Modes of Accessing Bicarbonate for the Regulation of Membrane Guanylate Cyclase (ROS-GC) in Retinal Rods and Cones

Abstract The membrane guanylate cyclase, ROS-GC, that synthesizes cyclic GMP for use as a second messenger for visual transduction in retinal rods and cones, is stimulated by bicarbonate. Bicarbonate acts directly on ROS-GC1, because it enhanced the enzymatic activity of a purified, recombinant fragment of bovine ROS-GC1 consisting solely of the core catalytic domain. Moreover, recombinant ROS-GC1 proved to be a true sensor of bicarbonate, rather than a sensor for CO2. Access to bicarbonate differed in rods and cones of larval salamander, Ambystoma tigrinum, of unknown sex. In rods, bicarbonate entered at the synapse and diffused to the outer segment, where it was removed by Cl--dependent exchange. In contrast, cones generated bicarbonate internally from endogenous CO2 or from exogenous CO2 that was present in extracellular solutions of bicarbonate. Bicarbonate production from both sources of CO2 was blocked by the carbonic anhydrase inhibitor, acetazolamide. Carbonic anhydrase II expression was verified immunohistochemically in cones but not in rods. In addition, cones acquired bicarbonate at their outer segments as well as at their inner segments. The multiple pathways for access in cones may support greater uptake of bicarbonate than in rods and buffer changes in its intracellular concentration.


Introduction
Bicarbonate is ubiquitous in the body; it is essential for pH regulation, and it provides a means for the disposal of CO 2 , a metabolic waste product. A third, emerging role is to accelerate the synthesis of cyclic nucleotides by soluble adenylate cyclase and by at least some membrane guanylate cyclases (for review, see Steegborn, 2014;Tresguerres et al., 2014;Sharma et al., 2016). ROS-GCs in the outer segments of retinal rods and cones are stimulated by bicarbonate, and given their critical role in phototransduction, bicarbonate must affect vision.
In darkness, relatively high levels of cGMP maintain cyclic nucleotide gated (CNG) ion channels in an open configuration. Influx of predominantly Na ϩ but also some Ca 2ϩ through the CNG channels depolarizes the membrane to support a steady release of neurotransmitter at the synapse. The entry of Na ϩ into the outer segment is matched by the exit of K ϩ through voltage-gated channels in the inner segment. This flow of ions, known as the circulating current or the "dark" current, is blocked when photon absorption by visual pigment in the outer segment activates a biochemical cascade that leads to cGMP hydrolysis and closure of CNG channels. The resultant hyperpolarization reduces neurotransmitter release at the synapse. Response recovery depends on cascade shutoff, but ultimately requires the restoration of cGMP and the reopening of CNG channels. The basal rate of cGMP synthesis in darkness is relatively low, but during the course of the photocurrent response, ROS-GC is subject to Ca 2ϩ -dependent regulation. The resting [Ca 2ϩ ] i in darkness is set by an equilibrium between entry of the ion through the CNG channels and its removal by Na ϩ / K ϩ ,Ca 2ϩ exchange. CNG channel closure in response to light upsets the balance allowing continued activity of the Na ϩ /K ϩ , Ca 2ϩ exchanger to lower [Ca 2ϩ ] i . GCAPs, neuronal calcium sensing subunits of the ROS-GC complex, stimulate cGMP synthesis by ROS-GCs at low [Ca 2ϩ ] and by doing so, restrict the size and accelerate the recovery of the electrical response to photons (for review, see Wen et al., 2014).
Bicarbonate is abundant in the retina, generated from CO 2 by carbonic anhydrases as a by-product of the extremely high metabolic activity sustained by rods and cones, particularly in darkness (for review, see Hurley et al., 2015;Narayan et al., 2017). Bicarbonate accentuates the role of ROS-GC in phototransduction. By stimulating the synthesis of cGMP by ROS-GC in darkness, it opens a greater fraction of CNG channels to confer a larger saturated response to light (Hare and Owen, 1998;Donner et al., 1990;Koskelainen et al., 1993;Duda et al., 2015). After photon absorption, bicarbonate amplifies the Ca 2ϩ -mediated, negative feedback control over the photoresponse provided by ROS-GC , affecting response kinetics and sensitivity (Lamb et al., 1981;Baylor et al., 1984;. The relatively low affinity of ROS-GCs for bicarbonate, with EC 50 s in the tens of mM range , raises the question: could gaseous CO 2 , which is always present with bicarbonate in aqueous solutions, actually be the more physiologically relevant modulator of ROS-GCs? If bicarbonate is the modulator, how does it access the ROS-GC inside the outer segments of rods and cones? The two questions are related, because CO 2 can freely pass through membranes but as a charged molecule, bicarbonate cannot. At least some cones but not rods express carbonic anhydrase (Musser and Rosen, 1973a, b;Parthe, 1981;Nork et al., 1990) to interconvert CO 2 and bicarbonate either to control [CO 2 ] i if CO 2 is the modulator or to form bicarbonate if it is the modulator. Immunohistological localization at the outer plexiform layer of sodium dependent, bicarbonate transporters, NBCn1 (Bok et al., 2003;Boedtkjer et al., 2008) and possibly NBCe2 (Kao et al., 2011;but see Hilgen et al., 2012), provide for additional bicarbonate uptake at the synapses of photoreceptors, if bicarbonate is the modulator. Bicarbonate might also enter and exit through gap junctions or through Clchannels (Qu and Hartzell, 2000). Bicarbonate is cleared from photoreceptors by Cl --dependent exchange (Koskelainen et al., 1993(Koskelainen et al., , 1994Kalamkarov et al., 1996). In the present study, CO 2 was tested as a modulator for ROS-GC in a reconstituted system and in salamander rods and cones. We further localized spatially the points of entry and egress for bicarbonate in salamander rods and cones and characterized their expression of carbonic anhydrase.

Animals
Larval tiger salamanders (Ambystoma tigrinum, Charles Sullivan and Wadelco), ϳ6 -10 inches in length, were kept at 12°C and fed earthworms every 10 -14 d. Sex of the salamanders was not determined. All care and use conformed to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and to a protocol approved by the Institutional Animal Care and Use Committee. For physiologic experiments, retinas from salamanders that were dark adapted overnight were isolated under infrared illumination and stored in a MOPS-buffered Ringer's solution on ice. In the initial experiments, the MOPSbuffered Ringer's contained: 58 mM NaCl, 2.5 mM KCl, 1 mM MgCl 2 , 1.5 mM CaCl 2 , 5 mM HEPES, 0.02 mM EDTA, 10 mM glucose, and 55 mM MOPS; pH 7.6. The [Cl -] in these experiments was ϳ65.5 mM. Later, the [MOPS] was reduced in some experiments with a concomitant increase in [Cl -] to ϳ90.5 mM: 83 mM NaCl, 2.5 mM KCl, 1 mM MgCl 2 , 1.5 mM CaCl 2 , 5 mM HEPES, 0.02 mM EDTA, 10 mM glucose, and 30 mM MOPS; pH 7.6.

Electrical recordings
After shredding a piece of retina, photocurrent responses to flashes were recorded from individual rods and cones (or from pairs of cones in a few special cases, see below) using the suction electrode technique with either outer segment (OS-in) or inner segment (IS-in) inside the pipette (Baylor et al., 1979;Makino et al., 1999). Except where expressly specified otherwise, all results are from green-sensitive rods or from red-sensitive cones. The pipette was filled with one of the MOPS-buffered Ringer's or with a corresponding low [Cl -] version in which NaCl and KCl were replaced with methanesulfonate salts (Sigma); pH 7.6. Records were low pass filtered at 30 Hz with an 8-pole Bessel filter (Frequency Devices) and digitized online at 400 Hz (Patchmaster v2x53, Heka). Traces were not adjusted for the delay introduced by low pass filtering. The records shown in the figures were subjected to additional digital filtering at 8 Hz (Igor Pro v7.02, Wavemetrics, Inc.). Recordings were made at room temperature, 19 -22°C.
Bicarbonate buffered Ringer's contained 25 mM or 50 mM HCO 3 in place of an equimolar amount of MOPS; pH 7.6. Although the solutions were kept in covered reservoirs, pH sometimes changed over a time scale of hours, so pH was measured after the recording session. A working range of 7.5-7.8 was deemed acceptable. Bicarbonate, bicarbonate plus carbonic anhydrase inhibitor, or carbonic anhydrase inhibitor alone were gradually introduced to cells by bath perfusion. Circulating current changed over the subsequent 10 -15 min and was typically measured 20 min after switching the perfusate. Two carbonic anhydrase inhibitors were used: acetazolamide (Sigma) and dorzolamide TCI (Tokyo Chemical Industry Company, Ltd.). Reversibility of treatment-induced changes in flash responses was generally expected except in experiments with low [Cl -] in the pipette, so for a given parameter, a ratio was taken of the magnitude during perfusion with bicarbonate divided by the average of the pretreatment and washed magnitudes. Cells for which parameters did not recover to within 20% of the starting value were generally not quantified. Flash response kinetics were determined for responses whose amplitudes were Ͻ0.2 of the maximum.

Biochemical assays of cGMP synthesis
Murine carbonic anhydrase Type II and bovine ROS-GC1 cDNA were cloned into a pBudCE4.1 vector (Thermo Fisher Scientific) at the NotI/XhoI sites of the EF-1a multiple cloning site and SalI/XbaI sites of the CMV multiple cloning site, respectively. The construct was verified by sequencing. COS cells originally purchased from ATCC (catalog #CRL-1651, RRID: CVCL_0224) and now maintained in the laboratory, were transfected with the vector using Lipofectamine (Thermo Fisher Scientific) according to the manufacturer's protocol. Zn 2ϩ at ϳ0.1 g/l was present in the growth medium to enhance carbonic anhydrase activity. Seventy-two hours after transfection, the cells were placed in an incubator with humidified air containing 15% CO 2 for 1 h. Cells were then washed with 50 mM Tris-HCl/10 mM MgCl 2 buffer; pH 7.4, scraped into 0.5 ml of the buffer, gently homogenized and centrifuged at 3000 rpm. The amount of cGMP in the supernatant was determined by radioimmunoassay (Nambi et al., 1982). Cells were immunostained for ROS-GC1 and carbonic anhydrase II to check for co-expression of both proteins in the transfected cells, as described below.
In experiments on a core catalytic domain fragment of bovine ROS-GC1, the coding sequence for the G817-Y965 region (numbering for mature protein according to Goraczniak et al., 1994) was amplified from bovine ROS-GC1 cDNA by PCR and cloned into the BglII/NcoI site of pET30Xa/LIC vector (Novagen) for expression in bacterial cells (BL21 RIL). The ROS-GC1 fragment was purified first on a Ni-NTA column and then on a Superdex 200 size exclusion column by FPLC. The effect of 30 mM bicarbonate on the rate of cGMP synthesis was assessed by radioimmunoassay (Nambi et al., 1982).
For labeling of retina, salamanders were dark adapted for ϳ30 min. After euthanasia, eyecups were isolated in dim ambient lighting and fixed in freshly prepared 4% paraformaldehyde in TBS at 4°C for 1 h, cryoprotected in 30% sucrose at 4°C, mounted in O.C.T. solution (Electron Microscopy Sciences), frozen at -80°C, and cut into 25-m sections using a Hacker-Bright OTF5000 microtome cryostat (Hacker Instruments and Industries Inc.). Sections were washed four times with PBS, incubated in 100% methanol at -20°C for 5 h and washed four to five times with PBS. Washed sections were incubated with rhodamine labeled peanut agglutinin antibody (Vec-tor Laboratories, catalog #RL-1072, RRID: AB_2336642) in TBS containing 1% Triton X-100 for 1 h at room temperature, washed several times with TBST, and incubated overnight with the anti-carbonic anhydrase II mouse antibody (Santa Cruz Biotechnology, Inc.) in TTBS containing 1% bovine serum albumin at 4°C, washed four to six times with TBST, incubated with Alexa Fluor 488conjugated secondary antibody (Jackson ImmunoResearch Laboratories) and washed with TBS. Rhodamine was excited with 550 -555 nm light and emission was detected at 575-580 nm; for Alexa Fluor, excitation and emission wavelengths were 493 and 519 nm, respectively.

Experimental design and statistical analyses
Statistical evaluations of the effect of cell type, recording configuration and bicarbonate concentration on physiologic parameters were made with a Kruskal-Wallis nonparametric test (Stata v11.2, StataCorp LLC) and Dunn's test for post hoc comparisons (Dinno, 2015). A repeated measures linear regression with circulating current as the dependent measure was performed with XTMIXED of Stata to take into account multiple measurements on the same cell, testing separately, the treatments applied to rods and cones; p Յ 0.05 was considered to be significant. Curve fitting to determine relative sensitivity to flashes was conducted using Igor Pro.
Biochemical assays to test for the effect of CO 2 were performed in triplicate and repeated three times. The effect of different conditions on the cGMP accumulation in COS cells, relative to that in COS cells expressing ROS-GC1 in air for each assay, was evaluated by an ANOVA with subsequent Bonferroni post hoc testing (Stata).
Biochemical assays on the core catalytic domain fragment were performed in triplicate and repeated twice. The effect of bicarbonate was assessed by a t test without assuming equal variances (Stata).

Bicarbonate sensing by ROS-GC
Tens of mM bicarbonate are required to increase production of cGMP in biochemical assays of ROS-GC1 catalytic activity , calling into question the identity of its true modulator; is it indeed bicarbonate or rather its precursor, CO 2 ? Perhaps ROS-GC1 responds to a much lower amount of CO 2 that exists in equilibrium with the bicarbonate in aqueous solution. As a test, COS cells were co-transfected with bovine ROS-GC1 and murine carbonic anhydrase II and assayed for cGMP synthetic activity in the presence of CO 2 . Co-expression of the two enzymes was verified immunohistochemically (Fig. 1). A representative experiment, repeated in triplicate, is shown in Figure 2. From four such experiments (ANOVA, F (7,16) ϭ 15.91, p Ͻ 0.00005), cGMP accumulation increased 3.5 Ϯ 0.5-fold (mean Ϯ SEM) in the cells co-expressing ROS-GC1 and carbonic anhydrase II when they were placed in a high CO 2 atmosphere, compared to cells expressing ROS-GC1 alone and tested in air (Bonferroni post hoc test, p Ͻ 0.002). The effect was blocked by carbonic anhydrase inhibitors; in the presence of 80 M acetazolamide, cGMP accumulation was only 30 Ϯ 10% (n ϭ 3) above the cGMP levels in cells incubated in the air only, and with 200 M dorzolamide, it was only 10% (n ϭ 1) higher. In preliminary experiments, a lower, 50 M concentration of acetazolamide was less effective in blocking cGMP accumulation. Cyclic GMP levels in COS cells expressing ROS-GC1 alone increased nonsignificantly by 13 Ϯ 2% (n ϭ 4) in the presence of high CO 2 . There were no significant effects in other control experiments conducted in air with either carbonic anhydrase co-expression or with carbonic anhydrase plus carbonic anhydrase inhibitor; the changes in levels were only ϩ2 Ϯ 1% (n ϭ 4), -1 Ϯ 1% with acetazolamide (n ϭ 3), and ϩ1% with dorzolamide (n ϭ 1), respectively. Since the increases in guanylate cyclase activity in cells coexpressing ROS-GC and carbonic anhydrase on exposure to CO 2 were similar to those observed with membranes of transfected COS cells or with retinal membranes  challenged with bicarbonate, we conclude that ROS-GC was targeted directly by bicarbonate and that gaseous CO 2 was its source.
To localize the interaction between bicarbonate and ROS-GC1, a soluble fragment consisting of the core catalytic domain from G817-Y965 was expressed in bacterial New Research cells and purified. The fragment formed a homodimer in solution, as verified by molecular weight markers during FPLC elution and gel electrophoresis performed without reducing agents. In accordance with its biochemical feature that in the absence of its fusion with the plasma membrane it retains reduced catalytic activity (Ravichandran et al., 2017), the fragment exhibited a low, basal activity of 9.0 Ϯ 0.5 pmol cGMP/min/mg prot (n ϭ 2) that increased to 31.0 Ϯ 1.6 pmol cGMP/min/mg prot (n ϭ 2) when 30 mM bicarbonate was present (two tailed t test with unequal variances, t (1.58) ϭ -13.12, p ϭ 0.0133). These experiments demonstrated direct bicarbonate binding to and stimulation of the core catalytic domain of ROS-GC1, as is the case for ONE-GC (Duda and Sharma, 2010), and ruled out the need for an intermediary protein.

Entry and exit of bicarbonate in rods
Given that bicarbonate is the modulator for ROS-GC and that it does not easily cross membranes, the points of entry for bicarbonate into photoreceptors were determined by mechanically dissociating retinas and making electrical recordings with either the outer segment (OS-in) or inner segment (IS-in) of a rod or cone pulled inside a tight-fitting, glass pipette. With OS-in recording, bicarbonate in the bath was accessible only to the inner segment and vice versa for IS-in recordings. The amplitude of the saturating flash response provided a measure of the circulating current in darkness. Addition of 50 mM bicarbonate to the bath increased the circulating current of four green-sensitive rods attached to large pieces of retina in OS-in recordings by 15 Ϯ 3%, similar to previous reports (Lamb et al., 1981;Duda et al., 2015). Large pieces of retina contained many other rods and cones along with inner retinal neurons and Müller glia. Isolated rods, re-corded with OS-in, do not typically respond to bicarbonate  or bicarbonate plus acetazolamide . To find out whether Müller cells were needed for uptake of bicarbonate into rods, two rods attached to small clumps of one to six other cells were recorded with OS-in. A cone was included in both clumps; intact Müller cells were not present in either clump. Circulating current was enlarged by 50 mM bicarbonate (Fig.  3A), and flash response kinetics were faster, indicating that bicarbonate uptake was preserved in clumps. We conclude that bicarbonate uptake by rods did not depend on contact with or close proximity to Müller cells.
Reasoning that access to bicarbonate must have required the synaptic region or spherule, which was intact for rods in large pieces of retina and in small clumps of cells but was usually broken off during the mechanical dissociation, we sought rare, isolated rods that retained their spherules (Fig. 4A). Although many more rods retain their spherule when dissociated with papain (Bader et al., 1978), we wished to avoid any complications arising from the enzymatic digestion of phototransduction cascade components (Aton and Litman, 1984). With OS-in, four out of four isolated rods with spherule indeed responded reversibly to bicarbonate with a circulating current that increased by 17 Ϯ 3%, an integration time of the dim flash response that decreased by 21 Ϯ 4% (n ϭ 3) indicating faster flash response recovery, and a sensitivity to flashes that was reduced by 52 Ϯ 10% (n ϭ 2; Fig. 3B-D).
In control experiments on 15 rods recorded with OS-in (Fig. 5A) and three rods recorded with IS-in, 2 M to 250 M acetazolamide or 500 M dorzolamide alone did not significantly change circulating current (ϩ1 Ϯ 2%, n ϭ 18), integration time (ϩ1 Ϯ 4%, n ϭ 12), or sensitivity (ϩ7 Ϯ 2%, n ϭ 11). The lack of effect of carbonic anhydrase Figure 2. CO 2 -induced stimulation of cGMP accumulation in COS cells co-expressing bovine ROS-GC1 and murine carbonic anhydrase II. Co-expression of carbonic anhydrase II with or without 80 M acetazolamide had no effect on cGMP accumulation in cells incubated in air for 1 h. But when the assay was conducted in a 15% CO 2 atmosphere, cGMP synthesis was much higher in the cells co-expressing carbonic anhydrase II. Asterisks indicate a difference (p Ͻ 0.0005) from every other condition, based on an ANOVA, F (5,12) ϭ 79.83 (p Ͻ 0.00005) and a Bonferroni post hoc test. CO 2 had no effect on cells lacking carbonic anhydrase II expression and the invigorating effect of CO 2 on cells with carbonic anhydrase II was blocked by 80 M acetazolamide. Error bars demarcate SEM, n ϭ 3.
inhibitor on five of these OS-in rods that were attached to a clump or to large pieces of retina meant that under our experimental conditions, rods did not take up appreciable quantities of bicarbonate produced endogenously by cones or Müller cells or any other retinal cells. In other control experiments, two isolated rods treated with bicarbonate plus acetazolamide at concentrations up to 32 M (Fig. 5B) and one rod treated with bicarbonate plus 100 M dorzolamide exhibited changes in circulating current similar to those obtained for treatment with bicarbonate alone. The results of both control experiments were con-sistent with the absence of an effect of carbonic anhydrase inhibitor on the phototransduction apparatus of rods .
Post-treatment washing of the rod with MOPS buffered Ringer's reversed the effects of bicarbonate (Figs. 3, 5B). An HCO 3 -/Clexchanger in the outer segment couples the entry of Clto the extrusion of internal bicarbonate (Koskelainen et al., 1994;Kalamkarov et al., 1996). Here we show that filling the pipette with a low Cl -Ringer's solution in OS-in recordings to suppress the HCO 3 -/Clexchanger trapped bicarbonate inside the rod and prevented circu- Dim flash responses, whose peak amplitudes were less than a fifth of the maximum, were scaled to their peak amplitudes. Integration time, given as the integrated area of the normalized dim flash response, decreased ϳ16% with bicarbonate. D, Loss in sensitivity to flashes of the rod in B with bicarbonate. Results were fit with a saturating exponential function: r/r max ϭ 1 -exp(-ki) where k is a constant equal to ln(2)/i 0.5 and i 0.5 is the flash strength eliciting a half-maximal response. lating current from subsiding to the pretreatment level after washing (Fig. 6A). Circulating current even increased during the time it took to clear the bath of bicarbonate. However, if a portion of the OS remained outside of the electrode allowing access to the higher Clin the bath during the wash (n ϭ 2) or if the rod OS was partially ejected for washout and then pulled back into the pipette for recording (n ϭ 2), then the effect of bicarbonate on circulating current was fully reversible (Fig. 6B). It was typically difficult to detect changes with 25 mM bicarbonate but with OS-in and low Clin the pipette, this treatment increased circulating current by 8 Ϯ 1% (n ϭ 5), shortened integration time of the dim flash response by 22 Ϯ 5% (n ϭ 4) and reduced sensitivity by 34 Ϯ 12% (n ϭ 3, not significant) with recovery only after exposing the outer segment to high Clin the bath. All rods in these experiments were attached to large pieces or clumps or if isolated, retained a synapse. We conclude that extrusion  by an HCO 3 -/Clexchanger located at the base of the OS, if not along its entire length, was the principal means for removing bicarbonate from the rod.

Access to bicarbonate in cones
Isolated cones or cones in a clump responded to 50 mM bicarbonate with an increase in circulating current in OS-in recordings (Fig. 7A), consistent with previous studies (Koskelainen et al., 1993;Duda et al., 2015). In the Duda et al. (2015) study, all cones were red-sensitive, single cones (Fig. 4B). Here, we extend observations to twin (Fig. 4C,D) and to double (Fig. 4E-G) cones, which are also red-sensitive (Attwell et al., 1984;Isayama et al., 2014) and to UV-sensitive cones (Fig. 4H). Twin cones consist of two members that are symmetric in morphology, whereas the two members in double cones differ in the sizes and shapes of their inner and outer segments (Walls, 1963). Usually only one member of a double or twin cone was recorded but there were a few recordings with both members inside the pipette. While salamander double cones are not electrically coupled (Attwell et al., 1984), the situation for twin cones is not known. For two double and five single red-sensitive cones, current increased 1.32 Ϯ 0.07-fold with 50 mM bicarbonate and for three UV-sensitive cones, it increased 1.18 Ϯ 0.08-fold. On average, the impact of bicarbonate on red-sensitive cones was greater than that on rods (p ϭ 0.032).
Rods lacking a spherule (synaptic zone) and recorded with IS-in did not respond to 50 mM bicarbonate (n ϭ 4), as reported in Duda et al. (2015). However, the circulating current increased in 18 red-sensitive cones with IS-in in the present study, including two double cones (Fig. 7B,C). In three of the cones, bicarbonate was applied a second time after washing with the same result. The relative increase in current brought about by bicarbonate was as large, if not larger, in cones with IS-in (1.45 Ϯ 0.05-fold) than with OS-in (Fig. 7A,B), although the difference did not quite reach significance (p ϭ 0.11). Halving the concentration of bicarbonate from 50 to 25 mM produced a smaller, 1.22 Ϯ 0.08-fold (n ϭ 7, p ϭ 0.015) effect in IS-in recordings (Fig. 7B,C). The comparisons were made with a rank order test (Kruskal-Wallis, 2 ϭ 14.08, 3 df, p ϭ 0.0028) followed by Dunn's post hoc testing. The circu-lating current in one cone increased from 7 pA in MOPS to 12 pA in 50 mM bicarbonate, then diminished to 9 pA in 25 mM bicarbonate and returned to 12 pA in 50 mM bicarbonate, providing further support that the lower concentration of bicarbonate was less effective.
The presence of carbonic anhydrase activity in some vertebrate cones (Musser and Rosen, 1973a, b;Parthe, 1981;Nork et al., 1990) suggests that CO 2 in equilibrium with exogenously applied bicarbonate could cross the plasma membrane and generate bicarbonate internally. To find out whether salamander photoreceptors express carbonic anhydrase, retinas were probed with peanut agglutinin lectin, which binds with high affinity to the saccharides in the cone extracellular matrix (Blanks and Johnson, 1984) and with an antibody raised against carbonic anhydrase II. Carbonic anhydrase II antibody and peanut agglutinin binding co-localized, verifying the carbonic anhydrase II expression in salamander cones (Fig.  8). Labeling was pervasive across the cones, inclusive of double and single cones, but was absent from rods.
Next, the internal conversion of CO 2 to bicarbonate in salamander cone OSs was tested by co-application of bicarbonate with a carbonic anhydrase inhibitor, acetazolamide or dorzolamide. Dorzolamide at concentrations of 100 -500 M failed to block or even diminish the increase in circulating current with bicarbonate in cones recorded with IS-in (n ϭ 5). Acetazolamide at concentrations as low as 1 M did prevent bicarbonate from increasing the circulating current in some cells (Fig. 9A), however, a reduced increase in current persisted in most cells. On analysis with a linear model with mixed effects (Wald 2 ϭ 48.12, 65 observations, p Ͻ 0.0005), the overall effect on circulating current of 50 mM bicarbonate plus acetazolamide at concentrations ranging from 1 to 160 M was an increase of 28 Ϯ 5% (n ϭ 17), less than the above 45% increase seen with bicarbonate alone on cones recorded with IS-in (p ϭ 0.002). In a subset of seven cones recorded IS-in and four cones recorded OS-in that were treated sequentially with bicarbonate and then with bicarbonate plus acetazolamide, increasing the concentration of acetazolamide improved its efficacy in reducing the effect of bicarbonate (Fig. 9B). Internal conversion of CO 2 to bicarbonate therefore contributed to the modulation of circulating current in cones.
In three out of three isolated cones recorded OS-in, treatment with 250 M acetazolamide but no added bicarbonate reduced the circulating current by 23 Ϯ 2% (Fig. 9C). Chloride concentration was 90.5 mM inside the electrode and in the bath for these experiments. Apparently, acetazolamide blocked the ongoing synthesis of bicarbonate from endogenous CO 2 that was otherwise high enough to stimulate ROS-GC activity in the cone outer segment.
There were no cases in which the circulating current during treatment with bicarbonate and carbonic anhydrase inhibitor fell below the pretreatment value. Therefore, some bicarbonate must have gained entry into cone OSs. One possibility may have been via reversed operation of a bicarbonate exchanger. The bicarbonate exchanger in cones imports bicarbonate when the retina is perfused with 6 mM bicarbonate in a low Cl -Ringer's solution to create an inward driving force on bicarbonate and an outward driving force on Cl - (Koskelainen et al., 1993). In our IS-in recordings, cones were treated with 25 mM rather than 50 mM bicarbonate to lower the inward driving force on bicarbonate and [Cl -] in the pipette was lowered to ϳ5 mM to reduce [Cl -] inside the cone. The [Cl -] in the bath remained high to set up an inward driving force on Clat the outer segment. Nevertheless, the bicarbonate-induced increase in circulating current (19 Ϯ 5%, n ϭ 11) was unaffected by the low [Cl -] in the pipette. Experiments conducted in which [Cl -] in the bath was increased from ϳ65.5 to ϳ90.5 mM to make reversed bicarbonate exchange even less favorable did not seem to make a difference, nor did the addition of dorzolamide to the bath. Perhaps 25 mM bicarbonate in the bath was still high enough to overcome the Clgradient and allow a net inward movement of bicarbonate by the exchanger.

Potency of carbonic anhydrase inhibitors
Acetazolamide was used in our early physiologic recordings at concentrations as low as 0.5 M, because 5 M blocked all carbonic anhydrase activity in histologic studies on retina from a number of species (Musser and Rosen, 1973a). The increased effectiveness of acetazolamide over tens of M in reducing the bicarbonateinduced increase in current in cones of the present study and the finding in our cell-based assays for cGMP synthesis, that 50 M acetazolamide failed to fully inhibit CO 2 stimulated, ROS-GC activity were surprising. To investigate further, we determined the dose-response relations to characterize the effectiveness of acetazolamide and dorzolamide in blocking the production of bicarbonate by carbonic anhydrase II expressed in COS cells that also expressed recombinant ROS-GC1. One experiment with each inhibitor is shown in Figure 10. In control experiments, ROS-GC1 activity was unaffected by inhibitors in the absence of CO 2 , with or without co-expression of carbonic anhydrase II (Fig. 2). Under the conditions of our experiments with COS cells, the IC 50 for acetazolamide was 37 Ϯ 12 M, n ϭ 2 experiments. For dorzolamide, IC 50 was 20 Ϯ 3 M, n ϭ 3 experiments.

Bicarbonate directly stimulates ROS-GC
In Caenorhabditis elegans, the membrane guanylate cyclase GY-9 functions as a CO 2 detector (Smith et al., 2013). The hypothesis that ROS-GC1 responds preferentially to CO 2 was ruled out in biochemical assays using a recombinant system. While the possibility that CO 2 had a marginal capacity to bind and stimulate ROS-GC1 cannot be ruled out, any slight increases in cGMP accumulation in ROS-GC1 expressing COS cells evoked by high CO 2 were most probably due to the spontaneous formation of bicarbonate that approached an equilibrium concentration of 4 mM, by the Henderson-Hasselbalch equation.   : 321, 597, 1300, 4530, 17,900, and 33,300 photons m Ϫ2 at 600 nm. Recorded with IS-in, but traces are inverted for ease of comparison with OS-in recordings. B, Dose-response relations for acetazolamide. Difference is given by the relative circulating current with 50 mM bicarbonate minus the relative circulating current with 50 mM bicarbonate plus acetazolamide for 11 cones subjected to both treatments. IS-in, black symbols; OS-in, gray symbols. Continuous line from linear regression has an r 2 value ϭ 0.41, p ϭ 0.033 for the slope. Dashed line shows the result that would be obtained if acetazolamide were to have had no effect. C, Reduced circulating current in an isolated cone with OS-in on treatment with acetazolamide alone. Flash strengths were: 246, 459, 1860, and 25,600 photons m Ϫ2 at 600 nm. Cells co-expressing carbonic anhydrase formed much more cGMP, consistent with Duda et al. (2018), except when a carbonic anhydrase inhibitor was present (Fig. 2) indicating that ROS-GC1 responded primarily to bicarbonate. Unexpectedly, dose-response determinations for COS cells (Fig. 10) yielded IC 50 values for acetazolamide and dorzolamide that were considerably higher than those for cell free preparations (for review, see Sugrue, 1996;Supuran and Scozzafava, 2007).
It may be argued that bicarbonate did not interact with ROS-GC1 directly, but instead bound to an accessory protein that then modulated the cyclase's catalytic activity. Mutagenesis experiments on an olfactory guanylate cyclase, ONE-GC, localized stimulation of enzymatic activity by bicarbonate to its core catalytic domain (Sun et al., 2009;Duda and Sharma, 2010), a segment that is highly conserved across membrane guanylate cyclases. We therefore expressed and purified a soluble fragment of ROS-GC1 consisting of the core catalytic domain and tested it for sensitivity to bicarbonate. The fragment formed a dimer with a relatively low basal rate of cGMP synthetic activity, but activity increased 3.4-fold with 30 mM bicarbonate. We conclude that bicarbonate binds directly to the core catalytic domain of ROS-GC1 to stimulate its enzymatic activity and that an accessory protein is not required.

Points of bicarbonate ingress and egress in rods
By varying the extent of mechanical dissociation of the retina in conjunction with suction electrode recording to restrict bicarbonate access to specific subcellular regions of rods, we localized bicarbonate uptake exclusively to the synapse. Potential pathways for entry at the spherule include electroneutral sodium bicarbonate co-transport by NBC3, also known as NBCn1 (Bok et al., 2003;Boedtkjer et al., 2008), electrogenic sodium bicarbonate cotransport by NBCe2 (Kao et al., 2011;but cf. Hilgen et al., 2012) and permeation through Ca 2ϩ activated Clchannels (Qu and Hartzell, 2000;MacLeish and Nurse, 2007). Ca 2ϩ activated Clchannels are opened by depolarization and by Ca 2ϩ , conditions favored in darkness. The identity of the channel(s) has yet to be clarified because immunolabeling for TMEM16A and TMEM16B yielded conflicting results (Yang et al., 2008;Stöhr et al., 2009;Mercer et al., 2011;Dauner et al., 2013;Jeon et al., 2013;Caputo et al., 2015). Our results exclude entry of bicarbonate into isolated rods through hemi-gap-junction channels in perinuclear regions of the inner segment (Custer, 1973;Zhang and Wu, 2004) under our experimental conditions. Internalized bicarbonate is removed by HCO 3 -/Clexchange at the OS because 4-4'-diisothiocyanatostilbene-2,2'-disulphonic acid (DIDS), an agent that blocks Na ϩ -independent bicarbonate transport, increases the circulating current in rods being exposed to bicarbonate. High [bicarbonate] and low [Cl -] around the OS actually cause rods to import bicarbonate by reversed action of the exchanger (Koskelainen et al., 1994;Duda et al., 2015). In the present study, bicarbonate was trapped inside rods for tens of minutes by lowering [Cl -] around the OS (Fig. 6) indicating that HCO 3 -/Clexchange in the OS was the principal means for clearing bicarbonate from the rod. Egress of bicarbonate through Clchannels at the spherule or through hemi-gap-junction channels may have been very slow and reversed action of the Na ϩdependent transporter may have been stymied by the low [Na ϩ ] i . The restricted location for egress at the outer segment ensures that bicarbonate would distribute throughout the cell and importantly, accumulate in a location where it would impact visual transduction by stimulating ROS-GC.

Additional pathways for accessing bicarbonate in cones
Cones, like rods, express Ca 2ϩ -activated Clchannels (Maricq and Korenbrot, 1988;Barnes and Hille, 1989;Mercer et al., 2011) and presumably express transporters at the synaptic zone. Unlike most rods, mechanically dissociated cones retain the synaptic zone and take up bicarbonate when recorded with OS-in (Fig. 7A). Larger effects were obtained with a lower (6 mM) dose of bicar- bonate in ERG recordings Koskelainen et al., 1993), perhaps because inner and outer segment pathways for uptake were invoked and removal was minimized.
Carbonic anhydrase activity is present in cones of several species (Musser and Rosen, 1973a, b;Parthe, 1981;Nork et al., 1990). The majority of human cones, probably red-and green-sensitive but not blue-sensitive cones, express carbonic anhydrase activity (Wistrand et al., 1986;Nork et al., 1990). Here, we show carbonic anhydrase II immunoreactivity in single and double redsensitive cones of salamander ( Fig. 8) but remain uncertain about the blue-sensitive cones because they comprise such a minor fraction of the photoreceptor population (Sherry et al., 1998;Chen et al., 2008). Conversion of endogenous CO 2 to bicarbonate inside cones was sufficient to increase ROS-GC activity in the outer segment, because circulating current was reduced by the addition of acetazolamide without added bicarbonate (Fig. 9C). While inhibition of carbonic anhydrase with acetazolamide prevented exogenously applied bicarbonate from increasing circulating current in some cones (Fig.  9A), block was incomplete in most cones ( Fig. 9B; see also Koskelainen et al., 1993) even for acetazolamide concentrations as high as 160 M. Photoreceptors are not known to express other carbonic anhydrases. Car- Figure 11. Pathways for bicarbonate trafficking through rods and cones. (1) Bicarbonate enters rods and cones at the synaptic zone through Ca 2ϩ -activated chloride channels and by Na ϩ -dependent transporters. It diffuses through the inner and outer segments and is extruded by an HCO 3 -/Clexchanger at the OS.
(2) In cones, bicarbonate may enter the OS by a reversed action of an HCO 3 exchanger whose operation seems to be less dependent on external Clthan the exchanger in rods.
(3) Alternatively, bicarbonate can be formed inside the cone by carbonic anhydrase when CO 2 crosses the membrane. CO 2 will pass through rods, but the conversion to bicarbonate is slow because rods do not express carbonic anhydrase.
bonic anhydrase XIV, which is included along with carbonic anhydrase II in some photoreceptor proteomes (Liu et al., 2007;Kwok et al., 2008;Skiba et al., 2013), may be a contaminant of Müller cells in which it is highly expressed (Nagelhus et al., 2005;Ochrietor et al., 2005). Regardless, acetazolamide is an effective inhibitor of both isoforms (for review, see Supuran and Scozzafava, 2007). The smaller increase in bicarbonate induced circulating current with acetazolamide in IS-in recordings meant that cones produced bicarbonate internally from CO 2 in equilibrium with bicarbonate in aqueous solution. These results help to account for the reduced photopic a-wave of the mammalian ERG particularly with bright flashes on administration of carbonic anhydrase inhibitor (Broeders et al., 1988;Odom et al., 1994;Sauvé et al., 2006) and indicate that it was not just due to acidosis. Dorzolamide inhibited recombinant murine carbonic anhydrase II in COS cells (Fig. 10B), but for reasons that are not known, it was ineffectual when applied to salamander cones. The ability of cone OSs to access bicarbonate in the presence of carbonic anhydrase inhibitor reveals the existence of yet another pathway for entry. Bathing the retina in low [Cl -] enhances bicarbonate uptake and produces an even larger circulating current in cones (Koskelainen et al., 1993), presumably by suppressing the exchanger or causing it to load bicarbonate. We lowered [Cl -] in the pipette in IS-in recordings to keep [Cl -] i low and lowered [bicarbonate] in the bath by half to reduce the driving forces for Clleaving the OS and for bicarbonate to enter. Somehow, bicarbonate still found its way into most cones to increase circulating current irrespective of whether carbonic anhydrase inhibitor was present. An HCO 3 -/Clexchanger is implicated because DIDS blocks the bicarbonate uptake (Koskelainen et al., 1993). It may be that bicarbonate exchange was less dependent on [Cl -] o in cones than in rods. We therefore propose at least three routes for bicarbonate entry into cone OSs (Fig. 11). The importance of each pathway may have differed across cells in our experiments in part due to variations in cone morphology and in part because mechanical dissociation led to partial loss of synaptic zones that affected bicarbonate entry, intracellular pH and [Cl -] in . These multiple pathways may have two functional consequences. First, it may enable cone OSs to accumulate greater levels of bicarbonate than rod OSs. Second, the expression of carbonic anhydrase, which can convert bicarbonate back to CO 2 , and the dependence of the directionality of bicarbonate movement by the exchanger on the bicarbonate gradient could enable cones to stabilize bicarbonate levels within their OSs rather than simply remove it.
Since bicarbonate influences cGMP synthesis by ROS-GC, changes in the production of bicarbonate in retina would be expected to affect visual function. Expression of carbonic anhydrase in red-and green-sensitive cones but not in blue-sensitive cones in humans (Nork et al., 1990) may explain why color vision changes after administration of a carbonic anhydrase inhibitor (Widengård et al., 1995;Leys et al., 1996). Disturbances in bicarbonate levels within the retina arising from defects in metabolism or transport would be expected to give rise to retinal disease due to dysregulation of pH and/or inappropriate ROS-GC activity. Knockout of NBC3 or NBCe2, which might block bicarbonate uptake into photoreceptors, causes a progressive decline in ERG amplitude as rods degenerate (Bok et al., 2003;Kao et al., 2011). The pathologic accumulation of cGMP in retinal rods and cones caused by mutations in phosphodiesterase, ROS-GC, or GCAPs results in degenerative retinal diseases (for review, see Iribarne and Masai, 2017). In such diseases, an overabundance of bicarbonate would be expected to accelerate the course of disease, whereas suppressed bicarbonate levels might have the opposite effect.