Impact of binding mode of low-sulfated hyaluronan to 3D collagen matrices on its osteoinductive effect for human bone marrow stromal cells

Characterization of sHA-modified fibrillary 3D Col matrices

In the present study, five different fibrillary 3D Col matrices were investigated: Pure 3D Col matrix (Col; non-crosslinked, non-modified), Col with adsorptively bound sHA (Col sHA_ads), crosslinked pure Col (Col_EDC), Col_EDC with adsorptively bound (Col_EDC sHA_ads) and with covalently bound sHA (Col_EDC sHA_cov) (see Table 1 for detailed information). The modification of fibrillary 3D Col matrices with sHA was performed either in parallel with EDC (1-ethyl-3-(3-dimethylamino propyl)-carbodiimide)-crosslinking or sequentially in order to prepare a combinatorial set of matrices with and without crosslinking as well as with adsorptive and covalent binding of sHA.

Initially, the influence of crosslinking and sHA modification on characteristic features of fibrillary 3D Col matrices as pore size, collagen fibril diameter and elastic modulus were examined using a well-established toolbox based on confocal laser scanning microscopy (cLSM) and colloidal probe force spectroscopy, respectively (Kalbitzer et al. 2015). The pore size of the Col matrices is known to depend on Col concentration. Herein, all matrices had a mean pore size in the range 3–4 µm (Figure 1A). The diameter of collagen fibrils within 3D Col matrices was previously shown to depend on the pH value and ion content of the fibrillation solution (Kalbitzer and Pompe 2018). In the different 3D Col matrices, fibril diameter was in the range of 0.7 µm for all fibrillary 3D matrices (Figure 1B). Neither crosslinking nor sHA-modification did alter collagen fibril diameter. EDC-crosslinking increased the elastic modulus of 3D Col matrices significantly from about 80 Pa to about 150 Pa (Figure 1C). The modification with sHA had no influence on the respective elastic modulus. These results are in line with previously published Col I matrices reconstituted with these protocols (Kalbitzer and Pompe 2018).

Figure 1:

Characteristics of 3D fibrillary Col matrices. Fibrillary 3D Col matrices were characterized immediately after reconstitution and modification with sHA. The topological parameters pore size (A) and fibril diameter (B) were calculated from cLSM-images of TAMRA-stained matrices. Elastic modulus (C) was determined by colloidal probe force spectroscopy. Results are shown as geometric mean ×/÷ GSD; n = 3. Significant differences related to Col are indicated with ***p < 0.001 and *p < 0.05.

On the route from precursors to osteoblasts, from bone marrow niche to bone, hBMSC are in contact with different soft respectively stiff environmental conditions. With the 3D fibrillary Col matrices, the microenvironment of hBMSC in the bone marrow niche can be mimicked in vitro. The stiffness of bone marrow was shown with <300 Pa (Shin et al. 2013) to be within our experimental range, well below the stiffness of mineralized bone up to 10 GPa (Rho et al. 1993). We know from earlier studies, and verified herein, too, that the substrates were stiffened by the cells during osteogenic differentiation (Vogel et al. 2020) resulting in an elastic modulus of around 250–1,500 Pa, already indicating a transition towards calcified bone regions (Engler et al. 2006). In this context it is interesting to note, that a too high stiffness as provided by e.g. hydroxyapatite- or calcium phosphate-based materials counteracted osteogenic differentiation of hBMSC in vitro and disturbed osteogenesis in vivo (Gröninger et al. 2020; Vahabi et al. 2012). Furthermore, we can state the fibrillary structure of Col matrices reasonable compares to natural bone, as the dimension of collagen fibrils are reported in the range of 0.02–0.5 µm depending on (tissue) age, collagen maturation stage and remodelling activity (Siadat et al. 2021; Starborg et al. 2013). In such a comparison one has to consider the type of analysis and possible drying and crosslinking artefacts for instance in electron microscopy studies.

The successful and homogeneous binding of sHA to the Col matrices was illustrated by cLSM using Atto565-labeled sHA (Figure 2A, upper panel). A uniform distribution of sHA over the whole thickness (about 500 µm) of the 3D Col matrices and the arrangement of sHA along the Col fibrils was seen confirming previous findings (Figure 2A, lower panel) (Kalbitzer et al. 2015; Sapudom et al. 2017). The comparison of fluorescence intensity of labelled sHA imply a higher amount of bound sHA on EDC-crosslinked Col matrices. To verify this qualitative impression and to quantify the amount of bound sHA, 3D Col matrices prepared with Atto565-sHA derivative were digested with papain after 1 day and 22 days of incubation in PBS (phosphate-buffered saline) at 37 °C. The result of fluorimetric sHA quantitation is shown in Figure 2B. At day 22 Col matrices contained about 2 µg adsorptively bound sHA/matrix. Significant more sHA was found in crosslinked 3D Col matrices, Col_EDC. Here, the amount of adsorptively and covalently bound sHA was about 6 µg sHA/matrix. Moreover, it was observed that adsorptively bound sHA was released over the duration of 22 days. In contrast, for covalent attachment of sHA we did not observe sHA release.

Figure 2:

Distribution and quantitation of sHA in fibrillary 3D Col matrices. The distribution of Atto565-sHA within the 3D Col matrices was determined after 24 h of storage in PBS at 37 °C and is shown in representative cLSM fluorescence images (A). The upper panel visualizes Atto565-sHA (red fluorescence) and the middle panel shows the collagen network in grey (images were taken in the reflection mode [RM]); and the lower panel shows merged images; scale bar 50 µm. The amount of sHA that stably remained in fibrillary 3D Col matrices were determined after 22 days of storage in PBS at 37 °C. Panel (B) shows the results of Atto565-sHA quantitation by fluorimetry after papain digestion of 3D Col matrices for 2 h at 60 °C after 1 day and 22 days. The results are shown as mean ± SEM; n = 8, values were normalized to pure Col matrices. Statistically significant differences related to Col sHA_ads for each time point are indicated with ***p < 0.001.

From these results, we can state that sHA is constantly available in the Col matrices during the cell culture studies, however, at different amounts and different binding mode. For later comparison of the effects of bound with those of dissolved sHA, we can calculate the sHA per ml cell culture medium bound in the Col matrices with about 2 µg sHA/ml for Col sHA_ads, and roughly 6 µg sHA/ml for Col_EDC sHA_ads and Col_EDC sHA_cov.

It is interesting to discuss, that sHA adsorbed on Col_EDC in higher amounts at a high stability, almost comparable to covalently bound sHA. We account this effect to an altered isoelectric point (IEP) of Col on the surface of the fibrils after EDC crosslinking in comparison to non-crosslinked Col via EDC. It is known that IEP of Col tightly regulates GAG and sGAG adsorption (Kalbitzer et al. 2015). On the other hand non-specific alteration of crosslinking of adjacent amino and carboxyl groups of amino acids by EDC is prone to alter the IEP behaviour (Nair et al. 2020), while well-known positively charged clusters important for sGAG binding remain preserved (San Antonio et al. 1994).

In sum it can be concluded sHA-modified matrices exhibit important features of the in vivo bone marrow in terms of microstructure and mechanics and present sGAG on constant amounts and different binding modes for cells.

Influence of sHA-modified 3D Col matrices on TNAP activity (early osteogenic differentiation marker)

To confirm equal seeding density and cell number on all five 3D Col matrices a MTS assay was performed first. This assay measures metabolic activity as an indication for cell number and viability. Supplementary Figure S1 shows that from 2 h after seeding (Supplementary Figure S2A) until day 4 (Supplementary Figure S2B) the cell number increased. Significant differences comparing the different 3D Col matrices were not evident. From a recent study it is known that about 25% of seeded hBMSC are invasive and started within 24 h to migrate into the 3D Col matrices; the invasion depth of 120–150 µm after 22 days only slightly depended on cross-linking status (Vogel et al. 2020).

For evaluation the osteoinductive potential of sHA modified 3D Col matrices, TNAP enzyme activity as an early osteoblast marker was analysed in hBMSC that were cultured on the different matrices for 11 days. On all five 3D Col matrices, hBMSC showed TNAP activity seen as deep purple signals as result of activity staining (Figure 3A). The microscopic analysis allowed only a qualitative evaluation as it is difficult to deduce total TNAP activity from dye intensity, in particular because the cells on the surface of Col_EDC sHA_cov slightly tend to aggregate. Therefore, TNAP activity was quantified after lysis of the entire matrices with an enzyme activity test. TNAP activity was significantly increased on all sHA-modified 3D Col matrices compared to pure 3D Col matrix (Figure 3B, left part).

Figure 3:

TNAP activity of hBMSC in fibrillary 3D Col matrices. hBMSC were cultured in OM/D on the different matrices for 11 days. Panel (A) shows representative images for TNAP activity staining (TNAP⁺-cells in purple); scale bar 200 µm. The activity of TNAP was quantified with an enzyme activity assay using p-nitrophenylphosphate as a substrate. (B) TNAP activity was determined in the different 3D Col matrices (left part) and with 3D Col matrices that were treated with 3 (middle part) and 10 µg (right part) sHA/ml. The bars show the total TNAP activity that were corrected by cell-free blanks and are given in percent related to Col (=100%). The additional effect that is achieved by the added dissolved sHA is shown in the white bars (B, middle and right part) with overlaid results from samples without extra sHA (B, left). The results are given as mean ± SEM; n = 4. Significant significances related to 3D Col matrix (left part) are indicated with **p < 0.01 and ***p < 0.001; significant differences of 3D Col matrices with added sHA (middle and right part) related to respective 3D Col matrices (left part) are indicated with ^## p < 0.01 and ^### p < 0.001; n.s. indicates no significant differences.

Roughly TNAP activity correlated with the amount of matrix-bound sHA (Figure 2). However, for Col_EDC matrices a slightly lower TNAP activity found with covalently attached sHA although the amount of sHA in the matrix was slightly higher compared to adsorptively attached sHA. This deviations were minor and might be attributed to differences in availability, as adsorptively bound sHA might be in part taken up intracellularly by the cells. In this context it is worthwhile to examine possible release and interaction pathways of matrix-bound sHA in comparison to dissolved sHA. Hence, hBMSC that are cultured with the different modified 3D Col matrices were treated with a solution of 3 and 10 µg of dissolved sHA/ml (the amounts were chosen slightly higher as the respective amounts of matrix-bound sHA). By adding dissolved sHA to hBMSC that were cultured with pure 3D Col matrices (non-crosslinked and crosslinked), sHA induced a clear, significant increase of TNAP activity (Figure 3B, middle and right). By adding dissolved sHA to hBMSC that were cultured with sHA-modified 3D Col matrices, sHA induced a slight, almost not significant increase of TNAP activity that was independently of the previous binding mode of sHA. For 10 µg sHA/ml solution no additional effect (compared to 3 µg sHA/ml) was observed pointing towards a saturation, as on both types of matrices already a high amount of sHA is present (Figure 2). In contrast, for Col matrix samples with lower amounts of sHA (Col sHA_ads) or no sHA (Col, Col_EDC,) present a dose-dependent increase in TNAP activity by dissolved sHA was seen.

The majority of hBMSC reside on the top of fibrillary 3D Col matrices (Vogel et al. 2020) so that the sHA amount bound in the entire 3D Col matrices will only partially be accessible for the hBMSC, hence, the amount of bound sHA has to be taken as the maximum available amount when comparing to the dissolved sHA. Degradation of non-crosslinked 3D Col by hBMSC was previously reported (Vogel et al. 2020), hence, Col I degradation and sHA release has to be assumed for Col sHA_ads samples. However, one should keep in mind, that putatively released adsorptive sHA could be available as dissolved sHA at maximum with concentrations of a few µg/ml in cell culture medium see Figure 2B and corresponding discussion. In 3D Col_EDC sHA_cov matrices, the EDC crosslinks between amino acid side groups (carboxyl and amino groups) of tropocollagen molecules as well as to sHA provided a high stability and resistance against enzymatic degradation (Slusarewicz et al. 2010), so that 3D Col_EDC matrices itself were not degraded by hBMSC (Vogel et al. 2020) and sHA release is not reasonable to be assumed. TNAP activity, however, increased with all sHA-modified 3D Col matrices independently of crosslinking and binding mode of sHA. In sum, we can conclude that matrix-bound sHA is at least similarly effective as dissolved sHA, which strongly suggests an extracellular mechanism for sHA-induced TNAP increase (see also discussion further below).

In order to support our arguments on the impact of extracellular sHA, hBMSC were cultured on Atto565-sHA-modified 3D Col matrices for 24 h and stained with calcein before cLSM analysis. For 3D Col_EDC sHA_cov a much lower intracellular uptake of sHA (red fluorescence) into the cell lumen, visualized by the calcein AM staining (green fluorescence), was found (Figure 4B) in comparison to Col_EDC sHA_ads (Figure 4A). For the latter the cells were able to internalize sHA in high amounts with many sHA clusters inside the cells. In contrast, for covalently bound sHA (Col_EDC sHA_cov) sHA was almost absent from the cell lumen, while tight interactions of sHA covered Col fibrils with the cell membrane were observed.

Figure 4:

cLSM images of hBMSC on Atto565-sHA modified 3D Col matrices. hBMSC that were cultured for 24 h in BM on Col_EDC sHA_ads (A) and Col_EDC sHA_cov, (B) prepared with Atto565sHA (red). The cells were stained with calcein (green) to visualize the cell lumen. cLSM images (top view – xy, upper panel; side view – z, lower panel) of calcein AM-stained hBMSC on matrices show that adsorptive bound Atto565-sHA (A) can be internalized by the cells, whereas covalently bound Atto565-sHA (B) remains outside. Nuclei were stained with DAPI (blue fluorescence); scale bars 50 µm (top view) and 20 µm (side view).

Influence of sHA-modified 3D Col matrices on calcium phosphate deposition (late osteogenic differentiation marker)

Following the findings on TNAP activity, the deposition of calcium phosphate by hBMSC as a late functional osteoblast characteristic after 22 days in culture was analysed. Both parameters correlate to each other: The phosphate which is released by TNAP activity is mandatory for cell-associated calcium phosphate deposition (Hempel et al. 2014).

Fluorescence-based Osteoimage™ assay indicated the presence of calcium phosphate deposits as green fluorescent signals on all 3D Col matrices Figure 5A (upper panel). The mineral deposits are found along the collagen network structure as seen in the lower panel (Figure 5A) visualizing the collagen network by reflection mode images. Other than with von Kossa or Alizarin S staining method (both common for calcium phosphate staining), the Osteoimage™ assay gave no background signals in cell-free control samples (Supplementary Figure S2).

Figure 5:

Calcium phosphate deposition by hBMSC in fibrillary 3D Col matrices. hBMSC that were cultured in OM/D on the different matrices for 22 days were stained with the Osteoimage™ assay as described. Calcium phosphate deposition within the matrices is seen as green fluorescence (A, upper panel); the collagen network (RM; reflection mode) is shown in grey (A, lower panel); scale bar 100 µm. (B) The elastic modulus of cell-laden 3D Col matrices was determined after 22 days in OM/D as described (see Figure 1 and the materials and methods section). The overlaid black/grey bars illustrate the initial stiffness values (see also Figure 1), the white bars the osteoinductive effects by the cells. The results are given as geometric mean ×/÷ GSD; n = 4. Significant differences related to 3D Col matrix are indicated with ***p < 0.001.

Furthermore, an increase in matrix stiffness was measured by force spectroscopy as congruent marker of osteogenesis (Figure 5B). This functional determination of osteogenesis in terms of 3D Col matrix stiffness increase caused by calcium phosphate deposition within the 3D Col matrices revealed similar results as the early marker analysis of TNAP activity. While the initial stiffness differences by EDC crosslinking seem to be minor, a stronger increase in matrix stiffness correlated with the presence of matrix-bound sHA and TNAP activity. Therefore, we argue that the stiffening of 3D Col matrices during osteogenic differentiation is not the reason but the result of sHA-supported mineral accumulation. It has to be noted, that an additional impact of sHA modification in dependence on binding mode of sHA was observed. While adsorptive sHA presentation at low and high amount stimulated an additional stiffness increase of the 3D Col matrices, the covalently bound sHA outperformed this stimulation. This finding again supports our above argumentation that a permanent extracellular presentation of sHA supports osteogenic differentiation of hBMSC.

Formation of ECM proteins by hBMSC on 3D Col matrices

As known from recent proteomics studies and in silico molecular modelling experiments, sHA affects the composition, assembly and remodelling of hBMSC- and osteoblast-derived ECM by electrostatic interaction with several ECM proteins (Ruiz-Gómez et al. 2019; Vogel et al. 2016). Many identified and significantly upregulated proteins play a crucial role for bone formation and regeneration as known for FN (Chatakun et al. 2014; Di Benedetto et al. 2015; Mathews et al. 2012; Raitman et al. 2017) and THSP1 (Amend et al. 2015; Shi et al. 2013). This was the motivation to investigate whether hBMSC, cultured on different 3D Col matrices, are able to enrich the matrices with endogenous formed ECM proteins and to ask whether the binding mode of sHA to 3D Col matrices has any effect on ECM protein synthesis. The formation of FN and THSP1 was analysed by Western blotting after culture of hBMSC on the 3D Col matrices for 8 days. The results of densitometric analyses of Western blots (Figure 6A) showed an increase of FN (Figure 6B) and THSP1 protein amount (Figure 6C) in the presence of sHA relative to Col matrices without sHA-modification. The strongest (about three-fold) induction of FN was seen for hBMSC cultured on Col sHA_ads. The induction of THSP1 was in the range of 1.5- to 2.5-fold with significant changes for Col sHA_ads and Col_EDC sHA_cov.

Figure 6:

Formation of fibronectin (FN) and thrombospondin-1 (THSP1) by hBMSC in fibrillary 3D Col matrices. hBMSC were cultured on the different matrices for eight days. Afterwards, the cells were lysed, and the lysates were applied to Western blot analysis using specific anti-FN- and anti-THSP1-1-antibodies. The membranes were re-blotted with antibodies against GAPDH and actin as internal loading controls. A representative Western blot is shown in (A). The densitometric evaluation of three independent Western blots is shown for FN (B) and THSP1 (C). The results, related to GAPDH and actin, are given as mean ± SEM in percent related to non-modified Col matrices (dotted line), n = 3. Significant differences versus controls are indicated with *p < 0.05 and **p < 0.01.

Discussion on mode of action of osteoinductive effect of sHA

The exact mechanism how sHA derivatives assert their osteoinductive effects is still not fully understood, however, from recent studies two modes of action can be hypothesized. For an extracellular mode of action, the interaction of sGAG, in particular of sHA with ECM components based on electrostatic interaction alters the assembly and fine structure of the ECM (Kii 2019; Raitmann et al. 2018; Wang et al. 2004) and in consequence outside-in signalling events mediated by the ECM. Furthermore, growth factor binding at matrix-bound sHA and differential receptor activation can be expected. The pro for this is that sulfated GAG and in particular sHA influences ECM composition and assembly as seen for FN, THSP, periostin and several extracellular enzymes as transglutaminase-2, lysyloxidase, matrixmetalloproteinase-2 and further the interaction of TIMP3 with its endocytosis receptor LRP1 (Kii 2019; Rother et al. 2016; Troeberg et al. 2014; Wang et al. 2004). Proteomics studies indicated that diverse integrins are influenced by sHA (Kliemt et al. 2013), but there was no indication that CD44 or other hyaluronan-relevant receptors as HARE/stabilin-2 (hyaluronan receptor for endocytosis), RHAMM (hyaluronan-mediated motility receptor) or LYVE1/XLKD1 (lymphatic vessel endothelial hyaluronan receptor 1/extracellular link domain containing 1) are involved in sHA action. An up-regulation of CD44 was found only with chondroitin sulfate (Schmidt et al. 2016). The second hypothesis assumes an intracellular action of sHA whereupon an uptake of sHA could take place by endocytosis or via organic anion transporters (OATP, Schneider et al. 2015). This idea is supported by sHA-induced regulation of Rab proteins (involved in endocytic processes; Kliemt et al. 2013) and by the fact that sHA interacts with many intracellular proteins as shown in a new proteomics study (Großkopf et al. submitted for publication). Here it was seen that once in the cell, GAG have many interaction partners in cytoplasm, nucleus, Golgi apparatus, and in cell membrane that are involved in differentiation process. The results of the present study support the first hypothesis of an extracellular mode of action, but did not exclude an additional involvement of intracellular mechanisms.

Summary

3D fibrillary Col matrices are a suitable in vitro cell culture system to study osteogenic differentiation of hBMSC. For the present study the matrices were modified with low-sulfated hyaluronan that previously revealed osteoinductive power when applied in solution or in mixture with collagen as 2D artificial ECM. The study aimed on examining the impact of presentation mode of sHA in terms of adsorptive versus covalent binding in biomimetic 3D Col matrices, which exhibit similar characteristics as the bone marrow in terms of pore size, fibril diameter, and elastic modulus. We found adsorptively bound sHA to be more strongly internalized by the cells whereas vanishing uptake of sHA was seen when sHA was covalently bound. Both adsorptively and covalently bound sHA induced an increase of early (TNAP activity) and late (calcium phosphate deposition, 3D matrix stiffening) osteogenic differentiation markers. While sHA effects on osteogenic differentiation of hBMSC are known and again are shown herein for soluble and adsorptively bound sHA, the novelty of the present study is that covalently bound sHA with almost no uptake by cells induced similar strong osteogenic effects providing strong arguments for an extracellular mode of action of sHA in osteoinduction. Upcoming in-depth studies will shed light on the influence of sHA on signal transduction events leading to accelerated osteogenic differentiation (effects on TNAP and mineralization).

Altogether fibrillary 3D Col matrices are demonstrated as a very helpful tool to study hBMSC differentiation in vitro by mimicking different physiological and biomedical relevant states and add information on the mechanisms of osteoinductive effects of sHA on hBMSC.

Preparation and modification of fibrillary 3D Col matrices

Col matrices were prepared on poly(styrene-alt-maleic anhydride) (PSMA)-functionalized glass cover slips (d = 13 mm, VWR) in 24-well plates (TCPS), Greiner Bio one) as previously described (Vogel et al. 2020). 40 µl of a 2.5 mg/ml Col solution (4.89 mg/ml isolated from rat tail (Corning, New York)/ml in 20 mM acetic acid and 250 mM phosphate buffer (pH value was adjusted to 7.5 with NaOH) were placed on a cover slip. The fibrillogenesis was performed at 37 °C for 75 min (95% relative humidity, 5% CO₂). With this approach, 3D matrices of 400–500 µm thickness were generated with reproducible pore size and fibril diameter (Kalbitzer and Pompe 2018; Sapudom et al. 2019).

1-ethyl-3-(3-dimethylamino propyl)-carbodiimide (EDC; Carbolution) crosslinking of 3D Col matrices was performed by incubation with freshly prepared EDC solution (20 mM in 0.1 M 2-(N-morpholino) ethanesulfonic acid buffer (MES buffer), pH 5) at 37 °C for 2 h. Matrix stiffness was increased roughly twice by crosslinking.

sHA was bound either adsorptively or covalently to Col matrices. For adsorptive binding, Col matrices were incubated with sHA (100 µg/ml in MES buffer) for 4 h at 37 °C. For covalent binding, Col matrices were at first incubated with sHA for 2 h. Immediate after removal of sHA solution, a freshly prepared EDC solution was added followed by an incubation step for 2 h at 37 °C. Nomenclature and properties of the resulting 3D Col matrices are summarized in Table 1. It has to be noted, that previous experiments with a wide range of sGAG concentrations (0.05–1 mg/ml) during binding to Col matrices showed that concentrations > 0.1 mg sGAG/ml did not lead to a higher amount of stably bound GAG after 1 day, but to a higher release during the first hours (Kalbitzer et al. 2015). Therefore, we used 0.1 mg/ml in the experiments herein, too.

All steps were performed sterile under the laminar flow. Col matrices were prepared in 24-well plates and stored in PBS at 4 °C until further analysis or in vitro experiments. Prior seeding of hBMSC, matrices were equilibrated overnight at 37 °C in basal medium (BM = DMEM (Biochrom, Berlin, Germany) with 10% heat-inactivated fetal calf serum (FCS, Th. Geyer, Hamburg, Germany), 20 U penicillin, and 20 µg streptomycin/ml (both from Biochrom).

Characterization of fibrillary 3D Col matrices

Col matrices were stained with 5-carboxytetramethylrhodamine (TAMRA) for topology analysis as previously described (Kalbitzer et al. 2015). For that purpose, Col matrices were incubated with TAMRA solution for 1 h at room temperature and washed with PBS. Images of Col matrices were taken using a confocal laser scanning microscope (cLSM) LSM700 (Zeiss) equipped with a 63× water objective. Images were acquired with 8-bit colour-depth, 1024 × 1024 pixels in resolution and a vertical stack size of 20 images at 5 µm distance (equivalent to 100 µm). The voxel size of the acquired images was 0.1 (x) × 0.1 (y) × 0.5 (z) µm. Pore size and fibril diameter were analysed using the Matlab protocol described previously with three images per sample (Kalbitzer and Pompe 2018).

For micromechanical characterization of Col matrices, the elastic modulus was determined by colloidal probe force spectroscopy as described (Kalbitzer et al. 2015). The force-distance curves were measured with a Nanowizard III (JPK Instruments, Berlin, Germany) and a 50 µm glass bead in diameter (Polysciences, Eppelheim, Germany) attached to a MLC triangular cantilever with a nominal spring constant of 60 nN/m (Bruker AFM, Camarillo, USO). A minimum 75 force-distance curves at three positions of each matrix in three independent experiments with an indentation of 3–7 µm were analysed for each sample.

Determination and distribution of sHA in fibrillary 3D Col matrices

The quantification of the amount of bound sHA in Col matrices was performed after 1 day and 22 days of incubation in PBS at 37 °C by fluorimetry of Atto565-labeled sHA as described (Kalbitzer et al. 2015). Briefly, sHA-modified fibrillary 3D Col matrices were prepared as described earlier but with Atto565-labelled sHA instead of sHA. The matrices were digested using a papain solution (0.02 mg papain from papaya latex/ml, 10 mM EDTA, 5 mM l-cysteine in 5× PBS) for 2 h at 60 °C. After digestion, solutions were analysed for fluorescence intensity at λ_ex = 535 nm and λ_em = 590 nm in a 96-well microplate reader (Infinite 200 PRO, Tecan). The fluorescence intensity was related to a standard curve (r > 0.99) obtained with Atto565-sHA. Values were normalized to pure Col matrices.

Using cLSM (LSM700, Carl Zeiss), the distribution of Atto565-labeled sHA derivative within Col matrices was visualized. Images were gathered with a 40×/NA1.2 water immersion objective (Carl Zeiss) in fluorescence mode for Atto565-sHA and in reflection mode for collagen network. Obtained images were 16 bit colour-depth with a resolution of 1024 × 1024 pixels and with a vertical stack size of 11 images (equivalent to 50 µm). The voxel width in xyz dimension was 0.21 × 0.21 × 5 µm.

In vitro-experiments with human bone marrow stromal cells (hBMSC) in 3D Col matrices

Human bone marrow stromal cells (hBMSC) were isolated from bone marrow aspirates collected from healthy Caucasian donors (males and females, average age 32 ± 5 years) at the Dresden Bone Marrow Transplantation Centre of the University Hospital Carl Gustav Carus. The study was approved by the local ethics commission (EK263122004, EV466112016). hBMSC were isolated as described (Oswald et al. 2004) and cell preparations of individual donors were not pooled and used in passage 3 to 5.

For the in vitro experiments, 7000 hBMSC/cm² were plated in basal medium (BM = DMEM with 10% heat-inactivated fetal calf serum and antibiotics) onto Col matrices (placed in 24-well plates). Medium was changed twice a week. To induce osteogenic differentiation, BM was replaced by osteogenic differentiation medium (OM/D = BM with 10 nM dexamethasone, 300 µM ascorbic acid, 10 mM β-glycerophosphate (Pittenger 2008) at day 4 after seeding.

Cell analyses were performed 2 h and four days after seeding (MTS assay), 24 h after seeding (calcein staining), at day 8 (Western Blot), at day 11 (TNAP activity), and at day 22 (calcium phosphate staining).

Metabolic activity

hBMSC were cultured for 2 h and 4 days onto five different fibrillary 3D Col matrix modifications. Metabolic activity was determined by MTS ([3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium inner salt) assay (CellTiter 96^® AQ_ueous One Solution Cell Proliferation Assay; Promega, Walldorf, Germany) as described (Vogel et al. 2020). Cell-laden matrices were incubated with 10% MTS solution for 2 h at 37 °C. The resulting MTS formazan was then quantified photometrically at 490 nm.

Fluorescence staining

hBMSC were cultured for 24 h onto Atto565-sHA-modified Col matrices and incubated with calcein AM (0.5 µl of a 4 mM stock solution/ml of PBS) for 30 min at 37 °C according to manufacturer’s instructions (LIVE/DEATH assay kit, Invitrogen, via ThermoFisher, Dreieich, Germany). The cell-permeable calcein AM dye is converted to green-fluorescent calcein by ubiquitous cellular esterases. cLSM analyses was performed using a cLSM Leica SP8 FALCON with a 63×/NA 1.2 water immersion objective in sequential image mode for calcein AM and Atto556 signal to minimize signal bleed through.

For visualization of calcium phosphate deposition, a fluorescence-based staining kit (Osteoimage™, Lonza, Switzerland) was used. hBMSC were cultured on Col matrices for 22 days in OM/D and fixed with 4% (w/v) PFA for 10 min. Staining with Osteoimage™ was performed according to the manufacturer’s instructions.

Determination of TNAP activity

TNAP enzyme activity was determined in cell lysates (lysis buffer: 1.5 M Tris–HCl, 1 mM ZnCl₂, 1 mM MgCl₂, 1% Triton X-100, pH 10) with p-nitrophenylphosphate as a substrate (Hempel et al. 2014). TNAP activity was calculated from a linear calibration curve (r > 0.99) prepared with p-nitrophenolate and is given in mU.

Western blot analyses

Western blot analysis was performed as previously described (Ruiz-Gómez et al. 2019). The SDS-PAGE (4% stacking gel, 5–15% resolving gel) was run and the proteins were transferred to nitrocellulose membranes (GE Healthcare, Freiburg Germany) by semi-dry blotting. Fibronectin (FN) and thrombospondin-1 (THSP1) proteins were analysed on one blot by separate incubation with specific antibodies (see above) followed by immunoreaction with horseradish peroxidase (HRP)-conjugated anti-IgG-antibodies (CST, via New England Biolabs, Frankfurt, Germany). Visualization of immune complex was performed by enhanced chemiluminescence detection (GE Healthcare) using a CCD camera system (MF-ChemiBIS1.6 via Biostep Jahnsdorf, Germany). For densitometric evaluation ImageQuant TL software (GE Healthcare) was used.

Statistics

The in vitro-experiments were performed with hBMSC from different (biological independent) donors each in duplicate. The number of replicates is indicated in the figure legends. Results are presented as mean ± SEM or geometric mean ×/÷ GSD (for colloidal force spectroscopy). Significant differences sets were calculated using Graph Pad Prism software release 8 using One-way ANOVA followed by Tukey post-test. p-values< 0.05 were defined as significant and are indicated in the diagrams.