Hippocampal networks on reliable patterned substrates

Abstract

Toward the goal of reproducible live neuronal networks, we investigated the influence of substrate patterns on neuron compliance and network activity. We optimized process parameters of micro-contact printing for reproducible geometric patterns of 10 μm wide lines of polylysine with 4, 6, or 8 connections at a constant square array of nodes overlying the recording electrodes of a multielectrode array (MEA). We hypothesized that an increase in node connections would give the network more inputs resulting in higher neuronal outputs as network spike rates. We also chronically stimulated these networks during development and added astroglia to enhance network activity. Our results show that despite frequent localization of neuron somata over the electrodes, the number of spontaneously active electrodes was reduced 3-fold compared to random networks, independent of pattern complexity. Of the electrodes active, the overall spike rate was independent of pattern complexity, consistent with homeostasis of activity. Lower mean burst rates were seen with higher levels of pattern complexity; however, burst durations increased 1.6-fold with pattern complexity (n=6027 bursts, p<0.001). Inter-burst interval and percentage of active electrodes displaying bursts also increased with pattern complexity. The extra-burst (non-burst or isolated) spike rate increased 4-fold with pattern complexity, but this relationship was reversed with either chronic stimulation or astroglia addition. These studies suggest for the first time that patterns which limit the distribution of branches and inputs are deleterious to activity in a hippocampal network, but that higher levels of pattern complexity promote non-burst activity and favor longer lasting, but fewer bursts.

Fig. 1

Steps in creating patterned growth of neuronal networks. A) entire stamp mask for a 6-connect pattern including moat region, B) higher magnification of A, C) MEA before stamping showing circular electrodes and leads for connection to the amplifier, D) fluorescence of PLL-FITC on stamped MEA substrate, E) neuron growth on stamped MEA substrate after 22 days in culture. Each square is 200 µm on a side.

Fig. 2

Treatment of MEA with oxygen plasma treatment and 3-GPS silane improves patterned polylysine transfer and cell survival. A) Fluorescent intensities of patterned lines of polylysine/PLL-FITC transfer by stamping onto electrode arrays prepared as indicated. Significance between bars with different letters p<0.01; bars with same letters are not significantly different (n = 4 arrays for each condition). B) Cell somata survival on patterned substrates after 22 days (n = 4 arrays for each condition, 18 pattern segments for each array). C) Cell somata survival on patterned substrates in relation to intensity of polylysine/PLL-FITC transfer. Note that survival is not a simple function of polylysine density (data combined from A and B). D) Untreated MEA substrate with poor cell survival at 22 days in vitro. E) Oxygen plasma and 3-GPS silane treated MEA substrate with good cell survival at 22 days in vitro. Phase bright objects are cell somata. Electrode spacing is 200 µm.

Fig. 3

Effect of polylysine dissolution time on dispersion and uniformity. A) Image of polylysine stock thawed 5 min. shows bright specs of aggregated polylysine and uneven brightness. B) Image after thawing for 60 min. shows uniformly bright field. Main graph shows the highest fluorescence intensity with the lowest variability in solution (as measured by the standard error) for 60 min. dissolution compared to shorter times (n = 5 glass coverslips for each condition, 4 fields of 0.145 mm² for each slip).

Fig. 4

Methods for removal of excess polylysine from stamps affect impression and subsequent cell growth on glass coverslips. PDMS stamps were submerged in 10% SDS for 20 min then, A) drying in a stream of nitrogen without any rinse. B) Water rinse and drying, C) Brief drying followed by water rinse and drying again. Only the last procedure resulted in acceptable growth of neurons (right panels).

Fig. 5

Substrate patterns of increasing complexity grow neurons equally well to 22 days in vitro. Acceptable neuron growth and compliance to patterns. A) 4-connect, B) 6-connect, C) 8-connect, D) Random, E) Average soma per pattern area follow trend of theoretical soma on pattern based on original plating density of 50 cells/mm² on patterned arrays and 500 cells/mm² on random arrays over the area of random MEA plating substrate available. No patterned MEA had more than 3% total soma grow off of the pattern (n = 2 arrays for each condition with each array analyzed in 10 fields of 0.4 mm² area).

Fig. 6

Spiking activity on patterns of increasing complexity. A) Spike waveforms comparing 6-connect patterned network bursts with random bursts. Random networks have longer lasting bursts (300ms window) but fewer bursts/min (60s window) compared to 6-connect patterned networks. B) Percent electrodes active is low for all patterns (4, 6, 8 connect) compared to random cultures (n= degrees of freedom for 73 arrays; ANOVA for connections). C) Average channel spike rate is unaffected by pattern complexity (n= degrees of freedom from 67 active arrays). D) Mean burst rate declines with pattern complexity (n= degrees of freedom for 400 active electrodes; ANOVA for connections), E) Burst duration increases with pattern complexity (n= degrees of freedom for 6031 bursts; ANOVA for connections). Results collected for all conditions combined: non-stimulated, stimulated, without added astroglia, with added astroglia. 4, 4-connect pattern; 6, 6-connect pattern; 8, 8-connect pattern; R, random network.

Fig. 7

Effect of stimulation and astroglia addition on burst characteristics. A–D) Unstimulated data (with and without added astroglia) show increased activity with pattern complexity. A) Percent of active electrodes (n=degrees of freedom for 33 arrays), B) spikes per burst (n=degrees of freedom for 2636 bursts), C) interval between bursts (n=degrees of freedom for 2207 bursts on channels with 2 or more bursts), D) isolated spike rate (n=degrees of freedom for 349 electrodes), E) with addition of astroglia, isolated spike rates declined with pattern complexity (n=degrees of freedom for 324 electrodes), or F) as a result of chronic stimulation during development (n=degrees of freedom for 413 electrodes). 4, 6, 8, R as in Fig. 6.

Fig. 8

Compared to random networks, patterned substrates produce low numbers of spike pairs cross-correlated within a 20 ms window on separate electrodes, averaged over number of arrays (n indicated). Asterisk, ANOVA F (1,49)=6.3, p=0.015. The correlated pairs were divided by the number of active electrodes for each array.