Heterogeneous spatial patterns of long-term potentiation in rat hippocampal slices
- 1Department of Physiology and Biophysics Program, University of Wisconsin - Madison, 1300 University Ave, Madison WI 53706, USA
- Corresponding author M. B. Jackson: Department of Physiology, University of Wisconsin - Madison, 1300 University Ave, Madison, WI 53706, USA. Email: mjackson{at}physiology.wisc.edu
Abstract
Although LTP (long-term potentiation) of synaptic transmission has received much attention as a model for learning and memory, its function within a neural circuit context remains poorly understood. To monitor LTP over an extensive circuit, we imaged responses in hippocampal slices using a voltage-sensitive dye. Following theta-burst stimulation, evoked optical signals showed an increase that lasted 40 min or more. Weak stimuli only potentiated the local area around the stimulating electrode, but stronger stimuli induced LTP over a wide area with a complex and non-uniform spatial pattern. The expression of LTP showed distinct peaks and valleys that depended on which axons were activated. Interestingly, the spatial distribution of LTP bore no relation to the spatial distribution of single-shock responses, but closely resembled the distribution of postsynaptic spikes evoked by theta bursts. Thus, postsynaptic spikes during induction constitute a critical determinant for the expression of LTP in intact circuits.
Long-term potentiation (LTP) of synaptic transmission has served as a valuable model in the investigation of the cellular basis of learning and memory (Bliss & Collingridge, 1993; Malenka & Bear, 2004). This form of synaptic plasticity has received a great deal of attention for more than three decades, but most research has focused on molecular and cellular mechanisms. Understanding how LTP contributes to learning and memory requires placing it within the context of neural networks and circuits. To do this one must follow the electrical activity of large numbers of cells simultaneously, and this task poses a serious challenge with currently available electrical recording techniques. Voltage imaging offers a means of studying functional activity broadly over a neural circuit (Grinvald et al. 1982; Salzberg, 1983; Grinvald et al. 1988; Bonhoeffer et al. 1989; Jackson & Scharfman, 1996; Nakagami et al. 1997; Tominaga et al. 2000; Jin et al. 2002), and thus provides a way to study LTP at a higher level of organization. This approach reveals the complexity of electrical activity in brain slices, and the changes in electrical activity during LTP can be detected with optical techniques (Saggau et al. 1986; Momose-Sato et al. 1999; Hosakawa et al. 2003; Aihara et al. 2005). In order to explore the spatial aspects of LTP expression in greater detail, and to map the expression of LTP within a functional circuit, we have initiated an imaging study of LTP in rat hippocampal slices using the voltage-sensitive absorbance dye RH482.
Different synaptic pathways generally have different forms of plasticity with fundamentally different properties. For a given pathway within a particular region, one might expect synapses to exhibit uniform behaviour. This hypothesis has yet to be thoroughly tested, but recordings from different locations indicated that LTP induced by high-frequency stimulation in the stratum radiatum (SR) tended to be much weaker near the stimulation site (Kauer, 1999). This could reflect the local action of inhibitory interneurons, which are activated together with the long-range excitatory Schaffer collaterals and commissural fibres. Voltage imaging can reveal spatial heterogeneity over a wide area, and thus provide insight into how hippocampal circuitry regulates the expression of activity-induced changes in function. The present study used voltage imaging to investigate the spatial distribution of LTP in the CA1 region of the hippocampal slice, and to relate this distribution to the spread of spikes and EPSPs evoked by electrical stimulation. We found that LTP expression varied within a slice. This spatial heterogeneity bears little resemblance to the distribution of responses to single shocks, but depends on which pathway was activated and on whether the induction stimulus elicited postsynaptic spikes.
Methods
Slice preparation
Male Sprague-Dawley rats (3–5 weeks old) were rendered unconscious by elevating levels of CO2, and decapitated, in accordance with the guidelines of the National Institutes of Health, as approved by the Animal Care and Use Committee of the University of Wisconsin - Madison. The brains were then removed and chilled in ice-cold cutting solution consisting of (mm): 124 NaCl, 3.2 KCl, 26 NaHCO3, 1.25 NaH2PO4, 1 CaCl2, 6 MgSO4, and 10 glucose, bubbled with 95% O2–5% CO2. Horizontal hippocampal slices 350–400 μm thick were cut with a tissue slicer (HR2, Sigmann Elektronik, Germany), selecting sections midway along the dorsal–ventral axis. A cut at the CA2 region was made to prevent epileptiform activity. Slices were stored in a vial of bubbled artificial cerebrospinal fluid (aCSF) at 29°C for 1 h. aCSF was identical to the cutting solution but contained 2.5 mm CaCl2 and 1.3 mm MgSO4. Slices were stained with bubbled aCSF containing 0.02 mg ml−1 of the voltage-sensitive absorbance dye RH482 (NK3630, Hayashibara Biochemical Laboratories, Okayama, Japan) (Momose-Sato et al. 1999; Chang & Jackson, 2003) for 1 h at 29°C, and returned to aCSF for later use. During recording, hippocampal slices were continuously perfused with bubbled aCSF in a submerged recording chamber and maintained at 29–31°C. (±)-2-amino-5-phosphonovaleric acid (APV) was applied by bath application in some experiments to test the role of N-methyl-d-aspartic acid (NMDA) receptors. All chemicals (other than RH482) were purchased from Sigma.
Electrophysiology
Stimulating and field potential (FP) recording electrodes were made from borosilicate glass capillaries (1.15 mm ID, 1.50 mm OD) filled with aCSF. Stimulating electrodes had 10–30 μm tips; recording electrodes had resistances of 2–5 MΩ. Slices were stimulated with 200 μs current pulses delivered by a monopolar constant-current stimulus isolator (Model A365, World Precision Instruments, Sarasota, FL, USA) to the SR every 30 s. The recording electrode was placed in the middle of the SR, 250–400 μm away from the stimulating electrode. A current pulse was applied 50 ms after the beginning of data acquisition. The theta burst stimulation (TBS) protocol used to induce LTP included three theta bursts (TB) separated by 10 s. Each TB consisted of 10 bursts at 5 Hz and each burst consisted of six pulses at 100 Hz. This induction protocol approximates that used to induce LTP 2, which has been shown to depend on IP3 receptor-sensitive Ca2+ stores (Raymond & Redman, 2006). Field potentials were recorded with an Axopatch 200B amplifier (Axon Instruments).
Voltage imaging and data analysis
The instrumentation follows Wu & Cohen (1993) and is very similar to the commercial photodiode array system Neuroplex-II from RedShirtImaging LLC (Fairfield, CT, USA) (Jackson & Scharfman, 1996; Demir et al. 1998). A set of 464 optical fibres was bundled into a hexagonal array, and each fibre was coupled to a Hamamatsu photodiode. Signals were amplified to 5 V nA−1 of photocurrent, low-pass filtered with a four-pole Bessel filter at 500 Hz, multiplexed, and digitized at a frame rate of 1.63 kHz with eight expansion boards (MSXB 002-02, Microstar Laboratories, Bellevue, WA, USA) and a DAP3200e/214 data acquisition board. Data were acquired in 500 ms segments, and traces were displayed as either single trials or averages of up to four trials.
An upright Reichert Jung Diastar microscope (Leica, Deerfield, IL, USA) was illuminated from below with a 100 W tungsten–halogen bulb driven by a Kepco ATE 36-30 DM power supply (Flushing, NY, USA). Illuminating light passed through a 702 ± 18 nm bandpass filter (Omega, Brattleboro, VT, USA), and transmitted light was collected with a 10 × Olympus UPlanApo objective (NA = 0.40). The centre-to-centre distance between neighbouring photodiode fields was 67 μm. An electronically controlled shutter was opened 200 ms before the start of data acquisition. Optical signals ΔI/I (where I is recorded light intensity) were presented as the change in transmitted light divided by the resting light intensity I/I measured before LTP induction. A movable mirror directed images to a CCD camera; images were captured with a frame grabber (Data Translation, Marlboro, MA, USA).
Data acquisition, signal processing and analysis were performed with a Pentium 4 computer running a program written in C++. Additional analysis and plotting were performed with the computer program Origin (Microcal, Northampton, MA, USA). We used a 3-point binomial temporal filter, in which signal amplitude S(t) was replaced by S(t − 1)/4 + S(t)/2 + S(t +1)/4, to reduce noise in the traces. The baseline drift of each signal was corrected by fitting the baseline to a third-order polynomial and subtracting this function; the baseline was taken as the time before the stimulus and 150 ms after the stimulus. Stimulus artifacts were removed from displayed FP recordings. Peak signals were used to create colour-encoded maps of responses and of LTP. The colours in pixels between centres of photodiodes were linearly interpolated. A Gaussian spatial filter with a space constant of half the interphotodiode distance reduced noise in these maps. LTP was defined as the increase in peak responses of optical signals. LTP data were only used if the pre-induction baseline remained stable for 20 min or more.
Analysis of optical EPSPs
The optical signals reported here generally consisted of EPSPs and spikes. In order to assess the magnitudes of these two
contributions, we fitted the EPSP component of an optical signal (oEPSP) to a computer-simulated function. The optical signal,
ΔI/I, is proportional to the membrane potential change within the physiological range of voltages (Wu & Cohen, 1993). Therefore:
where ΔVm is the deviation of the membrane potential from rest, averaged over all the cell membrane at a particular location, and β
is a constant. The membrane potential, Vm(t), is then determined by numerical integration of the differential equation:
where Cm is the membrane capacitance, gleak is the membrane leakage conductance, Eleak is the reversal potential of the leakage conductance (−65 mV), gsyn is the synaptic conductance, and Esyn is the reversal potential of the synaptic conductance (0 mV). The synaptic conductance was modelled as an α-function.
where t is time, and Gsyn and τα are free parameters. The simulated EPSP depends on τα, the product of β and Gsyn, and the ratios of Cm to gleak and of Gsyn to gleak. Varying Gsyn/gleak in the range of 0.01–20.0 has little impact on the time course of Vm(t), so we fixed it at a constant value of 1.0. The best fit was determined by varying τα, β×Gsyn, and Cm/gleak to minimize the root-mean-square error (see Fig. 1E). A narrow window was set around the spike to exclude this part of the record from the fit of the oEPSP. The spike amplitude
was taken as the peak amplitude minus the amplitude of the oEPSP at that time point.
Results
Spatial heterogeneity of LTP
Stimulation electrodes were positioned in the SR as indicated in Fig. 1A. A single pulse evokes optical signals over a broad area within the CA1 region of a slice (Fig. 1B), with the largest responses in the SR and stratum oriens (SO) (Fig. 1B and Ci). Although the voltage-sensitive dye stains the membranes of pyramidal cells, interneurons and glia cells, we attribute the optical signals primarily to voltage changes in pyramidal cells for the following reasons. In every instance optical signals were consistent with FP recordings from the same site. Figure 1E shows an example (compare FP with site 1), and this general parallel held in a large number of experiments. FP recordings in different regions of hippocampal slices are widely interpreted in terms of voltage changes in pyramidal cells (Andersen et al. 1980a). Interneurons are not expected to contribute significantly because of their lower density; pyramidal cells outnumber interneurons by an estimated ratio of 10 : 1 (Traub et al. 1999). Glia cells are numerous, but the dye we used, RH482, preferentially stains neurons over glia cells (Konnerth et al. 1987; Kojima et al. 1999). Furthermore, evoked responses in hippocampal glia cells have much smaller amplitudes and slower time courses as demonstrated by patch clamp recording (Bergles & Jahr, 1997), field potential recording (Diamond et al. 1998), and voltage imaging (Kojima et al. 1999).
Stimulation of the SR with theta bursts induced enduring increases in the magnitudes of both optical signals and FPs. Figure 1D shows plots of the time course of the slope of the field EPSP (fEPSP) and of the optical signals at four different locations (as pictured in Fig. 1A). After TBS, the amplitudes of the optical signals at all four locations increased, reaching a plateau within 10 min, and then remaining above baseline for the duration of the experiment (50 min), although the optical signal at the stimulation electrode (SE – location 3) declined slightly before stabilizing. The fEPSP slope and the optical signal at the same site both showed potentiation (location 1). Thus, the voltage-sensitive dye provides a signal that allows us to evaluate LTP throughout a slice.
Figure 1Ci shows a colour-encoded intensity map of the initial peak response to a single pulse, and Fig. 1Cii shows the map of peak responses to the same stimulus current 50 min after TBS. The post-TBS intensity map clearly shows larger responses extending over a wider area. Subtracting the peak intensity before TBS from that after TBS renders a map of LTP (Fig. 1Ciii). This map shows a spatially complex distribution, with seven distinct foci (white arrows). Figure 1Ci–Ciii illustrate a major finding of this study, that LTP shows a complex non-uniform spatial distribution, which bears little resemblance to the distribution of single-pulse responses.
The foci of strong LTP in the spatial map of Fig. 1Ciii can be highlighted by plotting the peak response amplitude along contours through a slice. A plot of response amplitude along the pyramidal cell somato-dendritic axis before TBS showed that responses were strongest in the SR (Fig. 2A, top, open squares). Responses in the SO were weaker, and the stratum lacunosum-moleculare (SLM) had the weakest responses. The weak responses in the SP cannot be compared with those in neighbouring SO and SR because the cell body layer has a lower density of plasma membrane, and the density of dye is therefore also lower. With a smaller fraction of light absorbed by dye the optical signals emanating from the SP will be reduced (Chang & Jackson, 2003), but the extent of this reduction is difficult to quantify. After TBS the responses grew over the entire contour (Fig. 2A, top, filled circles). Taking the difference between the pre- and post-TBS contours (Fig. 2A, below, filled triangles) showed that LTP was greatest in the SO, smaller but pronounced in the SR, and barely visible in the SLM.
Plotting responses along pathways perpendicular to that taken for Fig. 2A (i.e. through the SR or SO; contours were selected to pass through the LTP maxima in Fig. 1Ciii) revealed the spatial variations within layers (Fig. 2B and C). The profile of responses through the SR before TBS revealed a bell-shaped curve with a broad peak centred at the stimulating electrode (Fig. 2B, top, −2 min), and a plot along a contour through the SO showed the same qualitative behaviour (Fig. 2C, top, −2 min). As in the maps of Fig. 1Ci–iii, the profile of LTP expression along these contours (post- minus pre-induction; Fig. 2B and C, filled triangles below) did not resemble the profile of single-shock response amplitudes. These difference plots showed peaks where the amplitude plots were featureless, and these peaks corresponded well with the foci in the spatial map of LTP in Fig. 1Ciii (white arrows). In this experiment the SO contained four peaks (Fig. 2C, bottom), and the SR contained three (Fig. 2B). Thus, the spatial distribution of LTP exhibits a heterogeneity that was not evident in the distribution of responses to shocks, and this qualitative behaviour was seen in experiments on 43 slices.
LTP of optical signals depends on NMDA receptors
LTP detected optically depended on the activation of NMDA receptors. In the presence of the NMDA receptor antagonist APV (50 μm), electrical stimulation evoked large responses that spread over considerable distances (Fig. 3Aii), but TBS produced no LTP (Fig. 3Aiii). APV was then removed and the same TBS protocol induced LTP (Fig. 3Aiv). Figure 3B illustrates these results by plotting the time course of responses at two locations. Similar results were obtained in seven slices, and these experiments demonstrate that the LTP observed with voltage imaging shows the expected dependence on NMDA receptors.
Effect of stimulation intensity on the spatial distribution of LTP
Different stimulus currents produced qualitatively different spatial patterns of LTP expression. With 20 μA (n = 6), only the area around the SE expressed LTP (Fig. 4A, right), and this simple spatial pattern was seen in all six slices tested with this weak stimulus. The larger potentiation in the SO appeared primarily in the spike component (data presented below), indicating that neurons close to the SE became suprathreshold for action potentials after TBS. Similar results were seen with 40 μA (n = 7). LTP expression remained confined to the region around the SE, but the area was wider (Fig. 4B, right).
With 60 μA (n = 8), the initial pre-TBS responses in regions up to about 300 μm along a contour through the SO showed large responses (Fig. 4C, middle, green area in the SO). After TBS, the signals from these regions did not increase as much as more distant regions, implying that the responses within 300 μm were near saturation. As a result, more potentiation occurred at greater distances from the SE (Fig. 4C, right). (The issue of saturation will be examined in more detail below.)
With 80 μA (n = 9), TBS produced a heterogeneous spatial pattern of LTP expression (Fig. 4D, right). Regions showing more LTP generally were found within a few hundred microns of the SE or near the CA2 region, which was usually ∼1 mm from the SE. Thus, increasing the stimulus intensity did not simply enlarge the area of LTP expression, but increased the complexity of the LTP map and changed the qualitative appearance of the potentiated regions. For all stimulation intensities tested, areas with strong responses (Fig. 4, middle panels) and strong LTP expression (Fig. 4, right panels) were usually confined to the SR and the SO. The SLM showed weak responses and essentially no LTP.
Counting the number of peaks in the maps of LTP and plotting this number versus stimulus current shows the increase in the spatial heterogeneity of LTP (Fig. 4E). Peaks in LTP maps were generally taken as regions of at least three photodiodes where the peak signal was at least 6% higher than that of surrounding photodiodes. The total number of peaks increased with stimulus strength and reached a plateau at about five. (The dependence on stimulus was significant by one-way ANOVA; P < 0.005.) The increase in number of peaks was evident in both the SO and SR. Peaks in the SO and SR were generally in register (i.e. at mirror locations on either side of the SP), and this parallel has a simple explanation that neurons expressing strong LTP in the SR usually generated action potentials that invaded the basal dendrites in the SO.
Contribution of spikes to LTP
Spikes and EPSPs both contribute to optical signals, and the appearance of a rapid spike on top of the broader oEPSP corresponds closely with the appearance of the population spike inflection in a FP recording from the same site (Fig. 1E, compare FP top traces with optical traces from location 1). Spikes in the dendritic regions are generally somewhat broader than in the cell body layer, and the durations of dendritic spikes in our recordings of 3–4 ms are very similar to those recorded by patch clamping dendrites (Magee & Johnston, 1995; Spruston et al. 1995; Magee, 1999). By decomposing optical signals into these two basic components (see Methods) we can evaluate their roles in the LTP observed at different locations. (It should be noted that the spikes are postsynaptic as they always appear on top of the oEPSP. In the SR we can usually see a presynaptic axonal volley just before the onset of the oEPSP; data not shown.) To examine the contributions of spikes and oEPSPs we selected representative sites for detailed analysis from five experiments in which LTP was induced with 60 μA, a stimulus intensity strong enough to induce a spatially varied pattern of LTP (Fig. 4). In the SR, we chose one location at the SE, a second location 270 μm away (4 × the photodiode spacing of our detector), and a third location where LTP expression was maximal. These three locations were complemented by three locations in the SO directly across the SP from the locations selected in the SR.
For each location in the SR (Fig. 5A) and the SO (Fig. 5B), the overall signal, the spike component, and oEPSP component all increased significantly, except for the EPSP at the stimulation site in the SR. The changes in each parameter associated with TBS-induced LTP are shown in Fig. 5C and D. In general, spikes increased more than oEPSPs at locations expressing the greatest increase in overall signal. These results indicate that either the small changes in oEPSP amplitude bring a large number of cells above the threshold for action potentials, or that a change in dendritic excitably also contributes to the increase in spike amplitude (Daoudal et al. 2002; Wang et al. 2003; Frick et al. 2004; Fan et al. 2005). Conversely, at locations where LTP was weaker, the differences between the amplitude increases of spikes and oEPSPs were smaller and not significant. These results reflect a general trend that at sites showing the greatest LTP, spikes made the largest contribution.
Saturation of responses near the stimulating electrode
One form of spatial variation consistently seen with strong stimulation intensities was suppression of LTP near the stimulating electrode. To determine if this was due to saturation of responses, we varied the stimulation intensity and examined the responses near the SE and 270 μm away, in the SR and SO. Near the SE in the SR, the total response was close to saturation with stimulation intensities above 50 μA, while responses at the other three locations continued to increase with stronger stimulation (Fig. 6A). Decomposing responses into spikes and oEPSPs showed that both components increased with stimulus current, and reached a plateau as the responses saturated. The response curves of spikes were sigmoidal (Fig. 6B), reflecting the cooperative activation of action potentials by synaptic inputs. By contrast, the stimulus–response curves of oEPSPs were hyperbolic, with no sigmoidal character (Fig. 6C). Both oEPSPs and spikes near the SE saturated at lower stimulus currents than at more distant locations. Thus, near the site of stimulation, strong stimulus currents produce saturating responses that appear to occlude LTP, but 270 μm away the less than saturating responses could still be potentiated. This may be relevant to aspect of the spatial heterogeneity in the LTP maps such as Figs 1Ciii, and 4C and D, but leaves unexplained the peaks and valleys seen at greater distances from the site of stimulation.
Pathway specificity of LTP
The spatial heterogeneity of LTP could arise from the specific circuitry within a hippocampal slice or from a subtle spatial variation in the condition of the tissue. The uniformity in condition is a concern for work in a brain slice, where tissue recovers from the trauma of cutting. To address this possibility, we performed experiments with two stimulating electrodes. In these experiments, SE1 (red arrow in Fig. 7) was in the inner third of the SR (one third the distance between the SP and border with SLM), and near the subiculum. We placed SE2 (yellow arrow in Fig. 7) 600 μm away in the SR, from half-way to two-thirds of the distance to the SLM, and close to the CA2 region (Fig. 7A). In view of the parallel trajectories of SR axons (Andersen et al. 1980a), this positioning insured the stimulation of different pathways. Stimuli were applied alternately to SE1 and SE2 every 30 s (50 μA) and Fig. 7Bi and Ci display the spatial distributions of these two responses. TBS applied first to SE1 potentiated responses evoked by SE1 but not responses evoked by SE2 (Fig. 7Bii and Cii). One hour after the TBS at SE1, TBS applied to SE2 potentiated responses at sites where potentiation by SE1 failed. Figure 7Ciii displays the map of LTP induced by SE2, and Fig. 7Biii shows that responses evoked by SE1 were unchanged. The time course of responses illustrates these results by showing a location where the response to SE1 was potentiated by the first TBS at SE1 but not by the second TBS at SE2 (Fig. 7D, location 1). Likewise, responses at a second location show no change after the first TBS but were potentiated by the second (Fig. 7D, location 2). Thus, LTP expression was pathway-specific, with no detectable increases in responses evoked by stimulation at a site different from the site to which TBS was applied.
To evaluate these results systematically, we mapped areas in which LTP was within 50% of the maximal amount for a given experiment (green to red regions in Fig. 7Bii and Ciii). These areas showed no overlap in this case, and in a total of 19 experiments, the spatial distributions of LTP strongly overlapped in only 2 slices, partially overlapped in 10 slices, and failed to overlap at all in the remaining 7 slices. The fact that we can readily see LTP in many locations by applying TBS at one site, where TBS at another site failed to induce LTP, supports the argument for differences in functional circuitry rather than spatial heterogeneity in the health of slices as the underlying cause of spatial heterogeneity of LTP. It is important to note that in all of these experiments, the regions activated by the two electrodes prior to TBS always showed significant overlap. Thus, as in the comparison of the spatial distribution of LTP with the spatial distribution of responses to stimulation at a single site (Figs 1 and 2), the regions where responses were potentiated by TBS do not simply mirror the distribution of response amplitudes.
Spikes during TBS predict the spatial extent of LTP
Postsynaptic depolarization serves as a critical signal in the induction of LTP and postsynaptic action potentials play an essential role in providing this depolarization (Magee & Johnston, 1995; Spruston et al. 1995; Bi & Poo, 1998; Linden, 1999; Bi & Poo, 2001). We therefore examined the relation between the spatial distribution of LTP and the spatial distribution of spikes elicited by the TBS during induction. Figure 8 shows the results for two locations that failed to express LTP in response to TBS (location 1 in the SR and 2 in the SO) and two locations that showed potentiation (location 3 in the SR and 4 in the SO). Superimposed signals recorded 2 min before and 1 h after induction show how TBS affects the responses from each of these locations (Fig. 8B). The FP and the optical signal recorded at the same location did not change after TBS (Fig. 8B and D, FP and location 1). Signals at the adjacent location in the SO also did not change (Fig. 8B and D, location 2). The first pulse train of each TB failed to evoke spikes in the optical signals at locations 1 or 2, or in the FP recordings from location 1 (Fig. 8C). During the second and the third TB both the FP and the optical signals increased, but only transiently, and returned to baseline within 2 min. For the two locations expressing LTP in the SR and the SO (Fig. 8B, locations 3 and 4, respectively), spikes could be clearly identified during the first pulse trains of the second and the third TB (Fig. 8C). Analysis of optical signals at these two locations revealed that spikes contributed substantially to the amplitude increases (Fig. 8D, locations 3 and 4). We examined signals from several photodiodes in 76 different slices from 26 animals, and found that whenever an area showed LTP above a threshold of 6% of the maximum for that slice, spikes were always evident during TBS at those locations. Regions where spikes could not be seen during TBS induction generally failed to express LTP above our detection threshold.
Spikes during the second and third TB showed an excellent correlation with the expression of LTP. The spatial map of spikes during the first TB showed little spread or complexity, but with each successive TB the map of spikes grew and became more complex, ultimately resembling the map for LTP (Fig. 9A and B). The convergence of the TB spike map with the LTP map is illustrated by plotting spike amplitude versus LTP amplitude, point by point for the entire array of 464 detectors (Fig. 9C and D; points from a few bad detectors were omitted). The slopes of the best fitting lines are low for the spikes during TB1 but are clearly greater for TB2 and TB3. The trend is borne out by averaging the slopes for eight experiments with 60–80 μA stimulus currents (Fig. 9E). Some of these plots showed significant numbers of points in the upper left region (data not shown), indicating that postsynaptic spikes occurred at some locations with no LTP. However, these points were generally taken from locations very close to the SE, where saturation of responses could occlude LTP. These results show that the spatial variation in LTP reflects the spatial variation in spikes during later TBs. Thus, postsynaptic spikes predict the spatial pattern of expression of LTP.
Discussion
The present study confirmed that optical recording detects LTP in brain slices (Saggau et al. 1986; Momose-Sato et al. 1999; Hosokawa et al. 2003;Aihara et al. 2005), and extended this method to image the spread of enduring changes in excitability induced by TBS. With this detailed view of the spatial distribution of LTP, we were able to map its expression within the entire CA1 region, and make comparisons between different layers as well as between different locations within layers. These optical signals allow simultaneous assessment of synaptic plasticity over the full extent of the dendritic arbors of populations of neurons, and the decomposition of optical signals into spikes and oEPSPs offers a general approach to the problem of resolving changes in synaptic strength and dendritic excitability. Our experiments revealed clear spatial heterogeneity in the expression of LTP. The magnitude of LTP exhibits spatial variations intrinsic to a slice, and this raises the interesting question of how different pathways respond to a stimulus with different degrees of potentiation.
Spread of oEPSPs and spikes
Optical signals observed by voltage imaging reflect a sum of voltage changes contributed by spikes and EPSPs (Figs 1E and 8B) (Grinvald et al. 1982; Kasuga et al. 2003). FP recordings from the same site show the same components, and intracellular voltage recordings in this preparation are consistent with these two forms of population activity (Golding & Spruston, 1998; Gasparini et al. 2004). With weak stimulation in the middle of the SR, EPSPs dominated the optical signals. Locally evoked EPSPs in the SR were largely restricted to this layer, with little spread to dendrites in the adjacent SLM and SO. Thus, stimulation of Schaffer collateral/commissural fibres predominantly activated the proximal apical dendrites of pyramidal cells within this stratum (Andersen et al. 1980a). Stronger stimulation evoked larger EPSPs and spikes in the SR and the SO, but the responses in the SLM remained weak. Action potentials do not propagate effectively to the remote parts of apical dendrites in CA1 pyramidal cells (Spruston et al. 1995; Magee & Johnston, 1995). The distal apical dendrites have higher densities of various potassium channels (Magee, 2000; Reyes, 2001), and these could limit the spread of voltage changes into the SLM. Since dendrites in the SLM receive inputs from the entorhinal cortex through the temporoammonic pathway and dendrites in the SR receive inputs from Schaffer collaterals and commissural fibres, the electrical segregation of the distal and proximal apical dendrites suggests that these two compartments can process their respective inputs with some degree of independence (Magee, 1998; Remondes & Schuman, 2002; Bernard & Johnston, 2003; Nolan et al. 2004; Ang et al. 2005).
At locations expressing strong LTP, the spike component showed an increase that was significantly greater than the increase evident in the oEPSP. The increase in the spike component may indicate that a greater fraction of the postsynaptic cells at a location are driven past their firing threshold, or, alternatively, that the firing of cells at that location becomes more synchronous. With a dye that senses membrane potential, crossing the action potential threshold and generating a spike gives rise to a larger change in the optical signal than does a graded increase in an EPSP. However, in some experiments, increases in spike amplitude in the SR were not accompanied by detectable increases in the amplitude of oEPSPs, suggesting that in these cases the potentiation arose primarily from the modulation of voltage-gated channels (Daoudal et al. 2002; Wang et al. 2003; Frick et al. 2004; Fan et al. 2005). Regardless of the origin of the increase in spike signals, it is significant that TBS-induced LTP allows the same stimulus current to produce a larger population spike, indicating either increased spike synchrony or firing of a greater number of postsynaptic cells (Andersen et al. 1980b; Taube & Schwartzkroin, 1988; Chavez-Noriega et al. 1990).
Role of spikes in the induction of LTP
LTP depended critically on the generation of postsynaptic spikes during TBS. Areas without detectable spikes during TBS failed to express stable LTP. These observations support the hypothesis that LTP depends on postsynaptic spikes during induction. Regions near the stimulus electrode generated spikes during TBS without subsequent LTP, but this probably reflected the occlusion of LTP, either through saturation or some other mechanism. Elsewhere, the appearance of postsynaptic spikes during induction faithfully predicts the expression of LTP. Postsynaptic spikes provide postsynaptic depolarization, which enables Ca2+ to enter a cell and activate signalling cascades that modify synaptic strength and excitability (Grover & Teyler, 1990; Bliss & Collingridge, 1993; Malenka & Bear, 2004). Back-propagating action potentials play critical roles in providing this postsynaptic depolarization (Magee & Johnston, 1995; Spruston et al. 1995), but spikes originating in dendrites can also perform this function (Golding et al. 2002; Lisman & Spruston, 2005). Although in the present data we cannot distinguish dendritic spikes from back-propagating action potentials unequivocally, our results strengthen the case for an essential role of postsynaptic spikes in LTP (Magee & Johnston, 1997; Markram et al. 1997; Pike et al. 1999; Linden, 1999; Bi & Poo, 2001; Golding et al. 2002). The correlation between postsynaptic spikes and LTP thus suggests that the spatial complexity is a manifestation of a requirement for concurrent pre- and post-synaptic firing, as postulated by Hebb.
Spatial heterogeneity of LTP
The distribution of LTP in the CA1 region revealed several forms of spatial heterogeneity. Comparing different layers, we detected strong LTP in the SR and SO but not in the SLM. The absence of LTP in the SLM clearly results from the failure of evoked responses to spread to this layer, and indicates that even after the potentiation of synapses in the SR, the SLM remains functionally isolated (as discussed above). While responses in the SO are stronger, the large signal increases there cannot be attributed to the strengthening of synapses on basal dendrites. Since the activated Schaffer collateral/commissural fibres predominantly innervate dendrites in the SR (Andersen et al. 1980a), the increased signals in the SO probably resulted from more effective transmission of EPSPs and spikes from the apical dendrites into and through the cell bodies in the SP. The increase in the spike component in the SO could reflect both more pyramidal cells generating action potentials and increases in the amplitudes of dendritic action potentials. These changes would originate in the SR dendrites in the vicinity of the activated synapses, triggered by localized Ca2+ changes in dendritic shafts (Raymond & Redman, 2006). It will be interesting to test induction protocols that produce more extensive Ca2+ rises to see if this can potentiate synapses and dendrites over greater distances.
Increasing the stimulation intensity altered the spatial pattern of LTP (Fig. 4). Variations in stimulus frequency and in the statistical properties of stochastic stimuli have also been shown to influence the spatial pattern of both LTP and LTD in the CA1 region (Aihara et al. 2005). We found that increasing the stimulus current enlarged the size of the potentiated region and converted the map from a focus in the SR around the SE and an adjacent focus in the SO to a complex distribution with several local maxima. Strong stimulation saturated responses near the SE and occluded LTP. Although the release of a local extracellular signal that blocks LTP expression remains a possibility (Kauer, 1999; Aihara et al. 2005), saturation of both spikes and EPSPs is clearly an important factor in the suppression of LTP near the stimulating electrode. Thus, the evaluation of a contribution from local circuits will depend on controlling for saturation circuits and inhibitory interneurons are more likely to play a role in the other spatial features in the complex LTP maps. Although we attempted to address the role of inhibition by blocking GABAA receptors with picrotoxin or bicuculline, the resulting multi-spike activity made analysis and interpretation of the optical signals more difficult. Nevertheless, this remains an important possibility that we hope to address with pharmacological tools. It would be especially interesting to use pharmacological or genetic methods to target a specific group of interneurons, such as calbindin-containing bistratified interneurons, which have dendritic and axonal arbors in the SR (Dingledine et al. 1987; Freund & Buzsaki, 1996).
Since the LTP map so closely resembles the spike map during TBS, elucidating the mechanism underlying the spatial complexity of LTP may reduce the problem of determining how postsynaptic spikes are regulated and what forms of short-term facilitation allow synaptic inputs to bring cells above threshold. It is important to bear in mind that the spatial heterogeneity of LTP was not matched by the spatial heterogeneity in the response to a single shock applied prior to TBS. Thus, the distant local maxima in the distribution of LTP reflect the distribution of some other signal. The lack of overlap of LTP distributions for different stimulus sites in the same slice (Fig. 7) indicates that this other signal depends on TBS, and that it is not an ambient property of the slice. One possibility is that the distribution of LTP reflects the activity of another circuit parallel to the Schaffer collateral/commissural fibre pathway with a sparse and non-uniform projection field. This circuit would then facilitate the primary excitatory pathway during TBS to enable postsynaptic cells to spike and thus initiate potentiation.
The heterogeneous spatial distribution of LTP in the CA1 region of rat hippocampal slices implies an organization of hippocampal circuits specifically associated with synaptic plasticity. This organization could be related to the columns of neurons in the primary somatosensory cortex (Woolsey & Van der Loos, 1970; Petersen et al. 2003; Sun et al. 2006) and the primary visual cortex (Hubel & Wiesel, 1962; Grinvald et al. 1984; Bonhoeffer & Grinvald, 1991; Ohki et al. 2005). It is also possible that different ensembles of neurons, such as the ensembles responsible for theta and the gamma oscillations (Chrobak et al. 2000; Buzsaki, 2002; Remondes & Schuman, 2002), are activated and potentiated during TBS. The architecture revealed by the spatial heterogeneity in TBS-induced LTP may be an essential element in the selective modification of specific subsets of neural circuits during the storage of information by the nervous system.
Acknowledgements
We thank Camin Dean and Dan Johnston for critical comments on this manuscript. This work was supported by NIH grant NS30016, and the Epilepsy Foundation.
Footnotes
-
(Received 23 April 2006; accepted after revision 24 July 2006; first published online 27 July 2006)
- 2006 The Authors. Journal compilation © 2006 The Physiological Society
References
Figure 1. Spatial distribution of LTP A, experimental arrangement. A CCD image displays the hippocampal slice used in this experiment. A red dashed curve highlights stratum pyramidale (SP) and continuous red curves highlight the borders between strata. The red arrow indicates the stimulation site and the yellow arrowhead indicates the field potential recording site. The white arrow above points toward the CA2 region. The yellow dashed line indicates the section along which the profiles of responses and amplitude increases are shown in Fig. 2. SO, stratum oriens; SR, stratum radiatum; SLM, stratum lacunosum-moleculare; DG, dentate gyrus. B, spatial layout of traces (without spatial filtering) showing the response of each photodiode at the corresponding location. Layers and electrodes are indicated in A. Traces are averages of four trials. Each trace is 100 ms. C, spatial maps of responses and LTP with peak amplitude encoded as colour according to the scale in the lower left corner. Symbols and indicators are as in A. Ci, spatial distribution of peak responses evoked by 60 μA pulses before TBS. Cii, spatial distribution of peak responses evoked by 60 μA 50 min after TBS. Ciii, spatial distribution of amplitude increases as differences between Ci and Cii (LTP). White arrows indicate local maxima. D, time course of FP slope and optical signal amplitudes normalized to initial responses recorded before TBS. TBS was applied at the black arrow after obtaining a 20 min baseline. The optical signals were recorded at locations indicated in A. E, FP and optical signals from 4 locations (corresponding to sites indicated in A) before TBS (left), after TBS (middle), and superimposed (right). The dashed curves are fits of the EPSP component (Methods). Stimulation artifacts in FPs were removed. Notice that spikes contributed substantially to amplitude increases at locations 1, 2 and 4.
Figure 2. Plots of peak response amplitude and LTP along different axes A, profiles of amplitudes and amplitude increases along the somatodendritic axis (yellow dashed line in Fig. 1A). Top, profiles of responses before (□) and after (•) TBS. Bottom, profile of amplitude increases (LTP). B, profiles of peak amplitudes and amplitude increases along a contour through the SR. Top, profiles of responses before and after TBS. The black arrow indicates the stimulation site. Bottom, profile of amplitude increases, with three local maxima indicated by white arrows (as in Fig. 1Ciii). C, profiles of amplitudes and amplitude increases along a contour through the SO. Top, profiles of responses, as in B. Bottom, profile of amplitude increases, as in B.
Figure 3. LTP of optical signals requires activation of NMDA receptors A, spatial distributions of responses and amplitude increases in signals evoked by 75 μA pulses. Ai, an image of the slice with labels as in Fig. 1A. Aii, spatial distribution of peak optical signals before TBS (time point a in B). Aiii, spatial distribution of amplitude increases in the presence of APV (50 μm). Signals recorded at time points a and b in B were compared. Aiv, spatial distribution of amplitude increases after washing out APV. Signals recorded at time points c and d in B were compared. B, time course of optical signal amplitudes. Optical signals were recorded from locations indicated by 1 and 2 in Ai. Signals were normalized to the values prior to TBS (times a and c). Black arrows indicate the times when TBS was applied. The black bar indicates the presence of APV, which blocked LTP. After washing out APV, the second TBS induced LTP.
Figure 4. Stimulation intensity affects the spatial distribution of LTP A–D, peak response was encoded as colour and mapped as in Fig. 1C. The left column shows images of slices with labels as in Fig. 1A. The middle column shows the spatial distribution of response amplitudes before TBS. The right column shows the spatial distribution of LTP as amplitude increases 40 min after TBS. Stimulus intensities were: A, 20 μA; B, 40 μA; C, 60 μA; D, 80 μA. E, plot of the numbers of local maxima induced by different stimulation intensities. Mean ± s.d. is plotted with the number of experiments in parentheses.
Figure 5. Spike contributions at different locations before and after TBS to 60 μA current pulses A, responses in the SR, including total amplitude, spike amplitude, and EPSP amplitude at the stimulation site, 270 μm away from the stimulation site, and at the location expressing the strongest LTP. All amplitudes were normalized to the total amplitude at the stimulation site before TBS. All increases were significant except for the increase in the EPSP at the stimulation site. Data are shown as mean ± s.e.m. *Significant differences using t test at P < 0.05 (n = 5). B, responses in the SO at three locations as in A, with amplitudes normalized to the pre-TBS value at the stimulation site. C, normalized amplitude increases in the SR. The differences in amplitude increases between spikes and EPSPs were not significant at the stimulation site and 270 μm away. At the location with the strongest LTP expression, the amplitude increase in the spike component was significantly larger than that of the EPSP component (*). D, normalized amplitude increases in the SO. Similar to signals in the SR, the amplitude increase in the spike component was significantly larger than the increase in the EPSP component at the location expressing the strongest LTP (*). All values are means of 5 measurements.
Figure 6. Saturation of responses near the site of stimulation A, the total amplitude plotted versus stimulation current at the stimulation site in the SR, 270 μm away in the SR, across from the stimulation site in the SO, and 270 μm away in the SO (n = 7). Data were normalized to the response at the stimulation site to 100 μA and plotted as mean ± s.e.m.B, the spike component plotted versus stimulation current. The stimulation currents for half-maximal amplitude were about 40 μA near the stimulation site and about 60 μA 270 μm away. C, the oEPSP amplitude plotted versus stimulation current.
Figure 7. Spatial distributions of LTP induced by stimulation at different sites A, image of the slice employed in this experiment with labels as in Fig. 1A. The second stimulation electrode (SE2) is indicated by the yellow arrow. Labels 1 and 2 indicate sites for which the time course of optical signals is shown in D. Stimulus current was 50 μA for stimulation at both SE1 and SE2. B, spatial distribution of peak responses and LTP for stimulation at SE1. Bi, distribution of optical signals before the first TBS (time point a in D). Bii, LTP distribution after TBS applied to SE1. Signals recorded before the first TBS were subtracted from signals recorded 58 min after (time points a and b in D). Biii, LTP distribution after TBS applied to SE2. Signals recorded before the second TBS were subtracted from signals recorded 54 min after (b and c in D). TBS applied to SE2 failed to potentiate responses evoked by SE1. C, spatial distribution of peak responses and LTP for stimulation at SE2. Ci, distribution of optical signals before the first TBS. Cii, LTP distribution after TBS applied to SE1. Responses evoked by SE2 were not potentiated. Ciii, LTP distribution after TBS applied to SE2. This distribution differs from that in Bii. D, time course of optical signals recorded at locations 1 and 2, labelled as in A. Amplitudes were normalized to amplitudes recorded before the first TBS (time point a). Top, time course of optical signals at location 1 evoked by SE1 and SE2. The red and yellow arrows indicate TBS applied to SE1 and SE2, respectively. a, b and c indicate 2 min before, 58 and 114 min after the first TBS, respectively. At this location, only TBS applied to SE1 induced LTP. Bottom, time course of optical signals evoked by SE1 and SE2 at location 2. LTP was only induced by TBS applied to SE2.
Figure 8. LTP expression and spikes during induction A, spatial distributions of peak responses and response increases (stimulus current, 60 μA). Ai, image of the slice used in this experiment, with labels as in Fig. 1A. Aii, distribution of optical signals recorded 2 min before TBS. Aiii, distribution of LTP, computed by subtracting signals recorded 2 min before from 60 min after TBS. B, responses to single pulses recorded 2 min before (cyan) and 60 min after (red) TBS (average of 4 sweeps). FP is shown along with four optical signals recorded at the locations labelled in Ai. Black arrows indicate the time of stimulation. Note that FP and optical signals recorded at locations 1 and 2 did not increase after TBS while responses at locations 3 and 4 did increase. C, traces recorded during TBS (single sweep) and compared with control signals (cyan, the same as the cyan traces in B). The responses evoked by the first 6 pulses (at 100 Hz) of each theta burst (TB) are shown. Notice that no spike was seen in field potential and optical signals recorded at locations 1 and 2 during TBS while spikes were evident at locations 3 and 4. D, responses increased at locations 3 and 4 but not at locations 1 and 2. The FP slope and the optical signals recorded at locations 1 and 2 did not increase 60 min after TBS. At locations 3 and 4, spikes contributed substantially to the amplitude increases (arrowheads).
Figure 9. Distribution of peak responses, spike amplitudes during TBS, and LTP A, maps for an experiment with 60 μA stimulus current. Panel 1, responses to a single pulse (from Fig. 8Aii). Panels 2–4, spike amplitudes during the first, second, and third TB. Panel 5, LTP (from Fig. 8Aiii). B, maps for an experiment with 80 μA stimulus current. Panel 1, responses to a single pulse (from Fig. 4D). Panels 2–4, spike amplitudes during the first, second, and third TB. Panel 5, LTP (from Fig. 4D). In both A and B the first TBS triggered spikes close to the stimulating electrode, but the area spread and became more complex with the second and third TB, finally resembling the map of LTP. C, plots of spike amplitude during each TBS versus LTP for each point (each photodiode of the imaging device) from A. The lines are linear fits for the first (black), second (red), and third (green) TB, yielding R values of 0.60 ± 0.08, 0.85 ± 0.10, and 0.83 ± 0.10, respectively (R ± s.d., P < 0.0001). D, plots of spike amplitude versus LTP for the experiment shown in B, with linear fits coloured as in C. R = 0.35 ± 0.08, 0.86 ± 0.07, and 0.91 ± 0.07, respectively (P < 0.0001). E, slopes from linear fits such as in C and D (stimulus current = 60–80 μA, n = 8). A t test indicated significant increases for TB2 and TB3 versus TB1 (*P < 0.005).
