Controlling osteoblast morphology and proliferation via surface micro-topographies of implant biomaterials

Current research on surface modifications has yielded advanced implant biomaterials. Various implant surface modifications have been shown to be promising in improving bone target cell response, but more comprehensive studies whether certain implant surface modifications can directly target cell behavioural features such as morphogenesis and proliferation are needed. Here, we studied the response of primary alveolar bone cells on various implant surface modifications in terms of osteoblast morphology and proliferation in vitro. Analyses of surface modifications led to surface-related test parameters including the topographical parameters micro-roughness, texture aspect and surface enlargement as well as the physicochemical parameter surface wettability. We compared osteoblast morphology and proliferation towards the above-mentioned parameters and found that texture aspect and surface enlargement but not surface roughness or wettability exhibited significant impact on osteoblast morphology and proliferation. Detailed analysis revealed osteoblast proliferation as a function of cell morphology, substantiated by an osteoblast size- and morphology-dependent increase in mitotic activity. These findings show that implant surface topography controls cell behavioural morphology and subsequently cell proliferation, thereby opening the road for cell instructive biomaterials.


Results
Surface characterization. In order to characterize the differently modified zirconia surfaces with respect to their topographical and physicochemical properties, we employed scanning electron microscopy (SEM) and interferometry (IFM) to visualize and quantitatively grasp the surface topography, performed EDX analysis to describe the chemical composition and contact angle measurement to characterize the wettability of the biomaterial surfaces. SEM analysis revealed that the zirconia discs differed considerably in their surface structure, depending on the surface treatment ( Fig. 1a-f). Y-TZP surfaces that were not further processed after sintering showed homogeneously distributed grainy peaks on the micrometre level (Fig. 1a, SEM), whereas on the macroscopic level the surfaces were crossed by shallow grooves (Fig. 1a, insert, IFM reconstructed image). By contrast, surfaces in this group additionally coated with Ce-TZP showed a dense and grainy structure with undulated depressions (Fig. 1b). Coating of Ce-TZP samples with CaP resulted in a thin layer with net-like structures, which was homogeneously distributed over the zirconia-based surfaces (Fig. 1c) and which was not present on Y-TZP + CaP surfaces. The formation of the net-like structures on Ce-TZP surfaces can be explained by the combination of the coating technique and the substrate used. A suspension with CaP powder was sprayed on the surfaces of the sintered Ce-TZP substrates. When the sprayed droplets dried on the substrate surface, the CaP particles rearranged and accumulated at the edge of the droplet where a thin circular wall of particles was formed. The net-like structure was then formed by randomly adding additional overlapping walls from other droplets. In contrast, the Y-TZP substrate was not sintered and the porous nature of the substrate removed the water from the droplets by absorption. Thus, there was no rearrangement of the CaP particles during drying on the Y-TZP surface, which contributed to an even distribution of the CaP particles. Y-and Ce-TZP discs, which had been grinded after sintering, exhibited smooth surface characteristics at the low micrometre level, and the surfaces were divided into little rhombi by straight parallel grooves and protrusions (Fig. 1d). Regarding the blasted materials, SEM micrographs revealed that the topography of the only-sintered Y-TZP surface was changed from a surface with grainy peaks to a surface with smoother and denser characteristics, which was devoid of aforementioned structures, by blasting (compare Fig. 1a,e). On the contrary, the micro-texture of the blasted Ce-TZP surface hardly changed when compared with untreated Ce-TZP (compare Fig. 1b,f). On the  Quantitative surface characterization of zirconia biomaterials. Surface topography was evaluated by interferometry. Surface parameters describing the topography of implant surfaces were (a) S a (average surface height deviation amplitude), (b) S q (root-mean-square deviation), (c) S z (ten-point height of surface topography), (d) S sk (skewness), (e) S ds (density of summits), (f) S tr (texture aspect ratio) and (g) S dr (surface enlargement compared to a totally flat reference area). Data show mean values of three different areas per specimen ± SEM (n = 3). Statistically significant differences (p < 0.05, Tukey's HSD test) were marked with "a" if a Ce-TZP surface differed statistically significant from the corresponding Y-TZP surface, with "b" if Y-TZP + CaP, Y-TZPgr or Y-TZPsa differed from Y-TZP or if Ce-TZP + CaP, Ce-TZPgr or Ce-TZPsa differed from Ce-TZP, with "c" if Y-TZPgr or Y-TZPsa differed from Y-TZP + CaP or if Ce-TZPgr or Ce-TZPsa differed from Ce-TZP + CaP and with "d" if Y-TZPsa differed from Y-TZPgr or if Ce-TZPsa differed from Ce-TZPgr. surfaces suggest that the additional thin Ce-TZP layer filled the surface irregularities on the bulk Y-TZP discs after coating. By contrast, grinding resulted in Y-and Ce-TZP surfaces with similar topographical properties, which were characterized by a low surface enlargement of about 6% and a low peak density with S ds = 0.11/µm 2 (Fig. 2e). With regard to the latter parameter, blasting of Y-TZP exerted a similar smoothening effect, concerning the peak density, thereby yielding exactly the same S ds values of 0.11/µm 2 as determined for the grinded surfaces. Average surface roughness (S a and S q ) (Fig. 2a,b) and ten-point height of the surface (S z ) (Fig. 2c) were less different between the material groups, with the exception of sandblasted Y-TZPsa. As already mentioned above, sandblasting of Y-TZP produced large height deviations, which were here reflected by the significantly higher amplitude parameters S a , S q and S z for Y-TZPsa, when compared to the other materials. The negative skewness of all surfaces as measured by S sk (Fig. 2d) disclosed that all surfaces exhibited small peaks with comparatively deep and narrow valleys 36 . Based on the data obtained from SEM and IFM analysis it can be summarized that the applied surface treatments created distinct surface topographies, mainly related to the texture pattern (S tr ) and surface enlargement (S dr ). In detail, (i) coating of the bulk Y-TZP material with Ce-TZP and Ce-TZP + CaP caused a levelling of surface irregularities and hereby a reduction of the surface enlargement. (ii) Grinding yielded highly micro-anisotropic surfaces with smooth properties on the micrometre level in conjunction with a low degree of surface enlargement, whereas (iii) blasting increased the roughness, percentage of surface enlargement and isotropic properties of the treated surfaces. Analysis of the chemical surface composition by EDX (Table 1) showed that the main components of all test surfaces were zirconium with 43-64 mass% and oxygen with 26-35 mass%. Furthermore, 4-6 mass% carbon were found on all surfaces. As was to be expected, the Y-TZP surfaces additionally consisted of 4-4.6% yttrium. Furthermore, sandblasted Y-TZPsa surfaces contained 2.46 mass% aluminium while on all other Y-TZP surfaces only traces of aluminium could be found. This indicates that the detected aluminium on the Y-TZPsa originated from the blasted aluminium oxide particles and thus represents residues from the surface modification process. The Ce-TZP surfaces consisted of a ceria-stabilized zirconia-alumina-strontium hexa-aluminate composite and, accordingly, exhibited in addition to aforementioned zirconium, oxygen and carbon, 6.6-8.7% ceria and 3.3-4.7% aluminium, whereas strontium was found inconstantly in traces on the samples (not specified in Table 1). Calcium (Ca) and phosphorus (P) were exclusively detected on calcium-phosphate coated surfaces at a Ca/P ratio of 1.8-1.9, or more specifically, 4.26 mass% Ca and 2.36 mass% P for Y-TZP + CaP and 7.35 mass% Ca and 3.83 mass% P for Ce-TZP + CaP. All other material groups displayed only traces of phosphorus on their surfaces. Regarding surface wettability, contact angle measurement revealed that with exception of Y-TZP and Y-TZPsa, all test surfaces displayed a contact angle below 90° and thus can be classified as hydrophilic biomaterials 37 (Table 1). Within the hydrophilic biomaterial group the measured contact angles varied between 51.25° for the CaP-coated Ce-TZP to 85.93° for Ce-TZPsa. As surface wettability of implant materials is determined by the chemical composition and by structural surface features, the different hydrophilic properties of the test materials in the present work may originate from the various combinations of microtopography and surface chemistry.
Cell morphology. Morphology of AO was analysed by SEM and fluorescence microscopy of red labelled actin cytoskeleton at days 1 and 7. In order to describe and quantify the cell morphology on the different bioma- Table 1. (Physico-)chemical characterization of zirconia biomaterials. Elemental composition of the zirconia samples in mass% was measured by EDX and contact angles were calculated as a measure for surface hydrophilicity. Data show mean values ± SEM (n = 3 per group for EDX analysis and n = 10 per group for contact angle measurement). Statistically significant differences (p < 0.05, Tukey's HSD test) were marked with "a" if a Ce-TZP surface differed statistically significant from the corresponding Y-TZP surface, with "b" if Y-TZP + CaP, Y-TZPgr or Y-TZPsa differed from Y-TZP or if Ce-TZP + CaP, Ce-TZPgr or Ce-TZPsa differed from Ce-TZP, with "c" if Y-TZPgr or Y-TZPsa differed from Y-TZP + CaP or if Ce-TZPgr or Ce-TZPsa differed from Ce-TZP + CaP and with "d" if Y-TZPsa differed from Y-TZPgr or if Ce-TZPsa differed from Ce-TZPgr. www.nature.com/scientificreports/ terial surfaces, we measured the cell area, perimeter and the major and minor axis of the cell body. In addition, we calculated the hybrid shape parameters roundness, circularity and aspect ratio. As shown in Fig. 3, AO grown on Y-/Ce-TZP for one day displayed an elongated spindle-shaped morphology and actin fluorescence mainly located at the apical cell borders (Fig. 3a, arrow heads), while actin stress fibres were absent. A similar situation was observed for AO on CaP-coated and sandblasted surfaces (Fig. 3b,d). By contrast, AO on grinded surfaces appeared considerably elongated and flattened, with clearly visible actin stress fibres mainly aligned parallel to the cells' longitudinal axis within the entire cell body (Fig. 3c, SEM micrograph and fluorescence image, arrow heads).
The quantitative evaluation of the cell morphology, summarized in Fig  remained fairly constant and cell area (9,150 and 12,112 µm 2 ) even increased, thus suggesting that the elongated and spread cell shape persisted on these surfaces. In contrast, the other surfaces showed a stronger modulation of the aforementioned parameters at the analysed time points. In detail, on Y-/Ce-TZP and Y-/Ce-TZPsa cell area and perimeter increased, whereas circularity decreased from day 1 to day 7, roundness showed a trend towards lower values and the aspect ratio remained virtually unchanged. Concerning CaP-coated surfaces, spreading and shape parameters indicated that cell spreading on Y-TZP + CaP increased between day 1 and 7, whereas on Ce-TZP + CaP cell area and perimeter seemed to decrease. These findings, in conjunction with elevated values for roundness and circularity point to a rounding of the cells on Ce-TZP + CaP from day 1 to day 7.
Since quantitative morphometry demonstrated time-and biomaterial-dependent modulation of AO morphogenesis, we investigated possible correlations between cell shape, surface topography and wettability. For this purpose, we next performed a Spearman's test for correlation between aforementioned cell morphology parameters and interferometric surface parameters as well as contact angles. The Spearman's test for correlation provides values ranging from − 1 to 1, the so-called Spearman's rho (r s ), whereby values close to − 1 suggest an inverse correlation, values around 0 reject any correlation and values close to 1 indicate a strong correlation between two parameters. The results, summarized in Table 2, demonstrate that the topographic surface parameter S tr correlated significantly with cell aspect ratio (p = 0.001 for day 1 and p = 0.011 for day 7), roundness (p = 0.007 for day 1 and p = 0.007 for day 7) and circularity (p = 0.001 for day 1, day 7 not significant), whereat the aspect ratio exhibited an inverse correlation. These correlations reveal that the more uniform the surface structure in all directions, i.e. isotropic was, the less elongated the cells were, and that lesser or smaller cellular protrusions were found. Cell perimeter showed inverse correlation with S tr and S dr at day 7 (p = 0.046 for both correlations), indicating that isotropic surface properties and high surface enlargement reduced cell perimeter.
Intriguingly, despite the aforementioned variations in hydrophilicity of the different biomaterials, no correlation between the surface wettability and cell morphology could be found. This indicates that the structural surface properties, or more precisely the surface topography, appeared to superimpose the impact of surface wettability on cell morphology.
Cell proliferation. Since we detected biomaterial-associated differences in cell morphogenesis on the zirconia surfaces under study, we investigated how surface topography and/or chemistry influenced cell proliferation. Therefore, we next analysed the metabolic activity and proliferation of AO at days 1 and 7 on the zirconia discs. This was carried out by measuring the mitochondrial reduction of the redox indicator alamarBlue (AB) in the cell culture supernatant and the DNA content in the corresponding cell lysates.
As shown in Fig. 5, AB reduction (Fig. 5a) and DNA concentration (Fig. 5b) showed only slight differences between the samples at day 1, thus indicating that the initial attachment efficiency of the cells was comparable for all biomaterials. From day 1 to day 7 AB reduction increased significantly on all surfaces, albeit to differing degrees, with the highest values obtained on the grinded surfaces, namely Y-and Ce-TZPgr at day 7. www.nature.com/scientificreports/ Furthermore, it is striking that the Ce-TZP and Ce-TZPsa surfaces seemed to better support cell proliferation than their Y-TZP-based counterparts, as the AB reduction as well as DNA concentration showed a higher increase from day 1 to day 7 on these Ce-TZP surfaces with matched Y-TZP discs. More precisely, AB reduction on Ce-TZP/Ce-TZPsa at day 7 was 3.5/4.3 times higher than at day 1 and DNA concentration on these surfaces increased 4.0/4.2-fold during the experiment duration, while corresponding values for Y-TZP and Y-TZPsa were 3.0 and 3.2-fold for AB reduction, and 3.0 and 2.1-fold for DNA concentration. Such proliferation behaviour very likely arose from the different surface properties created by the Ce-TZP coating, mainly characterized by low surface enlargement (S dr ) values (see also Fig. 2), rather than from the chemical composition or hydrophilic properties, since on grinded surfaces, which displayed the smoothest properties on cellular level, cell growth showed a similar trend on Y-and Ce-TZP. The test for correlation between proliferation, surface topography and wettability substantiated this assumption, as the surface enlargement (S dr ) correlated inversely with the AB reduction at day 1 (p = 0.011) and with the DNA concentration at day 7 (p = 0.037), with the same trend at day 7 (p = 0.096) and 1 (p = 0.058), respectively (Table 3). Furthermore, the surface parameter S tr was also associated with cell proliferation at day 7, since Spearman's rho yielded values of − 0.62 (p = 0.115) for the AB reduction, and − 0.76 (p = 0.037) for the DNA concentration. This means that the more isotropic a surface was the lower was AO proliferation. Hence, the results of the statistical analysis demonstrate that surface properties including anisotropy and low surface enlargement better supported cellular proliferation on zirconia discs. In contrast, there was no correlation between the examined cell proliferation parameters and measured contact angles.
Correlations between cell morphology and proliferation. The data obtained from cell morphogenesis and proliferation experiments on the different zirconia surfaces revealed that AO with an elongated and  www.nature.com/scientificreports/ spread morphology at day 1 showed a higher increase in metabolic activity and DNA concentration, and thus cell number, up to day 7, compared to less spread cells with higher roundness and circularity values (compare grey and striped columns of Ce-TZP + CaP and Y-TZPsa with dark grey columns in Fig. 4 for morphology and the columns for these surfaces in Fig. 5 for DNA concentration). Therefore, values for AB-reduction and DNA concentration on Ce-TZP + CaP and Y-TZPsa surfaces were lower compared with e.g. grinded biomaterials at day 7. This observation suggests a direct link between cell morphology and proliferation. To visualize such correlation, we plotted the values detected for AB-reduction and DNA concentration versus the examined cell morphology parameters, i.e. cell area, perimeter, roundness, circularity and aspect ratio, and calculated the Spearman's rho for each parameter. Figure 6 shows the scatter diagrams with the calculated Spearman's rho (r s ). Scatter diagrams and Spearman's correlation test revealed a strong and statistically significant correlation of ABreduction as well as of DNA concentration with cell spreading as measured by cell area (Fig. 6a) and perimeter (Fig. 6b)  Hence, osteoblast spreading apparently favoured the metabolic activity and proliferation while circularity was associated with lower DNA concentration. Roundness (Fig. 6c) and aspect ratio (Fig. 6e) did not show statistically significant correlation with AB-reduction nor DNA concentration. Spearman's rho obtained for these correlations seemed to confirm that cell elongation favoured proliferation while cellular roundness decreased proliferation. Scatter diagrams furthermore revealed a switch in proliferation behaviour that occurred by surpassing a critical cell area of approximately 3,000-4,000 µm 2 or 400-500 µm cell perimeter, respectively.

Discussion
In our previous studies we demonstrated that implant surface topography and chemistry influence bone cell response in vitro and in vivo 6,38,39 , thereby pointing to cell instructive biomaterial properties, which emerge from implant-inherent features. Based on these findings, we here investigated whether specific topographical and/or chemical surface features of differently modified zirconia-based implant materials can be correlated to alterations in cell morphogenesis and proliferation. In this context, our experimental set-up utilized a causal approach, in which implant surface properties were used to induce various cell morphologies. Correlation analyses were subsequently applied to determine the geometrical aspects of morphology that were affected by the chosen surface properties. In addition they were employed to determine the relationship between osteoblast morphology and their proliferation behaviour. The decisive role of the induction of a certain cell morphology on other cell behavioural features like directing the differentiation of mesenchymal stem cells, has been shown by Kilian and co-workers 30 . In a comparable approach, Rolauffs et al. have also demonstrated that specific surface properties can be utilized causally for inducing specific aspects of cell behavioural functions 35 . With respect to surface topography, the characterization of the roughened zirconia surfaces by SEM and interferometry revealed that the applied surface modification methods created distinct differences in the microtopographies, which were mainly reflected by the spatial and hybrid parameters surface texture aspect ratio S tr and surface enlargement S dr . By combining IFM and SEM analysis for 3D-microtopography interpretation, we demonstrated that the micrometre-scale roughness parameter S a , an accepted predictive parameter for a successful osseointegration 5 and thus the most commonly applied topographical parameter to characterize dental implant surfaces was not suitable to describe the surface microtopography with which the cells interact, but rather seemed to reflect surface characteristics at a supracellular level. This was exemplified by the grinded Y-TZPgr materials, which exhibited a higher S a -value than the untreated Y-TZP, but, from the cell's viewpoint, displayed smoother surface characteristics (less rougher surfaces) at the low micrometre level than Y-TZP (compare SEM images Fig. 1a,d). Therefore, S a and/or R a may be unsuitable to describe an implant surface on a cell-instructive level, rendering them unreliable predictors for the outcome of cell morphogenesis and proliferation. www.nature.com/scientificreports/ www.nature.com/scientificreports/ To find out which of the examined surface structures influenced cell behaviour in terms of morphogenesis and proliferation, we next demonstrated that the surface texture aspect ratio S tr , a measurement for surface isotropy, affected almost all cell shape parameters, except cell area, under study, and thus overall AO morphogenesis. In contrast, the hybrid parameter S dr exerted its impact exclusively on cell spreading and the formation of cellular protrusions. However, osteoblast morphology remained unaffected by average surface roughness (S a ). Similar findings were reported by Anselme et al. who, in the context of an implant surface, stated that "human osteoblasts are more sensitive to the organization and morphology of the roughness rather than to its amplitude" 14 . Our data thereby showed that isotropic surface structures and high surface enlargement induced less elongated and spread cell morphologies with less and/or smaller cellular protrusions. On the contrary, AO on anisotropic surfaces aligned preferentially along the axis of the surface grooves, and displayed elongated and spread morphologies with pronounced actin stress fibre formation. The latter cell response on the grinded surfaces may be caused by contact guidance, i.e. the alignment of cells along anisotropic surface features, and the smooth microtopography, which provided large bio-adhesive areas for the cells. In both cases, namely anisotropic and smooth surface properties, human MSCs and osteoblasts have been shown to form stable and large focal adhesions with large actin bundles, substantiated by stress fibres [40][41][42] , as also observed in the present study.
With respect to AO proliferation, again both aforementioned surface parameters S tr and S dr were shown to be linked with the metabolic activity and cell number. Here it became apparent that anisotropic surfaces with low surface enlargement generally favoured cell proliferation, whereas the proliferation rate decreased with increasing surface isotropy. Thus, our results demonstrate that both surface parameters S tr and S dr appear to be suitable to predict AO behaviour with respect to morphology and proliferation on implant surfaces in vitro and could be considered as cell instructive surface features during implant processing. The lack of correlation between S a /S q and cell behavioural parameters found in this investigation together with our recently published report, which describes different cell and tissue reactions to implant surfaces with similar roughness values but different surface micro-architecture 6 , thereby add to the body of evidence that a surface specification by S a and/ or R a may not allow reliable conclusions on osteoblast behaviour.
As the applied ceramic implant surfaces exhibited different chemical properties, we further characterized the elemental composition and wettability of the biomaterial surfaces to assess the potential impact of these material parameters on osteoblast response. However, our analysis revealed no significant effect of surface chemistry and wettability on cell functions, since surfaces with different elemental composition but comparable topography, such as Y-TZPgr and Ce-TZPgr, showed similar osteoblast behaviour, and no correlation between surface wettability and examined cell function parameters could be proven. This suggests that the effect of surface topography on osteoblast function seemed to overcompensate the impact of elemental surface composition and surface wettability. This assumption is supported by our earlier work which demonstrates that osteoblast functions are rather controlled by surface architecture than by (physico-)chemical implant surface properties in vitro 38 . With respect to the CaP-coated Ce-TZP surfaces that yielded the least spread AO morphology over culture time and low proliferation, it appeared that again the distinct micro-topography with its net-like structures observed by SEM may have caused the aforementioned osteoblast response. This would also be in line with other studies, reporting that the biological activity of CaP coatings mainly depends on the topography and crystallinity of the coating and can therefore lead to conflicting results, namely positive as well as negative biologic effects, depending on the properties of aforementioned variables 43 .
Our correlation analysis not only revealed surface topography inherent characteristics of dental implant surfaces as being cell-instructive but also substantiated the assumption that osteoblast proliferation depends on cell morphology. Focusing on the latter issue, a rounded and less spread cell shape as found on the CaP coating coincided with a low proliferation, whereas a spread and elongated morphology on anisotropic grinded surfaces was linked to an enhanced proliferation. The afore described correlation of cell spreading and proliferation has already been reported for cells of other tissue origin, namely hepatocytes, endothelial and smooth muscle cells, showing that cell spreading is essential for the survival and proliferation of adhesion-dependent cells [26][27][28][29][30][31][32][33] . With respect to this issue, our data analysis thereby not only provides evidence for a direct correlation between AO morphogenesis and proliferation, but further revealed a switch in proliferation behaviour by surpassing a critical cell size, or more specifically cell area, of approximately 3,000 to 4,000 µm 2 . Intriguingly, such an increase in cell proliferation has also been observed for bovine and human endothelial cells 26,27 and human smooth muscle cells 32 , after exceeding a cell spreading area of approximately 1,500-2,200 µm 2 and 500 µm 2 , respectively. This correlation together with aforementioned previous reports corroborates the existence of a cell-innate cell area threshold, which may function as a proliferative stimulus 32 . From a more abstract viewpoint, this could imply that cell morphology-instructive biomaterial features can be systematically used as proliferation triggers. Regarding the cytoskeletal involvement in the regulation of cell adhesion, it has been demonstrated for endothelial cells that herein, the actin cytoskeleton is involved in the morphogenesis-related cell spreading, and that actin stress fibre formation, similar to that seen for our AO on grinded surfaces, increases the cells' mitotic activity 44,45 .
Taken together, for the here applied ceramic dental implant surfaces, that were subjected to commonly applied surface modification methods, topographical but not chemical/physicochemical surface parameters appeared to be cell instructive. This initial characterization of cell instructive parameters appears motivating, to further characterize implant-innate features which act cell instructive by controlling cell fate decisions. Thus, with respect to a putative translation of cell instructive parameters into the development of next generation biomaterials, cell instructive implant surfaces should for instance consider the aforementioned surface parameters (i) isotropy and (ii) surface enlargement. This is because if a biomaterial of interest should explicitly favour cell spreading and proliferation, the biomaterials surface features should include anisotropic properties in conjunction with low surface enlargement. These two parameters per se favour AO spreading, thereby considering that the aspect of spreading is a prerequisite for proliferation. In this context, the underlying molecular mechanisms of the material-cytoskeleton-crosstalk controlling cell fate and functions need further elucidation 46  www.nature.com/scientificreports/ At authoring this manuscript, techniques to imprint nanometre-scale features onto implant surfaces remained far from being integrated in commercially available implants and no established classification concepts of implant surfaces at the nanometre-level existed. The authors of this manuscript therefore recommend that future research focuses on the identification of topographical and/or physicochemical cell-instructive implant surface parameters at the nano-and micrometre-level and on their in vivo validation. This might render the design of implant surfaces with predictable influence on cell behaviour and subsequently osseointegration possible.

conclusions
The current study demonstrates that among the tested implant surface parameters, characterized for various zirconia-based implant biomaterials, texture aspect, and surface enlargement are significantly more effective than surface roughness or wettability in controlling morphogenesis and proliferation of AO in vitro. Regarding morphogenesis, calculation of cell area firstly reveals proliferation as a function of morphogenesis, and secondly, corroborates together with other reports the existence of a cell area-dependent threshold, which may function as cell-innate proliferation trigger. Moreover, the coincidence of a proliferation-favourable AO morphogenesis with actin stress fibres supports the notion that their presence can be used as cytoskeletal proliferation indicator. Based on the detected proliferation-supportive cellular responses towards certain implant surface features it can be concluded that they may in turn be used as cell behavioural predictors, which may be used as cell instructive parameters to be implemented in the next generation of "cell instructive" biomaterials.

Materials and methods
Implant materials and surface treatment. Biomaterial discs (15 mm in diameter, 1.5 mm thickness) used for cell culture were prepared from tetragonal zirconia polycrystals containing 3 mol% of yttrium oxide (Y-TZP) by uniaxial pressing. In order to compare the effects of the chemical and topographical surface properties on cell behaviour, different surface modification techniques were applied to the Y-TZP-discs. Biomaterials were grouped as follows: (i) sintered Y-TZP not further processed (referred to as Y-TZP), (ii) Y-TZP discs coated with a novel ceria stabilized zirconia-alumina-strontium aluminate nano-composite powder (referred to as Ce-TZP; see 47 for detailed description), (iii) Y-TZP and Ce-TZP discs coated with a thin layer of calcium phosphate (referred to as Y-TZP + CaP and Ce-TZP + CaP, respectively; CaP = calcium phosphate), (iv) grinded Y-TZP discs (referred to as Y-TZPgr; gr = grinded), (v) Y-TZPgr discs coated with Ce-TZP (referred to as Ce-TZPgr), (vi) alumina sandblasted Y-TZP discs (referred to as Y-TZPsa; sa = sandblasted), and (vi) Y-TZPsa discs coated with Ce-TZP (referred to as (Ce-TZPsa). The discs for cell culture experiments were sterilized using lowtemperature hydrogen peroxide gas plasma sterilization before performing the cell culture experiments.
Surface characterization. Surface topography of the implant materials was examined as already described in a previous report 6 by scanning electron microscopy (SEM, LEO435VP scanning electron microscope, Zeiss, Oberkochen, Germany) with the backscattered electron imaging mode and an accelerated voltage of 8.00-12.00 kV after sputter coating with gold-palladium for 60 s at 60 mA (SCD050; Balzers, Liechtenstein), and by light interferometry (IFM, MicroXAM 100 h; ADE; Phase Shift Technology, Tucson, USA). 3D reconstructed IFM images were created by the software MountainsMap Premium, version 7.4 (Digital Surf SARL, Besanҫon, France). The measuring area for IFM was set to 260 × 200 µm (n = 3 per group). Before parameter calculation, a digital (Gaussian) filter of 50 × 50 μm was applied to remove errors of form and waviness. Amplitude, spatial and hybrid parameters were obtained via IFM measurements to characterize the biomaterial surfaces in three dimensions (for detailed information about surface parameters see 48,49 . Surface amplitude was characterized by the parameters S a , S q , S z and S sk . The main roughness parameter S a describes the arithmetic mean value of the absolute surface asperity departures in μm and the root-mean-square deviation S q the standard deviation of S a . Although S a and S q describe similar surface features both parameters are listed in this work since S a is the most commonly used parameter to characterize surface roughness in medical science while S q is preferred by statisticians 48,49 . Furthermore, surface amplitude was measured by the ten-point height of surface topography S z which represents the mean value of the five highest peaks and five deepest valleys of a surface in micrometre and S sk as a measurement for the asymmetry of surface deviations from the mean plane. S sk adopts negative values if surface height distribution is shifted towards a predominance of valleys while positive values indicate that the surface height distribution has more peaks than valleys. Spatial parameters were the summit density S ds describing the number of peaks per area in 1/mm 2 and the texture aspect ratio of a surface designated as S tr . The latter parameter describes the topographic texture pattern and takes values between 0 and 1, with small values indicating strong anisotropy, i.e. less uniformity, and large values indicating uniform texture aspect in all directions 49 . The hybrid parameter S dr (developed interfacial area ratio) was examined to describe the surface enlargement compared to a totally flat reference area and is given in percent. Elemental composition of the biomaterial surfaces was analysed by energy-dispersive X-ray spectroscopy (EDX) (n = 3 per group). EDX was performed with a JSM-IT100 (JEOL Ltd, Tokyo, Japan). Magnification was set to 1,000, excitation energy to 15 kV and time of measurement to 100 s. Surface wettability of the zirconia discs was evaluated by static contact angle measurement (n = 10 per group). Contact angles of 2 µl water droplets were analysed using the Dataphysics OCA 10 optical contact angle measuring system (Dataphysics GmbH, Filderstadt, Germany) by measuring the angles of each drop against the surface.
Isolation and cultivation of primary osteoblasts. Alveolar bone osteoblasts (AO) were prepared from bone explants obtained during implant site preparation procedure at the Department of Prosthetic Dentistry, University of Freiburg, Germany. The collection and usage of the primary osteoblasts for scientific purposes was approved by the Ethics Committee of the Albert-Ludwigs-University, Freiburg, Germany (vote Nr. www.nature.com/scientificreports/ 411/08_121010) and informed consent was given by the patient. Research was performed in accordance with relevant guidelines and regulations. The bone specimens were dissected aseptically and cleaned in phosphatebuffered saline (PBS) to remove residues of blood and soft and/or bone marrow tissue. In order to ensure sterility of the bone explant cultures, the cleaned bone specimens were additionally sterilized in an iodide solution, subsequently washed with PBS. The bone explants and isolated cells were cultured in Dulbecco's Modified Eagle's Medium supplemented with 2% (w/v) glutamine, 10% (w/v) fetal calf serum and 50 µg/ml kanamycin, and maintained in a humidified 37 °C incubator with 5% CO 2 . All experiments were carried out with osteoblasts of passages 6 and 7.
Evaluation of osteoblast morphology. Cell morphology of AO on different zirconia surfaces under study was analysed by SEM and fluorescence-based actin cytoskeleton staining with Texas Red-X-labelled phalloidin (Life Technologies, Darmstadt, Germany) as described earlier in Altmann et al. 6 . In detail, cells were fixed with 4% formaldehyde (PFA) in PBS for 20 min at room temperature after 1 and 7 days of culture. In addition to actin cytoskeleton staining, osteoblasts were analysed by SEM at day 1. Therefore, cells were fixed and dehydrated in an ascending ethanol series (ranging from 30 to 100% ethanol, three times each for 20 min at room temperature), critical point dried (CPD030 Critical Point Dryer, Bal-Tec AG, Balzers) and sputter coated with gold-palladium for 60 s at 60 mA (SCD050, Balzers). For actin staining, the samples were fixed at day 1 and 7, treated with 2% (w/v) bovine serum albumin in PBS and 0.2% TritonX-100 in PBS for 15 min, and 2% (w/v) BSA in PBS for further 15 min. Actin labelling was then performed by incubating the samples for 30 min with phalloidin conjugate diluted 1:40 in PBS containing 0.5% (w/v) BSA. Nuclei were stained with 300 nM DAPI for 15 min. Optical evaluation was performed with the fluorescence microscope Biozero BZ-8000 (KEYENCE, Neu-Isenburg, Germany) and quantitative morphometric analysis (20 < n < 325) was carried out with the microscope software BZ Analyzer II from KEYENCE. AO morphology was estimated by measuring the cell area and perimeter which both provide information on the extent of cell spreading 50 , the major axis representing the long axis of the smallest rectangle drawn around the cell body and the minor axis, presenting the rectangle width. In order to measure changes in cell shape we used in analogy to Uynuk-Ool and coworkers 35 a panel of specific shape descriptors derived from the morphometrically collected data to calculate the hybrid morphology parameters roundness (4 × area / (π × major axis) 2 ), circularity (4 π × area / perimeter 2 ) and aspect ratio (major axis / minor axis). High values for the roundness parameter indicate a high degree of cell roundness, whereas a high aspect ratio points to an elongated cell shape. The parameter circularity describes the change from a circle with a large number of protrusions into a circle without protrusions, i.e. specifies whether cellular protrusions are formed or not. A high value indicates the absence of cellular protrusions.
Evaluation of metabolic activity and cell proliferation. The metabolic activity of AO cultured on the test surfaces was analysed by the alamarBlue (AB) metabolic assay (MorphoSys AbD, Düsseldorf, Germany). After 1 and 7 days of culture on the test surfaces the cells were incubated for 2 h at 37 °C with culture medium, supplemented with 10% (w/v) AB reagent (n = 6 per test group). Subsequently, the supernatant was analysed by fluorimetry according to the manufacturer's instructions (Tecan, Männedorf, Germany). The percentage of AB reduction in the samples was calculated using a 100% reduced AB control as reference. Proliferation of AO cultivated on the different zirconia discs was evaluated after 1 and 7 days of culture by performing DNA quantification of the samples with the Quant-iT PicoGreen (Invitrogen, Carlsbad, U.S.A.) assay according to the manufacturer's protocol (n = 6 per group). Prior to DNA quantification cells were washed once with PBS and lysed by a freeze-thaw cycle at − 80 °C in 400 μl TE-buffer (10 mM Tris-HCl, 1 mM EDTA, pH 7.5) 38 .
Statistical analysis. Differences between test groups were examined for statistical significance using the one-way ANOVA followed by a Tukey's HSD post hoc test for normally distributed data (according to the Shapiro-Wilk-test for normality) and the Kruskal-Wallis ANOVA followed by a Dunn's post hoc test for nonnormally distributed data. To test for a possible correlation between surface parameters, cell morphometry and proliferation, the Spearman's rank correlation test was applied. The Spearman's test for correlation provides values ranging from − 1 to 1, the so-called Spearman's rho (r s ), whereby values close to − 1 suggest an inverse correlation, values around 0 reject any correlation and values close to 1 point at strong correlation between two parameters. The deviation of the Spearman's rho from zero was examined for statistical significance using a twotailed t-test.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.