Biomedical Microdevices

, Volume 12, Issue 5, pp 935–948

Precise cell patterning using cytophobic self-assembled monolayer deposited on top of semi-transparent gold


  • Gaoshan Jing
    • Sherman Fairchild Center, Department of Electrical & Computer EngineeringLehigh University
  • Susan F. Perry
    • Department of Chemical EngineeringLehigh University
    • Sherman Fairchild Center, Department of Electrical & Computer EngineeringLehigh University

DOI: 10.1007/s10544-010-9448-8

Cite this article as:
Jing, G., Perry, S.F. & Tatic-Lucic, S. Biomed Microdevices (2010) 12: 935. doi:10.1007/s10544-010-9448-8


This paper reports a simple and effective method for cell patterning by using a self-assembled monolayer (SAM)-treated glass surface which is surrounded by semi-transparent gold coated with another type of SAM. Specifically, a hydrophobic SAM, derived from 1-hexadecanethiol (HDT), was coated on the gold surface to prevent cell growth, and a hydrophilic SAM, derived from 3-trimethoxysilyl propyl-diethylenetriamine (DETA), was coated on the exposed glass surface to promote cell growth. The capabilities of this technique are as follows: 1) single-cell resolution, 2) easy alignment of the cell patterns to the structures already existing on the substrate, 3) visualization and verification of the predefined cytophobic/cytophilic pattern prior to cell growth, and 4) convenient monitoring cell growth at the same location for an extended long term period of time. Whereas a number of earlier techniques have demonstrated the single cell resolution, or visualization and verification of the cytophobic/cytophilic patterns prior to cell growth, we believe that our technique is unique in possessing all of these beneficial qualities at the same time. The distinguishing characteristic of our technique is, however, that the use of semi-transparent Cr/Au film allows for convenient brightfield pattern visualization and offers an advantage over previously developed methods which require fluorescent imaging. We have successfully demonstrated the patterning of four different kinds of cells using this technique: immortalized mouse hypothalamic neurons (GT1-7), mouse osteoblast cells (MC3T3), mouse fibroblast cells (NIH3T3) and primary rat hippocampal neurons. This study was performed with a specific ultimate application—the creation of a multi electrode array (MEA) with predefined localization of cell bodies on top of the electrodes, as well as predefined patterns for cell extensions to grow in between the electrodes. With that goal in mind, we have also determined critical parameters for patterning of each of these cell types, such as the minimum size of a cell-adherent island for exclusively anchoring one cell or two cells, as well as the width of the cytophilic pathway between two islands that enables cell extensions to grow, while preventing the anchoring of the cell bodies. Additionally, we have provided statistical analysis of the occupancy for various sizes and shape of cell-anchoring islands. As demonstrated here, we have developed a novel and reliable cell patterning technique, which can be utilized in various applications, such as biosensors or tissue engineering.


Cell patterningSelf-assembled monolayers (SAMs)Single cell patterning

1 Introduction

Patterning mammalian cells has become essential for exploring the interactions between cells and their extracellular environment. These interactions play a critical role in many cellular functions, such as apoptosis (programmed cell death) (Chen et al. 1997), migration (Jiang et al. 2005) and differentiation (McBeath et al. 2004; Vunjak-Novakovic 2008). It has also been established that the pattern dimensions, themselves, can have an effect on cell attachment (Krsko et al. 2009; Tzvetkova-Chevolleau et al. 2009), growth (Kam et al. 1999), morphology (Hu et al. 2010), and long-term viability (Thissen et al. 2006). In order to pattern cells in specific geometries, a common approach is to create cytophilic (cell-attractive) regions separated by cytophobic (cell-repulsive) regions, so that cells are confined to defined regions.

Many patterning techniques have been developed, combining microfabrication, chemical surface modification and material science to pattern cells and proteins, such as microcontact printing (μCP) (Falconnet et al. 2006; Kane et al. 1999; Xia and Whitesides 1998), dip-pen nanolithography (DPN) (Piner et al. 1999; Salaita et al. 2007), inkjet printing for cell patterning (Roth et al. 2004), laser scanning lithography (Miller et al. 2006), parylene-based dry lift-off technique (Ilic and Craighead 2000) and microfluidic patterning (Chiu et al. 2000). For any cell patterning techniques based on chemical surface modification, it is crucial to generate cytophobic regions, successfully, in order to suppress the non-specific adhesion between the surface and cells. At present, microcontact printing (μCP), using polydimethylsiloxane (PDMS) stamps, is a commonly used technique to selectively pattern cytophobic and cytophilic molecules on a micrometer scale (Whitesides et al. 2001). One widely used cytophobic molecule is self assembled monolayer (SAM) derived from polyethylene glycol (PEG) (Smith et al. 2004). Due to its ability to resist protein adhesion, surfaces coated with PEG SAM resist cell growth and are thus cytophobic. Many different PEG surface-immobilization strategies have been successfully applied to pattern substrates. For instance, PDMS stamps have been used to imprint patterns of alkanethiols (which are protein adhesive) to boost cell growth. Then, thiolated PEG (protein repulsive) is used to backfill the open space, between the patterned regions, to prevent cell growth (Dike et al. 1999). Thus, cell-adherent regions, with defined shape and size, separated by cell repulsive islands, are created. Some researchers have also used PDMS stamps to print poly-lysine and then backfilled PEG silane (a protein-repulsive SAM) directly on a glass surface (Branch et al. 2000). Poly-lysine is a cell-attractive protein, and suitable for the growth of certain cell types, such as hippocampal neurons (Hickman and Stenger 1994).

In contrast to the majority of published studies in which cytophilic patterns are formed by printing, followed by generating cytophobic regions through backfilling, an alternative approach, negative patterning (Falconnet et al. 2006), is to create cytophobic patterns, first, followed by the generation of cytophilic regions. One example of this negative patterning is to stamp cytophobic octadecyltricholorosilane (OTS) SAM onto silicon wafers to create cell-free areas, while a backfill with 3-trimethoxysilyl propyl-diethylenetriamine (DETA) SAM is subsequently used to form the cytophilic patches (Falconnet et al. 2006; Kam et al. 1999; St. John et al. 1997).

For most of these techniques, the main challenges are 1) difficult alignment to features and/or structures already existing on the substrate (such as electrodes for multi electrode arrays (MEAs), for example), and 2) convenient cytophillic/cytophobic pattern visualization on the substrate, prior to cell culturing. Easy alignment is usually a trademark of lithographically-based processes (Falconnet et al. 2006; He et al. 2004). It is also noteworthy, that some MEA chips, capable of defining neural patterns, have been designed by combining a traditional MEA chip with the μCP technique, using a modified optical aligner (Chang et al. 2003; James et al. 2004), a technique that requires equipment modifications. The technique proposed in this study maintains the characteristic of simple alignment of the pattern to features existing on the substrate, without the need for extensive equipment modifications.

We have previously addressed the issues of alignment, and the simple, initial visualization of the cytophobic/cytophilic areas by developing a patterning method based on SAM-treated cytophobic gold and SAM-treated cytophilic silicon oxide (Jing et al. 2007). However, this method does not resolve another significant challenge, the formation of a pattern for cell growth which is easily visualized on a transparent substrate prior to cell culturing, to make it suitable for use in brightfield microscopy, with an inverted microscope, the preferred equipment used by the biomedical community for observing cells in vitro. The technique described here addresses this limitation.

Another important performance trait we sought to achieve, by our new technique, was single cell resolution. Single cell resolution patterning has been achieved by several patterning techniques, such as μCP (Chen et al. 1997; McBeath et al. 2004), using microfabricated magnetic microposts (Sniadecki et al. 2007), dielectrophoresis (Mittal et al. 2007), microneedles (Shibata et al. 2009) and acoustic picoliter droplet ejection (Demirci and Montesano 2007). However, this particular feature was rarely combined with the other performance characteristics (easy alignment and pattern visualization prior to cell culturing, as well as suitability for inverted microscope visualization). Yet, it is clear that the ability to achieve these diverse performance traits through a single technique would be extremely beneficial. Indeed, our technical performance requirements were developed in conjunction with the ultimate application of this cell patterning technique to the creation of an MEA in which there is single neuron/electrode correspondence, and in which the pathways for neuronal connection (ie neurite growth) between the electrodes is precisely defined. Our envisioned MEA (described in details in the later section) does not have capability to attract neurons to the electrodes, which was earlier executed by several investigators (Gray et al. 2004; Prasad et al. 2004), but relies, instead, on the favorable anchoring of a single neuron on top of the insulation layer surrounding an electrode opening. Therefore, we found it necessary to develop a cell patterning technique which satisfies all of the above requirements at the same time.

In this paper, we report the development of a technique which satisfies our requirements by allowing easy alignment to existing features on the substrate, while providing single cell resolution. Its additional, distinguishing benefit is that the use of semi-transparent Cr/Au film allows for convenient brightfield pattern visualization and offers an advantage over previously developed methods which require fluorescent imaging. By this method, a hydrophobic self-assembled monolayer, derived from 1-hexadecanethiol (HDT), was coated, specifically, on semi-transparent gold patterns fabricated on a glass substrate, to prevent cell growth (cytophobic). Then, a hydrophilic SAM, derived from 3-trimethoxysilyl propyl-diethylenetriamine (DETA), which serves as a chemical replacement for a cell attractive protein of poly-L-lysine (Hickman and Stenger 1994), was coated on the open glass surface surrounding the gold patterns to promote cell growth (cytophilic). This technique reduces the complexity of the cell patterning process using conventional methods. Namely, in other techniques using hydrophilic PEG SAM (protein repulsive and cytophobic) and hydrophobic alkanethiols SAM (protein adhesive and cytophilic) to pattern cells (Kleinfeld et al. 1988; Lamb et al. 2008), an additional step of soaking the substrate in protein to promote cell growth is required, prior to cell culture, for successfully patterning cells; this additional step is not required in our method. We have successfully demonstrated the patterning of four different kinds of cells on the cytophilic glass surface. Using this technique, we have accomplished all of the desired goals: single cell resolution, the alignment of the pattern to features and structures that already exist on the substrate, application of this technique to a variety of cell types and easy visualization of cell patterns under an inverted microscope.

2 Materials and methods

The sequence of individual steps describing our technique is presented in Fig. 1. First, we fabricated a gold pattern, on top of a glass wafer, by image reversal processing. A hydrophobic self-assembled monolayer, derived from 1-hexadecanethiol (HDT), was coated on the gold surface to prevent cell growth, and a hydrophilic SAM, derived from 3-trimethoxysilyl propyl-diethylenetriamine (DETA), was then coated on the exposed glass surface to promote cell growth. A variety of mammalian cell types then were cultured, in vitro, on the chip. Patterned cells, with or without subsequent fluorescent staining, were visualized by an inverted microscope. The details of individual steps are presented below.
Fig. 1

Scheme of the cell patterning technique

3 Substrate preparation

The semi-transparent Cr/Au film was patterned as follows: photoresist (Shipley 1818, Shipley Company Inc., Marlboro, MA) was spin-coated at 3000 rpm on a four inch diameter soda lime glass wafer (Mark Optics, CA) for 30 s with a ramp rate of 10,000 rpm/s to obtain a 2.2 μm thick film. After soft baking at 90°C for 60 s, a contact aligner EV620 (EV Group Inc., Albany, NY) was used to expose the photoresist with a dose of 80.4 mJ/cm2. The photomask used was made by a Heidelberg DWL66 direct laser writer (Heidelberg Instruments, Heidelberg, Germany) with a 10 mm working distance lens. The resolution of the photomask created in this fashion was 2 microns, which led to a corresponding 2 µm resolution of the resulting gold patterns.

The exposure step was followed by baking in ammonia for 90 min in an image reversal oven (Yield Engineering Systems, Inc, Livermore, CA). Then, the whole substrate was flood-exposed for 60 s by a contact aligner EV620 (EV Group Inc., Albany, NY). After developing in MF321 developer (Shipley Company, Inc., Marlboro, MA) for 90 s by a Hamatech-Steag Wafer Processor (Hamatech AG, Sternenfels, Germany), oxygen plasma was used to remove the photoresist residue. Next, 6 nm of chromium was deposited on the surface as an adhesion layer, followed by 16 nm of gold in an E-beam evaporator (CVC SC4500, CVC Product Inc., NY). After a lift-off process was performed using a Microposit Remover 1165 (Shipley Company, Inc., Marlboro, MA), a 2.2 microns thick layer of photoresist (Shipley 1818, Shipley Company, Inc., Marlboro, MA) was coated on the glass surface at 3000 rpm to prevent scratching during the following dicing process. Finally, the whole wafer was cut into segments (2.2 cm by 2.3 cm) using a K&S 7100 dicing saw (Kulicke & Soffa, PA). The die size was chosen so that it fit in a well of a six well culture plate (BD, NJ) during subsequent cell culturing.

Cytophobic and cytophilic areas were created on the samples by deposition of self-assembled monolayers (SAM). Before coating SAMs, the surface of the substrate was cleaned using acetone for 18 min and isopropyl alcohol (IPA) for 18 min in a bath sonicator (FS60H, Fisher Scientific, GA). This step served to remove the photoresist that had been used as a protective layer during dicing. Then, the substrate surface was cleaned with piranha solution (80% H2SO4 and 20% H2O2) for 3 h. Next, the substrate was immersed in a 10 mM solution of 1-hexadecanethiol (HDT) (Gelest Inc., PA) in ethanol for 18 h. This step was used to create hydrophobic regions on the gold surface of the substrate. Next, the substrate was immersed in a 23 mM solution of (3-trimethoxysilyl propyl)-diethylenetriamine (DETA) (Gelest Inc., PA) in methanol for 1 h. This step created hydrophilic regions on the glass region where the hydrophobic SAM had been excluded. After soaking, the substrate was carefully rinsed with ethanol and deionized water. The substrates were dried with a stream of nitrogen to remove any excess solvent.

4 Cell culturing

Four different kinds of mammalian cells (immortalized neurons, primary neurons, osteoblasts, and fibroblasts) were used to assess the effect of the SAM patterns on cell positioning and growth. Immortalized mouse hypothalamic neurons (GT1-7) cells were maintained in 25 cm2 flasks (Corning, GA) at 37°C in an incubator with humidified 8% CO2. Cells were grown in Dulbecco’s modification of Eagles medium (DMEM) (Invitrogen, CA) supplemented with 1 mM sodium pyruvate, 10 mM sodium bicarbonate, 2 mM L-glutamine, 10 mM Hepes buffer, 1% penicillin and streptomycin, 2.5 µg/ml fungizone and 10% fetal bovine serum (FBS, Invitrogen, CA) (Liposits et al. 1991). Immortalized mouse osteoblast cells, MC3T3 cells, were grown in minimum essential medium, α modification (α–MEM) supplemented with 10 mM sodium bicarbonate, 2 mM L-glutamine, 1% penicillin and streptomycin, and 10% newborn bovine serum (Invitrogen, CA). Immortalized, embryonic mouse fibroblasts, NIH3T3 cells, were grown in DMEM medium with 20 mM sodium bicarbonate, 2 mM L-glutamine, 1% penicillin and streptomycin, 2.5 µg/ml fungizone and 10% newborn bovine serum (Invitrogen, CA). The latter two cell lines were maintained in 25 cm2 flasks (Corning, GA) at 37°C in an incubator with humidified 5% CO2.

For patterning experiments, all cell lines were harvested by trypsinization, in 0.125% (w/v) trypsin solution at 37°C for 5 min. Following trypsinization, the cells were pelleted by centrifugation at 750 rpm (118 × g), for 5 min and re-suspended in the culture medium. The cells were plated at a density of 2 × 104 cells/ml in a Falcon six-well plate containing the cleaned and patterned substrates, with alternating hydrophobic (HDT) and hydrophilic (DETA) SAM regions. This density of cells was chosen based on the fact that the low density can lead to prolonged experimental duration, while higher density would result in the overgrowth of cells in short period of time. Substrates with semi-transparent gold patterns, without SAM coating, served as the controls. The cells were maintained under standard conditions until cell patterns could be observed by optical microscopy.

In addition to the immortalized cell lines, primary rat hippocampal neurons also were used to assess the effect of the SAM patterns on cell positioning and growth. The culturing procedure for primary hippocampal neurons was quite different from that of the cell lines and is described in details elsewhere (Brewer et al. 1993). Briefly, hippocampal neurons were extracted from the hippocampal tissue of embryonic day 18 rats and delivered in Hibernate E solution from Brainbits LLC (Brainbits, IL). The hippocampal tissues were dissociated at 30°C for 30 min in 2 mg/ml papain (Worthington, NJ) in hibernate E-Calcium solution (Brainbits, IL), then triturated and plated at a density of 160 cells/mm2 in a 35 mm2 Petri dish containing the cleaned and HDT/DETA patterned substrates. Plated neurons were maintained in neurobasal medium (Invitrogen, CA) containing B27 supplement (Invitrogen, CA), at 37°C in an incubator with humidified 8% CO2. After 4 days, one half of the medium was changed with Neurobasal/B27 medium containing 0.5 mM glutamine (Invitrogen, CA).

5 Cell imaging by epifluorescent/phase contrast microscopy

The patterned cells were rinsed with phosphate buffered saline (PBS) solution, and placed in serum-free medium (containing all of the components of the culture medium mentioned above, except FBS). A fluorescent stain, calcein AM (Invitrogen, CA), was added to the medium to a concentration of 2 µM and the cells were incubated at room temperature for 20 min. The stained cells were visualized with an Olympus IX70 inverted microscope (Olympus Corp., PA). Phase contrast images were also taken by the same inverted microscope. The scale bars for all relevant figures were calibrated with a stage micrometer (OB-M 1/100, Olympus Inc, NY), and scale bars were added with the SPOT imaging software (Diagnostic Instruments, MI). In some cases, the patterned cells were rinsed with phosphate buffered saline (PBS) solution for 5 min and preserved in 10% neutral buffered formalin solution for later analysis.

6 Cell imaging by scanning electron microscopy (SEM)

Scanning electron microscopy (SEM) was used to further analyze the cell patterns from the above experiments. Prior to performing scanning electron microscopy, fixation and dehydration were performed to maintain cell morphology. The fixation and dehydration were based on a standard procedure (Gray and Fedun 2006). Briefly, the patterned cells were rinsed with phosphate buffered saline (PBS) solution for 5 min and fixed in 10% neutral buffered formalin solution for 4 hrs. After fixation, the substrate (with patterned cells) was rinsed in DI water and dehydrated in a series of ethanol solutions (from 35% to 100%) followed by soaking in 100% hexamethyldisilazane (HMDS) (VWR, PA) for 5 min. After air drying for 30 min, the substrate was coated with 5.5 nm of gold layer by sputtering to facilitate SEM imaging.

7 Results and discussion

A single die with a semi-transparent gold pattern is shown in Fig. 2(a). In this design, there are over 500 different shapes, with variable dimensions, for investigating cell growth and connectivity. Three characteristic mask patterns (designed using L-EDIT software, Tanner Research Inc., Monrovia, CA) are illustrated in Figs. 2(b), (c) and (d). Using this mask, we obtained accurate information of the manner in which various shapes, with different dimensions, affect the cell growth. More importantly, these designs enabled the determination of the appropriate pattern size necessary to accommodate a single neuron cell body, rather than multiple cells, as well as the appropriate width of an interconnecting path necessary for facilitating neural connection (via neurites), while not allowing neural cell bodies to anchor.
Fig. 2

(a) A 2.2 cm by 2.3 cm single die with semi-transparent gold pattern (dark gray) on transparent glass substrate. (b) 16 circular islands with size “D” are linked by the interconnecting path with width “W” and length “L”. D changes from 10 μm to 50 μm. W changes from 0 μm to 50 μm (and W is no greater than D). L changes from 50 μm to 200 μm. (c) 16 square islands with size “D” are linked by the interconnecting path with width “W” and length “L”, which has the same dimension configuration as (b). (d) Lines 1 mm in length, with widths “D”, separated by 150 μm are designed. D changes from 2 μm to 100 μm. Pattern number is denoted by #

One main advantage of the technique described here, compared to other patterning methods, is the set of beneficial characteristics (single cell resolution, easy visualization of the intended pattern prior to cell culturing, possibility of alignment with pre-existing substrate features or structures) combined with its unique capability (to the best of our knowledge) to combine brightfield pattern visualization, thereby providing the option for additional fluorescent channels for subsequent analyses. Namely, because of the thinness of metal layer (22 nm thick), the gold layer is semi-transparent and a brightfield, inverted microscope can be utilized both to visualize the patterns prior to cell culture, and to monitor cell growth after cell culturing has been initiated. Since our gold layer is treated with a cytophobic, self-assembled monolayer, cell attachment to this region should be minimized, while the majority of cell attachment and growth should occur on the cytophilic transparent glass substrate. This technique permits us to obtain better quality images, at a higher magnification, compared to our previous cell patterning technique with an opaque substrate (Jing et al. 2007), which required visualization of cells using a compound microscope. We experimented on four kinds of cells: 1) immortalized, mouse hypothalamic neurons (GT1-7 cells), 2) MC3T3 osteoblasts, 3) NIH3T3 fibroblasts and 4) primary rat hippocampal neurons. All cell types were anchored and confined to the glass patterns coated with the hydrophilic SAM derived from DETA, whereas they did not anchor on hydrophobic gold surfaces coated with the SAM derived from 1-hexadecanethiol (HDT).

In this work, we have also determined the optimal dimensions of patterns on the substrate for four types of cells (see above) that would secure successful anchoring of individual cells on cytophilic islands (having in mind our MEA application), as well as the minimum cytophilic linewidth separating cytophobic regions which will assure cell separation when cells are fully confluent. This latter dimension was important, particularly for the GT1-7 immortalized cell line, as cells continue to divide and eventually will overgrow the entire substrate, given sufficient time. Our findings are described below.

8 Precise cell patterning with single-cell resolution

One goal of this experimental work was to obtain high quality imaging of single cell patterning, which was possible due to the development of our technique on semi-transparent substrates, thereby enabling the use of inverted microscopy and high magnification lenses.

The cell type whose patterning we have investigated in the most detail is the immortalized mouse hypothalamic neuronal (GT1-7) cell line. Therefore, first we will discuss their normal morphology and the morphology we have observed when cells were patterned on substrates of different shapes and dimensions.

Normally after trypsinization, cells retract their processes, round up and maintain a spherical morphology while suspended in the growth medium. The typical diameter of a suspended GT1-7 cell is approximately 10 μm (Jing et al. 2007). In contrast, when GT1-7 cells attach to a cell culture flask, they regain their typical neuronal morphology, with neural extensions as shown in Fig. 3(a). The largest dimension of some attached cells is 100 μm, as shown in Fig. 3(a). GT1-7 cells exhibit the same neuronal morphology whether they are cultured on an uncoated glass substrate or on a gold surface, as shown in Fig. 3(b). It should be noted that it is possible to observe cells which spread their cell bodies across both the untreated glass and gold regions because of the thinness of the metal layers, since they are both cytophilic and the heights of the underlying structures are very small. This is in contrast to what is sometimes observed from other techniques, in which surface topology, a result of thicker layers, can influence the growth and orientation of neurons in vitro, causing cells to grow along the edges of the structures (Craighead et al. 2001; Craighead et al. 1998).
Fig. 3

Phase contrast images of GT1-7 cells on (a) an untreated cell culture flask substrate and (b) on a glass substrate (light gray region) with semi-transparent gold patterns (deep gray region). (c) Two GT1-7 cells positioned on a cytophilic, DETA-treated 30 μm glass circle surrounded by HDT-treated gold. (d) One GT1-7 cell positioned on a cytophilic, DETA-treated, 20 μm glass circle surrounded by HDT-treated gold. (e) Four individual GT1-7 cells anchored on four (out of the possible sixteen) 10 μm wide squares of exposed glass. The surrounding surface was cytophobic SAM-treated gold. (f) SEM image of a GT1-7 cell anchored on a 10 μm wide square of exposed glass (deep gray region), surrounded by cytophobic SAM-treated gold (light gray region). (g) SEM image of two GT1-7 cells anchored on a 30 μm wide square of cytophilic glass (deep gray region), surrounded by cytophobic SAM-treated gold (light gray region). All images were obtained after 2 DIV

While GT1-7 cells could grow indiscriminately on uncoated gold or glass substrates, they were accurately confined to the predefined patterns generated by SAM-treated gold and glass substrate, as shown in Fig. 3(c). Our results show that DETA-treated features with a diameter of 30 μm, or greater, surrounded by HDT-treated gold supported multiple GT1-7 cells (Fig. 3(c)), whereas single cells were found located in circles with a diameter of 20 microns or less (Fig. 3(d)). Because of the thinness of the patterned metal layer, epifluorescence microscopy can be performed on the cells patterned using our technique, as shown in Fig. 3(e). Four GT1-7 cells were anchored on four (out of the possible sixteen) 10 μm square anchoring sites of exposed glass, surrounded by cytophobic SAM-treated gold. When we observe features as small as 10 micron by 10 micron squares, as in Fig. 3(e), it is difficult to obtain a sharp optical image, which indicates the cell’s reaction to pattern boundaries. However, through SEM, it was possible to analyze, in great detail, the manner in which cells react to such small features. From Fig. 3(f), we can see a GT1-7 cell which was anchored in a 10 μm wide square of exposed glass surrounded by cytophobic SAM-treated gold. In contrast, two GT1-7 cells shared a 30 μm wide square of exposed glass surrounded by cytophobic SAM-treated gold as shown in Fig. 3(g). As opposed to cells in a bigger island, as shown in Fig. 3(g), the morphological appearance of cell in the small island, as shown in Fig. 3(f) indicated a more rounded topological structure because of the size of cytophilic square is as small as the diameter of a GT1-7 cell in suspension.

Additionally, we have investigated the effect of the feature size on the attachment of neurons. Whereas this type of investigations is not new (Makohliso et al. 1998), we have focused on the patterns and types of cells that are of particular importance in our upcoming research. Specifically, we have investigated the relationship between the percentage of cell occupancy (PCO) and different sizes and shapes of cytophilic islands as shown in Fig. 4(a) and (b). The percentages are calculated for cell occupancy within each 16 islands without interconnecting path (shown in Fig. 2(b) and (c)), which include four different categories, 1) ≥ three cells, 2) two cells, 3) one cell or 4) no cells per island. Five different island arrays were examined for each of the categories. For instance, if there are six islands with more than three cells, five islands with two cells, three islands with one cell and two islands without cells, the PCOs for ≥ three cells, two cells, one cell and no cells categories are 37.5%, 31.25%, 18.75% and 12.5% respectively. We found that the trend of cell occupancy is similar for circular islands and square islands (see Fig. 4(a) and (b)). When the size of the cytophilic island is greater than 40 µm, more than three cells can be accommodated within one island. When the size of island is 30 µm, at most two cells can be accommodated in one island. For islands with the size smaller than 20 µm, only one cell can be accommodated in one island. This statement holds true for cell patterns after four days of cell growth which is the maximum time we observed cell patterns. We have never observed more than one cell patterned within the islands with characteristic size smaller than 20 µm. We can deduce that such a small area cannot support the attachment and growth of more than a single cell, nor induce proliferation of that cell. Because our technique is a passive way to pattern cells (cells, themselves, “find” the appropriate sites in which to survive), occasionally vacant islands can be observed even for islands as big as 50 µm.
Fig. 4

(a) Relationship between the percentage of cell occupancy (PCO) for GT1-7 cells and different sizes of circular cytophilic islands, from 10 microns to 50 microns in diameter. (b) Relationship between PCO for GT1-7 cells and different sizes of square cytophilic islands, from 10 microns to 50 microns in size

It was also observed that GT1-7 cells demonstrated different morphologies on cytophilic lines with different linewidths. When the linewidth was greater than 10 μm (for instance, a 40 μm line in Fig. 5(a)), GT1-7 cells were anchored inside of the cytophilic lines without crossing over to cytophobic gold regions, and were spread in a manner which is typically observed on a conventional culturing flask (see Fig. 3(a)). When the linewidth was 2 μm, as in Fig. 5(b), GT1-7 cell bodies still managed to anchor on such thin cytophilic lines and extend neurites along these lines even though the linewidth was much smaller than the diameter of GT1-7 cells in suspension. However, in contrast to cells on wider lines, the morphological appearance of cells on these small lines indicated a more rounded cell body with a more elongated extension, presumably to minimize contact with cytophobic surfaces.
Fig. 5

GT1-7 cells on DETA-treated cytophilic glass lines (light gray region) with widths of (a) 40 μm and (b) 2 μm, separated by 150 μm wide gaps of HDT-treated cytophobic gold (dark gray region) after 2 DIV

Currently, the smallest linewidth obtainable from our mask is 2 μm, so it is unknown whether GT1-7 cell bodies or their neurites can anchor on lines with a linewidth smaller than 2 μm. From our unpublished experimental results, using another cell patterning method with the help of an advanced lithographic technique, projection lithography, we have observed GT1-7 neurons and MC3T3 osteoblasts can anchor their bodies on thin lines as small as 0.5 μm.

In addition to the GT1-7 neuronal cell line, two more cell lines, MC3T3 osteoblasts and NIH3T3 fibroblasts, and one type of primary cells (rat hippocampal neurons), were also successfully patterned using this technique, as shown in Fig. 6. The relationships between the typical cell size, in suspension, and the minimum anchoring island dimension for a single cell, minimum island dimension for anchoring of two cells and minimum linewidth for cell anchoring are shown in Table 1. We found that cells from each of the cell lines can anchor their cell bodies on islands and lines smaller than their typical diameters, in suspension. In contrast, we observed the primary hippocampal neurons can only anchor their cell bodies on islands and lines that are equivalent or larger than the size of the cells in suspension. The successful demonstration of patterning other types of cells makes this technique a convenient and accurate method for tissue engineering.
Fig. 6

(a) MC3T3 preosteoblast cells (b) NIH3T3 fibroblast cells and (c) rat hippocampal neurons, on DETA-treated, cytophilic glass lines, 75 microns width (light gray region) separated by 150 μm wide gaps of HDT-treated gold (dark gray region) after 2 DIV

Table 1

Relationship between the dimensions of cytophilic regions to different cell types

Cell type

Observed range of cell diameters in suspension [µm]

Island dimension suitable for anchoring a single cell [µm]

Island dimension suitable for anchoring two cells [µm]

Island dimension suitable for anchoring more than three cells [µm]

Linewidth of a cytophilic interconnect preventing anchoring of a cell body [µm]

GT1-7 cells






MC3T3 cells






NIH3T3 cells






Hippocampal neurons






9 Effective separation of cell-growing regions

Another critical dimension we investigated was the appropriate cytophobic linewidth to prevent cell overgrowth after an extended duration of cell culturing (this portion of our research was focused on GT1-7 cells only). This is an important parameter for the proper design of biosensor devices which implement this cell patterning strategy for use with cell lines, as cell lines are often not contact-inhibited and continue to divide. After four days of culture, we have found that the minimum linewidth of a hydrophobic region necessary to effectively prevent GT1-7 cells from expanding beyond the designed cell growth region is 15 μm, as shown in Fig. 7(a). When the linewidth is smaller than 10 μm, as shown in Fig. 7(b), the cells will grow across the lines and become confluent over both cytophobic and cytophilic regions. When the linewidth is larger than 15 μm, we have not observed that cells will overgrow the hydrophobic region during the four days cell culturing period.
Fig. 7

GT1-7 cells on DETA-treated glass surfaces after 4 DIV. In both images, 40 μm gold square islands are linked by the interconnecting path with length of 100 μm and width of (a) 15 μm and (b) 10 μm

10 Long-term monitoring cell growth

Compared to other patterning methods (Falconnet et al. 2006), one main advantage of the technique described here is that specific patterns can be located conveniently, quickly, and unambiguously, due to clear, visible labels. This allows one to track the differences in cell growth at same location at different times. To do so, low initial cell concentration, 2 × 104 cells/ml, was applied in our experiment. In Fig. 8(a), GT1-7 cells growing on DETA-treated glass lines with widths of 75 μm separated by 150 μm wide gaps of HDT-treated gold were visible after two days of culture. Because of the short growing time, GT1-7 cells were distributed sparsely on DETA-treated glass regions. With the help of the gold reference label, it is easy to determine the location of the cells. In Fig. 8(b), the same region has been localized after four days of cell growth. Compared to Fig. 8(a), GT1-7 cells have covered the entire area of DETA-treated glass and the cell patterns are accurately matched to the designed shapes. For primary rat neurons, cell patterns can be maintained as long as four weeks because these cells do not divide and the cell number will not increase.
Fig. 8

(a) GT1-7 cells on DETA-treated glass lines (light gray region) with widths of 75 μm separated by 150 μm wide gaps of HDT-treated gold (dark gray region) after (a) 2 DIV and (b) 4 DIV

11 Patterned neural networks

Highly organized neural networks were obtained using this patterning technique. Achieving a pattern, as illustrated in Figs. 2(b) and (c), where single cells would be anchored on islands connected with pathways suitable only for neurite growth, but not for cell body attachment, was the ultimate goal of this research, because in this fashion, one-to-one, neuron-to-electrode correspondence would be achieved and it would be far easier to trace the origin of the electrical signals. In Fig. 9(a), GT1-7 cell bodies, occupying most of the 50 μm islands, form an organized network of neurites growing along 5 μm wide, cytophilic pathways. As was predicted from our earlier results (see Fig. 3, Table 1), more than one cell can occupy an island with characteristic dimension greater than 20 μm, as in Fig. 9(a) where an island is 50 μm square.
Fig. 9

(a) GT1-7 cell bodies occupying most of the 50 micron islands, with a network of neurites growing along 5 micron-wide, cytophilic pathways. (b) A single 1.5 cm by 1.5 cm MEA die. (c) Patterned GT1-7 neurons cultured on an MEA chip. (d) A typical spontaneous spike train of a GT1-7 neuron. (e) A single typical spontaneous spike of a GT1-7 neuron. (f) Patterned hippocampal neurons cultured on a MEA chip. (g) A typical spontaneous spike train of hippocampal neurons, obtained from the MEA electrode indicated by the white circle in (f). (h) Two typical spontaneous spikes from hippocampal neurons obtained by the MEA electrode indicated by the white circle in (f)

The ability to create organized cell networks using this technique, coupled with the convenient visualization of gold patterns under an optical microscope (making it simple to align the gold patterns to existing structures) can be exploited in the fabrication of multilayer biosensor devices. Using the cell culturing method described above, we have developed a multi electrode array (MEA) chip, shown in Fig. 9(b), to detect the neural activity of a patterned network. Here, patterned cytophobic gold, which defines the cell pattern after the cell culturing, is accurately aligned to the previously existing structures (electrodes) on the wafer, due to the inherent characteristics of the photolithographic process on which their creation is based.

A patterned, GT1-7 neural network was successfully cultured on the MEA chip shown in Fig. 9(c). The detailed microfabrication process has been described in detail elsewhere (Jing et al. 2009). Briefly, a thin film of Cr/Au/Cr is patterned on top of a glass wafer, which forms electrodes and interconnects. Then, PECVD oxide is deposited on top of the substrate, followed by the creation of openings (for vias and bond pads), which are etched into this layer using plasma etching. In the next step, a thin layer of Cr/Au is deposited, and afterwards patterned, on top of the PECVD oxide layer. This metal layer (when subsequently treated with cytophobic SAMs) facilitates the desired anchoring of the neurons on the cytophilic PECVD oxide sites (at predefined positions) and offers guidance to neurites in-between the neurons. The cell culturing process was identical to the previously described procedure.

One can see that this MEA supports anchoring of multiple neurons on cell anchoring sites. Using this device, spontaneous electrical signals with amplitude varying from 80∼180 µV were detected from the unpatterned GT1-7 neurons. A spontaneous spike train of a GT1-7 neuron is shown in Fig. 9(d) and a typical spontaneous spike of a GT1-7 neuron with the amplitude about 180 µV is shown in Fig. 9(e). Hippocampal neurons can also be patterned on the MEA chip surface as shown in Fig. 9(f). Similar neural signals with amplitude of approximately 80 µV were also obtained from the patterned hippocampal neural network shown as in Fig. 9(g) and (h). Neural signals with similar amplitudes and shapes have been acquired multiple times using our MEA system for GT1-7 neurons and rat hippocampal neurons. We are currently working on creating an MEA in which single neurons are anchored on cell permissive islands with electrodes underneath (to record electrical activity), and connecting lines to support neurite (but not cell body) growth.

12 Conclusion

We report a novel technique to pattern cells using alternating areas of cytophilic SAM-treated glass and cytophobic SAM-treated semi-transparent gold. Implementation of this method enabled us to achieve single-cell patterning resolution for four mammalian cell types: three types of immortalized cell lines (GT1-7 cells, MC3T3 cells and NIH3T3 cells) and a primary cell (rat hippocampal neurons). We have also demonstrated that this method (due to its photolithographic nature) offers the capability of easy alignment to the pre-existing features on the underlying substrate. Additionally, we were able to accomplish simple brightfield visualization of the cytophobic/cytophilic pattern prior to cell culturing.

Whereas some of these characteristics have been demonstrated before, this technique, which combines these characteristics, is unique. Also, compared to other techniques, one distinguishing advantage of our method is that it allows for convenient brightfield pattern visualization, which enables an additional fluorescent channel for subsequent analysis, as fluorescent imaging is not necessary to visualize the pattern, itself. This technique enables the design and fabrication of complex biosensor systems, which is demonstrated on our novel MEA chip which supports patterned neural networks.


This work was supported by National Science Foundation (CAREER grant ECS-0448886 and NER grant BES-0608742), as well as Pennsylvania Infrastructure Technology Alliance (Grant PA-DCED C000016682). The microfabrication part of this work was performed at the Cornell NanoScale Facility (CNF), a member of the National Nanotechnology Infrastructure Network, which is supported by the National Science Foundation (Grant ECS 03-35765). We also want to thank to Prof. Gregory Ferguson and his graduate student Joseph Labukas for useful advice.

Copyright information

© Springer Science+Business Media, LLC 2010