The effects of controlled nanotopography, machined topography and their combination on molecular activities, bone formation and biomechanical stability during osseointegration

The initial cellular and molecular activities at the bone interface of implants with controlled nanoscale topography and microscale roughness have previously been reported. However, the effects of such surface modiﬁcations on the development of osseointegration have not yet been determined. This study investigated the molecular events and the histological and biomechanical development of the bone interface in implants with nanoscale topography, microscale roughness or a combination of both. Polished and machined titanium implants with and without controlled nanopatterning (75 nm protrusions) were produced using colloidal lithography and coated with a thin titanium layer to unify the chemistry. The implants were inserted in rat tibiae and subjected to removal torque (RTQ) measurements, molecular analyses and histological analyses after 6, 21 and 28 days. The results showed that nanotopography superimposed on microrough, machined, surfaces promoted an early increase in RTQ and hence produced greater implant stability at 6 and 21 days. Two-way MANOVA revealed that the increased RTQ was inﬂu-enced by microscale roughness and the combination of nanoscale and microscale topographies. Furthermore, increased bone-implant contact (BIC) was observed with the combined nanopatterned machined surface, although MANOVA results implied that the increased BIC was mainly dependent on microscale roughness. At the molecular level, the nanotopography, per se , and in synergy with microscale roughness, downregulated the expression of the proinﬂammatory cytokine tumor necrosis factor alpha (TNF-α ). In conclusion, controlled nanotopography superimposed on microrough machined implants promoted implant stability during osseointegration. Nanoscale-driven mechanisms may involve attenuation of the inﬂammatory response at the titanium implant site.


Statement of Significance
The role of combined implant microscale and nanotopography features for osseointegration is incompletely understood.Using colloidal lithography technique, we created an ordered nanotopography pattern superimposed on screwshaped implants with microscale topography.The midterm and late molecular, bone-implant contact and removal torque responses were analysed in vivo .Nanotopography superimposed on microrough, machined, surfaces promoted the implant stability, influenced by microscale topography and the combination of nanoscale and microscale topographies.Increased bone-implant contact was mainly dependent on microscale roughness whereas the nanotopography, per se , and in synergy

Introduction
Commercially available and clinically used oral titanium (Ti) implants possess chemical and topographic surface variations [ 1 , 2 ].Despite generally successful long-term clinical performance of oral implants, it is also apparent that their surfaces have varying degrees of combined micro-and nanoscale features.The optimal shape, width, height, order and spacing of both macro-and nanoscale features on such surfaces remain to be established.
The surface topography of a Ti implant is considered a major determinant for the processes of osseointegration [1] .It is widely accepted that the microstructure of implants affects the bone response, which is supported by both in vitro and in vivo experimental data [ 3 , 4 ].Furthermore, cellular and molecular events at the bone interface with microroughened implants have been well documented [5][6][7] .
A role of isolated nano-features for bone cell performance in vitro has been established [8] .However, the relations among several, if not the majority, of the dimensions of nanoscale features and different properties of bone cells remain to be established.In vitro studies suggest that implants with nanofeatures stimulate the adhesion and proliferation of osteoblasts and induce osteogenic differentiation, matrix secretion and mineralization [ 9 , 10 ].Moreover, nanostructured implants have been shown to support in vitro growth and osteogenic differentiation of mesenchymal stem cells [ 10 , 11 ].Pertinent studies have typically employed nanoscale surface modification on two-dimensional (2D) commercially pure Ti, Ti alloys, ceramics or composite discs rather than on threedimensional (3D), clinically relevant Ti implants.Furthermore, in vitro conditions seldom represent the multifactorial and dynamic in vivo environment of the implant recipient bone.
The majority of studies exploring the influence of nanotopography on osseointegration have used Ti implants with combined nanoscale and microscale surface roughness; these studies have generally revealed a favorable bone response, typically with increased bone contact and interface strength [11][12][13][14][15] .Moreover, the descriptions of such combinatorial effects of nanoscale and microscale topographical properties have been indistinguishable from the effects of other surface properties (chemistry, surface energy and charge) that are inevitably altered by most surface modification techniques.Therefore, based on the current literature, it is difficult to draw firm conclusions about specific effects elicited by the nanotopography per se or about whether synergistic effects exist between the nanotopography and the underlying microroughness of the implant.
Recent attempts to gain mechanistic insights into the role of nanoscale vs microscale patterns for osseointegration in vivo were achieved by colloidal lithography in combination with a thin Ti coating of nanopatterned implants [ 16 , 17 ].Nanopatterning, as such, reduced the early recruitment and activity of inflammatory cells while enhancing early osteogenic activity and woven bone formation.Though these cellular and molecular events in response to nanostructured implants occurred during the initial-early healing phase, the role of such predetermined nanotopography on the events involved in the progression of osseointegration and the establishment of a mature and biomechanically stable interface has yet to be investigated.
In this study, 3D screw-shaped Ti implants were used, with and without predetermined surface topographies at the nanoscale and microscale.The in vivo molecular response at the bone-implant interface was investigated using quantitative polymerase chain reaction [18] in combination with removal torque (RTQ) measurements and histologic evaluation of the interface.The main aim of this study was to determine the role of controlled nanotopography on the molecular, histological and biomechanical outcomes of osseointegration and to explore whether the nanotopography effects, if any, are independent of the underlying machined microtopography.

Implants, nanopatterning and surface characterization
The implant design, nanopatterning and characterization were performed in the same batch that was used in a previously published study [16] .In brief, screw-shaped implants that were 2.3 mm in length and 2.0 mm in diameter were manufactured from grade IV Ti (Elos Medtech Pinol A/S, Denmark).Thereafter, four different surface modifications of the implants were made to produce implants with identical surface chemistry but different combinations of micro-and nanoscale topographies: polished (P), polished nanopatterned (PN), machined (M) and machined nanopatterned (MN).Surface polishing was achieved by an electrochemical process that uses a perchloric acid-based electrolyte (Sigma-Aldrich, Stockholm, Sweden) at a constant potential of 22.5 V. Nanopatterning was performed using colloidal lithography.First, the implants were soaked in a 5% wt/wt aluminum chloride hydroxide (Chlorohydrol, Summit Reheis, Huguenot, NY, USA) solution to form a sufficient positive surface charge.The positively charged implants were then soaked in a 2% wt/wt colloidal solution of negatively charged spherical polystyrene (PS) nanoparticles with a nominal diameter of 105 ± 5 nm (surfactant-free white PS latex solution, Invitrogen Corp, Carlsbad, CA, USA).This enabled a short-range ordered pattern of the polymeric particles to self-assemble on the implant surface due to electrostatic interactions.The adsorbed polymeric particles were further immobilized on the surface by heat treatment above the PS glass transition temperature in hot ethylene glycol (approximately 180 °C) for a few seconds.Afterwards, the size of the polymeric particles was reduced to 50 ± 7 nm in diameter by etching in a series of nitrogen and oxygen plasma (microwave plasma strip TePla 300PC, TePla AG, 150 W, 7-10 min).Furthermore, to achieve homogeneous chemistry on the implant surfaces, all implants (P, PN, M and MN) were sputter-coated (FHR MS150, 5 × 10 −5 mbar, 0.33 kW) with a 30 nm-thick Ti layer.Finally, implants were annealed at 500 °C (High Temperature Furnace, AWF 12/65, Lenton, Parsons Lane, Hope, UK) for five hours and kept in 70% ethanol until surgery [16] .

Surgery and implant retrieval
The local ethical committee for Animal Research in Gothenburg (Dnr 36/2012) approved the study, and the animal experiments followed the ARRIVE guidelines.A total of 48 male Sprague-Dawley rats (weight range 250-380 g) underwent an operation.Surgery was conducted under isoflurane inhalation general anesthesia (4.1%, air flow 650 mL/min) administered in a Univentor 410 unit and maintained by continuous isoflurane inhalation via a mask (2.3%, air flow of 450 mL/min).After shaving, cleaning with chlorhexidine (0.5 mg/ml), and injecting anesthetic (1 mL lidocaine with epinephrine; 10 mL/mL + 5 μg/mL) locally, the medial side of the proximal tibial metaphysis was surgically exposed.The implant recipient sites were prepared using round burs at a low drill speed under profuse irrigation of saline.A 1.6 mm diameter round bur was used to prepare the sites for the P and PN implants, whereas 1.6 mm followed by 1.8 mm diameter round burs were used for the M and MN implant recipient sites.The final size of the drill hole was decided based on pretests, considering a minor reduction in the implant diameters for the P and PN implants due to the electropolishing process.In order to exclude potential systematic errors related to biomechanical evaluation, each tibia received two implants following a predetermined schedule, ensuring randomization for the 4 implant types between proximal, distal, right and left tibia sites.Each rat received the four types of experimental implants, with 7-mm distance between the center of the two implants in each tibia in order to maintain safety margin between the adjacent sites.Thereafter, the surgical wound was closed using internal resorbable polyglactin suture (Vicryl 4-0) and external transcutaneous resorbable poliglecaprone suture (Ethilon Monocryl 4-0, Ethicon Inc, Sommerville, USA).All rats received subcutaneous buprenorphine analgesic (Temgesic 0.03 mg/kg) postoperatively, were allowed to move freely and were allowed to consume food and drink water ad libitum .
Implant retrieval was performed after 6, 21 and 28 days of insertion.At each time point, 16 rats were euthanized with an overdose of intraperitoneal sodium pentobarbital (Pentobarbital sodium vet; APL 60 mg/ml) injection.For RTQ and gene expression analysis, the skin and subcutaneous tissues were carefully opened and reflected, torque was measured and then the implants were unscrewed and preserved for either qPCR or SEM analyses.For histological and histomorphometric analyses, the implant with the surrounding bone was dissected en bloc and preserved for subsequent histological preparations ( Table 1 ).

RTQ measurements
For RTQ analysis, 10 rats per time point were used.After surgical exposure of the implants, a hexagonal screwdriver connected to the torque test machine was fitted into the implant internal hexagon [ 20 , 21 ].The tibia was held stable, using a small vise, in a horizontal direction, perpendicular to the direction of the torque screwdriver.For each measurement, torque against the rotation angle was recorded and monitored in real time.Once the breakpoint was obtained, measurements continued to be obtained under constant rotation until complete failure was achieved.

Histology and histomorphometry
After 6, 21 and 28 days, 6 rats were sacrificed for histological evaluation of intact bone-implants blocks.After retrieval, the bone-implant blocks were fixed and dehydrated prior to plastic embedding (LR White).The implants were divided along their long axis (EXACT® cutting and grinding equipment, EXACT® Apparatebau GmbH & Co, Norderstedt, Germany).Ground sections were prepared as previously described [22] .The ground sections were stained with 1% toluidine blue before histological examination under an optical microscope (Nikon Eclipse E600, Japan).Histomorphometric analysis was conducted by measuring the percentage of bone filling the thread (i.e., bone area percentage, BA%) and the percentage of bone in direct contact with the implant surface (i.e., bone-implant contact, (BIC%).Briefly, for all screw threads, the total thread area and full-length of the implant perimeter were measured using analytical software (NIS Elements 4.12, Nikon, Japan).The mineralized bone areas filling the threads and the lengths of mineralized bone in contact with the implant were determined.Thereafter, the BA% and BIC% were determined on each side of the implants and the values were averaged per implant.

qPCR
The qPCR analysis followed the same procedure as the previous study focused on the early phase of osseointegration [16] .After 6, 21 and 28 days, following the RTQ measurements, the same implants were completely unscrewed using a manual hexagonal screwdriver (10 specimens/implant type/time point; n = 10).Retrieval was performed while taking strict precautions for RNA preservation [23] .The retrieved implants, with the adherent biological material, were preserved in RNALater solution until analysis.Implant-adherent cells were homogenized in RLT buffer with β-mercaptoethanol and a TissueLyser ® instrument (Qiagen GmbH, Hilden, Germany).Thereafter, the samples were centrifuged at 16,0 0 0 g for 3 min, and the total RNA was extracted from the aqueous phase using an RNeasy Micro Kit (Qiagen GmbH).Total RNA was reverse transcribed into cDNA using a Grandscript cDNA synthesis kit (TATAA Biocenter AB, Gothenburg, Sweden).Validated assays of genes of interest, purchased from TATAA Biocenter, were used to target tumor necrosis factor-alpha (TNF-α), interleukin-1 beta (IL-1 β), alkaline phosphatase (ALP), osteocalcin (OC), cathepsin K (CatK), tartrate-resistant acid phosphatase (TRAP), receptor activator of nuclear factor-kappa B (RANK), receptor activator of nuclear factor-kappa B ligand (RANKL), osteoprotegerin (OPG) and caspase 3. Furthermore, prior to qPCR analysis, a reference panel was screened in randomly selected samples, representing all experimental groups and time points.The software geNorm [24] and Normfinder [25] were used to determine the best stable reference gene(s) for normalization.Based on the analysis, the most stable expression of reference genes was the combination of YHWAZ and GAPDH.qPCR analysis was then performed on all samples with assays targeting 9 different genes and the selected 2 reference genes.
Analysis was performed in a 10 μL reaction volume in duplicate on a CFX96 platform (Bio-Rad Laboratories, Inc., Hercules, USA) with TATAA SYBR Grand Master Mix (TATAA Biocenter AB).The quantities of the main genes were normalized to the arithmetic mean expression of the selected reference genes.Normalized relative quantities were calculated using the delta-delta C q method and 90% PCR efficiency (k * 1.9 cq ) [26] .

Backscattered electron SEM (BSE-SEM)
Selected implant-tissue blocks from the 28-day time point were polished using 40 0-40 0 0 grit silicon (Si) carbide paper and used for BSE-SEM imaging in a Quanta 200 environmental SEM (FEI Company, The Netherlands), operated at 20 kV accelerating voltage and 0.5 Torr water vapor pressure.

Statistics
The data was evaluated for normality using the Kolmogorov-Smirnov and Shapiro-Wilk tests.Both tests revealed that data was mostly not normally distributed, and hence non-parametric tests were employed.The Kruskal-Wallis test followed by the Mann-Whitney U test were applied to examine significant differences in RTQ, histomorphometry (BIC% and BA%) and gene expression among the different implant types, at each time point, and among the different time points for each implant type.Furthermore, a Spearman correlation analysis was conducted to determine potential associations between different genes denoting inflammation, bone formation and remodeling (data provided in Supplementary Results and Table S1).In addition, Pitman ś test was used to evaluate major temporal trends among the three time points.As no specific temporal trends could be observed, the time points were pooled, thereby allowing additional analyses with enhanced statistical power.The pooled data were also evaluated using two-way multivariate analysis of variance (MANOVA) to statistically determine whether the effects of nano and machine-(micro) topographies were dependent on each other and whether there were interaction effects between the two variables on all dependent variables (RTQ, histomorphometry and gene expression).Only dependent variables that showed significant differences in comparative tests of the pooled data were evaluated by two-way MANOVA.All statistical tests were conducted using IBM® SPSS® Statistics Version 25. p-values < 0.05 were considered statistically significant.In gene expression analysis, some samples were excluded, particularly from the 6-and 21-day time points, either due to undetectable C q values or a large discrepancy between the qPCR technical duplicates ( > 0.35 standard deviation) for the individual sample.An overview of the analytical methods, the number of animals and the number of statistically analyzed samples are presented in Table 1 .

Surface characterization
SEM evaluation showed successful removal of the microgrooves of the machined surface by the electropolishing procedure ( Fig. 1 ).However, electropolishing also resulted in reduction of outer diameter of the implant (by approximately 12%).Profilometer quantitative evaluation of implant roughness revealed that electropolishing decreased surface roughness parameter S q from 181 ± 11 nm to 56 ± 7 nm, which correspondingly influenced other surface parameters ( Table 2 ).Due to limited lateral resolution, the detection of nanopatterns on MN and PN surfaces was not feasible using optical profilometry.Therefore, these nanopatterns were qualitatively assessed using SEM ( Fig. 1 ) and further subjected to quantitative evaluation by image analysis and modeling.The SEM assessment showed that the semispherical nanostructures were uniformly covering the implant surfaces, having the same mean diameter (75 ± 8 nm) and distribution density (50 ± 3 particles/μm 2 ) on both MN and PN surfaces.The nanostructures covered about 18 ± 4% of the surface and increased the exposed surface area by 50 ± 11%.It was also found that the heat treatment of the sputtered Ti coating induced some nanoroughness.This nanoroughness, induced by heat treatment, was measured on Ti-coated polished Si wafers using AFM ( Table 2 ).After optimizing the sputtered Ti thickness, annealing temperature and ramping, the level of the heat-induced nanoroughness could be kept at S q = 1.4 nm that is considerably below the nanopattern height.
XPS revealed a predominance of Ti oxide (20% Ti and ∼46% O) and carbon ( ∼31%) on the differently prepared surfaces ( Table 2 ).High-resolution shifts of carbon suggest that 80% of detected carbon was related to hydrocarbon compounds, which are typical contaminants due to ambient air.Traces of other contaminants ( < 2%) were detected on all implant surfaces, in similar amounts.
Using TOF-SIMS, it was verified that the detected carbon was not due to PS [16] .Prior to annealing, the nanopatterned surfaces revealed approximately two times higher relative amounts of PSrelated ions as compared to flat controls without PS.After annealing, the detected amounts of PS-related ions were similar for all nanopatterned surfaces and controls, confirming identical chemical composition for all implant groups used in the study, despite the differences in microscale roughness and nanoscale topography.

RTQ analysis
Over time, all implant types, except PN, demonstrated a significant temporal increase in torque values from 6 days -21 days ( p = 0.002 for P , 0.3 for PN , 0.0 0 04 for M and 0.0 0 04 for MN), and no further temporal changes were observed between 21 and 28 days ( p > 0.05 ).All implants showed significantly higher torque values at 28 days than at 6 days ( p = 0.005 for P , 0.01 for PN , 0.0 0 04 for M and 0.004 for MN) ( Fig. 2 ).After 6 days, the highest RTQ was recorded for the MN group (MN vs. P ( p = 0.003 ), MN vs. PN ( p = 0.02 ), and MN vs. M ( p = 0.001 )) ( Fig. 2 ).After 21 days, the MN group showed higher RTQ values than both polished groups, with and without nanopatterning ( p = 0.01 vs. both PN and P) ( Fig. 2 ).At the late healing time point, 28 days, M implants achieved torque values comparable to those of MN implants ( Fig. 2 ).At this time point, a statistically significant difference was demonstrated for M implants versus PN implants ( p = 0.04 ) ( Fig. 2 ).
When the three time points were pooled for each implant type, both machined groups (M and MN) showed higher RTQ values than the polished groups (P and PN).MN implants demonstrated significantly higher values than P and PN implants (Supplemen-   tary Fig. S1A).Here, the PN group revealed the lowest RTQ values among all surfaces.

Histology and histomorphometry
After 6 days of healing, the tissue around all types of implants appeared well organized with no evident inflammatory infiltrates ( Fig. 3 ).At this time point, new bone formation was observed, which predominantly appeared as woven bone, with large, rounded osteocytes and a lack of lamellar structure.A considerable amount of newly formed woven bone covered with osteoid was observed de novo in the marrow part of the implant, which is between the implant threads ( Fig. 3 ).To a lesser extent, newly formed woven bone covered with thin layers of osteoid could be observed in conjunction with old trabeculae or fragments from cortical bone that had been displaced into the marrow during the drilling procedure and implant installation.Although qualitatively evaluated, a relatively higher proportion of osteoid appeared to be associated with MN implants.
After 21 and 28 days, the histological picture was comparable for all implant groups; more specifically, the woven bone had undergone remodeling to a considerable extent, resulting in mature lamellar bone surrounding the different types of implants ( Fig. 4 ).The osteocytes within this mature bone appeared elongated and smaller in size than those observed after 6 days.
Histomorphometric analysis showed that at 6 and 21 days, both machined groups (M and MN) were associated with a higher BIC than both polished groups (P and PN) ( Fig. 5 A).Nevertheless, a statistically significant difference in BIC% was only demonstrated for MN implants in comparison to P implants at the 6-day time point ( p = 0.01 ) ( Fig. 5 A).At 28 days, a significantly higher BIC was shown for MN implants than for PN implants ( p = 0.04 ).At this time point, PN implants also revealed a significantly lower BIC than P implants ( p = 0.02 ).Only P implants revealed a significant temporal increase in BIC over time, whereas the temporal changes in all other groups were gradual and did not reach statistical significance ( Fig. 5 A).
In contrast, a significant temporal increase in BA% was demonstrated from 6 days to 21 days, which remained unchanged at 28 days for all implant types ( Fig. 5 B).Apart from the finding that a significantly higher BA% was shown for P versus MN implants after 21 days ( p = 0.02 ), no statistically significant differences were found among groups.
After pooling the 3 time points for each implant surface, the lowest BIC% was found in the PN group, whereas the highest BIC% was observed in the MN group (Supplementary Fig. S1 B).The pooled data demonstrated higher BIC% for MN implants than for PN implants ( p = 0.004 ) (Supplementary Fig. S1B).No significant differences were observed for BA% among the different surfaces when the time points were pooled (Supplementary Fig. S1 C).

Gene expression analyses
The relative gene expression of cells adherent to the different implant types at each time point is presented ( Fig. 6 ).The gene data were divided into two panels, denoting major biological processes: 1) inflammation and apoptosis and 2) bone formation and remodeling.Generally, only a few significant differences were found among the implant groups or the time points.

Expression of proinflammatory and apoptosis genes
At 6 days and 21 days, the gene expression of TNF-α was at the highest level in the M group, whereas the lowest level was found in the MN group ( Fig. 6 A).Particularly at 21 days, MN implants demonstrated a significant 1.8-fold downregulation of TNF-α compared with M implants ( Fig. 6 A).Furthermore, at 28 days, a significant 1.4-fold reduction of TNF-α expression was found in cells adherent to PN implants in contrast to P implants ( Fig. 6 A).Temporally, whereas most of the surfaces revealed TNF-α downregulation after 28 days compared to 6 days or 21 days, the temporal expression of TNF-α on the P surface did not show significant temporal changes over time ( Fig. 6 A).No significant differences were found in the gene expression of IL-1 β ( Fig. 6 B) or caspase 3.
When gene expression data were pooled with respect to time points, TNF-α expression showed the lowest expression level in cells adherent to the MN surface, with a statistically significant difference in comparison to the P surface (Supplementary Fig. S2  A).Otherwise, similar to nonpooled data, no significant differences were observed in IL-1 β (Supplementary Fig. S2 B) or caspase 3 gene expression.

Expression of bone formation and bone remodeling genes
Among all analyzed genes related to bone formation and bone remodeling, the most prominent observation was found for RANKL gene expression after 6 days ( Fig. 6 ).Here, the lowest RANKL expression level was noted for MN implants, which had a significant 4-and 3-fold reduction in expression compared to M and PN implants, respectively ( Fig. 6 H).In contrast, in the absence of nanotopography, M implants demonstrated a significant 3-fold increase in RANKL expression compared to P implants ( Fig. 6 H).In addition to the modest significant temporal changes observed in RANK gene expression, no significant differences were observed in any of the other analyzed bone formation (ALP and OC) or remodeling genes (TRAP, CatK, OPG or RANKL/OPG ratio) ( Fig. 6 ).
When the gene expression data were pooled for the time points, RANKL revealed a surface-dependent gene expression response, with the lowest statistically significant expression level in the MN group compared to the expression level of all other surfaces (P, PN and M) (Supplementary Fig. S2 H).

MANOVA for main effects and interaction effects
MANOVA analysis revealed differential effects of nanotopography, machined topography and/or their interaction depending on the variables evaluated when pooling the time points (Supplementary Table S2, Supplementary Fig. S3).The RTQ variable revealed a significant main effect from the machined microtopography but not the nanotopography per se (Table S1).However, a significant interaction effect of the nano and machined microtopographies was  G, H, and I) and machined nanopatterned (MN) (J, K, and L) implants 6 days after implantation in rat tibiae.After 6 days of healing, the bone formed at the interface was mainly of the woven type, lined with osteoblastic seams, and had large embedded osteocytes.A considerable amount of osteoid could be observed in the marrow space, formed de novo between implant threads (black arrows).The newly formed osteoid was also found on the surfaces of the cortical bone (black arrowheads) and the bone fragments (white asterisks).The new bone was evidently formed via intramembranous ossification, and no cartilaginous tissue could be observed.At this time point, no evident inflammatory infiltrate could be observed, and the extent of new bone formation appeared comparable, irrespective of the implant type.D, H) implants after 28 days of implantation in rat tibiae.)The histologic picture after 28 days of healing appeared similar to that at 21 days, showing osseointegration with mature bone around the implants.Mature bone harbored osteocytes in elongated lacunae that appeared to run parallel to the implant surface.Only minor amounts of osteoid could be visualized (white and black arrows), likely as a result of ongoing bone remodeling.observed on RTQ (Supplementary Table S2, Supplementary Fig. S3 A).For the BIC variable, a significant main effect on BIC was only found for machined microtopography, whereas neither the main effect of nanotopography nor the interaction effect of nanotopog-raphy and machined topography were found to be significant on BIC (Supplementary Table S2, Supplementary Fig. S3 B).In contrast, both the nanotopography and the machined topography revealed significant main effects on the gene expression of TNF-α (Supplementary Table S2).Moreover, a significant interaction effect of the nano and machined microtopographies was demonstrated on the expression of TNF-α (Supplementary Table S2, Supplementary Fig. S3 C).No major significant main or interaction effects were found for the expression of RANKL (Supplementary Table S2, Supplementary Fig. S3 D).

BSE-SEM and RCE of retrieved implants
BSE-SEM of 28 day samples revealed that woven bone was largely replaced by remodeled mature lamellar bone within the implant threads for all implant groups ( Fig. 7 ).A few large osteocytes attributable to remnants of woven bone were nevertheless evident.Osteocytes adjacent to the implant surface were aligned parallel to the implant surface contour.Regions of ongoing remodeling were also noted within implant threads.
The osteocyte lacuno-canalicular network was directly observed via RCE ( Fig. 7 ).Osteocytes in bone within implant threads were arranged into highly interconnected networks with canaliculi extending ∼4-6 μm in the direction of the implant surface.

Discussion
The present study addressed the role of controlled nanotopography on three major aspects of osseointegration: (i) functional aspects, by measuring implant stability; (ii) morphological aspects, judged by the percentage of bone in contact with the implant surface; and (iii) molecular aspects, by determining the expression levels of genes involved in inflammation and bone regeneration in implant-adherent cells.Furthermore, the present study explored a previously unresolved issue of whether the effect of nanotopography is dependent on or independent of the underlying microtopography.A major observation was that the nanotopography superimposed on machined microtopography promoted an early increase in implant stability, as determined by RTQ analysis.This surfacedriven effect was confined to the early period of woven bone formation and the concomitant remodeling (6-21 days) typically observed in this experimental model.In line with this observation, an increased proportion of bone in direct contact with the implant surface was promoted by the same nanopatterned machined implant.This is in agreement with data from previous studies that demonstrated that BIC is a determinant and predictor of implant stability and RTQ [29] .The present data suggest that the enhanced BIC and RTQ of combined micro-and nanopatterned implants were due to enhanced osteogenic differentiation.Nevertheless, gene expression analysis of implant-adherent cells did not reveal higher osteogenic gene expression (ALP and OC) between 6 and 21 days in comparison to that of machined or other surfaces.On the other hand, an increase in osteogenic differentiation and the bone formation marker OC in cells adherent to MN implants was previously detected as early as 3 days after implantation [16] , indicating an early but transient pro-osteogenic effect of nanopatterned features, in vivo .Another explanation for the higher proportion of bone in contact with the combined micro-and nanopatterned implants is attributed to reduced osteoclastic resorption activities, leading to a net effect of more bone remaining in contact with the implant surface.This assumption is supported by the current observation of downregulated RANKL expression in cells adherent to the combined micro-and nanopatterned implant.In addition, a relatively lower expression, albeit not significant, of CatK was found in the present study.Downregulated CatK expression has previously been reported after 6 days in cells adherent in vivo to surfaces with nanostructures having identical shape, size and coverage as the surfaces in the present study [30] .Moreover, in search of the role of a smaller nanotopography pattern (15 nm nanopillars fabricated by self-assembly on polished titania), cocultures of osteoblasts and osteoclasts revealed enhanced osteogenesis-related gene expression and downregulation of osteoclastic genes, whereas an in vivo experiment in rabbit femora showed increased BIC for nanopatterned than for polished implants [31] .Together, the available data indicate that nanotopography in the ranges below 100 nm supports early osteogenesis and/or reduces osteoclastic activity in implantadherent cells.Furthermore, the reduced osteoclastic activity may at least in part be attributed to decoupling between osteoblasts and osteoclasts via the downregulation of the RANKL coupling factor.The sum of these activities results in an increased percentage of bone in contact with the nanopatterned surface and increased stability of the implant in the recipient site.
An important finding in the present study was the reduced TNF-α expression in implant-adherent cells on the combined micro-and nanopatterned implant, confirming previous observations of the downregulatory effects of nanotopography on the proinflammatory molecular response at the bone-implant interface [30] .This observation is of particular importance by virtue of previous data demonstrating an advantage of downregulation of proinflammatory cytokines in implant-adherent cells to achieve accelerated and stronger osseointegration [ 7 , 32 ].Although histology showed no major periimplant inflammation at any of the time periods, no further analyses of inflammatory cell types and their subsets were pursued in the present study.The cytokine profile of M1 macrophages is dominated by high secretion of TNF-α and Il-1 β.
Low secretion of TNF-α can lead to proinflammatory macrophages to transition to reparative macrophages (M2) [33] , which is associated with wound healing, osteogenesis and tissue remodeling [34] .
A recent in vitro study has shown that Ti surfaces with 80 nm nanotubes decrease the number and cell activity of M2 macrophages [35] .This modulation was highly dependent on the size of the nanostructures and more pronounced at larger nanotubes with a threshold of 80 nm.The latter finding is in agreement with previous observations in vivo (using an identical nanopattern as the one used in the present study) showing fewer periimplant CD163positive macrophages (surface marker of M2 macrophages) around nanopatterned surfaces compared with machined surfaces [30] .On the other hand, fewer CD68-positive macrophages (representing both M1 and M2 phenotypes) were found around similar nanopatterned surfaces in vivo [16] .Apart from mesenchymal stem cells, macrophages arrive early at the implant surface-bone interface.Interestingly, the gene expressions related to inflammation and bone remodeling revealed multiple associations, for example (TNFα and RANKL).The role of different nanoscale surface properties in macrophage polarization, cell-cell communication, interaction between inflammation, bone remodeling and osseointegration therefore warrants further research.From a scientific point of view, it is fundamental to discriminate the specific effect of the nanopattern from the underlying microscale surface roughness.To the best of our knowledge, the use of a controlled anodization process is the only other example of a controlled nanotopography that has been successfully translated as a semi-ordered nanopattern on cylindrical [36] and screw-shaped [37][38][39] implants and evaluated in vivo .When anodized nanotubular patterns were fabricated on polished screw-shaped titanium implants, the pattern with 30 nm diameter nanofeatures promoted higher BIC after early and late time points, whereas the 80 nm nanopattern revealed the opposite effect with reduced BIC in contrast to polished control [ 37 , 38 ].This finding is in agreement with finding in the present study where the semi-spherical nanopattern, of 75 nm diameter, on the polished surface was associated with lower BIC as compared to the corresponding polished surface without nanopattern.In contrast, when 30, 70 and 100 nm nanotubular patterns were superimposed on machined cylindrical titanium implants and evaluated in minipig calvaria model, the highest BIC was achieved with the 70 nm diameter nanopattern compared to the machined surface without nanopattern [36] .The latter observation is consistent with the current observation that machined nanopatterned (MN) screw implants promoted high BIC values.In addition, an indication that nanopatterns provide a synergistic effect with the underlying microscale roughness is provided in another in vivo study, in rabbit femur, where nanotubular pattern of 56 nm diameter superimposed on grit-blasted microrough screwshaped Ti-6Al-4V implants promoted higher BIC% in comparison to equivalent grit-blasted microrough screws without the nanopattern [39] .Taken together, these findings suggest that the effect of controlled nanotopography on BIC is largely influenced by the type of the underlying surface topography.
A major strength of this work was the possibility of discriminating the effects of controlled nanoscale topography and microscale roughness.According to the MANOVA model, although machined microroughness had a major effect on the RTQ, the significant interaction effect supports a combined influence of nanoscale topography and microscale roughness on implant stability in the recipient site.On the other hand, the degree of BIC was mainly dependent on microscale roughness, whereas nanotopography did not provide significant feature according to the MANOVA test.Furthermore, an important finding was the observed significant main effect of nanotopography on the gene expression of TNF-α, which not only provides additional evidence for the role of nanotopography per se in attenuating the inflammatory response but also reflects the sensitivity of analyzing the gene expression of implantadherent cells.
Most studies on the role of nanotopography for implants in bone have been performed in vitro whereas a minority of studies have been executed in different experimental animal models.To the authors ḱnowledge no studies have determined the role of nanotopography for osseointegration in humans.The clinical relevance of the results from the present study is not fully investigated and further studies are needed to assess the potential advantage if used clinically.We believe that a stronger bone adhesion using nano patterned implant surfaces would be an advantage in compromised conditions, e.g.poor bone quality and in grafted bone.

Conclusions
Well-defined nanotopography superimposed on microrough, clinically relevant, screw-shaped Ti implants promotes bone apposition and implant stability during the development of osseointegration.Furthermore, though increased stability is influenced by microscale roughness and the combination of nanoscale topography and microscale roughness, the increased BIC observed with the combined nanopatterned machined surface is mainly dependent on the microscale roughness.At the molecular level, the nanotopography, per se , and in synergy with microscale roughness, attenuates the expression of the proinflammatory cytokine TNF-α.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Scanning electron microscopy of polished (top) and machined (middle) implants with superimposed nanotopography before annealing.The bottom image shows the machined implant with superimposed nanotopography after annealing.

Fig. 2 .
Fig. 2. Removal torque analysis.Removal torque was measured at the interface between the recipient bone and polished (P), polished nanopatterned (PN), machined (M), and machined nanopatterned (MN) implants after 6, 21 and 28 days of implantation in rat tibiae.The column graphs show the mean and the standard error of the mean (n = 10).Statistically significant differences ( p < 0.05) are indicated with connecting bars and asterisks or lowercase letters.Connecting bars and asterisks ( * ) = statistically significant difference between two implant types at each time point; lowercase letters (a-g) = statistically significant difference between time points.

Fig. 3 .
Fig. 3. Histological evaluation after 6 days of implantation.Light micrographs of ground sections stained with toluidine blue showing the morphology of the bone interface with polished (P) (A, B, and C), polished nanopatterned (PN) (D, E, and F), machined (M) (G, H, and I) and machined nanopatterned (MN) (J, K, and L) implants 6 days after implantation in rat tibiae.After 6 days of healing, the bone formed at the interface was mainly of the woven type, lined with osteoblastic seams, and had large embedded osteocytes.A considerable amount of osteoid could be observed in the marrow space, formed de novo between implant threads (black arrows).The newly formed osteoid was also found on the surfaces of the cortical bone (black arrowheads) and the bone fragments (white asterisks).The new bone was evidently formed via intramembranous ossification, and no cartilaginous tissue could be observed.At this time point, no evident inflammatory infiltrate could be observed, and the extent of new bone formation appeared comparable, irrespective of the implant type.

Fig. 4 .
Fig. 4. Histological evaluation after 28 days of implantation.Light micrographs of ground sections stained with toluidine blue showing the morphology of the bone interface with polished (P) (A, E), polished nanopatterned (PN) (B, F), machined (M) (C, G) and machined nanopatterned (MN) (D, H) implants after 28 days of implantation in rat tibiae.)The histologic picture after 28 days of healing appeared similar to that at 21 days, showing osseointegration with mature bone around the implants.Mature bone harbored osteocytes in elongated lacunae that appeared to run parallel to the implant surface.Only minor amounts of osteoid could be visualized (white and black arrows), likely as a result of ongoing bone remodeling.

Fig. 5 .
Fig. 5. Histomorphometric analysis.The histomorphometric measurements of the bone-implant contact percentage (BIC%; A) and bone area percentage (BA%; B) were performed on the bone interface with polished (P), polished nanopatterned (PN), machined (M), and machined nanopatterned (MN) implants after 6, 21 and 28 days of implantation in rat tibiae.The column graphs show the mean and the standard error of the mean (n = 6).Statistically significant differences ( p < 0.05) are indicated with connecting bars and asterisks or lowercase letters.Connecting bars and asterisks ( * ) = statistically significant difference between two implant types at each time point; lowercase letters (a-g) = statistically significant difference between time points.

Fig. 7 .
Fig. 7. Scanning electron microscopy.(A-H) Backscattered electron (BSE) imaging shows the presence of mature, mineralized, remodeled, lamellar bone within the implant threads at 28 days after implantation.(I-L) Secondary electron (SE) imaging following resin cast etching reveals the presence of networks of intercommunicating osteocytes (Ot) and blood vessels ( * ) in close proximity to the implant surface (Ti).Polished (P): A, E, I and M; Polished with nano (PN): B, F, J and N; Machined (M): C, G, K and O; Machined with nano (MN): D, H, L and P. Scale bars in A-D = 500 μm, E-H = 250 μm, I-L = 100 μm, and M-P = 20 μm.

Table 1
Number of rats and samples used for the different analyses.

Table 2
Topographic and chemical characterization of implants. *