Mutations in the ubiquitously expressed splicing factor PRPF31 are found in 8% of patients suffering from autosomal-dominant Retinitis Pigmentosa (ad-RP; ). In our study, we used the zebrafish model to investigate the etiology of two RP mutations in the PRPF31 gene, AD5 and SP117. While both mutations cause the same disease with identical symptoms, our data suggest that this is caused by two fundamentally different mechanisms. We show that transgenic expression of SP117 leads to aberrant localization of the mutant protein in the rod cytoplasm rather than nucleus, but does not affect rod morphology or numbers. This implicates a loss-of-function for the SP117 mutation which results in reduced U4/U6.U5 tri-snRNP assembly and suggests that in patients carrying this mutation the disease phenotype is caused by haploinsufficiency. On the other hand, our studies using over-expression of AD5 strongly suggest that this variant has dominant-negative activity and its expression results in the degeneration of rod outer segments and late onset apoptosis (see below).
This study represents the first in vivo analysis to show how PRPF31 mutations lead to photoreceptor degeneration. We report aberrant splicing in pre-mRNA transcripts using an animal model for RP that is easily accessible for bioimaging, high throughput genetic approaches and drug screening.
Splicing factor deficiencies lead to RP
PRPF31 and at least four other RP related PRPF factors (PRPF3, PRPF8, PAP1, SNRN200) play essential roles in pre-mRNA splicing. Three hypothetical mechanisms are presently discussed to explain how mutations in general splicing factors can lead to photoreceptor specific degeneration. First, null mutations result in hyploinsufficiency by loss of function of the mutant protein or degradation of mutant mRNA by nonsense-mediated mRNA decay (NMD) [34–36]. As a consequence, the splicing machinery becomes compromised in its efficiency and hence cannot fulfil all its functions. Second, mutations lead to proteins with dominant-negative activities. These proteins may interfere with splicing and also potential other cellular activities, thereby leading to damage of the affected tissue. Third, mutations might lead to proteins forming insoluble and cytotoxic aggregates. Proteins affected by this class of mutations may affect the tissue by a combination of loss-of-function and dominant-negative effects. Consistent with the third mechanism, a missense mutation in PRPF31, A216P, has been reported to lead to insoluble protein that is deposited in the cytoplasm .
SP117 and AD5 mutations cause photoreceptor degeneration by different modes
PRPF31 is ubiquitously expressed and sufficient levels are crucial for organ maintenance. Consistently, homozygous Prpf31 knock-out mice are embryonic lethal , whereas heterozygous Prpf31 knock-out mice surprisingly display degeneration of the retinal pigment epithelium (RPE) . Thus, different PRPF31 thresholds either lead to degeneration or lethality. Reduced PRPF31 protein levels were detected in lymphoblastoid cells of RP patients, suggesting that haploinsufficiency is one of the possible causes for RP [11, 37].
The SP117 mutation in PRPF31 results in premature termination before the NLS and has been reported to confer gain-of-function toxicity in cultured mouse retina cells, leading to reduced rhodopsin expression and apoptosis [31, 32]. In our in vivo study, SP117 was found distributed along the rod cytoplasm rather than the nucleus, which is consistent with reports on mouse retina cell culture experiments. However, we found that expression of SP117 in zebrafish rods does not have detrimental effects on rod morphology or survival. Even over-expression at high doses in early embryos did not result in increased lethality. Instead, we found that SP117 protein is significantly less stable than wild-type or AD5 mutant Prpf31. This indicates that other than proposed from cell culture studies, mutant SP117 may not show any toxicity. We instead suggest that SP117 protein is non-functional, unstable and mislocalized, consequently leading to reduced tri-snRNP levels and hence a weakening of the spliceosome in a mutant situation. Thus, haploinsufficiency appears to be responsible for photoreceptor cell degeneration in human RP patients carrying the SP117 mutation.
Recently, it was reported that RP mutations in PRPF3 and PRPF8 exert dominant-negative activity in knock-in mice (Prpf3
+/T494M and Prpf8
+/H2309P) and induce degenerative processes that however surprisingly are found in the retinal pigment epithelium and not photoreceptors . With our study, we provide evidence that also AD5 acts in a dominant-negative fashion. The AD5 mutation has been intensively studied in several cell culture models. Using lymphoblastoid cells from RP patients, Rio Frio et al. could not detect any truncated AD5 protein by Western analysis and therefore concluded that AD5 leads to RP by haploinsufficiency upon nonsense-mediated decay of the mutant RNA . However, this study did not analyze retina tissue and the expression pattern in lymphoblastoid cells may not represent the situation in photoreceptor cells where symptoms manifest. On the other hand, in studies using cultured mouse retinal cells, AD5 expression has been reported to result in reduced rhodopsin expression, increased apoptosis and splicing defects in artificial minigenes [31, 32] very similar to the findings in our in vivo model.
We provide four lines of evidence supporting that AD5 exhibits dominant-negative effect. First, AD5 over-expression leads to early embryonic lethality in a dose-dependent fashion in zebrafish embryos. This lethality can be partially rescued by co-expression of wild-type prpf31. Second, AD5 over-expression aggravates the effect of Morpholino-induced Prpf31 knock-down. Third, AD5 expression in zebrafish embryos shows the same effect on endogenous prpf31 transcript levels, as well as other retina-specific transcripts as the Morpholino induced gene knock-down of prpf31 . Finally, rod-specific AD5 expression in stable transgenic fish results in the degeneration of rod outer segments and late onset apoptosis. Thus, we propose that mutant AD5 competes with endogenous PRPF31 for binding in the U4/U6.U5 tri-snRNP eventually leading to an impairment of the tri-snRNP and decreased splicing activity. Consequently, RP in patients carrying only one functional PRPF31 allele might be caused through a combination of dominant-negative effects and haploinsufficiency.
Different degrees of rod photoreceptor degeneration visualized in AD5 transgenic zebrafish
Prpf31 knock-out and A216P knock-in mouse have been generated, however, none of them showed degeneration in the photoreceptor layer, but instead in the retinal pigment epithelium [8, 33]. In contrast, our AD5 stable transgenic fish model revealed different degrees of rod degeneration as well as photoreceptor death (Figure 6E) very similar to the symptoms observed in human RP patients. Thus, our data provide an in vivo model in which the AD5 mutation of Prp31 leads to degeneration in photoreceptor cells. We found that early AD5 protein was restricted to rod nuclei. At later stages, however, AD5 was often found in the inner segments possibly as a consequence of nuclear membrane disintegration or alternatively deficient nuclear import. Dysfunctional rods with degenerated outer segments were observed as well as cells expressing apoptotic markers and lacking both outer and inner segments. The relatively low percentage (3.7%) of identified apoptotic rods in adult AD5 transgenic fish suggests that the induced rod degeneration is a long-term process, possibly reflecting the typical features of adult onset in RP patients. It is possible that compensatory mechanisms, e.g. the up-regulation of endogenous prpf31 and rod regeneration, account for the observed low percentage of apoptosis. Fish continuously generate photoreceptor cells throughout their life and it has been reported that chronic rod cell death stimulates rod genesis . This may suggest that an increased number of newly generated rods in AD5 transgenic fish could mask the actual extent of rod cell death.
Splicing defects in AD5 expressing zebrafish retinas
We have previously shown that a prpf31 knock-down in zebrafish leads to the down-regulation of retina transcripts, many of which are implicated in RP pathogenesis or components of the photoreceptor specific transcription factor network . Here, we report for the first time splicing defects in some of these retina-specific transcripts caused by transgenic expression of AD5 in vivo. Screening a small panel of photoreceptor specific genes, we identified splicing defects in gnat1, crx, rx1 and rx3. gnat1 encodes the rod specific form of transducin (Trα), an essential component in the rod phototransduction cascade [39–41]. For Gnat1, loss or deterioration of rod function was reported in homozygous knock-out mice  and in transgenic mice expressing mutant TrαG38D  and TrαQ200 L . crx encodes a cone-rod homeobox protein, a retina specific transcription factor important for maintaining photoreceptor function . Crx null mice failed to form outer segments of photoreceptor cells resulting in photoreceptor degeneration . Most notably, Crx is a known RP causing gene and Crx mutations have been identified in RP patients . Finally, retinal homeobox (Rx/Rax) genes, rx1 and rx3, are essential for eye development [44, 45]. Loss of Rx function prevents eye formation (anophthalmia) in fish and mice [46–48]. In humans, mutations in RX lead to anophthalmia or microphthalmia . Hence, all genes with splicing defects identified in this study are crucial for the formation and/or maintenance of photoreceptor function. This opens the possibility that aberrantly spliced gnat1, crx, rx1 and rx2 contribute to the rod degeneration observed in AD5 transgenic fish.
In addition to splicing defects, we also observed reduced transcription of several other photoreceptor-specific genes very similar to the situation observed in zebrafish prpf31 morphants . We suggest that it is the combination of wrongly spliced as well as diminished transcripts that eventually causes rod degeneration. For the future, our model is also excellently suited to study the effect of degenerative processes in rods on neighboring cells. This is important as in RP patients cones are secondarily affected in a process known as bystander-associated cell death .
Hypothesis for the retina-specific phenotype
The retina is a fast-metabolizing tissue which has a high demand for correctly spliced transcripts. Reduced levels of functional tri-snRNPs due to mutations in PRPF31 may still be sufficient for most of the general cell types, but not adequate for highly demanding photoreceptor cells. This could lead to a situation where reduced splicing efficiency leads to the accumulation of defects mostly in the retina, eventually resulting in a retina-specific phenotype while other tissues are not affected. Consistent with this, we recently showed that only low levels of Prpf31 are required for maintenance of general organ development while retina development requires more Prpf31 . Alternatively, individual mRNAs that are problematic to splice under such conditions could also be the primary cause for retina-specific defects . In the present study, we detected splicing defects in retinas expressing mutant Prpf31. Our observation that only single introns of individual retina transcripts are affected by aberrant splicing suggests that a partial Prpf31 deficiency might indeed selectively affect splicing of a distinct subset of transcripts. The fact that one of these transcripts encodes Rx1, a retina-specific transcription factor, implicates that this could affect a transcriptional network consequently leading to a photoreceptor specific phenotype. Noteworthy, our in vivo approach has limitations as it used rod specific expression of a Prpf31 mutant (for comparison of expression levels of mutant versus endogenous Prpf31, see Additional file 4, Figure S4). Therefore, we cannot exclude that there are splicing defects also in other tissues under conditions of general Prpf31 deficiency that obviously, however, do not interfere with the function of the respective organ. Definitely, more work is required in the future to determine the mechanisms that underlie this interesting phenomenon. In addition, as the very strong rhodopsin promoter is used to drive expression of the SP117 and AD5 RNAs, there is a possibility that the levels of mutant proteins, AD5 in particular, may contribute to the changes observed.