HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 98/2/769    most recent 00929.2006v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (2) Citing Articles via Google Scholar Google Scholar Articles by Weyand, T. G. Search for Related Content PubMed PubMed Citation Articles by Weyand, T. G.

Retinogeniculate Transmission in Wakefulness

Department of Cell Biology and Anatomy, Louisiana State University Health Sciences Center, New Orleans, Louisiana

Submitted 1 September 2006; accepted in final form 4 June 2007

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Despite popular belief that the primary function of the thalamus is to "gate" sensory inputs by state, few studies have attempted to directly characterize the efficacy of such gating in the awake, behaving animal. I measured the efficacy of retinogeniculate transmission in the awake cat by taking advantage of the fact that many neurons in the lateral geniculate nucleus (LGN) are dominated by a single retinal input, and that this input produces a distinct event known as the S-potential. Retinal input failed to produce an LGN action potential half of the time. However, success or failure was powerfully tied to the recency of the S-potential. Short intervals tend to be successful and long intervals unsuccessful. For four of 12 neurons, the probability that a given S-potential could cause a spike exceeded 90% if that S-potential was preceded by an S-potential within the previous 10 ms (100 Hz). Whereas this temporal influence on efficacy has been demonstrated extensively in anesthetized animals, wakefulness is different in several ways. Overall efficacy is better in wakefulness than in anesthesia, the durations of facilitating effects are briefer in wakefulness, efficacy of long intervals is superior in wakefulness, and the temporal dependence can be briefly disrupted by altering background illumination. The last two observations may be particularly significant. Increased success at long intervals in wakefulness provides additional evidence that the spike code of the anesthetized animal is not the spike code of the awake animal. Altering retinogeniculate efficacy by altering visual conditions undermines the influence inter-S-potential interval might have in determining efficacy in the real world. Finally, S-potential amplitude, duration, and even slope are dynamic and systematic within wakefulness; providing further support that the S-potential is the extracellular signature of the retinal EPSP.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The lateral geniculate nucleus (LGN) provides the strategic link between the retina and cortex, from which visual perception emerges. Coupling between the retina and the LGN is powerful and many, if not most, LGN neurons are dominated by a single retinal ganglion cell (Cleland et al. 1971; Levick et al. 1972; Mastronarde 1987, 1992; Usrey et al. 1999). This, along with the observation that the receptive fields of LGN neurons are extremely similar to those found in the retina, make a persuasive argument that the LGN functions as a "relay" of its retinal input. This relay is not always engaged; sometimes it is "on" and sometimes it is "off." The most promoted controller is state; the animal's level of alertness dictates retinogeniculate fidelity (reviewed in McCormick and Bal 1994; Sherman and Koch 1986; Singer 1977). It is also understood that control of fidelity by alertness is less a "switch" than a "knob." From this perspective, the LGN can be viewed as a variable gain amplifier (maximum gain 1.0) of its retinal input, with level of alertness controlling the knob or gain control. Thus gain would be low in an anesthetized (unconscious) cat with its eye propped open and high in an alert cat tracking a canary.

As simple and attractive as this idea is (it is the most promoted function of the LGN in textbooks), it ignores a few important details. Most significant is the transform. The LGN is not really a relay. Receptive fields of LGN neurons have a more powerful surround than their retinal counterparts, and this property is unlikely to simply be a nuance. Hubel and Wiesel (1961) made as persuasive an argument as any by showing that as the size of a flashing disc centered over the receptive field increased, the response of the retinal input decreased only slightly, whereas the LGN neuron was silenced (their Fig. 4). Other investigators have subsequently made more parametric investigations in paralyzed, anesthetized animals to show how the spatiotemporal attributes of stimuli (spatial frequency, contrast, temporal frequency) influence retinogeniculate fidelity (Cheng et al. 1995; Hamamoto et al. 1994; Kaplan et al. 1987; Mukherjee and Kaplan 1995). Their conclusion is that the LGN is a spatiotemporal filter of its retinal input.

View larger version (25K): [in this window] [in a new window]   FIG. 1. Relationship between S-potentials and lateral geniculate nucleus (LGN) spikes. A: a single 250-ms trace showing 9 S-potential failures and 19 LGN spikes. B: 9 failures and first 9 LGN spikes aligned to their initial rise. Note the excellent correspondence between the rising slope of the S-potential failures and the rising slope of the LGN spikes. Also, note variable delay between initial rise and LGN spike. C: traces in B have been truncated and the time base expanded to allow a better view of these critical events. Proposition here is that for a restricted but large set of LGN neurons, there is only one major retinal drive, it causes the S-potential [the retinal excitatory postsynaptic potential (EPSP)], and S-potentials drive nearly all LGN spikes. For this trace, and all traces that follow, the original analogue traces (digitized at 22.5 kHz) have been smoothed using a spline interpolation.

View larger version (28K): [in this window] [in a new window]   FIG. 2. Sample overlaid traces (10 each) from each of 12 S-potential–LGN pairs analyzed in this study. Traces are rank-ordered by efficacy. Traces were sampled from small regions of each file, so some may not exhibit much variability in amplitude. Overall, and in agreement with Wang et al. (1985), amplitude could vary by a factor of 2.

View larger version (41K): [in this window] [in a new window]   FIG. 3. Observed S-potential intervals for each of the 12 S-potential–LGN pairs analyzed in this study. As with Fig. 2, the order of presentation is by efficacy in driving action potentials.

View larger version (11K): [in this window] [in a new window]   FIG. 4. An LGN neuron that appears to be receiving 2 discernible inputs. A: sample trace (100 ms) showing many EPSP-like events, some of which appear to sum on each other at very short intervals (indicated by arrows). Compare this trace with others presented herein. B: superimposed traces of 10 EPSP-like events, all aligned to an initial slope shift. C: distribution of 1,173 inter-EPSP-like events taken near the location in the file of the traces shown in A and B. Majority of intervals were <5 ms, contrasting sharply with the intervals displayed in Fig. 3.

From this brief discussion, it should be appreciated that retinogeniculate fidelity, and thus LGN activity, turns on several factors. 1) Level of alertness, the most widely promoted, is important and numerous indirect studies indicate that retinogeniculate transmission improves as state moves from slow-wave sleep to wakefulness (Livingstone and Hubel 1981; Sawei et al. 1988), to arousal (Swadlow and Weyand 1985), and even possibly attention (O'Conner et al. 2002). 2) The spatiotemporal attributes of the stimulus are critical and expose the fragility of the relay label. The temporal summation of retinal activity described above adds a further complication to understanding how the first two factors interact. 3) Finally, oculomotor dynamics are also an influence in retinogeniculate transmission. The most studied of these is excitability around the time of eye movements (processes presumably involved with saccadic suppression). However, these effects may be minor compared with the other two (e.g., see Fig. 4 of Lee and Malpeli 1998; Noda 1975; Reppas et al. 2002).

Perception occurs when we are awake. When this simple fact is coupled with the widespread belief that wakefulness is an important determinant of retinogeniculate efficacy, it would seem of interest to study such transmission in a system unfettered by anesthesia and systemic paralysis. Although a number of studies (some cited earlier) have made indirect measurements of shifts in retinogeniculate fidelity associated with state, direct measurement of retinogeniculate fidelity has been done only a few times (Coenen and Vendrik 1972; Hirsch et al. 1983; Sakakura 1968). In the prior direct studies, the main comparison was to determine how fidelity shifted with state (sleeping and waking). There was little commentary on the stability or factors controlling fidelity within wakefulness. The following study examined some properties of retinogeniculate fidelity over time in awake, behaving cats. Within wakefulness, the LGN ignores half of the impulses generated by its main retinal input. However, the temporal distribution of the input was a powerful predictor of success and failure. Similar to what had been described in paralyzed, anesthetized animals, brief retinal intervals increased LGN spike probability and long intervals decreased spike probability. However, transmission in wakefulness has differences. Efficacy is higher, the duration of facilitating effects is briefer, long inter-S-potential intervals have a much better chance of generating LGN spikes, and it is relatively easy to briefly disrupt the facilitating influence of S-potential interval by shifting background illumination.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Strategy

The goal here is to describe retinogeniculate efficacy in the awake animal in the resting state. Historically, efficacy has been directly measured two ways: by determining the efficacy by which a particular retinal ganglion cell can drive an LGN cell to which it is connected (Cleland et al. 1971; Mastronarde 1987, 1992; Usrey et al. 1999) or by measuring the ratio of LGN spikes to S-potentials (e.g., Hubel and Wiesel 1961). Because obtaining connected pairs in the awake animal is not practical, the latter method was adopted. This was also the method used in the previous studies of retinogeniculate efficacy in wakefulness (Coenen and Vendrik 1972; Sakakura 1968). The S-potential is a distinct extracellular event and is thought to represent the retinal excitatory postsynaptic potential (EPSP) (Bishop et al. 1962b; McIlwain and Creutzfeldt 1975). Figure 1A illustrates a single trace record from the current study showing S-potentials that fail to produce action potentials and others that appear to trigger action potentials. That all of these events emerge from the same single source is reinforced by Fig. 1, B and C, in which the initial deviations from baseline for the first 18 events (nine spikes and nine "failed" S-potentials) in Fig. 1A are aligned. Figure 1B includes the peaks of the action potentials, whereas these events are truncated in Fig. 1C to show increased resolution of the initial deflections. The validity of this study depends on acceptance that all of these events emerge from a single source—the S-potential—and that this represents the retinally derived EPSP. Evidences for these assumptions are presented below. To measure efficacy, two other assumptions should also be true: that the S-potential represents the output of a single ganglion cell and that the S-potential drives all LGN spikes. Both of these latter assumptions need some qualifications that are explained below.

The S-potential is the retinal EPSP

The S-potentials illustrated in Fig. 1 are identical to those described by previous investigators who also had complementary intracellular records. These investigators noted the correspondence between the extracellular S-potential and the intracellular retinal EPSP (Hirsch et al.1983; Kato et al. 1971; McIlwain and Creutfeldt 1967). From extracellular records alone, there are good reasons for believing the S-potential is the retinal EPSP. Under some fortuitous conditions, connected pairs of retinal ganglion and LGN cells also include the S-potential. When the "driving" retinal ganglion cell produces an action potential, an S-potential always follows at a nearly fixed latency, from which an action potential may or may not appear (e.g., Cleland et al. 1971; Mastronarde 1987). The rising slope and amplitude of S-potential failures and the rising slope and slight inflection preceding LGN spikes fit well together, suggesting a sequence of EPSP to action potential (e.g., Fig. 1; cf. Bishop et al. 1962b).¹ Figure 2 shows superimposed traces of example S-potentials taken from the 12 LGN-S-potential pairs analyzed in this study. Note that these events show variability in both amplitude and duration. Amplitude variation proved to be systematic and, as subsequently presented, served to reinforce the idea that these events are EPSPs. Variability in duration appeared at least partly related to the recording situation; if the cell had much of a negative component, the S-potential was shorter in duration, whereas records that resembled intracellular events had longer durations.

The S-potential represents the output of a single retinal ganglion cell

This statement is true for a select population of LGN cells (12/15 studied here). Given the recording constraints, the most compelling argument is based on inter-S-potential intervals. If the S-potential reflects a single input, then it must necessarily display intervals consistent with a single neuron, and the distribution of intervals should correspond to those that others have observed among retinal ganglion cells. This appears to be the case for the12 neurons analyzed. Figure 3 illustrates the distribution of the observed S-potential intervals. The distribution of these intervals is similar to distributions of retinal ganglion cell intervals observed by others (e.g., Levine and Cleland 2001; Rathbun et al. 2007; Usrey et al. 1998).

The assertion that the S-potential represents the output of a single retinal ganglion cell for 12 of the pairs studied here is bolstered when contrasted with LGN cells that appear to receive multiple (presumably retinal) inputs. Figure 4 illustrates such a case. Figure 4A illustrates a sample trace that reveals several examples of S-potentials that appear to add to one another at intervals (arrows) shorter than any that appear in Fig. 3, and intervals incompatible with a single input. Figure 4B illustrates the individual traces of S-potentials aligned to their initial slope. Relative to Figs. 2 and 3, it is clear that many of these events have intervals <1 ms, an interval not observed in the intervals illustrated in Fig. 3 (in fact, 98% of S-potential intervals were >4 ms).

The S-potential drives all LGN action potentials

For the recording situation here, this statement is impossible to prove, but the assertion would appear true nearly all of the time in wakefulness. Under ideal conditions, one would identify the retinal ganglion cell that drove the LGN cell and show that 1) every retinal spike produced an S-potential that may or may not produce an LGN spike (a description of efficacy) and 2) every LGN spike occurred if and only if there was an antecedent S-potential caused by that retinal ganglion cell (a description of contribution). Such a situation may apply to a large fraction of X-cells (80%; Mastronarde 1987) and about 30% of Y-cells in the cat (Mastronarde 1992; but see Usrey et al. 1999). In the absence of paired recording, one is left with resemblance (e.g., Fig. 1). Lee et al. (1983) did a quantitative analysis in a single LGN cell in the monkey and found that 99% (2,480/2,505) of the spikes showed an inflection point consistent with being driven by the companion S-potential. Coenen and Vendrik (1972) commented that some LGN spikes failed to show an inflection, but such events were sufficiently rare to have no impact in determining efficacy. In the current investigation, an inflection could be observed about 95% of the time. Although this number is less than that observed by Lee et al. (1983), it is identical to the recent study by Sincich et al. (2007) in the monkey and does not significantly influence the results reported here.

Whereas Fig. 1 nicely illustrates the basic causal relationship between the S-potential and LGN spikes, Fig. 5 shows some of the problems associated with interpreting the relationship between S-potentials and LGN spikes. Figure 5A shows a single trace of nine S-potential failures and eight LGN spikes. In Fig. 5B, the onset of presumably all S-potential events are aligned, and one can appreciate the good correspondence of the rising slopes for all. In Fig. 5C, the vertical scale has been truncated to provide a better view of this correspondence. Of the eight spikes illustrated, two are significantly delayed relative to the onset of the S-potential. Variability in delay was previously described by Levick et al. (1972), although it is unlikely they observed the extensive range observed here (0.1 to 8+ ms; see following text). Figure 5D provides evidence that of the two delayed spikes, one is a delayed sodium spike (green) triggered by one of the aligned S-potentials, whereas the latter spike (red) is asserted to be a spike generated by a barely visible S-potential appearing 3.8 ms after the S-potentials aligned in Fig. 5C. The assertion that the green spike is delayed is based on the nearly vertical rise shown in the close-up of Fig. 5D, whereas the latter red spike appears to emerge from a second S-potential as evidenced by how well the initial rise conforms with the prior S-potential (black trace). All S-potential–LGN pairs showed delays to differing degrees. With the exception of one cell that appeared to have a different synaptic circuit (discussed in the following text), as best as can be determined, delays were <4 ms. Problems of discerning the existence of an S-potential embedded in a rising sodium spike were almost always related to short inter-S-potential intervals such as the example in Fig. 5D, or short interspike intervals that appeared to have no point of inflection on the second spike. The rarest event related to action potentials emerging out of a flat line with no inflection and following a long quiet interval.

View larger version (18K): [in this window] [in a new window]   FIG. 5. Problems of interpretation. A: a single 250-ms trace showing S-potentials and 8 LGN action potentials. B: alignment of the rising slopes of all but one event (explained below), with action potentials traces illustrated in red. C: detailed view of B. There appear to be 2 delayed spikes. D: detailed view of the onset of the 2 delayed spikes, superimposed with a single S-potential failure (black trace). Green trace is an action potential that emerges from a falling S-potential 2 ms previously (it is the 2nd to the last spike in C), and there is no indication that it is emerging from yet another S-potential. Red trace is an action potential that, when superimposed on the sample failure (black), suggests it emerges from an S-potential that is barely discernible (and that occurred 3.8 ms after the prior S-potential). For analysis, the red spike is assumed to be caused by the barely discernible S-potential, and the green trace is assumed to caused by an S-potential that occurred 2 ms before spike initiation.

There is a noted exception to the assertion that S-potentials drive all LGN spikes: "bursts," which represent the extracellular signature of the "low-threshold" (LT) calcium channel. Activation generates a large, but transient inward calcium current that also trips voltage-gated sodium channels, yielding a high-frequency burst of action potentials (Jahnsen and Llinas 1984). Thus under the right circumstances, a single S-potential would be capable of generating a brief train of LGN spikes (Hirsch et al. 1983; Lo et al. 1991; Wang et al. 1985). Using extracellular criteria based solely on interspike interval, we previously found that bursts constitute about 1% of spikes within wakefulness (Weyand et al. 2001). That incidence is even lower here, possibly because in the previous study we also observed that 50% of the bursts were associated with 14% of the neurons. None of the 12 neurons analyzed here was "bursty." Figure 6 shows an example of the unusual—two presumed LT bursts in wakefulness.

View larger version (28K): [in this window] [in a new window]   FIG. 6. Example of the rarely observed bursting in wakefulness. Shown are 2 bursts that appear remarkably similar to the low-threshold (LT) events described in intracellular records (e.g., Jahnsen and Llinas 1984). Top 2 traces show the horizontal and vertical eye position ("H" and "V"), and the next trace down shows the activity of the LGN neuron during a trial. Two bursts are indicated by the asterisks. Bottom trace: 2 bursts at higher temporal resolution. LT "spikes" are a large depolarization of greater duration than typical EPSPs onto which multiple high-frequency sodium spikes are superimposed.

All experiments were approved by the Institutional Animal Care and Use Committee at LSU Health Sciences Center. Adult cats were prepared for chronic single-cell recording using methods described previously (Malpeli et al. 1992; Weyand and Gafka 1998). Briefly, under barbiturate anesthesia, a permanent aluminum "crown" was affixed to the skull using anchoring screws and orthopedic cement. The crown would be used to rigidly hold the head during recording and training sessions. At this time, a Teflon-coated wire loop was also attached to the sclera of one eye. This loop would be used to determine gaze during the recording sessions using the magnetic search coil technique (Robinson 1962). The cat was given postoperative analgesics and at least 1 wk to recover from this surgery before training was begun using a food reward.

Training

After about 24 h of food deprivation, the cat was placed in a loose-fitting canvas bag and the head fixed to a Plexiglas frame. The bagged cat rested comfortably on a folded towel placed in a Plexiglas hemitube that was suspended by surgical tubing. The animal faced a dimly lit (0.5 cd/m²) screen on which images such as bars and square-wave gratings (contrast 0.4), as well as a 0.2° spot produced by a dimmed low-power laser could be projected. The cat was trained to look at the laser spot and, if it jumped to a new location, make a saccade to reacquire fixation on the spot. The cat was required to fixate the spot to an accuracy of 1.5° of the spot and to fixate for 1–3 s to obtain a reward of 1–2 g of the veterinary equivalent of beef-flavored baby food. The spot was used to control gaze and the projected bars and gratings could be used to probe the receptive field of the cell under study.

Second surgery

Once the cat was comfortable with working in the apparatus, the animal had a second brief surgery to implant a swiveling cannula and base assembly to allow easy microelectrode access to the LGN when the animal was awake. The cat was again anesthetized and a hole made through the orthopedic cement and skull overlying the LGN. The exposed dura was then slit and the cannula and base were cemented into place, with the cannula protruding approximately 6 mm into the brain. The cannula held a stylus that would be replaced with a microelectrode for the recording sessions. The cannula was held in place by a low-melting-point (46°C) alloy in the base. Touching this alloy with a heated solder iron allowed the cannula to swivel, and thus change electrode trajectory with minimal exposure of the brain to outside conditions. Such a procedure did not appreciably heat the cannula.

Recording/analysis

For recording, the stylus was removed and replaced with a sterilized tungsten-in-glass microelectrode protruding from a miniature microdrive. The microelectrode was slowly advanced through the brain until a single neuron in the LGN could be isolated. For this report, all neurons were in the dorsal A layers of the LGN, and successful isolation of S-potential–LGN pairs appeared to require higher impedance (1–1.5 M at 1 kHz) electrodes than used previously in the laboratory (e.g., Weyand et al. 2001). To distinguish normal extracellular recording from recording from these pairs, I will refer to situations in which pairs were isolated as "juxtacellular" recording. In such records (e.g., Fig. 1), the action potential is mostly positive, as was the S-potential. Probably because these electrodes were of higher impedance and recording was done in an awake animal, most large, isolated cells were lost within a few seconds. This report describes records obtained from 12 S-potential–LGN pairs held between 3 and 43 min. A summary of their basic properties (when known) is provided in Table 1. All of the neurons were relatively stable over this period, visually responsive (e.g., all gave a vigorous discharge to waving hand in front of cat's face), and had action potential durations of <1 ms. An initial segment-soma dendritic inflection point was usually not discernible unless the cell was injured (cf. Bishop et al. 1962b; Wang et al. 1985). Such records are not included here. Although the cells were stable over the period studied (see following text for additional analyses), some occasionally displayed high-frequency trains of spikes (10–50) that I attribute to mechanical stimulation of the membrane. Because the trains were <2% of total spikes, were episodic, did not appear to be mediated by S-potentials, and did not alter the stability of the temporal tuning characteristics subsequently described, they were excluded from analysis. Finally, some of the records contained occasional "doublets," i.e., two sodium spikes in rapid succession (500 Hz), with only the first spike having an S-potential. For purposes of analysis the second spike was omitted.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
S-potential interval powerfully predicts LGN spike probability

In wakefulness, S-potentials succeeded in driving action potentials about half the time. Figure 7 illustrates the overall efficacy of the 12 S-potential–LGN pairs analyzed. Average efficacy was 0.50, and the range was 0.23 to 0.72. Efficacy of retinogeniculate transmission varied directly with retinal activity. Figure 8 illustrates a sample continuous record of S-potential and LGN spike activity over a 60-s period. Although it is not surprising that the activity is coupled, efficacy (viewed here as the ratio between the two lines) appears to often vary with S-potential rate. High rates were successful and low rates were less successful. All 12 neurons exhibited this behavior to varying degrees, and this predictive relationship between retinal activity and efficacy is illustrated in Fig. 9. Each graph was created by determining the ratio of S-potentials that caused an LGN spike divided by the total number of S-potentials over each 4-ms epoch, and then fitting those data points to a sigmoid.² The horizontal line indicates the "naïve" probability line, the probability that any S-potential would cause an LGN spike without taking recency of prior S-potential into account. For most cells, inter-S-potential intervals of <20 ms indicated a higher than average probability that the LGN neuron would spike. Equally impressive was how S-potentials with long preceding periods of quiescence made an LGN spike an improbable event. For eight of 12 neurons, if an S-potential had not occurred in 50 ms, the next S-potential had <10% chance of triggering an LGN spike.

View larger version (7K): [in this window] [in a new window]   FIG. 7. Overall efficacy (LGN spikes/S-potentials) for the 12 LGN cells analyzed.

View larger version (21K): [in this window] [in a new window]   FIG. 8. Continuous (Gaussian-filtered) record of S-potential (red) and LGN (black) activity over a 60-s period illustrating not only the tight coupling between the S-potential and LGN activity, but also suggesting that efficacy was related to S-potential rate.

View larger version (41K): [in this window] [in a new window]   FIG. 9. S-potential interval predicts LGN spikes. Probabilities of LGN spikes were determined by dividing intervals into 4-ms bins and determining for each bin the ratio of S-potentials causing an LGN spike (successes) divided by all S-potentials in that interval (successes + failures). Points were then fitted to a sigmoid. Horizontal line is the overall efficacy of the S-potential in driving that LGN neuron, what might be considered the "naïve" probability. For each panel, the efficacy is stated in the top right, and the goodness of the data fit to a sigmoid are expressed by the "fit" term. These figures summarize interactions of >112,000 events (S-potential, LGN spikes).

View larger version (11K): [in this window] [in a new window]   FIG. 10. Predictive relationship of spike probability to S-potential recency is stable over time. A: 2 curves generated by splitting a data file in half, and then plotting spike probability as a function of time for the 1st half of the data (red curve) and 2nd half of the data (black curve); the 2 functions are nearly identical. B: histograms showing the results of doing this same procedure for all 12 LGN cells. Plotted is the "residual error," the difference between the 2 integrals plotted within cells. For example, in A this difference was –4.92 (negative because the curve constructed from the 1st half of the data had a faster drop than the data obtained from the 2nd half). Magnitude of these errors, which reflect stability of the functions across time, is small when compared with the errors observed when comparing across cells. C: error observed when comparing the curve from each cell with another (132 comparisons). Two conclusions: the "filters" were stable over time (essentially the same curve could be constructed from the 1st or 2nd half of the data), and a single canonical function linking LGN spike probability with S-potential recency appears unlikely.

View larger version (9K): [in this window] [in a new window]   FIG. 11. Visual stimulation yielded high efficacy (100% in this case). A: trace showing activity surrounding presentation of a rectangular "flash" (duration indicated by bar, rectangle was 4.8 x 3.5°) over the receptive field of an ON-center neuron. Note that S-potential failures (indicated by asterisks) were common during the periods before and after the visual stimulus. B: increased detail of trace immediately after flash.

The vast majority of the data were collected under stable conditions in which the cat faced a dimly lit screen and was intermittently required to fixate a small spot in the center of the screen. However, for several neurons changing background illumination had strong effects on activity. This proved to be a simple method to alter the temporal characteristics of retinogeniculate transmission. Figure 12 shows a dramatic example for an ON-center S-potential–LGN pair before and after darkening the recording chamber. The top trace shows a 250-ms trace of activity immediately preceding darkening the chamber. The next trace shows that immediately on darkening, no S-potentials or spikes occurred. The third trace shows the first five S-potentials to appear as the retinal activity returned. All five produced a spike. In fact, the first 19 S-potentials all produced spikes before a failure appeared. Efficacy remained high (80%) for 9 s despite an exceedingly low rate. The bottom trace reveals that reilluminating the chamber (indicated by bar) resulted in a "screaming" retina, but miserable efficacy. This effect is nothing more than the nonlinear gain control in transmission associated with light adaptation, but emphasizes the complexities of retinogeniculate transmission. The strong dependence of S-potential recency and efficacy observed in the light-adapted state essentially reversed: efficacy was high for long-intervals in the dark and low for the short intervals when the background illumination returned.

View larger version (14K): [in this window] [in a new window]   FIG. 12. Temporal dependence on efficacy can be fragile. Top trace: activity of a light-adapted ON-center S-potential–LGN pair just before darkening the chamber. Next trace shows the activity (none) immediately after darkening the chamber. Next trace shows the 1st 5 S-potentials that appeared as the retinal activity returned. All produced spikes, and for this example, the 1st 19 S-potentials to appear all drove spikes. Bottom trace: when the chamber is illuminated again, the S-potentials return at a very high rate. However, efficacy is poor and the S-potentials are small. Temporal facilitation does not apply here.

In the current experiments, the cats were awake throughout the recording sessions. Wakefulness is a broadly defined state that includes quiet wakefulness, arousal, and attention. Although not parametrically studied here, there was no evidence to suggest that manipulating vigilance had any additional influence on retinogeniculate efficacy. Figure 13 provides an example. The top panel shows three 10-s (Gaussian-filtered) traces indicating the average efficacy as the cat goes through trials in which the animal must hold fixation for 1 s to receive a food reward. The fixation period is indicated by the horizontal bar. Relative to periods preceding and following these trials in which the cat must attend to a central fixation point (because in each trial the cat executed a saccade to acquire the target), shifts in efficacy were unremarkable. The bottom panel of Fig. 13 shows two traces of a period surrounding one of the trials, illustrating that "failures" were just as high before and after as during the trial. For another S-potential–LGN pair whose receptive field was very close to area centralis, analysis of >40 trials requiring the cat to fixate yielded the same result. Efficacy increased during fixation, but that increase was linked to decreased inter-S-potential intervals (increased retinal activity) with no additional influence when the animal was required to fixate (and "attend," 2–4 s) versus the intertrial interval (6–8 s).

View larger version (25K): [in this window] [in a new window]   FIG. 13. Shifting levels of alertness have little influence on efficacy. Top: Gaussian-filtered traces of retinogeniculate efficacy over time for 3 rewarded trials in which the cat was required to fixate on a spot at the center of the screen for 1 s to obtain a reward (fixation period indicated by horizontal bar); the only interesting feature of fixation is that efficacy values were similar. Receptive field for this "ON-center" cell was about 4° eccentric and near the horizon. During fixation, the stimulus within the receptive field was at an unmodulated grayish-green low-contrast screen. Bottom: trace of period around one of the trials (efficacy of trial shown as black line at the top). Top trace: S-potential and LGN activity in the 2 s preceding the trial. Bottom trace: period from the onset of the fixation light to 0.5 s after the delivery of the reward. Note that the record includes the LGN spike (tall positive event), the S-potential (positive event 25% of amplitude of spike), as well as another (unknown) event.

Previous extracellular studies of S-potentials have made scant mention of amplitude variation. Levick et al. (1972) noted variation in amplitude associated with different retinal inputs, Wang et al. (1985) mentioned variation of nearly 50% was observed (although they never speculated on underlying causes), and Cheng et al. (1995) specifically stated that they did not observe amplitude variation. In this study, amplitude variation was systematic and consistent with what one would expect of the retinal EPSP. The retinogeniculate synapse is described as "depressing"; i.e., when the input is stimulated with high-frequency stimulus pairs, the second EPSP is smaller than the first. This phenomenon is known as paired-pulse depression (Chen and Rogehr 2000; Turner and Salt 1998). Figure 14 shows the direct relationship between increasing S-potential interval and amplitude. Although the relationship is far from perfect, it is highly significant statistically (P < 0.001). Considering how active membrane conductances are in wakefulness (cf. Hirsch et al. 1983), it is perhaps almost remarkable that this relationship was as easy to demonstrate here.

View larger version (12K): [in this window] [in a new window]   FIG. 14. S-potential amplitude increases with inter-S-potential interval. Plotted are the amplitudes of 770 failures, taken in the light-adapted state. Increased amplitude with increased EPSP interval is characteristic of depressing synapses, a property associated with retinogeniculate synapses in vitro.

View larger version (19K): [in this window] [in a new window]   FIG. 15. S-potential amplitude could vary with illumination level. Red traces are 8 superimposed "failures" extracted from records immediately before darkening the chamber. Black traces are 8 superimposed "failures" extracted from records immediately after darkening the chamber; the 2 sets do not overlap. Systematic and predictable S-potential amplitude variation such as this is consistent with the behavior expected of the retinogeniculate EPSP.

One LGN cell stood out from the other 11. This cell was heavily bombarded by the retina (65 impulses/s), yet was the least efficacious (23%). Whereas rising slope of the S-potentials was relatively constant among the other cells, it was (by comparison) more variable here (e.g., Fig. 2, bottom right). Delay in spike initiation from onset of the S-potential, something observed in all other cells, was taken to an extreme here; occasionally reaching 8 ms. Presenting a large (8° square) drifting grating over the receptive field revealed several unusual features of this cell. Figure 16A shows the activity surrounding presentation of the drifting grating (grating onset/offset indicated by arrows). Onset of the grating (shown in greater detail in Fig. 16B) results in sudden silence of the neuron and S-potential. Ironically, visual stimulation created the longest inter-S-potential intervals in this cell's record. The stimulus causes a phasic response (Fig. 16, A and B), although the initial response is not a spike, but a large S-potential. For this neuron, using this stimulus provided a method of producing the largest S-potentials observed in the records. Figure 16C shows an S-potential sequence from this trial. On the left, are the last 10 S-potentials preceding visual stimulation (black), followed by the first 20 S-potentials during visual stimulation (blue), followed by the first 10 S-potentials following visual stimulation (black). This figure shows that during visual stimulation, the long "pause" yields the largest S-potentials observed and, although variable, S-potential amplitude during visual stimulation was larger than that at other periods (0.786 vs. 0.625 mV, P < 0.001). The interpretation is that the duration of the interval preceding the S-potential predicts amplitude because this is a depressing synapse, and the overall increase in size is attributable to increased input resistance of the cell associated with using the drifting grating as the visual stimulus. This latter assertion arises from interpretation of the traces illustrated in Fig. 16D . The top trace shows the "background" activity taken from a section of the record a few seconds before visual stimulation (this cell was so active it was difficult to find sections without an S-potential or small negative spike for 20+ ms). The bottom trace shows activity about 60 ms after onset of visual stimulation. The bottom trace has more small, fast events, that by comparison are larger than those observed in the top trace. These are likely the same events in the top and bottom traces, just larger in the bottom trace. Events >50 µV are more frequent during visual stimulation (P < 0.001), an observation that would be consistent with a cell whose input resistance suddenly increased. The increased amplitude of S-potentials associated with visual stimulation are a product of increased inter-S-potential interval and increased input resistance. The increased noise is also (barely) discernible in Fig. 16, A and B. The effect was reliable and most pronounced at the beginning of visual stimulation.

View larger version (24K): [in this window] [in a new window]   FIG. 16. An unusual cell in which visual stimulation suppresses activity. A: trace of the activity of an LGN bracketed around a trial in which a large drifting grating was placed over the cell's receptive field (grating onset/offset indicated by arrows; stimulus was an 8 ° square, containing vertically oriented drifting square-wave gratings (0.23 cycles/deg, drift 2.86 Hz, stimulus contrast 0.40). For this trace, and at this timescale, note the high rate of activity before and after presentation of the grating, and the lack of activity during visual stimulation except at a particular phase of the grating. B: close-up of the period at the start of presentation of the grating (indicated by arrow) and the period immediately preceding. Three events of note. There is not a single S-potential for >0.2 s when the grating starts. When an S-potential does finally appear, it is large. "Fuzzy" activity associated with the initial portion of the visual response (to the right of the arrow) indicates the input resistance of the neuron has increased. C: S-potential amplitude shifts during visual stimulation. Shown are the initial 0.7 ms for the 10 S-potentials immediately preceding visual stimulation (black traces to the left), the first 20 S-potentials that appear during visual stimulation (blue), and the first 10 S-potentials that appear after the cessation of visual stimulation. S-potential amplitude is variable. Part of this variability is attributable to the time because the prior S-potential, and, in this case, part is attributable to the increased input resistance associated with the grating. For each of the blue traces, the large ones tend to be associated with the first S-potential that appears after a long interval. D: "noise" increases with visual stimulation. For this neuron, and at least this stimulus, an increased level of background activity was noticeable. Top trace (black) is taken from a 20-ms period before this trial, in which there were no S-potentials, action potentials, or negative spikes. Bottom trace (red) is taken from activity about 60 ms after the onset of the visual stimulus. Top trace is flatter, but the big difference is that the frequency of events >50 µV are increased during this period (P < 0.01). Such an observation is consistent with increased input resistance and/or a hyperpolarized membrane potential. As might be apparent in trace B, the effect was transient and most pronounced in the initial response to the grating.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

Retinogeniculate efficacy in wakefulness

Sakakura (1968) and Coenen and Vendrik (1972) previously studied retinogeniculate efficacy in awake cats, making the same assumptions about the circuits made here: LGN cells appear to be dominated by a single retinal input, and nearly all LGN spikes are caused by S-potentials. Sakakura (1968) obtained values from six neurons, but chose to provide quantitative data on only one (presumably representative) neuron. This cell had an overall efficacy of 0.55 in wakefulness, a number close to the average efficacy observed here. Efficacy measures from Coenen and Vendrik (1972), who used unanesthetized, but paralyzed cats, is a bit more difficult to extract. However, it does appear similar. During visual stimulation in wakefulness, they obtained efficacy of about 90+%. This corresponds well with the present results: visual stimulation delivers high S-potential frequency that results in high efficacy (e.g., Fig. 11). For epochs in which their cats were awake, but no visual stimuli present over the receptive field (the bulk of the data here), it would appear that their observations favor efficacy between 50 and 70%. This is higher, but not remarkably different from the present results.

Wakefulness improves retinogeniculate efficacy because increased activity of the ascending reticular activating system (ARAS) depolarizes LGN neurons, increasing the probability that a retinal EPSP will trigger an action potential (reviewed in McCormick and Bal 1994; Singer 1977). For the transition from slow-wave sleep to wakefulness, studies suggest efficacy improves by about 30–100% (Coenen and Vendrik 1972; Livingstone and Hubel 1981; Sakakura 1968). Because wakefulness itself is a broad category, it would be reasonable to expect additional increases in activity and/or efficacy as the animal moves from quiescence ("inattentive vision"; Bezdudnaya et al. 2006) to vigilance. However, such shifts appear to be modest or nonexistent unless coupled with visual stimulation (Bezdudnaya et al. 2006; Sawei et al. 1988; Swadlow and Weyand 1987).³ The present study appears consistent with this: shifts in vigilance did not yield shifts in activity or efficacy (e.g., Fig. 13). Anesthesia depresses the ARAS (i.e., the animal is "asleep"), depresses retinal ganglion cell sensitivity (McIlwain 1964), and would be expected to depress retinogeniculate efficacy. Overall, this expectation appears to be realized because the 50% efficacy observed here is roughly double the efficacy observed from merging studies in the cat. However, it is worth noting that variability across studies and across neurons is extensive. In the LGN of the anesthetized monkey, average retinogeniculate efficacy was identical to that observed here (Lee et al. 1985; Sincich et al. 2007). Whereas most connected retinogeniculate pairs appear well below 50% efficacy in anesthesia, Cleland and Lee (1985) describe a pair in the anesthetized cat operating at nearly 100%.

Inter-S-potential interval had a dramatic effect on retinogeniculate efficacy in wakefulness. Short intervals increased probability, long intervals decreased probability. Thus the facilitating temporal effects observed in anesthetized animals (Kara and Reid 2003; Levine and Cleland 2001; Mastronarde 1987; Rowe and Fischer 2001; Sincich et al. 2007; Singer et al. 1972; Usrey et al. 1998) extends into wakefulness as well. An obvious rationale for such temporal filtering is a mechanism for noise reduction. Constant low-frequency activity by the retina is ignored; high-frequency activity (brief interspike intervals) has a much better chance in evoking LGN spikes.

Whereas the general shape of functions relating recency of retinal events to LGN spike probability is similar in wakefulness and anesthesia, there appear to be at least three differences. One is that in wakefulness, short inter-S-potential intervals appear more successful in driving LGN spikes than under anesthesia (Mastronarde 1987; Sincich et al. 2007; Usrey et al. 1998). Efficacy exceeded 90% for four of 12 neurons for intervals <10 ms, and all cells were most efficacious at short intervals. Second, the duration of the facilitating effects was shorter in wakefulness, averaging 18.8 ms (point at which curves cross "naïve" efficacy lines in Fig. 9) versus 30+ ms reported in anesthetized animals (Mastronarde 1987; Rowe and Fischer 2001; Sincich et al. 2007; Usrey et al. 1998). Third, in anesthesia, LGN spike probability appears to converge on zero for long intervals (40 ms) between retinal spikes or S-potentials (Mastronarde 1987; Sincich et al. 2007; Singer et al. 1973; Usrey et al. 1998). This was not usually the case in wakefulness (Fig. 9), where probability hovered around 10% and, for two cells, was 40%. For all cells, about 12% of all S-potential intervals were 40 ms, and the average level of success was about 13%. From inspection of the data of prior published studies in anesthetized preparations, 1% efficacy at equivalent intervals would appear generous. This apparent dependence on summation to drive spikes in anesthesia (cf., Sincich et al. 2007) may be the most significant difference between these two states.

How powerful is the S-potential interval in predicting LGN spikes under natural conditions?

Despite early expectations of high fidelity of retinogeniculate transmission in wakefulness (e.g., Singer 1977), the failure of retinogeniculate efficacy to move to 100% simply because the animal is awake or alert is logical. The LGN sometimes "ignores" its main retinal drive because the LGN does something more significant than gating retinal input by state. The retinogeniculate transform appears to engage in judicious "pruning" (deletions) of the main retinal drive to create a different receptive field (e.g., a more powerful surround; Hubel and Wiesel 1961). The reasonable argument that the retinogeniculate transform would also include "insertions" from an anonymous source appears unlikely for this set of cells (discussed in METHODS).

As robust as temporal dynamics alone appears in predicting success and failure at the retinogeniculate synapse, there is good reason to believe the pattern of visual stimulation would significantly alter a simple decay function. The dramatic disruption obtained by altering background illumination demonstrated here (Fig. 12) is consistent with prior demonstrations that the spatiotemporal structure of visual stimuli is used to restructrure the LGN cell's receptive field from its main retinal drive (Cheng et al. 1995; Hamamoto et al. 1994; Hubel and Wiesel 1961; Kaplan et al. 1987).

Deciphering the "spike code" is perhaps the greatest challenge in neuroscience. The results here provide another caveat regarding the utility of reading the spike code in anesthetized animals. Although everyone agrees that anesthesia depresses LGN activity, and most agree that bursts are uncommon in wakefulness (Ruiz et al. 2006; Weyand et al. 2001), the current study indicates anesthesia sets a bias such that only retinal spikes preceded by relatively short intervals continue to cortex. The degree to which these long intervals contribute to visual perception is unknown. However, the spike trains that LGN neurons produce when the animal is exposed to time-varying natural images show a strong tendency to decorrelate; natural images induce the production of a broad range of interspike intervals. Such decorrelation occurs not only in the LGN of awake, behaving cats (Dastjerdi et al. 2003) and monkeys (Dong et al. 2005), but has been demonstrated in the LGN of the anesthethetized cat (Dan et al. 1996). The above indicates anesthesia provides a truncated, corrupted version of this decorrelation; long intervals have been chopped.

Methods: the S-potential and retinogeniculate transfer

These results not only provide support that S-potentials are the extracellular signature of the retinal EPSP, they also provide insight into retinogeniculate transmission in wakefulness. As first described by Levick and colleagues (1972), there is some variability in the delay between the onset of the S-potential and the generation of the action potential. Whereas it cannot be dismissed that part of this delay is attributable to damage, that the delay correlated with longer S-potential intervals and was stable argues that some of this variability is physiological. As described earlier, in vitro studies have characterized the retinogeniculate synapse as "depressing," i.e., paired stimulation of the optic tract at even modest frequencies (10 Hz) yields significant reduction of the second EPSP (Chen and Regehr 2000; Salt and Turner 1998). The data displayed in Fig. 14 provide support that these observations extend to in vivo conditions and wakefulness. That brief S-potential intervals increase LGN spike probability yet correlate with decreased S-potential amplitude may seem paradoxical. However, decreased amplitude of the EPSP at the retinogeniculate synapse can emerge from several documented sources: membrane depolarization, decreased transmitter release, receptor saturation, and receptor desensitization (Chen and Regehr 2000; Chen et al. 2002; Salt and Turner 1998). Among these, only depolarization would facilitate postsynaptic spike probability. The first EPSP would depolarize the membrane and a second EPSP that quickly followed would be smaller even if there were no decrease in transmitter release, receptor saturation, or receptor desensitization due to the decrease in driving force associated with the depolarization from the first EPSP. Due to temporal summation, the second EPSP, although smaller than the first, would provide sufficient depolarization to cross the spike threshold.

In principle, using the S-potential to measure the retinogeniculate transform is powerful and attractive. Despite a promising beginning (Bishop et al. 1958, 1962b; Hubel and Wiesel 1961), it has not been routinely used. Obtaining stable records of S-potential–LGN spike pairs is much more difficult than single-cell records, because one needs to be extremely close to the soma to pick up the S-potential. Even the records obtained here, although stable, could usually be held for only a few minutes, and the variable delay in spike initiation observed in all cells does raise concerns. Observing S-potentials at all is sometimes mysterious; one can be in the LGN and have a large nearly positive only spike, and yet no S-potential (e.g., Fig. 6 of an LT burst exhibited no apparent S-potentials; cf. Wang et al. 1985). The type of electrode may well influence what one records. Funke and Worgotter (1995) report that the glass-coated tungsten electrodes are particularly well suited for picking up S-potentials (presumably due to the combination of low capacitance of thick glass and the high conductance of tungsten wire). The inability to routinely observe S-potentials because the electrodes are too low in impedance, or because of other reasons, may also foster a prejudice to dismiss S-potentials as nonexistent or some second cell in the record. This is unfortunate. S-potentials are distinct events that obey a refractoriness and pattern of activity compatible with retinal ganglion cells. Their refractoriness with the action potential indicates they are a coupled event, greatly reducing the possibility that these events are a second cell. When the low-frequency portions of the signals are allowed to pass (e.g., <1 Hz here) S-potentials look like EPSPs, appear only when the LGN spike is large and mostly positive, disappear simultaneously with the LGN spike, and display more amplitude modulation than what one observes with action potentials; this modulation corresponds exactly to the kind of modulation one expects of the retinal EPSP (Figs. 14, 15, and 16).

	GRANTS