Autaptic inhibitory currents recorded from interneurones in rat cerebellar slices

Abstract

1. While the presence of autapses in the brain is indicated by a large body of morphological evidence, the functional role of these structures has remained unclear. To probe for autaptic currents, we have recorded current responses following short somatic depolarizing pulses in Cl⁻-loaded interneurones (stellate and basket cells) from rat cerebellar slices (animals aged 27-39 days).

2. In ≈20 % of the recordings, fluctuating inward current transients were obtained with a latency of 1.15-2.45 ms (measured from the peak of the depolarization-induced Na⁺ current; n = 10).

3. These transients were blocked by bicuculline and were sensitive to the extracellular Ca²⁺ concentration.

4. Assuming low release probability, as suggested by the high failure rate (0.65-0.92, n = 10), quantal sizes ranging from 21 to 178 pA (-70 mV; n = 10) were calculated from a variance analysis of autaptic current amplitudes.

5. We conclude that ≈20 % of interneurones have a functional autapse. Autaptic currents may inhibit firing of interneurones during high frequency discharges.

Autapses are synaptic contacts linking two parts of the same neurone. In electrophysiological recordings, autapses are manifest as synaptic signals which are directly elicited by somatic stimulation of the recorded cell. Autaptic signals have been described in ‘microculture’ preparations, where neonatal mammalian neurones are induced to grow on a restricted substrate (Furshpan, MacLeish, O'Lague & Potter, 1976), as well as in neurones of the buccal ganglion of Aplysia (where they were called ‘self-inhibitory synaptic potentials’: Gardner, 1977). In mammals, abundant morphological evidence for the existence of autaptic contacts has been obtained in various brain regions, for example the cortex (Van der Loos & Glaser, 1972), the striatum (Park, Lighthall & Kitai, 1980) and the cerebellum (Purkinje cells; King & Bishop, 1982). Paradoxically, no functional correlates of autaptic contacts have been reported until now in native vertebrate nervous tissue, raising the suspicion that such contacts may not be functional, or that they may have a transient existence before being recognised as erroneous and eliminated. Yet, electron microscopy studies suggest that autaptic contacts have a normal morphology. Recently, a series of morphological studies in the hippocampus and visual cortex have revealed that the extent of autaptic connections in these brain regions is much higher than previously suspected, that many inhibitory and excitatory neurones establish multiple autaptic contacts with themselves, and that these contacts survive into adulthood (e.g. Lübke, Markram, Frotscher & Sakmann, 1996; Tamas, Buhl & Somogyi, 1997; reviewed in Bekkers, 1997).

In this context, we have been interested to observe signals in interneurones (stellate and basket cells) of the molecular layer of cerebellar slices, which could only be accounted for by autaptic connections. This work is a description of these currents.

METHODS

General methods to obtain cerebellar slices, to identify interneurones in the molecular layer and to record from them, have been described before (Llano & Gerschenfeld, 1993). Rats (27-39 days old) were decapitated after cervical dislocation in deep anaesthesia (Metofane; Mallinckrodt Veterinary, Bray, Ireland), and slices were prepared as in the reference above.

Compositions of intra- and extracellular recording solutions

The intracellular recording solution contained (mM): 150 KCl, 4.6 MgCl₂, 0.1 CaCl₂, 1 EGTA, 10 potassium Hepes, 0.4 NaGTP and 4 NaATP. The external solution contained (mM): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃ and 10 glucose. In some experiments the concentration of Ca²⁺ was reduced to 1 mM while the concentration of Mg²⁺ was increased to 2 mM. Bicuculline methochloride (10 μM; Tocris Cookson, Bristol, UK) was used in other experiments to block GABA_A receptors.

Recording autaptic currents

Currents were recorded with a sample interval of 50 or 100 μs, and were filtered at a fifth of the sampling frequency. To obtain autaptic currents, axonal action potentials were elicited by applying short depolarizing voltage pulses in the soma of voltage-clamped interneurones. At interneurone-Purkinje cell synapses, this protocol elicits postsynaptic responses that are undistinguishable from those obtained following presynaptic action potentials before subjecting the presynaptic cell to voltage clamp (Vincent & Marty, 1996; Pouzat & Hestrin, 1997). The amplitude and duration of depolarizing voltage pulses were adjusted such that a regenerative voltage-dependent Na⁺ current would be registered in the soma, while the amplitude of the subsequent K⁺ current was kept at a minimal value. In this way, conductances linked to the axonal action potential had subsided by the time that autaptic currents were recorded. Stimulation frequencies were 0.2-0.5 Hz. Series resistance values during recording ranged from 10 to 35 MΩ, and were partially compensated (50-75 %).

All experiments were performed at room temperature (20-23°C).

Analysis of autaptic currents

Failures were identified from examination of individual sweeps. Means and associated standard deviations (s.d.) were calculated from the ensemble of failure sweeps (Fig. 1Aa and Ac). To confirm that failures were identified correctly, it was verified that the mean s.d. trace did not contain any time-dependent signal after the end of the Na⁺ currents. The mean failure trace was then subtracted from each individual sweep (Fig. 1B). The amplitude of individual subtracted traces was calculated as the difference between peak and baseline currents as means within windows (typically 0.5 ms long). The positions of the windows were adjusted from examination of the mean subtracted trace (Fig. 1Ab), and they were the same for all traces in a given experiment.

Figure 1. Demonstration of autaptic currents in cerebellar interneurones.

Aa, mean currents obtained in response to repetitive voltage pulses (holding potential, -70 mV; test potential, -30 mV; pulse frequency, 0.33 Hz). The two large inward current transients during and immediately following the pulse reflect activation of Na⁺ channels. In some trials, an inward transient current was observed starting about 2 ms after the end of the pulse, reflecting autaptic IPSCs, while in other trials, no autaptic current was recorded (failures). The continuous line is the mean of 100 consecutive current traces. The dashed line is the mean of the 69 traces in this series that were identified as failures. Ab, difference between general mean and failure mean, displayed with an expanded current scale. Ac, s.d. traces of the entire series and of the failures, showing that failures did not contain any fluctuating signal having kinetic properties similar to that of the mean autaptic IPSC. B, 23 successive superimposed failure-subtracted sweeps. C, amplitude distribution from the same results (400 sweeps; failure rate, 0.63; bin width, 5 pA; the highest bin contains 106 events). Failures can be unambiguously distinguished in the distribution of response amplitudes. The mean and s.d. value for these data (including failures) was 31 ± 48 pA. Noise s.d. was 3.7 pA. Inset, latency distribution for the same data (150 successful responses). Mean latency (counted from the peak of the Na⁺ current): 2.45 ± 0.32 ms (mean ± s.d.). Data from cell 1.

Stationary regions selected for quantitative analysis were derived from a covariance analysis of the amplitudes (serial correlogram; Perkel, Gerstein & Moore, 1967). To obtain the noise-corrected s.d. of the autaptic currents, the variance of peak amplitudes was measured during a period of stationary response, the variance of the noise was subtracted, and the square root of the difference was calculated. The variance of the noise was obtained from the amplitude distribution of the failures.

Latencies were measured from individual successful responses as the time between the peak of the Na⁺ current and the half-amplitude of the autaptic response.

To determine kinetic properties of autaptic currents, single failure-subtracted responses were aligned with respect to the half-maximum amplitude point. A rise time (10-90 %) and decay time (calculated from the mean of the two time constants of a two-exponential fit, weighted according to the relative amplitudes of the components) were calculated on the mean trace.

The confidence intervals in the Var/< I > column in Table 1 are the s.d. values of the (Gaussian) distributions of the variance to mean ratio obtained from 500 bootstrap replicates (Efron & Tibshirani, 1993).

Table 1.

Characteristics of autaptic currents

Cell number	< I > (pA)	S.D. (pA)	Failure frequency	Success mean (pA)	Var/< I > (pA)	Rise time (ms)	Decay time (ms)	Latency (ms)
1	(110) 31	(90) 48	(0.2) 0.65	(138) 89	75 ± 5	0.55	10	2.45
2	7.2	21	0.88	60	64 ± 4	0.50	86	2.05
3	3.7	10	0.90	37	27 ± 2	0.70	11	1.15
4	12	21	0.65	34	40 ± 4	0.55	7.7	2.05
5	12	32	0.82	67	82 ± 8	n.d.	7.1	2.05
6	10	15	0.74	38	21 ± 4	n.d.	10	1.50
7	12	25	0.91	133	56 ± 5	n.d.	6	1.15
8	11	41	0.92	138	150 ± 10	0.45	11.6	1.90
9	16	24	0.66	47	36 ± 3	0.55	15.2	2.00
10	10	43	0.91	111	178 ± 63	0.35	13.6	1.80
	12.5 ± 2.3	28 ± 4	0.80 ± 0.04	75 ± 13	73 ± 17	0.52 ± 0.04	10.1 ± 0.9	1.81 ± 0.13

Results were gathered as illustrated in Fig. 1. The second column indicates the mean of peak amplitudes of autaptic current including failures. The third column indicates the S.D. values from these measurements, after correction for the background noise. The sixth column shows the ratio of the (noise corrected) variance to the mean current, which gives an estimate of the quantal size. The bottom line indicates the mean ± S.E.M. of the results for 10 cells. In cell 1, there was a short period (about 5 min) at the beginning of the recording with high mean current and success frequency. These values are indicated in parentheses, but values from the later stationary period (about 1 h in duration) were used for the overall means. Such a downward trend was not observed in the other recordings. n.d., not determined.

RESULTS

Autaptic IPSCs in interneurones

The experimental protocol used to reveal autaptic currents is presented in Fig. 1. Recording pipettes were filled with a KCl-rich solution. Interneurones were held at -70 mV under voltage clamp. Short depolarizing voltage pulses (1 ms in duration and 40 mV in amplitude in the case of Fig. 1) were applied repetitively. These pulses elicited a sharply rising inward current which quickly returned to the baseline upon repolarization, representing the voltage-dependent Na⁺ current previously described with this preparation (Llano & Gerschenfeld, 1993). This current has an amplitude of 0.5 nA in the example shown. Following the pulse, a delayed inward current transient followed shortly after the Na⁺ tail current (Fig. 1Aa). Unlike the early Na⁺ current, this transient fluctuated widely from trial to trial (Fig. 1Ac), indicating that it could have a synaptic origin. It actually failed in a majority of the trials. Failures could be unambiguously separated from successful responses, as shown in Fig. 1Ac and C. Failure-subtracted traces are shown in Fig. 1B, and Fig. 1C illustrates the histogram of peak response amplitudes. The mean amplitude including failures was 24 pA (Fig. 1Ab). The transient currents had a mean latency of 2.45 ms after the peak of the early inward current (Fig. 1B; inset in Fig. 1C), with limited variability among trials (range, 1.8-3.4 ms).

These results suggest that short voltage pulses elicit synaptic currents similar to IPSCs recorded at many GABAergic brain synapses. Since the stimulated cell is inhibitory, and since latencies are short and reproducible, these currents are not generated by a polysynaptic pathway, but rather by the stimulated cell, or by another interneurone which would be directly stimulated by the recorded cell through an electrical junction. As further discussed below, electrical coupling occasionally occurs between interneurones, but is neither sufficiently prevalent nor sufficiently strong to provide a plausible external presynaptic source for the transient currents. Therefore, these currents are most probably autaptic, and will be referred to as such in the following.

Characteristics of autaptic currents

Quantitative results from autaptic currents obtained under the conditions of Fig. 1 (high intracellular Cl⁻, age group 27-39 days) are summarized in Table 1. Of a total of forty-seven cells that were investigated, ten (21 %) displayed clear autaptic currents. Mean current amplitudes including failures ranged from 3 to 31 pA; failure rates ranged from 0.65 to 0.92; and mean latencies ranged from 1.15 to 2.45 ms. The ranges overlap heavily or totally with those (6-447 pA, 0.15-0.94, 0.7-3.1 ms, respectively) obtained in a series of twenty paired recordings of interneurone-interneurone synapses (Kondo & Marty, 1998; data for 14- to 21-day-old rats; holding potential was -60 mV in that study instead of -70 mV here). Mean amplitudes had a tendency to be smaller for autaptic currents (12.5 ± 2.3 pA) than for ordinary synapses (56 ± 22 pA), but the difference was not significant (P > 0.05; Student's t test). Failure frequencies were on average higher for autapses (0.80 ± 0.04) than for ordinary synapses (0.62 ± 0.05, P < 0.001). Rise times were not statistically different in the two sets of data (1.81 ± 0.13 ms for autapses; 1.65 ± 0.12 ms for other synapses). Under the assumption of a low release probability, the ratio of the variance to the mean of the peak IPSC amplitudes can be used as an estimate of the mean quantal size. This assumption is likely to apply here in view of the high failure rate of autaptic currents. The numbers obtained in this manner ranged from 21 to 178 pA, again similar to the range 26-98 pA derived from more direct quantal analysis of paired recordings of interneurone-interneurone synapses (Kondo & Marty, 1998). It is also interesting to note that apparent quantal sizes vary markedly from one experiment to the other at autaptic contacts (Table 1), and also at interneurone-interneurone synapses (Kondo & Marty, 1998). These results are also consistent with the marked heterogeneity in miniature IPSC amplitudes in this preparation (Llano & Gerschenfeld, 1993; Auger & Marty, 1997; Nusser, Cull-Candy & Farrant, 1997).

Rise times (10-90 %) of autaptic currents averaged 0.52 ± 0.04 ms, and were not significantly different from those obtained in paired interneurone-interneurone recordings (0.62 ± 0.03 ms; Kondo & Marty, 1998). The decay of autaptic currents, like that of spontaneous IPSCs (Llano & Gerschenfeld, 1993), could be fitted either to one exponential or to the sum of two exponentials. The mean value across experiments of weighted decay time constants was 10.1 ± 0.9 ms (Table 1), similar to the value observed for spontaneous IPSCs (mean, 9.3 ms; data for 13- to 26-day-old rats; Nusser et al. 1997).

Autaptic currents are blocked by bicuculline

If the autaptic currents are, like IPSCs at interneurone- interneurone synapses, mediated by GABA_A receptors, they should be blocked by bicuculline. Bath application of 10 μM bicuculline indeed entirely blocked the autaptic currents (Fig. 2; n = 3), while the preceding Na⁺ current transients were not affected. The effects of bicuculline were readily reversible (Fig. 2).

Figure 2. Autaptic currents are blocked by bicuculline.

A, peak amplitudes of individual responses to successive trials in control solution, in the presence of 10 μM bicuculline, and after washing. Time is counted from the onset of whole-cell recording. B, representative single failure-subtracted traces during the three conditions. Data from cell 5.

Autaptic currents are sensitive to the external Ca²⁺ concentration

We next investigated the sensitivity of putative autaptic currents to modifications of the extracellular Ca²⁺ concentration. As exemplified in Fig. 3, lowering the Ca²⁺ concentration from 2 to 1 mM resulted in a reversible reduction of the mean autaptic response (to 12 % of the original value in the case illustrated) and in a significant increase in the frequency of failures. Reversible reductions to 46 and 39 % of the control were obtained in two other experiments. Thus, autaptic currents display a sensitivity to the extracellular Ca²⁺ concentration which is similar to that exhibited by ordinary synaptic currents.

Figure 3. Autaptic currents are sensitive to the external Ca²⁺ concentration.

A, peak amplitudes of individual responses to successive trials in control solution, in a low Ca²⁺ solution (1 mM), and after washing. Time is counted from the onset of whole-cell recording. B, mean traces from the three periods in A (mean of 50 sweeps). The timing of depolarizing pulses is indicated by short bars above the records. Out of 150 trials in the control period, 103 failures were found (failure frequency: 0.69), whereas in the same number of trials in the low Ca²⁺ solution, 141 failures were observed (failure frequency: 0.94), a highly significant difference (P < 10⁻⁶). The variance to mean ratio decreased from 75 ± 5 to 45 ± 6 pA. Data from cell 1.

Autaptic currents are not due to electrical junctions among interneurones

While the results presented so far strongly indicate that the signals originate at a direct GABAergic synaptic connection, they leave the possibility open that the depolarizing pulse in the recorded cell induces an action potential in a neighbouring interneurone and that the recorded signals are due to an interneurone-interneurone synapse. However, this scheme would require a very strong and consistent electrical coupling among interneurones, which is not supported by experimental evidence. In the abundant literature describing the anatomy of the cerebellum using electron microscopy, there is an isolated report by Sotelo & Llinas (1972) that gap junctions are present among stellate cells of several species including the rat. But there was no evidence of spread of neurobiotin staining from the recorded cell to its neighbours in previous studies (Vincent & Marty, 1996; Pouzat & Hestrin, 1997), whereas a clear spread is visible with the same technique among electrically connected Bergmann glial cells (Clark & Barbour, 1997). Furthermore paired recordings failed to reveal any correlation between the firing patterns of neighbouring interneurones (Vincent & Marty, 1996). Very occasionally (< 5 % of the experiments), biphasic current signals are recorded in voltage-clamped interneurones, presumably reflecting spontaneous firing of an interneurone which is coupled electrically to the recorded cell (L. Forti, personal communication). These currents have peak amplitudes of less than 20 pA at a holding potential of -60 mV. Direct injection of depolarizing currents into current-clamped interneurones during a period of 1 ms (a generous estimate of the duration of interneurone action potential duration) was performed to see whether such a current could elicit firing. It was found that interneurones would fire under these conditions only for input currents in excess of 500 pA (results not shown). Thus electrical junctions among interneurones are rare and, when present, are much too weak to insure firing synchrony. Likewise in the buccal ganglion of Aplysia, gap junction conductances are too low to account for autaptic potentials (Gardner, 1977). Overall we conclude that the recorded signals are not indirectly mediated by activation of another interneurone, and that they are bona fide autaptic currents.

DISCUSSION

Autaptic currents in cerebellar interneurones

The following findings support the identification of the spike-induced inward current transients reported in this work as autaptic GABAergic currents: (1) these signals fluctuate from trial to trial, as expected from synaptic currents; (2) they have a short and reproducible latency, indicating that they are not elicited by a polysynaptic path; (3) they are depressed by lowering the external Ca²⁺ concentration; and (4) they are reversibly blocked by bicuculline.

Morphological evidence from previous studies shows that axons and dendrites of interneurones are in a single plane, and that axons often loop around the soma in the dendritic tree before taking their definitive route parallel to the Purkinje cell layer, supporting the possibility of autaptic connections (Pouzat & Hestrin, 1997; Llano, Tan & Caputo, 1997). Indeed several regions of close axodendritic apposition, which are presumptive autaptic release sites, can be seen in each of the two cells shown in Fig. 5 of Llano et al. 1997. While definitive proof for the existence of autaptic contacts must await more thorough morphological studies employing electron microscopy, the evidence at hand supports the possibility of such contacts.

The present results contrast with those recently obtained in cortex pyramidal cells where, in spite of convincing evidence for the presence of multiple autaptic sites, autaptic EPSPs could not be detected (Lübke et al. 1996). In pyramidal cells, back propagating regenerative signals in the dendrites may obscure the autaptic signals (Lübke et al. 1996); such signals are unlikely to occur in the short, electrically compact dendrites of cerebellar interneurones.

Comparison between autaptic currents and intercellular synaptic currents

The parameters describing the latency, mean quantal size, rise time and decay kinetics of autaptic currents were all similar to those previously obtained for intercellular synapses. On the other hand, failure rates were higher on average, and mean amplitudes including failures were smaller for autaptic currents than for intercellular IPSCs. However, these discrepancies do not necessarily imply genuine differences between autapses and non-autaptic synapses. They could simply reflect the gap between the age groups of the two sets of data (14-21 days for intercellular synapses versus 27-39 days for autapses) since at interneurone-Purkinje cell synapses, the mean amplitude and success rate of IPSCs decreases dramatically from the first to the second of these age groups (Pouzat & Hestrin, 1997). Overall the results indicate that autapses have essentially the same functional properties as conventional synapses.

Functional significance of autaptic currents

We estimate that 21 % of interneurones have functional autaptic currents. One way to interpret the functional role of autapses is to see them as one component of local interneurone-interneurone inhibition. Following parallel fibre activation, interneurones are first excited, and then inhibited upon activation of the inhibitory interneurone- interneurone connections. Autapses should participate in this inhibition in proportion to the strength of the autaptic currents and to the fraction of autaptic versus intercellular inputs. Since on average autaptic currents are similar in size to intercellular synaptic currents, while each interneurone on average receives inhibition from 4.25 other interneurones (Kondo & Marty, 1998), the part of autapses in this inhibitory action is roughly 24 % for those interneurones that do have an autapse. When taking into account the proportion (21 %) of interneurones that have an autapse, this ratio drops to ∼5 %, indicating that the contribution of autapses is small.

A second possible role for autaptic currents is to regulate the firing of neurones (White & Gardner, 1981). In a series of cell-attached recordings performed in the age group 27-39 days under control conditions, we found that the spontaneous firing rate of interneurones is rather low, with a mean of 9.3 ± 0.6 Hz (mean ± s.e.m., n = 22). For such frequencies the mean spacing between spikes is much longer than the duration of autaptic IPSCs, so that no substantial effect of autapses on the firing pattern would be expected. However, under special circumstances such as, for instance, during activation by the metabotropic glutamate receptor agonist trans-ACPD (Llano & Marty, 1995), interneurones fire in bursts with intraburst frequencies of 40 Hz or more. It remains to be seen whether under such conditions the activation of autaptic currents participates in the regulation of the rate of firing.

Even if the functional role of autaptic currents turns out to be modest, the finding that they can be reliably measured in central nervous system slice preparations may bring considerable benefits for the study of basic mechanisms of synaptic transmission in the brain. Autapses can be used to study synaptic currents while controlling the composition of the pre- (and post-) synaptic cytosolic solution. The benefit of using autapses for such purposes should be high in slice preparations, because of the intrinsic difficulties of obtaining connected synaptic pairs in slices. Cerebellar interneurones are particularly attractive in this respect because of the short extension and simplicity of their axonal tree, which should facilitate the effective dialysis of test substances from the soma towards autaptic release sites. The striking similarities revealed in the present work between the properties of autaptic currents and those of interneurone- interneurone synaptic currents indicate that conclusions of quantitative analyses of release mechanisms could easily be transferred from the former to the latter preparation.