A point-mutation in the C-domain of CMP-sialic acid synthetase leads to lethality of medaka due to protein insolubility

Vertebrate CMP-sialic acid synthetase (CSS), which catalyzes the synthesis of CMP-sialic acid (CMP-Sia), consists of a 28 kDa-N-domain and a 20 kDa-C-domain. The N-domain is known to be a catalytic domain; however, the significance of the C-domain still remains unknown. To elucidate the function of the C-domain at the organism level, we screened the medaka TILLING library and obtained medaka with non-synonymous mutations (t911a), or single amino acid substitutions of CSS, L304Q, in the C-domain. Prominently, most L304Q medaka was lethal within 19 days post-fertilization (dpf). L304Q young fry displayed free Sia accumulation, and impairment of sialylation, up to 8 dpf. At 8 dpf, a marked abnormality in ventricular contraction and skeletal myogenesis was observed. To gain insight into the mechanism of L304Q-induced abnormalities, L304Q was biochemically characterized. Although bacterially expressed soluble L304Q and WT showed the similar Vmax/Km values, very few soluble L304Q was detected when expressed in CHO cells in sharp contrast to the WT. Additionally, the thermostability of various mutations of L304 greatly decreased, except for WT and L304I. These results suggest that L304 is important for the stability of CSS, and that an appropriate level of expression of soluble CSS is significant for animal survival.

www.nature.com/scientificreports/ transporters that are necessary for expression of Sia, including cmas, a gene for CSS. They serve as ideal vertebrate models because of their ease of handling, large numbers of progeny per generation, and in particular, their translucent embryos, which enable easy observations 21 . Furthermore, their stages of normal development have been well studied and described 22 . The greatest advantage, however, is that most reverse genetics approaches are applicable. In order to determine whether the C-domain of CSS plays a role at the organism level, we searched for medaka models with point mutations in the C-domain of the cmas through screening of the medaka Targeting-Induced Local Lesion IN Genome (TILLING) library 23 , and found a medaka strain with a single mutation in the C-domain of CSS (L304Q), which was lethal to young fry by three weeks old. Together with the severe phenotypes of the L304Q medaka, an induction of instability of the CSS molecule by this mutation demonstrated the significance of the C-domain at the organism and molecule levels.

Results
A point-mutation in the C-domain of CSS was lethal in medaka fish. To study the biological function of the C-domain of CSS using medaka as an animal model, we screened the medaka TILLING library with 5,760 mutant medaka strains 23 The medaka cmas (a gene for CSS) is located on chromosome 18, and encodes 459 amino acids (ENSORLG00000018421). The N-domain of medaka CSS (mdkCSS) contains five conserved motifs with high similarity to other vertebrate CSSs, while the C-domain of mdkCSS shows less homology to the C-terminal domains of CSS in other organisms (human, mouse, and rainbow trout) (Supp_FigS1). To identify medaka with a mutation in the C-domain, we targeted exons 6-7 of the cmas gene, screening the TILLING library by the high-resolution melting curve (HRM) assay 24 . This was followed by re-sequencing of the PCR fragments. Finally, only L304, considered to be a missense mutation, was found.
In order to determine the in vitro activity of this missense mutant mdkCSS from the TILLING library, recombinant L304Q mdkCSS (L304Q) protein and wild-type mdkCSS (WT) protein ( Fig. 1A) were expressed in Escherichia coli. The recombinant CSS protein showed a main band at 75 kDa (Fig. 1B, left), which was the molecular mass of mdkCSS with His-tag, S-tag, and thioredoxin-tag in front of the N-terminus of mdkCSS. The in vitro activities of crude mdkCSS proteins were performed with excess substrates, as described in the Methods section. As shown in Fig. 1B, right, L304Q had lower enzymatic activity, in relation to Neu5Ac, Neu5Gc, and KDN, than WT.
After the background mutations in L304Q medaka fish were removed by artificial insemination 25 and outcrossed with wild-type medaka fish (WT) for more than five generations, mdkCSS 304Q/Q (homo), mdkCSS 304L/Q (hetero), and mdkCSS 304L/L (WT) medaka were obtained by crossing heterozygous adult medaka (https:// shigen. nig. ac. jp/ medaka/ downl oad/ strain/ NBRP_ TILLI NG_ en. pdf ). Genotype of each medaka was determined by HRM assay and sequencing analysis (Supp_FigS2). The homo medaka could hatch at 9 dpf and appeared to grow normally until 14 dpf. After that, however, homo medaka fry began to be dead (Fig. 1C). The median lifespan of homo medaka was 19 dpf, which is the 1st fry stage in medaka development 22 , and very few homo medaka could survive until 50 dpf (Fig. 1C). On the other hand, the WT and hetero medaka fry could survive normally until adult fish, although 20% of them were lethal during early developmental stage (Fig. 1C). In any experiments below, the homo, hetero, and WT medaka that were obtained by crossing the same heterozygous adult after at least six generations were used. Therefore, we can safely say that they had strictly the same genetic background except for the point-mutation in question.
Sialylation was reduced in L304Q homo medaka. To understand whether the L304Q mutation in the C-domain of mdkCSS impaired enzymatic activity in vivo, the change in free Sia (a substrate of CSS) in L304Q homo medaka at 8 dpf was quantified by HPLC ( Fig. 2A). The quantity of free Neu5Ac accumulated in homo medaka was 50-fold higher than that in hetero and WT medaka fry. Furthermore, α2,6-sialic acid and α2,8-linked polySia epitopes in L304Q medaka fry were quantified by lectin-blotting with SNA lectin and western blotting using 12E3 antibody, respectively (Fig. 2B,C). The amount of both epitopes in homo medaka fry decreased after 8 dpf. It should be noted that α2,3-sialic acid epitopes were not obviously detected at 14 dpf, compared with the other two epitopes, as probed by MAA-lectin blotting (Supp_FigS3). The MAA epitopes were detected in all the homo, hetero, and WT embryos at 8 dpf, because they came from yolk materials that are known to be maternally derived, but not zygotically expressed (Supp_FigS3).
Cardiomyopathy was observed in L304Q homo medaka. It has been shown that aberrant sialylation causes dilated cardiomyopathy and myopathy 26 . Therefore, the heart development of L304Q homo medaka was observed from 0 to 9 dpf (the hatching stage). L304Q homo embryos appeared to be normal until 6 dpf (the heart development stage); however, from 7 dpf (pericardial cavity formation stage), there often observed such a phenomenon that blood could not completely be pumped out from their ventricles. One contraction out of three was often incomplete (Fig. 3B, homo-1 and 3), which led to the clogging of blood in the ventricle for a longer time in homo embryos than in WT subjects. When the patterns of ventricular contraction and expansion were monitored, ventricular arrhythmia was also observed in homo medaka (Fig. 3B). To confirm the protein level of myosin heavy chain (MHC) in medaka, western blotting using MF20 monoclonal antibodies, which recognizes the heavy chain of myosin II and have been used to highlight the myocardium of medaka at early developmental stage, was performed 27,28 (Fig. 3C), showing that the expression level of MHC protein in homo medaka fry at 14 dpf was much lower than that in WT. These results indicated that cardiomyopathy occurred in the homo fry, which may explain the lethality of the homo medaka.
To see how skeletal myogenesis was going, the body size was compared at 14 dpf (Fig. 4). The body length and body width of homo fry were significantly shorter than those of WT and hetero fry. However, no obvious Comparison of enzyme kinetics of L304Q and WT mdkCSS. The lethality of L304Q homo medaka led us to investigate effects of the L304Q mutation in the C-domain of CSS on enzyme properties. First, kinetic analysis of L304Q and WT mdkCSS was performed using purified soluble proteins. Recombinant mdkCSS proteins were expressed in E. coli and purified by affinity chromatography on a Ni 2+ chelating column. The purified mdkCSS proteins were then quantified by western blotting using an anti-mdkCSS antibody, and showed a main band at 75 kDa (Fig. 5A). For kinetic analysis, each reaction was performed at 25 °C with for 15 min using 0.2 µg of the purified recombinant mdkCSS protein, because these conditions allowed a linear increase in the formation of products. Neu5Ac was used as the substrate from 0 to 4 mM, and CTP was used at a saturating concentration of 5 mM. Based on the Lineweaver-Burk plot (Fig. 5B), the K m values indicated that the L304Q exhibited twice as high affinity for Neu5Ac as WT, and the V max values showed that the L304Q had three times slower reaction www.nature.com/scientificreports/ velocity than WT. The V max /K m values were thus determined as 0.031 min -1 for L304Q and 0.053 min -1 for WT (Fig. 5C), indicating that the efficiency of producing CMP-Sia of L304Q was 58% of that of WT. These results indicate that the L304Q mutation lowers the enzymatic activity of mdkCSS, but this reduction effect is not severe under conditions for kinetic analysis, i.e., in a short-time (15 min) incubation at 25 °C.

L304Q mutation in the C-domain of mdkCSS lowered the solubility in vivo.
In the course of preparation of the recombinant L304Q CSS using E. coli, we noticed that an yield of soluble protein for L304Q was significantly lower than that for WT, suggesting that L304Q is less soluble than WT. To understand whether the L304Q mutation impaired the solubility of CSS in vivo, myc-tagged mdkCSS proteins were transiently expressed in CHO cells. The CSS protein level was quantified by western blotting using anti-myc antibody (Fig. 6A) and www.nature.com/scientificreports/ the mRNA levels of mdkCSSs were quantified by RT-PCR (Fig. 6B). The mRNA levels of L304Q and WT were at the same level (Fig. 6B). On the other hand, the protein levels of L304Q and WT detected in the soluble and insoluble fractions were greatly different. For L304Q, the protein was mainly detected in the insoluble, but not soluble fraction; in contrast, for WT, mainly in the soluble, but not insoluble fraction. This was notably different between L304Q and WT. These results suggest that the L304Q CSS protein easily becomes insoluble after translation when expressed in CHO cells.
We then sought to investigate the intracellular localization of L304Q and WT in CHO cells. We found that unusually round-shaped cells were prominently increased in L304Q-expressing cells, while such abnormal-shape cells were very few in WT-expressing cells (Supp_FigS4A). The proportions of their cell numbers were 25% and 2% for L304Q-and WT-expressing cells, respectively. The abnormal-shape cells did not proliferate and appeared to be finally dead. On the other hand, the normal-shape cells showed much the same intracellular localization profiles for both L304Q-and WT-expressing cells (Supp_FigS4B). It is thus suggesed that the L304Q mutation has no effect on the nuclear-cytosolic localization property of CSS, and that L304Q CSS, once expressed, gradually induces protein aggregation to form the insoluble protein aggregates (Fig. 6A). The protein aggregates might cause the cell shape changes to abnormal-shape cells. However, further studies are necessary to gain further insight into the L304Q CSS-driven intracellular events.
The non-polar amino acid at position 304 of the C-domain was critical for thermostability of the activity. To gain more insight into the insolubility of L304Q CSS, Leu-304 was mutated into various amino acids using the Quick Change Lightning Site-Directed Mutagenesis method, and examined for the thermostability of the activity in vitro. Each His-tagged mdkCSS protein was expressed in E. coli cells, and its protein levels were analyzed by western blotting using anti-His antibodies (Fig. 7A). All the recombinant mdkCSS proteins were successfully prepared as a soluble enzyme, although their yields in the soluble fraction were different from each other. They were then pre-heated at 25, 35, and 45 °C for 15 min before the in vitro activity toward Neu5Ac was measured (Fig. 7B). When Leu-304 changed to a non-polar amino acid (Ile) the CSS activity remained the same as that in WT at any pre-heat temperatures tested. When Leu-304 changed to Ala, which contains a shorter hydrophobic side chain than Leu, the enzyme lost half of its activity once pre-heated at 45 °C for 15 min. When Leu-304 was changed to polar amino acids (Gln, Asn, and Ser), CSS completely lost its activity once it was pre-heated at 45 °C for 15 min (Fig. 7B). It was shown that the non-polar amino acid at the 304 position of the C-domain of mdkCSS was essential for maintaining the thermostability of the enzyme activity.
The thermostability of protein itself was further examined for the L304Q and WT CSS by the protein melting assay. The melting curves (Supp_FigS5) revealed that the melting points (Tm) for L304Q and WT CSS were 48.6 and 45.1 °C, respectively, indicating that the L304Q protein is slightly more thermo-stable than the WT. However, this Tm difference may not be significant, because the L304Q CSS activity prominently decreases at 35-45 °C below the Tm, while the WT CSS activity remains unchanged by sharp contrast. It might thus be considered that the L304Q mutation in the C-domain induces such structural changes that decrease the catalytic activity of the N-domain without changing the protein stability so much. 20 . To investigate if the oligomerization state of the L304Q CSS was different than the WT CSS, the native-PAGE experiments were performed (Supp_FigS6). The L304Q and WT CSSs gave much the same profiles with a major component at around 240 kDa on the native-PAGE, suggesting that the oligomerization status of the soluble L304Q CSS was not so different than the WT CSS.

Discussion
The L304Q medaka model provides the first insight into the biological significance of the C-domain whose presence is unique to vertebrate CSSs. Namely, the L304Q medaka show the 50% lethality at 19 dpf (Fig. 1C) and an impact of the mutation is serious in the survival of this organism. The accumulation of free Sia and the concomitant reduction of α2,6-Sia and α2,8-linked polySia epitopes (Fig. 2) indicate that L304Q medaka lose all CSS activity in vivo. This phenomenon appears to be contradictory to our kinetic data showing that the L304Q mutation has no severe effect on the enzymatic activity at least under conditions used for kinetic analysis, i.e., 15 min incubation at 25 °C (Fig. 5). Interestingly, however, when the L304Q protein is expressed in CHO cells at 37 °C, the L304Q protein is nearly absent in a soluble fraction, and instead accumulated in an insoluble fraction, while, by sharp contrast, WT protein is exclusively recovered in a soluble fraction (Fig. 6). Thus, the L304Q mutation brings about the insolubility under in vivo situations. Notably, in the L304Q-expressing cells, the abnormal-shape cell population prominently increases, probably because the insoluble protein aggregates might facilitate the cell shape changes through an unknown mechanism.
To understand how this point-mutation in the C-domain of mdkCSS changes the solubility of the enzyme, the thermostability of the CSS activity was investigated using a series of L304 substituted CSS mutants, showing that the non-polar long-side chain amino acids are critical for thermostability of the CSS activity (Fig. 7). Obviously, the L304Q CSS activity is highly temperature-sensitive more than the WT. However, the Tm values of the CSS proteins are much the same between them, i.e. 45-48 °C, with a slightly higher Tm for L304Q than WT. Therefore, it is difficult to conclude that the thermal instability of the L304Q CSS is related to its insoluble property in cell. It is rather considered that the L304Q might form those particular structures not only that decrease the catalytic activity of the N-domain at 35-45 °C, but also that allow undesired complex formation, leading to the easy accumulation in the insoluble fraction when expressed in CHO cells. Based on the results of the native-PAGE of purified CSSs (Supp_FigS6), it appears that the L304Q mutation does not seriously affect their homo-oligomer formation, as far as L304Q CSS is purified as a soluble form. It is thus suggested that the www.nature.com/scientificreports/ undesired complex of the L304Q CSS might be formed with other molecules than the CSS. It is notable to point out that Leu/Ile at this position, in addition to the adjacent conserved Leu residue, is highly conserved in other vertebrate CSSs (Fig. 8A), which suggest the importance of this Leu/Ile throughout evolution of the vertebrate CSS. Furthermore, the 3D structures of L304Q and WT mdkCSS are constructed from the crystal structure of the C-domain of mouse CSS 20 using MOE homology modeling (Fig. 8B). Leu-304 is located in the β4-sheet, which is situated in the interior of the C-domain of CSS, and neither at the boundary surface of the C-domain oligomer 20 . When Leu-304 is mutated to Gln, H's of NH groups at both of the side chain and the backbone of the Gln-304 residue form extra hydrogen bonds with O of the carbonyl group of backbone Gln-326 residue (Fig,  8B, magnified image). The formation of the extra hydrogen bonds may explain the slightly higher Tm of L304Q compared with WT. The amino acid sequence of the C-domain is homologous to a bacterial sugar phosphatase of the haloacid dehalogenase (HAD) superfamily that catalyzes carbonyl or phosphoryl group transfer reactions. However, neither mCSS nor its C-domain was shown to exhibit phosphatase activity in vitro 20 . The 3D structure of mCSS C-domain has been described as a functional platform for the dimerization of the catalytic domain; that is, the N-domain of CSS 20 . Taken together, it remains to be clarified how this point-mutation at the 304 position of the C-domain is related to the overall structural change of whole molecule leading to the enzymatic insolubility; however, this is the first structure-activity relationship study of the C-domain, which suggests the importance of the C-domain intactness. The lethality of L304Q medaka, together with the observation of the accumulation of free Sia and the extensive reduction of α2,6-Sia and polySia epitopes from 8 dpf, indicates that sialylation is essential for the development of medaka. It has been shown that mutations in Sia metabolism-related genes lead to various human diseases, such as muscular dystrophy, skeletal dysplasia, and intellectual developmental disorders [29][30][31][32][33] . Cardiovascular disorders have also been reported sporadically in patients with the UDP-GlcNAc 2-epimerase/ManNAc kinase (GNE) myopathy and SLC7A4 deficiency, and the cardiac phenotype has also been observed in the N-acetylneuraminate-pyruvate lyase gene (npl) morphant zebrafish 34 . Hyposialylation often causes enlargement of ventricles and weakness in ventricular contraction 26 . As shown in this study, the amount of myosin heavy chain in the hearts of L304Q mutant medaka is significantly reduced at 14 dpf (Fig. 3), indicating that cardiomyocyte production is affected by a sialylation level. Furthermore, the body length and width of L304Q homo medaka fry were also obviously smaller than WT and hetero fry at 14 dpf, indicating that skeletal myopathy also occurs in L304Q homo medaka fry without any behavioral anomalies. The abnormalities of heart motions, such as a clogging of blood in the ventricle and ventricular arrhythmia, observed in the L304Q homo medaka at 8 dpf may also be a sign of the impaired heart development. However, mechanisms by which the reduced sialylation level in L304Q mutant leads to the functional impairment of heart remain to be clarified. For other organs than heart, it is known that sialylation is important for podocyte maturation in mice, with altered sialylation impairing podocyte maturation 35 . However, the tissue morphology of kidney remains unchanged in WT and L304Q homo medaka fry, although the α2,6-Sia epitope disappears in L304Q homo, compared with that in WT medaka fry (data not shown). Further studies on kidney function are necessary in medaka.

Methods
Materials. Neu5Ac, CMP-Neu5Ac, isopropyl-1-β-D-galactopyranoside (IPTG), aprotinin, and leupeptin were from nacalai tesque (Kyoto, Japan). Neu5Gc was from Tokyo Chemical Industry Co. (Japan). KDN was prepared as described previously 36 . 1,2-dimethylenedioxybenzen (DMB) was purchased from Dojindo Molecular Technologies, Inc. (Kumamoto, Japan). Pre-stained Mw marker was obtained from Bio-Rad (Hercules, CA). Enhanced chemiluminescence (ECL) reagents and protein G-Sepharose resins were purchased from Amersham Biosciences (Piscataway, NJ). PVDF membrane (Immobilon P) was a product of Millipore (Bedford, MA). Molecular marker was purchased from Sigma (St. Louis, MO) and BIO-RAD. Ni-NTA-Agarose was purchased from QIAGEN (Valencia, CA). The recombinant enterokinase was a product of Novagen (Tokyo, Japan). An anti-polySia antibody, 12E3, which recognizes the oligo/polyNeu5Ac structure (DP ≥ 5) 37 , was generously provided by Dr. Tatsunori Seki (Tokyo Medical University, Tokyo, Japan). Pure Sambucus nigra lectin (SNA), which recognizes Siaα2 → 6Gal or GalNAc, were purchased from EY Laboratories, INC. (San Mateo, CA). Mouse anti-lectin antiserum to SNA was prepared as described below in "Preparation of anti-mdkCSS antibody". A pure Maackia amurensis lectin (MAA), which recognizes Siaα2 → 3Gal or GalNAc, were purchased from EY Laboratories, INC. (San Mateo, CA). Mouse anti-lectin antiserum to MAA was prepared as described below in "Preparation of anti-mdkCSS antibody". Anti-Myc monoclonal antibody (mAb.) 9E10 recognizing the Myc epitope (EQKLISEEDLN) was kindly provided by Dr. Gerardy-Schahn (Medizinisches Hochschule Hannover, Germany) and prepared as previously described 38 . The mouse monoclonal anti-His-tag antibody was purchased from Cosmo Bio Co. (Tokyo, Japan). The monoclonal anti-myosin 4 antibody (MF20) was purchased homo medaka at 8 dpf. Periodical changes of the blood cell density in the ventricle at the square V (in A) were measured as those of the light and shade intensity using ImageJ. The data from three individual medaka for WT and homo are shown. (C) The expression of myosin heavy chain (MHC) in WT and L304Q homo medaka at 14 dpf. The lysate was analyzed by the SDS-PAGE/western blotting using MF20 antibody, which recognizes the heavy chain of myosin II and have been used to highlight the myocardium of medaka (26)(left upper). GAPDH was also detected by anti-GAPDH antibody as the loading control (left lower). The MHC components were decreased in the homo medaka, which may indicate cardiomyopathy. **indicates p < 0.005. Full-length blots for (C) are presented in Supp_FigS10. www.nature.com/scientificreports/

Identification of mutations in the Medaka TILLING library by HRM analysis. The Medaka TILL-
ING library was previously prepared, and contains 5,760 mutant medaka strains induced by ethylnitrosourea (ENU) 23 . Mutations were identified by high-resolution melting curve analysis (HRM analysis) at the National Institute for Basic Biology, as described previously 24 . Exon 6 to exon 7 of the medaka CSS gene was amplified using the following primers: 5'-AGG TTC GCA TGT CTG TTT -3'(forward); 5'-AGC ATA TTT ATA GAA AAC TAG CAT GAA GTA -3'(reverse).
DNA sequencing. Five microliters of PCR products, identified as ENU-induced mutations by HRM assay, were purified using ExoSAP-IT. The purification process was as follows: 5 μL ExoSAP-IT solution was mixed with the PCR product, heated at 37 °C for 15 min, and re-heated at 80 °C for 15 min. One microliter of the purified PCR product was used as a template for sequencing analysis, which was carried out using BigDye Terminator version 3.1 (Applied Biosystems, Foster City, CA) and the ABI 3730xl sequencing platform.
Genotyping. Fin clips of selected medaka fish or larvae were fixed in 40 μL of methanol and lysed in an appropriate amount of protease K solution (10 mM Tris-HCl (pH 7.5), 10 mM EDTA, and 2 mg/mL proteinase K) prior to heating at 55 °C for 3 h. This was followed by denaturation through protease K activity at 95 °C for 10 min. After centrifugation, the supernatant of each sample was used as genomic DNA. 5'-AGG TTC GCA TGT CTG TTT -3'(forward) and 5'-AGC ATA TTT ATA GAA AAC TAG CAT GAA GTA -3'(reverse) were used to amplify exon 6 and exon 7 of the CSS gene by KOD-plus DNA polymerase (Toyobo, Japan), where the L304 mutation was located. Hetero and homo L304Q mutations were identified by HRM assays and confirmed by DNA sequencing analysis. www.nature.com/scientificreports/ an overnight culture were incubated in 500 mL of LA at 37 °C for 4 h, followed by induction as described above. The supernatant obtained from the sonically disrupted cells was mixed with an equal volume of 10 mM imidazole, 1.0 M NaCl, and 50 mM Tris-HCl (pH 8.0), and applied to an Ni 2+ chelating column (1 × 1.3 cm), which had been equilibrated with 5 mM imidazole, 0.5 M NaCl, and 50 mM Tris-HCl (pH 8.0). After washing with 30 mM imidazole, 0.5 M NaCl, and 50 mM Tris-HCl (pH 8.0), the column was eluted with 100 mM imidazole. The eluted fraction was desalted using VIVA SPIN20. SDS-PAGE, Coomassie Brilliant Blue staining, and western blotting were carried out with an anti-mdkCSS antibody (see above) or anti-His antibody.

Construction of Oryzias latipes CSS plasmids encoding the mutation identified from the TILL
Native-PAGE/western blotting. The purified recombinant L304Q and WT CSS proteins (500 ng) were mixed with 2-mercaptoethanol-free sample buffer for electrophoresis, loaded to native-PAGE (6% polyacrylamide gel) and electroblotted onto a PVDF membrane. The PVDF membrane was blocked with 1% skim milk in a phosphate-buffered saline, containing 0.05% Tween-20 (PBST), at 37 °C for 1 h. Thereafter, the membranes were incubated with anti-mdkCSS antibody (diluted 1:1000 ) at 4 °C for 16 h. Peroxidase-conjugated anti-mouse IgG + IgM-POD (0.4 μg/ml) (diluted 1:5000 ) were used as secondary antibodies. Color development was performed using chemiluminescent reagents (GE Healthcare), and the image was acquired by a LuminoGraph II (ATTO. Japan), and the intensity of targeted band was quantified by CS Analyzer4 (ATTO). HPLC was carried out on a JASCO HPLC system, using a 6.5 × 30 mm resource Q column. The column was eluted at a rate of 1.0 mL/min, with a linear gradient of 0 to 0.2 M NaCl in 20 mM Tris-HCl buffer (pH 8.0) for 10 min. The elution profile was monitored by measuring absorbance at 271 nm, and the peak area was integrated to determine the amount of reaction products. CMP-Neu5Ac was used for quantitative analysis, utilizing a calibrated standard curve, and concentrations of CMP-Neu5Ac, CMP-Neu5Gc, and CMP-KDN in the reaction mixture were calculated using a standard curve. Relative in vitro activity was reported as the amount of CMP-Sia divided by the amount of recombinant enzyme.

Assay for CMP-
Kinetic analysis. Enzyme concentration, incubation period, and CTP concentration were optimized for kinetic analysis. The former was determined with the BCA protein assay. Briefly, 0.25, 0.5, 1.0, 2.0 and 4.0 mM of Neu5Ac were incubated with 0.2 µg of the purified recombinant mdkCSS protein in 50 µL of 0.1 M Tris-HCl (pH 9.0), which contained 5.0 mM CTP, 20 mM MgCl 2 , and 0.1 mM sodium vanadate, at 25 °C for 15 min. The product was quantified by HPLC, as described above, and kinetic parameters were determined using doublereciprocal Lineweaver-Burk plots. Three independent experiments were performed. Thermostability measurement of the recombinant CSS activities. The purified recombinant CSS proteins were pre-heated at 25, 35, or 45 °C for 15 min before in vitro activity measurement. Relative in vitro activity of CSS was reported as the amount of CMP-Neu5Ac divided by the intensity of the recombinant protein, and the value of WT mdkCSS without pre-heating was set to 1.0. Three independent experiments were performed.
Protein thermal shift assay. The

Expression of mdkCSS in CHO cells.
WT and L304Q mdkCSS cDNA were subcloned into pcDNA4/ myc-His A vectors (Invitrogen) using an In-Fusion HD Cloning Kit (Takara Clontech). CHO cells were cultured in MEM-α (Wako, Japan) and supplemented with 100 units/mL penicillin G, 100 μg/mL streptomycin sulfate, and 10% fetal bovine serum (FBS). They were then placed in a 5% CO 2 and 95% air humidified atmosphere at 37 °C. Thereafter, CHO cells (1.5 × 10 5 ) were cultured in a 6-well plate overnight at 37 °C in 5% CO 2 and transiently transfected with 3 μg of each plasmid using the PEI-Max Transfection Reagent (Polysciences Inc., USA). The expression of CSS proteins was quantified by western blotting using anti-myc antibody.

SDS-PAGE and western blotting.
Cells and tissues were homogenized in phosphate buffered saline (PBS) containing 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride (PMSF), 50 mM sodium fluoride (NaF), 10 mM β-glycerophosphate, and 1 mM sodium o-vanadate. Protease inhibitors were also included (1 μg/ mL each of aprotinin, leupeptin, and pepstatin, 2 μg/mL of antipain, and 5 mM EDTA). After centrifugation www.nature.com/scientificreports/ at 15,000 × g for 15 min, at 4 °C, the supernatants were collected and used for further analysis. Each lysate was denatured in Laemmli buffer containing 5% (v/v) β-mercaptoethanol at 60 °C for 20 min. The samples were then electrophoresed using 7.5% or 10% SDS-PAGE and electroblotted onto a PVDF membrane. The following procedures were described in the section "native-PAGE/western blotting" with the exception that blocking was performed with 1% BSA, instead of 1% skim milk.

Immunostaining. CHO cells were grown on coverslips and transfected with each plasmid using PEI-Max
Transfection Reagent (Polysciences Inc., USA). The cells were then fixed with 4% paraformaldehyde at 25 °C for 8 min and permeabilized with 0.2% Triton X-100 in PBS for 15 min for immunostaining. After the cells were blocked with 1% bovine serum albumin (BSA) in PBS for 1 h, myc-tagged CSS was detected through incubation with anti-myc mAb.9E10 (1:500 dilution) at 37 °C for 1 h, and Alexa 488-conjugated anti-mouse IgG (Invitrogen, 1:1000 dilution in PBS) at room temperature for 30 min. For nuclear staining, cells were incubated with 1 μg/mL DAPI at 37 °C for 15 min. Images were obtained using a confocal laser scanning microscope (Olympus BX51).
Amount of sialic acid in medaka embryos. The tail tip cut from each medaka embryo was fixed in 40 μL of methanol and subjected to cell lysis in 25 μL of protease K solution (10 mM Tris-HCl [pH 7.5], 10 mM EDTA, and 2 mg/mL proteinase K) at 55 °C for 3 h. This was followed by denaturation of protease K activity at 95 °C for 10 min. After centrifugation, 1 μL of the supernatant was used for genotyping (See above). The remaining medaka embryo was homogenated as above (20 μL) and separated into two equal parts: one part was used for free Sia quantification and the other was used for bound Sia quantification. The free Sia in each sample was quantified using the DMB derivatization method 39 .
Digital video recording and analysis of the heart contraction. Each embryo in the chorion was immobilized in a hole made by 1.5% agarose. The heart contractions were recorded by a stereomicroscope (Olympus SZX12 DP80). Digital pictures were captured at maximum frame rate at a resolution of 1,360 × 1,024 pixels for up to 3 min and recorded in a PC using CellSens Standard software. Movies of heart movements were processed using ImageJ software. Contraction rhythms were measured based on alterations in the intensity of blood cell flow into and out of ventricle. Regions of interest (ROIs) in the ventricle were selected. The pixel intensities of the ROIs were digitized throughout the entire time series examined using ImageJ software.