The lipid moieties required for protein prenylation, FPP and GGPP, are key isoprenoid intermediates of the mevalonate/cholesterol synthesis pathway ( Figure 1 ). Briefly, acetyl-CoA is converted to HMG-CoA in a sequential manner catalyzed by ACAT1, ACAT2, and HMG-CoA synthase 1 (HMGCS1), which is then converted to mevalonate by HMG-CoA reductase (HMGCR) (13). HMGCR is the rate-limiting step in the mevalonate pathway and is the mechanistic target of statins, a commonly used drug used in the treatment of hypertension and hypercholesteremia (14). Mevalonate kinase (MVK) catalyzes the phosphorylation of mevalonic acid to phosphomevalonate and is subject to feedback inhibition by downstream isoprene intermediates like geranylpyrophosphate (GPP), FPP, and GGPP (15). Phosphomevalonate is converted to isopentenyl pyrophosphate (IPP) by phosphorylation and decarboxylation, catalyzed by phosphomevalonate kinase (PMK) and mevalonate pyrophosphate decarboxylase (MPDC), respectively (16). IPP undergoes reversible isomerization to dimethylallyl pyrophosphate (DMAPP). The condensation of IPP and DMAPP results in the 10-carbon isoprenoid geranyl pyrophosphate (GPP). Subsequent condensation of 1 or 2 more IPP units to GPP yields the 15-carbon and the 20-carbon isoprenoids FPP and GGPP, respectively (17). Key mevalonate pathway intermediates IPP, FPP, and GGPP feed into the synthesis of squalene, dolichols, ubiquinone, sterols and derivative steroid hormones, as well as protein prenylation (18). A schematic representation of isoprene synthesis through the mevalonate pathway is provided in Figure 1 .

After isoprenoid generation, irreversible prenylation is catalyzed by protein farnesyltransferase (FTase) or protein geranylgeranyltransferase type I (GGTase-I), which differ in target carboxy-terminal signal peptide recognition and the isoprenoid substrate (9). FTase catalyzes the addition of farnesyl pyrophosphate (FPP), while GGTase-I catalyzes the addition of geranylgeranyl pyrophosphate (GGPP, Figure 1 ) (8–10, 19, 20). The C-terminal recognition motif for prenylation is the CAAX (also referred to as Ca₁a₂X) domain, in which ‘C’ is the cysteine to be prenylated, ‘A’ is any aliphatic amino acid, and ‘X’ is an amino acid which determines enzyme specificity (8, 21, 22). The target motifs for FTase typically have serine, methionine, alanine, or glutamine at the X residue position, while GGTase-I usually recognizes leucine or phenylalanine. Some substrates may also be recognized by both enzymes as targets for prenylation (23–26). Recent studies have shown that proteins terminating in C(X)₃X and CXX may also potentially be substrates of FTase, where X represents any amino acid (27–29).

Following prenylation at the cysteine residue in the target motif, most prenylated proteins then undergo two additional modifications before transport to the cell membrane. First, the terminal -AAX residues are cleaved by the CAAX proteases Ras converting enzyme 1 (RCE1P) or zinc metallopeptidase (STE24). The resulting carboxylate then undergoes methylation by the S-adenosylmethionine-dependent isoprenylcysteine methyltransferase (ICMT) to form a post-translationally modified prenylcysteine methyl ester C-terminus on the protein (30, 31) ( Figure 1 ). The three steps (prenylation, proteolytic cleavage, and methylation) described above are necessary for function of most prenylated proteins but recently it has been found that heat shock protein chaperone Ydj1p in yeast undergoes prenylation without subsequent proteolysis or methylation in what has been deemed the “shunt pathway”, with evidence of these processing steps actually being deleterious to the protein’s function (32).

The other enzyme responsible for the transfer of 20-carbon geranylgeranyl-group from GGPP to the target motif is termed geranylgeranyltransferase type II (GGTase-II). GGTase-II is unique in that it can modify diverse C-terminal motifs from the Rab family of proteins including CCXX, CXC, or XXCC, where cysteine residues undergo geranylgeranylation and X represents any amino acid (33–36). GGTase-II also requires Rab escort protein (REP) for substrate recognition. REP binds to the substrate motif to be prenylated and presents the catalytic site of GGTase-II for prenylation (37). Recently, a fourth non-canonical prenyltransferase, GGTase-III, was identified which adds a second isoprenoid geranylgeranyl group onto a pre-farnesylated protein substrate with a -CC_FarAIM C-terminal motif (38, 39). Interestingly, a chemical proteomic study described the discovery of non-canonical prenylated protein, ALDH9A1, which lacks any classic prenylation motif (40). Moreover, this modification was found to be reversible and not inhibited by known prenyltransferase inhibitors suggesting the possibility of a yet-to-be identified prenyltransferase (40). Single cell RNA seq data analysis of the developing mouse retina (41, 42) suggests the expression of this non-canonical prenylated protein in majority of cell types in the developing retina. Taken together, the discoveries of non-canonical C-terminal sequences for FTase recognition, GGTase-III, and the possibility of other non-canonical prenyltransferases have challenged our understanding of this modification and have greatly expanded the potential pool of prenylated proteins.

The mechanistic understanding of the isoprene substrate (FPP vs. GGPP) specificity of prenyltransferases stems from structural studies, structure-function studies, peptide library studies, computational approaches, radiolabeling, and affinity tagging approaches (22, 43–45). Both FTase and GGTase-I are heterodimeric metalloenzymes consisting of an identical α subunit and distinct, homologous β subunits (46–48). The isoprene-binding site of both enzymes is constituted by the interface of the α and β subunits, and predominantly consists of amino acid residues of the β subunit. The isoprene binding pocket of both GGTase-I and FTase is lined with conserved aromatic residues. The selectivity of FTase for the shorter FPP (5 carbon) substrate is explained by the presence of bulky tryptophan and tyrosine residues of the β subunit (W102 and Y365), within the FPP binding pocket. This blocks the binding of the larger 20-carbon GGPP prenyl moiety (47).

To identify the prenyltransferase enzyme and target protein interactions responsible for substrate selectivity, a series of peptide substrates were co-crystallized with each prenyltransferase (26). This study revealed that the various CAAX sequences bind and interact along one face of the funnel-shaped peptide-binding site in both prenyltransferases except for the C-terminal X residue. FTase utilizes two complementary binding pockets for the X residue depending on the identity of that residue. For example, an X residue of S, Q and M interacts with a binding pocket comprised of Y131α, A98β, S99β, W102β, H149β, A151β and P152β, while F, L, H and N residues interact with a separate binding pocket comprised of residues L96, S99, W102, W106, A151 of the β subunit (26). GGTase-I differs greatly in its X residue specificity in that it is composed of one binding pocket which contains the residues T49, H121, A123, F174 of its β subunit. GGTase-I appears to prefer hydrophobic residues at the X position in its target substrates (26). Taken together, differential binding of prenyltransferases for the X residue in the CAAX domain, and the presence or absence of bulky residues in the isoprene-binding site of FTase and GGTase-I underlie the difference between protein farnesylation and geranylgeranylation.

The first-described example of prenylation defect-induced visual system dysfunction was X-linked choroideremia, wherein point deletion mutation in the gene CHM/REP1 (Chromosome X, location: Xq21.2), encoding Rab escort protein 1 (REP-1) was identified. REP-1 serves as a chaperone for several Rab proteins and is a crucial accessory protein for GGTase-II. Mutations in REP1 causes progressive mid-peripheral atrophy of the retina, RPE and the choroid, resulting in progressive vision loss, ultimately leading to blindness (60). More specifically, male patients initially present with night blindness in childhood, which then advances to peripheral visual field impairment and ultimately to total blindness later in life (61). Whereas female carriers are usually asymptomatic but may exhibit a unique speckling pattern on fundus autofluorescence imaging, and sometimes night blindness as well (62).

REP-1 chaperones unprenylated Rabs and presents the cargo to GGTase-II for geranylgeranylation (63, 64). REP-1 deficiency leads to the accumulation of unprenylated Rab proteins in the retina (65). Immunolabeling studies showed that REP-1 is expressed in both rods and cones, predominantly in the inner segment and the perinuclear cytoplasm (66). Interestingly, while CHM gene expression is found in extraocular tissues as well, the gene mutation predominantly affects the retina and choroid in patients. This is due to shared functional redundancies between REP-2, which may partially compensate for the geranylgeranylation of Rab GTPases in extraocular tissues (67). For example, it was shown that while Rab27a exhibits a similar affinity for REP-1 and REP-2, REP-1-Rab27a complex ( Figure 3 ) has a greater preference for GGTase-II compared to the REP-2-Rab27a complex in the retina (69).

The notion of CHM manifesting as a non-syndromic retinopathy has been challenged recently (70). Whole metabolomic analysis of plasma samples from 25 CHM patients versus age- and sex-matched controls showed plasma alterations in oxidative stress-related metabolites, in addition to alterations in tryptophan metabolism, leading to significantly elevated serotonin levels (70). Lipid metabolism was found to be disrupted in CHM patients with evidence of dysfunctional lipid oxidation, and dyshomeostasis in several sphingolipids and glycerophospholipid levels (70). Aberrations in sphingolipid metabolism has been implicated in neurodegenerative diseases, metabolic disorders, immune function, and cancer (71).

Several preclinical and clinical studies targeting CHM have yielded important insights into factors that affect developing effective therapeutic interventions (72–78). In vivo studies on CHM gene transfer utilizing HIV-based lentiviral vectors led to a decrease in the geranylgeranylated Rab protein load. However, this approach also showed limited transfection of PRs by lentiviral vectors (73). Alternatively, adeno-associated virus (AAV) vectors, such as adeno-associated virus serotype 2 (AAV2) mediated delivery of CHM have been shown to achieve efficient transduction of REP1 in PRs and RPE, with an acceptable safety profile (72, 79–81). However, recent results from phase III of the Efficacy and Safety of BIIB111 for the Treatment of Choroideremia (STAR) trial did not meet the primary endpoint of best-corrected visual acuity (BCVA) for FDA approval. The work highlights the need to consider the stage of disease, anatomical differences, surgical variability, and dose administration among participants (76). Gene therapy for choroideremia may still provide significant visual acuity gains if treated early. Episomal scaffold/matrix attachment region (S/MAR)-based plasmid vectors carrying the human CHM gene in CHM patient-derived fibroblasts and a CHM mutant zebrafish model showed some promise. This may provide an alternative delivery strategy for gene augmentation in CHM patients (82).

Several frameshift, missense, and nonsense mutations in the gene coding for Aryl hydrocarbon-interacting protein-like 1 (AIPL1) are reportedly associated with autosomal recessive Leber congenital amaurosis type IV (LCA4) (OMIM #604393) (111, 112). Patients with LCA4 typically present with early-onset severe visual impairment, nystagmus, and poor pupillary light responses compared to other LCA types (113, 114). Both scotopic and photopic Electroretinography (ERG) responses were severely diminished in rods and cones, respectively. Disc drusen, optic disc edema, and macular atrophy have also been noted (113, 114). Optical coherence tomography (OCT) imaging in older patients showed severe cone dystrophy (114).

AIPL1 is a specialized PR-specific co-chaperone required for PDE6 maturation and stabilization (115–117). Retinal degeneration caused by AIPL1 defects were described in two mouse models of LCA4 (115, 116). In Aipl1 knockout mouse model, rapid rod-cone dystrophy was observed by postnatal 3 weeks of age (115). While the AIPL1 h/h mouse model, carrying homozygous, hypomorphic mutant allele for AIPL1, causes diminished expression of AIPL1, resulting in relatively slower PR dystrophy beginning after 12 weeks of age with almost complete loss of PRs by 5 months of age (116). In both models, PDE6 expression was found to be drastically perturbed, suggesting an indispensable role of AIPL1 in PDE6 maturation.

The AIPL1 protein contains an N-terminal FK506-binding protein (FKBP) domain and a C-terminal tetratricopeptide repeat (TPR) domain with three tetratricopeptide repeats (111). Structural studies confirmed that the PDE6 prenylation is required for its binding to the FKBP domain of AIPL1 (91, 118). The TPR domain serves as a binding site for cytosolic HSP90 and the regulatory Pγ subunit of the PDE6 complex (119, 120). Interaction of PDE6 with the TPR domain of AIP1L also plays a key role in the chaperone-dependent folding and maturation of the PDE6 complex (117, 121). Intriguingly, AIPL1 appears to be colocalized predominantly from the synapse to the inner segment of PR cells, with an enrichment in the connecting cilium, while the PDE6 complex is eventually trafficked to the outer segment (122). This may suggest that AIPL1 chaperone function is limited to PDE6 maturation and does not extend to the trafficking of PDE6 to the outer segments. A recent transgenic mouse study showed that targeted mutation of PDE6A at the prenylation consensus cysteine target, does not affect its trafficking to the outer segment (123). Prenylation of PDE6C was found to be important for the formation of a stable ternary chaperone complex comprising of AIPL1, HSP90, and PDE6 subunits in HEK293 cells (123). Future work is required to determine if AIPL1 plays a role, if any, in a relay-type system where it offloads mature PDE6 complex to another trafficking chaperone like PrBP/δ.

Heterotrimeric G-protein transducin, Gt, is an essential signal transducer and amplifier in retinal PR cells for phototransduction (124). It is a membrane-bound G-protein comprised of α, β, and γ subunits, and is bound to a guanine diphosphate (GDP) molecule (124). Both α and γ subunits are lipid-modified, wherein the α subunit is acylated and the γ subunit is farnesylated, despite both subunits possessing a CAAX motif (125, 126). These modifications allow for outer segment membrane localization and subsequent protein interaction (127). Prenylation also enhances the binding of Gβγ to G-protein α subunits (128). The importance of Gγ subunit farnesylation was shown by Kassai and co-workers, who demonstrated that replacing the farnesyl group of Gγ with the more hydrophobic geranylgeranyl attachment rendered the Gβγ dimer incapable of undergoing light-driven translocation to rod outer segments (129).

Homozygous deletion of Gγ subunit (coded by the gene Gngt1) in mice showed rod atrophy starting at PN 1 month, with almost complete rod degeneration by 6 months of age and a significant decrease in the protein levels of Gα and Gβ, despite mRNA levels comparable to wild-type controls (130). In contrast, another study found only poor phototransduction signal amplification in intact rods and showed a decline in visual sensitivity of rods in a Gγ-deficient mouse model, without prompt retinal degeneration (131). The pivotal role of farnesylation of Gγ in mediating phototransduction was further confirmed in point mutant models where the Gγ farnesylation site was removed (132). This study showed that farnesylation was not essential for Gβγ dimerization ( Figure 5 ), but essential for localization to the rod outer segments and participation in phototransduction (132). Surprisingly, despite the clear functional role of farnesylated Gγ in the phototransduction cascade, there are no known mutations of Gngt1 linked with inherited retinopathies. This is perhaps attributable to the low frequency of mutation reported in this relatively small gene (225 bp open-reading frame) (5).

Rab small GTPases, which belong to the Ras superfamily, are key regulators of membrane trafficking and cell growth (133, 134). Rab28 carries a C-terminal CAAX motif that is farnesylated in contrast to the common Rab geranylgeranylation and was the first prenylated small Rab GTPase identified to be directly involved in an inherited retinal disease (135–137). Immunohistochemical analysis suggested that Rab28 colocalized at the PR basal body and the ciliary rootlet (137). Rab28 mutations cause cone-rod dystrophy 18 (CRD18) (OMIM #615374), presenting with central retinal atrophy and hyperpigmented appearance of the fovea. Both scotopic and photopic ERG responses were significantly diminished, suggesting decreased rod and cone response to light, respectively (136–138). A transgenic knockout mouse model of Rab28 showed Rab28 requirement for disc shedding by cone PRs and its subsequent phagocytosis. Loss of Rab28 function in knockout mice faithfully mimics the human CRD18 phenotype (139). Moreover, this study identified Rab28 protein interactors such as KCNJ13 and PrBP/δ (PRBP/δ and Rab28 complex, simulated in Figure 6 ). Autosomal recessive mutations of KCNJ13 leads to Leber congenital amaurosis type 16 (LCA16) (OMIM #614186), wherein KCNJ13 regulates phagocytic uptake of shed outer segments by the RPE (139, 140).

Autosomal dominant mutations in RPGR gene (Chromosome X, Xp11.4) account for over 70% of X-linked retinitis pigmentosa cases (XLRP) (OMIM #300455) worldwide (141). Two major spliced isoforms of RPGR have been identified in the mammalian retina: RPGR¹⁻¹⁹ and RPGR^ORF15 (142). Both protein isoforms share an identical N-terminal tandem repeat structure named RCC1-like domain (RLD), while the C-terminal domains are not conserved due to altered mRNA splicing. The C-terminus of RPGR¹⁻¹⁹ constitutes a target prenylation motif which is geranylgeranylated, whereas RPGR^ORF15 contains a Glu-Gly-rich low complexity region in the C-terminus (142, 143). The shared N-terminal RLD region may confer some functional redundancy between the two RPGR isoforms, while the differential prenylation of RPGR¹⁻¹⁹ may govern its differential subcellular colocalization and function. This has been demonstrated in vitro using transformed human RPE1 cells wherein the prenylated RPGR¹⁻¹⁹ isoform localized to the RPE primary cilia, while RPGR^ORF15 does not traffic to cilia (144, 145). Interestingly, rescue experiments in RPGR knockout mice showed that the RPGR^ORF15 isoform sufficiently rescued PR functioning (146, 147).

A report showed a decrease in RPGR¹⁻¹⁹ isoform, with concurrent upregulation in the RPGR^ORF15 isoform throughout eye development (147). Thus, suggesting a duality in the role of the two RPGR isoforms during retinal development vs.. functioning of the mature retina. However, the exact role of prenylated RPGR¹⁻¹⁹ in early retinal development has not yet been directly established. Recently, Zhang et.al., showed the effects of common missense mutations of the shared RLD on the RPGR protein interaction network, and demonstrated that the RLD domain is indispensable for interactions with PrBP/δ and INPP5E in RPE1 cells (148). Prenylation of RPGR¹⁻¹⁹ isoform is necessary for its ciliary targeting via PrBP/δ trafficking and that prenylation also enhances the RLD binding region for protein interaction (148). Future work should aim to gain the mechanistic understanding underpinning protein interaction disruption on retinal health and whether prenylation represents a dispensable modification in the setting of XLRP pathogenesis.

Mevalonate kinase (MVK) catalyzes the phosphorylation of mevalonic acid to phosphomevalonate. Autosomal recessive mutations in MVK causes metabolic disorder with poor genotype-phenotype correlation (149). It encompasses a severe form of the disease, mevalonic aciduria (MEVA) (OMIM #610377), and a milder periodic fever syndrome, hyperimmunoglobulinemia D syndrome (HIDS) (OMIM #260920) (150). HIDS is characterized by regular episodes of high fever and systemic inflammation in the affected pediatric patient population. In contrast, mevalonic aciduria patients present with chronic systemic inflammation, along with neurodevelopmental issues such as cerebellar atrophy, dystonia, and ataxia. MKD patients also present with gastrointestinal complications and retinal dystrophy (151–153). A study in 2013 showed that mutations in MVK causes late-onset non-syndromic RP, accompanied by mild MKD symptoms (154). It was thought that the accumulation of mevalonic acid may be responsible for the neuronal degeneration including, PR deterioration, observed in MKD patients (153). Two recent studies have shown that patients with MKD exhibit significant accumulation of unprenylated Rab proteins (155, 156). Therefore, the retinal dystrophy phenotype associated with MKD may arise from decreased de novo synthesis of isoprenoid metabolites (dolichol, cholesterol, FPP, and GGPP), as well as cellular insufficiency in prenylation of the target proteome.

Recent discoveries of novel prenyltansferases, targets motifs, and non-canonical mechanisms of prenylation, has increased the probability of discovering yet unidentified prenylopathies. Here, we briefly discuss the case of a rare autosomal recessive mutation that primarily affects the cis-prenyltransferase (CPT) complex involved in dolichol synthesis. The cis-prenyltransferase complex is a heterotetrameric complex comprised of the catalytic subunit dehydrodolichyldiphosphate synthase (DHDDS), and its interacting partner NUS1 or Nogo-B receptor (NgBR). The CPT complex catalyzes the serial cis-condensation of multiple IPP molecules to FPP, to ultimately generate dolichol, an isoprenoid metabolite which serves as the obligate glycan carrier, and is indispensable for protein glycosylation. We recently discussed the visual disorders pertaining to the dolichol synthesis pathway (157).

Recessive mutations in NUS1 leads to a systemic disorder including neurodevelopmental issues, retinal dystrophy, and optic neuropathy, unsurprisingly, given the requirement of dolichol synthesis and protein glycosylation in all cells (158). NUS1 performs several other critical roles in maintaining lysosomal homeostasis (159). Interestingly, a recent study showed that NgBR independently binds to farnesylated Ras protein, outside of its CPT function, and plays a key role in membrane localization of H-Ras ( Figure 8 ) (160). Moreover, NgBR binds not only to farnesylated H-Ras with its canonical CAAX (CVLS) motif, but also a recombinant geranylgeranylated H-Ras protein with GGTase-I target CAAX sequence (CLVL) (160). We have independently demonstrated that conditional ablation of Nus1 in the developing PRs, using a CRX-Cre mouse line, causes total PR and bipolar cell loss by PN 2–3 weeks (Ramachandra Rao et.al., unpublished results). However, the underlying degenerative mechanism remains to be understood. These findings highlight the need for investigating the additional chaperone-like roles of NgBR in binding and trafficking prenylated proteins in the retina.

We have provided a brief, tabular summary of the clinical features and molecular basis of known, above discussed prenylation-associated retinopathies in Table 1 . Defects in prenylation or the chaperone-mediated maturation and trafficking of key membrane-bound prenylated proteins essentially manifest as early rod-cone or cone-rod dystrophies, depending on the predominant affected cell type.

Computational approaches have been developed to enable prediction of prenylated proteome. Maurer-Stroh and Eisenhaber developed the algorithm PrePS to predict prenyltransferase substrates, based on data from experimentally derived prenylated motifs and prediction of new motifs acting as a substrate for farnesylation or geranylgeranylation (161). The training dataset consisted of 692 FTase and 486 GGTase-I substrates curated by literature survey as well as BLASTP analysis with known prenylated substrates against an NCBI database. In addition, an 11 amino acid sequence upstream of the prenyl cysteine was added to refine the algorithm. With this refinement of PrePS, the algorithm expands the rules which predict the prenylation of a substrate to a 15-amino acid sequence (162, 163). However, experimental determination of the target peptide prenylation is still necessary to determine true false positive rate of this predictive methodology. FlexPepBind is a structure-based computational modeling approach for predicting FTase substrates (162). This algorithm predicts peptide binding through a structure-based modeling approach by aligning different peptide sequences onto a template peptide-receptor complex. This structure-based predictive method was further experimentally validated using in vitro assays, wherein 26 of the 29 tested novel peptide were FTase targets suggesting a very low false negative prediction (162).

Computational approaches dependent on training datasets are expected to be biased against non-canonical, unidentified novel prenylation target motifs. To address this discrepancy, a recent study used yeast Hsp40 Ydj1p chaperone as a genetic reporter (164). Ydj1p is prenylated but is subject to the shunt pathway in which the prenylated protein does not undergo the downstream proteolytic processing and remains cytosolic. Prenylated Ydj1p produces a thermotolerant phenotype in yeast which serves as a rapid proxy to monitor protein prenylation status. Using this mutation-phenotype screen, the authors evaluated over 67,000 recombination events, correlating to about 93.5% of the total possible amino acid combinations which may code for CAAX motif. The hits identified from the Ydj1p-based genetic screen did not greatly overlap with those found in other screenings that utilized common targets such as Ras, suggesting that a much larger set of motifs may be serving as prenylation motifs than was previously thought (164). Only a small subset of prenylated motifs identified using the Ydj1p-based screening were identified as having a high probability of prenylation by FlexPepBind and PrePS with results of 27% and 7%, respectively. However, it is important to note that this discrepancy may also reflect different target peptide-specificities for the mammalian and yeast prenyltransferases enzymes.

Data mining methods using experimentally determined retinal proteome and transcriptome data may be highly effective in predicting the retinal prenylome, including non-canonical prenylation motifs. Applying targeted data mining to the transcriptome or proteome specific to the RPE and neural retina—separated into macular and peripheral regions—could be particularly powerful (165). These methods may help distinguish the prenylome specific to the human macula and peripheral retina and allow for better understanding of protein prenylation-associated vision disorders. Furthermore, pathway analysis tools like Ingenuity Pathway Analysis may reveal the potential cellular and retina-specific functions of these predicted prenylated proteins (166). The latest advancements in AI-driven prediction of protein structures, protein-protein interactions, and protein-ligand interactions such as AlphaFold2 and AlphaFold3, could provide preliminary computational validation of these predicted retinal prenylomes (167). It is important to acknowledge that novel hits arising from predictive, computational methodologies need to be experimentally validated using in vivo and in vitro approaches, as discussed below.

Fluorescence tagging of proteins has been extensively utilized to study prenylation at endogenous levels, providing insights into prenylated protein maturation and trafficking, and has also facilitated studies of prenyltransferase inhibitors since membrane localization is an easily measurable experimental endpoint. Diffuse fluorescence throughout the cell indicates cytoplasmic localization, which implies a protein either not undergoing prenylation or becoming prenylated without further processing as noted in the shunt pathway. For instance, the effect of proteolysis on membrane colocalization of K-Ras protein was determined using the above discussed chimeric approach (168). Using Rce1 ^+/+ and Rce1 ^-/- fibroblast cells transfected with a chimeric GFP-K-Ras construct, it was shown that fluorescence was localized to the plasma membrane in Rce1 ^+/+ cells but diffuse in Rce1 ^-/- cells, suggesting that proteolysis was necessary for Ras protein membrane localization (168).

An alternative approach for detecting protein prenylation inside mammalian systems that is not reliant on membrane localization as proxy endpoint, is called Protein Lipidation Quantification (PLQ). This method is derived from an established method of Micellar Electro-Kinetic Chromatography (MEKC) that allows selective partition of prenylated proteins into detergent micelles (using SDS) during capillary electrophoresis (169, 170). Detergent micelles serve as pseudo-stationary phase and exhibit differential electrophoretic migration compared to unprenylated free analyte. PLQ has been successfully used to separate prenylated full length fluorescent protein with a C-terminus CAAX motif in in-cell studies (171).

Chemical proteomic approaches facilitate the direct detection of prenylation in vivo and are based on the development of analogues of FPP and GGPP that are functionalized with biochemical affinity tags, immunogenic tags for detection by antibodies, and chemo-selective tags for bio-orthogonal labeling. Spielmann and coworkers showed the application of an immunogenic tag by using anilinogeraniol (AGOH) to detect FTase substrates (172). Anilinogeraniol is an analogue of an upstream precursor of FPP, which undergoes cellular kinase-dependent diphosphorylation to 8-anilinogeranyl diphosphate (AGPP), and where the third isoprene unit of FPP is replaced with an aniline moiety that then serves as an epitope for detection by Western blot (172, 173). Biotin-geranyl diphosphate (BGPP) has been generated for use as an affinity tag (174). BGPP only allowed for the efficient identification of GGTase-II substrates as the bulky nature of the biotin group was found to interfere with the protein substrate binding for FTase and GGTase-I. Engineered FTase and GGTase-I variants can utilize BGPP as a donor for protein modification (174). Lysates from cells expressing wild-type or the engineered prenyltransferase variants were incubated with BGPP, followed by avidin pull down and proteomic analysis, which allowed the identification of GGTase-II target proteins. Several Rab proteins such as Rab7 and Rab27, which were differentially prenylated in WT prenyltransferase expressing cells, were identified as GGTase-II substrates (174).

Distefano and coworkers designed alkynylated FPP analogue C15AlkOPP, which has been used to study protein prenylation inhibitors, labeling of proteins sensitive to human pathogens, and the delineation of the prenylome of Plasmodium falciparum, the causative agent of malaria (175–180). More recently, new alkyne-tagged isoprenoid analogues which closely mimic FPP (YnF) and GGPP (YnGG) allowed click-chemistry based detection of prenylated proteins via mass spectrometry in a human endothelial cell line (181). These analogues allow quantitative analysis of endogenous prenylated proteome since they do not alter the specific activity of prenyltransferases. Upon addition of the alkyne tagged analogue, proteins are captured via click CuAAC ligation to azide-containing fluor/biotin-tagged reagents. This facilitates the enrichment of prenylated proteins and subsequent analysis via LC-MS/MS. These analogues also proved useful in investigating protein prenylation in vivo. Intracerebroventricular (ICV) injection of the C15AlkOPP analogue in brains of a transgenic mouse model (APP/PSI) of Alzheimer’s Disease found the upregulation of several prenylated proteins in diseased mice (182). The development of these protein prenylation probes has greatly enhanced our understanding of protein prenylation biochemistry. Appropriate application of similar approaches in in vitro and in vivo models may prove fruitful in delineating the underlying degenerative mechanisms of prenylation-associated retinopathies.

Experimental determination of retinal prenylome in vivo is challenging since current prenylation screening assays utilize in vitro approaches. While proteomic and lipidomic strategies have proved to be powerful in refining our understanding of the prenylome, these approaches are expensive and labor-intensive. However, both the in vitro as well as the proteomic/lipidomic approaches described above may be highly relevant in defining the retinal cell type-specific prenylome. For instance, RPE prenylome and target-specific prenylation mechanisms may be easily investigated by applying such approaches in induced pluripotent stem cell (iPSC)-derived RPE cells, as well as primary RPE cells (183). Similarly, primary Müller glia and ganglion cell cultures may also be extremely useful in determining retinal cell type-specific prenylome. Identification of novel retinal cell type-specific prenylated hits may then be further validated in vivo using careful conditional gene ablation approaches.