Pyramidal cells accumulate chloride at seizure onset.

Acute slice preparation

Acute slice protocols were approved by the Boston University Animal Care and Use Committee. Transverse hippocampal brain slices (400 µm) were prepared as previously described ( Netoff et al., 2005 ) from juvenile (P10-P20) mice expressing GFP in somatostatin-positive interneurons under the control of the Gad1 (GAD67) promoter (strain FVB-Tg(GadGFP)45704Swn/J, Jackson Laboratories, Bar Harbor, ME) or from Clomeleon mice. After a 1hr incubation period, they were transferred to the recording chamber where they were bathed is artificial cerebrospinal fluid (ACSF, in mM, 126 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 2 MgCl 2 , 26 NaHCO 3 , 25 dextrose, 2mM CaCl 2 ). 50µM 4-aminopyridine was added to induce epileptiform activity. In some experiments 10µM acetazolamide was added inhibit carbonic anhydrase. To elicit seizures in Clomeleon acute slices, which did not spontaneously seize in 50 µM 4-AP, MgCl 2 was omitted from the ACSF (in addition to adding 50 µM 4-AP). All chemicals were obtained from Sigma-Aldrich (St. Louis, MO). Slices were initially visualized using oblique illumination.

Organotypic slice culture preparation

Organotypic slice culture protocols were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care. Roller tube type organotypic slice cultures were prepared as described by ( Gähwiler, 1981 ). Briefly, isolated hippocampi from P6–8 CLM-1 mouse pups were cut into 350-µm slices on a McIlwain tissue chopper (Mickle Laboratory Eng. Co., Surrey, United Kingdom). Slices were mounted in clots of chicken plasma (Cocalico Biologicals, Reamstown, PA) and thrombin (Sigma-Aldrich, St. Louis, MO) on poly-l-lysine-(Sigma-Aldrich) coated glass coverslips (Electron Microscopy Sciences, Hatfield, PA) and incubated in roller tubes (Nunc, Roskilde, Denmark) at 5% CO 2 , 36°C in 750µL Neurobasal-A growth medium with 2% B27, 500uM Glutamax, and 0.03 mg/mL gentamycin added (all from Invitrogen, Carlsbad, CA). Growth medium was changed every 7 days. Recordings were made in growth media, some with the addition of 10µM GABAzine (SR 95531).

Staining and imaging

Areas of interest were stained with Indo-1 AM (Invitrogen, Carlsbad, CA) using an adapted version of multicell bolus loading (Stosiek et al., 2003; but c.f. Garaschuk et al., 2006), in which a Picospritzer II (Parker Hannifin, Pine Brook, NJ) is used to inject dye through a glass pipette directly into the brain tissue. In this modified version of MCBL, a larger-tip pipette was used (~2MΩ when filled with KCl-based dye solution) to inject many sites for a shorter duration (~5s) than that previously described (1–2min). This resulted in the staining of a large area, with relatively low background staining. The dye used for these experiments, Indo-1, is a ratiometric calcium dye, emitting at a shorter wavelength when bound to calcium. However, the two-photon cross section of calcium-bound Indo-1 is so low that it is essentially invisible when using excitation light longer than 750nm ( Xu et al., 1996 ). We take advantage of this by imaging at 820nm (a wavelength that conveniently excites both GFP and calcium-unbound Indo-1) to get relative calcium measurements with high sensitivity. Because, in this configuration, fluorescence decreases when a cell is active, all Indo-1 fluorescence traces shown have been inverted for clarity. All images of calcium dynamics were acquired using Targeted Path Scanning ( Lillis et al., 2008 ). All images were acquired using a 20× 0.95 NA water immersion objective (Olympus, Tokyo, Japan). The calcium traces for each cell were then filtered using a 15 point median filter and a 10 point boxcar filter. To average calcium traces taken with different sampling rates (as is inherent to TPS, mean sampling rate for calcium data shown = 51.8Hz), traces were upsampled using the Matlab (The Mathworks, Natick, MA) function resample , which interpolates using a polyphase filter. In 4-AP, SLEs occurred approximately once every 4 minutes. To guarantee that a SLE would be captured, a four-minute scan was initiated 2–3 minutes after the previous SLE. I/E ratios were calculated, for each recording, by dividing the mean interneuron calcium trace by the mean principal neuron calcium trace. Clomeleon images were acquired using custom-designed software and the scanhead from a Radiance 2000MP (BioRad, Hemel Hempstead, UK), equipped with a 20× 0.95 NA water-immersion objective (Olympus, Tokyo, Japan), and PMTs with appropriate filters for YFP (545/30) and CFP (450/80). A SpectraPhysics MaiTai laser (Newport, Irvine, CA), set to 860nm, was used for two-photon excitation. Chloride concentrations were obtained by performing a calibration as previously described ( Glykys et al., 2009 ) using 10µM of the K + /H + ionophore nigericin, 100µM of OH − /Cl − antiporter tributyltin chloride, and known concentrations of extracellular chloride. YFP/CFP ratios for all selected cells were low-pass filtered at 0.5Hz and, because CFP and YFP bleach at different rates ( Kuner and Augustine, 2000 ), a linear trend was subtracted to remove the effect of differential bleaching. For the pre-ictal burst-triggered averaging shown in Figure 3 , chloride traces were resampled in the same manner described above for calcium traces. pH changes were measured by staining organotypic slice cultures with SNARF-1-AM by incubating them for >1hr in 10uM dye. Because of the spectral overlap between SNARF-1 and Clomeleon, wild type mice were used for pH imaging experiments. Images were acquired using TPS with an excitation wavelength of 820nm and PMT emission filters centered at 585 and 640. The ratio of light emitted at 585 and 640nm was recorded and linearly detrended (to correct for bleaching) and adjusted to baseline pH using a calibration strategy previously described ( Sheldon et al., 2004 ).

Electrophysiology

During imaging recordings, either a field potential recording (principal neuron spike delay.

Supplementary Material 01 Supplemental Figure: Network Analysis A) In representative calcium transients at SLE onset, it is apparent that cell 1 leads cell 2. B) This is evident in the cross-correlation trace between the two cells, in which the peak occurs at −40ms. C) For this example network, the numbers next to lines represent the peak of the cross correlation between the two cells. The direction of the arrow represents whether the peak came at a positive or negative time lag. Weighted degree is the sum of all peak correlations and represents how correlated the node of interest (green node in center) is with rest of the network (yellow nodes). In-degree is the sum of all “incoming” connections and represents the degree to which the node of interest follows the rest of the network. Similarly, out-degree is the sum of “outgoing” connections. For the example shown in A and B, since cell 1 leads cell 2, this connection will be counted towards in-degree for cell 2 and out-degree for cell 1. Finally direction-degree is out-degree subtracted from in-degree and represents the degree to which the cell of interest leads (negative values) or follows (positive values) activity in the network. D) To look at temporal dynamics of correlation structure in the network, we calculated cross-correlation using 1s windows shifted in increments of 0.1s. Running this analysis on the calcium imaging data shown in Figure 1 reveals a negative peak in the interneuron direction degree at the time of the pre-ictal burst. This suggests that the interneurons are leading activity in the network during the pre-ictal burst.