A mutation in Bicdl1 (BICD family-like cargo adaptor 1) leads to mild outer retina thinning and late retinal dysfunction in a mouse model.

Animals

All Bicdl1 KI (Bicdl1 knock-in) and wild-type (WT) male and female mice used in this study were on a C57BL/6J background. The number and ages of mice used are indicated for each experiment. All animals were maintained in a barrier animal facility at the University of Texas Southwestern Medical Center (UTSW Medical Center) on a 12-h -on/12-h -off light cycle and had free access to food and water. Before all procedures, the mice were anesthetized by intraperitoneal (IP) injection of a ketamine-xylazine cocktail (100 mg/kg - 5 mg/kg) and the eyes were dilated with a mixture (3:1) of tropicamide 1% solution (Alcon Laboratories, Inc. Fort Worth, TX) and phenylephrine hydrochloride 2.5% solution (Alcon, Inc. Lake Forest, IL). In the forward genetics pipeline, N-ethyl-N-nitrosourea (ENU) was used to induce random mutations in C57BL/6 founder male mice, which were then bred to C57BL/6J females (see below). Bicdl1 knock in (KI) mice were generated using a CRISPR-Cas9 genome editing technique (details below). The study protocols and procedures were approved by the UTSW Medical Center Institutional Animal Care and Use Committee (IACUC, animal protocol # 2015-G100937). We followed all guidelines released by the National Institutes of Health (NIH) and the Association for Research in Vision and Ophthalmology (ARVO) for animal care and animal use in ophthalmic and vision research.

Forward genetic pipeline: ENU mutagenesis and whole-exome sequencing

In our forward genetic pipeline, for each pedigree, we used ENU to induce random mutations in a C57BL/6J founder male mouse (generation 0, G0). The founder was bred with C57BL/6J females to produce G1 mice. Whole-exome sequencing (WES) was then performed in male G1 mice to identify all exonic mutations. These G1 males were then crossed with C57BL/6J females to generate G2, and G2 females were backcrossed with G1 males to generate G3 mice (Figure 1B in Wang et al.) [17]. To establish zygosity in G2/G3 mice, genotyping was performed at each of the mutated loci identified in G1 mice. Optical coherence tomography (OCT) imaging was then used to screen G2 and G3 mice for retinal pathology (Appendix 1). Linkage Analyzer software was then used to correlate the magnitude of quantitative OCT phenotypes with the genotype at each mutation site for all mice in the pedigree [17-21]. The output of this automated mapping is a Manhattan plot of genotype-phenotype associations. In the Manhattan plot, −log₁₀ p values (y-axis) for an association to a given phenotype are plotted versus the chromosomal positions of each mutations (x-axis) that had been identified in the G1 founders of each pedigree (Figure 1). In these plots, mice homozygous for the reference or WT allele are referred to as REF, mice heterozygous for the reference allele and variant allele are referred to as HET, and mice homozygous for variant or mutant allele are referred to as VAR. Mice that had not undergone mutagenesis (naive C57BL/6J mice) are referred to as WT (Figure 1).

Generation of Bicdl1 KI mice using the CRISPR-Cas9 system

For easy reference, we arbitrarily named the mutated Bicdl1 allele identified in the mutagenesis pipeline the “bargain allele”. CRISPR-Cas9 technology was used to generate Bicdl1 KI mice expressing a replacement allele that precisely replicates the bargain allele (Appendix 2). This allele contains a single base substitution of guanine to adenine at coordinate 115,808,212 bp of mouse chromosome 5 (Genome Reference Consortium Mouse Build 39), corresponding to the first nucleotide of Bicdl1 intron 4 (transcript variant 1 [NCBI reference sequence NM_001080808]; genomic sequence context: 5′-aaagacctacagGtactgggcagagaggct-3′, mutation site in uppercase). For both the bargain allele and the CRISPR-generated replacement allele, the predicted translation is a null protein product. Specifically, a single base pair substitution affecting a cryptic splicing site in intron 4 results in the insertion of 45 nucleotides from intron 4, leading to a shifted protein product beginning after amino acid 306 of the protein (which is normally 577 amino acids in length), and terminating after the inclusion of eight aberrant amino acids.

Bicdl1 KI mice harboring a replacement allele identical to the bargain allele (Appendix 2); mutated G to A at the splice donor site at coordinate 115,670,153 of GRCm38) on a C57BL/6J background were generated using CRISPR-Cas9 reagents at the Transgenic Center of UTSW Medical Center. The position of the G base to be mutated is 1nt from the 3′ end of exon 4 (ENSMUST00000055408.13 Bicdl1-201). The single guide RNA (sgRNA) sequence used was 5′TGA CAA AGA CCT ACA GGT ACT GG 3′. The guide was selected using the CRISPOR Design Tool. crRNA and tracRNA were annealed and mixed with Cas9 protein to form a ribonucleotide protein complex (RNP). The HDR template was 158nt with a silent mutation at the PAM site (TGG to TCG) to prevent the gRNA from cutting back the HDR knock-in template. The sequence of the HDR template was: 5′-CTG GAG CAT CGG CTC AGC GCC ACC TTG GAG GAG AAT GAC CTG CTT CAA GGG ACT GTG GAG GAG CTG CAG GAC CGG GTG CTG ATC CTA GAA AGG CAG GGC CAT GAC AAA GAC CTA CAG ATA CTC GGC AGA GAG GCT ACT GCA TAG AGC TAG GGC CAG AC-3′.

All reagents were synthesized by Integrated DNA Technologies (IDT; Coralville, IA). The mix of CRISPR reagents, RNP complex, and HDR template was delivered into the cytoplasm of fertilized one-cell eggs isolated from super-ovulated females via cytoplasmic injection or electroporation using a NEPA21 electroporator (Nepa Gene Co., Ltd., Chiba, Japan), with a concentration range of 50- 400 ng for Cas9 protein, gRNA, and HDR template. The surviving eggs were transferred into the oviducts of day 0.5 pseudo-pregnant recipient ICR females (Envigo, Indianapolis, IN) to produce putative founder mice. To generate the Bicdl1 KI line, we first crossed a Bicdl1 KI founder male mouse with a female C57BL/6J mouse. We then crossed the resulting Het females with male C57BL/6J mice again to generate F2 pups. After this, we used a Het × Het breeding approach to generate Bicdl1 WT and KI mice.

OCT acquisition and retinal thickness analysis

Mice were anesthetized one at a time, and the pupils were dilated before each experiment. GenTeal gel (Novartis, East Hanover, NJ) was applied to each eye to prevent corneal dryness. A Micron-OCT2 system (Phoenix-Micron Inc., Bend, OR) was used to acquire in vivo image-guided OCT images of both eyes of Bicdl1 KI mice and WT littermates. After gently contacting the gel-covered cornea with the camera eyepiece and centering the fundus image on the optic nerve head (ONH), we focused on the retina to acquire a B-scan. We obtained a half-size horizontal line scan superior to the ONH at a distance of two disc diameters from the center of the ONH. Total retinal thickness (TRT, measured from Bruch’s membrane to internal limiting membrane), outer nuclear layer (ONL) thickness, inner nuclear layer (INL), ganglion cell complex (GCC; top of the internal limiting membrane [ILM] to the top of the INL), and outermost retinal thickness (ORT, measured from Bruch’s membrane to external limiting membrane) were obtained from analysis of each OCT image using ImageJ (NIH, Bethesda, MD) as described previously [21,22]. For each parameter, three measurements were taken: one at the center of the OCT image and two at distances of 100 microns on either side of the center. These three measurements were averaged for statistical comparisons between Bicdl1 WT and KI mice. The number of mice used at different age groups is summarized as follows: 3-week-old (n = 7 WT and 5 KI), 2-month-old (n = 16 WT and 20 KI), 4-month-old (n = 16 WT and 17 KI), 6-month-old (n = 14 WT and 17 KI), 9-month-old (n = 19 WT and 19 KI), 12-month-old (n = 9 WT and 10 KI), and 15-month-old (n = 8 WT and 8 KI) per group.

Fundus photography and fundus spot grading

A Micron IV Retinal Imaging Microscope (Phoenix-Micron Inc., Bend, OR) was used to take color fundus images of both Bicdl1 WT and KI mice, as previously described [19,21]. Fundus photography and fundus spot grading were obtained at the following ages: 2-month-old (n = 16 WT and 20 KI), 4-month-old (n = 15 WT and 17 KI), 6-month-old (n = 10 WT and 14 KI), 9-month-old (n = 10 WT and 9 KI) and 15-month-old (n = 8 WT and 8 KI). The fundus photos were graded for the accumulation of white/yellow fundus spots using a modified version of a previously described grading system [20,21]. The scale is as follows: Grade 0: no spots, grade 1: 1-3 spots, grade 2: 4-10 spots, grade 3: 11 spots - up to a quarter of retina covered with spots, grade 4: up to half of the retina covered with spots, grade 5: more than half of the retina covered with spots. A score is recorded for each eye, and the scores from both eyes are added to give the final mouse score.

A separate Spectralis® OCT (Heidelberg Engineering, Heidelberg, Germany) was used to examine the anatomical correlate for fundus spots. Images were acquired according to the manufacturer protocols.

Preparation of retinal sections for histology, TUNEL assay, and immunohistochemistry

To prepare histological specimens, we used a modified freeze-substitution protocol [23], which helped us preserve the retinal anatomy and cellular morphology in retinal sections. Mice were anesthetized by IP injection of a ketamine/xylazine cocktail, and their left eyes were enucleated and immediately fresh-frozen in liquid nitrogen-cooled isopentane. The frozen eye samples were transferred to a pre-cooled freeze substitution solution of methanol 97%/ acetic acid 3% (v/v) and stored in a −80 °C freezer for at least 48 h. The eyes were gradually warmed to room temperature, transferred to 100% ethanol, and submitted to the UTSW Medical Center Histo Pathology Core for paraffin embedding using routine techniques. Retinal cross sections (5 µm) were prepared for hematoxylin/eosin (H&E) staining, terminal deoxynucleotidyl transferase-mediated UTP end labeling (TUNEL) assay, and immunohistochemistry staining with caspase-3, rhodopsin, cone arrestin, and Iba1.

H&E staining was performed on mouse sagittal brain sections and retinal cross-sections using standard automated stainer procedures. Briefly, following appropriate fixation for each tissue, samples were paraffin-processed, embedded, and sectioned by members of the UT Southwestern Histo Pathology Core. Paraffin sections were prepared concomitantly and examined by dark-field microscopy. Sections were stained for routine histopathologic evaluation by regressive H&E on a Sakura Prisma Plus x-y-z robotic stainer (Sakura Finetek USA, Inc., Torrance, CA) using Leica Selectech reagents (Hematoxylin 560 and Alcoholic Eosin Y 515, Leica Microsystems, Wetzlar, Germany). The H&E sections were imaged using a 20x objective lens (NA 1.0) on both sides of the ONH using a Leica DM2000 Upright Compound microscope (Leica Microsystems). The ONL thickness (in micrometers) and number of nuclei per ONL column (average of three columns) on the H&E images were measured at 500 µm intervals on both sides of the ONH (starting 300 µm from the edge of the ONH) using ImageJ.

The TUNEL assay was performed using the DeadEnd Fluorometric TUNEL system (Appendix 3) following the supplied protocol and previously published work [24]. Images were acquired using a Leica TCS SP8 confocal laser scanning microscope (Leica Microsystems) with a 25x objective lens (NA 1.0). The TUNEL-positive cells were quantified as the total number of TUNEL⁺ cells per mouse on the acquired images.

For caspase-3 staining, the retinal sections were deparaffinized and subjected to heat-induced epitope retrieval in citrate buffer (pH 6.0) at 95°C for 20 min. Following blocking of endogenous peroxidases using 3% hydrogen peroxidase and secondary antibody host-serum affinity using 5% normal goat serum (S-1000; Vector Laboratories, Newark, CA), sections were incubated at 4 °C overnight with a rabbit polyclonal antibody against cleaved caspase-3 (Appendix 3). Detection involved incubating the sections with a biotinylated secondary goat anti-rabbit IgG (H+L) (Appendix 3), and amplification with avidin-biotin complex and streptavidin peroxidase (SA-5004-1; Vector Laboratories). Visualization was performed using 3,3′-diaminobenzidine (DAB, Appendix 3), followed by nuclear counterstaining with hematoxylin. Images were acquired using a Keyence BZ-X1000 microscope (Keyence Corporation, Osaka, Japan) with a 20x objective lens (NA 0.75) and the caspase-3-positive cells were quantified.

Rhodopsin staining was performed on 9-month-old (n = 4 WT and 5 KI) and 19-month-old Bicdl1 (n = 5 WT and 5 KI) mice. Cone arrestin and microglia Iba1 staining were performed on 19-month-old Bicdl1 (n = 5 WT and 5 KI) mice. The additional paraffin-embedded retinal sections were deparaffinized, and antigen retrieval was performed using the heat-induced epitope retrieval (HIER) method [25] by heating in citrate buffer (pH 6.0) in a 95°C water bath for 30 minutes. The sections were then blocked and incubated overnight with anti-Rhodopsin or anti-cone arrestin antibody, followed by an appropriate secondary antibody (Appendix 3). DAPI was used for nuclear counterstaining. All sections were imaged using a 25x objective lens (NA 1.0) on a Leica TCS SP8 confocal laser scanning microscope (Leica Microsystems). Images of rhodopsin-stained midperipheral retina were measured on both sides of the ONH. The total area of rhodopsin staining was measured in ImageJ and divided by the total length of the retina in every image, thus providing a measure of the average thickness of the rhodopsin staining over the entire section. We report values normalized to WT. Cone arrestin-stained images were taken at the central region, mid-periphery, and periphery on both sides of the ONH. In sum, a total of 6 fields were imaged for each mouse (three antero-posterior regions/side x 2 fields/region). For analysis, fields were grouped according to retinal region. Specifically, two peripheral fields were captured per eye (one per side), and with five eyes from five individual mice per experimental group, a total of ten peripheral fields were analyzed for each mouse line (similar for the central and midperipheral fields). We counted all cone-arrestin+ cells in each field. Retinal sections were examined for Iba1+ subretinal microglia and imaged with the same microscope and settings.

RPE and retina flat mount preparation for microglia/macrophage staining and analysis

RPE and retina flat mounts were prepared from 9- and 15-month-old Bicdl1 WT and KI mice (n = 3 per group per age). After anesthesia via IP injection of a ketamine-xylazine cocktail, both eyes were enucleated and fixed in 4% paraformaldehyde (PFA) for 2 h. The fixed eyes were then dissected to remove the anterior segment (cornea, iris, and lens) and the retina was separated from the RPE-choroid-sclera. Both retinal and RPE-choroid-sclera flat mounts (hereafter referred to as RPE flat mounts) were prepared separately by making four radial cuts in the tissues, which were then flat-mounted on glass slides for immunostaining, as described previously [20,23]. In brief, the fixed retinal and RPE flat mounts were blocked in 5% donkey serum (v/v) with 0.3% Triton X-100 (v/v) & 0.2% bovine serum albumin (BSA) and 5% BSA (w/v) with 0.3% Triton X-100 (v/v), respectively, and then stained with anti-Iba1 (pan-microglia/macrophage marker), anti-Tmem119 (resident microglia-specific marker), and anti-CCR2 (marker for infiltrating monocyte-derived macrophages) primary antibodies followed by appropriate secondary antibodies (Appendix 3). A laser confocal microscope (see above) was used to image the paracentral and mid-peripheral regions in each of the four quadrants of each RPE flat mount. The images were analyzed using ImageJ, and the numbers of Iba1⁺ or Tmem119⁺ cells were counted separately in the paracentral and mid-peripheral regions of each RPE flat mount. The cell counts from these two regions were obtained for each eye and statistically compared between Bicdl1 WT and KI groups (n = 6 RPE flat mount per group).

Electron microscopy (EM)

The right eyes of 9-month-old mice (n = 3 WT and 4 KI) were collected for transmission electron microscopy. The samples were collected after the mice were deeply anesthetized by IP injection of ketamine-xylazine cocktail and processed for EM, as described before [18]. Briefly, enucleated eyes were fixed in 2% glutaraldehyde and 2% PFA in 0.2 M cacodylate buffer (v/v). After fixation, the cornea and lens were removed, and the resulting posterior cup was trimmed for EM processing. Then, 70 nm thin sections were cut and stained with 2% aqueous uranyl acetate/lead citrate (v/v) and imaged with a JEOL 1200EX II transmission electron microscope (JEOL USA, Inc., Peabody, MA) at the UTSW Medical Center Electron Microscopy Core. EM fields were imaged in a masked and standardized fashion (the whole sample was scanned, and every other EM field was photographed). Thickness measurements were obtained for Bruch’s membrane (BM), RPE basal infoldings, RPE cell body (excluding the microvilli), and from the bottom of the BM to the top of the RPE on each EM image at three equally spaced points using ImageJ. The average of these three measurements for each EM was reported for each parameter. The number of mitochondria in the apical and basal halves of RPE cells was also counted on each EM image.

Single-cell RNA sequencing (scRNA-seq) analysis and bulk RNA sequencing (bulk RNA-seq) confirmation

We extracted Bicdl1 expression in 90-day-old C57BL/6J retina from a publicly available scRNA-seq dataset deposited by Li et. al. [26], under GEO accession # GSE243413 (WT_P90 mouse retina, GEO accession # GSM7785354). Cells in the rod photoreceptor cluster were categorized as being high Bicdl1 expressors (“Bicdl1-hi”, expression >2.5), medium expressors (“Bicdl1-med”, expression between 1.5 and 2.5), or low expressors (“Bicdl1-lo”, expression >0 and <1.5). Ingenuity pathway analysis (IPA) was performed by comparing Bicdl1-hi with Bicdl1-lo cells. For confirmation purposes, bulk RNA-seq was done following our published protocol [19,21]. Briefly, retinas from deeply anesthetized 9-week-old Bicdl1 WT and KI mice (n = 4 per group) were collected. Three retinas (obtained from three different mice) were pooled to prepare each bulk RNA-seq sample. For each group we prepared 4 different pooled samples for total RNA isolation using a miRNeasy Tissue/Cells Advanced Micro Kit (Cat# 217684, Qiagen, Germantown, MD) according to the protocol supplied with the kit. After checking sample quality control, 1 µg total RNA/sample was submitted to the Next Generation Sequencing Core at UTSW Medical Center for bulk RNA-seq and pathway analysis.

Pathway analysis for both the scRNA-seq and the bulk RNA-seq data was performed using IPA on a set of differentially expressed genes meeting statistical cutoffs of p-value <0.05, and log₂FC > 0.58 (~ fold change >1.5). After applying filters (p < 0.001, |Z-score| > 2, and ratio > 0.15), the top upregulated and downregulated pathways (up to 20 for each direction, if available) were graphed using dot plots. Only five downregulated pathways met the criteria in each of the data sets (bulk RNA-seq and scRNA-seq). Thus, a total of 25 pathways were selected for each dataset (green highlights in Appendix 4 and Appendix 5). Violin plots were generated using the VlnPlot module in Seurat. Meanwhile, the volcano plots were generated using an EnhancedVolcano package in R.

Electroretinography (ERG)

A full-field ERG system (Celeris System, Diagnosis LLC, MA) was used to test the retinal function in Bicdl1 WT and KI mice at different age groups (12-, 15-, and 17-19-month-old). After dark adaptation overnight for at least 16 hours, scotopic ERG recordings were performed under a dim red light on deeply anesthetized mice, as detailed previously [19-21]. Briefly, after pupil dilation and application of GenTeal gel (Novartis, East Hanover, NJ), both eyes were contacted with electrodes for full-field stimulation, producing two-channel recordings with 10 sweeps for each of three different light stimuli (0.01, 0.1, and 1 cd⋅s/m²). Both the scotopic a-wave and scotopic b-wave were analyzed using Diagnosys Espion Software (Diagnosys, Lowell, MA). The photopic a-wave and b-wave were analyzed with the same system using two light stimuli (3 and 10 cd⋅s/m²).

Statistical analysis

Data are presented as the mean ± standard error of the mean (SEM). For all figures, groups of measurements were taken from distinct samples, rather than from repeated measurements. Sample sizes are included in each figure legend. Given the assumption of normal distribution of the data, comparisons between two groups were made using a two-tailed unpaired Student’s t-test. One exception was the quantification of fundus spots, which was based on a non-continuous, ordinal scale and required non-parametric Mann-Whitney U testA linear regression analysis was done to test trends in fundus spot accumulation over time using age in months as the independent variable. In all comparisons, a p-value statistic of < 0.05 is considered significant. Graphs and statistical analysis were done using GraphPad Prism 9.4.1 (GraphPad Software, La Jolla, CA) and Microsoft Excel 16.63.1 (Microsoft, Redmond, WA).

Association between a Bicdl1 mutation and outer retinal thinning in a forward genetic screen in mice

Third-generation (G3) descendants in a pedigree of mice carrying mutations induced by ENU in G0 male mice [17] were screened for retinal phenotypes [18,19,21]. A pedigree containing 80 G3 mice was imaged using the OCT2 imaging system. TRT, ONL, and ORT were measured using ImageJ, as described before [19]. Automated meiotic mapping and evaluation of candidate mutations, respectively, by the Linkage Analyzer and Candidate Explorer software [17,27] led to the identification of a point mutation in Bicdl1 (which we named “bargain allele” for easy reference). It was a substitution of guanine to adenine in coordinate 115,808,212 bp of the mouse genome (GRCm39), corresponding to the first nucleotide of Bicdl1 intron 4. This was associated with a mild but highly significant decrease in ORT and TRT in four homozygous mice (Figure 1A,C; p = 1.7 x 10⁻⁶ and p = 6 x 10⁻⁶, respectively). The significance threshold after Bonferroni correction was p < 0.0006. Of note, the scatter plot showed a very tight distribution for the ORT parameter in the homozygous mice, compared to the TRT parameter (Figure 1B,D). Moreover, mice heterozygous (HET mice in Figure 1B,D) for the Bicdl1 mutation did not show any difference in ORT or TRT compared to mice with no Bicdl1 mutated alleles (“REF” mice in Figure 1B,D).

Outer retinal thinning in CRISPR-generated Bicdl1 KI mice

Little is known about the relevance and role of Bicdl1 in retinal health. Therefore, to validate the effect of the Bicdl1 mutation on ORT and TRT phenotypes on a clean background (with no other ENU-induced mutations) and to enable subsequent phenotypic and molecular studies, we generated Bicdl1 KI mice having the same mutation that we identified in our forward genetics screen (a replacement allele – see Appendix 2). Ensembl, NCBI, and UniProt indicate the existence of several Bicdl1 splice isoforms. In most of these, including the 577-aa canonical isoform, the effect of the KI mutation is predicted to be “probably null”. There are two potential isoforms, predicted only by Ensembl (235-aa and 82-aa), that would not be affected by our mutation, but there is no published data supporting their existence. Bulk RNA seq data from the retina of the Bicdl1 KI mice indicated a 75% decrease in Bicdl1 transcript (Appendix 6), while expression of the housekeeping gene [28] RPP-30 was normal. This decrease in Bicdl1 transcript in Bicdl1 KI mice (despite the fact that there is only a point mutation in the DNA) suggests the activation of nonsense-mediated decay, supporting the conclusion that the mutation is indeed damaging.

A colony of Bicdl1 KI mice was established, and OCT images were obtained at 3 weeks and 2, 4, 6, 9, 12, 15 months of age. Representative images are shown in Figure 2A-F. The ORT, TRT, and ONL were measured using ImageJ. For simplicity and labeling purposes, the terms KI or Bicdl1 KI will be used to refer to homozygous Bicdl1^KI/KI mice, while the terms Het and heterozygous will be used to refer to Bicdl1^KI/+ mice. We documented a statistically significant decrease in ORT (3-10%) in Bicdl1 KI mice compared to the WT mice in all age groups (Figure 2G; p < 0.01 for each age group). Thinning of the ONL (3%) was first detected at 6 months of age (Figure 2H; p = 0.02). There was no difference in ORT, ONL, or TRT between female and male mice (Figure 2G-I). Heterozygous mice did not show any difference in ORT and ONL compared to WT mice (Appendix 7), confirming the data from the ENU mutagenesis screen (Figure 1B,D), indicating an autosomal recessive inheritance pattern, and strongly suggesting that the mutation causes a loss of function of the encoded protein.

To corroborate our findings of thinning in different retinal layers, retina sections were stained with H&E, and light microscopy images were obtained (Figure 3A-D). We demonstrated a statistically significant thinning of the ONL as measured by ONL thickness (up to 12%) and ONL nuclear count (up to 20%) in Bicdl1 KI mice compared to WT mice at 9 months (Figure 3E,F) and 19 months of age (Figure 3G,H). To look specifically at the rod photoreceptor outer segments, we performed immunohistochemistry of retinal sections using a rhodopsin antibody (Figure 4A-D). Measurements using ImageJ revealed a significant decrease in the average thickness of the rod outer segments in Bicdl1 KI mice compared to WT mice, both at 9 months (Figure 4E, 14% decrease, p < 0.05) and at 19 months of age (Figure 4F, 30% decrease, p < 0.05).

To determine whether the thinning of the ONL was due to apoptotic cell death, both a caspase-3 staining and a TUNEL assay were performed on retinal cross sections of 9 month and 19 month old mice. Interestingly, there were very few caspase-3 (Appendix 8) and TUNEL⁺ (Appendix 8) cells in both strains, and no significant difference in either assay.

Finally, analysis of the thickness of inner retinal layers (INL and GCC) using OCT and H&E did not show any significant differences between Bicdl1 WT and KI mice (Appendix 9).

Increased number of fundus spots in Bicdl1 KI mice corresponding to subretinal microglia

Increased numbers of fundus spots are commonly seen in mouse models of retinal degeneration in aged mice. This finding often correlates with the presence of activated subretinal microglia [19-21,23,29]. To determine whether this was the case in Bicdl1 KI mice, we took color fundus photographs at several time points (Figure 5A-J). A modified fundus spot scale was used for grading. Statistical analysis demonstrated a significant increase in the number of fundus spots at all time points after 2 months of age in the Bicdl1 KI mice compared to WT, up to a tripling of the fundus spot grading by 15 months of age (Figure 5K). The trend analysis demonstrated a significant increase in fundus spots with age in Bicdl1 KI mice but not in WT mice (Appendix 10).

In order to better characterize these fundus spots and understand their nature, we chose an eye with an asymmetric distribution of fundus spots and prepared an RPE flat mount, which we stained with the pan-microglia/macrophage marker Iba1 (Figure 6A-C). We found that the flat mount contained a similar distribution of Iba1⁺ subretinal cells adhered to the RPE compared to the fundus spots, suggesting that the fundus spots correlated with activated subretinal microglia/macrophages. In a second experiment, we imaged eyes with fundus spots using both fundus photography and a Spectralis OCT. We then used the infrared images from the Spectralis OCT to overlap the vessels of both imaging modalities. This allowed us to “register” both images and look for an anatomical correlate for the fundus spots on OCT. We found that the fundus spots correlated with subretinal hyperreflective spots located just above the RPE (Appendix 11). Retina sections stained with anti-Iba1 antibodies revealed Iba1+ cells localizing to that same location in the subretinal space (immediately above the RPE) of Bicdl1 KI but not WT mice (Appendix 11). We did not find any evidence for rosettes or true drusen. Finally, we studied the fundus photographs in more detail at early time points (2 and 4 months of age). Since there is a functional distinction between the superior and inferior retina in mice, including differences in photoreceptors, we decided to count fundus spots in each retina. As shown in Appendix 12, we found that the earliest increase in fundus spots in Bicdl1 KI mice occurred in the superior hemiretina.

RPE flat mounts were then prepared from 9-month-old mice and stained with anti-Iba1 (pan-microglial/macrophage marker) and anti-Tmem119 (specific for resident microglia) antibodies. We imaged paracentral and midperipheral areas in four quadrants on the RPE flat mounts. Representative images for the Iba1 staining, Tmem119 staining, and merged channels are shown in Figure 6D,E. Most of these cells were double-positive for Iba1 and Tmem119 staining (yellow cells in the merged-channel image). Bicdl1 KI mice displayed a statistically significant increase in the number of Iba1⁺ cells (p = 0.03, Figure 6F) and also Tmem119⁺ cells (p = 0.03, Figure 6G) in the paracentral zone compared to WT mice. Of note, in addition to being positive for Tmem119, these cells were negative for CCR2 (Appendix 13). Negative controls for RPE flat mounts are shown in Appendix 14.

We also observed migration of inner retinal microglia to the outer retina, specifically to the photoreceptor outer segments (POS), exclusively in Bicdl1 KI mice. These displaced microglia exhibited an activated ameboid morphology and stained positive for both Iba1 and TMEM119. In contrast, both WT and Bicdl1 KI mice demonstrated the presence of Iba1+/TMEM119+ microglia in the inner plexiform layer (IPL) and outer plexiform layer (OPL). These cells had a resting morphology, characterized by small cell bodies and long, branching extensions (Appendix 15).

Decreased retinal function in aging Bicdl1 KI mice

To determine if the outer retinal thickness changes observed in the Bicdl1 KI mice led to functional consequences, we performed ERG on 12-, 15-, and 17-19-month-old mice (Figure 7). Scotopic ERG results did not reveal any differences at 12 months of age. However, at 15 months of age, there was a trend towards a decrease in a-wave amplitude (p = 0.13; Figure 7B), and at 17-19 months of age, there was a significant decrease in the a-wave amplitude both with the 0.1 (14% decrease, p < 0.05) and with the 1.0 cd⋅s/m² stimulus (18% decrease, p < 0.01). No consistent differences were detected in the b-wave amplitude at any time point (Figure 7C). This is concordant with the fact that the retinal degeneration in Bicdl1 KI mice is relatively mild and primarily involves the outer retina. Of interest, we did not detect any consistent differences in the photopic ERG responses of Bicdl1 KI mice (Appendix 16). Moreover, cone arrestin staining (Appendix 16) did not show evidence of decreased numbers of cones in the KI mice compared to the WT mice. The inner segment/outer segment layer did appear thinner in KI mice (consistent with the outer retinal thinning seen in OCT), but the cones did not show any gross dysmorphic features.

Enrichment of oxidative phosphorylation and protein synthesis pathways in photoreceptors in Bicdl1 KI mice

With the goal of better understanding the role of Bicdl1 in the retina, we decided to analyze published scRNA-seq data from WT mice (C57BL/6J mice at P90, source data GEO:GSM7785354) [26], concentrating on rods and comparing cells showing high expression of Bicdl1 to those with low expression. Pathway analysis of the differentially expressed genes revealed enrichment of genes in multiple mitochondrial and protein synthesis-related pathways, which together accounted for 12 of the top 25 pathways (Figure 8A,B). This analysis suggested that perturbations in mitochondrial and translation pathways may be associated with differential Bicdl1 expression in rods.

To directly investigate gene expression changes resulting from Bicdl1 loss of function, we conducted RNA sequencing of Bicdl1 KI neuroretina and WT neuroretina in 9-week-old mice (Figure 8C,D). Although bulk RNA-seq yields gene expression data averaged across a population of cells, rods are by far the most common cell type in the retina, and Bicdl1 is highly expressed in rods. We found that 16 of the top 25 affected pathways were shared between the scRNA-seq and bulk RNA-seq results, and in the same direction of change for every single one of those. That included the top 5 pathways. Both analyses showed significant effects of decreased expression of Bicdl1 on four categories of pathways: energy metabolism-related pathways, protein synthesis-related pathways, oxidative stress and apoptosis-related pathways, and proliferation and regulation of transcription-related pathways (Figure 8, Appendix 4 and Appendix 5). We also found significant changes in the expression of several genes that might be associated with Bicdl1 (Appendix 17), including regulators of energy metabolism in photoreceptors (Ckb and Guk1), antioxidant enzymes (Gpx4, Sod1, Prdx2, Prdx4, Prdx5), and mitophagy-related genes (Fis1, Tomm6, Tomm7, and Tomm22). Among 15 dynein subunit proteins, we found significantly increased expression levels of both isoforms of the dynein light chain roadblock genes Dynlrb1 and Dynlrb2 in Bicdl1 KI mice versus WT mice [30]. Among the cargo adaptor proteins associated with the trans Golgi network [2,3,31,32], we found a significant increase in the expression of Ap4s1 and Ap1s1, decreased expression of Hook3, and no significant change in the expression levels of Bicd1, Bicd2, Hook2, and Ap3s1. The expression levels of Nudc, which is involved in rhodopsin transport [33], and Cfl1 were also increased (Appendix 17).

Decreased RPE mitochondria in Bicdl1 KI mice

Given the data suggesting an association of Bicdl1 to oxidative phosphorylation, the reported expression of Bicdl1 in RPE cells, and the fact that analysis of mitochondria is most reliable in RPE cells, we decided to perform an ultrastructural analysis of the RPE, including its mitochondria, in Bicdl1 KI vs. WT mice.

Representative EM images are shown in Figure 9A-B. Measurements of EM images using ImageJ were analyzed using either each EM field as a replicate (Figure 9C, D) or using the average of each eye (Figure 9E, F) as a test replicate. Analysis per field suggested that Bicdl1 KI RPE cells have thinner basal infoldings, a decrease in RPE thickness, and a decrease in the thickness of the Bruch’s membrane-RPE complex (Figure 9C). It also suggested a highly significant decrease in the number of basal, apical, and total mitochondria in the RPE in Bicdl1 KI mice (Figure 9D). Of note, when the analysis was repeated by mouse, rather than by EM field, the same trend for a reduction in RPE thickness parameters remained, but significance was lost (Figure 9E). Yet, the reduction in the number of mitochondria in Bicdl1 KI RPE remained significant even after this stricter analysis (Figure 9F).

Data statement

The data of this study are available from the corresponding author, Rafael Ufret-Vincenty, upon reasonable request.