Contribution of intraflagellar transport to compartmentalization and maintenance of the photoreceptor cell

Significance

Many cases of inherited blindness result from defects in protein trafficking to the light-sensitive ciliary compartment of photoreceptor cells called the outer segment. Like in other cilia, this trafficking relies on the process of intraflagellar transport (IFT). By acutely disrupting IFT in photoreceptors, we uncovered a major pathway for disposal of mislocalized proteins in these cells and revealed functional specializations of different types of ciliary transport protein complexes. Furthermore, our findings suggest a potential therapeutic strategy to ameliorate vision loss in patients suffering from these conditions.

Abstract

The first steps of vision take place in the ciliary outer segment compartment of photoreceptor cells. The protein composition of outer segments is uniquely suited to perform this function. The most abundant among these proteins is the visual pigment, rhodopsin, whose outer segment trafficking involves intraflagellar transport (IFT). Here, we report three major findings from the analysis of mice in which ciliary transport was acutely impaired by conditional knockouts of IFT-B subunits. First, we demonstrate the existence of a sorting mechanism whereby mislocalized rhodopsin is recruited to and concentrated in extracellular vesicles prior to their release, presumably to protect the cell from adverse effects of protein mislocalization. Second, reducing rhodopsin expression significantly delays photoreceptor degeneration caused by IFT disruption, suggesting that controlling rhodopsin levels may be an effective therapy for some cases of retinal degenerative disease. Last, the loss of IFT-B subunits does not recapitulate a phenotype observed in mutants of the BBSome (another ciliary transport protein complex relying on IFT) in which non-ciliary proteins accumulate in the outer segment. Whereas it is widely thought that the role of the BBSome is to primarily participate in ciliary transport, our data suggest that the BBSome has another major function independent of IFT and possibly related to maintaining the diffusion barrier of the ciliary transition zone.

Most vertebrate cells have a primary cilium (also known as a sensory cilium) that functions as a cellular antenna receiving a broad range of signals from the outside environment (see ref. 1 for a recent comprehensive review). Defects in cilia lead to a collective group of human diseases known as ciliopathies (2). Because cilia lack the capacity for protein synthesis, proteins must be synthesized in the cell body and trafficked to the cilium in a process involving intraflagellar transport (IFT) (3). IFT facilitates the bidirectional transport of material into and within the cilium powered by microtubule motors carrying three multisubunit adaptor complexes: the IFT-A and -B particles and the BBSome. Defects in IFT lead to impaired ciliary assembly and maintenance (4–8).

One striking example of the primary cilium is the outer segment of photoreceptor cells, which is responsible for performing the first steps of vision. Unlike other cilia, the outer segment is packed with layers of light-sensitive membranes containing a high concentration of the visual pigment, rhodopsin (9). As in other ciliated cells, outer segment proteins, including rhodopsin, are synthesized in the cell body (called the inner segment in photoreceptor cells) and subsequently trafficked to their final destination. The process of outer segment trafficking, best studied for rhodopsin, relies on IFT, along with several other molecular players (10, 11). Disruptions in IFT lead to severe degenerative phenotypes associated with rhodopsin mislocalization (12–25). Given a strong association between rhodopsin mislocalization and human retinal disease (26), some currently explored therapeutic strategies for rhodopsin-associated visual loss seek to reduce the levels of mutant rhodopsin accumulating in the cell (27).

In various IFT mutants and other animal models where rhodopsin is mislocalized from outer segments to the cell body, photoreceptor degeneration is accompanied by the accumulation of extracellular vesicles containing rhodopsin, which surround the inner segment of mutant photoreceptors (12, 14, 28–33) (see ref. 34 for a recent review). A similar vesicle accumulation occurs in animals transgenically expressing rhodopsin mutants that affect its trafficking (35–40). Notably, these vesicles are not released from the cilium, as documented in some other mutant photoreceptors (34) and frequently observed in other primary cilia (41), but are released from the inner segment plasma membrane (38). However, they do not contain the Na⁺/K⁺ ATPase normally residing in this membrane, suggesting that protein cargo(es) of these vesicles may be sorted prior to release (40). Importantly, normal photoreceptors also release extracellular vesicles containing rhodopsin from their inner segments (33), indicating that the vesicular release of mistrafficked proteins from photoreceptors is a fundamental process occurring in both health and disease.

In this study, we sought to analyze extracellular vesicle release from photoreceptor inner segments that is acutely induced by impairing IFT. Our main model was a conditional knockout of IFT20, which is unique among IFT subunits as it can exist both as a part of the IFT-B particle and in other protein complexes (42–44). It has been suggested that IFT20 plays a specific role in the trafficking of ciliary proteins, including rhodopsin, from the Golgi to the cilium (45–47).

Our results demonstrate that the loss of IFT20 leads to photoreceptor degeneration accompanied by a massive accumulation of extracellular vesicles budding from the photoreceptor inner segment. Using quantitative immunogold labeling, we showed that rhodopsin density in these vesicles is significantly higher than in the plasma membrane from which these vesicles originate. We next showed that vesicular release from IFT20 knockout photoreceptors is nearly ablated when rhodopsin is also knocked out and that reducing the level of rhodopsin expression in these photoreceptors markedly reduced the rate of their degeneration. These data reveal the existence of a sorting mechanism whereby mislocalized rhodopsin is recruited to and concentrated in extracellular vesicles prior to their release. This vesicular release may serve as a mechanism that allows these cells to cope with adverse effects of protein mislocalization caused by a range of mutations affecting the normal process of intracellular protein transport.

Another exciting result obtained in this study is that the loss of either IFT20 or another IFT-B subunit, IFT172, did not recapitulate a phenotype observed in mutants of the BBSome in which inner segment proteins abnormally accumulate in the outer segment (32, 48–54). Given that ciliary transport of the BBSome is dependent on IFT-B (3, 55), our data suggest that BBS proteins may serve another function related to maintaining the unique protein compartmentalization of the cilium, which is not dependent on IFT.

Results

Rod-Specific Knockout of IFT20 Causes Photoreceptor Degeneration.

To generate a rod-specific IFT20 knockout mouse, we crossed the Ift20^fl/fl mouse (56) with the iCre75 mouse in which Cre recombinase is expressed under control of the rhodopsin promoter (57). Light microscopy of retinal cross-sections from iCre75; Ift20^fl/fl and control Ift20^fl/fl mice showed relatively normal photoreceptor morphology up to ~P30, after which photoreceptors began to degenerate (Fig. 1). At P45, iCre75; Ift20^fl/fl retinas displayed an ~25% reduction in the amount of photoreceptor cells, and the outer segments of remaining cells were severely disorganized. By P60, outer segments were completely ablated, and by P90, nearly all photoreceptors were lost. Notably, cones appeared normal at the onset of rod photoreceptor degeneration, consistent with the IFT20 knockout being rod-specific (SI Appendix, Fig. S1).

Fig. 1.

Loss of IFT20 leads to severe photoreceptor degeneration that can be ameliorated by reducing rhodopsin expression. (A) Representative light microscopy images of control (Ift20^fl/fl), IFT20 knockout (iCre75; Ift20^fl/fl), and combined IFT20 knockout/rhodopsin hemizygous knockout (iCre75; Ift20^fl/fl; Rho^+/−) retinas at P30 through P90. OS: outer segment; IS: inner segment; ONL: outer nuclear layer. (Scale bar: 10 µm.) (B) Quantification of the number of photoreceptor nuclei in 100 µm segments of retinal sections located at 500 µm intervals from the optic nerve (ON) at each timepoint. For each genotype, three retinas from separate mice were analyzed. Two-way ANOVA with Tukey post hoc test was used to compare the nuclear counts (averaged across locations not including optic nerve) between genotypes at each timepoint. At P30, there was no statistically significant reduction in photoreceptor numbers in iCre75; Ift20^fl/fl retinas compared to Ift20^fl/fl retinas (P = 0.7477). At each subsequent timepoint, there was a statistically significant reduction in photoreceptor numbers in iCre75; Ift20^fl/fl retinas compared to Ift20^fl/fl retinas (P < 0.0001) and a statistically significant preservation of photoreceptors in iCre75; Ift20^fl/fl; Rho^+/− retinas (P < 0.0001). Error bars represent mean ± SEM.

To interpret the progression of this phenotype, we analyzed the time course of IFT20 loss in these mice. Because IFT20 is expressed by most retinal cell types (43), we used fluorescence-activated cell sorting (FACS) of dissociated retinal cells to isolate rod photoreceptors using an anti-CD73 antibody, which specifically recognizes rods in the retina (58) (SI Appendix, Fig. S2). DNA recombination in isolated cells was analyzed by PCR of the Ift20 gene region excised by Cre activity. A nearly complete Ift20 gene excision occurred by P11 (SI Appendix, Fig. S3 A and B), which is consistent with Cre expression in this line starting at ~P7 (57). We next used western blotting to analyze the level of IFT20 protein in sorted cells and found that it persists until ~P37 (SI Appendix, Fig. S3C). This was the oldest age we were able to reliably sort cells at, after which rapid photoreceptor degeneration made cells too fragile for reliable isolation. The presence of IFT20 protein in photoreceptors up to P37 shows that its lifetime is ~4 wk following Cre/lox recombination. After this time, photoreceptor degeneration begins, indicating the depletion of the IFT20 protein pool in these cells rapidly leading to their death. Notably, this time course of protein depletion is comparable with that in another recently analyzed model of conditional ArpC3 knockout using the same iCre75 line (59).

Rod-Specific Knockout of IFT20 Causes Mislocalization of Rhodopsin and Several Other Outer Segment Proteins.

Consistent with previous reports on the role of IFT20 in rhodopsin transport (46, 47), we found that rhodopsin was partially mislocalized from the outer segments of iCre75; Ift20^fl/fl rods analyzed at the onset of photoreceptor degeneration (Fig. 2) but not at earlier timepoints before IFT20 protein was depleted (SI Appendix, Fig. S4). We also investigated the localization patterns of several other representative outer segment proteins. Interestingly, some of them—guanylate cyclase 2, ABCA4, and, to a lesser degree, PRCD and the CNG channel—were also partially mislocalized, suggesting that their outer segment trafficking is dependent on IFT. On the other hand, we did not observe any change in the localization of some other proteins: peripherin-2, ROM1, and prominin-1. This suggests that either they employ alternative transport mechanism(s) or they are efficiently degraded when mislocalized. The former is consistent with reports that at least one of them, peripherin-2, utilizes an outer segment trafficking pathway distinct from that of rhodopsin (60).

Fig. 2.

Loss of IFT20 leads to mislocalization of rhodopsin and a subset of other outer segment proteins. Representative images of retinal cross-sections from control (Ift20^fl/fl) and IFT20 knockout (iCre75; Ift20^fl/fl) mice stained with antibodies against indicated outer segment proteins (green). Nuclei are labeled with Hoechst in blue in the left half of each panel. Mice were analyzed at P40 to P42. For each genotype, three retinas from separate mice were analyzed. Images shown are average intensity z-projections from z-sections with step size of 1 µm for a total of 10 µm thickness. OS: outer segment; IS: inner segment; ONL: outer nuclear layer. (Scale bars: 10 µm.)

Reduced Rhodopsin Expression Promotes Survival of IFT20 Knockout Rods.

Given that rhodopsin mislocalization is associated with photoreceptor degeneration (26), we questioned the extent to which the defects in rhodopsin transport underlie cell death caused by IFT20 knockout. We reasoned that, by reducing rhodopsin expression, we may ameliorate the pathology associated with rhodopsin mislocalization in this mouse. A useful model for testing this idea is the rhodopsin hemizygous knockout (Rho^+/−) mouse whose rods express ~50% of the normal rhodopsin content while remaining healthy (61). Therefore, we analyzed retinas of iCre75; Ift20^fl/fl; Rho^+/− mice and found that their photoreceptor degeneration was significantly slower than in iCre75; Ift20^fl/fl mice (Fig. 1). Particularly striking was the difference observed at P90 when iCre75; Ift20^fl/fl retinas had only a few cone nuclei remaining, while iCre75; Ift20^fl/fl; Rho^+/− retinas still retained 3 to 4 rows of photoreceptor nuclei. These data show that mislocalized rhodopsin is a major contributing factor to photoreceptor cell death observed in the rod-specific IFT20 knockout mouse and that attenuating the level of rhodopsin can promote photoreceptor cell survival in this mouse.

Inspired by this striking improvement of photoreceptor survival in the IFT20 knockout, we questioned whether reducing rhodopsin expression would have similar protective effects in other models of photoreceptor degeneration associated with rhodopsin mislocalization. Unlike IFT20, which can exist outside of the IFT particles, IFT172 is a constitutive subunit of the IFT-B particle (62). Mutations in IFT172 have been identified in patients with retinal degenerative disease (63), and the rod-specific IFT172 knockout mouse (iCre75; Ift172^fl/fl) causes photoreceptor degeneration associated with rhodopsin mislocalization (14). IFT172 protein was shown to be depleted at ~P28 in this knockout mouse (14), so we analyzed retinas of iCre75; Ift172^fl/fl; Rho^+/− and iCre75; Ift172^fl/fl mice at timepoints both before and after IFT172 protein depletion (Fig. 3). We found that the rapid photoreceptor degeneration occurring in IFT172 knockout retinas was significantly slowed down when rhodopsin expression was decreased. This finding suggests that attenuating the level of rhodopsin can promote photoreceptor cell survival in a variety of photoreceptor degenerative conditions in which rhodopsin is mislocalized.

Fig. 3.

Photoreceptor degeneration caused by the loss of IFT172 can be ameliorated by reducing rhodopsin expression. (A) Representative light microscopy images of control (Ift172^fl/fl), IFT172 knockout (iCre75; Ift172^fl/fl), and combined IFT172 knockout/rhodopsin hemizygous knockout (iCre75; Ift172^fl/fl; Rho^+/−) retinas at P25 and P40. OS: outer segment; IS: inner segment; ONL: outer nuclear layer. (Scale bar: 10 µm.) (B) Quantification of the number of photoreceptor nuclei in 100 µm segments of retinal sections located at 500 µm intervals from the optic nerve (ON) at each timepoint. For each genotype, three retinas from separate mice were analyzed. Two-way ANOVA with Tukey post hoc test was used to compare the nuclear counts (averaged across locations not including optic nerve) between genotypes at each timepoint. At P25, there were no statistically significant differences in photoreceptor numbers between genotypes (P = 0.0799). At P40, there was a statistically significant reduction in photoreceptor numbers in iCre75; Ift172^fl/fl retinas compared to Ift172^fl/fl retinas (P < 0.0001) and a statistically significant preservation of photoreceptors in iCre75; Ift172^fl/fl; Rho^+/− retinas (P < 0.0001). Error bars represent mean ± SEM.

Rod-Specific Knockout of IFT20 Leads to Release of Extracellular Vesicles From the Inner Segment.

We next performed ultrastructural analysis of iCre75; Ift20^fl/fl retinas during the early stages of photoreceptor degeneration between P30 and P45 (Fig. 4). In this time window, the major morphological phenotype consisted of progressive shortening and eventual loss of outer segments (Fig. 4A). Notably, the majority of remaining outer segments retained their typical cylindrical shape, a phenotype consistent with suppressed outer segment protein trafficking rather than a specific defect in photoreceptor disc formation. In contrast, inner segments of iCre75; Ift20^fl/fl rods appeared essentially normal, including the normal appearance of the biosynthetic membranes (Fig. 4B). However, we did observe a massive accumulation of extracellular vesicles surrounding inner segments (Fig. 4C). The inner segment origin of these vesicles was corroborated by observations of vesicles captured in the process of budding directly from the inner segment plasma membrane (Fig. 4D). This indicates that they can be classified as “microvesicles” or “ectosomes”. The release of extracellular vesicles from iCre75; Ift20^fl/fl rods is not a light-dependent phenomenon, as vesicles were also observed in dark-reared mice (SI Appendix, Fig. S5).

Fig. 4.

IFT20 knockout leads to outer segment loss and release of extracellular vesicles from inner segments. (A) Representative TEM images of control (Ift20^fl/fl) and IFT20 knockout (iCre75; Ift20^fl/fl) retinas at P30 through P45. (B) Representative high-magnification TEM images of the Golgi apparatus (magenta arrows) obtained at P37. (C) Representative high-magnification TEM images of extracellular vesicles (yellow arrows) adjacent to photoreceptor inner segments obtained at P37. (D) Representative high-magnification TEM images of extracellular vesicles in the process of budding off of the inner segment plasma membrane (yellow arrows) obtained at P45. For each genotype, three retinas from separate mice were analyzed. OS: outer segment; IS: inner segment. [ Scale bars: 5 µm (A), 1 µm (B and C), 0.2 µm (D).]

Rhodopsin is Concentrated in Extracellular Vesicles Prior to Their Release.

In the next set of experiments, we conducted immunogold labeling of iCre75; Ift20^fl/fl and control retinas with anti-rhodopsin antibodies (Fig. 5 and SI Appendix, S6). The specificity of rhodopsin labeling was evident from the dense labeling of outer segments of rods but not adjacent cones in both knockout and control retinas (Fig. 5 A and B). Consistent with observations from other animal models in which extracellular vesicles accumulate around inner segments (12, 29, 33, 35, 38, 64), the vesicles found in iCre75; Ift20^fl/fl retinas were labeled for rhodopsin (Fig. 5A). Occasionally, we also observed examples in which rhodopsin labeling was concentrated in small patches of the inner segment plasma membrane, potentially revealing an early stage of vesicle formation (Fig. 5C and SI Appendix, S7).

Fig. 5.

Rhodopsin is concentrated in extracellular vesicles. (A) Representative immunogold labeling of IFT20 knockout (iCre75; Ift20^fl/fl) retinal sections at P40 using the 1D4 anti-rhodopsin antibody. (B) Representative immunogold labeling of control (Ift20^fl/fl) retinal sections at P40 using the 1D4 anti-rhodopsin antibody. An extracellular vesicle lacking anti-rhodopsin labeling is marked with an asterisk. (C) Representative immunogold labeling of IFT20 knockout (iCre75; Ift20^fl/fl) retinal sections at P40 using the 4D2 anti-rhodopsin antibody. The boxed region (Left) is magnified on the right. A high local concentration of rhodopsin labeling is marked with an arrowhead. (D) Quantification of the number of gold particles per unit of membrane length. Analysis was performed on three pairs of IFT20 knockout (iCre75; Ift20^fl/fl) and control (Ift20^fl/fl) mice. Retinas from each pair were processed and analyzed side by side. To account for small variations in labeling efficiency across experiments, rhodopsin labeling density in all cellular compartments was normalized to OS labeling in the control retina of each pair (see data and analysis for each independent experiment in SI Appendix, Fig. S8http://www.pnas.org/lookup/doi/10.1073/pnas.2408551121#supplementary-materials). Error bars represent mean ± SEM. OS: outer segment; IS: inner segment; PM: plasma membrane (marked with arrows); EV: extracellular vesicle. [Scale bars: 0.5 µm (A and B), 0.25 µm (C).]

We compared the labeling density of rhodopsin in vesicles produced by iCre75; Ift20^fl/fl rods with that in the inner segment plasma membrane and outer segment discs. Because rhodopsin is a membrane protein, we expressed its labeling density as the number of gold particles per unit of membrane length. Strikingly, vesicles were labeled ~3-fold more densely than the inner segment plasma membrane from which they originate (Fig. 5D and SI Appendix, S8). The labeling of vesicles was as high as the outer segment disc membranes where rhodopsin comprises ~90% of the total transmembrane material, although this may reflect antibody binding saturation in one or both membranes highly packed with rhodopsin. These data provide strong evidence that rhodopsin becomes highly concentrated while being packed into the extracellular vesicles and support the existence of a sorting mechanism whereby selected proteins can be either enriched in or excluded from these vesicles.

We next analyzed rhodopsin immunogold labeling in the extracellular vesicles occasionally released by normal photoreceptors (Fig. 5B). As shown previously, about half of these vesicles contain rhodopsin (33). We found that the labeling density of rhodopsin-positive vesicles was the same as in vesicles from iCre75; Ift20^fl/fl mice (Fig. 5D and SI Appendix, S8). In contrast, the labeling density of the inner segment plasma membrane of control mice was very low, ~20-fold less than that of the vesicles. Thus, rhodopsin becomes highly concentrated during vesicle formation in healthy rods, as well.

Our finding that the loss of IFT20 leads to an increase in the rhodopsin labeling of the inner segment plasma membrane contradicts the previous report that the loss of IFT20 leads to an accumulation of rhodopsin in the Golgi but not in the plasma membrane (46, 47). Furthermore, our analysis of rhodopsin labeling in the Golgi of control and iCre75; Ift20^fl/fl rods revealed no difference (Fig. 5D and SI Appendix, S8). In considering this discrepancy, we should stress that, in previous reports, mice were analyzed two days after initiating Cre/lox recombination by tamoxifen, while our current data demonstrate that the IFT20 protein persists in these cells for at least twenty days following Cre/lox recombination (SI Appendix, Fig. S3). Such a lack of rhodopsin accumulation in Golgi in our experiments argues against a unique role of IFT20 in rhodopsin recruitment at the Golgi and rather suggests that the functional role of IFT20 in photoreceptors is no different than that of other IFT-B particle subunits.

Rhodopsin Knockout Greatly Suppresses Vesicular Release from IFT20 Knockout Rods.

To further interrogate the relationship between rhodopsin mislocalization and vesicular release, we questioned whether eliminating rhodopsin from IFT20 knockout rods would preclude extracellular vesicle release. We crossed iCre75; Ift20^fl/fl and rhodopsin knockout (Rho^−/−) mice and found that the massive accumulation of extracellular vesicles associated with IFT20 knockout was almost completely ablated in the absence of rhodopsin, with only a few examples of extracellular vesicles observed (Fig. 6). We further showed that the amount of accumulated vesicles in retinas of IFT20 knockout mice bearing a single copy of the rhodopsin gene was reduced to ~42% of IFT20 knockout mice expressing both copies of the rhodopsin gene (SI Appendix, Fig. S9). This correlation between the levels of rhodopsin expression and vesicle release further supports the concept that vesicular release from photoreceptor inner segments serves to dispose mislocalized rhodopsin.

Fig. 6.

Knockout of rhodopsin greatly suppresses vesicle release in IFT20 knockout retinas. Representative TEM images of rhodopsin knockout (Ift20^fl/fl; Rho^−/−) or double rhodopsin and IFT20 knockout (iCre75; Ift20^fl/fl; Rho^−/−) retinas obtained at P37. For each genotype, three retinas from separate mice were analyzed. Extracellular vesicles occasionally observed next to photoreceptor inner segments are marked by yellow arrows. (Scale bars: 1 µm.)

Loss of IFT Subunits does not Affect the Cellular Distribution of Syntaxin-3.

In the last set of experiments, we tested whether the knockout of IFT20 recapitulates another major phenotype observed in knockouts of various BBS proteins whose involvement in intracellular trafficking is dependent on IFT (3, 55). In addition to rhodopsin mislocalization and accumulation of extracellular vesicles, both relatively minor compared to IFT mutants, BBS mutants display a prominent outer segment accumulation of many proteins that are excluded from this cellular compartment in WT rods (48, 54). The commonly used marker for this abnormal protein accumulation is syntaxin-3, which in WT photoreceptors is localized to the plasma membrane of the cell soma and is excluded from the outer segment (32, 48–54). In contrast to a BBS mutant (Bbs10^−/−; Fig. 7A), the loss of IFT20 did not lead to any notable outer segment accumulation of syntaxin-3 (Fig. 7B). We next analyzed syntaxin-3 distribution in the rod-specific IFT172 knockout mouse. As in the case of the IFT20 knockout, we observed rhodopsin mislocalization and the release of rhodopsin-containing vesicles but not outer segment accumulation of syntaxin-3 in these rods (Fig. 7C and SI Appendix, S10).

Fig. 7.

Loss of IFT subunits does not lead to outer segment accumulation of syntaxin-3. (A) Representative images of retinal cross-sections from BBS10 knockout (Bbs10^−/−) and control (Bbs10^+/+) mice analyzed at P16-P18. Sections were stained with antibodies against syntaxin-3 (red) and rhodopsin (green). Nuclei are labeled with Hoechst in blue. (B) Representative images of retinal cross-sections from IFT20 knockout (iCre75; Ift20^fl/fl) and control (Ift20^fl/fl) mice analyzed at P40. (C) Representative images of retinal cross-sections from IFT172 knockout (iCre75; Ift172^fl/fl) and control (Ift172^fl/fl) mice analyzed at P28. In all panels, images shown are average intensity z-projections from z-sections with step size of 1 µm for a total of 2 µm thickness. For each genotype, three retinas from separate mice were analyzed. OS: outer segment; IS: inner segment; ONL: outer nuclear layer. (Scale bars: 10 µm.)

It was also demonstrated that the knockout of BBS4 leads to an enhanced level of protein ubiquitination in the outer segment (65), which was suggested to precede the removal of undesired proteins from the cilium by a BBS-dependent mechanism (65, 66). We observed a comparable increase in outer segment ubiquitin staining in Bbs10^−/− mice (Fig. 8), which shows that this effect takes place regardless of whether BBSomes are lost due to knockout of a core subunit [BBS4 (67)] or a chaperone required for BBSome assembly [BBS10 (68)]. In contrast, no such increase in protein ubiquitination was observed in IFT20 and IFT172 knockout mice (Fig. 8), consistent with no abnormal protein accumulation in their outer segments.

Fig. 8.

Loss of IFT subunits does not lead to outer segment accumulation of ubiquitinated proteins. Representative images of retinal cross-sections from control (Ift20^fl/fl, Ift172^fl/fl, Bbs10^+/+) and knockout (iCre75; Ift20^fl/fl, iCre75;Ift172^fl/fl, Bbs10^−/−) mice stained an antibody against ubiquitinated proteins (green). Nuclei are labeled with Hoechst in blue. Mice were analyzed at P16-18 (BBS10 knockout and control), P40 (IFT20 knockout and control), and P28 (IFT172 knockout and control). Images shown are average intensity z-projections from z-sections with step size of 1 µm for a total of 2 µm thickness. For each genotype, three retinas from separate mice were analyzed. OS: outer segment; IS: inner segment; ONL: outer nuclear layer. (Scale bar: 10 µm.)

Discussion

The first major finding of this study is that both healthy and diseased photoreceptors have the ability to concentrate mislocalized rhodopsin and dispose of it through the release of extracellular vesicles. This finding builds upon previous studies from Imanishi and colleagues who showed that photoreceptors can release mutant rhodopsin without releasing Na⁺/K⁺ ATPase, a protein normally found in the inner segment plasma membrane (38, 40). Together, these findings reveal the existence of a sorting mechanism in which inner segment proteins can be either enriched in or excluded from these vesicles. A graphical illustration of this phenomenon is presented in Fig. 9.

Fig. 9.

Model of rhodopsin recruitment and extracellular vesicle release. Rhodopsin normally localizes to the photoreceptor outer segment. However, rhodopsin can be mislocalized from this compartment due to occasional errors in its transport in normal photoreceptors or gross transport defects caused by certain mutations. This leads to accumulation of mislocalized rhodopsin in the inner segment plasma membrane, where it is recruited for release in extracellular vesicles. These extracellular vesicles accumulate between cells unless internalized by other cells.

Our results reveal several other interesting aspects of this vesicular release process. The finding that the rhodopsin labeling density of vesicles released by healthy and IFT20 knockout photoreceptors is comparable suggests that, in both cases, vesicular release is preceded by a sufficient number of rhodopsin molecules concentrating in a local site of the inner segment plasma membrane. The mechanism behind the recruitment of rhodopsin at this site remains a subject of future investigation. More generally, it is important to learn how the cell distinguishes mislocalized proteins from those normally residing in this membrane and whether the vesicle release mechanism applies to all proteins mislocalized from other subcellular compartments or to a specific protein subset.

Our second major finding is that the gross mislocalization of rhodopsin is a significant contributor to photoreceptor degeneration observed in the IFT knockouts, as knocking out one copy of the rhodopsin gene in both IFT20 and IFT172 knockouts greatly slowed the progression of photoreceptor degeneration. It remains to be seen whether attenuating the level of rhodopsin expression in other models of photoreceptor degeneration in which rhodopsin mislocalizes may be similarly beneficial. Should this be the case, the reduction of rhodopsin expression may serve as a mutation-independent therapeutic strategy to ameliorate photoreceptor loss in a range of inherited retinal degenerative diseases.

A related question is whether the vesicular release of mislocalized rhodopsin under pathological conditions promotes photoreceptor survival or accelerates photoreceptor cell death. While it is plausible that vesicular release of mislocalized rhodopsin is beneficial to the cell, as suggested in (38), the massive accumulation of vesicles in the interphotoreceptor space may be harmful to the retina. The extent of these vesicles’ clearance and the cell types responsible for such clearance also remain open to investigation. We previously showed that, at least in healthy retinas, photoreceptor cells can reabsorb these vesicles themselves (33). Other candidate cell types are Müller glia (69) and the retinal pigment epithelium (40). Under pathological conditions, it is likely that activated microglia migrating to the outer retina become involved, such as in the case of outer segment-derived vesicles from PRCD knockout photoreceptors (70).

The third major finding of this study is that the loss of IFT-B subunits does not phenocopy the phenomenon observed in all studied BBS mutants, which consists of a gross accumulation of non-outer segment proteins in this compartment (32, 48–54). Syntaxin-3 serves as a commonly used marker for studying this phenomenon (32, 48–54), whereas proteomic analyses showed at least a hundred other proteins being similarly mislocalized to the outer segments of BBS mutants (48, 54). The fact that cellular localization of syntaxin-3 is entirely normal in IFT-B mutants is highly unexpected because many consider the BBSome to be a third IFT particle (3) and, most importantly, ciliary transport of the BBSome is dependent on the IFT-B complex (3, 55). Mechanistically, the BBSome is reported to connect to the IFT-B particle through a protein called LZTFL1 (71, 72), whose mutation was analyzed in the first report of the outer segment protein accumulation phenomenon (48).

To seek a possible explanation for this difference between BBS and IFT-B mutants, we need to consider two potential scenarios under which syntaxin-3 accumulates in mutant outer segments. First, it is conceivable that under normal conditions, syntaxin-3 is able to leak into the outer segment but is efficiently removed by the BBSome. However, this is not consistent with the lack of syntaxin-3 accumulation in our IFT-B mutants given the dependency of BBSome transport on IFT-B. Furthermore, syntaxin-3 entering the outer segment is expected to freely diffuse across the membranes of newly forming discs, in which it would be captured upon disc enclosure. Once separated from the ciliary membrane in an enclosed disc, it could not be removed from the outer segment by BBSomes or any other IFT-related mechanism. For these reasons, and considering that syntaxin-3 is only one of more than a hundred proteins that accumulate in BBS mutant outer segments (48, 54), we favor an alternative possibility that the outer segment leakage of syntaxin-3 and other proteins is primarily a consequence of a defect in the barrier function of the ciliary transition zone, which maintains membrane compartmentalization in ciliated cells (73).

Two additional observations support the latter idea. First, leakage of syntaxin-3 and other inner segment proteins into the outer segment was documented in mice in which the barrier function of the ciliary transition zone is compromised (74–76). Second, the knockout of BBS1 in zebrafish results in both protein mislocalization and a change in the lipid composition of outer segment membranes (54)—a phenomenon consistent with a disruption in the diffusion barrier separating the inner and outer segments, but not with a mechanism involving directed protein transport across this barrier.

Taken together, these data suggest that BBS proteins may function independently of IFT-B in maintaining the integrity of the ciliary diffusion barrier. In fact, simultaneous loss of BBS and transition zone proteins affects the integrity of cilia far more dramatically than predicted from an additive effect of individual mutations (77, 78). The potential involvement of the BBSome in maintaining the integrity of the ciliary diffusion barrier in both photoreceptors and other ciliated cell types remains an exciting subject of future investigation.

Additional evidence for not fully overlapping functions of IFT and BBS proteins in photoreceptor cells may be found in the different clinical manifestations of inherited retinal degenerations arising from mutations in the corresponding genes. Whereas both lead to a broad range of systemic phenotypes (79, 80), the clinical course of retinal degeneration in patients with mutations in BBS genes tends to be somewhat different than that in patients with mutations in IFT genes. Patients with mutations in BBS genes often present with complaints of night blindness in childhood or early adolescence similar to typical retinitis pigmentosa; however, their central vision provided by cones is usually simultaneously affected, and they often have a concomitant bullseye maculopathy (81, 82). They lose central vision earlier than in typical retinitis pigmentosa, in which central visual acuity is maintained for many decades. In some patients with BBS mutations, a cone-only retinal degeneration may occur, or a retinal degeneration that involves solely the cones first, with photophobia as the primary symptom rather than night blindness (83). These patients have often lost most or all of their central, as well as peripheral, vision by their mid-20 s, unless they have a cone-only phenotype. In contrast, patients with retinal degeneration due to mutations in IFT proteins, such as IFT140 and IFT172, have a more typical retinitis pigmentosa clinical course, with night blindness and constricted peripheral vision in childhood, but retained central visual acuity until their 40 s or 50 s (63, 84). Along with our current findings, these differences in clinical manifestations call for further investigation into the contributions of IFT and BBS proteins to maintaining the functional compartmentalization of rod and cone photoreceptor cells.

Materials and Methods

Transmission Electron Microscopy (TEM).

Fixation and processing of mouse eyes for TEM was performed as described previously (85). Anesthetized mice were transcardially perfused with 2% paraformaldehyde, 2% glutaraldehyde, and 0.05% calcium chloride in 50 mM MOPS (pH 7.4). Enucleated eyes were fixed for an additional 2 h in the same fixation solution at room temperature. Eyecups were dissected from fixed eyes, embedded in 2.5% low-melt agarose (Precisionary), and cut into 200 µm thick slices on a Vibratome (VT1200S; Leica). Agarose sections were stained with 1% tannic acid (Electron Microscopy Sciences) and 1% uranyl acetate (Electron Microscopy Sciences), gradually dehydrated with ethanol, and infiltrated and embedded in Spurr’s resin (Electron Microscopy Sciences). 70 nm sections were cut, placed on copper grids, and counterstained with 2% uranyl acetate and 3.5% lead citrate (19314; Ted Pella). The samples were imaged on a JEM-1400 electron microscope (JEOL) at 60 kV with a digital camera (BioSprint; AMT). Image analysis and processing was performed with ImageJ.

Light Microscopy of Histological Sections.

Plastic-embedded blocks generated for TEM were sectioned through the optic nerve in 500 nm sections and stained with methylene blue for light microscopy as previously described (86). Images were taken with a confocal microscope (Eclipse 90i and A1 confocal scanner; Nikon) with a 60× objective (1.4 NA Plan Apochromat VC; Nikon) using Nikon NIS-Elements software. Image analysis and processing was performed with ImageJ. For the quantification of the number of photoreceptor nuclei, photoreceptor nuclei were counted in 100 µm boxes at 500 µm intervals from the optic nerve spanning 2,000 µm in each direction for three mice of each genotype, as previously described (87).

Immunofluorescence.

Anesthetized mice were transcardially perfused with a fixative solution containing 4% paraformaldehyde in 80 mM PIPES (pH 6.8), 5 mM EGTA, and 2 mM MgCl₂. Eyes were enucleated and postfixed in the same solution for two hours at room temperature. After fixation, dissected eyecups were embedded in 2.5% low-melt agarose (Precisionary) and cut by a Vibratome (VT1200S; Leica) into 100 µm thick slices. Agarose sections were blocked in PBS containing 7% donkey serum and 0.5% Triton X-100 for 1 h at room temperature before staining with primary antibody in blocking buffer overnight at 4 °C. For rhodopsin immunostaining, sections were incubated with primary antibodies for 3 d at 4 °C. After primary antibody staining, sections were washed three times in PBS and incubated with secondary antibody in blocking buffer for 2 h at room temperature. For anti-CD73 staining, the primary antibody is conjugated to APC, and no secondary antibody staining was performed. Finally, sections were washed three times in PBS, and nuclei were stained with 2 µg/ml Hoechst (Thermo Fisher Scientific) for 30 min at room temperature. In the case of PNA staining, sections were stained with 5 µg/ml PNA conjugated to Alexa Fluor 647 (Thermo Fisher Scientific) for 1 h at room temperature. Finally, sections were washed three times in PBS and mounted onto slides with Immu-Mount (Thermo) and coverslipped. Images were taken with a confocal microscope (Eclipse 90i and A1 confocal scanner; Nikon) with a 60× objective (1.4 NA Plan Apochromat VC; Nikon) using Nikon NIS-Elements software. Image analysis and processing was performed with ImageJ.

Immunogold Labeling.

Immunogold staining was performed as previously described (85). Eyes were fixed and cut into agarose sections as described above for regular TEM imaging. Sections were treated with 0.5% tannic acid, cryoprotected with 30% glycerol in 0.1 M sodium acetate, and freeze-substituted overnight in 4% uranyl acetate/95% methanol in a dry ice/ethanol bath. Sections were infiltrated with Lowicryl HM-20 (Electron Microscopy Sciences), slowly warmed to 4 °C, and polymerized for 3 d under UV light. Sections (70 nm) of the central retina were cut from embedded samples, placed on nickel grids, and treated with 10 mM citrate buffer (pH 6.0) containing 0.005% Tergitol NP-10 at 60 °C. Grids were blocked with 1% glycine in Tris-buffered saline (pH 7.6) containing 0.005% Tergitol NP-10 and incubated overnight with anti-rhodopsin antibodies. Grids were blocked with 1% donkey serum and incubated with donkey anti-mouse IgG conjugated with 6 nm colloidal gold (Jackson ImmunoResearch Laboratories). Grids were counterstained with 2% uranyl acetate and 3.5% lead citrate (Ted Pella). Samples were imaged on a JEM-1400 electron microscope (JEOL) at 60 kV with a digital camera (BioSprint). Image analysis and processing was performed with ImageJ.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.