Progress in treating inherited retinal diseases: Early subretinal gene therapy clinical trials and candidates for future initiatives

Abstract

Due to improved phenotyping and genetic characterization, the field of ‘incurable’ and ‘blinding’ inherited retinal diseases (IRDs) has moved substantially forward. Decades of ascertainment of IRD patient data from Philadelphia and Toronto centers illustrate the progress from Mendelian genetic types to molecular diagnoses. Molecular genetics have been used not only to clarify diagnoses and to direct counseling but also to enable the first clinical trials of gene-based treatment in these diseases. An overview of the recent reports of gene augmentation clinical trials by subretinal injections is used to reflect on the reasons why there has been limited success in this early venture into therapy. These first-in human experiences have taught that there is a need for advancing the techniques of delivery of the gene products - not only for refining further subretinal trials, but also for evaluating intravitreal delivery. Candidate IRDs for intravitreal gene delivery are then suggested to illustrate some of the disorders that may be amenable to improvement of remaining central vision with the least photoreceptor trauma. A more detailed understanding of the human IRDs to be considered for therapy and the calculated potential for efficacy should be among the routine prerequisites for initiating a clinical trial.

1. Introduction

After a relatively long journey of clinical and scientific progress, we have arrived at a time when therapeutic strategies are not only being discussed but also initiated for previously incurable orphan genetic retinal diseases. In the past three decades, there has been an ever-increasing list of causative genes associated with inherited retinal diseases (IRDs) and this list has been continually updated in reviews (for example, Iannaccone, 2005; Bramall et al., 2010; Wright et al., 2010; Verbakel et al., 2018) in addition to websites (Retnet, https://sph.uth.edu/retnet/; European Retinal Disease Consortium, https://www.erdc.info). Many reports have now tabulated ongoing and planned clinical trials of gene-based therapies for IRDs (for example, Trapani and Auricchio, 2018; Arbabi et al., 2019; Lee et al., 2019; Ramlogan-Steel et al., 2019). Missing in the current literature, however, are detailed explanations of the criteria used for selecting the first IRDs that have entered clinical trials. Also needed are assessments of what we have learned as a result of these major efforts by patients, doctors, administrators, regulators, and sponsors, and, when not successful, analysis of cause(s) of failure. In the present work, we review and discuss criteria for targeting an IRD for a clinical trial and make suggestions about which diseases (and stages of these diseases) warrant consideration as candidates for upcoming therapy.

2. Historical perspective: Mendelian genetic percentages to genes and mutations

By the early 1980’s, data on genetic percentages of IRDs from centers in different countries were already reported. The tallies represented first attempts to understand the genetic heterogeneity of these diseases and eventually led to linkage mapping of disease loci and finally gene and mutation identification. There were differences in percentages of the Mendelian inheritance types reported (Amman et al., 1965; Panteleeva, 1969; Jay, 1972; Bird, 1975; Fishman, 1978; Hu, 1982; Boughman and Fishman, 1983). For example, a common finding was that the largest subgroup of inheritance patterns, ranging from 36–92%, in most countries was autosomal recessive (AR) or presumed AR in the case of simplex and multiplex cases. In countries other than the UK, X-linked disease was ≤10%; UK results were 25%. There was a range of contribution of autosomal dominant (AD) inheritance from 5–39% depending on the population studied. In this premolecular era, questions arose as to the reason for such potentially interesting differences, such as different genetic pools (Fishman, 1978). At least some of the X-linked variation could be attributed to whether or not examiners evaluated the ophthalmoscopic appearance of mothers and other female relatives within a family with single or multiple affected men in one generation only. The heterozygous state of X-linked RP may be detectable by ophthalmoscopy as a patchy, and often more regional, pigmentary retinopathy or a tapetal-like reflex. A full blown disease, due to unfavorable lyonization, can also occur (Bird, 1975; Cideciyan and Jacobson, 1994).

Two North American centers that specialize in IRDs and originally collaborated to perform one of the first retinal gene therapy trials (Cideciyan et al., 2008) surveyed the genetic makeup of their patient populations. One center was the Center for Hereditary Retinal Degenerations (CHRD), first established at Bascom Palmer Eye Institute at the University of Miami and then relocated to Scheie Eye Institute at the University of Pennsylvania; CHRD is primarily an adult center specializing in IRDs. At the CHRD (henceforth named the Philadelphia center) between 1983 and 2019, 1656 patients diagnosed with an IRD were clinically and molecularly evaluated (male n=877; female n=779; mean age of first visit, 36.8 years; median age of first visit, 35.5 years; range of ages at first visit, 6 months – 89 years). The second center was the Department of Ophthalmology and Vision Sciences, Hospital for Sick Children, University of Toronto, Toronto, Canada, which is primarily a children’s center and a referral center for IRDs. At the Hospital for Sick Children (named the Toronto center) between 2005 and 2016, 1148 unrelated IRD patients underwent genetic testing (male, n=688; female n=460; mean age of first visit, 16 years; median age of first visit, 10 years; range of ages at first visit, 3 months – 77 years). From this 20 year difference in the mean age of first visit between the two centers, we could expect differences in the composition of IRDs; however, we would also expect similarities. There is variability in the age at which the disease is first detected for many IRDs, and this is not necessarily directly related to age of onset of symptoms or age of disease onset. For example, in IRD families with histories of affected members in various generations, proactive members may be examined and diagnosed pre-symptomatically. Acknowledging many caveats, the two sites generally serve different patient populations and we decided to compare the results.

In a survey of 300 IRD patients evaluated before 1990 at the Philadelphia center, AR/simplex/multiplex pattern of inheritance accounted for 63% (189 patients), and for the AD inheritance, it was 23% (70 patients); these figures are similar to many other series in that era. The X-linked mode of inheritance group was 14% (41 patients) which is between the higher number reported for the UK and the very low percentages of other countries (Figure 1A). The ocular fundi of female relatives of simplex males were routinely examined at the Philadelphia center, following the observations in UK studies (Bird, 1975). The decades following 1990 to the present day witnessed the discovery of >250 causative IRD genes (Duncan et al., 2018; Retnet, https://sph.uth.edu/retnet/). Once the causal genes began to be identified, collaboration occurred with many molecular scientists interested in gene discovery and, in more recent years, samples were sent to commercial laboratories to determine the molecular diagnosis of the IRDs. With a continuous effort to the present day, the number of unique genes that cause IRDs identified within our patient population has steadily increased. The genes unique to the Philadelphia center and to the Toronto center and those common to both centers are listed (Table 1).

Fig 1.

Mendelian patterns of inheritance, resolution of genotype over time and comparison of genotypes in IRD centers. (A-C) Data from the Philadelphia IRD center. (A) Pie chart showing the Mendelian patterns of inheritance in clinically-diagnosed patients seen prior to 1990. (B) Cumulative number of unique genes identified as causal over time. (C) The proportion of patients with unresolved and resolved genotypes in the current patient population and the comparison of Mendelian patterns of inheritance for patients with resolved and unresolved molecular diagnoses. (D) Comparison of the genes identified in the Philadelphia and Toronto cohorts that each cause disease in ≥2% of patients. Common top genes are given in the center. Blue, Philadelphia cohort; Orange, Toronto cohort. (E) Comparison of the top 5 disease-causing genes in the Philadelphia and Toronto cohorts along with percentages of the cohorts that are positive for each disease-causing gene. Blue (solid and hashed), Philadelphia; Orange (solid and hashed), Toronto. Solid bars represent genes that are in the top 5 for respective center; hashed bars indicate genes that are not in the top 5 for that center but still contribute to the overall cohort.

Table 1.

Disease-causing genes observed in the Philadelphia and Toronto IRD Centers

Genes unique to Philadelphia Center (8 genes)	Genes common to both Centers (58 genes)			Genes unique to Toronto Center (32 genes)
FLVCR1	ABCA4	CRX	PRPF31	AHI1	GRK1
IMPDH1	ADGRV1	DHDDS	PRPH2	ALMS1	IFT140
INPP5E	AIPL1	EYS	RDH12	ATF6	IMPG1
LCA5	BBS1	FAM161A	RDH5	BBS5	IMPG2
RLBP1	BBS2	GNAT2	RHO	BBS7	OFD1
RP1L1	BBS4	GUCA1A	RP1	BBS8	PDE6A
SCA7	BBS6	GUCY2D	RP2	BBS9	PDE6C
SNRNP200	BBS10	KCNV2	RPE65	BBS12	PNPLA6
	C1QTNF5	KLHL7	RPGR	BBS16	POC1B
	C2orf71	MAK	RPGRIP1	C8ORF37	PRPF3
	CDH23	MERTK	SPATA7	CA4	PRPF8
	CDHR1	MFRP	TIMP3	CDH3	RD3
	CEP290	MYO7A	TOPORS	CNGA1	SLC4A11
	CERKL	NMNAT1	TULP1	CNNM4	TMEM216
	CHM	NPHP5	USH1C	CYP4V2	TRPM1
	CLRN1	NR2E3	USH2A	DRAM2	USH1D
	CNGA3	opsin gene arraya	VMD2
	CNGB1	PCDH15	WDR19
	CNGB3	PDE6B
	CRB1	PROM1

^aOpsin gene array includes the OPN1LW and OPN1MW genes and the locus control region (LCR)

Of the 1656 patients who underwent genetic testing at the Philadelphia center, about half of the patients (48%) remain without a known genotype (Figure 1C); this is similar to previously published series investigating the molecular diagnoses of IRDs (Haer-Wigman et al., 2017; Carrigan et al., 2016; Ellingford et al., 2016; Bernardis et al., 2016; Huang et al., 2015; Glöckle et al., 2014). Of the patients with resolved genotypes, 59% (509 patients) are AR/simplex/multiplex; 24% (211) have an AD mode of inheritance; and 17% (146) have X-linked disease, indicating that the inheritance pattern composition of the genotypically-resolved patients remains generally the same through 2019 as compared to 1989. For patients whose molecular diagnosis remains unresolved through 2019, the overwhelming majority (80%, 635) has AR/simplex/multiplex disease; 16% (125) have AD disease; and 4% (30) have an X-linked mode of inheritance. The observations were comparable in the Toronto cohort. This is consistent with data from other groups that analyzed smaller IRD cohorts in which unresolved genotypes are composed primarily of autosomal recessive and sporadic cases (Audo et al., 2012; Bernardis et al, 2016; Huang et al., 2015). We learned that the likelihood of identifying a causal gene was increased if the inheritance pattern was known and if the phenotype was clearly defined.

Of the genes responsible for causing ≥ 2% of the IRDs, there were 11 that were common to both the Toronto and the Philadelphia cohorts (Figure 1D). The genes that comprise at least 2% of their respective cohorts account for 77% of the IRDs in the Philadelphia cohort and 63% of IRDs in the Toronto cohort. Of the top 5 genes in each cohort, only ABCA4, the most commonly-mutated gene in both cohorts, was shared. For Philadelphia, the remaining 4 top disease-causing genes identified were RHO, RPGR, PRPH2, and RPE65, whereas for Toronto, the remaining 4 were USH2A, VMD2, CHM and CNGA3 (Figure 1E). Both cohorts have patients with mutations in all of these genes, but there are differences between the populations in genetic composition of the patients. Many factors could be contributing to this, such as the ages of the patient populations, availability and methodology of genetic testing, and biases resulting from recruitment of certain populations stemming from research interests. The top disease-causing genes identified in Philadelphia and Toronto are similar to those in other surveys of inherited eye diseases or IRDs; ABCA4 and USH2A are among the top 3 genes in other reported cohorts (Haer-Wigman et al., 2017; Bernardis et al., 2016; Huang et al., 2015; Glöckle et al., 2014). In a recent estimation of the frequency of IRD genes, ABCA4, USH2A, RPGR, RHO and PRPH2 are listed as the most prevalent disease-causing genes in the United States (Stone et al., 2017).

A logical next step from clinical and then molecular diagnostics is to use this information to design and implement gene-based clinical trials in select IRDs. And this has been occurring.

3. From molecular genetic diagnoses to subretinal gene therapies

Milestones toward ocular gene therapy were present long before IRD causes were identified (Lee et al., 2019). Given the underlying molecular causes of IRDs, gene-based therapies were launched and there is now a list of clinical treatment trials of IRDs (Trapani and Auricchio, 2018; Ramlogan-Steel et al., 2019; Arbabi et al., 2019; Lee et al., 2019). What lessons have we learned from the first attempts to treat IRDs with a known genetic cause? To seek answers to this question, we provide a perspective on each of the 10 IRDs treated to date with subretinal gene therapy (Table 2) for which information is publicly available. Within these summaries, we will mention pre-clinical proof-of-concept studies, pre-clinical safety studies, clinical efficacy outcome measures and clinical natural history studies supporting each clinical trial. Our focus in this manuscript is on gene augmentation for specific IRDs; the interesting topic of gene therapies designed to target multiple IRDs is not covered (for example, Yu et al., 2017). There is also no attempt to discuss other promising therapies, such as stem-cell approaches, optogenetics or both (for example, Garita-Hernandez et al., 2019).

Table 2.

Subretinal gene therapy trials of inherited retinal degenerations

Mendelian Genetic Type	Disease	Gene	Cell(s) with primary genetic defect	Estimated Gene Mutation Frequency in the USa	Proof of Concept and Safetyb	Dissociation of Structure-Functionc	Clinical trial number (Phase)d
Autosomal Recessive	LCA2	RPE65	RPE (and cone)	1/576,667	c, m, nhp	+ for cones + for rods	NCT00643747 (Phase I/II)e NCT00481546 (Phase I) NCT00516477 (Phase I) NCT00821340 (Phase I)e NCT00749957 (Phase I/II)e NCT01208389 (Phase I/II) NCT01496040 (Phase I/II)e NCT00999609 (Phase III) NCT02781480 (Phase I/II)e
	RP38	MERTK	RPE	1/576,667	m, nhp	NP	NCT01482195 (Phase I)
	USH1B	MYO7A	Rod, cone (and RPE)	1/216,250	Complex (m), nhp	− for rods NP for cones	NCT01505062 (Phase I/II)
	STGD	ABCA4	Rod,cone (and RPE)	1/10,000	Complex (m), r, nhp	NP	NCT01367444 (Phase I/II)
	ACHM2	CNGA3	Cone	1/576,667	o, m, nhp	+/− for cones	NCT02610582 (Phase I/II) NCT02935517 (Phase I/II) NCT03758404 (Phase I/II)
	ACHM3	CNGB3	Cone	1/346,000	c, m, nhp	+/− for cones	NCT03001310 (Phase I/II) NCT02599922 (Phase I/II) NCT03278873 (Phase I/II)
	RP40	PDE6B	Rod	1/247,143	c, m	NP	NCT03328130 (Phase I/II)
	RP	RLBP1	RPE and Muller cells	1/1,730,000	m, nhp	+/− for rods +/− for cones	NCT03374657 (Phase I/II)
X-linked	CHM	CHM	Rod, cone (and RPE)	1/123,571	m	NP	NCT01461213 (Phase I/II) NCT02341807 (Phase I/II) NCT02077361 (Phase I/II)e NCT02553135 (Phase II)e NCT02671539 (Phase II) NCT02407678 (Phase II) NCT03507686 (Phase II) NCT03496012 (Phase III)
	XLRP (RP3)	RPGR	Rod and cone	1/36,042	c, m	− for foveal cones NP for extrafoveal cones NP for rods	NCT03116113 (Phase I/II) NCT03252847 (Phase I/II) NCT03316560 (Phase I/II)
Autosomal dominant	None to date

^aEstimated frequency from Stone EM, et al. Ophthalmology. 2017;124(9):1314-1331.

^bAnimal studies: c, canine; m, murine; nhp, non-human primate; o, ovine; r, rabbit

^cGradation:

− No dissociation

+/− Qualitative dissociation: Evidence for existence of photoreceptor cell of interest lacking normal function implying improvement possible

+ Quantitative dissociation: Colocalized measure of retinal photoreceptor structure and light sensitivity defining vision improvement potential

NP Not evaluated or unknown

^dIn order of trial start date

^eCompleted trials

3.1. RPE65-LCA:

The form of Leber congenital amaurosis (LCA) due to mutations in the RPE65 gene (responsible for the retinoid isomerohydrolase of the retinal pigment epithelium, RPE; Hamel et al., 1993, 1994; Marlhens et al., 1997; Redmond et al., 1998) was the first IRD where a pre-clinical gene therapy success was translated to the clinic. Scientific discoveries leading to understanding of the molecular pathways in the mammalian visual cycle occurred well over 20 years preceding the concept of human gene therapy trials. Of course, the discovered association of mutations in the RPE65 gene with LCA was critical (Hamel et al., 1994; Marlhens et al., 1997) and now we know that LCA is caused by many different genes and mutations therein (Kumaran et al., 2017). Preclinical studies using the canine disease model of RPE65-LCA yielded convincing proof-of-concept results; there was dramatic improvement of visual function after subretinal gene augmentation with adeno-associated virus (AAV) vector. This stimulated the interest of the more clinically-minded scientists; and, as such, translation to the clinic became imaginable (Acland et al., 2001, 2005; Cideciyan, 2010).

How did we prepare to perform this first-in-human clinical trial? Safety of the gene-vector product was studied in both affected canines and in normal non-human primates, and the data published (Jacobson et al., 2006a, 2006b). Of interest, these toxicity studies were performed without the use of systemic steroids. Efficacy outcomes to quantify the severely impaired vision with nystagmus in LCA had to be devised anew because traditional outcomes used for other eye diseases (Csaky et al., 2008) were not as helpful. Hence, the FST (full-field stimulus test) was developed (Roman et al., 2005, 2007). The natural history of the human disease was not known at the time the trials began. The advent of cross-sectional in vivo imaging by optical coherence tomography (OCT) would have permitted a pre-trial natural history to be performed but there was an urgent desire by investigators to translate to the clinic and natural history studies require patients, time and resources. What information was there about the human disease to be treated? There were brief clinical descriptions in the reports of gene causation (for example, Lotery et al., 2000) or visual function and retinal structure in rare cases (van Hooser et al., 2000). We knew that there was serious visual dysfunction and that electroretinograms (ERGs) were abnormal for rod- and cone-mediated signals at all ages; there could even be non-detectable ERGs in the first years of life (Jacobson et al., 2009a). To deliver a local subretinal gene delivery, however, it seemed necessary to have a strategy about where this injection should be placed in the retina. Although it would be far more convenient to design a trial with a standardized bleb location in every patient (thinking ahead to multi-center trials in late phase trials), targeting a region of retina without photoreceptors would not be sensible. Photoreceptor topography using OCT was thus performed (Jacobson et al., 2008a). Normal ONL thickness topography represents a composite measure of cone and rod photoreceptor nuclei: there is peak ONL thickness at the fovea due to the high cone density and a further area of high density is at 10–16° superior to the fovea, the rod hot spot (Curcio et al.,1990; Jacobson et al., 2008a; Figure 2A). In RPE65-LCA patients there was loss of photoreceptors even in the youngest patients studied and there were different patterns to the preserved ONL (Jacobson et al., 2008a). Examples of six young patients are shown (Figure 2B–G). Two 7-year-olds had abnormally thinned ONL; one patient maintained a central island of ONL surrounded by barely detectable ONL and more superior than inferior pericentral structure (Figure 2B). The other 7-year old had greater ONL thickness, albeit still reduced, and an inferior retinal island of greater abnormality (Figure 2C). Two other patients, both age 12, had thicker ONL in the superior and temporal retina compared with inferiorly. They differed in the degree of foveal and parafoveal ONL thickness (Figure 2D,E). A 13- and a 14-year old patient retained central islands of ONL. Beyond this central patch, there was very limited preservation of ONL in the scanned area in the 13-year old; but there were preserved islands beyond the arcades in the 14-year old patient (Figure 2F,G). Such findings led to the suggestion that it would be judicious for ONL mapping to be used to guide location of the therapeutic bleb (Jacobson et al., 2008a). Targeting subretinal gene delivery based on ONL mapping occurred in only one of the three initial clinical trials (Cideciyan et al., 2008).

Fig 2.

Photoreceptor layer thickness topography in young patients with RPE65-LCA (A). Left. OCT cross-sectional images through the horizontal (upper scan) and vertical (lower scan) meridians of a normal subject (icons at upper right of scans are the location of the image on a schematic fundus; ONL, highlighted in blue). Right. Normal ONL thickness topography with color scale above. Hatched area, location of the optic nerve head. (B-G) ONL thickness topography in 6 young patients with RPE65-LCA, illustrating that there can be abnormalities at these ages and differences between patients despite similar ages. All eyes are depicted as right eyes. F, fovea. Modified from Jacobson et al., 2008a, copyright by the Association for Research in Vision and Ophthalmology.

What did we learn from the first-in-human retinal therapy of an IRD? Several independent clinical trials were performed concurrently and others followed (Table 2). Common to all the trials were reports of safety issues that mainly related to the subretinal surgical procedure (Bainbridge et al., 2008; Maguire et al., 2008; Hauswirth et al., 2008; Cideciyan et al., 2008; Jacobson et al., 2012). Even in the early phase safety trials, efficacy was present in the retinal area of treatment. Was this predictable or just lucky? Evidence for dissociation of structure and function in RPE65-LCA patients across a wide age spectrum was provided before the clinical trial began. In other words, biochemical recovery of the visual cycle pathway could theoretically occur and activate residual photoreceptors (Jacobson et al., 2005).

To test this hypothesis, we determined the co-localized relationship between retinal structure and function in two patients (Cideciyan et al., 2008). Outer photoreceptor nuclear layer (ONL) thickness was 29–43% of mean normal thickness (Fig.3A). Patients with retinitis pigmentosa with degenerative photoreceptor loss but without RPE65 mutations would be expected to show sensitivity losses ranging from 0.7–1.1 log units (Figure 3B). Pretreatment conservative estimates of the loss of rod function for RPE65-LCA patients were 6.4–7.2 log units. After treatment, one patient showed a 2.3 log unit increase in rod sensitivity, which approached, but did not reach, the predictions of the simple photoreceptor degeneration (Figure 3B). Another patient showed a posttreatment visual gain of 4.8 log units and the result became no different from that expected of simpler diseases with only a degenerative component contributing to the vision loss (Figure 3B).

Fig 3.

Early effects on rod-mediated vision of gene augmentation therapy in patients with RPE65-LCA. (A) OCT cross-sectional images (grayscale) from a normal subject and two patients (P2, P1) with RPE65-LCA. The sections are taken from ~5 mm superior retina in the normal subject and P2; section in P1 is from ~5 mm inferior retina. Schematics of ONL and RPE layers are overlaid (in yellow) on the OCT and illustrate the interpretation of reduced outer retinal structure in patients compared to normal. (B) Relationship of ONL thickness and rod-mediated sensitivity loss in normal subjects (triangles) and the two patients before therapy (unfilled large symbols) and after (up to 90 days; filled large symbols) gene therapy. The effect of the therapy led to rod sensitivity increase of ~5 log units (P2) and ~2 log units (P1). Ellipse, normal variability; broken lines, model of the expected relationship between structure and function based on quantum catch. Modified from Cideciyan et al., 2008, copyright by the National Academy of Sciences.

The dissociation of structure and function in this type of LCA thus stands as a feature worth measuring in future candidate diseases before clinical trials are initiated as it directs methods used to establish outcomes. There was no consensus, however, about the longevity of the efficacy (Cideciyan et al., 2013a; Jacobson et al., 2015; Bainbridge et al., 2015; Maguire et al., 2019). Evidence from two different groups indicated that there was continued retinal degeneration and eventual localized loss of visual function after therapy (Cideciyan et al., 2013a; Jacobson et al., 2015; Bainbridge et al., 2015). A third group which was involved in the commercialization of the gene therapy product (Russell et al., 2017) used non-localized (global) visual function and functional vision measures (Deverell et al., 2017) to support the conjecture that efficacy was durable (Maguire et al., 2019). Recent studies in the dog model have shown that longevity of the efficacy depends on the number of photoreceptors present at the time of treatment. Retinas with ONL that is >63% of normal at the time of treatment show no progression of degeneration, while areas with less photoreceptors at the time of treatment show progressive degeneration (Gardiner et al., 2019). Most patients at the time of treatment have ONL thicknesses that are less than 40% of normal (Jacobson et al., 2005; Jacobson et al., 2008a; Jacobson et al., 2012).

Despite the complexities revealed by these early trials, RPE65-LCA has become a business model for pharmaceutical and biotechnology companies who would launch programs into rare-orphan IRDs and seek regulatory approval of future gene therapy products.

3.2. MERTK-RP:

A second severe IRD treated with subretinal AAV gene therapy was caused by mutations in the MERTK (MER Proto-Oncogene, Tyrosine Kinase) gene (Gal et al., 2000; D’Cruz et al., 2000). Like RPE65-LCA, this was a primary RPE disease; it was a recessive human IRD modeled by the well-investigated RCS (Royal College of Surgeons; rdy) rat with an RPE phagocytosis defect (Dowling and Sidman, 1962; LaVail, 2001). A genetically-engineered mouse model was also studied (Duncan et al., 2003). Various modes of therapy have been applied to the rodent models (LaVail, 2001), and gene augmentation was entertained as a possible treatment (Ali, 2004; LaVail et al., 2016). Proof-of-concept was performed in rodents with different viral vectors and suggested rescue of degeneration and halting of the phagocytosis defect leading to preserved ERG responses rather than loss of recordable ERGs (Vollrath et al., 2001; Smith et al., 2003; Deng et al., 2012; LaVail et al., 2016). What was the preparation for this clinical trial? Safety studies were performed in SD rats (Conlon et al., 2013); non-human primate studies were published along with the clinical trial results but very limited information was provided (Ghazi et al., 2016). Efficacy outcomes used in the trial mirrored those in the RPE65-LCA trials (Ghazi et al., 2016). Unlike in patients with RPE65-LCA, however, there was no evidence in the literature (Charbel Issa et al., 2009; Audo et al., 2018) or in baseline examinations of the six clinical trial patients indicating dissociation of function and structure (i.e. potential for improvement of vision). Considering the proof-of-concept rodent data, human natural history data, and a trial timeline for halting progression would have been of value instead of the expectation of short-term improvement. The two-year long trial (Table 2) showed no change (in either eye) of parameters used other than visual acuity: one patient was reported to have improved acuity which then regressed (attributed to cataract formation); another patient improved at the end of the trial; and a third patient reported improvement but this was relatively minimal and later lost, also complicated by cataract. These data suggest that greater attention to the candidacy of patients and the human disease for gene augmentation, along with modified outcome measures, may be needed to perform future trials in MERTK-RP. Even in such a well-researched mechanism as in this rodent model, details of the disease expression in humans cannot be simply assumed.

3.3. MYO7A-USH1B:

Proof-of-concept studies for Usher syndrome genotypes in general have been hampered by the lack of retinal degenerative disease expression in murine models; USH1B due to mutations in MYO7A was no exception. There has also been debate about the cellular site of disease in USH1B with the shaker1 mouse model showing defective melanosome distribution in the RPE (Liu et al., 1998). Is it the photoreceptors and/or RPE (Williams and Lopes, 2011)? The MYO7A gene is too large for the AAV cargo capacity, so it was delivered with an equine infectious anemia virus (EIAV)-based lentiviral vector. A light-induced degenerative phenotype was reported in the shaker1 mouse (Peng et al., 2011) and subsequent studies of safety and efficacy with the lentivirus were performed in this USH1B model. Independent of the proof-of-concept complexities, movements were made towards a gene augmentation clinical trial (Table 2). Safety studies were performed in the shaker1 murine model and in non-human primates. The macaque experiments produced uveitis to different degrees in the vector-administered eyes and there were retinal abnormalities attributed to the subretinal injection procedure (Zallocchi et al, 2014). Efficacy outcomes have been explored in detailed studies of patients with USH1B (Jacobson et al., 2009b; Jacobson et al., 2011; Sumaroka et al., 2016). The transition zones from preserved and functional retina to severe retinal degeneration with visual loss were defined by OCT and by chromatic rod and cone perimetry. These regions were suggested to be sites for subretinal injection of vector-gene (Jacobson et al., 2009b). A disease sequence was also proposed from a study of cross-sectional data (Jacobson et al., 2011); different rates of progression within the retina were noted and recommendations were made about how to monitor patients in natural history studies and treatment trials (Sumaroka et al., 2016). Evidence was previously presented that USH1B showed no dissociation of structure and function (Jacobson et al., 2008b), so a natural history of disease was indicated to determine the timeline for a trial testing an intervention. The proof-of-concept work also supported that the goal of a trial would be to slow the rate of degeneration. A trial began many years ago using subretinal injection of (EIAV)-based lentiviral vector and, to our knowledge, there has been no formal report of the results of this clinical trial of gene augmentation therapy in USH1B to date.

3.4. ABCA4-STGD:

The other disease group that had gene augmentation therapy with a lentiviral-based gene transfer vector is ABCA4-associated IRD (Kong et al., 2008; Binley et al., 2013). The molecular mechanisms and human phenotypes of this group of diseases have been well explored and therapeutic concepts other than gene augmentation have been proposed and initiated (Allikmets et al., 1997; Sparrow et al., 1999; Fishman et al., 2003; Sears et al., 2017). Prevalence of the disease group is relatively high, compared to some of the other autosomal recessive IRDs in early clinical trials (Table 2). Lipofuscin accumulation in mouse models of ABCA4-STGD has been identified but the retinal degeneration that characterizes the human disease has not been a main finding (Weng et al., 1999; Zhang et al., 2015). Safety studies in rabbits and non-human primates using subretinal EIAV-gene in macular-equivalent regions showed some uveitis compared to controls and other retinal abnormalities (Binley et al., 2013). Efficacy outcomes have been explored in detailed studies attempting to quantify the progression and estimate natural history of ABCA4-STGD with various parameters (Cideciyan et al., 2004, 2007a, 2009, 2012, 2015; Huang et al., 2014; Ervin et al., 2019). All evidence to date suggests that the goal of a gene therapy approach would be to halt progression of the disease and prevent future retinal degeneration; improvement in vision would not be expected. As such, a very clear notion of the progression rate in candidate patients and the stages of disease to include in various cohorts would be required. We are unaware of a formal report of results of the clinical trial of EIAV-gene augmentation therapy in ABCA4-STGD.

3.5. CNGA3- and CNGB3-ACHM:

The autosomal recessive IRD group achromatopsia (ACHM) has severe and retina-wide abnormality of cone photoreceptor-based visual function from birth and is known to be caused by several different genes (CNGA3, CNGB3, GNAT2, PDE6C, PDE6H and ATF6; Hassall et al., 2017; Michalakis et al., 2017; Hirji et al., 2018). CNGA3 and CNGB3 genes encode the two subunits of the cone cyclic nucleotide-gated (CNG) channel and represent the most common genetic causes of ACHM (Wissinger et al., 2001; Kohl et al., 2005). There are disturbances of visual acuity and color vision, photosensitivity and nystagmus; a small percentage of patients with autosomal recessive cone dystrophy due to CNGB3 and GNAT2 mutations have also been reported (Michaelides et al., 2003; Thiadens et al., 2010). ACHM is clinically and molecularly distinct from the X-linked IRD known as blue cone monochromacy (BCM), which is caused by mutations in the array of long- and middle-wavelength cone opsin genes. There are small and large animal models of both CNGA3 and CNGB3 mutations and proof-of-concept research has been performed for subretinal gene therapy using AAV (Carvalho et al., 2011; Komáromy et al., 2010; Banin et al., 2015; Gootwine et al., 2017). In these models, function is restored once the normal subunit is expressed in mutant cells; of note, age-dependent effects of the therapy have been documented in the CNGB3 canine model even when the number of cones are not significantly reduced (Komáromy et al., 2010, 2013) and the age effect can be abrogated with a combinatorial therapy approach (see below). Safety studies relevant to CNGB3 were performed in mice, dogs and non-human primates, and at certain dose levels, there was detectable uveitis (Ye et al., 2016a, 2016b, 2017). For CNGA3, non-human primates were studied and minimal inflammation was present at one month but not at 3 months post injection; other abnormalities were ascribed to the surgical procedure (Tobias et al., 2019). Efficacy has been considered despite the appropriate focus on safety in early trials planned or ongoing (Yang et al., 2014; Zelinger et al., 2015; Zobor et al., 2015, 2017; Langlo et al., 2016; Kahle et al. 2018). Although not specifically addressed or quantified, there are OCT scans, cone photoreceptor imaging and acuities in the literature indicating dissociation of structure and function at the fovea in patients without atrophic central retinal lesions. Qualitatively, there is potential for improvement of vision. Natural history studies of either genotype, to our knowledge, have not been reported but there are indicators (from data in the literature) that foveal ONL losses can be present in some patients while others only show apparent cone outer segment abnormalities. For one of the CNGA3 trials, preliminary safety results were provided (Reichel et al., 2017) with a submacular injection (Kahle et al., 2018). There have been no formal reports to date of the results of the ongoing CNGA3 and CNGB3 clinical trials of subretinal gene augmentation therapy.

While it is premature to consider complementary approaches to ACHM gene therapies, it is important to remember that, at least in the CNGB3 canine model, recovery of function in treatment of aged mutant animals required that cone photoreceptors be transiently deconstructed with ciliary neurotrophic factor (CNTF), i.e. to lose whatever outer segments were present and regrow new ones in the presence of both subunits of the CNG channel (Komáromy et al., 2013).

3.6. PDE6B-RP:

PDE6B was one of the first genes identified to cause forms of AR RP in mice (Bowes et al., 1990; Pittler and Baehr, 1991; Chang et al., 2007), humans (McLaughlin et al., 1993) and dogs (Aguirre et al., 1978; Farber et al., 1992; Suber et al., 1993). Rarely PDE6B mutations are associated with AD RP (Gal et al., 1994; Gao et al., 1996; Manes et al., 2014). PDE6B encodes the beta-subunit of rod cGMP-phosphodiesterase which is a key enzyme specific for phototransduction activation in rod photoreceptors (Lamb and Pugh, 2004). Consistent with the gene function, patients show early onset of night vision loss (Jacobson et al., 2007; Tsang et al., 2008; Tatour et al., 2019). Animal models show an extremely fast degeneration of the rods. Proof-of-concept studies with subretinal AAV in a murine model were mostly unsuccessful (Bennett et al., 1996) unless a slower degenerating mouse line was used and treatment initiated relatively early (Pang et al., 2008). Similarly, successful subretinal gene therapy in dogs was achievable only when injected at a very early time during post-natal development of the retina (Petit et al., 2012; Pichard et al., 2016). We are not aware of published preclinical safety studies but a human clinical trial is ongoing (Table 2). For a primary rod disease treated with gene augmentation of a rod-specific gene, clinical efficacy outcomes would have to include evaluation of rod function and rod structure but there has been no evidence of either in patients published to date (Danciger et al., 1996; Jacobson et al., 2007; Tsang et al., 2008; Ali et al., 2011; Kim et al., 2012; Tatour et al., 2019).

3.7. RLBP1-RP:

A spectrum of AR IRDs are caused by mutations in RLBP1 (Maw et al., 1997; Burstedt et al., 1999; Morimura et al., 1999; Eichers et al., 2002). Prevalence of RLBP1-RP is very low in diverse populations (Table 2); however, prevalence can be substantially higher in some remote geographical regions (Burstedt et al., 1999; Eichers et al., 2002). In addition to a range of retinal degenerative features in common with many other IRDs, RLBP1-RP patients demonstrate extreme slowing of the rate of dark-adaptation for both rod and cone photoreceptors (Burstedt et al., 2003, 2013). It has long been known that RLBP1 encodes cellular retinal-binding protein (CRALBP) which is an 11-cis-retinal blinding protein expressed in the Muller and RPE cells (Bunt-Milam and Saari, 1983); it is a critical visual cycle component. But the rod and cone dark-adaptation kinetic defects in patients have recently become more interpretable by the finding of direct CRALBP involvement in the retinal Muller cell visual cycle supporting the cones (Xue et al., 2015) as well as the canonical RPE visual cycle supporting the rods (Saari et al., 2001). Preclinical proof-of-concept studies with subretinal AAV targeted both the RPE and Muller cells in mice and showed improvements in the rate of dark adaptation in rods and cones (Choi et al., 2015). Preclinical safety studies performed in mice, rats and non-human primates showed intraocular inflammation and dose-dependent retinal thinning (MacLachlan et al., 2018).

Clinical efficacy outcomes and natural history have not been published but are reported to include measures of retinal structure and visual function including full-field dark-adaptation kinetics (Green et al. 2017; Ni et al., 2017). There have been no published reports from the ongoing clinical trial aiming to improve the rod and cone visual cycle defects and stabilize the retinal degeneration (Table 2).

3.8. REP1-CHM:

Choroideremia (CHM) has been recognized clinically for more than a century due to a distinctive funduscopic appearance in hemi- and heterozygotes (reviewed in Mitsios et al., 2018). The causative gene and understanding of pathophysiology are more recent discoveries (Cremers et al., 1990; Seabra et al., 1993; Tolmachova et al., 2006). The finding of independent site(s) of disease action (photoreceptors and/or RPE) in both the human patients (Syed et al., 2001; Jacobson et al., 2006c; Syed at al., 2013; Morgan et al., 2014) and in a murine model of this X-linked IRD adds complexity to the proof-of-concept data and how it relates to the human disease (Tolmachova et al., 2006, 2010, 2013). Safety studies were performed in wild-type mice and in vitro (Vasireddy et al., 2013). Efficacy outcomes have concentrated on central retinal parameters, considering the surgical convenience of a macular subretinal injection and the common mid- and late-disease stage showing relatively preserved central retinal structure and function but little or no peripheral vision (Heon et al., 2016). Psychophysical results have included best-corrected visual acuities and perimetry under fundus visualization (‘microperimetry’); objective results are OCT and AF imaging (MacLaren et al., 2014; Aylward et al., 2018; Dimopoulos et al., 2018; Xue et al., 2018a, 2018b; Lam et al., 2019). The natural history of CHM, whether from cross-sectional data or performed prospectively, agree that CHM is a severe IRD but there is retained (near normal) central vision into late decades of life (Heon et al., 2016; Aleman et al., 2017; Di Iorio et al., 2019). This slow progression has been cited as advantageous because it provides a wide window for administering the treatment (Sengillo et al., 2016; Dimopoulos et al., 2018: Pennesi et al., 2019; Mitsios et al., 2018), yet it complicates outcome assessments in the absence of robust natural disease history studies. The proof-of-concept work with subretinal AAV-CHM in the CHM murine model showed increased function by dark-adapted ERGs but not light-adapted ERGs when treated eyes were compared with the contralateral sham-treated eyes; there was also reduced or no change in responses at lower doses of this vector but the higher dose led to increases in ERG amplitude. Further studies are likely needed to reconcile some of the results in the treated murine model and what we know of the human disease (Tolmachova et al., 2013). Despite the many studies of CHM hemizygotes, there has not been a test of the hypothesis of potential dissociation of structure and function, as has occurred with some IRDs (Jacobson et al., 2005, 2008b, 2014; Cideciyan and Jacobson, 2019).

REP1-CHM patients have been treated in several clinical trials of subretinal gene therapy. There have been reports of improvement in visual acuity and improvements of rod and cone function (MacLaren et al., 2014; Edwards et al., 2016; Fischer et al., 2018), no significant changes between treated and untreated eyes (Dimopoulos et al., 2018; Fischer et al., 2019), and retained visual acuity in later stage disease when uninjected contralateral eyes show progression (Xue et al., 2018b). CHM has been stated to be an attractive target for gene therapy because it represents an unmet need and will reduce societal burden through decreased health care costs (Pennesi et al., 2019). Certainly, the X-linked inheritance and the ease of ophthalmoscopic diagnosis make this rare disease detectable in the clinic; mutations in the CHM gene, encoding REP1 (rab escort protein1), have been identified and the gene can be delivered to RPE and photoreceptors with AAV2 (Pennesi et al., 2019). The months-long recovery of these photoreceptor-rich foveas after the macular detachment from a central subretinal injection has been used as supportive evidence that such surgical procedures are now of limited concern and should open the door to further subretinal macular injections even in patients with normal acuities (MacLaren et al., 2014; Xue et al., 2018a, 2018b). But foveal thinning and other outer retinal alterations remain a vulnerability of the surgical procedure (Pennesi et al., 2019; Aleman et al., 2019). Efficacy outcome results from the available trial reports seem mixed with uncertainty as to what the therapy is intended to do; from a notion of improving vision, the goals of the therapy seem to have changed to stopping disease progression.

3.9. RPGR-RP:

X-linked RP is a severe and relatively common IRD (Bird, 1975; Jay, 1982; Fishman et al., 1988). Pedigree analyses and examination of heterozygotes led to the identification of many XLRP families across the world that enabled linkage and then identification of the gene (Wright et al., 1987; Meindl et al., 1996; Vervoort et al., 2000; Breuer et al., 2002). Clinical gene augmentation therapy trials have concentrated on the common XLRP subtype: mutations in the RPGR gene and especially ORF15 mutations which account for >70% of RPGR disease. There has been availability of both small and large animal disease models for RPGR-XLRP; comparisons of the various mutant mice have been made (Huang et al., 2012) and AAV gene augmentation has been performed in mutant mice (for example, Wu et al., 2015; Fischer et al., 2017). Key proof–of-concept data came from the canine RPGR-ORF15 disease model using subretinal AAV gene therapy; the results indicated rescue of photoreceptors and reversal of secondary effects, specifically retinal remodeling secondary to photoreceptor loss (Beltran et al., 2012). The experiments showed that the arrest of disease progression was feasible not only in early stages but also in later stages of retinal degeneration in the model (Beltran et al., 2015). Safety studies were performed in wild-type and mutant mice (Deng et al., 2015; Fischer et al., 2017; Song et al., 2018) and issues about instability of the ORF15 were addressed. Large animal safety studies have not been published to date. Efficacy, as with many forms of RP considered to be severe, has been characterized as having sufficiently rapid progression that it will be able to be detected on clinical trial outcomes (Martinez-Fernandez De La Camara et al., 2018). Indeed, there are sufficient objective and subjective parameters that could be used as secondary outcomes.

A highly cited outcome for the central retina of RPGR-XLRP patients is the OCT (EZ line width) and the corresponding cone-mediated visual sensitivity change that brackets the EZ line change from measurable to unmeasurable; this rate of change has been shown to be within a two-year time frame that would be acceptable to clinical trialists (Birch et al., 2013, 2015). This remarkable observation to permit monitoring change is confined to the central retina and, if trials are being designed for subretinal injection, then this injection must be macular. Also to be considered is the common macular degeneration in RPGR-XLRP patients that would make submacular injections depending on EZ width measures either not possible or not desirable (Charng et al., 2016). In addition, quantitative results at the fovea have shown a lack evidence for cone vision improvement potential (Jacobson et al., 2014).

Extracentral retinal regions with rod- and cone-mediated vision measured with chromatic psychophysical methods, however, should not be neglected and untreated (Charng et al., 2016); in other words, the surgical convenience of macular subretinal injections should not drive the clinical trial. For clinical trials aiming to halt progression of disease, a realistic duration of 4.5 years for rod vision and 6.1 years for cone vision have been estimated, and ways to try to reduce these to shorter durations have been suggested (Cideciyan et al., 2018). Natural history studies have mainly been motivated by determining the least time it would take to determine change in a clinical trial of RPGR-XLRP patients (Birch et al., 2013, 2015; Charng et al., 2016; Cideciyan et al., 2018). These may appear as widely disparate results but actually what is being reported is that early trials would have to reckon with the facts of variation within this population and if there was a surgical commitment to macular subretinal injections as a trial and central retinal parameters (mainly cone-mediated) to monitor, then excluding more than half of the patients because of macular degenerative disease and unmeasurable EZ width would need to be part of the entry criteria. A larger population, and not necessarily older cohort, would include patients with the common maculopathies but a clinical trial strategy to monitor extracentral rod and cone vision and accept a lengthier timeline to decision about efficacy (Charng et al., 2016; Cideciyan et al., 2018). The three ongoing clinical trials of RPGR-XLRP seem to have chosen the route of submacular injection of AAV-gene delivery, but no results have been reported to date (Table 2).

3.10. Summary:

The first ten subretinal gene augmentation therapy trials were performed in otherwise incurable IRDs associated with a range of visual abnormalities; all IRDs represent unmet needs and this is not a justification for treating one before another (Pennesi et al., 2019). Prevalence varied between ~1:10,000 and ~1:2,000,000 (Table 2) and was also not the main reason for performing these specific trials. What then was the reason for beginning the era of retinal gene therapy with these specific diseases? Convincing reasons for considering a human clinical trial should have been evidence in pre-clinical proof-of-concept and safety studies that the product could have a positive effect with minimum risk to the patient. Some of the ten disorders had more of such evidence, and some had less. In addition, pre-clinical proof-of-concept results would have to be matched with the potential for short term improvement of vision and/or arrest of progression that is measurable within the study duration in the patients enrolled. Was there indication that the patients (before the trials) had potential for improvement in vision? Only RPE65-LCA had definite pre-clinical evidence of improvement (Acland et al., 2001, 2005) and definite evidence obtained prior to trial onset of potential for improvement (Jacobson et al., 2005). Maybe not surprisingly, substantial improvement in vision did occur in the RPE65 trials. The only other patients studied quantitatively for potential for visual improvement among the 10 conditions were those with MYO7A-USH1B (Jacobson et al., 2008b) and RPGR-XLRP (Jacobson et al., 2014). Dissociation of structure and function was not present in the extra-foveal rods of MYO7A-USH1B and foveal cones of RPGR-XLRP. In the remaining diseases, there were no pre-trial quantitative studies in affected patients to test for structure-function dissociation that would signal the potential for improvement in vision with therapeutic intervention. In CNGA3- and CNGB3-ACHM, human phenotype studies were suggestive of retained but dysfunctional cones; however, quantitative predictions of the extent of cone function improvement remain unknown.

Was there indication that arrest of progressive retinal degeneration can be measured within the study duration? In the case of RPE65-LCA, natural history was measurable and there was evidence that the natural history was not modified despite large vision improvements (Cideciyan et al., 2013a; Jacobson et al., 2015). Data are less direct or not available in other IRDs. RPGR-XLRP has natural history data which present different time courses of degeneration depending on whether the central retinal structure or extracentral retinal function is considered. The former suggests a 2-year study duration (Birch et al., 2013) while the latter indicates 4.5–6 years of monitoring (Cideciyan at al., 2018) needed to determine slowing or halting of progression. To date, there are no reported results from the RPGR-XLRP trials. CHM was not examined in a prospective natural history study before subretinal clinical trials began, but there have been cross-sectional studies (prospective and retrospective). Most of these CHM studies emphasized the persistence of central retinal structure and function until late decades in life. Natural history of central and peripheral ABCA4-STGD disease are described (Cideciyan et al. 2009; Cideciyan et al., 2015; Strauss et al., 2019) but there have been no gene therapy results described to date.

The decision to begin most of the trials likely also had to do with the feasibility of AAV vector delivery, and the fact that the RPE65-LCA trial safely used AAV vectors for subretinal injections. Diseases with genes that could be within the carrying capacity of AAV were targeted; dual AAV vectors platforms were evolving but not ready for the clinic (Dyka et al., 2014, 2019). Some trials used EIAV based lentiviral vectors for MYO7A and ABCA4 which were too large for AAV. There was a competitive rush to initiate trials after the RPE65-LCA trial and this led to an assumption that uveitis was not a serious adverse event; systemic steroids before and after the subretinal procedure have almost become an ‘industry standard’ to dampen any potential uveitic response that would cause adverse events. This conventional wisdom needs to be reconsidered and any uveitis investigated for cause.

4. Moving towards intravitreal injections

The next series of trials should target IRDs with a goal of improving remaining central vision with the least photoreceptor trauma and the most predictable outcome, all the while seeking to improve the quality of life of the patients. There is no doubt that the RPE65-LCA trials were inspirational - not only to patients (with or without RPE65-LCA) but also to sponsors that were not previously interested enough in orphan genetic retinal diseases to invest in clinical trials of treatment. Some sponsors have assumed that RPE65-LCA provides a ‘one-size fits all’ template waiting to be followed from concept to commercialization, and the underlying principle in designing further trials has seemed to be ‘don’t change the formula that worked before’. Publications about investor expectations during the trial to predictions of cost-effectiveness after the clinical trial have stoked further interest (for example, Schimmer and Breazzano, 2015; Johnson et al., 2019). Now that there are 9 further subretinal gene augmentation trials without evidence of multi-log-unit improvements, it has become clear that choosing a disabling IRD and following an RPE65-LCA-type formula may not be sufficient to achieve success.

Although there are many apologists for the disruptive effects of surgically-induced macular detachments, this common method to deliver vector-gene to photoreceptors and RPE begs to become a relic of this early period of human gene therapy trials. Presently, the subretinal surgical technique has almost become a subspecialty unto itself. There are innovations and improvements in the method to reduce adverse events and this is to be applauded. The influx of ideas is due to the fact that the current list of gene augmentation trials has extended to include multi-center protocols and thereby multiple talented retinal surgeons (Davis, 2018; Davis et al., 2019). It has been stated that there are no alternative methods to deliver vector-gene to the outer retina (Davis et al., 2019); without further research and a detailed understanding of the diseases being treated, that will remain true. Intravitreal delivery has been stated to be only useful for inner retinal or optic nerve diseases (Davis et al., 2019). However, it is recognized that intravitreal delivery could reduce the retinal trauma from detachment that can complicate the goals of the therapy (Khabou et al., 2018; Miller and Vanderberghe, 2018). Intravitreal injections have the potential to transfect large retinal areas but experimental evidence in non-human primates indicates that penetration of current intravitreal vectors to photoreceptors would be mainly or only in the fovea (Dalkara et al., 2013; Boye et al., 2016; Khabou et al., 2018). There are problems that need to be addressed and then solved. For example, an inflammatory uveitic response to the current vectors is worth the research effort to refine the details of this mode of delivery; “more potent and better tolerated vector systems” are needed (Miller and Vandenberghe, 2018). Of course, there are also non-gene-therapy and/or non-gene-specific approaches involving systemic or ocular drugs, but these will not be considered here.

We propose a group of 6 IRDs that should be candidates for intravitreal gene augmentation therapy, once it is improved (Table 3; Figure 4). In order to evaluate which IRDs may be most appropriate candidates for intravitreal gene therapy, we considered criteria based on current knowledge of the diseases, feasibility of clinics to acquire outcome measurements and the lessons learned from gene therapy trials to date. The goal would be to improve foveal vision in general, and foveal visual acuity in specific.

Table 3.

Diseases warranting consideration of intravitreal gene augmentation therapy

Mendelian genetic type	Disease	Gene	Estimated gene mutation frequency in the USa	Proof of conceptb	Dissociation of structure-functionc	Comments
Autosomal recessive	ACHM2	CNGA3	1/576,667	o, m	+/− for cones	Qualitative evidence from literature of dissociation of structure-function
	ACHM3	CNGB3	1/346,000	c, m	+/− for cones	Qualitative evidence from literature of dissociation of structure-function
	EORP/LCA	TULP1	1/1,730,000	m model available	+ for cones	Small central island only with remaining ONL but insensitive
	Senior-Loken/LCA	NPHP5	1/346,000	c, m	+ for cones	Central island only and concern over anesthesia because of renal abnormalities and multiple previous anesthesias for renal transplants Dissociation of structure and function provend
	LCA6	RPGRIP1	1/865,000	m	+/− for cones	Qualitative evidence from literature of dissociation of structure and function
X-linked	BCM	OPN1LW, OPN1MW	1/576,667	m	+/− for cones	Qualitative evidence from literature of dissociation of structure and function

^aEstimated frequency from Stone EM, et al. Ophthalmology. 2017;124(9)-1314-1331.

^bAnimal studies: c, canine; m, murine; o, ovine

^cGradation:

+/− Qualitative dissociation: Evidence for existence of cell of interest lacking normal function implying improvement possible

+ Quantitative dissociation: Colocalized measure of retinal photoreceptor structure and light sensitivity defining vision improvement potential

^dSumaroka A, et al. Invest Ophthalmol Vis Sci. 2019;60(7):2551–2562; Downs LM, et al. Hum Mol Genet. 2016;25(19)4211–4226.

Fig 4.

Foveal ONL thickness and central visual function, as measured with best-corrected visual acuity (VA, represented as decimal). (A) OCT scans of a normal subject (upper panel) and patients with IRDs that show normal or close to normal ONL thickness and minimally reduced or even normal VA. (B) Patients with IRD that show foveal ONL thickness comparable to ones in panel A, but severely impaired visual function. Representative OCT scans are along the vertical meridian from inferior retina (I) to superior (S) through the fovea. ONL is highlighted in blue. Foveal ONL thickness is depicted by white double arrow in normal scan. Dashed line in ONL thickness bar graph is lower limit of normal (n= 11; ages 26–62). Dashed line in VA bar graph is normal VA (20/20 or 1.0 in decimal representation).

Given the understanding that intravitreal vector-gene delivery would be limited to foveal cones (at least initially), how can we predict which IRDs may be better targets for treatment? Are there any measurements that could be performed in the eye clinic with widely available clinical methodology that would permit such prediction? The two such parameters we identified are foveal ONL thickness and best corrected visual acuity (BCVA). Spectral domain OCTs are commonly performed in the eye clinic and a single horizontal (and possibly also a vertical scan) through the fovea should allow for a measurement of foveal ONL. For example, OCT instruments, such as the Spectralis (Heidelberg Engineering, Heidelberg, Germany), the Cirrus 6000 (Carl Zeiss AG, Oberkochen, Germany), the Avanti RTVue XR or RTVue-100 (both manufactured by Optovue, Inc., Fremont, California, USA) include manufacturer software with built-in tools to enable measurement of the foveal ONL manually, and in some devices, automatically. Given this measure of photoreceptor nuclear structure at the fovea, the other relevant measure is routine – BCVA. Foveal ONL thickness and BCVA results (represented as decimal) in normal subjects and in patients with IRDs are shown (Figure 4). IRDs which can present with normal ONL thickness and minimally reduced or even normal visual acuity are on the left (Figure 4A). The IRDs vary in severity of disease outside the fovea but share the feature of preserved foveal ONL and central visual function (as measured by acuity) at the stages depicted. The 6 retinopathies with potential for improvement are shown on the right (Figure 4B); these are not uniform in genotype or phenotype but all share near-normal foveal ONL thickness (at least in early stages of disease) but with relatively reduced visual acuity. These diseases could be considered for intravitreal gene therapy.

Retina-wide cone diseases with decreased visual acuity, such as ACHM and BCM, would be worthy of attention. The current subretinal approach to ACHM (macular injections) risks these fragile foveas. Sufficient photoreceptor structure is evident in published (and quantified) images of foveal structure in CNGA3 and CNGB3 forms of ACHM (Zelinger et al. 2015; Langlo et al., 2016, 2017). The time course of the maculopathy needs to be considered and disease stage as well as genotype need to be taken into account (Kahn et al., 2007). Given the relatively long time course of disease (Langlo et al., 2017), patients may favor waiting for development of lower-risk means to deliver a gene product in a clinical trial. There is also evidence for retained photoreceptor lamination at the fovea in BCM which suggests that this condition is also a candidate for intravitreal gene therapy (Cideciyan et al., 2013b; Sumaroka et al., 2018).

In the spectrum of IRDs that could benefit from intravitreal delivery of vector is the severe form of early-onset RP caused by mutations in the TULP1 gene (Jacobson et al., 2014). This rapidly progressive IRD can lead to an advanced level of visual loss with a residual central island even in early decades of life. Surrounding the small central island there could be little or no detectable photoreceptor lamination by OCT but foveal ONL structure can be intact. Measurements of TULP1 structure and function and comparison with data from other ciliopathies causing IRDs indicated insensitive function for the amount of photoreceptor structure remaining at theTULP1 foveas (Jacobson et al., 2014). Given the simplified clinical ‘screening’ method using foveal ONL and visual acuity described above, other early-onset IRDs at late disease stages could be candidates for treatment. Whereas a small central island of impaired vision would seem to be an unlikely target for a subretinal surgical procedure, proof of dissociation of structure and function and the availability of a less invasive intravitreal vector delivery may be worth considering.

Also deserving consideration for gene augmentation therapy are those forms of LCA that have not been approached in clinical trials to date. Two forms of LCA share the phenotype that leads to severe loss of peripheral retinal structure and function early in life but with retained central photoreceptors that are more dysfunctional than expected from OCT measures of structure. This pattern is common to patients with mutations in the RPGRIP1 and NPHP5 (IQCB1) genes. Of interest, this is also the phenotype of CEP290-LCA (Cideciyan et al., 2007b; Cideciyan et al., 2011; Cideciyan and Jacobson, 2019) which has shown evidence of visual improvement in an antisense oligonucleotide clinical trial (Cideciyan et al., 2019); the large CEP290 gene would not be a candidate for an AAV vector approach. RPGRIP1 has been studied in depth and is a key protein in the photoreceptor connecting cilium (Li, 2014; Raghupathy et al., 2017). There are murine models, and proof-of-concept research with subretinal AAV has been performed (Pawlyk et al., 2010). The form of LCA caused by NPHP5 mutations (one of the causes of Senior-Loken syndrome, SLSN) is associated with renal disease (Otto et al., 2005; Stone et al., 2011; Estrada-Cuzcano et al., 2011; Cideciyan et al., 2011). Dissociation between structure and function in NPHP5-LCA means there is opportunity for visual improvement with a gene therapy treatment. There are large and small animal models (Downs et al., 2016; Oh et al., 2017; Hanke-Gogokhia et al., 2018); and a cone-only Nphp5^−/−Nrl^−/− murine model has been treated with subretinal gene augmentation using an AAV vector (Hanke-Gogokhia et al., 2018). Preliminary studies in the dog NPHP5 model in which there is a dissociation of structure and function have shown recovery of cone function and improvement of rod function in response to subretinal gene augmentation (Hardcastle et al., 2018).

Although studies relating structure and function are not, to our knowledge, specifically reported in later stages of CHM with reduced visual acuity, it would also seem worthy to consider an intravitreal alternative, once available, to subfoveal injections. Improved acuity has been reported in such patients treated with exceptional surgical care not to cause excessive foveal stretching or macular hole formation; if the same result were possible with an intravitreal injection, it may eventually extend the limits of CHM gene therapy beyond institutions with a high level of surgical prowess (Lam et al., 2019). Those stages with excellent foveal ONL and visual acuity would not seem to be worthy targets for therapy (Figure 4A).

5. Honoring the commitment to resolve the unresolved genotypes

Many recent reviews of IRDs strongly suggest there should be continued pursuit of molecular genetic diagnoses because of the obvious value in understanding cause of disease as well as helping to clarify pathobiology (Duncan et al., 2016 AAO Recommendations; Duncan et al., 2018). Based on what our cohort reveals with regard to the Mendelian type of the genetically unresolved patients (Figure 1), we asked why AR/simplex/multiplex diagnoses are relatively high versus other modes of inheritance. AR disease makes up the majority of IRD cases, and although many genes that cause IRDs have been identified, it is also likely that many more genes that cause IRDs remain to be discovered. In the list of 9 genes that are ‘new or updated’ as per the RetNet database (Retnet, https://sph.uth.edu/retnet/, last update, October 8, 2019), 8 of the genes were associated with a recessive disease. It has been observed that the number of mapped and identified genes have generally increased linearly with time, and since technology aiding in gene discovery has only improved, it follows that new genes are rarer or more difficult to detect (Daiger et al., 2013). These rarer genes may account for a considerable portion of the unresolved AR/simplex/multiplex cases. Although grouped with the AR cases, some of the simplex or multiplex cases may in reality be autosomal dominant (for example, de novo or reduced penetrance) or X-linked (Schwartz et al., 2003; Branham et al., 2012; Glöckle et al., 2014; Haer-Wigman et al., 2017). Mis-categorization of inheritance pattern likely accounts for only approximately 10% of the unsolved simplex cases if our cohort is comparable to those in other reports (Haer-Wigman et al., 2017; Glöckle et al., 2014). Incomplete family history information or small family size may contribute to mis-categorization of inheritance type. The more rare, genetically heterogeneous, or unclear the phenotype, the less-likely the IRD will be molecularly resolved (Haer-Wigman et al., 2017; Bernardis et al., 2016). Also, prevalent genetic testing strategies, such as analyzing a panel of genes, single-gene testing or targeted testing of regions known to harbor mutations, may not be comprehensive enough to yield a definitive molecular diagnosis and can be cost-prohibitive to patients. Mutations in non-coding regions, large structural aberrations or variants that are not detected using current computational filtering methods likely also play a role in contributing to the high proportion of molecularly unresolved AR/simplex/multiplex patients (Cremers et al., 2018; Farrar et al., 2017). Although still imperfect, there is movement towards providing patients with more comprehensive testing options, such as whole exome sequencing, and these are becoming increasingly affordable. Resolution of molecular diagnosis is valuable for patients with IRDs as the diagnosis could affect clinical management and eligibility for future trials of treatment.

6. Future directions and conclusions

From the extensive preparation for the first human retinal gene therapy trials, there have been varying levels of preparation for later novel treatment trials for these incurable IRDs. Initially, there seemed to be reluctance by pharmaceutical and biotechnology companies to be involved in this new direction, possibly due to the tragic death of a participant in an early gene therapy clinical trial (Wilson, 2010; Smith and Byers, 2002; Somia and Verma, 2000). With time and healing, there has been greater interest. Companies are labeling themselves as gene therapy companies, and regular conferences of gene and cell therapeutics companies are now occurring throughout the year. Whereas the initial drivers for such therapies were mainly academic and clinical achievement, the field is transforming into a competitive marketplace for economic achievement (Schimmer and Breazzano, 2015; Johnson et al., 2019; Zimmerman et al., 2019).

In this manuscript, we advocated for a forward movement of therapy but one based on the realities of the initial results of IRD human clinical trials. We have learned that there are many prerequisites for entering a group of patients into a gene augmentation trial. Entry criteria should include at least the following: 1) sufficient proof-of-concept research; 2) evidence of safety preferably based on testing affected models of the disease that are faithful to the human condition being modeled and also acquiring non-human primate data; 3) a thorough understanding of the human disease expression and its natural history and not just an extrapolation from the model data; 4) informative efficacy outcomes based on the human diseases being treated even if these are not the conventional outcomes of previous non-IRD disease trials; and, 5) determining before the trial onset if the goal of the trial is an improvement of vision or a slowing of disease progression. A first step toward predicting the extent of functional improvement from treatment was recently taken using supervised machine learning (Sumaroka et al., 2019). Specifically, cone vision improvement potential in the CEP290 and NPHP5 forms of LCA was determined. Data from patients with RP and residual cone-mediated macular function were used to associate visual sensitivities to local photoreceptor structure from colocalized OCTs (Figure 5). Such predictive strategies regarding potential vision based on recordable cross-sectional OCT images could lead to measurement-based decisions of the maximal potential efficacy in clinical trials and guide decisions as to which retinal diseases (or disease stages) to be treated and whether to proceed with planned increases of dose in cohorts of a trial (Sumaroka et al., 2019).

Fig 5.

Predicted and measured sensitivities for CEP290-LCA patients with severe loss of visual acuity (VA) and fixation. The upper row shows OCT scans along the horizontal meridian across the fovea; VAs are given above the scans (LP=light perception, BLP=bare light perception, NLP=no light perception). Plots under scans are measured FST sensitivities with red and blue full-field stimuli in the dark-adapted state (symbols) and artificial intelligence predictions of localized retinal function based on residual retinal structure (lines). Gray regions depict normal cone-mediated sensitivity in dark-adapted conditions. Modified from Sumaroka et al., 2019, copyright by the Association for Research in Vision and Ophthalmology.

If there is no measurable improvement possible based on a predictive strategy such as supervised machine learning, then the intent of an intervention would likely be to slow the disease course. The category of IRDs without improvement potential must be reckoned with rather than ignored. Trials should not begin on an untested hypothesis of potential visual improvement and then be in the unenviable circumstance of not knowing when to end. If a hoped-for miracle result does not occur within weeks or months, there must be a plan. Indeed, there should be a plan before starting the trial.

Gene augmentation trials will obviously move forward at a greater pace if the targeted diseases can be treated with less risk of adverse events. Subretinal injections were an appropriate way to begin; but moving to another delivery system, such as intravitreal injections, should be sought for many of the IRDs. The inflammatory consequences of both types of delivery systems must also be accepted and then studied and understood as opposed to being masked by anti-inflammatory agents. It is obvious that further research is required and this is the investment currently needed. The next group of IRDs entering trials should not be slowed by potentially solvable problems that are simply not as immediately rewarding or interesting as making a media announcement that a sponsor is first to the finish line with a new drug.