Unidirectional Photoreceptor-to-Müller Glia Coupling and Unique K+ Channel Expression in Caiman Retina

Background Müller cells, the principal glial cells of the vertebrate retina, are fundamental for the maintenance and function of neuronal cells. In most vertebrates, including humans, Müller cells abundantly express Kir4.1 inwardly rectifying potassium channels responsible for hyperpolarized membrane potential and for various vital functions such as potassium buffering and glutamate clearance; inter-species differences in Kir4.1 expression were, however, observed. Localization and function of potassium channels in Müller cells from the retina of crocodiles remain, hitherto, unknown. Methods We studied retinae of the Spectacled caiman (Caiman crocodilus fuscus), endowed with both diurnal and nocturnal vision, by (i) immunohistochemistry, (ii) whole-cell voltage-clamp, and (iii) fluorescent dye tracing to investigate K+ channel distribution and glia-to-neuron communications. Results Immunohistochemistry revealed that caiman Müller cells, similarly to other vertebrates, express vimentin, GFAP, S100β, and glutamine synthetase. In contrast, Kir4.1 channel protein was not found in Müller cells but was localized in photoreceptor cells. Instead, 2P-domain TASK-1 channels were expressed in Müller cells. Electrophysiological properties of enzymatically dissociated Müller cells without photoreceptors and isolated Müller cells with adhering photoreceptors were significantly different. This suggests ion coupling between Müller cells and photoreceptors in the caiman retina. Sulforhodamine-B injected into cones permeated to adhering Müller cells thus revealing a uni-directional dye coupling. Conclusion Our data indicate that caiman Müller glial cells are unique among vertebrates studied so far by predominantly expressing TASK-1 rather than Kir4.1 K+ channels and by bi-directional ion and uni-directional dye coupling to photoreceptor cells. This coupling may play an important role in specific glia-neuron signaling pathways and in a new type of K+ buffering.


Introduction
Müller glial cells [1] serve numerous fundamental functions in the retina of vertebrates; many of these functions depend on potassium channels, responsible for a high potassium conductance of the cell membrane [2,3,4]. Although the electrophysiological membrane properties, as well as the main functions, of Müller cells are similar among the vertebrates, distinct inter-specific differences have been observed even between closely related mammals such as monkeys and humans [5]. To further investigate Müller cells functional diversity, possibly reflecting adaptations to specific retinal circuits, it is desirable to study Müller cells from different groups of vertebrates. A wide variety of mammalian Müller cells have been investigated (e.g., [6]); as well as fishes (elasmobranchs and teleosts: [7,8,9] and amphibians (salamanders and anurans: [9,10,11,12]. In reptilians, however, only Müller cells from the diurnal water turtle, Pseudemys scripta elegans, were characterized (e.g., [13,14,15,16]). Here we report a study of Müller cells from retinae of caiman (Caiman crocodilus fuscus), which has perfect night vision as well as vision in the bright daylight, with a large scale of adaptation to different light intensities. This ability is reflected by several morphological and functional idiosyncrasies in the caiman vision system [17]. Incidentally, crocodiles are closer related to birds (in which Müller cells were never studied electrophysiologically) than to the turtles (e.g., [18], and references therein) which makes the caiman an even more interesting subject of examination.
Radially oriented Müller cells span the whole thickness of the retina and conduct light to photoreceptors [19]. These cells contact all neuronal elements located within the retina. Spatial buffering of extracellular K + ions represents another most fundamental and extensively studied function of the Müller cell. In dark adapted retina, cells face large K + gradients, with K + concentrations ranging between 6-8 mM at the photoreceptor layer (i.e., at the distal part of Müller cell) and 2-3 mM at the vitreal surface where (i) Müller cell endfeet abut the vitreous body and (ii) complex ionic changes occur during light stimulation [20,21,22,23]. Specific spatial distribution of K + channels [24] allow Müller cells to redistribute K + ions from sites of high extracellular concentration to 'buffering reservoirs' such as the vitreous fluid or the intraretinal blood vessels, and thus prevent elevations of extracellular K + that may cause over-excitation of neurons with subsequent loss of information processing.
In the Müller cells and astrocytes of humans and of most animals studied, inwardly rectifying K + (Kir) channels, specifically Kir4.1 (Kcnj10), play a key role for glia-neuron interactions (for recent reviews, see [3,25,26,27]), being fundamental for example for glutamate clearance [28,29]. Genetic variations of Kir4.1 channels in humans and animals underlie severe disorders in the brain and in the retina, such as epilepsy, disruption of electroretinogram, glaucoma, stroke, ataxia, hypokalemia, hypomagnesemia, and metabolic alkalosis [27,30,31,32,33]. In addition, recently identified Kir4.1 mutations were found to result in autoimmune inhibition, contributing to pathogenesis of multiple sclerosis [34], hearing loss [35], autism [36] and seizures [30,33]. The mutated Kir4.1 protein is not inserted in the membrane or the channels are blocked, as revealed by the absence of representative potassium currents [37]. A loss of Kir4.1 channels function was also found in the retina and in the brain in diabetes [38], transient ischemia [39], and in trauma [40]; with deficit in Kir4.1 mediated permeability being linked to failures in neuronal function and neuronal cell death [3].
Surprisingly in the present study we found that Kir4.1 channels are absent in caiman Müller glial cells, which, however, express TASK channels. Furthermore, we observed a unique communication between photoreceptors and attached Müller cells, with unidirectional permeation of the fluorescent dye, sulforhodamine, from cones to Müller glial cells. We hypothesize that in the caiman retina, spatial K + buffering may involve both Müller cells and photoreceptors via trans-cellular K + fluxes. Preliminary results were reported in abstract form [57,58].

Animals
Experiments were carried out with IACUC approval and in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and according to institutional animal care and use guidelines. Adequate measures were taken to minimize pain or discomfort to experimental animals. Müller cells were isolated from retinae of adult caiman (Caiman crocodilus fuscus) as previously described [12,45,47].

Immunohistochemistry
Retinal sections from 8 caimans were used for the immunohistochemical studies. Eyes were perfusion-fixed in situ. Caimans (with a length of 50 cm to 120 cm on average) were immobilized on ice before being anesthetized. Animals were subsequently i.p. anesthetized with tiletamine HCl/Zolazepam HCl, 5 mg/kg before intraventricular perfusion of fixatives. The thorax was opened and a catheter was placed in the left ventricular chamber of the heart for vascular perfusion. Perfusion for immunohistochemistry contained 4% paraformaldehyde (as the only fixative) in phosphate buffered solution (PBS): NaCl 136.9 mM, KCl 2.7 mM, Na 2 HPO 4 10.1 mM, KH 2 PO 4 1.8 mM with pH 7.4 for 30 minutes. The right atrium was cut to allow outflow of perfusate. After perfusion, eyes were enucleated and pieces (0.5 x 0.5 mm) of the retina were cut and rinsed in fresh PBS.
For agar-embedding we used the isolated retinal pieces embedded in 3% agarose (w/v) in PBS as previously described [39,47] and agar cubes with tissue inside were cut by 80 mm sections. For cryo-embedding, after fixative perfusion the tissues were cryoprotected by immersion in 0.15 M sucrose in 0.1 M phosphate buffer, pH 7.4 (for 24 hrs), 0.5 M sucrose (for 24 hrs) and 0.8 M sucrose (for 48 hrs) and subsequently frozen at 260uC in liquid pentane and then stored in a 280uC freezer until next use. Cryostat sections of 25 mm cut using a vibratome (Leica VT1000S, Leica, Germany) were used.

Electrophysiology in retinal wholemounts
Caimans were immobilized on ice and dark adapted for 1 hr before being anesthetized as described above. Anesthetized caimans were sacrificed by decapitation followed by removal of eyes. Retinae isolated from eyes were treated with a mix of collagenase/dispase (2 mg/mL) and DNase-1 (0.1 mg/mL) in PBS (pH 7.4) for 30 min at room temperature. After treatment, retinae were washed in PBS and placed in a recording chamber for electrophysiological recording of the vitreal endfeet of Müller cells. The extracellular solution (ECS) to perfuse retinal tissue contained (in mM): NaCl 110, CaCl 2 2, MgCl 2 1, NaH 2 PO 4 1.25, NaHCO 3 25, D-glucose 25, KCl varied from 2.5 to 10 mM (substituted by NaCl to adjust osmolarity); pH 7.4, after aeration by carbogen (95% O 2 +5% CO 2 ). Electrodes were filled with intracellular solution (ICS) containing (in mM): K-gluconate 130, Na-gluconate 10, NaCl 4, HEPES 10, MgATP 4, phosphocreatine 4, NaGTP 0.3, pH adjusted to 7.2 with KOH/HCl. Spermine (0.25 mM; Sigma-Aldrich) was added to ICS in accordance with the finding that free spermine levels in Müller cells are in the submillimolar concentration range [62]. After filling with ICS, the micropipette resistance was ,8 MV. Voltage and current clamp recordings in whole-cell patch-clamp mode were performed using a MultiClamp 700A patch-clamp amplifier with a DigiData 1322A interface (Molecular Devices, Inc., Sunnyvale, CA, USA). The cells were kept at holding potential equal to equilibrium resting potential (to keep membrane current at zero level) and the cells were stimulated by a step to 2150 mV (for 120 ms) with following rising voltage ramp to +150 mV during 80 ms and then a step back to resting voltage. The pClamp 10 software package (Molecular Devices, Inc., CA) was used for data acquisition and analysis.
Note: retinal tissue stored for longer than 1 hr was not used. To study the coupling between Müller cells in retinal wholemounts, the fluorescent dye Lucifer Yellow (LY) 2 mM (Sigma-Aldrich) was added to the ICS as in [63].

Electrophysiology in isolated cells
Retinae were cut into pieces (0.5 x 0.5 mm) and rinsed in fresh Ca 2+ -Mg 2+ -free PBS solution (osmolarity was adjusted to 309 mOsm which corresponds to the osmolarity measured for vitreal liquid sampled from caiman eyes), then were transferred into Ca 2+ -Mg 2+ -free PBS containing papain (24 unit/ml) for 30 min, at 37uC. After washing and trituration in PBS and then in Dulbecco's Modified Eagle medium (DMEM), isolated cells were briefly rinsed in DMEM containing 0.001 mg/ml DNase-1 (Sigma-Aldrich, D-4263, from bovine pancreas). The cells were then washed in DNase-free DMEM medium, stored for 10 min on ice for sedimentation; then the supernatant was exchanged for fresh DMEM and the cells stored on ice. This procedure yielded Müller cells with characteristic morphology and preserved fine processes. We investigated two types of isolated cells: single Müller cells (SM) and Müller cells with attached photoreceptors (SMP). Cells were placed in the recording chamber, allowed to settle on the bottom of the chamber mounted on the stage of an inverted microscope (Nikon DIAPHOT 300, Nikon, Japan) with attached three-axis water hydraulic fine micromanipulator (MHW-3, Narishige, Japan) or an upright microscope (BX51WI, Olympus, Japan) with piezoelectric micromanipulators (MX7500 with MC-1000 drive, Siskiyou, Inc., Grants Pass, OR) used for positioning micropipettes during voltage-clamp and current-clamp recordings. The electrophysiological recordings were performed at room temperature. Electrodes for whole cell recording were pulled in four steps (using Sutter P-97 puller, USA) from hard glass (GC-150-10 glass tubing, Clark Electromedical Instruments, England). Electrodes were filled with intracellular solution (ICS) containing (in mM): KCl 130, MgCl 2 1, CaCl 2 1, EGTA 10, HEPES 10, Na 2 ATP 3, spermine HCl 0.25, pH adjusted to 7.2 with KOH/ HCl. They had resistances of 4-6 MOhm; after cell penetration, the access resistance was 10-15 MOhm, compensated by at least 75%. The extracellular solution (ECS) contained (in mM): NaCl 145, CaCl 2 2.5, MgCl 2 2, and HEPES 10; KCl varied from 2.5 to 10 mM (substituted by NaCl to adjust osmolarity). For recording from single isolated cells we used an Axopatch-200B amplifier with a CV-203BU headstage and a MultiClamp 700A amplifier with CV-7A headstage. DigiData 1200A and DigiData 1322A interfaces were used respectively for data acquisition and analysis (Axon Instruments, Molecular Devices, USA). High frequencies .
1 kHz were cut off and signals were digitized at 5 kHz. The pCLAMP-9 (Axon Instrument, USA) and pCLAMP-10 (Molecular Devices, USA) software packages were used for data acquisition and analysis.
To isolate conductances mediated by different potassium channels we used established pharmacological procedures: channels of TASK family were blocked by bupivacaine 0.1-1 mM [47,64], whereas Kir channels were blocked by 0.1-0.3 mM of barium [12,65,66]. ATP (3 mM) was used in the ICS to inhibit Kir 6.1/SUR1 (K(ATP)) channels [45]. Spermine (250 mM) was added to the ICS to block the outward component of Kir channels [67,68]. Addition of spermine in the ICS also helped to separate TASK-mediated currents from Kir currents [47]. At the concentrations used, barium has little effect on TASK channel currents [69,70,71]. The patched cells were kept at holding potential equal to equilibrium resting potential and then a step to 2150 mV (for 10 ms) with following rising voltage ramp to + 150 mV (for 16 ms) and back to resting voltage. To ensure the ionic nature of whole cell currents, we performed standard control experiments where KCl was substituted with CsCl in the ICS and in ECS; under these conditions, no currents were recorded (not shown). Inhibitors of Kir and 2P-domain channels, barium chloride and bupivacaine were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO, USA and Taufkirchen, Germany).

Fluorescent tracer diffusion studies
We analyzed the diffusion of the gap-junction channelpermeable dyes LY (see above) and sulforhodamine-B, dialyzed in whole-cell configuration in current clamp (zero current) mode. The patch pipettes were filled with ICS described above and containing 2 mM of LY or sulforhodamine-B. LY (Sigma-Aldrich, MW 457.25), is a green-to yellow-fluorescent negatively charged (-2 charges) dye, while sulforhodamine-B (Molecular Probes, MW 558.66) is an orange-to red-fluorescent non-charged water-soluble (polar) sulfonic acid tracer with strong absorption and good photostability [63,72,73,74,75].

Data Analysis
Data were analyzed using pCLAMP-10 (Molecular Devices, USA), Origin 8 software (OriginLab, Northampton, MA, USA) and are reported as mean 6 standard error of the mean. Significant differences between groups of data were evaluated using Student's t-test.

Glial markers in caiman Mü ller cells
First, we analyzed the presence of the established glial marker proteins, S100b (glial specific calcium binding protein), GS (glutamine synthetase), vimentin, and GFAP (glial fibrillary acidic protein) in caiman retinal Müller cells. In addition, we examined the localization of the alpha subunit of G-protein transducin (Gat1), a G-protein involved in phototransduction, in caiman photoreceptor cells. Immunoreactivity for all four glial proteins was identified in caiman Müller cells, and Gat1 was detected in caiman photoreceptor cells (Figs. 1, 2).
We found a robust expression of GFAP in all Müller cells, in somata, in processes and even in very fine distal branches (Fig. 1A). S100b (Fig. 1B) and vimentin (Fig. 2) were expressed throughout the length of all Müller cells, in all retinal areas. Prominent S100b and vimentin staining was observed in somata, distal processes and endfeet. Similarly, GS immunoreactivity was observed in all Müller cell compartments including somata and distal processes (Fig. 1C). In many vertebrate species, strong GFAP expression is only observed in cases of reactive gliosis whereas normal Müller cells are devoid of GFAP immunolabeling (e.g., [3]). However, exceptions have been described; for instance, in the retina of goldfish [76] and horse [6], the Müller cells are GFAP immunopositive under normal conditions. Thus, the GFAP expression of caiman Müller cells is suggested here to represent a species-specific feature, rather than a pathological event.
It has been previously demonstrated that the water channel AQP4 is mainly localized in the endfeet of Müller cells within the rat retina [3]. Using different primary antibodies, AQP4 staining was found basically at the inner limiting membrane which abuts to the vitreal Müller cell endfeet (figure S1).
Immunoreactivity for Kir6.1 was found in caiman Müller cells (Fig. 2C) whereas Kir6.2 was located to neurons (not shown). This differential localization has also been observed in amphibians and in guinea pig [43,44,45]. However, it has been shown that Kir6.1 expression alone is not sufficient to mediate K + currents; the sulfonylurea receptor subunit SUR1 is necessary as a partner of Kir6.1 to form functional K(ATP) channels in Müller glial cells [42,45], in other glial cells [41] and in non-glial cells [48]. The SUR1 immunolabelling was not detected in caiman Müller cells ( figure S2). Together, these data imply that, unlike in other vertebrate species studied so far, in caiman Müller cells the inwardly rectifying K + channels, Kir6.1 and Kir4.1, are not available to mediate K + currents, and to maintain the membrane potential of the cells, therefore, we studied other K + -channel localization.
Localization of two pore domain K + (2P) channels TASK-1, TASK-3, and TREK-2 To examine the putative presence of other potassium channels necessary to maintain the membrane potential of the cells [70] we used antibodies against three 2P-domain channels (TASK-1, TASK-3, TREK-2). Both TASK-1 and TASK-3 immunolabelling was detected in Müller cells, albeit with different intensity. Immunoreactivity for TASK-1 was strong; it was localized throughout the length of Müller cells, in endfeet, inner stem processes, somata, and distal processes (Fig. 2D); in contrast TASK-3 was rather weakly expressed only in distal processes of Müller cells (unpublished data). This pattern of TASK-1 expression throughout the Müller cells corresponds to what was reported for amphibian [46] and mammalian retinal Müller glia [47]. TREK-2 immunoreactivity was not detectable in caiman Müller cells (not shown). In summary, TASK-1 appears to be the dominant K + channel type in caiman Müller cells. Further electrophysiological recordings were made to prove this assumption.

Electrophysiological recordings from Mü ller cell endfeet in retinal wholemounts
In wholemount preparations of the caiman retina (Fig. 3A), Müller cells were patched at their endfeet at the top of the retina with photoreceptor side down. The average membrane potential (n = 23 cells) was 269.861.3 mV. Membrane potential was highly sensitive to extracellular K + , with an average depolarization of 26.562.0 mV (n = 15) in response to [K + ] o changes from 2.5 mM to 10 mM, close to the Nernstian shift for K + (,29 mV) (Fig. 3B), whereas the current-voltage curves in 2.5 mM as well as 10 mM [K + ] o showed linear properties.
Müller cells in the wholemount preparations can retain intercellular coupling thus affecting the space clamp [81]. To access the coupling in the wholemount preparation we injected the fluorescent dye, Lucifer Yellow, into the endfeet of the patchclamped cells. After removing the pipette the dye was found also in adjacent Müller cells (Fig. 3C, D), indicating cell-cell coupling. This coupling was not extensive, as in most of the cases only one additional cell was labeled; a similar weak coupling has been described to occur between rabbit retinal Müller cells [82]. When the gap junction blocker carbenoxolone (200 mM) was applied, the dye remained in the injected cell (Fig. 3E) and no dye-coupling was observed. Together, these data show that whole-cell currents cannot be reliably recorded from the cells in the wholemount  condition. To overcome this problem, we studied enzymatically dissociated single cells.

Electrophysiological recordings from isolated caiman Mü ller cells
Enzymatic dissociation yielded either completely or incompletely isolated (Fig. 4A, B, C) Müller cells. In the latter case, small numbers of cones (Fig. 4A) or rods (Fig. 4B) remained attached to a Müller cell. Whole-cell patch clamp recordings were made from completely isolated Müller cells (SM) as well as from Müller cells with adhering photoreceptor cells (SMP). The patch pipette was placed at the soma of the Müller cell if not stated otherwise.
Resting membrane potential in isolated Müller cells was more hyperpolarized than when recorded from endfeet in the retinal wholemounts. The average membrane potential recorded from SM in control ECS (K + 2.5 mM) was 282.5061.93 mV (n = 8) and SMP cells showed 284.2062.03 mV (n = 10). Under control conditions membrane potentials were very similar between the SM and SMPs (Fig. 4). Thus, the measured resting membrane potentials were close to the ideal (Nernstian) K + equilibrium potential, similar as reported previously from fish, amphibian, reptilian and mammalian Müller cells [3,12,16,65,66,83,84]. In addition, we tested cell responses to elevation of extracellular K + from 2.5 mM to 10 mM; SM depolarized from 282.5060.31 mV (control) to 254.7561.11 mV (high K + ; n = 8), (Fig. 4G). Similar depolarizations occurred in SMP, from 284.2062.03 mV (control) to 257.9061.46 mV (n = 10). If the Müller cells were patched at their endfeet, nearly identical values were measured.
Under bupivacaine, SM cells were depolarized from 2 79.8662.27 mV to 257.5064.72 mV (n = 8), whereas SMP cells were depolarized significantly less, from 284.2062.03 mV to 2 71.0063.76 mV (n = 10) (Fig. 4H). This difference in sensitivity to the TASK-1 blocker between SM and SMP suggests that the attached photoreceptors (Fig. 4A, B) may modify the Müller cell electrophysiology via cell-cell coupling. Bupivacaine (1 mM)  reduced the membrane currents evoked by voltage ramps in voltage-clamp mode (Fig. 4D, E). I/V-curves after the bupivacaine block show a minor remnant of bupivacaine-insensitive Kir currents. The addition of Ba 2+ , a K + channel blocker particularly effective for Kir channels, should thus reduce the difference in depolarization between SM and SMP cells. In fact, if Ba 2+ (0.1 mM) was applied together with bupivacaine (1 mM) the cells were further depolarized (Fig. 4F, H); SM cells to 2 30.8865.83 mV (n = 8) and SMP cells to 238.8063.90 mV (n = 10).
Taken together, isolated Müller cells showed a nearly perfect K + -sensitivity and also a sensitivity to blockers, unlike cells patched in retinal wholemounts (cf. Fig. 3); this is ascribed to imperfect space clamp control in the wholemount situation. The electrophysiological data from isolated cells support the immunohistochemical data by revealing bupivacaine-sensitive currents that may be ascribed to functional TASK-1 channels. Noteworthy, however, in cells with adhering photoreceptors the depolarization induced by bupivacaine was smaller than in single Müller cells, suggesting that Kir channels contribute to the membrane potential (Fig. 4H). This could be explained if Müller cells and photoreceptor cells were electrically coupled. Thus, we tested the isolated cell groups (Fig. 4A, B) for a possible dye coupling.

Coupling between cone and Mü ller cell
The electrophysiological and pharmacological data (Fig. 4D-H) suggest that Müller cells can transfer K + from and to adhering photoreceptors, via a coupling between the Müller cell cytoplasm and that of the photoreceptor cells. We, therefore, analyzed the diffusion of the gap-junction channel-permeable dye sulforhodamine-B. The dye was applied intracellularly in whole-cell configuration when holding the cell in the current clamp mode. We detected the dye-coupling in SMPs; however, the spread of the tracer is unidirectional, occurring only in the direction from the cone to the Müller cell (Fig. 5A, B). When we patched the soma of a Müller cell the tracer easily filled the entire Müller cell body, but never spread to the photoreceptors (Fig. 5B). When patching the inner segment of one of the adhering cones, the tracer diffused throughout the photoreceptor and through the entire adjacent Müller cell (Fig. 5A). We also used several other dyes such as Lucifer yellow, Alexa Fluor-488, Alexa Fluor-568, but these compounds were not permeable from one cell to another in any direction (data not shown). This difference may be caused by the fact that Lucifer yellow and both Alexa dyes are negatively charged molecules whereas sulforhodamine-B is a neutral molecule. When a cone inner segment was patched, the tracer spread throughout the entire photoreceptor and the attached Müller cell (Fig. 5A), but not into neighboring photoreceptors (Fig. 5A, insert). Thus, the observed uni-directional tracer diffusion from a cone to a Müller cell but not to other adjacent cones, and not from the Müller cell to a cone represents a unique cell-to-cell communication for relatively large, uncharged molecules. By contrast, small cations such as K + can probably permeate in both directions, as indicated by the electrophysiological data. We assumed that the movement of ions and sulforhodamine-B might be mediated by gap junctional coupling. To clarify this point, we performed an immunohistochemical staining for connexin 43 (Cx43).

Cx43-immunohistochemistry
In a study on rabbit retina, Müller cells were found to express Cx43 immunoreactivity [82]. Antibodies directed to Cx43 revealed robust, punctate labeling mainly in the outer caiman retina (Fig. 6B). Double-labeling of Müller cells by antibodies against glutamine synthetase (GS) and of cone photoreceptors by the lectin peanut agglutinin (PNA) showed that much of the Cx43 immunoreactivity can be ascribed to Müller cells and cone pedicles (Fig. 6).

Unique distribution of glial K + channels in caiman Mü ller cells
Here we provide the first morphological and functional characterization of Müller glial cells of the Spectacled caiman (Caiman crocodilus fuscus). Müller cells from crocodiles have been studied morphologically [85,86,87,88,89] but studies of the physiological characteristics of these cells were not yet performed. We show here that caiman Müller cells resemble Müller cells in some other vertebrate species by expressing classical glial markers, such as S100b, glutamine synthetase, GFAP (Fig. 1), and vimentin (Fig. 2B). Caiman Müller cells display a very negative membrane potential, close to the equilibrium potential for K + , similarly to that found in Müller cells of all other vertebrates studied so far (e.g, [3,4]), and their membrane potential was sensitive to small changes of K + concentration (Fig. 4G). However, we detected a striking peculiarity in the type(s) of K + channels mediating this K + conductance.
We provide several lines of evidence suggesting that unlike in Müller cells of other healthy vertebrates, functional Kir-type K + channels are not predominantly expressed in normal adult caiman Müller cells. First, Kir4.1 and Kir6.2 immunoreactivities were not detectable in caiman Müller cells. Second, the glial K(ATP) channel subunit, Kir6.1, was immunolabeled in the Müller cells but was not accompanied by its functionally essential SUR1 satellite subunit, [41,44,45]. Moreover, in the presence of ATP (in cytoplasm as well as in recording pipettes) the opening probability of Kir6.1 channels is extremely low [45,54]. In addition, the I/Vcurves recorded from enzymatically isolated single Müller cells (Fig. 4D, E) failed to display a Kir-like pattern (i.e., strong inward rectification near K + -equilibrium potential [67]). Therefore, Kirtype K + channels known to occur in Müller cells (and other glial cells) of other vertebrates are not available in caiman Müller cells; this suggests that other types of K + channels are responsible for maintaining the hyperpolarized membrane potential of these cells.
We found that TASK-1 2P channels were immunolocalized in Müller cells but not in other cells of the caiman retina (Fig. 2D). Membrane currents mediated by these channels are characterized by linear (or slightly rectified due to internal sodium block [99]) I/ V-curves typical for TASK channels [64,70,99] that are expressed specifically in Müller glia of different animals [47]. This fits to our observation that the I/V curves in caiman Müller cells were linear near equilibrium potential (Fig. 4D). Furthermore, a TASK blocker bupivacaine [47,64] inhibited outward currents almost completely (from 2.3 nA to ,0.15 nA at +75 mV (Fig. 4D, E,  dotted lines)). This indicates that TASK channels generate most of the K + currents in caiman Müller cells. It cannot be excluded that there are other (hitherto unknown) K + channels which are  maintaining a (depolarized but substantial) membrane potential after Ba 2+ and bupivacaine block (Fig. 4H) but certainly they cannot mediate large K + currents (Fig. 4F). The location of aquaporin-4 (AQP4) in the caiman retina can be sufficient for water transport. AQP4 has been found in the endfeet processes at the (i) ILM and (ii) around blood vessels in the vascular retina [100]. However, the caiman has an avascular retina with no vessels in the retinal parenchyma, and as we demonstrated with two different antibodies, AQP4 is localized only in the endfeet area of the ILM, where TASK1 is located as well (Fig. S1). This opens the questions: Is an AQP4/TASK1 assembly functional for water transport as has been shown for AQP4/Kir4.1 [100,101], or is the water transport due solely to K + -channel function [102]? Regardless, this is a separate avenue of research and does not fulfill the scope of the present study. Taken together, caiman Müller glial cells appear to be unique among the vertebrates studied so far, as their hyperpolarized membrane potential and high K + conductance, both essential for a wealth of glia-neuron interactions in the retina [3,4], rely upon TASK 2P rather than Kir-type K + channels.

Coupling of Mü ller glial cells
Here we show that caiman Müller cells, similar to Müller cells of other species [9,82,103,104], display somewhat limited coupling to their immediate cellular neighbors (Fig. 3C-F) and express the gapjunction protein, connexin 43 (Fig. 6). The Müller cell-Müller cell coupling may also contribute to the 'passive' I/V characteristic of the currents recorded in retinal wholemount preparations (Fig. 3B).
More strikingly, however, we observed a coupling between Müller cells and photoreceptor cells. This coupling appears to allow for a free bi-directional flux of small cations such as K + , as the electrophysiological properties recorded from enzymatically dissociated Müller cells were changed if the dissociation process failed to detach all photoreceptor cells from the Müller cell (Fig. 4). This coupling may be mediated by gap junctions formed by connexin 43 (Fig. 6). An expression of connexin 43 by Müller cells as well as neurons has been demonstrated before [105] but to the best of our knowledge, electrical coupling between Müller cells and photoreceptor cells is a novel finding. At present the functional role can only be speculated upon; however, it is tempting to suggest that spatial buffering of excess K + , one of the most fundamental functions of Müller cells [3] may involve direct K + fluxes between Müller cells and photoreceptor cells in the caiman retina.
We also found evidence for an uni-directional coupling between cone photoreceptors and Müller cells, allowing for the transfer of the non-charged tracer, sulforhodamine-B, from cones to Müller cells but not vice versa (and not between adjacent cones) (Fig. 5). A uni-directional coupling has earlier been described between different types of glial cells [106,107,108], but not between glia and neurons in retina. Now it remains to be clarified how such uni-directional coupling is established. The observed heterologous cone-to-Müller cell coupling appears to be a unique novel finding, particularly as it does not involve homologous cone-to-cone coupling (Fig. 5A). This might be indicative of a hierarchy of signaling between neighboring cells, useful, for instance, for molecular filtering of large size molecules but not of small cations.

Conclusion
This is the first functional study of retinal glial Müller cells from a representative of the crocodiles, the Spectacled caiman. The cells share many properties with Müller cells in other vertebrates but display several particular features, (1) a unique endowment of caiman Müller cells with K + channels, substituting Kir-like K + channels by TASK 2P channels; (2) electrical coupling of Müller glial cells with photoreceptors and (3) unidirectional tracer propagation from cones to Müller cells. It may be speculated that heterologous coupling between Müller cells and photoreceptor cells may allow for a specific cell-to-cell molecular signaling and modification of spatial buffering of K + ions.