INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Many synaptic systems in the mammalian CNS exhibit use-dependent, short-term depression (for reviews see Zucker 1989, 1999; Zucker and Regehr 2002). High levels of transmitter release probability in these synapses is believed to be a key factor governing the magnitude of depression (Dittman et al. 2000). However, the co-existence of use-dependent presynaptic processes that enhance transmitter release on millisecond-to-second time scales complicates interpretation of synaptic depression data (Magleby 1987; Zucker 1999). A number of recent studies have dealt with different aspects of synaptic maturation (e.g., Bellingham et al. 1998; Bolshakov and Siegelbaum 1995; Brenowitz and Trussell 2001; Chuhma et al. 2001; Iwasaki and Takahashi 2001; Mozhayeva et al. 2002; Pouzat and Hestrin 1997; Reyes and Sakmann 1999; Taschenberger and von Gersdorff 2000), but changes in use-dependent synaptic modulation with age remain incompletely understood. The present work examined changes during the first 12 days of postnatal life in synaptic transmission using a well-known spinal cord synaptic system that is representative of the anatomy of the majority of CNS connections.

The monosynaptic connections made by group Ia muscle spindle afferents directly onto alpha motoneurons were the first excitatory synapses in the CNS to be studied with intracellular recording in the adult cat (Coombs et al. 1955). For decades, this system served as a model for thinking about mechanisms of synaptic transmission in the CNS (see Burke and Rudomin 1977; Eccles 1964; Redman 1990). Dorsal root afferents establish synapses on motoneurons before birth in the rat (Kudo and Yamada 1987; Snider et al. 1992; Ziskind-Conhaim 1990) and make strong excitatory connections with homonymous motoneurons in neonatal mice (Lev-Tov and Pinco 1992; Mears and Frank 1997). Anatomical similarities with the cat provide strong inferential evidence that these synapses represent group Ia afferent terminations in rodents and that term will be used below.

Monosynaptic group Ia excitatory postsynaptic potentials (EPSPs) exhibit marked low-frequency synaptic depression in the neonatal rodent spinal cord when studied in vitro (Lev-Tov and Pinco 1992; Li and Burke 2001a; Seebach and Mendell 1996). In the present work, we systematically applied four paradigms to study activity-dependent synaptic depression between postnatal days 2 and 12 (P2-P12), which is a period during which rat pups show striking motor maturation (Glover 2000; Jiang et al. 1999). The paradigms were as follows: 1) paired-pulse depression at different intervals; 2) steady-state depression produced by trains of different constant frequencies; 3) depression during long sequences with irregular inter-stimulus intervals (Dobrunz and Stevens 1999; Markram et al. 1998; Sen et al. 1996); and 4) recovery following relatively high-frequency conditioning trains. Each of these paradigms provided a different but complementary sampling of the way in which activation history modulates synaptic transmission. The data were examined using a quantitative synaptic model that allowed provisional identification of changes in competing depression and enhancement processes that modulate short-term synaptic transmission (Li and Burke 2001a). We found that a combination of mechanisms gleaned from various studies of other central synaptic systems can predict the observed behavior of monosynaptic group Ia terminals in the spinal cord under all of the conditions tested. A preliminary report of this work has appeared in abstract form (Li and Burke 2001b).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Experiments were performed on the isolated spinal cords of neonatal Swiss-Webster mice at postnatal days 2 to 12 (P2-P12). Animal care and use procedures were in accord with the "Principles of Laboratory Animal Care" (National Institutes of Health Publication 86-23) and were approved by the NINDS Committee on Animal Care and Use. Most of the methods used have been described in detail elsewhere (Li and Burke 2001a).

Neonatal mouse pups were anesthetized by inhalation of methoxyflurane (Metofane) in a small chamber and then quickly decapitated, eviscerated, and transferred into a dissection chamber circulated with cold (4°C) artificial cerebrospinal fluid (ACSF) saturated with 95% O₂-5% CO₂. The composition of the normal ACSF was as follows (in mM): 128 NaCl, 4 KCl, 2 CaCl₂, 1 MgSO₄, 1 NaH₂PO₄, 25 NaHCO₃, and 30 glucose. The pH was 7.2~7.3 after saturating with 95% O₂-5%CO₂. In a few experiments, CaCl₂ was omitted (Fig. 1) or reduced to 0.8 mM. After ventral laminectomy, the spinal cord with intact dorsal (DR) and ventral roots (VR) was dissected free from dorsal root ganglia. After removing the dura and arachnoid membranes, the spinal cord was hemisected longitudinally with a tungsten needle. One hemicord was placed into a Sylgard-based recording chamber (volume about 7 ml). Flow of oxygenated ACSF through the chamber was controlled at 10-14 ml/min. The temperature of bath solution was kept constant at 24°C by a servo-controlled heater (TC-324B, Warner Instruments). The bathing solution was recirculated at all times, except when drugs were added or washed out. All experiments were done with the N-methyl-D-aspartate (NMDA)-receptor blocker (±)2-amino-5-phosphonovaleric acid (AP5; 100 µM) in the bathing solution to suppress spontaneous and evoked polysynaptic activity (see Li and Burke 2001a).

View larger version (30K): [in this window] [in a new window]   Fig. 1. Potentials produced by group Ia afferent arborizations in the motor nucleus are unchanged at frequencies between 0.125 and 8 Hz. Extracellular recording in the motor nucleus of the L5 ventral horn were averaged from 60 sweeps during L5 dorsal root (DR) stimulation at 0.125, 2, and 8 Hz with [Ca 2+]o = 2 mM (black traces; bath temperature 24°C). Afferent volleys entering the cord were merged with the end of the stimulus artifacts (stimulus duration, 100 µs). Subsequent sharp terminal potential (latency about 1.1 ms) was unchanged while synaptic field potentials (latency 1.5 ms) declined as stimulus frequency increased (black lines). Synaptic potentials disappeared about 10 min after Ca2+ was removed from bathing solution without changing terminal potential (thicker gray line). Vertical dashed lines indicated onset of terminal potential and synaptic field potentials. Estimated central synaptic delay was about 0.4 ms, indicating that initial excitatory postsynaptic potentials (EPSPs) were monosynaptic.

Dorsal and ventral rootlets of the L4 and L5 segments were drawn into the polyethylene suction electrodes for either stimulation or recording. Micropipettes for intracellular recording were made from 1.2-mm filament glass (WPI) drawn to produce DC electrode resistance between 40 and 70 M (model P-87, Sutter Instruments). Micropipettes were filled with 2 M K-acetate with 100 mM QX-314 (Alamone Labs, Jerusalem, Israel) added to suppress Na⁺ action potentials (Frazier et al. 1970) that otherwise interfere with measurement of EPSPs. In some experiments, the micropipette solution also contained 2% biocytin to label the recorded cells. Motoneurons were impaled from the ventro-lateral aspect of the hemisected cord and identified by antidromic invasion after VR stimulation (before QX-314 blockade of action potentials). Sharp micropipettes and current clamp conditions were used because the blind patch electrode approach provided inadequate sampling. In addition, group Ia synapses are widely distributed over the entire dendritic tree in adult cats (Burke and Glenn 1996) and the same situation is likely in rodents (Snider et al. 1992). Furthermore, the extensive dendritic trees of neonatal mouse motoneurons (Y. Li, G. Ascoli, and R. E. Burke, unpublished observations) make it unlikely that adequate space clamping can be obtained in these cells.

Stimulation and recordings

Mono- and polysynaptic EPSPs were produced by stimulating small DR filaments with trains of 10 pulses (duration 0.5 ms) at eight equally spaced frequencies from 16 to 0.125 Hz (intervals of 0.0625, 0.125, 0.25, 0.5, 1.0, 2.0, 4.0, and 8.0 s). Trains were separated by 2-min intervals, which were sufficiently long such that the first EPSPs in each train remained constant, indicating full recovery of evoked synaptic release (see Lev-Tov and Pinco 1992). Changing the ordering of train frequencies produced no changes in response patterns. The stimulus intensity was about 20% larger than that needed to produce the maximal amplitude of monosynaptic EPSPs. The viability of preparation was assessed by monitoring VR reflexes, which remained stable for 8-12 h in most of the experiments.

Regular trains produce data about paired-pulse and steady-state modulation at different frequencies. However, this approach required about 20 min to complete a single run of all eight train intervals and one or two repeats of selected frequencies to check stability, each separated by 2 min (Li and Burke 2001a). Even with stable membrane potentials, only two repetitions could be averaged to mitigate the variance in individual EPSPs (see Fig. 3). We therefore explored the use of irregular patterns of stimuli that covered the range of inter-pulse intervals used for regular trains. We found that sequences of 50 pulses with intervals that were randomly drawn from an alpha function distribution (x = t/ e^t/ with  = 0.8) provided comparable information (see APPENDIX). The range of intervals used in this study was between 0.11 to 5.9 s, with a preponderance of relatively short intervals (see Fig. 7A). A single cycle of 50 pulses with the interval distribution chosen for this work was complete in 82 s, making it possible to repeat the cycle three or more times without interruption (see Fig. 7B). Subsequently, responses with identical time histories from each cycle were averaged, excluding the first three responses from the first cycle.

We also tested the recovery of EPSPs at intervals of 0.5, 1, 2, 4, 8, or 16 s following conditioning trains of 10 pulse trains at 4 Hz (Fig. 9A). Conditioning trains were separated by 2-min intervals, repeating the series several times to permit averaging the recovery EPSPs at different intervals.

Data acquisition and measurements

Signals from suction electrodes on DRs and VRs were amplified with Cyberamp 380 amplifiers (Axon Instruments, Foster, CA) and band-pass filtered at 10-10 kHz. Intracellular potentials were amplified with an AXOCLAMP-2B (Axon Instruments) in current clamp (bridge) mode. Intracellular signals were low-pass filtered at 10 kHz and digitized at 10 kHz (16-bit resolution) by a multi-channel A/D converter (NBIO-16, National Instruments, Austin, TX). Custom designed software (LabVIEW programming language, National Instruments) was used in some preliminary experiments and later on Axograph software (Axon Instruments) was used to acquire and save the data in a Power Macintosh computer. Intracellular potentials and responses in VRs were also continuously recorded on a digital videotape recorder (VR-100 B, Instrutech, Great Neck, NY). Data analysis was done off-line using customized and commercial software packages. Statistical analyses were done using DataDesk, V5 (Data Description Institute).

Electrotonic potentials generated by DR afferent volleys were recorded by a suction electrode on a DR filament immediately adjacent to the stimulated rootlet to ensure that the afferent volleys produced by each stimulus pulse in the train were constant (see Fig. 3C in Li and Burke 2001a). In some of the present experiments in older mice (Fig. 1), extracellular potentials were averaged in the ventral using broken-tip electrodes filled with 2 M NaCl (Fig. 1; DC resistance, 5-7 M). Reflex responses in the VR usually exhibited parallel changes with monosynaptic EPSPs but will not be considered further in this paper.

Pharmacological substances

Drugs were introduced into the ACSF bathing solution in recording chamber via a gravity-fed line (flow rate:10-14 ml/min) for 10 min (usually 20 min) before subsequent tests were made to allow equilibration. Drugs used were as follows: (±)2-amino-5-phosphonovaleric acid (APV), 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX), bicuculline methiodide, and 2-hydroxysaclofen obtained from Sigma (St. Louis, MO) and Tocris (Ballwin, MO). APV was prepared as stock solutions of 30 mM in distilled water and stored at 4°C for several weeks. Bicuculline and 2-hydroxysaclofen were freshly made in distilled water.

Computer modeling

The mathematical model used in this study was developed based on observations in young mice (P2-P4; Li and Burke 2001a) and is described in that paper and in the APPENDIX. The model was implemented in a spread sheet (Microsoft Excel, V5) and in special-purpose programs written in Pascal (CodeWarrior, Metrowerks). A parameter search program that accepted inputs of experimental or test data provided the option of using a total of eight variations of the basic model system to extract parameter sets that fitted the input data with minimum root mean square (RMS) error (see APPENDIX). The programs are available from the authors on request.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

The present results are based on intracellular recordings of monosynaptic EPSPs from 148 antidromically identified lumbar motoneurons in hemisected spinal cords from neonatal Swiss-Webster mice at postnatal ages between 2 and 12 days (P2-P12). All experiments were done in vitro with 100 µM APV in the bathing solution, which was essential to suppress spontaneous background activity and stimulus-evoked polysynaptic potentials (see Fig. 1. of Li and Burke 2001a). The addition of APV produced, at maximum, a 10% reduction in EPSP amplitudes in P2-4 mice (Li and Burke 2001a). This reduction was only 2% in nine P10-12 cells in which EPSPs were measured before and after APV addition. These relatively small reductions were smaller than those found by Pinco and Lev-Tov (1993b) in the neonatal rat. There is evidence that the NMDA component of predominates at embryonic glutamatergic synapses (Ziskind-Conhaim 1990) but declines with postnatal age in rodents (Bellingham et al. 1998; Pinco and Lev-Tov 1993b). Addition of APV also produced, on average, a general hyperpolarization of motoneurons of 5-10 mV as well as reduction of background synaptic noise. Blockade of the NMDA component also eliminated one potential source of voltage-dependent uncertainty in EPSP amplitudes. Cells accepted for analysis had stable membrane potentials < 50 mV during data collection (see METHODS). Many cells had stable membrane potentials for over 2 h. Most preparations maintained stable monosynaptic reflexes in ventral roots for 10 h and were discarded when this sign of viability declined.

Initial EPSPs evoked by dorsal root stimulation are monosynaptic

Extracellular recordings in the ventral horn showed that the shortest latency EPSPs produced by dorsal root stimulation (presumably group Ia) have a synaptic delay of about 0.4-0.7 ms in P12 mice (Fig. 1) compared with 0.6-1.0 ms in P3 animals (Fig. 3C of Li and Burke 2001a). We conclude that the initial EPSPs produced by dorsal root stimulation at all postnatal ages studied are indeed monosynaptic (see also Mears and Frank 1997). The intraspinal delay between the entering volleys and onset of terminal potentials were somewhat shorter in P12 mice (about 1.2 ms; Fig. 1) than in P3 animals (about 1.5 ms; Fig. 3C of Li and Burke 2001a), which is consistent with increasing myelination in the more mature mice.

Unreliable action potential conduction in dorsal root afferents could contribute to apparent synaptic depression. To address this question, we examined the extracellular terminal and synaptic field potentials produced in the ventral horn during dorsal root stimulation at 0.125 or 8 Hz. In a previous study in P2-4 animals, we showed that afferent volleys and resulting terminal potentials were unchanged within this frequency range (Figs. 3 and 4 of Li and Burke 2001a). Increasing stimulation frequencies reduced the extracellular synaptic potentials, as expected, but did not change the preceding terminal potentials (Fig. 1). Bathing the cord in ACSF with 0 mM [Ca²⁺]_o eliminated the extracellular synaptic potentials without altering the amplitude of the preceding terminal potential. We conclude that conduction failures within intraspinal afferent arborizations are unlikely to account for the observed synaptic depression within the age and stimulation frequency ranges studied in this work.

Age-dependent changes in monosynaptic EPSPs

The first monosynaptic group Ia EPSP (R1) in motoneurons exhibited a progressive decrease in latencies between the stimulus pulses and EPSP onset from P3 to P10-12 (Fig. 2). The second (R2) EPSPs from trains of 10 pulses at 2 Hz, plus the averages of the last three responses (R8, 9, and 10, referred to as the "Tail" responses) also showed less depression with advancing age. The slopes of the EPSP rising phase of all responses progressively increased, suggesting greater synchronization of afferent volleys with advancing postnatal age (Brenowitz and Trussell 2001). At each age, the R2 and Tail responses had the same latency and rising slope as R1 when they were expanded to equal peak amplitude (see Fig. 5B of Li and Burke 2001a). Monosynaptic EPSP amplitudes were measured at the times denoted by the vertical dashed lines, slightly before the actual peaks and prior to the inflections that signal the onset of polysynaptic components (Fig. 2). Disynaptic group I EPSPs can be observed in the adult cat during locomotion but they always begin after the monosynaptic peak has been reached (Angel et al. 1996; Degtyarenko et al. 1998). Given the fact that all of the data in this paper were obtained at room temperature, we are confident that our data represents only the monosynaptic component.

View larger version (18K): [in this window] [in a new window]   Fig. 2. Monosynaptic EPSPs evoked during 2-Hz DR stimulus trains in 4 individual motoneurons show decreasing depression with increasing postnatal age [A: P3; B: P7; C: P10; D: P12; normal artificial cerebrospinal fluid (ACSF) plus 100 µM AP5, 24°C; see METHODS). Each set of traces shows single sweep records of the 1st (R1), 2nd (R2), and the average of the last 3 EPSPs (Tail) in each train. There was a steady decline in paired-pulse (R2) and steady-state (Tail) synaptic depression with advancing age, as well as shortening of central latencies (time from stimulus pulses to EPSP onsets) and of time from onsets to peaks (abscissae). Note presence of polysynaptic components with onsets later than the times for peak measurements (vertical dashed lines).

The population data for monosynaptic EPSP latencies (mean ± SD) at each postnatal age are summarized in Fig. 3. The decrease between P4 and P7 was marginally significant (P ~ 0.04; 1-factor ANOVA with post hoc Scheffe test), while the change between P4 and P10 were clearly significant (P < 0.001). Although there was a large range in the amplitudes of the first EPSP (R1; 5-25 mV) from cell to cell at each age, two-factor ANOVA analysis showed no significant differences with increasing age (P2 = 11.9 ± 5.3 mV, n = 17; P3 = 13.5 ± 5.8 mV, n = 37; P4 = 14.4 ± 6.5 mV, n = 6; P7 = 14.5 ± 4.6 mV, n = 6; P10 = 14.5 ± 5.5 mV, n = 14; P11 = 14.1 ± 4.9 mV, n = 18; P12 = 13.5 ± 5.9 mV, n = 16). The range of EPSP amplitudes appeared to depend on variations in the number of afferent fiber activated and the relative positions of the stimulated dorsal rootlets and the recorded motoneurons.

View larger version (16K): [in this window] [in a new window]   Fig. 3. Central latencies of monosynaptic EPSPs exhibit age-related reduction. Significant reductions in mean ± SD central latencies occurred between P4 and P7, as well as between P7 and P10 (2-way ANOVA with post hoc Scheffe test). Number of cells in each age group are indicated in parentheses.

Synaptic depression curves

The use of short stimulus trains at different frequencies provides data about paired-pulse (R2) and steady-state (Tail) depression. The interval between trains (2 min) was sufficiently long so that the trials were independent (Li and Burke 2001a). The examples in Figs. 2 and 4 show that there was progressively less short-term depression of both R2 and Tail responses with increasing age. Figure 4 also illustrates that the shapes of the depression curves during 2 Hz trains from the same four cells, each normalized by the amplitude of R1, changed with age. The decrease in EPSP amplitudes to steady state during 2-Hz trains became more gradual in the older mice (see also Fig. 9A). This aspect of curve shape resembles data from calyceal synapses in the developing chick (Brenowitz and Trussell 2001) and was captured by considering the relative amplitudes of the third response, R3, in addition to R2 and Tail amplitudes (Li and Burke 2001a).

View larger version (27K): [in this window] [in a new window]   Fig. 4. Depression curves during 10 pulse trains at 2 Hz (same 4 motoneurons as shown in Fig. 2.) show age-related differences in shape as well as reduced depression. Each set of responses was normalized by the amplitude of R1 in the respective trains. In addition to less general depression at older ages, the shapes of these curves change from the abrupt depression at R2 in the P3 motoneuron to more gradual declines at older ages. These shape changes are captured by the relative amplitudes of R2 (paired-pulse), R3, and Tail (steady-state) responses (dashed boxes). The subsequent figures describe frequency-dependent changes in depression by plotting relative R2 and Tail amplitudes against stimulus frequency or inter-pulse intervals. Changes in R3 curves were intermediate between those of R2 and Tail responses and were included in modeling fits, but are not shown for clarity.

We examined short-term synaptic depression at stimulus frequencies of 0.125, 0.25, 0.5, 1, 2, 4, and 8 Hz, which covered nearly the full range of frequency dependent depression of Ia EPSPs during first two postnatal weeks (Fig. 5). The normalized amplitudes of R2, R3 and steady state (Tail) responses were grouped into four age bins (P2-3, P4-7, P10-11, and P12). The data for R2 and Tail responses from the different age groups were significantly different from one another (P < 0.001, 2-factor ANOVA with post hoc Scheffe tests), except for the P2-3 versus P4-7 groups (P > 0.1).

View larger version (27K): [in this window] [in a new window]   Fig. 5. Age-related changes in frequency-dependent short-term depression. Curves are averaged ±SD paired-pulse (R2; A) and steady-state (Tail; B) depression curves from trains at different frequencies (note logarithmic abscissae) and postnatal ages. SD of the R2 data was generally larger than those of the Tail responses because the latter were averages of 3 responses in each curve, while the former were averaged from single EPSPs (see Fig. 4). The R2 curves from P2-7 animals exhibited a clear upward trend (relative facilitation) at train frequencies 1 Hz, while this trend was much less evident in the data from the P10-12 age groups. In contrast, the Tail response curves showed only monotonic decreases with increasing stimulation frequency at all ages.

Figure 5 depicts the averaged amplitudes of normalized R2 and steady-state responses against stimulus frequency for each age bin. Curves for R3 responses (not shown) were intermediate but had variances similar to those of R2 (see Li and Burke 2001a). In addition to decreased depression with age, the most striking difference was less relative facilitation of R2 responses at frequencies above 1 Hz in the older mice. This reduction was also evident in the curves for R3 (not shown). In contrast, the curves for Tail exhibited monotonic frequency-dependent depression at all postnatal ages.

Reducing [Ca²⁺]_o revealed frequency-dependent R2 facilitation in older mice

Reducing [Ca²⁺]_o reduces paired-pulse depression at group Ia synapses at relatively short inter-pulse intervals in neonatal rats (Pinco and Lev-Tov 1993a) and other central synapses (e.g., Dittman and Regehr 1998). We observed a corresponding exaggeration of relative high-frequency (>1 Hz) facilitation of R2 responses when [Ca²⁺]_o was reduced from 2.0 to 0.8 mM in P2-3 mice (Fig. 7A in Li and Burke 2001a). We tested for this effect in 12 motoneurons from P10-11 mice and found that relative facilitation of R2 responses was similarly exaggerated in the older animals when [Ca²⁺]_o was lowered from 2.0 to 0.8 mM (Fig. 6). The difference between the R2 depression curves at the two [Ca²⁺]_o concentrations was highly significant (P < 0.001; 2-factor ANOVA). This suggests that relative paired-pulse facilitation at higher frequencies is present in older mice at normal levels of [Ca²⁺]_o, even though it is much less obvious in R2 depression curves in older mice (Fig. 5A; see DISCUSSION).

View larger version (20K): [in this window] [in a new window]   Fig. 6. Reducing external Ca2+ unmasks high-frequency paired-pulse facilitation in older mice. Paired-pulse (R2) depression curves averaged from P10-11 motoneurons revealed an overt high-frequency R2 facilitation (shaded areas) when [Ca2+]o was reduced from 2.0 mM (filled circles; black line) to 0.8 mM (open squares; shaded line). Two-factor ANOVA showed that the 2 curves were significantly different (P < 0.001; bars indicate SD).

GABAergic presynaptic inhibition is not a factor

In an earlier study (Li and Burke 2001a), we reported that addition of the GABA_A receptor blocker bicuculline (10-20 µM) or the GABA_B antagonist 2-hydroxysaclofen (100 µM) into the bathing solution produced no changes in EPSP amplitudes or depression curves in P2-4 mice. However, maturation of presynaptic inhibitory pathways could in principle complicate interpretations in older mice (Lev-Tov et al. 1988; Peng and Frank 1989). We therefore examined the same drugs and doses in four P10-12 motoneurons and observed no changes in depression curves (data not shown). We conclude that GABA-related presynaptic inhibition plays little or no role in producing the observed EPSP depression curves during the first 2 wk of postnatal development.

Partial blockade of EPSPs does not affect depression curves

It seemed possible that nonlinear postsynaptic conductances, particularly in distal dendrites (e.g., Schwindt and Crill 1995), might contribute to activity-dependent EPSP modulation. Although blockade of NMDA currents by AP5 in the present experiments removed one potential source of voltage-dependent nonlinearity, other such complications are difficult to rule out. However, in our earlier work (Li and Burke 2001a), we found that synaptic depression curves were unchanged when EPSP amplitudes were reduced to 20-25% of control by submaximal doses of the AMPA receptor blocker CNQX. This was also done in one P11 mouse (n = 3 motoneurons). The normalized depression curves at several frequencies (as in Fig. 4), obtained before and during the development of partial CNQX blockade, remained essentially unchanged despite gradual reduction of absolute EPSP amplitudes to 10-20% of control. This negative results suggests that postynaptic nonlinearities do not contribute importantly to the present observations.

Responses to pseudo-randomized stimulation patterns

There has been recent interest in using irregular patterns of stimulation that mimic natural activation patterns to study short-term synaptic modulation (Abbott et al. 1997; Dittman and Regehr 1998; Dobrunz and Stevens 1999; Markram et al. 1998). In addition to certain practical advantages (see METHODS), this approach also samples activity-dependent synaptic modulation in a manner quite different from the use of regular trains. Different outcomes with the two methods might indicate the existence of time-dependent factors that are not revealed by using either approach alone. After assessment of a variety of possible distributions for inter-stimulus intervals using the synaptic model described below, we chose a sequence of 50 pulses randomly sampled from an  function distribution ( = 0.8) with a range of intervals that was highly skewed toward short intervals (Fig. 7A; see METHODS).

View larger version (28K): [in this window] [in a new window]   Fig. 7. EPSPs produced by repeated cycles of a pseudo-random irregular stimulation sequence were stable and repeatable during long (about 5 min) periods of stimulation. A: histogram showing the distribution of inter-pulse intervals in a sequence of 50 pulses depicted in the inset. Sequence of irregular stimulation intervals was selected by random sampling from an -function distribution with  = 0.8 s, which skews the distribution toward shorter intervals. The same sequence was used for all tests in this work. B: normalized EPSPs produced by 2 runs, each consisting of 3 repetitions of the same 50-pulse sequence in a single P3 motoneuron, plotted against the actual time in each sequence. Despite considerable variations in individual responses, the general pattern was similar in each cycle of a given run, and between the 2 runs in the same cell. C: same data plotted against inter-pulse intervals (note logarithmic scale) in each run (150 EPSPs per run). The vertical data scatter around the central tendency (light gray line fitted by hand) was similar for all frequencies and response amplitudes, suggesting that much of this variance depended on extraneous sources of noise. General shape of the data scatter indicates the presence of relative facilitation at inter-pulse intervals <0.5 s. D: 50 EPSPs within each of the 3 cycles in a run (i.e., preceded by the same time history) were averaged. This plot shows these averaged responses in the 2 runs plotted against one another. The 2 runs were very similar.

It was important to assure that the relatively long irregular trains did not produce additional systematic instabilities (Dobrunz and Stevens 1999). Figure 7B shows the distributions of EPSP amplitudes (normalized by the average amplitude of 5 control stimuli at 30-s intervals; see METHODS) from two runs of three cycles each in a typical P3 motoneuron. There was no systematic rundown evident in either trial and the two shared the same data space when plotted against cumulative time (Fig. 7B) or the interval immediately preceding the EPSP (Fig. 7C). This stability permitted us to average the normalized EPSPs from the three cycles in a run together, reducing the considerable variance evident in Fig. 7, B and C. Figure 7D shows that these averaged amplitudes were comparable in successive runs.

The irregular stimulation paradigm was applied to record EPSPs from motoneurons at different ages. Two-factor ANOVA tests indicated that the two age groups were significantly different (P < 0.001). Therefore data from six P2-3 and five P12 motoneurons were separately averaged and plotted against pulse number within each cycle in Fig. 8, A and B, respectively. The data from P12 motoneurons clearly exhibited less synaptic depression than those from P2-3 cells (note different ordinate scales). However, the patterns of data scatter were not identical when scaled to the same amplitude limits, as shown when the data were plotted against the inter-pulse interval immediately preceding each EPSP (Fig. 8C). The scatter of data points from P2-3 animals largely overlapped the region between the R2 and Tail depression found in regular trains (gray area; data as in Fig. 5 but plotted here vs. inter-pulse intervals). However, this was not the case for the scatter of irregular firing data from P12 animals, which lay below the R2 -Tail region from regular trains for pulse intervals >1 s. Thus the two approaches provide complementary but not identical sampling of synaptic behavior during repetitive activation.

View larger version (31K): [in this window] [in a new window]   Fig. 8. EPSP responses to a standard sequence of irregular stimulation intervals sequence differ with postnatal age. A and B: responses to the pseudo-random stimulation sequence shown in Fig. 7A, averaged ± SD from 6 P2-3 motoneurons (A) and 5 P12 cells (B) and plotted on the same ordinate scales. C: same average data replotted as functions of inter-pulse interval (abscissa; note logarithmic scale as in Fig. 7C; P2-3, ; P12, ). The data from P12 cells clearly exhibited less overall depression, but the general scatter pattern was similar to that from P2-3 animals. Dark gray region denotes the area between the R2 (Fig. 5A) and Tail (Fig. 5B) depression curves for P2-3 mice from constant frequency trains. The lighter, hatched region denotes the same region from the P12 data in Fig. 5.

Recovery from depression

The time course of recovery from activity-induced synaptic depression can provide useful constraints as to the mechanisms that might operate during short-term depression. Rapidly-decaying factors such as receptor desensitization and presynaptic inhibition are less likely to affect passive recovery curves than during repetitive stimulation (Dittman and Regehr 1998; Stevens and Wesseling 1998; von Gersdorff et al. 1997; Wang and Kaczmarek 1998). We examined the posttrain recovery curves in P3 and P12 motoneurons by delivering single test stimuli at various intervals (0.5, 1, 2, 4, 8, and 16 s) following conditioning trains of 10 pulses at 4 Hz (Fig. 9A; see METHODS.

View larger version (24K): [in this window] [in a new window]   Fig. 9. Time courses of EPSP recovery at various times following repeated 4-Hz, 10-pulse conditioning trains in 2 postnatal age groups. A: averaged ± SD normalized EPSPs during (left one-third, note fast time base) and after (right two-thirds, time base 20 times slower) from 18 P3 motoneurons and 7 P12 cells. Note the difference in shapes of the 2 conditioning trains (cf. Fig. 4) and the fact that the last 3 responses at the end of the train (i.e., the "Tail") were slightly larger than the 4th and 5th responses in the P3 data (see also Fig. 11A). Averaged responses during the recovery period (0.75-16 s) had larger SD because only 2 or 3 responses were obtained at each recovery interval, while 20 conditioning trains were averaged in each motoneuron. B: semilogarithmic plot of 1 minus the normalized amplitude of recovery EPSPs (ordinate) vs. recovery time (abscissa) from the averages for P3 and P12 data shown in A. Fits of a single exponential to these points gave nominal decay time constants of about 6.2 s for P3 and 4.9 s for P12. The dashed line superimposed on the P3 data denotes fit to the data using the synaptic model discussed in Fig. 10.

Figure 9A depicts averaged EPSPs during conditioning trains and recovery periods (note different time scales) from 18 P3 motoneurons and 7 P12 cells. Figure 9B shows a semilogarithmic plot of one minus the recovery data, which allows an estimate of the overall rate of EPSP recovery toward unity. The data points at zero time are the 10th pulses in the respective conditioning trains. From the slope of exponential curve fits to these data, we estimated that the overall recovery time constant was 6.2 s for the P3 data and 4.9 s for P12. These estimates were essentially the same using 20 pulse conditioning trains at either 4 or 8 Hz.

A presynaptic release model

Our observations suggest that presynaptic release mechanisms change during the first 2 wk of postnatal life but give no clear indication of what factors might produce such changes. In a previous paper (Li and Burke 2001a), we described a quantitative synaptic model based on data from P2-4 mice that includes a number of possible mechanisms that have been suggested to explain various features of short-term synaptic depression. This lumped synapse approximation ignores possible variations among the group Ia boutons that end on individual motoneurons (see Murthy et al. 1997). The model accurately predicted the synaptic depression data found with three external [Ca ²⁺]_o concentrations (0.8, 2.0, and 4.0 mM), as well as at two temperatures (24°C and 32°C), in P2-4 mice. The complete version (see APPENDIX) includes two independent presynaptic compartments, called N and S, that are depleted by release of some fraction, f, of the available N · S combinations. Each compartment is assumed to be renewed with independent exponential time constants, _N and _S. We assume that N represents the fraction of readily-releasable transmitter present at any moment, while S represents the available fraction of competent release sites (see APPENDIX).

The experimental data required that this simple depletion scheme be augmented by two activity- and time-dependent processes that enhance transmitter release by different mechanisms. The relative facilitation of R2 at higher frequencies (see Figs. 5 and 6) was accounted for by a factor P that enhances release with an increment P at each activation and decays exponentially with time constant _P. This modulation of release fraction f is consistent with the brief facilitation attributed to residual [Ca ²⁺]_i that follows each presynaptic action potential (Dittman and Regehr 1998; see reviews by Zucker 1989, 1999; Zucker and Regehr 2002).

To reconcile the observed paired-pulse with steady-state responses under all testing conditions, it was necessary to include a second enhancement factor, M, with increment M and decay time constant _M. Positive values of M result in increasing the rate of compartment N renewal by reducing _N, which has a disproportionate effect on the steady-state (Tail) depression. This mechanism was designed to approximate activity-dependent enhancement of renewal of readily-releasable transmitter that has been proposed for a number of central synapses (Stevens and Wesseling 1998; Wang and Kaczmarek 1998; Wu and Borst 1999; see also Smith et al. 1998). In young mice, the increments P and M changed in parallel with changes in [Ca ²⁺]_o, leading to the conclusion that both enhancement processes are Ca²⁺-dependent (Li and Burke 2001a).

Developmental changes in model parameters

Although the model summarized above provided a reasonable framework to interpret data from young (P2-4) mouse pups, it seemed possible that less elaborate versions could fit the data from older animals. Accordingly, we tested seven alternative combinations (1 or 2 presynaptic compartments, with or without the P and/or M processes) with the regular train data from each age range (Fig. 5), using a parameter search program that minimized the RMS error between observed and simulated responses (see APPENDIX). In all cases, these simpler variations produced larger errors than the full seven-parameter model (see Table 1 in Li and Burke 2001a) for the data from regular trains and from the irregular pulse sequence. The complete model accurately predicted population average data for both paired-pulse and steady-state depression data (Fig. 10), as well as the EPSP amplitudes produced by irregular stimulation sequences (Fig. 11) for all ages tested.

View larger version (17K): [in this window] [in a new window]   Fig. 10. Synaptic model fits (continuous lines) to the observed data (symbols and data as in Fig. 5) for R2 () and Tail () responses from mice in 4 age groups. The best-fit parameter sets (Table 1) produced the minimum root mean square (RMS) difference between the observed and calculated R2, R3, and Tail responses in 20 runs of the parameter estimation program (see APPENDIX)

View larger version (24K): [in this window] [in a new window]   Fig. 11. Synaptic model fits () to the averaged, normalized EPSPs () obtained during irregular stimulation sequences in 6 P3 (A) and 5 P12 motoneurons (B), plotted against time during the basic cycle of 50 intervals (see Fig. 7, A and B). As in Fig. 10, the parameters sets that gave these fits (Table 2) were those that produced the minimum RMS errors in 20 program repetitions.

The parameter sets that produced the best fits to the population data for each age range are given in Tables 1 and 2. With one exception, there was quite good agreement between the parameter sets extracted using regular train and irregular cyclic stimulation paradigms at the youngest and oldest ages, despite the differences in the ways that these paradigms sample the stimulus-response space (see Fig. 8C). The renewal time constants, _N and _S, as well as _P, showed surprisingly little changes with increasing postnatal age. On the other hand, model fits to both regular train and irregular data indicate that the baseline release fraction, f, and the increment P associated with the P process, both decreased systematically between P2 and P12. The M process decay time constant, _M, also declined in the older animals, although the change in the irregular pulse fits was smaller than that extracted from regular trains. The major discrepancy between the two experimental paradigms was in the estimates of M, which showed a systematic threefold increase with postnatal age in the regular train data but a decrease of about 30% between the P2-3 and P12 irregular data. Possible interpretations of these estimates are discussed in the DISCUSSION.

It proved to be difficult to use postconditioning recovery curves (Fig. 9) to estimate the full set of model parameters because this paradigm provided even less constraint for parameter estimation. Nevertheless, a modified parameter extraction program was applied to the data from 18 P3 motoneurons illustrated in Fig. 9. The model fit to these data (dashed line in Fig. 9B) showed departures from the simple exponential fit (solid line in Fig. 9B) that more closely matched the observed data points. This departure is due mainly to the relatively slow changes in _N produced by the M process. This factor also produced the small increase in normalized EPSPs evident during the second half of the conditioning train (Fig. 9A). No such increases were found during the P12 conditioning train (Fig. 9A), largely because of the faster decay of the M process in older mice (Table 1).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Short-term modulation of synaptic transmission at central synapses depends on simultaneous interaction of multiple factors, some of which decrease transmitter release while others facilitate it (Zucker 1989, 1999; Zucker and Regehr 2002). The present work used three experimental paradigms to document appreciable changes in short-term synaptic depression at in situ monosynaptic group Ia excitatory synapses on lumbar spinal motoneurons in mice through the first 2 wk of postnatal life. To our knowledge, this is the first study of this issue to use a quantitative model to dissect changes in the interacting processes that govern short-term synaptic modulation during postnatal development.

It was surprising that this important synaptic system exhibited so much residual low-frequency depression at older ages, despite the marked advances in motor abilities that occur between P2 and P12. Synaptic maturation during this period did not involve much speeding of the renewal of readily-releasable transmitter or fully capable release sites. The observed decrease in low-frequency depression can be attributed in part to a reduction in fractional transmitter release and in part to changes in a presynaptic process that speeds transmitter renewal.

Synaptic modeling

It is difficult to discuss the systematic changes described in this work without some kind of conceptual framework. Short-term synaptic modulation reflects a complex balance of simultaneous depression and enhancement processes that frustrate intuitive explanations (Zucker 1989, 1999; Zucker and Regehr 2002). Quantitative modeling can, in principle, identify such processes and dissect their interactions in producing experimental observations (e.g., Brezina et al. 2000; Dittman et al. 2000; Sen et al. 1996; Tsodyks and Markram 1997; Varela et al. 1997; Weis et al. 1999). The present model attempts to do this by combining mechanisms that have been suggested in different papers on a variety of central synapses into a single quantitative system.

Because of the complex anatomy of the group Ia synaptic system (Burke and Glenn 1996), the nature of the independent N and S compartments in the present model are unknown. However, it seems reasonable to assume that N represents the readily releasable transmitter pool postulated by many other models, while S is assumed to represent a pool of release-ready release sites. The assumption that N and S can vary independently is consistent with recent results (Stevens and Wesseling 1999) and implies that elements of S are not necessarily associated with elements of the readily-releasable transmitter pool (i.e., docked vesicles), but rather that this association is probabilistic (i.e., that docked vesicles can undock; Murthy and Stevens 1999). In a multiple-bouton system like Ia afferents, it is possible that the elements of S represent individual boutons. The S component can be viewed to represent an activity-dependent mechanism that reduces the baseline release fraction f (see Bellingham and Walmsley 1999; Wu and Borst 1999). Thus both the S and P factors modulate the baseline release fraction f, giving an effective instantaneous value f* = S · f(1+P) (see APPENDIX). This interpretation implies that changes in baseline f depend partly on the pool size of release-ready release sites (see Stevens and Wesseling 1999). With this view, the present model (APPENDIX) can be reduced to the classical formulation R = N · p (Liley and North 1953; see Zucker and Regehr 2002), but with both terms use-dependent.

Our experimental observations required the inclusion of two process that enhance transmitter output. The P process was designed to mimic the transient increase in evoked transmitter release that follows synaptic activation (paired-pulse facilitation), which is assumed to result from rapidly-decaying enhancement of release probability due to residual Ca²⁺ action (Atluri and Regehr 1996; Dittman and Regehr 1998; Zucker and Regehr 2002). The decay time constant, _P, found in the present work, is of the same order of magnitude found by these authors in other central synapses. Adding the P process to our model reproduced the relative facilitation of R2 and R3 at frequencies >1 Hz in the younger mice (Fig. 5; Li and Burke 2001a). However, there was a persistent discrepancy in fitting the steady-state depression curves that was not be mitigated by using a double-exponential time course for P process decay (see Dittman and Regehr 1998). Given recent evidence that Ca²⁺ entry indirectly activates a still-unknown mechanism that increases the rate of renewal of the RRP (Smith et al. 1998; Stevens and Wesseling 1998; Wang and Kaczmarek 1998), we added the M process to the model to give an exponentially-decaying modification of _N. This enabled the model to reproduce the entire time course of depression (Fig. 10), as well as departures of recovery curves from a simple exponential time course (Fig. 9B). In this regard, the present model has some similarity to that developed by Regehr's group (Dittman and Regehr 1998; Dittman et al. 2000). The parameters of the P and M processes changed in appropriate directions when [Ca²⁺]_o was altered in P2-4 mice (Li and Burke 2001a), providing evidence that both are related to Ca²⁺ entry (cf. Wu and Borst 1999).

Age-related changes in release fraction

The magnitude of short-term synaptic depression is positively correlated with estimates of the baseline magnitude of fraction release of readily releasable transmitter at many central synapses (Dittman et al., 2000; Zucker 1999). The estimated baseline release fraction, f, in P12 mice (0.2) was less than one-half that at P2-3, but this level is still high compared with other central synaptic systems that exhibit short-term enhancement (Dittman et al. 2000). The decline in f in neonatal group Ia synapses is consistent with developmental changes found at the more accessible calyx synapse preparation in the rat (Iwasaki and Takahashi 2001) and chick (Brenowitz and Trussell 2001). It remains to be determined whether this decline in release fraction represents maturational changes in the release mechanism itself, a general increase in the pool of readily releasable transmitter (Mozhayeva et al. 2002), or a combination of both (Zucker and Regehr 2002). The average absolute amplitudes of Ia EPSPs did not change with age in the present sample but this fact alone cannot be interpreted in terms of possible changes in quantal content or quantal size during development, as found in some other synaptic systems (Bellingham et al. 1998; Brenowitz and Trussell 2001; Chen and Regehr 2000; Iwasaki and Takahashi 2001). Nevertheless, the present results would be compatible with an age-related increase in the size of the readily releasable transmitter pool.

Although the group Ia afferent system is anatomically complex (Burke and Glenn 1996; Snider et al. 1992), the evidence from extracellular recordings in the ventral horn suggest that failure of action potentials invasion into presynaptic afferent arbors cannot account for the synaptic depression observed within the frequency range tested in these experiments (Fig. 1; Li and Burke 2001a; but cf. Streit et al. 1992).

Age-related change in the P and M processes

Assuming that the P process represents a direct enhancement of residual Ca²⁺ on transmitter release (Dittman and Regehr 1998), it seemed surprising that the relative enhancement of R2 responses became less evident in older mice with normal [Ca²⁺]_o (Fig. 5). Our earlier study in P2-4 animals demonstrated that the estimated decay time constant _P decreased from 0.14 to 0.08 s when the bath temperature was raised from 24°C to 32°C, with little change in P. This resulted in a shift of the break point in the R2 depression curve from about 1 to 2 Hz and a 40% reduction in the estimated _P (Fig. 10 and Table 3 in Li and Burke 2001a). It seemed possible that the apparent decrease in relative R2 facilitation at higher frequencies in older mice (P10; Fig. 5) could signal a developmental increase in the rate of residual Ca²⁺ decay, making it undetectable within the range of frequencies tested. However, the modeling results suggested that this was not the case (see Tables 1 and 2). Moreover, reducing [Ca²⁺]_o to 0.8 mM revealed an obvious relative R2 facilitation in the data from older mice with a similar frequency break-point (about 1 Hz; Fig. 6). The present modeling results (Tables 1 and 2) suggested that _P was similar at 24°C in mice of all ages tested. We therefore attribute the apparent diminution of high-frequency R2 facilitation to a reduction in the increment P (Tables 1 and 2), which presumably results from changes in the sensitivity of release mechanisms to residual Ca²⁺ after each presynaptic action potential (see Chuhma et al. 2001).

On the other hand, model fits to the regular train data (Table 1) suggested that the increment M increased with postnatal age while _M showed a systematic decrease. The latter was also evident in the fits to irregular sequences (Table 2) but M from these fits showed a decrease with age rather than an increase. For several reasons, we believe that the regular train estimates are more reliable. First, the M parameter is the least reliable of the seven in the face of noise in the input data (see APPENDIX and Fig. 12). In addition, the irregular train data does not contain information about steady-state (Tail) depression that is the most important constraint on the estimation of M process parameters. Finally, the regular train estimates were based on much larger samples than the irregular sequence estimates (see Table 1 and 2). Accordingly, we favor the interpretation that M steadily increases while _M decreases between P2 and P12.

View larger version (48K): [in this window] [in a new window]   Fig. 12. Histogram of the relative sensitivity of model parameters to Gaussian noise added to artificial data sets. Bars denote the slopes of linear relations between percent added (independent variable) noise and the coefficient of variation of the best fits to the noisy data (dependent variable). Data are arranged from left to right in order of increasing noise sensitivity. Parameters of the M process were the most error-prone in the face of noisy input data, while the release fraction f was by far the least sensitive parameter.

The nature of the activity-dependent process (or processes) that speeds renewal of readily-releasable transmitter is unknown. However, the present results suggest that the efficacy of the process (i.e., M) increases with postnatal age, even though the effect decays more rapidly. The ability to recognize it effects using a relatively simple paradigm (i.e., regular trains), as well as to gain some estimation of its magnitude and time course (albeit imperfect) may make it possible to design experiments to clarify the underlying mechanism(s).

Concluding comment

To our knowledge, there is no data on short-term modulation of group Ia transmission in the adult mouse. However, it is clear that Ia synapses in the adult cat exhibit relatively little short-term modulation at frequencies 100 Hz (Eccles 1964). We assume that there is continued maturation of this functionally important synaptic system in the mouse beyond 2 wk of age, as implied by ultrastructural changes in ventral horn synaptic boutons between P11 and adult rats (Vaughn and Grieshaber 1972). It would be technically formidable to study this system in an in vivo mouse preparation. Investigation of later stages in the postnatal development of the mouse spinal cord synapses likely will have to await further development of the in vitro approach that permits survival of spinal cord explants from older animals.