Regulation of synaptic transmission at the photoreceptor terminal: a novel role for the cation–chloride co-transporter NKCC1


  • Wen Shen,

    1. Department of Biomedical Science, Charles E Schmidt College of Medicine, Florida Atlantic University, Boca Raton, FL 33431, USA
    Search for more papers by this author
  • Lauren A. Purpura,

    1. Department of Biomedical Science, Charles E Schmidt College of Medicine, Florida Atlantic University, Boca Raton, FL 33431, USA
    Search for more papers by this author
  • Baoqin Li,

    1. Department of Biomedical Science, Charles E Schmidt College of Medicine, Florida Atlantic University, Boca Raton, FL 33431, USA
    Search for more papers by this author
  • Changlong Nan,

    1. Department of Biomedical Science, Charles E Schmidt College of Medicine, Florida Atlantic University, Boca Raton, FL 33431, USA
    Search for more papers by this author
  • Irene J. Chang,

    1. Department of Biomedical Science, Charles E Schmidt College of Medicine, Florida Atlantic University, Boca Raton, FL 33431, USA
    Search for more papers by this author
  • Harris Ripps

    1. Department of Ophthalmology and Visual Sciences, University of Illinois College of Medicine, Chicago, IL 60612, USA
    2. Whitman Investigator, Marine Biological Laboratory, Woods Hole, MA 02543, USA
    Search for more papers by this author

W. Shen: Department of Biomedical Science, Charles E Schmidt College of Medicine, Florida Atlantic University, 777 Glades Road, Boca Raton, FL 33431, USA. Email:

Key points

  • • The Cl uptake cotransporter, Na+–K+–2Cl type 1 (NKCC1), is expressed in the distal synaptic layer of the vertebrate retina, where photoreceptor terminals contact the dendrites of horizontal and bipolar cells.
  • • Light adaptation, dopamine and a D1 receptor agonist increased the expression levels of phosphorylated NKCC1, the active form of the transporter inserted in the cell membrane.
  • • Pharmacological blockage of NKCC1 with bumetanide increased the rod- and cone-mediated excitatory postsynaptic currents in horizontal cells.
  • • Inhibiting NKCC1 increased exocytotic membrane capacitance, intracellular Ca2+ levels, and voltage-dependent Ca2+ channel currents in both rod and cone terminals, all of which are associated with increased transmitter release.
  • • This study describes a new function of NKCC1, specifically the suppression of transmitter release at the photoreceptor terminals, a process that serves to prevent the depletion of glutamate and protect retinal neurons from its putative cytotoxic effects.

Abstract  The Na+–K+–2Cl co-transporter type 1 (NKCC1) is localized primarily throughout the outer plexiform layer (OPL) of the distal retina, a synaptic lamina that is comprised of the axon terminals of photoreceptors and the dendrites of horizontal and bipolar cells. Although known to play a key role in development, signal transmission and the gating of sensory signals in other regions of the retina and in the CNS, the contribution of NKCC1 to synaptic transmission within the OPL is largely unknown. In the present study, we investigated the function of NKCC1 at the photoreceptor–horizontal cell synapse by recording the electrical responses of photoreceptors and horizontal cells before and after blocking the activity of the transporter with bumetanide (BMN). Because NKCC1 co-transports 1 Na+, 1 K+ and 2 Cl, it is electroneutral and its activation had little effect on membrane conductance. However, recordings from postsynaptic horizontal cells revealed that inhibiting NKCC1 with BMN greatly increased glutamate release from both rod and cone terminals. In addition, we found that NKCC1 directly regulates Ca2+-dependent exocytosis at the photoreceptor synapse, raising the possibility that NKCC1 serves to suppress bulk release of glutamate vesicles from photoreceptor terminals in the dark and at light offset. Interestingly, NKCC1 gene and protein expressions were upregulated by light, which we attribute to the light-induced release of dopamine acting on D1-like receptors. In sum, our study reveals a new role for NKCC1 in the regulation of synaptic transmission in photoreceptors.



E Cl

Cl equilibrium potential


sodium-potassium-chloride co-transporter


outer plexiform layer


0.3% Triton X-100 in PBS


polymerase chain reaction


phosphorylated NKCC


quantitative polymerase chain reaction


starburst amacrine cell


Tris-buffered saline with 0.1% Tween




Membrane proteins that serve to transport ions, neurotransmitters and metabolites between extra- and intracellular compartments are essential for establishing the resting potential and response properties of neurons in the retina and throughout the CNS. Recent studies have examined the properties of the two isoforms of the Na+–K+–2Cl co-transporter (NKCC1 and NKCC2), members of a superfamily of electroneutral cation-chloride co-transporters that include K+–Cl (KCC2) and Na+–Cl co-transporters (NCC; Lauf et al. 1987). The two NKCC isoforms exhibit a similar stoichiometry, i.e. a thermodynamically coupled uptake of two chloride ions with one potassium ion and one sodium ion (Geck et al. 1980), and both are inhibited by the ‘loop’ diuretic bumetanide (BMN; Haas & Forbush, 1998).

NKCC1 is widely expressed in both developing and adult neurons. In developing neurons, where NKCC1 expression is high, GABA excites neurons and increases [Ca2+]i to promote neuronal growth and differentiation (Pfeffer et al. 2009; Cherubini et al. 2011). In mature neurons, expression of NKCC1 tends to be reduced, but is still expressed at relatively high levels on sensory neurons within some areas of the CNS (Delpire et al. 1999; Restrepo, 2005; Li et al. 2008; Delpire & Austin, 2010). Both NKCC1 and NKCC2 have been detected in the vertebrate retina (Gavrikov et al. 2006; Zhang et al. 2006; Dmitriev et al. 2007; Li & Shen, 2007; Li et al. 2008), although their cellular distribution and functional roles vary among species. In chick retina, for example, NKCC1 is expressed on the apical dendrites of retinal pigment epithelium cells, and its activity has been linked to the control of ocular growth (Crewther et al. 2008). In contrast, NKCC2 has been localized to the cell bodies and proximal dendrites of rabbit starburst amacrine cells (SACs), where, in concert with the KCC2 transporter, it governs the SAC responses to GABA and, ultimately, the directional selectivity of retinal ganglion cells (Gavrikov et al. 2006).

Interestingly, there is convincing evidence that NKCC1 is preferentially expressed in the distal synaptic layer within photoreceptor terminals and their second-order neurons, i.e. horizontal cells and some of the dendrites of rod-bipolar cells (Zhang et al. 2006; Li & Shen 2007; Li et al. 2008; Shen et al. 2008). However, despite the presence of NKCC1 at these sites, and its contribution to the local ECl in these neurons (Thoreson et al. 2003; Thoreson & Bryson, 2004; Shen, 2005), the effect of NKCC1 on synaptic transmission in the distal retina is largely unknown. In the present study, we have examined the cellular localization and functional role of NKCC1 transporter in the retina of tiger salamander, with particular emphasis on its modulatory effect on glutamate release in photoreceptors. Our results show for the first time that NKCC1 is an essential component in the regulation of neurotransmitter release at the synaptic terminals of photoreceptors.


Ethical approval

All procedures were performed in accordance with the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and approved by the University's Animal Care Committee.

Retinal slice preparation

Larval tiger salamanders (Ambystoma tigrinum), purchased from Kons Scientific (Germantown, WI, USA) and Charles Sullivan (Nashville, TN, USA), were maintained in aquaria with continuous filtration at 13°C under a 12 h dark–light cycle. Briefly, retinas were excised from eyes that were enucleated after the animals had been decapitated and double-pithed. Retinal slices were prepared in a dark room, under infrared illumination, using a dissection microscope equipped with powered Night-Vision scopes (BE Meyer Co., Redmond, WA, USA). The retina was placed in Ringer solution, and mounted on a piece of filter paper (Millipore, Bedford, MA, USA), with the receptor side up. The retina/filter paper was vertically cut into 250 mm sections using a tissue slicer (Stoelting Co., IL, USA). A single retinal slice was then mounted in a recording chamber and superfused with an oxygenated Ringer solution, consisting of (in mm): NaCl, 111; KCl, 2.5; CaCl2, 1.8; MgCl2, 1; Hepes, 5; dextrose, 10; pH 7.7. The recording chamber was placed on an Olympus (Tokyo, Japan) BX51WI microscope equipped with a CCD camera linked to a monitor. The microscope light transmitted through the condenser was filtered with an 850 nm infrared filter.

Whole-cell patch-clamp recording

Whole-cell recordings were obtained from photoreceptors and horizontal cells in the retinal slice using an EPC-10 amplifier and HEKA Pulse and PatchMaster software (HEKA Instruments Inc., Lambrecht/Pfalz, Germany). Patch electrodes (5–8 MΩ) were pulled with an MF-97 microelectrode puller (Sutter Instruments, Novato, CA, USA). Photic stimuli (3 s) were delivered from a red (660 nm peak emission) LED source focused in the plane of the retinal slice, and controlled by the output of the HEKA amplifier. For horizontal cell recordings, inhibitory inputs from glycine and GABA were blocked with a solution consisting of 2 μm strychnine, 100 μm picrotoxin and 10 μm NO-711. The electrodes were filled with an intracellular solution containing (in mm): potassium gluconate, 100; KCl, 10; MgCl2, 1; EGTA, 2.5; Hepes, 10; together with an ‘ATP regenerating cocktail’ consisting of ATP, 20; phosphocreatine, 40; creatine phosphatase, 2. For Ca2+ channel current recordings from rods and cones, the internal solution included (in mm): CsF, 90; tetraethylammonium (TEA), 10; MgCl2, 2; MgATP, 10; EGTA, either 5 or 1.5; Hepes, 20; adjusted to pH 7.2. The external Ringer solution was modified by the addition of CsCl2 (20 mm) and TEA (20 mm), which were substituted for an equal amount of Na+. Niflumic acid (10 μm) was added in order to block large conductance Ca2+-dependent Cl channels (pH 7.7). Capacitance recordings from rods and cones were made with the same internal and external solutions that were used for recordings of Ca2+ channel currents. The phase angle setting was made with Patchmaster software based on the EPC-10 amplifier circuitry, and verified with a model cell. Data were analysed and plotted using Patchmaster and Igor/Excel software (WaveMetrics, Portland, OR, USA).

A gravity-driven perfusion system was used to superfuse the preparation. A small-bore perfusion tube was placed 3 mm from the retinal slice, and was manually controlled for drug delivery during the experiments. BMN was used to block NKCC1 activation in retinal neurons. A stock solution of 100 mm BMN was dissolved in 1 m NaOH, and used at a final concentration of 50 μm, i.e. a 2000 times dilution from stock. All of the chemicals used in this study were purchased from Sigma (St Louis, MO, USA), & Tocris (Minneapolis, MN, USA).


Freshly enucleated eyes were fixed in 4% paraformaldehyde in Ringer solution for 30 min, and then rinsed extensively in Ringer solution. The eyecups were dehydrated and embedded in OCT compound (Ted Pella, Redding, CA, USA), frozen and sectioned at 10–14 μm. Frozen sections were collected on glass slides, air dried and stored at −80°C for less than a month. The frozen sections were rinsed with 0.1% Tween and 0.3% Triton X-100 in PBS (PBST-T), and treated with a blocking solution consisting of 10% normal goat or donkey serum and 0.1% sodium azide in PBST-T. The frozen sections were pretreated with 1% SDS for 1 min before applying the primary antibody. The primary antibody was prepared in 3% goat serum and 0.1% sodium azide in PBST-T, and the sections were incubated in this solution for 2 h at room temperature, after which they were washed with 0.1% sodium azide in PBST-T. They were then immersed in the secondary antibody, an Alexa 488-conjugated goat-rabbit IgG (Jackson Immunoresearch, West Grove, PA, USA), at a concentration of 1:600 for 30–45 min in darkness at room temperature. After rinsing in PBST (without Triton-X), the retinal sections were mounted on slides with Vectorshield (Vector Laboratories, Burlingame, CA, USA), and viewed on a confocal laser scanning microscope (Zeiss, LMS 700, Munich, Germany). Confocal images were acquired with a 40× oil-immersion objective, and processed with Zen software.

The phospho-antibody R5 (a gift from Dr Biff Forbush, Yale University School of Medicine) was used as the primary antibody at a dilution of 1:200 to detect phosphorylated (p)NKCC1. The antibody was raised in rabbit against a 16-amino acid peptide, Tyrh208-Argh223-Lys, in which Thr212 and Thr217 (in the human NKCC1 sequence) were incorporated directly during synthesis (Flemmer et al. 2002).

Western blot analysis

Retinal tissues were lysed in a 2× Laemmli buffer, and the total protein content was obtained from homogenates in the same buffer solution. The solution was heated for 10 min at 90°C, centrifuged at 4,000g for 10 min at 4°C, and the supernatant was collected. Protein concentrations were calibrated using a BCA™ Protein Assay Kit (Pierce, Rockford, IL, USA). Equal amounts of samples were loaded onto 3–8% Tris-acetate gels, and retinal proteins were separated by electrophoresis at 100 V for 1.5 h. The proteins were transferred to a Hybond-ECL Nitrocellulose membrane (GE Healthcare Bio-Sciences Corp., Piscataway, NJ, USA), which was then immersed in a block solution with 5% dry milk in a Tris-buffered saline with 0.1% Tween (TBS-T) buffer for 1.5 h at room temperature. pNKCC1 was detected with the R5 antibody used at a concentration of 1:5000 in 5% milk solution and incubated overnight at 4°C. The β-actin antibody was used as the loading control (1:200 dilution, overnight incubation at 4°C). The membranes were then washed in a TBS-T buffer and incubated for 45 min with a secondary antibody (HRP-conjugated goat anti-rabbit IgG; 1:5000 or infrared fluorescence IRDye 800CW-conjugated goat anti-rabbit). After further washes in TBS-T, positively stained bands were detected by a chemiluminescent blot assay with the ECL Plus Western blot reagent (GE Healthcare Bio-Sciences Corp) or an Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE, USA).

Real-time quantitative polymerase chain reaction (PCR)

RNA was extracted from salamander retinas collected from the animals after either 4 h dark or 4 h light adaptation. It was purified using the TRI Reagent solution (Sigma), and quantified by spectrometry. cDNA synthesis was performed using the ThermoScript Reverse Transcriptase (15U) and random hexamer primer (50 ng), dNTP (1 mm), dithiothreitol (5 mm) and RNaseOUT (40U; Invitrogen, Grand Island, NY, USA). Real-time PCR was then performed using LightCycler SYBR Green I master QPCR kit (Roche, Branchburg, NJ, USA) on the Stratagene Mx3005P QPCR system. The reactions were performed in a 96-well quantitative (Q)PCR plate with 20 μl of reaction buffer, containing cDNA (2 μl), forward primer (0.2 μm), reverse primer (0.2 μm), 1× LightCycler SYBR Green Master Mix and 2 mm MgCl2.

The specific primers for NKCC1 were designed as sense: 5′-CGG TGA AGT TTG CGT GGA T-3′ and anti-sense: 5′-CCA CCT CCT CTT ACA AAC CC-3′. An internal control was provided by the housekeeper gene β-actin, for which the sense primers were 5′-GTT GCT CCA GAA GAA CAT CC-3′ and the anti-sense primers were 5′-ACC TTC ATA GAT GGG CAC TG-3′. The expected length of the amplification product was 208 bp and 211 bp for NKCC1 and β-actin, respectively. The sequences of NKCC1 and actin are salamander specific, based on the information of our cloned sequences of salamander NKCC1 (unpublished data) and the salamander database website: Sal-Site. An initial activation for 10 min at 95°C was followed by 45 cycles of reactions (each cycle consisting of 20 s at 95°C, 20 s at 50°C, and 20 s at 72°C). Analysing the dissociation curves and observing the sizes of the PCR products on agarose gels confirmed the amplicons. The relative amounts of specifically amplified gene were calculated using the comparative Ct method, i.e. the number of cycles to reach the detection threshold (Livak & Schmittgen, 2001). Briefly, ΔCt values were calculated by subtracting the average NKCC1 Ct value from the average β-actin Ct value for dark- or light-adapted retinas, and the relative amount of NKCC1 mRNA was determined by raising 2 to the power of the negative ΔΔCt. The experiment was repeated three times with RNA isolated from four different salamanders (two light-adapted and two dark-adapted retinas).

Ca2+ imaging

Fluo-4AM (10 μg) was dissolved in 20 μl DMSO and 2 μl Pluronic F-127 for a final concentration of 5 μm. During a 45 min incubation period, Fluo-4AM was internalized and the acetoxymethyl (AM) esters were cleaved (hydrolysed) to form the free Fluo-4. After washing off the external Fluo-4AM solution, it was replaced with normal Ringer solution for calcium imaging. Fluo-4 is a highly sensitive Ca2+ indicator that is retained in the cytosol by high-affinity binding (Kd = 345 nm) to intracellular free Ca2+ ([Ca2+]i). The retinal slices were removed from the loading medium and placed in a recording chamber containing oxygenated Ringer solution, and Ca2+ imaging began after a 20 min recovery period. The Fluo-4 indicator was excited by a 480 nm light, and the emission signals passed through a dichroic filter cube and 520 nm filter. A Rolera-MGi Plus camera (Q-imaging) was used to collect the fluorescent signals. Frame images were taken every 5 s with an exposure time of 50 ms. A Lambda 10-2 unit (Sutter Instrument Co.), controlled by IP Lab 4.0 software, was used to open and close the filter shutter at each interval. Ca2+ signal changes in retinal slices were detected at room temperature in complete darkness.


Localization of pNKCC1 in salamander retina

The NKCC1 transporter is found in the cytosol and plasma membranes of cells both in its inactive and active (phosphorylated) forms. pNKCC1, when present in the plasma membrane, is responsible for Cl and cation uptake (Flemmer et al. 2002; Gamba, 2005). In previous studies, we showed that the R5 antibody is specific for pNKCC1, enabling us to detect and label pNKCC1 in amphibian, goldfish and mouse retinas by Western blotting and immunolabelling (Li & Shen, 2007; Li et al. 2008; Shen et al. 2008), and Fig. 1A shows an example of the Western results for the R5 antibody against pNKCC1 proteins in the samples from mouse brain, mouse retina and tiger salamander retina. A single protein band at molecular weight ∼160 kDa is present in each sample. These bands match the predicted molecular weight of pNKCC1, indicating that the R5 antibody specifically recognizes the protein in the murine brain and retina, as well as in the amphibian retina.

Figure 1.

Expression and localization of phosphorylated (p)NKCC1 
A, Western blots show that the phosphor-antibody specifically recognizes the 160 kDa pNKCC1 protein from salamander retina, as well as from mouse brain and retina. B, a confocal image of an anti-pNKCC1-labelled retinal section shows that the transporter is localized almost exclusively in the OPL. The adjacent light micrograph shows the laminar organization of the retina: ONL, outer nuclear layer of photoreceptor nuclei; OPL, outer plexiform (synaptic) layer; INL,  inner nuclear layer of second-order neurons; IPL, inner plexiform (synaptic) layer; GCL, ganglion cell layer.

Immunocytochemical localization of pNKCC1 in tiger salamander retina is shown in Fig. 1B. This confocal image shows that there is specific labelling by the R5 antibody in the outer plexiform layer (OPL), i.e. the region of the distal retina where photoreceptors make synaptic contact with second-order neurons (horizontal and bipolar cells). No pNKCC1 labelling was detected in the inner retina. This labelling pattern is consistent with NKCC1 antibody-labelling results from mammalian retinas (Zhang et al. 2006; Li et al. 2008). The results indicate that the localization of pNKCC1 is conserved in the neural retina of vertebrates. As reported in our previous study, pNKCC1 co-localizes with the synaptic vesicle protein SV2 in salamander photoreceptor terminals (Shen et al. 2008). Thus, the R5 labelling in Fig. 1B probably represents localization of pNKCC1 at both pre- and postsynaptic sites of the OPL, and suggests that this co-transporter might play an important role in signal transmissions between photoreceptors and their second-order neurons.

Photic regulation of NKCC1

In the suprachiasmatic nucleus, a region of the CNS involved in the control of circadian oscillations, NKCC1 protein levels in the dorsal and ventricular neurons vary in accordance with the daily circadian cycle, such that the responses to GABA swiftly shift between excitation and inhibition depending upon the time of day they are recorded; a similar variation was seen for the expression levels of NKCC1 (Choi et al. 2008). A circadian clock also governs the activity of the neural retina in vertebrates (Ribelayga et al. 2002; Witkovsky, 2004; Ruan et al. 2012). However, aside from sensitivity changes induced by the circadian clock, retinal responses are dependent upon the adaptive state of the retina, i.e. whether light- or dark-adapted. Although these conditions are distinctly different from the circadian control of subjective night and day, we considered the possibility that NKCC1 expression in amphibian retina might be affected by exposure to light.

To determine the influence of illumination on NKCC1 gene expression, we performed real-time qPCR on retinal samples collected from animals following 4 h of either dark- or light- adaptation. Figure 2A shows a representative agarose gel picture of the PCR products of 208 bp NKCC1 and 219 bp β-actin (internal control) derived from light- and dark-adapted retinas. The quantitative estimates of the relative abundance of the NKCC1 mRNA were determined with both semi-quantitative PCR and real-time PCR calculated from differences in the number of cycles to reach threshold (Ct). Analysing these values indicated that there was a fivefold greater amount of NKCC1 transcript from light-adapted retinas compared with dark-adapted retinas; this result was replicated four times in different salamanders during light and dark adaptation (Fig. 2B; = 4; < 0.0001). In addition, pNKCC1 protein levels in dark- and light-adapted retinal samples were quantitatively analysed using Western blot assays. The results (Fig. 2C) show that the amount of pNKCC1 protein was almost 1.5× higher in 4 h light-adapted retinas compared with that in 4 h dark-adapted retinas (= 17, < 0.001). These findings indicate that NKCC1 gene expression and protein levels of pNKCC1 are significantly enhanced by light. Moreover, the photic enhancement of pNKCC1 levels in the OPL was also seen in the antibody-labelled retinal sections from 4 h dark- and light-adapted animals (Fig. 2D).

Figure 2.

Quantitative analysis of NKCC1 gene and protein expressions in retinas from dark- and light-adapted animals 
A, agarose gels comparing transcript levels of NKCC1 gene expression in retinal samples collected from salamanders after 4 h of dark or light adaptation; β-actin was used as the internal control. B, bar graphs compare the real-time polymerase chain reaction (PCR) results of NKCC1 mRNA levels in dark- and light-adapted retinas (= 4). The two asterisks indicate that the differences are significant (**< 0.0001). C, Western blots of phosphorylated (p)NKCC1 protein levels in light- and dark-adapted retinas (= 17); the bar graphs indicate that the differences are significant (*< 0.001). D, immunoantibody labelling of pNKCC1 in the retinal sections from 4 h dark- and light-adapted animals. E, bar graphs from the Western blotting results of relative increases in pNKCC1 levels induced by dopamine (DA) and the D1 receptor agonist, SKF 38393 (SKF) compared with the pNKCC1 levels in untreated dark-adapted retinas (= 11, *< 0.001). Error bars in all figures represent the SEM. INL, inner nuclear layer; ONL, outer nuclear layer; OPL, outer plexiform layer.

Dopamine and pNKCC1 protein levels in the retina

In order to identify a possible pathway for the light-induced upregulation of NKCC1, we examined the effect of dopamine, a catecholamine whose concentration in the distal retina is increased as a result of light adaptation (cf. Witkovsky, 2004). Dopamine activates receptors that stimulate intracellular cAMP-dependent pathways (Silva et al. 1995), which may be involved in phosphorylation of NKCC1 proteins and thereby increase pNKCC1 (Lebel et al. 2009). A further rationale for testing the notion that dopamine might be involved in the photic regulation of NKCC1 was evidence that D1 dopamine receptors are expressed extensively in the distal retina of every vertebrate species tested, from avian to mammals (Hare & Owen, 1995; Veruki & Wassle, 1996; Nguyen-Legros et al. 1997; Mora-Ferrer et al. 1999).

We used intact retinas excised from animals that had been dark-adapted for 4 h, and exposed the tissues either to dopamine (500 μm) or to SKF 38393 (100 μm), a D1 receptor agonist, to stimulate dopamine receptor-mediated intracellular pathways. After 2 h in the drug solution, the tissues, which were kept in darkness, were placed into a lysis buffer and prepared for Western blot analysis; control retinas were kept in the dark for 2 h without dopamine or SKF 38393 treatment. β-Actin was used as an internal control. The results of this quantitative analysis indicate that, on average, pNKCC1 protein levels were substantially higher in retinas treated with dopamine compared with the control retinas (Fig. 2E). Activating D1 receptors with SKF 38393 also increased pNKCC1 levels in the retina, but to a lesser degree compared with dopamine (= 11, < 0.001). These results suggest that the release of dopamine and activation of D1 receptors may provide one of the mechanisms by which pNKCC1 levels are upregulated in the light-adapted retina.

The effect of BMN on the voltage–current properties of horizontal cells

Using a retinal slice preparation from light-adapted animals, we examined the effects of BMN, a selective inhibitor of NKCC1, on the voltage and current responses of horizontal cells. The presence of functional NKCC1 on horizontal cells probably accounts for earlier reports indicating that the intracellular Cl concentration in horizontal cells is significantly higher than in third-order neurons (Miller & Dacheux, 1983). Because this co-transporter is electrically neutral, i.e. there is no net transfer of charge, its inhibition by BMN should have no direct effect on the membrane conductance of horizontal cells. With inputs from the synaptic network blocked by 100 μm Cd2+ in the bath solution, the voltage-dependent channel activity in horizontal cells was recorded in whole-cell voltage-clamp mode with a high K+-gluconate solution in the recording electrodes. Figure 3A (left panel) shows current responses obtained with and without BMN (50 μm for 7 min) from cells activated by successive depolarizing voltage steps from −120 mV to +60 mV in 20 mV increments. The resultant current–voltage (IV) relationships are shown in the right panel of Fig. 3A and, as expected, the voltage dependency depicted in the averaged IV relationship was not significantly changed after blocking NKCC1 with BMN (= 10).

Figure 3.

Whole-cell recordings reveal the effects of BMN on horizontal cells 
A, in the presence of Cd2+, current recordings from a horizontal cell in a retinal slice, activated by multiple voltage steps from −120 mV to +60 mV are shown at the left. Averaged currents at each voltage step were measured at 5 ms from the end of the step in 10 horizontal cells, and plotted at the right. It is apparent that 50 μm BMN had a negligible effect on the horizontal cell activation curves when Ca2+-dependent network inputs were blocked by Cd2+. B, BMN depolarized the horizontal cell dark membrane potential, and increased the amplitude of the horizontal cell light response to either dim (0.8 lux) or bright (1.6 lux) red stimuli. The brighter stimulus shows more clearly the rapid and sustained components of the response. C, the same light stimuli were used to evoke EPSCs from horizontal cells at a holding potential of −60 mV. BMN increased the horizontal cell dark current by 74 + 24 pA (= 9). Light produced a sustained outward current in horizontal cells that reversed to a small overshoot at light offset (left). BMN increased EPSCs in horizontal cells and greatly increased the offset response (more so with the brighter light), but had little effect on the outward current. Note also that the addition of Cd2+ to the BMN solution (left panel) blocked all EPSCs, including that evoked by the BMN-induced presynaptic release of glutamate.

BMN increases glutamatergic input to horizontal cells

A very different situation was seen when we tested the effects of BMN on signal transmission at the photoreceptor–horizontal cell synapse. Horizontal cells are postsynaptic to photoreceptors, but they respond to GABA and glycine (Feigenspan & Weiler, 2004; Shen, 2005; Shen & Jiang, 2007), as well as to the glutamate released from photoreceptors. In order to isolate the effects of glutamate, the preparation was bathed in Ringer solution containing picrotoxin (100 μm), strychnine (2 μm) and NO-711 (10 μm), a cocktail that effectively blocked ionotropic GABA and glycine receptors, as well as GABA transporters that might increase membrane conductance.

Photic responses were elicited by a red (670 nm) LED that delivered 0.8 lux or 1.6 lux in the plane of the retina, and activated both rods and cones. As illustrated in Fig. 3B (control, left-hand panel), horizontal cells in darkness were slightly depolarized by the steady release of glutamate from photoreceptors. The dim light stimulus (0.8 lux) hyperpolarized the horizontal cells by suppressing glutamate release, and at light offset there was a transient depolarizing overshoot. The offset component is attributable to a rapid burst of glutamate release from cones that have a buildup of vesicles at the synaptic release sites during light exposure; these stored vesicles undergo rapid exocytosis at light offset (Jackman et al. 2009), producing the characteristic ‘off-overshoot’ in horizontal cells. The stronger the light intensity, the more hyperpolarized are cones, and the greater the number of vesicles accumulating in a releasable pool that undergoes exocytosis at light offset (Jackman et al., 2009). Thus, increasing the light intensity to 1.6 lux significantly increased the transient light offset response (right-hand panel of Fig. 3B, black trace).

Inhibiting NKCC1 activity with 50 μm BMN depolarized the dark membrane potential and increased the amplitude of the light response in horizontal cells elicited with either the dim or bright light stimulus (red traces, Fig. 3B). These effects developed slowly, and reached their maxima after 7 min in BMN. Similar findings were obtained from 19 horizontal cells in which the BMN-induced depolarization ranged from 5 mV to 20 mV. Both the dark membrane potential and the light response returned to control levels after removing the BMN (blue traces).

We also tested the effect of BMN on EPSCs in horizontal cells. Using the same internal and external solutions and light stimulation protocol as in Fig. 3B, horizontal cells were voltage-clamped at −60 mV in dark. A constant inward current was seen at the holding potential (control, Fig. 3C), due most likely to a tonic release of glutamate from photoreceptors. The dark currents were suppressed by both dim (0.8 lux) and bright (1.6 lux) light stimuli that hyperpolarized photoreceptors (control traces). At light offset, a larger, transient inward EPSC was evoked with the bright stimulus, which probably was a cone-dominated EPSC (Fig. 3C, black trace in right panel). The transient offset current was smaller and much slower in response to the dim stimulus (Fig. 3C, black trace in left panel); this feature is characteristic of the light response of a rod component that tends to depolarize slowly at light offset (see Fig. 5A).

Figure 5.

The effects of BMN on the photic responses of rod and cone photoreceptors 
A, dark potentials and light responses (intensity = 1.6 lux) recorded from rods and cones (= 12 for both) in gramicidin-perforated patch whole-cell recording, with and without 50 μm BMN show that inhibiting NKCC1 activity had no appreciable effect on the response properties of photoreceptors. B and C, in both rods and cones, BMN increased membrane capacitance (Cm) changes generated by a brief depolarizing voltage pulse (from −60 mV to −10 mV; 100 ms). However, the drug had no effect on Gs in photoreceptors (= 8 for both rods and cones), indicating that BMN caused more vesicles to fuse with the plasma membrane of photoreceptor terminals, leading to an increase of Cm. D and E, histograms show the BMN-induced average increase in capacitance resulting from exocytosis in rods and cones (means ± SEM). Black and red bars refer, respectively, to measurements before and after the addition of BMN.

BMN increased both the tonic dark currents and the transient offset EPSCs in horizontal cells (Fig. 3C), consistent with results of Fig. 3B showing that BMN depolarized horizontal cells. On average, BMN increased horizontal cell dark currents by −74 ± 24 pA (= 9). However, the effect of BMN on the peak offset EPSCs (arrow in Fig. 3C, right-hand panel) was highly variable, ranging from −57 pA to −219 pA (= 9). Most likely, the amplitude of the offset EPSCs in horizontal cells is affected by the number of vesicles that accumulate and are released at photoreceptor terminals. Furthermore, the effects of BMN on both the tonic dark currents and transient offset EPSCs were fully blocked by Cd2+, a voltage-gated Ca2+ channel blocker that inhibits Ca2+-dependent glutamate release from photoreceptors (Fig. 3C, left). These results provide support for our contention that by suppressing NKCC1 activity, BMN enhances glutamate input to horizontal cells.

The functional significance of the transient depolarization of horizontal cells at light offset, i.e. the ‘off-overshoot’, is not fully understood, but is thought to play a role in encoding different levels of contrast in cone vision (Struik et al. 2006; Rowan et al. 2010). We examined the effect of BMN on the ‘off-overshoot’ response in horizontal cells. A steady white background light (1 lux intensity) was presented to effectively saturate the rods, and the offset light responses were evoked by a series of 3 s red light stimuli of increasing intensity (1.2–2.5 lux) superimposed on the background light. In this way we could assess the influence of NKCC1 at the cone–horizontal cell synapse. Figure 4A (left) shows that as the contrast between flash and background increased, the magnitude of the ‘off-overshoot’ increased. A graph of the data from seven horizontal cells illustrates the linear relationship between these parameters (Fig. 4A, right). The effect of 50 μm BMN on the cone-driven off-overshoot is shown in Fig. 4B, in which we recorded the horizontal cell voltage change in response to a red stimulus (1.6 lux) delivered on the background illumination. This stimulus elicited a small ‘off-overshoot’ response that was greatly enhanced by BMN, reaching its maximum amplitude after ∼7 min in the BMN-containing Ringer solution; the effect was almost completely reversed after washing with control Ringer solution. As evident in the statistical plot of data from six horizontal cells, inhibition of NKCC1 activity by BMN caused the ‘off-overshoot’ responses to saturate at relatively low contrast (Fig. 4C) compared with the control in Fig. 4A. We suggest, therefore, that NKCC1, by its ability to modulate transmitter release from photoreceptor terminals, may participate in processes that govern the contrast response range in horizontal cells.

Figure 4.

BMN enhances the cone-mediated glutamate input in horizontal cells 
A, with rods saturated by the ambient background illumination, step-wise increases in the intensity of the light stimulus (1.2–2.5 lux) increased the cone-mediated transient ‘off-overshoot’ responses in horizontal cells (left). Averaged results from 7 horizontal cells show the linear relationship between light intensity and the peak of the ‘off-overshoot’ response (right). Error bars (SEM) for the brightest stimulus are within the data point. B, BMN produced an approximately threefold increase in the cone-mediated ‘off-overshoot’ response in horizontal cells, and the effect could be partially reversed by the Ringer wash. C, graph showing the peak offset responses elicited with increasing light intensities in the presence of BMN. Error bars for the brighter stimuli are within the data points. D, the reversal potential of the offset light-evoked EPSCs in horizontal cells from recordings at various voltages from −80 mV to +40 mV. The reversal potential at +20 mV is an indication that the EPSCs arise from activation of glutamate receptors.

The reversal potential of the BMN-enhanced light offset EPSCs in horizontal cells was determined in voltage-clamp recordings. Light offset currents were evoked by terminating a steady 1.6 lux stimulus, and recording the currents from horizontal cells held at voltages between −80 mV and +40 mV (Fig. 4D). The BMN-enhanced light offset currents were inward at negative voltages, reversed near +20 mV, and became outward at more positive voltages (= 4). This behaviour was similar to that seen in a previous study of glutamate EPSCs in salamander horizontal cells (Yang et al. 1998; Pang et al. 2008), and suggests that the BMN-enhanced light offset EPSCs in horizontal cells resulted from the activation of ionotropic glutamate receptors.

The foregoing results from both current- and voltage-clamp recordings indicate that inhibiting NKCC1 increases the glutamatergic input to horizontal cells, due most likely to an enhanced vesicular release of glutamate from both rods and cones.

The effects of inhibiting the activity of NKCC1 in light responses of photoreceptors

We studied the direct effect of BMN on photoreceptor light responses. In order to maintain the endogenous Cl concentration in photoreceptors during whole-cell recording, gramicidin-perforated patch-clamp recording techniques were performed on both rods and cones. Figure 5A shows the light-evoked voltage responses from dark-adapted rods and cones when exposed to the 1.6 lux red light stimulus. Both rods and cones were hyperpolarized by light onset, display a sustained voltage during the exposure, and re-polarize following light offset. However, the rate of re-polarization following light offset was grossly different. In agreement with previous studies (Wu, 1988), a slow re-polarization was typically seen in the rod light response, whereas a rapid off-overshoot potential was seen from cones. When comparing the dark potentials and light responses with and without BMN, it was apparent that light responses of both rods and cones were not significantly affected by BMN (Fig. 5A), consistent with the notion that NKCC1 is an electroneutral co-transporter that produces no direct change in the membrane conductance of the cells. Thus, the increased glutamate input to horizontal cells induced by BMN is not associated with a concomitant light-evoked voltage change in photoreceptors.

NKCC1 activity suppresses vesicle exocytosis

To obtain direct evidence that BMN increases the vesicular release of glutamate from photoreceptors, we measured the capacitance changes at the receptor terminals. For these recordings, a high Cs+ and low EGTA internal solution was used; the external Ringer solution contained 20 mm Cs2+, 20 mm TEA and 100 μm niflumic acid to effectively block the conductance of K+ channels and Ca2+-activated Cl channels. With these intracellular and extracellular solutions, the ECl value in the cells, derived from the Nernst equation, was about −55 mV. Although this value is below the typical ECl in photoreceptor terminals (Thoreson et al. 2002; Thoreson & Bryson, 2004), it may facilitate NKCC1 uptake of Cl.

Photoreceptors were voltage-clamped at −60 mV, and stimulated with a 1 kHz sinusoidal voltage that delivered a 30 mV peak-to-peak voltage change around the holding potential. HEKA software ‘Lock-in’ mode was used to measure changes in membrane capacitance (Cm), access conductance (Gs) and membrane conductance (Gm) for each sine wave during the recordings. Gs and Gm did not significantly change during the Cm recordings, indicating a stable control. The cell capacitances of dark-adapted photoreceptors ranged from 13.3 ± 15.8 pF in rods (= 8), and from 18.3 ± 25.2 pF in cones (= 8). Vesicle exocytosis was elicited by a 100 ms depolarizing step from −60 mV to −10 mV, and Cm changes were measured 30 ms after the depolarizing step to avoid gating charges. A 30 s interval separated the depolarizing pulses to allow photoreceptors to recover from the previous Ca2+ loading. As shown in the control recordings in Fig. 5B and C, the depolarizing voltage step elicited a 91 ± 18 fF (= 8) increase in the peak of the Cm recording in rods, and a 212 ± 34 fF (= 8) increase in cones, consistent with the range of values previously described in salamander cones in the slice preparation (Rabl et al. 2005). The increase in Cm results from the fusion of vesicles undergoing exocytosis at the photoreceptor terminals.

Assuming that the fusion of a single vesicle at the cone synaptic terminal causes a capacitance increase of about 0.056 fF (Thoreson et al. 2003), then the increase we observed in response to the 100 ms depolarizing step reflects the fusion of about 1625 ± 321 vesicles undergoing exocytosis from a single rod, and 3786 ± 607 vesicles from a single cone. However, after the addition of BMN, the depolarizing pulse elicited increases of 110 ± 11 fF (= 8) and 247 ± 14 fF (= 8) in the peak Cm of rods and cones, respectively. This indicates that after inhibiting NKCC1, approximately 1964 ± 196 vesicles and 4411 ± 250 vesicles underwent exocytosis at each rod and cone terminal, respectively. Thus, on average, 339 ± 125 more vesicles were released from rods, and 625 ± 357 more vesicles were released from cones compared with control. These results provide strong evidence that NKCC activity inhibits vesicle exocytosis in photoreceptor terminals.

NKCC1 negatively controls Ca2+ influx and [Ca2+]i levels in photoreceptor terminals

Because glutamate release depends mainly on the Ca2+ concentration in local regions of the photoreceptor terminals, we used the Ca2+ indicator Fluo-4 and Ca2+ imaging to investigate the effect of BMN on intracellular free Ca2+ ([Ca2+]i) changes in photoreceptor terminals. The [Ca2+]i rise in photoreceptor terminals was quantified by measuring increases in fluorescence intensity. The Fluo-4-loaded retinal slices were excited with 480 nm fluorescent light at a rate of 100 ms on and 5 s off. Figure 6A shows an example of a Ca2+ imaging recording from a retinal slice, in which [Ca2+]i changes in photoreceptor terminals were measured at the areas marked by the dashed line rectangle (see Normaski and fluorescent images). Over a period of 10 min, Ca2+ fluorescence intensity increased with application of 50 μm BMN, and returned to baseline levels when BMN was withdrawn (Fig. 6A, lower panel). The increased Ca2+ fluorescence intensity in response to BMN indicates that the inhibition of NKCC1 in photoreceptor terminals produces an increase in [Ca2+]i in photoreceptor terminals and, conversely, endogenous NKCC1 activity lowers [Ca2+]i in the terminals. It should be noted that the increase in [Ca2+]i in the OPL region in response to BMN occurred in 7 of the 12 retinal slices tested. We assume that the five retinas in which BMN produced no [Ca2+]i, change was due either to damage of the photoreceptor terminals during the preparation of the retinal slice, or perhaps to insufficient loading of the Ca2+ indicator.

Figure 6.

BMN increased [Ca2+]i and voltage-dependent Ca2+ currents in photoreceptor terminals 
A, Normasky and fluorescent images of a retinal slice loaded with Fluo-4 Ca2+ indicator. The rectangle (dashed line) indicates the area in which photoreceptor terminals are located and measurements made of [Ca2+]i changes. The time course of the BMN-induced [Ca2+]i increase is shown in the lower panel. B, a single voltage step depolarized the cone to −10 mV in whole-cell recording. In response to BMN, the voltage-dependent Ca2+ currents recorded initially (marked 1) increased gradually over a period of about 7 min (marked 2); = 9. ONL, outer nuclear layer of photoreceptor nuclei; OPL, outer plexiform (synaptic) layer; INL,  inner nuclear layer of second-order neurons; IPL, inner plexiform (synaptic) layer; GCL, ganglion cell layer.

The regulation by BMN of Ca2+ influx via voltage-dependent Ca2+ channels was studied in both rods and cones in voltage-clamp recording. The experiments were performed with patch pipettes containing a high Cs2+ concentration; the external solution was the same as when recording exocytotic capacitance. Cones were voltage-clamped at −60 mV, and activated by a series of depolarizing steps at 30 s intervals to allow Ca2+ channel activity to recover from the previous depolarization. The peak Ca2+ currents were generated when the photoreceptors were depolarized to −10 mV. Accordingly, we tested the effect of BMN on the amplitude of the peak Ca2+ currents evoked with 25 ms voltage steps to −10 mV. Figure 6B shows that 50 μm BMN gradually increased the Ca2+ current amplitude in a cone over a period of 7 min; similar results were obtained in recordings from nine rods.


The presence of NKCC1 co-transporters at numerous sites within the vertebrate nervous system provides ample justification for extensive study of their functional significance. This family of membrane proteins raises the intracellular ionic concentrations of Cl, K+ and Na+, whereas in NKCC1-null cells, the concentration of these ions is significantly lower (Delpire, 2000; Balena & Woodin, 2008). In the present and in previous studies, we have shown that NKCC1 is expressed almost exclusively in the distal synaptic region (OPL) of larval tiger salamander and adult vertebrate retinas (Li & Shen, 2007; Li et al. 2008), where photoreceptors transmit signals to horizontal and bipolar cells. Moreover, we found that the endogenous active form of the co-transporter (pNKCC1) can inhibit voltage-dependent Ca2+ channels and reduce [Ca2+]i in photoreceptor terminals, resulting in the suppression of glutamate release from both rods and cones. Conversely, we demonstrated that inhibition of active pNKCC1 with BMN increased voltage-dependent Ca2+ channel activity, [Ca2+]i levels and glutamate vesicle exocytosis at the photoreceptor terminals. Because NKCC1 activity is a function of the electrochemical driving force on Na+, K+ and Cl influx, we suggest that the mechanism underlying the NKCC1 regulation of Ca2+ channels may be related to these ionic changes in photoreceptor terminals. Indeed, Blumenstein et al. (2004) reported that increasing the intracellular concentration of Na+ suppresses voltage-dependent Ca2+ channels in both neurons and cell lines. Thus, NKCC1 activity may serve as a negative control of Ca2+ channel activity by the transport of cations and anions, clearly a novel mechanism for modulation of transmitter release.

Using molecular biological techniques, we found that NKCC1 expression was upregulated in response to photic exposure, and that dopamine and the D1 receptor agonist SKF 38393 mimicked the effect of light adaptation by increasing pNKCC1 protein levels. It seems likely that dopamine, via an intracellular pathway, activates cAMP pathways that promote phosphorylation of NKCC1 and/or increase insertion of the transporter into the plasma membranes of photoreceptor terminals.

We speculate that the enhancement of NKCC1 expression in the light-adapted retina may serve also as a negative control of the dopamine-mediated, circadian clock-regulated increase of transmitter release from fish cones that occurs in subjective day (Ribelayga et al. 2002). Although circadian regulation of NKCC1 levels has been reported in the suprachiasmatic nucleus (Choi et al. 2008), the mechanism involved in the regulation of NKCC1 gene expression is largely unknown, and there is as yet no evidence to indicate that a circadian clock regulates NKCC1 gene expression in retina. Other studies suggest that post-translational regulation of NKCC1 seems to be via protein phosphorylation, and results from Xenopus oocytes demonstrate that phosphorylation of serine and threonine residues at the N- or C-terminal domains of the NKCC1 transporter modulates its activity (Flemmer et al. 2002; Gamba, 2005). Although NKCC1 is phosphorylated in response to a number of physiological stimuli, such as changes in cell volume (Hamann et al. 2010) and traumatic injury (Kahle et al. 2008), the observation that pNKCC1 protein levels are subject to changes in ambient illumination is a new finding that extends our previous knowledge about the neural mechanisms of light and dark adaptation.

NKCC co-transporters serve to accumulate [Cl]i and maintain the ECl of neurons positive to their resting potential. In the vertebrate retina, NKCC1 is responsible for maintaining the high ECl values in amphibian rods and cones. Previous studies have reported that ECl is about −20 mV in rods and about −45 mV in cones (Thoreson et al. 2002; Thoreson & Bryson, 2004), indicating that the normal range of intracellular Cl concentration in photoreceptors is between 30 mm and 50 mm. Based on the stoichiometry of the NKCC1 co-transporter, entry of 10 mm Cl into photoreceptor terminals will be accompanied by a 5 mm rise in intracellular free Na+. A high concentration of [Na+]i close to Ca2+-sensitive release sites would promote the loss of [Ca2+]i from photoreceptor terminals (Matlib & McFarland, 1991; Dascal, 2004), and thus reduce transmitter release. This may explain why BMN inhibition of NKCC1 activity increases Ca2+-dependent glutamate release from photoreceptors.

Studies by Thoreson et al. (2000, 2002, 2003) have shown that Cl efflux through Ca2+-dependent Cl channels can suppress voltage-dependent Ca2+ channels in photoreceptors. Our results show that BMN slowly increased voltage-dependent Ca2+ channel currents and [Ca2+]i in photoreceptor terminals. Thus, inhibiting NKCC1 caused a lowering of intracellular Cl levels, a process that appears to be much slower compared with Cl efflux through chloride channels. On this view, BMN inhibition of NKCC1-mediated Cl uptake would gradually reduce ECl in photoreceptor terminals, and eventually halt Cl efflux through Ca2+-dependent chloride channels. Therefore, under normal physiological conditions, intracellular Cl accumulation by NKCC1 might also contribute to a reduction in glutamate release from photoreceptors.

At most synapses, neurotransmitter release is mediated by a Ca2+-triggered exocytosis of synaptic vesicles from presynaptic sites in nerve terminals. The major excitatory neurotransmitter in the CNS and retina is glutamate, which, in addition to signalling postsynaptic receptors, is a potent cytotoxin (Olney, 1982). This creates a unique problem for vertebrate photoreceptors, which, in darkness, continuously release glutamate at their synaptic terminals. As mentioned earlier, light hyperpolarizes the photoreceptors, suppresses the tonic discharge of glutamate and holds glutamate vesicles at their release sites until light offset, at which time there is a burst of vesicular release. It is likely that mechanisms exist for negative control of these active release systems at the photoreceptor synapse in order to regulate the extracellular concentration of glutamate and prevent its accumulation within the synaptic cleft. Processes that serve this purpose have been identified, including feedback by zinc ions that are co-released with glutamate to block voltage-gated calcium channels (Wu et al. 1993; Chappell et al. 2008), and glutamate uptake by Muller cells (Bouvier et al. 1991) and photoreceptor terminals (Rowan et al. 2010). Clearly, another such cytoprotective mechanism may be provided by the activity of NKCC1. We plan to determine whether NKCC1 is effective in this regard from the results of our ongoing studies of the NKCC1 knockout mouse. In conclusion, the findings presented in this study constitute an initial step towards uncovering the function and mechanism of NKCC1 activity in the processing of visual information.


Author contributions

W.S. designed the study and performed the electrophysiology experiments; B.L. contributed immunoantibody labelling data; L.A.P., C.N. and I.C. contributed data from biochemical and molecular studies, H.R. and W.S. analysed data and wrote the paper with inputs from all authors.


This study was supported by grants from the National Science Foundation (NSF, IOS-1021646, W.S.) and the National Eye Institute (NEI, EY 14161, W.S.).