Photoreceptor degeneration is the prominent characteristic of retinitis pigmentosa (RP), a heterogeneous group of inherited retinal dystrophies resulting in blindness. Although abnormalities in many pathways can cause photoreceptor degeneration, one of the most important causes is defective protein transport through the connecting cilium, the structure that connects the biosynthetic inner segment with the photosensitive outer segment of the photoreceptors. The majority of patients with X-linked RP have mutations in the retinitis pigmentosa GTPase regulator (RPGR) or RP2 genes, the protein products of which are both components of the connecting cilium and associated with distinct mechanisms of protein delivery to the outer segment. RP2 and RPGR proteins are associated with severe diseases ranging from classic RP to atypical forms. In this short review, we will summarise current knowledge generated by experimental studies and knockout animal models, compare and discuss the prominent hypotheses about the two proteins' functions in retinal cell biology.
Retinitis pigmentosa (RP) represents a heterogeneous group of retinal disorders characterised by a classic clinical triad of peripheral bone spicule formation, retinal blood vessel attenuation and pallor of the optic disc. The hallmark of RP is photoreceptor degeneration and, collectively, they affect 1:3000–1:7000 people, representing a leading cause of inherited blindness worldwide . The symptoms of RP typically begin in early childhood due to rod degeneration, with night blindness (nyctalopia) preceding a progressive constriction of the visual field. Secondary involvement of the central macula (cone degeneration) leads eventually to loss of central visual acuity and blindness. Age of onset and pattern of disease can vary widely in RP (even within the same family ), largely due to its heterogeneous genetic basis, with mutations in more than 50 genes so far identified as causing the non-syndromic form of disease (https://sph.uth.edu/retnet/). Inheritance can be autosomal dominant, autosomal recessive or X-linked, although rare digenic forms and mitochondrial inheritance have been reported . Furthermore, RP has been described as the ocular manifestation of several ciliopathy syndromes such as Usher, Joubert and Bardet–Biedl syndromes [3–5].
X-linked RP (XLRP) is a severe form of disease and remains untreatable. At least five genetic loci have been associated with XLRP, but to date only three genes have been identified [6–8]. The majority of XLRP cases are attributed to mutations in the retinitis pigmentosa GTPase regulator (RPGR; 70–90% of XLRP patients) and the retinitis pigmentosa 2 (RP2; 7–18% of XLRP patients)  genes. A deep intronic mutation in OFD1 (oral–facial–digital 1) gene has also been found to cause XLRP in one family , but this review will focus on the two most common genes.
Highly efficient protein transport is vital for photoreceptor health because of their unique structure: the biosynthetic inner segment (IS) is separated from the photosensitive outer segment (OS) by the connecting cilium (CC), a structure similar to the transition zone of primary cilia  (Figure 1A). Thus, all OS protein (synthesised in the IS) requires trafficking first to (and then across) the CC. RP2 and RPGR proteins are both localised to the CC and it is suggested that they play a role in sorting and trafficking to the OS . XLRP is thus largely considered a ciliopathy. In this review, we will briefly summarise current knowledge from biochemical data and animal models regarding the role of RP2 and RPGR in retinal cell biology. The elucidation of disease mechanisms is necessary to accompany recent advances in therapeutic techniques in helping patients with this debilitating condition.
Suggested roles for RP2 and RPGR in the photoreceptor CC.
RPGR disease mechanisms
Linkage analysis and deletion mapping first led to the positional cloning of RPGR at Xp21.1 20 years ago [6,10]. Responsible for the majority of XLRP, this alternatively spliced gene is predicted to code five protein variants . The highly repetitive, Glu-Gly-rich open reading frame 15 (ORF15) of the retinal-enriched isoform (RPGRORF15) is a mutational hotspot and is believed to be glutamylated prior to transport to the CC [60,61]. Non-sense mutations in this area predict premature truncations of the protein product (with loss of the conserved, basic C-terminal domain)  and account for the majority of disease . The repetitive domain is not under strict evolutionary control  and partial truncation of murine ORF15 does not appear to alter RPGR's function . Furthermore, in frame deletions do not necessarily lead to photoreceptor disease. That said, all human disease-causing mutations affect the RPGRORF15 variant. Thus, the RPGRORF15 isoform has been the focus of much research in an attempt to determine its role in photoreceptor biology and the underlying pathophysiology that results from the common human mutations. Animal models have succeeded in shedding light on RPGR's mechanism of action.
Both the constitutive (RPGREx1-19) and retinal-enriched (RPGRORF15) RPGR isoforms localise to the base of the photoreceptor CC between the IS and the OS [15,16]. Their differing expression in development, however, hints at distinct (if overlapping) functions . Conflicting reports from mammalian and zebrafish models [15,18,19] confuse our understanding of RPGR's role in photoreceptor development, but human RPGR/XLRP patients have apparently fully functioning retina at birth and so we must assume that any role in development is redundant. Several naturally occurring and knockout animal models exist for studying RPGR disease [15,19–23]. They suggest a role for RPGR in opsin transport in the mature retina [15,20,21]. Opsin mislocalisation resulting from compromised RPGR function leads to abnormal OS disc morphogenesis and both rod and cone degeneration [15,19–22]. This has been supported by post-mortem analysis of human retinal samples [24,25], and this opsin mislocalisation leads to a severe form of retinal degeneration in humans.
RPGR/XLRP causes a severe retinal degeneration, leaving patients blind from a young age. Steady visual acuity decline occurs over time , with two broad field loss patterns being noted. Within the gene, however, differing patient RPGR mutations cause notable phenotypic variability. While the vast majority of RPGR patients (95%) develop classic RP, cone dystrophy, cone-rod dystrophy, atrophic macular degeneration and systemic ciliopathies have all been documented [26–30]. Large patient cohorts have attempted to determine disease patterns, with genotype–phenotype concordance being observed regarding field loss patterns and electrophysiology  and worsening disease being documented as the mutation approaches the N-terminal end of RPGR . That said, dizygotic twins were shown to be discordant for RPGR/XLRP disease severity. Environmental factors and stochastic developmental influences may affect disease progression and genetic epistasis (‘modifier’ genes influencing disease) clearly plays a role, with recent sequencing of RPGR/XLRP patients showing SNPs in IQCB1 and RPGRIP1L affecting retinal function . Whatever the environmental and epistatic influences, it is misfunctioning of the RPGR protein that is the main stressor on the photoreceptors of patients with RPGR mutations and its function must be determined if novel therapies are to be developed.
Mechanisms of action
The opsin mislocalisation seen in animal models and patient autopsy specimens suggest that RPGR's function is to aid trafficking to, and then across, the photoreceptor CC. A full discussion of known RPGR interactions is out with the scope of this review (for a more comprehensive paper, see ref. ). Furthermore, novel post-translational modification mechanisms potentially involved in RPGR's function  that appear linked to retinal degeneration  will not be discussed. Instead, we will focus on the gathering evidence for three possible roles for RPGR in the photoreceptor CC (Figure 1B).
The RPGR-interacting protein 1 (RPGRIP1) localises to the photoreceptor CC [33–35] where it interacts with RPGR [35–37] and is proposed to tether RPGR to the ciliary membrane [34,38]. Recent work, however, showed that RPGRIP1's interaction domain with RPGR partially overlaps that of cGMP phosphodiesterase δ subunit PDE6δ (PDE6D) [39,40], a highly conserved prenyl-binding protein involved in shuttling proteins, including phototransduction proteins , between different membrane compartments. It is known to bind RPGR . While PDE6δ's exact binding site on RPGR has since been contested [43,44], this work suggested that RPGRIP1 may compete with and loosen PDE6δ's binding following RPGR-mediated delivery to the CC . This could simply be responsible for RPGR delivery to the CC, but could act to deliver prenylated proteins to the OS. Interestingly, PDE6δ-cargo binding is regulated by ADP ribosylation factor-like 2/3 (ARL2/3), thus promoting prenylated protein trafficking to the CC. The GTP-binding state of ARL3 is regulated by RP2 (see below).
RPGR's N-terminal (exons 1–10) encodes a regulator of chromosome condensation 1 (RCC1)-like domain (RLD) [10,29]. This RLD is predicted to act as a guanine nucleotide exchange factor (GEF), activating small GTPases that then facilitate intracellular signalling and trafficking pathways. While GEF activity for the originally predicted GTPase, Ran , has never been shown, RPGR was shown to activate RAB8  with human RPGR mutations perturbing this, resulting in RAB8 mislocalisation away from primary cilia. This is intriguing, as RAB8 has been suggested to regulate rhodopsin transport to the CC base , suggesting a model whereby RPGR-mediated activation of RAB8 facilitates rhodopsin trafficking. However, only one residue required for RCC1 GEF function is conserved in RPGR  and the β-hairpin extension required for GEF activity is not present , raising questions as to its role in activating RAB8. This finding needs further corroboration.
As mentioned, the C-terminal domain of RPGRORF15 is highly conserved  and appears critical for photoreceptor health. Given the majority of disease-causing RPGR mutations results in premature protein truncation and loss of the C-terminal, it would seem logical that it has an important function. The ORF15 basic domain interacts with the scaffold protein Whirlin (WHRLN) , known to play a role in cytoskeletal assembly in the ear [50,51] and photoreceptor. Whirlin's interaction with espin helps regulate periciliary membrane actin [52,53] and mutations in the gene are responsible for Usher syndrome (type 2D), a syndromic form of RP and deafness . An actin bundle connects the periciliary membrane complex with the basal body, along which the actin-based motor protein Myosin VIIa appears to travel [55,56]. Myosin VIIa contributes to rhodopsin transport into the CC , so it is interesting that RPGR appears to influence the turnover of actin , with increased actin polymerisation resulting from knockdown of RPGR. This fits with a model whereby RPGR interacts with the usherin protein network, facilitating rhodopsin delivery into the CC by mediating actin turnover. Of particular relevance to this proposed model is that actin depolymerisation results in lengthened, disorganised OS discs  reminiscent of the RPGR knockout (KO) mouse . Further work would be required to determine this proposed role.
RP2 disease mechanisms
RP2 was first identified by linkage analysis as one of the common XLRP genes . The RP2 protein consists of 350 residues and appears to be widely expressed. In the human retina, RP2 is predominantly localised to the plasma membrane of rod and cone photoreceptors, the retinal pigment epithelium (RPE) and other retinal cell types . Pools of RP2 have also been detected in the Golgi complex , the primary cilia (where RP2 is known to shuttle by an Importin-β2-mediated mechanism) [64,65] and the basal body of the CC in mouse photoreceptors . RP2 undergoes dual acylation at the extreme N-terminus, and these modifications are both necessary for plasma membrane localisation/cilia targeting [63,65–67] and are pathologically relevant . Animal models have been generated to help determine RP2's function in the photoreceptors.
Two RP2 knockout mice have been generated [69,70], which appear distinct in their retinal phenotypes. Originally, a mouse model encompassing an exon 2 deletion (leading to the absence of detectable protein product) was generated , resulting in retinal degeneration characterised by abnormal OS morphology, cone-specific ciliary membrane and axoneme elongation  and prominent mislocalisation of M-opsin. Photoreceptor function and morphology and M-opsin localisation in this model can be rescued for over 18 months by AAV-mediated gene augmentation . More recently, a novel Rp2 null mouse was generated by gene trap, which results in deletion of the Rp2 gene after exon 1 . These mice present with slowly progressive, late-onset cone-rod dystrophy and are seen to develop modest mislocalisation of prenylated proteins of the phototransduction cascade [namely cGMP phosphodiesterase PDE6 holoenzyme and rhodopsin kinase G protein-coupled receptor kinase 1 (GRK1)]  with correct localisation of transducin subunits and M-opsin. The phenotypical differences between the two mouse models could simply be attributed to the different experimental designs or may reflect the clinical heterogeneity that characterises patients with XLRP. However, it is also possible that the observed mislocalisation phenotypes are secondary manifestations of a more fundamental defect. That said, the cone-led dystrophy seen in both models is interesting given the clinical picture that characterises patients with XLRP.
Human RP2 mutations cause a severe, atypical form of RP with early macular involvement leading to central visual loss . Indeed, macular atrophy has been seen as the predominant feature in some case studies . Clinical studies suggest that RP2/XLRP patients have a faster loss of central visual acuity than those with RPGR mutations, despite comparable visual field loss and electroretinographic disease progression . RP2/XLRP patients are known to be myopic [31,73–75], although differences in the severity of myopia compared with RPGR/XLRP patients have been disputed [31,75]. The onset of nyctalopia appears similar between the two genotypes . Although ERG testing has shown rod degeneration to be the prominent feature of RP2 disease, the early cone involvement is in keeping with the findings in animal models. The underlying action of RP2, therefore, needs to be sought.
Mechanisms of action
Structurally, RP2 encompasses distinct domains, in particular an N-terminal β-helix domain and a C-terminal ferredoxin α/β-domain  that are thought to have distinct functional roles. The C-terminal domain shares weak homology with nucleoside diphosphate kinases (NDKs). However, it is unlikely that RP2 displays the canonical phosphotransferase activity of NDKs in vivo, since potential key catalytic residues are not conserved on RP2. Therefore, the functional role of this domain remains elusive, although it has been suggested that it forms the platform for protein–protein interactions [76,77].
On the other hand, the functional significance of the N-terminal domain has been studied in detail. It was noted very early that it shares 30% sequence homology and high structural similarity with tubulin cofactor C (TBCC), which acts as a GTPase-activating protein (GAP) in the α/β-tubulin heterodimer assembly [75,78]. RP2 can also act as a GAP for native tubulin , an interesting observation given the abnormal OS morphology and elongated ciliary membrane/axoneme seen in the Rp2 null mouse . However, it fails to replace TBCC in an in vitro reaction of tubulin heterodimerisation . Furthermore, TBCC lacks plasma membrane localisation in retinal cells ; therefore, it is unlikely that there is a functional redundancy between the two proteins in the retina. Of more certainty is RP2's function in regulating the activity of the Arl3.
The N-terminus of RP2 interacts specifically with Arl3 and has GAP activity towards this small GTPase . The RP2–Arl3-binding interface is a mutational hotspot for disease, highlighting the importance of this complex for retinal function. Arl3, a member of the Ras superfamily, was initially found in a screen for ciliary genes , and studies using knockout mice further support a role of Arl3 in cilia development and maintenance both in the kidney and the photoreceptors [82,83]. Arl3 is mainly cytoplasmic and has been described as a microtubule-associated protein . Structural studies have shown that GTP-bound Arl3 can regulate the retinal chaperones PDE6δ and Unc119 [84,85], which can bind and solubilise membrane-bound lipid-modified proteins (Figure 1A). Chaperone regulation is important for photoreceptor function as they facilitate shuttling of proteins of the phototransduction cascade, namely prenylated PDE6 holoenzyme and GRK1, as well as transducin subunits such as myristoylated Tα [86,87]. Indeed, PDE6δ ΚO mice show mislocalisation of PDE6 and GRK1 away from the OS of rods and cones , and deletion of Unc119 in mice and in a human patient leads to late-onset cone dystrophy . It is thought that the role of Arl3-GTP is to help release lipid-modified cargo in destination membranes, perhaps the trans-Golgi or the entrance of the ciliary membrane, and RP2, which catalyses GTP hydrolysis, can attenuate this by rendering Arl3 unable to interact with PDE6δ and Unc119  (Figure 1A).
Aside from trafficking of specific lipid-modified cargo, evidence exists for some other proposed roles for RP2, although their importance in the retina needs to be investigated. Cell culture studies suggest a role for RP2 in renal processes, in particular in intraciliary trafficking and apical secretion of the cystic disease protein Polycystin 2 , as well as in Golgi cohesion and global post-Golgi transport of vesicles [63,90]. Furthermore, it has been suggested that RP2 has exonuclease function in the base excision repair pathway of oxidative damage defence, via its NDK-like domain . Although the present study has not been independently confirmed, several recent studies have linked ciliopathies with proteins of the DNA damage response machinery [91,92]. In zebrafish, rp2 has been shown to be a maternal effect gene that is essential for embryonic development , although this claim is contradicted by the fact that rp2 knockout zebrafish can breed normally . Recently, Arl3 GTPase was found to be a novel binding partner and a positive regulator of the transcription factor signal transducer and activator of transcription 3. This finding implies that RP2 has a potential role in the regulation of transcription .
Fundamental questions remain unanswered in the RP2 field, namely whether the Arl3-interacting domain is exclusively important for retinal disease or if the C-terminus is also pathologically relevant. In addition, the regulation of the GAP activity of RP2 will be another important area of study and might provide the basis for pharmacological intervention that can accompany recent advances in gene therapy for treating XLRP.
XLRP due to RP2 and RPGR mutations is one of the most severe forms of RP. Because of the wide spectrum of phenotypic variability that exists for different mutations in patients and the frequently confounding results from experimental studies, the exact functions of the RP2 and RPGR proteins have still not been completely elucidated over the 20 years since their discovery. As such, it seems increasingly possible that RP2 and RPGR are multifunctional proteins with more than one role in retinal physiology. With recent advances in the development of novel therapeutics for retinal degeneration, it is reasonable to hope that XLRP will be treatable in the near future. In particular, cell replacement therapy (transplantation of stem cell-derived photoreceptor or RPE precursors into the retina) has been successfully applied to several rodent models of retinal degeneration  and is currently in human clinical trial for the treatment of age-related macular degeneration and Stargardt disease (reviewed in ref. ). Alternatively, gene augmentation therapy for Leber congenital amaurosis patients appears safe , although the effect on visual function has been disappointing, with retinal degeneration progressing despite good gene expression . That said, gene augmentation has successfully prevented or halted retinal degeneration in mouse and canine models of RPGR deficiency and an RP2 knockout mouse model [72,100,101]. Previous studies on mouse models have highlighted the need to carefully control RPGR levels in the retina, as both overexpression of RPGREx.1-19 (but not RPGRORF15) on a wild-type background  and administration of high doses of AAV vectors carrying full-length RpgrORF15 on an Rpgr null background  have detrimental effect on the retina. Due to these concerns, a thorough understanding of these proteins' functional roles is needed to avoid adverse effects and to promote the development of complementary strategies . Further experimental studies are essential to ensure that viable therapeutic strategies for XLRP will reach the clinic.
AAV, adeno-associated virus; ARL, ADP ribosylation factor-like; CC, connecting cilium; GAP, GTPase-activating protein; GEF, guanine nucleotide exchange factor; GRK1, G protein-coupled receptor kinase 1; IS, inner segment; KO, knockout; NDKs, nucleoside diphosphate kinases; OFD1, oral–facial–digital 1; ORF, open reading frame; OS, outer segment; PDE6, cGMP phosphodiesterase 6; RCC1, regulator of chromosome condensation 1; RLD, RCC1-like domain; RP, retinitis pigmentosa; RPE, retinal pigment epithelium; RPGR, retinitis pigmentosa GTPase regulator; RPGRIP1, RPGR-interacting protein 1; SNP, single nucleotide polymorphism; TBCC, tubulin cofactor C; XLRP, X-linked retinitis pigmentosa.
R.L. is funded by a PhD studentship from the College of Medicine and Veterinary Medicine, University of Edinburgh with additional support from the Medical Research Council (MRC). R.M. is funded by the Wellcome Trust and the Academy of Medical Sciences. T.H. is funded by the MRC Human Genetics Unit and a University of Edinburgh Chancellor's fellowship. This work was supported by grant funding [GR588] from RP Fighting Blindness (www.rpfightingblindness.org.uk).
We thank Matthieu Vermeren for critical reading of the manuscript.
The Authors declare that there are no competing interests associated with the manuscript.
These authors contributed equally to this work.