Glutamylation imbalance impairs the molecular architecture of the photoreceptor cilium

Using Ultrastructure Expansion Microscopy (U-ExM) that we recently adapted for retina imaging (Mercey et al, 2022; Gambarotto et al, 2018), we first assessed the nanoscale localization of tubulin PTMs along the rod photoreceptor cilium—the outer segment—in mouse retina with an expansion factor of 4.2× (Fig. EV1A,B). We focused on the longitudinal and transversal views comprising the basal body, the connecting cilium, where many retinopathy-associated proteins localize (Bachmann-Gagescu and Neuhauss, 2019), the bulge region that we recently described as a crucial compartment for membrane disc formation (Faber et al, 2023) and the distal axoneme. To map the distribution of different PTMs along the axoneme of the outer segment, we co-stained all microtubules using a mixture of anti α- and β-tubulin antibodies with a panoply of antibodies specific to tubulin PTMs: mono-glycylation (TAP952), acetylation (acetylated tubulin), glutamylation- and polyglutamylation (GT335 and PolyE) and detyrosination (detyrosinated tubulin) (Figs. 1A–E and EV1C).

We first mapped tubulin glycylation. This modification has been mainly found in motile cilia/flagella (Gadadhar et al, 2021; Grau et al, 2013) but is also present during primary cilia assembly (Rocha et al, 2014; Gadadhar et al, 2017). In the photoreceptor cilium, that can be considered as a specialized primary cilium, we found that glycylation is restricted to the connecting cilium and the bulge region, where it exactly lines microtubules (−0.6 nm shift relative to α- and β-tubulin staining) (Fig. 1A,F). In contrast, glycylation was mostly absent distally to the bulge and we confirmed that it was also absent from the basal body (Guichard et al, 2023). However, by oversaturating the signal, we found a reproducible faint signal that seems to localize at the level of subdistal appendages of the centriole (Appendix Fig. S1).

We next analyzed tubulin acetylation status in the outer segment. This modification takes place in the lumen of the microtubules (Fig. EV1C), acting on their mechanical properties (Eshun-Wilson et al, 2019). We found that acetylation is present from the basal body to the bulge region of the axonemal microtubules (Fig. 1B,F). However, similar to glycylation, the signal above the bulge was mostly absent or faint, suggesting that distal disorganized axonemal microtubules are less prone to be modified by glycylating or acetylating enzymes.

Next, we examined glutamylation localization with two different antibodies: GT335 raised against a synthetic peptide mimicking the glutamylation modification on a C-terminal tubulin tail; and PolyE, an antibody recognizing long glutamate side chains (>3 glutamates) (Shang et al, 2002; Rogowski et al, 2010) (Fig. 1C,D). For both antibodies, glutamylation staining reveals two localizations along the photoreceptor outer segment. The basal body is decorated with both glutamylation and polyglutamylation with the exact same localization as the tubulin, as previously shown in Chlamydomonas and human centrioles (Mahecic et al, 2020; Gambarotto et al, 2019) (Fig. 1C,D,F). However, the pattern changes drastically once the connecting cilium (CC) starts, with intense glutamylation (GT335 antibody) and polyglutamylation (polyE) signals more external to the microtubule wall and all along the CC, forming a sheath-like structure. We quantified this signal about 60 nm away from the microtubule center of mass (Fig. 1C,D,F), rather close to the membrane, where CEP290 has been recently localized (Mercey et al, 2022) (Fig. 1F). We also quantified the GT335 signal intensity at different locations and confirmed that the CC-associated GT335 signal is more intense compared to that of the centriole or the bulge (Fig. EV1D–F). The discrepancy between the position of tubulin and that of glutamylation suggests that this modification might decorate another protein(s) than tubulin, which are located closer to the membrane. One possibility is that we are also detecting the glutamylation of RPGR as previously observed (Sun et al, 2016) and also localized in this region (Takahashi et al, 2024). This external glutamylation signal is restricted to the connecting cilium, whereas the bulge and the distal axoneme exhibit glutamylation signal at the level of the microtubules, similar to the centriole. Moreover, unlike acetylation or glycylation signals, polyglutamylation seems to propagate more distally in the outer segment axoneme. To get a better impression of the precise localization, we also looked at the transverse view, where we confirmed the presence of a sheath-like glutamylation signal at the connecting cilium (Fig. 1C,D). Additionally, polyglutamylation staining was observed on the tubulin, a signal not detected by the GT335 glutamylation antibody (Fig. 1D). We also noticed that in some cases, PolyE accumulates at the base of the cilium, a signal that resembles IFT train accumulation in U-ExM (Appendix Fig. S2, white arrows) (Van Den Hoek et al, 2022). To further elucidate glutamylation levels in the photoreceptor cilium, we analyzed the GT335 signal at several early stages of outer segment (OS) development, specifically at P10, P14, and P22 (Fig. 1G–I). At P10, the GT335 signal was notably intense at centrioles and was already detectable along the developing OS microtubule axoneme. However, the signal was not continuous from the centriole to the distal cilium, as evidenced by a gap between the centriole and the connecting cilium (CC) (Fig. 1G, white arrow). Interestingly, at P10, the GT335 signal closely overlapped with the tubulin signal in width, but as development progressed to P14 and P22, the GT335 signal became wider and more intense. This suggests that the sheath-like localization observed is not initially present at early timepoints but instead emerges later during OS development (Fig. 1H,I).

Finally, we also assessed the profiles of detyrosinated tubulin and ∆2-tubulin. These two tubulin PTMs consist in the removal of the gene-encoded C-terminal tyrosine (detyr-tubulin), and the further cleavage of the penultimate glutamate residue (∆2-tubulin) of α-tubulin (Fig. EV1C). The staining pattern of detyrosinated tubulin closely mirrored that of glutamylation, and in particular for polyglutamylation as shown in the transverse view at the level of the CC (Fig. 1E). Detyrosinated tubulin is thus localized on microtubules, but also forms a sheath further away from them. Staining for ∆2-tubulin, by contrast, yielded a signal that was not distinct enough to conclude about the precise localization of this PTM (Appendix Fig. S3).

Altogether, these data reveal a complex distribution of tubulin PTMs along the photoreceptor outer segment. We found a specific enrichment at the level of the CC, where all PTMs analyzed are present, hinting at a prominent role of these modifications in this compartment.

Next, we wanted to assess whether the observed distribution of tubulin PTMs is conserved in human retina, as mutations of AGBL5, coding for CCP5, lead to retinitis pigmentosa in human (Fig. 2A–F). To do so, we expanded human retina from a healthy adult. We first noticed that the length of the CC is shorter in human photoreceptor cells as compared to mouse. Using either GT335 or POC5 (Mercey et al, 2022) antibodies as markers of the CC, we demonstrated that the human CC is less than half the length of the one in mouse, spanning about 650 nm vs ~1600 nm in mouse (Fig. 2G; Appendix Fig. S4). We also noticed that cone photoreceptors exhibit particularly long daughter centrioles that are more than 700 nm long, which could represent one of the longest centrioles observed in the human body (Appendix Fig. S4).

We show that glycylation signal (TAP952), while being faint, is present along the human CC and probably on the bulge, similarly to what we described in mouse (Fig. 2A). However, the signal was not strong enough to allow quantification. We also found that acetylation is decorating microtubules all along the OS, from the basal body to the bulge, but also distally, which was less obvious in mouse (Fig. 2B,F). Next, analyzing glutamylation (GT335 and PolyE) and detyrosination signals, we observed the same pattern as in mouse, forming a sheath outside of the microtubules at the level of the CC, and lining the microtubules at the basal body and the cilium above the CC (Fig. 2C–E). The measured diameters revealed the same range of distances to microtubules as in mouse (Figs. 2F and 1F), highlighting the conservation of PTM localization along the photoreceptor outer segment in mouse and human. In line with this, a recent paper using U-ExM in canine photoreceptor cells described a similar sheath localization at the level of the CC using GT335 antibody (Takahashi et al, 2024). In summary, U-ExM unveiled the conservation of PTM distribution along the photoreceptor cell OS but also highlighted structural organization differences between mouse and human photoreceptors cells.

Next, we investigated the physiological relevance of tubulin PTMs for photoreceptor cells. Indeed, as mutation in AGBL5, coding the deglutamylase CCP5, leads to retinitis pigmentosa in humans, it suggests that perturbation of glutamylation could lead to drastic consequences on photoreceptor survival. We thus analyzed retina of mice lacking CCP5 (Ccp5^−/−) and CCP1 (Ccp1^−/−). Absence of these two deglutamylating enzymes is expected to lead to hyperglutamylation (Rogowski et al, 2010). To test whether other PTMs than glutamylation could affect photoreceptor cells, we also assessed the consequence of loss of tubulin acetylation in Atat1^−/− mice.

After expansion, we assessed the integrity of the retina in mice aged from 8 to 18 months old using rhodopsin, a marker of the rod outer segment, as well as α- and β-tubulin staining, in comparison to control mice (Fig. 3A–D). Knockout of the deglutamylases CCP1 or CCP5 lead to a severe photoreceptor degeneration at 8 and 12 months, respectively (Fig. 3B,C). This is highlighted by the substantial decrease of ONL thickness compared to wild type (WT), where only a couple of nuclear rows are remaining (Fig. 3B,C,E). This phenotype was already described in pcd (purkinje cell degeneration) mice, bearing an inactivating mutation in the AGTPBP1 gene (coding for CCP1) (LaVail et al, 1982). At the level of the OS, rhodopsin is also highly impaired compared to the rod-like signal in the WT, suggesting disorganization of membrane discs. We confirmed this result in 7-month-old Ccp5^−/− using Electron Microscopy (EM), where we observed disorganized membrane discs, and a reduction of cell number to about half compared to WT (Fig. EV2B). Additionally, we show a mislocalization of rhodopsin in Ccp5^−/− and Ccp1^−/−, with a signal at the ONL-surrounding nuclei; a hallmark of photoreceptor degeneration (Fig. 3B,C, white arrows and insets). Interestingly, we noticed during dissections that retinas from Ccp5^−/− mice are thinner and more fragile compared to WT (Fig. EV2C). Of note, the other layers of the retina were not affected in these mutants. By contrast, we show that ATAT1 deficiency has no overall impact on photoreceptor survival, even in 18-month-old mice (Fig. 3D). Indeed, the thickness of the ONL is similar to control, with an intact rhodopsin staining (Fig. 3D,E).

We next focused the analysis on the photoreceptor outer segment, where PTMs are generally enriched (Fig. 3F–I). Given that CCP5 deficiency is linked to retinitis pigmentosa, and that we recently showed that FAM161A-associated retinitis pigmentosa RP28 is due to structural defects at the level of the CC (Mercey et al, 2022), we investigated whether this structural element was affected in deglutamylase mutants.

Using tubulin staining to reveal the architecture of the ciliary axoneme, we showed that WT photoreceptor OS have straight axonemal microtubule extending toward the distal part of the cilium (Fig. 3F). We also confirmed the presence of the bulge region, delineating the end of the CC (Figs. 3F and EV1A). By contrast, in the two deglutamylase mutants analyzed, the structure of the OS is impaired, mostly with open or disorganized axoneme on its distal end (Fig. 3G,H, white arrowheads). Interestingly, we found that for Ccp5^−/− and Ccp1^−/− mice, the structural defects were observed mostly above the CC, whereas the tubulin shaft seemed preserved at the CC, as it has been described in mice mutant for the bulge protein LCA5 (Faber et al, 2023) (Fig. 3G,H). Using EM in 10-month-old Ccp5^−/−, we confirmed that despite a severe membrane disc disorganization in the OS, CC seems mostly preserved (Fig. 3J). It should be also noted that more than 80% of the photoreceptor cells analyzed in Ccp5^−/− and Ccp1^−/− mutants are defective, highlighting the high penetrance of the degenerative phenotype in these mutants (Fig. 3K). By comparison, Atat1^−/− photoreceptor cells have no overall axonemal defects in old mice (Fig. 3F,I), demonstrating that lack of tubulin acetylation does not result in obvious alteration at the level of photoreceptor cells.

To gain mechanistic insights into photoreceptor degeneration linked to PTMs and in particular hyperglutamylation, we next assessed the molecular architecture of photoreceptors by U-ExM, with a specific focus on the outer segment (OS), in different PTMs mutants. We analyzed staining for different PTMs (glycylation, glutamylation and acetylation), the CC marker POC5 (Mercey et al, 2022), the bulge marker LCA5 (Faber et al, 2023), and the Intraflagellar Transport (IFT) component IFT88.

Since lack of both CCP1 and CCP5 affect the glutamylation status, we first examined the GT335 pattern along the photoreceptor OS. As expected, we observed a strong hyperglutamylation in Ccp5^−/− where the signal extends towards the distal part of the cilium, thus decorating the whole axoneme (Fig. 4A,G). Remarkably, a similar pattern was observed for acetylated tubulin, where the signal was prolonged distally (Fig. 4B,H). Low magnification images revealed that hyperglutamylation is not restricted to the OS but extends to the inner segment of the photoreceptor cells in Ccp5^−/− (Fig. EV3). Of note, in Ccp1^−/−, axonemes appear shorter than in Ccp5^−/−, probably explaining why hyperglutamylation and hyperacetylation towards the distal axoneme were less pronounced in these mutants. We also noticed that Ccp5^−/−, but not Ccp1^−/− display a strongly reduced glycylation signal along the axoneme, underpinning that hyperglutamylation caused by the loss of some of the deglutamylases could lead to reduced glycylation, as previously described (Grau et al, 2017) (Fig. 4C,I).

In both Ccp1^−/− and Ccp5^−/−, we confirmed a massive disorganization of the microtubules above the CC, with axoneme opening (Fig. 4A–F). Inside the CC, POC5 signal remains unaffected, suggesting that the CC inner scaffold is still present to maintain the cohesion of the microtubules in this compartment (Mercey et al, 2022), and that the observed photoreceptor cell degeneration is not due to structural defects of the CC. However, Ccp5^−/− photoreceptor cells exhibit an exacerbated CC, with the POC5 signal extending to more than 5 µm in some cases (compared to 1.5 µm in the control mice) (Fig. 4D,J). In contrast, CCP1 deficiency did not affect the length of the POC5 signal.

Next, we analyzed the bulge region, by investigating the distribution of LCA5 (Faber et al, 2023). We found that the deficiency of deglutamylases CCP1 or CCP5 leads to a highly reduced and diffused signal of LCA5, suggesting that the bulge region is lost, thus preventing membrane disc formation and presumably causing photoreceptor death (Figs. 4E,K and 3G,H).

As Ccp1^−/− and Ccp5^−/− display distal axoneme disorganization, rhodopsin mislocalization throughout the ONL, and lack of LCA5, we hypothesized that these phenotypes could be linked to intraflagellar transport (IFT) defects. Indeed, we recently showed that the bulge region, marked by LCA5, is crucial to organize IFT at the level of photoreceptor cell, and that the loss of the bulge is associated with defects in the IFT components localization (Faber et al, 2023). In WT, IFT88 accumulates both at the base of the cilium, where trains are formed, and above the CC, at the bulge (Fig. 4F, white arrowheads). By comparison, in both Ccp1^−/− and Ccp5^−/−, the IFT88 signal is greatly reduced above the CC and the cilium entry, highlighting the defects in trafficking towards the photoreceptor outer segment (Fig. 4F,L).

In parallel, we assessed the impact of alpha-tubulin acetyltransferase 1 mutants, which undergo a complete signal loss of acetylation (Appendix Fig. S5). We first show that these mutant mice have no structural defects at the level of the outer segment (Fig. 4A–F), confirming results obtained at the level of the whole retina (Fig. 3I). Moreover, these mutants exhibit no obvious changes in the glutamylation status of the outer segment (Fig. 4A,G). However, we noticed a reduced level of glycylation in Atat1^−/− mutants (Fig. 4C,I). The staining of the CC marker—POC5—revealed no difference in the CC length compared to WT (Fig. 4D,J), suggesting that this structure is intact. In line with this, the bulge marker LCA5 was not impacted by the loss of acetylation in the OS (Fig. 4E,K). Finally, in Atat1^−/− mutants, we show that the dual localization of IFT88 is conserved (Fig. 4F, white arrowheads) and that the intensity of IFT88 staining is similar to the WT at the base of the cilium (Fig. 4L).

Altogether, we demonstrate that defects in glutamylation, but not acetylation, strongly impact the architecture of the OS axoneme, that is associated with transport defects.

To better define the time course of the CCP5-associated retinal degeneration, we analyzed the anatomy of photoreceptor cilia in 3-, 7-, 10- and 12-month-old Ccp5^−/− mice. Using rhodopsin and tubulin staining as readouts for retina integrity, we found that degeneration starts at around 3 months, when ONL is already about 10 µm thinner than in the WT (Fig. 5A,E). ONL thickness progressively decreases in the following months to ultimately reach only 2 or 3 cell layers at 12 months (Fig. 5A–E, white double arrows), highlighting a progressive degeneration over the first year after birth. At 7 months, we noticed a decrease in the ONL thickness to about half of WT, which corresponds to an about 50% decrease in the number of cells observed per field of view with EM (Fig. EV2B). This continuous decrease of the ONL thickness is accompanied by mislocalization of rhodopsin around nuclei, a defect that becomes obvious at 7 months (Fig. 5A–D, white arrows). At the OS level, rhodopsin and tubulin signals reveal the overall absence of defects at 3 months, with membrane discs appearing correctly organized, the bulge region still present, and the distal axoneme straight and properly organized (Fig. 5A). However, from 10 months onwards, we observe a more and more pronounced disorganization of the OS, with opening of axonemes, shorter membrane stacks, and the presence of isolated rhodopsin patches outside the cell (Fig. 5C,D, white arrowheads).

We next analyzed glycylation and glutamylation levels at the OS during the time course of degeneration. Already at 3 months, we find that glycylation signal is strongly reduced concomitantly with highly increased glutamylation (Fig. 5F–I). This tendency further exacerbates over time, with barely any visible glycylation signal at 12 months. At this age, we measured the intensity of GT335 signal in the distal cilium and confirmed that the glutamylation propagates towards the distal part of the cilium in Ccp5^−/− mice (Fig. 5J–L). Interestingly, the external GT335 signal that forms a sheath-like structure at the level of the CC in the WT is lost in Ccp5^−/− mice, with most GT335 signal colocalizing with tubulin (Fig. 5J,K,M,N). This result suggests that rather than tubulin, another glutamylated protein generating this external signal at the CC is lost or relocalized in Ccp5^−/− mice, possibly participating in the collapse of the photoreceptor cell OS. In line with that, we tested RPGR, known to be glutamylated by TTLL5 and recognized by the GT335 antibody (Sun et al, 2016). We stained for this protein in WT and Ccp5^−/− mice (Fig. 5O,P). In the WT, RPGR seems to localize in both the CC and the bulge region. Interestingly, we observed a progressive decrease in RPGR signal in Ccp5^−/− mice, suggesting that the loss of the sheath-like GT335 signal could be associated with the loss of RPGR (Fig. 5Q). This protein being associated with IFT transport at the level of the CC (Hosch et al, 2011), our observation is consistent with the IFT defects observed in Ccp5^−/− photoreceptor cells. Finally, to confirm RPGR localization, we also stained RPGR in human photoreceptors, and observed that the signal clearly localizes at the level of the CC and the proximal part of the bulge, and external to the tubulin, where the GT335 sheath-like is observed (Fig. 5R,S), highlighting the conservation of this pattern.

To obtain complementary insights into the molecular mechanisms involved in CCP5-associated retinitis pigmentosa, we then analyzed CC and IFT markers during the time course of degeneration (Fig. 6A–F). At 3 months, both the CC length (POC5) (Fig. 6A,C) and the IFT enrichment at the basal body (IFT88) (Fig. 6B,D) are comparable to the WT, suggesting that the PTM imbalance (Fig. 5F–I) precedes the structural and functional defects of the OS. From 7 months onwards and gradually increasing up to 12 months, we find that axoneme disorganization is more and more pronounced, with distal axonemal microtubules mostly curled or broken (Figs. 6A,B and 5F,G red and white arrowheads, respectively) whereas CC are mostly preserved. IFT88 intensity is progressively reduced and ultimately absent from 12-month-old photoreceptor cells (Fig. 6B,D). In parallel, the length of the CC gradually increases (Fig. 6A,C). We confirmed this result in 12-month-old Ccp5^−/− mice using another CC marker, CEP290, which has been proposed to form the Y-links bridging microtubule doublets (MTDs) to the membrane periodically along the CC (Zhang et al, 2024; Louvel et al, 2023) (Fig. 6E,F). To assess the integrity of the CC axoneme at the ultrastructure level, we examined electron microscopy images of 7-month-old Ccp5^−/− photoreceptor cells, focusing on transverse sections of CC (Figs. 6G and EV4). In both WT and Ccp5^−/− CC, it was possible to identify the presence of Y-links (Fig. 6G, blue arrowheads). We also confirmed the presence of the inner scaffold (Mercey et al, 2022) (Fig. 6G, orange arrowheads), confirming that CC in Ccp5^−/− photoreceptor cells keep their structural integrity. However, in a few cases, we observed that the B-tubules appear open in mutant CC, a feature never seen in the WT, but this observation might also arise from the low contrast observed in the mutants (Fig. EV4, white arrowheads). Interestingly, electron microscopy images of 10-month-old Ccp5^−/− photoreceptor cells revealed the presence of Y-links, with a periodicity of 37 nm as previously observed (Zhang et al, 2024), on a distal portion of the cilium where the CC should be terminated, corroborating the extended CEP290 signal observed in expansion microscopy in the same mutants (Fig. 6H, black arrowheads).

We finally examined the localization of MAP9, which has recently been described to be associated with MTDs and partly dependent on glutamylation levels (Tran et al, 2024) (Fig. 6I,J). While in the WT, MAP9 signal decorates the CC, the signal is propagated toward the distal part of the cilium in Ccp5^−/− mice, similarly to what we observed for CEP290 and POC5. These results suggest that axonemal ultrastructure in the outer segment could be impaired in Ccp5^−/− mutant mice, with the presence of MTD-associated structures (inner scaffold and Y-links). To confirm this hypothesis, we compared 7-month-old WT and Ccp5^−/− mice using EM (Fig. 6K,L). In the WT OS, microtubules are packed into the membrane incisure cavity, with no round shape organization compared to the CC due to the absence of inner scaffold and Y-links (Fig. 6K, inset). By contrast, in Ccp5^−/− OS, we observed organized MTDs, with the presence of both inner scaffold and Y-links (Fig. 6L, inset), corroborating the extension of the CC observed with POC5 and CEP290 staining (Fig. 6A,E,H). We also noticed that in the mutant, the axoneme is mostly disconnected from the membrane discs, themselves highly impaired.

Altogether, we showed that Ccp5^−/− mice exhibit a slow and progressive degeneration of the photoreceptor cells characterized by the progressive decrease of glycylation and a concomitant increase of hyperglutamylation. At later stages, the signal of ciliary transport proteins is progressively reduced, in parallel with a loss of RPGR and a disorganization of distal axonemal microtubules. The loss of the bulge region accompanied by the exacerbation of the CC is presumably leading to the inability to form new membrane discs, ultimately causing photoreceptor cell death (Fig. 6M).