Pofut1 point-mutations that disrupt O-fucosyltransferase activity destabilize the protein and abolish Notch1 signaling during mouse somitogenesis

The segmental pattern of the vertebrate body is established via the periodic formation of somites from the presomitic mesoderm (PSM). This periodical process is controlled by the cyclic and synchronized activation of Notch signaling in the PSM. Protein O-fucosyltransferase1 (Pofut1), which transfers O-fucose to the EGF domains of the Notch1 receptor, is indispensable for Notch signaling activation. The Drosophila homologue Ofut1 was reported to control Notch localization via two different mechanisms, working as a chaperone for Notch or as a regulator of Notch endocytosis. However, these were found to be independent of O-fucosyltransferase activity because the phenotypes were rescued by Ofut1 mutants lacking O-fucosyltransferase activity. Pofut1 may also be involved in the Notch receptor localization in mice. However, the contribution of enzymatic activity of Pofut1 to the Notch receptor dynamics remains to be elucidated. In order to clarify the importance of the O-fucosyltransferase activity of Pofut1 for Notch signaling activation and the protein localization in the PSM, we established mice carrying point mutations at the 245th a.a. or 370-372th a.a., highly conserved amino-acid sequences whose mutations disrupt the O-fucosyltransferase activity of both Drosophila Ofut1 and mammalian Pofut1, with the CRISPR/Cas9 mediated genome-engineering technique. Both mutants displayed the same severely perturbed somite formation and Notch1 subcellular localization defects as the Pofut1 null mutants. In the mutants, Pofut1 protein, but not RNA, became undetectable by E9.5. Furthermore, both wild-type and mutant Pofut1 proteins were degraded through lysosome dependent machinery. Pofut1 protein loss in the point mutant embryos caused the same phenotypes as those observed in Pofut1 null embryos.


Introduction
The segmental feature of the vertebrate body is established in the embryo via the periodic formation of somites from the anterior unsegmented presomitic mesoderm (PSM). During somitogenesis, Notch signaling activation plays crucial roles to make the periodicity of somites and to define the segmental border in mice (Reviewed in [1,2]). In the PSM, Notch signaling is activated via cleavage of Notch1 upon the binding of its ligand Dll1 expressed on a juxtaposed cell. After Notch1 cleavage at the cell membrane, the intracellular domain of Notch1 (NICD) translocates into nuclei to activate target genes, which in turn suppresses Notch signaling activation in trans via activation of Lunatic fringe (Lfng). Thus, the periodical somite formation is controlled by cyclic activation of Notch signaling and negative feedback by Notch targets in the PSM (Reviewed in [1,2]). In order to activate Notch signaling in a group of cells periodically and synchronously in the PSM, cell surface amounts of Notch receptors and its ligands should be tightly regulated. It has been shown that subcellular localization of Notch receptors is controlled by several post-translational modifications, including proteolysis, ubiquitination, and glycosylation (Reviewed in [3][4][5]). Proteolysis of Notch receptors plays important roles for inducing endocytosis of receptors as well as ligands, which is essential for activation of the signaling [6,7]. If the Notch receptors do not interact with ligands, receptors are endocytosed constitutively, and recycled or degraded in lysosomes. Ubiquitination of Notch receptors is also important for controlling the endocytosis of inactivated receptors and subsequent trafficking [8,9]. The extracellular domain of Notch receptors is highly glycosylated during synthesis in the Endoplasmic Reticulum (ER) and trafficking through the Golgi. Glycosylation affects the maturation of Notch receptors, their trafficking, and their affinity for Notch ligands [10][11][12]. Previous observations demonstrated that Lfng plays a crucial role in synchronizing Notch signaling in the PSM [13,14]. Lfng functions to elongate ß1,3 N-acetylglucosamine to Ofucose attached to EGF repeats of the Notch1 receptor [15,16]. Taken together, glycosylation of Notch1 may be a critical modification for synchronous activation of Notch signaling.
The O-fucose attached to EGF repeats of the Notch1 receptor is transferred by Protein O-fucosyltransferase1 (Pofut1) [17,18]. Pofut1 has been shown to be localized to the ER [19], and is indispensable for Notch signaling activation [17,20,21]. The importance of O-fucose modification for Notch1 signaling activation was demonstrated by the reduced ligand binding ability of Notch1 in Pofut1 mutant ES cells without changing the cell surface amount of Notch1 protein [22]. Crystal structure analysis of Notch1 and its ligands also revealed that the O-fucose modification directly influences ligand binding [23,24]. Furthermore, an O-fucosylation site mutation in the EGF repeat of Notch1, which is implicated in ligand binding, was found to create a hypomorphic allele as the compound mutants with Notch1 deletion mutants displayed similar phenotypes with Notch1 null mutants [25]. On the other hand, subcellular localization of Notch1 protein was greatly altered in the PSM of Pofut1-null embryos [26], suggesting that Pofut1 functions may be different in each cell/tissue, and that the control of ligand binding affinity by O-fucose modification of the Notch1 extracellular domain may not be the only function of Pofut1 in the PSM. Similar abnormal subcellular localization of Notch proteins was reported in Ofut1, the Pofut1-homologue, mutant in Drosophila [20,27]. In Drosophila, there are different explanations for Ofut1functions in controlling Notch subcellular localization. One is that Ofut1 acts as a chaperone for Notch proteins in the ER [27], and the other is that it controls the endocytosis of Notch proteins [28]. In either case, each function is independent from the O-fucosyltransferase activity of Ofut1 as the function was rescued by expression of Ofut1 R245A [27] or the DXD-like motif mutant Ofut1 3G [28]. The Arg 245 of Ofut1/Pofut1 is conserved among species and has been shown as an important residue for enzymatic activity [22,27,29,30]. The DXD-like motif is found not only in fucosyltransferases, but in many other glycosyltransferases in different species as well [31], and has been reported as important for enzymatic activity [20,32].
It is possible that the important functions of Pofut1 as O-fucosyltransferase for periodical activation of Notch signaling in the PSM are masked by the phenotype that is induced by abnormal Notch1 localization in the Pofut1-null mutant. To investigate this possibility, we took advantage of the Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) /CRISPR-Associated Protein9 (Cas9) mediated genome-editing technique to establish Pofut1 point mutant mice that lack O-fucosyltransferase activity. This is the first report of mice lacking the enzymatic activity of Pofut1. By analyzing these mice, we unexpectedly found that the Pofut1 protein stability was markedly decreased in the mutants, and abnormal somitogenesis and mis-localization of Notch1 was observed, as in Pofut1-null mutants.

Mice
Mice carrying Pofut1-null (Pofut1 Δ/Δ ) and RBPJk-null (RBPJk Δ/Δ ) alleles were established previously [26,33]. B6C3F1 and MCH strain mice were purchased from CLEA Japan, Inc. Mice were kept in a specific-pathogen free (SPF) facility at 23 ± 2˚C, under 50 ± 10% humidity and 12-hr (8 am-8 pm) lights on, 12-hr (8 pm-8 am) lights off conditions. Mice were kept in cages larger than 130 cm 2 per mouse, and supplied CE-2 food (CREA Japan Inc) and water with 3 ppm sodium hypochlorite. The cages were changed every week. When mice were given surgery, they were anesthetized by intraperitoneal injection of 0.2 μg of 2,2,2-tribromoethyl alcohol (Avertin)/g body weight in advance. Mice were sacrificed with carbon dioxide gas or by cervical spine fracture dislocation. The monitoring of SPF conditions was conducted by the Central Institute for Experimental Animals with the core set every three months, and the full set once a year. All mouse experiments were approved by the Animal Experimentation Committee at the National Institute of Genetics (Permit Number: 27-13).
Transmission electron microscopy (TEM) and Immuno-TEM TEM was conducted according to the protocol previously published [34] with some modifications. Briefly, embryos for TEM analysis were dissected in ice-cold PBS, then immediately fixed with 2% formaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer (PB) for 2 hrs at room temperature, washed with 3% Sucrose /0.1M PB, post-fixed with 1% OsO 4 /0.1M PB, washed with distilled water, block-stained with 0.5% aqueous uranyl acetate, dehydrated with a series of ethanol, then embedded in Epon-mixture. The samples were trimmed with a glass knife, then cut to make ultra-thin sections with a Leica EM UC6. The sections were collected on grids, and stained with 2% uranyl acetate and Reynolds' lead solution. Immuno-TEM was conducted according to the protocol previously published [35] with some modifications. Briefly, embryos for Immuno-TEM analysis were dissected in ice-cold PBS, then immediately fixed with periodate-lysine-paraformaldehyde fixative for 2 hrs at room temperature, submerged in 10%, 20%, and 30% Sucrose/PBS at 4˚C, then embedded in O.C.T. compound (Sakura Finetek) and frozen. The frozen sections (10-μm thickness) were treated with 150 mM glycine/PBS, blocked with 3% BSA/PBS at room temperature for 1 hr, and incubated with anti-Notch1 (C20-R, 1:100, Santa Cruz Biotechnology) antibody three days at 4˚C. After washing the sections with PBS five times, they were incubated with goat anti-rabbit IgG HRP-conjugated antibody (1:200, Cell Signaling) for 2 hrs at room temperature. After washing with PBS, the sections were re-fixed with 0.5% glutaraldehyde/PBS for 5 minutes, washed with PBS, and incubated with 0.025% 3,3'-diaminobenzidine 4HCl (Dojindo) in 50 mM Tris-HCl pH 7.6 (DAB solution) for 30 min, then incubated with DAB solution containing 0.005% H 2 O 2 . The DAB reaction was stopped by washing the sections with PBS several times. Then, the sections were treated with 2% OsO4/0.1M PB for 1 hr, dehydrated with an ethanol series, and embedded with Epon-mixture. The samples were cut to make ultra-thin sections with a Leica EM UC6. The sections were observed, and pictures were taken with low (x10000) and high (x30000) magnification using the JOEL JEM 1010. The quantification of the subcellular localization was conducted as follows. Cells with at least partially visible nuclei in the low-magnification TEM images were counted, and subcellular localization was confirmed using highmagnification images of the same area. The localization was categorized as on the cell surface, on the rough-ER, or on vesicles not in the rough-ER. The cells containing vesicles not in the rough-ER without cell surface and rough-ER signal markers were counted as "other vesicles only". The ratios of each category were calculated for each embryo.
Different concentrations of vectors, sgRNA, hCas9 mRNA, and oligos were mixed in injection buffer (10 mM Tris-HCl and 0.1 mM EDTA, pH 7.5) and injected into the pronucleus of fertilized eggs in M2 medium (Sigma). The injected zygotes were cultured in KSOM (Millipore) at 37˚C under 5% CO2 until the 2-cell stage after 1.5 days. Thereafter, 20-32 2-cell stage embryos were transferred into the uterus of pseudo-pregnant MCH females at 0.5 dpc. Embryos were dissected 9 days after transfer or born naturally. Two independent mouse lines of Pofut1 R245A or Pofut1 3G were established, and these mice were back-crossed with C57BL6/J more than 4 times to reduce off target mutation effects, and no change in Pofut1 R245A/R245A or Pofut1 3G/3G embryo phenotypes was confirmed after back-crossing.

Western blot analysis
Whole embryos, ES cells, or 293T cells transfected with expression vectors using Lipofectamine 2000 (Invitrogen) were washed with PBS, and lysed with RIPA buffer (50 mM Hepes pH 7.4, 150 mM NaCl, 1% TritonX-100, 0.02% Sodium deoxycolate, and 1 mM EDTA) supplemented with Complete mini (Roche). The protein concentration was calculated with a Protein Assay (Bio-Rad) and samples were adjusted to be the same concentration. Sample buffer was added to the lysates and boiled for 5 min. Proteins were separated by running on an 8% SDS-PAGE gel, and transferred to an Immobilon-P membrane (Millipore). The membranes were blocked with 3% skim milk in TBS-T for 2 hrs at room temperature, and incubated with the following primary antibodies: anti-Notch1 (C20-R, 1:1000, Santa Cruz Biotechnology), anti-ß-catenin (rabbit polyclonal, 1:3000, Upstate Biotechnology), anti-Pofut1 ascites described above (1:2000), anti-ß-tubulin (TUB2.1, 1:5000, Sigma Aldrich), and HRP-conjugated anti-Flag antibody (Flag M2, 1:3000, Sigma Aldrich). For detecting endogenous Pofut1 protein, both #28-33 and #35-8 ascites were used for each experiment. The membranes were washed and incubated with HRP-conjugated anti-rabbit IgG antibody (1:3000, Cell Signaling) or HRP-conjugated anti-mouse IgG antibody (1:3000, Cell Signaling). The membranes were washed and incubated with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific) and the signals were detected with an ATTO Ez-capture MG image analyzer. Signal intensities of Pofut1 and ß-tubulin bands were measured with ImageJ, background signals subtracted, and Pofut1 intensity was divided by the corresponding ß-tubulin intensity. Then, each signal was adjusted with the wild-type sample (in different dose experiments: vehicle treated sample; in time course experiments: 0-hr time point) as 1.

Semi-quantitative real-time RT-PCR
Total RNA was isolated from whole embryos at E8.5 and E9.5 stages using an RNeasy mini kit (QIAGEN) according to the manufacturers' instructions. For the reverse transcription reaction, 0.5 μg of total RNA was used with SuperscriptIII (Invitrogen) and oligo-dT primers. To analyze the expression of Pofut1 and Hprt, the following primer sets were used, respectively. Pofut1 exon6 F (5'-TGAGAGCTACGTGTCAGAGATCCA-3') and exon7 R (5'-CACAGTTTCCAATGAAGTGGTCAG-3'). Hprt F (5'-TCCTCCTCAGACCGCTTTT -3') and R (5'-CCTGGTTCATCATCGCTAATC -3'). The house keeping gene Hprt was used as the "reference" gene. PCR reactions were carried out in 96-well microtiter plates in a 20-μl reaction volume with KAPA SYBRs FAST Universal 2X qPCR Master Mix (KAPABIOSYS-TEMS). A Thermal Cycler Dice Real Time System (TAKARA) was programmed for an initial step of 96˚C for 30 sec, followed by 40 cycles of 96˚C for 5 sec, and 60˚C for 30 sec. The relative expression level was calculated by the ΔΔCt method, and normalized with reference gene expression. Specificity of PCR amplification of each primer pair was checked in advance by running the PCR product on an agarose gel, and confirmed by melting curve analysis.

Establishment and culture of ES cell lines
Whole embryo culture E9.5 stage MCH embryos were harvested in M2 medium (Sigma) supplemented with 10% FBS, and rotation cultured in 75% rat serum, 25% DMEM (Invitrogen) with or without DAPT (Selleckchem) for 8 hrs.

Establishment of Poftu1 mutant mice deficient for O-fucosyltransferase activity via CRISPR/Cas9 mediated genome editing
In order to assess the importance of the O-fucosyltransferase activity of Pofut1, we established point mutation mice by CRISPR/Cas9 mediated genome editing. Two independent sites, Arg 245 and the DXD-like motif ERD, were chosen as target sites. The sgRNA for each target was selected using the "CRISPR design" tool (http://crispr.mit.edu) [40] (Fig 1A). To generate the 245 th arginine into alanine (R245A)-point mutant mice, we compared two approaches; one was injecting in vitro transcribed R245A sgRNA and hCas9 mRNA with single-strand oligo deoxynucleotides (ssODN) as a donor oligo [41], and the other was injecting the circular bicistronic expression vector pX330/R245A, and expressing R245A sgRNA and hCas9 mRNA with ssODN [37]. We also compared the targeting efficiency with long and short donor oligos, (A) Schematic of the two independent point mutation sites of the Pofut1 gene. The first mutation was introduced in the 5 th exon, and the resulting allele carried the R245A mutation. The second mutation was introduced in the 7 th exon and the resulting allele carried the ERD into GGG (3G) mutation. Each sgRNA coding sequence is capitalized and labeled in green. The protospacer-adjacent motif (PAM) sequence is capitalized and labeled in blue. Two donor oligos and one donor oligo were used for the R245A site and 3G site, respectively. The introduced mutation is indicated in red in the donor oligo sequence, and resulting mutant amino acids are shown below the boxes. Restriction enzyme which had 64 bp, and 34 bp homology arms, respectively, with a 2 bp mutation site in the middle of the ssODN ( Fig 1A). As Pofut1-null embryos exhibit embryonic lethality around E10, we conducted the F0 generation analysis by dissecting mutant embryos at E9.5. The efficiency of gene modification, including non-homologous end joining (NHEJ) mediated non-specific insertion and deletion (in-del) mutations, and homology-directed repair (HDR) mediated knock-in, was higher in the case of injecting transcribed RNA (9 out of 11 embryos) than injecting pX330/R245A (8 out of 17 embryos by injecting in 5 ng/μl, and 2 out of 8 embryos by injecting in 2 ng/μl) ( Table 1). Injecting pX330/R245A induced more mosaicism than injecting transcribed RNA. Knock-in of the point mutation occurred even using short donor ssODN (7 out of 18 embryos or pups) with comparable frequency to using long donor ssODN (3 out of 6 embryos) (Tables 1 and 2). We obtained 9 homozygous mutant embryos, including 7 embryos that had the correctly knocked-in point mutation in both alleles (3 embryos) or one of two alleles (4 embryos). Most of the homozygous mutant embryos (8 out of 9 embryos) phenocopied Pofut1 Δ/Δ embryos with differing ranges of severity, suggesting that Pofut1 R245A/R245A may display embryonic lethality. Indeed, we never obtained homozygous mutant pups when we tried to establish a Pofut1 R245A mutant line ( Table 2). Two Pofut1 R245A mutant pups were obtained by injecting the lower (2 ng/μl) concentration of pX330/R245A (Panels A-H in S1 Fig, Table 2). Some pups displayed several common phenotypes such as a small body, kinked tail, and coarse hair (Panels C and D in S1 Fig). Direct sequencing analysis of these pups revealed that the mosaicism, which may have been caused by introducing the mutation later than the two-cell stage. The pups with lower wild-type allele contribution, which were likely to contain homozygously mutated cells, corresponded with the severity of the phenotypes (Panels G and H in S1 Fig).
Next, we established Pofut1 3G mutant lines in which the DXD-like motif, tripeptide ERD, was mutated into GGG (3G) by injecting transcribed 3G sgRNA and hCas9 mRNA with donor ssODN (Fig 1A). We could not obtain homozygous mutant mice by F0 generation analysis at the E9.5 stage, presumably because of the lower activity of double-strand breaking by sites, MscI and StuI, which were created by the mutations, are shown with lines above the donor oligo sequence. PCR primers (R245A Fw, R245A Rv, 3G Fw, and 3G Rv) used for PCR genotyping are shown as arrows. (B) RFLP analysis of F1 embryo samples. Amplified PCR product using primers R245A Fw and R245A Rv, followed by digestion with MscI. The genotypes shown above were genotyping results by direct sequencing. (C) RFLP analysis of F1 embryo samples. Amplified PCR product using primers 3G Fw and 3G Rv, followed by digestion with StuI. The genotypes shown above were genotyping results by direct sequencing. (D,E) Direct sequencing results of PCR products using primers R245A Fw and R245A Rv, and primers 3G Fw and 3G Rv, respectively.
https://doi.org/10.1371/journal.pone.0187248.g001 Indicated RNA or vector and ssODN mixture was injected into zygotes, two-cell embryos were transferred into pseudo-pregnant females, and embryos were dissected 9 days after transfer at stage E9.5. Genotype of all embryos was analyzed by direct sequencing, and embryos that had the wild-type allele are indicated as Hetero.
the sgRNA and Cas9 for this locus. Instead, we obtained 4 heterozygous knock-in pups (Panels J and K in S1 Fig, Table 3). It has been reported that off-target mutations may occur using the CRISPR/Cas9 system. Previous studies demonstrated that loci with mismatches between sgRNA and target DNA sequences in the "seed sequence", which is 8-12 bp upstream of the PAM sequence, were tolerated for the double strand break by the CRISPR/Cas9 system [36,40,42,43]. We analyzed potential off-target sites that did not contain mismatches 8 bp upstream of the PAM with an NGG sequence or less than two mismatches in the entire 3G sgRNA sequence for mutations in the F0 generation Pofut1 3G pups by direct sequencing. However, we could not find any mutations (primer pairs used for this analysis are shown in S1 Table). No off-target sites that fall into the categories shown above were observed for R245A sgRNA.

Pofut1 R245A and Pofut1 3G proteins were down-regulated by posttranscriptional regulation
In order to examine the expression level of Pofut1 protein, we produced mouse monoclonal antibodies against Pofut1 peptides. After screening the clones, ascites from each clone were used for western blot analysis using whole lysate of Pofut1 +/+ , Pofut1 Δ/+ , and Pofut1 Δ/Δ embryos. We obtained 4 clones that can detect endogenous Pofut1 protein by western blot (Panels A-D in S5 Fig). Immunoreactivity of these antibodies for wild-type, and Pofut1 R245A and Pofut1 3G mutant proteins were examined using cell lysates that expressed Flag-tagged Pofut1 protein, and confirmed equivalent reactivity of these antibodies for wild-type and mutant Pofut1 proteins (Panels E-I in S5 Fig). These antibodies were used to examine Pofut1 protein expression in each mutant, and both Pofut1 R245A and Pofut1 3G protein levels were decreased in Pofut1 R245A/+ and Pofut1 3G/+ embryos, respectively, yielding approximately half the amount of Pofut1 protein than in wild-type (Fig 4A and 4B). Pofut1 R245A/R245A embryos at E8.5 expressed a little Pofut1 protein (Fig 4A), and there was almost no expression in E8.5 Pofut1 3G/3G embryos (Fig 4A). Pofut1 protein became undetectable in E9.5 Pofut1 R245A/R245A and Pofut1 3G/3G embryos (Fig 4B). Pofut1 mRNA levels in Pofut1 R245A/+ , Pofut1 3G/+ , Pofu-t1 R245A/R245A , and Pofut1 3G/3G embryos were not decreased (Fig 4C), suggesting that the Pofut1 protein level was regulated by post-transcriptional mechanisms. In order to determine if the decreased Pofut1 protein level was caused by an unexpected mutation by the CRISPR/Cas9 system, we amplified the entire coding sequence of Pofut1 cDNA from total RNA of each embryo, analyzed by direct sequencing, and observed no mutations other than the targeted mutations. The observed phenotype of Notch1 localization may be caused by the decreased protein level rather than the lack of O-fucosylation.

Pofut1 protein was degraded through a lysosome-dependent protein degradation pathway
In order to examine why Pofut1 R245A and Pofut1 3G proteins became unstable, we established ES cell lines from each Pofut1 mutant, and conducted biochemical analyses. Protein stability is mainly regulated by two mechanisms, the ubiquitin-proteasome pathway or the lysosomedependent pathway. If the decrease in Pofut1 protein was due to destabilization of the protein, inhibition of either pathway would stabilize the Pofut1 protein. After depleting feeder cells, each ES cell line was seeded and treated with proteasome inhibitor, MG132, or lysosome inhibitor, Chloroquine (CLQ), and then the Pofut1 protein amount in each ES cell line was examined by western blot. Pofut1 protein amounts in both wild-type and Pofut1 R245A/R245A ES cells were decreased when treated with MG132 regardless of the concentration (Fig 5A and 5B). On the other hand, Pofut1 protein amounts in both wild-type and Pofut1 R245A/R245A ES cells were increased upon CLQ treatment in a dose-dependent manner (Fig 5C and 5D). These inhibitor's inhibitory effects were confirmed by examining ß-catenin (Panel A in S6 Fig) and  S6 Fig), as these proteins were reported to be degraded through the proteasome-and the lysosome-dependent pathway, respectively [44,45]. Time course experiments showed that the Pofut1 protein amount was gradually increased by an unknown mechanism after seeding to plates (Fig 5E-5G). This increase was also observed in cells), and Pofut1 R245A/R245A (3 embryos, n = 34, 55, 55 cells  Pofut1 point-mutations that disrupt O-fucosyltransferase activity in mice decrease its own protein stability Pofut1 point-mutations that disrupt O-fucosyltransferase activity in mice decrease its own protein stability heterozygous cells (Fig 5E and 5G), regardless of treatment with or without DMSO as the vehicle control. MG132 treatment slightly reduced or did not affect Pofut1 protein levels in ES cell lines of any genotype, whereas CLQ treatment increased the Pofut1 protein level in wild-type and heterozygous ES cell lines (Fig 5E and 5G). In Pofut1 R245A/R245A ES cells, the Pofut1 protein level also increased upon CLQ treatment. In contrast, the Pofut1 protein level in Pofut1 3G/3G ES cells was very low, and up-regulation was never observed upon seeding or with CLQ treatment ( Fig 5G). As the Pofut1 protein level was partly recovered by CLQ treatment in Pofu-t1 R245A/R245A ES cells, we tried to examine O-fucosyltransferase deficient Pofut1 functions in CLQ treated ES cells by analyzing Notch1 maturation and localization. In WT ES cells, most Notch1 was cleaved at the S-1 position. In contrast, a larger fraction of full-length Notch1 was retained in Pofut1 R245A/R245A ES cells compared with wild-type ES cells (Panels B and C in S6

Relationship between Notch signaling activation and stabilization of Pofut1 protein
In order to analyze the relationship between the Notch signaling activation and Pofut1 protein destabilization, we first examined the Pofut1 protein amount upon gamma-secretase inhibitor, DAPT, treatment with E9.5 stage wild-type embryos, which inhibits Notch signaling activation. Although Notch1 cleavage for NICD production was completely blocked in the PSM of DAPT-treated embryos (Panel A in S7 Fig), the Pofut1 protein amount only slightly decreased (Panels B and C in S7 Fig). Consistent with this result, in the PSM of RBPjk mutant embryos, in which the Notch signaling mediated trans-activation is completely abolished, no significant change in the Pofut1 protein amount was observed (Panels D and E in S7 Fig). As the Notch signaling is activated in only some cells in the PSM and the activation is rapidly changed from posterior to anterior, even if the Pofut1 protein amount was altered by the Notch signaling activity, it may not be observed clearly in the PSM lysate. In order to observe the Pofut1 protein amount when Notch signaling was constitutively activated, we established activated-form Notch1 inducible cell lines using wild-type and Pofut1 R245A/R245A ES cells. The Pofut1 protein amount in wild-type ES cells, but not in the mutant ES cells, was slightly increased by the induction of the active-form of Notch1 protein (NICD) (Panels F and G in S7 Fig), suggesting that the Pofut1 protein amount may be regulated partially through Notch signaling activation. However, it is unlikely that lack of Notch signaling activation is the main cause of Pofut1 R245A protein reduction.

Pofut1 functions for Notch1 protein subcellular localization in the PSM
It has been shown that Pofut1 is indispensable for correct localization of Notch1 protein in the PSM, as Notch1 protein was abnormally accumulated intracellularly in the Pofut1 deficient PSM [26], similar to the Ofut1 mutant in Drosophila [27,28]. However, it was still unclear in which organelle Notch1 protein was accumulated and how the localization was regulated.
Although Notch1 protein partially co-localized with the ER marker KDEL in the Pofut1-deficient PSM, a substantial amount of Notch1 protein was observed intracellularly without colocalization with markers for the ER, Golgi, or endosomes by immunohistochemistry analysis (Panels M-P in S2 Fig and [26]). In this study, we conducted immuno-TEM analysis for detailed analysis of Notch1 protein localization, and found that the majority of Notch1 protein accumulated in the ER in the Pofut1-deficient PSM. The differing Notch1 localization with each method may be explained simply by a topological difference, as proteins with the ER retention signal, the KDEL peptide, generally localize in the lumen of ER, while we used an anti-Notch1 C-terminal antibody located outside of the ER to detect Notch1 protein localization. Another possibility is that ER-localized proteins may be distributed unevenly in the ER, and each ER marker labels only a portion of the ER structure in the cells, as reported previously [30]. Notch1 protein accumulation in the ER of Pofut1-deficient PSM suggests two possibilities regarding Pofut1 function. One is that Pofut1 has chaperone activity for Notch1 protein, and the other is that O-fucose modification of Notch1 by Pofut1 works in quality control of Notch1 protein in the mouse PSM, as reported in Drosophila [27,46]. We aimed to address these possibilities, and established two different point mutation lines, both of which disrupted essential residues for Pofut1 O-fucosyltransferase activity. However, unexpectedly, Pofut1 protein stability was markedly reduced in these mutants (Fig 4A and 4B), and caused the same phenotypes as in Pofut1 null mutants. Thus, it was not possible to clarify the importance of Pofut1 O-fucosyltransferase activity for Notch signaling in this work.
ER-resident chaperone dysfunction has been reported to induce ER-stress [47], and a swollen ER structure was observed in the cells with ER-stress [48]. In the Pofut1 mutant PSM, we did not observe any abnormalities in ER structure (Fig 3B and 3C), suggesting that, if Pofut1 works as a chaperone, it may only be for a few specific targets. One of the possible candidate proteins other than Notch1 is the Notch1 ligand Dll1. Delta was reported to be modified with O-fucose at the EGF-like repeats in its extracellular domain [49]. Therefore, it is likely that Pofut1 and Dll1 associate together in the ER. Furthermore, Dll1 subcellular localization was altered in the Pofut1 deficient PSM [26,50]. However, different from Notch1 localization, an adequate amount of Dll1 was observed on the cell surface of the Pofut1-deficient PSM [26,50]. It is possible that Dll1 mis-localization may have occurred due to the lack of transendocytosis, which generally occurs upon Notch1 and Dll1 interaction, caused by the markedly reduced cell surface amount of Notch1 protein. Further detailed analysis of Dll1 localization in the Pofut1 deficient PSM may reveal a novel Pofut1 function.

Regulation of Pofut1 protein stability
Unexpectedly, Pofut1 protein stability was markedly reduced in our mutants (Fig 4A and 4B). As the Pofut1 R245A and Pofut1 3G mRNA levels did not change compared with wild-type ( Fig  4C), it is likely that the protein reduction was regulated by post-transcriptional machinery. Even though there is a possibility that these mutations may disrupt Pofut1 protein conformation and decrease the stability, we think it is unlikely, at least for the Pofut1 R245A mutant, for the following reasons: First, we made two different types of point mutant mice, and both mutant proteins were similarly destabilized. Second, previous reports demonstrated that overexpression of Pofut1 R245A mutant proteins driven by exogenous promoters exhibited similar expression levels to wild-type protein in mammalian cells (Panels E-I in S5 Fig and [22]), Drosophila cells [20,27], and transgenic Drosophila [27,30]. Furthermore, physiological expression of Pofut1 R245A in Drosophila showed only a slight decrease of the protein [46]. Third, crystal structure analysis of the C.elegans Pofut1 homologue, CePofut1, revealed that the 240 th Arginine (R240) in CePofut1, which corresponds to the 245 th Arginine in mouse Pofut1, is the main residue that captures GDP-fucose directly and catalyzes it; thus, R240 is a key residue for O-fucosyltransferase activity [29]. The side chain of R240 is not likely to contribute to the conformation of the CePofut1 protein. On the other hand, the DXD-like motif is on an alphahelix, which composes part of the donor binding site; therefore, 3G mutants may disrupt Pofut1 protein structure. Indeed, analysis using the Protein 3D modeling software Phyre2 [51] supported these predictions, as Pofut1 R245A protein did not affect the 3D structure, but the DXD-motif mutation in Pofut1 3G protein changed the structure from an alpha-helix to a betasheet. Furthermore, we observed lower levels of Pofut1 3G protein when it was overexpressed in mammalian cells (Panels E-I in S5 Fig). We could observe some Pofut1 protein in R245A mutant embryos at E8.5 (Fig 4A), as well as in some conditions in ES cells (Fig 5A), but the protein level in 3G mutants was always very low (Figs 4 and 5). Taken together, destabilization of Pofut1 R245A protein may be induced by the mechanisms caused by the mutation in the Pofut1 O-fucosyltransferase enzymatic core domain. One of the possible mechanisms is that the mutation altered the affinity of Pofut1 protein for O-fucosylation acceptor proteins, and the interaction affected the degradation of Pofut1 protein. The other possibility is that Pofut1 activates some signaling pathway through its O-fucosyltransferase activity, and the downstream target in turn regulates Pofut1 protein stabilization. In the latter case, the most likely candidate of signaling is the Notch pathway because it was reported that the O-fucosyltration of Notch receptors by Pofut1 is important for Notch protein synthesis and trafficking, for increasing the binding affinity of Notch and its ligands, and for activation of Notch signaling [22-24, 46, 52]. We tried to evaluate this possibility, but inhibition of Notch signaling in the PSM using inhibitors or RBPjk mutant mice displayed no clear evidence for the necessity of Notch signaling activation for Pofut1 protein stabilization (Panels B-E in S7 Fig). However, there is still a possibility that cyclic activation of the Notch signaling averaged differential Pofut1 protein amounts in the PSM. We tried to examine the correlation between Pofut1 protein stabilization and Notch signaling activation in vivo by immunostaining the PSM with Pofut1 and anti-NICD antibodies, but unfortunately, none of the antibodies against Pofut1 could detect endogenous Pofut1 protein specifically. This was consistent with the western blot results with these antibodies in which several non-specific bands stronger than endogenous proteins were displayed (S5 Fig). Another possibility is that Pofut1 stability is regulated via the signal-sending cells that express Notch ligands. As subcellular localization of Notch1 is abnormal in the null and point mutants (Figs 2 and 3), it may affect the reverse signaling in the signal-sending cells independent of Pofut1 O-fucosyltransferase activity. Pofut1 O-fucosyltransferase activity may be important for activation of reverse signaling in the signal-sending cells because all Notch ligands contain EGF-like repeats, some of which were reported as Pofut1 targets [50,53]. However, the reverse signaling that is activated in the signal-sending cells upon Notch receptor and ligand interaction is largely unknown. Furthermore, over 100 proteins that contain EGF-like repeats with the O-fucosylation consensus sequence have been identified [54]. This suggests that there are more candidates of O-fucosylation acceptor proteins or signaling pathways activated by Pofut1 that may be involved in Pofut1 protein regulation. A future study is needed to identify the factors that are involved in regulating the stability of the Pofut1 protein.

Importance of O-fucosyltransferase activity of Pofut1 in somitogenesis
It has been reported that O-fucosylation of the Notch extracellular domain by Pofut1 is important for quality control of Notch protein and its interaction with ligands [22-24, 46, 52], and these functions are likely important for periodic Notch signaling activation during somitogenesis. Another fucntion of Pofut1 O-fucosyltransferase activity may be to regulate synchronous activation of Notch signaling during somitogenesis through Lfng and Dll3 functions. Both Lfng and Dll3 have reported inhibitory functions for Notch signaling activation during somitogenesis, but in different ways. Our recent observation revealed that Lfng plays a crucial role in synchronizing Notch signaling in trans through inhibition of Dll1 [13], but it was not clear whether the glycosyltransferase activity of Lfng is important for somitogenesis. Analysis of glycosyltransferase activity-deficient Lfng mutant mice with a similar strategy to this study would give insight on the importance of glycosyl-modification of Notch1 and Dll1 during somitogenesis. Dll3 is localized in the Golgi and interacts directly with full-length Notch1, inhibiting the signaling in cis by sending Notch1 to lysosomes before Notch1 processing [55,56]. A recent study reported that Dll3 was glycosylated by Pofut1, followed by Lfng. The Dll3 mutants, which have disrupted Pofut1 target sites, were unable to rescue the somitogenesis defects in Dll3 null mice [53]. These reports support the importance of Dll3 glycosylation for somitogenesis. Taken together, the glycosylation of Notch1, Dll1, and Dll3 by Pofut1 and Lfng may play a crucial role in precisely controlling the activation of Notch signaling during somitogenesis, but further studies are needed to reveal the underlying mechanisms.