Proteomic Dissection of Nanotopography-Sensitive Mechanotransductive Signaling Hubs that Foster Neuronal Differentiation in PC12 Cells

Neuronal cells are competent in precisely sensing nanotopographical features of their microenvironment. The perceived microenvironmental information will be “interpreted” by mechanotransductive processes and impacts on neuronal functioning and differentiation. Attempts to influence neuronal differentiation by engineering substrates that mimic appropriate extracellular matrix (ECM) topographies are hampered by the fact that profound details of mechanosensing/-transduction complexity remain elusive. Introducing omics methods into these biomaterial approaches has the potential to provide a deeper insight into the molecular processes and signaling cascades underlying mechanosensing/-transduction but their exigence in cellular material is often opposed by technical limitations of major substrate top-down fabrication methods. Supersonic cluster beam deposition (SCBD) allows instead the bottom-up fabrication of nanostructured substrates over large areas characterized by a quantitatively controllable ECM-like nanoroughness that has been recently shown to foster neuron differentiation and maturation. Exploiting this capacity of SCBD, we challenged mechanosensing/-transduction and differentiative behavior of neuron-like PC12 cells with diverse nanotopographies and/or changes of their biomechanical status, and analyzed their phosphoproteomic profiles in these settings. Versatile proteins that can be associated to significant processes along the mechanotransductive signal sequence, i.e., cell/cell interaction, glycocalyx and ECM, membrane/f-actin linkage and integrin activation, cell/substrate interaction, integrin adhesion complex, actomyosin organization/cellular mechanics, nuclear organization, and transcriptional regulation, were affected. The phosphoproteomic data suggested furthermore an involvement of ILK, mTOR, Wnt, and calcium signaling in these nanotopography- and/or cell mechanics-related processes. Altogether, potential nanotopography-sensitive mechanotransductive signaling hubs participating in neuronal differentiation were dissected.

Neuronal cells are competent in precisely sensing nanotopographical features of their microenvironment. The perceived microenvironmental information will be "interpreted" by mechanotransductive processes and impacts on neuronal functioning and differentiation. Attempts to influence neuronal differentiation by engineering substrates that mimic appropriate extracellular matrix (ECM) topographies are hampered by the fact that profound details of mechanosensing/-transduction complexity remain elusive. Introducing omics methods into these biomaterial approaches has the potential to provide a deeper insight into the molecular processes and signaling cascades underlying mechanosensing/-transduction but their exigence in cellular material is often opposed by technical limitations of major substrate top-down fabrication methods. Supersonic cluster beam deposition (SCBD) allows instead the bottom-up fabrication of nanostructured substrates over large areas characterized by a quantitatively controllable ECM-like nanoroughness that has been recently shown to foster neuron differentiation and maturation. Exploiting this capacity of SCBD, we challenged mechanosensing/-transduction and differentiative behavior of neuron-like PC12 cells with diverse nanotopographies and/or changes of their biomechanical status, and analyzed their phosphoproteomic profiles in these settings. Versatile proteins that can be associated to significant processes along the mechanotransductive signal sequence, i.e., cell/cell interaction, glycocalyx and ECM, membrane/f-actin linkage and integrin activation, cell/substrate interaction, integrin adhesion complex, actomyosin organization/cellular mechanics, nuclear organization, and transcriptional regulation, were affected. The phosphoproteomic data suggested furthermore an involvement of ILK, mTOR, Wnt, and calcium signaling in these nanotopography-and/or cell mechanics-related processes. Altogether, potential nanotopography-sensitive mechanotransductive signaling hubs participating in neuronal differentiation were dissected.

INTRODUCTION
Cells can sense nanotopographical cues deriving from the extracellular matrix (ECM), predominantly through integrin adhesion complexes (IAC), and the microenvironmental information is converted into alterations of the cytoskeletal organization and mechanics. Mechanosensitive signaling cascades and nuclear rearrangements eventually translate these modulations into cellular responses; this entire sequence is defined as mechanotransduction. Mechanosensing of the substrate nanotopography and the subsequent mechanotransduction strongly impact many physiological cellular behaviors, in particular in the context of cell differentiation (Geiger et al., 2009;Wang et al., 2009;Chen et al., 2014;Dalby et al., 2014;Jansen et al., 2015), but they might also influence pathophysiological processes, such as metastatic cell migration (Park et al., 2016).
Many aspects regarding the cellular capacity of sensing nanoscale topographical cues and how the information is integrated into complex mechanotransductive signals are still largely unknown (Chen et al., 2014;Dalby et al., 2014). Artificial nanostructured surfaces produced by diverse topdown microfabrication techniques (typical of the semiconductor industry) have been useful tools that helped to understand principal surface topography-related parameters controlling mechanotransductive processes (Chen et al., 2014;Dalby et al., 2014) which are hard to access in vivo. This approach has two major limitations. First of all, starting from simple surface motifs it is extremely difficult to reconstruct the morphological complexity of the ECM. Secondly, achieving an in-depth comprehension of the mechanotransductive processes and signaling requires systematic high-throughput and omic approaches (Cranford et al., 2013;Groen et al., 2016). Admittedly, many micro-/nanofabrication techniques have technical limitations hindering a feasible and cost-effective scaleup necessary to provide sufficiently large surface areas with a defined nanotopography (Chen et al., 2014). However, the yield of a reasonable amount of cellular material is mandatory for the implementation of omics approaches.
In this framework, we use the bottom-up nanofabrication method supersonic cluster beam deposition (SCBD) to quantitatively address the influence of nanoscale surface topography on mechanotransduction. SCBD permits the engineering of nanostructured surfaces with a reproducible nanoscale roughness parameter by assembling transition metal oxide clusters (Schulte et al., 2017), thereby realizing topographies that mimic ECM nano-features (Gasiorowski et al., 2013). SCBD can be applied efficiently to produce biocompatible substrates (made by titania or zirconia clusters) on large Abbreviations: AFM, Atomic force microscopy; ECM, Extracellular matrix; FA, Focal adhesion(s); FBS, Fetal bovine serum; FC, Focal complex(es); HS, Horse serum; IAC, Integrin adhesion complex(es); IPA, Ingenuity pathway analysis; min, Minutes; mM, Milli molar; NGF, Nerve growth factor; ns-Zr, Nanostructured zirconia; PCA, Principal component analysis; PLL, Poly-L-lysine; RMS, Root mean square; SAC, Stretch-activated channels; SCBD, Supersonic cluster beam deposition; SD, Standard deviation; TrkA, Tropomyosin receptor kinase A; TRPC, Transient receptor potential cation channels. macroscopic areas rendering it compatible with profound biological analyses, such as proteomic studies (Schulte et al., 2017).
Recently, using PC12 cells as a broadly accepted model system for neuronal differentiation, we demonstrated that appropriate biophysical stimuli; provided by the cellular interaction with nanotopographical cues of titania or zirconia surfaces produced by SCBD, promote neuronal differentiation processes (Tamplenizza et al., 2013;Schulte et al., 2016a). This potential of the cluster-assembled surfaces was also confirmed in hippocampal neurons (Schulte et al., 2016b). In both cellular models we applied label-free shotgun proteomics as an essential technique to examine the impact of the nanotopography on cellular differentiation processes since this quantitative approach can achieve simultaneously: (a) the identification of thousands of proteins isolated from a cellular model and (b) the quantification of each protein/phosphosite. It is therefore well suited to study differences in global protein expression between different samples, providing substantial information to delineate and profoundly understand cell signaling pathways and modulations of the cellular program (Toffolo et al., 2014;Zanotti et al., 2016), in particular also in the context of integrin-mediated mechanotransduction (Humphries et al., 2015;Robertson et al., 2015).
Neuronal differentiation is a specifically interesting biological process in this mechanobiological context. It is accompanied by drastic morphological and cytoskeletal rearrangements throughout the realization of neurites, dendrites and axons, strongly controlled by point contact-mediated neuron/ECM interaction (Myers et al., 2011;Flynn, 2013;Kerstein et al., 2015). In fact, in PC12 cells the extent of nanotopographytriggered differentiation (even in the absence of a biochemical stimulus) was comparable to the canonical differentiation mediated by NGF-induced TrkA activation. In each case, either NGF-or nanotopography-induced, the outcome was the outgrowth of neurites and a differentiated cell. Our previous study furthermore revealed that in the latter condition, complex mechanotransductive events were at the basis of cellular processes that lead to the onset of neuritogenesis and neuronal differentiation. However, at the proteome level we only compared the nanostructured surface with a roughness parameter R q of 15 nm root mean square (RMS) (which induced the strongest neuritogenesis, called ns-Zr15 hereafter) against a flat zirconia surface (which even after NGF stimulus impeded neuronal differentiation, named flat-Zr herafter). The cells on ns-Zr15 were found to have small IAC (predominantly focal contact (FC) size), few to none stress fibers and a low cell rigidity, contrary to the large IAC (focal adhesion (FA) size), stress fibers and an increased cellular rigidity on flat-Zr (Schulte et al., 2016a).
Besides these two conditions (ns-Zr15 and flat-Zr), the proteomic analyses in this study comprise instead PC12 in more versatile experimental conditions including a surface nanotopography with higher roughness, the biochemically NGFinduced canonical neuronal differentiation and manipulations that affect the biomechanical status of the PC12 cell. The characteristics of these additional experimental conditions evaluated in this extended proteomic analyses are (summarized in Figure 1): (1) The surface with an increased roughness R q of 25 nm RMS (ns-Zr25) has asperities that display subtle differences in diameter and dimension compared to ns-Zr15. This roughness induced neuritogenesis to a lower extent with respect to ns-Zr15 (Schulte et al., 2016a). (2 and 3) As canonical reference, representing the broadly studied biochemically triggered neuritogenesis and neuronal differentiation, PC12 cells on PLL-coated glass, in the absence (PLL) or presence of NGF (NGF), were introduced into the analysis. The cells exhibited large IAC (FA size), stress fibers and an intermediate cell rigidity without the NGF stimulus. The IAC size (to FC), stress fibers frequency and cellular rigidity decreased upon NGF-induced differentiation (Schulte et al., 2016a).
Moreover, the role of cellular biomechanics in this mechanotransductive sequence was approached by adding two conditions affecting the cellular tension: (4) Cells grown on ns-Zr15 in hypoosmotic medium to increase the cellular tension by osmotic swelling which counteracts the lower rigidity of the cells on ns-Zr15 and inhibits the nanotopography-triggered neuritogenesis (Schulte et al., 2016a). (5) Cells on PLL exposed to a short hyperosmotic shock (decrease in cellular tension), a treatment that morphologically triggered the outgrowth of neurites ( Figure S1).
A correlation of these alterations in cellular and topographical characteristics to changes in the PC12 neuronal proteome allowed us to obtain a deeper understanding of cellular nanotopography sensing and mechanotransductive signal integration. We were able to define potentially relevant surface nanotopography-sensitive mechanotransductive proteins and signaling networks/hubs that might play a key role in the signal FIGURE 1 | Representations of the cell morphology in the different conditions and a summary of the results presented in a previous publication (Schulte et al., 2016a). The figure summarizes the results of our previous publication (Schulte et al., 2016a) which provided the basis for the selection of the experimental conditions of the extended phosphoproteomic analyses of this work. In the upper row representations of the surface nanotopographies are displayed which were also used in this work (PLL-coated glass, flat-Zr, ns-Zr15, and ns-Zr25). Underneath example photos demonstrate the morphology of PC12 cells in the indicated conditions. In the table the impact of these different conditions on examined cellular parameter are recapitulated (FA, focal adhesions; FC, focal complexes; IAC, integrin adhesion complex; n.a., not analyzed).
integration driving, in this case, neuronal differentiation. For the sake of clarity (considering the many diverse conditions), in the main text we will talk more generally about signaling pathways and cellular processes affected or modulated in the different experimental settings, highlighting only a few particularly interesting proteins dissected from the global picture that introduce new aspects. Further examples of proteins and information on their to-date reported functions relevant in this context will be provided in the corresponding tables we refer to. The precise analysis of biomolecular events triggered by cell/nanotopography interaction, combined with the technical capacities of SCBD, constitutes the necessary foundation for efficient near-future exploitation of SCBD for versatile bio-applications.

Substrate Preparations
As a basis for all substrates standard microscope glass slides with the dimensions of 76 × 26 mm (surface area ∼20 cm 2 ) were used. On this carrier, we produced the cluster-assembled nanostructured films by supersonic cluster beam deposition (SCBD) of zirconia clusters obtained through a pulsed microplasma cluster source. Specific details on this bottomup nanofabrication approach can be found in Wegner et al. (2006) and Schulte et al. (2017). These cluster-assembled zirconia surfaces are given the abbreviation ns-Zr throughout the manuscript. The number after Zr indicates the roughness parameter R q . Two batches were fabricated for this work with roughness parameters of 15 nm RMS (ns-Zr15) and 25 nm RMS (ns-Zr25). The roughness and the morphological parameters have been systematically characterized by atomic force microscopy (AFM) (Podestà et al., 2015;Borghi et al., 2016). The capacity of SCBD to reliably cover large macroscopic areas with nanostructured films of a predefined roughness allowed us to perform the experiments on microscope glass slides with the dimensions of 76 × 26 mm (∼20 cm 2 surface area). This rendered possible the yield of sufficient cellular material from the different experimental situations to obtain the analyzed information, e.g., also data on the phosphorylation status of the proteins.
The flat zirconia surfaces (flat-Zr) with a roughness of ∼0.4 nm RMS were obtained with electron beam evaporation. For the canonical reference, the microscope glass slides were coated with poly-L-lysine (PLL) (Sigma-Aldrich, St. Louis, USA, Missouri) for 30 min at room temperature (RT), after cleaning with 70% ethanol and washing twice with PBS. This coating was done directly before plating the cells.
All substrates were sterilized with UV light for 10 min before seeding the cells.
For the experiments the PC12 cells were detached with 1 mM EDTA in HBSS and centrifuged at 1,000× g (5 min), washed with low serum medium (RPMI-1640 with all the supplements, but only 1% HS and without FBS), and centrifuged again at 1,000× g (5 min). Before plating the cells on the different substrate conditions, the cells were counted with an improved Neubauer chamber and then seeded with the concentration of ∼4,000 cells/cm 2 (after resuspension in RPMI low serum medium) onto the microscope slides that were placed into non-treated 4 well dishes with the dimensions 127.8 × 85.5 mm (Thermo Fisher).
For the NGF condition, the NGF stimulus (human NGF-β, Sigma-Aldrich) was added to the medium right after plating the cells making a final concentration of 50 ng/ml. For the ns-Zr15 hypo condition, the cells were re-suspended in RPMI low serum medium diluted 7.5/2.5 with deionised water (supplements were kept at the aforementioned concentrations) and pre-incubated in the hypoosmotic medium for 15 min before plating the cells eventually into the well, always in the hypoosmotic medium. For the PLL hyper condition, after the adhesion of the cells (1 h after plating) a hyperosmotic shock was applied to the cells (150 mM sucrose final concentration in the RPMI low serum medium) for 15 min, and washed once with RPMI low serum medium. The cells were left in RPMI low serum medium for the rest of the experiment.
The cells were left in the incubator for 24 h in all conditions. After washing twice with PBS, the cellular material was yielded for the proteomic analysis by scratching the cells from the microscope slides with cell scrapers (TPP, Trasadingen, Switzerland) in the presence of icecold PBS supplemented with protease (Roche, Basel, Switzerland) and phosphatase inhibitors (phosphatase inhibitor cocktail (Cell Signaling Technology), calyculine A (serine/threonine phosphatase inhibitor) 10 nM (Cell Signaling Technology), microcystin-LR 10 nM (Enzo Life Sciences).
For the inhibition experiments with SKF96365 (Sigma-Aldrich) and GsMTx4 (Alomone Labs, Israel), the resuspended cells were preincubated with the inhibitors (SKF96365 15 µM; GsMTx4 10 µM) in RPMI low serum medium (supplemented with 50 ng/ml NGF in the PLL +NGF condition) for 15 min in suspension before plating. The inhibitor treatment was maintained for 1 h, and then the medium was discarded and exchanged with new RPMI low serum medium (plus 50 ng/ml NGF in the PLL +NGF condition). For the rapamycin inhibition (Sigma-Aldrich), the cells were treated with the indicated rapamycin concentrations for the whole duration of the experiment. After 24 h, respectively 48 h for the rapamycin experiments, the morphology of the PC12 cells was recorded with an inverted Axiovert 40 CFL microscope Frontiers in Cellular Neuroscience | www.frontiersin.org (Zeiss, Oberkochen, Germany) equipped with LD A-Plan 20x/0.3 Ph1 or CP-ACHROMAT 10x/0.25 Ph1 objectives (both Zeiss) and the analysis was performed with ImageJ (NIH, New York, USA). Cells with neurites >10 µm were counted as differentiated and only neurites with a length >10 µm were considered for neurite length quantification. If cells have multiple neurites only the longest two were taken into the quantification, and in case of neurite branching the longest branch was measured. The neurite morphology was comparable between the canonical biochemically (NGF-)induced and the nanotopography-triggered neuritogenesis with 1.82 ± 0.42 neurites per cell for the first and 1.66 ± 0.21 for the latter (in total 160 differentiated cells for each condition were quantified from 8 independent experiments) ( Figure S2). In both cases the median was 2 neurites per cell and the vast majority of cells bore 1 or 2 neurites (together 82%, respectively 90%).
All the inhibition experiments were performed on coverslips with a diameter of 13 mm. The substrate preparation itself was the same as in the precedent section.

Shotgun Mass Spectrometry and Label Free Quantification
After reduction and derivatisation, the proteins were digested with trypsin sequence grade (Roche) for 16 h at 37 • C using a trypsin:protein ratio of 1:20. LC-ESI-MS/MS analysis was performed on a Dionex UltiMate 3000 HPLC System with a PicoFrit ProteoPrep C18 column (200 mm, internal diameter of 75 µm) (New Objective, USA). Gradient: 1% ACN in 0.1% formic acid for 10 min, 1-4% ACN in 0.1% formic acid for 6 min, 4-30% ACN in 0.1% formic acid for 147 min and 30-50% ACN in 0.1% formic for 3 min at a flow rate of 0.3 µl/min. The eluate was electrosprayed into an LTQ Orbitrap Velos (Thermo Fisher Scientific, Bremen, Germany) through a Proxeon nanoelectrospray ion source (Thermo Fisher Scientific). The LTQ-Orbitrap was operated in positive mode in data-dependent acquisition mode to automatically alternate between a full scan (350-2,000 m/z) in the Orbitrap (at resolution 60000, AGC target 1000000) and subsequent CID MS/MS in the linear ion trap of the 20 most intense peaks from full scan (normalized collision energy of 35%, 10 ms activation). Isolation window: 3 Da, unassigned charge states: rejected, charge state 1: rejected, charge states 2+, 3+, 4+: not rejected; dynamic exclusion enabled (60 s, exclusion list size: 200). Four technical replicate analyses of each sample were performed. Data acquisition was controlled by Xcalibur 2.0 and Tune 2.4 software (Thermo Fisher Scientific). Mass spectra were analyzed using MaxQuant software (version 1.3.0.5). The initial maximum allowed mass deviation was set to 6 ppm for monoisotopic precursor ions and 0.5 Da for MS/MS peaks. Enzyme specificity was set to trypsin, defined as C-terminal to arginine and lysine excluding proline, and a maximum of two missed cleavages were allowed. Carbamidomethylcysteine was set as a fixed modification, while N-terminal acetylation, methionine oxidation and Ser/Thr/Tyr phosphorylation were set as variable modifications. The spectra were searched by the Andromeda search engine against the rat Uniprot sequence database (release 29.05.2013). Protein identification required at least one unique or razor peptide per protein group. Quantification in MaxQuant was performed using the built in XIC-based label free quantification (LFQ) algorithm using fast LFQ. The required false positive rate (FDR) was set to 1% at the peptide, 1% at the protein and 1% at the site-modification level, and the minimum required peptide length was set to 6 amino acids.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Vizcaíno et al., 2016) partner repository with the dataset identifier PXD007644.

Statistical and Bioinformatics Analyses
Statistical analyses were performed using the Perseus software (version 1.4.0.6, www.biochem.mpg.de/mann/tools/). Only proteins/phosphopeptides present and quantified in at least 3 out of 4 technical repeats were considered as positively identified in a sample and used for statistical analyses. An Anova test (Permutation based FDR 0.05) was carried out to identify proteins/phosphopeptides differentially expressed among the different conditions. To tackle specific biological issues we then compared subsets of three proteomic data related to specific conditions, namely: [ns-Zr15, NGF, PLL], [ns-Zr15, NGF, flat-Zr], [ns-Zr15, NGF, ns-Zr25]. Therefore, proteins/phosphopeptides were considered differentially expressed if they were present only in one condition or showed a Post-hoc Bonferroni test p < 0.0167.
Regarding the proteomic data of ns-Zr15 hypo and PLL hyper which refer to peculiar cell conditions, the following comparisons were performed: [ns-Zr15, ns-Zr15 hypo], [PLL hyper and PLL], and [ns-Zr15 hypo, PLL hyper]. Proteins/phosphopeptides were considered differentially expressed if they were present only in one condition or showed a significant Welch t-test difference (cut-off at 5% permutation based FDR).
Bioinformatic analyses were carried out by DAVID software (release 6.7) (Huang et al., 2009), Panther software (Mi et al., 2017), ClueGO application of Cytoskape software (release 3.2.0) (http://www.cytoscape.org/), and Ingenuity Pathway Analysis (IPA R ) (QIAGEN Redwood City, www.qiagen.com/ingenuity) to cluster enriched annotation groups of Molecular Function, Biological Processes, Pathways, and Networks within the set of identified proteins/phosphopeptides. The compared data sets are indicated in the relative figures. Functional grouping was based on a Fisher Exact test p ≤ 0.05 and at least two counts.

Similarities and Differences between Biochemically and Mechanotransductively Promoted Neuronal Differentiation at the Protein Level
Focus on ns-Zr15, NGF, PLL, flat-Zr The versatile conditions included in the extended proteomic approach presented here are summarized in the introduction and in Figure 1. Altogether, they address different aspects of surface nanotopography of the substrate and/or biomechanics of the cell, integrating also information on the phosphorylation status of the proteins.
To dissect similarities and differences between the biochemically and mechanotransductively promoted neuronal differentiation at the proteome level we compared the data sets of ns-Zr15 (neuritogenesis-triggering cluster-assembled zirconia surface), PLL and NGF [canonical condition on PLL-coated glass, in the presence (NGF) or absence (PLL) of NGF]. The Venn diagram (Figure S3A), the work flow (Figure S3B), the Volcano plots ( Figure S3C) and the corresponding lists of differently expressed proteins (Tables S1-S6) are reported in the indicated Supplementary Information.
The comparison of ns-Zr15 vs PLL (Table S3) with ns-Zr15 vs flat-Zr (Schulte et al., 2016a) shows that the impact of ns-Zr15 on the protein expression profile is very similar (24 proteins altered in the same manner; marked X in Table S3) with respect to the two flat surfaces (flat-Zr and PLL).
Pointing instead more specifically toward the differences between ns-Zr15 and NGF (ns-Zr15 vs NGF, Table S2), the comparison reveals that 19 proteins (37%) are involved in cell proliferation and differentiation, 11 (22%) are receptors or players in signal transduction processes and 4 (8%) are related to Ca 2+ signaling. Moreover, the ClueGo analysis highlights that these proteins are mainly involved in neurofilament formation and assembly (e.g., vimentin, an intermediate filament protein

Similarities
Similarities and differences between biochemically and mechanotransductively promoted neuronal differentiation at the protein level Examples of proteins differentially expressed in the same manner in NGF vs PLL (Table S1) and ns-Zr15 vs PLL (

Zhou et al., 2010
Further information on a selection of proteins with versatile known roles in the regulation of neuronal functioning and neurogenic processes found to be differentially expressed in the same manner in NGF and ns-Zr15, compared to PLL.

Influence of the Surface Nanotopography Roughness on the Protein Expression
Focus on ns-Zr15, ns-Zr25, NGF, PLL As shown in Figure 1, the higher surface roughness R q 25 nm RMS (ns-Zr25) also induced neuritogenesis in PC12 but to a lower degree (Schulte et al., 2016a). The rationale of this phenomenon is not completely clear yet. Therefore, in addition to the mentioned ns-Zr15, we have included ns-Zr25 into this proteomic approach (Tables S7-S12). The Venn diagram and work flow for the comparison of NGF, ns-Zr15, ns-Zr25 are shown in Figures S4A,B, respectively.
Several proteins (7 out of 26; marked in light gray in Table S8) found to be upregulated in ns-Zr25 vs NGF (Table S8) are also upregulated in the comparison ns-Zr15 vs NGF (Table S2) or ns-Zr15 vs PLL ( Table S3), suggesting that the biological processes triggered by the cell/nanostructure interaction are partially similar, even if the roughness parameter is increasing.
The proteins upregulated only in ns-Zr25 vs NGF (Tables S8, S12), and not in the other conditions, are e.g., stress-induced proteins (such as CASC5, GPX3, A1M, and HSP90) and proteins involved in transport and trafficking. The data further demonstrates that the interaction of PC12 cells with higher roughness is accompanied by an increase of proteins related to formation/degradation of atherosclerosis plaques (APOB, SERPIND1), secretion, anti-inflammation activity and stress response (HMOX1, LOC681468, TXN, HMGN1, Cybasc3) while others are directly involved in gene expression control. Accordingly, the enrichment analysis of GO biological processes carried out by Panther on the proteins upregulated or only expressed in ns-Zr25 (Table 4) shows that the roughness increase triggers the expression of proteins involved in response to oxygen-containing compounds.
The proteome analysis of cells grown on ns-Zr25 also displays an increased expression of proteins involved in regulation of cell proliferation, differentiation and apoptosis, adhesion and trafficking, as well as intercellular signaling pathways. Some of these proteins indicate a strong neuronal differentiationpromotive effect also for this substrate [e.g., syntaxin 4 (Kennedy et al., 2010;Mohanasundaram and Shanmugam, 2010), clathrin (Heuser and Reese, 1973;Cosker and Segal, 2014), HMGN1 (Deng et al., 2013;Nagao et al., 2014), and SCN1B (Namadurai et al., 2015), details in Table 5], consistent with our results in primary hippocampal neurons where ns-Zr25 had the most significant effect on neuron differentiation and maturation (Schulte et al., 2016b). However, the induction of many stressrelated proteins suggests that the substrate situation is becoming suboptimal for PC12 cells leading to the altered expression of proteins that are essential for the regulation of neuronal survival [e.g., CREM (Mantamadiotis et al., 2002;Wu et al., 2012), NPM1 (Qing et al., 2008;Pfister and D'Mello, 2015); details in Table 5].
TABLE 2 | Differences between biochemically and mechanotransductively promoted neuronal differentiation at the protein level by comparing the conditions ns-Zr15, NGF, and PLL.   The best known function of Vbp1/prefoldin 3 is its involvement, as a chaperone, in the folding processes of actin and tubulin monomers. Furthermore, it has been reported to participate in the regulation of the RNA polymerase II, chromatin dynamics and transcription.

IAC/MECHANOTRANSDUCTION
Fat4 (protocadherin 4) up A hippo pathway-mediated role in neuronal migration and polarization for this protein has been reported recently. Interestingly and in line with our results, the Fat4 expression is controlled in a src-dependent manner by actin dynamics. The FAT4 protein is also thought to function upstream of YAP as a tumor suppressor, keeping cells from growing and dividing too rapidly or in an uncontrolled way. Further information on this protein regarding its role in calcium signaling-related processes can be found in the section down ADAM12 has been reported to be capable of regulating β1 integrin, the actin cytoskeleton and focal adhesion turnover. Kawaguchi et al., 2003;Eckert et al., 2017 Talin up Talin is a crucial protein in integrin activation and the initial phase of IAC formation, providing the first link between integrins and f-actin. It is therefore essential for the build-up of tension in lamellipodia. In a neuronal context, activated talin is necessary in the growth cones and enhances neuritogenesis.  NCoa2 (nuclear receptor coactivator 2)/GRIP1/TIF-2 up NCoA2 represents on the one hand a transcriptional coregulator for various nuclear receptors (e.g., androgen, estrogen, glucocorticoid, steroid receptor) and facilitates the access for DNA transcription by histone acetylation.
On the other hand it has been reported to be important for essential neuronal functions and dendrito-or synaptogenesis, in particular by coordinating neuronal and synaptic cargo and receptor (kainite, AMPA) trafficking. It can cooperate with other crucial proteins in these processes, such as SNAP25, 14-3-3 or liprin-α. The latter protein is furthermore known to be involved in the distribution and regulation of β1 integrin activation by stabilizing it in its inactive conformation at the cell membrane. Voegel et al., 1996;Wyszynski et al., 2002;Spangler and Hoogenraad, 2007;Essmann et al., 2008;Asperti et al., 2009Asperti et al., , 2010Selak et al., 2009;Dasgupta et al., 2014;Geiger et al., 2014;Heisler et al., 2014 Vesicle trafficking Ran(bp3) RanGAP1 up down The Ran GTPase (and its regulators) have versatile roles in neuronal cells, such as nucleocytoplasmic transport, trafficking, axon retrograde signaling and cytoskeletal dynamics. Yudin and Fainzilber, 2009 Trafficking Further information on a selection of proteins with prominent reported roles in the regulation of neuronal functioning and neurogenic processes and/or relevance for mechanotransduction differentially expressed comparing ns-Zr15 and NGF. Proteins differentially expressed in ns-Zr15vsNGF, either up-or downregulated or only expressed in the two conditions, were analyzed by IPA to identify possible networks.
In conclusion, it emerges that several critical proteins for membrane dynamics and configuration, integrin activation, IAC assembly and linkage to the f-actin are affected even by relatively subtle differences in the nanotopographical characteristics (Schulte et al., 2016a). This impact suggests a prominent role of the mentioned proteins in mechanosensing of topographical surface features.

Impact of the Cellular Interaction with the Neuritogenesis-Inducing Cluster-Assembled Zirconia Surface on Protein Phosphorylation
Phosphoproteomic Data of ns-Zr15, ns-Zr25, NGF, PLL, Flat-Zr The proteomic data presented here are complemented by an analysis of the phosphorylation state of proteins in the outlined experimental conditions, providing more profound information on the signaling pathways and potential specific key mediators (Tables S13-S22). A detailed analysis of individually identified phosphorylated proteins with interesting functions in the framework of this study is displayed in Table 7.
Frontiers in Cellular Neuroscience | www.frontiersin.org Proteins expressed in a differential manner in the comparion ns-Zr15vsns-Zr25 were analyzed by IPA to identify canonical pathways affected by the different nanotopography characteristics.
Flynn et al., 2012), septin-2 (Spiliotis and Nelson, 2006;Joo et al., 2007); details in Table 7]. Furthermore, several proteins essential in epigenetic and (post-)transcriptional regulation of gene expression are modulated at the phosphorylation level [e.g., KMT2D (Dhar et al., 2012), RtcB (Kosmaczewski et al., 2015), E2F4 (Persengiev et al., 1999); details in Table 7]. Regarding this latter aspect, it is noteworthy that lipin-1 phosphorylation is affected by the interaction with ns-Zr15. This phosphatidic acid phosphatase is important in lipid synthesis and SREBP-mediated transcriptional regulation (e.g., Fasn expression, see Table 1). Its phosphorylation is regulated by mTOR which thereby also controls its intracellular localisation and the lamin A-dependent nuclear organization (Peterson et al., 2011;Eaton et al., 2013). The alterations (regarding expression and phosphorylation levels) extend in a consistent manner our previous results (Schulte et al., 2016a), accentuating additionally the impact of the cell/nanotopography interaction on mechanotransductive processes and defining more precisely nanotopography-sensitive signaling hubs (Figure 2).
To find relevant patterns and specific differences in signaling processes related to the diverse conditions (PLL, NGF, ns-Zr15, ns-Zr25, and flat-Zr) a principal component analysis (PCA) was carried out on the corresponding phosphoproteomes. The analysis, applied to all the peptides found phosphorylated in these 5 conditions, reveals at a glance that the phosphoproteomes of NGF and ns-Zr15 cluster together (confirming again the common outcome of a differentiated cell). Flat-Zr and PLL instead are at the opposite ends of the plot (Figure 3A), suggesting that the cells on these two substrates behave very differently as far as protein phosphorylation concerns, in agreement with the other data reported so far.
If the same analysis is carried out focusing only on the sequence phospho-motifs present in the phosphoproteome data, a similar plot can be obtained (Figure 3B), but in this case a more evident separation can be observed between NGF and ns-Zr15, indicating that the kinases and phosphatases involved are, at least in part, different. Figure 3C reports in green all the substrate motifs that are more relevant in the PCA analysis, accounting for the 16% of all the phospho-motifs present in the phosphopeptides.
The enrichment analysis of these phospho-motifs, carried out by Panther and David, shows that there is a highly significant enrichment (p ≤ 0.05) of few signaling pathways in the cells on ns-Zr15 ( Table 8). The angiogenesis and VEGF signaling pathways are in line with the processes mentioned throughout this work as they comprise many players also involved in focal adhesion, MAPK and Ca 2+ signaling. In addition, the results suggest that the differences between PC12 cells grown on ns-Zr15 (compared to the NGF condition) could be partially ascribed to a modulation within the Wnt pathway (Table 8). An indicator is e.g., the downregulation of E-cadherin in the cells on ns-Zr15, considering the known crosstalk between (E-)cadherin cell adhesion and canonical Wnt signaling by release of β-catenin (Heuberger and Birchmeier, 2010). Interestingly, Wnt expression in PC12 cells leads to an upregulation of E-cadherin and a flat epithelial-like cell morphology associated with unresponsiveness to NGF-induced neuritogenesis (Bradley et al., 1993). In epithelial cells E-cadherin-mediated cell/cell adhesions are essential in mechanically connecting the intercellular actomyosin machineries to regulate tissue organization (Lecuit and Yap, 2015). Moreover, it has been demonstrated in human embryonic stem cells that the surface nanotopography has an impact on E-cadherin expression level . The potential impact of the substrate nanotopography on Wnt signaling in a neuronal setting is therefore an interesting issue for further investigations.
In the comparison ns-Zr15 vs ns-Zr25 (Tables S20-S22), apart from various already mentioned proteins, lamin A appeared as differentially phosphorylated. This protein is particularly interesting in the context of mechanotransduction representing one of the intermediate filaments that forms the interior of the nuclear envelope. It was found to be involved in the regulation of nuclear architecture/biophysics, chromatin organization, and transcription regulation at the end of mechanotransductive signaling cascades that influence differentiative processes (Swift et al., 2013). Its differential expression during adult neurogenesis TABLE 7 | Comparison of the conditions ns-Zr15, ns-Zr25, NGF, PLL and flat-Zr to analyse the impact of the cellular interaction with the neuritogenesis-inducing cluster-assembled zirconia surface on protein phosphorylation.

Associated to
Impact of the cellular interaction with the neuritogenesis-inducing cluster-assembled zirconia surface on protein phosphorylation Protein name Nidogen-1 The basement membrane protein nidogen-1 is known to be important in the regulation of hippocampal synaptic plasticity and network excitability.

Vasudevan et al., 2010
Brorin Brorin has been reported to favor neurogenesis and to inhibit astrogenesis contributing to axon guidance in the zebrafish forebrain.

Miyake et al., 2017
Galectin-8 Galectin-8 is a secreted ECM protein and matricellular modulator of cell adhesion that is bound by integrins which regulates cell adhesion and survival, promoting or inhibiting, dependent on whether it is present in a soluble or immobilized manner. Hadari et al., 2000;Zick et al., 2004 Ptprf (receptor-type tyrosine-protein phosphatase F)/LAR (leukocyte common antigen-related) The Ptprf/LAR receptor, a neuronal adhesion molecule essential in synapse maturation, is particularly interesting with respect to IAC, mechanotransduction and calcium signaling. The presence of this receptor in focal adhesions (FA) is controlled in a negative manner by myosin II-generated force and has been shown to have the capacity to regulate FAs (in mouse embryonic fibroblasts). It interacts in a CaMKII (Ca 2+ /calmodulin-dependent protein kinase II)-regulated way with liprin-α 1. The liprin-α 1/LAR interaction determines LAR distribution and therefore synapse morphogenesis. Moreover, Ptprf/LAR can be found, to a minor extent, tyrosine-phosphorylated in the adult brain. The function of this phosphorylation is yet unknown but could be important for the binding of Calcium homeostasis ArhGAP18 (Rho GTPase activating protein 18) ArhGAP18 has been shown to be involved, as negative regulator, in the control of RhoA activity and stress fiber formation by increasing the GTPase activity of Rho and stabilizing the RhoA-GDP inactive form. The protein, interacting with RhoA, has been described recently as YAP effector in the actomyosin-dependent regulation of tissue tension. A specific role of this protein in neurons has not been reported so far, but its expression level decreases in neurospheres during differentiation and it appeared as a gene associated with schizophrenia in a screening for single nucleotide polymorphisms. For another ArhGAP family member, ArhGAP15, a contribution in the neurogenesis of hippocampal neurons has been shown very recently. Gurok et al., 2004;Potkin et al., 2009;Maeda et al., 2011;Porazinski et al., 2015;Zamboni et al., 2016 Wnt signaling ASPM (Abnormal spindle-like microcephalyassociated protein) ASPM is known to contribute to the regulation of neuronal differentiation processes by actomyosin-dependent actions. As the protein name indicates, this protein is involved in the control of brain size and mutations of this protein can be responsible for developing the neural disorder microcephaly. ASPM is a positive regulator for Wnt signaling and its expression is essential for accurate neurogenesis. Furthermore, recently it has been found that the drosophila ortholog of this protein interacts with and regulates myosin II localization, thereby controlling neuroepithelium morphogenesis by mechanobiological events. Buchman et al., 2011;Rujano et al., 2013 Cofilin/destrin/ADF (actin depolymerising factor) Cofilin is essential for actin cytoskeletal organization by regulating the severing of f-actin and the turnover rate of actin and therewith, in the neuronal context, the actin retrograde flow in neurite growth cones of the developing brain. Phosphorylation negatively regulates its actin binding and thereby controls the f-actin homeostasis. Hawkins et al., 1993;Jovceva et al., 2007;Flynn et al., 2012 Septin-2 Septin-2 modulates actomyosin contractility by binding myosin II and recruiting regulatory proteins. Septin phosphorylation controls the assembly of septins into highly ordered polymers. Interestingly, septin-2 has been found to be phosphorylated in post-mitotic neurons. Spiliotis and Nelson, 2006;Joo et al., 2007   This transcription factor has been shown to play a role in neuronal differentiation and neuritogenesis. Persengiev et al., 1999 Rab23 (Ras-related protein 23) Rab23 participates to endocytic vesicle trafficking and is involved in the regulation of sonic hedgehog signaling in neural tube patterning. Eggenschwiler et al., 2001;Evans et al., 2003 Vesicle trafficking Selected proteins found to be differentially phosphorylated that have essential known functions with respect to processes relevant for mechanotransductive signaling and/or neuronal functioning and differentiation.
Altogether, many proteins that are important in neurogenic and/or mechanotransductive processes are differentially expressed and/or phosphorylated upon cellular interaction with the cluster-assembled zirconia surface that promotes neuritogenesis. Combining the analysis of our proteomic data with information available on these proteins and their functions, suggests a dynamic and complex modulation of an entire signaling network by the cell/nanotopography interaction that is in control of cellular behavior and fate, i.e., in this case neuronal differentiation. We were able to dissect potential nanotopography-sensitive key elements regulated within a mechanotransductive sequence, identifying many proteins that can be assigned to these principal categories: cell/cell adhesion, ECM and glycocalyx, cell/substrate interaction and IAC, integrin activation and membrane/f-actin linkage, integrin adhesion complexes, actomyosin organization/cellular mechanics and nuclear organization and transcriptional regulation (Figure 2).

Alterations in Cellular Processes and Signaling by the Modulation of Cellular Tension
Focus on Phosphoproteomic Data of ns-Zr15, ns-Zr15 hypo, PLL hyper In our previous paper we identified the alteration of the cellular nanomechanical properties as critical for the signal integration within the nanotopography-dependent mechanotransductive sequence that fostered neuronal differentiation. The interaction with the nanostructured surface dictated the IAC nanoarchitecture/dynamics and cytoskeletal organization in a manner that consequentially resulted in a softer membrane/cytoskeletal layer of the neuronal cell. Compensating this effect by a hypoosmotic gradient (causing cell swelling and an increase of cell tension) counteracted gradually the nanostructure-induced neuritogenesis on the morphological level (Schulte et al., 2016a).
The mechanotransduction dependency of the nanotopography-promoted differentiation was broadly validated by a proteomic comparison (Tables S23-S24) of PC12 cells interacting with the ns-Zr15 in the isoosmotic standard medium (ns-Zr15) or instead in the presence of hypoosmotic medium (ns-Zr15 hypo). Many proteins found to be expressed only in ns-Zr15 (Table S24) or to be downregulated in ns-Zr15 hypo (Table S23) can be classified as proteins involved in RhoGTPase-controlled cytoskeletal organization according to the IPA canonical pathways enrichment analysis (Figure 4). Consistent with the hypoosmotic manipulation of the membrane tension, clathrin-and caveolar-mediated endocytosis appeared among the five most affected pathways in this evaluation (Figure 4). These pathways are regulated by and respond to justifying their categorisation, can be found in the corresponding tables throughout the manuscript. Moreover, on the right a selection of proteins is listed with association to their functions in neuronal differentiation processes, particularly in neurite growth cones. modulations of the membrane tension and are crucially involved in cell volume and shape control (Raucher and Sheetz, 1999;Sinha et al., 2011;Gauthier et al., 2012). Intriguingly, mTOR signaling emerged as the strongest modulated pathway in this analysis (Figure 4). mTOR signaling represents a highly conserved pathway known to be an integrative master regulator of many cellular processes/pathways at the interface of intracellular and extracellular signals (Laplante and Sabatini, 2009), also in regard to neurogenic events (Garza-Lombó and Gonsebatt, 2016). Inhibition of mTOR(C1) with rapamycin had different effects on flat and nanostructured substrates. On PLL there was an increase of neurite outgrowth upon rapamycin inhibition both in the absence, or presence, of NGF. In the latter, the effect was additive to the NGFinduced increase, reproducing data reported by others (Parker et al., 2000). On ns-Zr15 instead no significant impact, neither promotive nor inhibitory, with respect to neurite outgrowth was observable (Figure 5). It can be speculated that the boosted The ns-Zr15vsNGF analysis of the phospho-motifs was carried out by Panther and David (p ≤ 0.05).
neuritogenesis on PLL is due to an induction of mTORC2 activation triggered by the rapamycin-mediated inhibition of mTORC1 as a negative feedback between the two mTORCs is known (Xie and Proud, 2013). Moreover, mTORC2 has been shown to be involved in the regulation of actin dynamics and morphology of neurons in a Rac/PAK-dependent signaling pathway that controls cofilin phosphorylation (Huang et al., 2013;Thomanetz et al., 2013). Very recently, it has been demonstrated in DRG neurons that topographical features can potentiate mTORC2 guiding neurite outgrowth (Thomson et al., 2017). On ns-Zr15 the mTORC2 might already be induced by the cell/nanotopography interaction and thus rapamycin treatment does not further affect neurite outgrowth. The altered cofilin phosphorylation (Table 7) is in line with this (Huang et al., 2013). The varying impact of mTOR(C1) inhibition by rapamycin depending on whether the cells interact with a flat or a nanotopographical surface makes mTOR signaling an interesting and promising target for further studies in this context but goes beyond the scope of this work. Moreover, prominent markers for (developing) neurons and neurite outgrowth [such as BASP1, MAP1B, or β-tubulin (TUBB5)] are strongly downregulated in the hypoosmotic condition. The same is true for many proteins involved in the actin polymerisation machinery and the cytoskeletal organization [(e.g., Capg, Arpc1b, 3 and 5, Capzb, fascin) which are crucial for the realization of neuritogenesis (da Silva and Dotti, 2002;Fletcher and Mullins, 2010); (Tables S23, S24)]. The alterations in the protein expression profile are largely mirror-inverted to those seen in the comparison ns-Zr15 vs flat-Zr (Schulte et al., 2016a). 37 proteins have an opposite expression level in these two comparisons (marked X in Table S23), whereas only 5 proteins are altered in the same way (marked x in Table S23).
On the other hand, a hyperosmotic shock applied to cells on PLL-coated glass (resulting in a decrease of membrane tension) led morphologically to the outgrowth of neurites (Figure 1,  Figure S1). The proteomic data (Table S25) disclosed that the neuritogenesis was accompanied by a modification of the protein profile similar to those found in ns-Zr15 vs flat-Zr (Schulte et al., 2016a). 39 proteins had the same alteration of the expression level (marked in x in Table S25). However, proteins known to be involved in IAC (abundant e.g., in the ns-Zr15 vs flat-Zr comparison; Schulte et al., 2016a) are basically missing here. In addition, 16 proteins also showed an opposite expression level modification (marked X in Table S25).
The PCA analysis carried out on the phosphopeptides differentially expressed in these conditions (ns-Zr15, ns-Zr15 hypo, NGF, PLL hyper, PLL) indicates that either hypoosmotic swelling on ns-Zr15, as well as the hyperosmotic shock on PLL, moves the profiles partially closer to the NGF condition. This emphasizes again that ns-Zr15 hypo basically lost its nanotopography-specific features, whereas PLL hyper has gained at least some characteristics of the NGF condition ( Figure 3D).
Overall, these proteomic data further reinforce that the modulation of the cellular nanomechanical properties is a key integrating signal causally linked to the change of the cellular program and the differentiation processes that we discussed in our previous work (Schulte et al., 2016a), with a potential involvement of mTOR signaling constituents.

Focus on Proteins Involved in Calcium Signaling and Homeostasis
Alterations of the integrin/ECM interaction (e.g., in growth cone filopodia; Kuhn et al., 1998;Gomez et al., 2001;Gui et al., 2006) and cellular biomechanics can modulate another important mechanotransduction-susceptible pathway; that is calcium signaling regulated by Ca 2+ influx passing mechanosensitive membrane channels (Vogel and Sheetz, 2006;Sukharev and Sachs, 2012). In turn, it has long been known that integrin/ligand binding is affected by divalent cations (also Ca 2+ ) (Zhang and Chen, 2012). Local changes in calcium concentration influence integrin adhesion dynamics in growth cones and axon guidance in a calpain/talin-dependent manner (Kerstein et al., 2017), mediated e.g., by Piezo1/Fam38A (McHugh et al., 2010). However, despite this acknowledged role of calcium signals in neuronal differentiative processes, the exact spatiotemporal regulation and impact of calcium signaling is rather complex and intricate with many details still elusive (Gomez and Zheng, 2006;Leclerc et al., 2012;Kerstein et al., 2013;Toth et al., 2016).
To study the involvement of mechanosensitive channel types in our experimental context, we used the inhibitors SKF-96365 for transient receptor potential cation channels (TRPC) and GsMTx4 for stretch-activated channels (SAC, such as e.g., Piezo). With respect to the canonical NGF-stimulated outgrowth our results were in line with findings published by others, i.e., the two inhibitors showed opposing effects on neurite outgrowth (Jacques-Fricke et al., 2006;Gottlieb et al., 2010;Kerstein et al., 2013). SKF-96365 impeded differentiation and neurite outgrowth, whereas GsMTx4 had a minor differentiationenhancing effect (Figure 6). This is consistent with the reported crosstalk between NGF/TrkA and TRPC-mediated calcium signaling (De Bernardi et al., 1996;Cohen et al., 2015). More interestingly and despite its independence of NGF/TrkA activation (Schulte et al., 2016a), the outcome was practically in the same range for the nanotopography-promoted neuritogenesis (Figure 6) suggesting a contribution of calcium signaling also in this mechanotransductively fostered differentiation.
Other proteins related to calcium signaling can be found upregulated comparing directly ns-Zr15 vs NGF (Table S2), such as Ahnak, Fat4 (protocadherin 4) and Prkcsh/PKC substrate 80K-H. Ahnak is a scaffolding protein that partakes in versatile cellular processes, many in fact related to calcium signaling  and/or membrane morphogenesis (together with S100A10) (Benaud et al., 2004;Lorusso et al., 2006). In the neuronal context, it is also a marker of enlargeosomes. These exocytic vesicles can contribute to the membrane supply for neurite outgrowth in a REST-regulated manner (Borgonovo et al., 2002;Racchetti et al., 2010;Schulte et al., 2010). Fat/Protocadherin 4 belongs to the calcium-dependent cadherin cell adhesion protein family and has been reported to be involved in the regulation of neuroprogenitor proliferation and differentiation upstream of YAP. Fat4 downregulation lowers differentiation of neuroprogenitors into neurons in the cerebellum (Cappello et al., 2013). Prkcsh/PKC substrate 80K-H, which colocalises with IP 3 R1, modulates IP 3 -induced calcium release and might therefore have a role in synaptic plasticity (Kawaai et al., 2009).
In summary, Ca 2+ signaling is important for both, NGFand nanotopography-triggered neuritogenesis, but the proteomic data suggest that the cell/nanotopography interaction might influence some specific proteins prominently involved in calcium homeostasis and/or signaling.

CONCLUSIONS
In recent years the relevance of microenvironmental and cellular mechanobiological aspects with respect to differentiation processes have become evident (Geiger et al., 2009;Wang et al., 2009;Dalby et al., 2014;Jansen et al., 2015), particularly in the neuronal context (Franze et al., 2013). However, many details regarding the mechanosensing of the microenvironment, the signal transmission, and integration within the mechanotransductive sequence remain elusive (Dalby et al., 2014;Humphries et al., 2015;Jansen et al., 2015). In previous publications we have introduced how ECM-like nanotopographical features of cluster-assembled titania and zirconia sufaces, produced by SCBD, can foster important events in neuronal differentiation (such as e.g., neuritogenesis; Schulte et al., 2016a,b, synaptogenesis and network maturation;Schulte et al., 2016b) through mechanotransductive pathways.
The presently adopted quantitative proteomic approach (associated to a systematic characterization) challenges PC12 cells with diverse experimental situations that address the impact of substrate nanotopography and/or cellular biomechanics on neuronal cell fate. The analyzed conditions comprised the canonical PC12 cell differentiation setting with a biochemical stimulus (PLL ±NGF; i.e., PLL and NGF), three different zirconia surface topographies with distinct nanoscale roughness parameters (flat-Zr, ns-Zr15, ns-Zr25) and treatments that affect Ubr4 (E3 ubiquitin-protein ligase)/p600 Ubr4 is involved in membrane morphogenesis and integrin signaling. It is required for neuronal survival in a calcium/calmodulin-dependent mechanism. The protein has versatile functions in the CNS, calcium signaling and cytoskeletal organization. The phosphorylation level of p600 at cyclin-dependent consensus site varies during cell cycle. The function of this phosphorylation, in particular in a neuronal context, is unknown, though. Nakatani et al., 2005;Belzil et al., 2013;Parsons et al., 2015 Neuronal survival, membrane remodeling, integrin signaling, cytoskeletal organization eIF3e (eukaryotic initiation factor 3 subunit) This tumor suppressor has been shown to contribute to the trafficking of CaV calcium channels and therefore the regulation regulation of calcium influx and intracellular calcium levels by these type of channels. Green et al., 2007 Transcription, trafficking S100A10 (S100 calcium-binding protein A10) This protein is, as the only member of the S100 family, not capable of Ca 2+ binding. It is nevertheless known to be implicated in calcium signaling, e.g., by regulating the trafficking of ion channels (also TRP). It participates also to the transport of neurotransmitter (receptors). Furthermore, together with AHNAK, it is involved in governing f-actin and cell membrane organization. Benaud et al., 2004;Rescher and Gerke, 2008;Jung et al., 2010 Trafficking, membrane remodeling, cytoskeletal organization Selected proteins differentially expressed after cell/nanotopography interaction that have reported functions in calcium signaling/homeostasis. the tensional state of the cell (ns-Zr15 hypo, PLL hyper). This approach enabled us to acquire an extensive molecular image of the processes and pathways that are sensitive to changes in the microenvironmental nanotopography and/or cellular nanomechanical properties, and to identify potential key elements therein. The data thus provide various starting points and indications on how the nanotopographical sensitivity is achieved and integrated into signaling pathways. We found various common, but also distinctive features in the protein expression and phosphorylation profiles comparing the canonical biochemically (NGF-)induced neuronal differentiation and the one triggered by the cell/nanotopography interaction. In this neuron-like PC12 cell model the mechanotransductive stimulus provided by an appropriate nanotopography is alone sufficient to achieve the necessary change in the cellular program that implements the neuronal differentiation. There are indications in the phosphoproteomic data that the nanotopographical stimulus is even more effective. A robust engagement of proteins involved in cell morphology, cellular assembly and organization, and cellular movement (see IPA, Table 3) was observed and many proteins with acknowledged roles related to IAC/mechanotransduction were altered. In addition, versatile proteins with tasks in neuronal functioning and differentiation have been identified to be modulated in the nanotopography setting; as well as several proteins essentially involved in epigenetic and (post-)transcriptional regulation during neuronal differentiation (Tables 1, 2, 5, 7). This is in line with the hypothesis of a potential epigenetic regulation of cell reprogramming and differentiation dependent on cellular mechanics and microenvironmental cues (such as in this case surface nanotopography) (Downing et al., 2013;Crowder et al., 2016).
We have seen on the morphological level that only specific roughness parameters of the cluster-assembled zirconia surfaces (for PC12 cells ns-Zr15) provide appropriate biophysical cues to gain full neuritogenesis. An increased nanotopography roughness (ns-Zr25) instead leads to an only partial effect on neuritogenesis (Schulte et al., 2016a). In line with this, the proteomic analysis demonstrated that there is a partial overlap in the molecular alterations between ns-Zr15 and ns-Zr25, compared to NGF. Yet, the comparison revealed and specified also some decisive differences regarding proteins important for integrin activation and cytoskeletal organization. Notably, the IPA emphasized the importance of IAC in the mechanosensing to interpret the distinct natures of the topographies as ILK (integrin-linked kinase) signaling is the only decreased pathway ( Table 6). This is striking because ILK signaling is essentially involved in determining the molecular architecture of IAC and its signaling is sensitive to variations in ligand spacing/density. The proteomic data suggests furthermore that the increase in roughness starts to cause cellular stress in this PC12 cell model.
In summary, many of the proteins found to be altered at the expression and/or phosphorylation level (Tables 2, 5, 7) can be associated with the following categories which all have high relevance with respect to mechanotransductive signaling: cell/cell adhesion, glycocalyx and ECM, integrin activation and membrane/f-actin linkage, cell/substrate interaction and IAC, actomyosin organization/cellular mechanics, and nuclear organization and transcriptional regulation. By integrating the dissected alterations into a potential context, a complex nanotopography-sensitive network with broad crosstalk opportunities crystallizes, capable of regulating the cell/microenvironment interface and consequentially cellular cytoskeletal mechanics and signaling, here in control of neuronal differentiation processes (Figure 2).
This comprehensive proteomic analysis insinuates that other pathways with strong correlation to mechanotransduction, such as Wnt (Table 8), mTOR (Figures 4, 5) and Ca 2+ signaling (Table 9, Figure 6), might also be affected by, and involved in, the nanotopography-triggered cellular processes. However, more profound future studies are required regarding these pathways.
Altogether, this proteomic-based analysis defined nanotopography-sensitive signaling hubs and key elements potentially important in the promotion of neuronal differentiation by nanotopographical cues. It delivered several interesting starting points to evaluate in more specific studies, and in a wider context with respect to the role of certain proteins in mechanotransductive signaling that regulates neuron development and maturation. In the framework of biomaterials that are based on nanoscale surface features, an in-depth understanding of the impact of cell/nanotopography interaction on cellular processes and fate is the indispensable prerequisite. An improved insight might help to harness and effectively control the potential of these biomaterials in biomedical applications. Vice versa, information obtained by advanced biomaterial approaches could provide conclusions for a better comprehension of the difficult to access in vivo mode of operation of microenvironmental and cellular mechanobiological processes, e.g., regarding epigenetic regulation.

AUTHOR CONTRIBUTIONS
The project was mostly conceived by CS and GT. EM, CS, and GT wrote the principal part of the manuscript and realized the figures. The proteomic approach and related data analyses were done by EM, SN, FG, AN, CS, and GT. CS performed the cell biological inhibition experiments and corresponding quantifications. CP executed the fabrication of the nanostructured surfaces by SCBD and flat-Zr by electron beam evaporation. CL and PM participated in the project conception/creation and the realization of the manuscript, contributing moreover reagents, materials, analysis tools and considerations regarding the nanotechnological aspects of the work.

ACKNOWLEDGMENTS
PM, GT, and CS acknowledge support from the European Union project "FutureNanoNeeds" grant "Framework to respond to regulatory needs of future nanomaterials and markets" (FP7-NMP-2013-LARGE-7). We are grateful for the contribution of Andrea Notarnicola to the experiments leading to Figure S1, Figure 6. We thank Justin Mayer for the proofreading of the manuscript.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncel. 2017.00417/full#supplementary-material     Table S1 | Proteins differently expressed comparing NGF vs PLL. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway, Reactome and molecular complexes (CORUM). Table S2 | Proteins differently expressed comparing NGF vs ns-Zr15. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway, Reactome and molecular complexes (CORUM). Proteins that are of particular interest regarding a potential connection of IAC/mechanotransductive signaling and neuronal differentiation processes in the nanotopography-induced setting are marked in gray. Table S3 | Proteins differently expressed comparing ns-Zr15 vs PLL. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway, Reactome and molecular complexes (CORUM). X = proteins differentially expressed in the same manner in ns-Zr15 vs flat-Zr and ns-Zr15 vs PLL. In gray, proteins differentially expressed in the same manner in NGF vs PLL and ns-Zr15 vs PLL. Table S4 | Proteins expressed only in NGF (with respect to the comparison PLL, NGF, ns-Zr15). List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). Table S5 | Proteins expressed only in ns-Zr15. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). Proteins that are of particular interest regarding a potential connection of IAC/mechanotransductive signaling and neuronal differentiation processes in the nanotopography-induced setting are marked in gray. Table S6 | Proteins expressed only in PLL. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). Table S7 | Proteins differently expressed comparing ns-Zr25 vs ns-Zr15. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). Table S8 | Proteins differently expressed comparing ns-Zr25 vs NGF. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway, Reactome and molecular complexes (CORUM). Proteins found to be upregulated in ns-Zr25 vs NGF and also up-regulated in the comparison ns-Zr15 vs NGF or ns-Zr15 vs PLL are marked in gray. Table S9 | Proteins differently expressed comparing ns-Zr15 vs NGF. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM).
Table S10 | Proteins expressed only in NGF (with respect to the comparison NGF, ns-Zr15, ns-Zr25). List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM).
Table S11 | Proteins expressed only in ns-Zr15 (with respect to the comparison NGF, ns-Zr15, ns-Zr25). List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM).
Table S12 | Proteins expressed only in ns-Zr25. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM).    Table S17 | Proteins differently phosphorylated in ns-Zr15 vs flat-Zr. The table lists all phosphorylated proteins annotated with Perseus software. Annotated terms include GOBP, GOMF, GOCC, Keywords, KEGG, CORUM, together with the enriched sequence motifs of phosphopeptides data set.
Table S18 | Proteins exclusively phosphorylated in ns-Zr15 (with respect to the comparison PLL, NGF, ns-Zr15, flat-Zr). The table lists all phosphorylated proteins annotated with Perseus software. Annotated terms include GOBP, GOMF, GOCC, Keywords, KEGG, CORUM, together with the enriched sequence motifs of phosphopeptides data set.    Table S22 | Proteins exclusively phosphorylated in ns-Zr15 (with respect to the comparison ns-Zr15, ns-Zr25). The table lists all phosphorylated proteins annotated with Perseus software. Annotated terms include GOBP, GOMF, GOCC, Keywords, KEGG, CORUM, together with the enriched sequence motifs of phosphopeptides data set. Table S23 | Proteins differently expressed comparing ns-Zr15 hypo vs ns-Zr15. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). X = proteins with an opposed alteration of the expression level comparing ns-Zr15 vs flat-Zr with ns-Zr15 hypo vs ns-Zr15; Y = proteins altered in the same way in ns-Zr15 vs flat-Zr and ns-Zr15 hypo vs ns-Zr15. Table S24 | Proteins expressed only in ns-Zr15 (with respect to the comparison ns-Zr15, ns-Zr15 hypo). List of the proteins annotated with Perseus software.
Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). Table S25 | Proteins differently expressed comparing PLL hyper vs PLL. List of the proteins annotated with Perseus software. Categorical annotation was supplied in form of Gene Ontology (GO) biological process (BP), molecular function (MF), cellular component (CC), keywords as well as in participation in a KEGG pathway and molecular complexes (CORUM). x = proteins with a similar alteration of the expression level comparing ns-Zr15 vs flat-Zr and PLL hyper vs PLL; X = proteins altered in an opposite manner in ns-Zr15 vs flat-Zr and PLL hyper vs PLL.