changes in nonlinear signal processing in rat hippocampus associated with loss of paired-pulse...

Changes in Nonlinear Signal Processing in Rat HippocampusAssociated with Loss of Paired-Pulse Inhibition

or Epileptogenesis

David Naylor

VA Greater Los Angeles Healthcare System, and Department of Neurology, UCLA Medical Center,Los Angeles, California, U.S.A.

Summary: Purpose: To study acute and chronic physiologicaleffects of perforant path stimulation using paired-pulse andnonlinear signal analysis techniques (Wiener kernel analysis).

Methods: Two to 3-month-old Wistar rats were implantedwith stimulating electrodes in the perforant path and recordingelectrodes in the granule cell layer. Loss of paired-pulse in-hibition was produced with 2 Hz continuous and 20 Hz(10 s/min) intermittent stimulation for periods of 1–15 min (0.1ms, 20 v pulses). Some animals received 30–60 min of stimu-lation, a model for status epilepticus/epileptogenesis. Re-sponses to paired-pulse or white noise inputs were recordedsequentially.

Results: Loss of inhibition with brief 1–3 min of stimulation,measured by increase paired-pulse ratio (P2/P1 ISI 40 ms) from0.25 (±0.27) pre- to 1.02 (±0.18) poststimulation (p < 0.001),lasted 43 (±15) min. For 30–60 min of stimulation, the paired-pulse ratios were 0.088 (±0.11), 1.59 (±0.036), 0.06 (±0.11),

0.82 (±0.22) for pre-, immediate post-, 1 week post-, and 1month poststimulation, respectively (p < 0.025). Compared toprestimulation values, Wiener kernel amplitudes for immedi-ate, 1 week, and 1 month poststimulation were 24% (±13%),72% (±17%), and 31% (±21%), respectively (p < 0.05). Wienerkernels 1 month poststimulation showed response prolongationwith increased opportunity for excitatory interactions of inputs(particularly those separated by 4 ms).

Conclusions: Brief perforant path stimulation causes sus-tained loss of inhibition in the dentate, possibly an early eventin the transition to status epilepticus. Stimulation for 30–60 mincauses chronic changes in paired-pulse and white noise (Wie-ner kernel) responses. Transient recovery occurs by 1 week, butlater new features appear (including delayed/late inhibition andpotential excitatory cross-talk) that might favor epileptic sei-zures. Key Words: Epilepsy—Hippocampus—Paired-pulseinhibition—Wiener analysis—Nonlinear.

Several animal models exist for the study of epilepsy[pilocarpine (1), kainate (2)]. Each model is character-ized by an initiating period of status epilepticus (SE) thatcan last >24 h and is followed by a “silent” period ofseveral weeks (3). Shortly thereafter, spontaneous sei-zures may be observed. Much investigation has focusedon the “silent” period, which is thought to represent atransition to the chronic epileptic state. In particular, twomain changes occur in seizure-damaged hippocampi: (a)abnormal recurrent excitatory circuits are generated by asprouting of excitatory mossy fibers (the axons of den-tate granule cells) onto granule cell dendrites of the innermolecular layer; and (b) disinhibition results from a se-lective loss of subtypes of interneurons [�-aminobutyric

acid (GABA)ergic cells with somatostatin, neuropeptideY (NPY), or parvalbumin], whereas other interneuronsare preserved (e.g., basket cells) (4). There is great un-certainty regarding the exact role of these two phenom-ena with epilepsy, and whether they are causative factorsor represent noncontributory or even compensatorymechanisms.

This study attempts to help address some of theseissues by characterizing progressive changes in physi-ologic indices from field recordings in vivo using theperforant-path stimulation model [derived from Sloviter,as described by Mazarati (5)]. Standard physiologic ap-proaches such as measurement of paired-pulse (PP) in-hibition, as well as more sophisticated nonlinear methodsof Wiener analysis, were applied to dentate gyrus pro-cessing of perforant-path inputs. The techniques wereapplied at various intervals after the initial convulsantstimulus to observe a progression of changes in the re-sponse properties of the hippocampus with initiation ofSE and with epileptogenesis during the “silent” period.

Address correspondence and reprint requests to Dr. D. Naylor at VAGreater Los Angeles Healthcare System and UCLA School of Medi-cine, Department of Neurology (127), 11301 Wilshire Blvd., Bldg 500,Rm 3256, Los Angeles, CA 90073, U.S.A. E-mail: [email protected]

Epilepsia, 43(Suppl. 5):188–193, 2002Blackwell Publishing, Inc.© International League Against Epilepsy

188

METHODS

The 2- to 3-month-old Wistar rats were implanted withstimulating electrodes in the perforant path with record-ing electrodes in the granule cell layer (see Mazarati fordetails on stereotaxic positioning). Circuitry changeswere assessed with standard stimuli such as pulses,paired pulses, or trains. In addition, Wiener analysis(which uses white noise or random-pulse stimuli) is aneffective technique for studying nonlinear systems andreveals attributes of circuitry processing not availablefrom the standard stimuli. Comparable to the expansionof a function into a power series, Wiener analysis ex-pands a functional (an operator on a function, for ex-ample, an integral) into a series sum of functionals. TheWiener kernels, hn, are the analogues of the power seriescoefficients (6). The first-order kernel, h1, is familiar asthe impulse response of linear signal analysis. Examplesof first- and second-order Wiener kernels are seen in Fig.1. In brief, the first-order kernel represents the linear(independent) effect from an input pulse at time t1 beforea response that occurs at t � 0. The second-order kernelrepresents the nonlinear (synergistic/antagonistic) inter-action from two pulses at t1 and t2 before the response att � 0. The full second-order kernel consists of a contourmap (Fig. 1E), and for presentation purposes, the diago-nal of the contour plot (along t1 � t2) is displayed (Fig.1iii). Filtered white noise from DC to 200-Hz andrandom-pulse stimuli were used with a typical stimulusepoch of 10–20 s. The standard deviation (�) of the whitenoise was adjusted from 50 up to 800 mV without evi-dence of injury to tissue or gliosis on postmortem in-spection.

To induce SE, some animals received from 30 to 60min of stimulation with 2-Hz continuous and 20-Hz (for10 s/min) stimulation with 20-V pulses (0.1 ms dura-tion). In addition, to explore the transition to SE, otheranimals were stimulated for periods ranging from 1 to 15min. Paired-pulse inhibition or Wiener kernels weremeasured every 5 min in these animals until recovery tobaseline physiologic response was noted. Those otheranimals receiving 30–60 min of stimulation (intended toinduce SE) were monitored for changes in paired-pulseresponses and Wiener kernels initially daily, and then atweekly intervals. Paired-pulse responses were assessedwith interstimulus intervals (ISIs) of 40 ms using 0.1-msduration spikes of 20–40 V.

RESULTS

Recovery of paired-pulse inhibitionPaired-pulse inhibition (ISI, 40 ms) is lost after 1 min

of continuous 2-Hz stimulation ending with 10 s of 20Hz. For 1 min of stimulation, the loss of paired-pulseinhibition could persist for >25 min. For longer stimulusperiods (3–5 min), the loss of paired-pulse inhibition

could persist for >40 min. In separate experiments, thefirst- and second-order kernels (as well as paired-pulseresponses) incrementally were assessed after 1 min ofstimulation. The kernels often did not return to baselinelevels until >1 h (even though the paired-pulse inhibitionrecovered by 25 min). Initially, a significant decline ofamplitude was noted in both first- and second-order ker-nels, and the attenuation is somewhat greater for negativeas compared with positive components, although bothare affected substantially. Some slowing of the kineticsof the negative component also was noted after 1 min ofstimulation.

EpileptogenesisPaired-pulse responses and first- and second-order

kernels were recorded before and at various times afterinduction of epileptogenesis using the perforant-pathstimulation (PPS) model for 30–60 min. Of the eightanimals undergoing PPS and providing initial results,five survived beyond 1 month, and of these, three dem-onstrated both strong paired-pulse and white-noise re-sponses. The severity of the seizure activity after PPStended to be less severe in the long-term (>1 month)surviving animals.

The pop-spike amplitude ratio for the second to firstpulses (ISI, 40 ms) in the long-term surviving animalswith good paired-pulse responses showed a mean ratio of0.088 (±0.11) before PPS, changing to 1.59 (±0.36) inthe immediate post-PPS period. Recovery of paired-pulse inhibition to a ratio of 0.06 (±0.11) at 7–12 dayswas subsequently followed at 4–6 weeks after PPS by aloss of inhibition with a ratio of 0.82 (±0.22) (p < 0.025;paired t test). This pattern of paired-pulse inhibition loss,recovery, then loss again is reflected in Fig. 1i. The pop-spike latency had a mean of 5.93 ms (±0.71) with keta-mine/xylazine anesthesia, 3.52 ms (±0.39) in awakeanimals before PPS, and 4.62 ms (±0.94) in the imme-diate post-PPS period. No significant change in pop-spike latencies from pre-PPS values was noted insubsequent periods after the immediate post-PPS periodin the surviving five animals.

Under ketamine/xylazine anesthesia, the peak re-sponse times (PRTs) of Wiener kernels were 7.67 ms(±1.30) and 9.19 ms (±1.99) for first- and diagonal ofsecond-order kernels, respectively. For awake pre-PPSconditions, the first- and second-order PRTs were 4.79ms (±0.38) and 6.59 ms (±0.90), respectively. Thesecond-order diagonal peak amplitude was 11.4 times(±8.2) higher under ketamine/xylazine anesthesia than inawake pre-PPS animals (p < 0.005), suggesting an in-crease of response sensitivity to perforant path electricalstimulation in the anesthetized animals. The immediatepost-PPS kernel peak amplitudes (Fig. 1B) were noted todecrease to 24% (±13%) and 28% (±17%) of pre-PPSfirst- and second-order diagonal values, respectively (p <

NONLINEAR PROCESSING IN EPILEPTOGENESIS 189

Epilepsia, Vol. 43, Suppl. 5, 2002

D. NAYLOR190


0.005). Recovery of amplitude over the subsequent days/weeks (Fig. 1C) was noted, with a return of amplitude to72% (±17%) and 81% (±24%), respectively, of pre-PPSfirst- and second-order diagonal peak amplitudes (p <0.001). However, at 4–6 weeks after PPS (Fig. 1D), theamplitudes again decreased to 31% (±29%) for first- and27% (±21%) for the second-order diagonal comparedwith pre-PPS values (p < 0.05).

The results for one animal shown in Fig. 1 demon-strate these trends and other features. For example, Wie-ner kernels before PPS generally show an increase in theamplitude of the kernels at higher variances of white-noise stimulation, and this increase appears to be pro-portionate for both first- and second-order kernels (Fig.1Aii and iii). This behavior was consistent with an in-crease of gain for higher stimulation levels (nonlinearamplification). At the highest stimulation levels (close tothreshold levels for production of a seizure), mild attenu-ation in amplitude might be observed (possibly relatingto a saturation or fatigue factor). The first-order kernelwas biphasic (high-pass), with an early positive compo-nent followed by an equally strong negativity (Fig. 1Aii).Strong paired-pulse inhibition was observed before PPS(Fig. 1Ai).

Several features emerge from the poststimulation ker-nels. First, in the immediate post-PPS period, the signifi-cant attenuation in the amplitude of both first- andsecond-order kernels was seen and associated with a lossof paired-pulse inhibition (Fig. 1B). Recovery of kernelamplitude and paired-pulse inhibition over the next weekwas noted as the kernels started to resemble the pre-PPSkernels again (Fig. 1C). However, at 1 month, the kernelamplitude was quite diminished, and loss of paired-pulseinhibition was seen. The qualitative features of the first-order kernels resembled those of the immediate post-PPSkernels (compare Fig. 1Bii and 1Dii). Those qualitativefeatures include the appearance of an earlier negativecomponent, apparently secondary to a decrease in thepositive component. In addition, the first-order kernel ismuch less biphasic and now appears less damped, withoscillatory features (Fig. 1ii, arrows) that also may beappreciated in the negative component at the end ofpaired-pulse responses (Fig. 1Di). The second-order ker-nel diagonal also shows a shift with the appearance of amore prominent late negativity along the diagonal (Fig.

1Diii). The emergence of other late features such as off-diagonal excitation is seen as well (Fig. 1Eii).

Seizure evolutionWhite-noise stimulation can induce a seizure if pre-

sented at significant amplitude of variance for adequateduration. For example, white noise of high variance wasable to initiate self-sustaining epileptiform activity thatpersisted beyond the 20 s of white-noise stimulation andcould evolve into a seizure. Changes during the onset ofa seizure were assessed by calculation of kernels duringsequential one-third portions of the 20-s stimulation pe-riod. Both first- and second-order kernels showed sig-nificant attenuation of amplitude as the seizure evolved.A delay of PRT for the negative component was noted aswell. Decreasing the variance of the white noise onlyslightly below that which was observed to induce a sei-zure was ineffective at provoking a seizure, even whenwhite-noise stimulus duration �90 s was used. However,minimally increasing the variance to the previously ef-fective level restored the ability to cause a seizure, sug-gesting an amplitude threshold phenomenon.

DISCUSSION

Wiener kernel analysis, using continuous white-noisesignals, revealed many attributes of hippocampal signalprocessing not readily apparent from standard techniquessuch as paired-pulse stimuli. Many consistencies oc-curred between Wiener analysis and paired-pulse resultssuch as the time course of loss/recovery of paired-pulseinhibition and kernel peak amplitudes, as well as thenegative component noted at 1 month after PPS insecond-order kernel diagonals that also was observed inpaired-pulse responses (Fig. 1D). The slower PRTs forsecond-order kernel diagonal as compared with first-order kernel responses may relate to the possibility ofcancellation of positive and negative inputs in odd-orderkernels, thereby allowing sharper-appearing waveforms.It also is possible that the more nonlinear inputs areslower [e.g., N-methyl-D-aspartate (NMDA)-receptor in-puts].

Wiener analysis techniques with random pulses havebeen used to probe perforant path–dentate projections(7,8), but the frequency of pulse delivery (and hencetemporal resolution) was limited by the requirement of

<FIG. 1. Paired-pulse response (row i), first-order kernel (row ii), and second-order kernel diagonal (row iii) for the perforant-pathstimulation (PPS) model of epilepsy with recordings made before PPS (column A), <1 day after PPS (column B), 1 week after PPS(column C), and 1 month after PPS (column D). The full second-order kernel contour plots (E) are shown for before PPS (Ei) and 1month after PPS (Eii) with time (T1, T2) along the axes and amplitude out of the plane (solid contour lines are positive, dashed arenegative). The line marked “diagonal” shows the orientation of the cut of a second-order kernel that is used to generate plots for row iii.Each trace for each graph on row ii and row iii represents a different standard deviation (�) of white-noise stimulation, with the lowest tohighest variance as var1 to var4, respectively. For column A, the variances in volts (V) are 0.2 V (trace 1), 0.3 V (trace 2), 0.4 V (trace3); column B, 0.2 V, 0.4 V, 0.6 V, 0.8 V; column C, 0.2 V, 0.3 V, 0.4 V, 0.5 V; and column D, 0.2 V, 0.4 V, 0.6 V, 0.8 V. The variancesof stimuli were adjusted between columns to optimize responses. Units for kernel amplitudes are in mV/(V·s) and mV/(V2·s2) for first- andsecond-order kernels, respectively (with uniform scaling across the series of graphs).



large-amplitude pulse stimuli. As a result, long-durationstimulus sessions were necessary, and interactions ofpulses generally were noted over hundreds of millisec-onds. In the present study, relatively mild continuouswhite-noise stimuli of brief duration were used, and thetime axis of the kernels was on the order of 40 ms. Insuccessive recording periods separated by a few minuteswith no intervening epileptiform activity, no significantchange in kernel shape was noted between recordingperiods, suggesting that the white-noise stimulus did nothave enduring effects on processing in the dentate gyrusbeyond the immediate stimulus period (for the stimulusintensities used). In addition, kernels in recently deadanimals were primarily first order (linear) with PRTs <1ms, providing a clear distinction of stimulus artifact andpassive tissue filtering from the signals of a physiologicbasis, as seen in Fig. 1.

In the present study, Wiener kernel analysis indicatedthat processing from the perforant path to the dentategyrus has many features of a signal amplifier (which mayshow fatigue at high levels approaching a threshold forseizure initiation). Further study will characterize the ex-act role between the ability of a stimulus to saturate thisamplifier and to initiate a seizure. It is unclear whetherthe amplification feature behaves in a graded manner orif a threshold phenomenon exists, with higher variancesignals being more likely to exceed that threshold.Analysis of mean square errors of kernel-predicted re-sponses compared with actual responses should helpclarify this. This signal-amplification property (strongsignals are further strengthened) may relate to importantfunctions in the hippocampus.

Very brief PPS (1 min) had very prolonged effects for�25 min on loss of paired-pulse inhibition and for >1 hon Wiener kernel responses. If the same field orienta-tions and generator responses seen with paired-pulsestimulation are active during the white-noise application,then it is reasonable to consider that positive and nega-tive components of kernels likely relate to excitatory andinhibitory activity, respectively. White-noise stimuli ofsufficient amplitude to provoke population spike re-sponses do so with the proper orientation as seen inpaired-pulse responses, supporting this postulate. As-suming this to be so, the negative (inhibitory) componentwas affected more than the positive (excitatory) compo-nent for PPS of 1-min duration. This early loss of inhi-bition with brief PPS may be an initial feature in theprogression to SE. Findings that GABAergic function isaltered (with decreased benzodiazepine-receptor sensi-tivity) by 45 min of status (9,10) are consistent with this.The longer recovery time detected by Wiener kernelmethods may relate to the use of much lower amplitudestimuli than for paired-pulse stimulation; the system mayhave earlier recovery to baseline responsiveness formaximal stimuli (such as high-voltage spikes) than for

more moderate amplitude (albeit continuous) stimuli. Forwhatever reason, the Wiener method appears to be moresensitive here.

By using the month after prolonged PPS (30–60 min)as a model for the “silent period” and epileptogenesis,several features were noted. First, the biphasic (high-pass) first-order kernel noted before PPS (Fig. 1Aii) be-came attenuated in amplitude and showed less consistentamplification features (with more pronounced negativi-ties and oscillatory features) at 1 month (Fig. 1Dii). Thepositive component was affected much more than thenegative component in first-order kernels, indicating dif-ferential as opposed to symmetric effects on positive andnegative components. The appearance in post-PPS ani-mals of an early negative (inhibitory) component (whichmay have been obscured by the robust positivity of pre-PPS recordings) also suggests that the first-order kernelfeatures may derive from distinct positive (excitatory)and negative (inhibitory) pathways that arrive in parallelto the dentate granule cells. These paths may be affecteddifferentially during epileptogenesis.

In addition, the diagonal of the second-order kernelreveals the emergence of a prominent negativity at ∼12ms (Fig. 1Diii) that was not noted before PPS (Fig.1Aiii). This later negativity also is seen in paired-pulseresponses (Fig. 1Di). Increased GABAergic inhibitionassociated with a loss of paired-pulse inhibition of popu-lation spikes has been described (11). Second-order ker-nel off-diagonal positive peaks (solid lines) at t1 � 13ms and t2 � 9 ms were noted at 1 month (Fig. 1Eii,arrows) and suggest that inputs with a 4-ms difference(e.g., 13–9 ms) may couple synergistically and providepotential excitation that did not exist previously. BeforePPS, an off-diagonal negativity (dashed lines) occurredat t1 � 9 ms and t2 � 5 ms (Fig. 1Ei, arrows) and in-dicated an antagonistic (inhibitory) interaction betweeninputs of 4-ms differential. Some EEG phenomenon,such as “fast ripples,” have a similar interpeak time dif-ference and have been associated with epilepsy (12,13).The 200-Hz “ripples” are noted more often during slow-wave sleep as well (14). Therefore, the prolonged fea-tures noted during the post-PPS period provide newinteractions, coupling and cross-talk, among inputs to thedentate. This offers new potential for excitatory behavior(if inputs are delivered at appropriate intervals), whichcould favor progression to seizures. The kernels showthat excitatory and inhibitory components both are at-tenuated significantly in amplitude during this period,suggesting that the tonic balance between such compo-nents may not be so important as the dynamic and quali-tative interactions described earlier. Amplitudeattenuation of both inhibitory and excitatory postsynapticcurrents during the “silent period” has been noted in apilocarpine based model (15, and Coulter D, WONOEPpresentation, 2001).

D. NAYLOR192


It also is evident that, despite some recovery by 1week after PPS (Fig. 1C), the kernels at 1 month havemany features (including amplitude attenuation) similarto what was observed in the immediate post-PPS period(compare Fig. 1B and D). This may suggest that chroniccircuitry alterations associated with epileptogenesis maymimic many of the immediate effects of SE on circuitry(at least from a signal-processing perspective). For ex-ample, those cells that are injured or dysfunctional im-mediately, despite transient recovery, may include cellsthat are affected or lost in the long term.

REFERENCES

1. Olney JW, de Gubareff T, Labruyere J. Seizure-related brain dam-age induced by cholinergic agents. Nature 1983;301:520–2.

2. Ben-Ari Y, Lagowska J, Tremblay E, Le Gal La Salle G. A newmodel of focal status epilepticus: intra-amygdaloid application ofkainic acid elicits repetitive secondarily generalized convulsiveseizures. Brain Res 1979;163:176–9.

3. Cavalheiro EA, Leite JP, Bortolotto ZA, Turski WA, IkonomidouC, Turski L. Long-term effects of pilocarpine in rats: structuraldamage of the brain triggers kindling and spontaneous recurrentseizures. Epilepsia 1991;32:778–82.

4. Sloviter RS. Decreased hippocampal inhibition and a selective lossof interneurons in experimental epilepsy. Science 1987;235:73–6.

5. Mazarati AM, Wasterlain CG, Sankar R, Shin D. Self-sustainingstatus epilepticus after brief electrical stimulation of the perforantpath. Brain Res 1998;801:2513.

6. Lee YW, Schetzen M. Measurement of the Wiener kernels of anonlinear system by cross-correlation. Int J Control 1965;2:237–54.

7. Sclabassi RJ, Eriksson JL, Port RL, Robinson GB, Berger TW.Nonlinear systems analysis of the hippocampal perforant path-dentate projection. I. Theoretical and interpretational consider-ations. J Neurophysiol 1988;60:1066–76.

8. Berger TW, Eriksson JL, Ciarolla DA, Sclabassi RJ. Nonlinearsystems analysis of the hippocampal perforant path-dentate pro-jection. II. Effects of random impulse train stimulation. J Neuro-physiol 1988;60:1076–94.

9. Macdonald RL, Kapur J. Acute cellular alterations in the hippo-campus after status epilepticus. Epilepsia 1999;40(Suppl 1):S9–20;discussion S21–2.

10. Kapur J, Macdonald RL. Rapid seizure-induced reduction of ben-zodiazepine and Zn2+ sensitivity of hippocampal dentate granulecell GABAA receptors. J Neurosci 1997; 17:7532–40.

11. Sayin U, Rutecki PA, Mellanby J, Sutula TP. Gamma-vinyl GABAreduces paired pulse inhibition in the rat dentate gyrus in vivo andin vitro. Epilepsy Res 2001;44:109–17.

12. Bragin A, Engel J, Jr., Wilson CL, Vizentin E, Mathern GW.Electrophysiologic analysis of a chronic seizure model after uni-lateral hippocampal KA injection. Epilepsia 1999; 40:1210–21.

13. Bragin A, Engel J, Jr., Wilson CL, Fried I, Mathern GW. Hippo-campal and entorhinal cortex high-frequency oscillations (100–500Hz) in human epileptic brain and in kainic acid–treated rats withchronic seizures. Epilepsia 1999;40:127–37.

14. Draguhn A, Traub RD, Bibbig A, Schmitz D. Ripple (approxi-mately 200-Hz) oscillations in temporal structures. J Clin Neuro-physiol 2000;17:361–76.

15. Rice A, Rafiq A, Shapiro SM, Jakoi ER, Coulter DA, DeLorenzoRJ. Long-lasting reduction of inhibitory function and gamma-aminobutyric acid type A receptor subunit mRNA expression in amodel of temporal lobe epilepsy. Proc Natl Acad Sci USA 1996;93:9665–9.



changes in nonlinear signal processing in rat hippocampus associated with loss of paired-pulse...

Documents