Simultaneous Generation of Gradients with ... - ACS Publications

Jul 18, 2013 - We quantitatively investigated the response of axon growth rate and ..... We counted axons as positive response to the gradient when th...
0 downloads 0 Views 5MB Size
Download PDF
Article pubs.acs.org/ac

Simultaneous Generation of Gradients with Gradually Changed Slope in a Microfluidic Device for Quantifying Axon Response Rong-Rong Xiao, Wen-Juan Zeng, Yu-Tao Li, Wei Zou, Lei Wang, Xue-Fei Pei, Min Xie, and Wei-Hua Huang* Key Laboratory of Analytical Chemistry for Biology and Medicine, Ministry of Education, College of Chemistry and Molecular Sciences, Wuhan University, Wuhan 430072, China S Supporting Information *

ABSTRACT: Over the past decades, various microfluidic devices have been developed to investigate the role of the molecular gradient in axonal development; however, there are very few devices providing quantitative information about the response of axons to molecular gradients with different slopes. Here, we propose a novel laminar-based microfluidic device enabling simultaneous generation of multiple gradients with gradually changed slope on a single chip. This device, with two asymmetrically designed peripheral channels and opposite flow direction, could generate gradients with gradually changed slope in the center channel, enabling us to investigate simultaneously the response of axons to multiple slope gradients with the same batch of neurons. We quantitatively investigated the response of axon growth rate and growth direction to substrate-bound laminin gradients with different slopes using this single-layer chip. Furthermore, we compartmented this gradient generation chip and a cell culture chip by a porous membrane to investigate quantitatively the response of axon growth rate to the gradient of soluble factor netrin-1. The results suggested that contacting with a molecular gradient would effectively accelerate neurites growth and enhance axonal formation, and the axon guidance ratio obviously increased with the increase of gradient slope in a proper range. The capability of generating a molecular gradient with continuously variable slopes on a single chip would open up opportunities for obtaining quantitative information about the sensitivity of axons and other types of cells in response to gradients of various proteins.

A

the gradient-reading mechanism is essential to predict the response change of axons; however, there are very few studies using these devices that could provide quantitative information about the response of axons to molecular gradients with different slopes. Notably, Dertinger et al. have quantified the effect of gradients that differed in slope on axonal specification using a microchannel network similar to a “Christmas tree’’10 by changing the width of the gradient generation channel and the molecular concentration of inlets. Subsequently, several groups have proposed laminar-based gradient generators capable of changing the gradient slopes by altering the experimental conditions such as flow rate and driving pressure of the inlets as well as the physical shape of the device.24−28 However, gradients with different slopes using these devices were generated, respectively, on different chips. A test using a number of different chips may bring more experimental errors and expend large volumes of expensive biological reagents. Further, it is hard for these gradient generators to flexibly change the gradient slope in a continuously variable manner, limiting the ability to simultaneously measure the sensitivity of

xonal specification and guidance are critical processes for establishing correct neuronal circuitous pathfinding in the developing nervous system. Many surface-bound and soluble diffusion guidance factors have been identified for axonal development.1−4 These chemoattractive or chemorepulsive factors influence the growth rate and growth direction of axons by generating various gradients.1,5 Several studies have suggested that the difference of gradient slopes would influence the growth direction response of axons, and the axon growth cones have an extreme sensitivity in response to the gradients with different slopes.6,7 Over the past decades, the flexible design and fabrication of microfluidic devices (μFDs) have facilitated establishment of the microenvironments tailored to particular neuronal structures.8,9 Various microfluidic devices have been developed in investigation of neuronal functions, such as axonal guidance,10−18 axonal function,19,20 and injury and regeneration.21,22 Microfluidic devices based on laminar flow or diffusion could generate gradients with multiple substances or multiple shapes.23 These gradient generators have been used to investigate the role of the molecular gradient in the process of axon development.10,13,16 The previous findings contribute to understanding qualitatively that how various gradients direct axonal specification and growth along the appropriate pathways to their particular destinations. Quantitative understanding of © 2013 American Chemical Society

Received: May 10, 2013 Accepted: July 18, 2013 Published: July 18, 2013 7842

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

Table 1. Dimensions of Chips Used in This Work main channel width/height (μm) device type gradient generation chip H1 H2 H3 cell culture chip

peripheral channel inlet

peripheral channel outlet

center channel

interconnecting grooves length/width/height (μm)

100/50 100/50 100/50

1000/50 1000/50 1000/50 1000/100

1000/50 1000/50 1000/50

200/50/5 200/50/10 200/50/15 800/10/2

SU-8 (50 or 100) on the silicon wafer patterned with the thinner interconnecting grooves followed by exposure through a high-resolution transparency mask (25 000 dpi). Step 3: the obtained masters were cured at 160 °C for 30 min on a hot plate to further cross-link the material. Step 4: the usable masters were placed in a Petri dish, and the PDMS mixture (oligomer/curing agent mass ratio 10:1) was poured over the masters. Step 5: the masters were cured at 75 °C for 2 h to obtain a fully cross-linked PDMS replica-molded after removing bubbles. Molecular Dynamics Simulation of Gradients. A finite element analysis (COMSOL Multiphysics 3.5a) software was used to ensure that the designed microfluidic device was able to generate the expected gradients. A simpler three-channel geometry was first generated in AutoCAD software and then imported into the COMSOL interface. The simulations presented here used the Navier−Stokes equation for incompressible flow and convection−diffusion equation. As the boundary conditions, a laminar inflow was set at the inlets, zero pressure at the outlets, and no slip on the walls. Mesh density of the entire domain was set to extrafine. Stationary analysis was performed to reduce the time necessary for model calculating. Generation of Substrate-Bound Cue Gradients. The gradient generation PDMS chips were rinsed with ethanol to extract un-cross-linked PDMS monomers for 30 min and dried at 75 °C for 2 h. Both the surface of the PDMS channel and acid-cleaned glass substrate were treated with oxygen plasma (Harrick Scientific, Ossining, NY) for 1 min and bound together to form irreversibly sealed microfluidic channels. Then, the microfluidic channels were filled with PLL solution (100 μg/mL, Sigma) immediately, and the PDMS chips were sterilized under UV light for 30 min in an incubator overnight. Laminin (50 μg/mL, Sigma, L2020) and BSA (5%) solutions were perfused through the inlets of two peripheral channels, respectively, by syringe pumps (Lange, China) at a flow rate of 50 μL/h for 10 h, to establish substrate-bound laminin gradients. Then, the microfluidic channels were rinsed with PBS and stored in an incubator until neurons were plated. Immunofluorescence Staining of Substrate Gradient. As-prepared laminin gradients were fixed with 4% paraformaldehyde for 20 min. Subsequently, rabbit antilaminin antibodies (1:100, Sigma, L9393) were incubated with fixed substratebound laminin overnight at 4 °C, which was preblocked by 3% BSA in PBS at 37 °C. The microchannels were washed three times with PBS to remove surplus antilaminin and incubated with secondary antibodies labeled with FITC (antirabbit IgG, 1:100, Pierce, American) for 1 h at 37 °C. After that, the microchannels were washed three times before observation. Integration of the Multilayer Microfluidic System Generating Gradients of Diffusion Cue. All of the inlet and outlet holes needed for the lower gradient generation chip and the upper cell culture chip were punched on the upper

the axons growth cone in response to gradients with different slopes. In addition, these laminar-based generators would generate unnatural shear stresses and flush out the essential secreted factors, restricting quantification of the axonal response to gradients of diffusible cues. Here, we propose a novel laminar-based microfluidic device enabling simultaneous generation of multiple gradients with gradually changed slope on a single microfluidic chip. This device consists of three larger channels (two peripheral channels and a center channel) that are connected via thinner interconnecting grooves. The peripheral channels with two particularly asymmetric designs (gradually changed in width, opposite flow direction of two fluids) could generate gradually changed fluid pressure difference, therefore enabling generation of multiple gradients that gradually differed in slope in the center channel. Hence, it is possible to investigate simultaneously the response of axons to multiple slope gradients with the same batch of neuron cells. We investigated quantitatively the response of axon growth rate and growth direction to the substrate-bound laminin gradients with different slopes using this single-layer chip. Furthermore, we compartmented this kind of gradient generation chip and a cell culture chip by a porous membrane29,30 to investigate quantitatively the response of axon growth rate to the gradient of a diffusible axon guidance factor, netrin-1.3,31

■

EXPERIMENTAL SECTION Reagents. Poly-L-lysine (PLL), laminin, antibodies against laminin and nerve growth factor (NGF) were purchased from Sigma (St. Louis, MO), DMEM/F-12 medium for cell culture was purchased from GIBCO (U.S.A.), L-glutamine was purchased from Amresco (U.S.A.), 3′,6′-di(O-acetyl)-4′,5′bis[N,N-bis(carboxymethyl)-aminomethyl] fluorescein, tetraacetoxymethylester (calcein-AM), for cell staining, was purchased from Dojindo laboratory (U.S.A.), and semaphorin 3A and netrin-1 were purchased from the R&D Systems (U.S.A.). All other chemicals unless specified were reagent grade and used without further purification. Design and Fabrication of the Microfluidic Device. Poly(dimethylsiloxane) (PDMS) chip channels were designed in AutoCAD software and fabricated using soft lithography technology,32 all the dimensions of channels as shown in Table 1. Both of the two types of PDMS chips were composed of main channels and interconnecting grooves. The masters were fabricated by spin-coating with negative photoresist (SU-8 Microchem, U.S.A.) using a two-layer fabrication technology described previously.33 Briefly, step 1: the interconnecting grooves were patterned by spinning a thinner layer (2 or 5 or 10 or 15 μm thick, respectively) of SU-8 (2 or 5 or 15) on a cleaned silicon wafer followed by exposure through a highresolution chromium mask (defined by electron beam lithography). Step 2: the main channels were then patterned by spinning a thicker layer (50 or 100 μm thick, respectively) of 7843

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

Figure 1. Microfluidic device capable of simultaneous generation of slope-tunable gradients. (A) Schematic design of the microfluidic device; the ∗ represents the interconnecting grooves, the ∗∗ and ∗∗∗ represent the inlets and outlets of the main channels, respectively. (B) Simulation of the gradient generation principle, the arrows showing the flow direction of fluids. (C) Photograph of a microfluidic device; the fluidic channels were filled with blue and red dye tracers. (D) Simulation of different gradient distributions on devices with the different heights of interconnecting grooves (50 μm in width). (E) Normalized fluorescence intensity profiles (fluorescein isothiocynate labled dextran, 500 kDa) in the center channel; the curves with different colors represent different gradient generation regions; regions I−VIII in the three profiles correspond to the regions I−VIII in panel D.

PDMS using a biopsy punch before integrating the multilayer microfluidic device. Both the lower and upper PDMS chips were rinsed with ethanol to extract un-cross-linked PDMS monomers for 30 min and dried at 75 °C for 2 h. The integration of the lower gradient generation chip and upper cell culture chip was performed through a polyester membrane which has been silanized by APTES (3-aminopropyltriethoxysilane).34,35 The polyester membrane (24 mm diameter, 0.4 μm pore size) was cut from the traditional Transwell (product no. 3450, Corning, U.S.A.) providing supports for neuron attachment and growth. Briefly, step 1: the Transwell surface was treated with oxygen plasma for 1 min to generate hydroxyl groups followed by placing in 1% v/v aqueous APTES solution to generate surface amine for 30 min. Step 2: the Transwell was rinsed with DI water for 10 min and dried under a stream of nitrogen, and then the polyester membrane was cut from the Transwell. Step 3: the polyester membrane and upper PDMS channel surfaces were treated with oxygen plasma, respectively, for 1 min followed by bonding the surface-activated polyester membrane and upper PDMS channel together to form a “polyester membrane−upper PDMS chip” composite. Step 4: the composite and a lower PDMS channel surface were treated with oxygen plasma, respectively, followed by bonding together to form an integrated multilayer microfluidic system. The upper channels were filled with PLL (100 μg/mL), and the lower channels were filled with PBS, and the multilayer

systems were sterilized under UV light for 1 h and in an incubator overnight. Then, the upper channel was filled with laminin (10 μg/mL) for neurons differential followed by filling with cell culture medium overnight until neurons were plated. Generally, hippocampal neuron suspensions with an approximate density of 2 × 106 cells/mL were plated into the upper main channel, and all the chambers were filled with culture media.33 The lower layer gradient of netrin-1 (100 ng/mL, R&D Systems, U.S.A.) was established by the lower layer channels after culturing neurons for 3 days to test the axonal response to molecular gradient. Hippocampal Neurons Preparation and Culture. Primary hippocampal neurons were isolated from SD rat embryonic day (E)18 embryos and dissociated using the protocol previously described.36 Briefly, the hippocampus was dissected from brain, treated with 0.125% trypsinase for 15 min at 37 °C, followed by adding the culture medium with serum to terminate the action of the trypsinase enzyme. Cells with the desired density were plated in the microfluidic channel and then placed in an incubator for 20 min allowing hippocampal neurons to attach on the substrate. The neurons were maintained in primary culture medium of DMEM/F-12 (GIBCO) supplemented with 10% fetal bovine serum, 10% horse serum (Invitrogen), 2% B27 (GIBCO), 100 U/mL penicillin, streptomycin, and 3% L-glutamine. Nerve growth factor (50 ng/mL) was added to the medium for neurons 7844

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

Figure 2. Quantitative analysis of the gradients gradually changed in slope. (A) Normalized fluorescence intensity profiles of substrate-bound laminin gradient in center channels. (B) Representative fluorescence micrographs of gradient generation regions showing different gradient distributions with different interconnecting groove heights; there was a uniform concentration of molecules along the red dashed line. (C) The profile showing the distributions of gradient slopes on three chips with different interconnecting groove heights. Error bars represent standard deviations, n = 3. We divided the center channel of the chip (5 and 10 μm in height of interconnecting grooves) into two parts, named “left” and “right”(panel B). (D) Fluorescence micrograph and (E) normalized fluorescence intensity profiles of a specific gradient generation region; the height of interconnecting grooves is 10 μm.

pressure difference drove different proportions of the fluids from the two peripheral channels into the center channel via the interconnecting grooves, forming stable gradients in about 20 min. Hence, our microfluidic device was capable of generating multiple gradients that gradually differed in slope on a single chip in a stable and reproducible manner. The gradients generated by the continuous convective flow could be maintained at a steady state indefinitely (Figure 1C). Furthermore, we could modify the distribution of gradients on a single chip by altering the dimension of the interconnecting grooves. Simulation analysis demonstrated that different profiles of gradient distributions could be easily obtained by changing the height or width of the interconnecting grooves (Figure 1D and Supporting Information Figure S1, only the center channels are shown for clarity). We quantitatively compared the gradient distributions on different chips by analyzing fluorescence intensities of FITClabeled dextran (fluorescein isothiocynate labeled dextran, Mw = 500 kDa) (Figure 1E) and rhodamine B labeled dextran (Mw = 70 kDa) (Supporting Information Figure S2). The results clearly indicated the ability of chips with different heights of interconnecting grooves in generating different profiles of gradient distributions. Generation and Quantitative Analysis of SubstrateBound Gradients. Laminin is a critical local substrate-bound molecule in directing axonal development of hippocampal neurons.2 We qualified the laminin gradients by immuno-

survival. Finally, the chip was placed in an incubator at 37 °C, 5% CO2 for neurons culture. Fluorescent Imaging and Data Analysis. Fluorescence micrographs of gradient generation and hippocampal neurons labeled with calcein-AM were captured by using an inverted fluorescence microscope (AxioObserver Z1 fluorescent microscope with camera and incubation system, ZEISS, Germany). Normalized fluorescence intensity profiles of gradient generation in the center channel were analyzed by using Image J (MacBiophotonics) software. An Image J plug-in named Neuron J was used to trace and quantify the elongated micrograph of axons. SPSS 16.0 (SPSS Inc.) was used to perform statistical data analysis.

■

RESULTS AND DISCUSSION

Simultaneous Generation of Multiple Gradients on a Microfluidic Device. Previous studies have demonstrated that variable combinations of different fluid pressures of the inlets could generate gradients with different slopes.26,27 The laminarbased microfluidic device proposed in this work contains three larger channels (two peripheral channels and a central channel, and the dimensions of the channels are described in Table 1) that are connected via thinner interconnecting grooves (Figure 1A). The peripheral channels with two particularly asymmetric designs (gradually changed in width, opposite flow direction of two fluids) generated gradually changed difference in fluid pressure in the center channel (Figure 1B). Gradually changed 7845

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

Figure 3. Quantitative analysis of axonal response to laminin gradients differed in slope. (A) The hippocampal neurons labeled with calcein-AM contacting with three different substrate environments after 48 h in culture (a, uniform BSA; b, laminin gradient; c, uniform laminin). (B) Definition of the angle α of axon growth according the gradient generation line. The sectors that formed the basis for the classification in positive and negative responding neurons are indicated. (C) Statistical data of the length of axons contacting with three different substrate environments after 48 h in culture on a single chip (the height of interconnecting grooves is 15 μm) (one-way ANOVA; **p < 0.05; n.s., not significant). Error bars represent standard deviations, n = 36, 54, and 30, respectively; similar results were obtained in other chips. (D) The representatively statistical guidance ratios of hippocampal neuronal response to laminin gradients with different slopes after 24 h in culture: region I (s = 0.14%), region IV(s = 1.31%), and region VIII (s = 0.00%) in the right part, and region VII (s = 1.69%), region VIII (s = 1.73%) in the left part. Error bars represent standard deviations; data of every slope were collected from six replicated samples.

fluorescence staining of the substrate-bound laminin using antilaminin antibodies (Figure 2A). Gradients gradually changed in slope could be generated from region I to region VIII on a single chip, and there is a uniform concentration distribution of laminin along the red dashed line representing the gradient generation dividing line (Figure 2B). We defined the slope “s” as the fractional change in concentration across 10 μm.7 We divided the center channel of the chip (5 and 10 μm in height of interconnecting grooves) into two parts named “left” and “right” (Figure 2B) to analyze the slope change, considering there were different distribution tendencies of gradient slopes in the two parts. The gradient slopes on the chip with interconnecting grooves height of 5 μm were very low in the left part and had a gradually increased distribution from region I (s = 0.48%) to region VIII (s = 2.44%) in the right part. The gradient slopes on the chip with interconnecting grooves height of 10 μm had a gradually increased distribution from region I (s = 0.02%) to region VIII (s = 1.73%) in the left part, and a parabolic-like distribution in the right part (the slope is 0.14% on region I, the maximum slope is 1.31% on region IV, and the minimum slope is 0.00% on region VIII). The gradient slopes on the chip with the interconnecting grooves height of 15 μm had also a parabolic-like distribution (the minimum slope is 0.00% on regions I/II/VIII, and the maximum slope is 0.79% on region

V). Obviously, there would generate laminin gradients with gradually changed slope on single microfluidic chips (Figure 2C). A representative fluorescence micrograph (Figure 2D) and normalized fluorescence intensity profiles (Figure 2E) in a specific gradient generation region on a chip have been shown. The results demonstrated that this device could generate gradients with continuously variable slopes in a narrowed region, indicating its potential to measure the sensitivity7 of axonal response to various factors gradients with slightly changed slope. Quantitative Analysis of Axonal Response to Substrate-Bound Laminin Gradients. Previous reports suggested that the chemoattractive or chemorepulsive factors distribute in a graded way and influence the growth rate and growth direction of axon by generating various gradients.1,5,10,37 To study how the substrate-bound gradient influence the axon growth rate, we quantitatively compared the length of axons cultured on laminin gradient substrates generated by microchips with interconnecting grooves height of 15 μm. The gradient slopes on these substrates had a parabolic-like distribution, producing three typical different substrate regions (uniform BSA, laminin gradient, and uniform laminin) on a single chip. We plated hippocampal neuron clusters17,38,39 on the chip to alleviate the neurons connecting with each other after 48 h in 7846

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

Figure 4. Multilayer microfluidic system enables studying axon guidance by diffusible cues under a microenvironment free of unnatural shear stresses. (A) Scheme of the multilayer microfluidic system fabricated with an upper PDMS layer and a lower PDMS layer sandwiching a porous polyester membrane; the ∗ represents the upper microgrooves, and the ∗∗ represents the cell culture channel. (B) Integration of the multilayer microfluidic system; all of the inlet and outlet holes needed by the lower gradient generation chip and the upper cell culture chip were punched on the upper PDMS. (C) Normalized fluorescence intensity profile (rhodamine B labled dextran, 70 kDa) of the lower PDMS center channel. (D) Fluorescence and phase contrast microscopic merged image of a specific gradient generation region; the ∗ represent the upper microgrooves, and the lower fluidic channels were perfused with DI water and rhodamine B labeled dextran.

uniform BSA substrate (one-way ANOVA; n.s., not significant). Taken together, the data demonstrate that laminin distributing in a graded way as a local presentation of growth-promoting substrate-bound molecule can effectively accelerate the axons growth during the process of neuronal polarity. In addition to growth rate, we found that the axons were asymmetrically distributed on the substrate with laminin gradient (Figure 3A, part b; statistical data is shown in Supporting Information Figure S4). Further, we investigated quantitatively the response of the axon growth direction to multiple laminin gradients that differed in slope using microchips with the interconnecting grooves height of 10 μm. Compared with the substrates obtained from the device with 15 μm height interconnecting grooves, this kind of device could generate a substrate gradient with wider slope distribution (from 0.00% to 1.73%) that was more suitable for investigating the response of axon growth direction to gradients in a larger range of slopes. We counted axons growth as positive response to the laminin gradients when the angle (α) between the line connecting the cell body and the axonal tip lay between −80° < α < 100° (the angle was divided as previously described), and counted axons growth in the opposite direction as negative response. We defined the guidance ratio to quantify the biased growth direction response as the number of positive axons minus the number of negative axons, normalized by the total number of axons.7 The guidance ratio was counted when one of the neurites grew rapidly to become an axon after 24 h in culture.40

culture (Figure 3A). We counted axons as positive response to the gradient when the angle (α) between the line connecting the cell body and the axonal tip lay between −65° < α < 115°, which is the direction of increasing concentration of laminin corresponding with the gradient generation line. We counted axons growing along the direction of decreasing concentration of laminin as negative response (Figure 3B). We made statistics on the dendritic length of isolated neurons (n = 100) after 48 h in culture40 (Supporting Information Figure S3). Since most dendrites (67%) are shorter than 40 μm, and almost all dendrites are shorter than 70 μm, therefore, neurites longer than 80 μm labeled with calcein-AM (the immunostaining did not identify neuronal axon and dendrite if neurons were cultivated for less than 3 days41) were included in the statistical analysis of axonal length. The statistical analysis (Figure 3C) demonstrated that there was no significant difference in the length of axons contacting with a uniform substrate (uniform BSA or uniform laminin) between positive and negative response. However, when the neurons contacted with a laminin gradient, their axons growing up gradient (positive) were apparently longer than axons growing down gradient (negative) [one-way analysis of variance (ANOVA); **p < 0.05; n.s., not significant]. Axons contacting with a uniformly higher laminin substrate were not significantly longer than the axons growing along the positive of a laminin gradient (one-way ANOVA; n.s., not significant). Moreover, axons growing along the negative of a laminin gradient were also not significantly longer than the axons contacting with a 7847

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

Figure 5. Response of axons to netrin-1 and statistical analysis of data. (A) The upper chip used to compartment neuronal somas and axons: axons extend into the different microgrooves after several days in culture; the arrows represent fluid flow direction in the lower chip. (B) Fluorescence micrograph of hippocampal neurons labeled with calcein-AM after 4 days in culture in the upper channel with constant flow of culture medium in the lower channel. (C) Fluorescence micrograph of hippocampal neurons labeled with calcein-AM after 10 days in culture in the upper channel with constant flow of netrin-1 in the lower channel; the flow rate is 5 μL/h (a, without netrin-1; b, netrin-1 gradient; c, uniform netrin-1). (D) Statistics of the axonal length in response to netrin-1. Error bars represent standard deviations, n = 34, 30, 24, respectively (one-way ANOVA; **p < 0.001; n.s., not significant).

Sema3A and netrin-1 are diffusible factors which have important functions in maintaining hippocampus formation.3 Previous findings demonstrated that netrin-1 is involved in promoting outgrowth of axons and required to attract efferent axons during the development of hippocampus,3,31 while Sema3A (one of the class 3 semaphorins) has a strongly repulsive guidance effect.3,42 Several studies have also demonstrated that axon growth rate was controlled by a combination of the gradient parameters. It has been discussed previously that axons sometimes grow further when growing up a gradient than growing down a gradient or on a uniform concentration of proteins.37,39,43 We have quantitatively investigated the effect of substratebound gradient on axon growth using a single-layer chip. However, a laminar-based gradient generation device with constant flow would generate unnatural shear stresses which markedly cause damage to the neurons, therefore limiting our ability to quantify the response of axon growth to the gradient of diffusible cues. In this study, we compartmented the gradient generation channels and cell culture channels by a porous polyester membrane29,30 to alleviate the damage from fluidic shear stress. The multilayer microfluidic system is composed of an upper compartmented neuron culture layer and a lower

We analyzed the isolated neurons whose polarities were established obviously. Statistics of the data were made as a function of gradient slopes from region I to region VIII of the right part and from region VII to region VIII of the left part where the gradient slopes were altered remarkably. Our data showed that the axon guidance ratio displayed obvious difference when neurons were cultured on the substrate-bound gradient with different slopes. There was a very low guidance ratio when the axons contacted with a uniform BSA or laminin substrate (regions I, VI−VIII of the right part, Figure 3D), in spite of the absolute concentration of substrate-bound laminin. Neurons cultured on the region with relatively steeper gradient (Figure 2C) have higher axonal guidance ratio (regions III−V of the right part and regions VII−VIII of the left part, Figure 3D). The results demonstrate that the axon growth cone is very sensitive in sensing the information of gradients with different slopes, indicating that gradient slope would be a key factor in determining axon guidance for axon pathfinding decisions. Multilayer Microfluidic System Generating Gradient of Diffusible Cues. In addition to substrate-bound gradients, there are secreted signaling proteins forming diffusion cue gradients in vivo that influence neuronal development.1 Both 7848

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

■

Article

CONCLUSIONS In this study, we introduced a novel microfluidic device with unique designs capable of generating multiple gradients with gradually changed slope for quantifying the axon response to substrate-bound laminin gradients. Furthermore, we compartmented the gradient generation chip and a cell culture chip by a porous membrane to investigate quantitatively the response of axon growth rate to the gradient of diffusion factor netrin-1. Our experimental data indicated that surface-bound laminin gradient could effectively accelerate neurite growth and enhance axonal formation, and the axon guidance ratio obviously increased with the increase of gradient slope in a proper range. Furthermore, we found axons grow further when growing up a gradient than when growing down a gradient or on a uniform concentration of soluble factor netrin-1. Our findings represent direct evidence that how the laminin influences the neuronal polarity and netrin-1 influences the neurites growth. Taken together, this work demonstrated the capability of the microfluidic device we developed in generation of multiple gradients simultaneously in a single chip to quantify the axonal response to substrate-bound or diffusion cues. In addition, our device has the potential to obtain quantitative information about the sensitivity of axons and other types of cells in response to gradients of various proteins.

gradient generation layer sandwiching a porous polyester membrane (Figure 4A). Neurons were plated in the cell culture channel (1 mm wide, 100 μm high), and the axons would extend into the microgrooves (10 μm wide, 2 μm high) after several days in culture. This compartmented culture method could alleviate the interfering to axons response from other axons. The sandwiching layer is a transparent, thin (10 μm) membrane with 400 nm pores permitting diffusion of biomolecules from the lower channels to the upper channels. This multilayer microfluidic system (Figure 4B) could bear a flow rate up to 500 μL/h without any leakage (Supporting Information Figure S5). The normalized fluorescence intensity profiles (rhodamine B labeled dextran, Mw = 70 kDa) in the center channel generated by the lower layer have been shown as Figure 4C. Three typically different microenvironment (regions I−II, without netrin-1; regions III−VI, netrin-1 gradient; regions VII−VIII, uniform netrin-1) of diffusion cue were generated on a single multilayer chip. The fluorescence and phase contrast microscopic merged image of a specific gradient generation region is shown as Figure 4D. Response of Axons to Neurotrophins and Statistical Analysis of Data. The axons extending into different microgrooves in the upper layer would lie in different microenvironments when the lower layer was perfused with two different molecular fluids (Figure 5A). First, we plated neurons at a lower density to test if the fluid in the lower layer affects the development state of neurons in the upper channel. The fluorescence micrograph of hippocampal neurons labeled with calcein-AM showed a well-grown state (Figure 5B), indicating the sandwich design could effectively alleviate the damage to neurons from fluidic shear stress. Then, we investigated the response of hippocampal axons to a Sema3A gradient, and there were very few axons extending in the microgrooves with a gradient of Sema3A (Supporting Information Figure S6) after being cultured for 14 days, indicating the neuron axons in the upper layer could respond to the molecules diffused from the lower layer channels. Finally, we studied the response of hippocampal axons growth rate to a netrin-1 gradient by quantitatively comparing the extending length of axons. Fluorescence micrographs of hippocampal neurons in the upper layer after 10 days in culture with a constant flow of netrin-1 in the lower layer for 7 days are shown in Figure 5C. The axonal response was statistically significant. The axons extended straight up a netrin-1 gradient further than they extended into microgrooves with a uniformly higher concentration of netrin-1 or without any netrin-1 (one-way ANOVA; **p < 0.001). The axons extending into microgrooves with a uniformly higher concentration of netrin-1 displayed no significant difference in the length of axons contacting without any netrin-1 (one-way ANOVA; n.s., not significant) (Figure 5D). The results suggested that contacting with a gradient of netrin-1 induces a rapid increase of the neurites growth; these findings are also in agreement with the growth-rate model proposed by Mortimer et al.37,44 They indicated that the growth-rate modulation as an alternative mechanism required by axons may operate together with the immediate and biased mechanism14,45 in response to a gradient signal. Our experimental data using the microfluidic system provide the direct and quantitative information of the effects of netrin-1 gradient on axon growth, which is important to understand the growth-rate modulate mechanisms.

■

ASSOCIATED CONTENT

S Supporting Information *

Figures S1−S6. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 86 2768752149. Fax: 86 2768754067. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Basic Research Program of China (973 Program, no. 2012CB720603), the National Natural Science Foundation of China (nos. 31070995 and 81071227), the Program for New Century Excellent Talents in University (NCET-10-0611), and the Program for Changjiang Scholars and Innovative Research Team in University (IRT1030).

■

REFERENCES

(1) Tessier-Lavigne, M.; Goodman, C. S. Science 1996, 274, 1123− 1133. (2) Esch, T.; Lemmon, V.; Banker, G. J. Neurosci. 1999, 19, 6417− 6426. (3) Skutella, T.; Nitsch, R. Trends Neurosci. 2001, 24, 107−113. (4) Yaron, A.; Zheng, B. Dev. Neurobiol. 2007, 67, 1216−1231. (5) Mortimer, D.; Fothergill, T.; Pujic, Z.; Richards, L. J.; Goodhill, G. J. Trends Neurosci. 2008, 31, 90−98. (6) Isbister, C. M.; Mackenzie, P. J.; To, K. C. W.; O’Connor, T. P. J. Neurosci. 2003, 23, 193−202. (7) Rosoff, W. J.; Urbach, J. S.; Esrick, M. A.; McAllister, R. G.; Richards, L. J.; Goodhill, G. J. Nat. Neurosci. 2004, 7, 678−682. (8) Millet, L. J.; Gillette, M. U. Trends Neurosci. 2012, 35, 752−761. (9) Soe, A. K.; Nahavandi, S.; Khoshmanesh, K. Biosens. Bioelectron. 2012, 35, 1−13. 7849

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850

Analytical Chemistry

Article

(10) Dertinger, S. K. W.; Jiang, X.; Li, Z.; Murthy, V. N.; Whitesides, G. M. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 12542−12547. (11) von Philipsborn, A. C.; Lang, S.; Loeschinger, J.; Bernard, A.; David, C.; Lehnert, D.; Bonhoeffer, F.; Bastmeyer, M. Development 2006, 133, 2487−2495. (12) Lang, S.; von Philipsborn, A. C.; Bernard, A.; Bonhoeffer, F.; Bastmeyer, M. Anal. Bioanal. Chem. 2008, 390, 809−816. (13) Wang, C. J.; Li, X.; Lin, B.; Shim, S.; Ming, G. L.; Levchenko, A. Lab Chip 2008, 8, 227−237. (14) Mai, J.; Fok, L.; Gao, H.; Zhang, X.; Poo, M. M. J. Neurosci. 2009, 29, 7450−7458. (15) Bhattacharjee, N.; Li, N.; Keenan, T. M.; Folch, A. Integr. Biol. 2010, 2, 669−679. (16) Millet, L. J.; Stewart, M. E.; Nuzzo, R. G.; Gillette, M. U. Lab Chip 2010, 10, 1525−1535. (17) Shi, P.; Nedelec, S.; Wichterle, H.; Kam, L. C. Lab Chip 2010, 10, 1005−1010. (18) Millet, L. J.; Stewart, M. E.; Sweedler, J. V.; Nuzzo, R. G.; Gillette, M. U. Lab Chip 2007, 7, 987−994. (19) Cohen, M. S.; Ghosh, A. K.; Kim, H. J.; Jeon, N. L.; Jaffrey, S. R. Chem. Biol. 2012, 19, 179−187. (20) Kim, H. J.; Park, J. W.; Byun, J. H.; Poon, W. W.; Cotman, C. W.; Fowlkes, C. C.; Jeon, N. L. ACS Chem. Neurosci. 2012, 3, 433− 438. (21) Hosmane, S.; Tegenge, M. A.; Rajbhandari, L.; Uapinyoying, P.; Kumar, N. G.; Thakor, N.; Venkatesan, A. J. Neurosci. 2012, 32, 7745− 7757. (22) Li, L.; Ren, L.; Liu, W.; Wang, J. C.; Wang, Y.; Tu, Q.; Xu, J.; Liu, R.; Zhang, Y.; Yuan, M. S.; Li, T.; Wang, J. Anal. Chem. 2012, 84, 6444−6453. (23) Kim, S.; Kim, H. J.; Jeon, N. L. Integr. Biol. 2010, 2, 584−603. (24) Holden, M. A.; Kumar, S.; Castellana, E. T.; Beskok, A.; Cremer, P. S. Sens. Actuators, B 2003, 92, 199−207. (25) Irimia, D.; Geba, D. A.; Toner, M. Anal. Chem. 2006, 78, 3472− 3477. (26) Keenan, T. M.; Hsu, C. H.; Folch, A. Appl. Phys. Lett. 2006, 89, 1141031−1141033. (27) Sip, C. G.; Bhattacharjee, N.; Folch, A. Biomicrofluidics 2011, 5, 022210. (28) Ren, L.; Liu, W.; Wang, Y.; Wang, J. C.; Tu, Q.; Xu, J.; Liu, R.; Shen, S. F.; Wang, J. Anal. Chem. 2013, 85, 235−244. (29) Huh, D.; Fujioka, H.; Tung, Y. C.; Futai, N.; Paine, R.; Grotberg, J. B.; Takayama, S. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 18886−18891. (30) Song, J. W.; Cavnar, S. P.; Walker, A. C.; Luker, K. E.; Gupta, M.; Tung, Y. C.; Luker, G. D.; Takayama, S. PLoS One 2009, 4, e5756. (31) Barallobre, M. J.; Del Río, J. A.; Alcántara, S.; Borrell, V.; Aguado, F.; Ruiz, M.; Carmona, M. A.; Martin, M.; Fabre, M.; Yuste, R.; Tessier-Lavigne, M.; Soriano, E. Development 2000, 127, 4797− 4810. (32) Xia, Y.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37, 550− 575. (33) Park, J. W.; Vahidi, B.; Taylor, A. M.; Rhee, S. W.; Jeon, N. L. Nat. Protoc. 2006, 1, 2128−2136. (34) Sunkara, V.; Park, D. K.; Hwang, H.; Chantiwas, R.; Soper, S. A.; Cho, Y. K. Lab Chip 2010, 11, 962−965. (35) Tang, L.; Lee, N. Y. Lab Chip 2010, 10, 1274−1280. (36) Kaech, S.; Banker, G. Nat. Protoc. 2006, 1, 2406−2415. (37) Mortimer, D.; Pujic, Z.; Vaughan, T.; Thompson, A. W.; Feldner, J.; Vetter, I.; Goodhill, G. J. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 5202−5207. (38) Sorkin, R.; Gabay, T.; Blinder, P.; Baranes, D.; Ben-Jacob, E.; Hanein, Y. J. Neural Eng. 2006, 3, 95−101. (39) Bagnard, D.; Thomasset, N.; Lohrum, M.; Püschel, A. W.; Bolz, J. J. Neurosci. 2000, 20, 1030−1035. (40) Tahirovic, S.; Bradke, F. Cold Spring Harbor Perspect. Biol. 2009, 1, a001644. (41) Cάceres, A.; Banker, G. A.; Binder, L. J. Neurosci. 1986, 6, 714− 722.

(42) Good, P. F.; Alapat, D.; Hsu, A.; Chu, C.; Perl, D.; Wen, X.; Burstein, D. E.; Kohtz, D. S. J. Neurochem. 2004, 91, 716−736. (43) Hansen, M. J.; Dallal, G. E.; Flanagan, J. G. Neuron 2004, 42, 717−730. (44) Mortimer, D.; Feldner, J.; Vaughan, T.; Vetter, I.; Pujic, Z.; Rosoff, W. J.; Burrage, K.; Dayan, P.; Richards, L. J.; Goodhill, G. J. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 10296−10301. (45) Zheng, J. Q.; Felder, M.; Connor, J. A.; Poo, M. M. Nature 1994, 368, 140−144.

7850

dx.doi.org/10.1021/ac4022055 | Anal. Chem. 2013, 85, 7842−7850