Although previous studies of Huntington's disease (HD) have addressed many potential mechanisms of striatal neuron dysfunction and death, it is also known, based on clinical findings, that cortical function is dramatically disrupted in HD. With respect to disease etiology, however, the specific molecular and neuronal circuit bases for the cortical effects of mutant huntingtin (htt) have remained largely unknown. In the present work, we studied the relationship between the molecular effects of mutant htt fragments in cortical cells and the corresponding behavior of cortical neuron microcircuits by using a novel cellular model of HD. We observed that a transcript-selective diminution in activity-dependent brain-derived neurotrophic factor (BDNF) expression preceded the onset of a synaptic connectivity deficit in ex vivo cortical networks, which manifested as decreased spontaneous collective burst-firing behavior measured by multielectrode array substrates. Decreased BDNF expression was determined to be a significant contributor to network-level dysfunction, as shown by the ability of exogenous BDNF to ameliorate cortical microcircuit burst firing. The molecular determinants of the dysregulation of activity-dependent BDNF expression by mutant htt seem to be distinct from previously elucidated mechanisms, because they do not involve known neuron-restrictive silencer factor/RE1-silencing transcription factor-regulated promoter sequences but instead result from dysregulation of BDNF exon IV and VI transcription. These data elucidate a novel HD-related deficit in BDNF gene regulation as a plausible mechanism of cortical neuron hypoconnectivity and cortical function deficits in HD. Moreover, the novel model paradigm established here is well suited to further mechanistic and drug screening research applications.
HD is an autosomal dominantly inherited neurodegenerative disorder resulting from a CAG repeat expansion mutation leading to the abnormal elongation of a polyglutamine tract in htt, a multifunctional cellular protein (Bates et al., 2002). Specific clinical manifestations of HD involve motor and cognitive impairments, including chorea, depression, and/or difficulties in decision-making. At the cellular level, HD seems to arise from neurotoxicity involving a state of cellular dysfunction preceding cell death. During this process, intracellular exposure to mutant htt and N-terminal htt fragments containing the expanded polyglutamine domain lead to protein aggregation, abnormalities in cellular signaling and trafficking, and the dysregulation of gene expression (Luthi-Carter, 2007).
Although it is established that HD brain neuropathology involves the striatum and cerebral cortex, the specific impact that mutant htt has on local versus network-level neurophysiology remains less clear. A great deal of emphasis has been placed on understanding changes in the behavior of corticostriatal synapses, focusing particularly on dysfunction and death of striatal cells (Cepeda et al., 2007). In contrast, specific electrophysiological abnormalities of cortical microcircuits arising from the presence of expanded polyglutamine proteins are much less well understood. The mechanisms underlying cortical dysfunction are of high interest because the cognitive and personality changes in HD could readily be explained by abnormal intracortical circuits. Motivated by this hypothesis, we set out to determine the relationship of molecular and synaptic changes in cortical microcircuits exposed to mutant htt fragments as a basis for inferring HD-related events.
Although many animal and cell line-based models of HD exist, there is a need for additional cellular models of HD in which mutant htt's effects on neuron-specific functions can be studied. Such models would be important for achieving a better mechanistic understanding of disease etiology and improving the prioritization of lead compounds and candidate drug targets identified in high-throughput screens (Heitz et al., 2008). Having noted the absence of a cellular model for simultaneous measurements of molecular and synaptic effects of mutant htt in cortical neurons, we established a new system to be able to address these issues. Here, we report results based on the implementation of a novel in vitro cortical neuron model of HD and the combined use of this model and extracellular multielectrode array (MEA) recordings. At the molecular level, the model allowed us to elucidate a novel activity-dependent aspect of mutant htt-related gene dysregulation. MEA recording data also demonstrated the model's utility in elucidating early disease-related changes in neuronal network behavior. We further show how this system might be used to identify pharmacologic agents that overcome specific deficits in neuronal connectivity.
Materials and Methods
Cortical cells from embryonic day 16 to 19 Wistar rat embryos (Charles River Laboratories, Les Oncins, France) were prepared and cultured as described previously (van Pelt et al., 2004). Cells were plated on MEAs and/or culture dishes, with prior surface coating by polyethylenimine (10 mg/ml; Fluka, Buchs, Switzerland) and laminin (0.02 mg/ml; Invitrogen, Carlsbad, CA) in neurobasal medium (MEAs) or poly-l-lysine (0.01%) in water (culture dishes). Plating density was 1500 cells/mm2 for MEAs, allowing for quick network formation and high multichannel electrophysiological recordings efficiency (L. Gambazzi, H. Markram, and M. Giugliano, unpublished observations). A lower density of 500 cells/mm2 was used for culture dishes to facilitate cell counting and morphological analysis. Medium containing Neurobasal, 2% B-27 supplement (Invitrogen), and 10% horse serum (Invitrogen) was changed three times per week by removing 0.7 ml and adding 1 ml of fresh medium. Where indicated, recombinant human BDNF (R&D Systems, Minneapolis, MN) at a concentration of 50 ng/ml was added to the medium twice a week starting from in vitro day 4. Although concentrations of exogenous BDNF treatment have varied between 2 and 100 ng/ml in the literature, a concentration of 50 ng/ml has been shown to be neuroprotective in a striatal lentiviral model of HD (Zala et al., 2005). Moreover, this concentration assures that the twice-weekly addition to the medium is sufficient for achieving a chronically increased level of BDNF. Recordings were performed at least 24 h after medium changes except as indicated (for experiments with the trkB inhibitor K252a).
Lentiviral Vector Production and Infection.
Plasmids encoding the first 171 amino acids of mutated htt (htt171-82Q) or wild-type htt (htt171-18Q) under the control of a tetracycline-regulated element promoter or the tetracycline-regulatable transactivator (tTA1) under the control of a phosphoglycerate kinase promoter were used to prepare self-inactivating lentiviral vectors (Fig. 1A) as described previously (Régulier et al., 2003). Lentiviral particles were resuspended in phosphate-buffered saline (PBS) + 1% bovine serum albumin, and the particle content of viral batches was assessed by p24 enzyme-linked immunosorbent assay (RETROtek; Gentaur, Paris, France). On the day after cell plating, cultures were infected with lentiviruses at ratios of 150 ng of p24/1000 cells (tTA1) and 120 ng of p24/1000 cells (htt171-82Q or htt171-18Q) (Fig. 1B). Although the use of full-length htt models may reproduce additional features of HD etiology, the use of N-terminal htt fragments has been an important and widely used strategy for approximating the effects of mutant htt.
Commercial MEAs with planar TiN substrate electrodes (Multi Channel Systems, Reutlingen, Germany) were used for ex vivo chronic multisite extracellular recordings. Substrate electrodes had a diameter of 30 μm (Fig. 1F) and were organized in an 8 × 8 square grid with 200-μm spacing. Cell maintenance and electrophysiological recordings were performed at 37°C, 9% O2, 5% CO2, and 65% relative humidity (Potter and DeMarse, 2001), inside a low-humidity incubator comprising an electronic-friendly environment (Jouan IG750; Thermo Fisher Scientific, Waltham, MA). MEA electrodes had an impedance of 100 kΩ (in PBS). Electronic amplifiers (Multi Channel Systems) had a standard gain of 61.6 dB, between 10 and 3 kHz (filters ± 60 dB/dec), and large input-impedance (1011 Ω in parallel to 10 pF).
MEA Data Acquisition and Analysis.
After 5 min of accommodation time after the mounting of each MEA inside the recording setup, spontaneous electrical activity from the cultured neurons was detected, amplified, and recorded at each of the 60 substrate extracellular electrodes simultaneously for 30 min at 20 kHz/channel. Recorded traces consisted of multiunit activity, directly related to the firing of action potentials by the neurons in the proximity of each electrode. MCRack software (Multi Channel Systems) was used to acquire and store the data (9 Gb/session, split into ∼2-Gb files), which were processed off-line channel by channel. Raw voltage waveforms (Fig. 1F) were digitally filtered between 150 and 2.5 kHz and fully rectified. The occurrence of an action potential at a given electrode was identified by a peak-detection algorithm, based on the crossing of an adaptive threshold, as in the LimAda algorithm. Recorded events included the time stamp of the spike, the index of the electrode where it was detected, and the 200 ms before and 800 ms after the corresponding raw voltage trace. The occurrence and duration of epochs of synchronized firing (i.e., bursts of spikes; see Fig. 1, F and G) were identified and detected as in van Pelt et al., 2004, by postprocessing the spike-time histograms (STHs). This made it possible to distinguish between epochs of asynchronous spiking activity and epochs of population-wide bursting. During each recording session, the overall number of bursts, the statistics of the interburst time intervals, and the number of individual spikes that composed each burst were used as observables of the network-level emerging spontaneous activity of the cultured neurons (Fig. 2 and see Fig. 5, A–D). Data are expressed as the mean ± S.E.M. Student's t test and the Kolmogorov-Smirnov test were used to assess the significance (p < 0.05) of differences in averages and cumulatives, respectively. Pairwise correlations between intraburst spike trains were evaluated as the peak amplitude of the spike-triggered cross-correlograms over a fixed time window (varied in the range 1 ms to 1 s).
The conclusions reported here are based on repeated observations collected by MEAs on 14 htt171-82Q- and 23 htt171-18Q-expressing (sister) cultures.
Immunostaining and Image Analysis.
Cultures were washed with PBS and fixed in 4% paraformaldehyde (Fluka) or methanol (Merck, Darmstadt, Germany) for PSD95 staining for 10 min at 4°C. Cultures were incubated with NeuN (1:500; Millipore Bioscience Research Reagents, Temecula, CA) or 2B4 (1:500; Millipore, Zug, Switzerland), PSD95 (1:200; Thermo Fisher Scientific), and neuronal class III β-tubulin Tuj1 (1:2000; Covance Research Products, Emerville, CA) antibodies in PBS containing 10% normal goat serum (Invitrogen) and 0.1% Triton X-100 (Fluka). Cultures were rinsed three times in PBS and then incubated for 1 h with fluorescent goat anti-mouse secondary antibodies (for NeuN, 1:1000 Cy3-conjugated antibody from Jackson ImmunoResearch Laboratories, Inc., West Grove, PA; for 2B4, 1:1000 Alexa Fluor 488-conjugated antibodies from Invitrogen; for Tuj1, 1:1000 Alexa Fluor 594-conjugated antibody from Invitrogen). Where indicated, cultures were stained with Hoechst 33342 dye (Invitrogen). Images of NeuN-labeled cultures (n = 5 or 6 per condition) were acquired by using a BD Pathway 855 microscope (BD Biosciences, San Jose, CA) under nonsaturating exposure conditions, using the same acquisition settings for all samples in a given experiment. Images of PSD95-, 2B4-, and Tuj3-stained cultures were acquired with a laser-scanning confocal microscope (TCS-SP2 AOBS; Leica, Wetzlar, Germany), using the same acquisition settings for all samples in a given experiment unless specifically noted otherwise. Counting of NeuN-positive nuclei and PSD95-positive punctae was performed with ImageJ software (National Institutes of Health, Bethesda, MD) applying intensity and size thresholds.
TrkB Inhibitor Experiments.
Experiments were initiated on cultures prepared on MEAs (n = 9) starting at 14 DIV. Spontaneous activity was recorded for 30 min in three sessions. The first comprised recordings of all cultures 5 h before initiating pharmacologic treatments. The second comprised recordings of all cultures 45 min after treatments with TrkB inhibitor K252a (Sigma-Aldrich, St. Louis, MO) at a concentration of 10 nM to three cultures, at 100 nM to three separate cultures, and maintaining three untreated cultures as controls. A third session recorded the behavior of all three groups of cells 1 day later (at 15 DIV). Analyses of MEA recordings were performed as indicated above.
Mathematical Model of a Cultured Neuron Network.
A simplified spike-rate model accounting for the patterned electrical activity emerging in populations of cultured neurons was defined and computer-simulated to interpret the electrophysiological recordings. We described the firing rate, R(t), of the ensemble of cortical neurons in terms of the single-cell f-I curve and recurrent connectivity (Giugliano et al., 2008; La Camera et al., 2008). The model replicates some of the features characterizing the population bursts as transient irregular outbursts in R(t). Full details and the source code are provided as a ModelDB entry (https://senselab.med.yale.edu/ModelDB; accession number 125748).
BDNF RNA Measurements.
Neuronal stimulation was performed 1.5 weeks after infection with lentiviruses. To reduce endogenous synaptic activity and prevent calcium entry through N-methyl-d-aspartate receptors, cortical neurons were pretreated for 30 min with 1 μM tetrodotoxin (Alexis Corporation, Lausen, Switzerland), 100 μM d-(−)-2-amino-5-phosphonopentanoic acid (Sigma-Aldrich), and 40 μM 6-cyano-7-nitroquinoxaline-2,3-dione (Sigma-Aldrich). To increase the specificity of the stimulations, neurons not subjected to stimulation were exposed to 10 μM nifedipine (Sigma-Aldrich) and 20 μM N-[2-(p-bromocinnamylamino)ethyl]-5-isoquinolinesulfonamide·2HCl (H-89) (Calbiochem, San Diego, CA). Where indicated, neurons were stimulated for 90 min with 10 μM forskolin (Calbiochem) and 30 mM KCl. Total RNA was extracted with a RNeasy kit (QIAGEN, Valencia, CA), and 800 ng of total RNA was used for cDNA synthesis using a high-capacity cDNA RT kit (Applied Biosystems, Foster City, CA). Quantitative real-time PCR was performed on a 7900HT Real-Time PCR System with SDS 2.3 software (Applied Biosystems) using Power SYBR green PCR master mix (Applied Biosystems). Relative expression (V) was calculated by normalization of BDNF to β-actin expression as described in Zucker et al., 2005. All quantitative PCRs were performed in triplicate. Primers used in this study were based on previous assays (Chen et al., 2003; Kobayashi et al., 2008) and are listed in Table 1.
To create a physiologically and molecularly tractable system for assessment, we used lentiviral gene transfer to develop a novel in vitro model of HD-related cortical dysfunction. This involved using HIV-1-derived vectors to express the first 171 amino acids of wild-type or mutated human htt under the control of a tetracycline response element-containing promoter (Rudinskiy et al., 2009; Fig. 1, A and B) to achieve high transduction efficiency and high levels of transgene expression in rat primary cortical neurons. The expression levels of huntingtin obtained with this approach are much higher than in transgenic mice but still lower than in acutely transfected cells or inducible cell lines. Immunodetection with an antipolyglutamine antibody (2B4; see Materials and Methods) showed that cultures expressing htt171-82Q accumulated the transprotein in the nuclear compartment (Fig. 1C) and developed neuritic htt inclusions (Fig. 1D) within 10 to 14 days after lentiviral transduction, whereas cultures expressing htt171-18Q exhibited only weak and diffuse, primarily non-nuclear, labeling (Fig. 1C). As assessed by the numbers of NeuN-positive nuclei, the neuronal viability in htt171-82Q-expressing cortical cultures showed no diminution compared with htt171-18Q-infected controls up to 3 weeks in vitro, but showed a progressive loss of neuronal cells at later time points (≥3.5 weeks) (Fig. 1E).
By replacing conventional cell culture dishes with glass-based arrays of 60 substrate microelectrodes (MEAs) (Marom and Shahaf, 2002; van Pelt et al., 2004), we monitored the effect of mutant htt fragment expression on cortical neuron network activity. Whereas htt171-82Q- and htt171-18Q-expressing cells were indistinguishable at early time points (≤2 weeks), we began to see disturbances in collective neuronal spiking activity in htt171-82Q-expressing cells between 2 and 3 weeks in vitro (Fig. 2). The fact that the collective firing behavior in htt171-82Q- and htt171-18Q-expressing cultures was similar initially suggests that the htt171-82Q protein does not have a major effect on ex vivo neuronal development. Direct inspection of raw MEA voltage recordings revealed a prominent temporal organization of spontaneous electrical activity in the form of bursts of spikes in htt171-82Q- and htt171-18Q-expressing cells (also in uninfected cells; see below). These properties are similar to those described for ex vivo neocortical preparations by other investigators (for review see Marom and Shahaf, 2002). After each recording session, single-unit activity was detected at each electrode by an adaptive threshold algorithm (see Materials and Methods). The occurrence of individual spike times for each MEA electrode was then represented as a raster plot and summarized as STHs (Figs. 1, F and G and 2A). These histograms estimate the instantaneous firing rate across all recording sites binned in constant time intervals (≤1 s). PBs were then identified as increased synchronous firing epochs and detected as peaks in the STH (van Pelt et al., 2004) (Fig. 2A). The PB duration was conventionally defined and estimated as the interval during which the instantaneous firing rate persisted above 5% of its local peak amplitude, computed by means of the raster plot for better temporal accuracy.
As expected, our system was able to detect mutant htt fragment effects on cortical neurons. The mean spike and PB numbers and the distribution of the interburst intervals (IBIs) differed significantly between htt171-18Q and htt171-82Q cultures (Fig. 2, A and B), with spikes and PBs being significantly less frequent in htt171-82Q neuronal networks (Fig. 2, B and D). In contrast, the average fraction of asynchronous (i.e., interburst) firing showed no difference between the two conditions (Fig. 2B). These observations suggest a specific impairment (or impairments) of network communication underlying PB ignition and not a change in cell excitability in the HD model. This stereotypical character of PBs was further confirmed in both the disease and control conditions by comparing the cumulative distributions of PB duration and the IBI (Fig. 2, C and D). In addition, the overall number of action potentials participating in each PB (i.e., the intraburst firing) revealed no significant difference between htt171-82Q- and htt171-18Q-expressing cells (Fig. 2E).
Taking into account the fact that htt171-82Q expression was not accompanied by cell loss or decreased intraburst firing, we interpreted the observed electrophysiological dysfunctions as not being the result of decreased spiking excitability or reduced electrode spike transduction. These interpretations were further supported by the fact that both synchronous/bursting and asynchronous modes of firing demonstrated similar features in the mutant and wild-type cultures (Fig. 2). The simplest interpretation of these observations was provided by computer-simulating a mathematical model of an ex vivo cortical network (see Materials and Methods). In this model, irregular PBs emerge as spontaneous transitions between a “resting” (i.e., asynchronous firing) state and an “excited”, transiently self-sustaining (i.e., synchronous firing) state (Giugliano et al., 2008). Although impairments in intrinsic excitability, such as the spike-frequency adaptation mediated by calcium-dependent outward currents, would interfere with PB by affecting its duration, the model predicts that the frequency (not the duration) of spontaneous PB occurrence depends on the recurrent excitatory connectivity and the efficacy of the synaptic coupling between neurons.
We also evaluated pair-wise correlations among spike trains as an indirect measure of the degree of functional connectivity in the network (Perkel et al., 1967). Assessed across all possible MEA electrode pairs, we found weak correlations between the firing activity in both htt171-82Q and htt171-18Q cultures. However, pairwise correlations were significantly smaller in the HD model than control (Fig. 2F), whereas no qualitative difference was observed when the (decreasing) relationship between correlation and interelectrode pair distance was evaluated (Fig. 2G).
Overall, the above modeling and recording results suggested that a discrete subset of electrophysiologic measures differentiate htt171-82Q- and htt171-18Q-expressing cells. Specifically, the rate of PB occurrence and the spike-train correlations are significantly decreased in htt171-82Q-expressing cells compared with htt171-18Q-expressing cells, consistent with a deficit in synaptic connectivity rather than a decrease in intrinsic excitability. This is also accompanied by a lower number of total spikes in the htt171-82Q-expressing cultures. In contrast, the spontaneous emergence of PBs [stereotypical of recurrent glutamatergic synaptic interactions in random networks (Marom and Shahaf, 2002)] and PB spreading to the entire network demonstrated similar characteristics in htt171-82Q and htt171-18Q cells. Likewise, the synaptic and/or cellular dynamics underlying PB termination (generally termed activity-dependent fatigue by intracellular ion accumulation or neurotransmitter ready-releasable pool exhaustion), was also similar in htt171-82Q and htt171-18Q cultures (as evidenced by the number of spikes/PB and the PB duration remaining unaffected).
We next considered the molecular mechanism of the electrophysiologic phenotype in the mutant htt-exposed cortical cells. One possible explanation for the observed decrease in PB firing behavior would be an underlying defect in synaptic reinforcement. Because BDNF is an important regulator of synaptic plasticity and loss of BDNF function has been previously implicated in HD, we assessed whether a deficiency of BDNF expression could explain the observed cortical microcircuit phenotype.
We first determined whether BDNF signaling through TrkB receptors was an important determinant of PB firing in our cortical neuron-based system. We addressed this question by assessing whether short-term application of the TrkB inhibitor K252a would suppress PB activity in the MEA cultures. In agreement with our expectations, short-term K252a treatment resulted in a dose- and time-dependent loss of PB firing (Fig. 3), without influencing neuronal viability (as determined by NeuN counts; data not shown). These results support the hypothesis that decreased BDNF activity might underlie the observed diminution of function observed in htt171-82Q cells.
To address this hypothesis, we first determined whether BDNF expression was altered in htt171-82Q-exposed neurons in a timeframe consistent with its driving the observed electrophysiologic abnormalities. Therefore, BDNF expression was measured in both resting and depolarizing conditions at time points preceding or coincident with altered network behavior. It is noteworthy that the earliest detected change was a deficit in activity-dependent induction of BDNF expression, which occurred at 10 days in vitro (Fig. 4A), i.e., before time points at which htt171-82Q- and htt171-18Q-expressing cells could be differentiated on the basis of PB firing. These results suggested that a deficit in activity-dependent BDNF gene regulation might be responsible for diminished cortical microcircuit activity in HD cells.
To determine whether BDNF levels could influence the network behavior in the predicted direction, we tested whether increasing BDNF availability could reverse the abnormal PB firing behavior observed in HD cells. Consistent with our hypothesis, adding exogenous BDNF to the culture medium normalized the PB frequencies and IBI intervals of htt171-82Q-exposed cells, leading to significantly increased PB frequencies and decreased IBIs compared with non-BDNF-treated sister cultures (Fig. 5). Indeed, the restoring effect on PB timing by BDNF was readily apparent (Fig. 5A) and extended to both the IBI (Fig. 5C) and duration distributions (Fig. 5B). Quantified in terms of numbers of individual action potentials, BDNF did not completely restore the size of individual PBs, however (Fig. 5D). Comparing samples of individual PBs recorded in treated and untreated htt171-82Q-exposed sister cells, their distribution is significantly shifted toward equally long PBs but slightly less populated by individual action potentials (see Figs. 2E and 5D). Taken together, these results support the idea that decreased activity-dependent BDNF expression is a mediator of the cortical microcircuit hypoconnectivity in HD-affected cells.
To further address the effect of impaired activity-dependent BDNF expression on synaptic connectivity in our model, we considered possible changes in the relative numbers of synapses in htt171-82Q cells (Fig. 5, E and F). These were assessed by quantifying numbers of PSD95-positive structures as an indicator of postsynaptic specializations. Indeed, htt171-82Q-expressing cells demonstrate a significant decrease in PSD95-positive bouton-like punctae, which is reversed by BDNF treatment. These results suggest that the structural organization of synapses is one parameter of cortical microcircuit connectivity that may be affected by BDNF availability in HD cells.
We subsequently considered the mechanism by which mutant htt fragments dysregulated activity-dependent BDNF expression. Previous studies have shown that increased NRSF/REST activity at the BDNF exon II promoter could mediate an HD-related decrease in BDNF expression (reviewed in Zuccato and Cattaneo, 2007). Furthermore, it has been shown that NRSF/REST can also regulate the transcription of BDNF exon I and regulation has been inferred for the transcription of BDNF transcript III (reviewed in Zuccato and Cattaneo, 2007). We therefore undertook exon-specific BDNF RNA analyses to determine whether an NRSF/REST-related mechanism could account for the decreased activity-dependent BDNF expression observed in our HD model system. We present these results adopting the rat BDNF gene nomenclature proposed by Aid et al. (2007). Although BDNF I, IIC, and III mRNAs showed an activity-dependent induction, the levels of these RNAs were not significantly different between htt171-82Q- and htt171-18Q-expressing cells (Fig. 4C). These results suggest that another mechanism accounts for polyglutamine-related differences in activity-dependent BDNF expression. Upon examining additional BDNF transcripts, we discovered that differences in activity-dependent BDNF gene regulation could be attributed specifically to the differential expression of BDNF IV and VI mRNAs (Fig. 4B). [It should be noted that transcript VI has also been referred to as transcript V in the rat, mouse, and human BDNF gene structure (see Pruunsild et al., 2007) as is the case in previous HD studies (Benn et al., 2008; Zuccato et al., 2008)]. These results indicate the existence of NRSF/REST-independent effects of mutant htt fragments may cause an early deficit in activity-dependent expression of BDNF RNA by cortical neurons. Moreover, this novel aspect of BDNF gene dysregulation meets the temporal and pharmacologic criteria to support its involvement in abnormal cortical microcircuit behavior.
Although the striatum has been more noted for its early degeneration in HD, the cerebral cortex also shows significant atrophy and has been implicated in striatal neuron death through evidence of increased glutamatergic tone, leading to excitotoxicity, and decreased neurotrophic support (Zuccato and Cattaneo, 2007). Likewise, corticocortical connections are also important for cortical network behavior and cortical output, and specific neuronal populations and the corpus callosum show significant atrophy in HD (Rosas et al., 2008, 2010; Thu et al., 2010). Therefore, studies of cortical neuron dysfunction have great potential relevance for both cortical and striatal effects of HD. In addition, synaptic effects of mutant htt are more easily studied in an excitatory neuron system where activity spontaneously arises without chemical or electrical intervention. The present model therefore comprises a rational and unique system for further analysis and dissection of HD-related effects on synaptic function and screening of synapse-targeted therapeutics.
The spontaneous activity emergence and transition from uncorrelated to clustered (i.e., PB) firing activity studied here is a very common feature of ex vivo neuronal networks (Van Den Pol et al., 1996). In the present work, we considered mutant huntingtin's effects at time points within the first 2 to 3 weeks in vitro because we aimed to study early network-level dysfunction in the absence of the confounding effects of cell loss. Because the spontaneous firing behavior of htt171-18Q and htt171-82Q cultures arises similarly at early time points and diverges subsequently, we believe that the polyQ-related differences in electrophysiological phenotype are not caused by effects on neuronal differentiation ex vivo.
The implication of abnormal glutamate neurotransmission in HD is a long-standing issue (Fan and Raymond, 2007). Thus, in some respects, it is surprising that mutant htt-exposed cortical neurons would show decreased rather than increased excitatory synaptic activity. This is in contrast to a simple excitotoxicity theory of HD-related neurodegeneration based on dysregulated cortical glutamate release or increased striatal responses to physiologically normal levels of transmitter. However, our results and previous reports of the lack of enhancement of excitotoxic phenomena in HD systems (Fan and Raymond, 2007) did not rule out that a more complex dysfunction of glutamatergic output from cortical neurons might result in excitotoxic sequellae in the cortex or striatum in vivo.
A tight relationship between glutamatergic neurotransmission and network-level spontaneous activity has been defined in cortical cultures previously (for a review see Marom and Shahaf, 2002). At the single-cell level, BDNF has been shown to acutely potentiate glutamatergic transmission by acting postsynaptically, for example, via regulation of glutamatergic receptor subunit trafficking (Nakata and Nakamura, 2007; Yoshii and Constantine-Paton, 2007). Our data are consistent with these previous observations by showing that BDNF is an important determinant of burst-firing behavior and regulates PSD-95-positive synaptic bouton number in primary cortical neurons.
Although BDNF normalized one major aspect of cortical microcircuit dysfunction in our HD model system (PB firing), it did not completely reverse polyQ htt effects. Specifically, the number of total spikes remained lower in htt171-82Q cells despite BDNF treatment. The lower spike number in htt171-82Q cells also persisted despite a BDNF-mediated increase in PSD95-positive boutons. The reduced spike occurrence frequency could be attributed to abnormalities in voltage-gated sodium channels (Oyama et al., 2006), intracellular calcium dynamics (Bezprozvanny, 2007), potassium channels (Hodges et al., 2006), and even toxic voltage-independent increased membrane permeability (Kagan et al., 2001), all of which comprise previously identified HD-related phenomena. Nevertheless, the stereotypical features of PBs (Fig. 2, B, C, and E), and the altered spike-timing cross-correlations (Fig. 2, F–G) are best explained by synapse-related dysfunctions.
Our observed behaviors of htt171-82Q-expressing neurons are also in line with previous observations of HD brain. On the molecular level, we and others have shown changes in the expression of synaptic proteins (Morton and Edwardson, 2001; Luthi-Carter et al., 2003; Hodges et al., 2006; Smith et al., 2007). Electrophysiologic data have also shown impairments in cortical function in HD mice (reviewed in Cepeda et al., 2007). Extracellular recordings in awake behaving animals also found significant reductions in the rate and synchrony of spontaneous burst firing in the cortices of HD compared with wild-type mice (Walker et al., 2008); moreover, the same study reported no change in burst duration. All of these findings are consistent with our observations. On the other hand, recordings in cortical slices from R6/2 mice by patch-clamp electrophysiology (Cummings et al., 2009) show a more complex profile of connectivity deficits. In the latter study, both excitatory and inhibitory synaptic projections onto cortical pyramidal neurons were defined during two stages of the disease: an asymptomatic stage (P21) and a stage considerably after emergence of overt behavioral symptoms (P80). At P80, Cummings et al. (2009) report seizure-prone cortical microcircuitry (characterized by increased efficacy, release probability, and connectivity probability in glutamatergic synapses and weaker efficacy, release probability, and connectivity probability in GABAergic synapses), rather the opposite of what we detected in vitro. However, in recordings before P21, the same group showed that the synaptic abnormalities are approximately reversed (down-regulated excitation and up-regulated inhibition). At first approximation, this early stage parallels our results and predictions in vitro; therefore, it is possible that our HD cortical model may best represent early-stage disease-related dysfunction. Another possible connection is provided by changes in BDNF expression (see below), which might explain the phenotype of later-stage mice as decreased BDNF-mediated facilitation of GABAergic neurotransmission (Hong et al., 2008; Sakata et al., 2009). Taken together, the present data suggest a complex dysregulation of neuronal firing behavior and circuitry in the HD cortex, with both BDNF-driven and other components.
The present study provides new evidence that decreased BDNF expression, particularly activity-dependent BDNF expression, is an early and relevant event underlying PB dysfunction. HD-related decreases in BDNF expression and trafficking have been explored in previous work (for a review see Zuccato and Cattaneo, 2007). The effect of another polyglutamine disease protein, mutant atrophin-1, on depolarization-dependent BDNF promoter activity has also been explored in an in vitro system (Miyashita et al., 2005). The present data are consistent with previous reports, yet several aspects of the present work extend the current understanding in novel and potentially important ways. The first is to elucidate and emphasize the important effects that BDNF deprivation may have on cortical microcircuits (rather than striatal neuron survival) in HD. The second is to show that the activity-dependent component of BDNF expression is affected by mutant htt. The third is to provide evidence that the earliest deficit in BDNF gene dysregulation occurs via a non-REST/NRSF-mediated mechanism (a finding that is paralleled by previous exon-specific BDNF transcript analyses, which show deficiencies in the orthologous BDNF transcripts in R6/2 mouse and human HD brains, which are caused by differences in nomenclature conventions, referred to as transcripts IV and V in these studies (Benn et al., 2008; Zuccato et al., 2008).
Although mechanisms of activity-dependent transcription from the exon VI promoter have not been extensively studied, several mechanisms have been implicated in the activity-dependent transcription of exon IV (for a recent review see Greer and Greenberg, 2008). These include the phosphorylation and DNA binding of the CRE-binding protein, the recruitment of CRE-binding protein-associated cofactors including CBP and TORC1, the phosphorylation-mediated release of transcriptional repression imparted by the methylated DNA binding protein MeCP2, direct promoter DNA demethylation, and positive regulatory activity of MEF2C. Given the important role attributed to the activity-dependent component of BDNF expression, these known mechanisms of exon IV regulation pose hypotheses meriting evaluation in HD.
Beyond understanding specific mechanisms of htt toxicity, there is also a pressing need for HD models that can be used to rapidly and accurately screen candidate drugs capable of rescuing neuronal function. The majority of currently used in vitro HD models are designed to examine cell survival or protein accumulation abnormalities, which may represent late stages of disease or be of uncertain relevance to addressing clinically meaningful endpoints. Moreover, the implementation of cortical cells for this assay not only facilitates the study of patterned network activity (which arises spontaneously in excitatory neurons), but also provides a system complementary to striatal cells in which to examine and potentially remediate mutant htt's effects. Although animal models of HD may be more faithful to the complex cellular and network relationships underlying human disease, their throughput rate for drug screening is low. Here, we present the possible alternative implementation of an in vitro HD model coupled to MEAs as a simple way to probe neuronal network function upstream of neuronal cell death. Electrophysiological MEA recordings seem ideally implemented for secondary screening purposes, comprising a rapid and repeatable longitudinal assay. Here we show proof-of-concept evidence that our system can successfully detect a disease-modifying effect [whose therapeutic benefit has been shown previously in transgenic HD mice (Cepeda et al., 2004; Lynch et al., 2007; Gharami et al., 2008; Giralt et al., 2009)]. In a broader context, one can also envisage the utility of MEA-integrated disease models for studying and developing therapies for a wide range of neurological conditions.
We thank Drs. M. Pignatelli, G. La Camera, and J.-R. Cardinaux for helpful discussions; Drs. J. Van Pelt, B. R. Miller, W. L. C. Rutten, A. Tscherter, J. Streit, M Chiappalone, V. Pasquale, and S. Martinoia for critical comments on an earlier version of the manuscript; and M. de Fatima Rey, S. Garcia, and K. Antoniello for excellent technical support.
This work was supported by the European Sixth Framework Program NeuroNano [Grant NMP4-CT-2006-031847], the European Sixth Framework Program Early-Stage Researchers Training Program in Nervous System Repair, the Telethon Action Suisse, a Commission for Technology and Innovation award from the Swiss National Science Foundation, the Ecole Polytechnique Fédérale de Lausanne, the University of Antwerp (Nieuwe Onderzoeks Initiatieven Bijzonder Onderzoeksfonds), and the Belgian Interuniversity Attraction Pole.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- Huntington's disease
- brain-derived neurotrophic factor
- population burst
- interburst interval
- multielectrode array
- spike-time histogram
- phosphate-buffered saline
- neuron-restrictive silencer factor
- RE1-silencing transcription factor
- days in vitro
- tetracycline-regulatable transactivator
- polymerase chain reaction
- postnatal day.
- Received February 22, 2010.
- Accepted July 12, 2010.
- Copyright © 2010 by The American Society for Pharmacology and Experimental Therapeutics