On and off sublaminae in the lateral geniculate nucleus of the ferret

Like the retinal ganglion cells from which they receive their input, most relay neurons in the lateral geniculate nucleus have ON- or OFF- center receptive fields with antagonistic surrounds. In the cat, neurons with these two types of receptive fields are anatomically intermingled, even though the ON and OFF systems are functionally segregated. In the ferret, there is a sublamination of the retinal input to lateral geniculate nucleus laminae A and A1. We have investigated the function of this sublamination by making microelectrode recordings and have found that each sublamina consists of geniculate neurons of a single center type.

In higher mammals, information about lightness and darkness in the retinal image appears to be conveyed from the eye to the brain in parallel channels of ONand OFF-center retinal ganglion cell axons (Kuffler, 1953). In the retina, segregation of center types is accomplished by the arborization of depolarizing bipolar cells to contact ON-center ganglion cells within a different sublamina of the inner plexiform layer from that in which hyperpolarizing bipolar cells contact the OFF-center cells (Famiglietti et al., 1976). These ON and OFF channels remain at least largely separate through the synapses of the lateral geniculate nucleus (LGN) (Hubel and Wiesel, 1961;Cleland et al., 1971).
Reports from Schiller's laboratory of a strict laminar segregation of center type in the LGN of the tree shrew (Conway et al., 1980) led us to wonder whether the subtle bisublamination of the main (A) laminae of the ferret LGN, noted by Sanderson (1974) and Guillery and coworkers (Linden et al., 1981) and termed "leaflets," might provide a substrate for the maintenance of separate ON and OFF channels. Microelectrode recordings now reveal that each leaflet consists of geniculate cells of a single center type.
l This work was supported by Grants ROl-EY-02874 and K04-EY-00213 from the National Institutes of Health, by research fellowships from the Sloan Foundation and the National Foundation-March of Dimes, and by a predoctoral fellowship from the National Science Foundation.
We wish to thank Dr. Simon LeVay for comments on an earlier version of this manuscript and Ms. Sheri Strickland for technical assistance.
' To whom correspondence should be addressed, at Department of Physiology, S-762, University of California, San Francisco, CA 94143.

Materials and Methods
Animals. Eight adult ferrets (Mustela putorius furo) were obtained from Marshall Research Animals, Inc., North Rose, NY, and housed in the University of California vivarium for 1 to 8 weeks. For all but two experiments, animals were females weighing between 700 and 850 gm; the others were l.l-to 1.2-kg males.
Labeling retinal afferents. In three ferrets, retinal afferents to the LGN were labeled autoradiographically by intravitreal injections of 200 to 2000 PCi of [3H]proline (Amersham TRK.439, specific activity 40 Ci/mmol). In these cases, an unlabeled or extremely lightly labeled gap was clearly seen between two tiers of label in both laminae A and Al. Following Sanderson (1974), we took the gap in the labeled afferents to indicate the dividing line between the sublaminae or "leaflets." Figure 1 shows such a labeled section, cut coronally from the side contralateral to the injected eye.
Inspecting such labeled sections along with neighboring sections stained by a Nissl method eventually allowed us to discern the locations of the leaflets in most stained sections without reference to the labeled ones. Although the ferret's geniculate nucleus lacks a cell-poor interleaflet plexus, knowledge of the form taken by the leaflets in labeled sections at each level guided us in picking out the subtle differences in neuronal architecture which mark the boundaries of the leaflets. A comparison of the labeled section shown in Figure 1 with the Nissl-stained section from another animal shown in Figure 24 will make clear that it was possible to do this. On the most caudal coronal sections and the most lateral parasagittal sections, determining the borders of the leaflets on unlabeled sections by reference to our atlas series of labeled sections was sometimes difficult; rostrally and medially 1944 Stryker and Zahs Vol. 3, No. 10, Oct. 1983 near the border with the medial intralaminar nucleus, it was sometimes impossible.
Retinal afferents in two other ferrets were labeled by injecting the vitreous humor of one eye with 0.25 mg of wheat germ agglutinin conjugated to horseradish peroxidase (Sigma, catalogue no. L9008) 3 to 5 days before perfusion, and reacting brain sections with tetramethyl benzidine according to the protocol of Meslaum (1978). Figure 3 shows examples of sagittal sections labeled using this procedure.
Microelectrode recording. Eight ferrets were prepared for physiological recording using techniques which are conventional for studies of the cat visual system (Shatz and Stryker, 1978). Briefly, the ferret was initially anesthetized with a mixture of acepromazine and ketamine (0.04 and 40 mg/kg, i.m., respectively) to allow cannulation of the femoral vein and trachea. Thiopentol sodium (20 mg/kg, i.v.) was employed to continue anesthesia throughout the rest of the surgery. The ferret was placed on a heating pad which kept body temperature at 38°C and fitted in a modified kitten stereotaxic instrument; then, the scalp, skull, and dura overlying the LGN were opened. Plastic contact lenses (2.7 to 2.9 mm base curve, plano) were fitted to focus the eyes on a tangent screen subtending 80" at a distance of 57 cm. This screen could be rotated to an angle of 60" to the animal's midsagittal plane when necessary for plotting eccentric receptive fields. Focus, determined by the front surface power of the contact lens, was checked by retinoscopy, and the most appropriate lens was selected. The projections of the two optic discs were then plotted on the tangent screen using a reversing beam ophthalmoscope. Barbiturate infusion was discontinued and anesthesia was maintained by ventilating the ferret with 75% nitrous oxide, 25% oxygen at a rate and volume which maintained peak inspiratory pressure at 1.5 kPa and end-tidal carbon dioxide at 3.8 to 4.3%. Neuromuscular blockade was then induced using pancuronium bromide (0.1 mg/ kg-hr).
Lacquered tungsten microelectrodes (Hubel, 1957) were driven down to the LGN on a vertical axis in five ferrets or on an axis in the parasagittal plane inclined 40 to 45" pointing anterior from the vertical in three ferrets. These electrodes had conical exposed tips tapering from 10 to 15 pm diameter to a sharp point over a length of 20 to 30 pm. With impedances of 1 to 2 megohm at 120 Hz, they were designed to record multiunit activity, although they would sometimes isolate single units. Typically, 2 to 5 different spikes more than 4 times greater than the noise level (50 pV) could be discerned. Such electrodes were advantageous for revealing clusters of units having similar response properties. At most sites within the LGN, all of the units recorded with such electrodes did have similar response properties (Zahs and Stryker, 1982;see Fig. 4, B and C, below): they were driven from the same eye and were of the same center type.
Assignment of recording sites to leaflets. One or more electrolytic lesions were made along the course of each electrode penetration by passing -4 to -6 PA for 4 to 6 sec. Nissl-stained sections containing the microelectrode penetrations were projected at a magnification of x 47, and tracings, including all sublaminar boundaries, were made without knowledge of or reference to the physiological findings. It was then possible, by reference to microdrive readings at the microlesions and at the entry point into the LGN, to reconstruct the locations of all recording sites on these tracings. This procedure located the point of transition between responses from the two eyes accurately at the border between laminae A and Al. Only after assigning each recording site to the C laminae or to one of the leaflets of laminae A and Al (or to the border zone between leaflets) were the physiological ON or OFF responses taken account of and noted on the tracings.
Two penetrations were made into caudal and medial regions of the LGN in which we were unable to identify the borders of the leaflets anatomically. Although ON and OFF responses were segregated here as they were elsewhere, data from the 28 recording sites along these penetrations are excluded from the present analysis.
Data from the initial three of the eight animals used in this study were not available for this correlation because the first two were used only to obtain stereotaxic coordinates and the third was sectioned in the wrong plane for reconstruction of the penetrations.

Results
Anatomy and physiology of LGNs cut in coronal section. The pattern of labeling of retinal afferents to the contralateral LGN is shown in Figure 1. This and subsequent coronal sections are shown with dorsal up and medial to the right. The two tiers of label in lamina A were found to be nearly continuous, and the overall pattern of labeling was similar to that described by Guillery (1971) and Linden et al. (1981). The principal difference between our results and theirs was that it was necessary for us to cut the brains coronally or parasagittally rather than horizontally, so that entire electrode tracks could be traced in one or a few sections.
A typical vertical penetration into the LGN is illustrated in Figure 2. When the electrode first passed into the nucleus, a region of high spontaneous activity and strong visual responses was invariably encountered. Histology later showed this region to be the C laminae of the LGN. The unit clusters here were frequently responsive to both eyes, depending on the eccentricity of the receptive fields in the visual field, and were usually responsive to both the onset and offset of a spot flashed in the center of the receptive field area. The single units isolated at these sites were always monocularly driven and of either ON-or OFF-center type, suggesting that in these laminae units of different types were intermingled within the resolving distance of the multi-unit microelectrode. Some unit clusters (38 of 138), however, were predominantly or exclusively of one sign of contrast, raising the possibility that a detailed organization may exist in these laminae beyond the resolution of our methods.
After a variable distance, the electrode next encountered a region of distinctly lower spontaneous activity and dramatically different response properties. Here, the unit clusters were driven exclusively by one eye and, in 225 of 245 cases, exclusively by either the onset or offset The Journal of Neuroscience ON and OFF Sublaminae in Ferret Lateral Geniculate Nucleus    Vol. 3, No. 10, Oct. 1983 of illumination in the receptive field. All single units 100~pm intervals along each penetration. As the penetraisolated at such sites shared these properties of the tion continued downward and receptive fields progressed cluster and were found to have approximately circular down the visual field (Zahs and Stryker, 1982), the eye receptive fields with mutually antagonistic center-sur-and center type could change or could remain the same round organization similar to that found in the cat (Hu-for distances greater than 1 mm. The penetration shown be1 and Wiesel, 1961).
in Figure 2, for example, passed briefly through a region We noted the eye and center type of unit clusters at of contralateral eye OFF responses before entering an    tion raises the possibility that the sublaminae are not completely continuous. We were unable to confirm this possibility with physiological recordings, for the elec-in Figure 5. The more dorsal penetration nearly followed trode never passed through such a site in a labeled the lines of projection of the visual field so that receptive animal. It is possible that this occasional discontinuity field position changed very little along its course. The is artifact due to incomplete labeling of or damage to the sequence of response types in such penetrations was retinal ganglion cells or axons which would supply input invariably as follows: first, C laminae, with high sponto the unlabeled area.
taneous activity and usually mixed responses; then, un- Figure 4 shows a vertical electrode penetration in less the receptive fields were in the monocular segment parasagittal section. This penetration passed through of the visual field, the outer leaflet of lamina Al containhigh spontaneous mixed ON and OFF activity in the C ing ipsilateral eye OFF-center units followed by the inner laminae and OFF responses in the outer leaflet of lamina leaflet of Al containing ipsilateral ON-center units; and A before encountering ON responses in the inner leaflet finally, the outer leaflet of layer A with contralateral of lamina A. It remained within this leaflet until nearly OFF-center units followed by the inner leaflet with ONthe end of the penetration, when it re-entered the outer center units. The electrode usually then passed into the leaflet of lamina A, again encountering OFF responses. perigeniculate nucleus in which responses and receptive Figure 4, B and C, shows poststimulus histograms for field positions were much different. unit clusters, each containing at least four units, at two sites along this penetration. Note that only an ON re-Discussion sponse is seen in Figure 4B gest strongly that the terminal arbors of retinal afferents The sublaminar distribution of 383 recording sites in are also segregated by center type; indeed, discovering 41 penetrations in five ferrets is shown in Table I. Note the functional correlate of the stratification of the retinal that within the A laminae, mixed ON and OFF responses afferents was the original goal of this study. were encountered primarily at recording sites located at the borders of leaflets. Only 8% of recording sites had These findings shed little light on whether ON-and OFF-center cells are segregated in the C laminae. Our any contribution from responses of the opposite center multi-unit electrodes found only 38 of 138 sites to be type.
The arrangement of the sublaminae was made clearer predominantly or exclusively of one center type, but such electrodes may sample over a distance that is considerby making penetrations as nearly normal to the genicu-able compared to the width of possible sublaminae in the late laminae as could be managed without removing the C laminae. tentorium. Examples of two such penetrations are shown Sublamination may be a general feature of organiza-Stryker and Zahs Vol. 3, No. 10, Oct. 1983 ----.. . Figure 5. Two electrode penetrations inclined in the parasagittal plane drawn onto a photograph of a Nissl-stained parasagittal section through the LGN. Note the sequence of responses encountered as the electrode passed through the leaflets. The arrowhead indicates the site of the electrolytic marking lesion. Other conventions as in Figure 4. tion by which different response properties can remain segregated from one another through stages of synaptic relay. In the LGN, a strict lamination or sublamination of ON-and OFF-center units like that in the ferret has been observed in the tree shrew (Conway et al., 1980) and mink (LeVay and McConnell, 1982). But sublamination may be covert; the retina was studied for many years before the function of ON and OFF sublamination in the inner plexiform layer was noted. If such sublamination were not strict, it would be difficult to discover with single-unit recording techniques. Perhaps a messy or incomplete sublamination may underly the intermingled response types found in the LGN of other animals, as results from the monkey suggest (Schiller and Malpeli, 1978).
Several questions remain about the sublamination observed in the ferret's LGN. (I ) How is the sublamination set up in development? LaMantia and Guillery (1982) have found that binocular enucleation in early postnatal ferrets prevents or disrupts even the much clearer cellular lamination which distinguishes lamina A from Al, while leaving the C laminae distinct. This finding suggests that the afferents do have a role in establishing or maintaining lamination of the A laminae.
(2) What are the projections of the different sublaminae; are they kept separate up to the level of the cortex? (3) Are the sublaminae themselves stratified with respect to X and Y input? These questions are currently under study.
Note added in proof. Since this paper went to press, a partial stratification of OFF responses in the LGN of the cat has been reported (Bowling, 1983). Thus, in the tree shrew, rhesus monkey, mink, ferret, and now, cat, OFF activity is always concentrated in the external (nearer the optic tract) half of geniculate laminae. At the same meeting, Norton et al. (1983) reported ON and OFF stratification in layer IV of the striate cortex of tree shrews.