C-subfamily ATP binding cassette transporters extrude the calcium fluorescent probe fluo-4 from a cone photoreceptor cell line

Whole transcriptome sequencing has revealed the existence of mRNAs for multiple membrane transporters in photoreceptors. Except for ATP binding cassette (ABC) member A4, involved in the retinoid cycle, an understanding of the function of most transport proteins in photoreceptors is lacking. In this research paper, extrusion of fluo-4, a Ca2+ fluorescent probe, from 661W cells, a cone photoreceptor murine cell line was studied with online fluorometry and immunocytochemistry. Fluo-4 efflux was temperature dependent, required ATP but not extracellular Na+, was not affected by pH in the range 5.4–8.4, and followed saturating kinetics with a Km of nearly 4 μM, suggesting it was effected by ABC type transporters. A panel of antagonists showed an inhibitory profile typical of the C subfamily of ABC transporters. Immunofluorescence staining was positive for ABCC3, ABCC4 and ABCC5. These experimental results are compatible with fluo-4 being extruded from 661W cones by one or a combination of C-type ABC transporters. Examination of physicochemical descriptors related to drug membrane permeability and ABC substrate binding region further suggested efflux of fluo-4 by C-type ABC transporters. Possible functions of this transport mechanism in photoreceptors are discussed.


Introduction
Mouse cone photoreceptor (PR) cell line 661W originated from a retina tumour generated by selectively targetting the expression of the oncoprotein large tumour antigen of the simian virus SV40 to PRs (al-Ubaidi et al. 1992). 661W cells express cone PR markers, including those of the cone cilium, but also a few markers specific to retinal ganglion cells, while typical rod markers are absent (Tan et al. 2004;Wheway et al. 2019). They have been useful for in vitro modelling of cone PRs in studies of retinal degenerative disease, oxidative stress, neuroprotection and phototoxicity (see for instance Crawford et al. 2001;Kunchithapautham and Rohrer 2007;Jackson et al. 2016).
Membrane transporters are divided in two superfamilies: (i) Solute carrier transporters (SLCs) and (ii) ATP binding cassette transporters (ABC). A total of 458 SLCs distributed in 65 families are annotated to date in the human genome (Hediger et al. 2019). SLCs leverage energy from chemical or electrochemical gradients to move solutes in a facilitated (gradient is from the same species transported) or in a secondary/tertiary fashion (where the gradient of a co-or counter-transported solute is coupled to the translocation of other substrate). Most SLCs are categorised as influx or uptake carriers but flux direction usually depends on gradient structure, circumstance that has been taken advantage of for experimental design in transmembrane transport research (Lin et al. 2015). In contrast, the 49 ATP binding cassette transporters (ABCs), subdivided in 9 families, are primary active transporters that use ATP to move substrates against a concentration gradient, typically in an efflux direction (Morrissey et al. 2012;Nigam 2015). Mouse genome has homologous Abcs and Slcs, capitalisation of the gene names 1 3 indicating murine versus human all-caps genes, with similar properties .
Some of the 500 + membrane-bound transporters show a broad substrate specificity, promiscuously transporting dozens or even hundreds of different endogenous metabolites and xenobiotics greatly affecting drug disposition (Nigam 2015). Among the SLC superfamily, the most important broad substrate transporters described so far comprise SLC21/SLCO/OATP family (organic anion transporter polypeptides), SLC22-OAT/OCT family (organic anion transporters/organic cation transporters) and SLC47A-MATE family (multidrug and toxin extrusion). For instance, in vivo, SLCO1B1 mediates the uptake of statins, antibacterials, anticancer drugs, HIV protease inhibitors, antihypertensives in addition to endogenous metabolites such as eicosanoids or steroids (Liu 2019). The main ABCs involved in multidrug transport are ABCB1/P-gp/MDR1 (P-glycoprotein, multidrug resistance), many of the members of the ABCC/MRP (multidrug resistance-associated protein) 13-member subfamily, ABCG2/BCRP (breast cancer resistance protein) and ABCB11/BSEP (bile salt export pump, sometimes called sister P-gp). ABCB1/P-gp substrates are numbered in the hundreds. The broad specificity picture is similar for ABCCs and ABCG2. There is a large overlap among the substrates of each of the multiespecific ABCs but differences in substrate selectivity exist. For instance, P-gp substrates are mostly lipophilic organic uncharged molecules or organic cations, BSEP has high affinity for conjugated bile acids and for a subset of the substrates of P-gp, whereas ABCCs tend to transport organic anions, amphotereous molecules and glutathione, sulphate and glucuronide conjugates, while ABCG2 prefers lipophilic substances but can also move sulphate-conjugates as well as some unconjugated organic anions (Ambudkar et al. 1999;Mao and Unadkat 2005;Deeley et al. 2006;Takara et al. 2012). These proteins play an important role in drug disposition, drug-drug interaction and pharmacotherapy resistance to the extent that recommendations have been issued by regulatory agencies for specific testing in drug development (EMA Committee for Human Medicinal Products 2012; FDA Center for Drug Evaluation and Research 2020).
Most multidrug-transporting proteins are expressed in the kidneys, liver and, importantly for pharmacotherapy, in the tightly sealed blood-organ barriers of brain and other organs, permitting the passage of small hydrophilic and other larger molecules across the vasculature to and from the parenchyma of brain (blood-brain barrier), placenta (blood-placenta barrier) or retina (blood-retina barrier -BRB) (Nigam 2015). Moreover, cancerous cells commonly overexpress multidrug transporters underlying chemotherapy resistance to cancer treatment (Cui et al. 2015).
In the retina, PRs express the transporter ABCA4 involved in ciliary retinal recycling. Mutations to this gene cause Stargardt disease, an early onset macular dystrophy (Molday 2015). The human proteome atlas (proteinatlas. org) (Uhlén et al. 2015) describes mRNA expression of multispecific ABCs and SLCs in PRs, namely ABCC5, ABCC1, ABCC4, ABCC2, SLCO3A1, SLCO4A1 and SLC22A5. Their role in cones and rods physio-pathology is as yet unknown. Furthermore, a whole-transcriptome RNA sequencing performed on 661W cells found mRNA transcripts of Slc22, Slc47, Abcg2, Abcc1, Abcc5, Abc1b, Abcc4, Abcc3 and Abcb11 (Wheway et al. 2019) pointing to the likely possibility of functional expression of these transport proteins in the cell line. Elucidating their function in 661Ws is surely useful to contribute to clarify the role of multispecific transporters in photoreceptor pathophysiology.
Fluorescence dyes, including calcium indicators, have been the tool of choice to study membrane transporter physiology (Wang et al. 2001(Wang et al. , 2003Sauna et al. 2004;Zhou et al. 2008). We found that the Ca 2+ fluorescent probe fluo-4 was rapidly and efficiently removed from 661W cones' cytosol. We hypothesised that one or several of the broad substrate specificity transporters were responsible for fluo-4 clearing. In the present work, we have characterised the kinetics of F4 efflux, endeavoured to identify the membrane transporter(s) that evacuate the Ca 2+ probe and discussed potential functions in PR cells.

Materials and methods
Cell culture 661w cell line (a generous gift of Profs. N. Cuenca and V. Maneu) was grown in Dulbecco's modified Eagle medium (DMEM, SantaCruz Biotechnology) supplemented with 10% foetal bovine serum (Sigma-Aldrich, Madrid, Spain), 100 μg/ml penicillin and 100 u/ml streptomycin. Cultures were maintained in a controlled gas (95% air, 5% CO 2 ), humidity (95%) and temperature (37 °C) environment. Proliferation rate was assayed by allowing cells plated at several densities to grow for 1 to 3 days in 96-well plates. Then cells were incubated with resazurin that is reduced to fluorescent resorufin by cell dehydrogenases. Fluorescence intensity is proportional to the number of cells. Doubling time was approximately 26 h. This result was in line with a previous 661W growth rate estimation (Crawford et al. 2001).
Appropriate number of cells was seeded in 75 cm 2 flasks to reach confluency every 2-3 days when they were trypsinised, counted and plated in 96-well black with clear bottom plates for fluorescence (Corning Incorporated, USA) at a density of 50,000 cells/well. All experiments were carried out 24 h after plating.

Fluorescence measurements
Cells were incubated with fluo-4-AM (Thermofisher, USA) at 37 °C in buffer containing 137 mM NaCl, 4 KCl, 1 MgCl 2 , 2 CaCl 2 , 10 glucose and 10 HEPES-NaOH (pH 7.4) for varying times and concentrations as specified in the 'Results' section. Fluorescence recordings were carried out in triplicate wells in a Fluostar Optima spectrophotometer (BMG Labtechnologies, Germany). Excitation wavelength filter was a bandpass 480/10 nm, and emitted fluorescence was recovered through a 520/10 filter. Fluorescence was blanksubtracted and expressed as a percentage of intracellular or of total overall F4 fluorescence as follows: 1. % of intracellular fluorescence obtained by cell lysis: 2. % of total overall fluorescence resulting from the addition of intracellular and extracellular fluorescence recorded from the loading buffer.
In some experiments, intracellular F4 concentrations were interpolated by plugging blank-subtracted fluorescence values in the equation of a [F4]-fluorescence intensity calibration line, obtained measuring fluorescence of known concentrations of a potassium salt of F4 readily soluble in aqueous buffer (see Results).

Immunofluorescence
Mouse monoclonal antibody against ABCC1 (sc-53130) was from Santacruz Biotechnology. Policlonal antibodies raised in rabbit reacting with ABCC2 (E-AB-17352), ABCC3 (E-AB-68030), ABCC4 (E-AB-60680) and ABCC5 (E-AB-12566) were purchased from Elabscience (USA). Secondary fluorescent conjugates against mouse (antibody-CF488A) and rabbit (antibody-CF568) were from Sigma-Aldrich (Madrid-Spain). Immunocytochemistry was performed according to standard procedures with the following specifics: 661W cells were plated in polyl-lysine coated glass coverslips at low density for 24 h. Then they were fixed with cold 4% para-formaldehyde for 10 min. A mild permeabilisation was carried out with 0.03% triton-X-100 in phosphate buffered saline (PBS) followed by a 1-h blocking step at room temperature in 10% goat serum in PBS. Primary antibodies were incubated for 1 h at room temperature in 1% goat serum-PBS. Fig. 1 Fluo-4 intracellular retention increases with sulfinpyrazone concentration. Panel (a) Experimental recording; each trace averages 3 wells and is shown with standard errors superimposed. Intracellular fluorescence intensity was measured by cell lysis with TX. Once maximal intensity was reached, EGTA was superfused chelating Ca 2+ and quenching F4 fluorescence. Arrowheads signal superfusion times. Note increasing sulfinpyrazone concentrations (shown next to each corresponding trace, in μM) augmented dye retention. Bottom-most trace is a blank recording without F4-AM addition. Control wells, incubated only with F4-AM, displayed a small increase about 13% of maximal fluorescence obtained co-incubating with sulfinpyrazone. Panel (b) Sulfinpyrazone concentration-response curve obtained from 4 experiments similar to panel A. Intracellular fluorescence is expressed as the percentage of the total F4 fluorescence obtained by adding intracellular and extracellular intensities. Standard error bars smaller than filled circles not shown. Half-maximal concentration (IC 50 ) was 94.18 ± 7.63 μM. a.f.u., arbitrary fluorescence units Primary dilution was 1:100. Three 5-min PBS washouts interspersed between different solutions. Secondary antibodies were incubated for 1 h at room temperature at a concentration of 1:500. Nuclei were counter-stained with 4', 6-diamino-2-phenylindole (DAPI) (Sigma-Aldrich), 1:1000, for 5 min. As controls, coverslips were treated with the same protocol but (i) without primary antibody and normal secondary antibody incubation or (ii) with normal primary antibody incubation and without secondary antibody. Condition i had slightly more background fluorescence and was used to configure maximum light intensity and exposure settings so that no fluorescence was observed. These settings, or lower exposure/intensities, were used to take microphotographs of the stained cells in a Leica DM4000 conventional fluorescence microscope.

Data analysis
Half-maximal inhibitory concentrations (IC 50 ) were obtained by fitting the responses to different concentrations of antagonist to the logistic-Hill equation: where C is the concentration of the inhibitor and h the Hill or slope coefficient. The experiments presented here measure dye retention; hence, higher antagonising effect resulted in a greater fluorescence signal. where V (efflux speed) and C (substrate concentration) were obtained from recordings and V max (the asymptotical V = V max C Km + C maximum speed) and Km (the Michaelis-Menten coefficient) were fitted from the experimental data.
A p-value < 0.05 was considered as the limit of statistical significance.

Fluo-4 is extruded from 661W cells by a membrane transporter
661W cells were incubated with 5 μM fluo-4-AM (F4-AM) for 1 h at 37 °C. F4-AM is a non-fluorescent cell-permeable molecule; once in the cytosol the acetoxymethilester (AM) moiety is cleaved by cell esterases transforming F4-AM into the fluorescent dye fluo-4 (F4), which is membraneimpermeable and undergoes a large fluorescence increase upon Ca 2+ binding (Minta et al. 1989). After the incubation period, the external loading buffer was transferred to a clean plate and replaced by an identical solution without dye. Cells were taken to a spectrophotometer equipped with solution injectors and rapidly permeabilised by superfusion of the non-ionic detergent triton-x-100 (TX). There was barely a small increase in fluorescence (Fig. 1a, control trace). However, wells with recovered loading buffer displayed large values of fluorescence intensity (not shown). A similar experiment was carried out in the presence of the widespectrum membrane transporter blockers probenecid (1 mM) or sulfinpyrazone (500 μM) throughout. In this instance, cell permeabilisation resulted in a large increase in fluorescent intensity (henceforward intracellular fluorescence). Loading buffer displayed now low fluorescence. Injection of 100 mM of the Ca 2+ chelator EGTA a few minutes after TX decreased fluorescence to blank (no probe) levels (Fig. 1a). Measurement of fluorescence of a buffer solution containing 2 mM Ca 2+ and 5 μM F4-AM in cell-free wells was only slightly above blank conditions verifying that the acetoxy-methylester (AM) form of F4 is non-fluorescent.
Intracellular signal increased dose-dependently with sulfinpyrazone concentration (Fig. 1b) while loading buffer fluorescence decreased. Half-maximal concentration of dye extrusion inhibition by sulfynpirazone (IC 50 ) was nearly 100 μM. IC 50 for probenecid was 350 μM. Furthermore, experiments carried out in the absence of extracellular Na + , which was equiosmolarly substituted by N-Methyl-D-Glucamine, displayed similar results, i.e. low F4 retention in control conditions and high intracellular signal with probenecid/sulfinpyrazone. Buffer pH of 5.4, 6.4 and 8.4 showed similar results to control experiments at pH 7.4 (data not shown) suggesting that this range of pH was not affecting efflux.
Taken together this experiments demonstrate that: (1) F4-AM enters 661W cell cytoplasm where it is hydrolysed to its fluorescent non-permeable form (F4); (2) F4 is then extruded from the cell into a buffer rich in Ca 2+ ; (3) Since F4 is impermeable to the cell membrane, it must have been extruded from the cell by a mechanism other than passive diffusion; (4) The mechanism by which F4 is extruded from the cell is inhibited by probenecid and sulfinpyrazone in a dose-dependent form, suggesting a membrane transport enzyme; and (5) F4 efflux is Na + and pH-independent.
Probe clearance from cytosol is a common problem in experiments with fluorescent dyes. It is often tackled by using wide spectrum inhibitors of plasma membrane transporters such as probenecid, due to the difficulty of pinpointing a specific transporter among dozens of potential options (Di Virgilio et al. 1988). The lack of specificity of these inhibitors together with the high concentrations used may difficult experimental interpretation. For instance, probenecid blocks OAT (Enomoto et al. 2002) and ABC tranporters (Gollapudi et al. 1997), but it also inhibits p2X7 purinoceptor (Bhaskaracharya et al. 2014;Bartlett et al. 2017), and activates transient receptor potential V2 (Bang et al. 2007). Following we further characterise F4 efflux in order to identify its molecular effector(s).

Transport mechanism is temperature-dependent
To test whether F4 extrusion rate depended on temperature, 661W cells were incubated for 1 h with 600 μM sulfinpyrazone and 5 μM F4-AM at 37 °C; Then, loading buffer was exchanged by clean solution (no dye, no sulfinpyrazone) at 4 °C and introduced in the spectrofluorometer prewarmed or pre-cooled to a specific temperature with the machine's temperature control device, or ice (lab temperature was 22 ± 2 °C) and fluorescence intensity was recorded for around 1 h at 30-s intervals. Temperatures tested were 17, 27 and 37 °C. Rate of fluorescence increase was higher the higher the temperature (Fig. 2a, b). Thus, at 17 °C, the steepest slope in the recording was 0.18 ± 0.04%/min (data were normalised to total intracellular fluorescence); at 27 °C, rate of fluorescence increase was 0.66 ± 0.07%/min. The fastest rate of 4.25 ± 0.38% F4 extrusion per min, recorded at 37 °C, was more than 6 times faster than at 27 °C and 24 times as fast as at a temperature of 17 °C. This experiment shows that increasing temperatures augment the efficacy of F4 efflux.

F4 traslocation outside the cytosol depends on ATP availability
Both probenecid and sulfinpyrazone are able to inhibit transporters that are ATP-dependent and non-ATP-dependent. To discern what superfamily of transporters was responsible for F4 extrusion in 661W cells, we performed experiments depleting cell reserves of ATP. This manipulation should stop F4 extrusion if dye removal consumed ATP; otherwise, no effect was expected. The main ATP synthesising machinery of most cells is located in the mitochondria; it comprises an electron transport chain (ETC) that creates a H + gradient which is used by the ATP-synthase to synthesise ATP from ADP and Pi. A less efficient way of fabricating ATP uses glycolysis.
In this experiment, we swapped glucose for 2-deoxiglucose in the buffer (2-deoxi-buffer) to impair glycolysis. 661W cells were pre-incubated for 30 min with 2-deoxibuffer and one of three drugs which were present at all times during the experiment: (i) oligomycin that inhibits ATP-sinthase; (ii) FCCP, a protonophore that dissipates the mitochodrial H + gradient; (iii) sodium azide, and ETC uncoupler that binds to complex IV. We used three structurally unrelated drugs to make as sure as possible that any effect was not the result of interactions of the drugs with the transport mechanism itself. After drug-incubation, cells were loaded for 1 h with fluo-4-AM at 37 °C. Finally, loading buffer was removed and both, buffer and intracellular fluorescence intensity, measured at 37 °C.
F4 extrusion was little affected by loading with 2-deoxibuffer alone. Oligomycin inhibited F4 extrusion as did FCCP and azide. Furthermore, greater efflux inhibition was achieved with higher concentrations of each drug, clearly pointing to concentration-dependence (Fig. 3). This experiments show that F4 is extruded from 661W cells by an ATPdependent transport mechanism.

F4 efflux exhibits saturating kinetics
The relation between initial efflux rate (V 0 ) and substrate concentration, i.e. [F4] c , was measured loading cells for different times with 5 μM or 10 μM F4-AM at 37 °C. 1 mM sulfynpirazone was present during loading to stop F4 efflux. We reasoned that the longer the cells were exposed to F4-AM, the more probe would diffuse into the cytosol; more time there would also be for cytosolic hydrolysis of the AM from F4. Hence, different [F4] c would be available for transport. Loading buffer was exchanged with clean cold buffer and the plate introduced in the spectrophotometer prewarmed at 37 °C. Fluorescence intensity was recorded for around 70 min at 30-s intervals; then, cells were lysed to quantify total intracellular fluorescence (Fig. 4a). Blanksubtracted maximal fluorescence was used to interpolate [F4] c from a fluorescence intensity-[F4] calibration line (Fig. 4a, inset). Efflux speed was measured as the slope of a line fitted to the steepest region of the recording previous to TX lysis. Figure 4b shows the V 0 -[F4] c results of several experiments. Efflux rate tends to saturate at higher [F4] c . Fitted Michaelis-Menten model yielded an apparent Km of 3.95 ± 0.67 μM. This figure is similar to previous Km findings for transport of Fluo-3, a closely related F4 analogue, by MRP1 (Km = 12 μM) and MRP2 (Km = 3.7 μM) (Keppler et al. 1999).

Pharmacological profile of fluo-4 efflux suggests an MRP effector
The experiments described above strongly suggest that F4 efflux from 661W cones was effected by membrane transporters of the ABC superfamily. Possible candidates being multispecific drug transporters P-gp, MRP family, BCRP and BSEP. Pharmacological identification however is not straightforward because: (a) Transporter promiscuity, often a drug, is substrate of several transporters (Dohse et al. 2010); (b) There is a lack of specific inhibitors, plus if high enough concentrations of inhibitors are used, specificity is lost (Wang et al. 2001;Tiwari et al. 2009); (c) In many instances, efficacy of inhibition appears to be related to the pair substrate-inhibitor used, reflecting the existence several different affinity binding sites within a given ABC transporter (Polli et al. 2001;Wang et al. 2001Wang et al. , 2003. Hence, a panel of drugs was tested to sketch an antagonist profile of the ABCs responsible for F4 efflux in 661W cells. Maximal antagonist concentrations were chosen based on the literature. If high inhibition was obtained (that is, a strong intracellular signal was achieved in assays as those of Fig. 1), a dose-response curve was obtained and IC 50 calculated. We also verified that the compounds used did not display any absorbance/fluorescence that could interfere with F4 fluorescence. We term all drugs used here as 'inhibitors' even though some may be substrates for one or several of the pumps, exerting competitive inhibition against F4 transport. Table 1 details blockade achieved by the drugs. Thus, uricosurics probenecid and sulfinpyrazone maximally inhibited efflux as did non-steroidal anti-inflammatory indomethacin, sulfonylurea glibenclamide and chloride channel blocker NPPB. All of them were typical blockers of C-type ABCs (Zhou et al. 2008). Oestrogen receptor modulator tamoxifen, a P-gp antagonist (Wang et al. 2001), achieved a dye retention slightly above controls at 100 μM (18%). Similar low effect (around 30%) had 30 μM fumitremorgin C, a potent and specific ABCG2 antagonist at submicromolar concentrations (Mao and Unadkat 2005). Anti-tuberculous rifampicin, which effectively antagonises BSEP at 10-30 μM (Byrne et al. 2002), had little effect (around 30% F4 retention) at 100 μM on 661W cells. Tyrosine kinase inhibitors (TKI) imatinib and nilotinib reports inform of submicromolar to low micromolar IC 50 s against P-gp and ABCG2-mediated efflux (Dohse et al. 2010;Haouala et al. 2011); Used here at concentrations as high as 100 μM, imatinib and nilotinib afforded around 40% retention of F4. Previous studies demonstrated that calcium channel blockers are also substrates/ inhibitors of ABCs. For instance, verapamil antagonised [ 3 H]-digoxin efflux mediated by MDR1/ABCB1 (IC 50 around 13 μM) while dihydropiridine nitrendipine produced only a modest inhibition (IC 50 around 70 μM) (Takara et al. 2002). Conversely nitrendipine was a strong inhibitor of BCRP/ABCG2-mediated transport of Hoechst33342 and verapamil had no effect (Takara et al. 2012). In the present study, probed at 100 μM, verapamil, nitrendipine and three other dihydropyridines, isradipine, nimodipine and nifedipine, showed around 20% inhibition suggesting that neither P-gp nor BCRP was involved in F4 extrusion from 661W cones. BSEP and P-gp are inhibited by anti-hypertensive reserpine (Wang et al. 2003;Nagy et al. 2004) and by isradipine (Pedersen et al. 2013;Jouan et al. 2016). Neither of them had inhibitory effect on 661W F4 efflux. Finally, antibiotic ivermectin blocks efflux by all three, P-gp, ABCCs and ABCG2, with IC 50 s ranging from 0.5 to 4 μM (Didier and Loor 1996;Lespine et al. 2006;Jani et al. 2011) as does glibenclamide (Golstein et al. 1999;El-Sheikh et al. 2007;Zhou et al. 2008). Here 100 μM ivermectin achieved 28.9 ± 4% inhibition while 100 μM glibenclamide fully blocked F4 eflux.
In summary (Fig. 5), 5 contrasted ABCCs inhibitors fully antagonised F4 efflux from 661W cells, while high concentrations of blockers specific for one or various other families of ABCs had little or no effect. This inhibitory profile is compatible with type C ABC transporters being the effectors of F4 efflux in 661W cells.

Immunofluorescence labels multiple C-type ABCs
The C-family of ABC transporters comprises 13 members, of which ABCC1 through ABCC5 can be categorised as multispecific. We performed secondary immunofluorescence assays against those 5 ABCCs. 661W cones were plated in polylisine-coated glass coverslips for 24 h, fixed and immunostained as detailed in Methods. ABCC1 and C2 staining was negative (not shown). Antibodies against ABCC3, C4 and C5 showed intense membrane patches and intracytoplasmatic staining (Fig. 6). Subcellular MRP immunolabelling has been reported in the cytoplasm (Lee et al. 2000) and several cell organelles, such as the rough endoplasmic reticulum (Sugawara et al. 1995), mitochondria (Roundhill and Burchill 2012;Roundhill et al. 2016), nucleus and lysosomes (Rajagopal and Simon 2003). This assay confirms the presence of C-type ABC transporters in 661W cone membrane.

Discussion
The aim of this work was to identify the mechanism of F4 efflux from 661W cones. Being F4 virtually cell-membrane impermeable, we hypothesised it was extruded by a multispecific membrane transporter. F4 extrusion efficiency increased with temperature, relied on ATP, showed saturating kinetics and was inhibited by type C ABC transporter antagonists and not by ligands of other multidrug transporters. Furthermore, 661W cones were immunolabelled by antibodies against several MRPs. We concluded that one or more MRPs were extruding F4 from 661W cones.
Structural studies have shown the location of the entry pathway to multidrug ABCs: ABCB1 ligands entrance opens towards the inner leaflet of the lipid bilayer (Aller et al. 2009); ABCC-and ABCB11-interacting compounds Fig. 5 Venn diagram summarising antagonist specificity for different ABCts and potency of F4 efflux inhibition in 661W cones. Compound names have been placed inside sets representing each a subtype/ subfamily of the multispecific ABC transporters studied. Each compound is also categorised according to inhibitory effect strength: low, up to around 30% F4 retention (smaller font size); medium, around 40-60% inhibition (intermediate font size); and high blockade (larger font size). Note all highly effective blockers fall into the ABCC set are recruited from within the aqueous phase of the cytoplasm (Johnson and Chen 2017;Wang et al. 2022), whereas ABCG2 allocrites (i.e. ABC transporter interacting compounds (Egido et al. 2015)) reach the binding site both from within the membrane inner leaflet and from the cytosol (Taylor et al. 2017). Hence, extracellular allocrites need to partition into the lipid membrane and/or cross to the cytoplasmic side. Uncharged organic ligands readily partition into the membrane according to size and lipophilicity (measured by logP) (Seelig 2007). High logP substances rapidly enter the membrane while slowly partitioning into the aqueous cytoplasmic side. Hence, they tend to build up in the membrane as suggested by the invertend 'V-shaped' relation between logP and biological activity (Kubinyi 1977). LogD 7.4 takes into account charged and uncharged species at pH 7.4. Higher dissociation results in lower values of logD 7.4 , while drugs that do not dissociate have similar logP and logD 7.4 . From this diffusional standpoint, higher logP/logD 7.4 compounds reach higher concentrations in the cell membrane facilitating interaction with ABCB1 and/or ABCG2. The 11 compounds in Table 2 with logD 7.4 > 2 had weak to null antagonising effect, suggesting F4 was not extruded by P-gp or BCRP.
A combination of hydrophobic and polar regions gives a multidrug ABC binding site preference for certain ligands roughly outlined as follows: P-gp site is mostly hydrophobic with some acidic patches so it favours binding of hydrophobic and weakly cationic compounds (Aller et al. 2009); ABCC1 and ABCC5 (and presumably other ABCCs) cavity has positively charged and hydrophobic regions favouring organic anion binding (Sager et al. 2012;Johnson and Chen 2017). ABCG2 binding region has a mixture of hydrophobic Fig. 6 Positive immunostaining for ABCC3, ABCC4 and ABCC5. 661W photoreceptors were labelled with antibodies against MRP1-5 and counterstained with the nuclear dye DAPI. Staining was positive for MRP3 (top row -A), MRP4 and MRP5 (middle rows -B, C). Bottom row (D) shows microphotographs of a control assay without primary antibody but normal secondary antibody incubation. Left column, counterstain with DAPI (blue -1); centre column, labelling of ABCC transporter (red -2); right column, merge (3). Arrowheads point to membrane patches of MRP. Scale bar: 5 μm and polar groups that interact with both hydrophobic and hydrophilic charged ligands (Taylor et al. 2017;Jackson et al. 2018). The ABCB11 pocket has hydrophilic and hydrophobic regions to recognise complementary portions of bile-acid-like compounds (Wang et al. 2022). Co-applied allocrites compete for binding and substrate/inhibitor character is determined in terms of concentrations and relative affinity. F4 is a highly negative anion (− 5 or − 3 if bound to Ca 2+ , Table 2) that may be transported by ABCCs, ABCG2 and/or ABCB11. F4 transport by ABCCs has been described (Sauna et al. 2004). Highly charged hydrophilic compounds are also good substrates for ABCG2 (Egido et al. 2015). All − 1 anions were full inhibitors. Negative charge hinders interaction with ABCB1, but anions could compete with F4 for binding to ABCCs, ABCG2 and ABCB11, which can accommodate negatively charged ligands. Neutrally charged compounds effectively inhibit P-gp and BCRP (Egido et al. 2015). High concentrations of dihydropyridines and ivermectin, neutrally charged at pH 7.4, did not inhibit F4 efflux. Ivermectin has been shown to inhibit ABCC1, C2 and C3 (Lespine et al. 2006), but its affinity is greater for ABCB1 and ABCG2 (Lespine et al. 2006;Jani et al. 2011); its lack of inhibition here may be accounted by F4 being transported preferentially by MRP4 or MRP5 over MRP1, 2 or 3. It is of note the strong effect of MRP5 inhibitor NPPB (Pratt et al. 2005), suggesting that MRP5 may be important for F4 extrusion in 661W cells. Rifampicine pK a s close to 7.4 point to the coexistence of several electrically charged species including a fraction of zwitterion, which are good activators of ABCG2 (Egido et al. 2015); however, 100 µM rifampicine, which also interacts with BSEP, barely had any inhibitory effect. Similarly, FTC, a lipophilic ABCG2 inhibitor of nanomolar potency, did not inhibit F4 efflux (Mao and Unadkat 2005). Imatinib, nilotinib, reserpine, tamoxifen and verapamil tend to be + 1 at pH 7.4. Low positive charge and high lipophilicity suggest interaction with ABCB1 and inhibition of ABCG2 (Egido et al. 2015). TKIs showed mild inhibition of F4 extrusion while verapamil, reserpine and tamoxifen barely interfered. TKI concentration used here, around 100 times greater than IC 50 s described to block ABCG2 and P-gp (Dohse et al. 2010;Haouala et al. 2011), suggests that at higher concentrations, TKIs loose specificity and are capable of antagonising MRP transporters (Dohse et al. 2010). In summary, F4 physicochemical properties suggest eflux by MRPs but also by BCRP or BSEP and not by P-gp. The five fully effective inhibitors could compete with F4 for the potential transporters; however, all other compounds, which strongly interact with one or more of ABCB1, ABCB11 and/ or ABCG2, exerted no inhibition on F4 efflux, suggesting that F4 is extruded from 661W cells by C-type ABC organic anion transporters.
The order of mRNA abundance in 661W cones according to Wheway et al. (Wheway et al. 2019) is as follows: Abcg2 ≅ Abcc1 ≅ Abcc5 > Abcb1b ≅ Abcc4 ≫ Abcb1a ≫ Abcc3 ≫ Abcb11 ≫ Abcc2. The highest expression levels are of Abcg2, Abcc1 and Abcc5 followed by Abcb1b and Abcc4, the rest comparatively having much lower number of transcripts. Our immunofluorescence assays on 661Ws were positive for ABCC3, C4 and C5 while staining for ABCC1 and ABCC2 were negative. Even though immunolabeling shows MRP protein is present, there is a discrepancy with the mRNA data, since mRNA levels for MRP1 in 661Ws are high and the immunolabelling was negative; also MRP3 number of transcripts is low and the staining was positive. This discrepancy might be ascribed to negative regulation of mRNA translation or to a rapid protein turnover; both these processes accounting for large variations in the correlation between mRNA abundance and protein expression (Maier et al. 2009).
It is worth examining transcriptomics of human photoreceptors. Human cones (hCones) have significant levels of ABCC1, ABCC4 and ABCC5 mRNA. Human rods (hRods) express a wider variety of transporter mRNAs including ABCC1, ABCC2, ABCC3 and ABCC5 and small quantities of P-gp and BSEP (proteinatlas.org). ABCC5 mRNA is the most abundant by far, around 9 and 14 times, in hCones and hRods respectively, the mRNA expression levels of next in abundance ABCC1. To put these figures in context, ABCA4, the transporter involved in retinoid cycling in PR Table 2 Compound physicochemical properties Data were obtained from online databases drugbank.com and chemspider.com. Where more than one pK a was available, they are separated by a semicolon as follows: strongest acidic pK a ; strongest basic pK a . n.a., not available discs, has twice/four times as much mRNA as ABCC5 in hCones/hRods. Incidentally, ABCA4 is an importer (Molday 2015) which rules out its participation in 661W F4 efflux. Should functional MRPs be confirmed in human PRs they may be important in reactive oxygen species (ROS) and cyclic guanosine monophosphate (cGMP) regulation, as is discussed next.
Because of the constant exposure to light, a generator of ROS, and their high metabolic rate, PRs and the retina at large must continuously detoxify large quantities of ROS (Wangsa-Wirawan and Linsenmeier 2003;Punzo et al. 2012). Reduced glutathione (GSH), a major antioxidant in the retina, inactivates oxidants in reactions catalysed by glutathione transferase and glutathione peroxidase, generating oxidised glutathione (GSSG) and glutathione-S conjugates (GS-conj). Glutathione peroxidase also catalyses the conversion of GSH and hydrogen peroxide to water and GSSG. PRs and retina express high concentrations of GSH and GSH processing enzymes (Singh et al. 1984;Beatty et al. 2000). MRPs remove endogenous toxic metabolites by extruding GSH, GSSG and GS-conjs (Leier et al. 1996;Cole and Deeley 1998). The view is suggested hence that glutathione adducts are effluxed from PR cytosol to the extracellular space where gamma-glutamyl transferase, retinal pigmented epithelium and Muller cells help recycle and/or dispose of them. Too large an imbalance between generation and extrusion/elimination of oxidants, GSH, GSSG and GS-conj may result in increased oxidant stress and cell death (Roh et al. 2007;Ballatori et al. 2009). Indeed retinopoathies such as age-related macular degeneration, diabetic retinopathy and retinitis pigmentosa are associated with an increase in retinal ROS (Beatty et al. 2000;Punzo et al. 2012;Du et al. 2013). Thus, MRP modulation could be a potential target in achieving an enhanced resistance to oxidative insult in PRs.
ABCC4 and ABCC5 extrude cyclic nucleotides including cyclic guanosine monophosphate (cGMP) (Sager and Ravna 2009) which is instrumental for transduction of luminous/dark signal into a glutamate release pattern at the cone/ rod-bipolar cell synapse. Free cytosolic cGMP concentration ([cGMP] c ) in PR cilium is tightly regulated, at around 1-5 μM (Zhang and Cote 2005). As MRP5 extrudes cGMP with a Km of approximately 2 μM (Jedlitschky et al. 2000;Sager and Ravna 2009), it could be an additional participant in maintaining appropriate [cGMP] c levels, should ABCC5 be located in ciliary outer segments. Moreover, [cGMP] c dysregulation producing increased levels of the nucleotide is associated with PR cell death in a dozen rodent models of hereditary human blindness (Farber and Lolley 1974;Xu et al. 2013;Arango-Gonzalez et al. 2014); In many of these models, [cGMP] c elevations occur both in cilia and somata of cones/rods (Arango-Gonzalez et al. 2014).
[cGMP] c causality in PR toxicity was proved by the protection afforded by reducing cGMP levels (Yang et al. 2020;Das et al. 2021).
Taken together, these studies suggest that, in pathologically elevated [cGMP] c conditions, MRP5 and/or the lower cGMP affinity MRP4 may play a protective role by clearing cytotoxic cGMP overflow from PRs. The question arises of whether MRP4/MRP5 modulation or induction in PRs might be a useful therapeutic tool in excess [cGMP] c -related retina diseases.
In conclusion, the first functional evidence of a multispecific ATP-dependent transporter in a PR cell model is presented here: 661W cones clear cytosolic F4 to background levels. Experimental data strongly suggests F4 efflux in 661Ws is mediated by type-C ABC transporters. If MRPs are confirmed functioning in PRs from primary cultures, explants and in vivo, it is suggested that they play a role in the regulation of oxidation metabolites and cGMP levels in PR physiology and retinal disease. Future research in this direction may benefit from the use of 661W cones in combination with genetic manipulation and other techniques to clarify the role MRPs and other ABCs play in PR (patho) physiology and their potential utility as therapeutic targets.