Nanowire Assisted Mechanotyping of Cellular Metastatic Potential

Nanotechnology has provided tools for next generation biomedical devices which rely on nanostructure interfaces with living cells. In vitro biomimetic structures have enabled observation of cell response to various mechanical and chemical cues, and there is a growing interest in isolating and harnessing the specific cues that 3D microenvironments can provide without the requirement for such culture and the experimental drawbacks associated with it. Here, a randomly oriented gold coated Si nanowire substrate with patterned hydrophobic–hydrophilic areas for the differentiation of isogenic breast cancer cells of varying metastatic potential is reported. When considering synthetic surfaces for the study of cell‐nanotopography interfaces, randomly oriented nanowires more closely resemble the isotropic architecture of a natural extracellular matrix. In the study reported here, the authors show that primary cancer cells preferably attach to the hydrophilic region of randomly oriented nanowire substrate while secondary cancer cells do not adhere. Using machine learning analysis of fluorescence images, cells are found to spread and elongate on the nanowire substrates as compared to a flat substrate, where they mostly remain round. Such platforms can not only be used for developing bioassays but also as stepping stones for tissue printing technologies where cells can be selectively patterned at desired locations.


Introduction
Important cellular functions like adhesion, migration, regeneration, apoptosis, proliferation, polarization, among others are influenced by the mechanical cues provided by extracellular matrix (ECM), the non-cellular tissue component surrounding imaging and interdependent microenvironmental factors, micro-and nanopatterned substrates have gained significant attention. For instance, while micropatterned surfaces have been used to precisely control cell adhesion and spreading, [17,18] nanoscale gratings have been leveraged to restrict the movement and clustering of specific receptors to alter signaling from these receptors. [19] Along with surface topography, surface chemistry [20,21] and wettability [22,23] play an important role in regulating cellular phenotypes and behaviors. Specifically, cell morphology and adhesion have a strong correlation with the wettability of a surface. [24] Increased or decreased contact area of cells in solution on a surface influences cell shape and strength of cell-substrate adhesion, resulting from the hydrophilic or hydrophobic nature of this surface. The cytoskeleton, a dynamic supramolecular structure governing important cellular functions, adapts its organization to accomplish force balance between the cell and the surface. Surface wettability determines how much of the surface is exposed to culture media leading to adsorption of adhesive cellular proteins. Lower interfacial tension between the liquid and the substrate surface may lead to less adsorption of adhesion-mediating proteins, resulting in adhesion failures for cells. [24] Thus, along with nanotopography, surface wettability is an important consideration to determine cell surface interactions.
Among various topographies in the size range that closely mimic natural ECM fibrils, [25,26] vertical nanowire-based systems have been one of the most widely explored architectures for studying cellular behavior, [27,28] probing intracellular biomolecular changes [29,30] or in situ drug [31] or gene delivery. [32] Vertical nanowire-assisted intracellular neural recording [33,34] and piconewton force measurement in neural growth cones [35] have been reported. Nanowires combined with transistors have been used for biosensing applications. [36] Modulation of cell adhesion by super hydrophobic and hydrophilic surfaces [37] has been utilized for cell patterning in bioassay. [38] Micropatterning of cells was enabled by a combination of topography and super hydrophilicity, allowing cells to concentrate on the hydrophilic region. [39] While most prior studies on cell-nanostructure interfaces have been based on vertical nanowires, to the best of our knowledge, there are no other studies related to the cell-surface interactions of randomly oriented nanowires, which better resemble naturally occurring ECM organization. [40,41] Here we report a novel substrate with swaths of randomly oriented hydrophobic-hydrophilic gold-coated silicon nanowires (Au/ SiNWs) to sort human breast cancer cells of different stages of metastatic progression. In addition to advantages which include high yield, low cost, and facile fabrication, randomly oriented Au/SiNWs are cyto-compatible with mammalian cells. [42,43] Moreover, their large surface-to-volume-ratio enables greater cell-surface interactions, which allow for sensing and probing of cellular function by electrical and optical means. [42][43][44] The spreading, elongation and concentration of primary tumor cells on the hydrophilic nanowire surface without any protein coating demonstrates the potential of a robust biomimicry platform for tissue printing applications. Surprisingly, secondary cancer cells isolated from lung and bone metastases do not/minimally adhere to either the hydrophilic or the hydrophobic region of the Au/SiNWs substrate. This paves the way for nanowire-based bioassays for diagnosing and monitoring tumor progression.

Results and Discussion
Randomly oriented SiNWs were obtained by plasma enhanced chemical vapor deposition (PECVD) on silicon (Si) wafer and successively coated with gold (thickness of 120 nm) by thermal evaporation.
The forest of Au/SiNWs is intrinsically highly hydrophilic and was converted to super-hydrophobicity by treatment with the plasma of fluorocarbon (CHF 3 ) for 90 s (details in Experimental Section). Figure 1a shows photographs of the wettability difference between the untreated (left) and treated area (right) on the same Au/SiNWs substrate. The two regions with different wettability were obtained by using a contact mask that protected half of the sample during fluorocarbon treatment. The water droplet on the uncoated region was well spread, wetting a large area of the surface, while on the other side, the droplet sitting on the treated region showed a contact angle of 155° (Figure 1b Before introducing the Au/SiNW substrate into cell culture, a micro-pattern consisting of hydrophobic-hydrophilic regions was fabricated on the NW mat by using a two-step process as summarized in Figure 2a. First, the AuSiNW surface was modified through fluorocarbon plasma to achieve hydrophobic properties (Figure 2a, 2). Next, the hydrophobic-hydrophilic micro-pattern was fabricated by selective removal of fluorocarbon coverage using an oxygen (O 2 ) plasma-based process through reactive ion etching (RIE) (Figure 2a, 4). To pattern the hydrophilic regions, a shadow mask (24 circular holes with a size and a separation of 500 µm each) was placed in contact with the sample. For a practical demonstration of the selective hydrophobic−hydrophilic patterning, we dispensed water droplets on the patterned substrate which resulted in the droplets rolling off hydrophobic regions and preferentially settling on hydrophilic regions (Figure 2b). Adherent cells have an affinity towards hydrophilic surfaces. [39] Thus, in biological contexts, it is expected that cells will tend to grow and concentrate preferentially on hydrophilic regions.
Two different cell lines, parental human breast cancer cells MDA-MB-231 (P231) and lung-metastatic cells (LM231), which were derived from P231 in an orthotopic mouse tumor xenograft model in which LM231 was isolated from spontaneous lung metastases, [45] were tested in our study alongside a third bone-selective metastatic derivative MDA-MB-231-1833 (BoM). Our recent investigations had revealed distinct differences in the biophysical and morphological properties of the P231 and LM231 cells. Briefly, LM231 cells were found to be more motile and less stiff, highlighting their invasiveness as compared to P231 cells. [46] Furthermore, LM231 cells displayed a larger area and higher cell mass as compared to P231 cells. [47] Based on these recent findings, we sought to investigate if P231, LM231, and BoM cells would interact differently with the hydrophobichydrophilic nanowire substrate.
Constitutively tdTomato-expressing P231 and LM231 cells were cultured on the substrate with patterned Au/SiNWs areas. BoM cells were also cultured on the patterned substrate. In the same dish a flat silicon wafer was placed as a control surface at the same time. None of the substrates were modified with collagen or poly-L-lysine, which are known cell adhesion promoters. The culture protocol is provided in the Experimental Section.
The substrates were fixed 48 h after plating and observed using a fluorescence microscope. The fluorescence image of hydrophobic-hydrophilic Au/SiNWs seeded with P231 cells (Figure 3a) shows that the cells grew on the hydrophilic portion of the pattern at high density (lower right corner), whereas cells were nearly absent from the hydrophobic section (upper left corner). A minimum number of LM231 and BoM cells, on the other hand, were observed in the hydrophilic circles (upper right corner of Figure 3b and upper left corner of Figure 3c, respectively) and could not be found at all on the hydrophobic side (lower left corner and right corner). Uniform distributions of P231, LM231, and BoM cells were observed on Si wafer (Figures 3d, 3e, and 3f, respectively) confirmed that cells were uniformly dispersed in the solution prior to plating. These findings support the notion that the observed differential adhesion resulted from combined effects of substrate topography and wettability.
ECM remodeling in the primary tumor microenvironment is strongly correlated with metastatic progression. [48,49] Biochemical and biophysical changes occur in the ECM in close proximity to the tumor, leading to metastasis. For example, increased protein deposition [50] and increased stiffness of the ECM [51] were found to contribute to primary tumor progression. However, the engrafted cancer cells have poor efficiency of forming a secondary tumor owing to the unfavorable microenvironment at the distant site, [52] often requiring local cells at the invaded site to secrete additional ECM proteins to facilitate colonization. [53] Such difference suggests that cell-ECM interaction of primary cancer cells may be different than that of secondary cancer cells, which is in agreement with our observation on Au/SiNWs.
A closer inspection of the fluorescence images reveals that the morphology of P231 cells grown on the hydrophilic Au/ SiNWs versus that on the planar substrate were different from each other. Although neither substrate was coated with collagen or poly-L-lysine, P231 adhered to and grew on both. However, fluorescence images show that P231 on the hydrophilic Au/ SiNWs (Figure 4a) exhibited 1.872-fold greater spreading and 1.39-fold more elongation compared to that of cells grown on control Si (Figure 4b). SEM of dried P231 cells on the hydrophilic Au/SiNWs (Figure 4c) and flat Si wafer (Figure 4d) confirmed these findings. The cells grew on both surfaces indicating that they were healthy and the difference in morphology was promoted by the nanostructuring of the Au/SiNWs. Similar observations have been reported earlier, where cells were grown and elongated along the ridges of nanostructured grooves. [54][55][56] Some other studies report opposite observations [57] indicating a strong dependence of the observed phenomenon on cell type and surface modifications. [26,58] To further characterize the differential morphologies based on the substrate, we used VAMPIRE (Visually Aided Morpho-Phenotyping Image Recognition) analysis [59,60] to elucidate subtle differences within cell morphologies. Frequently, cell shapes vary greatly while keeping broad morphological parameters such as the area and aspect ratio preserved. VAMPIRE analysis helped us to tease out the subtle differences in cell shapes while ensuring a high throughput of the approach. For the analysis, fluorescence images were pooled to collect more than 1500 and 2000 P231 cells on Au/SiNWs and Si wafer, respectively. The histogram in Figures 4e and 4f depict the percentage of different shape modes for each condition. A dendrogram of representative shape modes is shown in Figure 4g. Approximately 50% of the cell shape was elongated with a circularity of 0.66 on Au/SiNWs, whereas round shapes (circularity: 0.87) were dominant in the case of flat surfaces. Furthermore, although it is outside the scope of our current study, our observations indicate that the degree of surface roughness may also play a role in determining cellular morphology on such substrates (Supporting Information).
Finally, we assessed the viability of the P231 cells on hydrophilic Au/SiNWs. At 24 h post seeding, cells on Au/SiNWs and plain Si wafer were stained to assay for live and dead cells. For the live cell assay, Calcein AM diluted in phosphate buffered saline (PBS) was used. To probe dead cells, we used Hoechst 33342, which stains nuclei regardless of whether cells are alive or dead. Cells stained with Hoechst only were dead. The percentage of live and dead cells was assessed using ImageJ as shown in Figure 5a and b. Figure 5c quantitatively displays the percentage of live and dead cells on Au/SiNWs as compared to Si wafer. Au/SiNWs had a slightly reduced live-to-dead ratio (81.6% live cells) as compared to Si (96.3% live cells). The slightly lower survival rate of cells on Au/SiNWs versus Si may have been caused by the complex topography of Au/SiNWs substrate, where dead cells may have been retained on its surface. Overall, 81.6% cell survival on Au/SiNWs substrate suggests that it is intrinsically inert and does not impose cytotoxicity on cells.

Conclusion
In this report, we demonstrate a patterned hydrophobic-hydrophilic Au/SiNWs platform that is able to differentiate isogenic breast cancer cells of two different stages of metastatic progression. The primary cancer cells remained alive and  www.afm-journal.de www.advancedsciencenews.com selectively adhered to the hydrophilic region of the nanowire substrate. Furthermore, the cells exhibited distinct morphologies on Au/SiNWs compared to their counterparts on a flat surface, indicating that randomly oriented nanowire substrates can be used to modulate cellular phenotypes. With further development, such platforms can be used in synthetic biology or for tests to evaluate the metastatic capacity of cancer cells. Chemical functionalization or integration with plasmonically active materials can provide added functionality to the nanowire platform for detailed mechano-chemical studies in the future.

Experimental Section
Fabrication of Au/SiNWs: Au catalyzed SiNWs were produced by plasma enhanced chemical vapor deposition (PECVD) on Si wafer. To initiate NW growth, a 2 nm thick Au film was evaporated on the substrate. SiNW growth was performed with SiH 4 and H 2 as precursors at a total pressure of 1 Torr and flow ratio of SiH 4 /(H 2 +SiH 4 ) fixed to 1:10. The substrate temperature was kept constant at 350 °C. A 13.6 MHz radiofrequency with a fixed power of 5 W was used to ignite the plasma. The Au coverage on the SiNWs was obtained by thermal evaporation.
Fabrication of Hydrophobic-Hydrophilic Surface on Substrate with Au/ SiNWs: The micro-pattern consisting of hydrophobic-hydrophilic surface regions was fabricated on Au/SiNW substrates by using a two-step process. First, to obtain a hydrophobic surface the Au/SiNW substrate was exposed to a plasma of fluorocarbon (CHF 3 ) and Ar for 90 s at a total pressure of 100 mTorr and a flow ratio of CHF3/(CHF3+Ar) fixed to 60%, at room temperature and RF power of 50 W. [61] Next, the hydrophobichydrophilc micro-pattern was fabricated by selective removal of fluorocarbon coverage using an oxygen plasma-based process through a reactive ion etching (RIE) technique. To select hydrophilic regions, we used a shadow mask in contact with the sample that had a matrix of 24 circular holes with a size and a separation of 500 µm each.
Cell Culture on the Au/SiNWs: 10% fetal bovine serum (Corning Cellgro), 100 U mL −1 of penicillin were added to Roswell Park Memorial Institute (RPMI) 1640 with L-glutamine (Corning Cellgro) prior to use for culturing P231 and LM231 cells. MDA-MB-231-1833 boneselective metastatic derivative, BoM, was kindly provided by Dr. Joan Massagué (Memorial Sloan Kettering; New York, NY). [62] For these cells, Dulbecco's modified Eagle medium media was used instead of RPMI as per the provider's recommendation. The cultures were maintained in a humidified sterile environment in the presence of 5% CO 2 at 37 °C. Trypsin-EDTA was used to detach the cells from the cell culture dish and for seeding them at a density of ≈10 6 onto the substrate in a 60-mm culture dish. Au/SiNWs were exposed to UV light for 24 h for sterilization and dipped in cell culture media for a few minutes prior to adding cells.
For fixing the cells, substrate was washed with PBS at room temperature twice before it was immersed in formalin for 15 min. Post fixation the substrate was washed again with PBS and stored in the same solution for imaging.

SEM Sample Preparation:
The cells on the substrate were gradually dehydrated by exposing them to 50%, 60%, 70%, 80%, 90%, and 100% ethanol for 15 m each. The sample was dried in a critical point dryer before SEM imaging was performed.
Fluorescence Staining: For fluorescence staining, substrates were gently rinsed with PBS at room temperature. Calcein AM (Thermo Fisher Scientific) and Hoechst 33342 (Thermo Fisher Scientific) were diluted in PBS. The dye mixture was added to the culture dish, using just enough to cover the substrates and left at room temperature in the dark for 20 min. Post staining substrates were washed with PBS and imaged using a fluorescence microscope. The BoM cells were also stained by Calcein AM for imaging.
Imaging: Fluorescence images were obtained using a Leica microscope with a 10× objective. For imaging of cells expressing tdTomato fluorescent protein, or stained with Calcein and Hoechst, red, green, and blue filters were used, respectively. For high-resolution fluorescence imaging, a Leica SP8 confocal microscope with a 40× water immersion objective was used.
Image Analysis: ImageJ software was used for the analysis of fluorescence images. A cell counter plugin was used within the ImageJ software for counting the cells in the live/dead assay. Contact angle of a water droplet on the Au/SiNW substrate was also measured in ImageJ.
Shape Mode Analysis: For the shape mode analysis, the fluorescence images were segmented using CellProfile. [63] The segmented cells were then subjected to VAMPIRE analysis. [59,60] In this analysis, the images were rotationally aligned, and boundary coordinate points were extracted. Thereafter, principal component analysis was applied on the shape-coordinates and different shape modes were extracted. Every cell was then represented as a weighted sum of different shape modes. The cells were finally binned to a shape mode that carried the maximum weight. Cell spreading was calculated by dividing the average area of cells on Au/NWs by that on Si. 1.39-fold elongation was calculated by dividing the aspect ratio (major axis/minor axis: fitting the cell to an ellipse) of cells on Au/NWs by that on Si. Circularity was calculated using the formula (4 * pi * area)/(perimeter) 2 . The area, aspect ratio and circularity were averaged over 300 cells with shape modes 5 and 6 for Au/SiNWs substrates and shape mode 3 for Si.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.