Expression of NCBE in the Mouse Retina
The sodium-driven chloride bicarbonate exchanger NCBE is broadly expressed in the brain, including choroid plexus, cortex, olfactory bulb, cerebellum, brainstem, spinal cord, and the retina 
. We first examined if the retina of NCBE-deficient mice shows gross morphological changes. However, H/E stainings of NCBE WT () and NCBE KO retina () revealed no obvious morphological changes in the NCBE KO retina. Somata sizes of retinal neurons and widths of the plexiform layers appeared normal ().
NCBE is strongly expressed in the mouse retina.
To further characterize NCBE expression in the retina, we stained retina sections of NCBE KO and WT mice with a polyclonal antibody against NCBE 
. NCBE was predominantly present in both plexiform layers of the retina (). Photoreceptors and their terminals were devoid of label (Fig. S1
) whereas some cell membranes in the distal and proximal inner nuclear layer (INL) were stained for NCBE, presumably representing bipolar (, arrows) and amacrine cell somata (, arrowheads). No NCBE staining was found in NCBE KO retina sections, indicating that the NCBE antibody was specific for the mouse retina (). However, as the antibody used in our study is expected to detect all different splice variants, it remains unclear which variants are expressed in the retina.
NCBE is Differentially Expressed in Bipolar Cell Compartments
Double stainings for NCBE and the vesicular glutamate transporter 1 (VGluT1), a marker for bipolar cell terminals 
, showed that all axon terminals of ON and OFF bipolar cells in the IPL were stained for NCBE (not shown). As murine bipolar cells comprise five different OFF and seven different ON bipolar cell types 
, we double-labeled NCBE WT retina sections with NCBE and specific bipolar cell markers to analyze the NCBE expression in individual subtypes.
Type 1 and 2 OFF bipolar cells can be labeled by antibodies against NK3R 
. Double staining with NCBE showed strong NCBE expression on dendrites and axon terminals of NK3R-positive OFF bipolar cells (). OFF type 2 cells can also be stained by antibodies against synaptotagmin II (ZNP-1) 
and double labeling revealed that somata and axon terminals of ZNP-1-positive type 2 cells intensely expressed NCBE (not shown).
NCBE is expressed in bipolar cells and amacrine cells.
OFF bipolar cell types 3A and 3B can be distinguished based on their immunoreactivity to antibodies against the ion channel HCN4 and the protein kinase A regulatory subunit IIβ (PKARIIβ), respectively 
. Double stainings for HCN4 and NCBE showed that NCBE was expressed predominantly in dendrites (, arrow) and axon terminals (, arrowhead) of OFF bipolar cell type 3A. However, NCBE was not expressed in PKARIIβ-labeled dendrites (, arrows) and somata (, arrowheads) of OFF bipolar cell type 3B, confirming existing evidence that type 3A and 3B cells differ in membrane protein composition 
Type 4 OFF bipolar cells are immunoreactive to calsenilin (CSEN), a calcium-binding protein 
. Double-labeling revealed that NCBE was expressed in dendrites (, arrows) but not in somata (, arrowheads) of type 4 OFF bipolar cells. As antibodies against CSEN and PKARIIβ also label amacrine cells, NCBE expression on axon terminals of OFF bipolar cell types 3B and 4 could not be determined with these markers alone. However, VGluT1- and NCBE-labeled retina sections showed full colocalization in layer 2 of the IPL (not shown), in which the terminals of type 3B and 4 OFF bipolar cells stratify, confirming that all OFF bipolar cells expressed NCBE in their terminals.
To analyze NCBE expression on ON bipolar cells, we double-stained the retina with antibodies against NCBE and the G-protein subunit Gαo
, which is a marker for rod and cone ON bipolar cells 
. However, dendrites (, arrows) and somata (, arrowheads) of ON bipolar cells were devoid of label. This was confirmed by double stainings for NCBE and PKCα, a marker for rod bipolar cells 
. Again, no NCBE immunoreactivity was found on dendrites of rod bipolar cells (, arrows).
In contrast, PKCα-labeled axon terminals of rod bipolar cells were intensely labeled with NCBE (, arrows). This was confirmed using antibodies against CaB5, a marker for type 3, 5 and rod bipolar cells 
. CaB5/NCBE double staining also revealed NCBE on axon terminals of type 5 ON bipolar cells (data not shown). Additionally, NCBE localized to type 6 ON bipolar cell axon terminals (, arrows) as shown by double staining with antibodies against ZNP-1 
Taken together, we show that a deletion of NCBE did not alter the morphology of the retina and that NCBE was expressed in OFF bipolar cell dendrites and in ON and OFF bipolar cell axon terminals.
NCBE is Expressed in Starburst Amacrine Cells
Beside the expression of NCBE in bipolar cells, we also found NCBE-labeled cells in the proximal INL (), presumably on amacrine cells. To examine NCBE expression in these interneurons, we tested colocalization of NCBE with specific amacrine cell markers.
Choline acetyltransferase (ChAT) is predominantly expressed by starburst amacrine cells in the mouse retina. These cells are mirror-symmetrically organized with somata in the INL and the ganglion cell layer (GCL) and are involved in direction selectivity of ganglion cells 
. Double labeling revealed that somata of ChAT-labeled starburst cells in the INL and GCL express NCBE (, arrows). Also the dendrites in the cholinergic layers 2 and 4 of the IPL were occasionally labeled for NCBE (, arrowheads).
In the mouse retina, antibodies against calretinin label amacrine and ganglion cells 
. NCBE immunoreactivity was found on calretinin-labeled dendrites in the cholinergic layers 2 and 4 of the IPL (, arrowheads) and on calretinin-labeled somata in the proximal INL (, arrows). These cells may again represent starburst cells as these cells are also calretinin-positive 
. We found no NCBE immunoreactivity on calretinin-labeled ganglion cell somata or on dendrites in layer 3 of the IPL.
To test for NCBE expression in horizontal cells, we used antibodies directed against calbindin, which also label amacrine and ganglion cells 
. Double staining revealed that horizontal cells (, arrow) and ganglion cell somata do not express NCBE, whereas calbindin-labeled amacrine and displaced amacrine cells do (, arrowheads).
NCBE is Coexpressed with KCC2
Because NCBE may function as a pHi
regulator and chloride extruder 
in the retina, we compared the immunoreactivity pattern of NCBE with immunoreactivity pattern of KCC2, another neuronal chloride extruder 
. KCC2 is involved in the generation of an axo-dendritic chloride gradient in retinal ON cone bipolar cells 
and plays a role in the direction selectivity of starburst amacrine cells 
. Maximum () and single scan projections () of double stainings for KCC2 and NCBE showed that both transporters are fully colocalized on the same bipolar cell compartments: OFF bipolar cell dendrites and ON and OFF bipolar cell axon terminals.
The expression of NCBE in the mouse retina is summarized in . Our results revealed that NCBE was differentially expressed in bipolar cell compartments and is also expressed in starburst amacrine cells. NCBE was not expressed in horizontal cells () and photoreceptors (not shown). Whether NCBE was expressed in ganglion cell dendrites could not be determined unequivocally. Interestingly, NCBE was expressed in the same retinal cell compartments as KCC2. Thus, NCBE may play a similar role in maintaining a low [Cl−]i and contribute to direction selectivity in the retina. As NCBE may also regulate pHi in retinal neurons, these findings prompted us to analyze the physiological consequences of NCBE deletion.
Summary of the NCBE expression in the mouse retina.
Electroretinography in NCBE-deficient Mice Indicates ON Bipolar Cell Dysfunction
The functional status of the retina in health and disease may be examined via electroretinography (ERG). Although the ERG is a mass response, several protocols varying stimulus intensity and frequency as well as the light environment (background) allow to obtain detailed insights in the functionality of rod and cone photoreceptors and their downstream neurons as long as transient and not spiking signals are generated 
. Here, we compared ERGs of NCBE KO and WT mice at ages of 12 months (PM12) to avoid potential confounding by the differences in birth weight mentioned below. shows the scotopic single flash ERG responses from dark-adapted NCBE KO (red) and WT (black) mice stimulated with increasing light intensities (−4.0 to 1.5 log cd*s/m2
). We observed no difference in the scotopic single flash ERG a-wave, an initial negative deflection after light stimulation, between NCBE KO and WT mice (quantification , lower part). Since the a-wave, when reaching saturation before the onset of the b-wave, reflects rod photoreceptor function in mice, this finding indicates that lack of NCBE had no measurable effect on the maximal output of rod photoreceptors, which is in agreement with a lack of NCBE expression in photoreceptors of WT mice.
Electroretinography in NCBE-deficient mice indicates ON bipolar cell dysfunction.
Scotopic b-wave amplitudes of NCBE KO mice were smaller than those in NCBE WT mice at rod-specific intensities (below −2.0 log cd*s/m2) and at higher stimulus intensities, at which both rod and cone photoreceptors are activated (, upper part). Also, the b-wave latencies of NCBE KO mice were increased for all intensities under scotopic conditions and the waveforms were prolonged (, inset). Together with the unchanged a-wave (, lower part), this indicates either an impaired synaptic transmission to bipolar cells or a problem in the ON bipolar cells themselves, for instance a reduced excitability.
The slight amplitude reduction plus waveform prolongation was also observed in the scotopic flicker ERGs (0.5 to 30 Hz) at a constant intensity of −2 log cd*s/m2 (). In particular, the waveform prolongation led to a reduced flicker fusion frequency, i.e. NCBE KO flicker amplitudes were only mildly decreased at low stimulus frequencies (0.5 to 3 Hz) compared to WT, but rapidly declined at higher frequencies (5–10 Hz), and were practically zero at frequencies above 10 Hz ().
The cone system responses were assessed with the photopic single flash ERG (). Similar to scotopic conditions, NCBE KO b-wave amplitudes were considerably decreased at high stimulus intensities (1 and 1.5 log cd*s/m2) and b-wave latencies were increased compared to WT mice ().
Finally, we also obtained some information about cone ON and OFF systems via scotopic (dark-adapted) flicker ERGs (0.5 to 30 Hz) at a constant intensity of 0.5 log cd*s/m2 (). In this paradigm, the responses are dominated by the rod system below about 3–5 Hz, by the cone ON system from about 5–15 Hz, and by the cone OFF system at higher frequencies. We found small, but consistent amplitude reductions in the frequency ranges attributed to the rod and cone ON system, respectively, but not in the range attributed to the cone OFF system (). This is surprising given the strong NCBE expression in OFF bipolar cells.
In addition to ERG recordings of PM12 mice, we also recorded ERGs of NCBE KO and WT mice at the age of 4 weeks (PW4). The results (Fig. S2
) were similar to that of NCBE KO PM12 mice shown here. Furthermore, we morphologically evaluated the PW4 mice in vivo,
immediately after ERG recordings, and confirmed that the gross organization of the retina was not altered due to a genetic deletion of NCBE (Fig. S3
). Yet, the weight of NCBE KO mice was significantly reduced compared to the WT at PW4 (not shown). Thus, we decided not to use this data here, as the observed differences in ERG amplitudes and latencies may have resulted from a developmental delay in NCBE KO mice in this group. However, as we could reproduce the results in the PM12 mouse group and such adult NCBE KO have a normal weight, lifespan 
and eye size, impairments in PM12 ERGs most likely originate from altered retinal signal transmission rather than developmental defects.
Ganglion Cell Responses of NCBE-deficient Mice Show Temporal Changes
The intense expression of NCBE on bipolar and amacrine cells together with the impaired ERGs indicated that a deletion of NCBE may alter retinal signal processing, especially in the ON pathway. To test this, we recorded extracellular ganglion cell light responses and compared them between genotypes.
First, we compared the light responses of ON-transient ganglion cells. For all ganglion cells measured, different response parameters were extracted from peri-stimulus time histograms (PSTHs; see Methods; ): response amplitude (A1), response duration (A1τ2), and time to peak (L1). shows single PSTHs of representative ON-transient ganglion cells from NCBE KO (red) and WT mice (black) responding to stimuli with a constant light intensity (6.9 cd*s/m2) and increasing spot size (75–1,700 µm; ), a constant spot size (300 µm) and increasing light intensity (−4 to 2 log cd*s/m2; ), and a constant spot size (300 µm) and light intensity (6.9 cd*s/m2) with increasing temporal frequencies (1–15 Hz; ), respectively.
Light responses of ON-transient ganglion cells are longer and faster in NCBE-deficient mice.
At a constant light intensity (6.9 cd*s/m2
), response duration of NCBE KO ON-transient ganglion cells was increased for spot sizes above 175 µm (A1τ2
; , n
30) and differed significantly from the NCBE WT (n
31; for ANOVA values, please refer to the legend of ), pointing to more sustained responses in KO cells. The time to peak of NCBE KO ON-transient ganglion cells was significantly decreased for stimuli with increasing spot sizes (L1
; ). Also, in the NCBE KO, response amplitudes were larger for stimuli >300 µm (A1
, ). Thus, spatial tuning is altered in NCBE KO mice. However, the average spot size eliciting the maximum firing rate was 300 µm for both genotypes and was in line with the suggested receptive field center size of ganglion cells 
This spot size (300 µm) was used for stimulations with increasing light intensity. Under these conditions, response amplitudes were not significantly different (not shown) but NCBE KO ON-transient ganglion cells responded significantly longer (A1τ2; ) and faster (L1, ). As the impaired ERG b-wave rather suggested a slowing of responses, we stimulated ganglion cells with a flicker series (1–15 Hz; ). The PSTH parameter A1 was taken as a measure for the cells’ response. However, response amplitudes did not differ between NCBE WT and KO mice (). These data suggest that ON-transient ganglion cells respond stronger, longer and faster in NCBE-deficient mice despite the profound changes found in ON bipolar cell responses by ERGs.
Next, we compared the light responses of ON-OFF ganglion cells in both genotypes. Single PSTHs of representative cells are shown in . ON response components were characterized by the same PSTH parameters as described before. To characterize the OFF response component, response duration (A2τ2), time to peak (L2) and response amplitude (A2) were extracted from PSTHs.
Light responses of ON-OFF ganglion cells are slower in NCBE-deficient mice.
and OFF (A2τ2)
response durations were not significantly changed between NCBE KO (n
15) and WT ON-OFF ganglion cells (n
16) when the spot size was increased (; for ANOVA values, please refer to the legend of ). Interestingly, while response amplitudes did not differ between genotypes (A1
; ), the ON response component was slower in NCBE KO mice whereas the OFF response components was not (L1
Again, we used a spot size of 300 µm to stimulate ganglion cell receptive field centers and varied light intensity. Under these conditions, ON and OFF response durations of NCBE KO ON-OFF ganglion cells were significantly increased (A1τ2, A2τ2;
). As NCBE KO cells showed slower ON responses to light stimuli (L1; , top), while OFF responses did not differ from NCBE WT (L2, , bottom), we also analyzed the responses of NCBE KO ON-OFF ganglion cells to flicker stimuli () and used the maximal response amplitude (A1) as a measure. NCBE KO ON-OFF ganglion cells were able to follow temporal frequencies <5 Hz comparably to WT cells. However, ON-OFF ganglion cells from NCBE KO showed decreased responses to temporal frequencies between 5–10 Hz () and did not respond to stimuli above 10 Hz while WT cells followed flicker stimuli up to 15 Hz (). These results are consistent with the ERG recordings, which showed similar frequency differences between genotypes ().
In summary, PSTH analyses of ganglion cell light responses revealed that NCBE KO ON-OFF ganglion cells were slower than WT cells. In contrast, NCBE KO ON-transient ganglion cells were faster and showed more sustained responses than NCBE WT ganglion cells.
Both Visual Acuity and Contrast Sensitivity were Impaired in NCBE KO Mice
To test whether the altered ganglion cell responses impair general visual function in NCBE KO mice, we measured both visual acuity and contrast sensitivity using a virtual-reality optomotor system 
. Visual acuity was quantified by increasing the spatial frequency of the moving full contrast gratings until an optokinetic response could no longer be determined. Our experiments revealed a small but significant difference (p
-test) between the visual acuity of NCBE KO and WT mice (): the highest spatial frequencies eliciting an optomotor response were 0.346±0.008 cycles/degree in NCBE KO (n
7) and 0.379±0.017 cycles/degree in NCBE WT mice (n
Impaired visual performance of NCBE-deficient mice.
shows contrast sensitivity values of NCBE KO and WT animals, measured in the optomotor setup and plotted as a function of spatial frequency. Contrast sensitivity of WT mice was clearly superior to the performance of the mutants. Statistical analyses confirmed that genotype had a significant impact on contrast sensitivity (p≤0.01, t
-test): for instance, at a spatial frequency of 0.064 cycles/degree, NCBE KOs had a contrast sensitivity of 8.4±0.59 (corresponding to 11.9% contrast, n
7) while WT littermates had a contrast sensitivity of 11.7±0.40 (or 8.6% contrast, n
5). Contrast sensitivity in NCBE KO mice was reduced by a factor of 1.3 to 1.5 (at the different spatial frequencies) compared to WT littermates indicating that NCBE mice need approximately 1.5-fold higher contrasts in visual stimuli to trigger the optomotor reflex. Contrast sensitivity curves peaked at a spatial frequency of 0.064 cyc/deg for both NCBE KO and WT mice, as previously described for C57BL/6-mice