Mesenchymal stem cells remodeling of adsorbed type I collagen – effect of collagen oxidation

The effect of collagen type 1 (Col I) oxidation on Adipose Tissue Derived Mesenchymal Stem Cells (ADMSCs) remodeling is described as a model for acute oxidative stress. Morphologically, remodelling was presented by a mechanical rearrangement of adsorbed FITC-Col I and a trend for its organization in fibril-like pattern – a process strongly abrogated in oxidized samples, but without visible changes in cell morphology. The cellular proteolytic activity was quantified in multiple samples utilizing fluorescence de-quenching (FRET effect). In the presence of ADMSCs a significant increase of native FITC-Col I fluorescence was observed, almost absent in the oxidized samples. Parallel studies in cell-free system confirmed the enzymatic de-quenching of native FITC-Col I by Clostridial collagenase, again showing significant inhibition in oxidized samples. The structural changes in the oxidized Col I was further studied by Differential Scanning Calorimetry: an additional endotherm at 33.6°C along with the typical for native Col I at 40.5°C with sustained enthalpy (∆H) was observed in oxidized samples. Collectively, it has been evidenced that remodeling of Col I by ADMSCs is altered upon oxidation due to the intrinsic changes in the protein structure, thus presenting a novel mechanism for the control of stem cells behaviour toward collagen.


Introduction
Recent studies show that mesenchymal stem cells (MSCs) are strongly involved in the process of extracellular matrix (ECM) remodeling apart from their principal role in the various regenerative routes [1][2]. MSCs, often referred to as adult stem cells reside in most tissues assuring their repair upon damage and ageing [3], thus attracting great interest for most cellbased therapies [4]. Particularly Adipose Tissue Derived MSCs (ADМSCs) draw notable association, folding, secretion, self-assembly, and progressive cross-linking [19]. The major PTMs depends on the oxidation of lysine and proline residues, being critical factors for the structural and biomechanical functions of Col I fibrils, strictly regulated by several sequential processes inside and outside the cell. The oxidation of lysine and proline may occur upon distinct environmental changes [19], for example as a part of the physiological collagen processing like fibrillogenesis and osteocalcification. It might be also a component of the oxidative stress [20,21], that is dependent on the production of reactive oxygen species (ROS) including free radicals and peroxides [20]. The oxidation of collagen (also other ECM proteins) caused by ROS further modulates the qualities of the ECM by altering its production, turnover, and modifications, finally, having strong impact on the cell-ECM interaction [22].
Despite the extensive investigations on the role of oxidative stress in collagen genes expression and collagen turnover related to various chronic diseases, the studies in vitro utilizing direct cellular models are rather sparse [27][28][29]. Though it is clear native collagen undergoes intensive remodeling by cells, and particularly by fibroblasts, the specific role of MSCs was rather poorly investigated [16]. The effects of oxidative stress caused by ROS and accompanying regeneration of the injured tissues are also sparsely studied [30].
In this study, we aimed to learn more about the effect of Col I oxidation (as an in vitro model for the acute oxidative stress) on its remodeling by stem cells. For that purpose, we visualized morphologically the adsorbed Col I with the adhering ADМSCs and further developed a system for the quantification of observed cellular proteolysis using specially designed fluorescent probes.

Results
In this study we aimed to compare the biological response of ADMSCs adhering onto native and pre-oxidized Col I attempting to mimic the conditions of acute (short term) oxidative stress that occur in vivo, assuming that it might have a significant effect on the stem cells behavior. For that purpose native Col I was labelled with FITC to easily follow it's processing. For the preparation of oxidized FITC-Col I (FITC-Col I OXI) we used a previously described protocol [31]. The remodeling of adsorbed FITC-Col I was investigated employing two approaches, morphological and quantitative. For the morphological approach, ADMSCs were cultured on glass coverslips pre-coated with either native FITC-Col I or FITC-Col I OXI for 24 h. Then the samples were stained with Rhodamine-phalloidin and Hoechst to view simultaneously the substratum (green), the actin cytoskeleton (red) and the intact nuclei (blue), using the respective channels of the microscope. The second, quantitative approach was based on the de-quenching of FITC-Col I caused by the cellular proteolytic activity giving a proportional rise of the fluorescent signal (FRED effect) [32]. To confirm the de-quenching effect and for the comparison of native and oxidized Col I susceptibility to proteolysis, a separate experiment with collagenase from Clostridium Histolyticum (CH) were conducted.

Overall design of the experiments
A basic scheme of the study is presented on Figure 1. Substrates were coated with FITC-Col I or FITC-Col I-OXI (A). Cell adhesion was the next step (B), performed for 2 h in serumfree medium (to assure the attachment of ADMSCs to Col I only). Afterwards, 10% serum was added to each sample and cells were further cultured up to 24 h (C). The subsequent steps were different for the morphological estimation and for the quantitative evaluations of ADMSCs proteolytic activity, as detailed below.
2.2 Morphology of ADMSCs adhering on native or oxidized FITC collagen  difference in the morphology of the adsorbed proteins was found in this magnification. In the presence of cells, however, some discrete patterning of the adsorbed protein layer was observed, particularly stronger on native FITC Col I samples (C vs. G), suggesting a distinct proteolytic activity. On oxidized Col I samples only a homogenous accumulation of the protein beneath the cells was a typical find suggesting a rather missing proteolytic activity (C, D). Collectively these morphological data suggest that oxidized Col I substrates are barely remodelled by ADMSCs.

Quantitative measurement of Col I degradation
Col I was labeled with FITC according to Doyle 33 with some modifications (see Methods section), basically directed to strictly adjusting the pH of samples to pH 9 assuring maximal FITC binding. In these conditions part of the fluorophore become day-quenched (possessing lowered fluorescence because of the FRET effect) [32] and putatively de-quench (increasing the fluorescence) upon proteolytic degradation of the carrying protein. Indeed, as we found in a preliminary study (Fig 3) our FITC-collagen was dayquenched, increased its fluorescence in the presence of collagenase CH, thus providing opportunity for the quantitative studies on collagen proteolysis. It has to be noted here that upon adsorption both FITC-Col I and FITC-Col I-OXI presented a sufficiently bright fluorescent background supporting the morphological investigations shown above.
On the other hand, in a parallel experiment aiming to compare the degradation profiles of native and oxidized FITC-collagen using the same collagenase CH degradation system, we found that while the native FITC-Col I significantly de-quenches upon proteolytic digestion for 60 and 120 min (compared to time 0 min) in collagenase CH solution ( Figure 3). As shown on the next Figure 4, comparing RPU for native (blue line) and oxidized samples (green line). at 60 and 120 min, this effect was much less pronounced for the oxidized FITC-Col I probe.  to further confirm the significant increase of the fluorescence in the group of native FITC-Col I samples, but not in the oxidized ones, meaning that FITC-Col I-OXI is less susceptible to collagenases treatment. These data approve that our FITC-Col I preparations might be used for the quantification of ADMSCs proteolytic activity.

De-quenching of FITC-Col 1 by ADMSCs -effect of oxidation
According to the initial diagram (Fig 1)

DSC analysis of FITC-Col 1 -effect of oxidation
To evaluate the level of structural changes in the oxidized Col I molecules we investigated the effects of oxidation on the thermal stability of FITC-Col I by DSC analysis.
DSC measures the heat capacity of samples as a function of temperature and provides information about the thermal stability and putative structural changes in the molecule. DSC curves were used also to calculate the thermodynamic parameters, such as melting temperature (Tm), the total transition enthalpy (∆H total) and half-widths of transition (T ½).
As expected, the maximum of the heat absorption of native FITC-Col I was observed at 40.5°C (TM-main) ( Figure 6, Table 2). As a result of the oxidation, however, the thermogram results into splitting of two well-resolved transitions with melting temperatures at 33.6°C (TMpre) and 40.1°C (TM-main), respectively ( Table 2)., which confirms our previous investigation on calf skin collagen Type I suggesting certain changes in collagen structure upon oxidation [31].
To further analyse this observation, we compared the enthalpy and the half width of transition of native and oxidized samples ( Table 2). It should be noted that the value of total enthalpy (∆Htotal) after oxidation was very close to that of native collagen sample indicating only a discrete structural change in the collagen molecule upon oxidation. This was further confirmed by slightly higher half-width of transitions in oxidised sample as compared to native one.  * TM-main -temperature of the main transition; TM-pre -temperature of the additional pretransition event; ∆Htotal -total transition enthalpy;  TM-main½ -half-width of the main transition; TM-pre ½ -half-width of pre-transition.

Discussion
In vivo, cellular microenvironments are constructed of ECM proteins and proteoglycans where Col I play a major role, being under constant turnover to maintain tissue homeostasis [2,9,[14][15][16]. In most forms of cellular activity, the cells tend to remodel their adjacent microenvironment via mechanical reorganization and proteolytic degradation [34,35] [43]. In this respect, the implication of native collagen matrices is a challenging approach as collagen is a natural ECM protein with easy tuneable properties [40,44]. However, in vitro stem cells easily lose their self-renewal and multi-lineage differentiation potential during cell doubling.  [42,50]. Oxidative stress is generated when living cells are unable to neutralize excessive ROS, or are incapable to recycle the oxidized biomolecules.
Particularly the (over) oxidation of proline (enzymatic and non-enzymatic) was shown to contribute to fibrotic processes in heart and amyloid formation [51-53].
Cell-substratum interaction is a complex process that is bi-directional and dynamic, mimicking to a certain extent the physiological interaction of cells with the ECM.
Consequently, the adhering cells tend to rearrange adsorbed ECM components [15,[35][36][37][38]. To assure that cells attach exactly to collagen, as they may use also other adhesive proteins (like fibronectin, vitronectin, fibrinogen, etc.) [35,[36][37][38], we used collagen pre-coating followed by 2 h cell adhesion in a serum-free medium (Figure 1) before the serum was added (step B). Thus, we avoided the competitive effect of other serum proteins. In addition, the collagen was fluorescently labelled that made easier its morphological visualization as detailed above and the subsequent quantifications of cellular proteolytic activity. Related to our experimental conditions ADMSCs visibly recognize the native and oxidized Col I equally well, as evidenced by the lack of difference in both adhesion and overall cell morphology (Figure 2A and E).
However, a significant alteration in their ability to reorganize the oxidized FITC-Col I was observed ( Figure 2 F and G), which is basically a novel observation presumably reflecting an intricate stem cell behaviour in an oxidizing environment. As the reorganization is known to require cellular proteolytic activity, [15,36,54 ]a similar effect might explain the restoring capacity of protease inhibitors on the extracellular collagen fibril deposition in human MSCs 2.
The ECM however undergoes also proteolytic remodelling, which is a mechanism for the removal of the excess ECM, a process often approximated with remodelling [15]. It has to be noted, however, that cell-dependent ECM remodeling includes likewise the process of ECM organization and fibrils formation, which is critical for their functionality and for the interaction with other cells [35,36]. It is generally agreed that oxidative stress in vivo is characterized by an impaired ratio between lowered collagen synthesis and accelerated degradation [54]. The direct data on the degradability of oxidized collagen, however, are rather controversial. Chronic exposure to ROS causes an accumulation of damaged collagen and its fragmentation, which are more susceptible for proteolytic enzymes. Collagen crosslinking however, inhibits its degradation [55]. The disproportional collagen metabolism itself impairs cell-matrix interactions, which stimulates MMPs production [55][56]. Therefore, considered in a broader sense, the capacity of cells to repair or to replace ECM proteins following acute oxidation is likely to be an important predictor of how well cells are able to respond to oxidative stressors [54].
The quantitative aspects of the in-vitro cellular proteolysis are still insufficiently settled.
One approach to quantify such activity that we employed here was to measure the increasing An important observation concerning the proteolytic de-quenching of FITC-Col I is that it works well upon collagenase CH treatment giving significant rise of the total fluorescence. However, this de-quenching was substantially inhibited in oxidized FITC-Col I samples, meaning that oxidized collagen is more resistant to enzymatic degradation. We hypothesize that the effect of bacterial collagenase CH is connected with the specific binding in the active site and strong preference for glycine in P3 and P1′, proline at P2 and P2′ (according to proteases classification) in the cleavage site 58. It is notable that the proline is particularly vulnerable to oxidation by metal ion generated ROS and can be disproportionally modified by oxidation [59-62]. Thus, the specific cleavage site could be lost.
Unequivocally also sound the data for ADMSCs induced enzymatic remodelling of FITC-Col I, characterized by a significantly higher de-quenching of samples (+cells) compared to the samples (-cells). Interestingly, the fluorescence of native FITC-Col I increased with about 11 %, which is higher than the effect of CH collagenase (about 5%) suggesting an involvement of more active MMPs secreted by ADMSCs. On the other hand, the ability of ADMSCs to dequench the substratum associated FITC-Col I suggests that adsorbed collagen is sensitive to the peri-cellular proteolysis. The same tendency, however, was observed in the supernatants meaning that the proteolysis continued also in the medium, in fact targeting the spontaneously released collagen. It has to be noted, however, that the fluorescent signal in the medium is approx. 4 times higher than the signal measured from the substratum, meaning that only about 25 % of Col I remains there after 24 h of incubation, valid for both native and oxidized samples. The reason for this unexpectedly high spontaneous desorption of FITC-Col I might be attributed to the withdrawal of equilibrium between adsorbed and released Col I due to its initially low concentration in the medium (after washing) [63]. The competitive Vroman effect [64] of serum proteins also cannot be excluded assuming that 10% of serum was added to the system after ADMSCs adhesion (see Fig 1). The Vroman effect dictates that the protein of highest mobility adsorbs first on the substratum but later are replaced by less mobile proteins that have a higher affinity for the surface 64. The later might come from the serum, but also secreted by the cells. The competitive effect of some small amino acids like glycine is also an acceptable mechanism [65]. Nevertheless, this result leads to the assumption that the Nevertheless, our results raise the question: why oxidized collagen is less sensitive to remodeling, both mechanical and enzymatic? We anticipate that the process of collagenolysis depends on multiple interactions of mammalian collagenases with different exosites, serving to align the active site of collagenase which possesses a strong preference for the cleavage site perfectly matching the repetitive amino acid sequence of the native Col I molecules [72]. They are located near the peptide bond that will be cleaved, along with the local unwinding of the triple helix [72,73]. Any changes in the structure of collagen, as a result of oxidation, could prevent the proper aligning of collagenase and subsequently the degradation of type I collagen. Moreover, it was shown that the proline and hydroxyproline abundance in certain positions impacts the conformation of collagen molecule even its affinity to integrin receptors [74]. It may be a hint to understand the morphological observations for absent ADMSCs collagen reorganization in oxidized samples known to require integrin activity 15,18 . The missing difference in the overall cell morphology however suggests that such altered integrin activity is unlikely, which endorsed the view that some intrinsic structural changes in the collagen molecule upon oxidation are responsible for the effect.
In order to confirm the putative structural changes in Col I molecule, we performed DSC analysis comparing the native and oxidized samples ( Figure 6). In fact, this was a follow up study from our previous investigation on calf skin collagen denaturation profile [75] and the effect of oxidation [31]. Here we confirm that the native FITC collagen type I undergoes similar change in the thermal transition under heating, i.e. a single cooperative peak at 40.5 o C.
Thermal denaturation of oxidised collagen however resulted in a splitting of the main transition into two well-resolved transitions, i.e. along with the above typical collagen endotherm, a new transition at 33.6 °C appears. Interestingly, the enthalpy (∆H), which reflects the energetically aspect of the transition, was found to be similar to those of native FITC-Col I.
∆H is basically dependent on the fraction of native protein in the solution. [76] Consequently, if this fraction is less in the total protein, ∆H would drop down correspondingly [77], which is obviously not the case as we did not find such a significant change. To be noted that the transition half-widths (Figure 6 B) were still close to the native sample, which is an indication that the observed pretransition is caused by a rather discrete kind of damage to the collagen molecule. This allows us to speculate that upon oxidation in acute conditions collagen molecule undergoes mostly intrinsic reorganization causing a lowered transitional temperature and does not go into the denatured state [75,76]. This again points to the possibility that the low digestibility of collagen in an oxidizing environment depends on the distinct changes in the collagen structure, but not because of its denaturation.
Collectively, all these data lead us to believe that we have encountered a novel mechanism for the control of MSCs behaviour via subtle changes in their oxidative environment, presumably valid also for other cell systems.

Collagen preparation
Collagen type I (Col I) was isolated from rat tail by standard procedure combining acetic acid extraction and salting out with NaCl as described elsewhere [78][79]. Rat tails were

Fluorescent labeling of collagen
FITC labelled collagen was prepared according to the modified protocol of Doyle [33].
Briefly, 4 ml of Col I solution in 0.05M acetic acid (2.5 mg/ml) was titrated with 0.1M M borate buffer pH (9.0) and mixed with 50μL FITC dissolved in DMSO at a concentration of 1 mg/ml, then incubated at room temperature for 90 min in dark. The reaction was stopped by 0.05M Tris buffer (pH 7.4) and the excess of FITC was removed by intensive dialysis versus 0.05M acetic acid. Aliquots of FITC labelled Col I was stored at +4 0 C for up to 3 months.

Collagen oxidation procedure
FITC-Col I oxidation was performed according to the previously described protocol by incubating collagen solution (2 mg/ml) in 0.05M acetic acid, pH 4.3, with freshly prepared 50μM FeCl2 and 5mM H2O2 for 18 hours at room temperature [31]. The reaction was stopped with EDTA at a final concentration of 10mM. The excess of oxidants was removed by intensive dialysis against 0.05 M acetic acid.

Cells
Human ADMSCs of passage 1 were obtained from Tissue Bank BulGen using healthy

Statistical analysis
Data was analyzed by using SPSS Statistics for Windows, Version 23.0 (Armonk, NY: IBM Corp). All quantitative results were obtained from at least four samples for analyses.
Descriptive data was compared using Chi-Square and Mann-Whitney U tests. Comparison of differences between groups was performed by non-parametric test of Friedman; pair wise comparison was performed using post-hock analyses by Dunn-Bonferroni. Data were expressed as mean ± standard deviation (SD). Difference with p < 0.05 was considered to be statistically significant.

Conclusions
Both morphological and quantitative approaches demonstrate that the native Col I undergo significant remodelling by stem cells. A completely novel observation, however, is that the oxidized collagen in acute conditions cannot be remodelled by ADMSCs, both mechanically and enzymatically, confirmed quantitatively by the absent proteolytic dequenching of FITC-Col I. Parallel studies in the cell-free system show that the oxidative environment generally alters the enzymatic de-quenching of FITC-Col I. DSC analysis confirm that all these effects depend rather on the intrinsic structural changes of the oxidized Col I molecules than on the altered functionality of ADMSCs.
Institutional Ethics Committee Statement: The