Proc. Natl. Acad. Sci. USA Vol. 90, pp. 6533-6537, July 1993 Neurobiology lonotropic non-N-methyl-D-aspartate agonists induce retraction of dendritic spinules from retinal horizontal cells [plasicity/sh retna/a-amno-3-hydroxy-5-methyl-4-isoxazoleproplonic acid/(±- )-trans-1-amino-1,3-cyclopentanedicarboxylic acdd/6-cyclo-7-nitroquinoxafine-2,3-done] RETO WEILER* AND KONRAD SCHULTZ Department of Neurobiology, University of Oldenburg, P.O. Box 2503, 2900 Oldenburg, Germany Communicated by John E. Dowling, April 1, 1993

ABSTRACT Horizontal cells invaginate the photorecep- this synaptic plasticity is -45 min, and recent evidence tors in the retina and form reciprocal synaptic connections in suggests the involvement of protein kinase C (4). The neu- the cone pedicles. In fish retina the pattern of synaptic con- ronal pathway mediating the retraction of the spinules seems nections is plastic and modulated by the ambient light condi- to include an excitatory amino acid receptor. Glutamate and tions. Numerous dendritic spinules protrude from the terminal its analog kainate (KA) are able to induce retraction of horizontal-cell dendrites into the cone pedicle when the retina spinules in a light-adapted retina when injected into the is light-adapted and are retracted during dark adaptation. The vitreous humor (5). Both glutamate and KA were active when retraction of spinules can be induced during maintained ilHu- horizontal cells were synaptically isolated with cobalt, sug- mination by an injection of the putative cone transmitter gesting a direct action on horizontal cells or cones, but the L-glutamate or its analogue kainic acid into the vitreous humor. type of receptor involved was not further analyzed. It is now The formation and the retraction ofspinules have a time course well established that two of the excitatory amino acid recep- of minutes. Activation of protein kinas C through phorbol tors, the N-methyl-D-aspartate (NMDA) receptor and the esters initiates the formation ofspinules, but the retraction has metabotropic quisqualic (QU) receptor, through their link to not yet been linked to a specific second messenger. Herein we internal calcium concentration and coupling to phospholipase report that physiological concentrations of the glutamate an- C, respectively, contribute to the modulation of neuronal alogs quisqualic acid and a-amino-3-hydroxy-5-methyl-4- networks (for review, see ref. 6). isoxazoleproplonic acid induce retraction of spinules during This paper shows that the glutamate-induced retraction of maintained illumination. (±)-trans-l-Amino-1,3-cyclopen- dendritic spinules in horizontal cells is not mediated by tanedicarboxylic acid, an agonist for the metabotropic quis- metabotropic QU receptors or by NMDA receptors. Instead qualic acid receptor, was without effect on spinule retraction. the data suggest the involvement of an ionotropic a-amino- N-Methyl-D-aspartate and L-2-amino-4-phosphonobutyric 3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)- acid, agonists at other types of glutamate receptors, were also sensitive receptor. without any effect. The effects of the active agonists persisted when synaptic transmission was blocked. In the presence ofthe MATERIALS AND METHODS ionotropic quisqualate receptor antagonist 6-cyclo-7-nitro- All experiments were done with carp of body length of =15 quinoxaline-2,3-dione, the effects of all active agonists were cm. The fish were kept in aerated tanks under a controlled blocked. These results demonstrate that activation of ionotro- light-dark cycle of 12 h. For an experiment the fish was taken pic quisqualate receptors on the horizontal-cell membrane can out ofthe tank in the morning and for 4 h put in a smaller tank induce dendritic spinule retraction, a process associated with that was illuminated with a 100-w tungsten lamp. All drugs dark adaptation. were injected into the right eye of an animal under anaes- thesia with MS 222. The left eye served as a control and Horizontal cells are interneurons ofthe retina contributing to received an injection of Ringer solution. Standard Ringer the processing of light stimuli in the outer plexiform layer. solution, which was also used as a solvent, consisted of 90 They are postsynaptic to photoreceptors, and functionally mM NaCl, 2.5 mM KCI, 0.5 mM MgCl2, 2.5 mM CaCl2, 25.0 similar horizontal cells are electrically coupled through gap mM NaHCO3, and 10.0 mM glucose and was bubbled with junctions. Their major output is an inhibitory feedback onto 95% 02/5% C02, maintaining the pH at 7.3. A small incision cone photoreceptors that contributes to the spatial resolution was made into the sclera and drugs were delivered into the of more central retinal neurons and to color opponency (for vitreous humor using a Hamilton syringe with a blunt needle reviews, see refs. 1 and 2). There is considerable electro- tip. The following drugs were used: L-glutamic acid, L-as- physiological and morphological evidence that the feedback partic acid, NMDA, QU, KA, L-2-amino-4-phosphonobu- pathway is plastic and modulated by the ambient light con- tyric acid (L-AP4), AMPA, (+)-trans-l-amino-1,3-cyclopen- ditions. Although the exact site of the transmitter release is tanedicarboxylic acid (ACPD), 6-cycyo-7-nitroquinoxaline- not demonstrable morphologically due to the nonvesicular 2,3-dione (CNQX). L-Glutamate, L-aspartate, NMDA, nature ofthis process in horizontal cells, there is considerable AMPA, and L-AP4 were obtained from Sigma; KA and electrophysiological and morphological evidence that the CNQX were from Cambridge Research Biochemicals (Valley feedback from horizontal cells to cones involves the spinules Stream, NY) and ACPD was from Tocris Neuramin (Briston, on the terminal dendrites of the horizontal cells. For exam- England). The concentration of the drug in the injected ple, during dark adaptation, the spinules are retracted and red-green biphasic horizontal cells lose the depolarizing Abbreviations: AMPA, a-amino-3-hydroxy-5-methyl-4-isoxazole- responses to long wavelength stimuli (3). The time course of propionic acid; ACPD, (t)-trans-1-amino-1,3-cyclopentanedicar- boxylic acid; NMDA, N-methyl-D-aspartate; L-AP4, L-2-amino-4- phosphonobutyric acid; CNQX, 6-cyclo-7-nitroquinoxaline-2,3- The publication costs of this article were defrayed in part by page charge dione; QU, quisqualic acid; KA, kainic acid; SPR, spinules per payment. This article must therefore be hereby marked "advertisement" ribbon. in accordance with 18 U.S.C. §1734 solely to indicate this fact. *To whom reprint requests should be addressed. 6533 Downloaded by guest on September 24, 2021 6534 Neurobiology: Weiler and Schultz Proc. Natl. Acad Sci. USA 90 (1993) volume and the respective volume (5-10 ,ul) were chosen to which appear as large pale profiles in a cross section at the give, in accordance with the size ofthe eyeball and under the level of the cone pedicles (Fig. la), had numerous spinules, assumption of an even distribution within the eyeball, the identified by their size and the associated membrane densities desired estimated concentration at the retinal level. Concen- (7). In a given section, typically two to three spinules were trations given throughout the result section are estimated found per synaptic ribbon (Fig. 2). Such spinules were not ocular concentrations based on the above procedure. present in the right retina of the animal if glutamate (20 mM) After the injection the animal recovered from anaesthesia was injected 45 min prior to enucleation (Fig. lb). Glutamate within 5 min, was brought back into the illuminated tank, and not only induced the disappearance of spinules but also allowed to swim for 45 min. Thereafter, the animal was affected the pattern of the horizontal-cell terminal dendrites decapitated and the eyes were enucleated and hemisected. within the cone pedicles. Under glutamate treatment, the The retina was carefully removed and immersed into the dendrites were less dispersed within the pedicle and more fixative [1% paraformaldehyde/2.5% (vol/vol) glutaralde- confined to its center. This centralization is similar to what hyde/3% (wt/vol) sucrose/0.05 M sodium phosphate, pH happens during dark adaptation and was also seen in parallel 7.3] and fixed overnight at 4°C. Conventional protocols were control experiments where the retinas were dark-adapted for used to obtain ultrathin sections from the resin-embedded 45 min (Fig. lc). Radial sections of the cone pedicles con- retinas. Cone pedicles were analyzed in tangential and radial firmed the overall confinement of the horizontal-cell termi- sections from the central retina. A quantitative evaluation of nals to the center of the pedicle during drug treatment (Figs. spinule expression was achieved by counting the number of 1 e andf). In glutamate-treated retinas, some ofthe horizontal spinules per ribbon (SPR) in a minimum of 30 randomly terminals showed patchy and fluffy membrane densities distributed cone pedicles from three to seven animals (7). along their membranes but pronounced membrane densities Standard deviations were included in the histograms. associated with spinules were not present. An injection of aspartate (20 mM) prior to enucleation also RESULTS affected the ultrastructural organization within the cone In all light-adapted control retinas that received an injection pedicle. Horizontal-cell terminals were shifted toward the of 10 ,ul ofRinger solution, horizontal-cell terminal dendrites, basal center of the pedicle (Fig. lg), but spinules were still

FIG. 1. Electron micrographs of cone pedicles. (a-d) Tangential sections. (a) Light adapted retina. (b) Glutamate (20 mM)-injected retina from the same light-adapted animal. (c) Dark-adapted control retina. (d) AMPA (200 ,uM)-injected light-adapted animal. (Bar = 0.5 ,um.) (e-g) Radial sections. (e) Light-adapted control retina. (f) Glutamate (20 mM)-injected light-adapted retina. (g) Aspartate (20 mM)-injected light-adapted retina. hc, Horizontal-cell terminal; sr, synaptic ribbon; arrows in a, spinules; arrowheads in b and d, patchy membrane densities. (Bar = 1 ,m.) Downloaded by guest on September 24, 2021 Neurobiology: Weiler and Schultz Proc. Natl. Acad. Sci. USA 90 (1993) 6535 3.0

FIG. 2. Bars: open, from control retinas; solid, from retinas FIG. 3. Bars: open, from control retinas; solid, from retinas injected with the agonist; hatched, from retinas injected with the injected with the indicated agonists [NMDA (100 uM, n = 4), AP4 agonist under blockade of the synaptic transmission with cobalt (5 (1 mM, n = 4), KA (50 MAM, n = 6), or QU (50 MM, n = 7)]. mM); L, light adapted; D, dark adapted. Light- and dark-adapted data are averages from another set of experiments (n = 16), as are eyes (n = 3; data not shown), which makes it unlikely that glutamate (Glu, 20 mM, n = 6) and aspartate (Asp, 20 mM, n = 6) activation of NMDA receptors contributed to spinule retrac- data. tion. present although their overall number was reduced compared The effectiveness of KA and QU, but not of NMDA or to that of a light-adapted control (Fig. 2). The above effects AP4, suggests the involvement of QU receptors. Two types of glutamate and aspartate were achieved with estimated of QU receptors, ionotropic and metabotropic, can be dis- concentrations of the two agonists ranging from 5 to 50 mM. tinguished by the selective agonists AMPA and ACPD, Lower concentrations were without consistent effects and respectively (6, 11). AMPA (10 ,uM) reduced the number of higher concentrations induced clear signs of tissue damage spinules and 100 MM AMPA had an effect comparable to that (e.g., vacuoles and swellings). Although the above concen- of 20 mM glutamate (Fig. 4). At a concentration of 200 ,uM trations are estimates based on the calculations and assump- (Fig. ld), AMPA induced almost complete retraction and the tions and do not take into account the powerful uptake number of spinules was the same as in a dark-adapted retina. systems for excitatory amino acids existing within the retina, Some of the horizontal-cell terminals had patchy and fluffy the effective range coincided with the range found to be membrane densities along their membranes, similar to the ones seen with 20 mM glutamate in Fig. lb. Replacement of effective in electrophysiological experiments (8). Fig. 2 sum- calcium in the Ringer solution with 5 mM cobalt did not block marizes the results obtained with glutamate and aspartate and the effect of AMPA, and the difference between the light- for comparison the SPR value of a dark-adapted retina is adapted control and the treated retina was highly significant. included. The difference between the glutamate-treated ret- In contrast, ACPD at concentrations of 10 ,uM to 500 MM was inas and the light-adapted control retinas was highly signif- without any effect on spinules (Fig. 4). icant (P < 0.001). The SPR values in the aspartate-treated The above results suggest the involvement of ionotropic retinas do not differ from the light control. However, it is but not metabotropic QU receptors. CNQX has been estab- important to mention that, although the number of spinules lished as a fairly specific antagonist for the ionotropic did not significantly differ, their morphological arrangement (AMPA-sensitive) QU receptor (12). CNQX (20 ,uM) signif- did. This included a pronounced centralization of the hori- icantly blocked the effect of 100 uM AMPA, 20 mM gluta- zontal-cell terminals within the cone pedicle typical for mate, 20 mM aspartate, and 50 MuM QU on spinule retraction dark-adapted controls (Fig. lg). The same experiments were (Fig. 5), but even at concentrations as high as 100 MM, it did repeated in the presence of 5 mM cobalt, which synaptically not prevent the retraction of spinules during dark adaptation. isolates horizontal cells in the carp retina (9). The effects of glutamate were also observed in the presence of cobalt, whereas the effects of aspartate were not (Fig. 2). Injection ofNMDA (10MuM to 50 mM) did not cause spinule retraction in a light-adapted retina, and L-AP4 (100 ,uM to 10 mM) was also ineffective (Fig. 3). As reported (5), KA caused retraction of spinules. At a concentration of 10 MM, KA reduced the SPR value to that found in dark-adapted control retinas. QU (10-100 ,uM) also induced retraction of spinules. At a concentration of 50 MuM, QU was comparable to gluta- mate at 20 mM. The effects of KA and QU were not reduced in the presence of 5 mM cobalt. To test whether the failure of NMDA to cause spinule retraction was due to the block of NMDA-gated channels at the relatively hyperpolarized mem- brane potentials under experimental conditions, experiments were performed in which the NMDA was given with a low dose (1 MuM) of QU, which caused a depolarization of horizontal cells (10) and caused about a half-maximal retrac- FIG. 4. Bars: open, from control retinas; solid, from retinas tion of spinules. The combination of 1 ,uM QU and 100 ,M injected with the agonists; hatched, from retinas injected with the NMDA was injected into both eyes. In addition, the specific agonist under blockade of the synaptic transmission with cobalt (5 NMDA antagonist D-AP5 (100 ,uM) was injected into the left mM). Agonists were AMPA (100 ,uM, n = 5) and ACPD (200 ,uM, n eye. There was no difference in SPR values between the two = 6). Downloaded by guest on September 24, 2021 6536 Neurobiology: Weiler and Schultz Proc. Natl. Acad. Sci. USA 90 (1993) 3.0 The observation that the effect of the active agonists with the exception of aspartate (see below) persisted in the pres- ence of cobalt suggests that these agonists act directly on horizontal cells. Cobalt blocks synaptic transmission from photoreceptors to second-order neurons and thus synapti- 2.0 cally isolates them (9). The observed effects must, therefore, ct either result from direct action of the agonists on horizontal cells or from an action on cones that nonsynaptically triggers

co the morphological changes. The previously reported obser- 1.0 vation that the effects of KA on spinule retraction can be blocked by piperidinedicarboxylic acid (5), which blocks the depolarizing effect of KA on horizontal cells without affect- ing the light responses of cones (25), is in favor of a direct action. However, this does not rule out the possibility that Glu Asp QU AMPA there is also an action on cone pedicles. Indeed the sensitivity of the effects of aspartate to the presence of cobalt suggests FIG. 5. Bars: solid, from retinas injected with the agonists; such an involvement. Since aspartate depolarizes horizontal crossed-hatched, from retinas from the same animals injected with cells in the intact carp retina but not isolated horizontal cells the indicated agonist together with the antagonist CNQX (20 ,uM). (8), it most likely acts via an indirect pathway, which might Agonists were glutamate (Glu, 20 mM, n = 3), aspartate (Asp, 20 also be responsible for the effects on spinules. It is interesting mM, n = 3), QU (50 AM, n = 4), and AMPA (100 ,uM, n = 4). to note that the effects of aspartate differed from those of all other active agonists. Whereas aspartate was not a potent DISCUSSION signal for the dissolution of spinules, it was a potent signal for The results of this study suggest that spinules, formed on the shifting ofthe horizontal-cell dendrites within the pedicle. terminal dendrites of horizontal cells during light adaptation, Spinules are characterized by their shape and the associated are retracted after activation of ionotropic QU receptors. A membrane densities (7). Our results indicate that the active quantitative comparison ofthe different agonists is difficult to agonists, while being a potent signal for the dissolution ofthe obtain. The method used in this in vivo study does not allow shape of the spinules, were less potent in affecting their precise measurements of the concentrations at the retinal membrane densities. In all experiments the electron micro- level. The diffusion within the vitreous and the uptake by graphs revealed patchy membrane densities along the mem- neurons and glial cells of the retina might vary for different branes of the horizontal cell terminals. The location and the drugs, thus making a statement about the retinal concentra- features of these patches suggest very much that they are tion impossible. Despite these limitations, the results suggest identical to the membrane densities associated with spinules. the following order of potency: KA > AMPA > QU > This would imply that their formation and disintegration is glutamate. Aspartate did not significantly affect the SPR partially independent from the shape-forming process. A value but affected the morphology. NMDA, AP4, and ACPD similar result has recently been found from studies on the were completely ineffective. contribution of actin polymerization to spinule formation (22). There is abundant electrophysiological evidence for the The effects of glutamate and its agonists on spinule retrac- existence of non-NMDA glutamate receptors on fish hori- tion cannot be explained simply by their depolarizing action, zontal cells. Administration of either L-glutamate or L-aspar- since depolarization by elevation of the external potassium tate to the fish retina depolarizes horizontal cells and reduces concentration did not cause spinule retraction (5). Another their light-evoked responses (8, 9, 13-15). In the intact retina, explanation for the observed effects of glutamate and its relatively high concentrations (in the millimolar range) are active agonists might be that the AMPA receptor is linked to needed due to the presence of powerful uptake systems for a second messenger system, in addition to its direct action on these amino acids. The sensitivity for analogs such as KA or ion channels. Recent evidence obtained by N. Osborne QU, which are not taken up, is about 250 times higher (8). (personal communication) and in our laboratory suggests that Saturation of the uptake systems increases the sensitivity to AMPA increases significantly the formation of inositol glutamate and aspartate '15 times (16). In isolated fish trisphosphate in the rabbit and fish retina. There is evidence horizontal cells, L-glutamate but not L-aspartate elicits large that the enzyme phospholipase C can be activated by calcium depolarizations, and KA and QU are 10 times more potent (26) and calcium entering through the AMPA-gated channel than L-glutamate (8, 10, 17-21). (27) might trigger such an activation. The fact that activation It has recently been shown that the formation of spinules of AMPA receptors causes morphological changes at synap- requires the polymerization of actin (22) and activation of tic sites further demonstrates how intimately interrelated protein kinase C (4). Light and dark adaptation are both ionotropic and metabotropic effects of agonists may be. accompanied by changes in the phosphorylation pattern of phosphoproteins separated from horizontal cells (23). It We thank Dr. J. McReynolds for critical reading ofthe manuscript. seems very likely that the process of retraction involves This work was supported by a grant from the Deutsche Forschungs- cellular events coupled to second messenger systems. This gemeinschaft. suggests that the observed effects of glutamate and its agonists might be due to the activation of a metabotropic 1. Kaneko, A. (1987) Jpn. J. Physiol. 37, 341-358. receptor linked to the phospholipid pathway and generating 2. Dowling, J. E. (1987) The Retina (Harvard Univ. Press, Cam- inositol trisphosphate. However, the metabotropic QU re- bridge, MA). ceptor agonist ACPD was without effect, whereas the iono- 3. Weiler, R. & Wagner, H.-J. (1984) Brain Res. 298, 1-9. AMPA 4. Weiler, R., Kohler, K. & Janssen, U. (1991) Proc. Natl. Acad. tropic QU receptor agonist induced spinule retraction Sci. USA 88, 3603-3607. at low concentrations. The effect of AMPA was blocked by 5. Weiler, R., Kohler, K., Kirsch, M. & Wagner, H.-J. (1988) the ionotropic QU receptor antagonist CNQX, which has Neurosci. Lett. 87, 205-209. little affinity for the metabotropic QU receptor (24). It is 6. Schoepp, D., Bockaert, J. & Sladeczek, F. (1990) Trends therefore unlikely that the effect of AMPA resulted from the Pharmacol. 11, 508-515. activation of the metabotropic QU receptor. 7. Wagner, H.-J. (1980) J. Neurocytol. 9, 573-590. Downloaded by guest on September 24, 2021 Neurobiology: Weiler and Schultz Proc. Natl. Acad. Sci. USA 90 (1993) 6537

8. Ariel, M., Lasater, E. M., Mangel, S. C. & Dowling, J. E. 18. Ishida, A. T. (1984) Brain Res. 298, 25-32. (1984) Brain Res. 295, 179-183. 19. Lasater, E. M., Dowling, J. E. & Ripps, H. (1984) J. Neurosci. 9. Kaneko, A. & Shimazaki, H. (1975) J. Physiol. (London) 252, 4, 1966-1975. 509-522. 20. Perlman, I., Knapp, A. G. & Dowling, J. E. (1989) Brain Res. 10. Ishida, A. T., Kaneko, A. & Tachibana, M. (1984) J. Physiol. 487, 16-25. (London) 348, 255-270. 21. Tachibana, M. (1985) J. Physiol. (London) 358, 153-167. 11. Barnard, E. A. & Henley, J. M. (1990) Trends Pharmacol. 11, 22. Weiler, R. & Janssen-Bienhold, U. (1993) J. Neurocytol. 22, 500-507. 129-139. 12. Honore, T., Davies, S. N., Drejer, G., Fletcher, E. J., Jacobsen, P., Lodge, D. & Nielsen, F. E. (1988) Science 241, 701-703. 23. Janssen-Bienhold, U., Nagel, H. & Weiler, R. (1993) Eur. J. 13. Murakami, M., Ohtsu, K. & Ohtsuka, T. (1972) J. Physiol. Neurosci., in press. (London) 227, 899-913. 24. Murphy, S. N. & Miller, R. J. (1989) Mol. Pharmacol. 38, 14. Sugawara, K. & Negishi, K. (1973) Vision Res. 13, 977-981. 671-680. 15. Wu, S. M. & Dowling, J. E. (1978) Proc. Natl. Acad. Sci. USA 25. Slaughter, M. M. & Miller, R. F. (1983) Science 219, 1230- 75, 5205-5209. 1232. 16. Ishida, A. T. & Fain, G. L. (1981) Proc. Natl. Acad. Sci. USA 26. Baird, J. G., Challiss, R. A. J. & Nahorsky, S. R. (1991) Mol. 78, 5890-5894. Pharmacol. 39, 745-753. 17. Dowling, J. E., Lasater, E. M., VanBuskirk, R. & Watling, 27. Gilbertson, T. A., Scobey, R. & Wilson, M. (1991) Science 251, K. J. (1983) Vision Res. 23, 421-432. 1613-1615. Downloaded by guest on September 24, 2021