Intimate functional interactions between TGS1 and the Smn complex revealed by an analysis of the Drosophila eye development

Trimethylguanosine synthase 1 (TGS1) is a conserved enzyme that mediates formation of the trimethylguanosine cap on several RNAs, including snRNAs and telomerase RNA. Previous studies have shown that TGS1 binds the Survival Motor Neuron (SMN) protein, whose deficiency causes spinal muscular atrophy (SMA). Here, we analyzed the roles of the Drosophila orthologs of the human TGS1 and SMN genes. We show that the Drosophila TGS1 protein (dTgs1) physically interacts with all subunits of the Drosophila Smn complex (Smn, Gem2, Gem3, Gem4 and Gem5), and that a human TGS1 transgene rescues the mutant phenotype caused by dTgs1 loss. We demonstrate that both dTgs1 and Smn are required for viability of retinal progenitor cells and that downregulation of these genes leads to a reduced eye size. Importantly, overexpression of dTgs1 partially rescues the eye defects caused by Smn depletion, and vice versa. These results suggest that the Drosophila eye model can be exploited for screens aimed at the identification of genes and drugs that modify the phenotypes elicited by Tgs1 and Smn deficiency. These modifiers could help to understand the molecular mechanisms underlying SMA pathogenesis and devise new therapies for this genetic disease.

Studies in human cells have defined the role of TGS1 in maturation and trafficking of small RNAs. In human cells, there are two TGS1 isoforms, a long isoform (TGS1-LF) that contains the methyltransferase domain of the enzyme, and a short (TGS1-SF) isoform that consists only of the C-terminus of the protein. TGS1-LF is present in both the cytoplasm and the nuclear Cajal bodies (CBs) and regulates trafficking of both snoRNAs and snRNAs; TGS1-SF is restricted to the CBs where it specifically interacts with snoRNAs [1]. In the nucleus, the monomethylated 5' cap of snRNAs binds the cap-binding complex (CBC) that mediates their export to the cytoplasm through an interaction with the CRM1 and PHAX export factors [11]. Once in the cytoplasm, snRNAs associate with the Sm protein complex that physically binds TGS1 through its SmB component [2,12]. The seven Sm core proteins assemble into a heteroheptameric donut-shaped multiprotein structure that binds the U1, U2, U4 and U5 snRNAs, forming four of the five snRNP subunits of the major spliceosome [13]. The assembly of the Sm-snRNA particles is chaperoned by the survival of motor neurons (SMN) complex, which includes SMN, Gemin2-8 and Unrip/STRAP [14]. Following the snRNA interaction with the Sm and SMN complexes, TGS1 hypermethylates the MMG cap of snRNAs, and the TMG-snRNPs are reimported into the nucleus [1,2]. TGS1 has also been implicated in the regulation of the telomerase RNA moiety. In S. cerevisiae, TGS1catalyzes TMG cap formation on TLC1 (the RNA component of S. cerevisiae telomerase), and its loss causes telomere lengthening and an increase in telomere silencing [15]. TGS1 also mediates TER1 (the RNA of S. pombe telomerase) hypermethylation in S. pombe. However, loss of TGS1 in this yeast affects TER1 processing and stability, resulting in telomere shortening [16]. We have recently found that TGS1 hypermethylates human telomerase RNA (hTR) and that TGS1 deficiency increases both the mature hTR level and telomerase activity, leading to telomere elongation [4]. Thus, while TGS1 mediates the formation of a TMG cap on telomerase RNA in both yeasts ad humans, the effects of this post-transcriptional modification on telomerase regulation appear to be species-specific.
In humans, an impairment of SMN function causes Spinal Muscular Atrophy (SMA), a devastating recessive disorder characterized by motor neuron loss, progressive paralysis and death [17]. The human genome harbors two SMN genes, but SMN2 does not produce a sufficient amount of protein to compensate for homozygous SMN1 mutations found in SMA patients. Although the exact mechanisms through which SMN deficiency disrupts motor neuron function have not been fully elucidated, one of the most accredited hypotheses is that loss of SMN reduces the snRNP levels in neural cells, resulting in splicing defects in mRNAs with critical roles in motor neuron function and maintenance [18][19][20][21][22]. SMN targets include mRNAs encoding two negative regulators of the abundance of p53 [23], a key driver of motor neuron death [24]. Another SMN target, Stasimon, plays a dual role by preserving the function of the sensory-motor circuit and by restricting phosphorylation-mediated p53 activation [25]. SMN also plays splicing-independent functions that are thought to contribute to the SMA phenotype. These functions include regulation of axonal localization and transport of mRNPs [26][27][28][29], proper development and regeneration of the neuromuscular junction [30], neural stem cell division and differentiation [31,32], integrity and function of skeletal muscle [33], and prevention of transcriptional stress and DNA damage [34][35][36].
The Drosophila TGS1 gene is part of a bicistronic locus that also includes modigliani (moi), a gene that encodes a telomeric protein required to prevent telomere fusion [8,9]. Drosophila TGS1 (henceforth dTgs1) is an essential gene [8,9] that mediates snRNA hypermethylation [7] and interacts with the Gemin3 subunit of the SMN complex in the two-hybrid assay [37]. It has been reported that dTgs1 depletion affects fly motor behavior, eliciting phenotypes reminiscent of those observed in Smn-deficient animals [8,37]. Here, we show that dTgs1 physically interacts with all subunits of the Drosophila Smn complex, and that a human TGS1 transgene fully rescues the lethal phenotype of dTgs1 null mutants. In addition, we demonstrate that both dTgs1 and Smn are required for Drosophila eye development, and that the two genes cooperate to ensure viability of retinal progenitor cells. Importantly, we show that overexpression of dTgs1 partially rescues the eye defects caused by Smn deficiency, while Smn overexpression ameliorates the eye phenotype elicited by dTgs1 depletion. Thus our work establishes a new Drosophila model that can be exploited for a variety of functional studies on both dTgs1 and Smn.

Analysis of the Drosophila moi-dTgs1 bicistronic locus
dTgs1 is part of a bicistronic locus that also includes moi, which encodes a protein required for telomere capping (Fig 1A). Previous work based on the analysis of rescue experiments of mutations and deficiencies involving the moi-dTgs1 bicistronic locus has suggested that the two genes play independent functions [8,9]. To provide a definitive proof for this independency we generated several moi and dTgs1 mutations using CRISPR/Cas9 mutagenesis. The moi mutation we characterized (moi 2 ) contains a frameshift mutation resulting in an early stop codon (Fig 1A and 1B).
Homozygous moi 2 animals are lethal, dying in third instar larval/early pupal stages; they exhibit~3 telomeric fusions per larval brain cell (n = 200) and are fully rescued by a moi+ transgene constitutively driven by the tubulin promoter ( Fig 1B). The two CRISPR/Cas9-induced Tgs1 mutations (dTgs1 R1 and dTgs1 R2 ) result in early stop codons in the dTgs1 sequence ( Fig 1B). To characterize these mutations we generated a polyclonal antibody against a C-terminal fragment of dTgs1 (See Materials and Methods). Western blotting analysis showed that (B) Transgenic constructs and mutant combinations used in this study. For each combination is reported the viability or the lethal stage dTgs1 R1 and dTgs1 R2 homozygotes exhibit a strongly reduced dTgs1 band compared to wild type ( Fig 1C). Homozygotes for each of these mutations die as second instar larvae but are viable in the presence of either a ubiquitously expressed dTgs1 FL transgene including both moi + and dTgs1 + or an inducible UAS-GFP-dTgs1 transgene driven by Actin-Gal4 (Fig 1B). In contrast, dTgs1 R1 and dTgs1 R2 homozygotes, constitutively expressing the wild type moi transgene (moi + ), still die as second instar larvae ( Fig 1B). Importantly, the moi 2 /dTgs1 R1 and moi 2 / dTgs1 R2 heterozygous flies are fully viable, indicating that the protein products of the moi and dTgs1 genes are functionally independent. Collectively, these genetic and cytological analyses indicate that dTgs1 is an essential Drosophila gene that functions independently of moi. Previous work indicated that overexpression of a UAS-Tgs1-3XHA transgene, driven by the ubiquitous daughterless-Gal4 driver, is lethal [37]. In contrast, we found that overexpression of UAS-GFP-dTgs1 driven by the Actin-Gal4 ubiquitous driver is fully viable (Fig 1B). Also viable are flies carrying either one or two copies of the dTgs1 FL construct (tubulin-moi-dTgs1) ( Fig  1B). We do not understand the reason for this discrepancy, which might depend on the type of UAS construct used, the driver, or both.

dTgs1 physically interacts with the Smn complex
Previous work in human cells has shown that snRNAs bind the Sm proteins in the cytoplasm, with the assistance of the SMN complex that acts as a molecular chaperone [38]. Following the interaction with the Sm and SMN complexes, the snRNAs are hypermethylated by TGS1 and reimported into the nucleus [1,2,39]. During this process TGS1 directly interacts with the SMN protein and the SmB component of the Sm complex [12]. These findings prompted us to investigate whether dTgs1 interacts with the fly Smn complex. To identify the dTgs1 interacting partners we incubated extracts from 0-3 hour old embryos expressing GFP-dTgs1 with GFP-TRAP-A, and subjected them to affinity purification-mass spectrometry (AP-MS). The fly line used for this experiment carried a UAS-GFP-dTgs1 transgene on one of its second chromosomes and an Actin-GAL4 driver on the other, and was homozygous for the null mutation dTgs1 R2 , so that the flies of this line express GFP-tagged dTgs1 but not the endogenous dTgs1 protein. GFP-dTgs1 was efficiently purified (Fig 2A) and was the most abundant protein in precipitates (Fig 2B and 2C, S1 Table). To select bona fide dTgs1-binding partners we applied stringent filtering against a database of non-specific interactors as described by [40]; we selected only proteins IDs that were not found in the negative control and with a score above 50, and ranked them by mean area. Using this criterion, we identified 11 specific dTgs1 interactors that include Smn and four Gemins: Gem2, Gem3 and the protein products of the duplicated genes Gem4a (Glos or CG2941) and Gem4b (CG32786) [41,42] (Fig 2B and 2C). Peptides corresponding to Gem5 were also detected in precipitates, but with an overall score well below our cut-off ( Fig 2C; S1 Table). Thus, dTgs1 co-purifies with most components of the Drosophila Smn complex.
To confirm and extend these results we also performed AP-MS from 0-3 hours embryos expressing Smn-GFP. For this experiment we used embryos from mothers carrying two wild type copies of the Smn gene and homozygous for an Smn-GFP transgene placed under the control of the tubulin promoter [43]. Using the same criterion described above for dTgs1, Smn was the most abundant protein in precipitates, along with 11 specific interactors, again (L2, lethal in second larval instar; L3, lethal in third larval instar; P lethal in the pupal stage), and the moi and dTgs1 proficiency (+, normal; +/-reduced;-virtually null). (C) Western blots of extracts from second instar larvae homozygous for either the dTgs1 R1 or the dTgs1 R2 mutation, showing a strong reduction in dTgs1 compared to wild type larvae at the same developmental stage (CTR). The dTgs1 protein in mutant larvae is likely to be maternally supplied. Actin is a loading control. https://doi.org/10.1371/journal.pgen.1008815.g001

PLOS GENETICS
including Gem2, Gem3, Gem4a and Gem4b (Fig 2B and 2D, S2 Table). In addition, Smn-GFP co-purified Lsm11, the Cap binding protein (Cbp80) and dTgs1 (Fig 2D). Gem5 and Lsm10 were also present in precipitates but fell just outside of our stringent cut-off criteria ( Fig 2D). These findings are in good agreement with recent AP-MS results obtained from Drosophila embryos expressing Smn-FLAG [41]. Collectively, our results provide strong evidence that dTgs1 physically associates with the Drosophila Smn complex in vivo, suggesting a parallel functional interaction (Fig 2B, 2C and 2D).

Mutations in dTgs1 cause neurological phenotypes
Previous studies have shown that loss-of-function mutations in the Drosophila Smn gene result in a variety of phenotypes, including alterations in the sensory-motor neuronal network, abnormal neuromuscular junctions, and defective locomotion [44][45][46][47][48][49]. In addition, we have recently shown that Smn depletion in neurons results in unexpanded wings and unretracted ptilinum [43]. The ptilinum is a head muscle required to break open the operculum of the puparium, and is normally retracted after fly eclosion. The post-eclosion events are regulated by the bursicon neuropeptide and a specific set of neurons [50,51], suggesting that the wing expansion/ptilinum phenotype could be a consequence of improper functioning of the underlying neural circuit.
The finding that dTgs1 interacts with all Drosophila Smn components prompted us to ask whether its loss results in a phenotype comparable to that observed after Smn depletion. To analyze the role of dTgs1 in Drosophila we took advantage of moi 1 , a point mutation within the Drosophila moi-dTgs1 bicistronic locus that affects the function of both moi and Tgs1 (Fig 1A  and 1B). moi 1 homozygotes are viable in the presence of a moi + transgene, and both in the presence and in the absence of this transgene, they exhibit a substantially lower level of dTgs1 (~50%) compared to moi 1 /+ heterozygotes or moi 1 homozygotes bearing the dTgs1 FL transgene ( Fig 3A). Thus the moi 1 mutation is also hypomorphic for dTgs1 and will be henceforth designated as dTgs1 m1 . We also exploited a rescue construct, dTgs1 CD , encoding the entire moi-dTgs1 sequence but carrying point mutations within the dTgs1 methyltransferase domain [9] (Fig 1A and 1B). dTgs1 CD rescued the lethality of dTgs1 m1 homozygotes ( Fig 1B), but did not complement the dTgs1 R1 and dTgs1 R2 null mutations, indicating that the dTgs1 protein encoded by this construct is not functional.
Adult flies homozygous for dTgs1 m1 and bearing either the moi + or dTgs1 CD transgene showed a high proportion (20% and 16%, respectively) of individuals displaying defects in wing expansion and ptilinum retraction (Fig 3B and 3C). These defects were strongly reduced in flies homozygous for dTgs1 m1 but bearing a dTgs1 FL transgene ( Fig 3C). Thus, the wing expansion and the ptilinum phenotypes are caused by an impairment of the dTgs1 function. We also investigated whether dTgs1 mutant larvae exhibit locomotion defects, measured as frequency in contraction rates (peristalses). We found that dTgs1 m1 homozygous larvae bearing a moi + transgene exhibit a reduction in the frequency of peristalses compared to both dTgs1 m1 / + heterozygotes and dTgs1 m1 homozygotes bearing the dTgs1 FL rescue construct ( Fig 3D).

Tissue-specific silencing of dTgs1
To further explore the role of dTgs1 we used flies bearing the transgenic construct UAS-dTgs1 RNAi (abbreviated as dTgs1-RNAi), already described by [37]. To check the efficiency of this RNAi construct we determined the dTgs1 protein level in larval brains carrying (i) dTgs1-RNAi and the Actin-Gal4 driver in a wild type background and (ii) dTgs1-RNAi and Actin-Gal4 in a dTgs1 m1 /dTgs1 + background; these brains showed 60% and 70% reduction in dTgs1 compared to control brains carrying dTgs1-RNAi and no driver, respectively ( Fig 4A). The dTgs1 reduction was even stronger (~80%) when RNAi against dTgs1 was performed in flies heterozygous for the dTgs1 R1 mutation (Fig 4B). To assess the role of dTgs1in different tissues, we crossed dTgs1-RNAi-bearing flies with flies carrying various Gal4 drivers with different tissue specificities ( Fig 4C). The dTgs1-RNAi transgene in the presence of either the ubiquitously expressed Actin-Gal4 driver or the how24B-Gal4 mesodermal driver (targeting muscles) induced lethality in third instar larval/early pupal stages ( Fig 4C). The pan-neuronal drivers elav155-Gal4 and nsyb-Gal4 did not induce lethality or visible morphological phenotypes (Fig 4C), such as wing expansion defects, which were instead observed when a UAS-Smn RNAi construct is expressed by the same drivers [43]. The results of these RNAi experiments are in line with those previously reported by [37].

A human TGS1 transgene rescues the phenotypes elicited by mutations in Drosophila Tgs1
We next asked whether, and to what extent, a human TGS1 gene (henceforth designated as hTGS1) has the ability to functionally substitute for its Drosophila homolog. We generated a fly line bearing a wild type human TGS1 gene fused with the UAS promoter (UAS-hTGS1), and a second line carrying a similar construct in which the human TGS1 gene was mutated in the catalytic site (UAS-hTGS1 CD ; Fig 4D). To generate hTGS1 CD , we substituted the residues S763 and Y771 within the highly conserved Motif IV of the hTGS1 methyltransferase domain [52] (Fig 4D). This amino acid stretch is important for substrate binding and identifies the TGS1 catalytic center. Specific mutations in the S and W residues of this motif suppress the TGS1 catalytic activity in yeast, humans and Drosophila [5,9,52]. Western blotting analysis showed that UAS-hTGS1 and UAS-hTGS1 CD are expressed at similar levels in larval brains in the presence of an Actin-Gal4 driver ( Fig 4E). We then generated dTgs1 R1 and dTgs1 R2

PLOS GENETICS
homozygotes bearing the Actin-Gal4 driver and either UAS-hTGS1 or UAS-hTGS1 CD . The wild type hTGS1 transgene, but not hTGS1 CD , rescued the lethality of both dTgs1 mutants ( Fig 1B).
We next used the human TGS1 transgene (which is not targeted by our Drosophila RNAi construct) to ask whether dTgs1-RNAi affects the expression of both dTgs1 and moi. We constructed Actin-Gal4>dTgs1-RNAi and how24B-Gal4>dTgs1-RNAi flies (see Fig 4C) bearing

PLOS GENETICS
UAS-hTGS1 and found that the expression of the human transgene fully rescues the lethality caused by dTgs1 depletion. We also analyzed brain cells from actin-Gal4>dTgs1-RNAi larvae heterozygous for the dTgs1 R1 mutation. Although the brains of these larvae displayed an 80% reduction in the Tgs1 level (Fig 4B), they did not show telomeric fusions (200 cells examined). Collectively, these results confirm that dTgs1 plays a moi-independent function. Most importantly, they demonstrate a remarkable evolutionary conservation of the TGS1 function, and that the dTgs1 catalytic site is essential for fly viability.

The roles of dTgs1 and Smn in eye development
We next examined the roles of dTgs1 and Smn in the morphogenesis of the Drosophila compound eye, a remarkably patterned structure highly suitable for the analysis of cell viability and proliferation. To ascertain a possible role of dTgs1 in the eye development we exploited the dTgs1-RNAi construct and the eyeless-Gal4 driver (ey-Gal4), which efficiently drives expression of UAS-RNAi constructs in the eye imaginal disc [53]. eyeless is an essential Drosophila Pax6 gene that specifically promotes growth of the eye imaginal disc and is required for eye progenitor cell survival and proliferation [54,55].
When we expressed the dTgs1-RNAi construct under the control of the ey-Gal4 driver, we found that 92% of the eyes (n = 1,280) from these RNAi flies displayed a reduction in size with defects ranging from total eye absence to severe or mild eye reduction. We subdivided the eyes into 5 classes, ranging from class 0 characterized by the complete absence of the eye to class 100 that comprises essentially normal eyes; the sizes of the eyes in classes 25, 50, and 75 were < 25%, < 50%, and < 75% the size of wild type eyes, respectively (see Fig 5A and Materials and Methods for details on eye size estimation). To assess the specificity of the effects of dTgs1 downregulation on eye development we generated flies bearing the ey-Gal4 driver, dTgs1-RNAi and either the UAS-hTGS1 or the UAS-hTGS1 CD construct. We found that hTGS1 expression rescues the eye phenotype (n = 1,698 eyes examined), abrogating the 0-50 size classes and inducing a substantial increase in the proportion of eyes with a regular size (from 8% to 53%) (Fig 5B). In contrast, the expression of hTGS1 CD did not mitigate the eye phenotype (n = 344 eyes) of ey-Gal4>Tgs1 RNAi flies (Fig 5B). These results strongly suggest that the defective eye development observed in dTgs1 RNAi flies is specifically due to loss of the dTgs1 hypermethylase activity.
To assess the role of Smn in eye development we generated ey-Gal4> Smn-RNAi flies, using a previously characterized UAS-Smn-RNAi construct [43]. When this construct is expressed under the control of the Actin-Gal4 driver, RNAi larvae die in the third instar stage, and the abundance of the Smn protein in larval brains is reduced to 30% of the wild type level [43]. We found that ey-Gal4 >Smn-RNAi-induced Smn downregulation in flies heterozygous for the Smn X7 deficiency (that removes only the Smn gene) strongly affects eye development. Examination of 1,232 eyes from these Smn RNAi flies did not reveal cases of complete eye absence, but 82% of the eyes were substantially reduced compared to wild type (12% class 25, 24% class 50 and 46% class 75) (Fig 5C). Interestingly, some of the eyes we included in class 100 appeared slightly deformed and larger than the wild type eyes. In addition, 20% of the normal-sized eyes had an irregular surface and showed protruding areas.
In both ey-Gal4>Tgs1-RNAi and ey-Gal4>Smn-RNAi flies displaying a reduction of the eye also the antennae were strongly affected. We observed a variety of defects ranging from an abnormal number of antennae (usually 1 or 3) to the apparent transformation of the eye tissue into head epidermis or antennal tissue. The degree of eye-to-antenna transformation ranged from some antennal tissue protruding from the eye to the complete transformation of an eye into an antenna (Fig 5D). Similar homeotic fate transformation phenotypes have been . Class 0%, eye absent; classes < 0.25%, < 0.50% and < 0.75%, eyes with sizes below 25%, 50% and 75% of the wild-type eye size (class100%). (B) Eye sizes in flies expressing UAS-dTgs1 RNAi and the indicated transgenes driven by ey-Gal4 (ey>). The histogram bars represent the percentages of eyes falling into each size class shown in (A). The ey>dTgs1RNAi>hTGS1 and previously described in mutants causing cell proliferation defects in the eye imaginal disc and when eye specification genes such as ey, twin of eyeless (toy), eyes absent (eya), sine oculis (so) or dachshund (dac) are downregulated or misregulated [56][57][58].
To substantiate our data on the eye and antennal phenotypes we analyzed the eye-antennal imaginal discs from ey-Gal4>Tgs1-RNAi and ey-Gal4>Smn-RNAi third instar larvae. Fixed imaginal discs were stained with both anti-Caspase and anti Elav antibodies and counterstained for DNA with DAPI. Elav staining identifies the photoreceptor neural cells placed posteriorly to the morphogenetic furrow [59], and Caspase-3 forms brightly fluorescent aggregates in correspondence to the apoptotic cells [60] (Fig 6). We found that the eye imaginal discs of ey-Gal4>Tgs1-RNAi third instar larvae are significantly smaller (Fig 6A and 6B) and exhibit a reduced proportion of Elav positive tissue compared to their wild type counterparts (Fig 6A and 6C). In addition, these discs displayed an area enriched in apoptotic cells associated with bright caspase signals (Fig 6A and 6D; see methods for quantitation of apoptotic signals); this area was located anteriorly to the morphogenetic furrow and involved retinal progenitor cells not stained by anti-Elav antibodies. We did not observe caspase signals in wild type discs. This is consistent with previous work [61] showing that apoptosis in the wild type eye disc does not occur in third instar larvae but only later in mid-pupal stages to eliminate the excess of pigment cells [62,63].
The imaginal discs from ey-Gal4>Smn-RNAi larvae heterozygous for the Smn X7 deficiency showed defects that correlate with the eye phenotype observed in these animals. Indeed, the discs of these larvae showed a greater variability in size compared to the wild type discs ( Fig  6A and 6B). Smn RNAi disc sizes had a normal distribution like that of wild type discs, but many of them had sizes that were either 2 standard deviations (SDs) below (11%) or above (19%) the mean size of control discs (Fig 6A and 6B). The Smn RNAi discs showed a modest but significant reduction in the proportion of Elav positive tissue compared to wild type discs (Fig 6A and 6C). They displayed areas with retinal progenitor cells undergoing apoptosis and did not show apoptotic signals associated with Elav-stained cells (Fig 6A and 6D). In summary, Tgs1 or Smn knockdown in the eye imaginal disc results in apoptosis of retinal progenitor cells.

Functional interactions between dTgs1 and Smn in Drosophila eye development
To further explore the functional relationships between dTgs1 and Smn we first performed a phenotypic analysis of flies with eye-targeted dTgs1 silencing simultaneously overexpressing Smn. We generated ey-Gal4>dTgs1-RNAi flies bearing a ubiquitously expressed Smn-FLAG construct under the control of the tubulin promoter (tub-Smn) described by [43]. A comparison of these flies (n = 1,248 eyes) with the ey-Gal4>Tgs1-RNAi flies (n = 1,280 eyes), or the ey-Gal4>Tgs1-RNAi flies expressing hTGS1 CD (n = 344 eyes), showed that the expression of Smn-ey>dTgs1RNAi> hTGS1 CD carry a dTgs1RNAi construct and either the UAS-hTGS1 or the UAS-hTGS1 CD transgene, all driven by ey-Gal4; the ey>dTgs1 RNAi; tub-Smn flies express an Smn-FLAG transgene under control of the ubiquitous tubulin promoter. � , p< 0.05; �� , p< 0.01; ��� , p< 0.001; ns, not significant; two-way ANOVA. Note that the eye defects elicited by dTgs1 RNAi are partially rescued by hTGS1 but not by hTGS1 CD , and that the dTgs1 mutant phenotype is ameliorated by Smn overexpression. (C) Eye sizes in flies carrying the ey-Gal4 driver and either the UAS-SmnRNAi construct only (ey>SmnRNAi) or both UAS-SmnRNAi and UAS-GFP-dTgs1. The histogram bars represent the percentages of eyes falling into each size class shown in (A). �� p< 0,01, ��� p< 0,001, ns, not significant; two-way ANOVA. Note that the eye defects elicited by Smn RNAi are partially rescued by overexpression of dTgs1. FLAG substantially ameliorates their eye phenotype, reducing the percentage of very small eyes (classes 0 and 25) from 37% to 16% and increasing class 100 from 8% to 27% (Fig 5B).
We next analyzed the effect of dTgs1 overexpression in Smn deficient-eyes. We constructed ey-Gal4>Smn-RNAi flies heterozygous for the Smn x7 deficiency and compared their eye phenotype (n = 616 eyes) with that of flies of the same genetic constitution but bearing a UAS-GFP-dTgs1 construct (and thus overexpressing dTgs1) (n = 374 eyes). We found that overexpression of dTgs1 significantly counteracts the eye developmental defects caused by Smn silencing, reducing the 25 and 50 eye classes from 12% to 0% and from 25% to 7%, respectively, and increasing class 100 from 17% to 44% (Fig 5C).
Finally, we analyzed flies bearing the ey-Gal4 driver and both the Tgs1-RNAi and the Smn-RNAi constructs. 98% (n = 600) of these flies died at the pupal stage, and the rare surviving adults displayed severe eye defects. Interestingly, 90% of the ey-GAL4>dTgs1-RNAi>Smn-RNAi late lethal pupae manually extracted from the puparium showed a failure in development of the head structures (Fig 5E). This phenotype is not surprising, as headless flies have been previously observed in mutants in genes that initiate both eye specification and proliferation such as the two fly Pax6 genes ey and toy [55,64]. Since we used an ey-Gal4 driver to express the Tgs1-RNAi and the Smn-RNAi constructs, it is likely that simultaneous downregulation of these genes disrupts eye progenitor cell proliferation, leading to a headless phenotype. Thus, these results indicate that dTgs1 and Smn cooperate in the pathways involved in eye and head development.

The evolutionary conservation and the functional role of Tgs1
Drosophila dTgs1 is part of a bicistronic locus that also includes modigliani (moi), a gene required to prevent telomeric fusions (TFs). Previous complementation analyses with suitable transgenes suggested that the two genes play independent functions [8,9]. Here, we have provided a series of new data that support this conclusion. Specifically, we have used the CRISPR/ Cas9 technology to introduce early stop codons in the dTgs1 and moi coding sequences, and shown that these lethal mutations fully complement for viability and fertility. In addition, we have shown that the long isoform of the human TGS1 transgene (hTGS1) fully rescues the lethality of flies homozygous for dTgs1 null mutations. Notably, hTGS1 also rescues the lethality caused by RNAi-mediated depletion of dTgs1, and rescued larvae do not exhibit TFs in brains. This result not only provides further support to the functional independence of moi and dTgs1, but clearly shows that RNAi against dTgs1 does not disrupt the moi mRNA, validating the use of RNAi for specific impairment of the dTgs1 function.
Previous work has shown that TGS1 has high degree of functional conservation. The eukaryotic TGS1 proteins have different sizes and different degrees of global homology. However their C-terminal parts, which contain the methyltransferase catalytic domain, are structurally and functionally similar [52]. For example, loss of S. cerevisiae TGS1 is complemented by a wild type human TGS1 gene but not by a TGS1 variant with mutations in the catalytic site [5]. Similarly, the growth inhibition phenotype caused by mutations in S. cerevisiae TGS1 is rescued by a wild type Arabidopsis TGS1 gene but not by a gene carrying mutations in the methyltransferase catalytic domain [6]. In line with these results, we have demonstrated that (A). The number of discs analyzed, the mean disc area and descriptive statistics are reported below the graph. (C) Box plots showing the quantification of the Elav-labeled area in the discs described in (B), expressed as a percentage of the total disc area. The lines inside the boxes indicate the median; box boundaries represent the first and third quartiles; whiskers are 1.5 interquartile range. Descriptive statistics is reported below the graph. (D) Scatter dot-plots showing the quantification of the number of Caspase-3 positive foci per disc, in the discs described in (B). The mean area of the Caspase-3 foci and descriptive statistics are reported below the graph. https://doi.org/10.1371/journal.pgen.1008815.g006

PLOS GENETICS
dTgs1 and hTGS1 carrying mutations in the catalytic domain (dTgs1 CD and hTGS1 CD ) are unable to rescue the phenotypic consequences of dTgs1 deficiency. These results show that a human TGS1 gene can substitute for its Drosophila ortholog. They also indicate that the lethality of dTgs1 mutant and dTgs1 RNAi flies, and the defective eye development elicited by targeted dTgs1 depletion, are specifically due to loss of the dTgs1 hypermethylase activity.

The biochemical relationships between Tgs1 and Smn
Previous studies on yeast and human cells have shown that TGS1 directly binds the SMN protein and associates with the SmB subunit of the Sm complex. It has been proposed that SMN recruits TGS1 favoring its engagement with the m7G-capped snRNA bound to the Sm ring, so as to allow hypermethylation of the cap [2,12]. The TMG cap and the Sm ring are thought to act as a bipartite nuclear-localization signal that mediates nuclear import of the snRNP particles [13,[65][66][67]. Work in Drosophila has shown that dTgs1 interacts with the Gemin3 subunit of the SMN complex in a two-hybrid assay [37]. Interestingly, the fly SMN complex contains only four bona fide Gemin proteins (Gem2, Gem3, Gem4/Glos and Gem5/Rig) [41,42,68]. The existence of potential homologs of Gem6, 7 and 8 postulated by Lanfranco et al. [42] was not confirmed in a subsequent study [41]. Our AP/MS analyses of embryo extracts showed that the 11 most abundant dTgs1-interacting proteins include Smn and four Gemins (Gem2, Gem3, Gem4a and Gem4b). Conversely, the reciprocal AP-MS, using Smn-GFP as a bait, detected Gem3, Gem2, Gem4a and Gem4b, Lsm11, the cap binding protein Cbp80 and dTgs1, all within the top 11 interactors. Although Gem5 and other Sm complex subunits, such as SmD2 and SmD3, were not enriched in either AP-MS, they were present in the precipitates (S1 and S2 Tables), and therefore a specific biochemical association between these proteins and Tgs1 or Smn cannot be ruled out. Noteworthy, a physical interaction between TGS1 and SmB has previously been demonstrated in both human and yeast cells [12]. Moreover, consistent with our AP/MS results, Gem5 was the less abundant precipitate among the Gemins that co-purify with Smn-FLAG [41]. Collectively, our findings reveal a strong physical association between dTgs1 and the Smn complex subunits that has not been detected in previous studies on any organism. This suggests an intimate functional relationship between dTgs1 and Smn.

The functional relationships between Tgs1 and Smn
We have shown that homozygotes for dTgs1 null mutations die in the second larval instar, and that hypomorphic mutations in dTgs1 are defective in wing expansion and ptilinum retraction, a phenotype that likely reflects a perturbation in the neural circuit that coordinates the posteclosion performance [50,51]. We also found that third instar larvae homozygous for dTgs1 m1 hypomorphic mutations exhibit a reduction in the frequency of peristalses compared to controls. These results are consistent with previous studies showing that dTgs1 mutant or RNAi flies exhibit abnormal larval and adult locomotor behaviors [8,37].
Locomotion phenotypes similar to those observed in dTgs1 mutants have been previously seen in flies carrying mutations in Smn or expressing RNAi constructs against Smn. The unexpanded wing phenotype has been also observed in flies where Smn was specifically silenced in neurons using RNAi [43]. Smn mutant larvae are defective in the sensory-motor neuronal network and exhibit reduced muscle growth and defective locomotion [44,48,49,69]. These phenotypes have been attributed to reduced biogenesis of snRNAs leading to defective splicing of a subset U12 intron-containing RNAs, altering the expression of genes required for motor circuit function such as Stasimon [20]. Other studies suggested that splicing defects in other RNA types such as those encoded by genes involved in stress response could contribute to the fly neurological phenotype [21,70]. Studies in Drosophila have also shown that the Smn function is not restricted to motor neurons. For example, Smn is required for stem cell division and differentiation [31], and for maintenance of proper organization of nuclear compartments in both nurse cells and oocytes [71]. Whether these functions of Smn depend on its role in snRNP biogenesis and splicing is currently unclear.
Here, we have documented another Smn function not involved in motor neuron maintenance. We have shown that both Smn and dTgs1 are required for Drosophila eye development. eyeless-Gal4-driven RNAi against each of these genes resulted in high frequencies of flies with reduced eye sizes. RNAi against dTgs1 produced small eye discs containing a reduced proportion of Elav-positive differentiated cells compared to wild type. The eye discs of Smn RNAi flies were greatly variable in size; some of them were bigger and other smaller than their wild type counterparts. They also showed a modest but significant reduction of the proportion of Elavpositive cells. Both dTgs1 and Smn RNAi discs displayed frequent apoptotic cells in the anterior disc area that contain undifferentiated retinal progenitor cells but not in the posterior area containing Elav-stained cells. These results suggest that both dTgs1 and Smn are required to prevent cell damage leading to apoptosis of retinal progenitor cells. When either gene is downregulated, precursor cells die leading to a reduction in the differentiated retinal cells and to an overall reduction of the eye size. The few large discs and eyes observed after Smn RNAi might be the consequence of compensatory proliferation or apoptosis-induced proliferation of imaginal cells, two phenomena that compensate for disc cell loss or damage [72,73 and references therein].
We have also demonstrated that Smn and dTgs1 interact genetically in the control of eye disc development. Overexpression of Smn substantially ameliorates the eye phenotype caused by dTgs1 depletion, while dTgs1 overexpression partially rescues the eye defects elicited by Smn downregulation. In addition, double RNAi against both dTgs1 and Smn exacerbates the phenotypes observed in single RNAi flies, leading to severe eye defects and frequent headless individuals unable to emerge from the pupal case. These results suggest that dTgs1 and Smn cooperate in a pathway that controls survival of retinal precursor cells. We do not know the precise roles of these genes in this pathway but hypothesize that the eye mutant phenotypes are the consequence of defective snRNP biogenesis and pre-mRNA splicing. There is ample evidence that Smn controls these processes [38,74], and previous studies have shown that dTgs1 is required for cap hypermethylation of U2 and U4 snRNAs [7]. Since the protein products of dTgs1 and Smn strongly interact, we postulate that overexpression of one of these genes can partially compensate for downregulation of the other, and that inhibition of both genes can worsen the situation. However, assuming that both genes control pre-mRNA splicing in the eye imaginal disc, their specific pre-mRNA targets remain to be determined.
It should be noted that Tgs1 is likely to interact with several additional RNA species besides the snRNAs. For example, human TGS1 cooperates with SMN for cap hypermethylation of selenoprotein mRNAs [3]. TGS1 also trimethylates snoRNPs and regulates their traffic between the Cajal body and the nucleolus [4,75]. Finally, it has been reported that TGS1 is required for primary micro RNAs (pri-miRNAs) trimethylation and maturation in human quiescent cells [76]. Thus, failure to hypermethylate RNA species other than snRNAs could contribute to the eye defects observed in Tgs1 and Smn RNAi flies. The partially different eye phenotypes observed in these RNAi flies are likely to reflect the roles of both common and specific RNA targets of Tgs1 and Smn.
The Drosophila eye as model system for identifying modifiers of the Tgs1 and Smn loss of function phenotypes The Drosophila model has been extremely useful to identify modifiers of the Smn loss of function phenotype. As mentioned earlier, molecular analysis of Drosophila Smn mutants led to the identification of Stasimon, an Smn target gene whose expression in neurons restores the neurological defects elicited by Smn mutations [20]. Stasimon orthologs share a conserved function and ameliorate the Smn loss of function phenotype also in zebrafish and mice [20,25]. Other modifiers of the Smn-dependent phenotype have been isolated for their interaction with hypomorphic Smn mutations. For example, combinations of a specific Smn hypomorphic allele a with a large number of insertional mutations affecting approximately 50% of the fly genes (the Exelixis collection) led to the identification of 17 enhancers and 10 suppressors of the Smn mutant phenotype, most of which affected the neuromuscular junction (NMJ) [44]. The C. elegans orthologs of 12 of these genes were able to modify the Smn loss of function defects in the worm [77]. The same study showed that the invertebrate orthologs of Plastin 3 (PLS3), a human SMA modifier, also modify the Smn loss of function phenotype in both C. elegans and Drosophila models [77] More recently, we analyzed the roles of the Drosophila and C. elegans orthologs of the human WDR79/TCAB1 gene, which encodes a protein that interacts with SMN and controls the biogenesis of several RNA species. Downregulation of these genes in flies and worms resulted in locomotion defects similar to those elicited by Smn depletion. In addition, we found that WDR79 overexpression ameliorates the locomotion defects caused by Smn depletion and vice versa [43]. Collectively these results indicate that the Smn interacting genes identified in Drosophila are conserved across species, and reinforce the idea that Drosophila is a well-suited model organism for detecting Smn loss of function modifiers.
We have recently shown that TGS1 mutations cause an increase in the hTR level and telomerase activity in human cells, leading to telomere elongation [4]. Telomerase insufficiency and abnormally short telomeres can cause dyskeratosis congenita, pulmonary fibrosis and disorders associated with human aging [78]. The dTgs1 loss-of-function eye model could provide an important contribution to research on both SMA and telomere-related pathologies, as it allows an easy detection of chemical and genetic modifiers of the eye phenotype. Identification of such modifiers might help devising new approaches to the treatment of these genetic disorders.

GFP-TRAP-A isolation of GFP-Tgs1 and Smn-GFP
For AP/MS experiments we used embryos from females expressing GFP-dTgs1 from the pPGW-Tgs1 construct induced by Act-Gal4, or expressing Smn-GFP from a tub-Smn-GFP transgene (see above for description of the transgenic stocks). Batches of 0-3 h old embryos laid by cages of 1-10 days-old flies were dechorionated, weighed, flash frozen in liquid nitrogen and stored at -80˚C. For MS analysis, the following procedure was undertaken:~0.4 g of frozen embryos were homogenized in 1.5 ml of C buffer (50 mM HEPES [pH 7.4], 50 mM KCl, 1 mM MgCl2, 1 mM EGTA, 0.1% IGEPAL CA-630, protease inhibitors (Roche). Extract was clarified through centrifugation at 10,000 g for 10 min, 100,000 g for 30 min, and 100,000 g for a further 10 min. Clarified extract was incubated with 15 μl GFP-TRAP-A beads or blocked agarose beads (bab-20) (Chromotek) equilibrated in C Buffer for 2 h at 4˚C. Tgs1 GFP/GFP-TRAP-A beads were then washed 4 times with ice-cold C buffer and stored at -20˚C.

Mass spectrometric analysis
Mass spectrometric analysis was undertaken by the Bristol Proteomics Facility (http://www. bristol.ac.uk/biomedical-sciences/research/facilities/proteomics/), Samples were run~1 cm into the separating region of an SDS-PA gel, cut as a single slice and subjected to in-gel tryptic digestion using a DigestPro automated digestion unit (Intavis Ltd.). The resulting peptides were fractionated using a Dionex Ultimate 3000 nanoHPLC system in line with an LTQ-Orbitrap Velos mass spectrometer (Thermo Scientific). In brief, peptides in 1% (vol/vol) formic acid were injected onto an Acclaim PepMap C18 nano trap column (Dionex). After washing with 0.5% (vol/vol) acetonitrile 0.1% (vol/vol) formic acid peptides were resolved on a 250 mm × 75 μm Acclaim PepMap C18 reverse phase analytical column (Dionex) over a 150 min organic gradient, using 7 gradient segments (1-6% solvent B over 1 min, 6 15% B over 58 min, 15-32% B over 58 min, 32-40% B over 3 min, 40-90% B over 1 min, held at 90% B for 6 min and then reduced to 1% B over 1 min) with a flow rate of 300 nl min−1. Solvent A was 0.1% formic acid and Solvent B was aqueous 80% acetonitrile in 0.1% formic acid. Peptides were ionized by nano-electrospray ionization at 2.1 kV using a stainless steel emitter with an internal diameter of 30 μm (Thermo Scientific) and a capillary temperature of 250˚C. Tandem mass spectra were acquired using an LTQ-Orbitrap Velos mass spectrometer controlled by Xcalibur 2.1 76 software (Thermo Scientific) and operated in data-dependent acquisition mode. The Orbitrap was set to analyse the survey scans at 60,000 resolution (at m/z 400) in the mass range m/z 300 to 2000 and the top twenty multiply charged ions in each duty cycle selected for MS/MS in the LTQ linear ion trap. Charge state filtering, where unassigned precursor ions were not selected for fragmentation, and dynamic exclusion (repeat count, 1; repeat duration, 30 s; exclusion list size, 500) was used. Fragmentation conditions in the LTQ were as follows: normalized collision energy, 40%; activation q, 0.25; activation time, 10 ms; and minimum ion selection intensity, 500 counts. The raw data files were processed and quantified using Proteome Discoverer software v1.2 (Thermo Scientific) and searched against the dmel-all translation-r5.47 database using the SEQUEST (Ver. 28 Rev. 13) algorithm. Peptide precursor mass tolerance was set at 10ppm, and MS/MS tolerance was set at 0.8Da. Search criteria included carbamidomethylation of cysteine (+57.0214) as a fixed modification and oxidation of methionine (+15.9949) as a variable modification. Searches were performed with full tryptic digestion and a maximum of 1 missed cleavage was allowed. The reverse database search option was enabled and all peptide data was filtered to satisfy false discovery rate (FDR) of 5%.

Bioinformatics filtering of MS data
For stringent filtering, MS results were filtered by removing protein IDs with (i) single peptide hits, (ii) <20% peptide:protein coverage and (iii) overall MS Scores of <50. Remaining IDs were cross-referenced against an accumulated database of "false-positives"; MS data used as negative controls were accumulated from eight independent control GFP-TRAP-A experiments, each using extracts from~0.4 g 0-3 h embryos expressing GFP fusions to proteins in which a bait protein was not precipitated, as described by [40]. Any protein ID that was not identified in negative control list and with a score above 50 is presented in Fig 2. The full datasets are provided in S1 and S2 Tables.

Larval locomotion analysis
Larval locomotion analysis was described previously [43]. Briefly, locomotor activity was measured by counting the number of peristaltic contractions per minute of third instar larvae on a surface of a 1% agarose gel in a Petri dish; measurements were repeated ten times. To obtain unbiased measurement of locomotion parameters, larvae were blind-tested by three experimenters. The significance of multiple comparisons was evaluated with One Way Analysis of Variance. The Tukey's test was performed as Post-Hoc Test to determine the significance between every single group. (P<0.01 was considered significant).

Microscopic analysis of Eye-antennal imaginal discs
Third instar larval eye-antennal imaginal discs were dissected in ice cold PBS and fixed in 4% paraformaldehyde in PBS at room temperature for 30 min. After fixation, the tissues were washed in PBS 0,3% Triton X-100 (PBS-T) for 3 x 20 min and blocked with PBS-T and 5% BSA for 30 min. Discs were incubated overnight at 4˚C with anti-Cleaved Caspase3 (1:300; Cell Signaling Technology) and mouse anti-Elav [1:100; Developmental Studies Hybridoma Bank (DSHB)], washed 3 x 20 min in PBS-T, and then incubated for 1 hr at room temperature with Cy3-conjugated anti-rabbit (1:300, Life Technologies) and FITC-conjugated anti-mouse (1:100, Jackson Laboratories). All samples were mounted in Vectashield with DAPI (Vector) to stain DNA and reduce fluorescence fading. Images were acquired using an Axio Imager M2 fluorescence microscope (Zeiss, Germany). Image z-stacks were acquired with an Axiocam 512 (Zeiss) monochromatic camera and Apotome 2 (Zeiss); for each image15 z-stack were acquired at 0,5 micrometer Z step. All images were acquired with the same parameters. Fluorescence signals were quantified using Zen 2.5 Pro software (Zeiss, Germany). Measurements of the eye-antennal disc areas and the ELAV stained areas were performed with the Zen 2.5 Pro software (Zeiss, Germany) on Maximum intensity projections z-stacks. Caspase3 positive foci were also quantified using Zen 2.5 Pro software. Fluorescence threshold was setup starting PLOS GENETICS from the basal fluorescence, and spots were considered positive starting from 3 times the basal fluorescence. We considered only spots with areas greater than 1 μm but less than 60 μm and calculated the mean value of their fluorescence signals. Statistical significance was calculated using the Kruskal-Wallis test with GraphPad Prism.

Evaluation of the eye phenotype
The eye size of RNAi flies was evaluated by comparison with the wild type eye (Oregon-R). Abnormal eyes were assigned to one of the following classes: class 100 that comprises eyes of normal size, and classes 25, 50, and 75 that include eyes with sizes < 25%, < 50%, and < 75% the size of the wild type eye, respectively. Eyes were classified by visual inspection performed independently by at least two researchers. When classification was not clear-cut, the eye was assigned to the higher class in the evaluation of RNAi phenotype without rescue construct (e.g., an eye of dubious class 50 was assigned to class 75), and to the lower class in the presence of a rescue construct.

Generation of an anti-dTgs1 antibody
To generate an anti-dTgs1 rabbit polyclonal antiserum, the dTgs1 sequence coding for aa 335-466 was cloned into the pGEX vector to produce a peptide fused to GST. The GST-fusion peptide was expressed in BL21-CodonPlus Competent Cells-Agilent, and purified by incubating crude lysates with glutathione sepharose 4B (Amersham), as recommended by the manufacturer. Rabbit Immunization was carried out at Agro-Bio Services (www.agro-bio.com). The specificity of the antiserum was confirmed by Western blotting on protein extracts from homozygous dTgs1 null mutants and control flies (Fig 3C). In control flies, the antiserum detects a protein of the expected MW for dTgs1 (~60 KDa) that is strongly reduced in mutants.
Supporting information S1 Table. Full MS datasets from a GFP-dTgs1 AP-MS experiment. Extracts from 0-3 hr embryos expressing GFP-dTgs1 were subjected to affinity purification using GFP-TRAP-A beads, followed by Mass Spectrometry. (XLSX) S2 Table. Full MS datasets from an Smn-GFP AP-MS experiment. Extracts from 0-3 hr embryos expressing Smn-GFP were subjected to affinity purification using GFP-TRAP-A beads, followed by Mass Spectrometry. (XLSX)