Tuning neuron adhesion and neurite guiding using functionalized AuNPs and backfill chemistry

The adhesion of neurons depends on the interplay between attractive as well as repellant cues in the cell membrane and adhesion ligands in their cellular environment. In this study, an easy and versatile strategy is presented to control the density of cell binding sites embedded in a cell repulsive environment attached to a solid surface. Gold nanoparticles modified by positively charged aminoalkyl thiols are used as artificial neuron adhesion ligands. The density of the nanoparticles and their environment is varied by applying either no backfill, poly(ethylene glycol)-silane, or octyltrichlorosilane backfill. By this means the chemical composition of both cell attractive adhesion ligands and surrounding repellant cues is tuned on the nanometer scale. Primary rat cortical neurons are cultured on these particle modified surfaces. The viability and neuritogenesis of neurons is investigated as a function of particle density and background composition. A strong dependence of neuron viability on both averaged particle density and backfill composition is found in particular for intermediate particle packing. At high particle densities, the kind of backfill does not affect the cell viability but influences the development of neurites. This knowledge is used to enhance the guiding efficiency of neuron adhesion to more than 90% on nanopatterned surfaces.


Introduction
The control of cell adhesion is of increasing interest for the investigation of fundamental aspects of cell development such as cell differentiation, cell growth, tissue development and regeneration.5][6] A close and stable contact between neuron and device surface is important to ensure a small seal resistance and an optimal communication between them. 719][20] As chemical adhesion cues mainly cell adhesive molecules (CAMs), 21,22 neurotrophic factors, 23 ligands of specic guidance receptors, 24,25 extra cellular matrix proteins such as laminin and bronectin or selected oligopeptides of those like RGD and IKVAV are used.[26][27][28][29][30] Also corollaries of the specic surface composition such as the wettability 31 and surface charges 32 are inuencing cellular behavior on biomaterial surfaces. It has een recognized that in particular positively charged domains of peptides assist the adhesion of neurons.33 This nding was supported by the observation that also positive charged amino groups associated with surface bound synthetic molecules can promote the adhesion and growth of neurons.[34][35][36][37] In vivo studies showed that positive charges aid the regeneration of nerves. 38 Many ell membranes possess a high number of glycosylated proteins which are contributing to the formation of a negative net charge at the extracellular part of the membrane. The itercellular recognition and adhesion is regulated by these negatively charged carbohydrates.For neurons, the polysialylation of neural cell adhesion molecule (NCAM) facilitates cell migration and plasticity by regulating the repulsion between adjacent cells.39 The electrostatic interactions between the negatively charged glycocalyx and cationic adhesion cues has been used to immobilize and pattern cells on solid surfaces.2][43] It has been demonstrated that nanoscale objects are able to inuence cellular properties and processes such as cell shape, cell adhesion, alignment, differentiation and others.[44][45][46] The reason for the strong impact of nanoobjects on cellular features is that many cell functions are controlled by proteins which possess nanoscale dimensions.Therefore, it has been proposed that nanoscience may help achieve a greater precision and parallelism in electrical and chemical sensing of cell/neuron activity. 47Protein patterns as small as 75 nm were generated by microcontact printing and used to control the adhesion of primary neurons and directed outgrowth of neurites. 48J. H. Huang et al. arranged gold nanoparticles with varied interparticle spacing as carriers of single peptide guidance factors in patterned arrays on substrates by diblock copolymer micelle nanolithography.49 Their studies indicate that the spatial organization of the extracellular environment is important in regulating cell-extra cellular matrix (ECM) interactions and thus cell adhesion.21,26,41,[50][51][52][53] In addition to the chemical inuence of nanoparticles on cells, they allow to obtain additional analytical information about the investigated system due to their optical properties.54,55 We recently developed a simple and efficient strategy to deposit randomly close-packed AuNPs on a solid surface by electrostatic interactions. 56n this article we investigate the variation of the surfaces composition at nanometer scale by changing both the density of the particles and the chemical composition of the particle surrounding.The AuNPs were functionalized by 11-amino-1-undecanethiol (AUT), a molecule that possesses a positive charge under neuron culture conditions.The amino-functionalized gold nanoparticle (AF-AuNP) can be considered as cationic anchor spots for glycosylated plasma membrane proteins similar to widely used cationic polyamino acids like polylysines, however, with strongly conned size and control over density and local distribution.Samples homogeneously modied with poly-Dlysine (PDL) are used as control to evaluate the maturation of the neurons.We investigate the inuence of the densities of surface bound particle on the viability of primary rat cortical neurons.In addition, we modied the surrounding background of the AF-AuNP to vary the chemical contrast between the particle and their environment.This implicates that we control the lateral distribution of cell attractive and cell aversive materials at the nanometer scale and thereby tune the cell adhesive properties of the solid surface on micron and millimeter scale.We investigate the inuence of three common background materials: negatively charged SiO 2 , 2-[methoxy(polyethyleneoxy) 6-9-propyl] trichlorosilane (PEG silane), and octyltrichlorosilane (OTS).The bare, hydrophilic SiO 2 was supposed to have cell repellent properties since it possesses an excess of negative charges at culture conditions and should therefore repulse the glycocalyx of the neurons. PEG sane was chosen as backll molecule because of its excellent protein repulsive property to minimize unspecic interaction between proteins and inorganic surfaces.49,57,58 In addition, we used OTS as backll material since it was demonstrated that hydrophobic surfaces inhibit cell adhesion. 59We found that rat cortical neurons can adhere to AF-AuNPs functionalized surfaces and the cell viability at high particle densities was even better than for PDL control samples.Furthermore, we observed that not only ligand density is critical for neuronal adhesion 49 but also the chemical contrast between the AF-AuNPs and environment play a vital role in attachment of neurons on solid surfaces and the outgrowth of neurites.We used the gained knowledge to optimize the guiding efficiency of micro-and nano-patterned particle modied surfaces.The guiding of neuron adhesion and neurite outgrowth is not only required for the construction of dened in vitro neural networks but also for the alignment of neurons with solid-state devices to realize bioelectronic hybrid systems.[60][61][62][63] For this purpose various patterning techniques been used such as photolithography and microcontact printing.64,65 These patterning techniques have in common that they generate a local pattern with homogenous chemical composition on micrometer scale.This corresponds to a kind of "on" and "off" situation for the adhesion of cell and their processes from the nanometer scale perspective.Based on a large area top down nanoimprint lithography we generated locally dened chemical templates for the bottom up selfassembly of nanoparticle on these template patterns.Thereby, we control the lateral distribution of cell attractive and cell aversive materials at the nanometer scale.By this means we can enhance the chemical contrast between particle free and particle containing areas which results in a guiding efficiency for rat cortical neurons of more than 90%.

Characterization of particle modied surfaces
The preparation and characterization of the AuNP modied surfaces has been described in detail in previous publications. 56,66Just briey, a plasma activated SiO 2 surface is modi-ed by aminopropylsilane molecules and citrate stabilized AuNP are immobilized on this surface via electrostatic interactions.Subsequently the particle decorated surface is activated in an oxygen plasma and the particle surrounding SiO 2 surface is modied by octyltrichlorosilane (OTS) or 2-[methoxy-(polyethyleneoxy)6-9-propyl] trichlorosilane (PEG silane) from vapor phase.Finally, the sample is immersed into an 11-amino-1-undecanethiol solution to modify the surface of gold particles (Scheme 1).
In the following we report on the characterization of sample preparation steps that have not been part of previous reports as for instance the backlling.Therefore, we rst of all characterized the wettability of silicon surfaces before and aer modifying the surface by a backll layer.We used sessile drop contact angle measurements to obtain static contact angles.
Aer O 2 plasma activation of the bare silicon surface, we obtained a perfectly hydrophilic surface with a contact angle of q < 10 .Aer the PEG silane gra, the static contact angle increases however the surface remained hydrophilic with a contact angle of 44 AE 1 similar to previous reports. 67,68At contact angles smaller than 50 ethylene glycol surfaces become adsorption resistant against hydrophilic and hydrophobic peptides. 57For the OTS silanization we observed a change of the surface wettability from hydrophilic to hydrophobic and an increase of the static contact angle to (82 AE 2 ). 69urthermore, we evaluated the successful attachment of the 11-amino-1-undecanethiol to AuNP decorated Si/SiO 2 surfaces with and without PEG silane backll by means of X-ray photoelectron spectroscopy (XPS).In both cases, we prepared samples with a high density of particles of about 500 particles per mm 2 (see below).For both surfaces we observed a high content of Si, oxygen, and Au which are originating from the sample substrate and the AuNP core, respectively, Table 1.The signals of these elements decreased aer AUT binding due to the formation of a molecular layer on the gold particles attenuating substrate signals.In addition we found nitrogen as well as sulfur which can be assigned to the formed AUT shell on the AuNP.The sulfur and nitrogen signals are relatively weak since the particles decorate just parts of the sample (less than 20%) and the information depth of this technique is much larger than the length of AUT.Furthermore, the S2s signal is attenuated by the alkyl chain capping the thiol group.The fact that the sulfur and nitrogen signal can be observed also aer application of a PEG backll layer to the AF-AuNP free areas of SiO 2 shows that the backll does not impair the formation of particle associated gold thiolate bonds.

Particle density assays
Citrate stabilized AuNP were immobilized from colloidal solutions of different concentration on a self-assembled monolayer (SAM) of aminopropylsilane.With increasing particle concentration, the density of particles on the surface gradually increases as described previously. 66By this simple method we were able to tune the density of particles from 4 AE 2 to 541 AE 20 particles per mm 2 with a coverage changing from 0.1 to around 20% by simply diluting the particle solution from 0.0001 to 0.01 wt%, correspondingly.The SEM images reveal a random distribution of the AuNPs on silica surface (Fig. 1) at different concentrations.The particles are well dispersed.Aggregation of two or more particles occur rarely.By comparing the particle density before and aer backll, we found that about 5% of the particles were released from the surface due to the backll treatment.

Neuronal culture assay
Rat embryonic cortical neurons were cultured on AUT modied AuNPs with different particle densities and different backgrounds to investigate the inuence of the particle distance and  chemical background on the adhesion and viability of neurons.
We used a serum free neuron basal medium for cell culture in the scope of this work.Since we intend to control the adhesion of electrogenic neurons and the outgrowth of their neurites we aim to avoid a domination of the culture by glia cell aer long culture times.To identify the cell type, we performed an immunostaining to distinguish neuron from glial (results not shown).We found only very few glia cells on the sample aer 9 DIV of all surface treatments tested.The ratio of glia vs. neuron was typically smaller than 0.07 indicating a successful suppression of glia proliferation.The adhered neurons were co-stained with calcein AM (green) and ethidium homodimer (orange) to indicate healthy and dead cells, respectively, aer 5 days in vitro (DIV).We observed an enhanced viability of the neurons with increasing particle density for all samples tested for PEG silane backll, Fig. 1.For low particle density, most cells appear as small red dot which indicates cell death and a detachment of the cell from the sample surface.With higher particle density an increasing number of vital cells stained in green are visible.The samples with the highest particle density provided accordingly the best neuron culture conditions among the tested samples.
For a qualitative assessment of the neuron viability on the different particle modied surfaces we determined the relative live cell density for all samples and plotted this quantity vs. the particle density.The relative live cell density is calculated by dividing the live cell density of a particle containing sample by the live cell density of the corresponding PDL control sample.From the resulting plot remarkable differences become apparent for samples with different particle environments, Fig. 2. Samples that have been modied with particles but without backll facilitate neuron adhesion even for the lowest particle density.With increasing number of particles the relative live cell density rises steeply, however, saturates at approx.1.7.This means that not only the live cell density increases with the particle density but also that the density of vital neurons is higher on surfaces with high particle density than on PDL control samples.The nal neuron viability for samples with high particle densities is about three times higher than the cell density without any particle.
1][72] Only very few cells can survive on PEG backlled substrate but in an unhealthy state.With increasing particle densities the relative live cell density increases until it saturates.For each specic particle density, the neuron viability was larger for samples without backll than with PEG backll.Also the saturation occurs at higher particle densities for backll modied samples.In the case of a hydrophobic OTS backll, we found a poor neuron adhesion for the lowest particle density which increases only marginally until the cell viability steeply increased at a particle density of approx.500 particle per mm 2 .Remarkably, the nal relative neuron density is very similar for all surface modications independent of the used backll.
To evaluate the effectiveness of the backll, we plated neurons on bare, PEG silane, and OTS modied SiO 2 surfaces without any AF-AuNP.The bare SiO 2 surface does not facilitate a complete suppression of neuron adhesion (Fig. 3

top le).
Almost the same relative live cell density was found (0.5) for these samples as for AF-AuNP surfaces with low particle densities 4 particle per mm 2 .In addition we observed, that the neurons are able to spread from particle containing (500 particle per mm 2 ) to particle free areas of the surface without clear border between these domains (Fig. 3

top right).
The number of vital cells is higher on the particle modied regions, however, the neurons manage to mature also on bare  SiO 2 .On the contrary, no vital cells could be observed on bare PEG silane modied SiO 2 (Fig. 3 middle le).For samples modied with AF-AuNP and PEG silane backll we observed a sharp border between particle containing and particle free areas (Fig. 3 middle right).Neither somas are able to adhere nor do neurites manage to grow out from particle containing to particle free areas.OTS modied samples showed a similar behavior (Fig. 3 bottom).Apparently, both PEG silane and OTS efficiently suppress neuron adhesion while SiO 2 only reduces partially cell viability although no neuron binding ligands have been applied to the surface.To evaluate the inuence of the backll onto the neural development and neurite outgrowth, we cultured neurons at 9 DIV on AF-AuNP modied surfaces without backll, with PEG silane, or OTS backll.All samples possessed a high particle density of approx.500 particles per mm 2 .Subsequently, we stained the neurons for axons and dendritic neurites by rabbit anti-MAP2 primary antibodies and mouse anti-Tau1 antibodies, respectively.
The neurons cultured on backll free particle surfaces show a matured development with soma diameters in the range of 20 to 30 mm (Fig. 4 top).Each neuron possesses in average 6 short dendritic neurites which are typically shorter than 100 mm and stained orange.The axons stained in green have a length of more than several hundred microns.At this stage of development the neuron axons tend to loop back and to form networks of crossing axonal extensions.The morphological appearance of the neurons is similar to those of neurons cultured on PDL control samples or on extra cellular matrix proteins at glia containing co-culture conditions. 73Neurons cultured on surfaces modied by AF-AuNP and PEG silane backll exhibit a development that is similar to the backll free AF-AuNP samples, however the number of processes and length of axons are slightly smaller, Table 2. Neurons plated on OTS backlled samples have a different appearance.Although the number of neurons adhered to the surface as well as the number of neurites are comparable to the other two samples, length of the neuron processes is signicantly smaller.An analysis of the branching of neurites revealed that the inuence of AF-AuNP on the formation of higher order branches is small.
To further evaluate the adhesion of the cortical neurons on particle modied surfaces we investigated the samples by scanning electron microscopy.The high resolution capabilities of this technique allows a detailed imaging of the spreading of the neurite processes on the surface and simultaneous mapping of particles and neurons.Exemplary SEM images of neurons at 9 DIV cultured on AF-AuNP surfaces are shown in Fig. 5.
The cell somas tightly spread on the particle lm with many processes extending to the cell environment.The neurites themselves possesses many sub-micron sized lopodial processes that are contacting the surface of nanoparticles.Cross sections of the sample surface/neuron interface obtained from FIB cutting of xated cells demonstrate that the particles are not internalized by the neurons and still associated to the sample surface even aer long culturing time in vitro.This observation further conrms that the backll does not impair the binding strength of the particles to the sample surface.The electrostatic interaction between AF-AuNPs and the APTES functionalized SiO 2 surface is strong enough to hold the particles on the surface even aer cells interact with the particles in a cell  culture medium (Fig. 5 bottom). 74Also the top view SEM images do not reveal any evidence of a displacement (particle free areas next to neuron extensions) or accumulation of particles due to mechanical actuations of the cells.We did not observe any signicant difference of the surface/ particle/neuron interface for all kinds of backlls investigated by us.
In our previous work we used micro-and nanostructured AF-AuNP patterns for guiding of neuron adhesion.We achieved the guiding of neurons by establishing an adhesion contrast between particle associated positively charged amino groups and the SiO 2 substrate with predominant negative surface charges at neuron culture conditions.By this means we achieved a guiding efficiency of about 71%. 56The results described above, however, revealed that the difference of the viability of neurons on particle free SiO 2 surfaces is just a factor of three smaller than for surfaces with a high density of particle associated binding ligands.Furthermore, we found that the chemical contrast to AF-AuNP can be enhanced by applying a backll to particle surrounding areas.Therefore, we evaluated the guiding efficiency for the same AF-AuNP pattern as described previously however this time in combination with an OTS backll.The particle surface patterns were established on oxidized silicon samples by a process using nanoimprint lithography (NIL), particle immobilization, and backlling, Scheme 2.
The pattern consisted of 8.2 mm-wide stripes separated by gaps of 11.6 mm.The stripes are sub-structured in micron-sized squares of 10.1 mm Â 10.1 mm and line packages.The latter are composed of nano-sized lines of 200 to 400 nm widths and 300 nm gaps in between.The micron-scaled squares should provide nodes for a conformal adhesion of the neurons, whereas the features consisting of nano-scaled lines were adapted to the size of neurites and their extensions.Aer plating cortical neurons onto the patterned surface, the cells preferably adhere onto those areas that are decorated by high density AF-AuNP.At 3 DIV, a distinct majority of the neurons possess a bidirectional neurite outgrowth with processes longer than 100 mm.A branching of neuritis orthogonal to the particle stripes and crossing of the particle free areas occurs relatively seldom although the distance between the stripes is just 11.6 mm.The cells are well aligned to the particle pattern, Fig. 6.The guiding efficiency, represented by the ratio of neuron covered surface area on the pattern/the total neuron covered surface area, was higher than 90%.Thus, the guiding was signicantly enhanced in comparison to earlier reports based on the optimized composition of particles and particle backll.The guiding efficiency was determined accordingly to a previously reported procedure. 56heme 2 A UV-transparent polymer mold is fabricated by a hot embossing replication process from a Si-mold (a-c).A two-layer imprint resist is applied to Si/SiO 2 samples and imprinted under UV illumination (d).The residual and underlayer resists are removed by a dry etching process (e and f).The opened areas are functionalized by amino-terminated silane and subsequently the resist stack was removed by lift-off (g).Next, aminosilane patterns are decorated with citrate-stabilized AuNPs (h).Finally, particle-free SiO 2 surface domains are modified by backfill molecules and neuron adhesion ligands are attached to the particles (i).
Fig. 5 Focus ion beam cutting of cell after 9 DIV on the AF-AuNPs substrate without backfill: top-neurons after fixation, middle-amplified images of part of neurons after fixation, bottom-cross section of neurons after FIB cutting.

Discussion
In this report we describe the tuning of the chemical composition of solid surfaces on the nanometer scale by controlling the density of AF-AuNP and their surrounding matrix.The amino-functionalized nanoparticles act as anchor point for the attachment of cell adhesion ligands.The successful immobilization and modication of the nanoparticles has been evaluated by SEM and XPS, respectively.Together with the observation that the particles remain in place even aer extended culture time proves that backlling from vapor-phase does not impair the binding of AF-AuNP.
Since the density of surface associated particles can be varied, we are able to control the density of binding cues on the surface.This has a signicant impact on the viability of the neurons.At the lowest densities only a few particles (surface coverage of 0.1%) can be observed on an area of 1 Â 1 mm 2 .However, each of these particles carries several thousand aminoalkyl thiol molecules.Assuming a coverage of 5.5 AUT per nm 2 (ref.75) on the particle surface we obtain a ligand density of 3.1 Â 10 3 AUT molecules per particle if we take only the upper, solution exposed semi-sphere of the particle into account.The Debye length is about 1 nm for cell culture media, which leads to an effective screening of electrostatic interactions.Therefore, we can assume that the range of interactions of an individual particle is restricted to its hydrodynamic radius of about 11 nm.Consequently, only a small number of highly charged but locally conned adhesion spots are available at sample surfaces with low particle densities.Astonishingly, a relatively high number of vital neurons per unit area was observed for SiO 2 samples with low particle densities and without backll.This indicates that the neurons manage to adhere at particle free areas of the SiO 2 surface.This assumption was conrmed by the spreading of neurons from particle containing to particle free areas, Fig. 3.The presence of particle associated adhesion ligands enhances the neuron viability, however is not mandatory for cell survival.The isoelectric point of silicon dioxide can be typically found at pH values of 2-3. 76hus, it can be assumed that the SiO 2 surface carries a negative net charge at cell culture conditions.It has been reported that the adhesion of cells to negatively charged surfaces is rather limited due to electrostatic repulsion with the cell membrane, which is negatively charged too (À18 Â 10 À3 C m À2 ). 77However, it is also an established fact that the hydrophilic properties of the SiO 2 surface support the adsorption of solution-borne substances, 1 which may originate from the neurons themselves.For instance neurotrophins and growth factors can get associated to negatively charged glycoforms of cell adhesion molecules.Such solution-borne species apparently impair the cell aversive properties of SiO 2 .
Applying a backll to the particle environment improves the control over the surface composition by preventing undesired adsorption of interfering binding cues.The effectiveness of the 2-[methoxy(polyethyleneoxy)6-9-propyl] trichlorosilane and octyltrichlorosilane backll coating was demonstrated by particle free control experiments where a clear suppression of neuron adhesion was achieved.The viability of neurons is very low even for samples modied by AF-AuNP at density lower than 30 particles per mm 2 in the presents of PEG-based backlls.The number of surface tethered adhesion ligands is not sufficient to support maturation to vital neurons.With increasing particle density, the relative live cell density rises until it levels off at about 1.5 to 1.7.Under these saturation conditions, neurons exhibit a matured development with long as well short, branched neurites at 5 DIV for no backll and PEG silane backll.The size of the soma, number of neurite processes, as well as the length of the axons are similar to neurons plated on a PDL control culture and correspond well to a neuronal development of stage 4. 73 It can be assumed that the neurons nd a sufficient number of surface associated binding cues for their development.Also samples modied with OTS backll reaches neuron densities values similar to PEG silane and no backlled samples at high AF-AuNP densities.It is remarkable, that we nd about the same relative live cell density for particle densities higher than 500 particles per mm 2 , independently of the employed backll.Together with the observation that cell density asymptotically converges towards this value suggest that the number of receptors that get associated to the particles saturates.
To better understand the nature of the electrostatic interactions between cell and sample surface, we estimated the cationic charge density of the sample by multiplying the number of aminoalkyl thiol adsorbed to one AuNP with the respective particle density.We assume that the backll prevents the adsorption of charged, solution-borne compounds.Considering only the AF-AuNP associated ligands, we obtain a surface charge density of 24 Â 10 À2 C m À2 for 500 particles per mm 2 .This value is about one order of magnitude higher than the surface charge density of 15 Â 10 À3 C m À2 determined for HEK293 cells. 3The difference between sample associated positive charges and negative cell surface charges suggests that an excess of positive charges is required at the sample surface to support the neuron adhesion.In fact, a particle density of about 30 AuNP per mm 2 would be sufficient to compensate surface charge density of a cell, however are not sufficient to achieve a saturation of the live cell density, Fig. 2. One possible reason for this nding is the unequal distribution of charges on both surfaces.In the case of AuNP, a large number of amino-ligands (see above) is bound to a single anchor spot with a hydrodynamic radius of 11 nm.The short Debye length of the cell culture media leads to an effective screening of electrostatic interactions.On the contrary, the density of cell adhesion molecules (CAMs) on the cell membrane is in the range of 500 to several thousand per square micrometer. 22The CAMs exhibit glycoforms with attached glycans.These glycans can carry a high number of negative charges as in the case of NCAM associated polysalic acid (PSA).However, also these charged entities are screened by the high concentration of salt ions which connes the range of possible electrostatic interactions.Due to the ionic screening under culture conditions just a limited number of the CAMs would be able to bind to the surface if the particle density is as low as 30 AuNP per mm 2 .This suggests that a certain number of homogeneously distributed AuNP is required to facilitate a sufficient binding of glycosylated CAMs which nally enables good cell viability.This is seemingly the case at 500 randomly distributed 20 nm AuNP per mm 2 for all types tested backll, which is in the same order with the approximate density of CAMs on the cell membrane.Although the size of the particles is in the same range of the most glycosylated proteins it cannot be excluded that several proteins bind to the same particle.Furthermore, it is likely that several types of glycosylated CAMs are involved in the binding process, since the electrostatic interactions are rather unspecic.
Although we found similar maximum cell viability for all types of particle background we observed distinct differences for the maturation of the cells.In particular the development of neurites is different for samples with OTS backll in comparison to the other samples.The hydrophobic alkyl-silanes effectively suppress neuronal adhesion and development even at relatively high densities of AF-AuNP (300 particles per mm 2 ).At maximum particle densities (500 particles per mm 2 ) we nd neuron viability and number of neurites similar to samples with no or PEG backll but the length of the axons is signicantly reduced to about 1/3.The tendency to develop shorter neurite processes is probably associated to the hydrophobic passivation of the particle surrounding matrix for protein adsorption.Apparently neuron adhesion and neurite outgrowth depend differently onto the chemical background of cationic adhesion ligands.A selective dependence of axonal features on particle density has been recently observed also by Thelen et al. for dorsal root ganglia (DRG). 53During path nding as well as for adhesion the neuron can integrate over large distances and overgrow cell aversive sample areas with the length of several tens of microns. 7owever, small neurite extensions which possess oen a width smaller than 200 nm, Fig. 5, are more sensitive to variations of the surface composition on nanometer scale.
It turns out that the control of the composition of particle surrounding area is a powerful tool to tune the properties of surfaces on nanometer scale.The inuence of the particle composition on the local neuron adhesion and neurite outgrowth was demonstrated previously. 56However, a high guiding accuracy is hampered by an undened composition of the particle surrounding.On average, we nd a live cell density of 30% at bare SiO 2 without surface tethered adhesion ligands in comparison to samples with high particle densities.The composition of backll free SiO 2 surface areas may change over time due to adsorption, desorption, and surface diffusion of undened solution-borne compounds. 78This may strongly affect the reproducibility of the neuron adhesion properties of samples without backll and has direct impact on guiding attempts.The adhesion contrast between particle free and particle decorated areas is only moderate.
The backll increases the adhesion contrast for the neuron adhesion.Here, we dene the adhesion contrast (C) as ratio of the difference between relative live cell density of cell attractive surface (A) and relative live cell density of cell aversive surface (R) divided by the relative live cell density of cell attractive surface, eqn (1) A value of C ¼ 0 indicates no difference in adhesion between the two surface domains whereas C ¼ 1 represents a perfect adhesion contrast.This cell type specic qualitative parameter allows to estimate the appropriateness of a material boarder to restrict the spreading of cells across this boarder.From Fig. 2, we determined a chemical contrast of C ¼ 0.53 for AF-AuNP on backll free SiO 2 ; a value of C ¼ 0.87 for AF-AuNP with PEG backll; and a value of C ¼ 1 for AF-AuNP on OTS modied SiO 2 .Therefore we used the relative live cell density at 500 particle per mm 2 as A and the relative live cell density at 10 particle per mm 2 for the respective types of backll as R. The latter represents the averaged particle density observed for areas between the particle patterns, Fig. 6. 66 The high contrast obtained by the application of a backll resulted in an improvement of the guiding efficiency from 71% for backll free nanoparticle patterns to 92% for backlled but otherwise identical surfaces.It should be noted, that the high adhesion contrast was achieved for a surface that was entirely covered with particles.Solely the density of the particles varies from 10 particle per mm 2 to 500 particle per mm 2 .The local density and distribution of nanometer sized domains of cell attractive and cell aversive cues controls the adhesion of cortical neurons and their maturation.

View Article Online
Milli-Q water, then dried with N 2 ow.The cleaned substrates were activated with oxygen plasma at 200 W for 2 min (100-E plasma system, Technics Plasma GmbH).Thereaer, a monolayer of 3-aminopropyltriethoxysilane (APTES, 99%, Sigma) was formed on the surface by using vapor deposition in a glove box.Citrate stabilized AuNPs (0.01 wt%, Sigma) were deposited onto the substrates overnight.

Ligand immobilization
11-Amino-1-undecanethiol hydrochloride (AUT) (Sigma) was dissolved in absolute ethanol.Aer removal of citrate with oxygen plasma for 2 min, the substrate was soaked into 0.5 mM ligand solution overnight.Aerwards, the substrate was rinsed with copious of Milli-Q water and dried with N 2 -ow before further use.For the modication of the nanoparticle background, the samples are modied with octyltrichlorosilane or 2-[methoxy(polyethyleneoxy)6-9-propyl] trichlorosilane right aer citrate removal.Before ligand modication, the substrate was washed with dehydrated toluene and ethanol to remove any physically absorbed silane molecules.Then, the substrates were modied with 0.5 mM AUT dissolved in ethanol.

Characterization
The composition of the substrate was investigated with XPS to evaluate binding of the 11-amino-1-undecanethiol linked to gold nanoparticles, (PHI 5000 VersaProbe II, ULVAC-PHI).The silicon substrates were washed with acetone, 2-propanol and Milli-Q water, dried with N 2 .Right aer activation with oxygen plasma for 2 min, the samples were silanized with OTS and PEG silane respectively.The contact angles of the above two surfaces together with the bare silicon substrate were measured with a contact angle measurement system (contact angle measurement instrument OCA 20, Dataphysics) by the Sessile drop method with 5 mL Milli-Q water to estimate their wetting properties.

Cell culture
Rat embryonic cortical neurons were obtained and cultured as described by Brewer et al. with slight modication. 79Embryos were dissected from pregnant Wistar rat at 18 days gestation.Cortices were dissected from the embryonic brains and mechanically dissociated by trituration in 1 mL ice-cold HBSS (without calcium and magnesium), 0.035% sodium bicarbonate, 1 mM pyruvate, 10 mM HEPES (Sigma, Germany), 20 mM glucose (Sigma, Germany), pH 7.4, with a re-polished, silanized Pasteur pipette.The cell suspension was diluted 1 : 2 in HBSS (with calcium and magnesium), 0.035% sodium bicarbonate, 1 mM pyruvate, 10 mM HEPES (Sigma), 20 mM glucose (Sigma), pH 7.4.Nondispersed tissue was allowed to settle for 3 min and the resulting supernatant was centrifuged at 200 g for 2 min.The pellet was resuspended in 1 mL neurobasal medium with 1% B27, 0.5 mM glutamine, and getamycin per hemisphere.A fraction of the cells were counter stained with trypan blue and counted in a Neubauer counting chamber.The remaining cells were diluted in supplemented neurobasal medium.Diluted cells were plated onto the substrates at 10 000 cells per cm 2 .The medium was exchanged twice every week.All reagents were purchased from Invitrogen unless otherwise noted.

Immunochemistaining
Cells were characterized by uorescence microscopy aer live dead staining and antibody staining.For live dead staining, cells were incubated in a solution of calcein AM (live cells staining) and ethidium homodimer (dead cells staining) in PBS for 10 min.For antibody staining, cells were rinsed twice with pre-warmed PBS (pH 7.4) at 37 C, and xed with 4% paraformaldehyde solution for 7 min at 4 C. Aer rinsing cells for three times with PBS, the xed cells were permeabilized with 0.1% Triton X in blocking buffer for 15 min at room temperature and were again washed three times with PBS.Non-specic binding sites were blocked with 2% bovine serum albumin (BSA) and 2% goat serum in PBS overnight.To distinguish neurons from glia cells, the samples were immersed into mouse anti-MAP2 primary antibody (neuron specic) and rabbit anti-Glial Fibrillary Acidic Protein (GFAP) primary antibody with a working concentration of 1 : 500 in BSA blocking solution for 90 min and washed three times with PBS.Subsequently, the cells were incubated with secondary antibody (Alexa Fluor 488 anti-mouse and Alexa Fluor 568 anti-rabbit) together with DAPI for another 90 min.Finally, the substrates were washed three times with PBS and once with water before mounting in Dako Fluorescent mounting medium.Fluorescent imaging of stained samples was carried out with an AxionInager Z1 (Zeiss).Images were taken with an AxioCam MRm camera (Zeiss) and processed using AxioVision 4.7 soware (Zeiss).In the case of axon and dendrite specic staining, rabbit anti-MAP2 primary antibody and mouse anti-Tau1 primary antibody were used.Alexa Fluor 488 anti-mouse (Invitrogen) and Alexa Fluor 568 anti-rabbit (Invitrogen) were used as second antibody.The remaining processes were the same as for neurons and glials staining.Neurolucida soware was used for cells counting and cell viability measurements.

Focus ion beam cutting and scanning electronic microscopy characterization
Cells were xed for focus ion beam (FIB) cutting as follows: cells were rinsed twice with pre-warmed PBS (37 C) and then xed with 3.2% glutaraldehyde in PBS (pH 7.4) for 2 h.Samples were rinsed twice with PBS and subsequently dehydrated with increasing concentration of ethanol (30%, 50%, 70% for 10 min respectively, 90%, 95%, 100% each 5 min for three times, 100% for storage).The FIB cutting was done with a Helios Nanolab 600i apparatus (FEI).Cells were embedded in a 1 mm-thick platinum layer and subsequently cut by a Ga-ion beam at a beam current of 2 nA and an acceleration voltage of 30 kV.SEM images were performed by Gemini 1550 (Leo/Zeiss) using an inlens detector.SEM images with different particle densities were processed and counted with the aid of ScionImage soware. 66

Conclusions
In summary, we demonstrate in this work that both neuron viability and maturation can be efficiently inuenced not only by changing the density of gold nanoparticles on a SiO 2 surface but also alternating the chemical composition of their environment on nanometer scale.The AuNP were modied by aminoalkyl thiol molecules and act as adhesion spots for the immobilization of neurons.The particle surrounding was altered from hydrophilic SiO 2 , to protein aversive poly(ethylene glycol)-molecules, and hydrophobic alkylsilane molecules.Increasing densities of amino-functionalized nanoparticles enhance the adhesion of vital neurons and saturate at about 500 AuNP per mm 2 which is in the same order as the density of glycosylated cell adhesion molecules on cell membranes.The density of the particle associated positive charges is about one order of magnitude larger than the previously reported surface charge density of the membrane glycocalyx.Nevertheless, the viability of neurons on these samples is higher than on homogenously covered PLL samples.Moreover, the backll also inuences the development of neurites on the surface.In particular the length of the neurites was reduced by 2/3 for samples with hydrophobic backll in comparison to samples with hydrophilic particle environment.By modifying the particle's surrounding matrix, the cell adhesion can be suppressed for low particle densities but allows healthy neuron maturation for dense particle packing.The enhanced chemical contrast between particle free and particle containing areas resulted in an increase of the guiding efficiency for rat cortical neurons to more than 90% in contrast to previous results.By means of surface tethered nanoparticles the composition of solid surfaces can be tuned on nanoscale to match the number and distribution of various types of adhesion molecules in the cell membrane.This approach can contribute to the investigation the complex processes involved in neuron adhesion on the scale of adhesion complexes and to meet demanding challenges of neuroscience and bioelectronics such as inducing polarity, controlling neuronal co-culture, and prevention of glia proliferation.Moreover, this technique can be used as versatile tool to engineer technical surfaces to enhance cell viability, reduce immune responses, and improve the cell/device communication of bioelectronic hybrid systems.

Fig. 1
Fig. 1 Top: scheme of SiO 2 surface modified with gold nanoparticles at different densities.Middle: SEM images of SiO 2 surface modified with gold nanoparticles at different densities (scale bar: 200 nm; number: AuNP per mm 2 ).Bottom: live (green) and dead (orange) staining images of primary rat cortical neurons seeded at div6 on AF-AuNP modified SiO 2 surfaces with PEG backfill (scale bar: 100 mm).

Fig. 2
Fig. 2 Dependence of the live cell density of primary rat cortical neurons on the concentration of gold nanoparticles.The cells were seeded at 9 DIV on SiO 2 surfaces modified by amino-functionalized particles: bluewithout backfill, redwith PEG backfill, and blackwith OTS backfill.

Fig. 6
Fig. 6 Guiding of neurons on AF-AuNPs confined in line package pattern: top-AF-AuNPs decorated line package pattern fabricated by NIL (scale bar: 5 mm), bottom-neuronal guiding on AF-AuNPs line package pattern with octyltrichlorosilane backfill (scale bar: 50 mm).

Table 2
Neurites number and axon length comparison for different surfaces after 6 DIV Neurites number Axon length (mm) PDL 7.2 AE 0.9 380 AE 67 Particles without backll 6.1 AE 0.8 368 AE 69 Particles with PEG backll 5.4 AE 0.9 302 AE 59 Particles with OTS backll 6.2 AE 0.9 130 AE 76