Useful Surface Parameters For Biomaterial Discrimination

MARINA ETXEBERRIA, TOMAS ESCUIN, MIQUEL VINAS, AND CARLOS ASCASO Doctoral Student, Department of Dentistry and Department of Pathology and Experimental Therapeutics, Dentistry School, University of Barcelona, Barcelona, Spain Associate Professor, Laboratory of Prosthetic Dentistry, Dentistry School, University of Barcelona, Barcelona, Spain Department of Pathology and Experimental Therapeutics, Medical and Dentistry Schools, University of Barcelona, Barcelona, Spain Department of Public Health, Medical School, University of Barcelona, Barcelona, Spain


Introduction
Surface features significantly condition many technological and biomedical applications of biomaterials (Ham and Powers, 2014). Surface roughness and surface wettability can significantly determine major aspects of biological interactions and, subsequently, allow to predict the eventual failure or success of an implantprosthetic treatment (Park et al., 2012;Gittens et al., 2013). Surface modification strategies attempt to modulate the surface properties of biomaterials in order to affect cell-substrate interactions and improve the overall biological response (Ivanova et al., 2010). Furthermore, in order to accomplish this purpose a detailed characterization of surface topography must be achieved.
Characterization of surface roughness is complex as it depends on both the intrinsic properties of the material and manufacturing procedures and conditions (De Chiffre et al., 2000). In an attempt to have a more extensive and clear description, a wide variety of surface roughness parameters (RPs) have been developed. This has been termed as "the parameter rash" by Whitehouse (Whitehouse, 1982). Nevertheless, inconsistencies have been reported when describing surface topographies, in part due to the lack of standardized methods. Nowadays a wide set of parameters are being used; however, it seems that there is an urgent need to reduce the number of parameters in order to achieve a general standardization to facilitate comparisons and reduce cost.
The parameter reduction method is effective at selecting the RP to represent a surface (Nowicki, 1985;Ros en, 2008;Ham and Powers, 2014). This method is based on the analysis of strong and weak correlations between RPs; correlated RPs highlight the similarity between them; conversely, non-correlated RPs underline the difference among them. Highly correlated RPs are redundant and thereby one can be selected to represent the whole group. In contrast, poorly correlated RPs provide complementary information being thereby best discriminating between materials (Nowicki, 1985).
Progress in nanotechnologies has led to the development of nanometer resolution technologies allowing research and visualization at a scale in which interactions between bacterial cells and biomaterials' surfaces occur. Atomic Force Microscopy (AFM) is the most powerful tool for topographical characterization at the nanometer and sub-nanometer scales (Binnig et al., 1986;Dorobantu and Gray, 2010). AFM topography imaging is non-destructive and widely used in life sciences which provides high-resolution characterization of surface topography, biomolecules, membranes, and cells at the nanoscale. White Light Interferometry (WLI) is a type of computerized optical interference microscopy. Its use has rapidly widespread as a quality control of microscale engineering processes and has been used to analyze surface roughness and cell adhesion at the microscale (Hove et al., 2007). This method has been shown to be fast, non-destructive and accurate. The combination of both techniques has been proposed to improve the measuring efficiency of AFM for the surface characterization of biomaterials (Tyrrell et al., 2004;Guo et al., 2011).
To characterize surface structure, the present study examined six different dental materials for implant abutment manufacture using an atomic force microscope (AFM) for high resolution analysis, white light interferometry (WLI) and the drop-sessile-water method. From both methods for measuring surface roughness, amplitudinal roughness parameters were determined, which are so far the most cited surface parameters for surface characterization (Ivanova et al., 2010;Gittens et al., 2013;Webb et al., 2013). These are obtained from the height values of a given profile (denoted by R) or surface (denoted by S). The aim of this study was to attempt the combination of surface parameters resulting in an optimum surface description.

Specimen Preparation
Disks 10 mm in diameter and 2 mm thick were manufactured (n ¼ 16) from six different implant abutment materials. The tested materials were: cast cobalt-chrome (Co-Cr), direct laser metal soldered (DLMS) Co-Cr disks, Titanium grade V disks, Zirconia (Y-TZP) disks, E-glass fiber-reinforced composite, and polyetheretherketone (PEEK). The disks were manufactured as previously described (Etxeberria et al., 2014).

Cast Cobalt-Chromium Disks
Acrylic resin (pattern resin 1 LS, GC Corp.) disks of the desired final shape were fabricated and casted by induction (Ducatron S erie 3 UGIN'Dentaire. Seyssins. France) using Co-Cr (Wirobond C 1 alloy, BEGO, Bremer Goldschl€ agerei Wilh. Herbst GmbH and Co. KG, Bremen, Germany). After casting, the sprues were eliminated with the aid of carbide discs at low speed. The castings were sandblasted with 110 mm aluminum oxide particles (Korox 1 , Bego, Bremen, Germany) under three bar pressure to remove oxide films and residual investment.

DLMS Cobalt-Chromium Disks
The disk shaped specimens were designed in a 3D software package and saved in an industry standard stereolithography (STL) format. The standard DLMS (direct laser metal soldering) manufacturing method by EOSINT M 270 (EOSINT 270 GmbH Electro Optical Systems, Munich,Germany) was used to fabricate the disks.
Both the cast and the DLMS Co-Cr disks were polished in three stages: (a) using a hard rubber disk at 15,000 rpm; (b) then with a soft rubber disk at 15,000 rpm, and finally (c) using a soft brush with a polishing paste at 1400 rpm. Each polishing phase lasted 90 seconds.
All disks were handled by their lateral walls not to damage the disks' surfaces. In addition were gently cleaned using a cotton pellet with ethanol and dried under warm dry air.

Atomic force microscopy
The surface topographies of the tested materials were characterized at the nanoscale using AFM Park Systems,Korea). Images with the areas of 5 Â 5 mm 2 were scanned in the standard non-contact mode. The probe was supported on a rectangular-shaped cantilever tip (tip radius: < 10 nm, f¼ AE 300 kHz, spring con-stant¼ AE 40 N/m, silicon coating). The scan rate was 0.6 Hz and the resolution 256 Â 256 pixels. Representative roughness parameters S Min , S Max , S Mid , S Mean , S pv , S q , S a , S z , S sk , and S ku described in Table I were calculated from the roughness values obtained by AFM and processed by XEI image processing software (Park Systems).

White Light Interferometry
The surface topographies of the tested materials were characterized at the microscale using a white light interferometer microscope (LeicaSCAN DCM3D, Leica Microsystems, Switzerland). A 50 Â /0.50 Mirau objective was utilized. The threshold was set to 1.0% and the Gaussian filter to 25 mm. Vertical scanning interferometry mode images with the areas of 250.64 Â 190.90 mm 2 were obtained. Image data-analyses were performed using Leica map DCM 3D, version 6.2.6561 (Leica Microsystems, Switzerland) and R p ,R v , R z , R t , R a , R q , S a , S z , and S q roughness parameters described in Table I were calculated.

Surface Wettability
External water contact angles were analyzed with the sessile-water-drop method at room temperature Gittens et al., 2013). A 10 mL drop of MilliQ-quality water was placed onto the center of each specimen using an injector. Digital photographs were taken (Nikon D70) and the determination of the external contact angle was done using IMAT software (CCIT, Barcelona, Spain). Two contact angles (u left and u right ) per disk were obtained.

Statistical Analysis
The surface nanoroughness, microroughness and wettability data did not follow a normal distribution. Therefore, a non-parametric ANOVA statistical analysis was carried out for data comparisons. Quantitative data analysis including the median, minimum, and Right contact angle maximum were computed for each parameter. Spearman's rank correlation coefficient was used to express the degree of pair-wise association among nanoroughness parameters, microroughness parameters, and wettability. In order to identify statistical differences among the materials, Kruskal-Wallis and Mann-Whitney U test were performed with the Bonferroni adjustment according to the number of tests performed. Total data were analyzed in SPSS 21.0 to provide descriptive statistics and to perform nonparametric testing. Statistical analysis was performed with Statistical Package for the Social Sciences (Version 21.0; SPSS., Inc, Chicago, Illinois). Hypotheses were contrasted with an alpha error of 5% and estimations with 95% confidence level.

Results
Overall results of the measurements on the surfaces are summarized in Tables II-IV. Tables II, III and IV describe the median, minimum and maximum values computed for each surface parameter carried out of 16 estimations per each material.
Correlation coefficient calculations are presented from Table V to Table VII : among nanoscale roughness parameters and wettability (Table V); among microscale roughness parameters and wettability (Table VI) and  finally Table VII summarizes the correlations between nanoscale and microscale roughness parameters. At the nanoscale, roughness parameters showed poor correlations however three clusters of parameters are differentiated. A highly correlated group (r>0.86) comprised by S a -S max -S min -S pv -S q -S z in addition to S mean -S mid and S sk -S ku groups that are weakly correlated (r ¼ 0.29 and r ¼ À0.32 respectively) among themselves. Contrary to the nanoscale, at the microscale all the parameters are correlated (r>0.58). Nevertheless, two subgroups are slightly differently related by their correlation degree: the profile roughness parameters and the surface roughness parameters. Contact angles (u left and u right ) are highly correlated (r ¼ 0.97) among themselves regardless of the scale. Wettability did not correlate with any of the nanoroughness parameters in contrast it showed a weak and negative correlation with microroughness parameters. Correlation analysis of nano and microescale parameters in Table VII presented few and weak correlations.
Results of Kruskal-Wallis (p<0.01) and Mann-Whitney U test (p<0.003) (Tables II, III) show that S a roughness parameter exhibited the highest discrimination power at both scales.

Characterization of the Tested Materials
Results of the characterization of the analyzed materials showed that FRC was found to be the roughest   while DLMS Co-Cr resulted the smoothest. Zirconia was shown to be the most hydrophilic whereas FRC resulted the most hydrophobic material. Finally, in Figure 1 a graphic representation of the discrimination of the materials according to the selected parameters is described.

Discussion
Several attempts have been made to establish a set of surface parameters giving the optimum surface description for the discrimination of materials (Stout et al., 1994;Crawford et al., 2012;Webb et al., 2013). However, the statistical dependence of the surface parameters has rarely been analyzed. The present study is in agreement with previous studies that state that the commonly used parameters to characterize biomaterials are redundant (Stout et al., 1994;Crawford et al., 2012;Webb et al., 2013). A set of six parameters giving the highest discriminatory power (S a ,S ku, and S mid at the nanoscale, S a and S z at the microscale and u right ) where selected out of 21 parameters to represent the whole group of parameters.
The poor correlations exhibited among the nanoscale surface parameters are in agreement with previous studies (Ros en et al., 2008). However the strong correlations displayed by the S a -S max -S min -S pv -S q -S z cluster of parameters means that the determination of one of the parameters automatically leads to the definition of the others. S mean -S mid and S sk -S ku groups are not correlated thereby they provide additional complementary information. The present results may be explained by the fact that all the highly correlated parameters are height descriptors, S mean , S mid are normality height descriptors and S sk , S ku describe the spatial surface topography. Regarding the most correlated group, the criteria for selecting the parameter to represent the group was based on the most sensitive parameter in the materials discrimination which was  found to be S a (Table II). From the less correlated groups, the criteria for selecting the parameter was the lesser correlation of parameters. Hence, a preliminary set of three independent parameters, S a , S mid , and S ku was selected. S a (or its counterpart R a ) is one of the most commonly used parameters to quantify surface topography (Whitehead et al., 2005;Truong et al., 2010;Crawford et al., 2012). It quantifies the "absolute" magnitude of surface heights but in contrast, is insensitive to the spatial distribution of the heights. Similarly to previous studies, our results highlight that the S a value is insufficient for the surface discrimination of biomaterials at the nanoscale and spatial surface descriptors are needed for an optimized surface characterization (Ivanova et al., 2010;Webb et al., 2012). In practical terms, kurtosis values describe the shape of the distribution of the heights; (i.e., normal distributions have kurtosis value of three while sharper distributions have higher values and rounded distributions have lower). In the present study, DLMS Co-Cr and PEEK showed the smoothest surfaces at the nanoscale obtaining kurtosis values > 3 compared to the rest of materials, which had values of < 3. On the other hand, the zero value for S sk , (skewedness is a measure of the symmetry of height distribution) reflects symmetrical height distribution and these results are corroborated by a zero value for S mean . This may be explained by the fact that the materials underwent polishing procedures. It is evident that these parameters (kurtosis and skewedness) are material-dependent and that either one or the other or both should be addressed depending on the required information (Crawford et al., 2012). To the authorś knowledge, S mid has not been addressed before.
The applicability of the first subset of parameters has also played a role in determining bacterial adhesion. Thereby, in the study of Webb et al. the S a , S q , and S max parameters gathered similar bacterial counts in contrast to S sk and S ku (Webb et al., 2013).
In contrast to the nanoscale, at the microscale, all the roughness parameters are correlated (Table V) nevertheless, profile values are slightly differently related by  their correlation degree to surface values. These findings are comparable to previous studies (Nowicki, 1985;Ros en et al., 2008;Ham and Powers, 2014) however, the different correlation values obtained by Ham et al. is due to the different averaging methods. In their study the mean of three calculations was computed while in ours the median of 16 calculations. Due to the fact that all the parameters are correlated, the selection of the best set of roughness parameter for is hindered. Therefore, S a was selected to represent the whole group of parameters for being the most sensitive parameter on the pair-wise material discrimination at the microscale (Table III). This result is confirmed by recent studies which recommended the selection of S values as they are obtained from the surface and thus are more representative compared to those obtained from the profile (Webb et al., 2013). In the present study, S z shows the lowest correlation value with S a and with the rest of parameters and thus could be considered as a useful complementary roughness parameter. The efficiency of both parameters determined as the average and the maximum values has been widely used for material discrimination (Gorlenko et al., 1981;Nowicki, 1985;Gittens et al., 2013). Thus at the microscale subset, the two selected parameters are S a and S z .
The few and weak correlations encountered among nano and microroughness (  is of paramount concern to include two different scales. These results are in agreement with previous authors' recommendations of using optical measuring methods such as white light interferometry to expand the AFM measuring range and to improve roughness measuring efficiency (Tyrrell et al., 2004;Guo et al., 2011). Therefore, in general S a -S ku -S mid at the nanoscale and S a -S z at the microscale are not correlated being confirmed the complementarity of both groups of parameters. As first described by Wenzel, an intimate relationship between surface roughness and wettability exists (Wenzel, 1949). Nevertheless, this correlation was not observed at the nanoscale. Likewise in a recent study, no correlation was found between roughness and wettability at the nanoscale (Gittens et al., 2013). While wettability values did not correlate with nanoroughness parameters, they correlated poorly with microroughness parameters. The negative correlation encountered indicates that as the roughness value increases the external angle contact value decreases and vice versa. This is explained by Wenzel's method that states that roughness induces hydrophobicity (Wenzel, 1949) and has been confirmed in previous studies (Gittens et al., 2011;Webb et al., 2013).
The selected parameters are efficient in characterizing and differentiating between materials and the obtained characterizations are in agreement with previous studies (Rosentritt et al., 2009;Ivanova et al., 2010;Adbulmajeed et al., 2014;Ourahmoune et al., 2014;Kim et al., 2015). FRC exhibited the highest roughness value among the tested materials values but in the range of previous studies (Tanner et al., 2003;Garoushi et al., 2009;Adbulmajeed et al., 2014). In contrast, DLMS Co-Cr obtained the lowest roughness value. This finding is in agreement with recent studies that support the notion that the powder additive manufacturing layer by layer improves the surface compared to the conventional casting methods (Oyag€ ue et al., 2012;Castillo-Oyag€ ue et al.,2013). However, this finding is not in accordance with a recent study where the average roughness value of DLMS was significantly higher compared to cast Co-Cr. One explanation could be differences in the composition of the used metal alloys (Kilicarslan and Ozcan, 2012).
Previous studies have shown that smooth microscale surface characteristics (R a less than 0.1 mm) have minor influence on the surface wettability of a surface (Busscher et al., 1994;Adbulmajeed et al., 2011). In those studies smooth surfaces displayed contact angles that ranged between 60˚and 86˚and the differences of the contact angles were related to the surface chemistry. The present study, is in agreement with those studies since all the smooth surfaces investigated (all the materials except for FRC) showed contact angle values within this given range. Accordingly, rough FRC surfaces showed a contact angles above 86˚showing a similarity in the trend claimed by Wenzel (Wenzel, 1949). Wettability values in general are also in agreement with previous studies except for the Zirconia which showed the highest contact angle value. This may be explained by the fact that the surfaces were rougher than in previous studies (Att et al., 2009). On the other hand, FRC showed the lowest wettability, which may be explained by the influence of fibers on the wettability behavior of composite materials (Adbulmajeed et al., 2014).
The limitations of measuring devices may introduce errors during data acquisition which may reflect on the final surface characterization. For instance, even the very sharp tip of an AFM shows limitations, and the optical methods are limited when recording small wavelength components. In addition to this, filtering techniques should be considered with care.
Correlation tests can be carried out to systematize the choice of a set of parameters when multiple parameters have to be reduced. The selection of parameters should be founded on the results of the degree of correlation between the multiple parameters and the required properties regarding their application site. This set of parameters was efficient in differentiating between six types of materials at the nano and microscale. The adoption of this proposed set of parameters will enable universal comparisons.

Conclusions
The present study proposes six parameters for characterizing biomaterial surfaces: S a -S ku -S mid at the nanoscale, S a -S z at the microscale and one angle contact value are suggested for surface characterization.
Roughness quantification at two different scales gave complementary information.
Wettability was not correlated with nanoroughness. In contrast, it was correlated with rough surfaces at the microscale.