above, which are identical from the Grb2 link down to ERK, crucially depend on Hsp90, shown to be essential for Raf-1 activation (Cutforth et al., 1994; van der Straten et al., 1997). Consequently, since there are no known Hsp90-independent or Raf-1-independent mechanisms of ERK activation on native collagen I matrix, and as was seen in our experiments, osteogenic differentiation of on this matrix critically depends on elevated levels of Hsp90. It should be mentioned that in the context of Raf activation in cells on native collagen I matrix, the role of Hsp110 is complementary to that of Hsp90 in the following manner. Hsp70, also elevated in cells as a part of cellular stress response elicited by the exposure to native collagen I matrix, is capable of binding to and titrating Bag-1, another factor participating in Raf-1 activation (Song et al., 2001; Yamagishi et al., 2004). Hsp110 prevents the Hsp70/Bag-1 interaction, thus assisting in activation of Raf-1 (Yamagishi et al., 2004). In light of the above, the observed attenuation of the expression of Hsps90 (and of Hsp110) may be responsible for the decline of osteogenic potential seen during the expansion of bone marrow stromal cells on native collagen I matrix. Such a decline could result from one of the following two processes: (i) attenuation of stress response mentioned above and (ii) cell-mediated remodeling of native collagen I matrix into denatured conformation. Additional research is required to distinguish between these two possibilities.

On denatured collagen I, which lacks triple helical structure and where binding sites for triple helix-specific α₁β₁ and α₂β₁ integrins are not functional, interactions of mesenchymal stem cells with matrix occur through the engagement of integrins α_Vβ₃ with “cryptic” binding sites obscured in the native conformation (Davis, 1992). Potential ERK activation pathways initiated from α_Vβ₃ can be divided into two classes – Hsp90-dependent and Hsp90-independent. The Hsp90 (and Raf-1)-dependent α_Vβ₃ pathway duplicates an α₁β₁-initiated pathway: α_Vβ₃ – Cav1-…- Raf – Mek – ERK (Wary et al., 1996; 1998). However, as we demonstrated, due to insufficient levels of Hsp90, probably sequestered for housekeeping duties, this pathway is inactive in cells on denatured collagen I matrix under normal circumstances, where, therefore, Hsp90-independent mechanism is utilized. The results obtained suggest that, despite the lack of pp2A activation, the elevation of Hsp90 levels by external manipulations in mesenchymal stem cells on denatured collagen I matrix enables Hsp90-dependent pathway which appears to be dominant and mutually exclusive with Hsp90-independent one. Despite commonality of α_Vβ₃ − initiated Hsp90-dependent pathway with α₁β₁/α₂β₁–initiated ERK activation pathways, Hsp90-dependent osteogenic differentiation of mesenchymal stem cells on denatured collagen I matrix is not as efficient as that on its native counterpart. This can be explained by low levels of Raf-1 on denatured collagen I matrix resulting from the lack of α₂β₁-initiated activation of protein phosphatase 2A (Sanders et al., 2004; Chetoui et al., 2005), which, when it occurs as for example in cells on native collagen I matrix, increases the amount of available Raf-1. This interpretation is consistent with the observation that the engagement of α₁β₁ integrin is significantly less efficient in ERK activation than that of α₂β₁ (Sanders et al., 2004).

The class of potential Hsp90-independent pathways of ERK activation initiated from α_Vβ₃ can be further divided into categories of MEK-dependent and MEK-independent. The latter one (Rucci et al., 2005; Noon et al., 2005; Tapinos et al., 2005; Blanco-Aparicio et al., 1999; Saxena et al., 1999; Gomez et al., 2002), can be excluded because of our previous results that clearly indicate MEK dependency for osteogenic differentiation of bone marrow stromal cells on denatured collagen I matrix (Mauney and Volloch, 2009). This leaves the category of Hsp90-independent but MEK-dependent pathways. This category can be further separated into two groups, those activating only ERK but not p38 (Short et al., 2000; Wen-Sheng et al., 2006) and those simultaneously activating ERK and p38 (Kaneki et al., 1999; Hagemann et al., 2001; Franklin et al., 2000; Mittelstadt et al., 2005):

$${\text{α}}_{\mathrm{V}}{\text{β}}_3-\dots \text{-}Lc{k}_{{\searrow }_{ZAP-70-p38}}^{{\nearrow }^{MEK-ERK}}\text{\hspace{1em}and, possibly,}{\text{α}}_{\mathrm{V}}{\text{β}}_3-\dots \text{-}MEKK{1}_{{\searrow }_{SEK-p38}}^{{\nearrow }^{MEK-ERK}}$$

The results obtained exclude the former group and indicate that the Hsp90-independent, α_Vβ₃ integrin-triggered ERK activating pathway operating in cells on denatured collagen I under normal conditions is of a type that simultaneously activates ERK and p38 such as those diagrammed above. Upon stress-mediated elevation of Hsp90 levels, this pathway is being excluded and replaced by an available α_Vβ₃ integrin-initiated Hsp90-dependent pathway, which activates ERK but not p38 (Klekotka et al., 2001) resulting in a substantial drop of adipogenesis.

In this context, the role of Hsp90 in regulation of osteogenesis is that of an “enabler”, a pivotal lever facilitating efficient osteogenic differentiation by enabling the operation of Hsp90-dependent α₁β₁/α₂β_1-initiated ERK activation pathways on stress response-eliciting native collagen I matrix. The results obtained suggest that Hsp90 is a limiting factor in Hsp90-dependent osteogenesis, consistent with observations of Hsp90 being a limiting factor in ERK activation in other systems (Caraglia et al., 2005). Indeed, elevation of Hsp90 levels either by mild thermal stress or by exposure to mixed matrices results in a significantly increased extent of osteogenic differentiation indicating that elicitation of stress response by any means and consequent elevation of Hsp90 lead to the increase of osteogenesis, probably through the increase in ERK activation. When structurally mixed collagen I matrices were used, a more robust cellular stress response than on homogeneous native collagen I matrix was elicited. The more robust stress response indicates stronger stress, in this case more mechanical force generated presumably by matrix contraction. An obvious possible explanation, albeit one of many, is the increased rigidity of intercalated structurally mixed matrix. Whether this is indeed the mechanism responsible for the phenomenon observed, and if it is how it operates, remains to be established. The observations made suggest that, at least with respect to collagen I in native and denatured conformations, the most efficient cellular stress response-assisted differentiation of mesenchymal stem cells may, in fact, occur on mixed matrices. Differentiating mesenchymal stem cells may be exposed to mixed matrices consisting not only of different structural conformations of collagen I, but of different types of collagen some of which can be, in addition, glycosylated as, for example, in chondrogenesis. Such matrices might have considerable influence on differentiation of mesenchymal stem cells into various lineages, and their effect should be further investigated.

The key role of Hsp90 in supporting efficient osteogenesis seen on native collagen I matrix can be illustrated by observations made with native collagen IV matrices. Native collagen IV consists of both, collageneous triple helical domains as well as non-collageneous domains. The engagement of mesenchymal stem cells with this matrix occurs through α₁β₁ and α₂β₁ integrins (three and two binding sites per molecule respectively) as well as through α_Vβ₃ integrin (one binding site per molecule in the non-collageneous domain). As was mentioned above, native collagen IV matrix doesn’t undergo cell-mediated contraction. Consistent with the above proposed role of matrix contraction in elicitation of cellular stress response, levels of Hsp90 in mesenchymal stem cells on this matrix are low, about the same as on denatured collagen I matrix, and could not support α₁β₁/α₂β₁–initiated Hsp90-dependent osteogenic differentiation. It means that osteogenic differentiation of mesenchymal stem cells on native collagen IV matrix is driven by α_Vβ₃ integrin engagement with a single site per molecule at the noncollageneous domain (versus, for example, seven α_Vβ₃ sites per molecule on denatured collagen I matrix); accordingly osteogenesis occurs at very low level.

As was mentioned above, both myogenesis and adipogenesis of satellite stem cells are positively regulated by p38. However, in contrast to adipogenesis, myogenesis of satellite stem cells require elevated levels of β-catenin, shown to be necessary and sufficient for myogenic differentiation (Petropoulos et al., 2002). It appears that physiologically levels of β-catenin might be elevated in response to cellular stress elicited by the exposure to collagen I matrix, the major matrix component in skeletal muscle tissue, in its native conformation. It could be suggested, therefore, that upon recruitment of satellite stem cells, their fate as myogenic or adipogenic cells might be decided in large measure by a conformation of collagen I matrix they encounter. The immediate pathways involved in matrix-mediated stress-induced elevation of nuclear beta-catenin remain to be elucidated. In case of fluid shear stress they were shown to include an increase in the phosphorylation of GSK-3beta and Akt, as well as a reduction of the levels of beta-catenin sequestered by N-cadherin (Norvell et al., 2004). Whether these mechanisms are also involved in satellite stem cells on native collagen I matrix remains to be tested.

Reduction of levels of cellular stress response components elevated by the exposure to native collagen I matrix and involved in regulation of differentiation programs, namely Hsp90 in mesenchymal stem cells and β-catenin in satellite stem cells, can explain adipogenic shift which occurs physiologically in aging musculoskeletal tissues. Such a reduction can be effected either trough general physiological attenuation of stress response seen in aged individuals (Heydari et al., 1994) or/and as a result of the accelerated conversion of native collagen I matrix into denatured conformation due to an aging-related increase in secretion of MMPs capable of cleaving triple helical structure of native collagen (Wang et al., 2002; Wang and Rokkatta, 2002; McNulty et al., 2005) coupled with a decrease in the expression of TIMPs in aging tissues (Joronen et al., 2000; Lindsey et al., 2005). In this context, it should be mentioned that irregular reduction in cellular stress response either through its attenuation or through aberrations in collagen I structure can be involved in numerous pathologies associated with increased adipogenesis (Rodriguez et al, 2000; Sokolov et al., 1992; Willing et al. 1994; Roughley et al. 2003; Dollery et al., 2005; Berria et al., 2005; Kamei et al., 2004; Lecker et al., 2004; Li et al., 2004).

The main contribution of the present study is the introduction of a novel concept which proposes that developmental fate of mesenchymal stem cells, satellite stem cells, and possibly of other progenitor cells is regulated by structural state of collagen I matrix which plays, alongside differentiation stimuli, a decisive role in the selection of differentiation pathway through conformation-specific, α₂β₁-mediated elicitation of cellular stress response or lack thereof. Importantly, the results obtained suggest that different manifestations of cellular stress response, i.e. its distinct constituents, regulate differentiation into diverse lineages. Thus, augmentation of Hsp90 levels enables the operation of the efficient α₁β₁/α₂β₁-driven ERK activation pathways hence facilitating osteogenesis and suppressing adipogenesis of mesenchymal stem cells, whereas myogenesis of satellite stem cells appears to be promoted by the native collagen I matrix-elicited activation and nuclear translocation of another stress response component, β-catenin, shown to be essential and sufficient for skeletal myogenesis (Petropoulos et al., 2002). In the present study we analyzed the effects of stress response elicited by the exposure of cells to collagen I matrix in the native structural conformation. It is plausible that the elicitation of cellular stress response by other physiological factors might also play an active role in the determination of differentiation lineage. The results obtained suggest a potentially large scope of stress response involvement in regulation of differentiation into diverse lineages. It could be hypothesized, for example, that yet another stress response component, Hsp70, shown to be an interactive partner of the chondrogenic transcription factor SOX9 (Marshall and Harley, 2001), might play a regulatory role in chondrogenesis of mesenchymal stem cells. In conclusion, the present study proposes a novel concept of integral role of cellular stress response in differentiation of mesenchymal and possibly other types of stem cells into diverse lineages; it suggests new therapeutic approaches and indicates novel tissue engineering strategies.

EXPERIMENTAL PROCEDURES.

Abbreviations.

EP, early passage cells; LP, late passage cells; NC I, native collagen I matrix; DC I, denatured collagen I matrix; TCP, tissue culture plastic; ERK. Extracellular signal-regulated kinase; p38, p38 mitogen-actvated protein kinase; Raf or Raf-1, serine/threonine-specific kinase in ERK activating cascade; Hsp, heat shock protein; 17-AAG, 17-Allylamino-17-demethoxygeldanamycin, inhibitor of Hsp90; BSP, bone sialoprotein; MGP, matrix GLA protein; Runx2, runt-related transcription factor 2; MMP, matrix metalloproteinase; TIMP, tissue inhibitor of metalloproteinase. OS, osteogenic stimuli; AD, adipogenic stimuli.

Bone marrow stromal cells preparation.

Human bone marrow stromal cells were obtained from commercially available bone marrow aspirates (Cambrex, Walkersville, MD) from five male donors; similar trends in cell responses were observed with all donors. Cells were expanded on tissue culture plastic to passage 1 (P1) utilizing previously reported methods (Altman et al., 2002; Mauney et al., 2004; 2005). Briefly, whole bone marrow aspirates were plated at 8–10 μl aspirate/cm² on 185 cm² tissue culture plates and cultivated until confluency (~12–14 days) in 40 ml of expansion medium consisting of Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (FCS), 100U/ml penicillin, 100 μg/ml streptomycin, 0.1 mM nonessential amino acids, and 5 ng/ml of basic fibroblast growth factor (bFGF) (Life Technologies, Rockville, MD). Cultures were maintained in a humidified tissue culture incubator at 37°C with 5% CO₂ and 95% air. Bone marrow stromal cells were selected on basis of their ability to adhere to tissue culture plastic (TCP); non-adherent hematopoietic cells were removed during medium replacement after approximately 7 days in culture. Medium was changed twice per week thereafter. P1 cells were frozen in liquid nitrogen at 5 mln/ml in 90% fetal calf serum and 10% DMSO and utilized in subsequent experiments. Viable cell recovery after thawing was over 80%. Differentiation potentials of so obtained cell populations for both osteogenic and adipogenic lineages have been verified both in vitro and in vivo (Mauney et al., 2006; 2007). P1 and P2 mesenchymal stem cells were designated “early passage cells”, EP. Early passage cells were expanded to passage 8, P8, on denatured collagen I matrix; each passage corresponded to 3 population doublings (cells were plated at 1/8^th of the confluence and replated upon reaching confluence, thus constituting a passage). During the same time, cells on NC I and TCP reached only passage 6 or 7 due to their lower proliferation rate (Mauney et al., 2004). These cells were designated “late passage cells”, LP. Since experiments were carried out simultaneously, P8 on DC I were compared with P6–P7 on NC I and TCP.

Preparation of collagen matrices.

Collagen I solutions were prepared and utilized to cast collagen films using methods previously described (Volloch and Kaplan, 2002). Briefly, sterile rat tail-derived collagen type I (Roche, Indianapolis, IN; cat. #1179179) was dissolved at 0.5 mg/ml in 0.1% acetic acid and either maintained in its native conformation (NC) or denatured (DC) by incubation at 50°C for 8 hrs as previously reported (Volloch and Kaplan, 2002); at these conditions collagen was shown to be denatured and have only a small degree of fragmentation (Volloch and Kaplan, 2002).. To prepare films, native or denatured collagen solutions were added to 96 well tissue culture plates at 250 μg/cm², dried under a laminar flow hood for 4–5 days and lyophilized for 1–2 days to enhance film stability during cell cultivation. For preparation of mixed matrices, solutions of native and denatured collagen I were combined in defined proportions prior to film casting. Similar trends in cell responses were observed when rat tail collagen was compared with human placenta-derived collagen type I (Sigma-Aldrich, cat. #C7774) or when films were cast in 6 or 24 well tissue culture plates. Reported results were obtained with rat tail-derived collagen, except when human placenta-derived collagen type IV (Sigma-Aldrich, cat. #C7521) was used and compared with human collagen I mentioned above.

Osteogenic and adipogenic differentiation of bone marrow stromal cells.

Cells on native or denatured collagen I matrices were treated with either OS or AD or maintained as untreated controls as previously described (Mauney et al., 2004; 2005), and analyzed for osteogenic or adipogenic markers. Briefly, cells were plated at 5×10³ cells/cm² on native or denatured collagen I matrices in medium consisting of DMEM supplemented with 10% FCS, 100U/ml penicillin, 100 μg/ml streptomycin and 0.1 mM nonessential amino acids. 12 hours later either osteogenic (OS) or adipogenic stimulants (AD) were added and subsequently replaced with each medium change every 3–4 days. Osteogenic stimulants included 100 nM dexamethasone, 10 mM β-glycerophosphate, and 0.05 mM L-ascorbic acid-2-phosphate as previously reported (Mauney et al., 2004). Adipogenic stimulants consisted of 0.5 mM 3-isobutyl-1-methyl-xanthine, 1 μM dexamethasone, 5 μg/ml insulin, and 50 μM indomethacin as previously reported (Mauney et al., 2005). Unless otherwise stated, experiments continued for 14 days. Cultures were maintained as described above and medium exchange was carried out twice per week. Parallel control cultures were maintained similarly in the absence of stimulants.

Continual thermal stress.

Cultures were maintained as described above, except in a designated incubator adjusted to 39°C. No changes in the rate of proliferation or in cell behavior were observed.

Inhibition of Hsp90.

The inhibitor, 17-AAG, 17-Allylamino-17-demethoxygeldanamycin, a derivative of geldanamycin, was obtained from InvivoGen, San Diego, CA and used at 100nM as previously described (Waza et al., 2005).

Antibodies to α₁β₁ and α₂β₁ integrins.

Antibodies for α₂β₁ integrin (MAB1998Z ) and for α₁β₁ integrin (MAB1913Z) were obtained from Chemicon an used in saturating concentrations with early passage mesenchymal stem cells. Cells were incubated in suspension in the presence of an antibody for 30 min prior to plating and maintained for four days on native or denatured collagen matrices in the presence of antibodies or as untreated controls.

Oil Red-O Staining and Frequency of Lipid Accumulating Cells.

Following 14 days of AD treatment, cultures were analyzed for the frequency of lipid accumulating cells by Oil Red-O staining as previously described (Mauney etal.,2005). Briefly, cultures were fixed in 4% neutral buffered formalin for 12 hours and stained for 45 minutes with a filtered 60% Oil Red-O solution in PBS, made from a stock solution of 0.70g Oil Red-O powder in 200 ml of isopropanol as previously reported (Stewart et al., 2004). Cultures were then washed with PBS to remove background Oil Red-O stain and photomicrographs were taken with a Zeiss Axiovert S100 light microscope and a Sony Exwave HAD 3CCD color video camera utilizing Scion Image software. The number of Oil Red-O positive cells was then manually counted in each culture well to determine the frequency of lipid accumulating cells.

Real time RT-PCR Analysis.

Markers of interest were analyzed by real time RT-PCR as previously described (Mauney etal.,2004; 2005). Briefly, RNA was extracted using Trizol reagent (Life Technologies), and cDNA were synthesized using High-Capacity cDNA Archive kit (ABI Biosystems) following the manufacturers instructions. Reactions were performed and monitored using an ABI Prism 7000 Sequence Detection System (Applied Biosystems). The PCR master mix was based on AmpliTaq Gold DNA polymerase (Applied Biosystems). For osteogenic differentiation, cDNA samples were analyzed for bone sialoprotein (BSP), matrix GLA protein (MGP), and core binding factor alpha-1, (cbfa1) expression; for stress-response, heat shock proteins 90 (hsp90), 110 (hsp110), 27 (hsp27), and 70 (hsp70) were evaluated. Analysis was performed using commercially available primers and probes from ABI Biosystems Assays-on-Demand™ Gene Expression kits (BSP, cat. #Hs00173720_m1; MGP, cat.# Hs00179899_m1; cbfa1, cat.# Hs00231692_m1; hsp90, cat.# Hs00743767_sH; hsp110, cat.# Hs00198379_ml; hsp27, cat.# Hs00356629_g1; hsp70, cat.# Hs00271244_s1) following the manufacturer’s instructions; the housekeeping gene GAPDH was analyzed using primers and probes as previously described (Frank et al., 2002). Data analysis was performed using the ABI Prism 7000 Sequence Detection Systems version 1.0 software (Applied Biosystems, Foster City, CA). For each cDNA sample, the Ct value was defined as the cycle number at which the fluorescence intensity of each target gene was amplified within the linear range of the reaction. Relative expression levels for each gene of interest were calculated by normalizing the quantified gene of interest transcript level (Ct) to the GAPDH transcript level (Ct) as described previously (2^ΔCt formula, Perkin Elmer User Bulletin #2).

Muscle Satellite Stem Cells (SSCs).

Human satellite stem cells, passage 1, were obtained from Cambrex and expanded to passage 2 according to the manufacturer’s instructions. Passage 2 SSCs were plated at 90% confluency on native and denatured collagen I matrices, cultured for 8 days in DMEM/F-12 with 2% fetal calf serum, and analyzed for the nuclear content of active (unphosphorylated) β-catenin.

Protein Isolation and Immunoblotting of Nuclear β-catenin.

Following trypsinization, SSCs were suspended in Hepes/NaCl buffer (20mM, pH 7.5; 20mM) and lysed by addition of equal volume of 1% TritonX100. Nuclei were collected by centrifugation, washed, resuspended in Hepes/NaCl buffer, lysed by addition of equal volume of 2 × gel sample buffer (Invitrogen) and loaded on 10% polyacrylamide-SDS gel immediately after incubation at 100°C for 5 min. Western blots were analyzed using a mouse monoclonal primary antibody against human nonphosphorylated β-catenin followed by incubation with a goat anti-mouse secondary horseradish peroxidase conjugated antibody (cat. ##05–601 and 12–349,Upstate Cell Signaling Solutions). Bands of interest were visualized using ECL reagents. Equal loading of protein samples was ascertained by Ponceau BS (Sigma-Aldrich) staining of western blots.

Statistical Analysis.

All measurements were collected with N=3–5 independent samples per data point and expressed as means ± standard deviations. Data were analyzed with Microsoft Excel software utilizing a Student’s two tailed t-test assuming equal levels of variance. Statistically significant values were defined as p<0.05.