- Email: [email protected]
Modulation of neural network activity by patterning
Biosensors & Bioelectronics 16 (2001) 527– 533 www.elsevier.com/locate/bios
Modulation of neural network activity by patterning John C. Chang a,1, Gregory J. Brewer b,2, Bruce C. Wheeler a,* a
Department of Electrical and Computer Engineering, Beckman Institute, Uni6ersity of Illinois at Urbana-Champaign, 405 N. Mathews A6enue, Urbana, IL 61801, USA b Departments of Neurology and Medical Microbiology and Immunology, Southern Illinois Uni6ersity School of Medicine, PO Box 19626, Springfield, IL 62794 -9626, USA
Abstract Using neuronal cultures on microelectrode arrays, researchers have shown that recordable electrical activity can be influenced by chemicals in the culture environment, thus demonstrating potential applicability to biosensors or drug screening. Since practical success requires the design of robust networks with repeatable, reliable responses understanding the sources of variation is important. In this report, we used lithographic technologies to confine neurons to highly defined patterns (40 mm wide stripes); in turn these patterns gave us a measure of control over the local density of neurons (100– 500 cells/mm2). We found that the apparent electrical activity of the network, as measured by the fraction of electrodes from which signals were recordable, increases 8–10-fold with greater local density. Also, average-firing rates of the active neurons increased 3 – 5-fold. We conclude that patterned networks offer one means of controlling and enhancing the responsiveness of cultured neural networks. © 2001 Elsevier Science B.V. All rights reserved. Keywords: Patterning; Recording; Hippocampal; Network activity
1. Introduction Recent research has shown that neurons can be grown in culture so as to respond in a dose dependent manner to chemicals by changing their firing pattern (Gross et al., 1997; Morefield et al., 2000). This observation has led to the suggestion that neural networks can serve as chemical sensors. To fully exploit the concept of a neuron-based biosensor, however, the variables controlling the sensor behavior must be thoroughly explored. Variables such as cell-type, cell density, cell plasticity, and cell interaction should be reasonably controlled to manipulate important sensor properties, such as robustness and repeatability. To control robustness, one could alter the cell density as it is known that hippocampal neurons survive better at * Corresponding author. Tel.: + 1-217-333-3236; fax: +1-217-2445180. E-mail addresses: [email protected] (J.C. Chang), [email protected] (G.J. Brewer), [email protected] (B.C. Wheeler). 1 Tel.: + 1-217-244-2692. 2 Tel.: + 1-217-785-5230; fax: +1-217-524-3227.
high densities, because they secrete a greater amount of glutamine (Watanabe et al., 1998). Alternatively, glia can modulate the network baseline activity through the glutamine that they supply to the neurons (Huelsmann et al., 2000), or different cell-types may be selected to respond better to a specific stimulus (Morefield et al., 2000). However, methods for controlling the sensitivity and repeatability of the sensor seem less clear, because sensitivity to chemicals changes with the area of growth (Gross et al., 1997), spontaneous activity pattern changes with network size (Gross, 1994), and response changes with exposure history of the sensor (Gross et al., 1997). While these results are strong indications that neuronal cultures can serve as biosensors, they also underscore the need for further understanding of the underlying biological mechanisms, e.g., development and plasticity, in order to create robust, reliable and repeatable sensors. In order to further our understanding of the dependence of neural activity on experimental characteristics, we are exploring the potential for the use of networks grown in patterns. Previously, we have shown that patterned hippocampal neurons develop electrical activity (Chang et al., 2000). In this report, we have com-
0956-5663/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 5 6 - 5 6 6 3 ( 0 1 ) 0 0 1 6 6 - X
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
528
pared the differences in network activity between random and patterned neural networks (parallel lines) grown in a serum- and glia-free culture. Our data indicate that the neurons establish connections with one another and that patterning enhances the network activity.
2. Methods
2.1. Array preparation The electrode arrays used are obtained from MultiChannel Systems (Reutlingen, Germany) consisting of 60 titanium nitride electrodes of 10 or 20 mm diameter, occupying all but the corner locations of an 8-by-8 square grid with vertical and horizontal separation of 200 mm. The electrodes are insulated with 0.5 mm silicon nitride and have impedance ranging between 200 and 400 kV at 1 kHz. New arrays were patterned (as described below) without additional preparation. Recy-
cled arrays were first scrubbed with commercial dishwashing detergent (Dawn) using a rubber policeman, immersed overnight in a Alconox solution ( 1.5 g/200 ml water), and rinsed the following day with running deionized water (DI) for 30+ min before patterning. Fig. 1 shows the wide-lined pattern, overlaid on an image of an electrode array. The pattern consists of alternating 40 mm wide lines of poly-D-lysine (PDL; light regions) and 60 mm wide lines of presumed silicon nitride (dark region). The patterning process is slightly modified from the photoresist process reported by Chang et al. (2000). Prior to photoresist application, the arrays are sonicated in acetone for 4 min and rinsed under running DI for 15 min or more. The photoresist (AZ5214, Hoechst-Celanese) is spin-coated (5.5 krpm) for 30 s, baked at 110 °C for 90–120 s, exposed to UV through the mask (filtered at 320 nm, 300 mW/cm2) for 15 s, and developed for 30–60 s. The arrays are then cleaned by oxygen plasma (300 W, 500 mTorr) for 1 min and adsorbed with PDL (100 mg/ml DI) for 3 h. After aspirating the PDL solution from the arrays, the
Fig. 1. An array overlaid with patterns (white-40 mm foreground, dark-60 or 200 mm background) added by Photoshop.
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
arrays are sonicated in acetone for 4 min to remove the photoresist and disinfected in 70% ethanol in DI before storage in a 10 cm poly-styrene petri dish (generic). An autoclaved PDMS (Sylgard 184, Dow-Corning) culture ring (1.5-cm diameter) is rinsed in 75% ethanol in DI, aspirated dry, and placed onto the array, all performed within a sterile laminar-flow hood before the culture is started.
2.2. Cell culture The cultures are prepared as described by Brewer et al. (1993). Briefly, the embryonic hippocampal cells are harvested from E18 embryos and mechanically dissociated. The cells are plated at 200 cells/mm2 in Neurobasal/B27 medium (Life Technologies) containing 25 mM glutamate and 0.5 mM glutamine. Cultures are incubated at 37 °C in 9% O2 and 5% CO2 (Forma Scientific). Each week, one-half of the medium is changed with Neurobasal/B27 medium containing 0.5 mM glutamine.
2.3. Acti6ity recording For recording, the arrays are removed from the incubator, imaged (Nikon Diaphot), seated in the MEA-1060 Amplifier (Gain 1000X; MultiChannel Systems), and heated to 37 °C by the accessory heater from MultiChannel Systems. Activity was monitored in the original culture medium or in Hanks balanced salt solution with 1.8 mM CaCl2 and without MgCl2. The output of the amplifier is fed into a multi-channel spike detector (Plexon Instruments Inc., Dallas, TX) with its associated software. Analog output is also fed into a VCR to sample the active channels four channels at a time. At the time of recording, the number of active electrodes is identified, and the analysis of the recordings is performed using an offline-sorting program. Recorded action potentials are digitized at 40 kHz for 1.5– 3 min intervals using the Digitizer (Plexon Instruments Inc.) and sorted by extracting the signals at 1.5 times the maximum background noise followed by Kmeans clustering of the resulting principal component clusters to identify the individual units. The unit spike counts are then converted into frequency and averaged to obtain a mean and standard deviation.
2.4. Data analysis For both cultures, effective cell densities are determined from the phase-contrast image of the array center (: 1.5×1.5 mm2 centering on the electrodes) obtained at the time of recording. Each cell is identified as a dark spot with a bright halo. Cell clusters whose constituents cannot be identified were ignored, because the cluster center tends to be silent, while clusters whose
529
constituting cells can be identified were included. The effective on pattern density is then calculated as (cells on foreground stripe/total foreground area). If an image cannot be obtained or is unclear, the culture is ignored to avoid ambiguous results. Statistical analysis of the activity is performed using Student’s t-test for two independent groups with unknown population variances. The groups are further assumed to be normally distributed with unequal variances. The significance level was set at 5%.
3. Results
3.1. Random network acti6ity le6el Despite apparent good culture growth and a moderate density of neurites traversing over the electrodes (Fig. 2(a)), spontaneous activity was rare in random cultures plated at 200 cells/mm2. The age of these networks at recording ranged from 8 to 40 days in vitro (DIV). On average, the cultures yielded 0.891.93 (N= 10 cultures) spontaneously active electrodes, ranging from zero (predominant) to six active electrodes, with effective cell density of 4739 121 cells/mm2. The inactive cultures also resisted chemical stimulation by magnesium free balanced salt solution containing calcium without magnesium. This result is in stark contrast with that of patterned neuronal networks.
3.2. Patterned network acti6ity le6el The networks of patterned neurons initially developed by growing on the patterned PDL lines and extended neurites along the interface of PDL and silicon nitride surfaces. As the cultures age, they lose the pattern (occurring after culture day 8) due to the extension of neurites to neighboring PDL lines. Recordings were made between 11 and 51 DIV in the growth medium, and our observations suggested that effective cell density dictated the level of detectable activity in these patterned cultures. The cultures (Fig. 2(b)) yielded 1097.8 (N= 8) active electrodes, and the activity was completely blocked by 1.8 mM of Mg2 + and partially returned upon washout of the medium (Fig. 2(c)), indicating the role of synaptic transmission in the spontaneous activities of the network. Furthermore, the activity showed no correlation with culture age. The difference in the activity levels (between random and patterned cultures), as measured by number of active electrodes, was statistically significant (P= 0.005; t=3.6; d.f.= 16) with patterned cultures having greater level of activity. For the active units detected, the average-firing rate was 0.439 0.35 Hz in random culture (10 units sorted from seven active electrodes; max-
530
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
Fig. 2. Phase-contrast images of the cultures at the time of recording. (a) Random culture recorded around 14 DIV. (b) Patterned culture recorded on 27 DIV. (c) Sample of the recorded action potentials (1.25 s): top trace is before MgCl2 addition, middle is with added 1.8 mM MgCl2, and the bottom is after washout.
imum firing rate of 1.3 Hz) and 2.139 3.0 Hz in patterned culture (32 units sorted from 24 active electrodes; over one-third of the units had firing rates in excess of 1.3 Hz). This difference is statistically significant (P=0.005; t= 3.1, d.f.=32). The results are summarized in Table 1. We further hypothesized that the activity levels of the random and patterned cultures are influenced by the effective cell density and tested this by plotting the activity level (Fig. 3) versus the cell density. The effective cell density of the patterned cultures ranged from 130 to 500 cells/mm2, and the number of active electrodes increased rapidly with cell density above 250 cells/mm2 (Fig. 3). This suggests that patterning may enhance the network activity level by increasing the effective cell density.
4. Discussion Using a serum-free hippocampal neuronal culture, we have demonstrated an enhancement of both network and cellular activity level over that of a culture of randomly distributed neurons by patterning the neurons into parallel lines. Two possible reasons for this enhancement are discussed here: geometrical effects and the concentration of synaptic inputs. Indeed, the patterns here yield a 2–3-fold increase in local cell density, as compared to density averaged over the array area, so that should lead to an increased likelihood that activity would be sensed. However, in random cultures there are a substantial number of neurons with neurites and somata near (e.g., within 15 mm) electrodes (Fig. 2(a)), suggesting that proximity to
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
531
Fig. 2. (Continued)
the electrodes was insufficient to enhance observed network activity. Indeed some relatively quiet random cultures had densities greater than the local densities of more active patterned networks. Instead, it seems that basal network activity determines the observed activity level. Our data show that patterning results in enhanced network activity (percent active electrodes) as a func-
tion of increasing the effective cell density, and that the increase is particularly striking and supra linear for densities above 250 cells/mm2. It is as if a threshold is passed by the higher concentration established by patterning. Therefore, we believe that other mechanisms are involved in modulating the activity of the neuronal network.
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
532 Table 1 Summary of data and statistics
Random Patterned a b
Active electrodes (m9 S.D.)a
Unit firing frequency (Hz; m 9S.D.)b
0.8 91.9 10 97.8
0.439 0.35 2.13 9 3.0
P = 0.005, t=3.6, d.f. = 16. P = 0.005, t= 3.1, d.f. = 32.
Both cultured and in vivo hippocampal pyramidal neurons develop highly branched dendritic trees and an axon that extends for hundreds of microns with numerous branches. These neurons have a large number of inputs, implying that firing rates would be low in culture conditions in which the number of inputs is low (as is true in most dissociated cultures, such as reported here). Hence, cell-types naturally receiving a large number of inputs in vivo may require synchronized depolarization of their inputs in vitro in order to reach the firing threshold. However, there is evidence that neurons adjust their firing threshold and sensitivity according to the level of input they receive (Bear, 1996). An interpretation of our data consistent with these ideas is that, beginning at 250 cells/mm2, our cultured hippocampal neurons are receiving the minimum needed input for readily measurable spike activity. The level of network activity should be dependent on cell-type. For instance, bipolar neurons from the dorsal root ganglion fire readily both in vivo (Berne and Levy, 1993) and in random, serum-free cultures (Manos et al., 1999). In contrast, hippocampal pyramidal neurons fire at low rates in vitro and, at least in our work, at even lower rates in serum-free, glia-free culture.
Our data suggest that once active, the firing rate of cells is independent of density, at least over the densities tested. While there is a difference between the low-density random networks and the higher density patterned networks, there is no trend within the patterned network pool. Despite good growth in some random cultures (474 cells/mm2), there was little activity. Further experiments are needed to understand how synapse density influences cellular and network activity. A thorough understanding of network modulation by patterning should be important in the design of cellbased biosensors. By controlling the location and the activity level of the network, we may enhance the detectability of the sensor response and modulate the sensitivity of the sensor to certain chemicals of interest. In addition, with a mixture of cells, we may construct network sensors that respond to multiple chemicals in uniquely different fashions. Consequently, we believe that the understanding of network modulation by patterned growth needs to be pursued to provide better design principles for cell-based biosensors. A shortcoming of our experimental design is the broad range of culture ages at which the recordings were taken. However, plots of both firing rates and number of active electrodes showed no trend with culture age. Hence, we believe this limitation does not affect our basic result – that geometric constraint on neurons increases likelihood of recorded activity. It is known, however, that synaptic density becomes relatively dense by and continues to increase past culture day 8 (Fletcher et al., 1991; Ma et al., 1998; Ravenscroft et al., 1998; Yang et al., 1999), which would suggest a temporal change in firing rate within the respective combinations. Future work will address the temporal evolution of activity. Acknowledgements The authors would like to thank John Torricelli for his technical assistance in culturing the hippocampal neurons and the National Institute of Health for financial support (R21 RR13320). JC would also like to thank NIMH for an MD/PhD fellowship. References
Fig. 3. Plot of network activity against effective cell density. At densities above 250 cell/mm2, a rapid increase of activity is seen with an increase of cell density. At densities below 250 cell/mm2, local active neurons may be present. The error bars displayed are standard errors.
Bear, M., 1996. A synaptic basis for memory storage in the cerebral cortex. Proc. Natl. Acad. Sci. USA 93, 13453 – 13459. Berne, R., Levy, M., 1993. Physiology. Mosby, St. Louis, MO. Brewer, G.J., Torricelli, J.R., Evege, E.K., Price, P.J., 1993. Optimized survival of hippocampal neurons in B27-supplemented neurobasal, a new serum-free medium combination. J. Neurosci. Res. 35, 567 – 576. Chang, J., Brewer, G.J., Wheeler, B.C., 2000. Microelectrode array recordings of patterned hippocampal neurons for four weeks. J. Biomed. Microdev. 2 (4), 245 – 253.
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533 Fletcher, T.L., Cameron, P., De Camilli, P., Banker, G., 1991. The distribution of synapsin I and synaptophysin in hippocampal neurons developing in culture. J. Neurosci. 11 (6), 1617 –1626. Gross, G.W., 1994. Internal dynamics of randomized mammalian neuronal networks in culture. In: Stenger, D.A., McKenna, T.M. (Eds.), Enabling Technologies for Cultured Neural Networks. Academic Press, San Diego, CA, pp. 277 –317. Gross, G.W., Harsch, A., Rhoades, B.K., Gopel, W., 1997. Odor, drug and toxin analysis with neuronal networks in vitro: extracellular array recording of network responses. Biosens. Bioelectron. 12 (5), 373 – 393. Huelsmann, S., Oku, Y., Zhang, W., Richter, D.W., 2000. Metabolic coupling between glia and neurons is necessary for maintaining respiratory activity in transverse medullary slices of neonatal mouse. Eur. J. Neurosci. 12, 856 –862. Ma, W., Liu, Q.-Y., Jung, D., Manos, P., Pancrazio, J.J., Schaffner, A.E., Barker, J.L., Stenger, D.A., 1998. Central neuronal synapse formation on micropatterned surfaces. Dev. Brain Res. 111, 231 – 243. Manos, P., Pancrazio, J.J., Coulombe, M.G., Ma, W., Stenger, D.A., 1999. Characterization of rat spinal cord neurons cultured in
533
defined media on microelectrode arrays. Neurosci. Lett. 271, 179 – 182. Morefield, S.I., Keefer, E.W., Chapman, K.D., Gross, G.W., 2000. Drug evaluations using neuronal networks cultured on microelectrode arrays. Biosens. Bioelectron. 15, 383 – 396. Ravenscroft, M.S., Bateman, K.E., Shaffer, K.M., Schessler, H.M., Jung, D.R., Schneider, T.W., Montgomery, C.B., Custer, T.L., Schaffner, A.E., Liu, Q.Y., Li, Y.X., Barker, J.L., Hickman, J.J., 1998. Developmental neurobiology implications from fabrication and analysis of hippocampal neuronal networks on patterned silane-modified surfaces. J. Am. Chem. Soc. 120 (47), 12169 – 12177. Watanabe, M., Ohe, Y., Katakai, K., Kabeya, K., Fukumura, Y., Kobayashi, I., Miyamoto, K., Ishikawa, K., 1998. Glutamine is involved in the dependency of brain neuron survival on cell plating density in culture. NeuroReport 9, 2353 – 2357. Yang, Y., Quitschke, W.W., Vostrov, A.A., Brewer, G.J., 1999. CTCF is essential for up-regulating expression from the amyloid precursor protein promoter during differentiation of primary hippocampal neurons. J. Neurochem. 73, 2286 –2298.
Modulation of neural network activity by patterning John C. Chang a,1, Gregory J. Brewer b,2, Bruce C. Wheeler a,* a
Department of Electrical and Computer Engineering, Beckman Institute, Uni6ersity of Illinois at Urbana-Champaign, 405 N. Mathews A6enue, Urbana, IL 61801, USA b Departments of Neurology and Medical Microbiology and Immunology, Southern Illinois Uni6ersity School of Medicine, PO Box 19626, Springfield, IL 62794 -9626, USA
Abstract Using neuronal cultures on microelectrode arrays, researchers have shown that recordable electrical activity can be influenced by chemicals in the culture environment, thus demonstrating potential applicability to biosensors or drug screening. Since practical success requires the design of robust networks with repeatable, reliable responses understanding the sources of variation is important. In this report, we used lithographic technologies to confine neurons to highly defined patterns (40 mm wide stripes); in turn these patterns gave us a measure of control over the local density of neurons (100– 500 cells/mm2). We found that the apparent electrical activity of the network, as measured by the fraction of electrodes from which signals were recordable, increases 8–10-fold with greater local density. Also, average-firing rates of the active neurons increased 3 – 5-fold. We conclude that patterned networks offer one means of controlling and enhancing the responsiveness of cultured neural networks. © 2001 Elsevier Science B.V. All rights reserved. Keywords: Patterning; Recording; Hippocampal; Network activity
1. Introduction Recent research has shown that neurons can be grown in culture so as to respond in a dose dependent manner to chemicals by changing their firing pattern (Gross et al., 1997; Morefield et al., 2000). This observation has led to the suggestion that neural networks can serve as chemical sensors. To fully exploit the concept of a neuron-based biosensor, however, the variables controlling the sensor behavior must be thoroughly explored. Variables such as cell-type, cell density, cell plasticity, and cell interaction should be reasonably controlled to manipulate important sensor properties, such as robustness and repeatability. To control robustness, one could alter the cell density as it is known that hippocampal neurons survive better at * Corresponding author. Tel.: + 1-217-333-3236; fax: +1-217-2445180. E-mail addresses: [email protected] (J.C. Chang), [email protected] (G.J. Brewer), [email protected] (B.C. Wheeler). 1 Tel.: + 1-217-244-2692. 2 Tel.: + 1-217-785-5230; fax: +1-217-524-3227.
high densities, because they secrete a greater amount of glutamine (Watanabe et al., 1998). Alternatively, glia can modulate the network baseline activity through the glutamine that they supply to the neurons (Huelsmann et al., 2000), or different cell-types may be selected to respond better to a specific stimulus (Morefield et al., 2000). However, methods for controlling the sensitivity and repeatability of the sensor seem less clear, because sensitivity to chemicals changes with the area of growth (Gross et al., 1997), spontaneous activity pattern changes with network size (Gross, 1994), and response changes with exposure history of the sensor (Gross et al., 1997). While these results are strong indications that neuronal cultures can serve as biosensors, they also underscore the need for further understanding of the underlying biological mechanisms, e.g., development and plasticity, in order to create robust, reliable and repeatable sensors. In order to further our understanding of the dependence of neural activity on experimental characteristics, we are exploring the potential for the use of networks grown in patterns. Previously, we have shown that patterned hippocampal neurons develop electrical activity (Chang et al., 2000). In this report, we have com-
0956-5663/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 5 6 - 5 6 6 3 ( 0 1 ) 0 0 1 6 6 - X
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
528
pared the differences in network activity between random and patterned neural networks (parallel lines) grown in a serum- and glia-free culture. Our data indicate that the neurons establish connections with one another and that patterning enhances the network activity.
2. Methods
2.1. Array preparation The electrode arrays used are obtained from MultiChannel Systems (Reutlingen, Germany) consisting of 60 titanium nitride electrodes of 10 or 20 mm diameter, occupying all but the corner locations of an 8-by-8 square grid with vertical and horizontal separation of 200 mm. The electrodes are insulated with 0.5 mm silicon nitride and have impedance ranging between 200 and 400 kV at 1 kHz. New arrays were patterned (as described below) without additional preparation. Recy-
cled arrays were first scrubbed with commercial dishwashing detergent (Dawn) using a rubber policeman, immersed overnight in a Alconox solution ( 1.5 g/200 ml water), and rinsed the following day with running deionized water (DI) for 30+ min before patterning. Fig. 1 shows the wide-lined pattern, overlaid on an image of an electrode array. The pattern consists of alternating 40 mm wide lines of poly-D-lysine (PDL; light regions) and 60 mm wide lines of presumed silicon nitride (dark region). The patterning process is slightly modified from the photoresist process reported by Chang et al. (2000). Prior to photoresist application, the arrays are sonicated in acetone for 4 min and rinsed under running DI for 15 min or more. The photoresist (AZ5214, Hoechst-Celanese) is spin-coated (5.5 krpm) for 30 s, baked at 110 °C for 90–120 s, exposed to UV through the mask (filtered at 320 nm, 300 mW/cm2) for 15 s, and developed for 30–60 s. The arrays are then cleaned by oxygen plasma (300 W, 500 mTorr) for 1 min and adsorbed with PDL (100 mg/ml DI) for 3 h. After aspirating the PDL solution from the arrays, the
Fig. 1. An array overlaid with patterns (white-40 mm foreground, dark-60 or 200 mm background) added by Photoshop.
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
arrays are sonicated in acetone for 4 min to remove the photoresist and disinfected in 70% ethanol in DI before storage in a 10 cm poly-styrene petri dish (generic). An autoclaved PDMS (Sylgard 184, Dow-Corning) culture ring (1.5-cm diameter) is rinsed in 75% ethanol in DI, aspirated dry, and placed onto the array, all performed within a sterile laminar-flow hood before the culture is started.
2.2. Cell culture The cultures are prepared as described by Brewer et al. (1993). Briefly, the embryonic hippocampal cells are harvested from E18 embryos and mechanically dissociated. The cells are plated at 200 cells/mm2 in Neurobasal/B27 medium (Life Technologies) containing 25 mM glutamate and 0.5 mM glutamine. Cultures are incubated at 37 °C in 9% O2 and 5% CO2 (Forma Scientific). Each week, one-half of the medium is changed with Neurobasal/B27 medium containing 0.5 mM glutamine.
2.3. Acti6ity recording For recording, the arrays are removed from the incubator, imaged (Nikon Diaphot), seated in the MEA-1060 Amplifier (Gain 1000X; MultiChannel Systems), and heated to 37 °C by the accessory heater from MultiChannel Systems. Activity was monitored in the original culture medium or in Hanks balanced salt solution with 1.8 mM CaCl2 and without MgCl2. The output of the amplifier is fed into a multi-channel spike detector (Plexon Instruments Inc., Dallas, TX) with its associated software. Analog output is also fed into a VCR to sample the active channels four channels at a time. At the time of recording, the number of active electrodes is identified, and the analysis of the recordings is performed using an offline-sorting program. Recorded action potentials are digitized at 40 kHz for 1.5– 3 min intervals using the Digitizer (Plexon Instruments Inc.) and sorted by extracting the signals at 1.5 times the maximum background noise followed by Kmeans clustering of the resulting principal component clusters to identify the individual units. The unit spike counts are then converted into frequency and averaged to obtain a mean and standard deviation.
2.4. Data analysis For both cultures, effective cell densities are determined from the phase-contrast image of the array center (: 1.5×1.5 mm2 centering on the electrodes) obtained at the time of recording. Each cell is identified as a dark spot with a bright halo. Cell clusters whose constituents cannot be identified were ignored, because the cluster center tends to be silent, while clusters whose
529
constituting cells can be identified were included. The effective on pattern density is then calculated as (cells on foreground stripe/total foreground area). If an image cannot be obtained or is unclear, the culture is ignored to avoid ambiguous results. Statistical analysis of the activity is performed using Student’s t-test for two independent groups with unknown population variances. The groups are further assumed to be normally distributed with unequal variances. The significance level was set at 5%.
3. Results
3.1. Random network acti6ity le6el Despite apparent good culture growth and a moderate density of neurites traversing over the electrodes (Fig. 2(a)), spontaneous activity was rare in random cultures plated at 200 cells/mm2. The age of these networks at recording ranged from 8 to 40 days in vitro (DIV). On average, the cultures yielded 0.891.93 (N= 10 cultures) spontaneously active electrodes, ranging from zero (predominant) to six active electrodes, with effective cell density of 4739 121 cells/mm2. The inactive cultures also resisted chemical stimulation by magnesium free balanced salt solution containing calcium without magnesium. This result is in stark contrast with that of patterned neuronal networks.
3.2. Patterned network acti6ity le6el The networks of patterned neurons initially developed by growing on the patterned PDL lines and extended neurites along the interface of PDL and silicon nitride surfaces. As the cultures age, they lose the pattern (occurring after culture day 8) due to the extension of neurites to neighboring PDL lines. Recordings were made between 11 and 51 DIV in the growth medium, and our observations suggested that effective cell density dictated the level of detectable activity in these patterned cultures. The cultures (Fig. 2(b)) yielded 1097.8 (N= 8) active electrodes, and the activity was completely blocked by 1.8 mM of Mg2 + and partially returned upon washout of the medium (Fig. 2(c)), indicating the role of synaptic transmission in the spontaneous activities of the network. Furthermore, the activity showed no correlation with culture age. The difference in the activity levels (between random and patterned cultures), as measured by number of active electrodes, was statistically significant (P= 0.005; t=3.6; d.f.= 16) with patterned cultures having greater level of activity. For the active units detected, the average-firing rate was 0.439 0.35 Hz in random culture (10 units sorted from seven active electrodes; max-
530
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
Fig. 2. Phase-contrast images of the cultures at the time of recording. (a) Random culture recorded around 14 DIV. (b) Patterned culture recorded on 27 DIV. (c) Sample of the recorded action potentials (1.25 s): top trace is before MgCl2 addition, middle is with added 1.8 mM MgCl2, and the bottom is after washout.
imum firing rate of 1.3 Hz) and 2.139 3.0 Hz in patterned culture (32 units sorted from 24 active electrodes; over one-third of the units had firing rates in excess of 1.3 Hz). This difference is statistically significant (P=0.005; t= 3.1, d.f.=32). The results are summarized in Table 1. We further hypothesized that the activity levels of the random and patterned cultures are influenced by the effective cell density and tested this by plotting the activity level (Fig. 3) versus the cell density. The effective cell density of the patterned cultures ranged from 130 to 500 cells/mm2, and the number of active electrodes increased rapidly with cell density above 250 cells/mm2 (Fig. 3). This suggests that patterning may enhance the network activity level by increasing the effective cell density.
4. Discussion Using a serum-free hippocampal neuronal culture, we have demonstrated an enhancement of both network and cellular activity level over that of a culture of randomly distributed neurons by patterning the neurons into parallel lines. Two possible reasons for this enhancement are discussed here: geometrical effects and the concentration of synaptic inputs. Indeed, the patterns here yield a 2–3-fold increase in local cell density, as compared to density averaged over the array area, so that should lead to an increased likelihood that activity would be sensed. However, in random cultures there are a substantial number of neurons with neurites and somata near (e.g., within 15 mm) electrodes (Fig. 2(a)), suggesting that proximity to
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
531
Fig. 2. (Continued)
the electrodes was insufficient to enhance observed network activity. Indeed some relatively quiet random cultures had densities greater than the local densities of more active patterned networks. Instead, it seems that basal network activity determines the observed activity level. Our data show that patterning results in enhanced network activity (percent active electrodes) as a func-
tion of increasing the effective cell density, and that the increase is particularly striking and supra linear for densities above 250 cells/mm2. It is as if a threshold is passed by the higher concentration established by patterning. Therefore, we believe that other mechanisms are involved in modulating the activity of the neuronal network.
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533
532 Table 1 Summary of data and statistics
Random Patterned a b
Active electrodes (m9 S.D.)a
Unit firing frequency (Hz; m 9S.D.)b
0.8 91.9 10 97.8
0.439 0.35 2.13 9 3.0
P = 0.005, t=3.6, d.f. = 16. P = 0.005, t= 3.1, d.f. = 32.
Both cultured and in vivo hippocampal pyramidal neurons develop highly branched dendritic trees and an axon that extends for hundreds of microns with numerous branches. These neurons have a large number of inputs, implying that firing rates would be low in culture conditions in which the number of inputs is low (as is true in most dissociated cultures, such as reported here). Hence, cell-types naturally receiving a large number of inputs in vivo may require synchronized depolarization of their inputs in vitro in order to reach the firing threshold. However, there is evidence that neurons adjust their firing threshold and sensitivity according to the level of input they receive (Bear, 1996). An interpretation of our data consistent with these ideas is that, beginning at 250 cells/mm2, our cultured hippocampal neurons are receiving the minimum needed input for readily measurable spike activity. The level of network activity should be dependent on cell-type. For instance, bipolar neurons from the dorsal root ganglion fire readily both in vivo (Berne and Levy, 1993) and in random, serum-free cultures (Manos et al., 1999). In contrast, hippocampal pyramidal neurons fire at low rates in vitro and, at least in our work, at even lower rates in serum-free, glia-free culture.
Our data suggest that once active, the firing rate of cells is independent of density, at least over the densities tested. While there is a difference between the low-density random networks and the higher density patterned networks, there is no trend within the patterned network pool. Despite good growth in some random cultures (474 cells/mm2), there was little activity. Further experiments are needed to understand how synapse density influences cellular and network activity. A thorough understanding of network modulation by patterning should be important in the design of cellbased biosensors. By controlling the location and the activity level of the network, we may enhance the detectability of the sensor response and modulate the sensitivity of the sensor to certain chemicals of interest. In addition, with a mixture of cells, we may construct network sensors that respond to multiple chemicals in uniquely different fashions. Consequently, we believe that the understanding of network modulation by patterned growth needs to be pursued to provide better design principles for cell-based biosensors. A shortcoming of our experimental design is the broad range of culture ages at which the recordings were taken. However, plots of both firing rates and number of active electrodes showed no trend with culture age. Hence, we believe this limitation does not affect our basic result – that geometric constraint on neurons increases likelihood of recorded activity. It is known, however, that synaptic density becomes relatively dense by and continues to increase past culture day 8 (Fletcher et al., 1991; Ma et al., 1998; Ravenscroft et al., 1998; Yang et al., 1999), which would suggest a temporal change in firing rate within the respective combinations. Future work will address the temporal evolution of activity. Acknowledgements The authors would like to thank John Torricelli for his technical assistance in culturing the hippocampal neurons and the National Institute of Health for financial support (R21 RR13320). JC would also like to thank NIMH for an MD/PhD fellowship. References
Fig. 3. Plot of network activity against effective cell density. At densities above 250 cell/mm2, a rapid increase of activity is seen with an increase of cell density. At densities below 250 cell/mm2, local active neurons may be present. The error bars displayed are standard errors.
Bear, M., 1996. A synaptic basis for memory storage in the cerebral cortex. Proc. Natl. Acad. Sci. USA 93, 13453 – 13459. Berne, R., Levy, M., 1993. Physiology. Mosby, St. Louis, MO. Brewer, G.J., Torricelli, J.R., Evege, E.K., Price, P.J., 1993. Optimized survival of hippocampal neurons in B27-supplemented neurobasal, a new serum-free medium combination. J. Neurosci. Res. 35, 567 – 576. Chang, J., Brewer, G.J., Wheeler, B.C., 2000. Microelectrode array recordings of patterned hippocampal neurons for four weeks. J. Biomed. Microdev. 2 (4), 245 – 253.
J.C. Chang et al. / Biosensors & Bioelectronics 16 (2001) 527–533 Fletcher, T.L., Cameron, P., De Camilli, P., Banker, G., 1991. The distribution of synapsin I and synaptophysin in hippocampal neurons developing in culture. J. Neurosci. 11 (6), 1617 –1626. Gross, G.W., 1994. Internal dynamics of randomized mammalian neuronal networks in culture. In: Stenger, D.A., McKenna, T.M. (Eds.), Enabling Technologies for Cultured Neural Networks. Academic Press, San Diego, CA, pp. 277 –317. Gross, G.W., Harsch, A., Rhoades, B.K., Gopel, W., 1997. Odor, drug and toxin analysis with neuronal networks in vitro: extracellular array recording of network responses. Biosens. Bioelectron. 12 (5), 373 – 393. Huelsmann, S., Oku, Y., Zhang, W., Richter, D.W., 2000. Metabolic coupling between glia and neurons is necessary for maintaining respiratory activity in transverse medullary slices of neonatal mouse. Eur. J. Neurosci. 12, 856 –862. Ma, W., Liu, Q.-Y., Jung, D., Manos, P., Pancrazio, J.J., Schaffner, A.E., Barker, J.L., Stenger, D.A., 1998. Central neuronal synapse formation on micropatterned surfaces. Dev. Brain Res. 111, 231 – 243. Manos, P., Pancrazio, J.J., Coulombe, M.G., Ma, W., Stenger, D.A., 1999. Characterization of rat spinal cord neurons cultured in
533
defined media on microelectrode arrays. Neurosci. Lett. 271, 179 – 182. Morefield, S.I., Keefer, E.W., Chapman, K.D., Gross, G.W., 2000. Drug evaluations using neuronal networks cultured on microelectrode arrays. Biosens. Bioelectron. 15, 383 – 396. Ravenscroft, M.S., Bateman, K.E., Shaffer, K.M., Schessler, H.M., Jung, D.R., Schneider, T.W., Montgomery, C.B., Custer, T.L., Schaffner, A.E., Liu, Q.Y., Li, Y.X., Barker, J.L., Hickman, J.J., 1998. Developmental neurobiology implications from fabrication and analysis of hippocampal neuronal networks on patterned silane-modified surfaces. J. Am. Chem. Soc. 120 (47), 12169 – 12177. Watanabe, M., Ohe, Y., Katakai, K., Kabeya, K., Fukumura, Y., Kobayashi, I., Miyamoto, K., Ishikawa, K., 1998. Glutamine is involved in the dependency of brain neuron survival on cell plating density in culture. NeuroReport 9, 2353 – 2357. Yang, Y., Quitschke, W.W., Vostrov, A.A., Brewer, G.J., 1999. CTCF is essential for up-regulating expression from the amyloid precursor protein promoter during differentiation of primary hippocampal neurons. J. Neurochem. 73, 2286 –2298.