TTLL1 and TTLL4 polyglutamylases are required for the neurodegenerative phenotypes in pcd mice

Polyglutamylation is a dynamic posttranslational modification where glutamate residues are added to substrate proteins by 8 tubulin tyrosine ligase-like (TTLL) family members (writers) and removed by the 6 member Nna1/CCP family of carboxypeptidases (erasers). Genetic disruption of polyglutamylation leading to hyperglutamylation causes neurodegenerative phenotypes in humans and animal models; the best characterized being the Purkinje cell degeneration (pcd) mouse, a mutant of the gene encoding Nna1/CCP1, the prototypic eraser. Emphasizing the functional importance of the balance between glutamate addition and elimination, loss of TTLL1 prevents Purkinje cell degeneration in pcd. However, whether Ttll1 loss protects other vulnerable neurons in pcd, or if elimination of other TTLLs provides protection is largely unknown. Here using a mouse genetic rescue strategy, we characterized the contribution of Ttll1, 4, 5, 7, or 11 to the degenerative phenotypes in cerebellum, olfactory bulb and retinae of pcd mutants. Ttll1 deficiency attenuates Purkinje cell loss and function and reduces olfactory bulb mitral cell death and retinal photoreceptor degeneration. Moreover, degeneration of photoreceptors in pcd is preceded by impaired rhodopsin trafficking to the rod outer segment and likely represents the causal defect leading to degeneration as this too is rescued by elimination of TTLL1. Although TTLLs have similar catalytic properties on model substrates and several are highly expressed in Purkinje cells (e.g. TTLL5 and 7), besides TTLL1 only TTLL4 deficiency attenuated degeneration of Purkinje and mitral cells in pcd. Additionally, TTLL4 loss partially rescued photoreceptor degeneration and impaired rhodopsin trafficking. Despite their common properties, the polyglutamylation profile changes promoted by TTLL1 and TTLL4 deficiencies in pcd mice are very different. We also report that loss of anabolic TTLL5 synergizes with loss of catabolic Nna1/CCP1 to promote photoreceptor degeneration. Finally, male infertility in pcd is not rescued by loss of any Ttll. These data provide insight into the complexity of polyglutamate homeostasis and function in vivo and potential routes to ameliorate disorders caused by disrupted polyglutamylation.


Introduction
Protein polyglutamylation and de-glutamylation, which are catalyzed, respectively, by tubulin tyrosine ligase-like (TTLL) and cytosolic carboxypeptidase (CCP) enzymes, contribute to diverse biological processes and their mutation underlie rare neurodegenerative disorders. Polyglutamylation consists of the successive enzymatic addition of glutamate molecules to an acceptor glutamic acid within a substrate protein. During polyglutamate chain initiation, the first glutamate is added to the gamma-carboxyl group of a glutamic acid residue. Subsequently, polymer elongation occurs through glutamate addition to the alpha-carboxyl group of this initiating glutamic acid, and continues through iterative alpha linked additions [1]. This is a dynamic process where glutamate polymerization is catalyzed by several TTLL family members, whereas degradation of the polymer is catalyzed by the 6-member CCP family [1][2][3]. A range of proteins undergo polyglutamylation [4], with tubulin being the first identified and most extensively studied [1,5,6]. Indeed, the substrate specificities of TTLLs and CCPs were mainly determined with tubulin as substrate [1][2][3]7,8]. TTLL4 and TTLL5 catalyze the formation of initiating γ-carboxyl linkages with preferences for β-and α-tubulin, respectively. TTLL1 and 11 catalyze the formation of α-carboxyl chain-elongating linkages primarily in αtubulin. TTLL7 catalyzes both initiation and elongation of the polyglutamate side chain of βtubulin [1,9]. CCP5 uniquely catalyzes removal of the gamma-carboxyl linked branch point glutamate of tubulins [2,10]. The other 5 CCP family members all cleave α-carboxyl linked glutamate residues on tubulin [2,3] although they can be distinguished to some extent by their kinetic properties and synthetic substrate preferences in vitro [10,11] and their non-redundant function in vivo [11][12][13][14].
Despite its discovery 30 years ago, only relatively recently has a link been established between disrupted polyglutamylation and neurodegeneration in vivo. This emanated from the recognition that the zinc metallocarboxypeptidase, Nna1 (a.k.a. CCP1) [15], which is mutated in the recessive Purkinje cell degeneration (pcd) mouse [16] is a deglutamylase capable of degrading polyglutamate side chains [2,7,8]. In addition to male infertility, pcd mice are characterized by progressive and selective neurodegeneration of cerebellar Purkinje cells, olfactory bulb mitral cells, thalamic neurons, spinal motor neurons and retinal photoreceptors [17,18]. In parallel, tubulin polyglutamylation levels are increased in the central nervous system and testis of pcd mice [2,7,10]. Mutations of Agtpbp1, the gene encoding Nna1/CCP1, have also been shown to cause lower motor neuron degeneration in sheep [19] and in humans cause an infantile-onset, progressive, and severe neurodegeneration [18,20,21].
A picture is emerging of complex relationships amongst the anabolic and catabolic enzymes involved in polyglutamylation, their substrates and neurodegenerative phenotypes. Elevated tubulin glutamylation, as observed in pcd mice, is associated with neurodegeneration. However, the converse is not true as loss of TTLL1 function results in reduced glutamylation but no neurodegeneration [6] and targeted over-expression of Nna1, CCP4 or CCP6 in Purkinje cells, which degrades polyglutamate chains does not result in their degeneration [11]. These data imply it is longer, rather than shorter, polyglutamate chains that are deleterious to neurons. The role of the balance between anabolism and catabolism of glutamate chains in neurodegeneration is exemplified by the finding that loss of TTLL1 function, which reduces polyglutamylation, rescues Purkinje cell degeneration in pcd mice [7,22].
Many TTLLs are expressed in brain, some at levels similar to, or higher than TTLL1 [5,23,24]. This begs the question of whether additional TTLLs also contribute to neurodegeneration in pcd mice. For example, as TTLL1 mediates elongation of glutamate chains on αtubulin it might be anticipated that loss of an initiator TTLL for this tubulin isoform such as TTLL5 might produce a similar outcome. Here we test null alleles of different TTLLs for their ability to modify tubulin polyglutamylation and rescue degeneration of retinal photoreceptors, olfactory bulb mitral cells and cerebellar Purkinje neurons as well as testicular structure and function in pcd mice.

Polyglutamate homeostasis and Ttll expression patterns
Glutamylation is a complex balance between glutamate addition by the TTLL family of enzymes and their elimination by the CCP family of carboxypeptidases ( [1,2,7,8,14] and Fig 1A). Thus, glutamylation is dependent both on the levels of the various enzymes in particular cells and tissues, their catalytic properties (e.g. initiators or elongators) and preferences for different substrates (e.g. α-vs β-tubulin). The glutamylation homeostatic process is depicted in Fig 1A for αand β-tubulins. It should be borne in mind that substrate preferences have been largely established in vitro and less is known about their enzymatic properties and specificities in vivo.
To determine which TTLL glutamylases might contribute to the pcd phenotype, we examined their expression by quantitative RT-PCR in cerebellum, olfactory bulb and testis which have abnormalities in pcd mice [25] as well as cerebral cortex, kidney and liver that have no reported deficiencies [17,25] (Fig 1B). In general, expression of Ttlls was highest in testis followed by the three brain regions and lowest levels were observed in kidney and liver. Of note, Agtpbp1, the gene encoding Nna1/CCP1 and is mutated in the pcd mouse, is highly expressed in brain compared to non-neural tissues and is even present at high levels in cerebral cortex which has no reported degeneration.
Loss of Ttll1 or Ttll4, but not Ttll5, 7, or11, attenuates Purkinje cell degeneration in pcd mice TTLL1,2,4,5,6,7,9,11 and 13 are proven glutamylases whereas TTLL3, 8 and 10 are glycylases and TTLL12 has no known enzymatic activity [1,26,27]. Of the authenticated glutamylases, expression of Ttll2, 6 and 13 are at the limit of detection in neural tissues and Ttll9 has relatively low expression ( Fig 1B). Reviewing the Allen Mouse Brain Atlas and published literature [5,24,28] Ttll1, 4, 5, 7, and 11 are expressed at various levels in cerebellar Purkinje cells and mitral cells of the olfactory bulb that degenerate in pcd mice whereas Ttll9 has marginal or no expression in either neuronal type and no null strain was available to us. Therefore, we focused on Ttll1, 4, 5, 7 and 11 and crossed null alleles of each (S1 Fig) onto a pcd background. Absence of transcripts in all null strains was confirmed by RT-PCR of total RNA from various tissues (Figs 2A, 2B and S2).
All strains of mice were first tested for locomotor performance on an accelerating rota-rod and subsequently sacrificed and survival of Purkinje cells and levels of polyglutamylation in cerebellum determined by immunofluorescence and western blotting, respectively. In pcd mice, only loss of Ttll1 or Ttll4 restored locomotor activity to levels comparable to wild-type mice (Fig 2C and 2D). No significant improvement of rota-rod performance was seen with null alleles of Ttll5, 7 or 11 in pcd mice (S2 Fig).
Degeneration of Purkinje cells in pcd begins around day 17 after birth and progresses over several months and affects some lobes before others [17,25]. Purkinje cell integrity was assessed using immunofluorescence with an anti-calbindin antibody (Fig 3). Low power images reveal that cerebellum of pcd ( Fig 3B) is smaller than that of wild-type ( Fig 3A) or Initiator and elongator TTLLs differentially prefer α-tubulin (gray) or β-tubulin (brown) [1,5]. CCP5 uniquely cleaves the branching point glutamate, whereas CCP1, 2, 3, 4 and 6 degrade the α-carboxyl linked glutamates in the chain. (B) Expression profiles for Ttlls and Agtpbp1 (gene encoding Nna1/CCP1). Expression was determined using qRT-PCR on total RNA from adult cerebellum (CB), cerebral cortex (CX), olfactory bulb (OB), testis (TS), kidney (KD), and liver (LV). Data represents mean and SEM of transcript copy numbers/ng total RNA from 3 independent mice. Authentic polyglutamylases are indicated in black whereas Ttll3, 8 and 10 (blue) are glycylases [1]and Ttll12 (red) is of unknown activity [1]. https://doi.org/10.1371/journal.pgen.1010144.g001 Ttll1KO,pcd ( Fig 3C) or Ttll4KO,pcd mice ( Fig 3D). In addition, there is widespread loss of Purkinje cells in 2-month-old pcd mice (Fig 3B and 3B'). In contrast, many calbindin-positive Purkinje cells are evident throughout all cerebellar lobes examined in Ttll1KO,pcd (Fig 3C and  3C')  That loss of TTLL1 rescues Purkinje cell death in pcd was known [7,22], but loss of TTLL4 protecting these cells is a novel finding. As Purkinje cell loss is progressive in pcd, we considered the possibility that TTLL4 loss might slow but not prevent their death. Therefore, we compared 10-month-old wild type and Ttll4KO,pcd mice ( Fig 4A). The size of cerebellum in Ttll4KO,pcd mice (Fig 4Ab) is not different from wild-type of the same age (Fig 4Aa). Again, many calbindin-positive Purkinje cells were observed in all lobes examined (Fig 4Ab and  4Ad), indicating that the rescue is relatively long term.
We also wanted to compare the degree of Purkinje cell protection conferred by TTLL1-or TTLL4-deficiency. Anatomical reconstruction of the cerebellum would be a daunting undertaking, so we sought a quantitative orthologous method. Previously, we identified many markers of Purkinje cells using molecular profiling [29]. Therefore, we examined RNA expression of three such markers [30][31][32] as a surrogate for Purkinje cell numbers in 2 months old mice. Three marker genes were chosen to minimize the possibility that expression of a single marker Ttll4 transcripts in cerebellum, brain, eye, and testis of respective KO mice. (C and D) Gender balanced littermates of each genotype (n = 4-10/genotype) at 7 weeks of age were tested on a standardized accelerating rota-rod for five consecutive days. The latency to fall in seconds (mean ± SEM) was not significantly different between WT and Ttll1KO,pcd (C) or Ttll4KO,pcd (D) groups (one-way ANOVA p >0.05), whereas pcd mice were markedly impaired (C and D) (p < 0.05).
https://doi.org/10.1371/journal.pgen.1010144.g002 might be directly affected by glutamylation. Expression of all three markers was almost undetectable in pcd cerebellum (Fig 4B), reflecting the massive loss of Purkinje cells. In contrast expression of the three markers was statistically the same in wild-type (Fig 4Ba and 4Bb), Ttll1KO,pcd (Fig 4Ba) and Ttll4KO,pcd cerebellum (Fig 4Bb). This indicates that the degree of Purkinje cell rescue by loss of Ttll1 in pcd is indistinguishable from that conferred by loss of Ttll4 at this age.

Loss of TTLL1 or TTLL4 variably affect tubulin polyglutamylation in pcd cerebellum
In pcd, tubulin is hyperglutamylated [2] and loss of TTLL1 is reported to reduce glutamylation levels in cerebellum in parallel with its ability to rescue Purkinje cell degeneration [7]. Therefore, we compared tubulin glutamylation levels in cerebellum by immunoblotting of cerebellar extracts using the GT335 (detects gamma-linked branching glutamate [33]) and polyE (detects 3 or more consecutive glutamate residues residing at the C-terminus of a chain [1,6,34]) antibodies. In agreement with previous studies [2,7], tubulin glutamylation is increased in pcd mouse cerebellum and is markedly decreased in Ttll1KO and Ttll1KO,pcd cerebellum ( Fig 5A). In contrast, loss of TTLL4 has no marked effect on tubulin polyglutamylation either alone or in pcd mice ( Fig 5B). We also quantified the levels of glutamylation and confirm the reduction in Ttll1KO animals and the lack of statistically significant difference between pcd and Ttll4KO,pcd mice (Fig 5A and 5B). Nevertheless, the current analysis does demonstrate that TTLL4 loss of function attenuates Purkinje cell loss up to at least 10-months of age. In addition, we show that although loss of TTLL7 does not spare Purkinje cells in pcd mice, it does reduce polyglutamylation signal (S2C Fig). During revision of this manuscript a study appeared that also demonstrated reduced tubulin polyglutamylation but failure to rescue Purkinje cells in Ttll7KO,pcd mice [35]. These authors also showed that while TTLL1 is considered an α-tubulin preferring enzyme, it can glutamylate β-tubulin in the presence of TTLL7 [35].
A caveat with immunoblotting is that it is a bulk assay in which Purkinje cells are a relatively minor component and it is possible that there are changes in glutamylation in Purkinje cells that are masked by signal from other cellular components. To address this, we performed immunostaining of cerebellum with the GT335 antibody. As Purkinje cells degenerate in adult pcd, we first examined GT335 staining in postnatal day 19 (P19) mice, prior to Purkinje cell loss. In wild type mice Purkinje cells had prominent GT335 staining, whereas the granule cell layer was only weakly stained ( Fig 6A). In pcd and Ttll4KO,pcd mice, there was no obvious difference in the GT335 staining of Purkinje cells compared to wild-type although the granule cell layer was more prominently stained in both (Fig 6B and 6C). To establish specificity of staining we absorbed the antibody with porcine brain tubulin and saw a marked reduction in immunofluorescence (S4 Fig).
In adult wild-type mice, Purkinje cells have prominent GT335-immunoreactivity particularly in the cell soma and principle dendrites ( Fig 6D). In contrast, there are low levels of GT335 immunoreactivity in the granule cell layer. In adult pcd mice, Purkinje cells are absent but the granule cell layer is more intensely stained than in wild-type mice ( Fig 6E). In adult Ttll4KO,pcd mice, Purkinje cells are preserved and are GT335-immunoreactive ( Fig 6F), although they are not noticeably different from the wild-type. However, the granule cell layer is still intensely GT335-positive ( Fig 6F). This suggests the increase in GT355 immunoreactivity on western blots (most likely α-tubulin, based on comparison of its migration with other immunoblotting studies in pcd [2]), represents the contribution from granule neurons. Furthermore, our data indicate that Ttll4 loss does not affect the glutamylation of α-tubulin in granule neurons. In sum, loss of Nna1/CCP1 elicits increased tubulin glutamylation in granule neurons and is independent of TTLL4. However, there are no marked changes in GT335 immunoreactivity in Purkinje cells.

TTLL1 and TTLL4 also contribute to degeneration of olfactory bulb mitral cells in pcd mice
Other neuronal populations degenerate in pcd mice [17,25] and we next determined whether other TTLLs and especially TTLL1 or TTLL4 play any role in this process. Mitral cells of the olfactory bulb degenerate in pcd mice [36] albeit on a slower time course than Purkinje cells [17]. Mitral cells are output neurons central to odorant perception whose dendrites form synapses with olfactory nerve axons in structures called glomeruli and whose axons project via the lateral olfactory tract to the olfactory cortex [37]. We assessed mitral cell survival using Tbr2 immunostaining as a marker (Fig 7) [38]. In 5-month-old wild type mice, Tbr2-positive cells are present in two layers ( (A) Representative western blot of polyglutamylated tubulins in cerebellar lysates from wild-type (WT), pcd, Ttll1KO, and Ttll1KO,pcd mice detected using GT335 and polyE antibodies. In pcd, there is increased immunoreactivity to both GT335 and polyE antibodies compared to wild-type. However, GT335-and polyE-immunoreactive bands were greatly reduced in Ttll1KO and Ttll1KO,pcd. (B) In contrast to Ttll1KO, loss of Ttll4 function did not alter the basal or elevated polyglutamylation levels in pcd. (C and D) Quantitative analysis of intensity of GT335 or polyE immunoreactive bands using LiCOR with normalization to α-tubulin levels. Whereas Ttll1KO caused a significant reduction in GT335 and polyE signals, loss of Ttll4 had no significant affect. The bars represent the mean ±SEM of 4-6 animals of each genotype. �� p < 0.01; ��� p < 0.001; NS: Not significant. To determine whether loss of TTLL4 function slows rather than prevents mitral cell degeneration, we assessed survival in 7-and 10-month-old wild-type and Ttll4KO,pcd mice. As in younger animals, two bands of Tbr2-positive cells are present throughout the olfactory bulbs of both wild-type and Ttll4KO,pcd mice (Fig 8A and 8B), and many Tbr2-positive cells are present in the MCL of both genotypes at 7 and 10 months (Fig 8). Therefore, as with Purkinje

TTLL1 and TTLL4 differentially contribute to photoreceptor degeneration in pcd mice
We next examined whether knocking out any Ttll attenuated photoreceptor degeneration in pcd mice. Retinae of 5-month-old mice were immunostained for the rod-specific rhodopsin, and nuclei visualized by DAPI. In wild-type mice, the outer nuclear layer (ONL), which comprises the cell bodies of photoreceptors, is 10-12 nuclei deep (~51 μm) and rhodopsin immunoreactivity is restricted to the outer segment layer (OSL) (Fig 9Aa and 9Ae). In pcd, the ONL is much thinner (~15 μm) comprising 2-3 nuclei (Fig 9Ab), reflecting substantial photoreceptor loss. In addition, in pcd retinae the OSL is thinner and rhodopsin immunoreactivity is observed not only in the OSL (Fig 9Ab) but also in the ONL (Fig 9Af). The mis-location of rhodopsin is considered a pathological finding indicative of impaired trafficking of proteins from the photoreceptor cell body to the outer segment via the specialized connecting cilium [39,40]. In Ttll1KO,pcd mice, the thickness of the ONL is similar to wild-type values (~10-12 nuclei,~45 μm), and the length of the OSL is also comparable to that of wild-type (Fig 9Ac) and there is no mis-location of rhodopsin in the cell body (Fig 9Ag). Therefore, at 5-month of age the photoreceptor degeneration in pcd mice is largely prevented by Ttll1 loss-of-function. In the retinae of Ttll4KO,pcd double mutants, the ONL was about 5-7 nuclei in depth (~31μm), which is more than that of pcd but less than wild-type (Fig 9Ad). The OSL in Ttll4KO,pcd retina is also substantially broader than pcd (Fig 9Ad). However, some retention of rhodopsin in cell bodies was evident (Fig 9Ah). Therefore, TTLL4-deficiency partially attenuated or slowed photoreceptor degeneration in pcd mice.
Rhodopsin retention in the cell body is a sign of aberrant protein transport but may also be the secondary consequence of outer segment degeneration. Therefore, we examined the localization of rhodopsin in retina at age P30, when retina differentiation and maturation are just complete and prior to photoreceptor loss [17,41]. At P30, the depth of the ONL was comparable between the wild-type and pcd retinae (Fig 9Ba and 9Bb). However, rhodopsin was already evident in the cell body of pcd photoreceptors (Fig 9Bf). These results suggest that rhodopsin transport is directly impaired by Nna1 dysfunction in pcd mice. In Ttll1KO,pcd mice no mislocalized rhodopsin was evident (Fig 9Bg). On the other hand, mis-localized rhodopsin was evident in Ttll4KO,pcd retina, although it was less prominent than in pcd mice (Fig 9Bh). Therefore, the aberrant localization of rhodopsin observed in Ttll4KO,pcd mice at 5-month of age is already evident at P30 and confirm eliminating Ttll4 does not fully rescue the deficit in rhodopsin transport in pcd mice.
Besides TTLL1 and TTLL4 loss only TTLL5-deficiency had any influence on photoreceptor degeneration in pcd (Fig 9C). Mutations of TTLL5 have been linked to retinal disease in humans [42] and a Ttll5 mutant mouse is reported to undergo late (~20 months-old) photoreceptor degeneration [43]. At 5-month of age, we did not observe overt loss of photoreceptors in Ttll5KO mice. However, Ttll5KO,pcd mice lose almost all photoreceptors and only a single layer of cell bodies survives in the ONL (Fig 9Ca). Thus, loss of anabolic TTLL5 synergizes with loss of catabolic Nna1 to promote photoreceptor degeneration.

Neither Ttll1KO nor Ttll4KO rescues male infertility in pcd mice
The pcd male is infertile due to defects in spermatogenesis associated with testicular structural anomalies and cell death in the germinal epithelium [10,44]. TTLL1 (the gene targeted here) is the catalytic subunit of a multiprotein complex [26] and a mutation in another component of the complex (PGs1) behaves as a TTLL1-loss of function allele [24,45]. The PGs1mutant also exhibits male infertility and has testicular anomalies [45]. Progeny testing showed that the Ttll1KO males used here were also infertile. Cell death in the germinal epithelium was characterized by the presence of pyknotic and/or multinucleated giant cells, the latter being a specific form of germ cell degeneration caused by the fusion of spermatids (Fig 10, compare panel Aa with panels Ab, Ac, Dd and Ff; arrows indicate dying cells). Given that depletion of Ttll1 or Ttll4 ameliorated neuronal degeneration in pcd mice, we assessed whether knocking out Ttll1 or Ttll4 could also rescue male infertility.
Hematoxylin-eosin staining revealed a striking cell loss in the germinal epithelium of Ttll1KO testes (Fig 10 Ac and 10Ac's, arrows indicate dying cells). These morphologic changes were more severe than those reported for PGs1 mutants [45] and even those of pcd (Fig 10Ab  and 10Ab'). The Ttll1KO,pcd testis, although still abnormal, had substantially improved overall morphology and reduced cell loss in the germinal epithelium (Fig 10Ad and 10Ad') compared to either Ttll1KO or pcd. As expected, the sperm count of Ttll1KO mice was significantly lower than that of the wild-type ( Fig 10B). However, despite the improved testicular cytoarchitecture, the sperm count from Ttll1KO,pcd double mutants was not significantly different from that of either Ttll1KO or pcd (Fig 10B).
In contrast to Ttll1 mutants, progeny testing showed that male (and female) Ttll4KO mice were fertile. Indeed, the morphology of the testes and sperm counts from Ttll4KO mice were comparable with that of wild-type littermates (Fig 10Ae, 10Ae', 10Aa, 10Aa' and 10B). The testicular structure in Ttll4KO,pcd double mutants (Fig 10Af and 10Af') was improved compared with pcd mice, but substantial numbers of dead cells were still evident (Fig 10Af' arrows). Despite the improved testicular morphology, the sperm count from Ttll4KO,pcd was still as low as that of pcd mice (Fig 10B).

Discussion
A growing body of work has demonstrated that appropriately regulated polyglutamylation is required for neuronal survival across species [2,18,19] and also contributes to axonal regeneration [15,46] and spermatogenesis [10,27,45,47]. In this study we investigated the contribution of anabolic TTLL enzymes to polyglutamate homeostasis and degenerative phenotypes in pcd mice that are deficient in the catabolic Nna1/CCP1 enzyme [2,16]. We show that TTLL1 and TTLL4 are unique among the polyglutamylases as their loss counteracts to varying degrees the degenerative phenotypes in cerebellum, olfactory bulb and retinae of pcd mice (this study and [7,22]); providing evidence that the degeneration of different types of neurons in pcd mice likely share a common molecular pathogenesis involving hyperglutamylation.
An important question is the identity of the relevant substrate(s) for TTLL1 and TTLL4. The preference of TTLL1 for α-tubulin and TTLL4 for β-tubulin suggests that hyperglutamylation of both tubulin subunits is required for degeneration in pcd. However, this notion raises several questions. TTLL4 and TTLL7 are both expressed in Purkinje cells [5,28] and both initiate glutamylation of β-tubulin yet loss of TTLL7 does not counteract Purkinje cell degeneration in pcd (this study and [35]). Moreover, loss of TTLL7 reduces tubulin glutamylation in pcd (this study and [35]), whereas TTLL4 loss has no discernible effect either in bulk western blots or immunofluorescence analyses (Figs 5 and 6). This questions whether hyperglutamylation of β-tubulin contributes to the pcd phenotype and indeed whether tubulins are the relevant substrate for TTLL4 in this model. Non-tubulin substrates for polyglutamylation have been identified [4,43,48] but whether any of these contribute to degeneration is unknown. A similar specificity paradox exists for TTLL1 and TTLL5 which are both expressed in Purkinje cells [28]and prefer α-tubulin [1,26], yet only TTLL1 loss prevents degeneration (Fig 3). There are caveats for interpreting these data. First, substrate specificities and catalytic properties of TTLLs should be more accurately viewed as preferences that were established largely in vitro [1,26] and with much less being known about their selectivity in vivo in specific cell types. Indeed, a recent study found that TTLL1, which is generally considered an elongator for αtubulin, may function as an elongator on beta-tubulin in vivo [22,35]. Second, as proposed in the tubulin code, the various tubulin isoforms may be glutamylated at multiple sites [49,50] which could have distinct functional outcomes depending upon the pattern of residues glutamylated the lengths of the chains and the status of other tubulin post-translational modifications such as glycylation [51]. A further possibility is that critical glutamylation changes are confined to specific regions of the cell. To assess these parameters in vivo with single cell and sub-cellular specificity is challenging with the reagents and technologies currently available. Therefore, the conclusions that hyperglutamylation of β-tubulin is pathogenic and that it is the critical target substrate for TTLL4 in vivo requires further investigation.
If there is a common underlying pathogenic mechanism involving hyperglutamylation several observations remain to be explained. First, Nna1/CCP1 is broadly expressed in CNS neurons [16] and its elimination in pcd results in hyperglutamylation of tubulin throughout brain, yet only specific subsets of neurons degenerate [17]. Second, the tempo of degeneration of different cell classes varies, with some dying early, such as Purkinje cells whereas others, such as olfactory bulb mitral cells degenerate on an intermediate time scale and yet others, such as retinal photoreceptors, die progressively throughout a more protracted timeline [17,25]. How and whether these properties reflect differences in the balance of polyglutamate homeostasis (e.g. levels of various TTLLs and CCPs) or levels and subcellular distributions of different tubulin isoforms or other substrates in various cell classes remains to be fully established. Our systematic genetic analysis of TTLL dependencies in pcd mice provides perspectives on these issues.
Purkinje cells and mitral cells have relatively high levels of both TTLL1 and TTLL4 and both TTLLs contribute to degeneration in pcd. Therefore, intrinsic differences in the expression of TTLL1 and TTLL4 in distinct neuronal populations may be one determinant of their relative susceptibility to damage in response to perturbations that promote hyperglutamylation. Such intrinsic differences in TTLL1 and TTLL4 expression could also underlie the tempo of neuronal degeneration. Experiments are underway to test this hypothesis.
The degeneration of retinal photoreceptors highlights additional characteristics of the involvement of polyglutamylation in this process. Retinal photoreceptors appear particularly sensitive to changes in glutamylation status as their degeneration occurs not only in pcd, but also in mice or humans deficient in CCP5 [52,53] or TTLL5 [42,43] whose losses have no overt effects on other neurons [43]. In this instance sensitivity may relate to the specialized cellular architecture of photoreceptors. The rod outer segment of photoreceptors has no intrinsic translational capacity and proteins necessary for outer segment structure and function must be transported from the inner segment via a specialized connecting cilium [40]. The correct structure and function of the connecting cilium is essential for photoreceptor integrity and mutations in genes encoding its constituent proteins underlie several retinal diseases [39]. Furthermore, polyglutamylated tubulin is known to be associated with the axoneme of the connecting cilium [51] as well as cilia in general, centrioles, basal bodies, and flagella [54] and play a role in the proper formation of ciliated structures and trafficking along the cilia [55][56][57]. Indeed, hyperglutamylation of tubulin is associated with altered trafficking of vesicles and organelles in neurons of pcd mice although a causal relationship to degeneration has yet to be established [22]. The early-onset mis-localization of rhodopsin to the photoreceptor cell body is regarded as an indicator of impaired trafficking through the connecting cilium [40] and is a prominent feature of the pcd retina both prior to and during receptor degeneration (Fig 9 and  [51]). Our finding that TTLL1 deficiency not only attenuated photoreceptor loss but also eliminated the mis-localization of rhodopsin prior to receptor degeneration in pcd indicates it restores normal levels of transport through the connecting cilium and that impaired transport is the underlying cause of the degeneration. This notion is further supported by our finding that TTLL4-deficiency does not provide the same level of attenuation of photoreceptor loss as TTLL1 loss and only partially improves rhodopsin mis-localization (Fig 9). Given the ongoing impairment of rhodopsin trafficking in pcd,Ttll4 KO mice it is probable that its loss of function slows rather than prevents photoreceptor degeneration in pcd mice. It also indicates that if TTLL1 and TTLL4 act in a common pathway the two are not equipotent, at least in the retina.
Mutations of the X-linked retinitis pigmentosa GTPase regulator (RPGR) cause photoreceptor degeneration [58]. Furthermore, a specific isoform of RPGR is a substrate for TTLL5 [43] and mutations in TTLL5 cause retinal dystrophy in humans [42,59] and slow photoreceptor degeneration in mice leading to the proposal that glutamylation of RPGR is required for its normal function [43]. We did not observe any noticeable photoreceptor loss in 5-month-old TTLL5-null mice, which is unsurprising as the reported loss was in mice of~20-months of age [43]. However, 5-month-old Ttll5KO,pcd double mutants had a more severe photoreceptor loss compared to pcd alone (Fig 9C). Together this highlights two points: first, hyperglutamylation of one substrate can act cooperatively with hypoglutamylation of a different substrate to elicit photoreceptor degeneration; second, the tempo of degeneration is slower for hypoglutamylation compared to hyperglutamylation.
In testis, all enzymes involved in polyglutamylation are highly expressed [5] (Fig 1), potentially reflecting the importance of tubulin polyglutamylation in, for example, sperm flagellum formation. Despite their abundance, only the anabolic TTLL1, TTLL5, and TTLL9 [27,45,47], and the catabolic Nna1/CCP1 and CCP5 [3,10,44] play key roles in spermatogenesis. In this study, Ttll1 deficiency caused the most striking defects in testicular structure and morphology, worse even than pcd (Fig 10Ab and 10Ac). The abnormal morphology was accompanied by very low sperm counts (Fig 10B). In Ttll1KO,pcd double mutants, testicular morphology was improved compared to that of either single KO (compare Fig 10Ad with Fig 10Ab and 10Ac), however the marked deficit in sperm numbers was not rescued (Fig 10B). Therefore, TTLL1 and Nna1/CCP1 counteract each other in some but not all aspects of testicular function. In contrast to Ttll1KO mice, Ttll4KO males are fertile with no obvious testicular anomalies and normal sperm counts (Fig 10Ae and 10B). Nevertheless, knocking out Ttll4 also rescued much of the aberrant testicular morphology in pcd (Fig 10Ab) but failed to rescue sperm counts ( Fig  10B). This suggests that TTLL4 also generates substrates for Nna1/CCP1, but its function is dispensable in testis. Thus, the products of multiple TTLL polyglutamylases concomitantly contribute to male infertility in pcd mice.
Tubulin polyglutamylation levels in testis are generally low [10]. During male gametogenesis, exquisite regulation of tubulin posttranslational modification is required in multiple events such as spindle cleavage in mitosis and meiosis, sperm flagellum formation, trafficking in germ and Sertoli cells, and manchette formation and dismantlement [60]. It is conceivable that TTLL activity may be required in a transient manner during different events of spermatogenesis. This might be achieved through transcriptional regulation of their expression and potentially allosteric mechanisms as well as by the levels and activity of CCPs that metabolize their substrates. In contrast to markedly impaired male reproductive capacity, female reproduction is essentially normal in all TtllKO strains examined. This is despite the fact that tubulin posttranslational modification plays a role during oocyte maturation [60] and motility of cilia in the Fallopian tube epithelium is required for female reproduction [61].
This study identified the functional relationships between TTLL polyglutamylases and Nna1 in the nervous and male reproductive systems and revealed that only TTLL1 and TTLL4 generate the Nna1 substrates essential for neuronal degeneration. Further characterization of the substrates may provide insight into whether disrupted polyglutamylation plays a broader role in neurodegeneration and male infertility and present therapeutic opportunities.

Generation and genotyping of TtllKO and TtllKO,pcd double mutants
The heterozygotes of each Ttll were bred with pcd heterozygotes to generate Ttll,pcd double heterozygotes, which were further inbred to obtain individual TtllKO,pcd double knock-out mice, wild-type littermates, and other intermediate genotypes. Genotyping of pcd was done as described previously [16,29]. Genotyping the wild-type and TtllKO alleles was performed by PCR using the primers listed in S1 Table. Genotyping was also confirmed at the RNA level using RT-PCR on total RNA extracted from cerebellum, cerebral cortex, testis, or eye. First strand cDNA was generated using Superscript III kit and amplified with primers that target the deleted region of the respective KO alleles (primers listed in S2 Table). β-actin was amplified as a loading control using primers described previously [11].

Histology and immunohistochemistry
The procedures for histological analyses of cerebella, retina, olfactory bulbs, and testes were conducted as described [44,63,64]. A rabbit anti-calbindin D-28K antibody (Chemicon, Temecula, CA, USA) and a rabbit anti-Tbr2 antibody (Abcam, Cambridge, MA, USA) were used at the dilution of 1:500 to visualize Purkinje cells and olfactory bulb mitral cells, respectively and immune complexes were revealed using a peroxidase-conjugated anti-rabbit kit and diaminobenzidine tetrahydrochloride (DAB) substrate (Vector Labs, Burlingame, CA, USA). After immunostaining, sections were counterstained with hematoxylin (Sigma-Aldrich, St Louis, MO, USA).
Retina. For 5-month old mice, eyes were dissected, fixed in 4% paraformaldehyde, and embedded in paraffin after sequential dehydration. Immunofluorescence was performed on 5 μm sections. For postnatal day 30 (P30) mice, the retinae were peeled off from eyes and fixed in 4% paraformaldehyde before embedding in OCT embedding medium (Sakura Finetek, Torrance, CA, USA). Immunofluorescence was performed on 12 μm sections. A mouse anti-rhodopsin antibody [41] was used at the dilution of 1:500 to determine the localization of rhodopsin, which is visualized using Alexa Fluor 488 goat anti-mouse IgG (1:500). Images were taken with a Zeiss LSM 710 NLO Confocal Microscope or an Olympus BX60 equipped with DP71 camera. Sections from retina that included the optic nerve were chosen for comparison.

Microarray hybridization and analysis
Cerebella of 2-month-old mice were collected and RNA was isolated according to the manufacturers' protocol (TRIzol, ThermoFisher). RNA quality was confirmed by analysis on the Agilent 2100 Bioanalyzer. Total RNA was processed in Hartwell Center for Bioinformatics & Biotechnology at St. Jude Children's Research Hospital according to the Affymetrix GeneChip eukaryote two-cycle target labeling protocol. Biotin-labeled cRNA was added to a hybridization cocktail and then processed automatically on Mouse Genome 2.0_2 or Mouse Clariom S array. Normalized transcript measures were generated from scan intensity files using the RMA algorithm. The array data was quantitated for three Purkinje cell enriched transcripts (Pcp2, Ppp1r17 and Car8) that were identified previously [29] in wild-type; pcd; Ttll1KO; Ttll4KO; Ttll1KO,pcd, and Ttll4KO,pcd cerebella.

Sperm count
Sperm from epididymis and vas deferens were counted using the protocol described previously [65] with minor modifications by including sperm from vas deferens in counting.

Rota-rod test
To assess motor coordination, balance, and motor learning, gender-and age-matched littermate mice were tested on a rota-rod (San Diego Instruments, San Diego, CA) with an accelerating speed (0 to 40 rpm in 4 min and then hold constant speed for an additional min) as described previously [11,63], and the latency of the mice falling from the rod was scored as an index of their motor coordination. When both locomotor coordination and motor learning were assessed, the test was conducted for 5 consecutive days.

Statistical methods
The latencies to fall in the rota-rod test were expressed as mean ± SEM (in seconds) and were analyzed for statistical significance using One-way analysis of variance (ANOVA) with repeated measures followed by Bonferroni's multiple comparison test or Student's t-test for comparison between samples at the same time point. The level of significance was set at p < 0.05. In all other experiments, Student's t test was used to compare independent samples for statistical significance. Significance was set at p of < 0.05. Student's t test was performed using Microsoft Excel software.