THE JOURNAL OF COMPARATIVE NEUROLOGY 2965984313 (1990)
Burst Generating and Regular Spiking Layer 5 Pyramidal Neurons of Rat Neocortex Have Different Morphological Features YAEL CHAGNAC-AMITAI, HEIKO J. LUHMANN, AND DAVID A. PRINCE Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, California 94305
ABSTRACT Intracellular recordings were obtained from pyramidal neurons in layer 5 of rat somatosensory and visual cortical slices maintained in vitro. When directly depolarized, one subclass of pyramidal neurons had the capacity to generate intrinsic burst discharges and another generated regular trains of single spikes. Burst responses were triggered in an all-or-none manner from depolarizing afterpotentials in most bursting neurons. Regular spiking cells responded to electrical stimulation of ascending afferents with a typical EPSP-IPSP sequence, whereas IPSPs were hard to detect in bursting cells. Orthodromic activation of the latter evoked a prominent voltage-dependent depolarization that could trigger a burst response. Intracellularly labelled bursting and regular spiking cells were located in layer 5b, but had distinctly different morphologies. Bursting neurons had a large pyramidal soma, a gradually emerging apical dendrite, and an extensive apical and basal dendritic tree. Their axonal collateral arborization was predominantly limited to layers 5/6. In contrast, regular spiking cells had a more rounded soma with abruptly emerging apical dendrite, a smaller dendritic arborization, and 2 to 8 ascending axonal collaterals that arborized widely in the supragranular layers. Both bursting and regular spiking cells had main axons that entered the subcortical white matter. These data show that some subgroups of pyramidal neurons within the deeper parts of layer 5 of rat cortex are morphologically and physiologically distinct and have different intracortical connections. Bursting cells presumably function to amplify and synchronize cortical outputs, whereas regular spiking output neurons provide excitatory feedback to neurons a t all cortical levels and receive a more effective orthodromic inhibitory input. These data support the hypothesis that differences in gross neuronal structure, perhaps even the subtle differences that distinguish subclasses of neurons in a given lamina, are predictive of underlying differences in the type and distribution of ion channels in the nerve cell membrane and connections of cells within the cortical circuit. Key words: visual cortex, somatosensory cortex, anatomy, physiology, epilepsy
Studies of mammalian cortical function have often emphasized patterns of neuronal connectivity as determinants of the properties of the system (Hubel and Wiesel, '62; Mountcastle, '57; for reviews, see Martin, '84; Parnavelas, '84). However, it is increasingly apparent that the operation of neuronal networks cannot be accurately predicted without knowledge of the properties of their individual elements (Llinas, '88). Even within the simplest cortical module, anatomically distinct neuronal types and subtypes can be categorized on the basis of features such as size, shape, patterns of dendritic arborization, position, connections, and transmitter content. Electrophysiological studies have suggested that different classes of cortical neurons may o 1990 WILEY-LISS, INC.
possess different intrinsic membrane properties that affect fundamental characteristics of their behavior, such as responses to synaptic inputs and the pattern of their spike output (McCormick et al., '85; Connors and Kriegstein, '86). These data support the hypothesis that variations in neuronal structure are predictive of underlying differences in the type and distribution of ion channels in the neuronal cell membrane (for review, see Prince and Huguenard, '88). One Accepted February 7,1990. Y. Chagnac-Amitai's present address is Department of Physiology, Ben Gurion University, Faculty of Health Sciences, P.O. Box 653, Beer-Sheva, Israel.
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS corollary of this proposal is that cells found to have different intrinsic membrane properties will be shown to be distinct neuroanatomically. To an extent, this has already been demonstrated for some of the general classes of neocortical pyramidal neurons. Experiments in vivo have shown that pyramidal tract neurons with fast versus slow axonal conduction velocities have different biophysical properties (Takahashi, '65; Calvin and Sypert, '76), and different structural characteristics (Deschihes et al., '79; Sakai and Woody, '88). Recent studies of cortical pyramidal neurons in vitro have indicated that subclasses with different firing behaviors can be distinguished. In earlier studies from this laboratory, a small population of pyramidal cells with intrinsic burstgenerating properties was found in guinea pig sensorimotor cortex (Connors et al., '82; McCormick et al., '85; see also recent report of burst-generating neurons in visual cortex, by Montoro et al., '88). In addition to this distinctive firing pattern, which seemed to be a rarity among neocortical cells, a point of great interest was the restricted laminar location of these neurons to layer 4 and superficial 5 (Connors et al., '82). Other work suggested that synchronized bursts in this population of cells might be important in the initiation of interictal epileptiform discharges in neocortical slices, since the latter appear to begin in the same lamina (Connors, '84; for review of the role of intrinsic burst generation in epileptogenesis see Prince and Connors, '86). Because of the small numbers of these intrinsically bursting neurons encountered in previous studies, an analysis of their electrophysiological properties and possible morphological differentiation from other neurons within the same cortical lamina was not possible. Recently, another population of bursting neurons was identified in layer 5b of the mouse barrel cortex (Agmon and Connors, '89). The proportion of intrinsically bursting cells among the recorded population in this lamina is apparently larger than in layer 4 to upper 5. We have further investigated the firing properties, synaptic potentials, and morphology of intrinsically bursting and nonbursting cells in layer 5b. Results indicate that intrinsically bursting and regular spiking subclasses of pyramidal neurons in this lamina have distinctly different morphological characteristics as well as different connections within the cortical circuit. Portions of these results have been published in abstract form (ChagnacAmitai et al., '88).
MATERIALS AND METHODS Tissue preparation and recording techniques Sprague-Dawley rats (about 1month old, 100-150 g) were deeply anesthetized with intraperitoneal sodium pentobar-
Abbreviations ACSF DAP EPSP GABA HRP IB IPSP LY PT
RS SMC
vc
artificial cerebrospinal fluid depolarizing afterpotential excitatory postsynaptic potential gamma aminobutyric acid horseradish peroxidase intrinsically bursting inhibitory postsynaptic potential Lucifer Yellow pyramidal tract regular spiking sensorimotor cortex visual cortex
599
bital, decapitated, and their brains rapidly removed and placed in chilled (5"C), oxygenated artificial cerebrospinal fluid (ACSF). A block of parietal tissue was dissected and glued to the platform of a vibratome (Lancer series 1000). Slices 400 to 450 pm thick were cut in a coronal plane from a region corresponding to the sensorimotor cortex (Zilles et al., '80; Zilles and Wree, '85) and placed in a chamber whose temperature was maintained at 35 1°C. The slices were kept a t the fluid/gas interface and continuously superfused with ACSF (0.8ml/min). The atmosphere over the slices was warmed and saturated with humidified 95% 0,/5% CO,. The ASCF contained (in mM) NaC1, 124; KC1, 5; CaCl,, 2; MgSO,, 2; NaHCO,, 22; dextrose, 10; NaH,PO,, 1.25; and, when saturated with 95 % 0,-5 % CO,, had a pH of 7.2. Microelectrodes were filled with 4 M potassium acetate and had resistances of 120 to 170 MO. Intracellular recordings were obtained after slices had been incubated for at least 1 hour in the chamber. Layer 4 and the border region between layers 5 and 6 could be visualized as clear bands in transilluminated or tangentially lit slices (Agmon, '88). We confined our microelectrode penetrations to the upper part of the lower band, which was located about two-thirds of the distance from the pia to the white matter. Analysis of counterstained sections from slices containing labelled neurons showed that this area corresponded to layer 5b. Single extracellular stimuli (180 to 200 psec duration) were delivered through a monopolar sharpened tungsten wire or a bipolar stimulating electrode situated in the underlying white matter or near the pial surface in the same radial column as the recorded cell. Data were recorded on magnetic tape (0 to 10 KHz) and digitized with an ISC-16 system (RC electronics) on a personal computer. On the basis of electrophysiological criteria and in agreement with a previous report (McCormick et al., '85), we distinguished between regular spiking cells (RS) and intrinsically bursting neurons (IB). Whereas the former responded to depolarizing current injection with one or several single action potentials, whose frequency was dependent on the level of membrane depolarization (e.g., McCormick et al., '85), IB cells generated a burst of at least three action potentials arising from a slow membrane depolarization that appeared to be due to a summation of depolarizing afterpotentials (DAPs) of successive spikes. We also observed cells that generated "doublets" at the onset of the depolarizing pulse, i.e., pairs of closely spaced action potentials. This pattern was observed most frequently in neurons impaled with electrodes containing Lucifer Yellow (LY) and may have been induced by lithium salts diffusing into the cell (for further discussion of this problem see Tseng and Haberly, '89) or due to different intrinsic firing properties (e.g., Calvin and Sypert, '76). For purposes of our analysis, we classified neurons showing a doublet firing pattern as RS cells. Neither IS neurons, nor those that fired an initial doublet, could be converted to IB neurons by increases in the amplitude of the depolarizing current pulse (up to 1nA).
Staining, immunohistochemistry, and data analysis Intracellular labelling was accomplished by impaling neurons with thin-walled microelectrodes filled with 5 to 10% Lucifer yellow (Sigma) in 1.0 M lithium acetate (resistances of 100 to 150 MO). The dye was injected by passing hyperpolarizing current pulses (1.0 to 1.5 nA, 600 msec duration) at 0.5 Hz for 5-20 minutes. Only 1or 2 cells were injected in each slice to assure the clear identification of
Y. CHAGNAC-AMITAI ET AL.
600
A
.
-.
I
1
-I
11 nA
2
1L ---5l-
20 ms
i
3
L 1 2 0 mV 1 1 . 4 nA
1
-
20 ms
2
Fig. 1. Characteristic responses of bursting and nonbursting pyramidal neurons to depolarizing intracellular current injection. A1,2: Suprathreshold current pulses of increasing amplitude in a regular-spiking cell evoke a train of action potentials that shows marked spike frequency adaptation. B: Responses of an intrinsically bursting neuron to current pulses of increasing amplitude (1-3). B1: Burst complex evoked a t threshold is preceded by an initial fast spike. B2,3:A larger amplitude current pulse alternately evokes trains of spikes (B2) and spikes followed by bursts (B3). Prominent depolarizing afterpotentials (DAPs)
follow each action potential in the spike train (B2, arrow). The burst complex includes spikes of varying amplitudes and durations that may appear in an all-or-none manner (arrow in B3). C: Another intrinsically bursting neuron in which the burst is evoked at long latency by a threshold current pulse ( C l ) and latency becomes shorter as current intensity increases (superimposed sweeps of C2). Calibrations following C for B and C. In this and all subsequent figures, intracellular recordings are shown in the upper traces and current pulses in the lower traces.
each neuron. Slices containing injected neurons were fixed with 4% paraformaldehyde in 0.1 M phosphate buffer overnight, and then cleared with dimethyl sulfoxide for 20 minutes. Cells were viewed and photographed through a Leitz microscope equipped with epifluorescence, and slices that contained well-labelled neurons were processed for immunohistochemistry. Slices were resectioned at 50 to 60 I m on a freezing microtome, and the resulting sections were rinsed, blocked, and placed for 1day in a primary antiserum (1:500) raised against LY (Taghert et al., ’82). Further processing utilized the avidin-biotin-peroxidase method (Vectastain ABC kit, Vector laboratories) followed by cobalt intensification and a 3-3’-diaminobenzidine (DAB, Sigma) reaction with H,O, using standard techniques. A Nissl counterstain was used to allow examination of the cell’s location in relation to the cortical layers.
Well-filled, anti-LY processed neurons were reconstructed with the aid of camera lucida drawings of serial sections (magnification 160 or 250 x ). Some measurements were made by using a Zeiss MOP3 digitizing tablet. IB and RS neurons were compared in terms of the following parameters: somatic area; width of the apical dendrite a t a level 80 pm from the center of the soma; and dendritic branch patterns by using a Sholl (’56) analysis. No attempt was made to apply corrections for tissue shrinkage or other distortions, since the goal was to compare IB and RS neurons rather than to obtain absolute measurements.
RESULTS A total of 73 intrinsically bursting (IB) and 28 regular spiking (RS) cells were recorded in layer 5 of both sensori-
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS TABLE 1. Electrophysiological Properties of Bursting and Regular Spiking Laver 5 Pvramidd Neurons' V, (mV) 73.5 + 4.3 (22) 67.3 + 6.3 (6) 72.0 + 3.3 (9) 73.7 i 3.0 (7)
IBneurons SMC IBneruons VC RS neurons SMC RS neruons VC
R, (MQ) 30.2 + 13.8(23) 36.2 i 8.9 (6) 31.1 * 13.7 (10) 33.6 i 16.4 (7)
1st spike amplitude (mV) 89.7 * 9.5 (23) 83.6 i 5.3 (6)' 95.4 f 9.8 (10) 92.1 i 6.1 (7)
'Restingmembranepotential (V,),inputre&tance (R,),andamplitudeoffiratspikeevokedby depolarizing pulse for intrinsically bursting (IB)and regularly spiking (Rs)neurons of sensorimotorcortex (SMC) and visual cortex (VC), expressedas mean i standard deviation. 'Signifcant (t-test,P < 0.01) difference in 1st spike amplitude between IB and RS neurons in
vc.
motor cortex (SMC) and visual cortex (VC). All cells had stable resting potentials more negative than -60 mV for a t least 10 minutes, an input resistance of a t least 20 MQ, and action potential amplitude of a t least 75 mV. IB neurons generated a complex of 3 or more action potentials riding on a depolarizing envelope, following injection of a depolarizing current pulse (Fig. 1B1; Connors et al., '82; McCormick et al., '85). RS cells fired a train of single fast action potentials to an injection of depolarizing current pulses (Fig. 1A). The RS cells did not seem to differ from neurons with similar properties recorded in other layers and previously described in studies from this laboratory (Connors et al., '82; McCormick et al., '85). Since our efforts were focused on an analysis of the properties of IB cells, most RS cells that were impaled were not studied, and the relative number of such neurons in our sample is not representative of their true incidence in layer 5. We estimate that IB cells constituted 50-60% of our impalements in somatosensory cortex and around 30% in visual cortex, when penetrations were made into this lam-
ina. Some electrophysiological properties of 51 randomly selected neurons are shown in Table 1. There were no significant differences between the mean resting membrane potential and input resistance of regular spiking versus bursting neurons. As previously noted (McCormick et al., '85), the spike amplitude of IB neurons was slightly lower than that of RS cells.
Characteristics of burst generationin neurons of layer 5b One prominent characteristic of all of the 73 IB cells studied in layer 5b was a depolarizing afterpotential (DAP) that followed each spike (Fig. 1B2, arrow). A significant proportion (ca. 50%) of RS cells also generated DAPs and about 30% of these neurons generated a spike doublet but by definition never a full burst, even when graded increases in the amplitude of depolarizing current pulses were applied. In each IB neuron, the burst developed after an initial fast spike. Figure 2B shows a cell in which the same depolarizing current pulse alternately evoked a burst and a single action potential. Superimposition of two such alternating responses shows that the full burst complex emerged from the peak of the DAP. A similar relationship is shown in the neuron of Figure 1B in which a burst discharge is triggered after the first ( B l ) or second (B3) fast spike. In this neuron, as in the cell of Figure 2B, identical current pulses delivered at 0.5 Hz (Fig. 1B2, 3) could evoke either single spikes or spikes and burst complexes that developed as all-or-nothing events. When evoked by brief current pulses, burst comDlexes in such neurons outlasted the depolarizing pulse (Fig. 4A).
--/
-
1
L
J
2
60 1
J
L
2
--f
4
L
20 mV 1 nA
J 2 0 ms -
10 ms Fig. 2. Intrinsically bursting neuron with variable firing pattern.
A1,2 Burst complex containing some small spikes (arrow in A2) develops after the first evoked action potential. A 3 - 4 Slight increase in current amplitude. Pairs of single spikes followed by DAPs are evoked
alternately with responses in which a burst follows the second action potential (cf. A3 and A4). B Same neuron as A. Superimposed sweeps to show that the first spike of the burst is triggered from the peak of the DAP. Note faster time base in B.
Y. CHAGNAC-AMITAI ET AL.
602
1
-
J
2
20 ms
2 0 mV 1 nA
4
-
40 ms
Fig. 3. Neuron that generates a graded rather than all-or-none burst complex. Low intensity current pulses evoke single action potentials followed by prominent DAPs (1,2).As current is increased, DAPs increase in amplitude and summate to generate the slow envelope of the burst (sweeps 3,4).
In 13 of the 73 bursting neurons, the slow events underlying burst generation appeared to be graded. For example, in the cell of Figure 3, prominent DAPs followed single action potentials evoked by depolarizing current pulses (traces 1-3). When the current was increased to more than 0.5 nA, the DAPs of the first four spikes summated to create a depolarizing envelope underlying a burst complex (Fig. 3, trace 4). Although the graded bursts generated in cells of this sort are somewhat different from the more typical all-or-none events that are shown in the neurons of Figures 1 and 2, recordings in all neurons strongly suggested that summated DAPs were an important component of the slow depolarization underlying burst generation. The amplitude of the slow envelope underlying the burst varied between 10 and 30 mV (measured from the base of the first spike in the complex to the base of the spike a t the peak of the envelope). Depolarization sufficient to reach spike threshold usually evoked a burst consisting of several fast spikes which decreased in amplitude and increased in duration during the course of the burst (Figs. 1-3). Small increases in the amplitude of the triggering pulse usually did not affect the stereotyped pattern of spike generation, but did shorten the latency for burst triggering (Fig. lC2). Progressive inactivation of burst spikes was prominent in neurons where higher amplitude underlying depolarizations were evoked (e.g., Figs. 1B1,2B, 4A). In some neurons, small amplitude, broad spikes appeared at the end of the burst (Fig. 2A, arrow; Fig. 2B), or on the falling limb of larger amplitude spikes (Fig. lB3, spike preceding the last spike of burst), suggesting that more than one spike initiating zone was present. The lower amplitude spikes had similarities to the high threshold calcium spikes that occur in the late portion of bursts in hippocampal CA3 pyramidal neurons (Wong and Prince, '78), and in other cells (Llinas and
Yarom, '81a,b). Burst discharges were followed by slow afterhyperpolarizations lasting up to 100 msec, whose amplitude increased a t more depolarized membrane potentials (Fig. 4A). We examined the effect of altering the membrane potential (V,) on burst generation by injecting DC depolarizing or hyperpolarizing currents across the membrane, and evoking bursts from various membrane potentials with long or short depolarizing current pulses. Bursts could be evoked a t hyperpolarized membrane potentials when the depolarizing current pulse was sufficiently strong. In a few neurons, burst generation was blocked when V, was held positive to about -60 mV. This suggested that a low threshold calcium spike might contribute to burst generation, a finding similar to that reported in some bursting neurons of layer 4-upper 5 (Connors et al., '82; McCormick et al., '85). In all other IB neurons tested in this way, however, the burst could be evoked when the membrane potential was held depolarized, close to the firing level (Fig. 4A3). About 90% of the IB neurons generated a burst followed by a train of spikes when depolarized from resting membrane potential by a prolonged current pulse (Fig. 5B). About 10% of IB cells were capable of generating repetitive burst discharges during prolonged depolarizing pulses that kept V, close to the firing level (Fig. 5A; see also Agmon and Connors, '89). In these instances, repetitive bursts might arise from a depolarized V, without preceding hyperpolarizations (Fig. 5A3).
Synaptic responses of intrinsically bursting and regular spiking layer 5 neurons We compared the responses of RS and IB neurons to orthodromic activation produced by stimulation of either white matter or the subpial region in the same radial column
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS
A
603
Lkk J l I -
-72
r..................... l
-63
...........................
-57
J--L
1
............................
I I
2
-75-4
3
I
1
2
120 mV 1 nA 20 ms
1 2 0 mV 1 1 . 4 nA 40 ms
Fig. 4. Effects of membrane potential on burst generation in two neurons. A Burst evoked by a brief current pulse increases in duration as the membrane potential is depolarized from -72 mV (sweep 1) to -63 and -57 mV (sweeps 2 and 3). Note the increase in amplitude of the burst after-hyperpolarization as the membrane potential becomes more positive. B Bursts evoked by long current pulses in another neuron also increase in duration as membrane potential is depolarized by direct current as in A.
as the impaled cell. Such stimuli evoked typical sequences of excitatory and inhibitory postsynaptic potentials (EPSPs and IPSPs) in RS cells, as previously reported (Connors et al., '88). A t threshold, an isolated, short-latencyEPSP could be elicited (not shown). As the stimulus intensity was increased, early and late IPSP components emerged, that had reversal potentials of about -70 and -85 mV, respectively (Fig. 6A). The pattern of synaptic responses to stimuli delivered at the same sites was significantly different in IB neurons, as is shown in the example of Figure 6B. At low stimulus intensities, a short latency EPSP could be evoked (Fig. 6B, -67 mV trace). As the stimulus intensity increased, a threshold was reached at which the burst complex was triggered, a t variable latencies, as an all-or-none event (not shown). The orthodromically evoked burst itself was similar in configuration to the burst evoked by direct depolarization of the same neuron. A small depolarization of the membrane potential might render a previous subthreshold synaptic stimulus effective in evoking the burst complex (Fig. 6B, -60 mV trace). When this occurred, a slow ramp-like depolarization led from near the peak of the EPSP to the first spike of the burst (Fig. 6B, arrow). Paradoxically, stimuli of increased intensity, delivered with the neuron at the same membrane potential, would fail to trigger the burst, and instead evoke a large-amplitude subthreshold EPSP (Fig. 6C), or (if threshold was reached) a single action potential (Fig. 6D). The disappearance of the burst at higher stimulus intensities suggested recruitment of an underlying inhibitory conductance that shunted the burst-related depolarization. However, in most of the IB neurons it was not possible to demonstrate the fast and slow IPSP, even by evoking synaptic responses at depolarized
V,s. A typical EPSP-IPSP sequence could be clearly demonstrated in only a few bursting cells. In most IB cells, a prominent voltage-dependent depolarization was evoked by synaptic activation at resting potential. Examples are shown in Figure 6C,D. Subthreshold (Fig. 6C) and threshold (Fig. 6D) stimuli for eliciting an action potential from the short latency EPSP, evoked a slow depolarization that arose shortly after the peak of the EPSP and was independent of spike generation. The amplitude and duration of this response component were variable from stimulus to stimulus and increased as the membrane was depolarized from the resting potential towards the spike threshold. When suprathreshold stimuli were applied, a short latency spike was evoked from the initial EPSP component, followed by a prolonged (5100 msec), slow depolarization with a peak latency of 50-75 msec (Fig. 6D, -62 mV trace, arrow). The onset of the slow depolarization was influenced by the stimulus intensity, so that it arose close the peak of the EPSP with stimuli that were close to threshold (Fig. 6B, -60 mV trace) and from the repolarizing limb of the EPSP following more intense stimuli (Fig. 6D, -62 mV trace). This finding suggested the presence of a superimposed short latency IPSP. The graded subthreshold slow depolarizations appeared to merge into the ramp-like potentials that triggered delayed spikes or bursts (Fig. 6B-D).
Anatomical identification of intrinsically bursting and regular spiking neurons Cells were impaled with LY-containing electrodes following penetrations into the deeper portions of layer 5, using
Y. CHAGNAC-AMITAI ET AL.
604
J
I
I
3
1
3 JI
7
100 ms Fig. 5. A: Examples of repetitive burst-firing patterns in three layer 5 bursting neurons during prolonged depolarizing current pulses. A1,2: The initial burst in these neurons contains 4 and 3 spikes, respectively, followed by repetitive firing consisting of single spikes, doublets, or bursts of three spikes. A3: Neuron fired repetitive bursts of up to 6
spikes and frequent doublets during the 1 second depolarizing current pulse. Note that membrane did not repolarize to resting level between bursts. B1-3: Firing pattern in another neuron where a 1 second depolarizing current pulse of increasing intensity (top to bottom segments) evoked an initial burst followed by a train of spikes.
landmarks described above. Spike firing patterns were characterized within 2 minutes of impalement by passing depolarizing current pulses, and neurons were then classified as IB or RS cells. This rapid determination of electrophysiological behavior was made before any clear effect of Li+ injection on spike activities could be noted. In agreement with previous reports (Mayer et al., '84), alterations in spikes occurred within a few minutes after beginning the application of hyperpolarizing current pulses, a t which time some broadening of action potentials and a prolongation of the burst complex was noted. The dye injection seemed to facilitate generation of repetitive burst discharges during prolonged current pulses in some cells, but no neuron that was identified as a RS cell subsequently developed a burst firing pattern. Thus, recordings with LY-filled electrodes allowed clear differentiation of IB and RS neurons. The recordings obtained were not used for electrophysiological analysis.
25 labelled neurons were similar in their soma-dendritic features, as will be described in detail below (see Table 2), and all were situated in the lower half of layer 5. Most somata were pyramidal in shape (Figs. 7A, 8A, 9). Somatic area averaged 749.3 2 95.1 pm2 (mean 2 standard deviation; n = 13), and the average maximum diameter was 37.6 3.1 pm (n = 13) (see Table 2). When somata were viewed in counter-stained sections, it was clear that these neurons were among the largest in this area of cortex. The apical dendrite emerged from the cell body in a gradual, tapering manner, making it difficult to establish the boundary between the soma and the ascending dendrite. The diameter of the apical dendrite, measured 80 pm from the center of the cell body, averaged 8.5 1.6 pm (n = 13). From its origin, the apical dendrite took a straight course towards the pial surface, giving off oblique collateral branches within layer 5 , and bifurcating in layers 2/3 to give off a terminal tuft. In a few cells, the apical dendritic bifurcation took place in a lower lamina (Fig. 12A) but further subdivision was always above layer 4. All apical dendritic shafts were covered with spines. The basilar dendrites originated from the soma in 6 to 8 main branches and arborized extensively (Figs. 7A, 8A), creating an almost spherical perisomatic field that was up to 600 pm in horizontal diameter, and extended from mid-layer 5 to the upper part of layer 6.
Morphological characteristicsof intrinsically bursting neurons Twenty-five IB neurons were stained, including 21 from sensorimotor and 4 from visual cortical slices. Thirteen neurons were reconstructed for morphometric analysis. All
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS
605
C
0
D
B
-62 -67 -72
0
0
I
50 ms Fig. 6. Synaptic responses of regular spiking (A) and burstgenerating (B-D) pyramidal cells in layer 5. A Regular spiking neuron in which orthodromic stimulation (filled dot) evokes typical hyperpolarizing fast (*) and slow (*+) IPSPs when stimuli fall at depolarized membrane potentials (upper trace). B-D: Responses of an IB neuron to orthodromic stimulation of increasing intensity. B Threshold stimulus evokes an EPSP at resting membrane potential (-67 mV) in a bursting cell. Same stimulus delivered at membrane potential of -60 mV evokes a short latency EPSP with a slowly rising depolarization from its peak, leading to a burst response at a latency of about 60 msec. C: Increase in stimulus intensity to 1.5x threshold evokes a larger amplitude subthresh-
old EPSP. The falling phase of the EPSP becomes progressively more prolonged as the membrane potential is depolarized (cf. -72, -67, and the two -62 mV traces of C). At depolarized membrane potentials (-62 mV) a depolarizing voltage-dependent component of varying amplitude is apparent (arrow) which may reach threshold to evoke an action ~ evoke an EPSP potential. D: Suprathreshold stimuli ( 2 . 5 threshold) which triggers an action potential followed at depolarized membrane potentials (-62 mV) by a slow voltage-dependent component (arrow). Spikes are truncated. Resting potentials -70 mV and -67 mV for cells A and B-D, respectively.
labelled neurons into the white matter, and in one cell the axonal course could be traced through the neostriatum and into the internal capsule (Fig. 8B). All of the IB neurons that were labelled in rat visual cortex were located in layer 5 and had a morphology that was very similar to the cells described above. One of 13 reconstructed burst-generating cells had an axonal arborization pattern that differed from the above. In this neuron (Fig. 12A) a more extensive local arborization of axons could be traced, with 6 identified collaterals that emerged from the main axon, curved toward the cortical surface along an almost vertical course, and TABLE 2. Anatomical Parameters of Regular Spikingand Bursting Neurons' terminated with only a few additional branches in layers 2/3. Average Average apical Average maximum soma area dendritic width* A single anatomically similar cell was identified by Landry soma diameter et al. ('84) among pyramidal tract neurons in the rat motor (um) (m2) (wm) cortex, and this special axonal arborization has been de749.3 i 95.1 8.5 2 1.6 37.6 i 3.1 IB (n = 13) scribed by Donoghue and Kitai ('81) for large pyramidal 463.3 i 160.7 4.6 2 1.9 28.1 i 5.3 RS (n = 9)
Burst Generating and Regular Spiking Layer 5 Pyramidal Neurons of Rat Neocortex Have Different Morphological Features YAEL CHAGNAC-AMITAI, HEIKO J. LUHMANN, AND DAVID A. PRINCE Department of Neurology and Neurological Sciences, Stanford University School of Medicine, Stanford, California 94305
ABSTRACT Intracellular recordings were obtained from pyramidal neurons in layer 5 of rat somatosensory and visual cortical slices maintained in vitro. When directly depolarized, one subclass of pyramidal neurons had the capacity to generate intrinsic burst discharges and another generated regular trains of single spikes. Burst responses were triggered in an all-or-none manner from depolarizing afterpotentials in most bursting neurons. Regular spiking cells responded to electrical stimulation of ascending afferents with a typical EPSP-IPSP sequence, whereas IPSPs were hard to detect in bursting cells. Orthodromic activation of the latter evoked a prominent voltage-dependent depolarization that could trigger a burst response. Intracellularly labelled bursting and regular spiking cells were located in layer 5b, but had distinctly different morphologies. Bursting neurons had a large pyramidal soma, a gradually emerging apical dendrite, and an extensive apical and basal dendritic tree. Their axonal collateral arborization was predominantly limited to layers 5/6. In contrast, regular spiking cells had a more rounded soma with abruptly emerging apical dendrite, a smaller dendritic arborization, and 2 to 8 ascending axonal collaterals that arborized widely in the supragranular layers. Both bursting and regular spiking cells had main axons that entered the subcortical white matter. These data show that some subgroups of pyramidal neurons within the deeper parts of layer 5 of rat cortex are morphologically and physiologically distinct and have different intracortical connections. Bursting cells presumably function to amplify and synchronize cortical outputs, whereas regular spiking output neurons provide excitatory feedback to neurons a t all cortical levels and receive a more effective orthodromic inhibitory input. These data support the hypothesis that differences in gross neuronal structure, perhaps even the subtle differences that distinguish subclasses of neurons in a given lamina, are predictive of underlying differences in the type and distribution of ion channels in the nerve cell membrane and connections of cells within the cortical circuit. Key words: visual cortex, somatosensory cortex, anatomy, physiology, epilepsy
Studies of mammalian cortical function have often emphasized patterns of neuronal connectivity as determinants of the properties of the system (Hubel and Wiesel, '62; Mountcastle, '57; for reviews, see Martin, '84; Parnavelas, '84). However, it is increasingly apparent that the operation of neuronal networks cannot be accurately predicted without knowledge of the properties of their individual elements (Llinas, '88). Even within the simplest cortical module, anatomically distinct neuronal types and subtypes can be categorized on the basis of features such as size, shape, patterns of dendritic arborization, position, connections, and transmitter content. Electrophysiological studies have suggested that different classes of cortical neurons may o 1990 WILEY-LISS, INC.
possess different intrinsic membrane properties that affect fundamental characteristics of their behavior, such as responses to synaptic inputs and the pattern of their spike output (McCormick et al., '85; Connors and Kriegstein, '86). These data support the hypothesis that variations in neuronal structure are predictive of underlying differences in the type and distribution of ion channels in the neuronal cell membrane (for review, see Prince and Huguenard, '88). One Accepted February 7,1990. Y. Chagnac-Amitai's present address is Department of Physiology, Ben Gurion University, Faculty of Health Sciences, P.O. Box 653, Beer-Sheva, Israel.
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS corollary of this proposal is that cells found to have different intrinsic membrane properties will be shown to be distinct neuroanatomically. To an extent, this has already been demonstrated for some of the general classes of neocortical pyramidal neurons. Experiments in vivo have shown that pyramidal tract neurons with fast versus slow axonal conduction velocities have different biophysical properties (Takahashi, '65; Calvin and Sypert, '76), and different structural characteristics (Deschihes et al., '79; Sakai and Woody, '88). Recent studies of cortical pyramidal neurons in vitro have indicated that subclasses with different firing behaviors can be distinguished. In earlier studies from this laboratory, a small population of pyramidal cells with intrinsic burstgenerating properties was found in guinea pig sensorimotor cortex (Connors et al., '82; McCormick et al., '85; see also recent report of burst-generating neurons in visual cortex, by Montoro et al., '88). In addition to this distinctive firing pattern, which seemed to be a rarity among neocortical cells, a point of great interest was the restricted laminar location of these neurons to layer 4 and superficial 5 (Connors et al., '82). Other work suggested that synchronized bursts in this population of cells might be important in the initiation of interictal epileptiform discharges in neocortical slices, since the latter appear to begin in the same lamina (Connors, '84; for review of the role of intrinsic burst generation in epileptogenesis see Prince and Connors, '86). Because of the small numbers of these intrinsically bursting neurons encountered in previous studies, an analysis of their electrophysiological properties and possible morphological differentiation from other neurons within the same cortical lamina was not possible. Recently, another population of bursting neurons was identified in layer 5b of the mouse barrel cortex (Agmon and Connors, '89). The proportion of intrinsically bursting cells among the recorded population in this lamina is apparently larger than in layer 4 to upper 5. We have further investigated the firing properties, synaptic potentials, and morphology of intrinsically bursting and nonbursting cells in layer 5b. Results indicate that intrinsically bursting and regular spiking subclasses of pyramidal neurons in this lamina have distinctly different morphological characteristics as well as different connections within the cortical circuit. Portions of these results have been published in abstract form (ChagnacAmitai et al., '88).
MATERIALS AND METHODS Tissue preparation and recording techniques Sprague-Dawley rats (about 1month old, 100-150 g) were deeply anesthetized with intraperitoneal sodium pentobar-
Abbreviations ACSF DAP EPSP GABA HRP IB IPSP LY PT
RS SMC
vc
artificial cerebrospinal fluid depolarizing afterpotential excitatory postsynaptic potential gamma aminobutyric acid horseradish peroxidase intrinsically bursting inhibitory postsynaptic potential Lucifer Yellow pyramidal tract regular spiking sensorimotor cortex visual cortex
599
bital, decapitated, and their brains rapidly removed and placed in chilled (5"C), oxygenated artificial cerebrospinal fluid (ACSF). A block of parietal tissue was dissected and glued to the platform of a vibratome (Lancer series 1000). Slices 400 to 450 pm thick were cut in a coronal plane from a region corresponding to the sensorimotor cortex (Zilles et al., '80; Zilles and Wree, '85) and placed in a chamber whose temperature was maintained at 35 1°C. The slices were kept a t the fluid/gas interface and continuously superfused with ACSF (0.8ml/min). The atmosphere over the slices was warmed and saturated with humidified 95% 0,/5% CO,. The ASCF contained (in mM) NaC1, 124; KC1, 5; CaCl,, 2; MgSO,, 2; NaHCO,, 22; dextrose, 10; NaH,PO,, 1.25; and, when saturated with 95 % 0,-5 % CO,, had a pH of 7.2. Microelectrodes were filled with 4 M potassium acetate and had resistances of 120 to 170 MO. Intracellular recordings were obtained after slices had been incubated for at least 1 hour in the chamber. Layer 4 and the border region between layers 5 and 6 could be visualized as clear bands in transilluminated or tangentially lit slices (Agmon, '88). We confined our microelectrode penetrations to the upper part of the lower band, which was located about two-thirds of the distance from the pia to the white matter. Analysis of counterstained sections from slices containing labelled neurons showed that this area corresponded to layer 5b. Single extracellular stimuli (180 to 200 psec duration) were delivered through a monopolar sharpened tungsten wire or a bipolar stimulating electrode situated in the underlying white matter or near the pial surface in the same radial column as the recorded cell. Data were recorded on magnetic tape (0 to 10 KHz) and digitized with an ISC-16 system (RC electronics) on a personal computer. On the basis of electrophysiological criteria and in agreement with a previous report (McCormick et al., '85), we distinguished between regular spiking cells (RS) and intrinsically bursting neurons (IB). Whereas the former responded to depolarizing current injection with one or several single action potentials, whose frequency was dependent on the level of membrane depolarization (e.g., McCormick et al., '85), IB cells generated a burst of at least three action potentials arising from a slow membrane depolarization that appeared to be due to a summation of depolarizing afterpotentials (DAPs) of successive spikes. We also observed cells that generated "doublets" at the onset of the depolarizing pulse, i.e., pairs of closely spaced action potentials. This pattern was observed most frequently in neurons impaled with electrodes containing Lucifer Yellow (LY) and may have been induced by lithium salts diffusing into the cell (for further discussion of this problem see Tseng and Haberly, '89) or due to different intrinsic firing properties (e.g., Calvin and Sypert, '76). For purposes of our analysis, we classified neurons showing a doublet firing pattern as RS cells. Neither IS neurons, nor those that fired an initial doublet, could be converted to IB neurons by increases in the amplitude of the depolarizing current pulse (up to 1nA).
Staining, immunohistochemistry, and data analysis Intracellular labelling was accomplished by impaling neurons with thin-walled microelectrodes filled with 5 to 10% Lucifer yellow (Sigma) in 1.0 M lithium acetate (resistances of 100 to 150 MO). The dye was injected by passing hyperpolarizing current pulses (1.0 to 1.5 nA, 600 msec duration) at 0.5 Hz for 5-20 minutes. Only 1or 2 cells were injected in each slice to assure the clear identification of
Y. CHAGNAC-AMITAI ET AL.
600
A
.
-.
I
1
-I
11 nA
2
1L ---5l-
20 ms
i
3
L 1 2 0 mV 1 1 . 4 nA
1
-
20 ms
2
Fig. 1. Characteristic responses of bursting and nonbursting pyramidal neurons to depolarizing intracellular current injection. A1,2: Suprathreshold current pulses of increasing amplitude in a regular-spiking cell evoke a train of action potentials that shows marked spike frequency adaptation. B: Responses of an intrinsically bursting neuron to current pulses of increasing amplitude (1-3). B1: Burst complex evoked a t threshold is preceded by an initial fast spike. B2,3:A larger amplitude current pulse alternately evokes trains of spikes (B2) and spikes followed by bursts (B3). Prominent depolarizing afterpotentials (DAPs)
follow each action potential in the spike train (B2, arrow). The burst complex includes spikes of varying amplitudes and durations that may appear in an all-or-none manner (arrow in B3). C: Another intrinsically bursting neuron in which the burst is evoked at long latency by a threshold current pulse ( C l ) and latency becomes shorter as current intensity increases (superimposed sweeps of C2). Calibrations following C for B and C. In this and all subsequent figures, intracellular recordings are shown in the upper traces and current pulses in the lower traces.
each neuron. Slices containing injected neurons were fixed with 4% paraformaldehyde in 0.1 M phosphate buffer overnight, and then cleared with dimethyl sulfoxide for 20 minutes. Cells were viewed and photographed through a Leitz microscope equipped with epifluorescence, and slices that contained well-labelled neurons were processed for immunohistochemistry. Slices were resectioned at 50 to 60 I m on a freezing microtome, and the resulting sections were rinsed, blocked, and placed for 1day in a primary antiserum (1:500) raised against LY (Taghert et al., ’82). Further processing utilized the avidin-biotin-peroxidase method (Vectastain ABC kit, Vector laboratories) followed by cobalt intensification and a 3-3’-diaminobenzidine (DAB, Sigma) reaction with H,O, using standard techniques. A Nissl counterstain was used to allow examination of the cell’s location in relation to the cortical layers.
Well-filled, anti-LY processed neurons were reconstructed with the aid of camera lucida drawings of serial sections (magnification 160 or 250 x ). Some measurements were made by using a Zeiss MOP3 digitizing tablet. IB and RS neurons were compared in terms of the following parameters: somatic area; width of the apical dendrite a t a level 80 pm from the center of the soma; and dendritic branch patterns by using a Sholl (’56) analysis. No attempt was made to apply corrections for tissue shrinkage or other distortions, since the goal was to compare IB and RS neurons rather than to obtain absolute measurements.
RESULTS A total of 73 intrinsically bursting (IB) and 28 regular spiking (RS) cells were recorded in layer 5 of both sensori-
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS TABLE 1. Electrophysiological Properties of Bursting and Regular Spiking Laver 5 Pvramidd Neurons' V, (mV) 73.5 + 4.3 (22) 67.3 + 6.3 (6) 72.0 + 3.3 (9) 73.7 i 3.0 (7)
IBneurons SMC IBneruons VC RS neurons SMC RS neruons VC
R, (MQ) 30.2 + 13.8(23) 36.2 i 8.9 (6) 31.1 * 13.7 (10) 33.6 i 16.4 (7)
1st spike amplitude (mV) 89.7 * 9.5 (23) 83.6 i 5.3 (6)' 95.4 f 9.8 (10) 92.1 i 6.1 (7)
'Restingmembranepotential (V,),inputre&tance (R,),andamplitudeoffiratspikeevokedby depolarizing pulse for intrinsically bursting (IB)and regularly spiking (Rs)neurons of sensorimotorcortex (SMC) and visual cortex (VC), expressedas mean i standard deviation. 'Signifcant (t-test,P < 0.01) difference in 1st spike amplitude between IB and RS neurons in
vc.
motor cortex (SMC) and visual cortex (VC). All cells had stable resting potentials more negative than -60 mV for a t least 10 minutes, an input resistance of a t least 20 MQ, and action potential amplitude of a t least 75 mV. IB neurons generated a complex of 3 or more action potentials riding on a depolarizing envelope, following injection of a depolarizing current pulse (Fig. 1B1; Connors et al., '82; McCormick et al., '85). RS cells fired a train of single fast action potentials to an injection of depolarizing current pulses (Fig. 1A). The RS cells did not seem to differ from neurons with similar properties recorded in other layers and previously described in studies from this laboratory (Connors et al., '82; McCormick et al., '85). Since our efforts were focused on an analysis of the properties of IB cells, most RS cells that were impaled were not studied, and the relative number of such neurons in our sample is not representative of their true incidence in layer 5. We estimate that IB cells constituted 50-60% of our impalements in somatosensory cortex and around 30% in visual cortex, when penetrations were made into this lam-
ina. Some electrophysiological properties of 51 randomly selected neurons are shown in Table 1. There were no significant differences between the mean resting membrane potential and input resistance of regular spiking versus bursting neurons. As previously noted (McCormick et al., '85), the spike amplitude of IB neurons was slightly lower than that of RS cells.
Characteristics of burst generationin neurons of layer 5b One prominent characteristic of all of the 73 IB cells studied in layer 5b was a depolarizing afterpotential (DAP) that followed each spike (Fig. 1B2, arrow). A significant proportion (ca. 50%) of RS cells also generated DAPs and about 30% of these neurons generated a spike doublet but by definition never a full burst, even when graded increases in the amplitude of depolarizing current pulses were applied. In each IB neuron, the burst developed after an initial fast spike. Figure 2B shows a cell in which the same depolarizing current pulse alternately evoked a burst and a single action potential. Superimposition of two such alternating responses shows that the full burst complex emerged from the peak of the DAP. A similar relationship is shown in the neuron of Figure 1B in which a burst discharge is triggered after the first ( B l ) or second (B3) fast spike. In this neuron, as in the cell of Figure 2B, identical current pulses delivered at 0.5 Hz (Fig. 1B2, 3) could evoke either single spikes or spikes and burst complexes that developed as all-or-nothing events. When evoked by brief current pulses, burst comDlexes in such neurons outlasted the depolarizing pulse (Fig. 4A).
--/
-
1
L
J
2
60 1
J
L
2
--f
4
L
20 mV 1 nA
J 2 0 ms -
10 ms Fig. 2. Intrinsically bursting neuron with variable firing pattern.
A1,2 Burst complex containing some small spikes (arrow in A2) develops after the first evoked action potential. A 3 - 4 Slight increase in current amplitude. Pairs of single spikes followed by DAPs are evoked
alternately with responses in which a burst follows the second action potential (cf. A3 and A4). B Same neuron as A. Superimposed sweeps to show that the first spike of the burst is triggered from the peak of the DAP. Note faster time base in B.
Y. CHAGNAC-AMITAI ET AL.
602
1
-
J
2
20 ms
2 0 mV 1 nA
4
-
40 ms
Fig. 3. Neuron that generates a graded rather than all-or-none burst complex. Low intensity current pulses evoke single action potentials followed by prominent DAPs (1,2).As current is increased, DAPs increase in amplitude and summate to generate the slow envelope of the burst (sweeps 3,4).
In 13 of the 73 bursting neurons, the slow events underlying burst generation appeared to be graded. For example, in the cell of Figure 3, prominent DAPs followed single action potentials evoked by depolarizing current pulses (traces 1-3). When the current was increased to more than 0.5 nA, the DAPs of the first four spikes summated to create a depolarizing envelope underlying a burst complex (Fig. 3, trace 4). Although the graded bursts generated in cells of this sort are somewhat different from the more typical all-or-none events that are shown in the neurons of Figures 1 and 2, recordings in all neurons strongly suggested that summated DAPs were an important component of the slow depolarization underlying burst generation. The amplitude of the slow envelope underlying the burst varied between 10 and 30 mV (measured from the base of the first spike in the complex to the base of the spike a t the peak of the envelope). Depolarization sufficient to reach spike threshold usually evoked a burst consisting of several fast spikes which decreased in amplitude and increased in duration during the course of the burst (Figs. 1-3). Small increases in the amplitude of the triggering pulse usually did not affect the stereotyped pattern of spike generation, but did shorten the latency for burst triggering (Fig. lC2). Progressive inactivation of burst spikes was prominent in neurons where higher amplitude underlying depolarizations were evoked (e.g., Figs. 1B1,2B, 4A). In some neurons, small amplitude, broad spikes appeared at the end of the burst (Fig. 2A, arrow; Fig. 2B), or on the falling limb of larger amplitude spikes (Fig. lB3, spike preceding the last spike of burst), suggesting that more than one spike initiating zone was present. The lower amplitude spikes had similarities to the high threshold calcium spikes that occur in the late portion of bursts in hippocampal CA3 pyramidal neurons (Wong and Prince, '78), and in other cells (Llinas and
Yarom, '81a,b). Burst discharges were followed by slow afterhyperpolarizations lasting up to 100 msec, whose amplitude increased a t more depolarized membrane potentials (Fig. 4A). We examined the effect of altering the membrane potential (V,) on burst generation by injecting DC depolarizing or hyperpolarizing currents across the membrane, and evoking bursts from various membrane potentials with long or short depolarizing current pulses. Bursts could be evoked a t hyperpolarized membrane potentials when the depolarizing current pulse was sufficiently strong. In a few neurons, burst generation was blocked when V, was held positive to about -60 mV. This suggested that a low threshold calcium spike might contribute to burst generation, a finding similar to that reported in some bursting neurons of layer 4-upper 5 (Connors et al., '82; McCormick et al., '85). In all other IB neurons tested in this way, however, the burst could be evoked when the membrane potential was held depolarized, close to the firing level (Fig. 4A3). About 90% of the IB neurons generated a burst followed by a train of spikes when depolarized from resting membrane potential by a prolonged current pulse (Fig. 5B). About 10% of IB cells were capable of generating repetitive burst discharges during prolonged depolarizing pulses that kept V, close to the firing level (Fig. 5A; see also Agmon and Connors, '89). In these instances, repetitive bursts might arise from a depolarized V, without preceding hyperpolarizations (Fig. 5A3).
Synaptic responses of intrinsically bursting and regular spiking layer 5 neurons We compared the responses of RS and IB neurons to orthodromic activation produced by stimulation of either white matter or the subpial region in the same radial column
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS
A
603
Lkk J l I -
-72
r..................... l
-63
...........................
-57
J--L
1
............................
I I
2
-75-4
3
I
1
2
120 mV 1 nA 20 ms
1 2 0 mV 1 1 . 4 nA 40 ms
Fig. 4. Effects of membrane potential on burst generation in two neurons. A Burst evoked by a brief current pulse increases in duration as the membrane potential is depolarized from -72 mV (sweep 1) to -63 and -57 mV (sweeps 2 and 3). Note the increase in amplitude of the burst after-hyperpolarization as the membrane potential becomes more positive. B Bursts evoked by long current pulses in another neuron also increase in duration as membrane potential is depolarized by direct current as in A.
as the impaled cell. Such stimuli evoked typical sequences of excitatory and inhibitory postsynaptic potentials (EPSPs and IPSPs) in RS cells, as previously reported (Connors et al., '88). A t threshold, an isolated, short-latencyEPSP could be elicited (not shown). As the stimulus intensity was increased, early and late IPSP components emerged, that had reversal potentials of about -70 and -85 mV, respectively (Fig. 6A). The pattern of synaptic responses to stimuli delivered at the same sites was significantly different in IB neurons, as is shown in the example of Figure 6B. At low stimulus intensities, a short latency EPSP could be evoked (Fig. 6B, -67 mV trace). As the stimulus intensity increased, a threshold was reached at which the burst complex was triggered, a t variable latencies, as an all-or-none event (not shown). The orthodromically evoked burst itself was similar in configuration to the burst evoked by direct depolarization of the same neuron. A small depolarization of the membrane potential might render a previous subthreshold synaptic stimulus effective in evoking the burst complex (Fig. 6B, -60 mV trace). When this occurred, a slow ramp-like depolarization led from near the peak of the EPSP to the first spike of the burst (Fig. 6B, arrow). Paradoxically, stimuli of increased intensity, delivered with the neuron at the same membrane potential, would fail to trigger the burst, and instead evoke a large-amplitude subthreshold EPSP (Fig. 6C), or (if threshold was reached) a single action potential (Fig. 6D). The disappearance of the burst at higher stimulus intensities suggested recruitment of an underlying inhibitory conductance that shunted the burst-related depolarization. However, in most of the IB neurons it was not possible to demonstrate the fast and slow IPSP, even by evoking synaptic responses at depolarized
V,s. A typical EPSP-IPSP sequence could be clearly demonstrated in only a few bursting cells. In most IB cells, a prominent voltage-dependent depolarization was evoked by synaptic activation at resting potential. Examples are shown in Figure 6C,D. Subthreshold (Fig. 6C) and threshold (Fig. 6D) stimuli for eliciting an action potential from the short latency EPSP, evoked a slow depolarization that arose shortly after the peak of the EPSP and was independent of spike generation. The amplitude and duration of this response component were variable from stimulus to stimulus and increased as the membrane was depolarized from the resting potential towards the spike threshold. When suprathreshold stimuli were applied, a short latency spike was evoked from the initial EPSP component, followed by a prolonged (5100 msec), slow depolarization with a peak latency of 50-75 msec (Fig. 6D, -62 mV trace, arrow). The onset of the slow depolarization was influenced by the stimulus intensity, so that it arose close the peak of the EPSP with stimuli that were close to threshold (Fig. 6B, -60 mV trace) and from the repolarizing limb of the EPSP following more intense stimuli (Fig. 6D, -62 mV trace). This finding suggested the presence of a superimposed short latency IPSP. The graded subthreshold slow depolarizations appeared to merge into the ramp-like potentials that triggered delayed spikes or bursts (Fig. 6B-D).
Anatomical identification of intrinsically bursting and regular spiking neurons Cells were impaled with LY-containing electrodes following penetrations into the deeper portions of layer 5, using
Y. CHAGNAC-AMITAI ET AL.
604
J
I
I
3
1
3 JI
7
100 ms Fig. 5. A: Examples of repetitive burst-firing patterns in three layer 5 bursting neurons during prolonged depolarizing current pulses. A1,2: The initial burst in these neurons contains 4 and 3 spikes, respectively, followed by repetitive firing consisting of single spikes, doublets, or bursts of three spikes. A3: Neuron fired repetitive bursts of up to 6
spikes and frequent doublets during the 1 second depolarizing current pulse. Note that membrane did not repolarize to resting level between bursts. B1-3: Firing pattern in another neuron where a 1 second depolarizing current pulse of increasing intensity (top to bottom segments) evoked an initial burst followed by a train of spikes.
landmarks described above. Spike firing patterns were characterized within 2 minutes of impalement by passing depolarizing current pulses, and neurons were then classified as IB or RS cells. This rapid determination of electrophysiological behavior was made before any clear effect of Li+ injection on spike activities could be noted. In agreement with previous reports (Mayer et al., '84), alterations in spikes occurred within a few minutes after beginning the application of hyperpolarizing current pulses, a t which time some broadening of action potentials and a prolongation of the burst complex was noted. The dye injection seemed to facilitate generation of repetitive burst discharges during prolonged current pulses in some cells, but no neuron that was identified as a RS cell subsequently developed a burst firing pattern. Thus, recordings with LY-filled electrodes allowed clear differentiation of IB and RS neurons. The recordings obtained were not used for electrophysiological analysis.
25 labelled neurons were similar in their soma-dendritic features, as will be described in detail below (see Table 2), and all were situated in the lower half of layer 5. Most somata were pyramidal in shape (Figs. 7A, 8A, 9). Somatic area averaged 749.3 2 95.1 pm2 (mean 2 standard deviation; n = 13), and the average maximum diameter was 37.6 3.1 pm (n = 13) (see Table 2). When somata were viewed in counter-stained sections, it was clear that these neurons were among the largest in this area of cortex. The apical dendrite emerged from the cell body in a gradual, tapering manner, making it difficult to establish the boundary between the soma and the ascending dendrite. The diameter of the apical dendrite, measured 80 pm from the center of the cell body, averaged 8.5 1.6 pm (n = 13). From its origin, the apical dendrite took a straight course towards the pial surface, giving off oblique collateral branches within layer 5 , and bifurcating in layers 2/3 to give off a terminal tuft. In a few cells, the apical dendritic bifurcation took place in a lower lamina (Fig. 12A) but further subdivision was always above layer 4. All apical dendritic shafts were covered with spines. The basilar dendrites originated from the soma in 6 to 8 main branches and arborized extensively (Figs. 7A, 8A), creating an almost spherical perisomatic field that was up to 600 pm in horizontal diameter, and extended from mid-layer 5 to the upper part of layer 6.
Morphological characteristicsof intrinsically bursting neurons Twenty-five IB neurons were stained, including 21 from sensorimotor and 4 from visual cortical slices. Thirteen neurons were reconstructed for morphometric analysis. All
MORPHOLOGY, PHYSIOLOGY OF LAYER 5 PYRAMIDAL CELLS
605
C
0
D
B
-62 -67 -72
0
0
I
50 ms Fig. 6. Synaptic responses of regular spiking (A) and burstgenerating (B-D) pyramidal cells in layer 5. A Regular spiking neuron in which orthodromic stimulation (filled dot) evokes typical hyperpolarizing fast (*) and slow (*+) IPSPs when stimuli fall at depolarized membrane potentials (upper trace). B-D: Responses of an IB neuron to orthodromic stimulation of increasing intensity. B Threshold stimulus evokes an EPSP at resting membrane potential (-67 mV) in a bursting cell. Same stimulus delivered at membrane potential of -60 mV evokes a short latency EPSP with a slowly rising depolarization from its peak, leading to a burst response at a latency of about 60 msec. C: Increase in stimulus intensity to 1.5x threshold evokes a larger amplitude subthresh-
old EPSP. The falling phase of the EPSP becomes progressively more prolonged as the membrane potential is depolarized (cf. -72, -67, and the two -62 mV traces of C). At depolarized membrane potentials (-62 mV) a depolarizing voltage-dependent component of varying amplitude is apparent (arrow) which may reach threshold to evoke an action ~ evoke an EPSP potential. D: Suprathreshold stimuli ( 2 . 5 threshold) which triggers an action potential followed at depolarized membrane potentials (-62 mV) by a slow voltage-dependent component (arrow). Spikes are truncated. Resting potentials -70 mV and -67 mV for cells A and B-D, respectively.
labelled neurons into the white matter, and in one cell the axonal course could be traced through the neostriatum and into the internal capsule (Fig. 8B). All of the IB neurons that were labelled in rat visual cortex were located in layer 5 and had a morphology that was very similar to the cells described above. One of 13 reconstructed burst-generating cells had an axonal arborization pattern that differed from the above. In this neuron (Fig. 12A) a more extensive local arborization of axons could be traced, with 6 identified collaterals that emerged from the main axon, curved toward the cortical surface along an almost vertical course, and TABLE 2. Anatomical Parameters of Regular Spikingand Bursting Neurons' terminated with only a few additional branches in layers 2/3. Average Average apical Average maximum soma area dendritic width* A single anatomically similar cell was identified by Landry soma diameter et al. ('84) among pyramidal tract neurons in the rat motor (um) (m2) (wm) cortex, and this special axonal arborization has been de749.3 i 95.1 8.5 2 1.6 37.6 i 3.1 IB (n = 13) scribed by Donoghue and Kitai ('81) for large pyramidal 463.3 i 160.7 4.6 2 1.9 28.1 i 5.3 RS (n = 9)
