Fam151b, the mouse homologue of C.elegans menorin gene, is essential for retinal function

Fam151b is a mammalian homologue of the C. elegans menorin gene, which is involved in neuronal branching. The International Mouse Phenotyping Consortium (IMPC) aims to knock out every gene in the mouse and comprehensively phenotype the mutant animals. This project identified Fam151b homozygous knock-out mice as having retinal degeneration. We show they have no photoreceptor function from eye opening, as demonstrated by a lack of electroretinograph (ERG) response. Histological analysis shows that during development of the eye the correct number of cells are produced and that the layers of the retina differentiate normally. However, after eye opening at P14, Fam151b mutant eyes exhibit signs of retinal stress and rapidly lose photoreceptor cells. We have mutated the second mammalian menorin homologue, Fam151a, and homozygous mutant mice have no discernible phenotype. Sequence analysis indicates that the FAM151 proteins are members of the PLC-like phosphodiesterase superfamily. However, the substrates and function of the proteins remains unknown.

. (a) Multiple sequence alignment (MSA) of representative members of FAM151 family and PLClike phosphodiesterases. This MSA was generated with the program T-Coffee 28 using default parameters and slightly refined manually. The final superfamily alignment was generated using a combination of profileto-profile comparisons 26 and sequence alignments derived from structural superimpositions using DALI 29 , for those families whose tertiary structure is known, such us: Phospholipase D toxins (PDB-IDs: 4Q6X, 1XX1, and 3RLG), GDPD enzymes (PDB-IDs: 1YDY, 3QVQ, 2OTD, and 1VD6) and PLC (Phosphoinositide phospholipase C) enzyme (PDB-ID:1DJY). Families are indicated by coloured background to the left of the alignment: FAM151, Phospholipase D toxins, GDPD and PLC enzymes are indicated in red, pink, blue and green, respectively. The limits of the protein sequence conserved regions included in the alignment are indicated by flanking residue positions. Secondary structure predictions 30 were performed for the FAM151 family and these are consistent with X-ray determined secondary structures of the PLC-like phosphodiesterases. Alphahelices and beta-strands are indicated by cylinders and arrows, respectively. The alignment was presented with

Results
Menorin, FAM151A and B are members of the GDPD/PLCD superfamily. We analysed the amino acid sequence of FAM151B to identify similarities with proteins of known function. A search of the UniRef 50 database 6 shows the FAM151 family to be widely distributed in animals, including C.elegans, as noted above, and Drosophila, and is annotated in Pfam as containing the domain of unknown function (DUF) 2181. We used the remote homology detection server HHPred 7 using a profile from multiple FAM151 alignments to find more divergent homologues. HHPred searches with FAM151 family matched the Phospholipase D domain of a spider (Sicarius terrosus) toxin (PDB-ID: 4Q6X) with a highly significant E-value of 1.4 × 10 −5 8 . Moreover, in support of this, the next most statistically significant matches were to additional members of the Phospholipase D and glycerophosphodiester phosphodiesterase (GDPD) families (Fig. 1). These proteins are all members of the PLC-like phosphodiesterase superfamily of enzymes. PLC-like phosphodiesterases contain a TIM barrel fold, first described in triosephosphate isomerase but found in many proteins, and contain an evolutionarily conserved active site ( Fig. 1) crucially also conserved in menorin and FAM151A and B. These different enzymes hydrolyse phosphodiester bonds of a large number of different substrates, ranging from Glycosylphosphatidylinositol (GPI) protein anchors (in GDPD family) to sphingolipid and lysolipid (in Phospholipase D spider toxins [8][9][10][11] .
FAM151B has a single DUF2181 domain, as does menorin. In FAM151A the DUF domain is duplicated, but the critical active site residues in the C-terminal domain are missing suggesting that this second domain lacks enzymatic activity. Most vertebrates have orthologues of both FAM151A and B, but Medaka is one exception and appears to have only FAM151B.
The expression patterns of the two Fam151 paralogues are quite different. Fam151a expression in the mouse is undetectable in most tissues, but is highly expressed in the intestine, kidney and spleen. In contrast, Fam151b is found expressed at low levels in most tissues including the retinal pigmented epithelium (RPE), retina, iris, ciliary body, lens and cornea. (BioGPS: www.biogps.org).
Fam151b knockout mice exhibit early and rapid photoreceptor loss. Fam151b mutant mice containing a LacZ reporter inserted into intron 2 and a deletion of exon 3 (Fam151b tm1b(EUCOMM)Hmgu hereafter Fam151b KO ) were derived as described in the Methods. The insertion includes a polyadenylation site, and the deletion results in any mRNA from which the insertion has been spliced out having a frameshift and termination of translation 15 codons further. Phenotyping homozygous mutant mice through the IMPReSS pipeline identified a degenerative retinal phenotype at 15 weeks, characterised by patchy pigmentation of the retina, compared to the evenly coloured pink retina of control mice. We examined homozygous mice earlier, at 11 weeks, and also observed a patchwork pattern from fundal imaging indicative of retinal degeneration, not present in wild type litter mates (Fig. 2a). Histological analysis of the eyes revealed a severe reduction in the length of the outer segments of the photoreceptors, and in the number of the nuclei of the photoreceptor cells, found in the outer nuclear layer, indicating a substantial loss of photoreceptors. (Fig. 2b). The other retinal layers appear to be unaffected. To test for retinal function we carried out electroretinogram (ERG) analysis on the mutant mice, which confirmed a loss of photoreceptor function, shown by a greatly reduced scotopic a-wave (Fig. 2c).
We analysed Fam151b mutant and control eyes at several ages to determine when the loss of photoreceptor cells occurred. Eyes were taken at postnatal day 11 (P11) and examined by histology (Fig. 3). Mutant mice at this age had comparable numbers of nuclei in the outer nuclei layer to wild type littermates, and the remainder of the retina also appears normal. The mutant eyes appear to develop normally prior to eye opening at P14. We stained the retinal sections to label glial fibrillary acidic protein (GFAP), a widely used marker of retinal stress 12 , and found no upregulation of this protein at P11.
We compared Fam151b KO/KO and wild type littermates at P15. As shown in Fig. 3, mutant mice still have comparable numbers of photoreceptor nuclei to wild type littermates. However, at this stage, there is an upregulation of GFAP expression, indicative of retinal stress. By P21 Fam151b KO/KO mice have a dramatically reduced outer nuclei layer (Fig. 3) indicating a loss of photoreceptor cells along with GFAP upregulation.
We performed ERG on P15 mice to assess whether the photoreceptors present have any function prior to their degeneration. As shown in Fig. 4a, Fam151b KO/KO mice exhibit a significantly reduced scotopic (dark adapted) a-wave when stimulated with a high intensity light source (10 cd.s/m 2 ) although at standard light intensity, in both dark adapted and light adapted eyes, (3 cd.s/m 2 ) whilst the a-wave amplitude appears reduced it does not reach significance at p = 0.0534 and 0.487931 for dark and light adapted eyes respectively. This suggests that some of the photoreceptors are able to respond to a light stimulus, however the majority cannot and thus produces a lower a-wave amplitude. This may mean that the photoreceptor cells develop and can function normally but degenerate very quickly after eye opening. By P21 there is little to no a-wave amplitude when stimulated with both 10 cd.s/m 2 and 3 cd.s/m 2 flashes (Fig. 4b). This is replicated in the light adapted response showing that cones are also being lost. the program Belvu using a colouring scheme indicating the average BLOSUM62 scores (which are correlated with amino acid conservation) of each alignment column: red (>3), violet (between 3 and 1.5) and light yellow (between 1.5 and 0.5) 31 . Sequences are named according to their UniProt identification. (b) Structural superimposition of two members of the PLC-like phosphodiesterases superfamily: GDPD and PLCD1. Active site residues are labelled and coloured according to their reference protein sequences: human FAM151B, GDPD (PDB-ID: 3QVQ) and PLCD1 (PDB-ID:1djy) in red, blue and green, respectively. Structures are presented using Pymol (http://www.pymol.org). (2020) 10:437 | https://doi.org/10.1038/s41598-019-57398-4 www.nature.com/scientificreports www.nature.com/scientificreports/ the retinal pigment epithelium of Fam151b KO/KO mice appears healthy. The retinal pigment epithelium (RPE) is important for the survival of photoreceptor cells as it provides support and recycles components of photoreceptor specific pathways 13 . The early and rapid loss of photoreceptor cells seen in the Fam151b KO/ KO mice may be indicative of an RPE defect, so we examined two key RPE markers. Both RPE65 and ZO-1 are typical RPE and epithelial cell markers respectively and are both present and localised as expected. RPE65 is a membrane-associated RPE specific enzyme necessary for the generation of 11-cis-retinol in the retinoid cycle. Antibody staining on sections of 2 month old retinal tissue showed clear expression and localisation to the RPE layer in mutant mice (Fig. 5). ZO-1 localises to the tight junctions between cells. ZO-1 antibody staining on 2 month old flattened whole RPE, from which the rest of the retina had been carefully removed, enabled us to view the localization of ZO-1 at the tight junctions in both mutant and control RPE (Fig. 5).
P15 retinal sections stained for the light sensitive rod and cone opsins, rhodopsin and M/L-opsin respectively, showed that in mutant Fam151b KO/KO opsins localised to the outer segments of photoreceptor cells (Fig. 5).
photoreceptor loss is not due to light toxicity. Mice open their eyes at about P14, just before we observed increased GFAP expression and reduced retinal function, raising the possibility that photoreceptor function may be lost due to light toxicity. To test this we removed pups, along with their mothers, from a 12 hour light/dark cycle and placed them in a constant dark environment for 1 week from P13, about one day prior to eye opening, until P20 and immediately thereafter analysed their ERG. We compared ERGs from Fam151b KO/KO mice kept in a dark environment with wild-type littermates kept in the same environment and with Fam151b KO/ KO mice and wild-type littermates kept in a normal 12 hour light/dark cycle (Fig. 4c). The mutant mice maintained    , Menorin, in C. elegans we looked at the patterning of nerves in embryonic skin by neurofilament staining. Developing nerves within the skin were visualised and their branch angles measured. No significant difference was found between Fam151b KO/KO mice and their wild-type littermates. (Fig. 6).
Fam151a mutant mice. We asked if mutations in the paralogous gene, Fam151a, had any detectable phenotype and in particular if they resulted in a similar mutant retinal phenotype. Previously, mice homozygous for a targeted mutation in Fam151a had been bred and phenotyped as part of the IMPC effort and no eye defect was reported at 15 weeks of age. To confirm this, using CRISPR, we generated several mice carrying Fam151a mutations. We selected one, which had a complex double deletion of 19 and 6 bp, flanking a 16 bp section in exon 1, predicted to cause a premature stop codon in exon 2 (Fig. 7a) which showed a loss of Fam151a protein in a Western performed on kidney samples (Fig. 7b), and bred homozygous mutant mice which were aged alongside wild-type and heterozygous littermates.
At one year, ERG analysis of the mutant mice found no decrease in retinal response to a light stimulus; both a-waves and b-waves were similar to those observed in the wild-type litter mates (Fig. 7c). Gross retinal www.nature.com/scientificreports www.nature.com/scientificreports/ morphology also appeared normal on fundal imaging, and histological analysis showed no loss of photoreceptor cells, or increased presence of GFAP staining (Fig. 7d-f).
During the IMPC analysis of the Fam151a mutant mice 1 of 16 homozygous mutant mice was described as having abnormal heart morphology; an enlarged heart. We measured the heart weight relative to body weight of our Fam151a mutant mice and found no significant difference between Fam151a KO/KO mutants and their wild-type and heterozygous litter mates (Fig. 7g).

Fam151a/Fam151b double mutant mice do not exhibit a worsened phenotype.
We asked if the paralogous Fam151 genes compensated for each other in viability or gross phenotype in the mutant lines. To address this Fam151a and Fam151b mutant strains were bred together to produce Fam151 KO/+ /Fam151b KO/+ double heterozygotes which were then used in matings and offspring were observed at Mendelian ratios indicating that the mice were viable. We also observe the expected number of each genotype at weaning (Chi-squared (8 degrees of freedom) p = 0.3712) ( Table 1). We examined the gross retinal morphology of the retinas through fundal imaging at 3 weeks. Double homozygous mutants showed the same degree of retinal degeneration when compared to Fam151a +/+ Fam151b KO/KO littermates, which we confirmed with histology, showing no increase in the loss of photoreceptor cells in the double homozygous mice (Fig. 8). Mice were also maintained until one year of age and no abnormal phenotype, other than retinal degeneration, was observed.

Discussion
Loss of photoreceptors is a leading cause of blindness, understanding the genetic causes and pathology of the cell death is essential for developing therapeutics and possibly for developing ways of preventing the disease. Mouse models are an excellent resource in this area of research as they often recapitulate the human disease and therefore can be used to test the effectiveness of treatments and preventative strategies. www.nature.com/scientificreports www.nature.com/scientificreports/ Here we show that the Fam151b gene is essential for photoreceptor survival in the mouse. Retinal function also appears to be compromised with a reduced scotopic ERG response on eye opening in mice deficient for Fam151b. With such an early phenotype a diminished production or localisation of light sensitive proteins might be expected. However, rhodopsin and M-opsin staining in photoreceptors is still present and localised to the outer segments. These proteins are still observed at later stages, at week 11 two layers of photoreceptors remain with correct localisation of opsins, indicating that after the initial surge of cell death the surviving cells retain some aspects of normal cellular architecture. To date no patients with mutations in either FAM151 paralogues have been described. Clear loss of function mutation in FAM151B are on the whole rare, and show a pattern similar to other retinal disease-causing genes. Light is well known to cause or further exacerbate retinal degeneration [14][15][16][17] . Photoreceptors must constantly renew essential pathway components and remove toxic products. Because of this and due to the timing of the retinal degeneration, we investigated whether light caused photoreceptor loss in the Fam151b KO/KO mice and found this not to be the case. Mice kept in a dark environment from before their eyes were open showed a similar rate of degeneration. It is assumed that prior to eye opening some light may leak through the eyelids, but as no GFAP staining was observed in P11 retinas this is unlikely to be the cause of degeneration.
The molecular function of menorin/FAM151A and B is unknown. The C.elegans protein, SAX-7, interacts with menorin to control the dendritic branching of sensory neurons 5 . Mutants of the mammalian homologue of Sax-7, L1CAM, have been described as having abnormal axon guidance and retinal ganglion cells, as well as other neuronal problems [18][19][20][21] . This phenotype has similarities to that seen in C. elegans mutants, and has implicated L1CAM in the precise guidance of axons. However this differs from the phenotype we have observed in Fam151b KO/KO mice. Furthermore, the branch pattern of nerves located in the skin was analysed and no abnormality was observed in mutant mice (Fig. 6). In addition, loss of dendritic complexity in humans has been linked to behavioural defects, autism and schizophrenia 22,23 . Analysis of the Fam151b KO/KO mutant mice in the ImPress protocols found no abnormal behavioural phenotype.  www.nature.com/scientificreports www.nature.com/scientificreports/ Given the strong statistical significance of profile comparisons (HHpred), and the concordance of known and predicted secondary structures and conserved active site residues (Fig. 1), we have shown that the FAM151 proteins are members of the PLC-like phosphodiesterase superfamily of enzymes. However, the remarkable substrate  www.nature.com/scientificreports www.nature.com/scientificreports/ diversity of its phosphodiesterase homologues, means our computational analysis is unable to predict the substrate of the FAM151 family. Ultimately, identification of substrates and products will enhance our understanding of dendritic branching in nematode and retinal maintenance in mammals.

Material and Methods
Sequence analysis. The computational protein sequence analysis began by performing a JackHMMER iterative search 24 beginning from the human FAM151B protein sequence, against the UniRef50 database 6 . The FAM151 family is widely distributed in animals, including nematodes (C. elegans; UniProt: O45879) and hexapods (D. melanogaster; UniProt: Q9VLU9). This reproduces the family phyletic distribution reported in Pfam (Family: DUF2181/PF10223) 25 . Next, we took advantage of profile-versus-profile (HHpred) 7 to search PDB70 database for more divergent FAM151 homologues using as input a profile generated from the multiple protein sequence alignment of FAM151 family. The PDB70 database contains profile hidden Markov models (HMMs) for representative sequences, clustered to 70% maximum pairwise sequence identity to reduce redundancy, drawn from the PDB (Protein Data Bank) 26 .

Mice.
Mice were generated at MRC Harwell from ES cells with a targeted "knock-out first", tm1a, mutation in Fam151b, and subsequently crossed with CRE recombinase expressing mice to generate mice with a LacZ insertion in intron 2 and a deletion of exon 3 (the allele is Fam151b tm1b(EUCOMM)Hmgu , hereafter Fam151b KO ). The ES cells are of C57BL/6N origin, and the resulting mice initially maintained by crossing with C57BL/6NTac mice, and offspring selected for absence of Cre recombinase. The mutants were subsequently crossed with and maintained on a C57BL/6J background, and genotyped to ensure removal of the Crb1 mutation originating from the C57BL/6N strain.
Fam151a KO/KO mice were made by microinjection of CRISPR gRNA targeting exon 1 and Cas9 protein into the pronucleus of fertilized eggs (Table 2). Sequencing identified a mouse with a 19 and 6 bp deletion surrounding a 16 bp section in exon 1. gRNAs were chosen based on low off-target predictions and founder mice were bred with C57BL/6J mice first before breeding on.
All animal work was approved by the University of Edinburgh internal ethics committee and was performed in accordance with the institutional guidelines under license by the UK Home Office.
immunohistochemistry. Mice were culled and eyes were enucleated and placed into Davidson's fixative (28.5% ethanol, 2.2% neutral buffered formalin, 11% glacial acetic acid) for 1 hour (cryosectioning) or overnight (wax embedding). For cryosectioning eyes were removed from Davidson's fixative and placed into 10%, 15% and 20% sucrose in PBS for 15 mins, 15 mins and overnight respectively. Eyes were then embedded using OCT cryopreservant and kept at −80 until sectioned. For wax preservation eyes were removed from Davidson's fix and placed successively into 70% ethanol, twice in 70% 80% xylene, then 90% paraffin and finally twice in 100%; paraffin each for 45 mins.  www.nature.com/scientificreports www.nature.com/scientificreports/ Heamatoxylin and Eosin staining was performed on 8 µm paraffin tissue sections and imaged on a Zeiss Brightfield microscope.
For wholemount staining of RPE eyes were enucleated and immersed in 2% PFA for 3 minutes, and washed twice in PBS for 5 minutes. The rest of the retina was dissected from the RPE by removing the cornea and lens and carefully peeling the retinal tissue off to leave only the RPE attached to scleral tissue. Radial incisions were made in order to lay the RPE flat and methanol was added slowly and the tissue was placed at −20 °C for four hours. Staining was then performed on the RPE using the ZO-1 antibody (33-9100, Thermo Scientific) 1:100 in blocking buffer. All staining was performed on an n of 3 for each genotype.
For wholemount staining on embryonic skin at embryonic day 16.5 mice were collected and culled by schedule 1 approved methods. Limbs were removed and the body was placed in 4% PFA at 4 °C overnight. The bodies were then washed three times in PBS for five minutes at room temperature and transferred to 100% Methanol at −20 °C for 4 hours. The skin was then removed and rehydrated in Methanol/PBS-T (PBS with 0.2% triton X-100) mixtures of 75%, 50% and 25% for five minutes each. The skin was then stained with Neurofilament (2H3, DSHB) antibody 1:400 in blocking buffer. electroretinography. All mice undergoing an ERG were dark adapted overnight prior to the procedure, and experiments were carried out in a darkened room under red light using an HMsERG system (Ocuscience). Mice were anesthetised using isofluorane and pupils were dilated through the topical application of 1% w/v tropicamide before being placed on a heated ERG plate. Three grounding electrodes were used subcutaneously (tail, and each cheek) and silver embedded electrodes were placed upon the cornea held in place with a contact lens. The standard International Society for Clinical Electrophysiology of Vision (ISCEV) protocol was used which recorded scotopic responses before a 10 min light adaption phase in order to record photopic responses 27 . 3 and 10 cd.s/m 2 light intensity scotopic responses were used for analysis. Data was analysed using Graphpad Prism and compared by unpaired t-test with Welch's correction.