1. Tetanizing protocols were used to test whether glycinergic inhibiti
on undergoes long-term plasticity in vivo. For this purpose we studied
the inhibition evoked disynaptically in the teleost Mauthner (M) cell
by stimulation of the posterior branch of the contralateral VIIIth ne
rve. The advantage of this experimental design is that the inhibition,
which is mediated by identified second-order commissural interneurons
, is not contaminated by parallel excitation. 2. The VIIIth-nerve-evok
ed inhibitory postsynaptic potentials (IPSPs), which are generated at
the level of the soma, are depolarizing in Cl--loaded M cells. After V
IIIth nerve tetanization, these IPSPs exhibited potentiation lasting >
30 min in 23 of 31 cells. The maximum enhancement measured 5-10 min af
ter the onset of the tetanization averaged 100 +/- 19% (mean +/- SE).
In contrast, the non-''tetanized'' collateral IPSP induced by antidrom
ic stimulation of the M axon did not increase significantly, suggestin
g synaptic specificity of the potentiation. 3. Single-electrode voltag
e-clamp studies of Cl--loaded M cells indicated that this plasticity i
s due to an increased synaptic conductance that occurs without obvious
modifications of the kinetics or voltage dependence of the inhibitory
postsynaptic currents. 4. The synaptic conductance and its changes du
ring potentiation were quantified by measuring the inhibitory shunt of
the antidromic spike while recording with potassium-acetate-filled el
ectrodes. For this purpose the ratio, r', of the inhibitory to resting
membrane conductances, was calculated using the expression (V/V') - 1
, where V and V' are the amplitudes of the control and the test antidr
omic spikes, respectively. This ratio was called fractional conductanc
e. Measured at the peak of the expected VIIIth-nerve-evoked IPSP, r' i
ncreased by 114 +/- 18% (n = 46). Again the collateral inhibitory cond
uctance was not modified. 5. Because there are two synapses in the inh
ibitory pathway, it became important to determine whether modification
s of the second-order inhibitory junctions contribute to the overall p
otentiation. Several experimental procedures were used for this purpos
e. 6. The input-output relationship at the inhibitory synapses was det
ermined by comparing the size of the presynaptic volley and r'. The fo
rmer was recorded intra- or extracellularly as a monophasic positive p
otential, the so-called extrinsic hyperpolarizing potential, which inc
reases in parallel with the strength of VIIIth nerve stimulation. In 1
2 experiments where the presynaptic volley was unaffected by the tetan
ization, suggesting lack of involvement of the first relay, r' neverth
eless increased in amplitude by 79 +/- 14%. 7. In another eight experi
ments, the volley was increased by the conditioning protocol, indicati
ng potentiation at the first-order excitatory relay. When the input-ou
tput function of the inhibitory synapses was examined over a wide rang
e of test stimulus intensities, its slope was 50 +/- 20% steeper at 30
min posttetanization. This enhancement could occur when only a few co
mmissural interneurons were tetanized by low-intensity trains. 8. Dire
ct evidence for plasticity of the inhibitory synapses was obtained wit
h paired recordings from identified commissural interneurons and Cl--l
oaded M cells. Stability of Cl- loading was monitored using the amplit
ude of the collateral IPSP as a control. Eight of 23 pairs exhibited p
otentiation, with an average IPSP increase of 70 +/- 20%, the longest
recording session lasting 24 min after tetanization. 9. Taken together
, these results demonstrate that inhibitory synapses undergo a form of
plasticty comparable with that termed long-term potentiation (LTP) at
excitatory junctions. 10. Evidence suggests that a minimum level of M
cell cytoplasmic Ca2+ is necessary for the induction of the inhibitor
y LTP: depolarizations applied to the M cell in conjunction with repea
ted test stimulations of the contralateral VIIIth nerve resulted in a
persistent potentiation. This observation is consistent with an earlie
r finding that postsynaptic injection of a Ca2+ chelator blocks induct
ion of this potentiation. 11. Although quantal analysis was not system
atically performed, results obtained using the coefficient of variatio
n and constraint deconvolution methods were in close agreement and sug
gested that the inhibitory LTP is mainly expressed presynaptically. 12
. The functional meaning of this form of plasticity is discussed in th
e context of its role in the control and adaptability of reflex activi
ties rather than within the more conventional scheme of learning and m
emory. Consideration of network design, properties of the M cell's aff
erent inputs, and factors involved in acoustic localization lead to th
e suggestions that inhibitory LTP contributes to the direction and ori
entation of the sound-evoked startle response triggered by the M cell.