INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

The concept of "selective silencing," in which specific groups of neurons are reversibly inhibited (not killed) by expression of a foreign gene, has appeared in several studies. Viral injections (Bensadoun et al. 2000; Naldini et al. 1996) or space- and time-limited transgenics (J. Chen et al. 1998) can deliver such a silencing gene to areas or specific cell types in the nervous system in ways that cannot be achieved with diffusible molecules such as TTX. Selective genetic ablation of specific neurons has also been demonstrated using diphtheria toxin (Drago et al. 1998; Francis et al. 2000). The ability to similarly target a fully reversible, nonlethal silencer in vitro or in vivo has enormous implications to studies of behavior, neural computation, and development.

Obvious choices of silencing genes in a Hodgkin-Huxley neuron are the K⁺ and Na⁺ channels. If K⁺ current can be increased or Na⁺ current decreased, the amplitude of action potentials will be proportionally lessened and the firing threshold raised. Previous silencing studies have focused on K⁺ channels because of their diversity and the ease with which they may be cloned and mutated (Ehrengruber et al. 1997; Johns et al. 1999; Nadeau et al. 2000; Sutherland et al. 1999). However, this approach has been subject to severe complications. K⁺ efflux is directly linked to apoptosis (Yu et al. 1997) and neurons expressing K⁺ channels through which outward current flows show massive death within hours of gene expression (Nadeau et al. 2000).

Expression of strong inward rectifiers may be less lethal, although this is not certain: previous in vitro studies have used either drug-activated (Ehrengruber et al. 1997) or inducible channels (Johns et al. 1999) and have not attempted to express constitutive current for 24 h. Constitutive expression during development in vivo does not appear to lead to cell death (Sutherland et al. 1999) but instead to compensatory mechanisms that change the long-term phenotype of infected neurons from hypoexcitable to hyperexcitable. Effects are seen on both endogenous Na⁺ and K⁺ channel expression (Eustache and Gueritaud 1995; Nadeau et al. 2000; Sutherland et al. 1999).

Due to these complications, K⁺ channels are not ideal silencing agents. A genetic silencer of Na⁺ channels is hence desirable because complete pharmacologic blockade with TTX is well understood and, in many neuronal types, may be maintained chronically without cell death or upregulation of TTX-resistant channels. However, the ideal approach to genetically suppressing Na⁺ current has remained unclear. Mutagenesis of  subunit DNA is difficult due to its large size and propensity for recombination (McPhee et al. 1998; Yarov-Yarovoy et al. 2001), and it is not known what mutations, if any, might result in a dominant negative phenotype. Overexpression of the inactivation gate may block Na⁺ channels, but it is not likely to be specific (Patton et al. 1993). Creating a vector to express antisense RNA poses the same problems as working with the original  subunit gene.

Rather than mutate the Na⁺ channel directly, in this study we suppress its transcription in cultured hippocampal neurons by infecting them with a virus encoding the zinc finger transcription factor neuron restrictive silencer factor (NRSF/REST). The low cytotoxicity of lentiviral vectors allows us to monitor the course of infection over 2 wk. Previous work with NRSF has focused on cell lines and on measurements of levels of gene transcription, demonstrating that NRSF silences gene expression by binding to a neuron restrictive silencer element (NRSE) in the promoter. Transcriptional repression of the Na⁺ channel type II (NaCh II) promoter by NRSF has been demonstrated in PC12 cells (Tapia-Ramirez et al. 1997).

However, it is impossible to predict from these data how neurons will be affected by overexpression of NRSF. Primary neurons are very different from cell lines, and levels of gene expression are often very poor correlates of physiology, especially over a time course of many days to weeks. Differentiated neurons express isoforms of Na⁺ channel other than NaCh II, whose balance may be shifted by activity enhancement or deprivation. To further complicate the picture, functional NRSEs have been identified in the promoters of many neuron-specific genes apart from NaCh II, namely synapsin I, brain-derived neurotrophic factor (BDNF), the AMPA receptor subunit GluR2, and others (Brene et al. 2000; Myers et al. 1998; Schoenherr et al. 1996). This wide variety of genes makes complex feedback loops likely: direct effects on factors such as BDNF may lead to secondary alterations in excitability (Rutherford et al. 1997) that in turn affect NRSF expression. Neurons may also possess compensatory mechanisms that alter the effects of direct gene repression over a longer time course.

We obtained the rather surprising result that overexpression of NRSF in cultured neurons leads to a benign phenotype, very similar to that seen with chronic TTX application. Effects were sensitive to intracellular cAMP concentration but relatively insensitive to NRSF expression levels. We thus conclude that the physiological effects of NRSF in terminally differentiated hippocampal neurons are similar to those of Na⁺ channel blockade. Repression of neuronal phenotype or cell death are not seen even with high levels of overexpression, making this a potentially useful silencing gene. Our findings also implicate NRSF in the feedback loops that exist between excitability and receptor subunit expression and suggest further experiments to distinguish between direct and indirect effects.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Molecular biology

The entire mouse NRSF coding sequence was excised as a 3.4-kb BamHI-EcoRI fragment from the plasmid pCS2-NRSF (Shimojo et al. 1999) (gift of D. J. Anderson, Caltech). To create the bicistronic lentiviral construct pHR'-NRSF-CITE-GFP, this sequence was cloned into the plasmid pHR'-ROMK1-CITE-GFP (Nadeau et al. 2000; Naldini et al. 1996), from which ROMK1 had been excised with BamHI and EcoRI. To create the enhanced construct pHR'-WRE-NRSF-CITE-GFP, containing 500 bp of an enhancer sequence from the woodchuck hepatitis virus, NRSF-CITE-GFP was excised from pHR'-NRSF-CITE-GFP with BamHI and SmaI and inserted into pHR'WRE (Zufferey et al. 1999) with BamHI and EcoRI, blunted on the 3' end with T4 DNA polymerase. Lentiviruses were generated from these constructs, and from the control construct encoding GFP alone (enhanced GFP, Clontech, Palo Alto, CA). Viruses were used to infect hippocampal neurons as described (Nadeau et al. 2000). Details of all constructs are given in Fig. 1.

View larger version (7K): [in this window] [in a new window]   Fig. 1. Constructs used to express neuron restrictive silence factor (NRSF) in hippocampal neurons in culture. All fragments shown are in the vector pHR' (Zufferey et al. 1999); they are flanked by the 5' and 3' UTRs of HIV-1. A: pHR'-WRE-NRSF-CITE-GFP, containing the CMV promoter; the full NRSF coding sequence; a cap independent translation enhancer (CITE) element allowing for transcription of 2 independent genes; the full coding sequence of enhanced green fluorescent protein (EGFP); and the woodchuck responsive element (WRE) enhancer. The construct without the enhancer, pHR'-NRSF-CITE-GFP, is identical until its termination at the NotI site following the EGFP. B: pHR'-GFP, the control construct which encodes EGFP alone. The EGFP sequence is identical to that in A.

Hippocampal cultures

Neurons were prepared from E18 Wistar rats as described (Nadeau et al. 2000). Plating and feeding medium was Neurobasal with B27 supplement, with 500 µM Glutamax, 25 µM glutamate, and 5% horse serum (all from Life Technologies, Gaithersburg, MD); serum-free cultures were also prepared by omitting serum from both plating and feeding medium. These cultures have been shown to be 70-90% glutamatergic; in studies of synaptic transmission, GABAergic interneurons were excluded based on morphology (Benson et al. 1994). Such visual identification of neurons was verified using paired recordings and shown to be >95% accurate; GABAergic neurons appeared large and round in mature cultures, while glutamatergic cells were smaller and took a variety of shapes (unpublished data). In all other recordings, neurons were chosen at random. GFP was visualized by fluorescence microscopy on a Nikon inverted microscope illuminated by a 100 W Hg lamp, using an EndowGFP double band-pass filter (exciter 470/40; dichroic 495 LP; emitter 525/50). The objective was a ×40 air with NA = 0.5.

Immunocytochemistry

Antibody staining to Na⁺ channels was with polyclonal rabbit anti-NaCh I, II, III, and 6 (Chemicon, Temecula, CA). Cultured cells were fixed for 30 min in 4% paraformaldehyde, permeabilized for 2 min on ice with 1% Triton X-100, and preincubated for 30 min in PBS with 5% goat serum. Primary antibody was diluted 1:50 in the same solution and incubated with gentle shaking for 2 h at room temperature or overnight at 4°C. The dish was washed five times and incubated with fluorescent secondary antibody (Cy-3 conjugated goat anti-rabbit, Jackson ImmunoResearch, West Grove, PA) for 60 min at 37°C.

Antibody staining to NRSF was performed with mouse monoclonal antibody, 12C11 (gift of D. J. Anderson, Caltech). This hybridoma supernatant was not diluted. Antibody staining to TuJ1 was with mouse monoclonal anti--tubulin (Babco, Richmond, CA), diluted 1:200; antibody staining to synapsin was with mouse monoclonal anti-synapsin I (Chemicon), diluted 1:100. All other procedures were as in the preceding text, except that the fluorescent secondary for mouse primaries was rhodamine-conjugated anti-mouse IgG (Chemicon).

Images were taken with the same optics used to detect GFP expression described in the preceding text. Both Cy-3 and rhodamine were visualized with a Texas Red filter (exciter 560/55; dichroic 595 LP; emitter 645/75; Chroma Technologies, Brattleboro, VT). Negligible background was produced by these red secondary antibodies when observed with the GFP filter.

In addition, quantification of immunostaining was performed on a Zeiss LSM-510 using both FITC (488 nm) and rhodamine (568 nm) laser lines, with a ×40 Plan-Neofluar water-immersion lens with NA = 0.9 (Zeiss). At least two independent preparations were used for each stain, and at least four random sections of the dish were chosen for imaging. Image processing and quantification was performed with NIH Image 1.62.

Whole cell recording

All recordings were performed at room temperature, using borosilicate dot glass capillaries (Sutter Instruments, Novato, CA) pulled to a tip resistance of 5-7 M. Drugs were bath-applied or perfused continually through flow pipes of 250-µm ID mounted ~500 µm from the recorded cell. The dish was washed with 4-6 ml of control saline between recordings. Signals were recorded with an Axopatch 200A amplifier and sampled by a Digidata 1200 (Axon Instruments, Foster City, CA) at 20-100 kHz and filtered at 1-10 kHz.

Controls were from sister dishes, mock-infected or infected with GFP alone. All conclusions were drawn from data taken from at least three independent preparations. Data from both control groups were pooled when no effects of GFP infection were detected on resting membrane potential, spike threshold, input resistance, or EPSC amplitude and frequency. Cells were eliminated from analysis if the holding current was >1 nA. For Na⁺ current quantification and studies of synaptic currents, series resistance and capacitance were compensated on-line at 80% with a lag of 60 µs, and cells were eliminated if series resistance exceeded 20 M or changed by >20% over the course of the recording.

Excitatory postsynaptic currents (EPSCs) were determined to result from activation of AMPA receptors by demonstrating complete block with the addition of 20 µM 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX, RBI, Natick, MA); however, recordings were performed in the absence of CNQX. All mEPSCs were recorded in the presence of 1 µM TTX. Evoked GABA responses were elicited with 0.1 mM GABA.

Analysis

Current and voltage commands and data acquisition were performed using pCLAMP 8.0 (Axon). Data were analyzed with AxoGraph 3.5 (Axon). Na⁺ currents were quantified by subtracting traces taken in 1 µM TTX from those without TTX. The activation curve was estimated by fitting a straight line to the chord conductance of depolarizing voltage steps from a holding potential of 89 mV. Spike threshold was taken as the point at which the derivative of the membrane potential exceeded 10 V/s. Statistical tests were two-tailed Student's t; P values of < 0.05 were taken as significant.

GFP-infected and uninfected controls were pooled when GFP infection demonstrated no effects on active or passive membrane properties (Nadeau et al. 2000; in preparation).

For EPSC analysis, we defined a template function for each data file, comparing potential events to this function using the event detection module of AxoGraph. This procedure rejects events that do not have a smooth rise and decay phase and that occur too close to another event (<10 ms). The minimum signal to noise ratio was set to 5. Averages was taken with baselined events aligned on the point of fastest rise; mean ± SE number of events per cell was 60 ± 10. A linear fit to EPSCs at negative potentials was used to estimate the predicted EPSC at positive potentials if the responses were ohmic; the rectification ratio was defined as the measured mean EPSC at +40 mV divided by this predicted value.

Action potential simulations were performed using NEURON5 (available for free download from Yale University), using a hippocampal pyramidal cell model (Migliore et al. 1999). Experimental input parameters were temperature, internal and external solution composition, resting membrane potential, and applied i-clamp protocol.

Solutions

All chemicals were products of Sigma unless stated otherwise. For Na⁺ current quantification and measurements of AMPA EPSCs, the internal solution contained (in mM) 125 CsF, 5 NaF, 10 KCl, 10 Tes, and 10 EGTA, adjusted to pH 7.4 with KOH and 250 mOsm with sucrose; the bath solution contained (in mM) 135 NaCl, 5 KCl, 1 MgCl₂, 1 CaCl₂, 5 CoCl₂, 5 CsCl, and 1 Tes, adjusted to pH 7.4 with NaOH and 260 mOsm with sucrose. Calculated junction potential (pCLAMP 8.0) was 9 mV, and all reported potentials are adjusted for this value.

For quantification of GluR2 subunit composition, the internal solution contained (in mM) (Liu and Cull-Candy 2000) 95 CsF, 45 CsCl, 10 HEPES, 10 EGTA, 2 NaCl, 2 ATP-Mg, and 1 QX314 (Tocris Cookson, Ballwin, MO), 5 TEA,1 CaCl2, 0.1 spermine, adjusted to pH 7.3 with CsOH. The bath solution contained (in mM) 125 NaCl, 10 HEPES, 2.5 KCl, 2CaCl₂, 1 MgCl₂, and 10 glucose, adjusted to pH 7.4 with NaOH and 260 mOsm with sucrose. Calculated junction potential (pCLAMP 8.0) was 10 mV, and all reported potentials are adjusted for this value.

For all other recordings, the internal solution contained (in mM) 100 K-gluconate, 10 HEPES, 3 Na₂phosphocreatine, 1.1 EGTA, 3 MgATP, 0.2 NaGTP, 5 MgCl₂, and 0.1 CaCl₂, adjusted to pH 7.2 with KOH and 250 mOsm with sucrose. The bath solution contained (in mM) 110 NaCl, 10 HEPES, 5.4 KCl, 1.8 CaCl₂, 0.8 MgCl₂, and 10 glucose, adjusted to pH 7.4 with NaOH and 260 mOsm with sucrose. Calculated junction potential (pCLAMP 8.0) was 14 mV, and all reported potentials are adjusted for this value.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Verification of expression and neuronal markers

Hippocampal neurons infected with the bicistronic lentivirus NRSF-CITE-GFP become visibly fluorescent within 48 h and increasingly fluorescent for several days afterward (Fig. 2 A and B). Seventy to 90% of neurons express the GFP marker, and a similar percentage show positive immunoreactivity to monoclonal anti-NRSF, localized to the nucleus as is expected for the full-length protein (Shimojo et al. 2001) (Fig. 3 A and C). Controls infected with GFP alone show a similar pattern of green fluorescence, but their endogenous NRSF levels are insufficient to produce positive staining under these conditions (Fig. 3B). All neurons stain positively for -tubulin and synapsin, and no reduction in intensity from either of these markers is seen at any time after NRSF or GFP infection (synapsin, Fig. 3 B and E; tubulin data not shown). Time points tested were 48 h, 2 days, 4 days, 6 days, with two independent preparations per time point; all p values relative to control >0.2. Staining for Na⁺ channels and tubulin was quantified as integrated total fluorescence per cell (normalized to cell area), while synapsin staining was quantified by counting discrete puncta per unit length of red-conjugated anti-synapsin on a GFP background. All cultures in to these studies were >10 days in culture (dic) at the time of infection, ensuring mature levels of expression of neuronal markers.

View larger version (61K): [in this window] [in a new window]   Fig. 2. Effects of NRSF infection on survival and morphology. Scale bar = 20 µm. A: mature hippocampal cultures 30 days in culture (dic), 12 days postinfection (dpi) with NRSF-CITE-GFP. All of the neurons in this field of view show GFP expression (top, left, and right); there are also fluorescent glia (bottom center). Infection efficiency for neurons averages 70-90% (Nadeau et al. 2000; unpublished data). B: neurons retain normal morphology under phase contrast. Neurons can be easily distinguished from glia in these cultures by the 3-dimensional appearance of their somata and processes under phase contrast microscopy. C and D: fluorescent and phase contrast images of neurons infected with NRSF immediately postplating (6 dic, 6 dpi). Fluorescent neurons are swollen and blebbed. A normal-appearing neuron (red arrow) does not fluoresce. E and F: fluorescent and phase contrast images of neurons 10 dic, 6 dpi, showing somal flattening and disintegration of processes with beading in the fluorescent cell (white arrow). An uninfected cell (red arrow) demonstates normal morphology for this age. Infected neurons cannot be patch clamped. G and H: neurons 10 dic, 6 dpi, raised in 50 ng/ml BDNF. An uninfected cell (red arrow) shows normal processes; an infected cell (white arrow) has a normal soma but has failed to extend neurites. This cell showed no synaptic activity but was otherwise electrophysiologically normal, with a resting membrane potential of 44 mV. Its uninfected neighbor displayed a normal pattern of 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX)-sensitive spontaneous excitatory postsynaptic currents (sEPSCs).

View larger version (29K): [in this window] [in a new window]   Fig. 3. A: cultures infected with NRSF show positive immunostaining with anti-NRSF; the staining is localized to the nucleus, so neurons and glia cannot be distinguished in the fluorescent image. B: neurons infected with GFP alone show no anti-NRSF staining and no bleed-through of GFP fluorescence into the red channel (same gain and exposure time as A). C: overlay of staining from A with the phase contrast image allows distinction of neurons (black arrows) from glia. D and E: immunoreactivity to synapsin I is unaffected by NRSF infection for 6 days (see Fig. 6D). D: NRSF, 4 dpi. Binding of anti-synapsin I is quantified by counting the red or yellow puncta per unit length (yellow indicates coincidence of green and red channels). E: GFP control. NRSF puncta per unit length relative to GFP, 1.0 ± 0.1. Glial cells (white arrows) express GFP but show no synapsin staining. F-H: immunoreactivity to NaCh II decreases over time (all panels photographed with the same gain and exposure time; also see Fig. 6D): uninfected neurons (F); NRSF (G), 4 dpi. Fluorescence per cell (normalized to cell area) has decreased to 59 ± 5% of the control value (NRSF, n = 11 cells; control, n = 12; P < 0.005); NRSF (H), 7 dpi. Fluorescence does not differ significantly from background (n = 3 dishes for NRSF). I and J: Immunoreactivity to Na6 is weak and does not change with NRSF infection. Images taken with gain increased 16-fold over panels (F-H). I: control neurons, 21 dic; J: NRSF, 21 dic, 6 dpi. Fluorescence per cell, normalized to cell area, is 90 ± 10% of the control value (n = 10 NRSF; n = 12 controls).

Electrophysiology

FRESHLY PLATED NEURONS ARE KILLED BY NRSF OVEREXPRESSION. Hippocampal neurons may be infected with GFP-only lentivirus directly after plating, with a high rate of transduction and no adverse effects on cell morphology, electrophysiology, or life span (Nadeau et al. 2000). Active and passive membrane properties remain indistinguishable from controls for 30 days after infection (unpublished data).

SURVIVAL AND WHOLE CELL PROPERTIES: DIFFERENTIATED NEURONS. Differentiated neurons with dendritic arbors, 7 dic, survive infection with NRSF for 8-12 days without obvious change in morphology or viability (Fig. 2, A and B). When cells eventually die, they may show disintegration of processes but never display the abnormal somata or beading of early-NRSF cells. From 5 to 7 dpi, their passive membrane properties are identical to those of uninfected and GFP-infected neurons except for a twofold decrease in input resistance at hyperpolarized potentials that disappears on application of Co²⁺. At more depolarized potentials, those relevant to action potential firing, there is no significant change in inward or outward K⁺ current until 7 dpi. At this point, mean outward current drops by >40% (Table 1).

View larger version (40K): [in this window] [in a new window]   Fig. 4. NRSF infection causes hyperexcitability lasting several days. All traces are shown without series resistance and capacitance compensation. A: voltage step protocol, with test potentials between 154 and +46 mV (holding at 94 mV). B: in a control neuron infected with GFP, the individual voltage steps are well clamped and inward and outward currents are apparent. C: a neuron infected with NRSF (4 dpi) shows escape from voltage clamp at both hyperpolarized and depolarized potentials. Large swings of membrane potential are evident during holding, and the neuron escapes clamp altogether at 124 mV (solid arrow). At depolarized potentials, unclamped action potentials become apparent (dashed arrow). It is impossible to perform accurate compensation or to draw conclusions about the magnitude of the Na+ current under these conditions. D: at 6 dpi, the Na+ current is visibly reduced but the hyperexcitability persists. E: the addition of 1 mM Co2+ to the bath restores NRSF neurons to qualitatively normal voltage clamp behavior (6 dpi). F: at 7 dpi, NRSF-infected neurons lose all Na+ channel activity and are effectively silent. They also show reduced outward K+ currents at this stage (Table 1, Fig. 6C). G: a control neuron chronically exposed to low levels of TTX (200 nM for 9 days) shows similarities to the NRSF phenotype.

View larger version (16K): [in this window] [in a new window]   Fig. 5. NRSF-infected neurons show high levels of spontaneous activity but reduced action potential peak. A: held in i = 0 mode, control neurons (GFP) show small spontaneous membrane potential changes, which do not lead to firing. B: NRSF neurons 6 dpi (NRSF) show continuous firing. C: in 20 µM CNQX (NRSF CNQX), all activity is suppressed (GABAergic events are rare and not seen on this trace). D: spontaneous firing is blocked by 1 mM Co2+ (NRSF Co2+), but small potential changes remain. E: in neurons incubated in 200-500 nM TTX for 1 wk (TTX), a similar pattern to the NRSF cells is observed. F: shorter time scale of maximum evoked spikes from NRSF-infected (left) and control (right) neurons in i-clamp mode. These example neurons demonstrated values close to those of the means in Table 1: difference in AP amplitude A, 23 mV; difference in threshold T, 3.1 mV; negligible changes in width and AHP. G: NEURON simulation of a pyramidal neuron action potential, with input of all relevant experimental parameters, shows similar effects obtained by reducing mean Na+ conductance <gNa> by a factor of 2 (from 0.12 to 0.06 pS/cm2). A = 18 mV, T = 3 mV; changes in width and AHP are smaller than the experimental variability of these parameters (see Table 1).

SUPPRESSION OF NA⁺ CURRENT. Because of the fast large currents associated with Na⁺ channel activation, quantifying them by voltage-clamp presents particular difficulties. However, recording media containing Co²⁺ and Cs⁺ may be prepared that isolate the Na⁺ current and reduce its magnitude as well as suppressing hyperexcitability ("Na⁺ current medium"; see METHODS; Fig. 6A) (Stoll and Galdzicki 1996; Zhang et al. 1998). The resulting currents reverse at E_Na and do not exceed 120 pA/pF (Fig. 6B). Peak current occurs at 10 mV (Fig. 6B), consistent with previous studies (Desai et al. 1999), and at least two sets of voltage-step measurements from the same neuron were taken during the recording to ensure reliability of Na⁺ current measurements. Neuronal size and input resistance were comparable (Table 1, Fig. 6C); neurons with high access resistance were discarded. These precautions maximized the accuracy of voltage clamp, and current densities recorded from hippocampal neurons under these conditions have been shown to agree with other methods of Na⁺ channel quantification, such as saxitoxin binding and RT-PCR (Stoll and Galdzicki 1996). In our study, there was an excellent correspondence between peak Na⁺ current and binding of anti-NaCh II antibody (Fig. 6D).

View larger version (35K): [in this window] [in a new window]   Fig. 6. NRSF expression causes cAMP-sensitive suppression of Na+ channel function. A: Na+ currents in hippocampal neurons. Same voltage-step protocol as in Fig. 4A. The Na+ current was evoked from voltage steps of 149 to +51 mV, with holding at 89 mV, in medium containing Co+ and Cs2+; 1 µM TTX was then applied, and the traces subtracted to yield the TTX-sensitive component. B: peak Na+ current density vs. voltage measured in NRSF-infected neurons over the course of 7 days. At 2 dpi, Na+ current density in NRSF neurons (, n = 9) is slightly but not significantly larger than in uninfected or GFP-infected controls (, n = 29). By 4 dpi (, n = 20), the current is reduced significantly with respect to controls (P < 0.002); there is no further significant decrease at 6 dpi (, n = 13). By 7 dpi, all detectable Na+ currents have vanished (, n > 20, error bars smaller than symbol). C: in the same neurons, outward currents do not change until 7 dpi. Total steady-state outward currents are shown for NRSF 4-6 dpi (NRSF, , n = 33), controls (cont, , n = 38), and NRSF 7 dpi (NRSF 7 days, , n = 14). No significant differences exist between NRSF neurons <7 dpi and controls at any membrane potential. D: decrease in peak current (, I (Na) peak, - - -) and current at 0 mV (, I (Na) 0 mV, ) relative to control is paralleled by reduced immunoreactivity to anti-NaCh II (, NaCh II). Immunostaining was quantified as whole cell fluorescence due to secondary antibody, normalized to cell area, using 6 neurons from each of 2 independent preparations. Synapsin I immunoreactivity (; quantified as shown in Fig. 3, D and E) does not change throughout the time period observed. E: probability plot of Na+ current density for controls (, n = 38) and NRSF 4-6 dpi (, n = 33) shows a significant decrease in the number of NRSF neurons expressing the highest levels of functional Na+ channel. F: activation curves for controls () and NRSF neurons () are identical (when only 1 symbol appears, the values overlap). G, conductance; the x axis is membrane potential.

NA⁺ CURRENT SUPPRESSION IS ACCOMPANIED BY INCREASED SIZE OF AMPA MEPSCS. Previous studies have identified increased AMPA receptor sensitivity as a major component of the hyperexcitability that follows activity blockade (Turrigiano et al. 1998). In fact, neurons exposed to 200-500 nM TTX for 2-7 days display the same Co²⁺- and CNQX-sensitive hyperexcitability in voltage clamp mode as do NRSF-infected cells (Figs. 4G and 5E). Hence, quantification of gluamatergic synaptic events may not only shed light on the NRSF phenotype but suggest analogies to other forms of activity alteration.

View larger version (23K): [in this window] [in a new window]   Fig. 7. AMPA hyperexcitability results from increased size of miniature EPSCs (mEPSCs; recorded in the presence of 1 µM TTX). A: AMPA mEPSCs recorded from GFP-infected neurons (GFP); GFP-infected neurons exposed to 200-500 nM TTX for 2 days (GFP TTX 2 days) or 9 days (TTX 9 days); neurons infected with NRSF-CITE-GFP for 3 days (NRSF 3 days), 5 days (NRSF 5 days), or 7 days (NRSF 7 days). Between 7 and 9 days after addition of either NRSF virus or TTX, no synaptic events can be resolved. B: average waveform on a shorter time scale, showing increased mEPSC amplitudes for TTX and NRSF conditions. Significant differences (P < 0.05) exist for control (n = 10) vs. TTX (n = 14; 128 ± 10% of control) and control vs. NRSF 4-5 dpi (n = 7; 140 ± 20%). At 3 dpi, the NRSF amplitudes show a tendency to increase (n = 5; 114 ± 11% of controls), but this difference is not statistically significant. C: the probability distribution of mEPSC amplitudes in NRSF-infected neurons shows a shift between 3 ()and 5 dpi (); at the later time, the distribution is the same as in TTX-exposed neurons (). "Percent" indicates percentage of events having a given amplitude; the x axis is the absolute value of the mEPSC. D: kinetics do not change significantly with either TTX or NRSF expression. Traces show averages scaled to the NRSF 5 dpi average.

View larger version (42K): [in this window] [in a new window]   Fig. 8. Downregulation of the AMPA GluR2 subunit and the GABA-A 2 subunit can be measured electrophysiologically. A: GluR2-containing EPSCs, found in all cultured hippocampal neurons with pyramidal morphology, show linear or outwardly rectifying behavior in the presence of intracellular polyamines. Top: holding potential +40 mV; bottom: 40 mV. B: after 48 h of incubation in 200-500 nM TTX, EPSCs switch from outward to inward rectification. C: faster time scale of mean EPSC in control neurons, showing similar kinetics but differing amplitudes at the 2 potentials. Reversal potential is 0 mV (not shown). Mean rectification ratio, 1.1 ± 0.07, n = 9. D: faster time scale of mean EPSC in TTX-exposed neurons (mean ratio, 0.50 ± 0.03, n = 6). Determining the direct effects of NRSF on GluR2 expression is difficult under these conditions, and so the experiment was not performed for NRSF-infected neurons. E: in a control neuron, response to applied GABA (left) is faster and larger in the presence of 2 µM DZ (right; double lines indicate several washings to remove desensitization). F: although an NRSF neuron (5 dpi) shows smaller and slower evoked currents consistent with activity deprivation (data not shown), potentiation in response to diazepam (DZ) persists.

GABA HYPOEXCITABILITY IS NOT DUE TO DIRECT SUPPRESSION OF NRSF GENES. Currents evoked by direct application of GABA to infected and uninfected neurons show a significant GABAergic hypoexcitability. Evoked responses are extremely variable among cells, but a reduced peak and mean are seen in the NRSF neurons (NRSF, 210 ± 30 pA, range 30 to 510 pA, n = 14; control, 300 ± 50 pA, range 50 to 1,400 pA, n = 28; see also Fig. 8, E and F). Again, this is consistent with activity deprivation, but unlike the case of the AMPA receptor, we were able to exclude direct effects of NRSF on the transcription of receptor subunits. The GABA-A receptor 2 subunit is an NRSF target gene, and receptors with partial or complete repression of this subunit show altered responses to diazepam (DZ) (Maric et al. 1999). In one study, injection of antisense oligodeoxynucleotides to 2 reversed the modulation of DZ, from excitatory to inhibitory, in cortical neurons (Malatynska et al. 2000). However, in our study, all neurons 4 to 6 dpi that responded to GABA showed potentiation with 2 µM DZ. Peak GABA-evoked currents increased 32 ± 5% in control neurons (n = 9, range 10-100%), and 48 ± 15% in NRSF neurons (n = 7, range 15-185%; difference with controls not significant; Fig. 8, E and F). While the variability of these results is very large, the sign of the difference is an all-or-nothing effect that allows a qualitative conclusion to be drawn.

QUALITATIVE ALTERATIONS ARE SEEN WITH ALTERED CAMP LEVELS. Neurons to which the cAMP antagonist Rp-cAMPS is added along with NRSF virus show complete absence of Na⁺ currents at 2 dpi; there is no effect of this drug on uninfected or GFP-infected cells (Fig. 9A). Similar suppression is observed in neurons grown in serum-free culture conditions (not shown; n > 20 NRSF neurons showed no Na⁺ current, n > 20 control neurons showed normal Na⁺ current). No synaptic events are detectable (n > 10; n > 10 controls showed normal patterns of EPSCs), and cell death occurs within 3-4 days. BDNF (20 ng/ml) will prevent death but does not restore Na⁺ current (not shown; n > 10).

View larger version (28K): [in this window] [in a new window]   Fig. 9. A: Na+ current suppression is enhanced by lowering cAMP and reversed or prevented by raising cAMP. Currents are shown as fractions of the current density measured in matched drug-exposed controls. The cAMP antagonist Rp-cAMPS (100 µM) causes all measurable Na+ current to disappear by 2 dpi (Rp, , n = 5, error bar smaller than symbols). [Currents in Rp-cAMPS-exposed controls (n = 5) showed no decrease in Na+ current relative to unexposed controls, P > 0.5]. NRSF neurons at 6 dpi exposed to 0.5 mM cpt-cAMP because 4 dpi (cAMP6, , n = 13) show Na+ currents that are slightly but not significantly larger than in matched controls (n = 8). At 12 dpi, treated at 7 dpi after they had presumably lost all Na+ currents, cells retain significant Na+ currents (cAMP12, , n = 7), although they are significantly reduced to ~70% of the control currents (n = 9). B: a neuron grown in serum free culture conditions and infected with NRSF for 2 days shows reversal of DZ modulation (see RESULTS).

VIRAL ENHANCER DOES NOT AFFECT THE MAGNITUDE OR TIME COURSE OF THE RESULTS. It has been proposed that subtle changes in NRSF levels may be responsible for different neuronal phenotypes (Palm et al. 1998). In addition, the decrease in Na⁺ current we observe over time could be due to NRSF protein build-up as lentiviral expression levels have been shown to plateau after >1 wk (Nadeau et al. 2000). To test this hypothesis, we performed comparative sets of experiments with NRSF-CITE-GFP and a virus bearing the woodchuck hepatitis virus enhancer, NRSF-CITE-GFP-WRE. This 500-bp sequence increases lentiviral expression five- to eightfold (Zufferey et al. 1999) and led to remarkably enhanced GFP production in our experiments. GFP became visually detectable 24-48 h earlier with NRSF-CITE-GFP-WRE than with NRSF-CITE-GFP, and at 4 dpi showed a 2.4 ± 0.3-fold increase in whole cell fluorescence (n = 10 cells each). While NRSF and GFP are not a fusion protein in this construct, levels of retroviral expression of genes coupled through a CITE have been shown to be very tightly correlated (Liu et al. 2000; Nadeau and Lester, unpublished data).

NRSF is not significantly upregulated in TTX-exposed neurons

The similarity in physiology between NRSF-infected and TTX-exposed neurons provokes the hypothesis that NRSF plays a role in the response of normal cells to activity deprivation. Using immunohistochemistry, we were unable to detect an increase in NRSF protein levels in TTX-exposed neurons (data not shown). The monoclonal antibody used was raised to the N terminus of the protein and thus should be able to detect an increase in either full-length NRSF or truncated isoforms (Z. F. Chen et al. 1998).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

This work supports the general conclusion that genetic silencing of the Na⁺ channel represents a feasible strategy. The long-term survival, health, and lack of compensation in our neurons are notable. While suppression of Na⁺ channels does lead to synaptic changes and eventual reduction of K⁺ current, there is no corresponding restoration of Na⁺ channel function or action potential firing. Thus a transcriptional lack of NaCh II does in fact translate into a lack of Na⁺ current; while this may seem obvious, it is not at all the case for other neuronal excitability genes such as K⁺ channels.

It may also not prove to be true for other neuronal subtypes, such as those from the dorsal root ganglion, which express a wide variety of Na⁺ channels whose expression levels are sensitive to variations in input (Dib-Hajj et al. 1999). Furthermore, many types of neurons are more sensitive to activity blockade than those of the hippocampus. While mature hippocampal neurons tolerate a week of TTX exposure with little effect on morphology and survival (Kossel et al. 1997), other neurons die as a result of activity blockade, notably those from the spinal cord (Bergey et al. 1981) and retinal ganglion (Lipton 1986). Experiments with neurons from other brain regions are currently in progress.

However, for many in vitro or in vivo experiments in the hippocampus, NRSF may prove an ideal silencing gene. It especially lends itself to experiments in which a partial silencer is desired, since a significant reduction in Na⁺ current persists for 3-4 days without adverse effects on any other measured physiological parameters. After a week, Na⁺ current is completely suppressed, and full silencing may be studied for 5 days thereafter before death occurs. Studies in vivo will have to be performed to determine whether these parameters remain applicable, but it is likely that NRSF will prove at least as benign as it is in vitro, given the known sensitivity of cultured neurons to viral infection and alterations in excitability (Blomer et al. 1996; Ho et al. 1995; Slack and Miller 1996).

NRSF's effects are also prevented and/or reversed by a moderate rise in intracellular cAMP levels, one that has little physiological effect on normal cells. Thus at least in cultured systems, NRSF is reversible or inducible. This provides a useful alternative to viral vectors with inducible promoters, which demonstrate high background, repressor toxicity, or other problems (including difficulty of expressing >1 gene) (Kafri et al. 2000; Strathdee et al. 1999; Nadeau and Lester, unpublished results).

In cases where NRSF's effects on other genes are not of concern, manipulation of cAMP levels provides a method for making a "tunable" silencer. Concentrations of cAMP may be varied over time, creating different levels of silencing in infected populations of a dish, slice, or multielectrode array (Wong 1998).

The observed results are very different from what existing data lead one to predict. Because of its large number of target genes, NRSF might be expected to be "overkill" as a neuronal silencer, repressing dozens of neuron-specific functions. What is remarkable is that, in the presence of normal levels of cAMP, all direct effects of NRSF overexpression in mature hippocampal neurons could be attributed to suppression of NaCh II. Infected neurons show progressively smaller peak Na⁺ currents and a concurrent amplification of glutamate sensitivity and reduction of GABA sensitivity, which resemble those seen in response to activity blockade by TTX. The time course of the synaptic changes is consistent with their being a secondary effect, as they do not become significant until 5 days after infection.

Of course, other genes may be affected in ways that are not revealed by our experiments, although functionally significant downregulation of several NRSF target genes was excluded in cultures with normal intracellular cAMP. No dysfunction of synapses or altered morphology was seen, indicating little if any suppression of synapsin and tubulin; immunostains to these markers also showed no measurable decrease. Potentiation of GABA-evoked currents by DZ remained normal during NRSF expression. Direct effects on GluR2 were impossible to resolve because of demonstrated indirect suppression of this AMPA receptor subunit by activity deprivation.

Future experiments will show whether NRSF is in fact as clean of a silencer as it seems. Rapid silencing and lethality in freshly plated, differentiating neurons suggests a qualitatively different process occurring during development. NRSF is currently thought to consist of two repressor domains (Huang et al. 1999), where the N-terminal domain represses GluR2 and NaCh II by association with histone deacetylase 2 (HDAC); the C-terminal domain is HDAC independent but is also capable of repressing NaCh II (Andres et al. 1999). Thus in developing cells more genes may be accessible to repression; in fully developed cells, only those in permissive chromatin regions or those susceptible to C-terminal domain repression may show decreases. Further experiments to elucidate this may include expression of the N and C termini independently.

In mature neurons, levels of repression may be determined more by balance of NRSF isoforms than by HDAC association. Several isoforms of the NRSF protein exist, including a truncated form REST4 that contains five zinc fingers rather than the usual nine. This form acts as a dominant negative, able in the presence of PKA to de-repress the repression caused by full-length NRSF on the vesicular acetylcholine transporter (VAChT) gene in PC12 cells (Shimojo et al. 1999). Our observed dependence on cAMP may be a result of alternative splicing to produce REST4, occurring in response to full-length protein expression. If so, it is remarkable that the cells are able to counteract even the highest levels of viral overexpression. Detailed time courses and experiments with artificially transduced REST4 will further clarify this mechanism.

The phenotype changes qualitatively as cAMP levels fall. We have shown two results that provide evidence that NRSF is working at promoter sites other than the Na⁺ channel in cultures with lowered cAMP levels: lethality preventable by BDNF and altered DZ responses indicating 2 subunit downregulation. The known effects of BDNF on synaptic excitability (Rutherford et al. 1997) make it increasingly difficult to separate direct and indirect effects of NRSF.

Apart from identifying a potentially useful silencer, this study is the first to quantify physiological responses of living neurons to NRSF, and it raises many questions about the feedback loops involved in regulation of excitability in health and disease. Many of the genes shown to be regulated by activity and to influence excitability in turn, such as GluR2 (Brene et al. 2000), GABA-A receptor subunits (Penschuck et al. 1999), and BDNF (Rutherford et al. 1997), are also NRSF target genes. While our immunocytochemical methods do not detect an increase in NRSF in TTX-exposed neurons, this is does not necessarily mean that NRSF does not play a role in the phenotype of chronic activity blockade. Several studies have shown that efficacy of binding of NRSF to NRSEs is a more important variable than protein levels (Kojima et al. 2001; Lee et al. 2000; Yoo et al. 2001). NRSF levels in normal neurons may change little with activity, but the protein's effectiveness as a transcriptional regulator may be sensitively modulated by means of the cAMP-sensitive balance between NRSF and REST4 as well as by other posttranscriptional and/or posttranslational mechanisms yet unknown.

Expression of dominant negative REST4 in cultured neurons will be the first step toward determining whether NRSF plays a role in altered excitability states in the hippocampus. These experiments are currently in progress. Work with cell lines and/or transgenics will also enable effects on individual genes to be elucidated, as techniques such as RT-PCR and Western blot are extremely difficult to perform on primary neuronal cultures, which contain on the order of only 10⁴ neurons and a poorly understood mixed glial population. However, cell lines do not display the array of genes and feedback loops that are found in differentiated neurons, so cell line experiments will do little to identify secondary and tertiary mechanisms. When gene chip technology becomes readily available, it may help to deconvolve the many direct and indirect effects of this transcription factor.

A role may also be sought for NRSF in animal models of disease. When NRSF's function is up-regulated during seizures, cAMP is also high (Palmer et al. 1979), so the effects of the transcription factor will be limited and unlikely to lead to neuronal death. A reduction in cAMP levels, as occurs with anoxic injury, trauma, and perhaps Alzheimer's disease (Ohm et al. 1991) causes the suppression of a variety of NRSE-containing genes and possibly cell death. When pathology is severe, perhaps even endogenous NRSF levels will be sufficient to suppress gene function and lead to neuronal silencing and degeneration. Studying the relationships between NRSF levels, cAMP concentration, and pathologic states may reveal critical roles of this protein in excitotoxicity, neurodegeneration, or both. NRSF should be considered as a possible regulator of BDNF levels in many forms of neurotoxicity, especially in structures such as the hippocampus where NRSF levels are relatively high under normal conditions.