Role of monocarboxylate transporters in regulating metabolic homeostasis in the outer retina: Insight gained from cell-specific Bsg deletion

Abstract

The neural retina metabolizes glucose through aerobic glycolysis generating large amounts of lactate. Lactate flux into and out of cells is regulated by proton-coupled monocarboxylate transporters (MCTs), which are encoded by members of the Slc16a family. MCT1, MCT3, and MCT4 are expressed in the retina and require association with the accessory protein basigin, encoded by Bsg, for maturation and trafficking to the plasma membrane. Bsg^−/− mice have severely reduced electroretinograms (ERGs) and progressive photoreceptor degeneration, which is presumed to be driven by metabolic dysfunction resulting from loss of MCTs. To understand the basis of the Bsg^−/− phenotype, we generated mice with conditional deletion of Bsg in rods (RodΔBsg), cones (Cone∆Bsg), or retinal pigment epithelial cells (RPEΔBsg). RodΔBsg mice showed a progressive loss of photoreceptors, while ConeΔBsg mice did not display a degenerative phenotype. The RPEΔBsg mice developed a distinct phenotype characterized by severely reduced ERG responses as early as 4 weeks of age. The loss of lactate transporters from the RPE most closely resembled the phenotype of the Bsg^−/− mouse, suggesting that the regulation of lactate levels in the RPE and the subretinal space is essential for the viability and function of photoreceptors.

1 |. INTRODUCTION

The retina, composed of the neural retina and the retinal pigment epithelium (RPE), is among the most metabolically active tissues in the body and relies on glucose to support the high energy demands of visual transduction.^1,2 The RPE transports glucose from the choroid to the outer retina via GLUT1 transporters expressed in its basolateral and apical membranes.³ In the outer retina, glucose is primarily metabolized through aerobic glycolysis, a process that produces large amounts of lactate.^3,4

Transport of lactate within and out of the retina is mediated by monocarboxylate transporters (MCTs). MCTs differ in their affinity for lactate, which facilitates the directional flux of lactate and thus regulates metabolic coupling in the outer retina.¹ In particular, MCTs 1–4 have been characterized as proton-coupled lactate transporters^5,6 that are encoded by the SLC16a family, of which there are 14 known members. The neural retina expresses Slc16a1 (encoding MCT1) and Slc16a3 (encoding MCT4), while the RPE expresses Slc16a1 and Slc16a8 (encoding MCT3). The RPE expresses MCTs in a polarized manner with MCT1 expressed on the apical membrane and MCT3 expressed on the basolateral membrane. This allows for the transepithelial transport of lactate from the outer retina to the choroidal vessels.^7–12

MCT1, 3, and 4 are nonglycosylated integral membrane protein with 12 membrane-spanning domains. These transporters form a heteromeric complex with basigin (BSG), a single pass, highly glycosylated membrane protein that is a member of the immunoglobulin superfamily and encoded by Bsg.^5,6 There are two isoforms of BSG, BSG1 and BSG2, which arise by an alternative splicing event.⁵ Assembly of MCTs and BSG in the endoplasmic reticulum is required for proper maturation and trafficking to the plasma membrane. Previous work from our laboratory and others has shown in vitro that in the absence of one subunit, the other subunit is targeted for degradation.^10,13–15 This is similar to other heteromeric transporters, such as the Ca²⁺-ATPase and amino acid transporters, where it has been shown that in the absence of the accessory protein, the transporter is targeted for degradation.^16,17

The importance of MCTs in maintaining lactate homeostasis in the outer retina and normal visual function was demonstrated by the phenotype of the Bsg^−/− mice. In this model, in the absence of BSG, MCT1, MCT3, and MCT4 were not trafficked to the plasma membrane and were instead targeted for degradation.¹⁴ In the Bsg^−/− mice, retinal development appeared normal, with proper lamination and outer segment (OS) elongation, but this normal retinal structure failed to reveal abnormalities in photoreceptor function that were apparent in the electroretinogram (ERG).^18,19 The ERG abnormalities could be attributed to nutrient deprivation of photoreceptors from the loss of the lactate shuttle between photoreceptors and Müller cells.^14,20 Alternatively, the abnormalities could be a secondary effect due to the inability of the RPE to transport lactate and protons causing changes in pH and osmolarity in the subretinal space.^10,21,22 Lactate generated by the neural retina is a crucial substrate for oxidative phosphorylation (OXPHOS) for the RPE. When the oxidative capacity of RPE is disrupted, dedifferentiation occurs, along with subsequent photoreceptor cell death.^23–25

Since Bsg is expressed throughout the retina, it has not been possible to discern whether the retinal phenotype of the Bsg^−/− mice was from the loss of MCTs in photoreceptors, RPE, or both.¹⁴ To understand the role of lactate transporters in supporting outer retinal function, we generated mouse lines with genetic deletion of Bsg from rod photoreceptors, cone photoreceptors, or the RPE. We report that the impact of Bsg deletion differs significantly between rods, cones, and RPE cells and that the RPE-specific deletion most closely mirrors the phenotype of the systemic Bsg^−/−, highlighting the importance of lactate transport in maintaining the homeostasis of the outer retina.

2 |. MATERIALS AND METHODS

2.1 |. Animal handling

Mice carrying a floxed Bsg allele²⁶ were crossed to transgenic mice expressing Cre recombinase in rods (B6;SJLPde6b⁺ Tg(Rho-iCre)1Ck/Boc JAX stock #015850)²⁷ or cones (Tg(Opn1mw-cre)1Asw)²⁸ or RPE (C57BL/6-Tg(BEST1-cre)1Jdun/J)²⁹ to generate mice lacking Bsg in rods (Bsg^flox/flox; Rho-iCre; hereafter RodΔBsg) or cones (Bsg^flox/flox; coneCre; hereafter ConeΔBsg) or RPE (Bsg^{flox/ flox}; Best1-Cre; hereafter RPEΔBsg) and control littermates (Bsg^flox/+Cre positive; Bsg^flox/floxCre negative; hereafter control). To generate mice lacking Bsg in both rod and cone photoreceptors, ConeΔBsg and RodΔBsg were crossed to generate Cone/RodΔBsg. To generate MCT4^−/− mice, MCT4^+/− mice were purchased from Taconic Bioscience. The animals were backcrossed for 10 generations to C57Bl/6J (Jackson) mice and MCT4^+/− mice were used for breeding to obtain knock out and wild-type littermates.³⁰ The animals were genotyped by PCR using primers listed in Table 1. Neural retinas from 1 to 2-month-old NRL^−/− and C57BL/6J wild-type mice were provided by the laboratories of Dr. Anand Swaroop (National Eye Institute) and Dr. Jean Bennett (University of Pennsylvania). All animal procedures were conducted with the approval of the Institutional Animal Care & Use Committees of Thomas Jefferson University, the Louis Stokes Cleveland VA Medical Center, or the Cleveland Clinic, and conformed to the ARVO statement for use of animals in ophthalmic and vision research.

TABLE 1.

List of primers used for qPCR and genotyping

Genes	Forward (5′−3′)	Reverse (5′−3′)
Bsg1	GGA ATG CTC CAA ACG ACA GC	AGT AAG GTG GTT GCG GTC TG
Bsg2	GGC GGG CAC CAT CCA AA	CCT TGC CAC CTC TCA TCC AG
Nptn	AAC CAG CTG GGC CAA TGA A	ACC AAA GAG GTC CAA GCA GAA
Emb	ATC GCT TAC GTG GGG GAT TC	GAG CGT CAA TGG GAA CCT GT
Rplp0	AGATTCGGGATATGCTGTTGGC	TCGGGTCCTAGACCAGTGTTC
BSG Flox	GTA TAT GTG CTG CCG AAG CGA G	CAA AGC AGG TGG ACA GCC TAA TCT
Rho-iCre	TCA GTG CCT GGA GTT GCG CTG TGG	CTT AAA GGC CAG GGC CTG CTT GGC
Cre	ACTGGGATCTTCGAACTCTTTGGAC	GATGTTGGGGCACTGCTCATTCACC
Best-Cre	ATG CCC AAG AAG AAG AGG AAG GTG TCC	TGG CCC AAA TGT TGC TGG ATA GTT TTT A
MCT4	GCA GCG CAT CGC CTT CTA TC	GTG TCA AGC TTA TGC CTG TC

2.2 |. Western blot

Eyes were enucleated following an overdose of ketamine (100 mg/kg) and xylazine (10 mg/kg). Neural retina and RPE/choroid were isolated, and protein were extracted using 100 µL per retina and 50 µL per RPE/choroid of Pierce RIPA buffer (Cat#89900, Radioimmunoprecipitation Assay Thermo Scientific, Rockford, IL) with Halt Protease Inhibitor (Cat# 78420, Thermo Fisher Scientific, Waltham, MA). Neural retina samples were homogenized first with an 18-G Sterile needle (Cat#305195, BD PrecisionGlide Needle, Franklin Lakes, NJ) and then a 25-G Sterile needle (Cat#305124, BD PrecisionGlide Needle). Samples were placed on ice for 30 minutes with intermittent vortexing. The protocol for enriching the RPE protein or the RPE/choroid protein isolation has previously been described.³¹ In short, eyecups were isolated and four cuts were made to flatten each eyecup before it was placed in RIPA buffer and agitated to release the RPE. The supernatants were transferred to new tubes. The tubes were flicked every 10 minutes for 30 minutes after which samples were centrifuged at 14 000×g for 30 minutes, and the supernatants removed for protein quantification by BCA Protein Assay Kit (Cat#23225, ThermoFisher). A total of 5 or 15 µg (retina) or 5 µg (RPE/choroid) protein was loaded on 4%–12% NuPage Bis-Tris Protein gels (NP0321BOX, Invitrogen, Carlsbad, CA) or 10% NuPage Bis-Tris Protein gels (Cat# NP0301BOX, Invitrogen). Gels were run for 50 minutes at 200V per manufacturer instructions. Gels were transferred electrophoretically onto EMD Millipore Immobilon-P PVDF Transfer Membranes (Cat# IPVH00010, EMD Milipore, Burlington, MA) at 20V for 1 hour. Membranes were incubated for 1 hour at room temperature in blocking buffer (5% low fat powdered milk in Tris-buffered saline with 0.1% Tween20 [TBST]) and incubated overnight with antibodies at 4°C in blocking buffer TBST. Membranes were washed three times with DI water and incubated for 1 hour with secondary antibody in blocking buffer at room temperature. Blots were developed using chemiluminescence (Cat# 34075, SuperSignal West Dura, Thermo Fisher Scientific) on FluorChem M Protein Simple (San Jose, CA) detection system.

2.3 |. PNGaseF treatment

Retina lysates from control and NRL^−/− mice were prepared as described above. Lysates were treated with PNGaseF following the manufacture’s protocol (Cat# P0704S, New England BioLabs, Ipswich, MA). Briefly, 10 µg of protein and 1 µL of glycoprotein denaturing buffer were combined in a 1.5-mL tube and brought up to 10 µL of volume with deionized water (dH₂O). Proteins were denatured at 100°C for 10 minutes and centrifuged for 10 seconds at 10 000×g at 4°C. After addition of 2 µL of GlycoBuffer 2, 2 µL of 10% NP-40, 1 µL of PNGaseF, and 6 µL of dH₂O tubes were incubated in a 37°C water bath for 1 hour. SDS sample buffer was added to the tubes, and proteins were resolved by SDS-PAGE and western blot as described above. Blots were probed with BSG antibody.

2.4 |. Co-immunoprecipitation

The procedure for co-immunoprecipitation was previously described.¹⁵ In short, retinas were lysed in ice-cold lysis buffer (50 µL/retina [25 mM HEPES buffer, pH 7.4], 150 mM NaCl, 5 mM MgCl₂, 1% CHAPS detergent) containing protease inhibitor Halt Protease Inhibitor (Cat# 78420, Thermo Fisher Scientific) in each tube. The lysate was spun down at 14 000×g at 4°C for 30 minutes. Afterward, 10 µL of each tube was set aside for lysate control for immunoblotting. The remaining lysate was divided equally into two tubes to be incubated with 1 µL of MCT1 or MCT4 overnight at 4°C. Then 25 µL of Pierce Protein A/G Magnetic Beads (Cat#88802, Thermo Fisher Scientific) was added to the samples for 1 hour. The samples were isolated by using DynaMag-2 Magnet (Cat#12321D, Thermo Fisher Scientific). The samples were washed in lysis buffer, and then resuspended in 2×LDS sample buffer and analyzed by immunoblotting.

2.5 |. Tissue preparation and immunofluorescence

Mice were euthanized, and the eyes were enucleated and immediately placed into 4% paraformaldehyde (PFA) (Cat#15710, Electron Microscopy Sciences, Hatfield, PA) for 2 minutes. The eyes were placed in 1× PBS and the cornea was removed. Eyecups with the lens intact were placed in 1 mL of 4% PFA for 2 hours at room temperature on a rocker. The lens was then removed, and the remaining eyecup was washed in 1× PBS three times. The eyecup was put through a sucrose gradient from 5% to 30% in a stepwise manner. Samples were frozen in Neg-50 Frozen Section Medium (Cat# 6502, Thermo Fisher Scientific), and cryosections (10-µm thick) were collected and stored at –80°C.

For immunofluorescence confocal microscopy, frozen sections were brought to room temperature, and a hydrophobic barrier was drawn around the samples on the slide using an ImmEdge Hydrophobic Barrier PAP Pen (Cat# H-400, Vector Laboratories, Burlingame, CA). 1× PBS was used to wash away the Neg-50 Frozen Section Medium and sections were blocked in PBS with 5% BSA and 0.1% Tween 20 (PBST) for 1 hour at room temperature. Antibodies were diluted in PBST containing 1% BSA (see Table 2 for antibody dilutions), and sections were incubated with the primary antibodies overnight at 4°C. The slides were washed three times in PBST and then incubated with the appropriate secondary antibodies diluted in 1% BSA PBST for 30 minutes. Slides were then washed three times in PBST and once in PBS, then incubated with 4′,6-diamidino-2-phenylindole (DAPI) for 10 minutes to label nuclei. Slides were washed twice in PBS and coverslips were mounted with Gelvatol. Slides were imaged on an LSM 780 NLO laser scanning microscope (Carl Zeiss, Oberkochen, Germany) using ApoPlan ×63/1.4 objective and EC NeoPlan ×10/0.3 objective.

TABLE 2.

List of antibodies used in this study and dilution

Antibody	Company	Catalog	WB dilution	IF dilution
BSG(CD147)	Santa Cruz	Sc-9757	1:5000	1:500
MCT1	Philp Lab	(7–12)	1:2000	1:250
MCT3	Philp Lab	(7–12)	1:5000	–
MCT4	Philp Lab	(7–12)	1:2000	–
β-actin	Cell Signaling	4967	1:2000
β-tubulin	Millipore Sigma	T4026	1:5000	–
Cre	Millipore	MAB-3120	–	1:100
Cone Arrestin	Millipore	AB15282	–	1:100
ZO-1	Zymed Laboratories Inc.	61-7300	–	1:100
Bovine anti-goat IgG-HRP	Jackson Immuno Research	115-475-207	1:2000	–
Goat anti-rabbit IgG-HRP	Jackson Immuno Research	111-035-144	1:2000	–
Alexa Fluor Donkey anti Rabbit 488	ThermoFisher Scientific	A21206	–	1:500
Alexa Fluor Donkey anti Goat 555	ThermoFisher Scientific	A21432	–	1:500
Alexa Fluor Donkey anti Mouse 546	ThermoFisher Scientific	A10036	–	1:500

2.6 |. RNA and cDNA

Neural retinas were isolated and placed immediately into 1-mL TRIzol (Cat# 15596026, Thermo Fisher Scientific) after which they were homogenized with a 1-mL syringe with an 18-G needle (Cat# 305195, BD PrecisionGlide Needle) then with a 25-G needle (Cat#305124, BD PrecisionGlide Needle). RNA was extracted according to manufacturer specifications. RNA was quantified on NanoDrop (Cat#ND-1000, Thermo Fisher Scientific). RNA (1 µg) was reverse transcribed to 20-µL cDNA using EcoDry Premix ([oligo dT] Cat# 639543 Takara Bio USA, Mountain View, CA).

2.7 |. Quantitative PCR

cDNA of mRNA derived from 1 µg of total RNA was used for RT-PCR. mRNA (1 µg) was amplified from 1 µg of total RNA. qPCR was done using 10 ng of cDNA, and PowerUp SYBR Green Master Mix (Cat#A25742, ThermoFisher Scientific) on a QuantStudio 5 Real-Time PCR System (Cat#A28139, ThermoFisher Scientific). The PCR reaction was done according to manufacturer’s protocol, using primers listed in Table 2. In brief, reactions were heated to 50°C for 2 minutes and held to 95°C for 10 minutes. Samples were then denatured at 95°C. Cycle threshold (C_t) values were generated by the software and normalized to RPLP0, and 2^−(ΔCt) values were used to compare gene expression.

2.8 |. Flatmounts for RPE

The procedure for RPE flatmounts were performed as previously described.³ Eyes were isolated from euthanized mice and fixed in 4% PFA for 8 minutes in a 96-well plate, after which they were transferred to a well with 1x Dulbecco’s phosphate-buffered saline (DPBS) (Cat# 21–030-CM, Corning, Corning, NY). Extraocular muscles and connective tissues attached to each eye were removed using fine forceps and scissors. The anterior segment of the eye was removed using a razor blade and the retina was dissected using fine forceps. The posterior eyecup was fixed again in 4% PFA for 8 minutes and subsequently washed with 1× DPBS. The fixed eyecup was permeabilized in 0.3% Triton X-100 in 1× DPBS for 15 minutes and blocked with 5% BSA and 0.1% Triton X-100 for 1 hour. The eyecups were incubated in primary antibody overnight in a well in 1% BSA 0.1, %Tween 20, 0.1% Triton X-100 and washed in 1× DPBS (Table 2). Eyecups were then incubated at room temperature with secondary antibody in 1% BSA, 0.1% Tween20, and 0.1% Triton X-100 for 1 hour and then washed in 1× DPBS and transferred to a well with DAPI and 0.1% Triton X-100 for 1 hour. Eyecups were finally washed in 1× DPBS and were placed on glass slides. A hydrophobic pen was used to draw a circle around the eyecup, and 1× DPBS was used to slightly suspend the eyecup with the RPE side facing upward. To flatten the eyecup, 4–8 radial cuts were made, and a Kimwipe was used to gently wick away the 1× DPBS. Gelvatol was used as a mounting medium for the coverslip. Flatmounts were imaged using a Nikon (Melville, NY) Eclipse E800 fluorescent microscope using the Plan Fluor 10×/0.30 or 40×/0.75 objectives.³

2.9 |. Confocal scanning laser ophthalmoscopy (cSLO) and spectral domain optical coherence tomography (SD-OCT)

Mice were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) and eyes were anesthetized with 0.5% proparacaine HCl ophthalmic solution (NDC: 17478-263-12 Akorn, Lake Forest, IL). Pupils were dilated with 1% tropicamide eye drops (NDC: 17478-102-12 Akorn). Ocular eye shields^32,33 and Systane Ultra Lubricant Eye Drops (Alcon Laboratories Fort Worth, TX) were used to keep eyes hydrated. cSLO images were obtained using a Spectralis HRA+OCT (Heidelberg Engineering, Franklin, MA). Mice were positioned with the optic nerve in the center of the image using a 55° field of view (FOV) lens and imaged with two different modes, Infrared reflectance (IR) and blue autofluorescence (BAF). Following cSLO, SD-OCT imaging was conducted using a Bioptigen Envisu R2210 system (Leica Microsystems, Buffalo Grove, IL). A 50- FOV objective lens was used to image the posterior pole, which provided a 1.4-mm diameter view of the mouse retina. Orthogonal B-scans (1000 A-Scans/B-scan) of the horizontal and vertical meridians were collected with the optic disk centrally located and averaged over 15 frames. ImageJ 1.52i was used to analyze images for outer nuclear layer (ONL) thickness and inner segment and OS thickness (photoreceptor layer [PL]), and to quantify BAF images. Animals were imaged between 4 and 35 weeks of age.

All images on the cSLO were acquired with the auto-normalization activated, which provided the best contrast. The normalization feature performed a histogram stretch on every image collected normalizing to the brightest and darkest pixels in image and assigning them to the 255 and 0 gray intensity levels, respectively. As a result, images without substantial autofluorescence appear bright since the lipofuscin autofluorescence originating from the RPE is the brightest entity within the image. The background in the control and mutant images gets enhanced or suppressed, respectively, relative to the brightness of objects within the field of view. Mice without any appreciable autofluorescence signal will appear bright due to the auto-normalization correction, but in images with numerous bright objects, the background gets pushed down and appears darker. These bright objects are referred to as hyperfluorescent foci (HF).

2.10 |. Lactate assay

Krebs-Ringer-Bicarbonate (KRB) was made containing (98.5 mM NaCl, 4.9 mM KCl, 1.2 mM KH₂PO₄, 1.2 mM MgSO₄7H₂O, 20 mM HEPES, 2.6 mM CaCl-2H₂O, 25.9 mM NaHCO₃). All components were dissolved in deionized H₂O and 5% CO₂ was used to bring KRB to a 7.4 pH. KRB was filtered through a 0.22-µm membrane. Glucose was added to KRB solution and the final concentration was 5 mM. KRB with 5mM of glucose was aliquoted into 48-well plates (Cat# 3548, Corning, NY) and placed into a 37-C cell incubator under 5% CO₂ before dissection and the rest of KRB with 5 mM of glucose was placed on ice in a 35-mm dish for the retina dissections. Neural retinas were dissected from mice and immediately put in KRB with 5 mM of glucose. Single retina was placed in a well of a 48-well plate that was previously placed in the 37-C cell incubator. Lactate released was measured at 30 minutes using the Lactate Reagent (Cat# 73510, Trinity Biotech, Bray, Co Wicklow, Ireland) according to manufacturer protocol and graphed relative to control.

2.11 |. ERG—methods and analysis

Control and mutant mice were tested at ages ranging from 4 to 35 weeks. After overnight dark adaptation, mice were anesthetized with ketamine (80 mg/kg) and xylazine (16 mg/kg) diluted in 0.9% saline. Eyedrops were used to dilate the pupils (1% tropicamide [NDC: 17478-102-12], 2.5% phenylephrine HCl [NDC: 17478-201-15], 1% cyclopentolate [NDC: 17478-096-15], all Akorn) and to anesthetize the corneal surface (0.5% proparacaine [HCl NDC: 17478-263-12], Akorn). Three electrodes were used to record ERGs. A thin stainless steel wire served as the active lead³⁴ and was coated at the end with 1% carboxymethylcellulose. Two Grass needle electrodes (Astro-Med, West Warwick, RI) served as the reference and ground leads and were placed in the cheek or tail, respectively. Strobe flash stimuli were presented in an LKC (Gaithersburg, MD) ganzfeld, first to the dark-adapted eye (−3.6 to 2.1 log cd s/m²) and then superimposed upon a steady 20 cd/m² background field and after a 5-minute light adaptation period (−0.8 to 2.1 log cd s/m²). ERGs were band-pass amplified (0.03 to 1000 Hz), averaged, and then stored using an LKC UTAS E-3000 signal averaging system.

The amplitude of the a-wave was measured at 8 ms after flash onset from the prestimulus baseline. The leading edge of the a-wave to a 1.4 log cd s/m² flash was measured using the equation:

$$P3\left(i,t\right)=\left(1-\exp \left(iA{\left(t-td\right)}^2\right)\right)RmP3$$ (1)

where RmP3 is the maximum response amplitude, A is a measure of sensitivity, and td is the delay in phototransduction.³⁵

The amplitude of the b-wave was measured from the a-wave trough to the b-wave peak or, for low luminance flashes that do not elicit an a-wave, from the baseline to the peak of the b-wave. The function relating b-wave amplitude to flash luminance has two limbs.³⁶ We fit the first limb of this function using the Michaelis-Menten equation³⁷:

$$R/R_{\max }=L/\left(L+K\right)$$ (2)

where R is the amplitude of the a- or b-wave; R_max is the maximum amplitude of the a- or b-wave; L is the flash energy (log cd s/m²), and K is the flash energy that elicits an amplitude of half R_max (half-saturation coefficient).

2.12 |. Statistical analysis

Unpaired two-tailed Student’s t-tests were done between two samples to determine the P values. One-way ANOVA was done when comparing more than two genotypes, and two-way ANOVA was done when comparing more than two different variables. Bonferroni’s correction was done to determine the final P value. P ≤ .05 was considered significant. All data in figures are mean ± SE N ≥ 3.

3 |. RESULTS

3.1 |. Generation and validation of the cell-specific Bsg knockout mouse

There are two splice variants of the Bsg gene, Bsg1, and Bsg2 (Figure 1A). Bsg1 includes exon 1A, which encodes for an additional Ig-domain that is not present in Bsg2. Previous studies have shown that Bsg1 is preferentially expressed in photoreceptors, while Bsg2 is widely expressed in most other cells.³⁸ Bsg floxed mice were generated by insertion of floxed sites on either side of exon 1.²⁶ Targeted deletion of exon 1 ensured that neither splice variant of Bsg would be expressed (Figure 1A).

FIGURE 1.

Validation of mouse models with cell-specific deletion of Bsg in photoreceptors or RPE. A, Diagram of floxed Bsg gene and the splice variants Bsg1 and Bsg2. LoxP sites were placed on either side of exon 1 so crossing the Bsg^flox/flox mice with the cell-specific Cre recombinase transgenic lines resulted in loss of expression of both Bsg1 and Bsg2. Exon 1A, present in Bsg1 but not Bsg2, is highlighted in red. B, Western blot of detergent-soluble lysate from control retina, NRL^−/− retina, and control RPE untreated (left) or after PNGaseF treatment (right). Note that rBSG1 is absent from the cone-rich NRL^−/− retina and cone BSG1 (cBSG1) and rod BSG1 (rBSG1) from control retinas have different mobilities on SDS-PAGE in the untreated control samples but not after PNGaseF treatment. BSG2 between the neural retinas and the RPE also appear differentially glycosylated, but not after PNGaseF treatment. C, Western blot of detergent soluble retinal lysates from control, RodΔBsg, ConeΔBsg, and NRL^−/− mice shows the rod-specific deletion of rBSG1 in RodΔBsg and cone-specific deletion of cBSG1 in ConeΔBsg. Western blot confirms the deletion of BSG2 from RPEΔBsg RPE. In B) and C) blots are representative of SDS-PAGE and blots performed on lysates from at least three animals per genotype

BSG1 and BSG2 levels were examined in neural retina and RPE by preparing detergent soluble lysates from control or NRL^−/− mouse eyes for western blot analysis (Figure 1B). In control neural retinas, the most prominent BSG1 band had a molecular weight of ~55 kD, similar to what was previously reported.³⁸ A second lighter band was observed above the prominent BSG1 band and was the only BSG1 band detected in lysates prepared from the Nrl^−/− neural retina (Figure 1B left). The different mobilities of the BSG1 bands detected in rod-dominant control and cone-like photoreceptor dominant NRL^−/− neural retinas suggest the two bands observed in the control retinas correspond to a rod-specific isoform (rBSG1) and a cone-specific isoform (cBSG1). The differences in band intensity for rBSG1 and cBSG1 in control retinas were consistent with the ratio of rods to cones, as rod photoreceptors make up a majority of the photoreceptor population.^39,40 To determine whether the difference in mobility of rBSG1 and cBSG1 resulted from differential glycosylation, lysates from control and Nrl^−/− neural retinas were treated with PNGaseF to remove the N-linked oligosaccharides. After PNGaseF treatment, BSG1 from control and Nrl^−/− neural retinas migrated with the same molecular weight, indicating that BSG1 is differentially glycosylated in rod and cone photoreceptors (Figure 1B right). BSG1 was not detected in the RPE (Figure 1B). Differences were also found in the relative mobility of BSG2 between the neural retina and the RPE. After treatment with PNGaseF, the mobility of BSG2 was the same in the neural retina and the RPE (Figure 1B right).

To understand the importance of MCTs in supporting the metabolism of photoreceptors and the RPE, Bsg was selectively knocked out of rods (RodΔBsg), cones (ConeΔBsg), or RPE (RPEΔBsg) by crossing Bsg^flox/flox mice with cell-specific Cre transgenic animals, as described in the Materials and Methods. We confirmed the cell-specific deletion of Bsg by western blot analysis of lysates prepared from control and cell-specific knockout mice (Figure 1C). The control neural retinal lysates yielded bands for rBSG1, cBSG1, and BSG2. Western blot analysis of RodΔBsg neural retinal lysates showed the rBSG1 band was absent, while the cBSG1 and BSG2 bands remained. In ConeΔBsg lysates, rBSG1 and BSG2 were present, while the cBSG1 band was diminished. Control RPE had a single BSG2 band, which was not detected in RPE lysates prepared from RPEΔBsg mice. These western blots confirmed the cell-specific deletion of BSG in the RodΔBsg, ConeΔBsg, and RPEΔBsg mouse lines.

3.2 |. Characterization of the phenotype of RodΔBsg mouse

In the RodΔBsg neural retina, qPCR showed that Bsg1 was decreased, while Bsg2 and other members of the Ig superfamily were unchanged (Figure 2A), reflecting the expression of Bsg2 in other cells in the neural retina including endothelial, Müller glia, astrocytes, and inner retinal neurons,^14,38 and the expression of Bsg1 being restricted to photoreceptors as previously reported.³⁸ We used immunofluorescence confocal microscopy to examine the distribution of BSG and MCT1 in frozen retinal sections of eyes from control and RodΔBsg mice (Figure 2B). In the control retina, BSG and MCT1 co-localize in the inner segments, ONL, and the outer plexiform layer. When we examined the labeling pattern of BSG and MCT1 in RodΔBsg retinal sections, there was loss of BSG and MCT1 labeling of the plasma membrane of rod inner segments, while the heteromeric complex was still detected in the inner segments of cone photoreceptor cells (Figure 2B, asterisks). To confirm that the remaining BSG was from cone photoreceptors, RodΔBsg sections were co-labeled with BSG and cone arrestin (Supplemental Figure 1, asterisks). BSG labeling in the inner segments colocalized with cone arrestin, confirming BSG expression in cone photoreceptors of control and RodΔBsg. BSG staining in cones became more prominent in RodΔBsg compared to control, as rods account for a majority of the photoreceptors.^39,40 The remaining MCT1 and BSG labeling in the RodΔBsg ONL reflects the presence of these proteins in Müller cells and cones (Figure 2B). Some MCT1 staining was observed in inner segments lacking BSG (Figure 2B, arrows), suggesting that MCT1 synthesis continues in rods, but cannot be properly trafficked to the plasma membrane in the absence of BSG, as previously shown in vitro and in vivo.^{10,11,13–15}

FIGURE 2.

rBSG1 is required MCT1 and MCT4 in rod photoreceptor cells. A, qPCR showing reduced levels of Bsg1 in RodΔBsg retina without compensatory increases in levels of other Ig superfamily members. Bars indicate average ± SEM for N = 4 mice. B, Immunofluorescence of BSG (red) and MCT1 (green) of frozen sections of eyes from control and RodΔBsg mice. Loss of BSG1 resulted in loss of MCT1 in inner segments. MCT1 was still detected in rods (arrows) but was not trafficked to the plasma membrane (Scale bar = 25 µm). Asterisks indicate cones which retain both BSG1 and MCT1. Data are representative of N = 3 experiments. C, Western blots of detergent soluble lysates from RodΔBsg retinas confirms genetic deletion of Bsg from rod photoreceptors targets MCTs for degradation. Blots are representative of N = 8 mice. D, Western blot quantification of MCT1 and MCT4 compared to β-tubulin of control and RodΔBsg samples. E, Comparison of Log₂(CPM) values of Slc16a1 (MCT1) and Slc16a3 (MCT4) between rods and cones at P28 from GSE74660. F, Immunoprecipitation of MCT1 and MCT4 shows enrichment of rBSG1 in control as compared to RodΔBsg retina. Data are representative of N = 3 experiments

Western blot analysis of neural retina lysates from control and RodΔBsg mice showed a decrease in both MCT1 and MCT4 (Figure 2C,D). While our lab had previously demonstrated that MCT1 was expressed in photoreceptor cells, we were surprised to find MCT4 was also decreased in the retinal lysates prepared from RodΔBsg mice, as MCT4 was reported to be restricted to Müller glial cells and inner retina.⁴¹ This finding suggested that MCT4 was expressed in rod photoreceptors, so we queried publicly available RNA-seq data (GSE74660) for rod and NRL^−/− photoreceptors.⁴² The data showed high levels of Slc16a1 expression (MCT1) in both rods and cones, while Slc16a3 expression (MCT4) was higher in rods than in cones (Figure 2E).

To determine whether MCT4 protein was present in rod photoreceptor cells, we performed co-immunoprecipitation assays to determine if rBSG1 co-immunoprecipitated with MCT4. Detergent soluble retinal lysates were prepared from control and RodΔBsg mice and incubated with or without antibodies against MCT1 or MCT4. We found that rBSG1 co-immunoprecipitated with both MCT1 and MCT4 in control neural retina lysates, but not in RodΔBsg lysates (Figure 2F). These results confirm that rods express both MCT1 and MCT4 and that each transporter forms heteromeric complexes with rBSG1.

3.3 |. Progressive photoreceptor degeneration in RodΔBsg retina

In vivo imaging was used to monitor longitudinal structural changes in retinas of RodΔBsg mice. BAF-cSLO images of RodΔBsg retinas show an age-related increase in hyperfluorescent foci (HF) in the subretinal space (Figure 3A). HF have been correlated with activation and migration of microglia and inflammatory monocytes into the subretinal space.^43–45 The appearance of HF in the RodΔBsg retina correlated temporally with altered photoreceptor OS morphology and significant thinning of the ONL and PL layers, observed via SD-OCT (Figure 3B,D,E) and histology (Figure 3C). These changes were not observed at 4 weeks of age (Figure 3D,E,F) but were already significant at 9 weeks of age and became more pronounced with age. By 35 weeks of age, the RodΔBsg ONL and PL layer was only ~60% of control. HF were not seen in 30-week-old control retinas but were pan-retinal in RodΔBsg retinas at that age. We also noted HF in mice at 35 weeks expressing only the iCre transgene, which may indicate a toxic effect of Cre recombinase for older rods. H&E stained histological sections also revealed disorganized OSs and the presence of cells within the OS layer of the older RodΔBsg mice (Figure 3C) that likely correspond to the HF seen by AF-SLO and are presumed to be inflammatory monocytes based on positive staining for Iba1 (data not shown ).

FIGURE 3.

Age-related structural changes in RodΔBsg retina. A, Blue-light autofluorescence cSLO imaging of control and RodΔBsg eyes and F) quantification of hyperfluorescent foci (HF), number of HF increased with age in RodΔBsg retina (average (±SEM) of N ≥ 3 mice per genotype). B, SD-OCT images of control and RodΔBsg eyes. C, Histology of paraffin section of eyes from control and RodΔBsg mice (scale bar indicates 50 µm). D, E, quantification of thickness of ONL and PL layers from SD-OCT images. Data points indicate average (±SEM) of N ≥ 3 mice

3.4 |. Diminished ERGs in RodΔBsg mice

We used ERGs to evaluate overall retinal function in RodΔBsg mice. Figure 4A (left) presents representative ERGs obtained from RodΔBsg and control littermates at 17 weeks, an age where we see marked anatomical changes in the RodΔBsg retina (Figure 3). These responses were obtained under dark-adapted conditions and reflected primarily rod-mediated activity. The RodΔBsg responses contain the major ERG components, mainly reflecting the activity of rod photoreceptors (a-wave) or rod bipolar cells (b-wave). The amplitude of the overall RodΔBsg response is clearly reduced, with the magnitude of this reduction comparable across the stimulus range examined (Figure 4A right) and consistent with the loss of photoreceptors observed in vivo by SD-OCT and in histological sections (Figure 3).

FIGURE 4.

Age-related changes in ERGs of RodΔBsg retina highlights importance of lactate transporters on rod photoreceptor function. A, Representative dark-adapted ERGs (left) and summary luminance-response functions from control and RodΔBsg littermates at 17 weeks of age. Values to the left of each pair of waveforms indicates flash strength in log cd s/m2. Calibration indicates 500 µV and 100 ms. B, Values of Rmax and RmP3 from control and RodΔBsg littermates at the ages indicated. C, Comparison of the leading edge of the a-wave of 4-week-old mice. For each mouse, amplitude was normalized to the a-wave trough. D, Values of A from control and RodΔBsg littermates at the ages indicated. E, Representative light-adapted ERGs (left) and summary luminance-response functions from control and RodΔBsg littermates at 17 weeks of age. Values to the left of each pair of waveforms indicates flash strength in log cd s/m². Calibration indicates 100 µV and 100 ms. F, Values of Rmax for the cone ERG at the ages indicated. Data points indicate average (±SEM) for 4–7 mice

To follow ERG changes with age in the RodΔBsg mice, we fit Equations (1) and (2) to the leading edge of the a-wave and the b-wave luminance-response function, respectively. At 4 weeks of age, the a-wave amplitude parameter RmP3 was comparable in RodΔBsg and control mice (Figure 4B); RmP3 values of RodΔBsg mice decreased steadily at later ages (Figure 4B) with a time course that mirrored the loss of rod photoreceptors (Figure 3D). The b-wave amplitude parameter R_max was also reduced in RodΔBsg mice (Figure 4B), but the difference relative to control did not change with age to the extent seen for RmP3. At later ages the a-wave reductions were greater than the b-wave reductions, indicating that loss of MCT1 and MCT4 is more detrimental to light-stimulated OS responses than to the transmission of the visual signal from photoreceptors to bipolar cells, despite the loss of MCT1 and MCT4 in the outer plexiform layer of rod photoreceptors.

To examine the kinetics of the rod response, we superimposed the leading edge of normalized individual waveforms evoked by a high luminance flash (Figure 4C).⁴⁶ These waveforms were obtained from 4-week-old mice before the onset of ONL degeneration (Figure 3D). The kinetics of the RodΔBsg a-waves were slower than those of control responses, indicating that phototransduction gain was decreased in the absence of rBSG1 (Figure 4C) and a similar decrease was observed at all ages examined (Figure 4D). A similar analysis was not reported for the Bsg^−/− mouse.^18,19

In comparison to the changes noted for the dark-adapted ERG, cone ERG waveforms were similar between RodΔBsg and control mice at 17 weeks of age, and the cone ERG luminance response functions overlapped (Figure 4E). Cone ERGs of RodΔBsg mice decreased at later ages (Figure 4F), a change that is likely secondary to the progressive loss of rod photoreceptors (Figure 4D–F), which is seen in other models where the primary defect involves rods.⁴⁷

3.5 |. ERGs are not diminished in the MCT4^−/− mouse

Since rod photoreceptors were found to express both MCT1 and MCT4, we could not distinguish which of these might play a more significant role in defining the phenotype observed in RodΔBsg mice. To address this, we characterized the retinal phenotype of MCT4^−/− mice (MCT1^−/− do not survive⁴⁸). Western blot analysis showed that MCT4 was not detected in the MCT4^−/− neural retina, while levels of rBSG1, cBSG1, and MCT1 were unchanged (Figure 5A). Since MCT4 is associated with a high rate of glycolysis and lactate production,^49,50 we examined lactate efflux in neural retinas isolated from control, MCT4^−/−, and RodΔBsg mice. There was no difference in lactate efflux between control and MCT4^−/− mice (Figure 5B). In comparison, lactate efflux was reduced by 20% in the RodΔBsg retina (Figure 5B). We examined overall retinal function in the MCT4^−/− mice at 6 months of age, where a clear phenotype was observed in RodΔBsg animals (Figures 3,4). ERGs of MCT4^−/− mice were comparable to control under both dark-adapted (Figure 5C) and light-adapted (Figure 5D) conditions. These results indicate that the RodΔBsg phenotype does not reflect a dependence of rods on MCT4 and suggests that the phenotype is indicative of an essential role for MCT1 in rod photoreceptors.⁵¹

FIGURE 5.

Retinal phenotype of the MCT4^−/− mouse A) Western blot analysis of control, MCT4^−/−, and RodΔBsg retinas. Blots are representative of N = 3 mice. B) Lactate efflux from control (N = 6), MCT4^−/− (N = 3), and RodΔBsg retinas (N = 3). Luminance-response functions for the major components of the C) dark-adapted and D) light-adapted ERGs obtained from control and MCT4^−/− mice. Data points indicate average (±SEM) for 12 mice

3.6 |. Slow and incomplete loss of cone ERG amplitude in ConeΔBsg mice

Due to the low density of cone photoreceptors in the mouse retina, western blot and IHC did not easily detect changes in MCT1 levels in ConeΔBsg mice (Figure 1C). To visualize the knockout of Bsg from cone photoreceptors by the Opn1mwcre transgene, ConeΔBsg mice and RodΔBsg mice were crossed to generate Cone/RodΔBsg mice. We co-stained BSG and cone arrestin, to confirm cone-specific knockout in Cone/RodΔBsg histological sections by the Opn1mwcre (Supplemental Figure 1). In comparison to control and RodΔBsg sections, there was no co-localization of BSG and cone arrestin in these sections, confirming cone-specific knockout of Bsg by the Opn1mw-cre. This finding, coupled with the western blot data (Figure 1C), demonstrates the Opn1mw-cre transgene effectively deleted Bsg expression from cone photoreceptors.

To confirm the deletion of Bsg from cone photoreceptors resulted in loss of MCT1 in cone photoreceptors, we compared the distribution of BSG and MCT1 labeling by immunofluorescence confocal microscopy on histological sections from RodΔBsg mice and Cone/RodΔBsg. BSG and MCT1 co-localized in cone inner segments of RodΔBsg mice (Figure 6A, asterisks), but were not detected in the inner segments of rods or cones in sections of eyes from the Cone/RodΔBsg mice (Figure 6B). The remaining staining of BSG and MCT1 in the ONL and inner segments of Cone/RodΔBsg mice must be from Müller cells, as Bsg was deleted from both rods and cones (Figure 6B, Supplemental Figure 1).

FIGURE 6.

Age-related changes in ERGs of ConeΔBsg retina. A, Immunofluorescence confocal microscopy of frozen sections of RodΔBsg and B, Cone/RodΔBsg eyes with BSG (red) and MCT1 (green) confirms BSG and MCT1 are not expressed in cone photoreceptors of Opn1mwcre transgenic mice (scale bar = 20 µm). C, Representative dark-adapted ERGs (left) and summary luminance-response functions (right) obtained from control and ConeΔBsg littermates at 17 weeks of age. Values to the left of each pair of waveforms indicates flash strength in log cd s/ m². Calibration indicates 500 µV and 100 ms. D, Values of Rmax and RmP3 from control and ConeΔBsg littermates at the ages indicated. E, Representative light-adapted ERGs from ConeΔBsg at 17 weeks of age. Values to the left of each pair of waveforms indicates flash strength in log cd s/m². Calibration indicates 100 µV and 100 ms. F, Values of Rmax for the cone ERG at the ages indicated. Data points indicate average (±SEM) for 6–9 mice

To examine the functional impact of deleting Bsg from cones, we used ERGs, which provide a sensitive measure of cone dysfunction,^52–54 to examine the ConeΔBsg mice. Under dark-adapted conditions, we noted no difference in the overall ERG waveform or amplitude (Figure 6C,D), indicative of spared rod pathway function at all ages examined. Under light-adapted conditions that isolate the cone pathway, ERGs of ConeΔBsg mice were only slightly reduced as compared to those obtained from control littermates at 17 weeks of age (Figure 6E). Across the age range examined, we noted that responses of ConeΔBsg mice were comparable to control at 4 and 9 weeks of age and were only slightly reduced at the later ages examined (Figure 6F). The reduction did not appear to become more pronounced with age, and Rmax values for the cone ERG luminance-response functions retained 84% of control values at 35 weeks of age (Figure 6F).

3.7 |. Characterization of RPEΔBsg mice demonstrates an essential role for MCT1 and MCT3 maintaining the integrity and function of RPE and photoreceptor cells

The RPE expresses BSG2 (Figure 1B), which is required for maturation and trafficking of MCT1 and MCT3 to the apical and basolateral RPE membranes, respectively.^10,11,14 BSG2 is targeted to the basolateral membrane of the RPE through its association with MCT3 and to the apical membrane of the RPE through its association with MCT1, that lacks a basolateral sorting signal.¹¹ Since the Best1-Cre used here results in variable levels of Cre expression,^3,29 we restricted our analysis to RPEΔBsg mice with high levels of Cre expression (Figure 7A). Western blot analysis of RPE/choroid lysates from RPEΔBsg mice showed reduced levels of BSG2 as compared to control, as well as reduced levels of MCT1 and MCT3 (Figure 7B). At 17 weeks of age, SD-OCT images did not show thinning of the ONL or inner nuclear layer (INL) but showed areas of retinal detachment and OS disruption (Supplemental Figure 2). In older mice, at 43 weeks of age, BAF-cSLO showed increased HF in the RPEΔBsg retina (Figure 7C) while SD-OCT imaging revealed areas where the RPEΔBsg was detached (Figure 7D, red asterisk). RPEΔBsg mice had a thinned ONL, disrupted and hypertrophic OSs, as well as inflammatory cells that were found in the subretinal space (Figure 7E). RPE flatmounts from RPEΔBsg revealed multiple abnormalities, including enlarged and irregularly shaped cells consistent with RPE stress⁵⁵ (Figure 7F).

FIGURE 7.

RPE-specific deletion of BSG1. A, Immunofluorescent labeling of flatmounts of RPE from control and RPEΔBsg mice label with antibodies to BSG (green) and Cre (red) (scale bar = 100 µm). B, Western blot of lysates from control and RPEΔBsg RPE. C, BAF-cSLO demonstrates the increase in hyperfluorescent foci in RPEΔBsg. D, SD-OCT images of control and RPEΔBsg retina in 10-month-old RPEΔBsg. Red asterisk indicates retinal detachment. E, Representative H&E stained retinal cross sections of control and RPEΔBsg at 26 weeks of age (scale bar = 25 µm). F, Disorganized and enlarged RPE cells in RPEΔBsg flatmounts stained with ZO-1. All images are representative of N = 3 experiments (scale bar = 100 µm). Representative dark-adapted G) and light-adapted H) ERGs (left) and summary luminance-response functions right obtained from control and RPEΔBsg littermates at 17 weeks of age. Values to the left of each pair of waveforms indicates flash strength in log cd s/m². Calibration indicates 500 µV and 100 ms. Data points indicate average (±SEM) for 30 control and 6 RPEΔBsg mice

ERGs were obtained from RPEΔBsg and control mice to determine the impact of Bsg deletion from the RPE on outer retinal function. RPEΔBsg mice had reduced ERGs under both dark-adapted (Figure 7G) and light-adapted (Figure 7H) conditions. These changes were noted in the youngest mice examined (4weeksold) and became more pronounced in older mice. Even at 17 weeks of age where there was no loss of ONL or INL, there was a severe reduction in ERG a- and b- wave amplitude, highlighting the importance of lactate transporters in the retina. Of all of the models examined in this project, the ERG changes noted in RPEΔBsg most closely matched those reported in the Bsg^−/− model highlighted in Figure 8.^{18, 19}

FIGURE 8.

Comparison of ERG a-wave changes in Bsg^−/− cell-specific Bsg knockout mice. Summary of a-wave changes across the age range examined for RodΔBsg (Figure 4B), ConeΔBsg (Figure 6D), RPEΔBsg (Figure 7G), and Bsg^−/−19

4 |. DISCUSSION

Mice with a global genetic deletion of Bsg are infertile and have memory, immune, and visual deficits.¹⁸ In the Bsg^−/− mice, light- and dark-adapted ERGs are severely reduced as early as postnatal day 17, although photoreceptor cell degeneration is not apparent until after 8 weeks of age.^18,19 Previous studies from our lab linked the early visual deficits in the Bsg^−/− mice to impaired trafficking of MCTs to the plasma membrane, resulting in impaired lactate transport within and out of the retina.¹⁴ Findings from these studies demonstrated the overall importance of MCTs to maintaining the function and viability of photoreceptor cells but did not provide insight into which cells were driving the phenotype, as Bsg is expressed throughout the retina.

In our current study, we found the loss of photoreceptor function of the RPEΔBsg mouse most closely resembled the previously reported phenotype of the Bsg^−/− mice. In Figure 8, ERG a-wave data from the Bsg^−/− mice¹⁹ was compared with the cell-specific Bsg knockout mice reported here. The reason that the deletion of Bsg from the RPE resulted in an early reduction of ERGs that mirrored the Bsg^−/− phenotype^18,19 was most likely because of alterations to the microenvironment due to the inability of the RPE to regulate the lactate levels in the subretitinal space, impacting pH and osmolarity of the subretinal space (Figure 8 and Supplemental Figure 2). In contrast, the phenotypes of RodΔBsg and ConeΔBsg mice appeared much later and were less severe than was reported for Bsg^−/− mice.^18,19 The RodΔBsg showed early reduction of ERG activity at 9 weeks of age, with the reduction in the a-wave correlating to rod photoreceptor death. These findings suggest that the dark-adapted ERG phenotype of the Bsg^−/− was not due to energy deficits, but rather from alterations to the RPE. Additionally, ConeΔBsg did not show early loss of light-adapted (cone-driven) ERGs, suggesting that unlike rods, cone photoreceptors do not rely on lactate transport to maintain metabolic homeostasis and that ERG changes reported for Bsg^−/− mice^18,19 were driven by changes in the RPE.

4.1 |. The loss of Bsg in the RPE closely mimics the Bsg^−/− phenotype

The RPE supports photoreceptor cell function including the visual cycle and the regulation of the transport of metabolites in and out of the subretinal space to meet the metabolic demands of the outer retina.^1,56 Glucose is the primary metabolic substrate of the outer retina where it is metabolized predominantly through aerobic glycolysis.⁴ Excess lactate contributes to the metabolic coupling between the outer retina and the underlying RPE. The lactate generated by the retina is taken up by the RPE via MCT1, and then is either utilized for OXPHOS by the RPE to spare glucose for the outer retina^3,24,57 or it is transported across the basolateral membrane of the RPE by MCT3 to the choroidal capillaries.¹⁰ The transepithelial transport of lactate out of the retina by the RPE maintains the lactate levels, pH, and osmolarity of the subretinal space.^{10, 21, 22}

In RPEΔBsg mice, the loss of both MCT1 and MCT3 would be expected to prevent the transport of lactate, H⁺, and H₂O out of the subretinal space, thus altering the microenvironment of photoreceptors.²² The retinas of RPEΔBsg exhibited a decrease in both rod and cone function as early as 4 weeks of age. At 17 weeks of age, the ONL and INL did not show changes in thickness, suggesting the ERG phenotype seen in RPEΔBsg and the Bsg^−/− mice was driven by the changes in the microenvironment.^10,21,22 In vivo imaging by BAF-cSLO and histological sections showed invasion of immune cells into the subretinal space and disruption of OS structure. SD-OCT images from 17- and 43-week-old mice showed retinal detachment, consistent with fluid accumulation in the subretinal space as a consequence of the loss of MCTs, which coordinate the transport of fluid and lactate out of the subretinal space.²¹

In our previous study of the Slc16a8^−/− (MCT3^−/−) mouse, the morphological and functional changes were not as severe as the RPEΔBsg mice,¹⁰ which suggests that MCT1 expressed in the apical membrane of RPE, plays a significant role in regulating the lactate and H⁺ levels in the subretinal space. MCT1 may plays an important role in maintaining the close association between the RPE and photoreceptor cells. SD-OCT imaging of the RPEΔBsg mice consistently showed detachment between the RPE and photoreceptors. This wasn’t observed in histological sections of Slc16a8^−/− mice, suggesting that MCT1 transport activity contributes to the movement of fluid out of the subretinal space.^21,22 This hypothesis is consistent with a recent analysis of GWAS data linking SNPs in Slc16a1 (MCT1) with an increase in macular thickness. The same GWAS study, implicated MCT3 in macular thickening, as well as age-related macular degeneration (AMD)(GCST006976).⁵⁸ Our current studies and the finding from the GWAS study exemplify the importance of MCT1 and MCT3 for maintaining homeostasis in the outer retina.

The combined loss of MCT1 and MCT3 from the RPE may have a more severe phenotype than the Slc16a8^−/− mice due to the multiple roles of MCT1 in the RPE. Lactate uptake from the subretinal space through MCT1 is utilized by the RPE to fuel OXPHOS, so glucose is spared for the outer retina. The RPE relies on OXPHOS to maintain its differentiation, as disruption of RPE mitochondria, or the TCA cycle enzyme has been linked to dedifferentiation of the RPE.^23,24 Loss of MCT1 in the RPEΔBsg mice could lead to nutritional deficits in the RPE, altering its oxidative capacity and shifting the balance between OXPHOS and glycolysis. This could lead to atrophy of the RPE and alterations to RPE polarity, resulting in fluid accumulation in the subretinal space. In order to clarify the role of MCT1 in the RPE, future studies will characterize the retinal phenotype of MCT1 knock out of the RPE.

4.2 |. The expression of MCT4 and glycolytic enzymes in rod photoreceptors

It has been reported in the literature that photoreceptor cells are dependent on aerobic glycolysis to support their anabolic and catabolic metabolism required for visual transduction and OS renewal.⁵⁹ Key enzymes involved in aerobic glycolysis, such as HK2, PKM2, and LDHA, have been found to be expressed in photoreceptors (Figure 2D,E).^59–64 In this study, we found rod photoreceptors express both MCT1 and MCT4, which are required for aerobic glycolysis.⁴⁹ Cone photoreceptors may also express both MCT1 and MCT4, but this will need to be further validated.

Previously, Slc16a3 (encoding MCT4) expression had not been reported in neurons except under stressful conditions, for example, following ischemic stroke.⁶⁵ To our knowledge, this is the first time Slc16a3 expression has been reported in neurons under homeostatic conditions. Characterization of MCT4^−/− suggested that retinal MCT4 may not be needed under normal conditions, as ERG and lactate efflux did not change (Figure 5). Slc16a3 (MCT4) has been shown to be under the transcriptional control of HIf1α in different tissues.^66,67 We speculate that due to the dense population of mitochondria in the inner segments, the mitochondria may generate metabolic intermediates or induce areas of low oxygen in the ONL that may stabilize HIf1α resulting in an upregulation of MCT4 expression, along with other glycolytic genes.^68,69

In retinal disease models, MCT4 may play a larger role since photoreceptors under stress are more reliant on aerobic glycolysis for survival.^1,47,61,64 It has been suggested that upregulation of aerobic glycolysis in photoreceptors is required to preserve rod and cone photoreceptors under stressful conditions.^61,64 MCT4 may be required for lactate efflux under stress conditions, as MCT4 can export lactate in high-lactate environments.⁵⁰ It would be interesting to see the response of the retina in the MCT4^−/− mouse under conditions of retinal stress, for example, after detachment or pharmacological insults, to understand the role of MCT4 in the retinal stress response.

4.3 |. Lactate transporters are important to maintain rod photoreceptor activity and viability

Lactate is generated as the end-product of aerobic glycolysis, so we expected to find a decrease in lactate efflux in retinas from RodΔBsg mice. Lactate efflux measured in ex vivo assays from RodΔBsg retinas was only 20% less than the controls. Since rod photoreceptor cells constitute over 50% of the cell population in the neural retina,^39,40 we expected a higher decrease from RodΔBsg retinas. Similarly, the deletion of HK2 or PKM2 from rod photoreceptors resulted in a decrease in lactate efflux of less than 20% and 14%, respectively.^61,64 This suggests that rod photoreceptors may not be the primary lactate-producing cells in the neural retina, and that they both produce and utilize lactate, as previously reported.²⁰

Despite the loss of lactate transport in RodΔBsg mice, the retinas developed normally as measured by OCT and ERGs at 4 weeks of age. However, the kinetics of the leading edge of the ERG a-wave were slowed (Figure 4C). Since rod photoreceptors are unable to either export or import lactate, this could be and early sign of metabolic dysregulation that is then followed by cell death (Figure 3D). Imported lactate could serve as a source of carbon for the TCA cycle that could be utilized to generate intermediates required for rod photoreceptor function.²⁰ Alternatively, the inability to export lactate could cause a buildup of lactate and H⁺ in rods, leading to a decrease in the flux of glycolysis and an increase in the generation of reactive oxygenated species, causing cellular stress and eventual cell death. It is likely that a contribution of both scenarios are involved, as most cells rely on both OXPHOS and glycolysis, balancing the redox state and the energetic needs to maintain its metabolic homeostasis.⁷⁰

It is interesting to note that while RodΔBsg showed an early functional phenotype at 4 weeks of age, ConeΔBsg exhibited only a mild phenotype at 17 weeks of age. This suggests that cone photoreceptors do not rely on lactate transport and aerobic glycolysis to maintain metabolic homeostasis. Based on findings from this study, and those from previous studies where metabolic enzymes were deleted in rods or cones or RPE,^3,61–63,71 it is tempting to speculate that rod photoreceptors are more sensitive to changes in aerobic glycolysis, while cone photoreceptors are reliant on OXPHOS and other substrates to maintain their metabolic needs.⁷² This idea is consistent with our recent finding that glucose deprivation in the outer retina had a greater impact on rod than cone viability.³ Future studies will investigate the possible metabolic differences between rod and cone photoreceptors and how the RPE meets the metabolic demands of both rods and cones.

5 |. CONCLUSIONS

This study highlights the importance of MCTs in the RPE for maintaining lactate homeostatsis in the outer retina to support the function and viability of photoreceptors.

Abbreviations:

BSG

basigin

ERG

electroretinogram

INL

inner nuclear layer

MCT

monocarboxylate transporter

OCT

optical coherence tomography

ONL

outer nuclear layer

OS

outer segment

RPE

retinal pigment epithelium