The ciliopathies are a collection of disorders with overlapping clinical manifestations that include renal cystic disease, polydactyly, retinal degeneration and defects of the central nervous system1
, and provide a useful model to investigate both the effects of variation at a single locus and the potential epistatic interactions between alleles at functionally related loci. This is because several ciliopathy-causing genes can either cause a distinct recessive form of ciliary disease or contribute modulators of penetrance and expressivity. In some instances, multiple allelism at a single locus can partially explain phenotypic variability. For example, null mutations in NPHP3
, while hypomorphic NPHP3
alleles cause Nephronophthisis (NPHP)3
. Similarly, hypomorphic CEP290/NPHP6
mutations are observed frequently in Leber congenital amaurosis (LCA) in both humans and in a mouse model4,5
, whereas loss-of-function mutations lead to a constellation of ciliopathies that include MKS, NPHP, Bardet-Biedl Syndrome (BBS) and JBTS6-9
without a clear explanation for the phenotypic variation.
Consistent with the suggestion of second-site modifiers, several alleles are reported to modify ciliary phenotypes; a hypomorphic mutation in MGC1203
contributes to the overall penetrance and expressivity of BBS10
, while heterozygous variants in CEP290
contribute to the phenotypic severity and pleiotropy of NPHP11
. Finally, likely additive effects between alleles in bona fide
BBS and MKS loci produce hybrid phenotypes of the two clinical entities7
A paradigm emerging from the above studies is that genes mutated in one ciliopathy may contribute alleles across the entire clinical spectrum and that the assessment of total mutational load across the ciliary proteome12,13
will help dissect phenotypic causality and variability. Towards this end, we have initiated two complementary strategies – interactome studies on ciliary proteins and medical resequencing of known ciliopathy genes across a cohort of patients with diverse phenotypes (MKS and JBTS as severe, BBS as intermediate, NPHP and LCA as mild).
We first investigated RPGRIP1L
, under the hypothesis that in addition to its association with MKS and JBTS14,15
, it might also contribute alleles to other ciliary disorders. We screened 166 unrelated patients of northern European descent (17 MKS, 15 NPHP (with and without extra-renal phenotypes), 34 BBS, 100 LCA) and discovered 10 different novel nonsynonymous changes in 24 individuals (). We also detected two different common variants G1025S (rs2111119) and D1264N (rs3213758), both of which are present at similar allele frequencies when compared across our ciliopathy cohort, a cohort of 96 control individuals of northern European descent, and 60 CEPH controls from HapMap (data not shown).
Total mutational load in northern European ciliopathy patients with RPGRIP1L variants
None of the observed alleles are sufficient to explain disease manifestation under a Mendelian model. However, most alleles were either unique to our patient population, or were enriched compared to ethnically matched controls, and mapped within known functional domains of RPGRIP1L (; Suppl. Fig. 1
However, the rarity of most alleles precluded us from delineating their impact on the clinical phenotypes. One notable exception, A229T, is sufficiently frequent to empower such a study. To assess allele neutrality without a preconceived model of inheritance, we performed Transmission Disequilibrium Testing (TDT); we screened 145 BBS families for which DNA was available from both parents. Focusing on families in which only one of the two parents was an A229T heterozygote, we identified 18 informative trios and detected significant over-transmission of the threonine-encoding allele in patients (15/18 transmissions; 83.3%, p<0.01).
Despite the modest number of trios, the TDT suggested that the A229T change is not neutral. To probe this possibility further, we re-evaluated our original resequencing data. We found a surprising enrichment of the 229T allele in northern European patients in whom retinal degeneration is mandatory for diagnosis, such as in BBS or Senior-Loken Syndrome (SLS) (5.9%; 4/68 chromosomes, and 15%; 3/20 chromosomes respectively; , compared to a 2.8% allele frequency in unaffected northern Europeans; ). We therefore wondered whether this variant might contribute to the retinal defect, which is a frequent, but not ubiquitous, ciliopathy phenotype. If this is true, then comparisons between NPHP (no retinal degeneration) versus SLS cohorts (NPHP and retinal degeneration) should be informative. Even though each of these phenotypes is rare in the general population, we were able to study two northern European cohorts: 66 unrelated patients with isolated NPHP and 84 patients with SLS. Remarkably, 229T was absent from the NPHP group, but was detected at an allele frequency of 5.4% in SLS patients (). The small size of this cohort precluded meaningful statistical analysis. Nonetheless, these data suggested that the absence of 229T might have a protective effect from RP.
Frequency of RPGRIP1L A229T (c.685G>A; rs61747071) in ciliopathy patients of northern European descent
To investigate this possibility, we queried whether the 229T allele frequency is significantly different in a larger case-control cohort of northern European ciliopathy patients (Suppl. Table 1
). We categorized our patients into two groups, with retinal phenotypes being the sole distinguishing factor. Both groups exhibited significantly different 229T allele frequencies compared to controls (n=3016 control alleles); the allele was absent from the non-retinal degeneration group (n=230 alleles; p=1.92E-03) but was enriched in the retinal degeneration group (combined: 4.5%, n=974 alleles, p=6.45E-03; ). Furthermore, in agreement with our earlier data, we observed significantly different 229T allele frequencies between the non-retinal pathology and retinal degeneration ciliopathy groups (p=7.35E-05; ). To evaluate the statistical significance of our findings further, we performed permutation analyses where the observed genotypes in the entire sample were assigned randomly without replacement in the three groups (ciliopathy +RP, ciliopathy −RP, controls). After 107
independent random assignments, we counted the number of times that the differences in minor allele frequency in each group matched or exceeded the actual differences. This study provided additional evidence that the association data did not occur by chance (). Importantly, we found A229T to be transmitted through at least three different haplotypes (three patients contained unique RPGRIP1L
SNPs in cis
with 229T), while the frequency of 229T was indistinguishable across cohorts collected at different sites (Suppl. Table 1
), arguing against a stratification artifact.
Our genetic data suggest a role for 229T in the development of RP in ciliopathy patients. However, the low frequency of this allele in the population and its apparent modest effect in our cohorts limit the power of genetic studies. We therefore evaluated the potential consequences of A229T to protein function.
As a first test, we performed a genetic rescue experiment using an in vivo
model. We have shown previously that suppression of ciliary and basal body proteins in zebrafish leads to measurable phenotypes in mid-somitic embryos, some of which are caused by defects in Wnt signaling16,17
, a phenotype also observed in other experimental ciliopathy models18
. The ability of human mRNA to rescue somitic phenotypes has proven useful for establishing the pathogenic potential of alleles by comparing the efficiency of rescue of wild type (wt) human messenger RNA (mRNA) compared to mRNA bearing the missense variant(s) under investigation7
. We identified a single partial zebrafish transcript that, upon reciprocal BLAST analysis, showed the highest amino acid similarity between zebrafish and human (73% similarity, 55% identity), to which we designed a splice-blocking morpholino (MO). The MO was injected into wild-type (wt) embryos at the two-cell stage (50-100 embryos per injection) and scored at the eight to nine somite stage. Consistent with data from other ciliopathy genes7,10,16,17
, we observed dose-dependent gastrulation defects that included a shortened body axis, thin somites with broad lateral extensions, minor kinking of the notochord, underdeveloped anterior structures and tail extension abnormalities, all of which enabled us to categorize embryos into two classes based on objective criteria (Class I or Class II; ; Suppl. Fig. 2a
). These phenotypes resulted from loss of rpgrip1l
, since RT-PCR and sequencing analysis of embryos showed 10 ng of MO to cause ~50% reduction of rpgrip1l
mRNA (Suppl. Fig. 2b
). Importantly, the phenotypes could be rescued by co-injection of 50 pg of wt human mRNA (embryos scored blind to injection cocktails; , Suppl. Table 2
Functional assessment of the RPGRIP1L A229T variant in vivo
We therefore proceeded to compare the efficiency of rescue between human mRNA encoding each of 229A and 229T. In addition, we included two known pathogenic missense alleles, T615P and A695P, that cause JBTS14,15
as positive controls, as well as the neutral G1025S variant as a negative control. Embryos were co-injected with a cocktail of MO and each human mRNA (n=88-115 embryos per allele) and scored blind to the injection cocktail (Suppl. Table 2
). We found that the 229T rescue gastrulation phenotype was significantly different from wt mRNA rescue (p<0.0001; , Suppl Table 2
). The assay was specific, since each of T615P and A695P were scored as pathogenic, but had no off-target effects; the rescue of the 1025S-encoding mRNA was indistinguishable from the 1025G message ().
To substantiate these assessments, we measured embryonic structures labeled in situ
, and myoD
riboprobes. On whole embryo flat-mounts age-matched to eight to nine somites, we calculated the ratio of the width (w) spanning the distal ends of the 5th
somites to the length (l) of the notochord as demarcated by adaxial cell labeling (n=8-20 embryos measured per variant). Consistent with our live embryo scoring, comparison of the mean w/l ratios between wt and mutant rescues indicated that both positive control alleles and A229T, but not G1025S, are pathogenic (; Suppl. Table 3
As a third line of in vivo
evidence, we examined 5-day rpgrip1l
morphant larvae (n=52-78 embryos/injection). Consistent with other ciliopathy morphants19
, we observed tail extension abnormalities in 35% of rpgrip1l
morphants, which was rescued by co-injection of wt human RPGRIP1L
message (). However, mRNA encoding 229T was unable to efficiently rescue the tail phenotype (p=0.031), consistent with both the T615P and A695P rescues (p=0.001 and 0.004 respectively). These observations, in addition to the absence of tail phenotypes in morphants rescued with G1025S mRNA, substantiated further the specificity of our assay ().
We next turned to the question of why A229T might have the observed effect on the retinal degeneration phenotype. We noted that mutations in one of the NPHP genes, NPHP5
, are also associated with an increased incidence of SLS; all tested patients with NPHP5
null variants exhibit retinal degeneration compared to a 10% incidence of retinopathy in patients with mutations in other NPHP-associated genes20
. NPHP5 interacts with RPGR20
, a ciliary protein mutated in 70-80% of X-linked RP and 20-25% of simplex RP21-24
. We therefore hypothesized that RPGRIP1L might also interact biochemically with RPGR and that the A229T allele might have a measurable effect on the affinity of this interaction. As part of our ongoing efforts to understand the pathobiology of RPGR, we had performed yeast two-hybrid (Y2H) screens of human retinal cDNA as prey using full-length RPGRORF15
as bait and identified the carboxyl-terminal domain of RPGRIP1L (RPGR-interacting domain; RID); this interaction was validated by known-bait and known-prey analysis using different domains of RPGR as bait (). To corroborate the findings, we generated an anti-human RPGRIP1L antibody (Suppl. Fig. 3
) and performed co-immunoprecipitations (IP) using bovine retinal extracts. In reciprocal IP experiments with RPGRIP1L or RPGR antibody, we detected the expected bands on immunoblots corresponding to RPGR or RPGRIP1L, respectively (). Co-transfection of COS-1 cells followed by co-IP demonstrated that Xpress-tagged RPGR and GFP-RPGRIP1L proteins are part of the same complex (). We then tested A229T and three reported JBTS variants14,15
that are shown to be pathogenic in our zebrafish assay. Triplicate experiments showed that the GFP-tagged A229T mutant protein showed considerably lower (~80%) ability to immunoprecipitate RPGR and has a profound effect on RPGR binding compared to the JBTS mutations (). We also found that RPGRIP1L localizes to the cilium and basal body of mouse IMCD3 cells and to the inner segment/connecting cilium of mouse photoreceptors (Suppl. Fig. 4
), in the same spatiotemporal context as demonstrated previously for RPGR4,23,24
. Hence, the abrogation of RPGRIP1L-RPGR interaction is likely to be relevant in photoreceptors.
RPGRIP1L interacts with RPGR and missense variant A229T abrogates this interaction
Second-site modifications are emerging as prevalent phenomena in Mendelian disorders. For example, PLS3 is a protective modulator of SMN1 dysfunction associated with Spinal Muscular Atrophy (SMA)25
; mutations in CNTF
modify the age of onset in amyotrophic lateral sclerosis (ALS)26
, while a mutation in MGC1203
can modify the global penetrance/expressivity in BBS10
. However, the penetrance and expressivity of tissue-specific defects within the context of syndromic disease pose a considerable challenge, especially for alleles of modest frequency in the general population and modest effect on the phenotype, which cannot be captured by traditional population-based genetic tools. Our data suggest that, in patients of northern European descent, the presence of the 229T allele of RPGRIP1L has predictive value for retinal degeneration. It will be important to consider the nature and position of mutations at the primary locus, since A229T probably exerts its effect in the context of mutations in specific ciliopathy proteins that can complex with RPGR in the retina; this is consistent with the previously proposed poison model of genetic epistasis27
It is possible that some of the other alleles reported herein might also exhibit a similar effect; however, their rarity will necessitate the evaluation of larger cohorts. Notably, in vivo
evaluation of each of the rare alleles using the zebrafish rescue assay suggested that most are pathogenic (, Suppl. Table 2
, Suppl. Table 3
, Suppl. Fig 5
) and that they can potentially interact with alleles at various sites of the ciliopathy genetic spectrum (). Also, destabilization of the RPGRIP1L-RPGR complex is unlikely to be the sole cause of retinal degeneration. RPGRIP1L also associates with RPGRIP1 (), null mutations in which cause early-onset, severe retinal degeneration28
. Three of the alleles found in our cohort, A1183G, R1236C and D1264Y map within the RPGRIP1-interacting domain of RPGRIP1L; hence, their contribution to disease might be mediated by loss of RPGRIP1 activity/function. Finally, RPGRIP1L and RPGR are members of a larger macromolecular complex whose composition and function will vary across tissue types and ciliary functions. Establishment of the content and the stoichiometry of such complex(es), coupled to the evaluation of ciliopathy alleles on the stability of the macromolecular complex in a tissue/function-dependent manner will enable the currently poor allelic stratification of variability in disease phenotypes.
Functional assessment of RPGRIP1L missense variants in an in vivo zebrafish model