Expression of fbl, ncl, and ubtf during X. laevis development

To investigate the role of the methyltransferase fibrillarin (Fbl) in Xenopus development, we first established its spatiotemporal expression during early embryogenesis. For comparison, we mapped the expression of two additional ribosome assembly factors, namely Ubtf, important in rRNA synthesis, and nucleolin (Ncl), a late-acting ribosome assembly factor.

We first analyzed the temporal expression of fbl, ncl, and ubtf and found all three genes to be expressed maternally (stages 2, 4, and 6), and throughout development, notably after the midblastula transition (MBT, stage 8) (S1 Fig).

Next, we analyzed the spatial expression of the selected genes by whole-mount in situ hybridization (WISH). We selected five developmental stages corresponding to critical time points during embryogenesis: segmentation (stage 4), gastrulation (stages 10–11), neurulation (stage 18), and early and late organogenesis (stages 25–26, and stage 32). Transcripts of all three genes were detected only at the animal pole of cleaving-stage embryos and at gastrulation (Fig 1A, stage 4 and stages 10–11), suggesting asymmetric distribution of the corresponding maternal transcripts. From neurulation on (stage 18), expression of fbl, ncl, and ubtf was observed in the developing neurectoderm. At the tailbud and early tadpole stages (Fig 1A, stages 25–26, and stage 32, respectively), transcripts were detected in the developing eyes, brain (both the midbrain and the hindbrain), and branchial arches. Expression of ncl and ubtf was also observed in the pronephros at these later stages.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Spatiotemporal expression of fbl, ncl, and ubtf in Xenopus laevis during development.

A, fbl, ncl, and ubtf mRNAs were detected in X. laevis during development by whole mount in situ hybridization (WISH). The stages and, where relevant, the animal and vegetal poles are indicated. The schematics highlight the labeled structures at stage 18 (neurulation) and at stages 25–32 (organogenesis). NCCs, neural crest cells; H, hindbrain; M, midbrain; E, eyes; P, pronephros; BA, branchial arches. B, fbl expression was analyzed by RT-qPCR in animal cap explants injected either with noggin (to induce differentiation to neural tissue detected with sox3) or with noggin and wnt8 (to stimulate differentiation to neural crest tissue detected with slug). Non-induced (non-injected) cap explants naturally differentiate to epiderm detected with krt12. As a control, expression of the inspected genes was normalized to gapdh. fbl was most abundantly expressed in neural crest tissue. C, Spatial distribution of fbl mRNAs established by WISH in E14.5 mouse embryos. The schematics indicate the orientation of the section plane (dashed line). v, vibrissae; m, mesenchyma; c, primordium of mandibular cartilage; r, retina; sm, submandibular glands.

https://doi.org/10.1371/journal.pgen.1010012.g001

We further analyzed the expression of fbl, the focus of this work, in animal cap explants. Animal cap cells are ectodermal cells forming the roof of the blastocoele. These cells are pluripotent and specify at that stage to form epidermis. We analyzed fbl expression by RT-qPCR in animal cap explants derived from embryos injected at their animal pole with either noggin mRNA, to induce their differentiation to neural tissue, or a combination of noggin and wnt8 mRNAs, to induce their differentiation to neural crest cells (NCCs) [43]. Expression of fbl appeared similar in neuralized and non-induced control caps, but it was higher in induced NCCs (Fig 1B). Fbl is thus likely to play an important role in NCCs. As controls, the expression of krt12 (encoding the epidermal marker keratin, EpK), sox3 (a neural marker), and slug (an NCCs marker) were determined. As expected, krt12 was found only in non-induced cells, sox3 was higher in induced neural caps, and slug was detected only in caps induced to form NCCs (Fig 1B).

We wondered whether the abundance of fbl transcripts in NCCs is an evolutionarily conserved feature. Previously, the distribution of fbl transcripts was studied in the mouse embryo, but the insufficient resolution of the analysis at the time did not allow revealing any specific enrichment in the brain area [44]. To address this, we carefully inspected fbl expression in E14.5 mouse embryos by sectioned in situ hybridization and found abundant fbl transcripts in the retina and the neural-crest-derived mesenchymal cells of the head (Fig 1C). This localization is consistent with that observed in frogs.

In conclusion, fbl, ncl, and ubtf appear to be expressed non-uniformly during Xenopus embryonic development, with strikingly stronger expression in neural crest cells and their derivatives. The spatial distribution of fbl transcripts is similar in mouse. Ribosomes are produced in every cell, and Fbl is essential to ribosome production ([45–47] and see below). Nonetheless, an elevated fbl transcript abundance in particular tissues may be associated with increased Fbl protein synthesis. If Fbl is limiting for ribosome synthesis, this increased production might alter the rate or pathway of pre-rRNA packaging, processing or modification.

Effects of Fbl, Ncl, and Ubtf depletion on X. laevis early development

To assess the involvement of Fbl, Ncl, and Ubtf in early embryogenesis, we performed knock-down experiments with antisense morpholinos (MOs) directed against the translation initiation site in their respective mRNAs. Embryos at the two-cell stage were injected unilaterally at the animal pole with 20 ng MO. Injected embryos were fixed at stage 42 and examined for morphological changes (Fig 2A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Morphological defects during Xenopus laevis development upon Fbl, Ncl, or Ubtf depletion.

A, Embryos injected unilaterally at the two-cell stage with the indicated morpholino (MO) were inspected under the microscope at stage 42. As a control, a non-targeting morpholino (Ctrl) was used. Two embryos are presented (#1 and #2), with the left side injected. With the Ctrl morpholino and those targeting ubtf, and ncl expression, two independent experiments were performed, the total number of embryos tested being 100, 94 and 77, respectively. With the morpholino targeting fbl expression, a total of 154 embryos were analyzed in three independent experiments. B, Quantification of the data presented in panels A and E. Percentage of embryos displaying reduced eye size, or craniofacial skeleton defects (head asymmetry). C, Cartilage staining with alcian blue revealed a defect in all the branchial cartilages, namely: ir, infrarostral; m, meckel; p, palatoquadrate; c, ceratohyal; b, other branchial cartilages. Non-dissected and dissected embryos are presented for comparison. Arrowheads highlight striking reduction, or near-complete loss of the branchial arches. D, Western blot analysis of Fbl-depleted and rescued morphants. Total protein was extracted from individual neurula embryos; one embryo equivalent was loaded per lane. The blots were probed with an anti-Flag antibody (top panel) or an anti-Fbl antibody (bottom panel). The anti-Flag antibody detects only the rescue constructs (Flag-Fbl-wt or Flag-Fbl-D238A). The asterisk (*) denotes detection of an aspecific band. The anti-Fbl antibody detects both the endogenous protein and the rescue constructs. The percentage of Fbl expressed with respect to the uninjected control is indicated. The western blots displayed are representative of four independent experiments. As loading control, the blots were probed for alpha-tubulin. E, Both the wild-type Flag-Fbl and the catalytically deficient Flag-Fbl-D238A restored formation of branchial arches to about half the normal size. Representative example of an embryo co-injected with fbl MO and either a MO-resistant wild-type rescue construct (top) or the same construct harboring the D238A mutation (bottom). A total of 77 and 65 embryos were analyzed in two independent experiments with the Flag-Fbl-wt and Flag-Fbl-D238A construct, respectively.

https://doi.org/10.1371/journal.pgen.1010012.g002

Microinjection of a MO specifically targeting Fbl, Ncl, or Ubtf mRNAs resulted in a similar and more or less pronounced alteration of the eye. Between 59 and 82.5% of the injected embryos displayed a reduced eye size accompanied, in 46–59% of cases, by head asymmetry indicative of craniofacial skeleton alterations (Fig 2A and 2B). The area occupied by the retinal pigmented epithelium (RPE) was measured and found to be reduced by 10–25% (S2 Fig and S1 Data). No such eye or craniofacial cytoskeleton alterations were observed in embryos injected with a non-targeting control MO (Ctrl). The morphology of the craniofacial cartilage was further investigated by alcian blue staining (Fig 2C). This further revealed in the morphants a striking reduction, or near-complete loss in the case of Fbl depletion, of the branchial arches (Fig 2C, arrowheads). The amount of residual endogenous Fbl after MO treatment was established by western blotting as ~31% (~one-third) of the normal level (Fig 2D, lane 2).

To prove that the observed developmental defects were caused by Fbl depletion, we performed rescue assays. When the fbl MO was co-injected with an in vitro-synthesized mRNA encoding Flag-tagged wild-type Fbl (Flag-Fbl-wt), the eye size, the area occupied by the retinal pigmented epithelium, the presence of the branchial arches, and the symmetry of craniofacial skeleton were all largely restored (Fig 2B, 2E and S2 Fig and S1 Data). Note that the rescue transcript was made resistant to MO interference by modifying its 5’ untranslated region (see Materials and Methods). Efficient expression of Flag-tagged-fibrillarin was confirmed by western blotting (Fig 2D, lane 3). In rescued Fbl morphants, about half-sized normal branchial arches were formed (Fig 2E, green arrowheads). This highly distinctive phenotype was never observed in embryos treated only with fbl MO. Such partial morphological rescue is compatible with the amount of Fbl detected in rescued morphants: counting both the endogenous residual protein and the Flag-tagged construct, this amounted to ~49% of the normal level (Fig 2D).

At this stage, it was interesting to probe if the methyltransferase function of Fbl is important for morphological rescue. Considering that the residual amount of endogenous Fbl in the morphants is about one-third of the normal amount, testing this possibility was always going to be challenging. Nonetheless, as an initial approach, we expressed the fbl-D238A mutation instead of the wild-type allele in our rescue assay. In the atomic resolution structure of the Archaeoglobus fulgidus fibrillarin-Nop5 complex bound to its cofactor and methyl donor S-adenosyl-_L-methionine (AdoMet), it has been shown that Asp-133 (equivalent to Xenopus laevis D238) is situated within 3.5 Å of the thiomethyl carbon of the bound AdoMet. This implies that it plays a role as a catalytic residue [48] (see S3 Fig for the position of the residue in the catalytic pocket of the human protein). When this residue was mutated into an alanine, the methylation activity of fibrillarin was indeed totally abolished in an in vitro methylation assay [49]. Interestingly, expression of Flag-Fbl-D238A in morphants rescued the phenotypes (eye size, craniofacial skeleton symmetry, and branchial arches) to the same extent as did Flag-Fbl-wt. In particular, half-sized branchial arches were formed (Fig 2B and 2E). This indicates that providing maturing embryos with an additional supply of Fbl (in this case ~10%, see Fig 2D, lane 4), even if it is not catalytically active, was sufficient to restore development partially.

We conclude that all three tested ribosome biogenesis factors, Fbl, Ncl and Ubtf, are important for eye and craniofacial skeleton development in Xenopus.

Fibrillarin depletion reduces expression of neural border and neural crest markers

To understand the origin of the eye and craniofacial skeleton defects observed upon Fbl depletion, we analyzed the distribution of a series of prototypic developmental markers. Embryos were injected unilaterally at the two-cell stage with 20 ng MO targeting fbl transcripts. The injected embryos were fixed at the neurula stage (stage 15–18) or at organogenesis (early tadpole stage 32) and analyzed by WISH to reveal expression of selected markers (Fig 3A and 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The distribution of developmental markers is affected upon Fbl depletion in Xenopus laevis.

A, Expression patterns of a series of representative developmental markers established by WISH in embryos asymmetrically depleted of Fbl. The embryos were observed at the neurula stage (stages 15–18) or at organogenesis (early tadpole, stage 32). ep, epidermis; np, neural plate; npb, neural plate border; ba, branchial arches; nc, neural crest; ot, otic placode; mhj, midbrain-hindbrain junction; Ip, lens placode; b, brain; e, eye; pn, pronephros; tg, trigeminal placode; p1, V1 interneuron progenitors of the spinal cord. The injected side of each embryo was identified by co-injecting lacZ mRNA. B, The altered distribution of twist and pax6 transcripts, observed upon Fbl depletion, was restored by co-injecting embryos with MO and an mRNAs encoding the wild-type protein (Flag-Fbl-wt) or the D238A (Flag-Fbl-D238A) mutant. Legend as in panel A. The injected side of the embryos was identified by co-injecting lacZ mRNA. In panels A and B, the penetrance of the phenotype (number of embryos with reduced expression of the marker vs. the total number of observed embryos) is indicated at the bottom right of each panel. C, Detection of apoptotic DNA fragmentation by TUNEL. Embryos were unilaterally injected with fbl MO alone or together with p53 MO. As controls, embryos were injected with a Ctrl MO or with a transcript encoding DMRT5, known to promote apoptosis. The number of embryos with increased apoptosis vs. the total number of observed embryos is indicated.

https://doi.org/10.1371/journal.pgen.1010012.g003

Expression of the neural plate border genes msx1 and hairy2 and of the neural crest specifier gene twist was reduced on the injected side of the embryos (Fig 3A). This was observed at both the neurula and early tadpole stages for hairy2 and twist, and at the neurula stage for msx1 (Fig 3A). In contrast, expression of krt12 (an epidermal marker), sox2 (a pan-neural plate marker), pax2 (forebrain), pax6 (eye), and also pax8 (otic placodes and pronephros) and prdm12 (cranial placodes) all appeared unaffected at the neurula stage. Of these, only pax6 was downregulated in later-stage embryos (stage 32).

Importantly, the downregulation of twist, observed in neurula embryos, and that of pax6, observed in the eye in tadpole embryos, were both rescued by co-injection of the MO-resistant wild-type fbl construct (Fig 3B). Co-injection of the fbl-D238A construct also restored proper expression of twist and pax6 at these stages (Fig 3B).

In order to provide initial mechanistic insights into the cellular basis of the phenotypes observed in frog upon fibrillarin knockdown, we assessed apoptotic DNA fragmentation by performing terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assays (Fig 3C). The fbl MO was injected unilaterally at the two-cell stage and TUNEL was applied to whole-mount embryos once they reached the neurula stage. TUNEL revealed important levels of DNA fragmentation in the injected sides of the embryos, particularly obvious at the neural plate, where fbl is normally highly expressed. This indicates massive apoptosis. Interestingly, the increased TUNEL staining was suppressed when the fbl MO was co-injected with an anti-p53 MO. This is consistent with the role of p53 in promoting apoptosis. As a control, embryos were unilaterally injected with a transcript expressing DMRT5, whose overexpression has been shown to promote apoptosis under these conditions [50].

Thus, Fbl depletion affects neural crest survival as a result of massive p53-dependent apoptosis. Altogether, our results also indicated that the head and eye defects of Fbl morphants are a likely consequence of altered cranial neural crest cell development, as these cells start to migrate into and become patterned within the facial primordia at the neurula stage, coinciding with the observed increased apoptosis.

Fibrillarin is required for efficient pre-rRNA processing in X. laevis

The involvement of fibrillarin in pre-rRNA processing has been established in a variety of eukaryotic models, ranging from budding yeast to human cells. In these models, the protein was shown to be important for the initial pre-rRNA cleavages leading to synthesis of 18S rRNA, the RNA component of the small ribosomal subunit [45–47]. These cleavages largely rely on the box C/D snoRNA U3, conserved in Xenopus and shown in this organism to be important for these processing steps [51,52].

To test whether Fbl is involved in ribosome biogenesis in Xenopus, total RNA extracted from MO-treated embryos was analyzed by northern blotting at the early tadpole stage (S4 Fig). Probing of the membrane with specific radioactively-labeled oligonucleotides revealed several phenotypes upon Fbl depletion: 1) accumulation of the primary ribosomal transcript (40S), 2) reduction of 20S pre-rRNA, a precursor of the 18S rRNA, and 3) reduced production of mature 18S rRNA (S4 Fig). These phenotypes are consistent with inhibition of early pre-rRNA cleavages and confirm an evolutionarily conserved role of Fbl in small ribosomal subunit biogenesis in Xenopus.

During the early steps of ribosomal subunit assembly, fibrillarin binds to nascent pre-rRNA transcripts. This early binding step must occur for subunit biogenesis to proceed faithfully (RNA processing) and, importantly, it is independent of its catalytic activity in methylation ([53]). In agreement with our catalytic mutant rescue analysis (Fig 2E), we conclude it is the involvement in pre-rRNA processing of fibrillarin which is responsible for the early developmental defects reported. Furthermore, ribosome biogenesis dysfunctions, including pre-rRNA processing inhibitions, are well-established as triggering nucleolar surveillance leading to p53 stabilization (discussed in [54]). The p53-dependent apoptosis observed by TUNEL (Fig 3C) is thus explainable by the involvement of Fbl in pre-rRNA processing.

Systematic mapping of 2’-O-methylated ribose in X. laevis ribosomal RNAs

Fibrillarin modifies pre-rRNAs, aided by box C/D snoRNAs acting as antisense guides [55]. In pioneering work, Prof. Maden in Liverpool, U.K. mapped the sites of 2’-O methylation on X. laevis ribosomal RNAs, using RNase finger-printing by 2-D thin-layer chromatography [56–58]. Although quite laborious, relying on the use of heavily radioactively-labeled RNAs, and offering only limited resolution, this approach enabled this investigator, nonetheless, to map 89 of the 98 2’-O methylation sites predicted at the time [56–58]. It remained unclear, however, where on the ribosome the nine remaining sites might lie, whether additional sites exist (the sensitivity of the technique was good but some sites might have escaped detection), and most importantly, what proportion of ribosomes in a population of particles is modified at each sites.

At this stage, although we had established it is the role of fibrillarin in pre-rRNA processing which is important for early development, we nonetheless felt it was interesting to reassess systematically in a quantitative fashion sites of ribosomal RNA 2’-O methylation. Such analysis might reveal undocumented variation at specific positions, possibly at particular stages of development.

To investigate systematically the 2’-O methylation of rRNAs in X. laevis, we applied RiboMethSeq, a deep sequencing-based mapping technique relying on differential sensitivity of 2’-O methylated nucleotides to mild alkaline hydrolysis [59,60]. RiboMethSeq does not only afford mapping rRNA sugar methylation with single nucleotide resolution, it is also particularly powerful for quantitative assessment [14–16]. Specifically, we wanted to establish a comprehensive repertoire of all sites of 2’-O methylation in X. laevis rRNAs (possibly identifying novel sites), to evaluate quantitatively the level of modification at each modified position (possibly identifying sites of hypomodification), and to test the possibility that individual sites might be differentially modified during development.

We selected for analysis six stages across early development: segmentation (stage 2), gastrulation (stage 10), neurulation (stage 16), early organogenesis (tailbud, stage 23 and tadpole, stage 32), and late organogenesis (stage 45). As RiboMethSeq requires only minute amounts of total RNA [60], for each stage analyzed we used RNA extracted from a single embryo. For consistency, four embryos were analyzed individually at each selected stage (Figs 4 and S5). Briefly, total RNA was extracted, fragmented by means of mild alkaline treatment, and converted to a deep sequencing library. Libraries were loaded onto a HiSeq1000 sequencer. Following alkaline fragmentation, RNA fragments starting or ending with a 2’-O methylated residue are strongly underrepresented because the phosphodiester bond adjacent to the modification is resistant to cleavage. Mapping of the millions of reads (15 to 25 million) onto the reference rRNA sequences made it possible to infer which position was methylated and the extent of modification at each site (methylation score) [60]. Note that for 28S rRNA, we observed several sequence differences between the reference X. laevis sequence in the database and the sequence of our animals. We therefore deposited a revised 28S sequence at NCBI under ref. MT122800. For the other rRNAs, the reference sequences and ours were the same.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. RiboMethSeq analysis of Xenopus 18S, 5.8S, and 28S rRNAs.

The methylation status of rRNAs was investigated systematically by RiboMethSeq at six developmental stages (stages 2, 10, 16, 23, 32, and 45). At each stage, RNA extracted from four individual embryos was sequenced independently. X axis, modified nucleotides. Y axis, methylation scores (see S1 and S2 Tables for values). In bold, positions hypomodified at all stages (using a cut-off of 0.75). Inset, representative pictures of embryos at each analyzed stage are displayed.

https://doi.org/10.1371/journal.pgen.1010012.g004

The results for 18S, 5.8S, and 28S rRNAs are displayed in Fig 4. The complete dataset is available in S1 and S2 Tables. Cluster analysis revealed, for each of the six developmental stages analyzed, clear clustering of the methylation scores obtained for the four individual embryos (S5 Fig), revealing that our dataset is extremely robust.

Several conclusions were drawn from our RiboMethSeq analysis:

Firstly, the analysis confirmed all 89 of the previously mapped 2’-O methylated positions.

Secondly, we identified 15 novel putative 2’-O methylated residues: one on 18S and fourteen on 28S rRNA (Fig 4 and S1 and S2 Tables). We confirmed ten of these new positions, as well as thirteen of the previously mapped ones, by use of an orthogonal mapping technique, namely: low dNTP primer extension detection [61] (Fig 5). The orthogonally-confirmed novel positions are the 28S nucleotides Am393, Am395, Am960, Gm1286, Am1886, Gm2398, Gm2943, Gm3050, Gm3305, and Gm3501 (Fig 5 and S1 and S2 Tables). The previously mapped positions which we confirmed orthogonally are the 28S residues Am963, Cm977, Am2309, Cm2326, Um2359, Cm2383, Am2869, Am2875, Am2911, Am2936, Gm3534, Gm3603, and Cm4013.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Orthogonal validation by low dNTP primer extension analysis of newly identified 2’-O methylated positions on Xenopus 28S rRNA.

Low dNTP primer extension has proved efficient for specifically detecting 2’-O methylated nucleotides [61]. Total RNA extracted from stage 23 X. laevis embryos was analyzed by primer extension using regular or low dNTP conditions with sixteen oligonucleotides (LD4277 to LD4292) designed to survey 23 positions on the 28S rRNA. A position was confirmed when the signal obtained with low-concentration dNTP was higher than that observed with the regular concentration. The analysis confirmed all 23 positions inspected (highlighted by red squares). Note that position 3501 was confirmed twice, once with oligonucleotide LD4288 and once with LD4289.

https://doi.org/10.1371/journal.pgen.1010012.g005

A bioinformatic search of the X. laevis transcriptome, assisted by the Snoscan software, allowed us to assign box C/D snoRNA guides to all the 2’-O methylated positions identified by RiboMethSeq (S3 Table). We identified a total of 49 novel X. laevis box C/D snoRNA candidates (S3 Table).

Thirdly, using as cut-off a methylation score of 0.75, we observed that numerous positions are not fully modified (Fig 4 and S1 and S2 Tables). This indicates that in X. laevis as in other investigated experimental models such as budding yeast, zebrafish, mouse, and human, cells produce heterogeneous populations of ribosomes with differentially modified ribosomes that might display specific translational properties (see Discussion).

Lastly, close inspection of the methylation scores at individual positions through development revealed that specific nucleotides are differentially modified as the embryos develop (Fig 4 and S1 and S2 Tables), with a trend towards reduced methylation levels at specific positions. At stage 45 (highlighted in purple in Fig 4), this was notably the case at one position in 5.8S, three positions in 18S, and eight in 28S.

Of particular interest is the 18S rRNA Um252 modification, a newly identified position in X. laevis which is not modified in human cells and which is hypomodified at all the stages inspected. Other positions of interest on 18S rRNA are: (1) Cm760, which is hypomethylated throughout development (interestingly, the corresponding position is also hypomethylated in cultured human cells [15]); (2) Um1761, which is hypomodified throughout development whereas its human counterpart is fully methylated [15]; (3) three additional positions which are hypomethylated only at stage 45 (Um393, Am477, Am555). The equivalent human positions are also hypomodified, at least in some cell lines investigated [14,15,17]. On the 28S rRNA, we noted that: (1) seven positions are hypomodified at all developmental stages analyzed (Am963, Am1848, Cm1850, Gm2895, Am2911, Am3018, Cm3565), the equivalent human positions of five of these also being hypomodified [14,15]; (2) eight positions which are hypomethylated only at stage 45 (Am393, Cm1525, Cm1907, Am2337, Cm3020, Gm3534, Am3680, Gm3746). In the human cell lines tested, the equivalent positions of three of these (Cm1881, Cm3869, and Am4571) are also hypomethylated [15,17].

Neural crest cell survival requires efficient ribosome biogenesis in Xenopus

With the emergence of ribosomopathies, a novel class of human syndromes caused by ribosome biogenesis dysfunction, investigators have become aware of the importance of understanding how tissue-specific diseases can involve a process as global as ribosome biogenesis [26]. In this work, we have studied the involvement of the box C/D snoRNA-associated methyltransferase fibrillarin in the early development of X. laevis. We have established the spatiotemporal expression of fibrillarin, revealing it to be abundantly expressed in neural crest cells and cells derived from them (Fig 1). This is also true of two other factors involved in distinct ribosome biogenesis steps, Ncl and Ubtf. We have shown that morpholino-mediated depletion of Fbl strongly impacts the survival of NCCs, with dramatic repercussions on eye and craniofacial skeleton formation leading to head dysmorphism caused by severe branchial arches alterations (Fig 2). We have corroborated these developmental defects by demonstrating altered expression of several markers of NCC formation (hairy2, msx1, and twist) and by detecting increased apoptosis at the neural plate (Fig 3).

Why are ribosome biogenesis factors more abundantly expressed in neural crest cells and derivatives?

A straightforward explanation might be that specific parts of the embryo rely on more abundant protein production to develop, and thus on more active ribosome biogenesis. Another possibility is that specific parts of the embryo have differential requirements for specific ribosome biogenesis factors, possibly to assemble distinct ribosomes. Although this second hypothesis is quite appealing, we do not particularly favor it at this stage, because the spatiotemporal expression profiles of the three ribosome biogenesis factors inspected in this work largely overlap and because these are core factors playing evolutionarily conserved essential roles in ribosome biogenesis. This view is strengthened by the observation that additional ribosome assembly factors, and also several ribosomal proteins, have been shown to share, during early development, expression patterns largely similar to that observed here for Fbl, Ncl, and Ubtf [36–39,62].

During development in the mouse, Fbl is essential at a very early stage, prior to embryo implantation [44]. The mechanisms involved and the precise spatiotemporal expression of Fbl in mouse embryos remained undefined. Here we have clarified the pattern of expression of Fbl at high resolution in E14.5 mouse embryos, finding it to be consistent with our observations on X. laevis (Fig 1C). In zebrafish, fibrillarin is preferentially expressed in the proliferating cells of the optic tectum, in the retina, and in the dorsal midbrain, and it is important for the maturation of these regions [47]. These observations on a fish are perfectly consistent with our own findings in X. laevis (Figs 1 and 3). The function of fibrillarin in pre-rRNA processing is conserved in yeast [45], human cells [46], and zebrafish [47], and now we show it to be conserved also in X. laevis (S4 Fig).

There are currently two models for explaining how a global process such as ribosome biogenesis can lead to tissue-specific disorders: the ‘specialized ribosomes’ model and the ‘ribosome concentration’ model. The specialized ribosome model posits that compositionally distinct ribosomes, such as those harboring more or less methylated residues, have differential translational capacities [63]. The ribosome concentration model proposes, rather, that the amount of functional ribosomes is what dictates which mRNAs are preferentially translated. In this case, the rationale is that naturally ‘less favored’ mRNAs will be even less translated when the intracellular ribosome concentration drops [64]. Both of these models have qualities and both mechanisms are likely to coexist in cells, each contributing to some extent to regulating the translational capacity. Because of the pre-rRNA processing defects observed and the phenotypic rescue with a catalytically defective allele, in the case of the Fbl-mediated developmental defects reported here, our data clearly favors the ribosome concentration model.

The majority of bones, cartilages, and connective tissues comprising the head and face are derived from neural crest cells. Craniofacial skeleton alterations are conspicuous in developing animals where ribosome biogenesis is inhibited. Typically, they are observed in fish and mouse embryos with phenotypes reminiscent of Treacher-Collins syndrome, a neurocristopathy caused by mutations in factors important for rRNA synthesis [42,65,66]. They are also observed upon fibrillarin depletion in Xenopus, as reported here, and in both Xenopus and fish models upon depletion of several other factors intervening at different steps of ribosome biogenesis [33,38]. Altogether, this strongly supports the notion that NCC survival is particularly sensitive to the capacity of cells to synthetize ribosomes.

Ribosomal RNAs are distinctly hypomethylated at late stages of Xenopus development

On the basis of our RiboMethSeq data, we report 103 methylation sites on X. laevis rRNA: 34 on 18S, 2 on 5.8S, and 67 on 28S (see S1 and S2 Tables). By comparison, human rRNAs harbor 112 positions: 42 on 18S, 2 on 5.8S, and 68 on 28S, most sites being conserved from frog to human (S1 Table). On 18S rRNA, X. laevis carries the Um252 modification, which is not conserved in human cells and which is hypomodified at all stages inspected. Human 18S rRNA contains nine modifications which are not conserved in X. laevis, namely Am159, Um172, Cm174, Um354, Cm621, Gm867, Cm1272, Gm1447, and Um1668. Among them, most are hypomodified in the human cell lines tested (Cm174, Um354, Cm621, Gm867, Cm1272, Gm1447 and Um1668). On 28S rRNA, X. laevis harbors four modifications which are not conserved in human cells: Gm432, Cm1113, Gm3534 and Cm4013. Only Cm3534 is hypomethylated at later developmental stages (32 and 45). Human 28S rRNA contains five modifications that are absent in X. laevis: Gm1316, Um1773, Cm2824, Gm3627, and Gm4618. Again, most of these are hypomodified (Gm1316, Um1773, Cm2824, Gm4618).

Of the 103 positions mapped here in X. laevis, fifteen have not been detected previously [56–58]. Importantly, using of a low dNTP primer extension assay, we have orthogonally validated 23 positions, including ten of the newly identified ones (Fig 5). Furthermore, upon searching the X. laevis transcriptome, we have identified box C/D snoRNA candidates for all the identified positions (S3 Table). We also note that one position identified historically [56,57], Um2326, was not confirmed by our RMS or our low dNTP primer extension analysis. This position is also absent in X. tropicalis. We conclude that this position was originally misassigned.

We have profiled rRNA sugar methylation in individual embryos at six developmental stages, concluding that the vast majority (87.6%) of positions are robustly installed, while a minority (12.4%) are hypomethylated (cut-off score: 0.75). In principle, fractional ribosomal RNA 2’-O methylation implies that cells produce a heterogeneous population of ribosomes. Whether these differentially modified ribosomes are all actively involved in translation remains to be established. Ribosomal RNA hypomethylation has been observed in cancer cells [14–17], and during development in fish [18] and mouse [19]. We now report it in Xenopus. The exact significance of substoichiometric rRNA methylation remains largely unknown.

We have also observed that the degree of hypomethylation increases substantially at later stages of development. This is particularly striking at stage 45, at three positions on 18S, one on 5.8S, and eight on 28S. Whether and how reduced methylation at these sites might affect translation, and possibly late stage of development, deserves future investigation.

The mechanisms of hypomethylation are not yet known. In particular, it remains unclear whether sites of hypomethylation are hardcoded or whether they simply reflect stochastic elements inherent in ribosomal subunit biogenesis. What we have learned so far with cancer cells is that not all sites vary to the same extent [15]. In particular, sites located at the periphery of a ribosomal subunit are more prone to variation than those located at the core of the ribosome, where the functionally important sites lie. Thus, only a limited number of positions are subjected to partial modification.

We have also learned that upon controlled and progressive depletion of fibrillarin in human cells, the sites which become less modified are largely those which are naturally hypomodified [15]. This clearly indicates that some sites are more easily modified than others. It also provides a simple mechanistic explanation of hypomethylation: under conditions of active cell proliferation, as in cancer cells or at specific developmental stages, the rRNA modification machinery might be ‘overwhelmed’ by increased ribosome production, and snoRNPs might be insufficient in number or might find it hard to access their binding sites within the complex folding 3-D structure of the maturing ribosomal subunits. In agreement with this hypothesis, profiling of rRNA 2’-O methylation in differentiated tissues, where cell proliferation is reduced as compared to cultured cells, shows most sites to be robustly methylated [18,19].

In a recent work, the team of Prof. Gall used RNA-sequencing data from the giant oocyte nucleus of X. tropicalis to annotate snoRNAs and to compare his findings with those pertaining to other vertebrates [67]. In their analysis, the authors identified in X. tropicalis 102 2’-O methylated positions and 99 box C/D snoRNAs. They predicted novel modified positions, some of which were investigated in a complex orthologous system consisting of a yeast strain deleted of the endogenous snoRNA guide and provided with an exogenous frog guide expressed from a plasmid. Methylation of the predicted substrate was then verified by low dNTP primer extension [67], which is non- quantitative. While snoRNP machineries are well conserved, the use of a heterologous system might not be optimal in all cases because of species specificities. It also raises questions regarding the copy number of the complementing exogenous snoRNA, particularly in cases of weak anti-sense guide elements which might not allow modification when expressed at physiological levels. By comparison, we have conducted a comprehensive survey of rRNA 2’-O methylation directly on X. laevis rRNAs extracted from individual developing animals. For this we have used the highly quantitative RiboMethSeq technique, subsequently assigning snoRNA guides to the known and newly identified positions.

In conclusion, we have shown in X. laevis that neural crest cells are particularly sensitive to ribosome biogenesis inhibition, and that in the late stages of embryonic development rRNA hypomethylation strikingly increases at specific positions. Further study is required to determine whether this developmentally regulated rRNA hypomethylation results in differential translation of particular mRNA transcripts important for controlling late developmental transitions.

Ethics statement

Animal housing and experimental protocols were approved by the CEBEA (“Comité d’éthique et du bien être animal”) of the IBMM-ULB and conformed to the European guidelines on the ethical care and use of animals.

Expression plasmids and morpholinos

The fbl, ubtf, and ncl expression vector were constructed by PCR from X. laevis pCMV-SPORT6-FBL (mRNAID: XM_018228388.1), pCMV-SPORT6-UBTF (mRNAID: NM_001101787.1), and pCR4-TOPO-NCL (mRNAID: NM_001088088.1), using primers described in S4 Table. The PCR products were cloned into the BamH I and Xho I sites of the pCS2 vector. To deplete Ubtf, fibrillarin, and nucleolin, we used an antisense morpholino specific to sequences directly upstream of their ATG. Morpholinos were purchased from GeneTools (Philomath, OR) (S5 Table).

Fibrillarin rescue constructs: We produced two fibrillarin rescue constructs, one encoding the wild-type allele, the other encoding the D238A mutation (see S3 Fig). mRNA was produced from each construct by in vitro transcription, it was capped, poly-adenylated, and purified on gel. The purified mRNA was injected directly into the embryos (see below). The rescue constructs were resistant to the morpholinos because the sequence upstream of the ATG was modified so as to lose its complementarity to the antisense oligo. To distinguish endogenously expressed fibrillarin from protein produced from the injected mRNAs, a Flag-tag was inserted in-frame at the amino-terminus of the rescue constructs. The wild-type construct was amplified with primers LD3036 and LD2749 and cloned into pCS2-Flag between the EcoR I and the Xho I site. The D238A catalytically deficient fbl mutant was generated by site-directed mutagenesis with the help of the QuikChange II site-directed mutagenesis kit (Stratagene) and primers LD3881 and LD3882. Oligonucleotides are listed in S4 Table.

Embryos, micro-injections, and X. laevis animal cap explant culture

Xenopus embryos were obtained from adult frogs by hormone-induced egg-laying. Embryos were generated by in vitro fertilization [68] and staged according to [69]. For in situ analysis, embryos were injected in one blastomere at the 2–4 cell stage (200 pg mRNA or 10 ng of morpholino per blastomere) and fixed at the neurula or tailbud stages. X. laevis wnt8 (50 pg/blastomere [70]), noggin (100 pg/blastomere [71]), Flag-fbl-wt (250 pg/blastomere), and Flag-fbl-D238A (250 pg/blastomere) mRNAs were synthesized with the mMessage mMachine kit (Ambion). To reveal the injected side, embryos were co-injected with lacZ mRNA (50 pg/blastomere).

In the animal cap assays, mRNAs were microinjected into the animal region of each blastomere of four-cell-stage embryos. Animal caps were dissected at the blastula stage (stage 9) and cultured in 1x Steinberg medium, 0.1% BSA until sibling embryos reached the neurula stage. All Xenopus embryo injections were performed on at least three batches of embryos.

The area occupied by the RPE was calculated for both eyes with Image J and the injected to non-injected side area ratio was calculated. RPE ratios were analyzed with the Kolmogorov-Smirnov test (an unpaired t-test, unparamatretric and specific for cumulative distribution comparisons).

Wild-type C57BL/6 mice were crossed to generate the embryos used in this study.

Lineage tracing, in situ hybridization, and cartilage staining

Xenopus embryos were fixed in MEMFA and stained with X-gal to visualize LacZ activity on the injected side before being processed for in situ hybridization. Whole mount in situ hybridization was performed on Xenopus embryos as previously described [68]. Synthetic digoxigenin-labeled antisense mRNAs probes were prepared using the T7 polymerase and pCS2-fbl, or pCS2-ubtf, or pCS2-ncl. NBT/BCIP was used as a color substrate. For in situ hybridization on mouse embryos, 20-μm cryostat sections of 4% paraformaldehyde-fixed, 30% sucrose/PBS-infused tissues frozen in gelatin (7.5% gelatin, 15% sucrose/PBS) were used. In situ hybridization experiments were performed as previously described [72]. The mouse FBL antisense probe was generated with the corresponding EST. The other antisense probes were generated with previously described templates encoding krt12, sox2, msx1, and hairy2 [73], twist [74], pax2, and pax6 [75], and prdm12 [76]. Cartilage staining of Xenopus embryos was performed on tadpole heads as described [36].

TUNEL assay

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) detects DNA breaks formed during apoptosis. Whole mount TUNEL staining of developmentally staged (neurula) control and morphant embryos was carried out as described in [77] with the following reagents: terminal deoxynucleotidyl transferase (TdT) (15 U/μl and 5x buffer, Thermofisher Ref. 10533065), digoxigenin-11-dUTP (25 nmol/25μl, Sigma Ref. 11093088910), Normal goat serum (Thermofisher Ref. 10000C), anti-dig alkaline phosphatase conjugate (Sigma Ref. 11093274910), nitro blue tetrazolium (Sigma Ref. 11383213001), and 5-bromo-4-chloro-3-indolyl phosphate (Sigma Ref. 10760994001). The control DMRT5 transcript was prepared as previously described [50].

Western blot analysis

Total protein was extracted from single Xenopus embryos lysed with a mini-pestle (Vwr, Ref. 431–0094) in liquid nitrogen in a 1.5-ml Eppendorf tube and resuspended in ice-cold lysis buffer (50 mM Tris pH 7.6, 10 mM EDTA, 150 mM NaCl, 1% Triton, supplemented with complete protease inhibitor, Roche Ref. 11697498001–1 pellet per 50 ml buffer) and incubated on ice for 10 min. The lysate was cleared by centrifugation (12,000 rpm, 10 min at 4°C). One embryo equivalent was loaded onto a 15% SDS-PAGE gel. The gels were transferred to nitrocellulose membranes and probed in TBS/0.1% tween /3% BSA. Probing was as follows: anti-fibrillarin (Abcam, Ref. ab5821) incubated overnight at 1:2000 dilution, followed by anti-rabbit HRP (Cytiva, Ref. NA934V) used at a 1:5000 dilution for 2 hours; anti-Flag (Sigma, Ref. F3165) incubated overnight at 1:2000 dilution followed by anti-mouse HRP (Jackson ImmunoResearch, Ref. 115-035-062) used at 1:5000 for 2 hours; anti-alpha-tubulin (Sigma, Ref. T5168) incubated overnight at a 1:5000 dilution followed by anti-mouse HRP.

RT-qPCR analysis

For RT-qPCR analysis of X. laevis animal caps, total RNA was extracted using the NETS method. The RNA samples were digested with RNase-free DNase I before RT- PCR. The amount of RNA isolated was quantified by measuring the optical density with a nanodrop spectrophotometer. The RNA samples were digested with RNase-free DNAse I before RT-PCR. cDNA was synthesized with the iScript cDNA synthesis kit (Biorad). Synthesized cDNA was amplified with the qScript One-Step SYBR Green qRT-PCR Kit (Quanta bio) with suitable primers (see S4 Table). Samples were normalized to X. laevis gapdh.

Pre-rRNA analysis by northern blotting

Total RNA was extracted from MO-treated and untreated embryos with Tri reagent. Pooled stage 32 embryos were used. 5 μg total RNA was analyzed by northern blotting as previously described [46]. The probes used for the hybridization are depicted in S4 Table.

Systematic mapping of ribosomal RNA 2’-O methylation by RiboMethSeq

RiboMethSeq analysis of rRNA 2’-O-methylation was performed as described in [60]. Briefly, ~100 ng of total RNA extracted from biological material was fragmented in bicarbonate buffer at pH 9.3 at 96°C, fragmentation time was adapted to obtain short RNA fragments of ~20–30 nt in length. Fragmented RNA was end-repaired, to insure compatibility with adapter ligation. Library preparation was done using NEBNext Small RNA kit, according to manufacturer’s recommendations. Barcoded libraries were loaded to the flow-cell and sequenced on Illumina HiSeq 1000. Raw reads were trimmed and aligned to the reference. Bioinformatics analysis was performed as described [60]. The complete data set is available at the European Nucleotide Archive under accession n°PRJEB42253.

Site-specific detection of 2’-O methylation by low dNTP primer extension

Primer extension was conducted as described in [78] on 5 μg total RNA extracted from stage 23 embryos. The primers used are listed in S4 Table.

SnoRNA identification

To find snoRNAs in RNA-seq experiments, we downloaded sequencing data from lariat intronic RNAs of X. laevis [79]. The FASTQ file was downloaded from NCBI (SRR7615230, project PRJNA479418) with the SRA Toolkit package (http://ncbi.github.io/sra-tools/, version 2.1.10): fastq-dump SRR7615230—outdir/path/to/dir. FASTQ files were converted to FASTA files with seqtk software (https://github.com/lh3/seqtk, version 1.0–1) and used as our next input: seqtk seq -a SRR7615230.fq.gz>SRR7615230.fasta. We used snoScan (version 0.9.1) [80] to detect box C/D methylation guide snoRNAs in X. laevis RNA sequencing experiments. The predicted snoRNA sequence is provided in FASTA format. The snoRNA sequence consists of four leading nucleotides followed by the box C sequence, the sequence complementary to rRNA, the D box (and D’ box) sequence, and two trailing nucleotides. snoScan computes a score based on probabilistic models of sequence features (terminal stem, box C, guide sequence, box D, etc.). When multiple snoRNAs were predicted for one methylation site, we chose the one with the highest snoScan average bit scores and with high similarity to a corresponding known snoRNA in X. tropicalis. To validate our results, we checked if we detected the same snoRNAs as already identified by the Rfam database [81] in X. laevis (https://rfam.xfam.org/). We aligned snoScan snoRNA candidates with Rfam snoRNAs with Clustal Omega [82] (https://www.ebi.ac.uk/Tools/msa/clustalo/) and compared sequence alignment scores.